Absolutism vs Comparativism About Quantity

Shamik Dasgupta
We naturally think that material bodies have weights, sizes, masses, densities, volumes, and charges; that there are spatial distances between them temporal durations between events involving them. These are all features that fall under the category of quantity. In this paper I discuss a question that arises for all quantities but which is best illustrated by the case of mass. The property of having mass is a determinable that appears to have two kinds of determinates. On the one hand, we naturally think that something with mass has a determinate intrinsic property, a property it has independently of its relationships with other material bodies. But we also think that things with mass stand in various determinate mass relationships with one another, such as x being more massive than y or x being twice as massive as y. My question is: of the intrinsic masses and the mass relationships, which are fundamental? According to a view I will call absolutism, the intrinsic masses are fundamental. Loosely speaking, the view is that the most fundamental facts about material bodies vis a vis their mass include facts about which intrinsic mass they possess. The absolutist does not deny that things with mass stand in determinate mass relationships, and she might even agree that those relationships are fundamental too. But more likely she will think that those relationships hold in virtue of the intrinsic masses: if my laptop is twice as massive as my cup, the idea is that this is (at least partly) in virtue of the particular intrinsic mass that they each possess.1 In contrast, comparativism is the view that the most fundamental facts about material bodies vis a vis their mass just concern how they are related in mass, and all other facts about their mass hold in virtue of those relationships.2
1 I say at least partly because an absolutist might say that it is in virtue of their intrinsic masses and various higher-order relations between those intrinsic masses (see for example Mundy [24]). But this issue will not concern us here. 2 Along, perhaps, with higher-order facts about those mass relationships, though again this issue does not concern us here. Thanks to David Baker for helpful suggestions about how to formulate these views.

I will describe both these views more precisely in Section 1, and as we will see an analogous issue arises for all quantities. Given the central role that quantities play in our understanding of the natural world, the question of absolutism vs comparativism is central to any inquiry into what the natural world most fundamentally consists in. Moreover there is more at stake in the issue than just our understanding of quantities, for comparativism brings with it a commitment to fundamental external relations and these have historically been treated with some suspicion. It is therefore surprising that the contemporary metaphysics literature contains relatively little discussion of the issue. In this paper I motivate and defend comparativism. But I am less interested in pressing that particular view as I am in formulating the issue and broadly surveying what I take to be the more important lines of argument for each position. The main consideration in favor of comparativism is developed at the end of the paper in Section 8 and is based on the idea that we can only ever observe mass relationships. More fully, the idea is that even if material bodies possessed the intrinsic masses posited by the absolutist, those intrinsic masses would be undetectable in a very strong sense: the structure of the physical laws governing our world would guarantee that they could never have an effect on our senses. As is well known, absolute velocity and absolute simultaneity are undetectable in this very same sense, and for this reason most contemporary metaphysicians and physicists call them redundant or superuous and dispense with them on Occamist grounds. My thought is that the same reasons should compel us to dispense with the intrinsic masses posited by the absolutist. More precisely, the Occamist principle I appeal to is that positing undetectable structure is a vice, in the sense that if one theory of the material world posits undetectable structure that another does not then all else being equal we should prefer the latter. All else being equal, then, we should prefer comparativism. Now, whether all else is indeed equal depends on whether there are stronger countervailing reasons to reject comparativism. So in sections 27, the bulk of the paper, I consider a number of potential objections to comparativism and argue that none are convincing; hence my preference for comparativism. The discussion will mostly be limited to the case of mass, and I will make the simplifying assumption that the correct physical theory of mass is a classical, Newtonian theory. The assumption is of course false, but much of the paper discusses rather general, philosophical considerations that do not depend on the physical details, and in those discussions this simplifying assumption is harmless. Considerations from physics are dis-

cussed towards the end of the paper, and there the assumption may be less harmless. But I continue to assume it because the considerations turn out to be conceptually rather delicate and so it is worth clarifying them in the context of this simplifying assumption before asking whether they generalize to more realistic physical theories. I think the considerations stand a good chance of generalizing to at least some other quantities and physical theories, but whether this so is a question for another time.

More on Absolutism and Comparativism

The absolutism vs comparativism issue per se has not received much discussion. To be sure, views that count as absolutist have been defended by Armstrong [2] and [3], Mundy [24], and Lewis [22]; and views that count as comparativist have been defended by Ellis [13], Bigelow and Pargetter [5] and [6], and Field [14] and [15]. But those discussions are often intertwined with other issues about quantites, so it will help to clarify the issue I have in mind before considering arguments either way. I understand both views in terms of the idiom of one kind of fact holding in virtue of another. As I use the phrase this is an explanatory idiom: to say that a fact holds in virtue of another is to say that the latter explains the former in a distinctively metaphysical sense. To illustrate, imagine asking what explains Europes being at war in 1939. A causal explanation might describe events during the preceding 50 years that led, say, Chamberlain to declare war. But there is another kind of explanation that would try to say what goings on in Europe at the time made it count as a continent at war in the rst place. Regardless of what caused the conict, someone in search of this second explanation recognizes that war is not a sui generis state and that there must therefore have been something about the continent that made it count as being a continent at war rather than (say) one at peace. A plausible answer is that it was at war in virtue of how its citizens were acting, for example that some were committing politically motivated acts of violence against others. As I use the phrase, an explanation of this second kind is a statement of that in virtue of which Europe was at war. I take this kind of explanation to be reasonably intuitive: regardless of the truth of this explanation of why Europe was at war, we seem to understand the claim reasonably well.3 Earlier I said that the absolutist thinks that the most fundamental facts
term ground has been used by Kit Fine and Gideon Rosen to describe this sense of metaphysical explanation (see Fine [16] and [17], and Rosen [27]). But since that term has also been used in other ways in the recent literature I have used the phrase in virtue of to try to minimize confusion.
3 The

about material bodies vis a vis their mass include facts about which intrinsic mass they possess. We can put this more precisely in terms of the above idiom: the view is that there are facts about which intrinsic mass a given material body possesses that are not explained (in the above metaphysical sense) in other termsin particular, that are not explained in terms of its mass relationships with other bodies.4 As I said earlier, the absolutist need not deny that things with mass stand in various mass relationships, but she will likely think that when they do this is at least partly explained (in this metaphysical sense) by the particular intrinsic mass that each body has.5 In contrast, the comparativist thinks that all facts about material bodies vis a vis their mass are explained (in this metaphysical sense) in terms of the mass relationships between bodies.6 This leaves open what kinds of mass relations are most explanatorily basic: they might be mass ratios such as an object being twice as massive as another, orderings such as an object being more massive than another, or even just linear structures such an object lying between two others in mass. But this in-house dispute will not matter for our purposes. For simplicity I will often assume that the comparativist under discussion is of the rst type, but nothing will hang on this.7 As stated, both views are claims about what actually holds in virtue of what and so are neutral on the relation between intrinsic mass and mass relationships in other possible worlds. One might of course argue that comparativism is necessarily true if true at all (mutatis mutandis for absolutism) but I will not take a stand on this issue here.8 I have not yet mentioned facts about mass in a particular scale, such as my laptops being 2 kgs, and one might consider this omission strange since it is this sort of fact that we most often express when talking about mass. But it is not immediately obvious whether this is ultimately a fact
might be because there are facts about which intrinsic mass a material body possesses that are unexplained. Or, if there are innite descending chains of explanations, it might be because any such fact is explained by another fact of the same kind; that is, another fact about which intrinsic mass something has. 5 Though this is not enforced: the absolutist might conceivably think that there are also unexplained facts about mass relationships along with unexplained facts about intrinsic masses. But I do not know of anyone who holds this view or of any reasons in its favor so I will largely ignore it in what follows. 6 Along, perhaps, with higher-order facts about the relations themselves; see footnote 2. 7 Baker [4] discusses which kinds of relations the comparativist should and should not think are explanatorily basic. He also discusses (without endorsing) the idea that the comparativist might introduce what he calls mixed relations. These are inter-quantity relations, for example my mass being twice your length. I will not discuss these mixed relations here. 8 Both views, I should say, are neutral as to the status of facts about the nomic relation between mass and other quantities. The issue just concerns the status of catergorical facts concerning the masses of things.
4 This

that holds in virtue of my laptops intrinsic nature or in virtue of its mass relationships, so to avoid begging questions it is best to state the absolutism/comparativism issue without mentioning these facts and leave their status as a further question (I discuss this question in Section 5). I have so far assumed that holding in virtue of is a relation between facts and I will continue to do so for ease of prose. But those wary of facts may express claims about what holds in virtue of what with the sentential operator because. Thus, when I previously said that Europes being at war holds in virtue of facts about the actions of its citizens at that time, one could re-state this without reference to facts as Europe was at war in 1939 because a battalion of troops marched to Normandie in 1939 and. . . 9 so long as because is understood in the metaphysical rather than causal sense. On this way of talking, the absolutist will assert things like My laptop is more massive than my cup because my laptop has the intrinsic mass M and my cup has the intrinsic mass M*. whereas the comparativist will deny such a claim. Moreover, I have so far made free reference to such things as intrinsic masses and mass relations. For example, the natural interpretation of the indented sentence above takes M to be a term referring to a property and the expression has the instrinsic mass to be a relational predicate holding of my laptop and the property M. But nominalists wary of properties and relations can also make sense of the absolutism vs comparativism issue by reading the indented sentence in another way. For example, they might take has the intrinsic mass M to be a primitive monadic predicate containing no referential devices at all. Still, in what follows I will continue to refer to properties and relations for ease of prose.10 Although I have focused on the case of mass, it should be clear that the same issue arises for other quantities too. For example, consider the case of spatial distance. When material bodies X and Y stand in the determinable
9 This way of expressing claims about what holds in virtue of what is suggested by Fine (see his [16] and [17]). Strictly speaking the right hand side will consist of a list of sentences rather than a conjunction, but this complication does not matter to us here. 10 This is not to say that the absolutism/comparativism issue is entirely independent of the issue of realism about properties. One might argue, for example, that the nominalist has a hard time being an absolutist since her vocabulary would then be required to include an innite number of primitive predicates, one for each determinate intrinsic mass. Still, logically speaking the issues are orthogonal and in what follows I will not be concerned with considerations that depend on a resolution to the question of realism about properties.

relation of being spatially related, there are two kinds of determinate relationships that they enter into. On the one hand, it is natural to think that X stands in a determinate distance relation to Y. But it is also natural to think they stand in various comparative spatial relations to other bodies, for example the relation of X being twice as far from Y as Z is from W, or even just of X being further from Y than Z is from W. But of the former fact about the distance between X and Y, and the latter fact about the comparative distance between X, Y, Z and W, which is fundamental? The absolutist says that the most fundamental facts about distance include facts of the former kind, while the comparativist says that all facts about distance are ultimately explained (in the metaphysical sense) in terms of facts of the latter kind.11 This is why I used the terminology absolutism and comparativism in the case of mass rather than intrinsicalism and relationalism: when we generalize to other quantities such as spatial distance, the facts that the absolutist takes to be most fundamental are themselves facts about relationships. As I said, my main reason for endorsing comparativism (developed in Section 8) is that the intrinsic masses posited by the absolutist would be undetectable in a very strong sense. The Occamist principle described earlier deems this to be a signicant mark against absolutism. However, this is not a decisive mark against the view and if there were stronger countervailing considerations against comparativism then absolutism would remain our best all-things-considered theory. So in Sections 27 I consider six kinds of objection to comparativismobjections from intuition, from modality, from semantics, from kilograms, from Humeanism, and from physicsand argue that none are compelling. All things considered, then, I believe that comparativism is the preferable view.

Objections from Intuition

Let us start with objections to comparativism from intuition. When rst introduced to the issue, absolutism strikes many as being the more intuitive and plausible view. If my laptop is more massive than my cup, it initially seems that this is (at least in part) because of their intrinsic masses. Moreover, comparativism conicts with an intuitive locality principle concerning mass, namely that given a connected region of spacetime R
absolutism/comparativism issue about distance should not be confused with the substantivalism/relationalism issue. The latter issue concerns the relata of spatial relations and asks whether they are, fundamentally speaking, material bodies or regions of space. Whichever way that dispute is resolved, we may then raise the absolutism/comparativism debate by asking whether, at the fundamental level, those relata stand in 2-place absolute distance relations or 4-place comparative relations.
11 The

composed of two sub-regions R1 and R2 , the fundamental facts about mass within R1 and the fundamental facts about mass within R2 together determine the fundamental facts about mass within R.12 This principle strikes us as intuitively correct, but the comparativist must deny it since the mass relationships within R1 and those within R2 do not determine the mass relationships between a body in the one region and a body in the other. So let us agree that absolutism is initially the more plausible view. One might then argue that this is a reason to think that it is true. But the last sentence conates a number of arguments. One unconvincing argument is that just as we are endowed with a reliable faculty of perception, we are also endowed with a reliable faculty of intuition which delivers the verdict that absolutism is true. To this we might well object that there is good reason to doubt that we have a faculty of this sort (for one thing, anatomists and neurologists have yet to nd anything corresponding to it). Here we need not deny that we have a faculty of intuition that is a reliable guide to math and logic; all we need insist is that we do not have one that delivers reliable verdicts about what the fundamental physical properties and relations are, for it is this that absolutists and comparativists disagree about. The denial of such a faculty should therefore be uncontroversial. Still, one might argue that the intuitive plausibility of absolutism carries epistemic weight without appealing to a faculty of intuition. For example, one might say that absolutism is a Moorean truth, a proposition in whose truth we are more certain than any premise used in an argument to the contrary. Or one might point out that absolutism is our starting point in the inquiry, and then argue for a principle of epistemic conservatism according to which our starting point is (defeasibly) justied merely by virtue of being our starting point. Either way, the upshot would be that our initial absolutist inclinations are epistemically signicant. In response, I do not object to the principle of epistemic conservativism or to Moorean approaches to some questions in philosophy. In particular, I concede that absolutisms initial plausibility is at least some reason to believe it. But any such reason is defeated by my Occamist argument to the contrary. To see this, consider the case of absolute simultaneity. While it is initially plausible that there is such a thing as simultaneity, most would agree that considerations from special relativity are enough to defeat any consideration from Mooreanism or epistemic conservativeness in simultaneitys favor.13 Now, I take the initial plausibility of simultaneity
to Eliot Michaelson for elucidating this locality principle for me. the Moorean, this shows that our belief in simulteneity is not, after all, more certain than the premises of any argument against it. To the epistemic conservative, it shows that while our belief in simultaneity may have been epistemically privileged by virtue of being our starting point, its privilege was not enough to ward off arguments to the contrary.
13 To 12 Thanks

to be at least as strong as the initial plausibility of absolutism. Therefore, since my reason to reject absolutism is the same as our reason to dispense with simultaneity (I leave it until Section 8 to make good on this claim), it will be be strong enough to defeat considerations from Mooreanism or epistemic conservativeness in absolutisms favor.

Objections from Modality

Another class of objections to comparativism appeal to modal considerations. The idea behind all the objections is that the comparativist cannot make sense of certain situations that seem, at least on the face of it, to be possible. I believe that those with absolutist inclinations are often moved by these kinds of considerations, so I will consider four such objections in some detail. The rst objection can be put aside reasonably quickly. The objection is that while it seems possible for everythings mass to double tonight at midnight, the comparativist cannot make sense of this since the mass relationships between things tomorrow would be exactly the same as they were today and so by the comparativists own lights there would have been no change. But this objection is not compelling, for the comparativist may claim that some of the fundamental facts about mass concern how something at one time is related in mass to something at another time. If so, she can make sense of the possibility after all. A second, more compelling objection along these lines appeals to the intuition that it seems possible for everythings mass to always have been double what it actually is. The possibility of uniform doubling under consideration here is not one in which everythings mass doubles at a particular time, but rather one in which the entire history of the universe is just like ours with the one exception that at any given time, everythings mass is double what it actually is at that time. The objection is then that while this intuitively seems to be possible, the comparativist can make no sense of it.14 But why think that the comparativist cannot make sense of the possibility of uniform doubling? At this point the objector needs to choose whether to express her objection with modal operators or in the framework of possible worlds. In the latter framework, the possibility of uniform doubling is represented by a possible world just like ours but with the one exception that everythings mass is double what it actually is. To argue that the comparativist can make no sense of such a world, the ob[12] objects to comparativism in this way. Hawthorne [18] outlines the objection and seems to take it to have some force, but he does not explicitly endorse it.
14 Eddon

jector would appeal to a plausible necessitation principle: that if a fact Y holds in virtue of a fact X, then every world in which X obtains is also a world in which Y obtains.15 This necessitation principle is standard in the literature on the in virtue of relation and is extremely plausible. For example, if the fact that Europe was at war in 1939 holds in virtue of facts about the actions of its citizens, then those actions are what accounts for and makes it the case that that Europe was in a state of war rather than (say) a state of peace. Consequently, it seems that any world in which the citizens of Europe act in that way must also be a world in which the continent is at war, just as the principle implies. Now, along with the necessitation principle comparativism implies that worlds agreeing on all the mass relationships agree on all facts about mass. But the so-called doubled world agrees with the actual world on all mass relationships; hence according to the comparativist it is not a world that differs from the actual world regarding any fact about mass and is therefore not a world in which everythings mass is doubled. That is the objection stated in the framework of possible worlds. It can also be stated using modal operators by formulating the necessitation principle in terms of modal operators instead. In what follows I will only discuss the possible worlds version of the objection just outlined. The details of the modal operator version differ subtly, but there is no space to discuss those details and in any case the main morals of our discussion in terms of possible worlds will carry over. How might the comparativist respond to the objection? One response is to become a modal realist in Lewis sense and say that the fundamental facts of the world are really facts concerning a plurality of concrete worlds. The comparativist may then think that the fundamental facts concerning mass relationships include how objects in different worlds relate to one another in mass. A comparativist of this sort will point out that the actual world and the doubled world disagree on their inter-world mass relations, and therefore the necessitation principle does not imply that they agree on all facts about mass. But I do not wish to rest my case on this response. Putting aside the unpopularity of modal realism, a more important worry is whether the modal realist can legitimately allow fundamental relationships between objects in different worlds. For example, the generalization
am simplifying here. As I describe in Section 5, I believe that the in virtue of relation is irreducibly plural in the sense that some facts Y can hold in virtue of some facts X even though no Y taken on its own holds in virtue of anything. If that is right, the necessitation principle should be rephrased with plural variables as follows: if some facts Y hold in virtue of some facts X then every world in which the X all obtain is also a world in which the Y all obtain. But this complication is not relevant to the current objection so I will ignore it here.
15 I

of this approach in the case of spatial distance is that fundamental spatial relations hold between bodies in different possible worlds, and one might argue that this conicts with Lewis account of a possible world as the mereological sum of spatio-temporally related things.16 So it is worth asking whether the comparativist can respond to the objection without appealing to fundamental inter-world mass relationships. I believe she can. In the rst place, she may argue that her failure to make sense of the possibility of uniform doubling is no real vice; and in the second place she may argue that she can, perhaps surprisingly, make sense of the possibility without fundamental inter-world mass relations after all. Both responses seem reasonable to me, so I will discuss each one. Start with the rst. The objection rests on the intuition that a uniform doubling of mass really is possible, but is this right? I nd that my inclinations here depend on my theoretical convictions: when absolutism strikes me as attractive it seems possible, but when I am in the grip of comparativism I feel that the possibility is a silly philosophical mistake. This should not be surprising, since the absolutist and the comparativist are both likely to agree that if absolutism is true then uniform doubling in mass is possible. Now, if our intuition that doubled worlds are possible rests on a prior belief that absolutism is true, the current argument would at best collapse into the argument from intuition just considered or, at worst, beg the question. So the question is whether we have an inclination to think that uniform doubling is possible that is independent of any prior belief in absolutism, and if so how strong that inclination is. I am not sure how one might go about answering this question, but an answer is crucial to the current argument. For now, then, it seems reasonable to take the issue of uniform doubling to be a case of spoils to the victor. The second response is that the comparativist can, perhaps surprisingly, make sense of the possibility of uniformly doubling without appealing to fundamental inter-world mass relationships. To see how let us turn to the third modal objection to comparativism, for our response to it will provide the materials needed to make sense of uniform doublings. So, putting aside for a moment the possibility of everythings mass being doubled, note that it is surely possible for just my laptop to have been twice as massive as it actually is. The third modal objection is that the comparativist cannot even make sense of this. On the face of it, this objection is far more powerful than the last. For while we might reasonably deny that it is possible for everythings mass to have been doubled we must surely agree that my laptop could have been more massive than it is. If the com16 Lewis himself thought that there can be no perfectly natural relationships between objects in distinct possible worlds. See his [21], p. XX.

10

parativist cannot even make sense of this latter possibility, that is a vice indeed. I am therefore surprised not to have seen or heard this argument expressed by those with absolutist inclinations. But why think that the comparativist can make no sense of the possibility? Again, the objection can be developed in the framework of possible or with modal operators, but let us stick to the former. The idea, then, is she can perfectly well make sense of a world W just like ours except that the mass-ratio between my laptop and all other things is double what it actually is. But without inter-world mass relationships, there is no fact of the matter as to whether W is a world in which my laptop is twice as massive as it actually is, or one in which my laptop is the same mass and everything else is half as massive as they actually are. Notice that if the actual bodies and those in W had intrinsic masses then the problem would not arise, for those intrinsic masses would determine mass relationships between bodies in the one world and bodies in the other. But without fundamental inter-world mass relationships, the comparativist has no resources to make a similar inter-world comparison.17 How might the comparativist respond? I believe she can accuse the objection of resting on an incorrect model of how a possible world represents my laptops mass and introduce a better model that allows her to make sense of the possibility in question. To see this, it will help to work with a specic model of how a possible world represents something de re of my laptop in the rst place. I will work with Lewis famous proposal that it does so not by containing my laptop but by containing one of its counterparts, though nothing hangs on this choice. Given this assumption, the world W introduced above can be re-described as a world containing a counterpart of every actual material body such that if my laptop is r times as massive as another body x, my laptops counterpart in W is 2r times as massive as xs counterpart in W. Now, consider an object other than my laptop, such as my cup. Notice that the mass relationships it enters into are extremely similar to those entered into by its counterpart in W. The only difference is that while my cup is (say) half as massive as my laptop, my cups counterpart in W is one fourth as massive as my laptops counterpart in W. Other than that, my cups mass relationships to all other bodies are exactly the same as its counterparts relationships to theirs. But this is not so with my laptop: the mass relationships it enters into differ systematically from those that its counterpart in W enters into by a factor of 2. So the comparativist may
situation is not improved by noting that comparativism is a contingent claim and allowing the material bodies in W to have intrinsic masses. For so long as the actual material bodies lack intrinsic masses, there remains no fact of the matter as to whether W is a world in which my laptop is twice as massive as it actually is.
17 The

11

say that it is in virtue of this asymmetry that W represents my laptop as being twice as massive as it actually is and everything else as being the same mass as they actually are. In effect, the comparativist just introduced a mass-counterpart relation in addition to the ordinary, Lewisian counterpart relation. Since my cup and its counterpart in W resemble one another with respect to their mass role, we call them mass-counterparts. And because my cups counterpart is also its mass-counterpart, W represents my cup as being the same mass as it actually is. Here the mass-counterpart relation is doing analogous work to Lewis counterpart relation: just as the latter is not identity but stands in for it when determining what a world represents de re, the mass-counterpart relation is not the same-mass-as relation but stands in for it when determining what a world represents about mass. Indeed, the comparativist can introduce a slew of mass-counterpart relations, one for each real number. My cups mass role resembles its counterparts mass role, so we call them mass1 -counterparts. My laptops mass role does not resemble its counterparts mass role in the same way, but since the mass ratios my laptop stands in are uniformly half those of its counterpart, their mass roles resemble each other modulo a factor of 2. As a result, the comparativist can call them mass2 -counterparts. More generally, she can say that x and y are massr -counterparts just in case xs mass role resembles ys mass role modulo a factor of r. And she may then propose the general principle that, relative to a counterpart relation and a set of mass-counterpart relations, W represents an actual object x as being r times as massive as it actually is just in case x has a counterpart in W that is also xs massr -counterpart. Relative to the mass-counterpart relations just described, this delivers the desired result that W represents my laptop as being twice as massive as it actually is. The ordinary counterpart relation is context sensitive, in the sense that the features of individuals relevant to determining whether they resemble one another, and therefore whether they are counterparts, are sensitive to the context in which the modal claim is made. Similarly, the comparativist can allow that those aspects of a bodys mass role relevant to determining what its massr -counterparts are depend on the conversational context. For example, a context in which mass relationships to my laptop are particularly salient might be one in which my laptops counterpart in W is also my laptops mass1 -counterpart (since the latter agrees with my laptop on all mass relationships to itself). Relative to this mass-counterpart relation, W represents my laptop as being the same mass as it actually is and everything else as being half as massive as they actually are! Although the discussion so far assumed Lewis own theory of de re

12

modality, the mass-counterpart theory just introduced is consistent with many other theories including ersatz ones. There is of course much more to say about it, but much of the ensuing discussion will resemble the literature on Lewis own counterpart theory so instead let me return to the second modal argument we left earlier and explain how the comparativist can use the mass-counterpart theory just introduced to make sense of the possibility of uniform doubling. The problem, remember, was that a uniformly doubled world agrees with our world on all mass relationships and so, according to comparativism and the necessitation principle outlined earlier, also agrees with our world on all facts about mass whatsoever and is therefore not a doubled world after all. But with mass-counterpart theory in hand, I believe the comparativist can accuse the objection of ignoring the distinction between worlds and possibilities. That is, she can respect the necessitation principle and concede that she can make no sense of a uniformly doubled world, but insist that she can nonetheless make perfectly good sense of the possibility of uniform doubling. To see how, note that it is a familiar fact that worlds and possibilities come apart in ordinary counterpart theory. To use Lewis example, I might have been either one of a pair of twins: I might have been the rst born, and I might have been the second born.18 Here we have one possible world with twins but two possibilities, one in which my counterpart is the rst born and the other in which my counterpart is the second born. Indeed, we have already seen an analogous distinction between worlds and possibilities in mass-counterpart theory: the world W discussed above represented two possibilities depending on which mass-counterpart relation we focused on, one in which my laptop is twice as massive as it is and one in which everything else is half as massive as they are. In his discussion of worlds and possibilities, Lewis also says that our counterparts need not always be in other worlds. When I consider the unhappy possibility of being my neighbor Fred, Fred himself (my worldmate) is acting as my counterpart and represents me as having all his properties.19 In this case the actual world, along with a certain counterpart relation, is representing a non-actual possibility for me. With mass-counterpart theory, the comparativist can model the possibility of uniform doubling analogously to how Lewis models the possibility of my being Fred, namely by using the actual world along with a suitable mass-counterpart relation. To see how, note that the possibility of uniform doubling is a possibility for all the material bodies taken together. If we let S be an ordered set of all those bodies, then accord18 See 19 See

Lewis [21], p. 231. Lewis [21], p. 232.

13

ing to Lewisian counterpart theory a world represents a possibility for those bodies by containing a counterpart of S. Well, surely S can be its own counterpart in normal contexts. Moreover, S can also be its own mass2 -counterpart in normal contexts. This is because Ss mass rolethe pattern of mass-relations entered into by the members of Sresembles its mass role perfectly modulo a factor of 2: put intuitively, the pattern of mass-relations are exactly as they would be were everything doubled in mass! And relative to these counterpart and mass-counterpart relations, our mass-counterpart theorysuitably generalized to apply to ordered setsimplies that the actual world itself represents the possibility of uniform doubling. So in this sense, uniform doublings of mass are possible even for the comparativist. This is not to deny that there is another sense in which uniform doublings of mass are impossible for the comparativist. What we just did, effectively, is distinguish between two modal notions. One is the notion that quanties over what I called possible worlds above and might be called strict possibility. This is the notion on which the above necessitation principle is true, that if a fact Y holds in virtue of a fact X then every possible world in which X obtains is also a possible world in which Y obtains. The comparativist should concede that she can make no sense of the strict possibility of uniform doublingsindeed as the third modal objection rightly points out the comparativist cannot even make sense of the strict possibility of my laptop being more massive than it actually is! But there is another modal notion that quanties over what I called possibilities above, where a possibility can be thought of as a triple of a possible world W, an assignment of objects to a counterpart in W, and an assignment of objects to a massr -counterpart in W.20 We might call this loose possibility. As argued above, the comparativist can then make perfectly good sense of the loose possibility of my laptops being more massive than it actually is, and indeed the loose possibility of uniform doublings. What the modal objections to comparativism need to establish, then, is not just that the envisaged situations (uniform doublings of mass, or just my laptop being more massive than it actually is) are possible in some generic sense but rather that they are strictly possible, for it is only this latter claim that the comparativist must deny. I have seen no attempt by absolutists to establish that they are strictly possible and I am skeptical that they can do so. At the very least, the burden is now on the absolutist to establish the required premise. For this reason I remain unmoved by
one is attached to the idea that it is possible that P iff there is a world in which P is true, one should feel free to call the possibilities possible worlds instead when using the loose notion. The nomenclature is not what matters here.
20 If

14

these modal objections to comparativism. There is a more general lesson here, namely that when engaging in modal reasoning (as these objections do) it is crucial to take great care to clarify the modal notion in use. For there may be many notions of possibility related to but distinct from strict possibility. If so, then when faced with the objection that she cannot make sense of this or that possibility the comparativist can always explore the response of saying that the envisaged situation is possible in one of these other senses even if it is impossible in the strict sense. Indeed I have no particular allegiance to the counterpart-theoretic response developed above and am open to resting my response on another alternative to strict possibility, if one can be developed. The counterpart-theoretic response is dialectically useful, though, since it nicely illustrates the existence of at least one of these other modal notions.21 It is worth mentioning a fourth kind of modal argument against comparativism, namely that the comparativist cannot make sense of a possible world containing just one massive body.22 In particular, one argument is that she cannot make sense of it having the determinable property of having mass, while a second argument is that she cannot make sense of it having any particular determinate mass (say, the mass of an electron). But in response to the rst argument, the comparativist can say that having the determinable mass consists in standing in determinate mass relationships, and then point out that the lone particle stands in such a relation with itself, namely the same-mass-as relation. And in response to the second objection, the comparativist can appeal to mass-counterpart theory and say that whether the particle counts as having the mass of an electron depends on the mass-counterpart relations allowed by the conversation in which the world is being discussed. I conclude, then, that objections to comparativism based on modal considerations are not convincing.
21 David Chalmers distinguishes two notions of metaphysical possibility that he calls prior and posterior possibility (he mentions the idea briey in footnote 5, p. 449 of his [7] and is developing the idea in other work). His notion of prior possibility is I think the same as the notion of strict possibility in use here. It is less clear to me how his notion of posterior possibility is related to the notion of loose possibility discussed in the text, though his notion of posterior possibility is certainly more general since the notion of loose possibility is (as dened above) restricted to the case of mass. 22 Thanks to Michaela McSweeney for helping me to appreciate the force of this argument. Similar arguments are given by Eddon [12].

15

The Objection from Semantics

One might instead try to refute comparativism on semantic grounds. To see how, recall Kripkes famous claim that we use the term meter with the stipulation that it is to refer to the length of the standard meter in Paris.23 The analogous view in the case of mass is that we use kilograms with the stipulation that it is to refer to the mass of that lump of platinumiridium alloy in Paris that serves as our standard of measurement, known as the International Prototype Kilogram (IPK). But the entity that the Kripkean theory takes the referent of kilogram to be, namely the mass of IPK, sounds suspiciously like an intrinsic property of IPK. After all, if the fundamental facts about mass were just facts about mass relationships, it is difcult to see what the mass of IPK could possibly be. So, the argument goes, if comparativism were true then kilogram would fail to refer and sentences like My laptop is 2 kilograms would fail to be true. Of course, the comparativist might bite the bullet and concede that kilogram sentences are not true. According to this error theory response, the term kilogram is similar to phlogiston: both were used with the stipulation that they are to refer to whatever entity satises some description, but since nothing answers to the description they both fail to refer. The main difference between the two cases is that there is a pragmatic reason to continue using the term kilogram that is lacking in case of phlogiston. For as long as our use of kilograms is governed by the inference rule a is r kilograms b is s kilograms Therefore, a is r/s times as massive as b we can use kilogram as a convenient way of storing and communicating information about mass ratios even if sentences containing it are not true.24 If this error theory sounds radical and unwarranted, compare it to the case of absolute simultaneity. If semantic investigation revealed that the truth of our ordinary talk requires there to be such a thing as absolute simultaneity but it subsequently turned out for reasons of physics and metaphysics that there is no such thing, we would have no qualms concluding that our ordinary talk is in error. Similarly, if the above semantic argument showed that the truth of kilogram sentences requires the truth of absolutism but it subsequently turned out for reasons of physics and
23 See 24 I

Kripke [19]. discuss the role of this form of inference in more detail in Dasgupta [10].

16

metaphysics that absolutism is false, we should have no qualms accepting the resulting error theory. But while this error theory is defendable, there is no need for the comparativist to adopt it since the semantic objection fails to establish that the truth of kilogram sentences requires the truth of absolutism in the rst place. In fact, it fails for two reasons. First, the Kripkean theory of reference-xing it presupposes is false. To see this, imagine reading in the Times that the French have been subjecting us to an illusion that makes IPK appear twice as massive as it actually is. Imagine that the article explains that the illusion has been systematic, so that whenever we used IPK to calibrate our measuring instruments, the calibration succeeded even though we were misled about the properties of the lump. So, if we were to put IPK on one of the many calibrated measuring instruments around the world, it would read 500 grams rather than 1 kilgoram. How would we report this discovery? Intuitively, by saying that we discovered the surprising fact that IPK is 500 grams! But the Kripkean theory predicts otherwise. Since the theory is that 1 kilogram is stipulated to refer to the mass of IPK whatever that mass is, it implies that the article should instead be reported as telling us that while the standard object is (of course) still 1 kg, all other material bodies are actually half the mass in kilograms that we previously thought they were. And this, I claim, is not how we would intuitively report it. But there is a second and perhaps more decisive reason why the semantic objection fails to establish that the truth of kilogram sentences requires absolutism. Even if we granted the Kripkean theory of reference, the argument is supposed to be that the entity to which kilogram is stipulated to refer, namely the mass of IPK, is not identical to IPKs mass relationships, and it concludes that the comparativist must say that there is no such thing. But this last step is a non sequitur. All the comparativist claims is that the most fundamental facts about mass are facts about mass relationships; it is perfectly consistent with this that there is such a thing as the mass of IPK which is not identical to any mass relationships, so long as any fact of the matter concerning it holds in virtue of facts about IPKs mass relationships. So the comparativist is free to agree that there is such a thing as the mass of IPK to which the term kilogram refers after all. The non sequitur exhibited by the semantic objection is vividly exemplied in the following case. Consider a physicalist who claims that all facts hold in virtue of facts concerning physical entities, and imagine an objector who says The term stock market refers to the stock market, but the stock market is not a physical entity; therefore your physicalism is false. In response, our physicalist will surely point out that the argument

17

misses its mark entirely: her view was never that everything is a physical entity but rather that all facts about the world hold in virtue of facts concerning physical entities. The comparativist can say exactly the same about the semantic objection to comparativism. In sum, I do not believe that there is semantic evidence that the truth of kilogram sentences requires absolutism to be true. But even if there were, I believe that the appropriate response for the comparativist would be to adopt an error theory about kilograms.

The Objection from Kilograms

Since comparativism holds that all facts about mass hold in virtue of mass relationships, one might naturally try to refute the view by nding a counterexample, i.e. a fact about mass that does not hold in virtue of mass relationships. For example, consider the fact that my laptop is 2 kgs. If one can argue that there are no mass relationships in virtue of which this obtains, one would naturally take oneself to have refuted comparativism. Absolutists should nd this strategy promising, for I believe that there are good arguments to the effect that my laptops being 2 kgs does not hold in virtue of any mass relationships. Unfortunately there is no room to discuss these arguments in full detail, but let me say something to motivate the idea.25 First, note than an absolutist has no problem accounting for my laptops being 2 kgs: she can say that it is either identical to, or else holds in virtue of, my laptops having a certain intrinsic mass. If absolutism were true, this would be an extremely plausible view. For if material bodies really did have intrinsic masses, it would be natural to think that terms of the form r kilograms would refer to those properties (even if the Kripkean view about what xes the referents of the terms is incorrect). If so, then it is almost irresistible to say that my laptops being 2 kgs is either identical to, or else holds in virtue of, its having a certain intrinsic mass; namely the one that is the referent of 2 kgs. But the fact that this account of my laptops being 2 kgs is so natural and satisfying shows that, at least intuitively, the mass relationships that it enters into are entirely irrelevant when it comes to explaining what makes it 2 kgs. My laptop stands in all sorts of mass relationships to standard objects in Paris and measuring instruments in Paruguay and electrons on Pluto, but the fact that the absolutists explanation is so satisfying shows that, intuitively, all these relationships are irrelevant to an explanation of its being 2 kgs. Therefore, the argument goes, whichever mass relationships the comparativist picks in order to explain its being 2 kgs, she will
25 These

arguments are developed in more detail in Dasgupta [10].

18

violate our intuitions as to what is relevant to explaining that fact. Of course, the comparativist might concede this and reply that revising our opinions about what is explanatorily relevant is a natural consequence of theoretical inquiry. To an extent, this reply is well taken. But all hands should agree that this would be a signicant revision of pre-theoretic belief and therefore counts as at least a point against her view. As I said, there is no space to develop this kind of argument in detail. But let us give the absolutist the benet of the doubt and suppose that there are no mass relationships in virtue of which my laptop is 2 kgs. Where does this leave the comparativist? Of course, one option would be to adopt the error theory described earlier, according to which there are no facts about mass-in-kilograms in the rst place. But I believe that there is a better option. The key is to recognize that the in virtue of relation is irreducibly plural, in the sense that a plurality of facts Y can sometimes hold in virtue of another plurality of facts X even though no Y when taken on its own holds in virtue of anything. Given this pluralistic conception of the in virtue of relation, the comparativist may take the set K of all kilogram facts and the set R of all facts about mass relationships, and propose that the members of K (plurally) hold in virtue of the members of R even though no kilogram fact taken on its own holds in virtue of anything. This view neatly sidesteps the problem of relevance discussed above, for R does not contain irrelevant information when it comes to explaining the members of K. To be sure, R contains irrelevant information when explaining my laptops being 2 kgs, such as information about its mass relationships to electrons on Pluto, but since K contains facts about how massive those electrons are in kilograms the relationships between them seem perfectly relevant when explaining Ks members all together. By adopting this position, the comparativist can then agree that there is a fact of my laptops being 2 kgs (contra error theory), concede that there are no mass relationships in virtue of which it obtains, and yet nonetheless insist that this is perfectly consistent with comparativism since it remains the case that all the facts about mass in kilograms when taken together as a plurality hold in virtue of the underlying mass relationships. There are many virtues of this view. One is that the comparativist can respect the intuition described earlier that facts about my laptops mass relationships are not part of what explains my its being 2 kgs. They are not part of the explanation of this fact, on this view, because the fact on its own has no explanation in the rst place! Another virtue is that it neatly explains why absolutism is initially the more intuitive and attractive view. For as we have seen, we have a strong intuition that my laptops being 2 kgs does not hold in virtue of its mass relationships to other bodies. Ac-

19

cording to the current approach, the absolutists mistake is just to take this to imply that its being 2 kgs must be explained in terms of its intrinsic nature, when instead the correct conclusion is that we can only explain facts about mass in kilograms when they are taken all together as a plurality. The absolutists mistake is therefore understandable, but a mistake nonetheless. Now I have not argued that this pluralist explanation of kilogram facts is satisfactory, and unfortunately there is no space to do so here.26 But at the very least, it is clear that if the in virtue of relation is irreducibly plural then refuting comparativism is signicantly more difcult than one might have thought. For it is then not enough for the absolutist to argue that my laptops being 2 kgs fails to hold in virtue of its mass relationships. In addition, she would need to argue that the plurality of kilogram facts taken together do not hold in virtue of the totality of mass relationships. Until she shows this, comparativism remains a live option.

The Objection from Humeanism

Humean Supervenience (HS), says Lewis, is the view that all there is to the world is a vast mosaic of local matters of particular fact, just one little thing and then another. . . We have geometry: a system of external relations of spatio-temporal distances between points. . . And at those points we have local qualities: perfectly natural intrinsic properties which need nothing bigger than a point at which to be instantiated. . . And that is all. There is no difference without a difference in the arrangement of qualities. All else supervenes on that.27 So stated, HS is inconsistent with comparativism since it asserts that the fundamental physical quantities like mass are intrinsic and are instantiated at single points of spacetime. But HS is supported in the literature by wide range of arguments. If it is inconsistent with comparativism, the argument would be, so much the worse for the latter. However, most arguments for HS are perfectly consistent with comparativism. To see this, note that HS is the conjunction of two theses: one stating that everything supervenes on the categorical nature of the physical world, and a second describing what the categorical nature of the physical world is like. The above quote focuses on the second and says that
26 I 27 Lewis

motivate and defend this pluralist explanation at some length in Dasgupta [10]. [22], p. ix.

20

the categorical nature of the world consists in the distribution of intrinsic properties across spacetime. And that second thesis is indeed inconsistent with comparativism. But much of the literature on Humean Supervenience focuses on the rst thesis, the view that everything elseincluding chances, causes, counterfactuals, minds, morals, etcsupervenes on the worlds categorical nature. And a brief glance at that literature reveals that none of the arguments depend on whether those underlying categorical facts consist in the instantiation of intrinsic properties (as Lewis says) or in the instantiation of comparative relations (as the comparativist says). Even if one is moved by those arguments, one may still adopt comparativism. To be sure, the second thesis does play a role in Lewis metaphysics. For example, he famously analyzes de re modals in terms of counterparts, and he says that objects are counterparts insofar as they resemble each other, and he says that resemblance is ultimately a matter of sharing intrinsic properties. But one can easily restate his view in terms friendly to the comparativist by allowing resemblance to ultimately be a matter of participating in the same pattern of relations instead. Indeed, Lewis only thought that resemblance was ultimately a matter of sharing intrinsic properties because he thought that it was a matter of sharing perfectly natural features, and the second thesis of HS states that those perfectly natural features are all intrinsic (save for geometric relations). Thus, if one gives up that second thesis in favor of comparativism and allows that some perfectly natural features are relations, it will follow from the rest of his system that resemblance is sometimes a matter of participating in the same pattern of relations after all. As a result, a large chunk of Lewis system remains essentially unchanged even if we endorse comparativism. This is not to say that all of Lewis views are easily recast in comparativist terms. Still, once one sees how many of them can be, the above quote stating that the categorical world consists in the distribution of intrinsic properties sounds less like an essential part of his view and more like a convenient working assumption.

Objections from Physics

The objections to comparativism considered so far have been broadly apriori, but is there empirical evidence against the view? If we could see the intrinsic mass had by a given material body or detect it with the help of mechanical devices, that would presumably count as empirical evidence for absolutism. But I will argue in Section 8 that if material bodies had the intrinsic masses posited by the absolutist, those intrinsic masses would 21

be invisible to the naked eye and undetectable by any physically possible device. Still, if the absolutist could show that intrinsic mass is indispensible to our best conrmed scientic theories, one might then think that empirical evidence conrming those theories would thereby count as empirical evidence that each material body has an intrinsic mass even if we cannot tell which particular one it is. To see how this idea might be developed, consider one simple law governing mass, f =ma, and let us pretend for simplicity that our best conrmed physical theory states that it is the only law governing the motions of material bodies. Now, consider a world W exactly like ours with the one exception that everythings mass is double what it actually is. One might argue that if the equation f =ma actually obtains then it does not obtain in W since doubling everythings mass while leaving their forces and accelerations unchanged would break the equality.28 Since W is just like ours in all mass relational respects, the argument would be that the truth of f =ma depends not just on the mass relations between things but also on which intrinsic masses they have. Therefore, empirical evidence conrming f =ma is ipso facto empirical evidence conrming absolutism. But the argument does not convince, for it depends on a controversial interpretation of the equation f =ma. Taken at face-value, it states a mathematical relationship between the numbers and vectors that represent force, mass and acceleration in a particular scale. But what does it state about the quantities themselves? There are at least two interpretations, one on which it states something about absolute quantities and another on which it states something about comparative quantities. On the rst interpretation it states that the absolute masses, accelerations and forces all line up in a specic way. For example, part of the content of the equation on this interpretation is that anything with the determinate intrinsic mass M which is accelerating at a determinate rate A will have a particular determinate force F acting on it, and so on for other determinate masses,
is where it helps to ignore any other laws specifying the force acting on each particle and pretend that f =ma is the only law governing our world. For we thereby sidestep the complication that with the identication of inertial mass and gravitational mass, mass plays a unique role in classical mechanics: not only is it a brake on acceleration as described in f =ma, it is also a determiner of the gravitational force between things as described in the inverse-square gravitational force law. The fact that it plays this dual role might tempt one to think that doubling everythings mass would preserve the truth of the classical mechanical laws, since the increase in gravitational forces would be counter-balanced by the increased brake effect experienced by each body. But even if this line of reasoning were sound, it would not generalize to other quantities. I consider f =ma in isolation from whatever force laws it might couple with precisely because we are looking for general considerations.
28 This

22

accelerations and forces. And, in the other direction, part of the content of the equation will be that anything with the particular determinate force F acting on it will either have the determinate mass M and be accelerating at the determinate rate A, or else will have the determinate mass 2 M and be accelerating at the determinate rate A/2 (where 2 M is the intrinsic mass that is twice M and A/2 is the determinate rate of acceleration that is half A), and so on for other pairs of determinate masses and accelerations. More generally, we might express this interpretation as follows: (L1) For any material thing x, (a) For any reals r1 and r2 , if x has mass r1 M and acceleration r2 A, then x has force r1 r2 F acting on it. (b) For any real r3 , if x has force r3 F acting on it, then there are reals r4 and r5 whos product is r3 , such that x has mass r4 M and acceleration r5 A. By contrast, the second interpretation of the equation takes it to state a connection between mass relationships, force relationships and acceleration relationships. For example, part of the content of the equation on this interpretation is that if a body x is twice as massive as another body y and an equal force is applied to both, then y will accelerate at twice the rate as x. More generally: (L2) For any material things x and y, (a) For any reals r1 and r2 , if x is r1 times as massive as y and is accelerating r2 times the rate of y, then x has r1 r2 times as much force acting on it than y. (b) For any real r3 , if x has r3 times as much force acting on it than y, then there are reals r4 and r5 such that r4 r5 = r3 , and such that x is r4 times as massive as y and is accelerating r5 times the rate of y.29 The argument under consideration assumed that the actual physical laws do not obtain in the world W in which all masses are doubled. We can now see that whether this is true depends on whether the fundamental law governing the quantities is (L1) or (L2). The assumption is true if the
is a very simplistic example of a comparativist rendering of a physical law. For a more realistic attempt to render physical theories in comparativist friendly terms, see Field [14]. I will discuss (L2) here because it is particularly easy to grasp but the discussion is intended to generalize to more realistic comparativist friendly laws such as those described by Field.
29 This

23

fundamental law is (L1), since in W the absolute masses, forces and accelerations line up differently than they actually do. But the assumption is false if the fundamental law is (L2), for (L2) only talks of mass relationships and those are the same in W as they are in the actual world. Indeed, it is clear that if the fundamental law is (L2), then whether it obtains does not depend on material bodies having the intrinsic masses posited by the absolutist, contra the argument under consideration. So in order to run the indispensability argument the absolutist must argue that the empirical evidence conrming f =ma is evidence that conrms (L1) and disconrms (L2). Only then would the intrinsic masses posited by the absolutist be indispensable to what is conrmed by the empirical evidence. But what evidence would favor (L1) over (L2)? The difference between the laws is this: (L1) implies that if the state of the world at present differed only in that everythings mass were double what it actually is, things would proceed to accelerate at half their actual rate.30 (L2) does not imply this because the mass ratios would be exactly the same in the doubled state and that is all that the law makes reference to. So the two laws issue these different predictions, but how could we test which prediction is correct? The obvious idea is to construct two isolated laboratories that are exactly alike at an initial time except for the fact that one is a doubled-mass version of the other. One might think that if the bodies in the doubled-mass laboratory proceed to accelerate at half the rate as the bodies in the other, this would conrm (L1) and disconrm (L2). But on further reection this experimental outcome is predicted by (L2) and would therefore not disconrm it. After all, the bodies in the doubled-mass laboratory are twice as massive as their counterparts in the other and are subjected to the same forces as those counterparts, and the experimental outcome is that they accelerate at half the rate as their counterparts. But this is exactly what (L2) would predict! The trouble is that the two laws make different predictions about what would happen if the entire world were doubled in mass, but when attempting to test which prediction is correct we can do no better than to compare different parts of the world (our two laboratories) and the laws make exactly the same prediction about what would then occur. How else might we obtain evidence that conrms (L1) and disconrms (L2)? Suppose we could see or detect which particular intrinsic mass, force and acceleration each material body has, and suppose it turned out that they were lined up in the way stated by (L1).31 Would this conrm (L1)
I am bracketing the effect that doubling everythings mass might have on the forces acting on things. See footnote 28. 31 I will argue in Section 8 that we could never see or detect such a thing. But I do not want to rest my argument here on the results of that section. So here I am supposing, per
30 Again,

24

and disconrm (L2)? No, because that observation is consistent with the hypothesis that material bodies have absolute quantities but are governed by (L2). For if that were the case then the absolute quantities are bound to line up in one way or another, it would just be a matter of accident (or perhaps, more specically, a matter of initial conditions) that they line up as they do whereas according to (L1) they would line up like that as a matter of law. It therefore appears that there is no possible evidence that would conrm (L1) and disconrm (L2). Under the pretense that f =ma is the only law of our best conrmed physics, then, the intrinsic masses posited by the absolutist are not indispensable to our best physics after all.32 The pretense is of course false, but it seems reasonable to hope that similar reasoning will apply in the case of better conrmed physical theories. But I leave the question of whether this is so for another time.33

8
8.1

In Favor of Comparativism
The Occamist Argument

Having surveyed a number of objections to comparativism, I nd none convincing. But is there any positive reason to be a comparativist? I believe there is. I will argue that if material bodies really did have the intrinsic masses posited by the absolutist, those intrinsic masses would be undetectable. Our Occamist principle says that it is a mark against a theory if it posits undetectable structure, so this is reason to prefer comparativism. This epistemic argument is not the only possible argument for comparativism. Some might favor comparativism purely on grounds of onimpossible, that we can make such observations just for the sake of argument. 32 Baker [4] argues that the idea that the fundamental laws of our world are comparativist-friendly laws like (L2) (or, more realistically, the kind of laws that Field expresses in his [14]) has serious problems. There is no space here to discuss his arguments in the detail they deserve but I hope to do so in further work. Very briey, the problems he describes concern modal properties of the laws such as whether they are deterministic. Now I argued in section 3 that when discussing comparativism it is important to think carefully about how modal claims about quantities are to be evaluated. There I argued in favor of a mass-counterpart interpretation of modal claims. I believe that if the comparativist endorses this mass-counterpart theory she has all the resources needed to respond to Bakers objections. But I hope to defend this idea in future work. 33 Field [14] has famously made a good start at expressing a portion of physics in purely comparativist terms. However, the current point does not depend on the success of Fields project, which was to express physics without reference to numbers, sets or other abstracta. For example, (L2) serves the comparativists purposes but it freely quanties over real numbers.

25

tological parsimony, the idea being that while everyone recognizes mass relations it is only the absolutist who goes further and posits extra intrinsic masses that are not explicable in terms of those mass relationships. Others might favor comparativism because it leads to a simpler axiomatization of a theory of mass.34 But my epistemic argument is particularly important. For if we could see or otherwise detect which intrinsic mass a given body hasif the intrinsic masses posited by the absolutist were really part of our datathen a comparativist theory that dispensed with those intrinsic masses would not be empirically adequate. And in that case comparativism should obviously be rejected outright purely on grounds of empirical inadequacy, and the proposed benets of ontological parsimony and theoretical simplicity would then be neither here nor there. So a hidden assumption in these other arguments for comparativism must be that the intrinsic masses posited by the absolutist are not detectable after all, which is precisely what my epistemic argument tries to establish. Much depends on what I mean by undetectable. If I used the term to include anything that we cannot see with the naked eye, our Occamist principle would recommend that we become radical scientic anti-realists and dispense with so-called theoretical entities such as electrons. But that is not how I use the term here. Instead, something is undetectable in my sense of the term if, roughly speaking, it follows from the structure of the laws of motion governing our world that it is physically impossible for it to have an impact upon our senses. Electrons are therefore perfectly well detectable in this sense because there are physically possible processes, such as those that occur in particle accelerators, by which the presence of an electron can be made to have an impact on our senses via its impact on (say) the movement of a dial or an image produced on a computer screen. In contrast, features like absolute velocity and absolute simultaneity are undetectable in my sense: even if they were real, it turns out that the laws of motion governing our world are set up in such a way as to guarantee that it would be impossible for them to ever have an impact on our senses. That is why they are considered to be redundant or superuous in modern physics, and most physicists and metaphysicians therefore believe on Occamist grounds that those features are not real after all. I will argue here that the same goes for intrinsic mass. The Occamist principle I use here is not vericationist: there is no claim that talk of these undetectable features is meaningless. It is just a principle of theory choice, a principle stating that (all else being equal) we should favor those theories without undetectable features. Nor does our
is an idea that Field emphasizes in his [15]. Thanks to a referee for encouraging me to think about this consideration.
34 This

26

Occamist principle say that we should always dispense with undetectable features. It just says that undetectable features are undesirable, so that all else being equalor at least near enough equalwe should prefer theories like comparativism that dispense with them. But in the previous sections I have been arguing that there are no overwhelming reasons to reject comparativism and therefore that all is indeed near enough equal after all. Our crucial premise, then, is that if material bodies really possessed the kind of intrinsic mass posited by the absolutist, those intrinsic masses would be undetectable in our sense of the term. How should we argue for this? There is a reasonably well known argument for the analogous claim in the case of absolute velocity, so let us rehearse it before applying it to the case of intrinsic mass.

8.2

The Case of Velocity

What is absolute velocity? We often talk of a material bodys velocity relative to another body: a car might have a velocity of 65 mph in a particular direction relative to the highway and 10 mph in the same direction relative to the train traveling alongside it. But how fast is it really going, independent of any material reference point? If there is an answer to this question, that is a statement of its absolute velocity. Now if there were such a thing as absolute velocity, why would it be undetectable? At least naively, one might think that the speedometer found in an ordinary car is a device that detects the vehicles absolute velocity. But we can argue that such a device at best measure relative velocity and that absolute velocity is undetectable after all.35 For in order to detect absolute velocity, there would need to be some physically possible process that, when initiated at t0 to measure the absolute velocity of a given body, will generate a readingan image on a computer screen, say, or the position of a needlethat indicates what that bodys velocity was at t0 . Moreover, the outcome that would be produced if the body were traveling at one velocity must be discernibly different from the outcome that would be produced if it had a different velocity, on pain of our not being able to tell what velocity a given outcome indicates.36 So, if we simargument that follows has received perhaps its clearest written expression in Roberts [26]. I heard similar arguments orally in seminars given by Tim Maudlin at Rutgers and David Albert at Columbia. However, all these theorists run the argument in importantly different ways. My presentation here is similar to my presentation in Dasgupta [8]. 36 At least, that is the ideal: in practice, we do not mind if the outcomes that would be produced by velocities differing only by some tiny amount are indiscernible. More accurately, then, what we require is that the outcomes would be discernible when the
35 This

27

ply wanted to measure whether a given body was in a state of absolute rest or absolute motion, the process would need to produce one outcome if the body was at rest at t0 for example an inscription of At restand a discernibly different outcome if the body was moving t0 an inscription of Moving, say. Finally, since the process is a physical process, the outcome produced will depend on the physical laws governing our world. Putting this all together, we can therefore say that absolute velocity is detectable only if there is a physically possible device which at a given time t0 has two properties: rst that, according to the laws, it will display At rest on a computer screen at a later time t1 iff it was presented with a body at rest at t0 ; and second that (according to the laws) it will display Moving on a computer screen at t1 iff it was presented with a body that was moving at t0 .37 But according to most of our best conrmed physical theories, it is physically impossible for a device to have both properties. For suppose I take a device with the rst property and present it with a body at rest at t0 , and it therefore displays At Rest at t1 . We can show that it does not have the second property by considering a world W just like ours with the one exception that at all times the absolute velocity of all bodies is ve mph greater in a certain specied direction. Now, W is a world in which the device is presented with a moving body at t0 , and yetsince the relative positions of all bodies at all times are (by construction) the same in W as they actually arethe device still displays At rest at t0 . But it turns out that according to our best physics, the laws of motion governing W are the same as those governing ours. Therefore, the behavior of the device in W represents how it behaves according to our laws of motion; hence it does not have the second property listed above. QED. To be clear, the argument here does not assume that absolute velocity can only be detected by a device with these two properties. The assumption is rather that if absolute velocity is detectable then it is detectable with a device with these two properties (even if it is also detectable in other ways too). The argument then is that it is not detectable with a device of this kind, therefore it is not detectable. I will discuss the assumption further in section 8.5.38 This is not to say that speedometers in cars are useless, for the argument here is consistent with the thesis that they detect the cars velocity
velocities differ by more than some amount x, in which case we say that the process measures absolute velocity up to an accuracy of x. 37 We use biconditionals here because we not only want each initial velocity to issue in a readable outcome, we also want each outcome to be uniquely associated with that initial velocity so that we know what the outcome indicates. 38 Thanks to a referee for urging me to clarify the assumption used here.

28

relative to a given body such as the road. All the argument shows is that they do not detect the cars absolute velocity. And of course if absolute velocity is undetectable our Occamist principle then recommends (all else being equal) endorsing a theory of spacetime according to which there is no such thing as absolute velocity.39

8.3

Undetectable by Us

Returning to the case of intrinsic mass, let us suppose for reductio that material bodies have the intrinsic masses posited by the absolutist. For similar reasons as just discussed in the case of velocity, I claim that which particular intrinsic mass each body has would be undetectable.40 The claim may initially sound implausible. After all, my laptop and my cup feel different in mass when I pick them up, so am I not thereby detecting the intrinsic mass of each item? Similarly, one might naturally think that an ordinary bathroom scale is a device that allows us to detect the intrinsic mass of the object placed on it by displaying its mass in the position of a needle. But it turns out that this is a mistake: just as a speedometers at best allow us to detect the velocity of a car relative to
may recognize this as one line of reasoning that can be used to extract metaphysical conclusions (in this case about the structure of spacetime) from facts about the symmetry or invariance groups of the fundamental laws of motion (in this case the relevant symmetry group being uniform velocity boosts). The general idea of using symmetries to motivate various metaphysical views is not new. See Earman [11] chapter 3, and North [25] for discussions of this idea with respect to the metaphysics of spacetime. It is also used (explicitly and implicitly) in contemporary discussions of diffeomorphism invariance in General Relativity and permutation invariance in Quantum Mechanics to motivate various structuralist metaphysical conclusions (see Ladyman [20] for a discussion of both these issues, and see Dasgupta [9] for a discussion of the former). Where I may differ from others is my belief that the best such arguments go by way of intermediary conclusions about what is undetectable. That is, the best arguments in my view rst use facts about symmetries to draw an intermediary conclusion to the effect that some putative feature is undetectable, and only then use the Occamist principle to draw the metaphysical conclusion that those undetectable features are not real. The question of whether symmetry arguments must take this epistemic form is vexed and not one I can discuss here, though I hope to do so in future work. In any event there is no need to settle the matter here for we can evaluate the argument in the text on its own merits regardless of whether there are other ways of drawing metaphysical conclusions from symmetries. Thanks to a referee for encouraging me to point out the connections between the discussion in the text and these other debates. 40 I should stress that the argument concerns the kind of intrinsic mass as posited by the absolutist; that is, the kind of intrinsic mass the possession of which is not explicable in terms of mass relationships. As I argued in section 5 the comparativist can agree that there are facts about my laptops mass in kilograms such as its being 2 kgs, and perhaps this is a fact about its intrinsic mass in some sense of the term intrinsic. But she insists that these facts hold in virtue of facts about mass relationships so my argument is not directed at these kinds of intrinsic masses.
39 Readers

29

the highway, so too picking things up or putting them on bathroom scales at best allows us to detect the mass relationships between things and not their intrinsic masses. The argument is a little more complex than in the case of velocity, but let us start by running an analogous argument and modify it as required. For convenience, let us suppose that the terms 1 kg and 2 kgs label particular intrinsic masses. Following the above discussion, we can then say that intrinsic mass is detectable only if there is a physically possible device which at a given time t0 has two properties: rst, that (according to the laws) it will display 1 kg on a computer screen at a later time t1 iff it was presented with a 1 kg object at t0 ; and second, that (according to the laws) it will display 2 kgs on a computer screen at t1 iff it was presented with a 2 kg object at t0 . We now argue that it is physically impossible for a device to have both properties. To this end, suppose I take a device with the rst property and present it with a 1 kg object at t0 and it therefore displays 1 kg on a screen at t1 . We show that it does not have the second property by considering a possible world W just like ours with the one exception that everythings mass is double what it actually is; that is, a world in which our device is presented with a 2 kg object at t0 but in which it nonetheless displays 1 kg at t0 . And we now make a key assumption, namely that the laws of motion governing W are the same as those governing ours. It then follows that the behavior of the device in W represents how it behaves according to our laws of motion; hence it does not have the second property listed above. QED. But is this key assumption correct? As before, let us suppose for simplicity that our best conrmed physical theory states that there is only one law of motion governing our world: f =ma. In the last section, we saw that this equation might express one of two things: a law governing absolute quantities like (L1), or a law governing comparative quantities like (L2). We also saw that if the actual law turns out to be of the second kind then it would obtain in W too, while if the actual law turns out to be of the former kind then it would not obtain in W (in the latter case W does not represent how our device behaves according our laws of motion and it remains open that it has the two properties after all). So the assumption that the law of W are the same as the actual laws amounts to the assumption that the laws are like (L2) and govern comparative quantities. Can the comparativist assume this? The assumption does not presuppose the falsity of absolutism, since the hypothesis that the laws govern comparative quantities is consistent with the idea that material bodies also have intrinsic quantities in virtue of which those comparative quantities obtain. But is it reasonable for the comparativist to believe that the as-

30

sumption is true? Recall that in the last section we argued that there is no possible evidence that would conrm the hypothesis that law governs absolute quantities but disconrm the hypothesis that it governs comparative quantities. Importantly, the argument did not assume that intrinsic mass is undetectable, so we can appeal to the result of that argument without begging questions. So, if the evidence does not settle the matter either way, what is it reasonable to believe? This is a delicate issue in epistemology concerning what it is reasonable to believe given certain evidence. An extremely permissive view of rationality would say that if the evidence does not settle whether p or notp, it is reasonable to believe either. On this view, the comparativist may reasonably believe that the actual law governs comparative quantities, in which case she can run the above argument as written. But one might worry that the argument would then be dialectically weak, since on this permissive view it would also be reasonable for the absolutist to believe that the laws govern absolute quantities in which case she will remain unmoved by the argument. Moreover, other views in epistemology would insist that if the evidence equally supports two views, we should remain agnostic between the two rather than believe either. Luckily there is no need to settle the issue here. For what our central argument above shows is that whether or not a device with two properties listed above is physically possible depends on which kind of law governs our world. More specically, the argument shows that a given device has those two properties only if our laws govern absolute rather than comparative quantities. Therefore, my evidence that a given device has those two properties can be no higher than my evidence that the laws govern absolute quantities rather than comparative quantities. But the conclusion of the last section was that there is no possible evidence that the actual laws govern absolute quantities and not comparative quantities. Therefore, regardless of the delicate epistemic question of what it is reasonable to believe given certain evidence, all hands should agree that I can have no evidence that a given device has the two properties listed above. And this conclusion is strong enough for our purposes. For even if a given device does in fact have the two properties and gives a reading of 1kg when I present it with a material body, the fact remains that if I have no evidence that the device has those two properties then the reading gives me no evidence as to what the bodys intrinsic mass is. So our conclusion is that even if absolute mass is in some sense detectable by the device it is not detectable by us, which is the point we were trying to establish all along. The argument here trades on the familiar point that the outcome of a measurement depends on three things: the value of the feature being mea-

31

sured, the initial state of the device being used to measure the feature, and the laws that govern the interaction between the feature and the device. If we know enough about the last two factors, we can use the outcome of the measurement to infer what the value of the feature was. But if we do not know enough about what laws govern our world, then we may not be in a position to make the inference. The argument here is that if we lack evidence as to whether our laws govern absolute or comparative quantities, then we have exactly this kind of handicap when trying to detect which particular intrinsic mass a given body has.

8.4

A Second Argument

Interestingly, one can argue for the same conclusion without relying on the premise that there is no evidence that the actual law governs absolute quantities rather than comparative quantities. Suppose, perhaps per impossible, that we were to acquire evidence that the actual law governs absolute quantities and that devices with the two properties listed above are therefore physically possible. One can argue that there would still be some evidential uncertainty as to the details of what that law is in such a way that we could never acquire evidence that a given device has the two properties. To see this, suppose we are given a device and are asked to determine whether it has the two properties listed above. One of those properties was that it will behave in a certain way according to the laws governing it, namely it will register 1 kg on a computer screen at t1 iff it is presented with a 1 kg object at t0 . Now suppose that, in fact, the device in front of us will behave like that according to (L1). And suppose that (L1) is the actual law, so the device really does have the property. Now (L1) does not obtain in W but something closely related does, namely the result of replacing all occurrences of M with 2 M (where this latter term refers to the mass that is double that of M). Call this law (L1*). Then what W shows is that if everythings mass were uniformly doubled then, according to (L1*), the device will not behave in the same way. Rather, it will register 1 kg on a computer screen at a later time t1 iff it is presented with a 2 kg object at t0 , rather than with a 1 kg object. So whether or not the device in front of us has the required property depends on which of two hypotheses is true: hypothesis H, which attributes to the device its actual mass and states that (L1) obtains, or H*, which attributes to the device the mass it has in W and states that (L1*) obtains. But one can argue that no evidence could possibly favor either hypothesis over the other. Let us assume that W is indiscernible from the actual world, in the sense that everything in W looks and feels and smells and 32

tastes just like it does in the actual world (I will discuss assumption in a moment). Then it follows that no empirical evidence would falsify either hypothesis. And (L1) and (L1*) are of exactly the same form: both are equally simple, elegant, unifying and explanatory. So neither hypothesis trumps the other on any theoretical virtue we use to choose between hypotheses that agree on the empirical data. But the device in front of us has the required property only if hypothesis H is true. Therefore, since there can be no evidence that would favor H over H*, there can be no evidence that the device has the required property. And as we just saw when discussing the second strategy, this means that even if a given device does in fact have that property, the readings it delivers will give us no evidence as to what a bodys intrinsic mass is and therefore intrinsic mass will remain undetectable by us.41 Admittedly, this line of argument is rather more involved than the one I initially outlined. But still, some comparativists might be interested in developing it.

8.5

Indiscernibility

We have assumed if intrinsic mass is detectable, it is detectable by processes that indicate mass with inscriptions of 1 kg and 2 kgs.42 This is not to deny that it might be detectable in other ways too; our assumption was just that if it is detectable at all, it must at least be detectable in this way. Our strategy was then to argue that intrinsic mass is not detectable in this way; hence it is not detectable at all. How plausible is the assumption? Borrowing an idea of Alberts, the comparativist might argue that it should be extremely plausible. For given anything D that counts as an intrinsic mass detector, it is presumably possible for us to decide in advance to record the result of the measurement produced by D by writing 1 kg or 2 kgs on a piece of paper depending on what the result is. If so, then the result of coupling D with our decision to record the outcome of D in that way constitutes a composite device that
41 This argument appeals to the assumption that the actual world and W are indiscernible, and one might worry that this begs the question by assuming that absolute mass is undetectable. But the worry is misplaced. At most, the indiscernibility of the two worlds just shows that absolute mass is invisible to the naked eye, but it remains open that it is possible to build measuring devices that might reveal them to us. The fact that the worlds have such similar laws is then being used to show that no such device is possible. 42 In full, remember, we assumed that intrinsic mass is detectable only if there is a physically possible device which at a given time t0 has two properties: rst, that (according to the laws) it will display 1 kg on a computer screen at a later time t1 iff it was presented with a 1 kg object at t0 ; and second, that (according to the laws) it will display 2 kgs on a computer screen at t1 iff it was presented with a 2 kg object at t0 .

33

detects mass with the inscriptions 1 kg and 2 kg, and our assumption is vindicated.43 Still, one might try to resist the assumption. For one might try to argue that there are physically possible process by which somethings intrinsic mass can have a discernible effect on the qualitative character of our experience, and that we can therefore detect which intrinsic mass a given body has by noticing what kind of qualitative experience we enjoy at the end of the measurement process. According to this objection, my argument so far only shows is that it would then be physically impossible for us to record the result of the measurement by writing 1 kg or 2 kgs on a piece of paper depending on what the result is. In response, the comparativist might emphasize that the scenario being envisaged is extremely implausible. For it seems compelling that whenever I am able to enjoy two discernibly different qualitative states, I am also able to produce some bodily movementinscribing 1 kg, say, or putting my arm in the airin response to the one experience and not the other. The objection must therefore deny what we take to be an obvious fact about our mental life. But there are two further responses that the comparativist might want to explore. One is to deny that differences in intrinsic mass has any effect on the qualitative character of our experience. To argue this, the comparativist might argue that the doubled world W would be indiscernible from the actual world, in the sense that everything would look and feel and taste and smell exactly the same as it actually does. But is this true? The question is subtle. In the case of velocity, we have empirical evidence that a boosted world would be indiscernible from the actual world based on our experiences in trains. For we have experienced reasonably small environments that are to some extent isolated from external interference and have noticed that they look the same while in motion as they do at rest in the station. Now this evidence is not conclusive, since it remains open that if everything were put in smooth motion there would be some discernible difference in the qualitative nature of our experiences (perhaps our visual eld would be tinged with yellow). But it seems that our best theory of what determines the character of our conscious states implies that the subjects state of absolute motion is not a relevant factor. Now in the case of mass we lack the same kind of empirical evidence that the doubled world W is indiscernible from the actual world, since we have had no experience of reasonably isolated environments that differ only in a doubling of mass. Nonetheless, there is some reason to believe that W would be indiscernible, for our best theory of what determines
43 See

Albert [1] for more on this way of thinking of about detectability.

34

the character of our conscious states seems to imply that intrinsic masses are not a relevant factor. Insofar as ones conscious life is determined by physical facts at all, it seems to be determined by the positions of particles composing ones brain (and perhaps the local environment). When a headache pill cures your pain, that is because it altered the positions of various particles in your brain, not because it made any of those particles more massive. Since W agrees with the actual world in all facts about particle positions, there is some reason to think it is indiscernible after all. A full discussion of this issue would take us too far into the philosophy of mind. But in any case the issue is not crucial, for a third response to the initial objection is to argue that even if intrinsic mass has an effect on the character of our experience, it would still be undetectable! To see this, suppose that the effect that intrinsic mass has on our experience means that our visual eld would be tinged with yellow in W. Does this mean that we can infer, on the basis of the character of our visual eld, which intrinsic mass each thing has? It does not, for to make that inference we would need to appeal to a hypothesis stating which intrinisic masses give rise to which sort of experience. But the hypothesis that one set of intrinsic masses (i.e. those instantiated in the actual world) give rise to my actual experiences and the hypothesis that another set of intrinsic masses (i.e. those instantiated in W) give rise to my actual experiences are both left open by my actual experience, and are both equally simple, elegant, explanatory, and so on; and so the fact that I am enjoying my actual experiences is not evidence for either hypothesis over the other! As a result, my enjoying these experiences does not put me in a position to infer which particular mass any given body has. So here are three responses available to the comparativist. I will not try to assess which response is most plausible, but I should emphasize that ones choice of response is likely to affect the extent to which the argument will generalize to other quantities. For example, if the comparativist makes the second response, resting her case on the fact that W is indiscernible, then the argument will only generalize to a given quantity if uniform transformations of its absolute values while leaving the comparative values xed results in an indiscernible world. And if the comparativist makes the third response, resting her case on the fact that the experiential difference in W leaves open which intrinsic masses are instantiated, then the argument will only generalize to a given quantity if the experiential difference in those transformed worlds still leaves open which absolute values are instantiated.

35

8.6

Inexpressible Ignorance

This, then, is my reason for preferring comparativism, at least in the case of mass. If material bodies had the intrinsic masses posited by the absolutist, those masses would be undetectable. Our Occamist principle states that all else being near enough equal, this is a mark against absolutism and in favor of comparativism. I argued in Sections 27 that there are no decisive objections to comparatativism, so all else appears near enough equal. This Occamist argument rests on the premise that the intrinsic masses posited by the absolutist would be undetectable, but there is a sense in which the state of ignorance we would be in with regards to which intrinsic mass each body has would be inexpressible. For how could we express it? If I said that I do not know whether everything is twice as massive as it actually is, I would be mistaken since I know very well they are not! The trouble is that I just described the non-actual possibility W in such a way that I can infer from my description that it is non-actual. To remedy this, I could try giving each absolute mass a name and describing the non-actual possibility in those terms. For example, suppose that Kripke was right that we use terms of the form n kgs with the reference-xing stipulation that it is to refer to the absolute mass that is n times that had by the standard kilogram in Paris. Since W differs in the mass in kilograms of my laptop, I could try saying that I do not know whether my laptop is 2 kgs or 4 kgs. But this is not clearly right either. By discovering that my laptop is twice as massive as the standard kilogram, and then appealing to the reference-xing stipulation, I can infer that the sentence My laptop is 2 kgs is true. And while knowing that a sentence is true does not imply knowing the proposition it expresses, many theorists would say that in this case I would know that my laptop is 2 kgs. So this is the sense in which my ignorance is inexpressible: there is a sense in which I do not know which absolute mass my laptop has, but there is no sentence s that expresses its absolute mass for which I can truly say I do not know that s. Those who identify knowing which with knowing that might now conclude that I am not ignorant of which intrinsic mass my laptop has after all, but this would be to let theoretical opinion obscure the phenomena. For above I described a clear and vivid sense in which the particular intrinsic mass had by my laptop lies beyond our epistemic grasp, and we can naturally express this by saying that we do not know which intrinsic mass it has in one sense of that term. If this state of ignorance cannot be analyzed in terms of knowing that, so be it: we would be guilty of theoretical prejudice if we concluded that there is no such state at all. This is not to deny that there is an important question as to what the state consists 36

in, but that is a question for another time.44

Conclusion

The question of absolutism vs comparativism has received relatively little discussion, and I consider this a signicant lacuna in our understanding of what the natural world fundamentally consists in. In this paper I have tried to clarify what the issue amounts to and describe where I see the major battle lines as lying. I believe that comparativism is probably the correct view for mass, but if I have not convinced you of that I hope to have shown that the issue is important and that there is interesting further work to do in the area.45

References
[1] Albert, D. The Technique of Signicables. (Manuscript). [2] Armstrong, D. 1978. A Theory of Universals: Volume 2. Cambridge: Cambridge University Press. [3] Armstrong, D. 1988. Are Quantities Relations? A Reply to Bigelow and Pargetter. Philosophical Studies 54. [4] Baker, D. Manuscript. Some Consequences of Physics for the Comparative Metaphysics of Quantity.
44 In this respect the case of absolute mass diverges from the case of absolute velocity, for our ignorance in the latter case is expressible: I can truly say I do not know whether I am at rest. Indeed, the case of absolute mass is more akin to the case of absolute location in space. For if there were such a thing as absolute space then worlds that differ only in a uniform shift of all matter three feet to right would look and feel and smell the same, and as in the case of velocity we can argue that no physically possible measuring device could reveal which particular region of space we are in. For this reason, our position in absolute space is undetectable. Still, as Maudlin [23] point out, there is nothing I can say to express what I am ignorant of, for it is clearly false to say that I cannot not know whether I am here or three feet to the right of here! In this regard I agree with Maudlin entirely. But he went on to argue that there is no sense at all in which I am ignorant of my location in space, and here I believe that he made the mistake described in the last paragraph of allowing theoretical prejudice to obscure the phenomena. I give some reasons to reject Maudlins view in my [9]. 45 Parts of this material were presented in November 2010 at California State University Los Angeles, and in January 2011 at Leeds University and the University of Oxford. Thanks to members of these audiences for all their helpful feedback. Thanks also to David Baker, Alexi Burgess, Kenny Easwaran, Maya Eddon, Kit Fine, Dustin Locke, Michaela McSweeney, Eliot Michaelson and David Plunkett and three referees for extremely helpful comments on earlier drafts of this paper.

37

[5] Bigelow, J. and R. Pargetter 1988. Quantities. Philosophical Studies 54: 287316. [6] Bigelow, J. and R. Pargetter 1990. Science and Necessity. Cambridge: Cambridge University Press. [7] Chalmers, D.J. 2012. Constructing the World. Oxford: OUP. [8] Dasgupta, S. 2009. Individuals: An Essay in Revisionary Metaphysics. Philosophical Studies 145: 3567. [9] Dasgupta, S. 2011. The Bare Necessities. Philosophical Perspectives 25: 115160. [10] Dasgupta, S. On the Plurality of Grounds. (Manuscript). Draft available at http://www.shamik.net/Research.html [11] Earman, J. 1989. World Enough And Space-Time. Cambridge, MA: MIT Press. [12] Eddon, M. Fundamental properties of fundamental properties. (Manuscript). [13] Ellis, B. 1966. Basic Concepts of Measurement. Cambridge: Cambridge University Press. [14] Field, H. 1980. Science without Numbers Princeton, NJ: Princeton University Press. [15] Field, H. 1985. Can We Dispense With Space-time? PSA 2: 3390. [16] Fine, K. 2001. The Question of Realism. Philosophers Imprint 1 (1): 130. [17] Fine, K. Guide to Ground. (Manuscript). [18] Hawthorne, J. 2006. Quantity in Lewisian Metaphysics. In Metaphysical Essays, pp. 229238. Oxford: Clarendon. [19] Kripke, S. 1972. Naming and Necessity. Oxford, UK: Blackwell. [20] Ladyman, J and D. Ross. 2007 Every Thing Must Go: Metaphysics Naturalized. Oxford: OUP. [21] Lewis, D. 1986. On the Plurality of Worlds. [22] Lewis, D. 1986. Philosophical Papers: Volume 2. Oxford: Oxford University Press. 38

[23] Maudlin, T. 1993. Buckets of Water and Waves of Space: Why Spacetime Is Probably A Substance. Philosophy of Science 60 (2): 183203. [24] Mundy, B. The Metaphysics of Quantity. Philosophical Studies 51: 29 54. [25] North, J. 2009. The Structure of Physics: A Case Study. Journal of Philosophy 106: 5788. [26] Roberts, J. 2008. A Puzzle About Laws, Symmetries and Measurable Quantities. British Journal for the Philosophy of Science 59: 14368. [27] Rosen, G. 2010. Metaphysical Dependence: Grounding and Reduction.

39

Individuals: An Essay in Revisionary Metaphysics

Shamik Dasgupta
New York University Department of Philosophy 5 Washington Place New York, NY 10003 email : shamik@shamik.net cell : +1-917-216-8681 fax : +1-212-995-4179

Abstract We naturally think of the material world as being populated by a large number of individuals. These are things, such as my laptop and the particles that compose it, that we describe as being propertied and related in various ways when we describe the material world around us. In this paper I argue that, fundamentally speaking at least, there are no such things as material individuals. I then propose and defend an individual-less view of the material world I call generalism.

Keywords: bundle theory, generalism, holism, individuals, velocity

We naturally think of the material world as being populated by a large number of individuals. These are things, such as my laptop and the particles that compose it, that we describe as being propertied and related in various ways when we describe the material world around us. To what extent is this natural conception true of the fundamental structure of reality? In this paper I propose a view on which it gets things entirely wrong. Fundamentally speaking, I claim, there are no such things as individuals. As we will see, my view has the surprising consequence that the world is, in a well-dened sense to be discussed later, radically holistic : that it is at root a unied whole rather than a collection of disparate parts. The view is best motivated, I believe, by facts about the structure of our best-conrmed laws of physicsfacts that have been well known to philosophers for quite some time, but for one reason or another have so far been overlooked. I proceed as follows. In Section 1, I argue against the view that the fundamental furniture of the material world includes a domain of individuals, a view I call individualism. On this view the most basic, irreducible facts about our world include facts about what individuals there are and how they are propertied and related to one another, such as a is F, b is G, a bears R to b where a and b are individuals, or primitive individuals as I will call them to underline their status on this view.1 My argument against individualism is based on what I show to be a striking analogy between primitive individuals and absolute velocity, by which I mean the velocity of a material body through absolute space rather than relative to another material body. Roughly speaking, the analogy is that our best physical theories imply that both are physically redundant and empirically undetectable. Contemporary orthodoxy counts this as reason enough to doubt that absolute velocity is real, and I will argue that consistency demands we adopt the same attitude about primitive individuals. If we reject individualism, what are we to replace it with? One view, the famous bundle theory, says that individuals are not primitive entities but are instead nothing other than bundles of properties. But in Section 2 I argue that the bundle theory is inadequate as a response to my arguments
Individualism is perhaps the most natural view about the structure of our world and is implicit in a large amount of contemporary philosophy. Any metaphysician who describes the fundamental structure of the world by describing what individuals there are and what they are like is, at least implicitly, an individualist. For example, see Lewis [19]. More explicit endorsements of individualism can be found in Wittgenstein [37] and Russell [28]; and, more recently, in Allaire [2], Armstrong [3], and Hawthorne and Sider [14]. Note that the individualist need not claim that all individuals are fundamental entities: she may think, for example, that chairs and tables are nothing other than collections of electrons and quarks.
1

against individualism. If we reject individualism for the reasons outlined in Section 1, we should not adopt the bundle theory in response. In Section 3 I propose my own view about the structure of the material world, a view I call generalism. While the bundle theorist constructs individuals out of properties, the generalist bypasses individuals and simply constructs facts out of properties instead. To my knowledge, generalism has not been discussed in the recent literature, but I argue that it is perfectly suited as a response to the arguments against primitive individuals. Insofar as the considerations in Section 1 give us good reason to reject individualism, then, they give us good reason to adopt generalism instead. But good reasons to adopt generalism need not be decisive reasons: whether we should ultimately adopt it depends on its overall theoretical virtues. And as we will see, generalism departs radically from our pretheoretic conception of the world in a number of ways, so this may be seen as a point against it. But in Section 4 I argue that generalism has just the theoretical virtues we normally demand of a view that revises pre-theoretic belief, such as explanatory adequacy, simplicity and, importantly, the resources to explain away our pre-theoretic convictions on its own terms. All things considered, then, I conclude in favor of generalism. Before we begin I should say that my methods may seem foreign to readers familiar with the literature on individuals, which tends to focus on purely apriori, modal considerations. For example, individualists have famously argued on apriori grounds that there is a possible world containing just two indistinguishable spheres, and they argue that the bundle theorist cannot make sense of such a world since their view implies that indistinguishable things are identical.2 In this paper I diverge from the literature by bracketing these sorts of issues and asking instead what we can learn from physics. Surprisingly, this question has been largely ignored in the literature. Whether apriori considerations have anything to add is a question I leave for another time. I should also add that my topic in this paper is restricted to the structure of the material world. When I ask whether there are any such things as primitive individuals, this is shorthand for asking whether the material world contains any such things. The structure of the non-material worldif there is anything answering to that descriptionis a topic for another occassion.

Against Individualism

As I said, I think we should reject primitive individuals for the same reason that contemporary orthodoxy rejects absolute velocity: our best physical theories imply that they are physically redundant and empirically unde2 The classic paper is Black [5]. For more contemporary literature in this vein see OLeary-Hawthorne [22], Zimmerman [38], and references therein.

tectable. But, as we will see, this charge is easily misunderstood in the case of individuals. So I propose to start by clarifying exactly what the charge amounts to in the case of velocity. After that I will argue that the same charge can be leveled against primitive individuals. Finally, I will say why their redundancy and undetectability give us reason to doubt their reality.

1.1

A Primer on Velocity

Consider a car cruising down the highway alongside a train. It is traveling at 55 mph relative to the highway and 10 mph relative to the train. But how fast it is really traveling, independent of any reference point? One popular response to this question is that there is no answer: there is no such thing as its absolute velocity, only its velocity relative to other reference points, none of which is privileged over the others as dening its real velocity.3 What motivates this response? These days it is not thought to be an apriori issue; rather, the idea is that we can learn from physics. An example will help. Consider Newtonian Gravitation Theory (NGT), by which I mean Newtons three laws of motion and his inverse-square gravitational force law. The orthodox view these days is that if we had reason to think that the laws of NGT were the true and complete physical laws of motion, that would give us reason to think that absolute velocity is not real.4 But why? The answer is that if the laws of NGT were the true and complete laws of motion, then if absolute velocity were real it would be physically redundant and empirically undetectable. Admittedly, some theorists emphasize the redundancy and others the undetectability, so there may be some disagreement over which of these is most important.5 But no matter: everyone agrees that NGT implies both the redundancy and undetectability of absolute velocity, and that one or both implications give us reason to doubt its reality. Later, I will argue that every physical theory seriously considered over the past 400 years has both implications about primitive individuals. But rst let me explain what the implications consist in.
Readers familiar with the substantivalism/relationalism debate should note that I allow reference points to include unoccupied inertial trajectories in substantival spacetime. 4 I am slurring over a subtlety here. The laws of NGT as written down by Newton himself make reference to absolute velocity, so what Newton wrote down would not be true if there were no such thing as absolute velocity. But I am using the phrase laws of NGT to refer to laws that can be expressed in dierent ways depending on what one takes to be the underlying metaphysics of the world they govern. We use the Schrodinger equation in quantum mechanics the same way, to refer to a law that will be formulated in very dierent ways depending on ones view about the ontology of a quantum mechanical world. The claim in the text, then, is that orthodoxy considers a theory that dispenses with absolute velocity and formulates Newtons laws without reference to it better than Newtons own theory. 5 Earman [10], p. 44, emphasizes the redundancy; Maudlin [21], p. 192, emphasizes the undetectability; while van Fraassen and Ismael [35] appear to emphasize both.
3

To start, what does it mean to say that absolute velocity is empirically undetectable? The concept is best illustrated by outlining a particular line of argument and asking what the argument shows. We are all familiar with the fact that sitting in an airplane in smooth motion feels just like sitting in one at rest on the runway, and this naturally motivates the idea that we cannot tell smooth motion from rest. But in fact this thought experiment is not conclusive, for it leaves open the possibility that we could build a measuring device to help us detect our absolute velocity. After all, we cannot detect a particles charge with the naked eye either; we need to build complex electrometers which detect charge for us and display it on a screen. So the question arises whether it is possible to build an absolute velocity measuring device. Surprisingly, the literature contains little explicit discussion of this question. But in fact there is a line of argument which shows that if the laws of NGT were the true and complete physical laws, there could be no such device. It will prove useful to rehearse the argument before applying it to the case of primitive individuals.6 Imagine trying to build a device that detects whether it is moving or at rest. What properties must this device have? At a minimum, it would need to be sensitive to whether it is moving or not. But it is worth being quite specic about what this means. Presumably, it would need to have a ready state in which it is set up to measure its velocity through space. And it would need to be built in such a way that if it were in its ready state at some initial time t0 and was then switched on while moving, it would whir away and then register this fact by, say, displaying Moving on a computer screen at some later time t1 . Correspondingly, if it were in its ready state at t0 and then switched on while at rest, it would then register this dierently, say by displaying At rest on the screen at t1 instead. But it is straightforward to show that if the laws of NGT were the true and complete physical laws, it would be impossible to build a device with these properties no matter how much funding were made available. The key fact about NGT is this: Given any two closed systems governed by (and only by) the laws of NGT, if at an initial time they dier only in facts about absolute velocity but are exactly the same in all other respects, including all facts about relative velocity, then they will continue to be exactly the same in all those other respects at all subsequent times.7
6 This line of argument is very rarely discussed explicitly in the philosophical literature, though see Roberts [27] for an exception. Feynman gestures at it in his [12], chapter 16. I have also heard versions of it in seminars given by David Albert and Tim Maudlin. 7 This slurs over some subtleties. First, the term closed systems is used as a dummyterm. The reader can think of closed systems as possible worlds, or as mathematical models, or as idealized laboratories with walls that insulate against eects from outside. Second, I leave open whether the initial time refers to an instant or a (perhaps inntesimally small) period of time. Incidentally, some theorists describe the indented claim as

This implies that any two closed systems like this will continue to agree at all subsequent times on all facts about the relative positions of things, including the pattern of pixels illuminated on computer screens. So, let us suppose that an absolute velocity measuring device in its ready state is switched on at t0 , and at t1 the word At Rest is displayed on its screen. Then no matter how the device was built, no matter how many years of technological development was invested in it, it follows immediately from the indented claim above that if the entire system had diered at t0 only in its absolute velocity, the device would still have displayed At Rest at t1 . So the device does not have the properties specied in the last paragraph. Notice how powerful this form of argument is: analogous reasoning establishes that no device can display our absolute velocity in pointer positions, or patterns of ink on computer printouts, or sounds emanating from speakers, or. . . . And it is hard to see how else a device could display our velocity, if not in these sorts of media. So the argument establishes, purely on the basis of the structure of the laws of NGT, that if they were the true and complete laws governing our world we would be precluded from ever being able to distinguish between closed systems that initially dier only in facts about their absolute velocities. This is what I mean when I say that absolute velocity is empirically undetectable according to NGT.8 Of course, one can imagine other laws according to which initial differences in absolute velocity would lead to subsequent dierences in interparticle dierences, and if those were the true and complete laws of motion we could perhaps build an absolute velocity measuring device that exploited this fact. But the laws of NGT are not like this, and that is why they imply that absolute velocity is empirically undetectable. Later, I will show that precisely the same sort of argument can be used to show that every physical theory we have taken seriously over the past 400 years implies that primitive individuals are empirically undetectable too. But rst we need to nish our primer on velocity. I have just explained why absolute velocity is empirically undetectable according to NGT, but the other charge is that it is physically redundant according to NGT. What does
the fact that NGT is symmetric under uniform velocity boosts (see for example Wigner [36]). More contemporary formulations of symmetry are somewhat dierent and avoid the complications just mentioned, but the formulation in the text is more intuitive and will suce for our purposes here. 8 I believe that this form of argument has been extremely inuential throughout the history of physics. For example, when formulating his Special Theory of Relativity, I would argue that Einstein was guided (in part) by the constraint that the correct laws of physics must imply, through reasoning much like that rehearsed, that absolute velocity is empirically undetectable. Admittedly, there is much more to say about what empirical undetectability really amounts to, but a full discussion of the issues involved would take us too far from the main thread of the current paper so I will use the rough characterization in the text for simplicity. I say more about what empirical detectability amounts to in my [9].

this mean? It means nothing other than the indented, italicized claim above. For the intuitive idea behind that claim is that according to NGT, mere dierences in absolute velocity do not give rise to any other dierences at later times. Note that this is in stark contrast to other quantities: an initial dierence in facts about mass would give rise (via F=MA) to dierences in acceleration and therefore to dierences in inter-particle distances at subsequent times, and this is precisely what devices that measure mass exploit. But dierences in absolute velocity have no such eect, and as we just saw this fact can be used to argue that velocity is empirically undetectable.

1.2

Primitive Individuals As Danglers

As a shorthand, call something a dangler if and only if it is physically redundant and empirically undetectable. I just argued that it is a consequence of NGT that absolute velocity is a dangler. I will now argue that (1) It is a consequence of every physical theory considered over the past 400 years that primitive individuals are danglers. This is the rst premise of my argument against individualism and in favor of my alternative metaphysics. Now, redundancy is relatively straightforward. Call facts that concern particular primitive individuals, such as a is F, b is G, a bears R to b, a6=b individualistic facts. And call facts that do not concern any particular primitive individual general facts. These include facts that can be expressed in predicate logic without constants (but with identity), such as (x)Fx, (y)Gy, (x)(y)Rxy, (x)(y)x6=y (I assume that the predicates just used do not include substitutes for constants, such as x Socratizes.) These facts may be taken to imply the existence of individuals, but they are not individualistic facts because they do not concern any particular primitive individual. The claim that primitive individuals are redundant to NGT can now be stated as follows: Given any two closed systems governed by (and only by) the laws of NGT, if at an initial time they dier only in their individualistic facts but are exactly the same in all other respects, including all general facts, then they will continue to be exactly the same in all those other respects at all subsequent times. For example, consider the following system: a primitive individual called Peter is at an initial time t0 propelled up in the air by a slingshot, only to fall by gravity back to Earth. And now consider a dierent system whose initial state at t0 diers only in the fact that a dierent primitive individual, 7

Paul, is slung. By hypothesis, we are to suppose that Peter and Paul have the same mass, shape, charge and so on. What the indented claim implies is that if the two systems both obey NGT, then Paul will make exactly the same trajectory through space as Peter. According to NGT, the identity of each particle makes no dierence to how the slingshot or the Earths gravitational eld aect it. So, just as with absolute velocity, it is a straightforward consequence of NGT that mere dierences in individualistic facts at an initial time do not give rise to dierences in the future evolution of inter-particle distances or any other dierence; in particular, they do they give rise to any dierence in general facts. We can imagine other laws in which the identities of the primitive individuals that initially compose a system has an eect on the future evolution of inter-particle distances.9 But the laws of NGT are not like that, and it is uncontroversial that none of the physical theories we have considered over the past 400 years are like that either.10 So primitive individuals are redundant to all our best physical theories in the same sense as absolute velocity is redundant to NGT. Of course, one might hold that the existence of primitive individuals is entailed by (non-trivial) general facts, and therefore that they are not redundant in some other sense. I will discuss this when I discuss how best to dispense with primitive individuals in Section 3. My claim here is just that primitive individuals are redundant in the sense dened above, namely that mere dierences in individualistic facts do not give rise to any other dierences at later times. To establish (1), we also need to argue that primitive individuals are empirically undetectable according to our best physics. It is worth noting here that there is a distinguished tradition according to which we do indeed lack a certain kind of epistemic access to primitive individuals. Russell, to take just one example, described them as an unknowable something and said that they cannot be dened or recognized or known.11 Part of the intuition is that since they are clothed in properties one can only ever see the properties and never the individual itself, much as I cannot see someones skin when they wear an anorak. But this is only metaphorical,
Earman discusses the possibility of laws of this type on pp. 6-7 of his [11]. Aristotles physics might be thought of as a theory of this sort: he thought that the universe had a distinguished center, with earth gravitating towards it and re away from it. 10 This statement of redundancy assumes that the laws are deterministic. The generalization to probabilistic laws is reasonably straightforward, but considerations of space prevent me from discussing it here. 11 Russell [30], p. 97; and Russell [29] respectively. Other proponents of this tradition include Locke, who described primitive individuals as an unknown support of those qualities (see his [20], II xxiii 2); and Hume, who described them as unknown and invisible (see his [16], p. 220). I should say that none of these authors called their subject matter primitive individuals, but I believe they were all talking about them.
9

and many contemporary philosophers think that Russell was misguided.12 I think Russell was onto something, but I believe a better way of making the point is to say that primitive individuals are empirically undetectable. As we will see, putting things this way avoids potential objections to Russells way of putting things. To argue that primitive individuals are empirically undetectable, let us use the case of absolute velocity as a guide. In that case, we started out by noticing that an airplane in smooth motion would look and feel and smell exactly the same if it were at rest on the runway, and this motivated the idea that we cannot tell smooth motion from rest. What is the analogous thought in the case of individuals? Well, imagine a situation in which there is a primitive individual is placed in front of you. Depending on what sorts of things primitive individuals are (or would be, if they were real!), this might be a situation in which you are in front of a chair, or an electron, or perhaps something else. To x ideas, let us suppose without loss of generality that it is a chair. And now imagine a situation in which everything is exactly the same except that a dierent primitive individual is in front of you. Suppose this dierent individual has exactly the same qualities as the actual chair in front of you: imagine it were colored the same, shaped the same, and so on. The analogous thought to that in the case of velocity is that the situation would look and feel and smell exactly the same to you: we cannot tell the dierence between situations that dier only in their individualistic facts.13 Now, as in the case of velocity, this sort of thought experiment is not conclusive because it leaves open that we could build a device that would allow us to distinguish between the two situations. But I claim that the redundancy of primitive individuals in all our best-conrmed physical theories means that if the laws of physics are anything like what we think they are, it is impossible to build such a device no matter how much funding we are granted. If this is right, then primitive individuals are empirically undetectable according to our best-conrmed physics in precisely the same sense as absolute velocity is empirically undetectable according to NGT. The argument is just like the one we ran in the case of absolute velocity. Consider a particular primitive individual , and imagine what property an -detector must have. Presumably, it would need to have a ready state in which it is set up to detect whether it is in front of . And it would need to be built in such a way that if it were in its ready state at some
See for example Armstrong [3], p. 956; and Sider [32], p. 289. Not all philosophers would agree with the way I put things here. For example, Campbell [7] defends a view on which the phenomenal character of an experience depends on which individual one is presented with, and he may therefore want to say that the two situations would look dierent. Nonetheless, he can agree that there is a sense in which the two situations are indistinguishable and that we cannot tell the dierence between them. This is, strictly speaking, all my arguments require, but for ease of prose I shall sometimes slur over this subtlety and talk of the situations as looking the same.
13 12

initial time t0 and were switched on in front of , it would register that fact by (say) making a dial swing to the right at some later time t1 ; while if it were switched on in front of something else the dial would swing to the left at t1 instead. But it follows immediately from the redundancy of primitive individuals in all our best-conrmed physical theories that this machine is impossible if anything like those theories is true. For let us suppose that an -detector in its ready state is switched on at time t0 , and at t1 the dial swings to the right. Then no matter how the device was built, no matter how many years of technological development was invested in it, the redundancy of primitive individuals in the laws of physics implies that if the entire system had diered at t0 only with respect to what primitive individual was in front of the device, the dial would still have swung to the right. So the device does not have the property specied earlier in this paragraph. So I claim that primitive individuals have the same status in contemporary physics as absolute velocity has in NGT: they are physically redundant and empirically undetectable. This outlines my argument in favor of my claim that: (1) It is a consequence of every physical theory we have considered over the past 400 years that primitive individuals are danglers. Now, contemporary orthodoxy thinks that if absolute velocity were a dangler, this would be a good reason to doubt its reality. Out of consistency, then, I claim that we should take ourselves to have good reason to doubt the reality of primitive individuals.

1.3

Dispensing with Danglers

But I have not yet said anything about why somethings being a dangler should give us reason to doubt its reality, so let me turn to this question now. Then I will discuss potential objections to my line of argument. If we discover that something is a dangler, I do not claim that this is a decisive reason to think that it is unreal. But I do think it is something of a vice if a theory about the structure of the world implies that there are danglers. Specically: (2) Consider two theories about the structure of the material world, and suppose you discover that the rst implies that the world contains a dangler while the other does not. All else being equal, it is rational for you to prefer the latter over the former. In Section 3 I will introduce generalism, a theory about the structure of the world that dispenses with primitive individuals. And in Section 4 I will argue that all else is equal, i.e. that it scores equally well on other theoretical virtues such as simplicity, explanatory adequacy, and so on. That will constitute my argument against individualism and in favor of generalism. 10

It is worth noting that (2) is assumed, perhaps implicitly, throughout the philosophy and physics literature. The case of absolute velocity in NGT is one example, but the case of absolute simultaneity in the Special Theory of Relativity (STR) is perhaps more striking. It is often remarked that according to the laws of STR there are no facts about absolute simultaneity i.e. facts about whether two events occur at the same timeonly facts about simultaneity relative to an inertial frame of reference. But this remark is not strictly speaking true: on the Lorenz formulation, the laws of STR govern the motions of particles through a classical space-time structure in which there are well-dened facts about absolute simultaneity. What is true, though, is that absolute simultaneity is a dangler in STR. The incorrect remark is so often made, I think, because (2) is so widely taken for granted. The justication for (2) is broadly speaking Occamist, and is nicely illustrated with the case of absolute velocity. Suppose the laws of NGT are the true and complete laws of our world, and therefore absolute velocity is a dangler. The redundancy of absolute velocity means that initial differences in absolute velocity do not give rise to any other dierence such as a dierence in inter-particle distances. So there is a sense in which the particular absolute velocity of a system at an initial time is not needed to explain any facts about subsequent inter-particle distances and so on: even if its initial absolute velocity had been dierent, all facts about inter-particle distances would have been the same. Does this mean that facts about the particular absolute velocity of a system are never needed in explanations? Not quite, because it remains the case that the initial absolute velocity of a system is required to explain its absolute velocity at later times. But this is where the undetectability comes in: for if facts about absolute velocities are empirically undetectable, then facts about the particular absolute velocity of a system at an initial time are not required to explain anything we can empirically detect ! Therefore, Occams razor has us prefer theories about the structure of the material world that dispense with them (all else being equal). Precisely the same goes for primitive individuals: the redundancy and undetectability of individualistic facts implies that they are not required to explain anything we can empirically detect. This, in broad outline, is the theoretical motivation behind (2). In addition, I would also add that the project of attempting to dispense with danglers has borne much theoretical fruit in the past (for example, in the theory of motion). So I think there is a pragmatic reason for continuing the project in the case of primitive individuals, if only to see where it leads. It is worth guarding against some potential misunderstandings. First, (2) only concerns theories about the structure of the material world, so it does not say anything about whether we should believe in immaterial souls or abstracta such as numbers, sets and so on. Second, (2) is not a rabid get rid of everything that is not physics principle sometimes found implicit in the literature. The property of being a table, for example, is a high-level 11

property not referred to in physics text-books; but this property is neither empirically undetectable nor physically redundant in the senses specied earlier. Therefore, (2) does not recommend dispensing with this property. Third, (2) does not say that it is meaningless to talk about danglers, or that it is part of the meaning of reality that empirically undetectable structure is not real. So it is not a vericationist principle, and is perfectly consistent with the claim that some danglers are real. It just says that we have reason to dispense with danglers, all else being equal.

1.4

Two Objections

So far I have motivated (1) and (2). If they are right, we should be interested in whether there is a theory about the structure of the material world that dispenses with primitive individuals, since (all else being equal) it would be rational for us to prefer it. But before I propose a theory of that sort, let me pause to consider two objections to my line of argument so far. The rst objection claims that experience presents us with a world full of individuals. We may agree with (1) that we are unable to empirically detect which particular ones they are; but, the objection goes, our experiences nonetheless present them as being there. The objection is that a theory that dispenses with primitive individuals cannot account for this datum and so should not be accepted; therefore, (2) is false. In reply, I can agree for the sake of argument that experience presents us with a world full of individuals, for this is no objection to (2). Remember, (2) just says that all else being equal, we are rational to prefer theories without danglers. If a theory that dispenses with primitive individuals cannot account for how the world appears, then all else would not be equal. At best, this objection points out is that there is an onus on me to show that the alternative metaphysics I propose can account for how the world appears. Once we have the view on the table, I will argue that it can. A second objection is that (1) is false. (1) implies that when faced with a primitive individual in front of you, say a chair, you cannot empirically detect which primitive individual it is. Against this, it may be pointed out that it is perfectly possible to know which chair it is. I think this sort of point convinced people that Russell was wrong when he said that primitive individuals were an unknowable something. Now as stated, the objection simply confuses the question of whether we can empirically detect which primitive individual it is, with the question of whether we can know what chair it is. To respond, I will argue that (i) even if we grant the objector that we can know what chair it is, it does not follow that we can empirically detect which primitive individual it is; and (ii) it is empirical detection, not knowing what, that matters to the Occamist argument outlined above. There are two things the objector may mean by knowing what. First, 12

suppose I ask whether you know what chair is in front of you. A natural response would be Sure, I know that it is the only chair I own, or Sure, I know that it is the one I sat on yesterday. The thought here is that if you know that the chair uniquely has as a certain property (as in the rst answer), or if you are able to re-identify the chair over time (as in the second answer), then you know what chair is in front of you.14 This may be right, at least in the colloquial sense of knowing what. But still, it does not follow that you can empirically detect what individual is in front of you. For even if you know that the chair in front of you uniquely has a given property and is the same chair you sat in yesterday, the individualist claims that there is a further fact of the matter concerning its identity. And what we argued earlier is that situations that dier only in this further facte.g. in which a dierent primitive individual is in front of you and was the chair you sat in yesterdayare indistinguishable if the laws of motion governing our world are anything like we think they are. Therefore, even if you know what chair is in front of you in this colloquial sense, it does not follow that you can empirically detect what primitive individual it is. In fact, precisely the same can be said in the case of absolute velocity. Suppose I ask whether you know what your velocity is. In some contexts, it may be appropriate to answer Sure, I know that my velocity is the same as it was yesterday. But this does not mean you can empirically detect absolute velocity. For even if you know that your velocity has not changed since yesterday, the believer in absolute velocity thinks there is a further fact of the matter concerning what your absolute velocity is. And we argued earlier that according to NGT, situations that dier only in this further fact are indistinguishable; therefore, it remains the case that absolute velocity is empirically undetectable according to NGT. Now the objector may point out that it is easy to acquire knowledge of this further fact of the matter concerning the identity of the individual in front of you: simply point at the chair and think the individual in front of me is that one ! The idea would be that the content of your belief so formed would depend on the identity of the primitive individual in front of you. As a result, if you had formed a belief like this in a situation that diered only with respect to the identity of the individual, you would have formed a dierent belief that is true of that other individual instead. Some theorists may therefore count this as knowing the further fact of the matter concerning the individuals identity. This is the second notion of knowing what chair it is that the objector may have had in mind. In response, I grant that this sort of knowledge of primitive individuals is possibleor at least would be, if there were such things! But it would not
See Strawson [33], pp. 1538 for a discussion of the importance of these everyday senses of knowing what.
14

13

count as empirically detecting what individual is in front of you. For it would remain the case that two scenarios diering only in their individualistic facts are indistinguishable, and therefore primitive individuals would still be empirically undetectable in our sense. This is not to deny that there is an important category of belief of the sort just discussed, often called de re belief in the literature. And if I take this category seriously, there is an onus on me to show how my alternative view can account for it if, as the view claims, there is at a fundamental level no re. Once my view is on the table I will argue that it can. For now, my point is just that the possibility of this sort of belief does not aect the truth of (1).15 So in neither sense of knowing what does knowing what chair it is imply having empirically detected what primitive individual it is. And I claim that it is empirical detection, not knowing what (in either sense), that matters to the Occamist argument against primitive individuals. For when a feature of the material world is empirically undetectable, it is impossible (given the laws) to distinguish between situations that dier with respect to facts about that feature. The Occamist razor I wish to wield simply says that we have reason to dispense with the putative structure that dierentiates between those situations. The fact that if individuals were real we could re-identify them over time or have de re beliefs about them is, according to this Occamist razor, neither here nor there. There is much more to say about the line of argument I have been pursuing.16 But I take the analogy with absolute velocity to suggest that there is something to it. If so, we should be interested in whether there is a defendable theory of the material world that dispenses with primitive individuals.

Against the Bundle Theory

But if individuals are not primitive entities, what are they? A natural suggestion is that they are nothing other than bundles of properties. This is the well-known bundle theory of individuals. Now, the bundle theorist does not think that any old collection of universals bundles an individual. Instead, she says that properties bundle an individual when and only when the properties are compresent. What does this mean? The bundle theorist is likely to take compresence as a primitive
One might argue that the same remarks apply to the case of velocity. Suppose you said to yourself I hereby let Bob name the real number that is my absolute velocity in miles per hour, and then thought my velocity through space is Bob miles per hour. Now, it is debatable whether this would really result in a similar type of knowledge as that just outlined in the case of individuals. But even if we grant that it would, it remains the case that scenarios diering only in facts about absolute velocity would be indistinguishable to you, and so it would remain the case that absolute velocity is empirically undetectable in our sense. 16 I try to say some of it in Dasgupta [9].
15

14

predicate. Intuitively, when the bundle theorist says that the properties F and G are compresent, she means to describe a situation that the individualist would describe by saying that F and G are instantiated by the same individual. So, when the individualist says that my laptop is black and solid, the bundle theorist says instead that blackness and solidness are compresent. Finally, if individuals are to be identied with bundles of universals, there must be a bundling operation that takes compresent universals and delivers another entity, the individual.17 Such is the standard presentation of the bundle theory. One would expect the bundle theorist to further clarify her view by supplying a logic for the compresence predicate and the bundling operation. But unfortunately the theory is rarely developed in detail, and as a result many questions are left open. Can relations be compressed with other properties? What is the adicity of the compresence relation? Are the compresence relation and the bundling operation one and the same? We are not told, and the view could in principle be developed in many directions.18 Fortunately, nothing I say here will depend on these details, but only on two uncontroversial facts about the view. First, when our bundle theorist says that individuals are bundles of properties, she means something rather specic by property. For since she wishes to dispense with primitive individuals, she cannot think that properties are themselves sets of primitive individuals. Nor can she think they are tropes or property-instances such as this very redness of my sweater, where this is a dierent entity than that very redness of my apple even if my sweater and my apple are exactly the same color. For one can argue that these are danglers just like primitive individuals: I cannot empirically detect whether my sweater has this very redness or another redness of exactly the same color. Instead, by property our bundle theorist will mean a universal : a quality that individualists would describe as being capable of instantiation by many things at once. On this conception of a property, the redness of my sweater is one and the same thing as the redness of my apple. The second fact about the bundle theory is that it implies a principle known as the Identity of Indiscernibles (IOI), which states that it is not possible for two individuals to share the very same universals.19 It implies this because the driving idea behind the view is that there is nothing more to an individual than the universals that bundle it; if the same universals
The bundle theory can be traced back to many historical philosophers including Russell [30] and [29]. Its contemporary advocates include Paul [24], and Cover and Hawthorne [8]. 18 Hawthorne and Sider [14] develop the view in more detail than most; however, the view they emerge with (though do not endorse) has more in common with the theory I develop in the next section. 19 I should say that the title Identity of Indiscernibles is sometimes used to name other related principles, but the details of these other principles need not concern us here.
17

15

could give rise to dierent individuals, this idea would be lost.20 Unfortunately, this commitment to IOI means that the bundle theory is unsatisfactory as a response to the problem of danglers. For consider a world that we would ordinarily describe as containing exactly two individuals both of which have the very same universals; specically, suppose that they each have the universal F and nothing else. Since she endorses IOI, the bundle theorist rules this world as impossible. But notice that the putative world can be characterized in purely general terms by the following sentence: (A) (x)(y)(Fx & Fy & x6=y) Therefore, by ruling this world as impossible, the bundle theorist puts a restriction on the sorts of general facts that can hold. To see why this is a problem, consider again the case of velocity. What is the standard response to the discovery that absolute velocity is a dangler in NGT? It is to adopt the view that Newtons laws are better formulated on a Galilean space-time structure instead. What is a Galilean spacetime structure? The details need not concern us here, but two things are important. First, in a Galilean structure there is no such thing as a particles absolute velocity, only its velocity relative to inertial frames of reference. And, second, a Galilean structure places no restriction on the pattern of relative velocities that a system of particles may display. This second point is crucial. If the arguments outlined in the last section were successful they established that absolute velocity is physically redundant and empirically undetectable in NGT, but they did not establish anything about relative velocity. Even if those arguments worked, for example, it remains open that relative velocity is empirically detectable according to NGT. So if one wants to dispense with absolute velocity because it is a dangler, it would be somewhat perverse to adopt a view that dispenses with it and, in addition, constrains the pattern of relative velocities that a system of bodies may instantiate. For example, suppose the new view implied that a world in which there are just two particles at rest relative to one another is impossible. Then we would be entitled to complain that we only wished to dispense with absolute velocities; we did not in addition want to rule out certain patterns of relative velocities as impossible. Returning to the case of individuals, the analogous arguments purported to establish that individualistic facts are physically redundant and empirically undetectable, but they did not establish anything about general facts.
It is uncontroversial that the bundle theory implies IOI. What is controversial is whether this counts against the view. See Hacking [13] and OLeary-Hawthorne [22] for arguments that it does not. I should say that some theorists appear to deny that the bundle theory implies IOI (for example, Rodriguez-Pereyra [26]), but on closer examination it invariably turns out that they are referring to views that identify individuals with bundles of property-instances or tropes. For the reasons just given in the text, this is not the sort of view under consideration here.
20

16

Even if the arguments were successful, for example, it remains open that general facts are empirically detectable. So if our aim is to dispense with primitive individuals because they are danglers, it would be somewhat perverse if our replacement view put additional constraints on what sorts of general facts can hold. But this is exactly what the bundle theory does! So the bundle theory is unsatisfactory as a response to the problem of danglers in just the same way as the strange theory of motion mentioned at the end of the last paragraph. This is by no means a conclusive objection to the bundle theory, but it does show that it is at best an inelegant response to the problem of danglers. A more elegant response would dispense with primitive individuals without constraining the sorts of general facts that can hold. Let us see if we can develop a view of that sort.21

Generalism

The bundle theorist went astray because she asked What are individuals, if not primitive entities? The natural answer was that they are nothing other than bundles of properties, and she was thereby led to a view that implies IOI and which therefore constrains what sorts of general facts are possible. I propose that we adopt a dierent strategy. Unlike the bundle theorist, we will not ask how to construct individuals out of more basic entities. Instead, we will simply ask for an account of the fundamental structure of the world that dispenses with primitive individuals but which allows us to make sense of the whole array of possible general facts, including that expressed by (A), and which therefore does not imply IOI. This approach mirrors our approach in the case of velocity, in which we develop a metaphysics of motion that dispenses with absolute velocity but places no restriction on the pattern of relative velocities that a system can display.22 How might we develop such a metaphysics? To clear the ground, let me briey put aside two ideas. One idea is to propose that individualistic facts supervene on general facts: if two possible worlds agree on all their general facts they agree on all their individualistic facts. The idea is that on this view, individualistic facts are determined by the general facts, at least on a
Just to clarify, my objection to the bundle theory is not that it implies IOI and we have apriori reasons to think that there are possible worlds that are counterexamples to IOI. Others object to the bundle theory in this way, but as I said in the introduction I wish to bracket apriori intuitions about what worlds are possible. Instead, my objection is that the bundle theory implies IOI and IOI is utterly unmotivated by my arguments against primitive individuals. 22 It may be interesting to note that in his seminal paper on this topic, Adams [1] claims that the denial of primitive individuals stands or falls. . . with a certain doctrine of the Identity of Indiscernibles (p. 11). I believe he said this because he fell into the same trap as the bundle theorist, namely of asking what individuals are, if not primitive entities. I hope to show here that another approach is possible.
21

17

modal sense of determined.23 To see why this suggestion will not do, consider the analogue view in the case of velocity. The analogue view is that facts about absolute velocity supervene on facts about relative velocity: if two possible worlds agree on all facts about relative velocities they agree on all facts about absolute velocity. Now this view is perfectly consistent with the view that absolute velocity is a real quantity in terms of which relative velocity is dened; all it says is that there is a (rather bizarre) modal connection between the two quantities. So in and of itself, this modal claim does nothing to dispense with absolute velocity. To dispense with absolute velocity, we instead need to adopt a theory about the structure of the actual world, namely a Galilean spacetime structure. This may imply certain modal claims, such as that all facts about motion supervene on facts about relative velocities, but it is not at root a modal claim. Similarly in the case of individuals. All the current suggestion says is that there is a certain modal connection between general and individualistic facts. But this modal claim is one that the individualist is free to agree with, so in and of itself it does nothing to dispense with primitive individuals. Instead, we want a theory about the structure of the actual world that dispenses with primitive individuals. This new theory may imply certain modal claims, such as that all facts supervene on general facts, but at root it will be a claim about the structure of our world and not just a modal claim. This naturally leads to another suggestion: that all fundamental facts are general. On this view, the fundamental facts of the world are of the form (x)Fx, (y)Gy, (x)(y)Rxy, (x)(y)x6=y and so on.24 This suggestion is closer to what we want because it is at least a claim about the structure of our world. But as it stands it is unacceptable. After all, we have been brought up to understand that quantiers range over a domain of individuals. So our natural understanding of the facts listed above is that they hold in virtue of facts about individuals, and it would therefore appear that we have made no progress. Instead, what we want is a clear articulation of the fundamental facts of the world that meshes with the idea just mooted, namely that the struc23 Lewis calls this view anti-haecceitism, and uses haecceitism to label the view that there are purely individualistic dierences between some possible worlds (see his [19], p. 221). This terminology is the norm in one wing of the literature, but I hesitate to use it because other wings of the literature use this terminology to denote dierent distinctions. 24 This may have been the view discussed (though not endorsed) by OLeary-Hawthorne and Cover [23], when they write that a theorist may insist that the full story about that world [can] be captured by general propositions, of the sort (x)(x is a sphere), (x)(y)(x is a sphere and y is a sphere and x6=y), and so on. (p. 12). In this quotation they are talking about a possible world, but the analogous view about the actual world sounds very close to the view under consideration.

18

ture of the world is fundamentally general. I call this approach generalism. Before I develop it, let us pause to ask what this mesh requires.

3.1

Desiderata on a New Metaphysics

Assume for the moment that all possible general facts can be expressed in the language of predicate logic with identity but without constants. Call this language PL. (I assume that PL contains no substitute for constants, such as predicates like x Socratizes). The generalist will claim that the fundamental facts of the world are those expressed by a dierent language G (for Generalism). The idea is that when we would ordinarily describe a situation as being one in which a given sentence p of PL is true, the generalist will describe it as being a situation in which, fundamentally speaking, a certain sentence g of G is true. And vice-versa: when the generalist describes a situation with g , we would ordinarily describe it with p. When sentences of PL and G are related in this way I will call them equivalent (which is not the relation of logical equivalence since the sentences come from dierent languages). Now, any metaphysician who paraphrases sentences of one language with sentences of another faces the question of what the relation of equivalence really consists in. This is an extremely delicate question, but to keep things tractable let me just state a necessary condition that the generalist puts on the relation: it must preserve the logical structure of each language. For instance, if a sentence g1 of G is equivalent to (x)(Fx & Gx), and if a sentence g2 of G is equivalent to (x)Fx, then it had better be the case that g1 logically implies g2 . More generally: Preservation: Suppose a sentence g1 in G is equivalent to a sentence p1 in PL, and a sentence g2 in G is equivalent to a sentence p2 in PL. Then g1 logically implies g2 if and only if p1 logically implies p2 . The generalist will claim that the fundamental facts of the world are those expressed in a language G satisfying this constraint. But there are two further things required of G if generalism is to mesh with the idea that the world is fundamentally general. First, we require that G is rich enough to make sense of all possible general facts: Sufficiency: Every sentence of PL is equivalent to a sentence of G. This is precisely where the bundle theory failed, for it could not make sense of a situation completely characterized by (A). But we also require that G describe no more structure than the structure of generality: Modesty: Every sentence of G is equivalent to a sentence of PL. This is precisely where individualism fails, for there is no sentence of PL equivalent to That very individual is red. So by satisfying these desiderata, 19

generalism attempts to sail between individualism and the bundle theory. In fact, generalism will stand to individualism in just the same way as a Galilean space-time structure stands to traditional conceptions of motion that make sense of absolute velocity. A Galilean space-time structure rejects absolute velocity but makes sense of all possible facts about relative velocity; likewise, a view satisfying these desiderata will reject primitive individuals but allow one to make sense of all possible general facts. Admittedly, I assumed above that all possible general facts can be expressed in PL. Given the expressive limitations of PL one may think that a more realistic assumption is that all possible general facts can be expressed in second-order logic, or perhaps an innitary rst-order logic of some degree. But rather than get sidetracked by this issue, I propose to continue with my (potentially) simplifying assumption as a working hypothesis in order to get a avor of the sort of view that emerges. The extension of the current approach to other languages is a project for another time.

3.2

An Algebraic Approach

Indeed, even with this assumption in place there is more than one way to satisfy these desiderata, and I will not argue which is the best. Instead, I will just outline one view and call it generalism, ignoring for simplicity that this is only one way to execute this strategy.25 The generalists ontology consists of a domain of properties, each with a given adicity. Properties of adicity greater than one might be better called relations, but I will use the term property to cover them all.26 Like the bundle theorist, our generalist thinks of these properties as universals rather than sets of individuals or tropes. To talk about these properties the generalist introduces to language G terms of the form P n, where n indicates the adicity of the property denoted. So, for example, we might use the term F 1 to denote the 1-place property of being a friend.27 According to the generalist, the domain of properties instantiates a certain algebraic structure. To talk about this structure, she introduces six expressions to G: &, , c, p, , and . Syntactically, they apply to terms to give a complex termmuch like the father of x in English applies to
25 I am not aware of a view like generalism being proposed in the literature. The closest I have come across is a view discussed very briey by Van Cleve in the last section of his [34]. It also bears some similarity with the ontological nihilism of Hawthorne and Cortens [15]. The language G described here is borrowed from a language discussed by Quine [25] and developed by Kuhn [18]. Burgess and Rosen [6] discuss the prospects of using a language like this to pursue nominalism in the philosophy of mathematics, but the literature contains little discussion of the prospects of using it to formulate a generalist metaphysics. 26 I should also say that the ontology will include properties of adicity zero which might more accurately be called states of aairs, but I will discuss this in more detail later on. 27 For ease of prose I will be rather sloppy about the use-mention distinction in what follows. I do not think this will lead to any serious confusion.

20

terms to give a complex termso we can call them term-functors. What do they mean? A complete and precise explanation requires some formal machinery, so to avoid getting bogged down I leave those details to the appendix. Here I will just impart the basic idea, in the hope that the benets of brevity outweigh the costs of imprecision. The most familiar term-functors are and &, which take properties and give their negations and conjunctions respectively. For example, if F 1 is the 1-place property of being a friend and G1 is the 1-place property of being generous, then F 1 is the 1-place property of not being a friend; (F 1 & G1 ) is the 1-place property of being a generous friend; and (F 1 & G1 ) is the 1-place property of being a friend that is not generous. The term-functors and are permutative expressions. They are used to achieve much the same eect as we achieve in PL by permuting the order of variables. For example, if L2 is the 2-place property of loving, which we ordinarily understand as holding between individuals x and y if and only if x loves y , then L2 is the 2-place property of being loved, which we ordinarily understand as holding between x and y if and only if y loves x. Thus, (L2 & L2 ) is the 2-place property of loving unrequitedly, which we ordinarily understand to hold between individuals x and y if and only if x loves y and y does not love x. What about c? Well, if L2 is the 2-place property of loving, which we ordinarily understand as holding between individuals x and y if and only if x loves y , then cL2 is the 1-place property of being loved by someone, which we ordinarily understand as being instantiated by an individual y if and only if someone loves y . Very roughly, c partially lls a property by stating, as we ordinarily say, that something instantiates its rst position. I will leave an explanation of the term-functors p and to the appendix, since they are not necessary to get an intuitive feel for the view. In addition to these term-functors, the generalist also helps herself to a 2-place term I 2 which denotes what we ordinarily think of as the relation of identity. Thus, (L2 & I 2 ) is the 2-place property of loving another, which we ordinarily understand as holding between individuals x and y if and only if x loves y and x6=y . So the generalist claims that there is a domain of properties that instantiates the algebraic structure described by these term-functors. But what is the world like, according to the generalist? When we ordinarily describe a situation as being one in which someone loves someone, how does the generalist describe it? To answer this, consider the 1-place property cL2 just described, and consider the result of applying another c to get ccL2 . This is a 0-place property that we might ordinarily understand as a state of aairs, namely the state of someone loving someone. Now ccL2 is a term and, as such, it does not state anything about what the world is like. To state this, the generalist introduces to G the primitive predicate x obtains, which applies to any 0-place term to produce a sentence. Thus, when we 21

would ordinarily say that someone loves someone, the generalist says that ccL2 obtains. Similarly, when we would ordinarily say that someone loves someone unrequitedly, the generalist says that cc(L2 & L2 ) obtains. The generalists view, then, is that the fundamental facts of the world all have the form P 0 obtains where P 0 is a 0-place property. The term P 0 may of course be complex, formed from more basic terms along with applications of the term-functors. Signicantly, this view satises our three desiderata: in the attached appendix I outline how one can dene a logic of consequence on sentences of G, and a relation of equivalence between sentences of G and PL, in such a way that Preservation, Sufficiency, and Modesty can all be met. Therefore, given any sentence p of PL, the generalist can produce an equivalent sentence of G that, according to her, expresses the fundamental nature of a situation that we would normally describe with p. As a result, the generalist is in no way committed to IOI, and in particular she can make sense of the possible world characterized by sentence (A) six pages back, a possible world that the bundle theorist could make no sense of.28 The generalist and the individualist therefore paint radically dierent pictures of the material world. The individualist tells us that there is a domain of individuals propertied and related in a certain way; while the generalist tells us that there are states of aairs that obtain, where these states of aairs are composed purely out of properties. Unlike the bundle theorist, the generalist makes no attempt to construct individuals out of properties. But it was never part of her remit to do so. The generalists innovation is to say that the question What are individuals, if not primitive entities? is the wrong question; that we should instead be asking for an account of the fundamental facts of the world that makes sense of all possible general facts without appealing to primitive individuals. By providing an account, she carves away primitive individuals while doing no damage to the space of possible general facts that can hold, just as a Galilean spacetime carves away absolute velocity while doing no damage to the pattern of relative velocities that can be instantiated by a system. In this way, generalism is perfectly suited as a response to the problem of danglers.29 Note that generalism is a claim about the structure of the fundamental facts, and so is neutral on the status of sentences that may be thought to presuppose a domain of individuals. For example, consider everyday sentences such as
Incidentally, she expresses (A) with the sentence cc(F 1 & pF 1 & I ) obtains. This is not to say that the generalist dispenses with danglers all together, for she may think that there is such a thing as absolute velocity! But the point is that she goes some way towards a dangler-less world-view by dispensing with primitive individuals.
28 29

22

Ernest Hemingway is coming to tea, Someone is coming to tea, and philosophical sentences such as Perceptual experiences present us with a world full of individuals. If S believes that Ernest Hemingway is coming to tea, then S believes a proposition that contains an individual as a constituent. Given any one of these sentences, the generalist might be an error theorist and say that it is false because it presupposes a domain of individuals that does not really exist; or she might instead be a ctionalist and say that it is literally false but true of a ction of individuals; or she might be a reductionist and say that it is true in virtue of the underlying facts expressed by G; or she might be a non-reductive realist and say that it is true because it expresses the same thing as a sentence of G! Other options are available too, but the generalist need not take a stand either way. I granted in Section 1 that there would be an onus on the generalist to account for the possibility of de re belief and the fact that perceptual experiences present us with a world full of individuals. We can now see how this onus might be discharged, for as we have just seen generalism is (in and of itself) perfectly consistent with the truth of these claims.

3.3

The World is One

Before moving on to defend generalism from objections it will be important to note that, unlike individualism, generalism is radically holistic. To see why, it helps to rst appreciate why individualism is atomistic. Consider a situation that an individualist would describe as containing exactly two primitive individuals, one of which is F and stands in relation R to the other one which is G, and suppose neither has any other properties. According to individualism, the fundamental facts about the situation are facts like a is F, b is G, a bears R to b And these facts together would characterize the situation entirely. Some individualists would disagree with this last statement. They would argue that other facts are also needed to completely characterize the situation, such as the fact that a and b are distinct, the universal fact that everything in the situation is identical to a or b, and perhaps negative facts concerning the properties a and b do not have.30 Fortunately, there is no need for us to take a stand on this issue: whichever facts the individualist adds to her initial list, she will still think that any situation can be characterized by a (potentially large) number of facts that together add up to give
30

See Russell [28] and Armstrong [4], chapters 5 and 6, for views of this kind.

23

the situation as a whole. Since the situation as a whole is thought to be decomposable into parts, it is natural to call this an atomistic metaphysics.31 For the generalist, though, things are very dierent. How would she characterize the situation described above? One impediment to answering this question is that language G, in which she expresses the fundamental structure of the world, is foreign to most of us. However, Modesty says that any sentence of G is equivalent to a sentence of PL. So when discussing facts that the generalist believes to be fundamental and which she would express in G, it is harmless to express them in PL instead as a convenience to ourselves. Given this convenience, the generalist would say that the above situation is characterized by the following fact: (*) (x)(y)(Fx & Gy & Rxy & x6=y)32

Again, generalists may debate whether we would also need a conjunct in the matrix to specify that everything in the situation is identical to x or y , or negative conjuncts specifying what properties are not instantiated. But we need not settle the matter: whatever conjuncts the generalist adds to the matrix, she will not think that the situation is decomposable into atomic parts. After all, if she tried to replicate the individualists decomposition, she would think that the atomic facts were something like: (**) (x)Fx, (x)Gx, (x)(y)Rxy33

But unlike the individualists list, this leaves open whether something F stands in relation R to something else thats G, as is stated in (*), or the other way round. And it does not help to add other facts to (**), such as that (x)(y)x6=y, for the resulting collection would still not determine whether or not the situation as a whole is as stated in (*). As a result, the generalist thinks that the situation must be characterized by stating the single fact (*) all in one breath. Unlike the individualist, she cannot decompose the situation into atomic parts the sum of which give the situation as a whole. That is why I call generalism a holistic metaphysics. The smaller matters of fact listed in (**) still hold in the situation, but the generalist takes them to be derivative from the one fact (*). This is a radical departure from pre-theoretic intuition. When I see my laptop and my cup on the table, I intuitively see the situation as being composed of many facts: my laptop being on the table, my cup being on the table, and so on. Add these facts up, I naturally think, and you get the entire situation. But according to generalism this is an illusion: the situation
Russell was, of course, the classic atomist. See his [28]. For those readers familiar with G, the generalist will state it this as the fact that cc(F 1 & pG1 & R2 & I 2 ) obtains. 33 In G we would express these as the facts that cF 1 obtains, that cG1 obtains, and that ccR2 obtains.
32 31

24

is fundamentally speaking a single whole. Indeed, generalism implies the striking claim that, fundamentally speaking at least, there is only One Great Fact that captures our entire world all at once! Contrary to pre-theoretic intuition, the world is not the logical sum of smaller matters of fact such as those listed in (**); rather, those facts are to be seen as derivative from the One Great Fact. According to the generalism, the world really is One.34

In Defense of Generalism

Let us recap. Individualism, remember, is the view that among the fundamental facts of the world are facts concerning what primitive individuals there are and what they are like. But I argued in Section 1 that (1) It is a consequence of every physical theory considered over the past 400 years that primitive individuals are danglers. I also argued that it is a point against a metaphysics if it implies that the world contains danglers. As I put it there, (2) Consider two theories about the structure of the material world, and suppose you discover that the rst implies that the world contains a dangler while the other does not. All else being equal, it is rational for you to prefer the latter over the former. I just now outlined generalism, according to which the world does not contain primitive individuals. Therefore, (1) and (2) imply that all else being equal, it is rational to prefer generalism over individualismor at least, it is if anything like contemporary physical theory is to be believed. Of course, all else might not equal: generalism diers in many respects from individualism, so it might turn out that the virtue of dispensing with individuals is not worth the cost. But in this section I will argue that (3) The benet that generalism enjoys over individualism in virtue of dispensing with danglers outweighs its putative costs. Together, (1), (2) and (3) constitute my argument in favor of generalism. To defend (3), I will consider a number of putative costs that generalism might be accused of incurring.
34 Schaer has recently discussed a view he calls Monism (see his [31]). On this view there is just one individual, the world object; or, at least, if there are other individuals they are to be thought of as derivative from the world object, which is the fundamental entity. Generalism and monism share some similarities, but notice that monism is, at least on the face of it, atomistic: the entire situation concerning the world object can be decomposed into the fact that it is F, the fact that it is G, and so on.

25

4.1

Generalism is horribly complex.

It is often said that simplicity is a theoretical virtue, and it may be complained that generalism fails dramatically on this count. One observation that may encourage this complaint is that most of us nd it dicult to understand the language G (at least at rst) since it contains term-functors that we are not familiar with. But does this mean that we should reject generalism? No: when we say that simplicity is a theoretical virtue, we do not mean that the theory should be easily understood. As Laplace put it, The simplicity of nature is not to be measured by our conceptions.35 What, then, does simplicity as a theoretical virtue amount to? One thing it might mean is that a theory is simpler if its ontology contains fewer kinds of fundamental entities. But generalism scores well on this point, since its ontology are just properties. Another thing it might mean is that a theory is simpler if it contains fewer primitive bits of ideology. But generalism scores well on this point too, for G contains just six term-functors and the one primitive predicate. Yet another thing it might mean is that a theory is simpler if it contains fewer danglersbut of course generalism does better than individualism in this regard! I do not claim that this is all that simplicity could amount to, but it is worth noting that at least on these points generalism does as well as individualism, if not better.

4.2

Generalism ies in the face of pre-theoretic belief.

A second complaint one might have against generalism is that it revises our pre-theoretic conception of the world. It is worth getting clear on the charge. If generalism revises pre-theoretic belief, this is not because it implies that sentences like Ernest Hemingway is coming to tea are false. As I stressed in the last section, generalism is consistent with the truth of sentences like these; it just says that they are not true of the fundamental nature of reality. Instead, generalism is revisionary because it is holistic in the sense just outlined, while our pretheoretic conception of the world is atomistic. But does this give us reason to reject generalism? Not at all, for inquiry sometimes delivers surprises. The claim that the earth orbits the sun was once revisionary, but that in itself did not give our ancestors reason to reject it. What might be true is that we should not revise pre-theoretic belief without reason. But in this case we have a reason, namely that primitive individuals are danglers; according to (2), this is a reason to favor generalism.
35

Laplace, Exposition du System du Monde, quoted in Gribbin [17], p. 298.

26

4.3

Generalism is holistic, and this is a serious cost.

Grant that if the holistic nature of generalism is objectionable, this is not because it ies against pre-theoretic belief. But might it be objectionable on other grounds? Hawthorne and Sider claim that it is a condition of adequacy on any metaphysics that it is atomistic.36 But they do not say why this is so, and it is not obvious to me why holism should be ruled out by at. One might worry that holism is bad because it rules out the possibility of local causal explanations. These are explanations that explain a given fact, say that a window broke, in terms of facts concerning the windows immediate vicinity, such as that a baseball collided with it.37 The worry would be that if the world is holistic in our sense one cannot appeal to such localized matters of fact. But this worry confuses the sense in which a causal explanation is local, which concerns the distance between explanans and explananda, with the sense in which generalism is holistic, which concerns the logical structure of reality. Remember, even if the world is holistic in our sense, there are nonetheless general matters of fact concerning the trajectories of baseballs around windows, and holism allows that these are causally relevant to the fact that the window broke. All holism implies is that the world is not the logical sum of such facts.

4.4

Generalism cannot explain our pre-theoretic conception.

Our pre-theoretic conception of the world is atomistic, but generalism is radically holistic. I just argued that this in itself is not a signicant cost for the view, for inquiry regularly leads to surprises. But can the generalist explain why our pre-theoretic conception is so deeply ingrained in us? If not, I think this would indeed be a signicant cost. To appreciate the importance of the worry, compare our ancestors discovery that the sun does not really rise, that really the earth rotates on its axis. It is easy for us to forget how radical a break with pre-theoretic belief this discovery once was. Now, this break from pre-theoretic belief was not in itself a reason for our ancestors to doubt the discovery. But if the view that the earth rotates could not explain on its own terms why it looks to us as though the sun rises, I think this would have been reason to doubt the view. Of course, in that case an explanation is easily given. But in our case the worry is that we are missing something analogous: an explanation by the generalist of why it seems to us as though our world is atomistic when really it is not. However, I think an explanation can be given, and I shall nish the paper by outlining it. The key is that thinking and reasoning about a holistic
See Hawthorne and Sider [14]. A precise characterization of local causation would require much more ne-tuning, but this will do for our purposes.
37 36

27

world is cognitively demanding, at least for creatures like us. The ability to think about the world atomistically, as if it contained a domain of primitive individuals, is pragmatically valuable. As a result, it is no surprise that we nd it natural to think about the world in this way. To see this, it is instructive to consider a ctional example. Consider a subject who thinks in a language of thought: for this subject to believe that p is for her to believe a mental sentence s which means that p. And let us suppose that her mental language is G. According to the generalist, the structure of her thought mirrors the fundamental structure of the world. Now suppose she is investigating the properties of a physical system, and at an initial time t0 her belief about the system consists in a belief in a mental sentence of the form P 0 obtains. Once again, an impediment to discussing language G is that it is foreign to most of us. So I will pretend for convenience that our subject thinks in PL instead (remember, PL is predicate logic with identity but without constants). Given this convenience, we may suppose that at the initial time t0 her belief about the system consists in a belief in the mental sentence (P1) (x1 ). . . (x10 )(Tx1 . . . x10 )

where T is some reasonably complicated predicate. Now suppose at t1 she discovers that the system is in fact a little more complicated and she comes to believe the mental sentence (P2) (x1 ). . . (x10 )(Tx1 . . . x10 & Fx10 )

Notice that to update her prior belief, she must come to believe this entire sentence. It is not sucient for her simply to believe, say, the sentence (x)Fx in addition to (P1), for those two sentences do not entail (P2). The blame lies in the holism: since the situation cannot be decomposed into a collection of bite-sized facts, an agent cannot update her belief about one bite-sized fact, add it to what she already believes, and expect to have a unied view about the situation. Instead, our subject must update her state of belief by believing a sentence that represents her entire world-view all at once ! Clearly, this places severe computational demands on belief revision.38 But what if she were to think about the world dierently? For each variable xi she could introduce into her language of thought a constant, written ai , which purports to refer to an individual. The beauty of this innovation is that the new language allows her think about the system as
I am slurring over some subtleties here. For one thing, there might be identifying descriptions that would make her life easier. But there is little reason to think that such descriptions will be available in general. And in any case that method of information storage is extremely impractical: if our agent were to discover that part of the identifying description is false she would have to revise every single belief about the thing she thinks satises the description!
38

28

if it were decomposable into bite-sized facts. For at t0 her belief about the system would have consisted in her believing many sentences such as (C1) T1 a1 , T2 a2 , . . . , T10 a10 where the conjunction of these is the sentence T(a1 . . . a10 ). And at t1 she could update her beliefs by simply coming to believe the sentence (C2) Fa10 where (C2) is being used, along with the sentences listed in (C1), to conveniently represent the state of the system represented by (P2). Much better! So I suggest that if a subject really did live in the holistic world described by the generalist, it would be pragmatically useful to introduce constants into her language of thought that purport to refer to individuals. The resulting language would not accurately represent the underlying holistic structure of reality, but it would make her life a whole lot easier. I just said that our subject uses (C2) and the sentences listed in (C1) to represent the state of the system represented by (P2); and, more generally, that she uses constants to represent the state of a holistic system in a convenient manner. It is worth asking what role the constants would need to have in order to perform this task. And the answer is: exactly the role they have in the introduction and elimination rules for the existential quantier in ordinary predicate logic! Recall that these rules are:

-Intro

P(a/x) = (x)P

-Exit

(x)P

P(a/x) Q

(where -Exit requires that a does not occur in (x)P, Q, or any undischarged assumptions). Now suppose again that at the initial time t0 our agent initially believes the sentence (P1). To represent the system with constants, she reasons as follows: 1 2 3 (x1 ). . . (x10 )(Tx1 . . . x10 ) T(a1 . . . a10 ) Fa10 Initial belief (P1) at t0 Assumption Empirical discovery at t1

She can then continue to update her beliefs about the system, using the constants for as long as she wants. To recover what she has learnt in a form that is free of constants, she just needs to use the above rules: 29

4 5 6

T(a1 . . . a10 ) & Fa10 (x1 ). . . (x10 )(Tx1 . . . x10 & Fx10 ) (x1 ). . . (x10 )(Tx1 . . . x10 & Fx10 )

2, 3, &-Intro 1, 25, -Exit 4, -Intro

where this conclusion is, notice, the sentence (P2). I take this to be a striking conuence: the very role that constants would need to play, if they were to allow her to navigate conveniently through a holistic world, is precisely the role that they are deemed to play in contemporary deductive logic! My claim, then, is that it would be no surprise if our agent started to think in a language that purports to represent an atomistic world populated by individuals: thinking in such a language makes her life easier. This does not amount to a complete explanation of why our natural conception of the world is atomistic: the above story made various assumptions, and I am not proposing that at some point in history we introduced constants into our mode of thought. But the story helps make it unmysterious why creatures like us would end up thinking about the world atomistically, even if we really lived in the holistic world described by the generalist.39 I have discussed a number of putative vices that generalism might be accused of having, and in each case I have argued either that the accusation is incorrect, or that the putative vice is not such a vice after all. There may be other vices I have not considered, but this at least supports the view that (3) The benet that generalism enjoys over individualism in virtue of dispensing with danglers outweighs its putative costs.

Conclusion

We naturally think of the material world as being populated with individuals, but I have argued that this conception is not true of the fundamental structure of reality. Individuals considered as primitive entities are objectionable on precisely the same grounds as absolute velocity: the physical laws of motion imply that both are danglers. This does not imply that we should dispense with primitive individuals: we still need a reasonable theory to adopt in its stead. But I outlined a theory that does without them and defended it from objections. In particular, I argued that it has the resources to explain on its own terms why our untutored conception of the world is so deeply ingrained in us. So perhaps we should stop taking our untutored conception at face-value and start taking seriously the possibility that we think that way because it is useful to think that way, and not because it is getting at the underlying truth.
I should emphasize that the above story is not about the semantics of constants in our agents language of thought. The claim is just that the pragmatic usefulness of constants explains why they exist in her mental language in the rst place.
39

30

Acknowledgements
Thanks to David Brian Barnett, Ned Block, Jonny Cottrell, Hartry Field, Kit Fine, Paul Horwich, Geo Lee, Farid Masrour, John Morrison, Karl Schafer, Jonathan Schaer, Stephen Schier, Michael Schweiger, Ted Sider, Peter Unger, Seth Yalcin, and participants of the NYU dissertation seminar during the spring of 2008 for extremely helpful comments on earlier drafts of this paper.

References
[1] Adams, R. 1979. Primitive Thisness and Primitive Identity. The Journal of Philosophy 76 (1): 526. [2] Allaire, E. B. 1963. Bare Particulars. Philosophical Studies 14: 17. [3] Armstrong, D. M. 1997. A World of States of Aairs. Cambridge: CUP. [4] Armstrong, D. M. 2004 Truth and Truth-Makers. Cambridge: CUP. [5] Black, M. 1954. The Identity of Indiscernibles. In his Problems of Analysis, Ithaca: Cornell University Press. [6] Burgess, J. and G. Rosen. 1997. A Subject with No Object: Strategies for Nominalistic Interpretation of Mathematics. Oxford: Clarendon Press. [7] Campbell, J. 2002. Reference and Consciousness. Oxford, UK: OUP. [8] Cover, J. and J. OLeary-Hawthorne. 1998. A world of universals. Philosophical Studies 91: 205219. [9] Dasgupta, S. In Preparation. Symmetry and the Undetectable. [10] Earman, J. 1989. World Enough and Space Time: Absolute versus Relational Theories of Space and Time. Cambridge, MA: MIT Press. [11] Earman, J. 2002. Laws, Symmetry, and Symmetry Breaking; Invariance, Conservation Principles, and Objectivity. PSA 2002 Symposia. [12] Feynman, R. 1963. Lectures on Physics: Volume 1. Reading, MA: Addison-Wesley Publishing Company. [13] Hacking, I. 1975. The Identity of Indiscernibles. The Journal of Philosophy 72: 249256. [14] Hawthorne, J. and T. Sider 2002. Locations. Philosophical Topics 30: 5376.

31

[15] Hawthorne, J. and A. Cortens. 1995. Towards Ontological Nihilism. Philosophical Studies 79: 143165. [16] Hume, D. 1978. A Treatise of Human Nature. Edited by P. H. Nidditch. Oxford: Clarendon Press. [17] Gribbin, J. 2002. Science: A History. London, UK: Penguin Books. [18] Kuhn, S. 1983. An Axiomatization of Predicate Functor Logic. Notre Dame Journal of Formal Logic 24 (2): 23341. [19] Lewis, D. 1986. On the Plurality of Worlds. Oxford: Blackwell. [20] Locke, J. 1997. An Essay Concerning Human Understanding. Edited by R. Woolhouse. London: Penguin Books. [21] Maudlin, T. 1993. Buckets of Water and Waves of Space: Why Spacetime Is Probably a Substance. Philosophy of Science 60: 183203. [22] OLeary-Hawthorne, J. 1995. The Bundle Theory of Substance and the Identity of Indiscernibles. Analysis 55: 1916. [23] OLeary-Hawthorne, J. and J. Cover. 1996. Haecceitism and AntiHaecceitism in Leibnizs Philosophy. No us 30 (1): 130. [24] Paul, L. 2002. Logical Parts. No us 36 (4): 578596. [25] Quine, W. V. 1976. Algebraic Logic and Predicate Functors. In his The Ways of Paradox and Other Essays, Second Edition, pp. 283307. Cambridge MA: Harvard University Press. [26] Rodriguez-Pereyra, G. 2004. The Bundle Theory is compatible with Distinct but Indiscernible Particuars. Analysis 64 (1): 7281. [27] Roberts, J. 2008. A Puzzle About Laws, Symmetries and Measurable Quantities. British Journal for the Philosophy of Science 59: 14368. [28] Russell, B. 1985. The Philosophy of Logical Atomism. LaSalle, Ill: Open Court. [29] Russell, B. 1948. Human Knowledge: Its Scope and Limits. New York: Simon and Schuster. [30] Russell, B. 1948. An Inquiry into Meaning and Truth. London: Allen and Unwin. [31] Schaer, J. 2007. From Nihilism to Monism. Australasian Journal of Philosophy 85 (2): 175191. [32] Sider, T. 2006. Bare Particulars. Philosophical Perspectives 20: 38797. 32

[33] Strawson, P. F. 1959. Individuals: An Essay in Descriptive Metaphysics. London UK: Methuen. [34] Van Cleve, J. 1985. Three Versions of the Bundle Theory. Philosophical Studies 47 (1): 95107. [35] van Fraassen, C and Ismael. 2003. Symmetry as a Guide to Superuous Structure. In Symmetries in Physics: Philosophical Perspectives pp. 37192, edited by K. Brading and E. Castellani. Cambridge, UK: CUP. [36] Wigner, E. P. 1967. Symmetries and Reections. Bloomington, IN: Indiana University Press. [37] Wittgenstein, L. 1933. Tractatus Logico-Philosophicus. New York: Harcourt, Brace. [38] Zimmerman, D. W. 1998. Distinct Indiscernibles and the Bundle Theory. In Metaphysics: The Big Questions, edited by P. van Inwagen and D. W. Zimmerman. Oxford: Blackwell.

33

Appendix on Language G
First I will dene the syntax of language G and give a more complete and precise explanation of what the term-functors mean. Then I will dene a notion of logical implication on G and a relation of equivalence between G and PL. Third, I will outline the proof that the relation of equivalence satises Sufficiency, Modesty, and Preservation, as required. Finally, I will respond to an objection to the way these tasks are carried out.

The Syntax of G
The vocabulary of G consists of a countable set of atomic terms, each associated with an integer n 0, including a term I 2 associated with the number 2. If a term of G is associated with the integer n, we call it an n-place term. The vocabulary of G also includes six term-functors &, , c, p, , and . The set of terms are then dened inductively as follows: 1. All atomic n-place terms are n-place terms. 2. If P n and Qm are n-place and m-place terms respectively, then (b) cP n is an (n-1)-place term unless n=0, in which case it is a 0place term. (c) pP n is an (n+1)-place term, and (d) (P n & Qm) is a max(n, m)-place term. 3. Thats all; nothing else is a term. Finally, the vocabulary of G includes a single predicate, x obtains. A sentence of G is an expression of the form P 0 obtains, where P 0 is a 0-place term.40 (a) P n, P n, and P n are n-place terms,

The Term-Functors
In the text I gave a partial explanation of what the term-functors mean by saying things like if L2 is the 2-place property of loving, which we ordinarily understand as holding between individuals x and y if and only if x loves y , then (L2 & L2 ) is the 2-place property of loving unrequitedly, which we ordinarily understand to hold between individuals x and y if and only if x loves y and y does not love x. Here I shall give a more precise and general explanation of what the term-functors mean in this form. But there are two points worth keeping
One could expand G to include sentential connectives, variables and quantiers ranging over properties, modal operators and so on, but I shall not explore these expansions here.
40

34

in mind. First, I am explaining what the term-functors mean in terms that we ordinarily understand, namely in terms of individuals. But of course this is just an explanatory convenience, and we should keep in mind that the properties are ultimately to be understood independently of a domain of individuals. For a discussion of the legitimacy of this explanatory technique, see the end of this appendix. The second thing to keep in mind is that in the above quotation from the text the if and only if is not a material bi-conditionalif it were, it would not allow me to pick out the intended properties. I shall not discuss exactly what bi-conditional it is, but I take it that we understand the quotation above well enough to be getting on with. Now for the explanation of the term functors. Let P n be the n-place property we ordinarily understand as holding of individuals x1 . . . xn if and only if (x1 . . . xn ). And let Qm be the m-place property we ordinarily understand as holding of individuals y1 . . . ym if and only if (y1 . . . ym ). Then 1. P n is the n-place property we ordinarily understand as holding of x1 . . . xn if and only if it is not the case that (x1 . . . xn ); 2. (P n & Qm) is the max (n, m)-place property that we ordinarily understand as holding of x1 . . . xk if and only if (x1 . . . xn ) and (x1 . . . xm ), where k = max (n, m); 3. P n is the n-place property we ordinarily understand as holding of x1 . . . xn if and only if (xn x1 x2 . . . xn1 ); 4. P n is the n-place property we ordinarily understand as holding of x1 . . . xn if and only if (x2 x1 x3 . . . xn ); 5. If n 1, then cP n is the (n-1)-place property that we ordinarily understand as holding of x2 . . . xn if and only if there is something x1 such that (x1 . . . xn ); otherwise cP n is the 0-place property P n; 6. pP n is the (n+1)-place property that we ordinarily understand as holding of x1 . . . xn+1 if and only if (x2 . . . xn xn+1 ). Finally, I 2 is identity, the 2-place property we ordinarily understand as holding between x and y if and only if x is identical to y .

Predicate Functor Logic

Our task now is to dene a relation of logical consequence on G and a relation of equivalence between G and PL in such a way that we can prove Sufficiency, Modesty, and Preservation. We shall reduce this task to one already accomplished by dening a closely related language of predicate functor logic I will call Q.41 There is already a well-dened relation of
The name is in deference to Quine, who brought it to the attention of philosophers in his [25]. The language Q has been explored subsequently by Kuhn [18].
41

35

consequence on Q and a relation of equivalence between Q and PL for which analogues of these three theorems hold, and we will use this fact to induce corresponding relations on G and prove our theorems. Syntactically speaking, the language Q is just the same as G with the exception that (i) it does not contain Gs single predicate x obtains, and (ii) its expressions are written in normal font rather than the bold font of G. The terms in G are, when written in the normal font of Q, to be understood as predicates instead. More precisely, the vocabulary of Q consists of a countable set of atomic n-place predicates, including a 2-place predicate I2 , and six predicate functors , &, c, p, and . An expression of the form Pn is an atomic n-place predicate of Q if and only if P n is an atomic n-place term of G; an expression of the form Pn is an n-place predicate of Q if and only if P n is an n-place term of G; and so on. We can dene a notion of logical implication on Q model-theoretically. Without loss of generality, we can assume that the predicates of Q are the predicates of PL. So we can give a semantics for Q using the standard models of PL (i.e. predicate logic with identity but without constants).42 These models are pairs of the form M = (D, v ), where D is a non-empty set and v is a function from n-place predicates (other than I2 ) to sets of n-tuples of D. Given a predicate Pn of Q, a model M = (D, v ) and a sequence d = (d1 , d2 . . . ) of elements of D, we say that d satises Pn in M , written d|=M Pn , if and only if 1. If Pn is atomic then (d1 ,. . . , dn ) v (Pn ), 2. If Pn = I2 then d1 = d2 , 3. If Pn = Qn then it is not the case that d|=M Qn , 4. If Pn = (Qm & Rr ) then d|=M Qm and d|=M Rr , 5. If Pn = cQn+1 then there is an x D such that (x, d1 , d2 ,. . . )|=M Qn+1 , 6. If Pn = pQn1 then (d2 , d3 ,. . . )|=M Qn1 , 7. If Pn = Qn then (d2 , d1 , d3 , d4 ,. . . )|=M Qn , and nally 8. If Pn = Qn then (dn , d1 , d2 ,. . . , dn1 , dn+1 ,. . . )|=M Qn . We can then say that Pn is true on M , written |=M Pn , if and only if d|=M Pn for all sequences d = (d1 , d2 ,. . . ) of elements of D. With this semantics in place, it is now easy to dene a relation of logical implication on Q and a relation of equivalence between Q and PL. First, the relation of logical implication can be dened as follows: a predicate Pn logically implies a predicate Qm , written Pn |=Qm , if and only if for any model M , if |=M Pn then |=M Qm .
42

The following semantics is borrowed from Kuhn [18].

36

Now for the relation of equivalence between Q and PL. Remember, the models M = (D, v ) are models of PL too. So let us assume that the notion of an n-ary formula (x1 . . . xn ) of PL being true on a model M , written |=M (x1 . . . xn ), is dened in the normal way.43 Then we can say that an n-ary formula (x1 . . . xn ) of PL and a predicate Pn of Q are equivalent if and only if for every model M , |=M (x1 . . . xn ) if and only if |=M Pn . For our purposes the important results are these: Q-Sufficiency Every n-ary formula of PL is equivalent to some n-place predicate of Q. Q-Modesty Every n-place predicate of Q is equivalent to some nary formula of PL.44 With these results in hand, our task is now easy.

Implication, Equivalence, and our Three Theorems

First, we use Q to dene a relation of consequence on G and a relation of equivalence between G and PL. Then we prove our three theorems. The relation of logical implication on Q naturally induces a relation between terms of G: P n Qm if and only if Pn |=Qm . In turn, this induces a relation of logical implication on sentences of G: P 0 obtains logically implies Q0 obtains, written P 0 obtains |= Q0 obtains, if and only if P0 Q0 ; that is, if and only if P0 |=Q0 . (The sign |= is being used for logical implication in both languages, but I take it that no confusion will result.) The relation of equivalence between Q and PL also induces a relation of equivalence between G and PL: a sentence P 0 obtains of G is equivalent to a sentence p of PL if and only if P0 is equivalent to p. Preservation is now easy to prove. Suppose a sentence P 0 obtains in G is equivalent to a sentence p in PL, and suppose a sentence Q0 obtains in G is equivalent to q in PL. We must show that P 0 obtains |= Q0 obtains if and only if p|=q . Now, P 0 obtains |= Q0 obtains if and only if P0 |= Q0 (by denition of implication in G). But P0 is equivalent to p and Q0 is equivalent to q (by our hypothesis and the denition of equivalence between G and PL), and therefore P0 |=Q0 if and only if p|=q (by model-theoretic reasoning). So P 0 obtains |= Q0 obtains if and only if p|=q , as required.
43 Roughly speaking, we say that the formula is true on the model if and only if it is satised by every sequence in the model. To dene the notion of a formula being satised by a sequence in the model, we need a way to associate free variables in the formula with objects in the sequence. Here we can simply use the convention that the ith free variable is associated with the ith object in the sequence. 44 To prove these, dene translation functions from n-ary formulas of predicate logic to n-place predicates of Q and vice versa, and then prove that the value of each function is always equivalent to the argument. For one example of translation functions like this see Kuhn [18].

37

Finally, Sufficiency and Modesty follow straight from Q-Sufficiency and Q-Modesty along with the above denitions.

Using a Domain of Individuals

Earlier in the appendix and in the text I explained what the term-functors of G mean by appealing to our ordinary understanding of properties being instantiated by a domain of individuals. But the generalist claims that individuals are not constituents of the fundamental facts of the world. So was that method of explanation legitimate? Similarly, I just used the semantics for Q to dene a relation of implication on G and a relation of equivalence between G and PL. But the semantics for Q used ordinary models of PL of the form M = (D, v ), where D is a domain of individuals. Was that legitimate? I think it was. The key is to distinguish between metaphysical and conceptual priority. Metaphysically, the generalist claims that the fundamental facts of the world are those expressed by G. But conceptually, I have explained the view to you in terms that you are familiar with, namely in terms of individuals. Once you are competent with G you can then, if you so wish, become a generalist by giving up your understanding of G in terms of a domain of individuals and simply taking the term-functors and predicate of G as undened primitives. There is no space to defend the cogency of the strategy here in full. I shall only plead innocence by association, for there are many other cases in which we adopt it. The case of velocity is, unsurprisingly, a good example. Our ordinary concept of the relative velocity between two material things is dened as the dierence between their absolute velocities. Once you are competent with the concept, you can then (if you so wish) dispense with absolute velocity and either think of relative velocity as a primitive concept or dene it in other terms. And this is precisely what we do when we adopt a Galilean conception of space-time.

38

Metaphysical Rationalism
Shamik Dasgupta
Draft of March 2012

Those who are ignorant of true causes, make complete confusion think that trees might talk just as well as menthat men might be formed from stones as well as from seed... B. Spinoza, The Ethics, 1p8s2 A metaphysical rationalist is someone who endorses the Principle of Sufcient Reason (PSR), the principle that everything has an explanation. The PSR can be understood in a number of different ways depending on the meaning of explanation and the range of the quantier everything, so metaphysical rationalists come in many different stripes. But every rationalist thinks that in some interesting sense there is an explanation (or reason) for everything. Her rationalism is then qualied as being metaphysical to distinguish it from other rationalisms (such as epistemic rationalism, a thesis concerning the sources of justication) but this qualication can now be dropped for brevity. Although rationalism of one form or another was embraced by such luminaries as Aquinas, Spinoza, and Leibniz (amongst others), it is fair to say that it has since largely fallen out of favor. Indeed my sense is that (with a few exceptions) contemporary philosophers tend to treat it as an antiquated view that need not be taken seriously today.1 This may be the result of two factors. First, the PSR (in one form or another) is often thought to have consequences that might be thought problematic, including necessitarianism (the view that every truth is necessarily true), the existence of God, and indeed the existence of every conceivable being.2 And second, there is a perceived lack of evidence in its favor. These two factors naturally lead to an attitude of (perhaps respectful) dismissal.
notable exceptions include Della Rocca ([13] and [12]) and Pruss [32], who have each argued for rationalism recently. Lin [29] and [30] has recently discussed rationalism in some detail and defends it from an important objection in his [29], though I do not know whether he ultimately endorses the view. 2 To be clear, it is (as we will see) somewhat controversial whether the principle does indeed have these consequences. For example, Lin [30] raises difculties with the claim that they follow from Spinozas understanding of the principle. Still, they are all consequences that have (at one time or another) been associated with the PSR.
1 Two

This attitude may be justied when it comes to some versions of the PSR. But I believe that there is a version that does not deserve it. One way that this version differs from others is that it is just concerned with one specic mode of explanation, what is nowadays referred to as metaphysical or grounding explanation. I will argue that this version avoids the problematic consequences just mentioned. It does not imply the existence of God or every conceivable being. It does imply necessitarianism of a certain kind, but I will argue that this kind of necessitarianism is not problematic. But is there any evidence in its favor? If there is, I confess that I have none to offer. But is this a reason to dismiss it? It is if rejecting the PSR is the default view, if the burden of proof is on the rationalist to establish her position but not on her opponent to establish hers. This might be true of versions of the PSR that concern causal explanation: absent strong evidence in their favor we should probably reject them. But there is an important difference between causal and metaphysical explanation that means that when it comes to the version I want to discuss there is no default view: there is a burden on both those who accept it and those who reject it to offer evidence for their view, so that absent evidence one way or the other we should remain agnostic. Since I do not know of much evidence either way I suggest that a humble agnosticism is the rational attitude. If this is right then the contemporary anti-rationalist bias is unwarranted. I stress that I do not know whether this version of the PSR is true and I will not try to demonstrate that it is true. My aim is just to formulate it (sections 15), explore its consequences (sections 612), and then argue that the contemporary bias against it is unwarranted (sections 1317). Still, this is a valuable exercise if only because the version of the PSR I develop is the anti-thesis to a metaphysical picture that is something of a contemporary dogma. This dogma has two components: that some aspects of the world are metaphysically brute or arbitrary, and that the world could in so many ways have been different (we will see how these are connected). Lewis doctrine of Humean Supervenience is the zenith (or nadir) of this dogma, but it is found to some degree in most contemporary metaphysics. The rationalist view defended here is an alternative picture. I believe that it resembles a Spinozistic view, but I am no scholar so I will develop it in contemporary terms without attempting to root it in history. Whether this rationalist alternative is true is a question for another time: here my aim is just to show that there is a respectable alternative. Even if we ultimately decide to stick with dogma, we will at least better know our opponent.

Metaphysical Explanations

Our rst task is to formulate the version of the PSR that I have in mind (sections 15). I said that the PSR can be understood in a number of dif2

ferent ways depending on the meaning of explanation and the domain of the quantier everything. Let us start with the notion of explanation. Some versions of the PSR might focus on causal explanation, perhaps stating that everything has a causal explanation.3 Other versions might focus on teleological explanation, perhaps stating that everything has a purpose. Yet another version interprets the reason in Principle of Sufcient Reason to be a practical reason and then states that God had a practical reason to create the world exactly as she did.4 Generalizing form these specic theses, some versions are disjunctive and state that everything has an explanation of either this or that sort.5 And one might be even less specic and just state that everything has an explanation of some kind or other.6 I will not discuss any of these versions of the principle. The version I will discuss specically concerns the notion of metaphysical explanation. This notion is often discussed under the label of ground. To illustrate, imagine explaining why Europe was at war in 1939. A causal explanation might describe a sequence of events over the preceding 50 years that led to the conict. But there is another kind of explanation that would try to say what goings on in Europe at the time made it count as a continent at war in the rst place. Regardless of what caused the conict, someone in search of this second explanation recognizes that war is not a sui generis state and that there must therefore be something about the continent in virtue of which it counted as being a continent at war. A plausible answer is that it was at war in virtue of various facts about how its citizens were acting, (e.g. some were committing politically motivated acts of violence against others). An explanation of this second kind is a metaphysical explanation, or (as it is often put) a grounding explanation. What is the logical form of a grounding explanation? Explanations are typically expressed with the sentential operator because: Europe was at war in 1940 because its citizens were acting in certain ways. So one view is that the logical form of a claim about ground is because where is a sentence, is a list of sentences, and because is read in the metaphysical rather than causal sense.7 Informally, the sentences in describe those aspects of the world that together (fully) explain its being
3 Pruss ([32], chapter 13) and Weaver ([40] and[41]) have recently argued for a principle roughly along these lines, namely that all contingent events have causes. 4 This might be along the lines of what Leibniz had in mind by the principle. 5 For example, one of the principles that Rowe nds in Clarke states (in effect) that the existence of every being is explicable either by the causal effect of another being or by its own nature. See Rowe [36]. 6 My sense is that Pruss [32] is interested in defending a principle along these lines (though he restricts it to contingent facts). Though as I said in footnote 2 he has also argued for more specic principles. 7 This is how Fine characterizes the logical form in his [17] and [19].

the case that . Alternatively one may reify and treat ground as a relation between facts. On this approach the logical form becomes the Xs ground Y where Y is a singular variable and the Xs is a plural variable, both ranging over facts.8 This second approach leads to easier prose so I will use it here. But strictly speaking I think of ground as a sentential operator so my reference to facts in what follows is merely a convenient shorthand. I will assume that ground is transitive in the following sense: if the Xs ground Y and if Y along with the Y*s ground Z, then the Xs along with the Y*s ground Z. This is not beyond dispute but the issue is not important for our purposes: if you deny that ground is transitive you can take me to be discussing the transitive closure of your notion. Ground is a philosophically important mode of explanation. When a physicalist says that mental states arise out of or are determined by physical states it is tempting to see this as a claim about grounds, about what explains our mental states in this metaphysical sense. Or when normative naturalists tell us that they can account for normative facts in purely natural terms it is tempting to understand this as a claim about what grounds the normative. One might of course question whether the word ground (so used) corresponds to anything real, just as one might raise the analogous Humean question about cause. But that issue is beyond the scope of this paper.9 Here let us take the notion for granted and ask what kind of PSR can be formulated in its terms.

On Innocence and Experience

One such principle is that everything has a ground. But that is not what I want to focus on. I am instead interested in the principle that everything has a special kind of ground. What is this special kind? Grounding explanations start with why-questions (asked in the metaphysical sense of why). Suppose I ask why there is a mountain here in front of me and you say that it is because some particles are arranged thus and so. And suppose that what you said is true. To what extent have you answered my question? Certainly there is a sense of explained in which you have explained why there is a mountain here. But of course the question then arises as to why those particles are arranged thus and
is how Rosen characterizes the logical form in his [35]. Strictly speaking, one might want to allow that ground is non-distributively plural, in the sense that sometimes a collection of facts are together grounded even though none have a ground on their own (see Dasgupta [9]). But I will ignore this complication here. 9 See Hofweber [24] for skeptical arguments against the notion of ground. See Raven [33] for a defense of the notion.
8 This

so in the rst place. Suppose you now tell me that there is no explanation, that this is just a brute fact. Then in another sense you have not explained why there is a mountain here after all. For you have not explained why the world is one that gives rise to a mountain rather than one in which the particles are arranged differently and give rise (say) to an ocean instead. That is, if the arrangement of particles itself has no explanation it remains unexplained why the world turned out like this (gesturing at the mountain and the particles). In that sense my original question of why there is a mountain here remains unanswered. Put guratively, it has not been explained why, out of all the logically or metaphysically possible ways that things could have turned out, it turned out to contain a mountain here. This is the kind of situation that the rationalist I have in mind rules out. The problem is not so much that there is an ungrounded fact about the arrangement of the particles. The problem is rather that because of this the original fact of the mountains being here has not been explained in the way that the rationalist wants. The desire for this special kind of explanation stems from a child-like curiosity in which one looks at the surrounding mountains and oceans and football matches and operatic performances and thinks Good grief, how come it all turned out like this? Being told that the mountain is here because some particles are arranged thus and so does not (even if true) answer the question. This child-like curiosity is easily lost with experience, when what once lled us with awe begins to seem mundane, but the rationalist takes the childs question seriously and expects an answer. We might recall our child-like innocence by thinking about cases of ground relating to the logical connectives. Suppose that the child learns that there is a red rose and asks why (in the metaphysical sense) this remarkable fact obtains. She might be told that since existentially quantied facts are (in general) grounded in their instances it is because this particular stem is a red rose. But even if the instance grounds the existential fact and so explains it in some sense there is clearly another sense in which the childs question remains unanswered. For she wanted to know (as it might be put) why, out of all the ways the world could have been, it turned out to be like this, a world that contains a red rose. It is no answer to this question that it is because this particular stem is a red rose!10 What would answer the childs question as to why there is a mountain here? Suppose we said that the mountains existence is grounded in the arrangement of particles, which itself is grounded in some physical eld, which in turn is grounded in something else. . . and so on without end. Would citing some non-terminating descending chain of grounds like this
10 Similar points apply to conjunctions, disjunctions, and universal quantications. Even if conjunctions are grounded in their conjuncts, merely stating those conjuncts does not answer the childs question of why the conjunction obtains.

answer her question? I do not think so. For her question is clearly not answered at the rst step in the chain when one describes the particle arrangements, since (as we have seen) she will just complain Yes, but why is the world like that? But the same goes for any step in the chain. So all we have in a non-terminating descending chain is innitely many bad answers. And it is doubtful (to say the least!) that putting innitely many bad answers together somehow produces a good answer. For it seems perfectly intelligible for the child to gesture at the mountain and the particles and the elds and all the subterranean material involved in the descending chain and ask why it all turned out like that, i.e. arranged in such a way as to give rise to a mountain.11 Nor, I think, would the childs question be answered if we grounded the mountains existence in a self-explanatory fact, i.e. a fact that grounds itself. For the situation is just a kind of non-terminating descending chain and so for the above reason the childs question appears to remain unanswered. And in any case it is not at all clear whether the proposed explanation is even coherent: it is dubious whether it makes any sense for a fact to explain itself in the very same sense that the arrangement of particles might explain the presence of a mountain. A natural thought is that to answer the childs question one must (at a minimum) explain why there had to be a mountain here. Anything less than this (the thought is) will result in the child wondering why the world happened to turn out the way it did. So one might think that a necessary condition on answering the childs question of why F obtains is that one states some necessary facts that ground F. But even if this is right it is not a sufcient condition. For suppose the question then arises as to why those necessary facts obtain and suppose one is told that there is no answer, that they are just brute. Then the child will complain that it has still not been explained why the world turned out (indeed had to turn out) to be like this, a world that gives rise to mountains. If the original explanation in terms of contingent particle arrangements did not answer the childs question, moving to necessary facts does not help.12 What then would a satisfactory explanation look like? The previous explanations fell short because whenever they grounded the mountains existence in some underlying facts (contingent or necessary) the question always arose as to why those underlying facts obtain. Insofar as this further question arises the child has the unsatised feeling that it has still not been explained why the world turned out to contain a mountain here. But this unsatised feeling will not arise if the mountains existence is explained in terms of facts for which this further question of why they obtain does not even arise in the rst place. Roughly speaking then let us
11 Bliss 12 This

[5], section 4, makes a related point. is why the modal description of the childs question was merely gurative.

call a fact substantive if it is apt for being grounded, if the question of what grounds it can legitimately be raised and admits of a sensible answer (an answer that either states its ground or else states that it has none). In contrast, a fact is autonomous if it is not apt for being grounded in the rst place, if the question of why it obtains does not legitimately arise. If we explain the mountains existence in terms of autonomous facts then the question as to why those underlying autonomous facts obtain does not even arise and so there is no further question as to why the world turned out like this, i.e. set up in such a way as to produce mountains. The childs question of why there is a mountain here will then be answered after all. This is not to say that there always is an answer to the childs question. My claim so far is just that there is a way of asking Why P? that we seem to understand and that is properly answered only by appeal to autonomous facts. The version of the PSR I will defend is that there is indeed always an answer to these childs questions, or more precisely that every substantive fact has an autonomous ground. But we will assess the view later, for now I am just establishing what the view is.

Autonomy

What does it mean to be autonomous? The notion is obviously central to the PSR under investigation so it is worth clarifying. I will not try to dene it (in part because I suspect that it has no denition). But something can be said to clarify the notion I have in mind, starting with a general characterization and then turning to examples. The basic idea is reasonably familiar in the case of causal explanation. Consider facts about how various particles are arranged. Some particle arrangements (we may assume) have a causal explanation. The arrangement that constitutes my laptop in its current state might be an example. Other particle arrangements may not. If I ask what causally explains why a certain group of particles were arranged in a certain way and if the denoted arrangement happens to have been the initial condition of our universe then the appropriate answer would be that there is no explanation, that it is a causally brute fact about the worlds initial state. But all particle arrangements are apt for causal explanation: the question of what causally explains how they came to be arranged like that can legitimately be raised even if the answer is in some cases nothing. Consider by contrast the fact that 1+2=3. Like the initial condition, this lacks a causal explanation. But there is a difference. If someone asked you what causally explains why 1 and 2 came to equal 3 you would not just say nothing, you would start talking about the notion of causation and the nature of abstract objects in an attempt to show that the question should not have been raised in the rst place. The particle arrangement 7

that happened to be the initial condition lacks a causal explanation even though it is a good question why they came to be arranged like that. The mathematical fact by contrast lacks one because it is not apt for being causally explained in the rst place. The result is a three-fold distinction. There are those facts that are apt for causal explanations but lack onee.g. the initial conditions. And there are those that are apt for causal explanations and have onee.g. the current arrangement of particles that constitute my laptop. And then there are those that are not apt for causal explanation in the rst placee.g. facts of pure arithmetic. My thought is that an analogous three-fold distinction can be drawn in the case of grounding explanations too. There are those that are apt for having a ground but lack one. These are the so-called fundamental or brute facts. And there are those that are apt for having a ground and have one. These are the so-called derivative facts. And nally there are those that are not apt for having a ground in the rst place. These are what I call autonomous. That is one route into the notion of autonomy. But it might help to triangulate onto the notion with a second analogy, an analogy with proof (this will also help us to nd examples of autonomy). Suppose one is doing standard, axiomatic set theory. One starts with a language containing the binary predicate of set-membership and one states some axioms governing it and proves some theorems. It is then useful to expand the language and introduce by stipulation a new word: Denition: x is a subset of y d f any member of x is a member of y. One can then express various claims about subsets, for example Proposition: There is an x and a y such that x is a subset of y but not vice-versa. and attempt to prove them from the axioms and the denitions. Now the question of whether the Proposition has a proof is clearly a good question regardless of whether it actually has one. Indeed this kind of question is the bread and butter of set theory. The same applies to the axioms of the system: the question of whether one of them is provable from the rest is a good question that can lead to remarkable advances even if the answer is nojust think of the attempt to prove Euclids 5th postulate! But the same is not true of the denition itself: the question of whether we can prove the denition from the axioms is clearly not a good question. If your homework buddy asked you whether you can prove the above denition you would take her to be confused about its status as a denition. You would not just say no, you would start talking about the role of the denitionthat its function is to introduce by stipulation a new term to our language, and so onso that your buddy comes to see that the request 8

for a proof is in some sense illegitimate. As we might put it, the denition is not apt for being proved. There may be other good questions to ask of the denition such as whether it is useful or corresponds to some prior concept we had in mind. But how it might be proved from the set-theoretic axioms is not one of them.13 So as in the case of causation we recognize a three-fold distinction. There are those truths that are apt for proof but lack one (if your derivation system is complete then these are just the axioms). Then there are those truths that are apt for proof and have one. And nally there are the denitions that are not apt for proof in the rst place. Autonomous facts are those that stand to ground as denitions stand to proof. What are some examples? The analogy with proof is suggestive. For a stipulative denition like the above states what a term means. And the worldly analogue is a statement of what something isa statement (that is) of its nature or essence. When we ask in the philosophy seminar room What is knowledge? and consider the answer that knowledge is true and justied belief, we are considering a claim about what knowledge is, about the essence of knowledge. And when we ask in a chemistry lab What is water? and consider the answer that water is a substance composed of H2 0 we are again considering a claim about what water is, about its essence. Regardless of whether these claims are correct we seem to understand them reasonably well. Let us follow Fine and take the logical form of these claims to be It is essential to x that where x is an item of any ontological category and is a sentence.14 I will take this to be synonymous to saying that it is in the nature of x that or that it is part of what x is that .15 So essentialist truths are the worldly analogue of denitions. And it seems to me that they do indeed stand to ground as denitions do to proof: the question of what grounds them strikes us as illegitimate in something
course in another axiomatic system the word subset might be dened differently, in which case the material equivalence x is a subset of y iff any member of x is a member of y might have a (non-trivial) proof from those other axioms and denitions. Moreover one might construct a logic of denition, i.e. a logic of the d f operator, and one might conceivably then show that the denition in the text is provable from the axioms governing d f . But the point is just that when doing standard axiomatic set-theory, the question of whether the denition in the text can be proved is intuitively a bad question. 14 This is how Fine regiments talk of essence in his [15]. 15 Note that it may only be part of what x is that , so we should not presume that uniquely species x. Nor should we presume that there is a collection of such that does uniquely specify x. It might be that the only answer to the question of what canary yellow is, is that it is a determinate shade. This is why I use the term essence rather than real denition, since the latter suggests that one can state what x in a way that uniquely species it.
13 Of

like the way that the question of how one might prove a denition does. For suppose that it is essential of knowledge that someone knows only if she truly and justiably believes. And suppose someone asks what explains this (in the metaphysical sense). In virtue of what (the question is) is it part of what knowledge is that someone knows only if she truly and justiably believes? It is difcult to know how to respond. One is tempted to say that this is just what knowledge is. . . but of course this is what we were asked to explain! In saying this one is most naturally heard not as trying to explain this fact about knowledge in any serious sense but rather as deecting the demand for explanation. Or suppose that it is essential of water that it is a substance composed of H2 O and suppose someone asks what explains this (in the metaphysical sense). The question is not why we should believe it or how water came to refer to what it does, there are good answers to both those questions. The question is rather: in virtue of what is it part of what water is that it is composed of H2 O? It is again hard to know what to say other than that is just what water is! And in saying this one is again most naturally heard as sidestepping the question rather than giving it a serious answer. Compare these essentialist facts to (say) the fact that some particles are arranged thus-and-so. This might (for all we know) be grounded in the undulations of a quantum wave-function in a massively high-dimensional Hilbert space.16 Or it might be a brute, ungrounded fact about the world. But either way we consider it a perfectly good and legitimate question why they are so arranged even if the answer turns out to be no reason. So we appear to distinguish the brute facts from the essentialist facts. Both are groundless, but there is a difference. The former are (by denition) apt for being grounded, it is just that they lack a ground. By contrast, the latter are not apt for being grounded in the rst place, in roughly the same sense that arithmetic facts are not apt for causal explanation and that denitions are not apt for proof. I am using autonomous to label the latter category, and brute or fundamental to mark the former. Thus the brute facts play the same role vis a vis metaphysical explanation as the initial conditions play vis a vis causal explanation and the axioms play vis a vis proof. The autonomous facts play a different role, one more analogous to the role that denitions play in proof.17 The claim (to be clear) is not that it is autonomous that someone knows
Albert [3] for a view like this. should say that the analogy with causation is (like any analogy) not perfect. I said that arithmetical facts are not apt for causal explanation but they also (arguably) do not causally explain anything. In contrast, even if essentialist facts are autonomous they certainly ground other facts. For example, a conjunction of two essentialist facts will (like any conjunction) be grounded in its conjuncts. Still, the point of the analogy is to bring out the similarity between arithmetic facts and essentialist facts vis a vis the question of what explains them regardless of what (if anything) they explain.
17 I 16 See

10

only if she truly and justiably believes. For that may be grounded in the fact that that is part of what knowledge is. More generally one might endorse what Rosen called Essential Grounding: If it is essential to x that , then because it is essential to x that .18 So the essentialist fact claimed to be autonomous is not the fact that but rather the fact that it is essential to x that . In each caseof ground, causation, and proofI have argued that we nd intelligible the notion that certain facts are not apt for ground or causal explanation or proof (respectively). I have done this by pointing out that we respond to certain questions in a distinctive way (e.g. that we respond not by answering them but by pointing out that they should not have been asked in the rst place). This has led at times to me glossing the notion in epistemic or cognitive terms, for example in terms of certain questions not making sense or being illegitimate. But the notion of autonomy is not dened in epistemic or cognitive terms. Our distinctive response to these questions is evidence that we nd the notion intelligible, but is not necessarily that in terms of which the notion is to be dened.19 To be sure, this evidence is not conclusive and one might still refuse to nd the notion of autonomy intelligible. In response I believe that the intelligibility of the notion can be further motivated by showing that it is of great theoretical utility, but that is a long story that there is no space to tell here.20 And though there is more to say about the notion, I think we can do quite a lot with the rough-and-ready characterization just given. So let us grant that the notion is intelligible for the sake of argument.
[35]. might propose a link between autonomy and some epistemic notion, for example that a fact is autonomous iff it is apriori or conceptual, but even if true this would be a strong thesis in need of motivation. Della Rocca (on one reading) defends something like the view that a fact is autonomous iff it is conceptual in his [12]. I am inclined to think that the case of water discussed above is a counterexample to this thesis but I will not discuss this further here. 20 The theoretic utility I have in mind is (in rough outline) this. One of the central theoretical roles for the notion of ground is to make sense of -isms like naturalism, phenomenalism, physicalism, and so on. The rough idea is to understand (say) physicalism as the view that all facts are grounded in purely physical facts. But there is a problem with this rough idea that the notion of autonomy can be used to solve. The problem arises when we ask what grounds those facts that detail how various (say) mental facts are grounded in physical facts. According to physicalism as understood above they must also have a purely physical ground, but it turns out (and this is part of the long story) that they do not. The solution is to see that our intuitive understanding of physicalism is a restricted claim: that all facts of a restricted kind are grounded in purely physical facts, where the grounding facts themselves are not of this restricted kind. Of course the challenge is then to say what the restriction is and why it is not ad hoc. I believe that we can meet the challenge by using the notion of autonomy. If this is right then the notion of autonomy plays a central role in making intelligible some of the most important questions of philosophy. I argue for all this in Dasgupta [10].
19 One 18 Rosen

11

Essence and Autonomy

Still, even if one grants that the notion of autonomy is intelligible, one might reject my suggestion that essentialist truths are autonomous. One might for example point out that ingenious philosophers have proposed general theories about essentialist truths, for example that if something is an essentialist truth then that is in virtue of its being a necessary truth of some sort. If essentialist truths are autonomous then the attempt to give this kind of theory is deeply confused in something like the way that attempting to prove a denition is. And even these theories are false (the objection is) they are surely not confused in that way.21 But if there is a confusion here then I believe that it is easily explicable. It arises from the following picture: one starts with the idea that there are two ways to have a propertyan essential way and an accidental wayand one then takes the essentialist facts about something to be facts concerning which properties it has in the essential way. On this picture the essentialist facts just appear to be facts concerning which properties are had in that way by a given domain of things. And such facts then look to be substantive if anything is: surely we can ask why (in the metaphysical sense) the given object has that property in the essential way. But this picture is deeply misleading: on my conception the essentialist facts concern what those things are in the rst place. It is not that there is some independently given domain and the essentialist facts are certain facts about what properties they have, it is rather that the essentialist facts specify what the domain is in the rst place. It is those kinds of facts that strike me as obviously autonomous.22 So my diagnosis is that those who have been tempted to provide grounds for essentialist facts have been in the grip of the picture just described. Insofar as we all mean the same thing by essence I claim that the picture is deeply misleading, albeit understandable. But it is also possible that the word essence has been used by others to mean something different in which case the above picture may be accurate of their notion and there may be no disagreement. Of course insofar as I am leaning on a particular reading of the term essence one might question whether the reading corresponds to anything real. That is a big question that is beyond the scope of this paper and (as with the notion of ground) I will be taking this particular notion of essence
to Tobias Wilsch for pressing this objection on me. course given my notion of an essentialist fact one may dene the notion of having a property essentially: x essentially has the property of being F d f it is part of what x is that x is F. In this way we can make perfectly good sense of the idea that there are two ways to have a property and that the essentialist truths are those regarding which which properties things have in the essential way. But we are misled if we forget that the primary notion here is that of a specication of what something is. This is why I call the above picture misleading and not false.
22 Of 21 Thanks

12

for granted. Our question is then what kind of rationalist view can be formulated in its terms. When I say that essentialist facts are autonomous it is important that by an essentialist fact I mean a statement of what something is in its most core respects. Following Fine we can then extend this notion in a number of ways. We can (for example) chain essences together to get mediated essences. If it is essential to knowledge that knowledge is true and justied belief, and if it is essential to truth that truth corresponds to the facts, then it is mediately essential to knowledge that knowledge is justied belief that corresponds to the facts. Now it is implausible that mediated essences are autonomous. For if it is mediately essential to knowledge that knowledge is justied belief that corresponds to the facts then that is presumably because of the two statements of (core) essence mentioned above. So when considering putative counterexamples to the idea that essentialist facts are autonomous it is important to make sure that they concern the core notion of essence, a statement of what something is in its most core respects, and not some extended notion.23 Perhaps the most pressing problem with the idea that essentialist facts are autonomous arises when we think of the similarities in essence amongst things of the same kind. Suppose one holds that it is essential to {Socrates} that it contain Socrates as a member. Presumably one will also think that it is essential to {Aristotle} to contain Aristotle as a member. Why this striking pattern? One might suggest that it is because each essentialist fact has a common ground: they are each grounded (at least in part) in the fact that it is essential to the set-membership relation that it is essential to any set that it has the members it has. Here I must of course disagree. But I need not deny that the pattern exists or say that it is entirely random. For the above fact about the essence of set-membership and the fact that {Socrates} and {Aristotle} are sets imply that {Socrates} essentially contains Socrates and that {Aristotle} essentially contains Aristotle. So there is a relation of implication here, though not one of ground. The resulting picture is not unreasonable. Suppose I want to tell you what something is. Once I tell you that it is a set it follows (from the nature of set-membership) that I must then tell you what its members are. But each statement of what the thing isthat it is a set, that it contains these membersis autonomous. This is by no means the end of the issue. But I am inclined to think that the thesis that all essentialist facts are autonomous is defendable. What about the converse, that all autonomous facts are essentialist facts? I have even less to say about this here. Indeed the resolution to
23 For more on the notion of mediated essence see Fine [15]. Koslicki [27] also distinguishes a number of other notions of essence. It is a good question whether one of them can be taken to ground the others but I cannot discuss it here.

13

this question will likely depend on very broad ranging considerations. Suppose one follows Augustin Rayo in recognizing a distinctive just is operator that is used to say things like For it to be the case that Tom is a bachelor just is for it to be the case that Tom is unmarried. And suppose one distinguishes these facts from essentialist facts. Then one might well think that such facts are also autonomous. So the issue depends on very general questions concerning what kinds of metaphysical tools one recognizes and there is clearly no way to settle it here. Other candidates for autonomy are conceptual truths. If a conceptual truth is understood to be something other than a statement of the essence of a given concept then this would also be an example of non-essentialist, autonomous fact. But again, the issue here clearly depends on broad questions about the nature of conceptual truth. For the sake of concreteness I will largely assume in what follows that all autonomous facts are essentialist facts, but this is a working hypothesis and not a considered view.

A Principle of Sufcient Reason

Returning to rationalism, our version of the PSR is (as I said) that every substantive fact has an autonomous ground. More precisely: PSR: For every substantive fact Y there are some facts, the Xs, such that (i) the Xs ground Y and (ii) each one of the Xs is autonomous. The claim is not just that all the remarkable facts about mountains and oceans that fascinate the child have a metaphysical explanation. The claim is rather that they have a metaphysical explanation of a very special kind, the kind that would answer the childs question of why they obtain. On the assumption that all autonomous facts are essentialist facts, this implies that every substantive fact has an explanation that bottoms out in facts about the essences of things. As an example of such an explanation, my daughter (who was 3 at the time) once asked me why the DMV was closed on Presidents day. The answer (it seemed to me) was that Presidents day is a public holiday, that public institutions are closed on public holidays, and that the DMV is a public institution. Arguably, these truths are essential to Presidents day, public holidays, and the DMV respectively. (For example, if by Presidents day we mean not the calendar date but a day that plays a functional role (so that, for example, it makes sense to consider moving Presidents day to another date) then it is plausible that part of that functional role is serving as a public holiday.) If that is right, then we have an explanation that bottoms out in purely essentialist and therefore autonomous truths, an explanation that (I claim) answers the childs question of why the DMV is closed on Presidents day. (I can report that my explanation satised 14

my 3-year-old, a high bar indeed!) The rationalists view is that everything substantive can be given this kind of explanation.24 The opposing view, fundamentalism, is that some substantive facts have no autonomous ground. This may be because a descending chain of grounds has no end. Or it may be because there is a domain of brute facts (i.e. substantive facts without grounds) that ground all else. On this second view the question of why there is a mountain here might be explained in terms of particle positions and those particle positions might be explained in terms of some eld, but at some point we will reach rock-bottom, a layer of substantive facts without grounds. Being substantive, these facts at the bottom are perfectly apt for being metaphysically explained, it is just that they have no explanation: things just are like that and that is all there is to say about the matter. The childs question of why the mountain exists therefore has (on this view) no satisfactory answer. This is what distinguishes rationalism from this second kind of fundamentalism: both think that some domain of groundless facts ground all else, but only the rationalist thinks that those groundless facts are of a kind that sufces to answer the childs questions. So our PSR implies that there are no brute facts, but it also implies that there are no non-terminating descending chains. This might be puzzling. For I initially glossed rationalism as the view that everything has an explanation and this might be thought to imply that every descending chain of ground is non-terminating. So one might complain that someone endorsing our PSR does not deserve the title rationalist. This complaint is incorrect. For one thing, the PSR has a long history of being interpreted in such a way that it would not be satised merely by a non-terminating chain.25 But more importantly the only groundless facts according to my version of the PSR are autonomous facts. And they are not apt for being grounded in the rst place so it would be perverse to complain that their being ungrounded somehow violates the gloss that everything has an explanation. Compare: I said earlier that I talk of ground as a relation between facts. It then follows that (say) my laptop does not have a ground simply because it is not the kind of thing that is apt for being grounded. But it would obviously be absurd to infer from this that my laptop is therefore brute and the PSR is violated! When we glossed rationalism as the view that everything has an explanation we meant
be clear, nothing hangs on whether the explanation discussed here is correct, or whether the facts appealed to in the explanation really are essentialist truths. The example just serves to illustrate what an explanation in terms of autonomous truths might look like. 25 Just think of Clarkes cosmological argument in which he insists (on the basis of some version of the PSR or other) that a non-terminating chain of dependence must itself have some external cause. It is debatable what version of the PSR he used but he clearly did not consider it satised merely by a non-terminating chain.
24 To

15

that everything that is apt for being explained (given the particular notion of explanation in play) has an explanation. For ususing (as we are) the notion of groundthis means that every substantive fact has an explanation. Which is precisely what my PSR implies.

From Rationalism to Necessitarianism

This then is the version of the PSR I want to discuss. Let us now explore its consequences (sections 612) before asking whether the contemporary bias against it is justied (sections 1317). As I said at the outset the PSR has sometimes been thought to imply necessitarianism (the view that every truth is necessarily true), the existence of God, and indeed the existence of every conceivable being. I will argue that our version has the rst consequence but not the latter two. It should not be surprising that our version implies necessitarianism. For the motiving idea behind our PSR is that for any substantive fact there is a satisfactory answer to the childs question of why it obtains. And (as I said earlier) it is natural to think that to answer the childs question one must at a minimum say why it must obtain. But is there a general argument that the PSR implies necessitarianism? One popular argument to this effect fails when applied to our version of the principle.26 The argument is this. Suppose for reductio that the PSR is true and that there are contingent facts. Let C be the conjunction of all these contingent facts. The PSR is then said to imply that there are some facts, the Es, that explain C. Now either all the Es are necessary facts or not, but the argument is that each disjunct leads to absurdity. Suppose rst that all the Es are necessary. Then since they explain C it follows that C is necessary. But C is not necessary since any conjunction with a contingent conjunct is contingent. Suppose now that not all the Es are necessary. Then one of the Es, call it e, is contingent. So e is a conjunct in C; and since the Es explain C it follows that the Es explain any conjunct in C; so the Es explain e; so e is part of an explanation of itself. But no contingent fact can be part of an explanation of itself. Q.E.D. But the argument fails on at least two counts when applied to our PSR. For one thing our PSR does not imply that there are some Es that explain C: if C is autonomous then our PSR allows that C has no ground. But even if we grant that there are some Es that explain C the argument still fails. The mistake is the inference from the claim that the Es explain C to the claim that the Es explain any conjunct of C: this inference is not valid
26 The argument is due to van Inwagen [39] and Bennett [4]. My presentation here is based on Della Roccas presentation in his [13]. van Inwagens argument differs a little from Della Roccas presentation but the moral of what I say here applies to van Inwagens original presentation.

16

when it comes to ground. For one of the paradigm examples of ground is the idea that conjunctions are grounded in their conjuncts: if A&B, then A and B together ground A&B. But clearly one may hold this plausible view without holding that A and B together ground A. One might (for example) hold that A and B are each groundless and yet also hold that they together ground A&B. So our rationalist may consistently hold that the Es ground C without thinking that the Es or indeed any subplurality of the Es ground any conjunct in C.27 This is not to deny that the argument may be valid when applied to other versions of the PSR. It may be true that if some Es causally explain why there was a hurricane and a snowstorm then those Es also causally explain why there was a hurricane (though I confess to being skeptical in this case too). But this is evidently not true of the notion of ground. Still, there is another argument that purports to establish that the PSR implies necessitarianism. The argument is roughly Spinozistic in origin: its strategy is to rst argue that the PSR implies that everything is explicable in terms of necessary facts and then argue that whatever is so explained must itself be necessary.28 Applied to our PSR the Spinozistic argument proceeds as follows. The rst premise is that all autonomous facts are necessary. And the second premise is that if the Xs ground Y then it is necessary that if the Xs obtain then Y obtains. From our PSR it then follows that all substantive facts are necessary.29 Both premises are plausible. The rst is particularly plausible if considered in the contrapositive: if a fact is contingentif it obtains but might not havethen surely the question of why it obtains can legitimately be raised. One might also argue for it on the basis of two sub-premises: that all autonomous facts are essentialist facts and that all essentialist facts are necessary. I will discuss this sub-argument further in due course. Some deny the second premise and think that even if the Xs ground Y it is nonetheless possible for the Xs to obtain without Y.30 But I maintain
27 In conversation van Inwagen suggested to me that the argument goes through if we accept the general principle that if the Es explain C and C necessitates C* then the Es explain C*. For since C necessitates each of its conjunctions it follows that the Es explain each conjunct. But if explains means ground we should reject this principle for independent reasons. For the arrangement of particles in front of me might ground there being a mountain here, and there being a mountain here necessitates that 2+2=4, but the arrangement of particles in front of me does not (I claim) ground the fact that 2+2=4. 28 I take it that this is (in roughest outline) Spinozas idea in The Ethics, 1p29: In nature there is nothing contingent, but all things, from the necessity of the divine nature, have been determined to exist and act in a certain way. At any rate, this is how Lin interprets this passage in his [29]. 29 Strictly speaking we also need the assumption that if it is necessary that if the Xs obtain then Y obtains, then if the Xs are necessary then Y is necessary too. But (ignoring complications regarding plurals) this is just the distributive axiom K which is a logical truth in all serious logics of necessity. 30 See for example Leuenberger [28].

17

that the premise is true. If Europes being at war in 1939 was grounded in how its citizens were acting, then those actions are what made it the case that Europe was at war and are that in virtue of which it counted as being a continent at war. If it were possible for them to act like that and yet be in a state of peace then presumably their acting like that was not the full story of what made the continent count as being at war after all (though it might have been a contributing factor). There are of course weaker explanatory relations denable in terms of ground such as the contextually relevant subset of Ys ground that would be appropriate to cite in answer to a question of why Y obtains (in that context, given what your audience knows and so on), and that contextually relevant subset may not necessitate Y. Perhaps those who deny the third premise are implicitly working with one of these weaker notions. But premise three, stating (as it does) a principle about ground, is true. There is more to say about this Spinozistic argument. But before we say it let us pause to consider what it does and does not purport to establish.

Relative Possibility

Let us rst be clear on the sense of necessity at issue. Consider the arguments second premise. It is controversial whether grounds logically or conceptually necessitate what they ground. The premise is more plausible when understood as saying that grounds metaphysically necessitate what they ground, and this is how I understand the premise here. I will say more about what this notion of metaphysical necessity amounts to in the next section but for now let us take it as given. So understood, the Spinozistic argument does not establish that every truth is logically or conceptually necessary, just that they are metaphysically necessary. Here is one place, then, where my PSR is weaker than various others. Della Rocca (at least on one reading) takes a complete explanation of a given truth to be a demonstration that it is a conceptual truth (or follows from conceptual truths).31 So his version of the PSR implies that every truth is a conceptual truth. But my version works with a different notion of explanation and so (at least as far as the Spinozistic argument goes) it does not imply anything like that strong. Still, my PSR is strong enough! My armchair exists. According to the Spinozistic argument, my rationalist then says that it is metaphysically necessary that it exists, that it could not (in the metaphysical sense) have failed to exist. Nor could it have been (say) 3 inches over the left. It had to exist and be exactly where it actually is!32
in particular his [12]. the metaphysical sense of had, of course; I will often drop this qualication now for ease of prose.
32 In 31 See

18

But it does not follow that it essentially exists or is essentially where it is. This is just to point out the now-familiar distinction between (metaphysical) necessity and essence. All that follows from the Spinozistic argument above is that its existence and location are necessitated by autonomous facts which are themselves necessary. Now given our working hypothesis we can assume that these autonomous facts are essentialist facts. But they need not be essentialist facts about the armchair itself. That is one view, but a second view is that its existence is grounded in facts about the essences of the underlying matter making it up. This might be the result of two claims: the (perhaps plausible) claim that the armchair exists because various bits of physical materialparticles, elds, whateverare arranged in a certain way, and the (perhaps less plausible) claim that it is essential to that physical material to be arranged like that. But never mind the details: the idea on this second view is that insofar as the chair exists necessarily this is not thanks to the essence of the chair itself but rather thanks to the essences of the underlying physical material that grounds its existence. Indeed on this second view very little may be essential to the chair itself. So this is a sense in which the chairs existence is contingent: whether it exists is left entirely open by its own essence. It is only from the essences of the underlying bits of physical material that its existence follows. What we just did is introduce a kind of relative necessity: necessity due to the essences of a certain class of things. More precisely let us say that it is necessary relative to some things, the Xs, that iff either (i) it is essential to one or more of the Xs that , or (ii) the fact that is grounded in facts about the essences of one or more of the Xs. The dual is a notion of relative possibility: it is possible relative to the Xs that iff it is not necessary relative to the Xs that . That is, if the essences of the Xs do not include and do not ground the fact that . The result is this. If the argument in the last section is valid then the rationalist must say that my chairs existence is necessary. But ner-grained distinctions are possible. She can consistently say that it is necessary relative to the underlying physical material that my armchair exists and yet also that it is possible relative to my armchair that it does not exist. The rationalist can even say that it is possible relative to everything we know of that it does not existno wonder then that its existence strikes us as contingent!33
notion of relative possibility resembles the notion of per se possibility that Adams [2] and Lin [29] both attribute to Leibniz (at least during some points of his career). On this view, a substance is necessarily F just in case it is essentially F. A substance is possibly F just in case it is not essentially not- F (Lin [29], p. 427). This allows Leibniz to say that it is possible in this per se sense that I do not exist, even though in another sense (the per accidens sense) I had to exist by virtue of the fact that God had to choose our world (it being the best of all possible worlds). But there are differences. First, the notion of relative possibility is more general: with per se possibility we only get what is possible for
33 This

19

Necessitarianism Revisited

So our PSR is consistent with contingency in some senses of the term. It may be logically and conceptually possible for my chair not to exist. And it may be possible relative to itself that it not exist. What our PSR implies (according to the Spinozistic argument) is just that it is metaphysically necessary that it exists. But what is meant by metaphysically necessary? On one view metaphysical necessity can be dened in terms of relative necessity. For example one might say that to be metaphysically necessary is to be necessary relative to the worldthat is, relative to all things taken together.34 Necessitarianism then becomes the view that every fact is either an essentialist fact or else grounded in essentialist facts. But now recall our working assumption, that the autonomous facts are all and only the essentialist facts. Given this assumption, necessitarianism then amounts to the claim that every fact is either autonomous or grounded in autonomous facts. . . a restatement of our PSR. So on this view of metaphysical necessity (and our working assumption) our PSR implies necessitarianism only because they are one and the same claim! And (unsurprisingly) the rst premise of our Spinozistic argument that our PSR implies necessitarianismthe premise that all autonomous facts are necessary becomes a logical truth. For it becomes the claim that all essentialist facts are either essentialist facts or grounded in essentialist facts, a tautology of the form All Fs are Fs or Gs. But one might think that there is a notion of metaphysical necessity that is not necessity relative to the world. The claim that our PSR implies necessitarianism then becomes a non-logical truth in need of argument. In particular our rst premisethat all autonomous facts are necessaryis on this notion of metaphysical necessity a non-logical truth. So is it true? It is hard to say without more details about the modal notion under discussion. But some schematic defense can be given. For (as mentioned earlier) the premise follows from the two sub-premises that all autonomous facts are essentialist facts and that all essentialist facts are necessary. The rst sub-premise is a working assumption, but what about the second? I believe that it is very plausible indeed given the notion of essence I am using hereat least it is hard to see how they could be contingent.
x relative to x, but with relative possibility we can relativize what is possible for x to any arbitrary collection of entities. And second, even when we relativize what is possible for x to x, relative possibility is stronger than per se possibility. For (loosely speaking) for it to be per se possible that x is F, being F just has to be consistent with the essence of x; while to be possible relative to x that x is F, being F has to be consistent with the essence of x and all that that essence grounds. Still, I take it that the two notions of per se and relative possibility are very much in the same ball-park. I will further discuss this connection in sections 1516 below. 34 This is (I take it) more or less Kit Fines idea that metaphysical necessity ows from the natures of things (see Fine [14]).

20

For a possibility is a way for things to be. But essentialist facts (as I understand them) detail what those things are in the rst place. In this sense then the essentialist facts are prior to the possibilities: the essentialist facts about the things give us the raw materials, as it were, and only then do the possibilities detail different ways for that raw material to be. For example the essentialist facts about {Socrates} tell me what the thing is, that it is a singleton containing Socrates. Once this is specied we can specify various possibilities for this singleton, for example that it exists along with {Aristotle} and that it does not. But if essentialist facts are prior to possibilities then it is not the case that they obtain in some and not in others. So they are not contingent. This idea that essentialist facts are prior to possibilities is encouraged by the analogy with stipulative denitions. For (at least on the pre-Quinean picture) denitions are prior to theories: one must rst dene ones terms and only then can one use them to express a theory. As a result it is incoherent for one and the same denition to be true in one theory and false in another.35 Of course things are different on the Quinean picture, on which words do not have meanings prior to being used in theories but instead acquire their meaning (insofar as that is the appropriate word) from their role in the theory. The worldly analogue of this Quinean view is perhaps the idea that essences are contingent, that (for example) the essence of this armchairwhat the armchair isdepends on what is actually true of it. There may be notions of essence that bear analogy with this Quinean picture but I am (by stipulation) talking about the worldly analogue of pre-Quinean denitions and so it seems to follow that they are prior to possibilities.36 Does this mean that essentialist facts are necessary, obtaining in all possible worlds? Not quite. Insofar as essentialist facts are prior to possibilities there is a sense in which they do not obtain at possibilities in the rst place. So we might distinguish between the necessary facts, those that obtain at all possibilities, and the essentialist facts that are prior to those possibilities. These latter might be better labelled as amodal. The fact that it is raining or it is not raining is (arguably) a fact of the rst kind, a fact that obtains in all possibilities.37 Of course we can then dene
are delicacies here regarding the individuation of terms. For it is of course possible that one and the same syntactic string receive two different denitionsthis is just a case of ambiguity. What is incoherent on this pre-Quinean picture is the idea that a denition of a word (understood as a semantically individuated entity) could be true in one theory and false in another. 36 At this point again one might again wonder whether the particular notion of essence I have in mind corresponds to anything real. But as I said in section 4 this is not a topic I can settle here. The dialectic in this paper is to take the notion as given and see what kind of rationalist view can be formulated in its terms. 37 This notion of amodality bears comparison with the notion of an unworldly fact that Fine discusses in his [18]. See also Hudson [22].
35 There

21

a broader notion of necessity by disjoining these notions of necessity and amodality. If something is necessary in this broad sense then it does not vary from possibility to possibility (either because it obtains at all possibilities or because it is amodal, or some combination of the two). In this broader sense the essentialist facts are then necessary after all. There is no need to decide which notion corresponds best to our intuitive notion of necessityindeed there may be no determinate answer to that question. It sufces to distinguish the notions and the corresponding necessitarian theses. In terms of the broad notion, necessitarianism becomes the view that no fact is contingent in the sense of varying from possibility to possibility. This can be argued to follow from our PSR by the two-premise Spinozistic argument I have been discussing. One might suspect that our PSR also implies the stronger claim that all facts are amodal.38 But I will not discuss this further here: the idea that no fact is contingent in the sense of varying from possibility to possibility is interesting enough to be getting on with so that is how I will understand necessitarianism in what follows.

Self-Explainers

So much for necessitarianism. The PSR has also (at times) been thought to imply the existence of God, or at any rate the existence of some necessary or self-explanatory being that might then be argued to have various divine or unnatural attributes. Does our PSR have these consequences? It depends on what is meant by necessary and self-explanatory. Our PSR does imply the existence of beings that satisfy these descriptions under some interpretations and in the spirit of exploring its consequences it is worth clarifying what these interpretations are. But none are theological or unnatural: I will argue that the atheist-cum-naturalist philosopher can nd our PSR agreeable. Let us start with the question of self-explainers. Does our PSR imply that there are facts that explain themselves in the very same sense that (say) an arrangement of particles might explain the existence of a chair? If it does then the natural question is how such a situation could be possible. One answer (favored by theists) is that the self-explanatory fact must concern some remarkable entity such as God. But another answer is that the situation is impossible, that it is an essential or conceptual truth about ground that nothing can ground itself. On this view the implication of
what the PSR implies (given our working assumption that autonomous facts are essentialist facts) is that all facts are grounded in an amodal base. So suppose that if the Xs ground Y and if the Xs are amodal, then Y is amodal too. Then the stronger necessitarian thesis follows.
38 For

22

self-explanatory facts would constitute a reductio of the PSR.39 But I see no reason to think that our PSR has this implication. If anything it implies the opposite, that no fact explains itself, since it requires that all substantive facts ultimately be grounded in autonomous facts which are groundless. It might be instructive here to see how Clarkes famous cosmological argument (to the effect that the PSR implies self-explainers) fails when it comes to our PSR. Clarke explicitly talked of dependence, but translated into our talk of ground his major premise was that if the PSR is true then any descending chain of ground either terminates in a self-explaining fact or else goes on and on explaining one fact in terms of another ad innitum. And he then argued (controversially) that even in the second case there must be some self-explanatory fact that explains the entire innite chain. But clearly the major premise is false when it comes to our PSR. Far from requiring that every descending chain of ground terminate in a self-explainer or else go on ad innitum, our PSR requires (on the contrary) that the chain terminates in autonomous facts, facts that have no ground. Now there is a loose and non-literal sense in which one might describe autonomous facts as self-explanatory. Perhaps this is all that is meant by others who talk of a facts being self-explanatory. Indeed my autonomous facts play a similar role in our PSR as the (literally) self-explanatory facts play for Clarke: they are the unmoved movers, facts not grounded in others but which give rise to all others. But my autonomous facts do not have the unnatural or divine property of being self-explanatory in any literal sense. Indeed for all that has been said they are not unnatural or divine at all: they are just essentialist facts, facts that we express when stating what (for example) knowledge is. If our PSR implies the existence of any divine or unnatural being, more must be said.

10

Essential Existence

Which brings us to the question of necessary existence. Does our PSR imply that there are necessary beings? And if so does it follow that these
39 Many who write about ground these days (for example Rosen [35]) assume that ground is irreexive even in the weak sense that no fact is even part of a ground of itself. Jenkins [25] by contrast argues that such cases of self-grounding are possible (and indeed actual). I myself take an intermediate view. It would be presumptious to declare that it is built into the very notion of causation there can be no closed time-like loops of causal inuence of the kind one might nd if time-travel is possible. And it strikes me as similarly presumptious to declare in advance of investigation that there can be no extended loops of ground in which A grounds B, B grounds C,. . . and Z grounds A (which, given transitivity, is a situation in which A grounds A). But it does strike me as built into the notion of ground that nothing immediately grounds itself without a detour through some other facts. So I am tempted to accept an irreexivity principle if it is understood to govern immediate ground, though perhaps not if it is understood to govern ground.

23

beings are somehow unnatural or divine? Well our PSR certainly implies (as we saw) that even such mundane things as my armchair exist necessarily if by necessarily we mean necessarily relative to the world. But this clearly does not imply that it is in any way unnatural or divine. A much stronger sense in which something exists necessarily is to exist necessarily relative to itself that is, for its existence to be part of or grounded in its own essence. Does our PSR imply that there are such things? One might suspect that it does. For suppose that my armchair is not such a thing. Since its existence is not autonomous (at least, not on our working hypothesis that autonomous facts are essentialist facts) our PSR implies that there are some facts, the Xs, that ground its existence. Now suppose (premise) that existence can only be explained in terms of existence. More fully, call a fact of the form x exists or of the form it is essential to x that x exists an existence fact. Then our premise is that if some facts ground the fact that y exists then at least some of them are existence facts. This is not wholly implausible: if my chairs existence is grounded in (say) facts about the arrangement of particles then some of these latter facts presumably include that those particles exist. The fact that those particles exist may then be explained in terms of the existence of some more basic matter, perhaps some kind of eld or what have you. But according to our PSR this cannot go on forever: the chain of ground must terminate in some autonomous facts. And according to our premise these must include a fact to the effect that it is essential of some given thing that it exists. So if my chair exists then there is some x (who knows what) such that it is essential to x that x existsthat is, an x that exists necessarily relative to itself.40 In response one might object to the premise.41 But I do not want to pursue that line here. So let me grant that the PSR implies that there are beings that exist necessarily relative to themselves. What follows? Does it imply that these beings are somehow unnatural or divine? I see no reason to think so. For something might be thought to exist necessarily in this strong sense but not thereby inherit these attributes.
40 This bears some resemblance to Aquinas cosmological argument in its insistence that a descending chain of explanations cannot go on forever. 41 One objection would be to propose that facts about the existence of particulars are grounded in facts expressed by a kind of predicate-functor language (see Hawthorne and Cortens [23] and Dasgupta [7] for developments of this idea). Indeed we might think that these predicate-functor facts are essentialist facts if we relax our logical form of claims about essence to include claims of the form

It is essential to being F that . where need not concern the existence of anything. On this view we can satisfy the PSR without accepting that there are things whose existence is part of or grounded in its own essence.

24

Natural numbers might be an example. For suppose that one believes that natural numbers exist and (moreover) exist necessarily. Necessarily, in what sense? Is the necessity of their existence thanks to their own natures or thanks to the nature of something else? It is hard to see what the something else could be. So a potentially attractive view is that they exist necessarily in the strong sense, that it is in their own nature to exist and instantiate the structure they do. Never mind whether this is correct, the point is that no one would take this to imply that numbers were divine. Space (or space-time) might be a better example. For suppose one endorses the substantival view that space-time exists independently of the material bodies and elds situated within it. On this view space-time is the stage on which the material history of the world unfolds. Now one idea traditionally associated with this view is that space-time is constant across possible worlds: there may be many ways for the material history of the world to unfold but it always unfolds on the same stage. On this view then space-time exists necessarily. But again, in what sense? Is the necessary existence of space-time due to the nature of space-time itself or to the nature of something else? Again, it is hard to see what the something else could becertainly not matter, since the whole point of substantivalism is that space-time is independent of its material constituents. So the substantivalist might naturally be led to the view that it is essential to space-time that it exist and instantiate the structure it does. Never mind (again) whether this view is correct, the point is that no one would take this to imply that space-time is somehow unnatural or divine.42 Indeed this last example indicates that our PSR is consistent with contemporary physicalism-cum-naturalism. The combined view would be that at rock bottom there is some kind of physical space (in the broad sense of the term) in whose nature it is to exist and instantiate some kind of structure the details of which will depend on the particular physical theory guiding the physicalist (if general relativity it might be a four-dimensional topological manifold instantiating a certain geometric form; see footnote 42). The view will then be that all other (substantive) factsabout my armchair, my mental states, my duties and obligationsare grounded in the existence and structure of that underlying physical space (which as we said is essential to it).43 The existence of my armchair and my mental states is then necessary not because of their own natures but because of the nature of the underlying physical material in which they are grounded.
view is perhaps easiest to motivate in the context of physical theories in which space-time has a xed metric structure. But even in theories in which the metric is dynamic the view is not unimaginable. Indeed Tim Maudlin once proposed when discussing General Relativity that space-time points bear their metric relations essentially (see Maudlin [31]). This not quite the view under discussion but it is related. 43 See my [10] for more on why the physicalist can legitimately restrict her view to substantive facts.
42 This

25

To be clear, I am not asserting that it is essential to numbers or spacetime that they exist and instantiate the structure they do. The point is just that these claims can be debated and yet not thought to entail the existence of anything unnatural or divine. Of course one might insist that anything in whose essence it is to exist deserves the title God. But I do not want to ght over the word: the important point is just that the claim that it is essential to something (e.g.space-time) to exist does not conict with the views of an atheist-cum-naturalist philosopher. Once again, much hangs on the particular version of the PSR under discussion. For suppose one were considering a version of the PSR that implies that it is an analytically or conceptually true, of some being x, that x exists.44 Then one might object that analytic or conceptual truths are never existence-entailing, or else insist that a being that exists as a matter of analytic or conceptual truth must be divine indeed! But our version of the PSR does not imply this: it just implies that it is an essential truth about something that it exists, and not all essential truths are analytic (it might be essential to water to be composed of H2 O, but this is not analytic).45 Still, the fact that our PSR implies that there are beings that exist necessarily relative to themselves might be thought problematic enough regardless of whether they are unnatural or divine. So let us ag this consequence as one to be discussed later when asking whether the contemporary anti-rationalist bias is justied.

11

Non-Existence and Plentitude

Spinoza offered a rather different argument from the PSR to the existence of God. He supposed for reductio that God does not exist. By the PSR, there must then be some explanation of this. He then argued that there could not be an explanation; contradiction. Hence God exists. Spinozas argument that there could be no explanation for the nonexistence of God involves some of the more idiosyncratic elements of his philosophy.46 But regardless of those details it is clear that a rationalist
version that Della Rocca discusses in his [12] may well be of this type. makes a similar point when discussing necessary existence in his [1], though for a somewhat different purpose. 46 His idea was that the explanation would either be internal or external. The latter would appeal to a substance distinct from God but this is ruled out by Spinozas view that distinct substances cannot affect each other. An internal explanation would explain Gods non-existence by citing some contradiction in Gods nature, much as we might explain the non-existence of a square circle by citing a contradiction in the nature of that shape. But Spinoza claims that God is not self-contradictory so there cannot be an internal explanation of Her non-existence either. Something like this argument can be found in Spinozas various proofs of proposition 11 in part 1 of his [38]. But I am no scholar so I may well be ignoring many subtleties in Spinozas view. For more on Spinozas views on this matter see Lin [30] and Della Rocca [11].
45 Adams 44 The

26

atheist must either nd some autonomous ground of Gods non-existence or else claim that Her non-existence is autonomous. The same goes for other non-existents. There are no humans over 9 feet tall, but this is a substantive fact so (by our PSR) it must have an autonomous ground. And I have no identical twin, but again this is substantive and so (by our PSR) must have an autonomous ground. If there is no autonomous ground then our PSR implies that these things exist after all. So if no autonomous grounds are found our PSR might be thought to imply the principle of plentitude, the principle that every conceivable being exists.47 So the challenge is to nd autonomous grounds for substantive facts about non-existence. One idea is to ground somethings non-existence in its own essence. But even if this is the right explanation in some cases (perhaps the non-existence of a square circle) this is unlikely to generalize. In particular, it is unclear whether God has this kind of essence. So let us ask what other kinds of explanations of non-existence might be available. It helps to note that this question also arises for some fundamentalists. For regardless of the fundamentalism/rationalism issue, some have been attracted to the view that negative facts are grounded in positive facts, i.e. that what is not the case is grounded in what is the case. This is one interpretation of the slogan that truth depends on being. More precisely, the thesis is that any fact of the form It is not the case that is grounded in facts that are not of that form. The negative fact that I am not a doctor will on this view be grounded in various positive facts about what I am, for example that I am a teacher, a father, . . . , perhaps along with a totality fact that those are all the facts about what I am.48 But facts about non-existence are negative facts. So anyone who thinks that truth depends on beingrationalist or fundamentalistthinks that facts about non-existence are grounded in positive facts. One might say for example that there are no humans over 9 feet tall because this human is less than 9 feet and that human is less than 9 feet and. . . (and one might add the totality fact that these are all the humans). And so the atheist may then ground the non-existence of God similarly: God does not exist because these are all the existents and they have these (non-divine) attributes. The point is that this explanation of non-existence is just as available to the rationalist as it is to the fundamentalist.49
47 For more on the relation between the PSR and the principle of plentitude in Spinoza, see Lin [30]. 48 On one view the totality fact is part of what grounds the negative fact. On another view it is not part of the the ground but is a necessary condition for the other positive facts to count as a ground for the negative fact. I will not try to decide between these views. 49 This is to give an external explanation of Gods non-existence, to use Spinozas expression (see footnote 46). Spinoza may have used a notion of explanation on which he

27

12

The Appearance of Bruteness

The following rationalist picture emerges. At rock bottom are facts about the essences of things. They may or may not include facts about the essences of non-existents but let us suppose (to build a clear picture) that they do not. Other positive facts are then grounded in those essentialist facts. And negative factsincluding those about non-existenceare then grounded in those positive facts. Since the atheist denies that God exists she will on this picture deny that there are positive facts about the essence of God. So when the theist says that it is essential of God that she is perfect the atheist will disagree. The atheist may think that there are negative facts about the essence of God, e.g. that it is not the case that it is essential of God that she is perfect. But these are (on this picture) grounded in the positive facts about the essences of things, none of which are about God. Does this satisfy our PSR? One might worry that it does not. For consider all the claims that the theist makes about the essence of God. They constitute (let us suppose) a logically consistent description. Indeed there are many logically consistent descriptions that might (for all we know) constitute somethings essence: one for me, one for God, one for my (poosible) identical twin, and so on. Why then did the rst turn out to describe something real but the latter two did not? On the current picture there seems to be no answer: it looks brute and inexplicable which of the many logically consistent descriptions that might constitute somethings essence are actually true of something that exists, in violation of our PSR. But there is a reply available. If the question is why facts about my constitutive essence obtain the answer is that they are autonomous and so are not apt for being grounded in the rst place. So there is no explanation of why they obtain but this does not violate our PSR. If instead the question is why these other claims about the essence of God do not obtain then this is just a request for an explanation of a negative fact, e.g. that it is not the case that it is essential of God that she is perfect. On the current view these are grounded in positive facts which are in turn grounded in autonomous facts about the essences of the existents. Either way, there is no violation of our PSR. The worry that there is something brute stems I think from the feeling that since the theists statement of the essence of God is logically consistent it surely could be true and describes something that could exist. This feeling is correct but one must be careful to identify the sense of contingency in play. It is not that the statement could be true in the metaphysical sense of could. It is rather that it could be true in a logical or epistemic sense.
is correct to say that there can be no external explanations. But for us, using the notion of ground, such explanations would appear to be commonplaceat least they are on the view that truth depends on being.

28

For all we know apriori, we might say, there are many ways the world could turn out, many essentialist statements that could obtain and would then provide the ground for all else. Some of these epistemic or logical possibilities include facts about the essence of God. But once we discover that these possibilities do not obtain we see (as rationalists) that in the metaphysical sense they could not obtain. To be sure, there may be some readings of explanation on which this response is not available. To explain something on some readings of the term might require showing that it is (or follows from) an apriori or conceptual truth. As I said, Della Rocca appears to be working with a notion of explanation rather like this.50 But (as we have seen) ground is not this kind of explanatory notion. Once again, the crucial move is open to us precisely because of the particular version of PSR under discussion.

13

The Anti-Rationalist Bias

So far I have stated a version of the PSR and explored its consequences. Let us now ask whether the contemporary bias against it is justied (I hope it is clear by now that there is a contemporary bias against it). My aim is to argue that it is not. What could justify this bias against rationalism? One objection is that the notions in terms of which it is formulated (such as ground, essence, and autonomy) are unintelligible. But if they are unintelligible then fundamentalism (as I have dened it) is also unintelligible and the entire debate (if there is one to be had) needs to be formulated differently. So in order to have this debate let us take these notions for granted. There are then (broadly speaking) two considerations one might raise against rationalism. One might (in the rst place) have reason to think that rationalism is false. But even if one does not one might (in the second place) think that rationalism is such a very strange view that fundamentalism is the default view, so that absent strong arguments for rationalism we should be fundamentalists. In section 17 I will argue that this second consideration is a mistake: neither view is default, each requires strong arguments to warrant our belief. But the bulk of what follows will focus on the rst consideration, the idea that we already have good reason to think that rationalism is false. What reasons? Not counterexamples: I doubt that the contemporary fundamentalist rests her position on a particular substantive fact that she believes to be brute or grounded in an innite descending chain. For I do not think we have rm beliefs as to what kinds of facts are brute. Fundamentalists may sometimes write as if (say) particle positions are brute but this is often a simplication used to sidestep difcult questions
50 See

Della Rocca [12].

29

about the interpretation of quantum mechanics and other eld theories. And in any case I doubt that a fundamentalist would wish to rest her position on this or that nal theory of the world turning out correct.51 So if there are reasons to think that rationalism is false they must be very general considerations, considerations that consist in showing that various implications of the PSR are false prior to settling on a nal theory of the world. Now our PSR (as we have seen) does not imply the existence of God or every conceivable being so one cannot object to our PSR on those grounds. But we have identied two potentially objectionable consequences: that there are things whose existence is part of or grounded in their own essence, and that everything is necessary. So I will try to show that these consequences are not problematic (sections 1416). It goes without saying that I will not be able to settle the matter in the space available, but I will try to outline the beginnings of a defense.

14

Against Essential Existence?

Let us start with the idea that there are beings whose existence is part of or grounded in their own essence. Is this problematic? Remember that essentialist facts need not be analytic so it is not in order to complain that analytic truths are never existence-entailing. Still, it might be thought problematic for the following reason. Call an essentialist fact antiAnselmianism if it is of the form It is essential to a that ( x)(a = x __ (x)). where the underscore may be lled by if, only if or iff; and Anselmian otherwise. Rosen and Kment have each suggested recently that all essentialist facts are anti-Anselmian.52 Let us work in a free logic in which (a) does not imply that a exists. Then there is a very strong sense in which Anti-Anselmian essences are not existence-entailing: it is not just that they do not imply as existence, they do not even imply as existence conditional on facts about other things. Suppose for example that it is essential to my armchair that ( x)(x = my armchair if x is composed of these particles arranged in this way). This is an anti-Anselmian essence. It does not imply that my armchair exists (in a free logic). But even if those particles are so arranged it still does not follow (logically) that there is anything composed of them and so it still does not follow that my armchair exists. Contrast this with the following
51 Note in this regard that genuinely chancy events (particle decays or wavefunction collapses, on some views) would not be counterexamples to our PSR. At best they are counterexamples to causal versions of the PSR, but it is consistent with the existence of genuinely chancy events that they have an autonomous ground. 52 See Rosen [34], who introduced the terminology. And see also Kment [26].

30

Anselmian essence: that it is essential to my armchair that it exists if these particles are arranged in this way. If those particles are indeed so arranged then it follows (even in a free logic) that my armchair exists. The objection to the idea that there are beings whose existence is part of or grounded in their own essence then consists of two premises: (i) that anti-Anselmian essences can never ground somethings existence, and (ii) that all essences are anti-Anselmian. But both premises can be questioned. In response to (i) one might point out that although an anti-Anselmian essence does not imply anythings existence it might nonetheless be sufcient to ground somethings existence. Still, I am inclined to accept (i) so let me focus on (ii). Why think that all essences must be anti-Anselmian? The Anselmian statement of the essence of my chair described above may or may not be true, but is there something about the very notion of essence that requires that it is false? I cannot see that there is. For a statement of xs essence (as I use the term) is a statement of what x is. And the Anselmian statement above seems to me a perfectly legitimate statement of what my chair is (whether or not it is true): it is something that exists if these particles are arranged in a certain way. So I see no reason to believe (ii). At least, the onus is on the fundamentalist to offer some argument in its favor. Of course my example of the Anselmian essence of my chair falls short of the kind of Anselmian essence that the rationalist might require, i.e. an essence of x that would ground xs existence. The point is here is just that the current argument against such essences, which appeals to (ii), is not convincing. Although (ii) is false I think we can diagnose why it might be attractive. The key is to distinguish between two statements. On the one hand is a statement of what x is, where x may be an item of any category (an object, a property, what have you). This is what I am calling a statement of xs essence. And on the other hand is a statement, for a given property, of what it is for an object to have that property. In principle, these are different things. To take a favorite example of Mark Johnstons, a statement of what red is might be different from a statement of what it is for a material body to have red as its surface color (the former might involve saying that red is a color, that red is a universal, and so on; the latter might involve talking about the material body having a disposition to reect light in a certain way). With this distinction in hand we can now see that an anti-Anselmian essence is a statement of what it is for an object to have a certain property, namely the property of being identical to a. Which is not a statement of as essence as I am using the term. Rosen and Kment may be right that statements of what it is to have the property of being identical to a always have that anti-Anselmian form, but it does not follow that essentialist facts as I am using the term must be anti-Anselmian. Thus those tempted by (ii) are, I suspect, using a different notion of essence than I am. 31

15

Intuitions and Conceivings

Let us turn then to the other controversial consequence of our PSR, necessitarianism. This is perhaps enough on its own to turn people against rationalism. But why think that necessitarianism is false? What exactly is wrong with the view that every truth is necessary? It is surprisingly difcult to say. Perhaps the rst thing that springs to mind is the objection from obviousness. It just seems obvious that my chair could have been a little over to the leftif the rationalist denies this she denies what we surely know to be true! But a number of responses are available. First, the rationalist might accuse the objection of using a modal intuition as evidence for modal truth and then point out problems with using intuitions as evidence. But I do not want to rest on this response. Insofar as the objection rests on intuition I am happy to think of an intuition as a belief in which one has enough condence to use as a premise in an argument. Presumably we are all condent indeed the chair could be a little over to the left! A second response is to deny that our modal intuition (or belief) has the advertised content. For (as we saw in section 7) the rationalist can agree that it is possible relative to the chair that it be elsewhere, that the nature of the chair does not settle one way or the other where it is. Indeed she can agree that it is possible relative to everything we know of that it be elsewhere. So she might say that our modal intuitions are tracking these relative possibilities in which case they do not tell against rationalism. At the very least the onus is on the objector to argue that our intuitions track the kind of possibility that would tell against rationalism and it is not at all clear that that is so. There is much to be said in favor of this response but again I do not want to rest on it. Instead I want to outline a more concessive response. Grant for the sake of argument that our modal intuition conicts with rationalism and that modal intuition is evidence for modal truth. Still, the rationalist can say that this modal evidence is subordinate to other evidence. Let me explain what I mean by example. Consider the question of whether water is composed of H2 0. This is in large part an empirical question to be settled in the laboratory. Now, prior to chemical investigation our ancestors presumably had the modal intuition (or belief) that water could exist without H2 0.53 If correct it would follow that water is not composed of H2 0. But what was the epistemic signicance of this modal intuition, for our ancestors? One view is that it was no evidence at all. But I am exploring a more concessive approach to modal intuition, so let us grant for the sake of argument that it was at least some evidence that
you doubt this, consider that you probably have the modal intuition (or belief) that water could exist without various super-heavy elements.
53 If

32

water is not composed of H2 0. Still, there is obviously a sense in which it does not carry the day. What does this mean? Two things. First, that empirical evidence (say, from chemical tests) trumps the modal intuition, in the sense that if empirical evidence conicts with modal intuition then the empirical evidence defeats the intuition (not the other way round). And second, that only empirical evidence is sufcient to settle the matter of the composition of water. Here I do not mean that the empirical evidence is indefeasible: all evidence is defeasible. I mean that only the empirical evidence is sufcient to render the question (of the composition of water) a closed research project not worthy of further serious investigation. No amount of modal intuition can settle the question in this sense, but empirical evidence can and has. When I say that evidence from modal intuition is subordinate to the empirical evidence, I mean that the latter trumps the former and that only the latter is sufcient to settle the matter.54 It is obvious to us in retrospect that when it comes to the question of the composition of water, modal intuition is subordinate to empirical evidence. But our ancestors could see this for themselves prior to collecting the empirical evidence. Even back then, before doing the science, it should have been considered bad intellectual practice to think that the question of the composition of water could be settled by modal intuition. Which is presumably why they bothered to do the science in the rst place! I claim that the same goes for questions about what grounds my chairs existence and location. These questions are in large part empirical, scientic questions. Settling them will involve settling empirical questions about the chairs ultimate physical composition and the structure of spacetime. The rationalist thinks that the answers will turn out to consist in facts that we take to be essential truths, the fundamentalist thinks that the answers will turn out differently. But it is a largely empirical question what the answer is. Now admittedly the rationalist answer goes against our modal intuitions, but what is the epistemic signicance of this? Not much. It is no more epistemically signicant (I claim) than our ancestors intuition that water could exist without H2 0. Even if we are concessive and grant that it counts as some evidence against rationalism, it is subordinate to empirical evidence. We know in advance that when it comes to this question of what grounds my chairs existence and location, empirical evidence trumps our modal intuitions and only empirical evidence can
am slurring over a subtlety regarding the nature of evidence. Suppose ones evidence is what one knows. And suppose one takes the initial evidence to be the modal proposition that water could exist without H2 0. Then the correct way to put the point is that when one learns (on the basis of the empirical evidence) that water is H2 0, one in part learns that the modal proposition was not evidence after all. If (on the other hand) one takes the initial evidence to be that the proposition that it is intuitive that water could exist without H2 0, then this might be true and known and therefore count as evidence even if water is H2 0. But these details do not matter for our purposes here.
54 I

33

settle the matter. So our modal intuition does not settle the matter and it would be a mistake to consider the question of rationalism vs fundamentalism a closed question not worthy of further investigation on the basis of those modal intuitions. And yet contemporary metaphysicians regularly make this mistake when they assume fundamentalism without so much as a passing comment! This is what I mean when I say that the contemporary anti-rationalist bias is unjustied. At the very least, anyone wishing to argue against rationalism by way of modal intuition must face the challenge of saying why our modal intuition against rationalism is any more epistemically signicant than (say) our ancestors modal intuition about water. The same goes for arguments from conceivability. For the fundamentalist might argue that it is possible for my chair to be elsewhere because one can conceive of it being elsewhere. Here the rationalist could in principle reply that conceivability is no guide to possibility. But the more concessive response is to grant that conceivability is some evidence for possibility but insist that evidence from conceivability is subordinate to non-modal evidence in the above sense. The virtue of these concessive responses is that one need not question the connection between possibility and conceivability, or between possibility and modal intuition. The exact nature of these connections are a matter of considerable controversy, but one might think them so central to the notion of possibility that one denies them only by changing the subject. The concessive response does not face this problem.55 This is not to say that intuitions or conceivings can never carry the day. Consider for example conceivability arguments against physicalism. Here one argues that one can conceive of physically duplicate situations that differ in various respects about consciousness, and on that basis one argues that the physical facts are not sufcient to explain consciousness. I believe that our conceptions can carry the day in this kind of case. Why the difference? Well, the question at issue here is whether various facts would (if they obtained) count a sufcient explanation for another fact. These are largely apriori questions: we can work out, in advance of knowing what water is composed of, whether the hypothesis that it is composed of H2 0
55 For more on the exact nature of the connection between conceivability and possibility see essays in Gendler and Hawthorne [21]. I should say that my concessive strategy here is neutral on the degree to which modal intuition is evidence for these empirical propositions. One view is that it is some evidence but not much, e.g. it justies slight changes in credence but not much. But the concessive response can even grant that it is strong evidence in the sense that it justies large changes in credence. On this view our ancestors may have been justied (on the basis of modal intuition) in having a very high credence that water is not H2 0. The point then is that having high condence does not entail considering the matter settled (in the sense discussed in the text). There is no need here to decide between these two versions of the concessive strategy.

34

would explain its uidity. The largely empirical question is which, out of all the possible facts that would (if they obtained) explain something, actually obtain. My claim is just that modal intuition and conceivability does not carry the day with regards to these latter questions. Perhaps our conviction that my chair could have been located elsewhere stems from the fact that we cannot even imagine what a rationalist explanation of its location would be like, or why we would believe it. Of course I do not have the correct rationalist explanation to hand. But it is worth noting in this regard that discovering that what we thought was contingent actually follows from underlying essences arguably has precedent in the history of physics. For consider the fact that the speed of light (in a vacuum) is c.56 In a classical, pre-relativistic context no particular (non-zero) speed is privileged and so it appears contingent and arbitrary that light travels at this speed. It could (metaphysically speaking) have had any number of speeds, it is at best a nomic necessity that its speed is c. But here are three discoveries that would lead one to conclude that it its speed follows from the essences of things and is therefore necessary. One rst discovers that one lives in a Minkowski space-time structure. In this structure there is a privileged set of trajectories, those of space-time interval 0 (otherwise known as the light-cone structure). One then discovers that light (in a vacuum) travels along these privileged trajectories. This (uncontroversially) sufces to explain why light has the speed it does. And the third discovery is that the previous two discoveries were discoveries about essence: that it is essential to space-time to have the Minkowski structure it has, and that it essential to light that it travels along those special trajectories. It follows that the speed of light is grounded in its own nature and that of the underlying space-time structure it lives in. In which case it is not contingent that it has that speed after all. The point here is not that this view of light is true (though I believe it is plausible independently of any rationalist considerations). The point is that it is not outrageous: it is a view that we can reasonably debate and which is not far from a view that many accept. So it is an example of a potential discovery in which we come to see that something that was intuitively contingent actually follows from its essence and that of the underlying physical space. The rationalist expects (insofar as we are truthtrackers) that this kind of discovery will become the norm and that in the limit we will see that even my armchairs position is like the speed of light. So much for intuition and conceivability. One might instead argue for contingency from the Humean slogan that there are no necessary connections between distinct existences. According to this slogan my existence, for example, should not necessitate the existence of the Eiffel Tower, contra necessitarianism. But on further examination it is unclear whether ratio56 Where

c = 299,792,458 meters per second.

35

nalism is at odds with the slogan after all. For why should we think that there are no such necessary connections? One reason might be that they would be brute and unexplained. If so then the idea behind the slogan is that there are no brute necessitities. But of course the rationalist agrees since she claims that there are no brute facts tout court! So at least on this interpretation of the slogan there is no conict between it and the kind of necessitarianism that the rationalist endorses. Perhaps a historically more accurate interpretation of the slogan has to do with conceivability: that we can always conceive of distinct things existing without each other. But then the argument collapses into the argument from conceivability discussed above.

16

The Theoretical Role of Contingency

What other reasons are there to believe in contingency? One might worry that one cannot be a free or responsibility agent if there is no contingency. This is a serious worry and I do not know how to answer it. But the worry applies just as much to the fundamentalist who believes that the laws are deterministic as it does to the rationalist. For the worry stems from the idea that we are not in control of what we do, regardless of whether this is because what we do is nomically necessitated by the past or whether it is metaphysically necessitated tout court. And indeed many solutions to the problem that are available to the fundamentalist determinist are also available to the rationalist. Frankfurt for example has argued that freedom and responsibility consist in various psychological mechanisms corresponding to motivation that can be present in a deterministic world.57 But the same mechanisms can be present in a necessitated world too so this solution to the problem is available to the rationalist. A somewhat different way of arguing for contingency is to argue that it plays an indispensable role in various successful philosophical theories and that we should therefore believe in it for that reason. This might appear to be a promising strategy, for recent philosophy is replete with modal analyses that require contingency for their success. Consider for example the project of modeling content with sets of possible worlds. Here I include the semantician who models the content of an utterance with a set of possible worlds, and the Bayesian epistemologist who models an agents doxastic state with a credence function from sets of possible worlds to the real numbers. Both projects require that there be very many distinct possible worlds, enough to generate the many distinct sets of worlds that serve as the objects of credence or the contents of utterances. Since both projects require contingency one might infer from their theoretical success to the falsity of rationalism.
57 See

Frankfurt [20].

36

In response, I deny the assumption that these projects use the notion of metaphysical possibility. Or to be more accurate, I deny that the best versions of these projects use the notion of metaphysical possibility. Why? Precisely because of the potential conict with rationalism under discussion! Consider for example Spinozas derivation of necessitarianism from (something like) the idea that all truths ow from the essence of God, a necessary being. Regardless of whether you think Spinoza was right about this, the idea that he would thereby not be in a position to do semantics or formal epistemology strikes me as absurd. But that is what these projects imply if they use the notion of metaphysical possibility. To put a different spin on the same point, if the semantician insists on using metaphysical possibility then the success of her theory of (say) descriptions would depend on the falsity of Spinozas metaphysics. A complete defense of her theory of descriptions would then require a refutation of Spinozas view. And this is absurd: surely the success of her theory of descriptions has nothing whatsoever to do with the truth or falsity of Spinozas theological beliefs! Just imagine how absurd it would be to teach a class on the semantics of descriptions and hear a student object on the basis of Spinozistic metaphysics. The absurdity does not stem just from the fact that Spinozas views are not widely held, for even someone who takes Spinozas view seriously should balk at the students objection. The absurdity stems instead from the fact that the question of Gods nature is entirely independent from the question of the meaning of descriptions. To be clear, the argument here is not that Spinozism is true, therefore possible worlds semantics is wrong. The point is rather that this argument would be valid if the semantician used the the notion of metaphysical possibility; but the argument clearly is not valid, hence she should not use metaphysical possibility.58 Of course if she uses the notion of logical or epistemic or conceptual possibility then none of these problems arise since these notions are not hostage to metaphysical fortune.59 But then the assumption of metaphysical contingency is not playing an indispensable role in these theories and there is no conict with rationalism after all. The same goes for other possible worlds analyses seen in the recent literatureof dispositions, of laws, of intrinsicality, of counterfactuals, etcthe success of which require that there be many possible worlds. In
58 The point here is therefore different from the familiar point that possible worlds analyses of content struggle with (say) mathematics. For the argument there is that mathematics is necessary, therefore possible worlds semantics is wrong in the case of mathematical belief. I am not making that kind of argument. Moreover in the case of mathematical belief the possible worlds theorist can always try to avoid the conict by modifying the theory to account for these outlier cases or restrict the scope of her theory to contingent propositions. But obviously neither of these responses are available if faced with someone who thinks that everything is necessary! 59 See for example Chalmers [6], who develops a version of Bayesianism that uses his notion of epistemic possibility.

37

each case we must apply the Spinoza test: could someone who believes for metaphysical or theological reasons that everything is metaphysical necessary make sense of the notion? The answer in all these cases seems clearly yes. If so, then the analysis in terms of metaphysical possibility is inadequate for the above reasons. There are then two options: either give up the modal analysis, or else persevere with the modal analysis but then explain what the relevant sense of possibility is, if not metaphysical possibility.60 Developing the relevant modal notion may be difcult, and this might explain why some semanticians insist (in person) that they really do intend to use the notion of metaphysical possibility and hold their theories hostage to metaphysical fortune. But I think this is a mistake: developing the relevant modal notion is a critical component to these modal analyses. Of course if ones starting point is the premise that metaphysical necessity refers to whatever plays a role in these kinds of analyses, then it follows straight from the success of (say) possible worlds semantics that necessitarianism is false. But the point of the above discussion is that the premise is false: everyone (both rationalists and fundamentalists) should agree that metaphysical necessity does not (or at any rate should not) play these theoretical roles.61 My strategy here is to distinguish between different modal notions and point out that rationalism only implies necessitarianism with regards to one of them. This means that necessitarianism is weaker than we might have thought. But it is not devoid of content or interest. For we all understand the modal notion at issue: it is the sense of necessary in which essentialist truths are plausibly necessary, and the sense in which the grounds necessitate what they ground. To use our paradigm examples, then, it is the sense in which it is necessary that water is composed of H2 0, and in which it is necessary that Europe was at war if its citizens were acting in various ways. It is in that sense that rationalism implies that everything is necessary, and that thesis is striking enough. If one insists that there must be a single notion of possibility that plays a multitude of roles in philosophy then my strategy is not available. But some reason should then be given as to why we should expect all these roles to be lled by one notion and I do not know of any. Indeed the above considerations suggest that it would be a mistake to insist on one notion.
60 I applied this Spinoza test to theses about the nature of space in Dasgupta [8] and favored the rst option of giving up the modal analysis. 61 Lin [29] uses something like this premise when arguing that Leibnizs notion of per se possibility is what we mean by metaphysical possibility, for what he explicitly argues for is that Leibnizs notion is suited to playing these various theoretical roles. And since much is per se possible, he concludes that Leibnizs rationalism does not imply necessitarianism. But since I disagree with the premise I am not wholly convinced of Lins conclusion.

38

17

Burden of Proof

No doubt there are other potential objections to rationalism. But there is no room to discuss them all. Suppose for the sake of argument that the rationalist can respond to them. What follows? Does it follow that the contemporary bias against rationalism is unjustied? Perhaps not, for one might argue that the burden of proof is very much on the rationalist. The rough idea here is that it is the rationalist who makes the strong and controversial claim, not the fundamentalist. The rationalist (after all) thinks that every substantive fact has an autonomous ground while the fundamentalist merely thinks that there is a substantive fact without one. Just think (then) about all the very many ways in which rationalism could fail: it could fail because this substantive fact has no autonomous ground or because that substantive fact has no autonomous ground or. . . . So the rationalist makes a stronger claim in the loose sense that a universal generalization over a large domain is stronger than its negation. And so one might think that fundamentalism is the default position: absent compelling reasons one way or the other we should put our faith in fundamentalism. That at least is the rough worry. But it is difcult to make precise. For the obvious idea is to appeal to an indifference principle to the effect that all logically independent possibilities are equally likely, and then point out that there are many more logically possible ways for fundamentalism to be true than rationalism. And there are famous problems with indifference principles like these (especially when applied to innite domains). So the rationalist might respond to the worry along these lines. But I do not want to rest on that response for it focuses on just one way of stating the worry. And (speaking personally) the worry remains even once problems with indifference principles are pointed out: there remains a feeling that because the rationalist makes this universal claim about how all facts of a certain kind are grounded, there is a burden on her to provide serious evidence in favor of her view. So let me grant that this is the case. Still, it only follows that fundamentalism is the default view if the fundamentalist does not also incur a burden to provide evidence in favor of her view. In stating the above worry we assumed that she does not, because we assumed that she does not make a universal claim about how all facts of a certain kind are grounded. But this, I think, was a mistake: in rejecting the PSR the fundamentalist will go on to make other universal claims about how all facts of a certain kind are grounded that the rationalist need not make. So if there is a burden on the rationalist to provide evidence for her view then there is (by parity of reasoning) also a burden on the fundamentalist to provide evidence for her view. The result is that neither view is the default view: absent evidence one way or the other, it would be a mistake to put ones faith in fundamentalism. 39

The line of thought in the last paragraph rests on the idea that the fundamentalist will go on to make various universal claims about how all facts of a certain kind are grounded. Why is this so? The key lies in an important difference between metaphysical and causal explanation. Let me say what the difference is and why it is important. Consider an arrangement of particles at a given time, perhaps the arrangement that currently constitutes my laptop. And let us suppose that the arrangement has a cause, in the sense that the physical state of the world at some prior time nomically determined that the particles would come to be arranged like this. Still, it did not have to be causedat least, not if the physical laws take anything like the standard dynamical form. For such laws are consistent with any instantaneous arrangement of particles being the initial condition of a (possible) physical system, all they describe is how such a system would evolve over time. Moreover nothing about the nature of the particular particle arrangement rules out its being an initial condition of a physical system. But ground is different in this respect. Suppose for example that there is a conference occurring, and suppose (as is plausible) that this is grounded in facts about how various people act (some wrote papers, others read them, all converged in a hotel to discuss, etc). Then it is not that the occurrence of the conference just happens to be grounded, it is not that it could have been brute. Rather, if there is a conference occurring then this has to have a ground. Something about the nature of ground or the nature of conferences implies that it is impossible for there to be brute conferences. Other examples abound. Facts about football matches, the nancial crisis, and operatic performances (to take just a few cases that spring to mind) are not only grounded but had to be grounded: it is impossible for them to be brute, ungrounded facts about the world. So the difference is this. In the case of ground we do not think that any old fact can be brute. Rather, we think that all brute facts are of a certain type. For now we can leave it open how this certain type is to be characterized, so long as it excludes facts like the occurrence of a conference. The important point is just that in the case of causation there is no such restriction: any particle arrangement can be a causally brute initial condition. To see why this matters, compare our version of the PSR with a causal version that states that every arrangement of particles has a causal explanation. Suppose that A asserts that this causal PSR is true. In asserting this she makes a universal claim about the causal status of all particle arrangements. If B then denies it and says that there are causally brute facts then there is no need for B to also say that those causally brute facts are all of some type. Indeed B may have no opinion about what those causally brute facts are like. So if we assume (as we are) that those making strong universal claims carry a burden of proof then it is just A and not B who carries that burden. So rejecting the causal PSR is indeed the default position. 40

But if A asserts that our grounding version of the PSR is true then the situation is different. For suppose B denies our PSR and says that there are metaphysically brute facts, i.e. substantive facts without grounds. Then (as we just saw) B will go on to say something about what those brute facts are like, i.e. that they are all of a certain type. Of course A agrees, but for her it is a vacuous claim since she denies that there are brute facts. For B (by contrast) it is a strong universal claim about a non-empty class. So both A and B assert a strong universal claim about some non-empty class. So if A incurs a burden of proof because she asserts a strong universal claim about a non-empty class, then (by parity of reasoning) so does B. So neither view is the default position, both views require evidence to warrant our belief. In this way the difference between metaphysical and causal explanation leads to this dialectical difference between the two versions of the PSR. I said that B (who rejects the PSR and believes that there are brute facts) will say that every brute fact is of a certain type. To get a feel for how strong this claim is it might help to describe some ways in which this type might be characterized. One idea is to follow Sider and introduce the primitive notion of somethings being natural.62 Perhaps the natural things include electrons and the property of being negatively charged while the unnatural things include football matches and the property of being a stable economy. Then Bs universal generalization might be construed as the claim that every brute fact contains only natural constituents.63 This is a very strong claim: it rules out all (logically possible) chains of ground that bottom out in things that are not perfectly natural like football matches and operas and conferences and the like. That this is a strong claim can be missed if by natural you read constituent of an ungrounded fact in which case it is a logical truth. But on Siders approach natural is a primitive notion not dened in terms of ground. So the italics express the strong claim that any brute fact is constituted just of entities of a certain (independently specied) kind. Alternatively, one might try to characterize Bs claim in terms of the notion of essence. When we noted that facts about conferences are necessarily grounded it may be have been tempting to think that this is due to the nature of conferences, i.e. that it is essential to conferences that facts concerning them are grounded. Conversely then one might think that the same goes for brute facts. Suppose that the fact about the charge of a given electron is brute. Then one might say that it is essential to being an electron and being negatively charged that if x is an electron that is negatively charged then this is a brute fact. So Bs universal generalization might be (roughly) construed as the claim that every brute fact is essentially brute in
Sider [37]. He uses the term fundamental but that term is already in use here so is likely to be misleading. Siders notion is (as he emphasizes) a generalization of Lewis notion of naturalness so my switch in terminology is not entirely out of place. 63 This is very close to Siders principle of purity.
62 See

41

this sense. Once again this is a very strong claim: it rules out all grounding explanations that bottom out in facts that are not essentially brute (like the occurrence of a conference). In any case there is no need to decide between these two ways of eshing out Bs claim. The point is just that she willon pain of admitting the possibility of brute conferences!make some strong claim of this kind. So if we grant that those making a strong universal claims carry a burden of proof then it follows that both the rationalist and her opponent B carry this burden and that neither is the default view. Now even if B carries a burden of proof it does not quite follow that the fundamentalist carries a burden of proof, for the fundamentalist (i.e. someone who rejects our PSR) is not forced to agree with B that there are brute facts. She might instead think that there is a non-terminating descending chain of grounds for every substantive fact. But this is (of course) another strong universal generalization! So the way to put the point is this. There are two ways in which our PSR could fail: either every substantive fact is grounded in a non-terminating descending chain, or there are brute facts and all of them are of a certain type. Each way consists in a strong universal generalization over a non-empty class. So if the rationalist incurs a burden of proof because she asserts a strong universal claim, then so does someone who rejects the PSR in either of these ways. Of course the fundamentalist might just make the disjunctive claim that our PSR fails in one of these two ways, and this disjunction is (in the logical sense) weaker than either disjunct. But still, it is hard to see why this would absolve her of the burden of showing that the disjunction is correct. The upshot is this. With the casual version of the PSR, it is perhaps correct to say that rejecting it is the default position. But even so, the difference between causal and metaphysical explanation means that when it comes to our grounding version of the PSR the dialectic is different. It would be a mistake to think that either the rationalist or fundamentalist position is default, so the burden is on the proponent of each view to provide evidence in its favor.

18

Conclusion

Our choice then is between two strong views, rationalism and fundamentalism. How are we to decide between them? We could of course wait until we settle on our best nal explanatory theory of the world and see which one it implies. But can we do better? Can we argue prior to settling on such a theory that one of these views must be correct? It is easy to be pessimistic. For each view is a strong and universal claim about grounding explanations. And I take it that neither is an an-

42

alytic or conceptual truth.64 So one might wonder what reason we could have to expect to nd either pattern in nature. What after all could such a reason be founded in? The deliverances of some miraculous faculty of insight by which we peer into the explanatory structure of the world? The worry is reminiscent of the problem of induction. For the claim that causal patterns observed in the past will repeat themselves in the future is a similarly strong and synthetic claim about the causal structure of the world, and it is famously difcult to answer the Humean worry that we cannot have reason to believe it without waiting to see how the future pans out. Any argument for rationalism or fundamentalism will have to confront these difcult questions in epistemology. For this reason I have left an argument one way or another for another time.65 Instead, the provisional attitude I recommend is one of humble agnosticism. Still, what I have tried to argue is that the anti-rationalist bias prevalent in contemporary metaphysics is unwarranted, an article of faith rather than reason.66

References
[1] Adams, R.M. 1983. Divine Necessity. Journal of Philosophy 80 (11) 741 752. [2] Adams, R.M. 1994. Leibniz: Determinist, Theist, Idealist. Oxford: OUP. [3] Albert, D.Z. 1996. Elementary Quantum Metaphysics. pp. 277284 in Bohmian Mechanics and Quantum Theory: An Appraisal, edited by Cushing, Fine and Goldstein. Dordrecht: Kluwer. [4] Bennett, J. 1984. A Study of Spinozas Ethics. Indianapolis: Hackett. [5] Bliss, R. Viciousness and the Structure of Reality. Philosophical Studies. Online First. [6] Chalmers, D. J. 2011. Freges Puzzle and the Objects of Credence. Mind 120 (479): 587635.
64 Here Della Rocca would disagree and would urge that the PSR is itself a conceptual truth (see his [12]). But this disagreement stems from him using a different notion of explanation than I am. 65 For some arguments for rationalism, see Della Rocca [13] and Pruss [32]. I am not wholly moved by those arguments but this is not the place to say why. 66 Versions of this paper were presented at the Carolina Metaphysics Workshop (June 2012), at the philosophy department colloquia at Rochester University and MIT (both in November 2012), and at the Princeton rst-year seminar in November 2012. Many thanks to all involved for those helpful discussions, especially to Dan Korman and Nat Tabris for their detailed comments on the paper. Thanks also to Elizabeth Barnes, Ricki Bliss, Darren Bradley, Jonathan Cottrell, Sinan Dogramaci, Michael Della Rocca, John Morrison, Jonathan Schaffer, Ted Sider, James Stazicker, and Michael Strevens for their valuable comments on previous drafts of this paper.

43

[7] Dasgupta, S. 2009. Individuals: An Essay in Revisionary Metaphysics. Philosophical Studies 145(1): 3567. [8] Dasgupta, S. 2011. The Bare Necessities. Philosophical Perspectives 25: 115160. [9] Dasgupta, S. On the Plurality of Grounds. Manuscript. [10] Dasgupta, S. The Status of Ground. Manuscript. [11] Della Rocca, M. 2002. Spinozas Substance Monism. In Spinoza: Metaphysical Themes, edited by O. Koistinen and J. Biro. Oxford, UK: OUP. [12] Della Rocca, M. 2003. A Rationalist Manifesto: Spinoza and the Principle of Sufcient Reason. Philosophical Topics 31 (1): 7593. [13] Della Rocca, M. 2010. PSR. Philosophers Imprint 10 (7): 113. [14] Fine, K. 1994. Essence and Modality: The Second Philosophical Perspectives Lecture. Philosophical Perspectives 8: 116. [15] Fine, K. 1995. Ontological Dependence. Proceedings of the Aristotelian Society (95): 269290. [16] Fine, K. 1995. The Logic of Essence. Journal of Philosophical Logic 24 (3): 241273. [17] Fine, K. 2001. The Question of Realism. Philosophers Imprint 1 (1): 130. [18] Fine, K. Necessity and Non-Existence. In his Tense and Modality, pp. 321354. Oxford, UK: OUP. [19] Fine, K. 2012. Guide to Ground. pp. 3780 in Metaphysical Grounding: Understanding the Structure of Reality, edited by F. Correia and B. Schnieder. CUP. [20] Frankfurt, H. 1971. Freedom of the Will and the Concept of a Person. The Journal of Philosophy 68 (1): 520. [21] Gendler, T. and J. Hawthorne. 2002. Conceivability and Possibility. Oxford: OUP. [22] Hudson, H. 2006. Brute Facts. Australasian Journal of Philosophy 75(1): 7782. [23] Hawthorne, J. (OLeary-) and A. Cortens. 1995. Towards Ontological Nihilism. Philosophical Studies 79(2): 143165.

44

[24] Hofweber, T. 2009. Ambitious, Yet Modest, Metaphysics. In Metametaphysics: New Essays on the Foundations of Ontology, edited by D. Chalmers, D. Manley, and R. Wasserman. OUP. [25] Jenkins, C. 2011. Is Metaphysical Dependence Irreexive? The Monist 94(2): 88-99. [26] Kment, B. Modal Concepts and Causal Inquiry. In preparation. [27] Koslicki, K. Essence, Necessity and Explanation. Forthcoming in Contemporary Aristotelian Metaphysics, edited by Tuomas Tahko. [28] Leuenberger, S. Grounding and Necessity. Manuscript. [29] Lin, M. 2012. Rationalism and Necessitarianism. Nous 46 (3): 418448. [30] Lin, M. The Principle of Sufcient Reason in Spinoza. Forthcoming in The Oxford Handbook of Spinoza. [31] Maudlin, T. 1988. The Essence of Space-Time. PSA: 8291. [32] Pruss, A. 2006. The Principle of Sufcient Reason: A Reassessment. CUP. [33] Raven, M. In Defense of Ground. Forthcoming in Australasian Journal of Philosophy. [34] Rosen, G. 2006. The Limits of Contingency. In Identity and Modality, edited by F. MacBride. OUP. [35] Rosen, G. 2010. Metaphysical Dependence: Grounding and Reduction. In Modality: Metaphysics, Logic, and Epistemology, edited by B. Hale and A. Hoffmann. OUP. [36] Rowe, W. L. 1975. The Cosmological Argument. Princeton, NJ: Princeton University Press. [37] Sider, T. 2011. Writing the Book of the World. Oxford: OUP. [38] Spinoza, B. The Ethics. [39] van Inwagen, P. 2002. Metaphysics. Westview Press. [40] Weaver, C. 2012. What Could Be Caused Must Actually Be Caused. Synthese 184 (3): 299317. [41] Weaver, C. A Church-Fitch Proof for the Universality of Causation. Forthcoming in Synthese.

45

On the Plurality of Grounds

Shamik Dasgupta
Draft of July 2013, comments welcome

Recent metaphysics has contained a good deal of discussion about the notion of ground. The notion is intuitive enough: it is sometimes said that Europes being at war in 1939 was grounded in the actions of its citizens, meaning (something like) that it was in a state of war in virtue of those actions, or that those actions are what made it the case that it was at war, or that its being at war is explained by those actions. Regardless of whether this claim about the wars ground is true, we have a reasonably strong pre-theoretic grasp of what it means. One reason why the notion of ground has commanded such attention is the idea that it is needed to formulate many core philosophical issues. Consider for example theses like materialism about consciousness, normative naturalism, and phenomenalism. These claim that certain factsabout conscious states, norms, and external objects (respectively) in some sense arise out of or are determined by or xed by various underlying factsabout my brain, or natural properties, or sense data (respectively). But how should this talk of determination or xing be understood? One might suggest that it be understood in terms of supervenience, or analysis, or identity. But a number of philosophers are attracted by the idea that it is best understood in terms of ground.1 On this view the above theses are theses to the effect that the material state of my brain grounds my conscious states, that the natural facts ground the normative, and that patterns of sense data ground the existence of external objects. I will not rehearse the reasons for preferring these grounding formulations. But given its potential philosophical importance, it is unsurprising that there has been much interest recently in the notion of ground itself. In this spirit, one aim of this paper is to argue that ground is irreducibly plural. Now it is already recognized that somethings ground can be a pluralityEuropes being at war in 1939 is a good example of something that is presumably grounded in a multitude of facts concerning the actions of its many citizens during that time. Those facts together are what explains its being at war even though none of them is a sufcient explanation in1 See

for example Fine [5], Rosen [22] and Schaffer [25].

dividually. But the literature uniformly assumes that what is grounded must be a single fact. Here I disagree and argue that what is grounded can be a plurality too: there can be cases in which they, the members of a plurality, are explained in more fundamental terms even though none of them admits of explanation on its own. Now if ground is irreducibly plural in this sense then this is important to know about. For (as I said) fans of ground are tempted to see much contemporary philosophy as attempting to establish whether facts of one type (say, the natural) are sufcient to ground facts of another (say, the normative). And a natural strategy of arguing in the negative is to argue that a given fact of the latter type cannot be grounded in facts of the former type. But if ground is irreducibly plural then this form of argument is invalid. For even if one were to successfully argue that there is no natural ground of the the fact that I ought not eat meat, it would remain open that the normative facts taken together have a natural ground in which case normative naturalism would be vindicated after all. As we will see, this invalid form of argument may be responsible for certain popular views in metaphysics, in which case it is important that the mistake be exposed. My claim that ground is irreducibly plural is a claim about the logical form of ground. It is the claim (to be claried below) that, logically speaking, ground is a binary relation plural in both positions: they are grounded in them. Of course the limit case is a plurality of one, so it may turn out (as it happens) that in each case of ground we have a single fact grounded on its own. Still, on my view the claim in each case would strictly speaking remain plural: that they (all one of them!) are grounded in them. However I believe that there are actual examples in which many facts are grounded together. Consider the individualistic facts, facts that concern particular individuals such as Saul Kripke is a philosopher Barack Obama is president This very book (pointing at the book on my table) is interesting I believe that these facts together are (plurally) grounded in the purely qualitative nature of the world even though none of them has a qualitative ground when taken on its own. Or consider facts about the mass-inkilograms of things, such as The president is 75 kgs The laptop is 2 kgs The book next to the laptop is 1/2 kg.

Once again, I believe that these facts are (plurally) grounded in the mass relationships between things even though none of them has such a ground when taken alone. Indeed the same goes (I claim) for distance-in-meters, time-in-seconds, preferences-in-utils, and other cases in which there are mathematical values of a given quantity in a given scale. In each case my view is that facts about the mathematical values in a given scale are plurally grounded in the underlying, scale-independent facts (about geometry, time, or preferences respectively). But I will focus on the case of individuals and mass-in-kilograms in what follows. The result is a structuralist view of individuals and kilograms respectively, since an account of any one member of the group is inevitably an account of them all. (I apologize for the term. Structuralism is already applied too widely. I will discuss the relation between my views and other views that go by the name as we go along.) The trouble is, neither the case of individuals nor the case of mass is an uncontroversial example of plural grounding. For one thing, each structuralist view implies a strong claim that there is no room to motivate here, namely that the world is fundamentally qualitative (in the rst case) and that mass is fundamentally relational (in the second). So I cannot very well argue that ground is plural just by pointing at these examples! And I do not know of uncontroversial examples to appeal to instead. So the argument will have to be indirect. I will start by arguing for the conditional claim that if the world is fundamentally qualitative then the individualistic facts are plurally grounded in the qualitative (sections 25). Then I will argue (on the basis of the very same kinds of considerations) that if mass if fundamentally relational then the kilogram facts are plurally grounded in those mass relationships (sections 68). Sections 910 then develop each structuralist view and respond to objections. This will not establish that either structuralist thesis is true (since I will have only motivated the conditional claims), but it will suggest that these structuralist hypotheses are coherent and intelligible and worth taking seriously. And so our view about the logical structure of ground should allow for these hypotheses: we should think that ground is irreducibly plural. Though I only argue for the conditional claims here, I have in other work argued for the antecedent of each conditional.2 So this paper completes the argument for structuralism in each case. The paper therefore has three distinct topicsthe nature of ground, the nature of individuals, and the nature of quantities like massbut each topic will inform the others.
2 See XXXX [reference omitted for blind review] for an argument that the world is fundamentally qualitative. And see XXXX [reference omitted for blind review] for an argument that mass and other quantities are fundamentally relational.

More on Ground

It is important to clarify the notion of ground at issue before we start. As I use the term ground is an explanatory notion: to say that X grounds Y is to say that X explains Y, in a particular sense of explains. To illustrate, imagine asking why Europe was at war in 1939. A causal answer might describe events during the preceding 50 years that led, say, Chamberlain to declare war on Germany. But there is another kind of answer that would try to say what goings-on in the continent made it count as a continent at war in the rst place. Someone in search of this second answer recognizes that being at war is not a sui generis state of affairs, so that if Europe was at war there must have been some underlying facts about the continent in virtue of which it counted as being a continent at war rather than (say) one at peace. Presumably it has something to do with how its citizens were acting, for example that some were committing various politically motivated acts of violence against others. An answer of this second kind is a statement of what grounds Europes being at war in 1939. Or imagine asking why there is a table in the room. A causal answer might be that someone recently placed it there. But there is another answer that recognizes that tables are not sui generis items and that if there is a table in the room then that must be in virtue of various other facts, for example that various micro-particles are arranged in a table-like way. Again, an answer of this second kind is a statement of what grounds the fact that there is a table in the room. We should distinguish between full and partial explanations. A single explosion in the Rhine might partly explain why Europe was at war, but it does not fully explain it. By ground I mean a full explanation.3 What is the logical form of a claim about grounds? Explanations are typically expressed with the sentential operator because: Europe was at war because its citizens were acting in various ways. So one standard view is that the logical form of a claim about ground is S because where S is a sentence, is a list of sentences, and because is read in the metaphysical rather than causal sense. Informally, the sentences in describe those aspects of the world that together explain its being the case that S. It is important that is a list and not a conjunction: we would like to make sense of the plausible view that conjunctions are grounded in their conjuncts, but if were a conjunction this would collapse into the view that a conjunction is grounded in itself.4
that a given fact may have more than one (full) ground. For example, if P and Q both obtain, then PQ is (fully) grounded in P, and also in Q. 4 This logical form of ground is suggested by Fine [5] and [6].
3 Note

Now, this logical form presupposes what I call a singularist view of ground, according to which any aspect of the world that admits of explanation can be explained on its own. I will be arguing that singularism is wrong and that sometimes a plurality of aspects taken together can be explained even though none of them can be explained when taken alone. But this pluralist view of ground is unintelligible given the above logical form, so the pluralist will instead take the logical form to be because where both and are lists of sentences.5 Informally, the aspects of the world described by the sentences in are explained, when taken together, by the aspects described by the sentences in , even though there is no presumption that each sentence in describes something that can be explained on its own. According to the pluralist, the singularist mischaracterized the logical form by generalizing from special cases in which the number of sentences in is one. My ofcial approach is to treat ground as a sentential connective, but it streamlines prose to treat it as a relational predicate that applies to facts. Since my talk of facts is just a convenient shorthand there is no need to say much about what facts are, but I will assume that they are reasonably negrained and that logically equivalent facts can be distinct.6 On this way of talking, a singularist will take the logical form of a grounding claim to be Y is grounded in the Xs where Y is a singular variable and the Xs is a plural variable, both ranging over facts.7 But the pluralist will instead let Y be a plural variable and replace is with are: she says of the many Ys that they are grounded in the Xs, with no presumption that each Y has a ground on its own. I make two assumptions about ground. The rst is that the grounded is metaphysically necessitated by its grounds. More formally: If because , then it is metaphysically necessary that if then where X is the conjunction of the sentences in the list X. This principle (or, more precisely, the singularist restriction of it) is endorsed in much of the recent literature on ground.8 And it has some plausibility: if Europes being at war in 1939 was grounded in the actions of its citizens at that time, then those actions are what made it the case that Europe was at war,
the singularist and the pluralist will likely allow the lists to be innite. a discussion of how ne-grained the notion of fact must be in the context of questions of ground, see Correia [3]. 7 This logical form of ground is endorsed by Rosen [22]. 8 See for example Fine [6] and Rosen [22]. For an extended argument in its favor see Trogdon [31].
6 For 5 Both

and are that in virtue of which it was at war. But then (the idea is) it must be impossible for those citizens to have acted in those ways and yet for Europe to have been at peace. To be sure, the principle is controversial and some have argued that it is false.9 But I will not discuss the issue here: in this paper the principle is an unargued assumption. I do not assume the reverse scheme, since there can be necessary connections without grounds: it is metaphysically necessary that if Obama exists then 2+2=4, but Obamas existence does not explain why 2+2=4. Nor do I assume that the grounded necessitates its ground, since a disjunction may be grounded in one of its disjuncts without necessitating it. My second assumption is that all parts of an explanation must be explanatorily relevant: if the Xs ground the Ys and x is one of the Xs, then x is explanatorily relevant to the Ys in the sense that x plays at least some role in making it the case that the Ys obtain.10 This assumption is natural in the case of causal explanation: even if the oil spill in the gulf was causally explained by BPs negligence, it is not causally explained by BPs negligence and the number of electrons in Alpha Centauri, for surely the latter is entirely irrelevant to the matter in the sense that it played no role in bringing about the spill. I assume the same for ground: even if Europes being at war is grounded in various facts about how its citizens acted, it is not grounded in those actions and the fact that 2 is prime, for the latter played no role in making the continent count as being at war and so is irrelevant to the matter. This requirement of relevance is widely endorsed in the recent literature. Indeed it is one of the central features used to distinguish ground from (say) metaphysical necessitation and logical consequence.11 To be clear, if X is explanatorily relevant to Y, this does imply that every explanation of Y appeals to X. For example, if P and Q both obtain then PQ is (fully) grounded in P, and also (fully) grounded in Q. So not every explanation of PQ appeals to P, but P is explanatorily relevant to PQ in the sense I have in mind. It is important that we do not dene a fact x to be explanatorily relevant to the Ys iff x is one of some Xs that ground the Ys, for then my assumption would become a tautology and loose its teeth. This is not the place to discuss whether the notion of relevance can be dened otherwise: for the purposes of this paper I take it to be another primitive alongside ground.12 My assumption is therefore a substantive principle linking two distinct
for example Leuenberger [14], Schaffer [26], and Skiles [28]. I say that x must be relevant to the Ys I mean this in the non-distributive sense: I do not assume that x must be relevant to each of the Ys individually or even to just one of the Ys. For the root idea is that explanans must be relevant to the explanandumso the natural extension of that root idea if ground is irreducibly plural is that explanans must be relevant to the (perhaps many) explananda in this non-distributive sense. 11 For example, see Fine [6] and Rosen [22], where this assumption is explicit and central to each authors conception of ground. 12 Fine [6] argues that ground and relevance cannot be dened in terms of each other.
10 When 9 See

notions. The assumption is not beyond doubt, but I will not defend it here: in this paper it is an unargued starting point. It is sometimes assumed that ground is transitive.13 What does this mean? We know what it is for a binary singular relation to be transitive, but what about a binary plural relation? One can formulate a number of transitivity-like principles, but I will not assume any of them here. Still, it will be useful to speak of the transitive closure of ground on one sense of the term. To this end let us stipulate that ground is transitive iff: (i) If the Xs ground the Ys, and the Ys along with the Y*s ground the Zs, then the Xs along with the Y*s ground the Zs, and (ii) If the Xs ground the Ys along with the Y*s, and the Ys ground the Zs, then the Xs ground the Y*s along with the Zs.14 Then call the transitive closure of ground the notion of derivative ground.

Individualism and Qualitativism

I now turn to my rst conditional claim: that if the world is fundamentally qualitative then the individualistic facts are plurally grounded in the qualitative facts (sections 25). What do I mean by individualistic and qualitative? I will not try to dene these terms, but roughly speaking a fact is individualistic iff whether it obtains depends on how things stand with a particular individual (or individuals), and qualitative otherwise.15 For example, facts of these forms are all individualistic facts: a is F a bears R to b a6=b where a and b are individuals, since whether they obtain depends on how things stand with a (and, for the latter two, with b). To illustrate some qualitative facts it will help to use the notion of a qualitative property. Roughly, a property or relation is qualitative if it does not concern any particular individual. For example, being seated, having a sister, and having two sisters are all qualitative: even if ones having these properties
13 Though not always. Rosen [22] remains agnostic, and Schaffer [27] and Tahko [30] have both argued against transitivity. See Litland [16] for a response to Schaffers arguments. 14 Thanks to Daniel Berntson for help in formulating this principle. 15 More precisely: a fact F is individualistic iff there are some Xs such that whether F obtains depends on how things stand with the Xs. But I will continue to use the more readable expression in the text. A complete denition would need to rene the notion of dependence, among other things, but I think the idea is clear enough for us to work with.

implies the existence of other individuals, they do not concern any particular individual. These contrast with non-qualitative properties such being Kripke and being Obamas sister, which concern the individuals Kripke and Obama respectively. Then qualitative facts include facts that can be expressed in the language PL of predicate logic with identity but without constants, such as (x)Fx (x)(y)(Fx & Gy & x6=y) (x)(Fx Gx) so long as F and G express qualitative properties. For whether the rst obtains does not depend on how things stand with any particular individual, it just depends on there being some individual or other that is F. Similar remarks apply to the others. If you agree with the bundle theorist that there are facts about which qualitative properties are compresent then these would also be qualitative facts. One might of course try to dene the distinction between individualistic and qualitative facts in more detail, but the intuitive idea glossed here is sufcient for our purposes.16 Now, of the qualitative and the individualistic, which are the more fundamental? A natural view is that the most fundamental facts of our world are individualistic facts about how a domain of individuals are propertied and related to one another, and that they are sufcient to ground (or at least derivatively ground) the qualitative facts. Let us call this individualism. In contrast, let qualitativism be the opposite view that the most fundamental facts are qualitative facts and that they are sufcient to ground (or at least derivatively ground) the individualistic facts. Qualitativists may disagree on what kind of qualitative facts one nds at the bottom level: the traditional bundle theorist says that they concern which monadic, qualitative properties are compresent, but other qualitativists may think that they are the facts that can be expressed in predicate logic with identity (but no constants). And other qualitativist views are possible too.17
would only add that I do not count tropes as qualitative properties. Also, while the individualistic facts above are all expressed with referring terms, qualitative facts may be expressed with referring terms too so long as they do not refer to particular individuals but (say) to qualitative properties. 17 I motivated another version of qualitativism in XXXX [reference omitted for blind review]. And a different version has been developed by Paul in a number of papers, including [20] and [19]. A variety of qualitativist views are explored in Hawthorne and Sider [10], though not endorsed. I have not mentioned views that eliminate individualistic or qualitative facts altogether. This is because the recent interest in ground is largely driven by the vision that the benets of eliminativist views can be enjoyed by more plausible views about what grounds what.
16 I

Individualism is perhaps the more natural position. Suppose that an individual X is both red and round. It follows that something is red and round. But it is natural to think that something is red and round because X is red and round, just as the individualist says. However, I favor qualitativism. Very briey, my reason is that if individualism were true then the individualistic facts of our world would lie beyond our epistemic ken. The rough idea is that our knowledge of the world is limited to knowledge of its qualitative nature and whatever is grounded in that qualitative nature, and since individualism implies that there are further facts of the matter as to which particular individuals lie behind those qualities it follows that those facts would be unknowable. A reasonable Occamist principle then recommends that we dispense with such epistemically inaccessible facts.18 However, my aim here is not to argue for qualitativism but instead to argue that if qualitativism is true then individualistic facts are plurally grounded in the qualitative, not one by one. Before arguing for this it is important to ward off two potential misconceptions as to what qualitativism is. First, it is often said that the traditional bundle theorist cannot make sense of situations in which distinct individuals have the very same monadic qualitative properties. An example of such a situation is the infamous Max Black world, a world in which there are just two spheres of iron 2 miles apart that share all their intrinsic qualitative properties (they are exactly the same size, shape, color, etc). If the traditional bundle theorists view is that each sphere is identical to the collection of its monadic, qualitative properties, then since (by hypothesis) both spheres have exactly the same such properties it follows that they are identical and so there are not 2 distinct spheres after all. But as I emphasized three paragraphs back the traditional bundle theorist is just one kind of qualitativist. Other qualitativists might think that the underlying qualitative facts are those expressed by predicate logic with identity (but no constants), in which case there is no difculty describing such a situation with (roughly speaking) something like the following: (x)(y)(Fx & Fy & x is two miles from y & x6=y)

18 Roughly this line of argument can be traced back to Leibniz. I develop it in some detail in XXXX [reference omitted for blind review]. Note that there is no epistemic problem for the qualitativist precisely because she thinks that individualistic facts are grounded in qualitative facts and so she denies that they are (in the relevant sense) further facts about the world beyond its qualitative nature. The epistemic premisethat knowledge is limited to the qualitative and what is grounded in the qualitativeis controversial (to say the least!) and deserves more discussion than I can give it here. But, just to be clear, it does not (without other assumptions) rule out de re or rst-personal knowledge, and it does not say that we acquire such knowledge by deriving it from qualitative knowledge. It just says that whatever is known in those cases must ultimately have a qualitative ground.

where F expresses the complete intrinsic qualitative nature of each sphere. Yet other qualitativists might think that the underlying qualitative facts have some other structure, but they might nonetheless think that they are rich enough to make sense of these kinds of situations.19 In any event, the potential misconception to ward off is the idea that qualitativism per se rules out such situationsit does not, even if some versions of it do.20 The second potential misconception to ward off is the idea that qualitativism is just anti-haecceitism (and likewise that individualism just is haecceitism). This is a mistake. For anti-haecceitism (at least as characterized in the recent literature) is a modal thesis, a thesis to the effect that there can be no difference in the way the world is individualistically without a qualitative difference. Admittedly, the term anti-haecceitism has been used for a number of related modal claims, some expressed with modal operators and others with quantication over worlds.21 But they are all modal claims, not grounding claims, and so none of them imply qualitativism for the reason that a necessary connection does not imply a connection of ground: as I said in section 1, if the Xs necessitate Y this does not imply that the Xs ground Y. Of course it follows from my assumption that the grounded is necessitated by its grounds that qualitativism implies anti-haecceitism (in at least one of its characterizations). So if you are an anti-haecceitist this might be because you are a qualitativist. But it might instead be because you hold independent theses about modality that imply anti-haecceitism, even though you are an individualist.
19 I develop a qualitativist view along these lines in XXXX [reference deleted for blind review]. 20 Thus qualitativism per se does not imply the Principle of the Identity of Indiscernibles (PII), the principle that indiscernible things are identical. This principle comes in a variety of different avors depending on which notion of discernibility is in use. Objects x and y are absolutely discernible (roughly speaking) iff there is a monadic qualitative property that x has and y does not. And objects x and y are weakly discernible (again, roughly speaking) iff x and y stand in an irreexive relation to one another. (These notions of discernibility were claried and brought to bear on issues in the philosophy of physics by Saunders [23].) The two spheres in the Max Black world are absolutely indiscernible since they share all their monadic qualitative properties, both intrinsic (e.g. being brown) and relational (e.g. being 2 miles from an iron sphere). But they are weakly discernible since they each stand in the irreexive qualitative relation of being two miles from with something. The point in the text (then) is that qualitativism does not imply the PII stated with the notion of absolute discernibility. Nor, we can now add, does it imply the PII stated with the notion of weak discernibility. For even if every qualitative relation that the two spheres stand in is one that each sphere stands in to itself, it remains the case that such a situation could be expressed in predicate logic with identity (but no constants). 21 Lewis ([15], Chapter 4) characterized anti-haecceitism as the view that any two possible worlds that agree qualitatively agree about what they represent de re of any given individual. Others (for example Pooley [21] and others from the philosophy of physics literature) characterize it as the view that possible worlds that agree qualitatively are identical. Yet others (including Skow [29]) characterize it with modal operators.

10

I believe this latter position was Lewis. He was an anti-haecceitist (in at least one of its guises) but was he a qualitativist? I suspect not. This is not altogether clear since he never wrote explicitly in terms of ground. But someone with qualitativist inclinations (regardless of whether they speak in terms of ground) would be likely to endorse the traditional bundle theory or develop some other theory of what the underlying qualitative facts are like, and Lewis never did this. Indeed he had the perfect opportunity to offer such a theory when telling us what a possible world is in Chapter 1 of On the Plurality of Worlds, where he could have said that a possible world is some kind of collection of qualitative universals. But instead he tells us that it is a mereological sum of individuals. So while Lewis is an anti-haecceitist, I see no evidence that he was a qualitativist. In my view his anti-haecceitism is best understood as following from his views about the nature of de re modality, not his views about nature of individuals. So qualitativism is an explanatory (and not a mere modal) claim. And like any explanatory claim it faces the challenge of showing that the explanans really are sufcient to explain the explanandumin this case that the fundamental qualitative facts really are sufcient to explain the individualistic facts. The anti-haecceitist faces no such challenge since she only asserts a modal connection and not an explanatory connection. But for the qualitativist, meeting the challenge is crucial: if she cannot meet it then she would have to be an eliminativist about individualistic facts and claim that there are no such facts! Insofar as this is an intolerable consequence, meeting the challenge is crucial for the success of qualitativism. It is this challenge that I take up here. My thesis is that the qualitativist does indeed face signicant difculties in meeting this challenge if she tries to ground each individualistic fact one by one, but that these problems dissolve if she grounds them plurally.

Finding Obama in a Qualitative World

To see why, recall that the qualitativist says that the individualistic facts are derivatively grounded in the qualitative nature of the world. This implies that there is a non-empty set S of individualistic facts that are grounded, and not just derivatively grounded, in the qualitative.22 It will help to work with an example so let us suppose (without loss of generality) that S contains the fact that Barack Obama exists. My opponent thinks that this fact when taken alone is grounded in some set Q of qualitative facts. But what could Q possibly be? I will argue that any candidate set Q that necessitates Obamas existence contains irrelevant facts; or, contrapositively, that if Q is restricted to facts that are relevant to his existence then it will not neground is transitive then S is the set of all individualistic facts, while if it is not then it may be a proper subset; but for our purposes all that matters is that S is not empty.
22 If

11

cessitate his existence. So my two assumptionsthat a ground must both necessitate and be relevant to what it groundspull in opposite directions and cannot be jointly satised. Then I will show that these problems dissolve if we plurally ground all the facts in S together. Let us start by trying to construct a candidate set Q. We might start with facts concerning the existence of something with just a few of Obamas qualitative properties, such as being born on a small island and being well educated. But this would clearly not be sufcient to ground Obamas existence since it is possible for someone to have those qualities and yet for Obama not to exist. So let us try adding to Q more facts about Obamas qualitative nature. To this end, let R be a bounded region of space-time containing Obama, perhaps the region lled by the entire history of our solar system. And let QR be the set of facts characterizing the entire intrinsic nature of R in its most fundamental qualitative respects. Does QR ground Obamas existence? No, because it does not necessitate his existence. For it is possible for there to be a region of space-time R* disjoint from R which agrees intrinsically with R in all its most fundamental qualitative respectsi.e. in which all the facts in QR obtainbut which contains different individuals. It might help to imagine (though this is not crucial to the argument) that R* is spatio-temporally far removed from R. Moreover it is possible for there to be such a region R* and yet for R to differ in such a way that Obama never existed: perhaps all we need to suppose is that his mothers egg was fertilized by a different sperm. Since all the facts in QR would obtain in this possibility, it follows that QR does not necessitate Obamas existence and hence does not ground his existence either. The possibility I describe here is not controversial. It is uncontroversial that in Moscow there could be an intrinsic duplicate of my laptopcall it l . And it is uncontroversial that l could exist unchanged even if my laptop differed in some intrinsic respect (perhaps it lost a key). I am just making an analogous modal claim about the spatio-temporal region R.23 This is not to deny that there are facts about the nature of R that could explain his existence. Essentialists about origins might explain his existence by the fact that a particular sperm fertilized a particular egg within R. Others might explain his existence in terms of some particular fundamental particles in R that compose him. I have no objection to these explaa counterpart theorist make sense of this possibility? Certainly she can. The possibility in question is represented by a world W in which there is a laptop x that is a perfect duplicate of mine and a laptop y that closely resembles mine but has lost a key. In suitable contexts both x and y will resemble my laptop enough to be its counterparts. In that case W stands in for two possibilities: one in which x represents my laptop and another in which y represents my laptop. The latter is a possibility in the duplicate laptop exists even though my laptop has lost a key. See Lewis [15], chapter 4, for more on this use of counterpart theory whereby one world stands in for many possibilities.
23 Can

12

nations, but they are not available to the qualitativist since they both appeal to individualistic facts. What the above argument shows is that nothing about the qualitative nature of R could ground Obamas existence.24 What then must be added to QR ? It is no use adding a fact that is necessitated by QR itself, such as generalizations that are grounded in QR , for the resulting set would still not necessitate Obamas existence (if it did, then QR would necessitate his existence on its own). But QR was a complete characterization of the intrinsic nature of our entire solar system in its most fundamental qualitative respects. So to necessitate Obamas existence we must add facts about the qualitative nature of the cosmos outside our solar system. We might for example add facts concerning the qualitative nature of some region in Alpha Centauri. But the problem is that even if the resulting set necessitates Obamas existence, those goings-on in Alpha Centauri seem entirely irrelevant when it comes to explaining his existence. Surely what happens in Alpha Centauri plays no role whatsoever in making it the case that Obama exists. If I asked What metaphysically explains Obamas existence? and you started talking about Alpha Centauri, I would likely assume that you had misunderstood the question since your answer would be too bizarre to take seriously! Remember, the kind of explanation at issue here is metaphysical, not causal. Facts about the goings on in distant regions of spacetime might be relevant to a causal explanation of how heavy elements came into existence and therefore what caused Obama to exist. But we are asking for a grounding explanation of Obamas existence and it is almost inconceivable that the correct answer could include the goings on outside our solar system.25 My premise is that these facts about the universe outside R are irrelevant to the matter.26 I will support the premise in section 4 below, but it is very plausible. For recall how natural it would be to explain Obamas existence in terms of facts about the particular fundamental particles that compose him, or the fact that a particular sperm fertilized a particular egg. This explanation is not available to the qualitativist, but the fact that it is
might think that the solution is to qualitatively ground those individualistic facts about the particles or the egg and sperm. But exactly the same problems will recur. This is just to say that the current argument does not depend on my decision to discuss the fact that Obama exists and will apply to other individuals too. 25 The question here is not what is relevant to whether a denizen of another possible world W is a counterpart of Obama. It may be that facts about the entire nature of the actual world and W are relevant to this question, but this question is not ours. Our question is what is relevant to an explanation of Obamas existence. 26 By facts about the universe outside R I include specic facts about particular regions, for example about particular electrons in Alpha Centauri. But I also include general facts such as that every region outside R has certain characteristics. The inclusion of the latter makes sense because they will (plausibly) be grounded in the former, so that if Obamas existence is grounded in the latter it will be derivatively grounded in the former. So, if the former are objectionable in an explanation of Obamas existence, then so too are the latter.
24 One

13

so natural shows that we take facts about goings-on outside our solar system to have nothing to do with the matter. If one resists my premise, one resists a very plausible starting point.27 I granted for the sake of argument that adding facts about Alpha Centauri to QR would result in a set that necessitates Obamas existence, but this was too concessive. Our previous argument that QR does not necessitate his existence made very few assumptions about R, so incrementally enlarging R does not address the underlying problem. So is there any set of qualitative facts that necessitates his existence? The most plausible suggestion is a complete qualitative specication of the entire cosmos, plus a totality fact to the effect that they are all the qualitative facts there are. Call this set QT . Does QT necessitate Obamas existence? I do not have a rm intuition either way so I am happy to concede that it does. Indeed since QT contains the totality fact our above argument cannot be used to show that it does not. Moreover, we are arguing that if qualitativism is true then the individualistic facts are plurally grounded in the qualitative, and it follows from qualitativism (and our assumption that the grounded is necessitated by its grounds) that the qualitative facts necessitate any given individualistic fact. So denying that QT necessitates Obamas existence is not dialectically available here. So QT appears to be the best candidate for a set of qualitative facts that necessitates Obamas existence. But of course QT contains facts about the qualitative goings-on in all corners of the entire cosmos, and surely most of those goings-on are entirely irrelevant to an explanation of why Obama exists. That is the basic idea: in attempting to nd a qualitative ground that plausibly necessitates Obamas existence we are forced to include facts that are irrelevant to the matter. Or put the other way: in zeroing in on the facts that are relevant to an explanation of his existence we nd that they no longer necessitate his existence. The above is an argument-scheme that can be lled in for different values of R: if you think (as I do) that the qualitative goings-on in Jupiter are explanatorily irrelevant to Obamas existence, you could take R to be a region that includes our planet and not much else and the argument would go through just the same. The argument has nothing to do with whether Obama is discernible from other things in the contemporary meanings of that term. For example, it has nothing to do with whether there is a (perhaps complex, highly relational) qualitative property that only he instantiates. For even if there is such a property the question remains whether its instantiation explains Obamas existence, and the argument is that there are constraints
be clear, one might ask two questions here. First, if qualitativism is true, are facts about Alpha Centauri relevant to explaining Obamas existence? And second, regardless of whether qualitativism is true, are facts about Alpha Centauri relevant? I think the answer to both questions is clearly no.
27 To

14

on explanation (necessitation and relevance) that suggest not.28 I said that these problems dissolve if we plurally ground individualistic facts in qualitative facts. How so? One simple proposal is to let IT be the set of all individualistic facts and let (as before) QT be the set of all qualitative facts, and say that the members of IT are (plurally) grounded in the members of QT even though no member of IT is grounded in any subset of QT . This a structuralist view of individuals, since it implies that an account of any one individual is inevitably an account of them all. But that is just one proposal and there are many details to argue about. Some might argue that the qualitative ground should include only certain kinds of qualitative facts such as existential generalizations or facts about how properties are bundled together. Others might insist that only a proper subset S of individualistic facts (e.g. facts about the fundamental particles) are plurally grounded in the qualitative and that other individualistic facts are grounded singularly in some members of S. Still others might want to say that structuralism is a necessary truth. But these are in-house arguments between theorists all of whom deserve to be called structuralists. Since the differences between their views will not matter here, I will focus on the simple proposal described above. A chief advantage of structuralism is that it avoids the difculties we faced when trying to ground Obamas existence on its own. For one thing, QT contains no irrelevancies when it comes to explaining the members of IT together. To be sure, QT does contain irrelevancies when explaining Obamas existence, such as qualitative facts about electrons in Alpha Centauri. But since IT contains individualistic facts about those very electrons the qualitative facts about them would appear to be perfectly relevant when explaining IT s members! And as we saw earlier (when discussing the idea that QT grounds Obamas existence) it is not implausible that QT necessitates all the individualistic facts IT . (Moreover, as we also saw, if one denies that QT necessitates IT then one denies qualitativisim, and here I am attempting to establish that if qualitativism is true then the individualistic facts are plurally grounded in the qualitative.) The problems we faced when trying to ground Obamas existence on its own therefore dissolve when we instead ground individualistic facts plurally. None of this implies that structuralism is the best form of qualitativism, since structuralism may suffer from problems of its own. Still, it is evidence in its favor. Of course if structuralism is true then there is a sense in which the
is an extensive recent literature on the question of whether various individuals are discernible from one another in the absolute or weak sense dened in footnote 16. For example see Ladyman and Ross [13] and references therein. These questions about discernibility are interesting but (as I said in section 2) they are not ours. I discuss the relation between qualitativism and these other various views about discernibility in my XXXX [reference omitted for blind review].
28 There

15

members of QT give rise to Obamas existence, even if they do not ground it. More precisely, let us say that some facts account for a fact Y iff there are some facts such that are (plurally) grounded in and Y is a logical consequence of . Then structuralism implies that QT accounts for Obamas existence. But accounts for is not a purely explanatory notionat least, not if (as I am assuming) relevance is required for explanationsince relevance is not preserved under logical consequence. If the question is what the members of QT explain, the structuralist says that they explain the members IT together but not individually. Distinguishing between ground and accounting is not splitting hairs: as I said in section 1, the requirement of relevance is one of the central features used to distinguish ground from related notions such as metaphysical necessitation and logical consequence. Indeed identifying the explanatorily relevant facts responsible for producing this or that aspect of the world is arguably the raison detre of the notion of ground. Can the virtues of structuralism mentioned above be replicated without taking ground to be plural? One might try using conjunctions in place of plurals. For the conjunction of all members of IT call this conjunction IT is (like any conjunction) grounded in its conjuncts. The structuralist then says that those conjuncts of IT are (plurally) grounded in the members of QT . It follows that IT is derivatively grounded in QT . One might then try replicating the virtues of structuralism without taking ground to be irreducibly plural by proposing that IT is grounded directly in QT without the detour through its conjuncts. This view shares the virtues mentioned above, but it is untenable. For even if IT is grounded directly in QT , it must also be grounded in its conjuncts (on pain of denying the evident truth that conjunctions are grounded in their conjuncts). And what then of those conjuncts? We cannot say of any conjunct that it is grounded qualitatively (on pain of falling foul of the arguments just given). And we cannot say that they are each grounded in the conjunction (on pain of moving in too tight a circle). So it looks like they must (on this view) be groundless. The result is that IT is radically overdetermined: it is grounded in its conjuncts, and it is also grounded in QT , even though the conjuncts are not grounded in the members of QT or vice versa. This is not an explanatory thesis that should be taken at all seriously. Alternatively, one might try replicating the virtues of structuralism, without taking ground to be plural, with a theory of facts. For example, suppose there were a fusion operation that takes a plurality of facts and delivers a single fact, their fusion. Then let I* be the fusion of the members of IT . (I* is not the conjunction IT , though they obtain under the same conditions.) Suppose one then claims that whole facts are prior to their parts, in the sense that the latter ontologically depend on the former. It follows that the members of IT ontologically depend on I*. And one might then say that I* (the single fact) is grounded in the members of QT , while 16

the members of IT have no ground. This would have the virtues of structuralism discussed abovethe members of QT arguably necessitate and are relevant to I*without taking ground to be plural.29 Should we accept this view? More needs to be said about it (about the fusion operation, and about the priority claim) before we can answer this. But I set it aside here because it is a solution to our problems that depends on a substantial theory of facts. As I said in section 1, my ofcial approach is to treat ground as a sentential operator, such that claims about what grounds what (typically) make no reference to facts. My reference to facts in the prose is just a convenient shorthand, eliminable in principle. So I cannot solve the problem at hand with a theory of facts: to do so would be to treat my reference to facts seriously and not as a mere convenience. I therefore leave this kind of solution as a possibility to be developed by someone who takes fact-talk more seriously than I do.

Cosmic Explanations

Above I appealed to the premise that qualitative goings-on outside our solar system are irrelevant to an explanation of Obamas existence. I said earlier why I nd the premise plausible, but it might be resisted. Indeed a qualitativist wedded to singularism about ground might take the moral of the argument to be that those qualitative facts about far ung areas of the cosmos are relevant to Obamas existence after all. So let me support the premise with some argument. Well, there is of course no knock-down argument to be had. We are engaged in an inference to the best explanation so the aim is to show that the structuralists explanation is better than the singularists. So let me point out various unattractive aspects of the singularists explanation (I will point out some more virtues of the structuralists explanation in section 9 when I discuss plural explanations in more detail). Start with the idea that Obamas existence is grounded in QT . This is (as I said earlier) perhaps the most plausible example of a set of facts that necessitates Obamas existence. But what about Romney? What grounds his existence? Suppose we say that it is also grounded in QT . Then we have the absurd conclusion that Obamas existence and Romneys existence have exactly the same ground! Here I do not assume that distinct facts always have a distinct ground: the facts PvQ and PvR might have a common ground, P. But it is unsurprising that the disjunctions have a common ground since they have a common constituent. My point is just that in the case of Obama and Romney, it is almost unbelievable that the explanation of why the one exists is exactly the same as the explanation of why the other exists. Surely (this is an unargued premise) if Obamas and
29 Many

thanks to an anonymous referee for bringing this view to my attention.

17

Romneys existence each have an explanation there must be some differentia: some facts that play a role in making it the case that Obama exists but no role in making it the case that Romney exists (and vice versa).30 Where might this differentia be found? There are two options: within R or outside of R. Suppose the latter. Then there are certain qualitative goings-on in far ung corners of the universe that are relevant to explaining Obamas existence but not Romneys. And this is absurd as an explanatory hypothesis (this is another unargued premise). It is one thing to bite the bullet and say that various qualitative goings-on in Alpha Centauri are relevant to explaining Obamas existence. I could perhaps be persuaded of that. But it is another thing to say that such goings-on play a role in making it the case that Obama exists but not that Romney exists, as if Obama has his very own plot of space-time in far ung corners of the universe that is partially responsible for his existence but not Romneys. As an explanatory hypothesis, the idea is (frankly) hard to take seriously. So the differentia must be found within R. This is perhaps the most plausible option available to the singularist, but it is nonetheless odd. For the intrinsic qualitative nature of R is not sufcient (on this proposal) to ground Obamas existence, facts outside R are needed too. So the view is that various qualitative goings-on outside R play a role in making it the case that Obama exists (and that Romney exists), but that various qualitative goings-on within R play no role whatsoever in making it the case that Obama exists. On the face of it this looks rather bizarreI nd it hard to see what kind of serious explanatory theory would grant an explanatory role to those far-ung goings-on but not to more nearby goings-on. Moreover it is far from clear that the proposed ground (on this proposal) would necessitate Obamas existence. For suppose that the qualitative goings-on within R that are said to be relevant to explaining Obamas existence are facts about the existence of something with just a few of Obamas qualitative properties, such as being born on a small island and being well educated. Suppose (that is) that it is just those facts within R plus the various qualitative goings-on outside R that are said to ground Obamas existence. Then the proposed ground would clearly fail to necessitate Obamas existence for the same reason that the proposals discussed in section 3 failed: it would be possible for something else within R to have those qualitative properties and yet for Obama not to exist. Indeed this worry about necessitation arises as soon as we retreat from QT . So the general problem might be put like this. We cannot say that Obamas existence is grounded in QT , else (by parity of reasoning) we would have to say that Romneys existence is grounded in QT too, in which
30 Of course the structuralist will admit that the same facts account for Obamas existence and Romneys existence, in the sense dened in the last section. But (as emphasized there) the notion of accounting for is not a purely explanatory notion. The point here is that Obamas existence must have a different explanation from Romneys.

18

case both have exactly the same ground, which is absurd. So we have to pare down QT to nd some core set of facts that is the ground of Obamas existence but not Romneys. But when we do so it is far from clear whether the proposed ground necessitates his existence (and in some cases the proposed explanation is hard to take seriously in the rst place). None of this is conclusive. I have just tried to indicate some difculties one encounters when searching for a qualitative ground of Obamas existence on its own. Insofar as structuralism avoids these difculties, that is a point in favor of structuralism. So I leave it as a challenge to the qualitativist who wishes to ground Obamas existence on its own to develop an account that avoids these difculties and is better than the structuralists. At this point one might reject my methodology. I appealed to premises about relevance and ground (e.g. my original premise that facts about Alpha Centauri are irrelevant to Obamas existence, and my premise in this section that Obamaa and Romneys existence have different grounds). But it might be objected that these premises cannot be used as evidence because I have given no theory of how justied belief or knowledge about relevance or ground is possible. It is true that I have offered no such theory, but to conclude that our beliefs about relevance and ground are of no evidential signicance is a gross over-reaction. If someone proposed that Europes being at war in 1940 was partly grounded in the number of electrons in Alpha Centuari one would reasonably reject the proposal since that obviously played no role in making it the case that Europe was at war. Somehoweven if we know not howour grasp of the fact of Europes being at war (perhaps along with rudimentary empirical knowledge) is enough to inform us that how its citizens were acting is relevant to explaining it and the number of electrons in Alpha Centauri is not. One can reasonably point this out without having a developed theory about how this is possible. I am just making similar points about Obamas existence. Moreover I do not claim that our beliefs about irrelevance and ground are indefeasible. It seems obvious that the number of electrons in Alpha Centauri is irrelevant to a causal explanation of the oil spill in the gulf, but there are empirical discoveries that could lead me to think that it is relevant after all (we might discover that someone formulated a plan to sabotage BPs machinery on the condition that Alpha Centauri contains more than n electrons, and then enrolled in astronomy class. . . ). It seems just as obvious that facts about Alpha Centauri are irrelevant to a metaphysical explanation of Obamas existence; but if I had good theoretical reasons to be a qualitativist and good reasons to think that the only way to then make sense of Obamas existence is to ground it in facts about the entire cosmos, I would consider accepting the surprising result that those facts about Alpha Centauri play a role in explaining Obamas existence after all. But this would be a radical revision of pretheoretic belief. And the point is that this radical revision is not required. The struc19

turalist has no need to revise her pretheoretic conviction that facts about Alpha Centauri are explanatorily irrelevant to Obamas existence, precisely because she denies that his existence (taken alone) has a qualitative ground in the rst place! Now one might say that this comes at the cost of rejecting singularism about ground, which was also a pretheoretic belief. But even if this was a pre-theoretic belief (which I doubt), this observation carries very little weight. For claims about the logical form of ground (like singularism) are highly abstract claims about the nature of explanation, and it is not at all clear why we should take our pretheoretic opinions about that sort of thing seriously. So structuralism saves the pretheoretic beliefs that matter.

The Inter-Dependence of All Things

It is worth comparing structuralism with other, related views. We already know that structuralism is not mere anti-haecceitism. For (as emphasized in section 2) the latter is just a modal claim, while the former is an explanatory claim. Structuralism is a version of qualitativism, since it says that the qualitative is sufcient to ground the individualistic (so long as we are careful to hear this plurally!). But it has an important point of agreement with individualism: namely, that a given individualistic fact like Obamas existence has (when considered on its own) no qualitative ground. Admittedly, it is tempting to infer from this point of agreement that individualism is true: we have (after all) an individualistic fact that cannot be qualitatively explained, which appears to be a counterexample to qualitativism! And indeed the inference would be valid if singularism about ground were true. Insofar as we have been in the grip of singularism, then, this might explain why individualism has traditionally been the more popular doctrine. But the inference is invalid, for even if a single individualistic fact has no qualitative ground, the individualistic facts together may (plurally) have a qualitative ground, just as the structuralist thinks. One might object that if structuralism says that there are ungrounded facts about individuals then it is not a version qualitativism after all. But this is mistaken. For the idea behind qualitativism is that everything arises out of purely qualitative facts, that (to use the popular metaphor) all God had to do when making the world was x the qualitative facts. And this is indeed the case according to the structuralist. It is just that those qualitative facts explain the individualistic facts all at once, not one by one. Perhaps the most familiar version of qualitativism is the famous bundle theory, on which each individual is identied with a set of compresent qualitative properties. This is rather different from structuralism. For while the bundle theorist sees a certain set of compresent properties and 20

says Here is Obama!, the structuralist sees no such thing. For the structuralist, no part of the qualitative nature of the world can be said to be responsible for Obamas existence. Is structuralism (as dened here) what an ontic structural realist like Ladyman has in mind when he talks of individuals whose identity and individuality are secondary to the relational structure in which they are embedded?31 It is hard to say, in part because this talk of identity and individuality are obscure in the extreme. Indeed a (personal) motivation for developing the idea of plural grounding is to make sense of the kind of metaphysical thesis that Ladyman might have in mind in clearer terms. But suppose that what Ladyman had in mind is that individualistic facts are grounded in facts about relational, qualitative structures. Still, it remains open that each individualistic fact on its own is grounded in that qualitative structure. In that case his view resembles the bundletheorists, insofar as they both agree that there is some qualitative body of fact responsible for each individualistic fact. This is precisely what the structuralist (in my sense of the term) denies. What I call structuralism perhaps resembles a view of Spinozas in Part I of The Ethics, at least on Garretts reading.32 Spinoza famously claims that the nite modesrocks, chairs, tablesfollow from the essence of God. But according to Garrett the correct reading is that they only follow from Gods essence when taken together: it is false of any single nite mode that it follows from Gods essence, but it is true of them all together that they follow from Gods essence. Substitute the qualitative nature of the world for Gods essence and understand the notion of following in terms of ground, and you have the structuralist view described above.33 It is worth emphasizing that structuralism is a claim of grounds, not of semantics. So it is consistent with a compositional semantic theory that assigns a truth-condition to each individualistic sentence on its own. For example structuralism is consistent with a compositional semantic theory on which Obama is sitting is true in English iff Obama is sitting, or iff
[13], p. 144. Garrett [7]. 33 There is more in common between my view and Spinozas. For if Garretts reading is correct then Spinoza endorses a many-from-one view: a view on which many things (the nite modes) follow from one (God). Now when I motivated qualitativism in XXXX [reference deleted for blind review] I argued that at the most fundamental level there is only one qualitative fact, the World Fact, that characterizes the qualitative nature of the whole cosmos all at once. On this view it is a mistake to see the World Fact as the logical sum of smaller qualitative matters of fact each of which characterizes just part of the worlds qualitative nature. When that view is combined with the structuralist view defended here the result is a many-from-one view on which the many individualistic facts are grounded together (but not one by one) in a single underlying source. Many-from-one views have not been popular in contemporary metaphysics, to put it mildly. But if I am right they can be motivated by an argument for qualitativism of the sort found in XXXX [reference deleted for blind review], together with the argument in this paper.
32 See 31 Ladyman

21

Obama instantiates the property referred to by sitting, or what have you. So the mere fact (if it is one) that there are correct semantic theories of this type is no threat to structuralism. What structuralism may imply is that there are no truth-conditions for a single individualistic sentence in fundamental (i.e. qualitative) terms. Whether structuralism implies this depends on what is meant by a truthcondition (equivalently: what is meant by the connective iff in a statement of truth-conditions). But even if it implies this there is no conict with the project of semantics, for it is no part of that project to state truthconditions in fundamental terms (if you doubt this, go and count how many semantic theories are stated in the language of quantum mechanics).34 Moreover it may nonetheless be possible to take a set of individualistic sentences together and state the truth-conditions for them in fundamental (i.e. qualitative) terms (again, I hedge because this all depends on what is meant by a truth-condition). The resulting semantics would be holistic, delivering a truth-condition for them without delivering one for any member of the set taken alone. Which is precisely the kind of semantics in fundamental terms that a structuralist would expect.

Absolutism and Comparativism

So much for individuals. Perhaps surprisingly, an analogous structuralist view can be motivated with similar arguments about what is on the surface a very different case, namely that of quantities like mass, charge, energy, temperature, length, and so on. I will focus on the case of mass, but the discussion generalizes to other quantities. Let us start by distinguishing two views about mass. The property of having mass is a determinable that appears to have two kinds of determinates. It is natural to think that something with mass has a determinate intrinsic property, a property it has independently of its relations to other material bodies. But it is also natural to think that things with mass stand in various determinate mass relationships with one another, such as x being more massive than y or x being twice as massive as y. Now of the intrinsic masses and the mass relationships, which are fundamental? According to a view I will call absolutism, the intrinsic masses are prior to the mass relationships. The absolutist does not deny that things stand in determinate mass relationships, she just says that those
point here amounts to Siders distinction between a linguistic semantics and a metaphysical semantics (see his [24]). The former is what gets done by linguists and contemporary philosophers of language, in which a semantics for (say) the term football match would not be expected to be given in terms of the underlying quantum mechanical states that make it up such matches. In contrast, a compositional semantics stated in those fundamental terms is what Sider calls a metaphysical semantics. In the text I am using semantics to refer to what Sider calls linguistic semantics.
34 The

22

relationships, and indeed all facts about the masses of material bodies, are derivatively grounded in facts about the particular intrinsic mass had by each body. If my laptop is more massive than my cup, the absolutist will say that this is because of the intrinsic mass that they each possess.35 In contrast, comparativism is the view that all facts about the masses of material bodies are derivatively grounded in facts about how they are related in mass to one another. Some comparativists will say that the most fundamental mass relations are ratio relations while others will insist that they are merely ordinal, but this in-house dispute will not concern us here.36 I favor comparativism. My reason is analogous to my reason for favoring qualitativism. The rough idea is that all we can ever observe are the mass relationships between things, for example that one body is more massive than another. If as the absolutist claims there are further facts of the matter concerning which particular intrinsic mass each body has facts that are not grounded in those mass relationshipsthen those facts lie beyond our epistemic ken. A reasonable Occamist principle then recommends that we dispense with such epistemically inaccessible facts.37 However my aim here is not to argue for comparativism but to argue that if comparativism is true, then certain facts about mass must be grounded plurally in mass relationships rather than one by one. I have in mind facts about mass in a given scale, such as that my laptop is 2 kilograms, that Beckham is 75 kgs, and so on. The comparativist faces the challenge of showing that mass relationships really are sufcient to explain these kilogram facts. If she cannot meet this challenge then she would have to be an eliminativist about kilogram facts and claim that there are no such facts. Insofar as this is intolerable, meeting the challenge is crucial to the success of comparativism. I will argue that the comparativist faces signicant difculties if she attempts to ground each kilogram fact in turn, but that these difculties dissolve if she grounds them plurally. Before we start, note that the absolutist can very easily explain each kilogram fact on its own. For if material bodies have the intrinsic masses posited by the absolutist it is plausible that terms of the form r kilograms would refer to those properties. If so, then it is almost irresistible to say (for example) that my laptops being 2 kgs is grounded in (or perhaps even identical to) its having a certain intrinsic mass; namely, that intrinsic mass that is the referent of 2 kilograms.
perhaps, facts about how the intrinsic masses themselves are related to one another. The details of the view can be cashed out in many different ways, but these differences will not matter in what follows. Absolutists include Armstrong [2], Eddon [4], and Mundy [17]. 36 For a more precise account of the distinction between absolutism and comparativism, see XXXX [reference omitted for blind review]. 37 Like the Occamist argument against individualism there is much more to say here. Some of it is said in XXXX [reference omitted for blind review].
35 And,

23

Finding Kilograms in a Comparative World

So the absolutist has what appears to be an attractive explanation of each kilogram fact on its own. Not the comparativist, though. To see this, consider the fact that my laptop is 2 kgs. If the comparativist tries to ground this fact in mass relationships, she must nd some set R of facts about mass relationships that explains its being 2 kgs. But what could R be? I will argue (as before) that any candidate set R that might necessitate my laptops being 2 kgs contains irrelevant information. Once again we have a case in which my two assumptionsthat a ground must both necessitate and be relevant to what it groundscannot be jointly satised. Let us start by constructing a candidate set R. The most obvious suggestion is to let R be the single fact that my laptop is twice as massive as the standard kilogram in Paris, often known as the International Prototype Kilogram (IPK). But the trouble is that this does not necessitate the fact that my laptop is 2 kgs. For it is possible for my laptop and IPK to both be twice as massive as they actually are, in which case my laptop would still be twice as massive as IPK and yet would be 4 kgs, not 2 kgs. What other mass relationships might necessitate my laptops being 2 kgs? It would not help to add to R facts about the mass ratio between my laptop and (say) 20 other benchmark items, since the same kind of argument shows that those relationships do not necessitate my laptops being 2 kgs either. But what if we let R contain facts about how my laptop is related in mass to all other bodies in the entire cosmos: would R then necessitate my laptops being 2 kgs? It is not clear. R would x the mass relationship between any two bodies, so the question is whether the entire cosmos could be exactly as it is in all mass-relational respects and yet differ in the mass of my laptop, and I do not have a clear intuition either way. However, we are arguing that if comparativism is true then the kilogram facts are plurally grounded, and it follows from comparativism (and our assumption that grounded is necessitated by its grounds) that the mass relationships necessitate any fact about mass. So denying that R necessitates my laptops being 2 kgs is not dialectically available. So assume that R would necessitate my laptops being 2 kgs. The trouble is that R would contain explanatorily irrelevant information. It would contain facts about the mass relationship between my laptop and electrons in Alpha Centauri, and (premise) these are irrelevant to explaining my laptops mass in kilograms. Surely its mass relationships to electrons in Alpha Centauri play no role whatsoever in making it the case that it is 2 kgs. This premise might be resisted, but it is very plausible. Moreover I do not claim that the premise is sacrosanct. If I had good theoretical reasons to be a comparativist and if the only way to then account for my laptops being 2 kgs were to ground it in its relations to electrons in Alpha Centauri, I might consider revising my beliefs about what is relevant. The point is 24

just that this would be a signicant revision of pre-theoretic belief. The premise can also be supported with argument. For recall how natural it was for the absolutist to ground my laptops being 2 kgs in terms of its intrinsic mass: that intrinsic property (if it had such a thing) would very nicely explain its being 2 kgs. This explanation is not available to the comparativist, but the fact that it is so natural suggests that we (pretheoretically) take facts about electrons in Alpha Centauri to be irrelevant to the matter. Indeed this last point might be turned into an objection to any appeal to mass relationships, even mass relationships to IPK. The argument would start with the observation that we nd the absolutists intrinsic explanation so satisfying. And it would argue that this is evidence that we (pretheoretically) take my laptops mass relation to any other body (including IPK) to be explanatorily irrelevant to why it is 2 kgs. Perhaps its relation to IPK is relevant to explaining the semantic fact that 2 kgs picks out the intrinsic mass it does. But not, according to this argument, the non-semantic fact of my laptops being 2 kgs. The argument has some appeal. After all, the absolutist could in principle appeal to my laptops relation to IPK when explaining why it is 2 kgs, but if she did then her resulting explanation would look decidedly odd. One would ask why she brought IPK into the picture when all that appears relevant is its intrinsic mass. If we accept this argument, then any comparativist explanation of its being 2 kgseven the initial suggestion in terms of its being twice as massive as IPKis objectionable on the basis that it appeals to what we pre-theoretically take to be irrelevant information.38 This is important. For one might have tried to rene that initial suggestion in light of the modal objection to it discussed earlier. One might have said that my laptops being 2 kgs is grounded in its being twice as massive as IPK actually is. Or one might have developed a view according to which it is impossible for IPK to have differed in mass at all: while the lump of metal in Paris could have been more massive (the idea would be) IPK should not be identied with that lump and is instead a co-located yet distinct object that has its mass essentially.39 The modal objection would have no force against either of these views, but according to the above argument both views are objectionable since they appeal to what we take to be irrelevant information, namely my laptops mass relationship to IPK.40
38 One respond to this argument by saying that if comparativism is true then the mass relations must be relevant, since they are all the comparativist has to work with. But this ignores the possibility of error theory. For the comparativist might concede the argument in this paragraph and conclude that since there is no grounding my laptops being 2 kgs in terms that she recognizes, there is no such fact. To say that the comparativist must ground its being 2 kgs in mass relationships is akin to claiming that being a witch must be explicable in natural terms, since natural facts are all we have to work with. 39 Thanks to Jack Spencer for bringing this view to my attention. 40 To be clear, both the rened views in the last paragraph are vulnerable to another (per-

25

As before, our belief that relations to IPK are irrelevant is not sacrosanct. If I had good theoretical reasons to be a comparativist and if the only way to then make sense of my laptops being 2 kgs were to ground it in its relationships to IPK, for example, I would consider revising that belief. But the virtue of plural grounding is that no revision is required.

Structuralism Redux

How so? One simple proposal is to let K be the set of all kilogram facts and let R be the set of all fundamental facts about mass relations, and then say that the members of K are plurally grounded in the members of R even though no member of K is grounded in any subset of R. Call this a structuralist view of kilograms, since an explanation of any kilogram fact is (on this view) inevitably an explanation of them all. As in the case of individuals there are many details to argue about: whether R should include only certain kinds of mass relations such as ratio relations, whether K should include only those facts concerning the mass in kilograms of the fundamental particles, and so on. But these are all in-house arguments between theorists all of whom deserve to be called structuralists. Since their differences will not matter here I will focus on the simple proposal described above. It is easy to see how structuralism generalizes to other scales and quantities such as pounds, miles, seconds and so on. An advantage of structuralism (as before) is that it dissolves the problems we faced when trying to ground my laptops being 2 kgs on its own. For one thing, R contains no irrelevancies when it comes to explaining the members of K. Sure, R contains irrelevancies when explaining my laptops being 2 kgs such as mass relationships between electrons in Alpha Centauri. But since K contains kilogram facts about those very electrons the relationships between them are entirely relevant when explaining Ks members! And as we just saw (when discussing the idea that R grounds my laptops mass in kgs) it is not implausible that R necessitates K. (Moreover, if one denies that R necessitates K then one denies comparativism, and here I am trying to establish that if comparativism is true then the kilogram facts are grounded plurally in the mass relations.) So our problems dissolve when we instead ground kilogram facts plurally. This does not imply that structuralism is the best form of comparativism, but it is evidence in its favor. Though structuralism is a version of comparativism, it has an imporhaps more decisive) objection. The objection is that they both appeal to the intelligibility of mass comparisons across different possible scenarios, and yet it is doubtful that this is intelligible to the comparativist. But the issue of cross-world mass comparisons is delicate and it would be distracting to discuss it here (I discuss it at some length in section 10). So for now I rest my objection to these views on the charge of irrelevance.

26

tant point of agreement with absolutism: namely, that a given kilogram fact has (when considered on its own) no relational ground. It may be tempting to infer from this point of agreement that absolutism is true, and indeed the inference would be valid if singularism about ground were true. Insofar as we have been in the grip of singularism, then, this might explain why absolutism has traditionally been the more popular doctrine. But the inference is invalid: even if the kilogram fact has no relational ground on its own, it may be that the kilogram facts together have a relational ground (just as the structuralist says). I have focused on the case of mass but I expect that the lessons here generalize to other cases in which we have a mathematical representation of worldly phenomena, for example a representation distance in meters, time in seconds, acceleration in meters-per-seconds-squared, rational preferences in utils, and so on. In all these cases I believe that the facts about the mathematical values in a given scale will be plurally grounded in the underlying facts that give rise to the numerical representations. If that is right then we have here a general method of approaching the metaphysics of numerical representation, not just mass. But I leave the generalization to other cases for another time.

Structural Explanations

So far I have argued that each structuralist view dissolves problems that we faced when attempting to ground each individualistic fact or each kilogram fact alone. But it is one thing to say that the views dissolve certain problems, it is another thing to show that the structuralists proposed explanation in each case really is explanatory, i.e. that the underlying mass relations in R really are sufcient to explain the plurality of kilogram facts in K, and that the underlying qualitative facts in QT are really sufcient to explain the individualistic facts in IT . Focus on the case of kilograms, where I think the worry is most pressing. The structuralist says that the total body of mass relations explains why my laptop is 2 kgs, my table is 10 kgs, my bed is 100 kgs, etc. But it is (logically) consistent with those relations that my laptop is 4 kgs, my table 20 kgs, my bed 200 kgs, etc. The worry is then that a mere description of the mass relations has missed something out and has not explained why the kilogram facts are as they are rather than (say) double what they are.41
is needed in formulating the worry. One way to put it is that R does not single out a particular material body as privileged and so does not x a unit kilogram. But this way of putting the worry is confused. For consider the standard absolutist view according to which a given material bodys being r kgs is grounded in its having a certain intrinsic mass. On this view the proposed groundsi.e. facts about which intrinsic mass each material body hasdo not single out any particular material body as privileged and so in that sense do not x a unit kilogram either. What then does it mean to x a unit?
41 Care

27

One might respond with abstract argument. For example, one might argue for the general principle that if some Xs are relevant to some Ys and necessitate those Ys then the Xs ground those Ys. For it is almost undeniable that the relations in R are relevant to the kilogram facts in K. And (as I said in section 8) we are assuming that R necessitates K. But even if this abstract argument has some merits, something more illuminating can be said to make the structuralists explanation compelling. The key is to recognize that the basic role of kilogram predicates in our language is to conveniently store information about mass ratios. Once that role is clearly in view, the structuralists explanation becomes compelling and the idea that something has been missed out evaporates. To show this, let me rst describe a ctional community in which their predicates are stipulated to play exactly that role and then show that the structuralists proposed explanation is compelling when it comes to explaining the facts that they express with their predicates. Then I will argue that our own community just like theirs in all relevant respects. Consider then a community that initially lacks kilogram predicates. Let us imagine that the only expressions they have with which to talk about mass are predicates of the form x is r times more massive than y, one for each positive real r. If comparativism is true then their language is complete in the sense that they can state all the fundamental facts about mass. Nonetheless, their language is somewhat impractical: if one of their citizens Rahul is hosting a pot-luck dinner and wants everyone to contribute half the amount of rice in his cupboard, the only way he can issue the request is to say something like Please could everyone bring half the amount of rice in my cupboard. Each guest would then need to visit Rahuls house before the party to measure out the right amount. It therefore behooves them to nd some way of attributing mass to things one by one, as it were, so that they could all have determined the right quantity of rice at home. The important thing is that these attributions should be coordinated so that they imply the mass-ratios that they are interested in. To this end Rahul introduces a slew of 1-place predicates of the form x is r Dravids into the language, one for each positive real r. His idea is that the ratios between real numbers can then be used to represent the mass-ratios between the material bodies. Since this is the primary function of these predicates the only thing that Rahul says when introducing them is that they are governed by the following inference rule: x is r Dravids
Presumably the idea is that the expression 1 kg is stipulated to refer to that intrinsic mass had by the IPK, and so the IPK is then said to be of unit mass on the kilogram scale. But if that is the question of how a unit is xed then it is a meta-semantical question about what determines the meanings of our words, not a question about what grounds the kilogram facts. So neither absolutism nor structuralism should not be expected to answer it.

28

y is s Dravids Therefore, x is r/s times as massive as y His idea is the members of his community should go forth and apply these predicates to material bodies in such a way that each application coheres with other applications made in their communitycoheres, in the sense that inferring by the above rule yields truths about mass-ratio. So a speakers primary aim in applying one of the predicates is just that her application coheres with a (perhaps weighted) majority of the other applications in her community. It does not matter how the practice gets going: the very rst speaker has free reign to apply any of the predicates to any object! But once this rst application is made the above inference constrains subsequent applications by other speakers. So the community is now faced with a coordination problem. But this is easily solved by distributing Dravid-measuring instruments to the population that are all calibrated with one another calibrated, in the sense that they are all designed to assign numbers to things in such a way as to cohere in the above sense.42 Having introduced these predicates Rahuls life is much easier. If his own Dravid-measuring instrument says that he has 2 Dravids of rice in his cupboard, he can simply ask each of his guests to bring 1 Dravid of rice and it will then follow (so long as the instruments are calibrated) that each guest will bring half the amount of rice in his cupboard, as desired. Importantly, note that in introducing his predicates Rahul said nothing about a standard object in terms of which the term 1 Dravid is dened or has its reference xed. All that matters (given what Rahul said) is that
practice will only work, note, if it is possible to assign numbers to material things in a coordinated manner. This is conrmed by a so-called representation theorem of measurement theory. Say that a function f from material things to real numbers represents mass-ratio iff the following is true: x is r times more massive than y iff f (x) = r. f (y). Then a representation theorem states that if the mass-ratios between things obey various constraints then there is at least one function that represents mass ratio. It follows that there exist applications of Dravid predicates to things that cohere in the sense mentioned in the text. The so-called uniqueness theorem then states that given any function f that represents mass ratio, (i) r. f also represents mass ratio for any positive real r, and (ii) every function that represents mass ratio can be written as r. f , for some positive real r. The representation and uniqueness theorems together imply that given any material body and any real number, there is a unique function that maps that body to that number and that represents mass-ratio. Which means that the rst speaker does indeed have free reign to apply any of the predicates to any object, sure in the knowledge that it is possible for the communitys other applications of Dravid predicates to cohere with that rst application. Now I just slurred over many details of the representation and uniqueness theorems. For one thing, these theorems are usually stated with respect to an underlying relational language that contain just two predicates: x is greater or equal in mass than y and a predicate for material composition. But this simplication is harmless for current purposes. For more realistic theorems and proofs see Krantz et al [11].
42 Their

29

their Dravid-measuring instruments are calibrated in the above senseit does not matter whether they are all calibrated with a special standard object. Of course if the community all agree that a particular bag of rice is 1 Dravid then they might put that bag in a (protected) public space and use it as a practical aid in calibrating their Dravid-measuring instruments. But there is no requirement that they dene or x the reference of 1 Dravid in terms of that bag. So, if they discovered that the bag is actually half as massive as they thought it was, they would not be required by the semantics of Dravid to continue to think that it is 1 Dravid come what may. Ratheras long as they were still condent that their Dravid-measuring instruments were calibratedthey would say that the bag is actually 0.5 Dravids (and they might then use some other object to help calibrate their devices instead). The point is that the bag would just be a (dispensable) practical aid used to further the primary goal of coordination. Now suppose that the Dravid predicates become deeply entrenched in Rahuls community, in the sense that the community has applied the predicates widely and there is a (perhaps weighted) core majority of those applications that cohere with one another. Then I claim that a structuralist explanation of Dravid factsi.e. of what is expressed by applications of Dravid predicatesis almost irresistible. To see this, suppose Rahul asserts one of those applications in the coherent core, say This brick is 2 Dravids. And suppose as comparativists we then ask Rahul what it is about the underlying mass-ratios that make the brick 2 Dravids. Then there would appear to be no answer. Remember, there is no privileged standard object, such that something can be said to be 2 Dravids in virtue of being twice as massive as it. Given the role of Dravid predicates in their language, it seems that if there is an explanation of why the brick is 2 Dravids it is that the bricks being 2 Dravids coheres with the mass-in-Dravids of other thingsbut since this answer appeals to the mass-in-Dravids of other things it is not an answer that is acceptable to the comparativist. So the bricks being 2 Dravids appears to have no purely mass-relational ground on its own. But now take the coherent core set of applications and add the as-yet unaccepted sentences of the form x is r Dravids that cohere with that core, one sentence for each material body x. The resulting set of sentences D is a complete representation of mass, in the sense that one could recover the entire mass relational nature of the world from its members by way of the above inference scheme. And now suppose Rahul asserts each member of D in turnThis brick is 2 Dravids, my table is 10 Dravids, David Beckham is 75 Dravids. . . and suppose that when he is done (!) we ask him to explain what makes all that the case. Well, since his primary aim in applying the predicates is just that his applications cohere in such a way as to represent the mass ratios between things, it is almost irresistible to say that what he said (when he asserted the members of D) is the case 30

because of the underlying mass relationships between the material bodies. Indeed, if the role of Dravid predicates is just to store and communicate information about mass-ratio, it is hard to see what else could possibly be needed to explain the mass-in-Dravids of things! This is structuralism through and through: the Dravid facts taken together are explained in terms of the underlying mass relationships, but no Dravid fact on its own has a mass relational ground. To be clear, this structuralist view of Dravids is not logically implied by the way Rahul uses his predicates, but it does strike me as almost irresistibleor at any rate the best explanation going. I believe that the community I just described is in all important respects ours: we use kilogram predicates just as Rahul uses Dravid predicates. The primary role of our kilogram predicates is just to conveniently store information about mass-ratio, so our primary aim in applying them is that our applications cohere with a (perhaps weighted) majority of the other applications in our linguistic community. Standard objects like the IPK in Paris are nothing other than practical aids at achieving global coordination. Once we see this, the structuralist explanation of what we express with kilogram predicatesi.e. the kilogram factsis just as compelling and irresistible as the structuralist explanation of the Dravid facts. What can prevent us from seeing this, though, are misguided theories about the role of standard objects like IPK, which invite us to think that each kilogram fact ought to have a ground on its own. For example, consider the view that x is r kilogram is dened to be true of an object x just in case x is r times more massive than IPK. This view encourages the idea that a given objects being r kgs has a ground on its own, namely in terms of its being r times more massive than IPK. But as Kripke famously argued, this view about kilogram predicates is false: it has the incorrect consequence that IPK is necessarily 1 kilogram.43 Or consider the Kripkean view that we use each term of the form r kilograms with the reference-xing stipulation that if it is to refer to anything, it is to refer to the mass that is r times that mass had by IPK.44 On this view IPK is not part of the semantics of kilograms but it is part of the meta-semantic theory about how the referent of r kilograms is xed. Still, this view also encourages the idea that each kilogram fact has its own ground, this time in terms of the intrinsic masses referred to by terms of the form r kilograms. But this Kripkean view is false. To see this, imagine reading in the Times that there is in fact no special lump of metal in Paris known as IPK and that the French created the illusion of such a lump with an elaborate system of lights and holograms. The article explains that the illusion was
43 See 44 Again,

Kripke [12]. see Kripke [12].

31

systematic, so that whenever we thought we were using IPK to calibrate various measuring instruments with one another the calibration succeeded even though we were misled about the existence of the lump. How would we report the discovery? Presumably just by saying that IPK (surprisingly) does not exist, and not much more. Importantly, if I had previously believed that my laptop is 2 kgs then I would not revise that belief in light of the discovery: I would continue to believe and assert that my laptop is 2 kgs even though there is no special lump in Paris. But the Kripkean view has difculty explaining this datum. For the view implies that if it turns out that there is no such thing as IPK then terms of the form r kilograms fail to refer, and it is then difcult to see why I would be inclined to continue to say that my laptop is 2 kilograms. A second case is perhaps more telling. This time, imagine reading in the Times that while there is such a thing as IPK, it turns out that the French have been creating an elaborate illusion designed to make us think that it is twice as massive as it actually is. Again, the article tells us that the illusion was systematic, so that the measuring instruments around the world that were calibrated with the help of IPK are indeed all calibrated with one other. The only surprise is that if we were to put IPK on any one of them we would get a reading of 500 grams, not 1 kg as expected. How would we report this discovery? Presumably by saying something like Wow it turns out that the standard kilogram in Paris is actually 500 grams! In particular, if asked how massive my laptop was I would be inclined to say It is 2 kgs, this article has no bearing on that question. But the Kripkean theory predicts otherwise. For that theory says that 1 kilogram is stipulated to refer to the mass of IPK whatever that mass is, so it implies that the article should instead be reported as telling us that while the standard object is (of course) still 1 kg, it turns out that my laptop is (surprisingly!) 4 kgs after all. And this is not how we would report it. I just discussed two views that give standard objects like IPK a central role in the semantics or meta-semantics of kilograms. Each view encourages the impression that each kilogram fact has a ground of its own. But each view is false. Seeing that they are false therefore removes obstacles to seeing the virtue of structuralist explanations. The correct view, I said, is that we use of kilogram predicates just like Rahul uses his Dravid predicates. The primary role of our kilogram predicates is just to conveniently store information about mass-ratio, so our primary aim in applying them is that our applications cohere with a (perhaps weighted) majority of the other applications in our linguistic community. Standard objects like the IPK in Paris play no semantic or meta-semantic role, they are just a practical aid in achieving global coordination. This view predicts our reactions to the two stories in the Times just discussed, which is evidence that it is true. And once we see that it is true, the structuralist explanation of kilogram factsi.e. the facts 32

we express with kilogram predicatesis just as compelling and irresistible as the structuralist explanation of the Dravid facts. You do not need to endorse this view of kilogram predicates in order to endorse structuralism. I have tried to motivate it just because I have found that it removes obstacles from appreciating the structuralists explanation (i.e. by minimizing the role of IPK in kilogram talk) and so makes it plausible that the underlying mass relations really are sufcient to explain the kilogram facts. But the truth of structuralism does not depend on it. I should emphasize that this view about how we use kilogram predicates is consistent with compositional semantic theories that assign truthconditions to each kilogram sentence on its own. With regards to Rahuls language, we might say that x is 2 Dravids is true in the language of Rahuls community iff x is 1 kilogram, or iff x is 2.2 pounds, or what have you. Indeed once their use of Dravid predicates became deeply enough entrenched, truth-conditions of this kind would appear to be highly plausible. And the existence of truth-conditions like this is consistent with the structuralists claim that no kilogram fact has a ground in purely massrelational terms (this is the analogue of the point I made in section 5 regarding structuralism about individuals). I have discussed the case of kilograms but I believe that roughly the same goes for individuals. Just as kilogram predicates are devices of measurement whose primary role is to conveniently store information about underlying mass-ratios, so too are singular terms devices of measurement whose primary role is to conveniently store the underlying qualitative world. And just as kilogram predicates fulll their function by being governed by the canonical inference described above, so too our singular terms are governed by the introduction and elimination rules for the existential quantier, rules which allow them to fulll their role of allowing us to conveniently reason about what is ultimately a purely qualitative world. I develop this view of singular terms in XXXX [reference omitted for blind review], but there is no space to discuss it here. Still, if it is right, then it helps us see that the structuralists explanation of individualistic facts in terms of qualitative facts is extremely plausible.

10

Modal Problems?

The last section argued that the structuralists proposed explanation of kilogram facts (and individualistic facts) is compelling. Still, one might think that it cannot be correct since it is subject to devastating problems. There is of course no space to consider every potential problem but let me discuss two that concern structuralisms modal implications.45
45 I focus on these just because they are the objections I have most often encountered when talking about structuralism.

33

The rst objection notices that if my laptop is in fact 2 kgs, it is nonetheless possible for it to have been 4 kgs and yet for everything elses mass in kilograms to have remained the same. The possibility of this independent variation is evidence (the objection goes) that my laptops mass in kilograms has a ground all on its own which can vary independently of the grounds of the mass in kilograms of other obejcts, contra structuralism.46 The second objection accuses the structuralist of not being able to make sense of any possibilities concerning mass in kilograms in the rst place. To see this, consider the possibility just mentioned of my laptop being 4 kgs instead of 2. Why think that the structuralist can make no sense of this? She can make sense of a world W that is just like ours with the one exception that the mass-ratio between my laptop and all other things is double what it actually is. But the worry is that on the structuralists own lights there is no fact of the matter as to whether W is a world in which my laptop is 4 kgs, or one in which it is 2 kgs and everything else is half the mass in kgs that they actually are. For the structuralist is a comparativist who thinks that all facts about mass are grounded in mass relationships, and the problem is that those mass relationships do not x how the bodies in the actual world are related in mass to those in W. And if there is no fact of the matter as to whether my laptop in W is more massive than my laptop actually is, the worry is, there can be no fact of the matter as to what its mass in kilograms is in W. I believe that both objections can be answered: we can make sense of possibilities concerning mass in kilograms (answering the second) and the way we do this will imply that my laptop could have been 4 kgs even if everything elses mass remained the same (answering the rst). So let us start with the second objection. One response is to endorse modal realism and say that the fundamental facts about the world are really facts concerning a plurality of worlds. The comparativist may then think that the fundamental facts concerning mass relationships include how objects in different worlds relate to one another in mass. But one might nd the idea of inter-world mass relations repugnant so let me outline another response that does not appeal to them.47 This response accuses the argument of using an incorrect model of how a possible world represents my laptops mass and introduces a better model that allows her to make sense of the possibility in question. First, how does a possible world represent something de re of my laptop in the rst place? Lewis famously said that it does so not by containing my laptop itself but instead by containing one of its counterparts. It does not matter for our purposes whether he was right about this, but let us assume that
to Richard Chappell and Brad Weslake for helping me appreciate the force of this objection. 47 The presentation here overlaps with my presentation in XXXX [reference omitted for blind review].
46 Thanks

34

he was so that we have a working model of de re representation in play. Given this assumption, the world W introduced above can be re-described as a world containing counterparts of my laptop and every other material body x such that if my laptop is r times as massive as x, my laptops counterpart in W is 2r times as massive as xs counterpart in W. Now, is there anything about W in virtue of which it can be said to represent my laptops being 4 kgs? Well, notice that the mass ratios that my laptop enters into differ systematically from those that its counterpart in W enters into, by a factor of 2. That is, if my laptop is r times more massive than another object x, then my laptops counterpart in W is 2r times more massive than xs counterpart in W. In contrast, consider any object other than my laptop, like my printer. The mass ratios it enters into are almost exactly the same as its counterpart in W. The only difference concerns its relation to my laptop: while my printer is (say) twice as massive as my laptop, my printers counterpart in W is the same mass as my laptops counterpart in W. So my laptop and my printer differ in this respect: my printers mass role is very similar to the mass role of its counterpart in W, but my laptops mass role is systematically different from that of its counterpart in W, by a factor of 2. So we might say that it is in virtue of this difference that W represents my laptop as being twice as massive as it actually is and everything else as having the same mass that they actually have. The structuralist can then piggy-back on this, for if W represents my laptop as being twice as massive as it actually is and if my laptop is actually 2 kgs, then we can take W to represent my laptop to be 4 kgs. In effect, we just introduced a mass-counterpart relation in addition to the ordinary, Lewisian counterpart relation. Since my printer and its counterpart in W resemble one another with respect to their mass role, let us call them mass-counterparts. And (the idea is) because my printers counterpart in W is also its own mass-counterpart, W represents my printer as having the same mass as it actually is. Here the masscounterpart relation is doing analogous work to Lewis counterpart relation: just as the latter is not identity but instead stands in for it when determining what a world represents de re, the mass-counterpart relation is not the same-mass-as relation but instead stands in for it when determining what a world represents about mass. And like Lewis counterpart relation, those aspects of an items mass-relational prole important to determining its mass-counterparts will presumably depend on the conversational context. With a bit of conversational coaxing we might engineer a lax enough context in which my laptops counterpart in W is also its own mass-counterpart; and relative to this mass-counterpart relation W will represent my laptop as being 2 kgs and everything else as having half the mass in kilograms that they actually have!48
48 I

develop this mass-counterpart theory in more detail in XXXX [reference omitted for

35

Lewis counterpart theory is often seen as a reduction of de re modality in terms of de dicto modality. We can similarly see the mass-counterpart theory just described as a reduction of modality concerning mass in kilograms in terms of modality concerning mass ratios. We can therefore distinguish between two senses of possibility: a fundamental sense that just concerns variations in mass ratios, and a looser sense that (also) concerns variations in mass in kilograms. So the structuralist should concede that in the fundamental sense of possibility the objections under discussion are well taken. But she can say that there is a looser sense of possibility whereby possibilities concerning mass in kilograms are explained in terms of possible worlds concerning mass ratios in the above way. And so she can agree that (in many contexts) it is possible in this loose sense for my laptop to have been 4 kgs even while all other things have the same mass in kilograms that they actually have, thereby answering the two objections. Earlier I assumed that grounds necessitate what they ground. Having distinguished these senses of possibility the question arises as to which notion of possibility makes this assumption true. Is it the fundamental sense or the reduced sense? And if the latter (context sensitive) notion, what are the relevant contexts? This is a deep question, but it is beyond the scope of the current paper so I will not try to settle it here. It sufces for current purposes to describe the relevant sense by pointing to paradigm examples of ground: that Europes being at war was grounded in the actions of its citizens, that the existence of a table is grounded (say) in the existence and arrangement of various particles, and so on. The relevant sense of necessity is then the sense in which those grounds intuitively necessitate what they ground. So it is the sense in which it is impossible for the citizens to act like that and be at peace, and in which it is impossible for the particles to be arranged like that and there not be a table, and so on. So when I said that the qualitative goings-on in our solar system do not necessitate Obamas existence or that my laptops being twice as massive as IPK does not necessitate its being 2 kgs, the claim was that they do not necessitate in that sense (whatever it is). Which seems true.49

11

Pluralism and Symmetry

I have argued that if the world is fundamentally qualitative then the individualistic facts are plurally grounded in the qualitative. I also argued that if mass if fundamentally relational then the kilogram facts are plublind review]. 49 Thanks to Juhani Yli-Vakkuri for a helpful conversation on this point. As it happens, I believe that the notion of necessity in question may well be Fines notion of necessity that applies to those truths that follow from the essences of things. I develop this idea a little in XXXX [reference omitted for blind review].

36

rally grounded in those mass relations. Both claims are conditional so the arguments did not purport to show that there are actual cases of plural grounding. But they do suggest that the consequents of the conditionals are coherent and intelligible hypotheses that are worth taking seriously. So our view about the logical structrure of ground should allow for them: we should be pluralists and think that the logical form of a claim about ground is irreducibly plural, i.e. that they are grounded in them. One consequence is that we need to take care when linking ground with fundamentality. Schaffer [25] says that a fact is fundamental iff it has no ground. But this is problematic if one also thinks that the fundamental facts are (pictorially speaking) those that lie at the bottom of the great chain of being, those that (as the metaphor goes) God had to determine when making the world. For if pluralism about ground is correct then a fact may have no ground but be part of a plurality of facts with a ground. In this case the fact would count as fundamental in Schaffers sense, but is not something that God had to determine when making the world and so (in the relevant sense) does not lie at the bottom of the great chain of being. If one wants to use the word fundamental to track facts at the bottom of the great chain, one should say instead that a fact is fundamental iff it is not one of a plurality with a ground. So we should distinguish two senses of fundamentality: one (Schaffers) tracks those facts without grounds, while the other tracks those facts at the bottom of the great chain. If pluralism about ground is correct, these two senses come apart. Conating these two senses can lead to mistakes. Indeed I suspect that the conation lies behind much of the attraction of individualism. Suppose you think that the existence of a given individual is not grounded in further individualistic facts. And suppose you think (correctly, in my view) that it has no qualitative ground either. It follows that its existence is fundamental in Schaffers sense. If you conate the senses of fundamentality, you will then think that its existence lies at the bottom of the great chain of being, that it is something that God had to determine when making the world. So you will think that individualism is true. But this reasoning equivocated on these senses of fundamentality. For even if its existence is fundamental in Schaffers sense and lacks a qualitative ground, it may nonetheless be one of a plurality of individualistic facts that together have a qualitative ground. If so, then its existence does not lie at the bottom of the great chain after all; contra individualism. (I also suspect that a similar mistake lies behind the attraction of absolutism.) So an important project, I think, is to identify occurrences of this kind of mistake, as I tried to do earlier in the case of individuals and kilograms. Other mistakes can stem from failing to recognize that ground is irreducibly plural too. I will nish by identifying one. It involves a famous argument against qualitativism that turns out to be unsound if pluralism about ground is true. The rst premise of the argument states that the 37

Max Black world discussed in section 2 is possible, i.e. that it is possible for there to be just two spheres of iron located 2 miles apart which share all their qualitative properties (they are of exactly the same mass, color, shape, etc). The second premise is that the qualitativist cannot make sense of this possibility. I will argue that the second premise is false if pluralism about ground is true.50 To see this, start by asking how the second premise is to be justied. One question is whether the qualitativist has the resources to describe the fundamental, qualitative facts of a Max Black world. As we saw in section 2, some qualitativists appear to be unable to do this. For example if the traditional bundle theorists view is that the underlying qualitative facts just concern which intrinsic, qualitative properties are compresent, it is difcult to see how she can describe a situation in which there are two individuals with the same such properties. But (again as we saw in section 2) other qualitativists have no problem with this. For example, a qualitativist might think that the fundamental qualitative facts are existentially general facts that can be expressed in predicate logic with identity but without constants, in which case the Max Black world can be described as follows: (x)(y)(Fx & Fy & x6=y)

where F expresses the qualities of each sphere.51 But still, even if the qualitativist can described the fundamental facts of a Max Black world, one might justify the second premise by arguing that there is no way to ground the individualistic facts about the two spheres in those underlying qualitative facts. This is close what Adams had in mind when he wrote that . . . the clearest way of proving the distinctness of two properties is usually to nd a possible case in which one would be exemplied without the other. In order to establish the distinctness of thisnesses [i.e. individualistic properties] from all suchnesses [i.e. qualitative properties], therefore, one might try to exhibit possible cases in which two things would possess all the same suchnesses, but with different thisnesses.52 Label one of the spheres A and the other B. Put in terms of properties, Adams observation is that A and B share their qualitative properties and yet sphere A has the individualistic property of being identical to A while
See Hawley [9] for a discussion of others. 51 Even those qualitativists such as myself, who do not wish to treat existentially general facts as fundamental, can nd other qualitative facts sufcient to describe the fundamental nature of a Max Black world. I say more about this in XXXX [reference omitted for blind review]. 52 Adams [1], p. 12
50 This is not the only available response to the argument.

38

B does not. This sufces to show that the individualistic property is distinct from any of As qualitative properties, which was Adams aim. To argue that the individualistic property is not grounded in any of As qualitative properties, we just add the assumption that if a property P is grounded in property Q, then necessarily anything with Q also has P. That is the argument put in terms of properties, but since we take ground to be a relation between facts let us reconstruct it in those terms. To this end, consider the fact that A exists and the fact that B exists. In what might each of these be grounded? Putting the possibility of plural grounding aside, there must be some fact about the distribution of qualitative properties that explains As existence, and likewise for B. But, one would argue, the qualitative facts that explain As existence must be different from those that explain Bs. After all, if one asked what explains As existence and got an answer, and then asked what explains Bs existence and got the very same answer, one would naturally want to reply Wait a minute, that was what explained As existence; what then makes this it the case that B exists? Now, since both spheres have many qualitative properties, one could try saying that As existence is explained by somethings being iron and spherical, and that Bs existence is explained by somethings being black and hard. But this would be utterly implausible: since A and B share all their qualitative properties, it would be a mystery why being black and hard explains Bs existence but not As. Therefore, the argument goes, nothing qualitative can plausibly be said to ground As existence and not Bs; and so As existence has no qualitative ground. The argument is therefore slightly different than that which was run against the traditional Bundle Theory. In that case, the Bundle Theory logically implied that the spheres were identical, contrary to hypothesis. In this more general case there is no such implication; instead, the charge is now that there is no plausible explanation of their existence. That is, I believe, the best defense of the second premise. How should qualitativists respond? Interestingly, they almost uniformly grant the second premise and instead deny the rst. Thus, the literature is full of qualitativists bending over backwards to show that we may plausibly deny the possibility of a Max Black world. For example, Hacking argues that a Max Black world can be re-described as a world in which there is just one sphere situated in a non-Euclidean space so tightly curved that it is 2 miles from itself.53 And Hawthorne argues that individuals can be multiply located in space, so that the Max Black world can be re-described as a Euclidean world in which a single individual A is located 2 miles from itself.54 There appears to be an implicit assumption, then, that to block the argument the qualitativist must deny that Max Black worlds are possible.
53 See 54 See

Hacking [8]. his [18], originally published under the name OLeary-Hawthorne.

39

But if pluralism about ground is true then this assumption is false. Even if we concede the possibility of Max Black worlds, the above argument for the second premise at best shows that neither As existence nor Bs existence has a qualitative ground on its own. But if pluralism is true then it remains open that the individualistic facts in the Max Black worldincluding As existence and Bs existenceare plurally grounded in the worlds qualitative nature even though none of them have a qualitative ground on their own, just as the structuralist states. As a result, the qualitativist may concede the possibility of Max Black worlds and yet deny that they are problematic for her view. Qualitativists should welcome this result, for there are compelling arguments based on plausible assumptions that Max Black worlds are indeed possible (for example, Adams argument from the possibility of two spheres that are almost qualitatively identical). A qualitativist who denies the possibility of Max Black worlds must therefore deny those plausible assumptions, but if pluralism is true there is no need for her to do so. Of course, this is by no means a full defense of qualitativism since there are other arguments to contend with. Still, it is an example of a case in which recognizing the possibility of plural grounding has a signicant, and perhaps surprising, impact on an issue in metaphysics. Whether other issues are similarly affected is a question I leave for another time.

12

Conclusion

The recent literature on ground has uniformly assumed what I call singularism, according to which the logical form of a claim of grounds is that this (a single fact) is grounded in them. I have argued that if certain assumptions about the fundamental nature of the world are granted then it is plausible that certain collections of facts are grounded plurally in the worlds underlying nature: they (the members of the collection) are grounded in them even though none of them admits of a ground of its own. Our view about the logical structure of ground should therefore allow for these hypotheses: we should think that ground is irreducibly plural. If this is right then it is important that we ensure that our metaphysical theorizing about is not implictly infected with singularist assumptions.55

References
[1] Adams, R. 1979. Primitive Thisness and Primitive Identity. Journal of Philosophy 76: 526.
55 Acknowledgements

omitted for blind review.

40

[2] Armstrong, D. 1988. Are Quantities Relations? A Reply to Bigelow and Pargetter. Philosophical Studies 54. [3] Correia, F. 2011. Grounding and Truth-Functions. Logique et Analyse 53: 251279. [4] Eddon, M. Fundamental Properties of Fundamental Properties. Forthcoming in Oxford Studies in Metaphysics: Volume 8. [5] Fine, K. 2001. The Question of Realism. Philosophers Imprint 1 (1): 130. [6] Fine, K. 2012. Guide to Ground. In F. Correia and B. Schnieder (eds), Metaphysical Grounding: Understanding the Structure of Reality. Cambridge: Cambridge University Press. [7] Garrett, D. 1991. Spinozas Necessitarianism. In God and Nature: Spinozas Metaphysics, edited by Y. Yovel, pp. 191218. [8] Hacking, I. 1975. The Identity of Indiscernibles. Journal of Philosophy 72: 249256. [9] Hawley, K. 2009. Identity and Indiscernibility. Mind 118: 101119. [10] Hawthorne, J. and T. Sider 2002. Locations. Philosophical Topics 30: 53 76. [11] Krantz, D., R. Luce, P. Suppes and A. Tversky 1971. Foundations of Measurement: Volume 1 New York: Academic Press. [12] Kripke, S. 1972. Naming and Necessity. Oxford, UK: Blackwell. [13] Ladyman, J. and D. Ross. 2007. Every Thing Must Go: Metaphysics Naturalized. Oxford: OUP. [14] Leuenberger, S. Grounding and Necessity. Manuscript. [15] Lewis, D. 1986. On the Plurality of Worlds. [16] Litland, Jon. On Some Counterexamples to the Transitivity of Grounding. Manuscript. [17] Mundy, B. The Metaphysics of Quantity. Philosophical Studies 51: 29 54. [18] OLeary-Hawthorne, J. 1995. The Bundle Theory of Substance and the Identity of Indiscernibles. Analysis 55: 1916. [19] Paul, L. A. 2002. Logical Parts. Nous 36: 57896. 41

[20] Paul, L. A. 2012. Building the World from its Fundamental Constituents. Philosophical Studies 158: 22156. [21] Pooley, O. 2005. Points, Particles and Structural Realism. In The Structural Foundations of Quantum Gravity, edited by Rickles, French and Saatsi. Oxford, UK: Oxford University Press. [22] Rosen, G. 2010. Metaphysical Dependence: Grounding and Reduction. [23] Saunders, S. 2003. Indiscernibles, General Covariance, and Other Symmetries: The Case for Non-Reductive Relationalism. In J. Renn (ed.), Revising the Foundations of Relativistic Physics. Kluwer. [24] Sider, T. Writing the Book of the World, manuscript. [25] Schaffer, J. 2009. On What Grounds What. In Metametaphysics, edited by D. Chalmers, D. Manley and R. Wasserman, pp. 34783. Oxford, UK: OUP. [26] Schaffer, J. 2010. The least discerning and most promiscuous truthmaker. The Philosophical Quarterly 60: 307324. [27] Schaffer, J. 2012. Grounding, transitivity, and contrastivity. In F. Correia and B. Schnieder (eds), Metaphysical Grounding: Understanding the Structure of Reality. Cambridge: Cambridge University Press. [28] Skiles, A. Against Grounding Necessitation. Manuscript. [29] Skow, B. 2008. Haecceitism, Anti-Haecceitism, and Possible Worlds. The Philosophical Quarterly 58: 98107. [30] Tahko, T. E. Truth-Grounding and Transitivity. Forthcoming in Thought. [31] Trogdon, K. Grounding: Necessary or Contingent? Forthcoming in Pacic Philosophical Quarterly

42

Symmetry as an Epistemic Notion (Twice Over)

Shamik Dasgupta
Draft of May 2013

Symmetries in physics are a guide to reality. That much is well known. But what is less well known is why symmetry is a guide to reality. What justies our inferences that draw conclusions about reality from premises about symmetries? I argue that answering this question reveals symmetry to be an epistemic notion twice over. First, these inferences must proceed via epistemic lemmas: premises about symmetries in the rst instance justify epistemic lemmas about our powers of detection, and only from those epistemic lemmas can we draw conclusions about reality. Second, in order to justify those epistemic lemmas, the notion of symmetry must be dened partly in epistemic terms.

1
1.1

Symmetry-to-reality reasoning
A rough introduction to symmetry

I said that symmetry in physics is a guide to reality. Let me illustrate what I mean with a simple example.1 Suppose that we have compelling evidence to think that the world is ultimately made up out of point particles with mass. And suppose that the way they move looks for all the world (given our evidence) like they are governed entirely by just two laws: f=ma and the inverse-square gravitational force law. Call these the laws of NG (Newtonian Gravitation). Now the laws of NG (like all laws) come associated with a set of symmetries. What does this mean? We are familiar with the idea that material objects can exhibit symmetries, for example my plate is (roughly) symmetric through straight lines that bisect its centre. But what does it mean for a law to exhibit symmetries? This is one of the central questions of this paper and I will not offer an answer until near the end. But we need
1 I hope that seasoned philosophers of physics will bear with me over the next 2 pages as I introduce what to them will be familiar terms. My policy is to write (as far as possible) so as to be accessible to a smart rst-year graduate student working in any area of philosophy.

a rough characterization to get us started. So for now let us think of a symmetry of a law as a transformation on physical systems that (at a minimum) preserves the truth of the law. The symmetries of NG include rigid spatial translations and rotations, rigid temporal translations, and uniform velocity boosts.2 For example, consider a possible closed physical system in which some particles are moving around in a way that conforms to the laws of NG. And now consider a second possible closed system just like the rst except that it is located 3 feet to the right of the rst. In this second system the particles are moving around just as they are in the rst with the one exception that at any given time any given particle is 3 feet to the right of where it is in the rst system. It is easy to prove that this second system also conforms to the laws of NG. After all, the masses of all its particles are exactly the same, as are their distances between one another and their accelerations (at any given time). And since the laws of NG govern just these quantities, the second system must therefore conform to NG if the rst one does. Now the operation that takes the rst system as input and yields the second as output is called a rigid spatial translation, and so what we have shown is that this translation preserves the truth of NG. And clearly all such translations (for arbitrary distances and directions) preserve the truth of NG too, and they are all said to be symmetries of NG. Similarly, consider a third possible closed system just like the rst except that it is moving at 5 mph towards the north relative to the rst. In this third system the particles are moving around just as they are in the rst system with the one exception that at any given time any given particle is moving 5 mph faster towards the north than it is in the rst system. It is easy to prove (in the same way as above) that this third system must also conform to the laws of NG if the rst one did. And the operation that takes the rst as input and yields the third is called a uniform velocity boost, and all such boosts (for arbitrary directions and speeds) preserve the truth of NG and are said to be symmetries of NG too. This is (as I said) very rough as it stands. At some point we must say more about what a physical system is (a possible world, or a part of a possible world, or a model, or a set of sentences, etc). And we will see that to count as a symmetry the transformation must preserve more than just the truth of the laws. But this loose characterization communicates the rough idea well enough for now. Now the symmetries just described preserve some quantities but not others. Any two possible closed systems related by a rigid translation or uniform velocity boost clearly agree on all facts about the distances between particles at times, their relative velocities, and so on. These quan2 These are sometimes called dynamical symmetries, e.g. in Earman [10], but here I drop the qualier for brevity.

tities are therefore called invariant in NG: their values are preserved by all of NGs symmetries. But other quantities are altered by a symmetry of NG. For example, suppose that these point particles are moving around in Newtonian space, in which each particle at any given time has an absolute position in that space and an absolute velocity through it. Then a (non-trivial) translation will change the absolute position of each particle and a (non-trivial) boost will change each particles absolute velocity. These quantities are therefore called variant in NG: if particles have these quantities then their values are altered by a symmetry of NG.

1.2

The symmetry-to-reality inference

The way in which symmetry is used a guide to reality can now be easily stated. The idea is that if a putative feature is variant in laws that we have reason to think are true and complete, then this is some reason to think that the feature is not real. So, if we had reason to think that the laws of NG were true and complete then the idea is that because absolute velocity is variant in NG (i.e. there are symmetries of NG that alter its values) we would also have reason to think that absolute velocity is not real and endorse a view of space and motion according to which there is no such thing. The resulting view might be that the point particles live in a Galilean spacetime structure in which there is no notion of absolute velocity. Or it might be a more radical relationalist picture on which there is no such thing as space or spacetime through which the particles are moving. But we need not settle on the specic metaphysics: the point is just that some such view is motivated. The line of reasoning generalizes to other physical theories too. Absolute simultaneity is (famously) variant in the laws of the special theory of relativity. And it is commonplace to think that, insofar as we consider those laws to be true and complete, we should think that there is no such thing as absolute simultaneity (perhaps by endorsing the view that spacetime instantiates a Minkowski structure). From cases like these we can abstract the form of these symmetry-toreality inferences as follows: (1) Laws L are the complete laws of motion governing our world. (2) Feature X is variant in L. (C) Therefore, X is not real. This form of inference is ubiquitous. Earmans famous principle SP1 according to which it is a condition of adequacy of ones metaphysics of spacetime that it not contain any variant featuresis (in effect) an endorsement of this inference, and he puts his principle to work throughout 3

his book.3 North agrees that this form of inference is a methodological principle in physics, a principle that she then uses to investigate the metaphysical implications of Hamiltonian and Lagrangian formulations of classical mechanics.4 And Baker writes that this use of symmetries has many metaphysical applications that he then explores.5 The inference is also implicit in the motivation behind recent structuralist theses about spacetime. For example, consider recent discussions of diffeomorphism invariance in General Relativity. These are symmetries that alter facts about which particular regions of the manifold instantiate which parts of the elds. And many have inferred from these symmetries that there must therefore be a sense in which the particular regionswhen considered as independent of the eldsare not real. On one reconstruction, this is just the symmetry-to-reality inference at work.6 So far I have focused on various symmetries that consist of spatiotemporal transformations (boosts, translations, etc). But it is also natural to consider other transformations too, for example uniform multiplications of mass (perhaps with an associated alteration of the gravitational constant or spatial distances). If such transformations can be shown to be symmetries of NG in the same sense that boosts and rigid translations are, the symmetry-to-reality inference would have us endorse a relational view of mass on which only the mass relationships that are invariant are real and any further notion of intrinsic mass is unreal.7

1.3

Two Questions

So symmetry-to-reality reasoning is ubiquitous. here I ask two (related) questions about it. First, how is it to be justied? I am asking this in the spirit of reconstructive epistemology, much as one might ask why we are justied in trusting the testimony of others, or why inductive inferences are justied, or why modus ponens is justied. The question is not whether they are good inferences: that is assumed at the outset. The question is why they are good inferences, i.e. what justies them.8 Nor does the question concern the sociological history of these
Earman [10], p. 46 for his statement of this principle. North [17]. 5 See Baker [1], p. 1157. 6 See Dasgupta [6], Ladyman and Ross [16], and Pooley [18] for just three examples. Sometimes the inference here goes via issues of determinism, as in the presentation of the Hole Argument in Earman and Norton [9]. But in Dasgupta [6] I argue that the detour via determinism is inessential to the matter, and I think that Ladyman and Ross agree. See Hoefer [13] for more on why determinism is not the crucial issue. 7 I develop this line of argument in Dasgupta [7]. 8 Belot [3] has recently argued that they are not justied. But (as I will discuss later) we disagree less than might be apparent. For what he argues is that if symmetry is dened in various ways then the inference is unjustied, and I agree with that. My question (as will
4 See 3 See

inferences or the psychological processes we undergo when making them. It is rather an epistemic question of justication. Why do premises about symmetries justify metaphysical conclusions about what is real? What is wrong, epistemically speaking, with favoring a Newtonian view of space over a Galilean one? What epistemic mistake would one be making? It is not immediately obvious. Someone who believes that the laws of NG are true and complete but who favors the Newtonian view of space over a Galilean one cannot be accused of logically inconsistent belief. This is just to say that these inferences are not valid. But they are often good, just as inductive inferences are invalid but (sometimes) good. I want to know what makes them good, when they are good. One might say with Earman that it is a condition of adequacy on a space-time theory that it does not contain features that vary under the symmetries of the laws. But this just restates our data that symmetry-toreality inferences are (sometimes) good: what I want to know is why this should be a condition of adequacy on a space-time theory. One answer might be that there is a basic epistemic norm favoring metaphysical theses that dispense with variant features, where by basic I mean that it cannot be justied in terms of further epistemic norms. Examples of basic epistemic norms might be perception (trust your senses), induction, and modus ponens. It is precisely because these norms are basic that the classic skeptical problems arise: if the skeptic asks why we should expect the future to resemble the past it is notoriously difcult to say anything that does not sound horribly question-begging! But it is wildly implausible that symmetry-to-reality inferences are epistemically basic in the same way that an inductive inference arguably is. Rather, one suspects that if a symmetry-to-reality inference is good that is because (i) variant features have some property (such as being redundant) and (ii) there is some more basic norm advising us to dispense with features with that property. Our question, then, is what this property is. The question becomes more pressing when one looks at what is often meant by symmetry. For the notion is often dened in purely formal, mathematical terms, so that whether a given transformation is a symmetry of a given set of laws just depends on the formal and mathematical features of those laws and their models. But why should those features of the laws have anything to do with metaphysics, with whats real? It is not at all obvious, at least not at rst glance. There is a question, then, as to what symmetry could mean, such that the symmetry-to-reality inference is justied. That is my second question.
be clear below) is what else symmetry could mean, such that the inference is justied.

1.4

Two Answers

I will argue for epistemic answers to both questions. With respect to the rst question, I will argue that we have reason to think that variant features are not real because (i) they are undetectable, and (ii) there is a more basic Occamist norm advising us to dispense with undetectable structure. As a result, the symmetry-to-reality inference when fully spelled out proceeds via an epistemic lemma concerning detectability, as follows: (1) Laws L are the complete laws of motion governing our world. (2) Feature X is variant in L. (3) Therefore, X is undetectable (from (1) and (2)). (C) Therefore, X is not real (from (3) and an Occamist norm that we dispense with undetectable structure). On this view the argument is driven by an Occamist norm that has nothing to do with symmetry per se but rather with detectability. Thinking about symmetries just gets us into a position to see that various putative features of the world are vulnerable to the Occcamist norm. The second question, then, becomes the question of what symmetry could possibly mean such that the inference to (3) is justied. I will argue that the standard denitions of symmetry in formal and mathematical terms leave that step unjustiable, but that it is justiable if symmetry is dened in broadly epistemic terms. That is why I think that symmetry is an epistemic notion twice over: once with respect to its denition, and once with respect to its role in metaphysical reasoning. In what follows I will focus almost entirely on the very simple case of NG and its symmetries like boosts and translations. It is sometimes thought that the symmetry-to-reality inference is easy to justify in this simple case and that the more interesting question is to what extent the inference generalizes to other theories and symmetries.9 But I believe that even in the very sanitized setting of NG the inference is considerably more involved than is sometimes recognized. So my approach here is to focus on NG, in the hope that by clarifying this simple case we will then be in a better position to see how it might generalize.
example, Belots [3] is for the most part an argument to the effect that the inference does not generalize. But he says very little about how the inference is to be justied in the very simple cases I have in mind.
9 For

2
2.1

Against Redundancy
Redundancy

Let me start by arguing against what is perhaps the most common alternative to my reconstruction of the inference. On this alternative, we have reason to think that variant features are not real because (i) they are redundant, and (ii) there is a more basic Occamist norm advising us to dispense with redundant features. The inference on this view then looks like this: (1) Laws L are the complete laws of motion governing our world. (2) Feature X is variant in L. (3*) Therefore, X is redundant (from (1) and (2)). (C) Therefore, X is not real (from (3) and an Occamist norm that we dispense with redundant structure). On this view we can draw metaphysical conclusions from symmetries without the detour through epistemic claims about what is undetectable. I suspect that this is the standard view (if any view deserves that title), and it has been defended explicitly by a number of people including Baker [1], Earman [10], and North [17] (to name just three). But I do not think this reconstruction of the inference works. For what does redundant mean? The term has many meanings. But if the above reconstruction is to work there must be a meaning such that (i) being redundant follows from being variant (so that the inference to (3*) is justied), and (ii) being redundant in that same sense gives us reason to think that the feature is not real (so that the inference from (3*) to (C) is justied). I will argue that there is no meaning that plays both these roles. Now it might seem impossible to argue for this until we determine what symmetry means, and hence what variant means, for until then we will not know what the real content of (2) is. But we can proceed by thinking just about absolute velocity. For uniform velocity boosts are paradigm examples of symmetries of NG, so whatever symmetry means absolute velocity had better turn out to be variant in NG. So if this reconstruction of the inference is to work there must be a meaning of redundant on which (i) absolute velocity is redundant in NG, and (ii) being redundant in this same sense gives us reason to think that the feature unreal. I will argue that there is no meaning that plays both these roles.

2.2

The Non-Redundancy of Absolute Velocity

Let us then canvass some different readings of redundant. On one reading a feature is redundant iff it makes no difference to how a system will 7

evolve over time. Baker appears to use this reading when defending this version of the symmetry-to-reality inference. He says that we should dispense with variant features because they are not difference-makers. And something is a difference maker, he says, if it makes some difference in how the state of a (physically possible) world evolves in time.10 Now it may be that if a feature is redundant in this sense then we have reason to think that it is unreal. But the trouble is that variant features are not necessarily redundant in this sense: even if the laws of NG are true and complete, absolute velocity is not redundant in this sense of the term. Why might one think that absolute velocity is redundant in this sense? Consider a physical system composed of point particles with masses moving around in Newtonian space in accordance with NG. Suppose at T1 it is in a state S1, and suppose that the laws of NG imply that by a later time T2 it will have evolved into a state S2. Now consider all the interparticle distances in S2. Symmetry considerations show that facts about absolute velocity in the earlier state S1 make no difference to those later inter-particle distances. After all, a uniformly boosted system is one in which (i) the earlier absolute velocities are all uniformly different, (ii) the later inter-particle distances are the same, and yet (iii) the laws of NG still obtain (by the fact that uniform boosts are symmetries of NG). It follows that the absolute velocities of the particles at T1 can vary uniformly in uncountably many ways and the laws of NG will still imply that the particles will end up standing in exactly the same inter-particle distances at T2 as they actually do. In this sense absolute velocity makes no difference to future facts about inter-particle distances. My guess is that this line of thought is what lies behind Bakers suggestion. But the absolute velocities of the particles at T1 will make a difference to the absolute velocities of the particles at T2! In a boosted system, the absolute velocities at T1 are all different and the absolute velocities at T2 are all correspondingly different too. So if there is such a thing as absolute velocity, it is a difference maker even though it is variant. So in this sense of redundant, a features being variant does not imply that it is redundant.11 Of course one might conclude on the basis of a symmetry-to-reality inference that absolute velocity is not real, and it would then be true that it is not a difference maker. But that would beg the current question: we are trying to establish how the symmetry-to-reality inference works in the rst place, so we cannot assume that absolute velocity is unreal. So much for difference-making. A second (and related) idea is that something is redundant iff it is dispensable from all explanations of physical phenomena. Now it may be that if something is redundant in this sense then we have reason to think that its not real. But the point just
10 Baker 11 A

[1], p. 1159. similar point is made in Sklar [21].

made shows that variant features are not necessarily redundant in this sense of the term: the absolute velocities of the particles at T1 will be vital to explaining their later absolute velocities precisely because if their current velocities were different then their later velocities would differ too. Of course if absolute velocity was undetectable then it would be dispensable from all explanations of detectable phenomena. And perhaps that would be reason to think that it is unreal. But to show that absolute velocity is redundant in this sense we would have to show that it is undetectable, and we are currently exploring justications of the symmetry-to-reality inference that do not proceed via this epistemic lemma. A third suggestion is that a feature is redundant iff its values are arbitrary.12 The idea is that because absolute velocity varies under the symmetries of NG there is no explanation as to why the absolute velocities of things are as they actually are rather than some boosted variation. But the suggestion is confused. If the idea is that the absolute velocity of a given particle can never be explained then the idea is false: we just saw that its absolute velocity at a given time will be explained (in NG) by its absolute velocity at prior times (and the forces operating on it). Perhaps the thought is that the absolute velocities at the initial condition are unexplained and arbitrary. If so, the charge is correct but would apply to the values of all features whether variant or invariant, and so cannot be that property of variant features that would justify dispensing with them.

2.3

Some Redundancies

We have discussed senses of redundant on which absolute velocity is not redundant. There are other senses on which it is redundant, but unfortunately they are not senses that give us reason to dispense with it. One idea is that a feature is redundant to some laws iff it is not needed to support those laws, where this means that one can formulate the laws without assuming that the feature is real. Perhaps this is what Earman had in mind when he wrote: The motivation [for the symmetry-to-reality inference] derives from combining a particular conception of the main function of laws of motion with an argument that makes use of Occams razor. Laws of motion, at least as they relate to particles, serve to pick out a class of allowable or dynamically possible trajectories. If [there are variant features], the same set of trajectories can be picked out by the laws working in the setting of a weaker space-time structure. The theory that [posits variant features] is thus using more space-time structure than is needed to
12 This

sort of idea obviously has its roots in the Principle of Sufcient Reason.

support the laws.13 It may also have been what North had in mind when she justied the symmetry-to-reality inference by saying we should go with the least structure. We should not posit structure beyond that which is indicated by the fundamental dynamical laws.14 And it may have been what Belot had in mind when he wrote Why was it (in hindsight, in one sense) a mistake for Newton to postulate absolute space? Because he thereby postulated more spacetime structure than was required for his dynamics.15 It is uncontroversial that absolute velocity is redundant to NG in this sense: it is possible to formulate the laws as governing the trajectories of particles through a Galilean space-time structure in which there is no such thing as absolute velocity.16 Admittedly, it is not immediately obvious that whenever a given feature is variant in some laws it is redundant to those laws in this sense. But let us grant that this is so for the sake of argument. The trouble is that it is not at all clear why a features being redundant in this sense is reason to think that it is not real. Why should the fact that I can describe the laws of motion without mentioning a given feature constitute evidence that the feature is unreal? It is worth expanding on this point. The kind of physical theory under discussion contains two components. First, it describes the fundamental physical ontology of the world, the kind of stuff that the physical world is made up out of. For example it might describe what kinds of particles or elds there are, what kind of spacetime structure they live in, and so on. Call this the theorys metaphysics. And second, the theory will describe how that aforementioned physical stuff behaves. These are the theorys laws. Now the remarkable thing about the symmetry-to-reality inference is that it uses facts about one half of your theory (the laws) to inform the other half (the metaphysics). But it is highly unobvious why this is reasonableafter all, they are different parts of the theory! Earmans idea is that if you can write the laws without mentioning some feature in the metaphysics, then that is reason to think that the feature is unreal. But it is not at all clear why this should be so: why think that every real feature must be mentioned when writing down one half of your theory? Keep in mind that even though one can formulate NG without mentioning absolute velocity, absolute velocity is nonetheless a difference maker
[10], p. 46; my emphasis [17], p. XX; second emphasis mine. 15 Belot [3], p. XX; my emphasis. 16 I am slurring over a difculty here. For one might argue that the laws when formulated over a Galilean space-time structure are different than when formulated over a Newtonian space. After all, the quantity of acceleration will be dened differently in each case. So if the statement in the text is true, that is because we are using a notion of sameness of law that would count them as being the same law even in light of these differences. But it would be distracting to elaborate on this here.
14 North 13 Earman

10

and indispensable to explanations of physical phenomena in an NG system, in the senses discussed above. Why then is the mere fact that one can formulate the laws without mentioning it a reason to think that it is unreal? The same point applies to another sense of redundancy. For one might say that a feature is redundant to some laws iff its values are irrelevant to whether the laws obtain. Absolute velocity is redundant to NG in this sense: the particular absolute velocities of things are irrelevant to whether the laws of NG obtain, in the sense that we can uniformly change them however we wish and the laws will remain true. Moreover this follows from the fact that absolute velocity is variant in NG. So perhaps it is this sense of redundancy that Earman had in mind in the above quote. But the trouble again is that it is not at all clear why being redundant in this sense is reason to think that a feature is unreal. Remember, a feature can be redundant in this sense and yet (like absolute velocity) be a difference maker and be indispensable to explanations of physical phenomena. All redundancy means, on this interpretation, is that the particular values of a feature from one half of our theory (the metaphysics) is irrelevant to whether the other half of the theory (the nomos) is true. But there is no obvious reason to think that every categorical aspect of the world must be relevant in this sense to whether the laws obtain. For these reasons I do not think that this non-epistemic reconstruction of the symmetry-to-detection inference, in terms of redundancy, succeeds. Perhaps its initial attraction rests on an equivocation between different notions of redundancy. It is true that absolute velocity is redundant in NG in the two senses just discussed: it is not needed to formulate the laws and its values are irrelevant to whether NG is true. And it may be true that being redundant in the earlier sensesnot being a difference maker, being dispensable from explanations of physical phenomenais reason to think that a feature is unreal. But the senses of redundancy are not the same. Now there is a sense of redundant on which (i) absolute velocity is redundant in NG, and (ii) its being redundant implies that it is not real. It is the sense on which a feature is redundant iff models that differ only in a uniform transformation of its values must represent the same possible world. It may well be that models related by a uniform velocity boost represent the same possible world and so it may be that absolute velocity is redundant in this sense. But the problem with this sense of redundancy for our purposes should be obvious: we would only be justied in believing that uniform velocity boosts must represent the same possible world if we already had reason to think that absolute velocity is not real. So the claim that velocity is redundant in this sense can hardly be used as a nonquestion-begging premise in an argument that it is not real. These points are all straightforward, so why are they often ignored? One reason might be the literatures focus on technical formulations of the matter, which tends to obscure the important epistemic questions. It 11

is easy to show that physical situations differing in a uniform velocity boost are related formally in various ways, and that the variant features are therefore redundant in some formal sense. But this is not enough. The question remains as to why a features being redundant in that formal sense is reason to believe that it is unreal. As we have seen, it is surprisingly difcult to answer the question. Indeed there is an independent consideration that raises a doubt about any non-epistemic reconstruction of the symmetry-to-reality inference. If we could see or detect absolute velocity, we should believe that it is real regardless of whether it is redundant in any of the above senses. So a necessary condition on our reasonably believing that it is unreal is that we do not believe that it is detectable. And from there one might argue that a necessary condition on our reasonably believing that it is unreal is that we believe that it is undetectable. If that is right then it is hard to see how any non-epistemic reconstruction of the inference could work.

3
3.1

From Symmetry to Detection

The Epistemic Approach

For these reasons I believe that the symmetry-to-reality inference must proceed via the epistemic claim of undetectability, as outlined earlier: (1) Laws L are the complete laws of motion governing our world. (2) Feature X is variant in L. (3) Therefore, X is undetectable (from (1) and (2)). (C) Therefore, X is not real (from (3) and an Occamist norm that we dispense with undetectable structure). But now I must say why this is a good inference.

3.2

The Occamist Norm

I will focus on the inference from (1) and (2) to (3), but let me say something briey about the inference from (3) to (C). Vericationists might say that it is analytic, that it is part of the meaning of reality that everything real is detectable. But we need not take such a strong line. All we need (roughly speaking) is a plausible Occamist norm to the effect that positing undetectable structure is an epistemic vice, in the sense that all else being equal (or near enough equal) we should prefer theories that do not posit undetectable structure. (This will be qualied as we go along, but this basic idea will do for now.) 12

The plausibility of this norm depends on what is meant by undetectable. If the term was used to include anything that we cannot see with the naked eye, the norm would recommend that we become radical scientic anti-realists and dispense with so-called theoretical entities such as electrons. But that is not how I use the term. Instead, something is undetectable in my sense if, roughly speaking, it is physically impossible for it to have an impact on our senses. Electrons are therefore detectable in this sense because there are physically possible processes, such as those that occur in particle accelerators, by which the presence of an electron can be made to have an impact on our senses via its impact on (say) the movement of a dial or an image produced on a computer screen. In contrast, the idea is that if the laws of NG are true and complete then it is physically impossible for the absolute velocity of any given body to have an effect on our senses. So understood, the motivation behind this Occamist norm is forthcoming. Roughly speaking, we should think that a given physical feature is real only if we have empirical evidence that it is real. If we can show that the feature is undetectable we will have shown that we do not (and cannot) have empirical evidence in the form of observations or measurements of it. It remains possible that dispensing with the feature yields a theory that has too many other vices to warrant belief, such as being too inelegant or complex. In that case we would have empirical evidence of sorts that the feature is real, in the sense that our all-things-considered best empirically conrmed theory implies that it is real. But that is a situation in which all else is not equal. When that kind of situation does not arisewhen the theories that dispense with the feature do not have those other epistemic viceswe will have no empirical evidence to think that the feature is real after all. And so in those situations we should endorse those theories that dispense with it. Which is what our Occamist norm states. Note, then, that if this how symmetry-to-reality inferences work then we can only draw the conclusion of the inference when we have the alternative theory in hand and have shown that all else is equal. This explains why it was rational for Newton to believe in absolute velocity even though he knew that it was variant in NG and knew that it was undetectable. The reason this was rational for him was that he had no good alternative theory to hand. He had good reason (his bucket argument) to think that relationalism was not empirically adequate. And relationalism was the only alternative view he knew of (he was not aware of Galilean space-time structures in which there is a well dened feature of absolute acceleration (as required by his bucket argument) but no absolute velocity). So for Newton, all else was not equal and he was rational to believe in absolute velocity.17
17 There

is a lesson here for contemporary structuralists, such as Ladyman and Ross [16],

13

3.3

From Symmetry to Detection

Let us now focus on the move from (1) and (2) to (3). How can we argue, just on the basis that a given feature is variant in what we take to be the true and complete laws of motion, that it is undetectable? Let us start simple, again with the case of absolute velocity. Why is it reasonable to infer that it is undetectable, merely on the assumption that it is variant? It would be easy to offer an argument from authority. Feynman himself writes the laws of Newton are of the same form in a moving system as in a stationary system, and therefore it is impossible to tell, by making mechanical experiments, whether the system is moving or not.18 And Earman writes that because Newtons laws of motion and gravitation have (Gal) as their dynamical symmetries, no feature of the lawlike behavior of gravitating bodies can be used to distinguish an absolute frame: in this sense, absolute space is unobservable.19 It is clear from the respective contexts that both authors are moving from a premise about the symmetries of NG to the conclusion that absolute velocity is undetectable. And yet even though the inference is commonplace, it is surprisingly rare to see any argument for itindeed, neither Feynman nor Earman say anything to justify it. So is a justication available? As it turns out, there is a reasonably well-known argument that does the trick.20 At a rst approximation it goes like this. Presumably a necessary condition on absolute velocity being detectable is this: that there is some physically possible process that, when initiated to measure the absolute velocity of a given body at t0 , will generate a reading at t1 an image on a computer screen, say, or the position of a needlethat indicates what that bodys velocity was at t0 . Moreover, the outcome that would be produced if the body were traveling at one absolute velocity at t0 must be discernibly different from the outcome that would be produced if it had a different absolute velocity at t0 , on pain of our not being able to tell what
who take the fact that diffeomorphisms are symmetries of General Relativity to suggest that There are no things. Structure is all there is (p. 130). For it is not enough to note that individuals points of the manifold are variant features and declare that they are therefore not real. That would be analogous to Newton declaring that there is no such thing as absolute velocity without a genuine alternative theory in hand, a move that we would rightly have regarded with suspicion. To motivate structuralism, one must present a clear theory of the fundamental structure of the material world without making reference to regions of the manifold, a theory that does well on other theoretical virtues such as simplicity, elegance, and so on. But contemporary structuralists have not presented such a theory. I say more about this, and make a start at presenting such a theory, in Dasgupta [6]. 18 Feynman [11], p. 15. 19 Earman [10], p. 48. 20 Roberts presents this kind of argument beautifully in his [19]. David Albert and Tim Maudlin have both presented variations of this argument in presentations, though not (as far as I know) in print. The presentation here follows my presentation in Dasgupta [7].

14

absolute velocity a given outcome indicates.21 So, if we simply wanted to measure whether a given body was in a state of absolute rest or absolute motion, the process would need to produce one outcome if the body was at rest at t0 for example an inscription of At restand a discernibly different outcome if the body was moving t0 an inscription of Moving, say. Finally, since the process is a physical process, the outcome produced will depend on the physical laws governing it. Putting this all together we can say that absolute velocity is detectable only if there is a physically possible device which at a given time t0 has two properties: i. according to the laws, it will produce one outcome at a later time t1 iff it was presented with a body at rest at t0 ; and ii. according to the laws, it will produce a discernibly different outcome at t1 iff it was presented with a body that was moving at t0 . But if the laws of NG are true and complete we can argue that it is physically impossible for a device to have both properties. For suppose I take a device with the rst property and present it with a body at rest at t0 , and it gives some outcome. For the sake of deniteness one can imagine that it displays At Rest on a screen at t1 . We can show that it does not have the second property by considering a world W just like ours with the one exception that it has been subjected to a uniform velocity boost, say, 5 mph to the north. Now we know three things about W. First, it is a world in which the device is presented with a moving body at t0 (rather than a body at rest). Second, we know (by construction) that the relative positions of all bodies at all times are the same in W as they actually are. So, if the device actually displayed At rest at t1 , it displays At rest at t1 in this boosted world too. More generally, it appears that a world differing only in an absolute velocity boost and thereby agreeing with the actual world in respect of the relative positions of all material bodies at all times would be indiscernible from the actual world. So no matter what outcome the device gives in the actual world, we know that it gives an indiscernible outcome in the boosted world. But because boosts are symmetries of NG, the third thing we know about W is that it is a world in which the laws of NG obtain. So the behavior of the device in this boosted world represents how it behaves according to the laws of NG; hence it does not have the second property listed above. QED. That seems straightforward enough in the case of velocity. But does the argument generalize to arbitrary laws and arbitrary features?
least, that is the ideal: in practice we do not mind if the outcomes produced by velocities differing only by some tiny amount are indiscernible. More accurately, then, what we require is that the measurement outcomes are discernible when the velocities differ by more than some amount X, in which case we say that the process measures velocity up to an accuracy of X.
21 At

15

Well, the argument went through providing that three things were true of the boosted situation. First, that it is a situation in which the variant feature (absolute velocity) differs. This is guaranteed by what it is to be variant. Second, that it is a situation in which the laws obtain. That much is guaranteed by the fact that symmetries preserve the truth of the laws. But the third thing that needed to be true of W is that it is (in some intuitive sense to be made clear) observationally equivalent to the original world, that at least to the naked eye it looks and feels and smells exactly like the original world. This is what we were trying to show when we pointed out that W agrees on all facts about the relative positions of things and (therefore) on what is displayed on a computer screen. So what we need, if the argument is to generalize to arbitrary laws and arbitrary features, is: Given any set of laws, any two systems related by a symmetry of those laws are observationally equivalent. Is this true? It is not clear. Indeed we do not know enough about what symmetry means to even start answering the question. I said at the beginning that a transformation is a symmetry of L only if it preserves the truth of L, but I also said that this is not a sufcient condition. So we are not in a position to begin asking whether the indented claim is true. The obvious strategy now is to reverse engineer. We need to ask what symmetry could possibly mean such that this claim is true.

4
4.1

The Meaning of Symmetry

The Framework

There is little agreement about the meaning of symmetry. Perhaps there is nothing wrong with this: there may be many related notions each playing a different theoretical role. Our task is just to nd a notion that makes sense of the symmetry-to-reality inference. We need to start by agreeing on a framework in which to talk about symmetry. What kind of thing is a symmetry? I said that a symmetry is a transformation on physical systems. But what is a physical system? A possible world? Part of a possible world? A set-theoretic model? There is a danger in using possible worlds or parts thereof. For suppose we think of symmetries as functions from possible worlds to possible worlds. And suppose it turns out that, necessarily, spacetime has a Galilean structure in which there is no such thing as absolute velocity. Then there are no (non-trivial) uniform velocity boosts since there are no worlds that differ only in a uniform velocity boosts! But surely even if that metaphysics of spacetime turns out to be true we should still be able to

16

talk about (non-trivial) uniform boosts being symmetries of NG.22 So let us work instead with set-theoretic structures that are used to represent (possible or impossible) physical systems. Now, what a given set-theoretic structure represents is not intrinsic to the structure but is due in part to how we choose interpret it (models, like words, are our tools not our masters). So, more precisely, let us work with a domain of structures and an interpretation of each structure, where this latter is a specication of what each structure represents. Then even if there is no such thing as absolute velocityeven if there is necessarily no such thingwe can still talk of a set of structures each of which represents a physical situation in which particles move through Newtonian space with well-dened absolute velocities. It is just that, so interpreted, the structures may represent incorrectly (they may even represent metaphysical impossibilities). I said that a symmetry is a transformation on physical systems, so if we represent physical systems with structures we can represent a transformation with a function from structures to structures. But we must distinguish transformations that are taken to be generated by a recipe, from transformations that are not. To illustrate, consider the vector space V of velocities, consider some velocity v in V, and consider the function f v on V that maps each vector v* to the vector v*+v. Then f v naturally induces a function Fv on structures, where Fv is dened to map each structure s to a structure Fv (s) in which the velocity of each thing is boosted by f v . The function Fv , then, is a uniform velocity boost. But it is just a function, i.e. a set of ordered pairs of structures, with no explicit mention of how it was generated. To mark this I will say that Fv is a bare transformation. In contrast, I will say that the pair < Fv , f v > is a generated transformation. A generated transformation is a function on structures (e.g. Fv ) together with a function on a property space (e.g. f v ) that species how the function on structures is generated. We must now choose whether to work with bare or generated transformations. I choose the latter. Their virtues will emerge as we go along. One virtue is that they allow a convenient denition of what it is for a feature to vary under a given transformation. Earlier I introduced this notion loosely, but given its central role in symmetry-to-reality reasonrelated reason to be wary of the possible worlds formulation of symmetry is that it leaves subsequent theorizing about symmetries hostage to metaphysical fortune. Consider for example a Spinozist who believes that every truth is necessarily true. Suppose this Spinozist believes this because of her theological view that all truths follow from the essence of God and that God is a necessary being. If we then dened symmetry to be a function from possible worlds to possible worlds then the Spinozist would be forced to say that there are no (non-trivial) symmetries of any theory since the only such function is identity! Yet surely this Spinozist could reasonably do physics and calculate the various symmetry groups of different theories, for these issues (intuitively) have nothing to do with her (perhaps quirky) religious views. See Dasgupta [6] for more on the signicance of this example.
22 A

17

ing we need to be more precise. First, distinguish the determinable feature (e.g. absolute velocity) from its determinate values (e.g. 1 kph to the north). For simplicity let us think of a determinable as any set of perfectly determinate properties. Then the rough idea is that a transformation varies a given determinable if the transformation is generated by a (non-trivial) bijection on the determinables determinates. More precisely: a determinable D varies under a generated transformation <T , t> iff t is a (non-trivial) bijection on D and T is the function on structures induced by t. Thus absolute velocity varies under the above generated transformation < Fv , f v >, because f v is a non-trivial bijection on the determinate velocities. To say that a determinable D varies under a transformation <T , t>, then, is not just to say that T alters some of Ds values in some physical systems. It is rather to say (roughly speaking) that T alters all of Ds values in all physical systems, and does so uniformly (i.e. as given by t). Our question, then, is: Which generated transformations count as symmetries of a given set of laws? There are (broadly speaking) three approaches to dening symmetry: formal approaches dene it in purely formal, set-theoretic terms; ontic approaches dene it with reference to certain privileged physical features, and epistemic approaches dene it in epistemic terms. Most denitions in the literature are either formal or ontic, but I will argue that only an epistemic denition can make sense of the symmetry-to-reality inference.

4.2

Formal Denitions

Say that a generated transformation <T , t> preserves a law L iff for any structure s, if s is a model of L then so is T (s). Then we know that a necessary condition on a transformation being a symmetry of L is that it preserves L. Suppose that we now say that this is also a sufcient condition. This would then be a purely formal denition. But this denition is inadequate, for it does not guarantee what is needed to underwrite the argument (from section 3) that variant features are undetectable. For that argument to go through it must be that given any set of laws, any two systems related by a symmetry of those laws are observationally equivalent. But there are many transformations that preserve a given set of laws but do not map systems to observational equivalents. To see this, pick two models of NG, m and m*, that differ in any way you likefor example, let m contain a single particle at rest and let m* contain 10 particles in motion. And consider the function F on structures that maps m to m* and vice-versa and which is identity on all other structures. This can be generated by considering the determinate property M that the entire system m instantiates (i.e. the property of containing a single particle at rest), and the corresponding determinate property M* that the entire system m* instantiates. The set {M, M*} is then a determinable property, 18

so if we let f be the only non-trivial bijection on it then we have the generated transformation F = < F, f >. Now F counts as a symmetry of NG according to the above denition, since (by construction) it preserves NG. But it clearly does not map systems to observationally equivalent systems. Put slightly differently, the determinable property of containing a single particle at rest or containing 10 particles in gravitational motion varies under this transformation F, so if F counts as a symmetry then according to the reasoning outlined in section 3 we should be able to infer that whether a system contains a single particle at rest or 10 particles in gravitational motion is undetectable; which is clearly false. The problem is that the reasoning in section 3 assumed that symmetries map systems to observationally equivalent systems, which F does not.23 If one worries that it was articial to consider a determinable with just two determinates, one can instead consider the set WP of all world properties, i.e. the set that contains M and M* and indeed all the other determinate properties that an entire physical system could instantiate. Each property in WP is a fully determinate way for an entire physical system to be. Now let WP NG be the subset of WP of world properties that are had by a system in which NG is true. Then there are obviously a host of non-trivial bijections g from WP to itself that map M to M* and vice-versa, and map elements of WP NG to elements of WP NG . The generated function <G, g> will then count as a symmetry of NG on the current denition, but will map m to m* and vice versa and so will not always map structures to observationally equivalent structures. One might try to rule out these examples by insisting that a symmetry be generated from a function on a set of fundamental properties, the thought being that the world properties in WP are not fundamental. But the trouble is that they may turn out to be fundamentalindeed that is the view of the monist who thinks that (fundamentally speaking) there is just one object, the cosmos, that has one of the properties in WP!24 One might instead try to avoid the problem with a different kind of formal denition. The idea is to think of symmetries as transformations on (representations of) instantaneous physical states rather than fourdimensional physical systems. Instead of requiring that the transformation
[3] gave a similar argument against this denition of symmetry. His argument was more straightforward because he was discussing the use of symmetries that simply identies structures related by a symmetry; since m and m* should clearly not be identied it follows that F cannot count as a symmetry. But the symmetry-to-reality reasoning I am discussing is more complicated than his, proceeding (as it does) via lemmas about variant features being undetectable. So the reason why F is a counter-example to the current denition of symmetry is a little more involved. 24 See Schaffer [20] for a contemporary defense of monism. To be clear, the point here does not depend on the truth of monism. The point is rather that the monist may legitimately wish to use the symmetry-to-reality inference to motivate various metaphysical theses. So an adequate denition of symmetry should be acceptable to the monist.
23 Belot

19

preserve the laws, we would require that it commutes with the laws. This means that given any initial state X, the state produced by rst evolving X in accordance with the laws and then transforming it is always the same as the state produced by rst transforming X and then evolving in accordance with the laws. Let us also require that it be a bijection on states. The resulting denition of symmetry has a distinguished history dating back at least to Wigner and is used by contemporary authors such as Baker.25 And it seems to avoid the kind objection raised above, for a function like F that is identity on all states but two will not in general commute with the laws and so will not in general count as a symmetry. Well, perhaps not in general, but we can arguably nd problematic cases. Consider an NG system A in which a single particle is at rest, and a second NG system B in which two mass-less particles are at rest some distance from one another. These are both static systems, so let S A be the state that system A is in at all times, and SB be the state that system B is in at all times. Now consider the function G that maps S A to SB and viceversa and is identity on all other states. This is a bijection that commutes with NG, but it is not a symmetry of NG. And the problem is not solved by requiring that a symmetry is generated from bijections on determinate properties. For G can be induced from a bijection on the determinate properties of entire states, in much the same way that F was induced by a bijection on the determinate properties of entire worlds.26 A formal denition of symmetry must x these problems using just the formal materials of set theory, model theory, and so on. I have not argued that this is impossible (in particular, the Wignerian approach can be developed in ways that I have not discussed here). But our notion of symmetry must imply that given any set of laws, any two systems related by a symmetry of those laws will be observationally equivalent. And it is (to put it mildly) extremely hard to see how any purely formal denition could have this consequence. Absent some reason to think that symmetry must be dened formally, then, let us put this approach aside.
Wigner [23] and Baker [1]. one does not recognize the possibility of mass-less particles, one could instead consider a physical theory in which particles have charges governed by the coulomb law. One can then let the particles in system B be massive, but let them also have charges that repel each other to exactly the same extent that their masses attract one another, so that they are at rest. Then, once again, g is a bijection that commutes with the laws of the theory, but is not a symmetry of the theory. One might of course patch up the proposed denition by requiring that a symmetry be a smooth transformation on states. But this is to introduce the notion of smoothness into the denition, which is not purely formal. The resulting denition is therefore ontic, and I will discuss ontic denitions shortly.
26 If 25 See

20

4.3

Ontic Denitions

How should we proceed? Think back to when we argued (briey) that two situations related by a uniform velocity boost are observationally equivalent. We noted that the situations would (by construction) agree on all the relative positions of things. Perhaps it was because of that that it seemed right to say that they are observationally equivalent. There is something compelling about this idea. As Bell remarked, in physics the only observations we must consider are position observations, if only the positions of instrument pointers.27 One might object that we should also consider color observations, like the color of litmus paper. But still, one might think that somethings color supervenes on the relative positions of things (its parts, the light reecting off it, etc). So the compelling idea is not that we can only observe relative positions, but that our observations supervene on relative positions. Preserve the relative positions, Bells idea is, and you have an observationally equivalent situation. The natural idea, then, is to require symmetries to preserve relative positions. More precisely, say that a determinable D is preserved by a generated transformation <T , t> iff s and T (s) agree on all the determinate values of D, for any structure s. Then the natural idea is to dene a symmetry of a law to be a (generated) transformation that preserves the law and also preserves relative position. It will then follow (assuming Bells idea) that given any set of laws, any two situations related by a symmetry of those laws will be observationally equivalent, as required. Perhaps you disagree with Bells idea. Still, you might think that there are some other determinable features F on which our observations supervene. As a shorthand, let us say that such a collection of features F x the data. Then the general idea is to dene a symmetry of a law to be a function that preserves the law and also preserves those features F. This general idea can be developed in two ways depending on what features F we pick. On one approach we allow F to include epistemic or observational features such as looking red to me or appearing from my perspective to be 2 feet. This results in an epistemic denition, so put these features aside until the next section. For now, restrict our attention to physical features like distance, mass, charge, spin, etc. (I will not try to dene physical, the idea is clear enough to work with.) The result is a denition of symmetry that requires a symmetry to preserve the laws and preserve certain privileged physical features. To mark this let us call them ontic denitions. Now an ontic denition can be developed in two ways, one de re and the other de dicto. On the de re approach we rst pick some physical features F (that we think x the data), and we then require a symmetry to preserve those features. On the de dicto approach (by contrast) we require
27 Bell

[2], p. XX. [[get page number]]

21

a symmetry to preserve physical features that x the data, whatever they happen to be. Thus the de re denition will say, of some features F, that a (generated) transformation <T , t> is a symmetry of a law L iff (i) T preserves L, and (ii) T preserves those features F. And the de dicto denition will say that T is a symmetry of L iff (i) T preserves L, and (ii) T preserves all features that x the data (whatever they are). The de re denition above is just schematic. For what are the privileged physical features F, that (by denition) are preserved by all symmetries? One might think there is an obvious answer: Bells relative positions. But on second thoughts the suggestion is still schematic. For what are relative positions? Spatial distances? Or spacetime intervals? Something else? A moments thought reveals that they cannot be spatial distances, for they are not preserved by the Lorenz symmetries of special relativity. So a proponent of the de re approach must propose some privileged physical features F. One might follow Earman and suggest that a transformation <T , t> counts as a symmetry of a law L iff T preserves L and t is a diffeomorphism on spacetime. This implies that T will always preserve features like smoothness and differentiability, so that (for example) if a given system s contains particles arranged in a smooth and differentiable line, T (s) also contains particles arranged in a smooth and differentiable line. So these features like smoothness and differentiability are (in effect) Earmans suggestion of the privileged features F. The vast majority of denitions of symmetry offered in the literature are, I believe, ontic denitions of one kind or another. Indeed this de re ontic denition of Earmans has been enormously popular. For example, Roberts [19] uses it as his denition of symmetry, and Belot [3] also takes it to be one of the standard denitions (he calls it the notion of a classical space-time symmetry). And even when Belot considers other denitions, they all turn out to require a symmetry to preserve various privileged physical features and so they all count as ontic denitions.28 Are ontic denitions adequate? One might complaint that Earmans denition is not sufciently general: it will only count as symmetries those functions that can be generated by a diffeomorphism on spacetime. This gives us symmetries like boosts and shifts, but it cannot give us the gauge symmetries of electodynamics. And on his denition one cannot even ask
28 Why

has Earmans denition been so popular? It might be due to Einsteins idea that

All our space-time verications invariably amount to a determination of space-time coincidences. If, for example, events consisted merely in the motion of material points, then ultimately nothing would be observable but the meeting of two or more of these points. . . coincidences between the hands of a clock and the points on a clock dial, the observe point-events happening at the same place at the same time. (Einstein [8], p. 177) Since diffeomorphisms preserve all such coincidences, the Earman denition of symmetry will, if Einstein is right, map physical systems onto observationally equivalent systems.

22

whether (say) uniform multiplications of mass (perhaps along with a corresponding multiplication of the gravitational constant) are symmetries of NG, since (again) this transformation cannot be generated from diffeomorphisms on space-time. But this complaint is not weighty, since it is not hard to generalize his denition to cover these cases. A better objection comes from Belot [3], who argues that ontic denitions are not extensionally adequate: they count as symmetries transformations that vary features that we would not want to consider unreal. But I believe that there is a more systematic problem with ontic denitions regardless of whether they are extensionally adequate. The problem is one of inferential circularity. In order to perform a symmetry-to-reality inference, I need to take what I believe to be physical laws and work out what their symmetries are. But according to an ontic denition of symmetry, in order to check whether a given transformation <T , t> counts as a symmetry of those laws, I rst need to know which physical features x the data so that I can check whether T preserves them.29 And the problem is that in many cases we discover which physical features x the data by engaging in symmetry-to-reality reasoning. Thus on the ontic denition it is hard to see how a symmetry-to-reality inference can ever get going. The objection rests on the claim that, in many cases, we discover which physical features x the data by engaging in symmetry reasoning. To see why this is plausible, imagine asking someone which physical features they think would need to be preserved in order to preserve the data. 150 years ago we would have all said that spatial distance is one such feature: if two physical systems differ (enough) with respect to the spatial distance between things, we will notice the difference. But it is crucial that this is not counted as something that xes the data, else Lorenz transformations will not count as symmetries of STR! And indeed these days we do not consider it a physical feature that xes the data. But why not? The reason, I claim, is that we engaged in symmetry reasoning. We took it as a premise that the Lorenz transformations are symmetries of STR, and from there inferred that the features that vary under those transformations (like spatial distances) are not real, and that (therefore) what we really see when we think we are seeing spatial distance is some other quantity like spacetime interval, or spatial distance relative to a frame, or what have you. In this way our belief as to whether a given feature (like spatial distance) xes the data is based on prior beliefs as to whether the feature is real, which in turn is based on prior beliefs about what the symmetries
29 On the de dicto approach this is just because I will not know which functions count as symmetries until I know which physical features x the data, while on the de re approach this is because I will not know what the denition of symmetry is until I know which features x the data. But either way, it follows from any ontic denition that in order to work out what the symmetries of a given set of laws are I rst need to know which physical features x the data.

23

of the physical laws are. So we cannot (on pain of inferential circularity) dene symmetries to be transformations that preserve features that x the data (in either the de re or the de dicto sense). To be clear, the objection is not that these ontic denitions are denitionally circular. It is just that they lead to an inferential circularity. Put otherwise, the objection is that they get the order of justication backwards: we often use premises about symmetries in order to work out which physical features x the data, so we cannot at the same time dene symmetries to be those operations that preserve features that x the data. Now one might reply that a de re ontic denition can avoid this circularity. For suppose that we just pick some physical feature and stipulate that a symmetry must (by denition) preserve it. Then we can easily know what the symmetries of a given law are without rst engaging in symmetry-todetection inferences. The above objection assumes that we require that a de re ontic denition appeals to features that we have reason to believe x the data: if we relax this requirement, the objection fails. Perhaps this was what Earman intended when he offered his de re ontic denition that requires symmetries to preserve differentiable structure: perhaps there was no thought to the effect that differentiable structure xes the data, perhaps he was just stipulating that the notion of symmetry must preserve it. But if this was his idea then the problem is that the resulting denition is objectionably arbitrary. For whatever physical feature we pick as our privileged feature F, it will follow just by virtue of the resulting denition of symmetry that it is impossible to run a symmetry-to-reality inference on F and conclude that it is unreal. This is because it will be built into the very denition of symmetry that F never varies under the symmetries of any law. Now it would clearly have been a mistake to have arbitrarily picked absolute velocity or spatial distance as such a privileged feature, for in that case uniform boosts and indeed Lorenz transformations would by denition not count as symmetries of any law. So what is different about the differentiable structure that Earman privileges, or indeed any other supposedly privileged feature F? What is so special about them, such that they are by denition immune to being rejected as unreal on the basis of the symmetry-to-reality inference? Without an answer to this question, the denition is intolerably arbitrary. And of course the obvious thing to say is that those privileged features singled out by a de re ontic dention are privileged precisely because they x the data. That is why they are immune to Symmetry-to-Reality reasoning, and why it is reasonable build those features right into the very denition of symmetry. But of course if we say this then we are back with the problem of inferential circularity. A similar problem of inferential circularity arises for Healeys denition of symmetries as those functions that preserve the laws and also

24

preserve the value of all measurable quantities.30 By measurable here he means what I meant by detectable above, i.e. not just directly observable features but also those we can detect with the help of measuring instruments. We may grant that Healeys denition is extensionally adequate, but it is not inferentially adequate for our purposes. For in the symmetryto-reality inference I have in mind we use symmetry considerations to work out what is measurable, so we must be able to work out what the symmetries of a given set of laws are prior to working out which quantities are measurable, and so we cannot dene the notion of symmetry in terms of measurability on pain of inferential circularity. The problem of inferential circularity also arises for a nal ontic denition that might be worth mentioning. One might try dening the symmetries of some laws to be those transformations that preserve the laws and which map structures to physically equivalent structures, where structures are physically equivalent if they are equally well suited to represent any given physical system. But the trouble with this should be obvious, for two structures that differ (say) in a uniform velocity boost will be equally well or ill suited only if there is no such thing as absolute velocity. If there is such a thing as absolute velocity and one of the structures represents a given physical system correctly then the other will represent it incorrectly. So in order to work out whether uniform velocity boosts are symmetries, on this denition, we rst need to work out whether there is such a thing as absolute velocity. Which (once again) gets the line of reasoning in the symmetry-to-reality inference precisely backwards. Most denitions of symmetry I have seen are either formal or ontic. If so, then the discussion here reveals an irony in the literature: the symmetry-to-reality inference is widely used, and yet the standard denitions of symmetry do not legitimate it. If there is anything going for the inference, there must be something implicit in the meaning of symmetry that is ignored by these denitions.

5
5.1

Epistemic Denitions
Taking observation seriously

The implicit ingredient, I think, is that symmetries must preserve the appearancesin some sense to be made clear. If we build this into our denition of symmetry it will follow by denition that given any set of laws, any two situations related by a symmetry of those laws are observationally equivalent. Which (remember) is what we need in order to underwrite the argument (from section 3) that variant features are undetectable. This is the epistemic approach to dening symmetry.
30 See

Healey [12].

25

Epistemic denitions can be developed in a number of ways. One might for example take as primitive a relation of observational equivalence between structures. Suppose we are in a position to know whether the relation holds between two structures without knowing anything about the underlying metaphysics of our world. Then one might simply say that a transformation <T , t> is a symmetry of L iff (i) T preserves L, and (ii) T maps structures to observationally equivalent structures. Still, this is not very satisfying. Rather than being a primitive relation, it is natural to think that the relation of observational equivalence holds between two structures in virtue of their intrinsic properties. So a better denition would identify the intrinsic properties that make for observational equivalence, and then dene a symmetry to be a transformation that (in addition to preserving the laws) preserves them. Of course, we should not fall into the trap of identifying some physical feature and say that structures are observationally equivalent when they agree on those. For this is what the ontic denitions did, and they fell to the problem of inferential circularity. Instead, observational equivalence must be dened in epistemic terms that do not depend on the underlying metaphysics, such that we can be in a position to know whether two structures are observationally equivalent prior to knowing anything (via symmetry-to-reality reasoning) about the metaphysics of our world. Did Ismael and van Frassen propose an epistemic denition when they focused on transformations that (i) preserve the laws, and (ii) preserve all qualitative features of every model?31 It depends on what they mean by a qualitative feature. They say that qualitative aspects of a physical situation correspond in our terminology to parameters which characterize that situation, and are directly accessible to us through perception.32 If this is an attempt to delineate some privileged physical features that are directly accessible through perception, then the approach is ontic and falls afoul of the problem of inferential circularity. For we typically come to know which physical features are directly accessible by engaging in symmetry-to-reality reasoning (150 years ago we might have said that spatial distance was directly accessible, but of course today we would not). So we must take care that in specifying the intrinsic properties that make for observational equivalence we do not appeal to some privileged physical structure. How this should be done depends on delicate issues in epistemology and the philosophy of perception. Here is not the place to decide those issues, but let me roughly outline two approaches to illustrate what an epistemic denition might look like.
and van Frassen [15], p. 379; my emphasis. I should emphasize that they did not present this as a denition of symmetry, their ofcial denition of the term was purely formal. But this is the concept that they claim is at work in the symmetry-to-reality inference, so it is their denition of symmetry as I am understanding the term. 32 Ismael and van Frassen [15], p. 375; my emphasis.
31 Ismael

26

5.2

How things look

Observationally equivalent situations look and feel and taste and smell and sound exactly the same. Suppose we take this seriously. Suppose that each physical system has a way it looks (where looks is now used a to express how things are presented to us across all sensory modalities). Then we might say that a transformation <T , t> is a symmetry of L iff (i) T preserves L, and (ii) T preserves the way things look.33 This is an epistemic denition. This ts well with our paradigm examples of symmetryrigid translations, uniform boosts, and so onwhich all map situations to situations that look the same. If your pasta sauce tastes garlicy, it would still taste garlicy if everything had been shifted 3 feet over to the right. Moreover, we know this independently of knowing the underlying metaphysics of the world, e.g. whether we live in a Newtonian or a Minkowski world. So this denition arguably avoids the problem of inferential circularity. This basic idea can be developed in a number of ways. One question is whether we should work with an absolute or relative notion of how things look. Things look different from different perspectives, suggesting that we should work with the notion of how things look relative to a given perspective, where a perspective is (perhaps) a location and an angle. But one might in principle try working with an absolute notion of how a given system looks, i.e. how it looks from nowhere. Suppose we work with the relative notion of how things look. Then what does it mean to preserve the way that things look from a given perspective? Consider a uniform shift of all matter 3 feet to the right. Given a perspective P, things will not look the same from P if everything were shifted over since everything would look further over to the right! For everything to look the same we must also shift the perspective 3 feet over to the right. In general, then, when asking whether a given transformation <T , t> preserves the way things look we must transform the perspective by t. (Here then is another reason to work with generated, rather than bare, transformations.) It may be that this idea of transforming the perspective by t is not well-dened for some generated transformations (the transformation F discussed in section 4.2 may be an example). But in that case we can stipulate that a generated function only counts as a symmetry if modifying a perspective by t is well-dened. This is perhaps the most obvious way of developing an epistemic definition, but it raises a number of questions. If I occupy a perspective then (given my poor eyes) things will look a good deal more blurry than if my eldest daughter occupied it. So perhaps we should work with the notion of how things look from a given perspective to a given person. But then relative to whom should our notion of symmetry be dened? Second, it
33 Perhaps

this is along the lines of what Ismael and van Frassen [15] had in mind.

27

is not clear that there is a fact of the matter as to how things look (to anyone) from various perspectives. Most perspectives are unoccupied, so the notion is presumably (something like) how things would look to someone were they to occupy it. But consider a very high energy system that a human body could not survive. Is there a fact of the matter as to how things would look to me in that situation? Perhaps these questions have reasonable answers. But let me outline another approach that might be less controversial.

5.3

Observation sentences

This other approach uses the venerable notion of an observation sentence. Suppose we run a trolley down a slope and measure its progress. We might record the result of the experiment by writing (1) The trolley came to a halt 20 seconds after being released. (2) The trolley travelled 2 meters. In recording these results we do not mean to commit ourselves to any particular view about the underlying metaphysics of space-time. Regardless of whether we live in a Newtonian space or a Galilean space-time or a Minkowski space-time, we want to leave open that these observation reports are correct. So these sentences must have correctness-conditions that place few constraints on the underlying metaphysics. If we live in a Newtonian space, (2) will be correct iff the spatial distance between the start point and the end point of the trolleys trajectory is 2 meters. If we live in a Minkowski space-time world, (2) will be correct relative to an inertial frame X iff the trolley travelled 2 meters relative to X.34 And so on. Are these correctness-conditions truth-conditions? That depends on broader issues in semantics. If one says that they are, then an utterance of (2) can be true even if uttered in a Minkowski world, though it would admittedly not perspicuously state the truth. But one might insist that (2) wears its truth-conditions on its sleeves. On this view (2) is true iff two points are a certain spatial distance apart, and so is false if there no such distances (as is the case in a Minkowski space-time).35 Still, there is no need to choose between these options. For even if one insists that utterances of (2) are false in a Minkowski world, one can still distinguish the good utterances of (2) that satisfy the above correctness conditions and the bad ones that do not. Which is all we need.
bears unpacking, so that it is clear what it is for x to be 2 meters from y relative to a frame. But the details do not matter for our purposes. 35 See Boghossian [4] for a defense of this approach to the semantics of location and motion in a relativistic world.
34 This

28

Now suppose one takes observation sentences like these to constitute our data. Then (roughly speaking) two structures in which the same observation sentences are correct are observationally equivalent. And so we might dene a symmetry to be a transformation that preserves the laws and also preserves which observation sentences are correct. The idea needs a little tweaking. Whether an utterance of (2) is correct when uttered in a Minkowski world depends on the frame of reference in which the sentence is uttered. So instead of asking which observation sentences are correct in a given structure, we should ask which observation sentences are correct in a given structure relative to a given index (e.g. a frame of reference or what have you). So, what the correctness-conditions determine is not a mapping from structures to sentences that are correct in that structure, but rather a mapping that takes as input a structure and an index and gives as output the set of sentences that are correct in that structure relative to that index. Let us call this mapping m. The indices are functioning like perspectives from the last proposal. So when we say that a symmetry must preserve which observation sentences are correct, we mean that given any structure s and index i, the result of transforming s and also transforming i similarly must result in no change to which observation sentences are correct. More precisely, then, let us say that a transformation <T , t> preserves the observation sentences iff for any structure s and any index i, m(s, i) = m(T (s), t(i)). (Here yet again we see the virtue of working with generated transformations.)36 So we might then say that a transformation <T , t> is a symmetry of L iff (i) T preserves L, and (ii) <T , t> preserves the observation sentences. This epistemic denition arguably avoids the problem of inferential circularity. In order to work out whether a transformation is a symmetry of a law, one does not need to rst work out (with symmetry-to-reality reasoning) what the underlying metaphysics of our world is, e.g. whether we live in a Newtonian or Galilean or Minkowski world. Instead, the idea would be that ones grasp of observation sentences like (1) and (2) allows one to determine its correctness-conditions in each kind of world, including what kinds of indices correctness is relativized to. These correctness-conditions determine the function m, and with m in hand it is then an apriori matter whether the transformation is a symmetry of the law. Moreover this denition arguably avoids the problems that faced the earlier epistemic denition in terms of how things look. Whether (1) and (2) are correct (relative to a given index) does not depend on which subject occupies the perspective of that index or whether she needs glasses. And the correctness conditions of sentences like (1) and (2) may be well dened even in situations in which there is no fact of the matter how things would look to a given subject (e.g. situations in which I would not survive).
36 As

before, it may not be the case that t(i) is well-dened for all transformations.

29

Still, the denition uses the notion of an observation sentence, a notion that some are loath to recognize. But it is worth pointing out that the current denition is compatible with Sobers deated notion of an observation sentence, on which a sentence counts as being observational only relative to a theoretical problem.37 Sobers notion might be acceptable even to those who do not wish to draw an absolute distinction between observational and theoretical sentences. In any event, this is not the place to mount a full defense of the notion of an observation sentence. The project of identifying the best denition of observational equivalence, and hence the best epistemic denition of symmetry, clearly involves deep issues in epistemology and the philosophy of perceptionissues that would take us too far aeld right now. So (to be clear) I do not claim that either of these denitions is correct, I outlined them just to illustrate what an epistemic denition might look like.

Symmetry as an Epistemic Notion (Twice Over)

Suppose then that some epistemic denition like this can be made out. The resulting epistemic reconstruction of symmetry-to-reality reasoning would go like this. We start by asking whether a given determinable property (e.g. absolute velocity) is real. To answer this, we ask whether it varies under the symmetries of the laws. Suppose it does. Then what this means (on an epistemic denition) is that there is a systematic way of altering its determinate values such that given any physically possible system, the result of altering the determinate values like that results in an observationally equivalent system in which the very same laws obtain. It then follows (by the reasoning outlined in section 3) that the feature is undetectable: there is no physically possible process by which we might discover which determinate values are actually instantiated. So, if there is an alternative theory that does without the feature and that scores well enough on other theoretical virtues, the Occamist norm tells us to believe it. Let me end by discussing the approach further and defending it from objections.

6.1

The Occamist Norm revisted

I discussed the Occamist norm earlier in section 3, but what I said there needs a little qualication. To see thisand to further understand the mechanics of the epistemic approachconsider the following case of David Bakers (personal communication). Imagine that our best theory says that there are two kinds of particles, Aons and Bons. They behave identically
37 Sober

[22].

30

except under very specic conditions C that have been created in laboratories, so we can detect whether something is an Aon or a Bon only by creating those conditions C. Now consider a transformation that changes all Aons that never enter conditions C into Bons and all Bons that never enter conditions C into Aons. By construction this transformation preserves the laws (since Aons and Bons behave identically outside conditions C) and maps models to observationally equivalent models. So it is a symmetry of the laws, according to the epistemic denition. What metaphysical conclusion does our symmetry-to-reality reasoning then recommend? If our method was to identify models related by a symmetry transformation (or, more precisely, say that they represent the same physical system) then we should conclude that there is no difference between a system in which an Aon never enters conditions C and a system in which a Bon never enters conditions C. This would be an unwelcome result.38 But this is not the method of the epistemic approach defended here. According to the epistemic approach, we must rst identify which features vary under the symmetries, and only then can we conclude that they are undetectable and therefore (all else being equal) unreal. So, which features vary under Bakers symmetry? Not being an Aon or being a Bon. To see this, consider the determinable that consists of the following determinates: {being an Aon, being a Bon}. And let t be the non-trivial bijection on this set. Then Bakers symmetry is not <T , t>, where T is the function on structures induced by t. Indeed <T , t> is not a symmetry, because it switches Aons and Bons that are in conditions C and hence does not preserve the laws (the laws state that Aons and Bons behave differently in conditions C). So this determinable does not vary under the symmetries of the laws, so our symmetry-toreality reasoning cannot be used to establish that whether a particle is an Aon or a Bon is undetectable or that these properties are unreal. To nd Bakers symmetry, we need to take the following determinable D: {being an Aon not in C, being a Bon not in C}. If we let the non-trivial bijection on this set be t* then Bakers symmetry is <T *, t*>, where T * is the function on structures induced by t*. So D is variant under the symmetries of the laws, and so (by the reasoning in section 3) we can conclude that it is undetectable. But this is the right result, since it is in effect dened to be undetectable: it is the determinable property of being an Aon or a Bon in
is not uncommon to see this methodof identifying models related by a symmetry transformationin the literature. For example it is the method employed by Ismael and van Frassen [15] on p. 380, when they appear to endorse always identifying models related by a symmetry transformation (in my sense of the term symmetry). So Bakers case would seem to be a problem for their view. They discuss a similar case on pp. 384-85 and respond by appealing to a principle of recombination that would prevent identifying the models. It is not obvious how this squares with what they say on p. 380, but I will not pursue this exegetical point here.
38 It

31

conditions where one cannot detect which! Still, we do not want to say that D or its determinates are unreal. Does our Occamist norm imply that we must say this? Not exactly: it just says (as I put it earlier) that positing undetectable structure is an epistemic vice, in the sense that all else being (near enough) equal we should prefer theories that do not posit such structure. So the norm does imply that it is an epistemic vice to posit D, but it only says that we should think that D is unreal if we have an alternative theory that does not posit D and is (near enough) equal on other epistemic virtues. But it might be objected that it is no epistemic vice at all to posit D. After all, in almost every theory we can cook up a determinable that is dened to be undetectable in much the same way that D is! So unless we want to accuse every theory of having this epistemic vice of positing undetectable structure, we need to qualify our Occamist norm. Why then might it be a vice to posit undetectable structure like absolute velocity, but not a vice to posit undetectable structure like D? Well, D is dened in terms of more basic properties: being an Aon, being a Bon, and not being in C. And none of these more basic properties are variant or undetectable in Bakers example. In contrast, velocity is either a fundamental property, or else is dened in terms of cross-temporal spatial distance, which is itself variant and undetectable (in NG). Either way, with velocity we have an undetectable property at the bottom level, whereas with D we only get an undetectable property by stitching together detectable properties. Our Occamist norm should be sensitive to this. To this end, call a property fundamentally undetectable iff (i) it is undetectable, and (ii) it is not dened in terms of detectable properties. Then instead of saying that it is a vice to posit undetectable structure, our Occamist norm should just say that it is a vice to posit fundamentally undetectable structure. Indeed this revised norm is what we should have endorsed all along. For my argument for the Occamist norm was (very roughly) that discovering that something is undetectable is discovering that we have no empirical reason to think that it is real. But on reection this is only true of fundamentally undetectable structure. We have plenty of empirical reason to think that (merely) undetectable features like D are real, for they are dened in terms of properties that are detectable and hence which we have empirical reason to think are real.

6.2

Observational Equivalence?

Perhaps the most controversial aspect of the epistemic approach is its appeal to the notion of observational equivalence. Is there such a notion that the epistemic approach can appeal to? One might deny that there is, for one of two reasons. One might deny that there is any intelligible notion of observational equivalence. Or one 32

might concede that the notion is intelligible but deny that it is suitable for the epistemic approach. The rst position is difcult to maintain. The argument that there must be some intelligible notion of observational equivalence proceeds from the familiar problem of testability.39 As Maudlin puts the problem, to be of interest, physical theories have to make contact with some sort of evidence, some grounds for taking them seriously or dismissing them. To see what he means, consider a theory about the trajectories of point particles through a (at) Minkowski space-time. How can we test the theory? It makes predictions about spacetime intervals between particles, but what observations would conrm those predictions? Something needs to be said about what observations correspond to various arrangements of particles in Minkowski spacetime, so that if the theory predicts a certain arrangement we know what observations would conrm that prediction. There are various candidate solutions. But any solution to this problem will tell us what observations correspond to each arrangement of particles in Minkowski spacetime. And so (the argument would go) this will induce a notion of observational equivalence: two arrangements will be observationally equivalent iff they correspond to the same observations. That (in very rough outline) is a reason to think that if a theory is testable then there is a notion of observational equivalence on its models. But even if this is conceded, the question remains whether there is a notion of observational equivalence that is suited to our epistemic approach, and the above argument does not establish that there is one. Indeed, Maudlins own solution to the problem of testability does not produce one. He continues the above quote as follows: in classical physics the evidential connection is made between the physical description and a certain class of local beables, such as the positions of macroscopic objects. Local beables are things like rocks and pointer needles. So, to see if a theorys predictions obtain (says Maudlin) we see if things like rocks and pointer needles are positioned in the way that is associated with that prediction. But if by position he means some privileged physical featuree.g. spatial positionthen his proposal will not do for our purposes. For our denition of symmetry will then collapse into an ontic denition and will fall to the problem of inferential circularity discussed in section 4. So the question remains as to whether there is a notion of observational equivalence that is suitable for our purposes. I have no general argument that there isI expect that the argument will have to be by example, i.e. by producing a workable notion (perhaps by developing one of the two notions of observational equivalence described in section 5). So I have not shown that there is a notion observational equivalence suitable for our purposes. My claim is instead that the legitimacy of symmetry-to-reality
39 Thanks

to Adam Caulton for bringing this line of argument to my attention.

33

reasoning depends on one. One might object quite generally to using a notion like observational equivalence in a metaphysical investigation to into which putative features of the world are real. For observation equivalence is a relative matter: two physical systems that are observationally equivalent for us may be observationally inequivalent for subjects with more rened perceptual capacities. Yet surely whether a putative physical feature is real does not depend on contingencies regarding the build of our eyes! So (the thought is) observational equivalence is unt for use in metaphysics. But this is confused. It is correct that which physical features are real does not depend on our perceptual faculties, but the epistemic approach does not deny this truism. The epistemic approach does imply that which physical features we should believe are real depends on which systems are observationally equivalent, and so depends in turn on our perceptual faculties. But this is correct. In theorizing we can do no better than use the tools we have. If other creatures have better tools they may be led to better theories, but this is hardly relevant to what we should believe.

6.3

Consequences

The epistemic approach has a number of consequences regarding the use of symmetries. Let me mention just two. One consequence is that symmetry-to-reality reasoning is more involved than one might have thought (or hoped). To see this, suppose that symmetry were given a formal or a de re ontic denition. Then the question of what the symmetries of a given law are can be settled by straightforward mathematical analysis. This gives the symmetry-to-reality inference a particularly analytic gloss: we just do the mathematical analysis andhey presto!we get metaphysical conclusions. It is perhaps because of this veneer of mathematical simplicity that symmetry-to-reality inferences are so popular and yet are given such little justication. But on our epistemic approach things are not so easy. To work out what the symmetries of a law are we need to work out which structures are observationally equivalent. How are we to work this out? The paradigm cases of uniform boosts and rigid shifts are deceptively simple. Experiences on trains convince us that things inside the train look the same whether at rest or in smooth motion, and the same in NYC as it did in Boston. So our reason for thinking that boosted and shifted situations are observationally equivalent is partly empirical. But in other cases the situation is far more controversial. Consider for example the question of whether uniformly multiplying everythings mass (perhaps along with a corresponding change to the gravitational constant or to spatial distances) is a symmetry of NG. Suppose we convince ourselves that it preserves the laws. To argue it is a symmetry, we need 34

to argue that structures related by such an operation are observationally equivalent. But we have nothing like the same kind of empirical evidence for that as we do in the case of uniform shifts or uniform velocity boosts! A reason to think that the situations would look the same would have to come from elsewhere, perhaps a functionalist account of consciousness. Thus on our epistemic approach, symmetry-to-reality reasoning involves not just mathematical analysis, but also considerations of perception and observation equivalence. A far cry from the purely mathematical gloss it is often given. A second consequence of the epistemic approach is that it disolves various questions that might otherwise look substantive. To take one example, consider Roberts [19], where he focuses on the inference from a features varying under the symmetries of the laws to the conclusion that it is undetectable (this is the move from (2) to (3) of the symmetry-toreality inference). He agrees that this is a reliable inference but he thinks that this is a remarkable fact in need of explanation. The central question of his paper is what explains it. Is the reliability of the inference remarkable? It is if one insists that symmetry is a purely formal or ontic notion. For in that case the inference draws a connection between the formalism of the laws and what we can detect. And such a connection would be remarkable indeed! Since Roberts works with Earmans ontic denition of symmetry, it is unsurprising that he considers the connection remarkable. But on my view the connection is hardly remarkable at all. The very notion of symmetry in play is in part an epistemic notion, with the notion of observational equivalence built into it. As I argued in section 4, if one does not work with this notion of symmetry the inference is not reliable in the rst place. So the reliable connection between symmetries and detection is not a connection between mere formalism and detection, but a connection between observational equivalence and detection. Which is hardly remarkable. At least, it is not in need of the kind of explanation that Roberts gives it.

Conclusion

I have argued that symmetry is an epistemic notion in two respects. First, the symmetry-to-reality inference proceeds by rst drawing an epistemic lemma. And second, for the inference to work the concept of symmetry must be dened partly in epistemic terms. I focused on the very simple case of boosts and shifts in NG. Even in this sanitized setting, I argued that symmetry-to-reality reasoning is epistemic in these two respects. Thus when investigating this inference in the context of more sophisticated theories, I suggest that we take this 35

epistemic approach as our starting point. This may reveal that the prospects for symmetry-to-reality reasoning are better than a reading of Belot [3] might suggest. For Belot starts out with various ontic denitions of symmetry and shows (correctly, I think) that in a number of cases they produce the wrong results. And he concludes that it appears that the sort of constraint that knowledge of the symmetries of a theory places on the range of reasonable interpretations of that theory may well be more modest than one might have hoped.40 Now if ontic denitions were in play in the paradigm, sanitized cases of boosts in NG, then Belot would have established the pessimistic claim that that sanitized reasoning does not generalize to more sophisticated settings. But if I am right then even in those sanitized settings the notion of symmetry at work is epistemic: there is simply no paradigm case of a symmetryto-reality inference in which a purely formal or ontic notion of symmetry is at work. In which case the reasoning in these sanitized settings may generalize after all. (Though if what Belot meant by the above quotation is that symmetry-to-reality reasoning is messier than one might have hoped, I agree entirely.)41

References
[1] Baker, D. 2010. Symmetry and the Metaphysics of Physics. Philosophy Compass 5 (12): 1157-1166. [2] Bell, J. S. 1993. Speakable and Unspeakable in Quantum Mechanics. Cambridge: CUP. [3] Belot, G. 2013. Symmetry and Equivalence. pp. XXXX in The Oxford Handbook of Philosophy of Physics, edited by R. Batterman. Oxford, UK: OUP. [4] Boghossian, P. 2006. What is Relativism? In Truth and Realism, edited by P. Greenough and M. Lynch. Oxford, UK: OUP. [5] Dasgupta, S. 2009. Individuals: An Essay in Revisionary Metaphysics. Philosophical Studies.
[3], p. XX. [[get page numbers from published version]] paper was a very long time in the making and went through several incarnations from 2007 to 2013. As a result I doubt that I can remember everyone who helped me along the way. I do know that David Baker and Jill North have given me invaluable and detailed comments on earlier drafts, without which this paper would be much worse than it is. Thanks to them, and thanks also to Gordon Belot, Adam Caulton, Marco Dees, Kit Fine, Richard Healey, and Robbie Hirsch for their feedbck. Some of this material was presented in September 2012 at the Structuralism in Physics and Mathematics conference at Bristol Universitythanks to the audience there for their comments. To those I have forgotten, my sincere apologies.
41 This 40 Belot

36

[6] Dasgupta, S. 2011. The Bare Necessities. Philosophical Perspectives [7] Dasgupta, S. 2013. Absolutism vs Comparativism about Quantity. Oxford Studies in Metaphysics. [8] Einstein, A. 1916. The Foundation of the General Theory of Relativity. Reprinted in English in The Principle of Relativity, 1952, edited by W. Perrett and G. B. Jeffrey. New York: Dover. [9] Earman, J. and J. Norton. 1987. What price substantivalism? The hole story. British Journal for the Philosophy of Science 38: 515525. [10] Earman, J. 1989. World Enough and Space Time. Cambridge: MIT Press. [11] Feynman, R. 1963. Lectures on Physics: Volume 1. Reading, MA: Addison-Wesley. [12] Healey, R. 2009. Perfect Symmetries. Britisih Journal for the Philosophy of Science 60: 697720. [13] Hoefer, C. 1996. The Metaphysics of Spacetime Substantivalism. Journal of Philosophy 93: 527. [14] Huggett, N. 1999. Space from Zeno to Einstein. Cambridge: MIT Press. [15] Ismael, J. and van Frassen. 2003. Symmetry as a Guide to Superuous Theoretical Structure. pp. 37192 in Symmetries in Physics: Philosophical Reections, edited by K. Brading and Elena Castellani. Cambridge, UK: CUP. [16] Ladyman, J. and D. Ross. 200X. Every Thing Must Go. [17] North, J. 2009. The Structure of Physics: A Case Study. Journal of Philosophy 106: 5788. [18] Pooley, O. Substantivalism and Haecceitism. Manuscript. [19] Roberts, J. 2008. A Puzzle About Laws, Symmetries and Measurable Quantities. British Journal for the Philosophy of Science 59: 14368. [20] Schaffer, J. 2010. Monism: The Priority of the Whole. Philosophical Review 119 (1): 3176. [21] Sklar, L. 1977. Space, Time and Space-Time. [22] Sober, E. 1990. Contrastive Empiricism. pp. 392412 in Scientic Theories, edited by C. Wade Savage. [23] Wigner, E. P. 1967. Symmetries and Reections. Bloomington, IN: Indiana University Press.

37

PHILOSOPHICAL PERSPECTIVES
Philosophical Perspectives, 25, Metaphysics, 2011

THE BARE NECESSITIES

Shamik Dasgupta
We talk of material bodies being situated in space and of events as occurring at times. And once we are taught to think four-dimensionally, we talk of bodies having trajectories through spacetime. But how seriously are we to take this apparent reference to space, time or spacetime? Very seriously, according to a view known as substantivalism. On this view, the material history of the world plays out in a container that exists independently of its material contents. Substantivalists may disagree about the structure of this container, but they all agree that there is such a thing. Is substantivalism true? This is in part an empirical question and some physical theories are considered more hospitable to substantivalism than others. So rather than ask whether substantivalism is true, many philosophers ask whether we should believe it conditional on the truth of this or that physical theory. Recent work has tended to focus on the issue as it arises in the context of the General Theory of Relativity (GTR) and in this context substantivalism is by far the received view. Against this consensus, my aim here is to argue that substantivalism in GTR faces far more serious difficulties than has been recognized and that an adequate substantival theory fit to deal with these difficulties has not yet even been formulated , let alone defended. I will end with some suggestions about how we might formulate such a theory. So my aim is not to argue that substantivalism is false but rather that its status as the received view is premature: even if some substantivalist thesis does turn out to be correct, we have not yet seen what that correct thesis is. My claim generalizes to substantivalism in the contexts of other physical theories, but in keeping with the recent literature I focus on the case of GTR. Why is substantivalism in GTR currently considered so unproblematic? I suspect it is because the objection to the view that is most often discussed is the Hole Argument, and this argument is widely agreed to fail. In particular, the Hole Argument makes assumptions about modality and determinism, and substantivalists have denied these assumptions by endorsing a modal view known as anti-haecceitism or else a certain conception of determinism (the former move

116 / Shamik Dasgupta

being the most popular). I agree that these are both reasonable responses to the Hole Argument. But the problem is that there are other serious objections to substantivalism that make no assumptions about modality or determinism at all. Defending substantivalism against these objections requires developing the view in a way that has not yet been done. The idea behind these other objections will be familiar to many readers. For in the context of classical physics it is well known that various substantival views (such as Newtons) bring with them a commitment to unobservable and physically redundant structure, and this is universally regarded as reason to reject those views in favor of a weaker substantival view that dispenses with the offending structure (such as Galilean substantivalism). I will argue that analogous objections apply to the received substantival view in GTR. This suggests that we should develop a weaker substantival view that stands to the received view just as Galilean substantivalism stands to Newtons view. My claim, then, is that the received substantival view in GTR is no better than the Newtonian view of space is in classical physics. Why have these objections that are familiar in other contexts been overlooked in the case of GTR? Perhaps because they are often presented as making modal assumptions that the substantivalist can deny by endorsing anti-haecceitism. So I will argue that when properly understood the objections make no essential appeal to modal assumptions and that even a substantivalist who endorses antihaecceitism is vulnerable to them. Indeed, one of the general morals of this paper is that modal considerations are largely irrelevant to the question of substantivalism, whether in the context of GTR or other physical theories. Admittedly, substantivalism has sometimes been characterized in modal terms. But I will argue that substantivalism (in any physical theory) is not at core a modal thesis but rather a thesis about ground . This much is often recognized in the literature. But what remains unappreciated is that if substantivalism is not a modal thesis then the best objections to it will not appeal to modal assumptions, and so modal claims like anti-haecceitism will not constitute responses to those objections. So the recent focus on modal considerations and the rush by substantivalists to endorse anti-haecceitism is, I think, something of a red herring. I have tried to write for metaphysicians new to the literature on substantivalism in GTR, and also for philosophers of physics who are unfamiliar with recent developments in metaphysics that I will make use of. The result is a longer paper than I would have liked. But since much depends on how the question of substantivalism is understood, it makes sense to build the issue from the bottom up. The paper is organized into four parts. Sections 12 argue that substantivalism is a thesis about ground. Sections 35 then discuss the Hole Argument and describe the standard substantivalist responses to it, all of which deny its assumptions about either determinism or modality. Sections 69 then show that substantivalism is vulnerable to other arguments that do not make those

The Bare Necessities / 117

assumptions about determinism and modality. Finally, Sections 1011 suggest how we might go about formulating a new kind of substantivalism in light of these other arguments.

1. Newtonian Substantivalism
What is substantivalism in GTR? It will help to abstract away from the technicalities of GTR and revisit the well known debate between Newtonian and Leibnizian views of space in classical physics. Once that issue is properly understood, it will be easy to carry it over into GTR. What then is the issue dividing the Newtonian and the Leibnizian? The Newtonian is said to think that space is a real substance: a three-dimensional, Euclidean object in which matter is situated. Against this substantivalist view, the Leibnizian thinks that reality of space as a substance is an illusion: all there really are, on this view, are bits of matter spatially related in certain ways, and any talk of space itself is to be understood in these terms.1 But what exactly is the disagreement between these two positions? Not whether space exists, for the Leibnizian may agree that space exists so long as she insists that it is somehow constructed or derived from the pattern of spatial relationships between material bodies. Instead, the disagreement is whether space is in some sense independent of the matter situated within it. This is a familiar point. As Sklar characterizes it, Newtonianism is the view that space can be said to exist and to have specified features independently of the existence of any ordinary material objects.2 But how is this notion of independence to be understood? It is often understood in modal terms. Thus, the issue is sometimes said to be whether it is possible for space to exist without matter, with the Newtonian affirming and the Leibnizian denying this possibility. Alternatively, the issue is sometimes said to be whether it is possible for the entire material history of the world to have unfolded three feet to the right of where it actually did, again with the Newtonian affirming and the Leibnizian denying the possibility.3 But this does not adequately characterize the issue. For one of the main arguments for the Newtonian position is the famous Bucket Argument, according to which inertial effects can be explained only if the motion of a material body is understood to be motion relative to space rather than relative to other material bodies. Whatever the merits of this argument, the fact that the Newtonian uses it shows that the core of her view is that there are facts about a bodys motion through space that are in some sense over and above the totality of facts about its motion relative to other bodies. But someone might agree with the Newtonian on this point and yet have other views about modality that have nothing to do with space per se but that nonetheless force her to deny the above possibilities that were said to be definitive of the Newtonian position.

118 / Shamik Dasgupta

For example, she might be a Spinozist who believes that everything true is necessarily true. The Spinozist I have in mind believes this not because of anything she believes about space but rather because she believes that all truths follow from the essence of God and that God is a necessary being.4 This theorist would clearly deny that it is possible for space to exist without matter and deny that a uniform shift of all matter is possible, simply because her (perhaps quirky) views about religion imply that nothing non-actual is possible! Still, her Spinozist views have nothing to do with the issue that divides the Newtonian and Leibnizian and so need not prevent her from being led to think by the Bucket Argument that space is independent of matter in the relevant sense. Therefore, we should not define the Newtonian position as a belief in the above possibilities. Or, to take a perhaps less extreme example, a Newtonian might also be a Quinean and deny that there is any such thing as de re possibility and necessity. On this view it makes no sense to say, of an object, that it is possibly F or necessarily G. The Quinean I have in mind believes this not because of anything she believes about space per se but rather because of general considerations about the logic and metaphysics of modality. Such a theorist would clearly not assert that uniform shifts are possible, for that would be to make a de re modal claim: it would be to say, of space, that it is possible for matter to be distributed over it in a certain way. But once again, her (perhaps outdated) views about the nature of modality have nothing to do with the nature of space per se, so she may perfectly well be convinced by the Bucket Argument that space is independent of matter in the sense that is important to the Newtonian. So a belief in the possibility of uniform shifts should not be definitive of the Newtonian position. If there is a grain of truth in the modal characterization of the views, it is that the Leibnizian view of space implies that rigid spatial shifts are impossible since the spatial relationships between bodies would be the same in the shifted scenario as they actually are, while the Newtonian view is consistent with their possibility. But since a Newtonian may have other views about modality which imply that the shifts are impossible, such as Spinozism, it is a mistake to define the issue in the modal terms described above. How then should the issue be understood? There is a growing literature in metaphysics that develops a variety of non-modal notions that may be of some use. These include ground, ontological dependence, truth-making (at least, in some of its incarnations), and Siders notion of a metaphysical semantics, amongst others.5 We might therefore understand the Leibnizian as claiming that facts about space are grounded in facts about matter, that space ontologically depends on matter, that truths about space are made true by facts about matter, or that sentences about space have a metaphysical semantics in terms of notions to do with matter. For our purposes any of these would do. But for the sake of definiteness it will be useful to pick one of them and I pick the characterization in terms of ground.6 What does ground mean? Following Fine, I use it to label a distinctively non-causal, metaphysical mode of explanation: to say that X grounds Y is just

The Bare Necessities / 119

to say that X explains Y in this metaphysical sense.7 To illustrate, imagine being asked what explains Europes being at war in 1939. A causal explanation might describe a sequence of events over the preceding 50 years that led Chamberlain to declare war on Germany. But there is another kind of explanation that would try to say what goings on in Europe at that time made it count as a continent at war in the first place. Regardless of what caused Chamberlain to declare war in 1939, someone in search of this second explanation recognizes that its being at war was not a sui generis state of the continent and that there must therefore be something about it in virtue of which it counted as being a continent at war. Here we might say that it was at war in virtue of how its citizens were acting, for example that large numbers of people mobilized and fired guns at each other. As I use the term, an explanation of this second kind is a statement of what grounds Europes being at war in 1939. I take this kind of explanation to be reasonably intuitive: regardless of the truth of this claim about what grounds Europes being at war, we seem to understand it reasonably well. Now ground is an explanatory notion, and explanations are typically expressed with the sentential operator because: Europe was at war in 1939 because its citizens were acting in certain ways. This suggests that the logical form of a claim about ground is: S because where S is a sentence, is a list of sentences, and because is read in the metaphysical rather than causal sense.8 Alternatively, one might treat ground as a relation between facts. On this approach, the logical form of a grounding claim becomes the Xs ground Y where Y is a singular variable and the Xs is a plural variable, both ranging over facts.9 Officially I endorse the former option and think of ground as a sentential operator, but it is often convenient as a shorthand to treat it as a relation between facts so I will do that here. I assume that ground is transitive and irreflexive.10 I also assume the principle of Necessitation that if the Xs ground Y, then it is metaphysically necessary that if all the Xs obtain then Y obtains too.11 This principle is standard and plausible: if it were possible for the citizens of Europe to act in a certain way in 1939 and yet for the continent to be in a state of peace, then those actions would not be the complete explanation of why the continent counted as being in a state of war. However, it is important in what follows to note that the reverse principle is not true since there can be necessary connections without grounds. For example, it is metaphysically necessary that if Obama exists then 2 + 2 = 4, but it is not the case that 2 + 2 = 4 because Obama exists. Claims of ground are therefore stronger than claims of necessary connection: the former imply the latter (by Necessitatation), but not vice-versa.

120 / Shamik Dasgupta

I suggest that we understand the issue that divides the Newtonian and the Leibnizian in terms of ground. On this approach, the Newtonian view is that facts about the spatial relationships between material bodies are grounded in facts about which regions of space they are situated in and how those regions are spatially related. For example, if asked to explain (in the metaphysical sense) why two material bodies m 1 and m 2 are 3 meters apart, the Newtonian may answer that m 1 is located at a region of space r 1 , m 2 is located at a region of space r 2 , and r 1 is 3 meters from r 2 . She need not claim that these latter facts about space are groundless. She might for example ground the fact that r 1 is 3 meters from r 2 in terms of facts about the distances between regions in between r 1 and r 2 , and perhaps so on ad infinitum. Or, if she is a Spinozist, she might ultimately ground all facts, including those about space, in the essence of God. What is important is that on her view those facts about space are not grounded back in the spatial relations between material bodies.12 I believe that it is this thesis that Sklar implicitly had in mind when he said that the Newtonian takes space to be independent of matter. In contrast, the Leibnizian reverses the order of explanation. Insofar as she recognizes facts about space she will explain them (in the metaphysical sense) in terms of facts about how material bodies are spatially related to one another. Thus, the Leibnizian takes on the challenge of accounting for inertial effects ultimately in terms of the spatial relations between bodies, and the considerations rehearsed in the Bucket Argument show that this is no easy task. As I understand them, Newtonian and Leibnizianism are each necessarily true if true at all. Thus it is more accurate to describe the Newtonian as claiming that, as a matter of necessity, the spatial relationships between bodies are grounded in facts about space; and similarly for the Leibnizian.13 This means that the Leibnizian view is not the Newtonian views negation. But discussing the corresponding contingent claims introduces needless complications, so it is convenient to discuss these two necessary claims for simplicity since the main morals carry over.14 We can now draw a distinction that will be crucial in what follows. Consider two theorists: a Leibnizian, and a Newtonian who is also a Spinozist. By the principle of Necessitation, it follows straight from Leibnizianism that that the uniform spatial shifts discussed earlier are impossible. To be sure, the Newtonian Spinozist agrees that the shifts are impossible, but only the Leibnizian thinks that this follows from the correct answer to the question of substantivalism. For the Newtonian Spinozist, the impossibility of shifts has nothing to do with the question of substantivalism per se and instead follows from entirely independent factsin this case, facts about religion and modality. To mark the distinction, I will say that the Newtonian Spinozist makes a bare modal claim.15 One of my central claims in what follows will be that in the case of GTR, subtantivalists have tended to defend their view with a bare modal claim when what is really needed is a modal claim that follows (by Necessitation) from a view about the nature of spacetime.

The Bare Necessities / 121

Admittedly, some participants to the substantivalism debate may find this talk of ground metaphysical in the pejorative sense. One source of skepticism might be my reference to abstracta like facts and grounding relations between them. But if so the skepticism is unwarranted, for as emphasized above my talk of facts is just a shorthand: grounding claims are strictly speaking nothing other than explanatory claims using the word because which involve no reference to facts at all. To my mind, that mode of explanation is just as intuitive and no more problematic than causal explanation. But this is not the place to mount a full defense of ground, and in any case I am happy for the notion to be judged by its fruits. Skeptics can therefore read this paper as asking what progress can made in the substantivalism debate if we accept the notion. If progress is made, this is reason to take the notion seriously.

2. Substantivalism in GTR
How does the issue just characterized carry over to the case of GTR? Well, the models of GTR are triples M = (M , g, T ), where M is a four-dimensional differentiable manifold of points, and g and T are a metric tensor field and a mass-energy tensor field, respectively, defined on M .16 Now these models are mathematical objects used to represent physical systems, not physical systems themselves. What then are those physical systems composed of ? This is a matter of considerable controversy, and the question of substantivalism in GTR is just one part of it. Taking the models at face value, it is natural to interpret M as representing a physical manifold of regions and g and T as representing some kind of physical object or structure distributed over the physical manifoldlet us call them the physical metric and mass-energy fields to distinguish them from their mathematical representatives. The question of substantivalism is then the question of the status of the physical manifold of regions represented by M . But as before, the interesting question is not whether such a manifold exists but rather whether it exists independently of other elements of the system. Again, this is a reasonably familiar point. As Hoefer characterizes it, the issue is whether the manifold exists independently of material things . . . and is properly described as having its own properties, over and above the properties of any material things that may occupy parts of it.17 But how is the notion of independence to be understood? As before, I propose to understand the issue in terms of ground. Thus, all parties to the debate may agree that physical systems governed by GTR contain a physical manifold of regions. The question is whether facts about that manifold are grounded in other facts about the system, or vice-versa. I take substantivalism in GTR to be the latter view. Let us spell this out a little. The issue is the status of facts about the manifold, but what sort of facts are these? We can represent them abstractly as facts of the form (r 1 , r 2 , . . .)

122 / Shamik Dasgupta

where the r i are regions of the manifold and describes them. I place very few restrictions on . In some cases will describe the manifolds intrinsic topological structure, but in other cases will describe how the physical metric and massenergy fields are distributed over the manifold. Exactly what describes in these latter cases depends on ones metaphysics of physical fields. If you think that a physical field is an instantiation of a property at each region or a pattern of relationships between the regions, then would express those properties and relations. If instead you think of a physical field as an extended object with parts, then would describe which parts of the fields are located at each region r i .18 However one understands them, these facts about the manifold play the same role for the substantivalist in GTR as facts about space (i.e. its geometry and where material bodies are located in it) played for the Newtonian. In both cases, these facts are said to ground all other facts about the physical system. For the Newtonian, those other facts concerned the spatial relationships between material bodies. What are the other facts for the substantivalist in GTR? They are many and varied, but most perspicuous are facts that characterize the space-time interval between different parts of the physical mass-energy field.19 The substantivalist will claim that they are grounded in (i.e. explained by, in the metaphysical sense) facts about the manifold, including how the fields are distributed over it. That is the core substantivalist view, but of course it may be developed in all sorts of ways. To take just one (extreme!) example, our substantivalist need not think that facts about the manifold are groundless: if she is a Spinozist she may ground them in the essence of God. The important point is that she does not ground them back in terms of the space-time intervals between bits of matter.20 The issue I just described is clearly an extension of the original issue of Newtonianism vs Leibnizianism into the context of GTR. But it is not the only extension. For example, the original Newtonian vs Leibnizian debate is sometimes characterized as a debate about the nature of space, yet I have not said anything about what space or spacetime is in GTR. One could of course ask whether spacetime in GTR is to be identified with the physical manifold alone or the manifold along with the physical metric field, but that is not the question I ask here.21 No doubt there are yet other questions about GTR systems that are an extension of some aspect of the Newtonian vs Leibnizianism issue, but here I focus on the issue described above.

3. The Hole Argument

So far I have argued that substantivalism in GTR should be understood as a claim about ground . Even though it is rarely stated explicitly in these terms, I believe that it is what theorists like Hoefer implicitly have in mind when they describe substantivalism as the thesis that the manifold exists independently of

The Bare Necessities / 123

material things. And as I said at the outset, substantivalism is the received view these days. I now turn to motivating my central claim, namely that substantivalism faces far more serious difficulties than has been recognized and that an adequate response to these difficulties has not yet been formulated, let alone defended. Now one difficulty with substantivalism that has been widely discussed is brought out in Earman and Nortons Hole Argument. Indeed, most contemporary substantivalists have developed their view primarily as a response to that argument. So my aim in the next three sections is to describe exactly what those responses are so that we understand what kinds of substantivalist views constitute the received wisdom. Then in Sections 69 I will argue that these received views are all vulnerable to other objections. What then is the Hole Argument? Earman and Nortons basic idea was that substantivalism implies that GTR is indeterministic. They took this to be a bad consequence and reason to think that substantivalism is false. What does it mean to say that a theory is deterministic? Standard definitions these days are given in terms of possible worlds. For example, a theory T might be said to be deterministic iff any two possible worlds in which T is true and which agree in all respects at one time agree in all respects at all later times. Now in GTR this definition should really be rephrased in a relativistically acceptable way by avoiding reference to times. Moreover, even in classical contexts this definition is subject to certain kinds of counterexamples.22 But none of these issues are relevant to our purposes so it pays to ignore them and work with this definition for simplicity. Why is substantivalism thought to imply that GTR is indeterministic? To describe the argument we need the notion of a diffeomorphic shift. To this end, recall that the intended models of GTR are of the form (M , g, T ). For simplicity we can assume that all intended models have the same manifold of points M , though of course different models will differ in the fields g and T . Now, take a model M = (M , g, T ) and consider a diffeormophism d :M M , a bijection from M to M such that it and its inverse are both differentiable. Such a function on the manifold induces a function on models, whereby a model M = (M , g, T ) is mapped to a drag along structure d (M ) = (M , d (g), d (T )), where d (g) is a tensor field that results from dragging the value of g at each point in M along to its image under d . Formally: for all x in M , the value of the field d (g) at the point d (x) is g(x). The field d (T ) is defined similarly. Such a function on models is called a diffeomorphic shift. Now I called d (M ) a structure rather than a model, since for all I have said there is no reason to think that it will satisfy the laws of GTR. But it turns out that d (M ) is a model of GTR iff M is, for any model M and any diffeomorphism d . Let us call this property of GTR diffeomorphism invariance.23 Diffeomorphic shifts are somewhat analogous to the rigid spatial shifts I mentioned while discussing Newtonianism and Leibnizianism. For we can think of a rigid spatial shift as a function on classical models that maps each model to

124 / Shamik Dasgupta

one that differs only in the fact that all the matter is displaced by some uniform distance in some uniform direction. In some sense, the most direct analogue of this in the case of GTR would be the result of just dragging the mass-energy field over the manifold. But in GTR the metric field varies from model to model and the equations of GTR state how it relates to the mass-energy field. So just dragging the mass-energy field around will not (in general) preserve the equations of GTR since the resulting mass-energy field will (in general) be out of kilter with the metric field. If we are interested in functions that preserve the truth of the equations of GTR, we must consider functions of the form d (M ) in which the metric field is dragged along with the mass-energy field. Returning to the Hole Argument, consider an arbitrary model M = (M , g, t) of GTR and a diffeomorphism d that is identity on all of M save for a bounded region R, the hole. Let us call any such diffeomorphism a hole diffeomorphism. By diffeomorphism invariance, d (M ) is also a model of GTR. Since d was identity outside R, M and d (M ) are exactly the same outside R. The only difference is that in d (M ) the fields g and T have been shifted around together within R. The idea is then to use M and d (M ) to show that GTR is indeterministic. After all, if R is chosen wisely it will only contain events that occur after a specific time, in which case the models would agree in all respects at that time and diverge thereafter. But the definition of determinism was given in terms of possible worlds, and yet M and d (M ) are models. What then is the link between models and possible worlds that might complete the argument? This issue has been the source of much controversy. Earman and Norton used the premise that if substantivalism is true, then given any model M and any diffeomorphism d , M and d (M ) represent distinct possible worlds W and d (W) respectively.24 By writing d (W), we are now using d to represent a function on worlds that is induced by a diffeomorphism on the physical manifold (as opposed to when we write d (M ), where d is used to represent a function on models that is induced by a diffeomorphism on the mathematical manifold M ). Intuitively, d (W) differs from W only in the fact that the physical metric and mass-energy fields have been uniformly dragged over the physical manifold in accordance with the diffeomorphism on the manifold. Since d is a hole diffeomorphism, this means that the fields are dragged around within the hole but left alone outside. The idea behind Earman and Nortons premise, then, is that if (as the substantivalist claims) the manifold exists independently of the fields, it must surely be possible to redistribute the fields over the manifold in this way. It is important to understand exactly how W and d (W) differ. To this end, let us distinguish between individualistic and qualitative facts. I suspect that the distinction resists being defined in other terms, but it is reasonably intuitive and can be illustrated with examples. Roughly speaking, a fact is individualistic when it concerns a particular individual. For example, consider the fact that this very book (pointing to a book on my table) is blue. This is an individualistic fact since

The Bare Necessities / 125

it concerns that particular book. In contrast, a qualitative fact does not concern any particular individual at all. The fact that there exists a book on my table is qualitative because it does not concern any particular book and would obtain even if a different book was on the table. More generally, qualitative facts include all those facts that can be expressed in predicate logic with identity but without constants, such as (x) Fx (x) (Fx & (y) (Fy y = x)) so long as F and G express qualitative properties. Of course, this now raises the question of what a qualitative property is, but again the idea is reasonably intuitive: qualitative properties do not concern any particular individual. Thus, the property of being green is qualitative, but the property of being Socrates is not. Note that the property of being a sibling is qualitative: having it might imply the existence of other individuals, but it is qualitative because (roughly speaking) your having it does not depend on any particular person being your sibling and just depends on your having some sibling or other.25 Now I said that W and d (W) differ only in the fact that both physical fields have been uniformly dragged over the manifold. Importantly, both worlds agree on all qualitative facts about the distribution of the fields over the manifold, for example that there exist various regions propertied and related in certain ways. They disagree only on individualistic facts of the form (r 1 , r 2 , . . .) concerning which particular regions within the hole play which qualitative roles. To complete the Hole Argument, then, we pick a time t to the past of the hole. By construction, W and d (W) agree on all facts at t and yet diverge thereafter, namely with respect to those facts within the hole just mentioned. By the above definition of determinism, it follows that GTR is indeterministic. Why is this a problem for the substantivalist? Earman and Norton do not take it to be apriori that GTR is deterministic. Instead, their idea is that if GTR is indeterministic this should follow from facts about the physics, not from metaphysical theses such as substantivalism. As they put it, determinism may fail, but if it fails it should fail for reasons of physics . . . .26 This raises the question of what makes something a reason of physics rather than a reason of metaphysics, but I will not discuss this aspect of their argument further. We can now represent their argument explicitly as follows: (H1) If substantivalism is true then given any model M and any hole diffeomorphism d , M and d (M ) represent distinct possible worlds W and d (W) respectively. (Premise)

126 / Shamik Dasgupta

(H2) W and d (W) are both worlds in which GTR is true. (By the fact that GTR is diffeomorphism invariant)27 (H3) If W and d (W) are both worlds in which GTR is true, then GTR is indeterministic. (By the definition of determinism) (H4) Therefore, if substantivalism is true then GTR is indeterministic. (By (H1), (H2) and (H3)) (H5) (H4) is unacceptable: indeterminism should not follow from a metaphysical theory such as substantivalism. (Premise) (H6) Therefore, substantivalism is false.

4. Standard Responses
The Hole Argument is the only serious objection to substantivalism in GTR that has received much attention recently. And it is widely agreed to fail, since most theorists either reject (H1) or (H3). As a result, substantivalism in GTR is considered reasonably unproblematic. But in the next two sections I will try to get as clear as possible about how substantivalists have responded to the Hole Argument, for I will then argue that their resulting views are vulnerable to other objections. Substantivalists who deny (H3) concede for the sake of argument that M and d (M ) represent distinct possible worlds W and d (W) in which GTR is true. But they deny that it follows that GTR is indeterministic because they think that the definition of determinism used by Earman and Norton is too strong. They claim that since W and d (W) agree in all qualitative respects and diverge only with respect to individualistic facts, this should not count as a failure of determinism. For our purposes there is no need to discuss the details of their alternative conception of determinism; it suffices to say that it implies that GTR is deterministic even if M and d (M ) represent distinct possible worlds W and d (W).28 But by far the more popular response is to deny (H1) and insist that the substantiavlist need not take M and d (M ) to represent distinct possible worlds W and d (W) after all. Now if (H1) is taken literally, this response is obviously correct. For a substantivalist might grant that there are distinct worlds W and d (W) but simply deny that they are represented by M and d (M ) respectively. After all, models are just representational tools, not our masters, so even a substantivalist who recognizes distinct possible worlds W and d (W) is free to ignore one of them and use M and d (M ) to represent the same world if she pleases! But equally obviously, pointing this out is futile. For once it is granted that there are distinct worlds W and d (W), indeterminism follows regardless of how they are represented (at least, it follows on the definition of determinism that Earman and Norton use).29 This shows that Earman and Nortons premise (H1) is ill formulated. Since the definition of determinism is put in terms of possible worlds, the important

The Bare Necessities / 127

question is whether the substantivalist must say that there are distinct possible worlds like W and d (W) regardless of whether she chooses to represent them with M and d (M ) respectively. Thus the first premise should really be (H1*) If substantivalism is true then given any GTR world W and any diffeomorphism d on the physical manifold of W, there is a distinct possible world d (W). and the argument goes through just as it did before.30 The popular response, then, is to deny (H1*). Indeed, to call this response popular may be an understatement. As Pooley puts it, there is close to a consensus in the philosophical literature on Earman and Nortons hole argument that there is nothing anti-substantival about denying that there can be such distinct possible worlds.31 Given the discussion in Sections 1 and 2, it should be obvious that (H1*) is false. For as we saw there, the substantivalist might also be a Spinozist who thinks that everything true is necessarily true, in which case she will deny that diffeomorphic shifts are possible because she denies that anything non-actual is possible in the first place! Such a theorist holds a perfectly consistent view: that facts about the manifold ground other physical facts about a GTR system (this is her substantivalism), but that all facts are ultimately grounded in the essence of God, a necessary being, so that nothing non-actual is possible (this is her Spinozism). Now Spinozism is of course not a popular view these days (to put it mildly!). But there are many popular views about modality that imply that diffeomorphic shifts are impossible even if substantivalism is true. For example, consider the modal thesis of anti-haecceitism, which we can state as the view that qualitatively identical possible worlds are identical in all respects. If anti-haecceitism is true, there do not exist worlds like W and d (W) that differ only in a diffeomorphic shift (since if there were such worlds they would be qualitatively identical). So if anti-haecceitism is true, substantivalism does not imply that GTR is indeterministic after all (at least, not in the way that Earman and Norton claimed). Unsurprisingly, then, the popular substantival response to the Hole Argument is to endorse anti-haecceitism. Now the substantivalist might endorse anti-haecceitism as a brute thesis about modality. Alternatively, she could propose deeper theses about the nature of modality that imply anti-haecceitism. For example, Brighouse writes that what a substantivalist should say about the way we individuate space-time points or regions across possible worlds is that we individuate according to qualitative similarity.32 Now this talk of individuating points across possible worlds is widespread but can mean several things. One thing it could mean (which was not Brighouses intention) is the view that each point p has an individuating qualitative role, by which I mean a qualitative role such that, necessarily, a point q instantiates the role iff q is identical to p. If there were diffeomorphically shifted worlds, certain qualitative roles would be instantiated

128 / Shamik Dasgupta

by one individual in one world and a distinct individual in the other. The view that points have individuating qualitative roles therefore implies that diffeomorphic shifts are impossible. Thus, the substantivalist could endorse the modal view that regions have individuating qualitative roles and thereby deny that GTR is indeterministic.33 Another view about modalitythis time one that Brighouse may have intended to encompass by her talk of individuationis counterpart theory. On this view, no individual exists in more than one possible world. However, an object in one world is a counterpart of an object in another iff they resemble each other in certain (contextually determined) qualitative respects. An object x is then said to be possibly F (in a context) iff there is a possible world W that contains an individual y which is F and which is (in that context) a counterpart of x. Now on this view, the possibility of a diffeomorphic shift would be witnessed by a qualitatively identical possible world. But, as Brighouse puts it, the counterpart of any given point in any of the qualitatively indiscernible worlds will have all the same qualitative properties as that point has, in which case the possible world will not represent the possibility of a shift after all.34 I should say that it is not entirely clear whether Brighouse is using Lewis own counterpart theory or modifying it for her own purposes.35 But the issues here are tricky, and since I have no desire to object to this strategy of responding to the Hole Argument let me concede that some version of counterpart theory implies that diffeomorphic shifts are impossible. Butterfield (1988) also uses counterpart theory to similar effect, and Teller agrees that this is a legitimate move in his Teller (2001). These are not the only modal views that imply that diffeomorphic shifts are impossible. But there is no need to describe every conceivable view. It is enough to point out with Pooley, that it is an almost universal consensus on the Hole Argument that (H1*) is false: the substantivalist may endorse one of these modal views that imply that diffeomorphic shifts are impossible and thereby deny that GTR is indeterministic. But while this may be close to a consensus, what has not been sufficiently appreciated is that the substantivalist might claim that diffeomorphic shifts are impossible for two very different reasons. On one approach, the claim follows from her views about the nature of the manifold. But on another approach, it is a bare modal claim, a claim that does not follow from the nature of the manifold but rather from independent facts about the nature of modality. The distinction here is crucial to what follows, so let me say more about it. Recall that the substantivalist grounds facts about the space-time intervals between bits of matter in facts about the manifold of the form (r 1 , r 2 , . . .) where r 1 , r 2 , . . . are particular regions of the manifold. These are individualistic facts since they concern those particular regions r 1 , r 2 , . . . . But the substantivalist may wish to go on and ground these individualistic facts in qualitative facts. Let

The Bare Necessities / 129

us call the view that they are qualitatively grounded thin substantivalism, and the view that they are not thick substantivalism. Like substantivalism in general, we will take each variety to be a necessary claim. Now if thin substantivalism is true then it follows straight from the principle of Necessitation (the principle that the ground metaphysically necessitates what it grounds) that qualitatively identical worlds agree on all individualistic facts about the manifold and therefore that there do not exist worlds like W and d (W) that differ only in a diffeomorphic shift. Thick substantivalism does not have this implication: since it holds that individualistic facts about the manifold are over and above the qualitative facts in the sense of not being grounded in them, it allows that qualitatively identical worlds like W and d (W) can differ just with respect to which regions of the manifold play which qualitative role. But must the thick substantivalist think that there are distinct possible worlds like W and d (W)? Not at all, for she might endorse any of the modal views just discussed, such as Spinozism or anti-haecceitism, as a bare modal claim, a claim that has nothing to do with the nature of the manifold per se but which implies that shifts are impossible. It is obvious that if the thick substantivalist is also a Spinozist, her denial that diffeomorphic shifts are possible follows not from her specific views about the nature of the manifold but rather from independent views about religion and modality. Like the thin substantivalist, this theorist agrees with Pooleys consensus that diffeomorphic shifts are impossible. But while the thin substantivalist takes this claim to follow (by Necessitation) straight from her views about the nature of the manifold, this thick substantivalist takes it to be a bare modal claim. What might be less obvious is that the thick substantivalist may endorse any of the above anti-haecceitistic views as a bare modal claim too. For example, she may endorse counterpart theory as a thesis about the nature of modality, a theory about what it is for a given individual to be possibly F or necessarily G. This theorist holds a perfectly consistent view: that individualistic facts about the manifold ground other physical facts about a GTR system (this is her thick substantivalism) but that counterpart theory is the correct theory of what it is for an object to have a modal property. To be sure, her counterpart theory may imply that individualistic facts are necessitated by qualitative facts. But this does not mean that she believes the former to be grounded in the latter, for as we saw earlier modal claims do not imply grounding claims (mathematical facts are necessitated by the fact that Obama exists, but the former are not grounded in the latter). Thus even though her views about modality imply that diffeomorphic shifts are impossible, she nonetheless insists (as a thick substantivalist) that individualistic facts about the manifold are something over and above the qualitative facts in the sense that the former are not grounded in the latter. Counterpart theory may differ from Spinozism in many respects (including its plausibility!), but one point of similarity is that the thick substantivalist can endorse it as a bare modal claim in order to respond to the Hole Argument.

130 / Shamik Dasgupta

Similarly, the thick substantivalist may endorse the modal view that regions of the manifold have qualitative individuating roles as a bare modal claim. She will then deny that diffeomorphic shifts are possible, but will nonetheless insist that individualistic facts are something over and above the qualitative facts in the sense that the former are not grounded in the latter. These then are the two ways that a substantivalist can deny the possibility of diffeomorphic shifts that I want to distinguish. On the one hand, she might be a thin substantivalist in which case the impossibility of diffeomorphic shifts follows (by the principle of Necessitation) from her view that individualistic facts about the manifold are grounded qualitatively. On the other hand, she might be a thick substantivalist in which case the impossibility of diffeomorphic shifts is a bare modal claim and follows from her independent views about the nature of modality. So far I have not said anything about which approach the theorists described above took. But I will argue in the next section that thin substantivalism remains undeveloped. Thus, insofar as substantivalists have presented clearly formulated theories on which diffeomorphic shifts are impossible, they are thick substantivalist views on which the impossibility is a bare modal claim. Then in the subsequent sections I will argue that thick substantivalism is subject to serious objections. The charge will be that if individualistic facts about the manifold are something over and above the qualitative facts (as the thick substantivalist insists), they are redundant and undetectable in GTR in just the same sense that absolute velocity is redundant and undetectable in classical physics. I will argue that this is true regardless of whether diffeomorphic shifts are possible, and that this is good reason to reject thick substantivalism. That will constitute my argument for my central claim, namely that an adequate substantival view has not yet even been formulated and that substantivalisms status as the received view is premature.

5. The Elusive Thin Substantivalist

I want to argue that no workable version of thin substantivalism has yet been proposed. Even though some authors may appear at first to have thin substantivalist leanings, on further examination no thin substantivalist theory is forthcoming. Admittedly, this is somewhat delicate since most participants to the debate have not explicitly written in terms of ground, the notion in terms of which I defined thin substantivalism. So I will first say what would have to be done in order to propose a workable thin substantivalist theory, and then argue that nothing like that has been done. We may start by noting that out of thick and thin substantivalism, thick substantivalism is surely the default substantival view. After all, it is overwhelmingly natural to suppose in general that qualitative facts are grounded in individualistic facts. We naturally think that there being a book on the table is grounded

The Bare Necessities / 131

(i.e. explained, in the metaphysical sense) by the fact that this very book is on the table. The latter individualistic fact is, we like to think, that in virtue of which the former obtains. So the the onus is very much on the thin substantivalist to motivate her view. Moreover, my description of thin substantivalism so far is no more than the vaguest of hand gestures. I described it as the view that individualistic facts about the manifold are grounded qualitatively, but this says nothing about what kinds of qualitative facts are taken to provide the ground. So before motivating her view the thin substantivalist must tell us what the view is. Specifically, she must (1) clearly articulate what the underlying qualitative facts are like, and (2) show that they are sufficient to explain (in the metaphysical sense) individualistic facts about the manifold. Until she does this, there is no thin substantival view on the table to assess. It is important to understand what is required to complete task (1). One way to attempt it would be to endorse the traditional bundle theory of individuals, according to which all facts about the world are ultimately grounded in facts of the form F, G, . . . are compresent where F, G, . . . are intrinsic, monadic, qualitative properties and compresent is a primitive, plural predicate. Here we are being told explicitly what kinds of items the world contains at the fundamental level (qualitative properties), and what can be true of them (they can be compresent). Now, when substantivalists discuss the traditional bundle theory they usually reject it. The typical complaint is that it cannot make sense of symmetric physical systems in which more than one individual has the same intrinsic, monadic, qualitative properties. The famous Max Black world in which there are two qualitatively identical spheres of iron 3 meters apart is often used as an example of such a system. I agree that this is good reason to reject the traditional bundle theory; my point here is just that it is at least a theory that attempts task (1). As we will see, this is more than can be said for other proposals that might sound, on the surface, like an expression of thin substantivalism. Note that the thick substantivalist who endorses anti-haecceitism faces no task comparable to (1). For as we saw earlier, qualitative facts include all those facts that can be expressed in the language PL of predicate logic with identity but without constants, so long as the predicates express qualitative properties.36 So an anti-haecceitist might express her view as follows: any two worlds that agree on all facts expressible in PL agree simpliciter. Regardless of whether it is true, it is a view on the table that we can now assess. But the thin substantivalist does not have it so easy. For suppose she proposes that all individualistic facts are grounded in facts expressible in PL.37 The trouble is that it is not clear what this could mean. For it is arguably analytic of the existential quantifier that existentially quantified facts are grounded in their instances.38 Indeed, this

132 / Shamik Dasgupta

understanding of the quantifier is arguably implicit in the standard Tarskian semantics for PL. So if we are now being told, say, that the fact that (x)Fx is fundamental, I need to be re-taught how to interpret PL. Which parts of (x)Fx are referential? Which are predicational? Or does it have some other structure altogether? The bundle theorist did better in this regard, since she clearly told us what kinds of things the world contains and what can be true of them. The trouble with the current suggestion is that we are being told nothing similar. This does not refute the suggestion, but it does show that much more needs to be said before a view has even been stated. Given what it takes to complete task (1), it is reasonably clear that no thin substantival view has yet been proposed. I do not claim that it cannot be done in Section 11 I will make a start at formulating such a viewjust that we have yet to see what a developed thin substantivalist view looks like. This is most clear with the substantivalists discussed in the last section. For while Brighouse, Butterfield and Pooley all take great care to develop their preferred version of anti-haecceitism, they do not engage in task (1). This is not to say that they are committed to rejecting thin substantivalism. Indeed, Pooley explicitly considers the possibility that anti-haecceitism follows from some thesis about what grounds individualistic facts about the manifold. My point is just that this thesis about ground is not developed. Other substantivalists make comments that sound like they might have thin substantivalism in mind. But if so, they seem not to have appreciated what it takes to develop such a view since they have not made any attempt at task (1). For example, Hoefer says that we should reject the ascription of primitive identity to space-time points.39 Now I am not entirely sure what he means by this, and sometimes when clarifying his view he seems to be proposing a bare modal thesis such as anti-haecceitism.40 But at other times he seems to have something like thin substantivalism in mind. If so, it is striking that he does not develop a positive theory of what the underlying qualitative facts are like and rests for the most part on negative claims such as that quoted above. Similarly, Baker says that we should reject the possibility of [haecceitistic] facts, and by a haeeceitistic fact I take it that he means an individualistic fact without a qualitative ground.41 But again, this is a purely negative claim and not a positive theory of what the underlying qualitative facts are like. Am I demanding too much of the would-be thin substantivalist? I think not. After all, consider what we demand from the Leibnizian in a classical context. We characterized Leibnizianism as the view that all facts about space are grounded in facts about how material bodies are related to one another, but that was just a vague hand gesture and not a specific proposal. What we need from a would-be Leibnizian is (1) a specification of what kinds of relational facts she takes as basic, and (2) a demonstration that those relational facts are indeed sufficient to explain (in the metaphysical sense) all facts about spaceor at any rate all facts about space needed to make sense of classical physics. It is precisely when we suspect that she is not able to do this (perhaps for the kinds of

The Bare Necessities / 133

considerations outlined in the Bucket Argument) that we reject Leibnizianism. Thus it is clearly not sufficient for a would-be Leibnizian to make the negative statement I reject Newtonian space and leave it at that. This point is particularly important if ones argument for Leibnizianism is not that its competitors are somehow incoherent but rather that they score badly on a particular theoretical virtue such as ontological parsimony. For even if its competitors score badly on that one virtue, we should only endorse Leibnizianism if it does better than its competitors overall, once all virtues are considered. But obviously we can have no idea whether Leibnizianism does better until we have the theory on the table to assess! Similarly in the case of thin substantivalism. We can have no idea whether thin substantivalism is our all-things-considered best theory until we develop the theory and see what it looks like. Purely negative statements to the effect that there are no primitive identities or haecceities are simply not enough to go on, but those sorts of remarks are all we get from Hoefer and Baker.42 Other people who sound like they might have thin substantivalism in mind are the so-called structuralists. But again, when they express their views they make no attempt at task (1). For example, Ladyman and Ross said that the structure of spacetime [should be] accepted as existent despite its failure to supervene on the reality of spacetime points, and that we should think of spacetime points as entities whose identity and individuality are secondary to the relational structure in which they are embedded.43 These sound at first like the remarks of a thin substantivalist, but we are never told what kinds of qualitative facts constitute that underlying relational structure. At other times, structuralists say that spacetime points are individuated by their role in a qualitative structure. But what does this mean? On one interpretation, it means that regions of the manifold have the individuating qualitative roles discussed in the last section. But as we saw, that is a purely modal claim that can be endorsed by the thick substantivalist, not a statement of thin substantivalism. If this talk of individuation is intended to be a statement of thin substantivalism, we need to be told what the underlying qualitative facts are like before there is a view on the table to assess. At yet other times, structuralists have expressed their view by talking about invariances. To see what the idea is here, consider rational choice theory in which an agents preferences over options are represented by a utility function assigning a number to each option. It is sometimes said that a utility representation is invariant under positive monotonic transformations, and this often taken to communicate something about the underlying psychological facts: namely that an agent can meaningfully be said to prefer x over y, but not to prefer x twice as much as y. With this in mind, a would-be thin substantivalist might attempt to express her view that the underlying facts about the manifold are all qualitative by saying that the models of GTR are invariant under diffeomorphism, i.e. that for all models M and diffeomorphisms d , M and d (M ) represent the same possible world.44

134 / Shamik Dasgupta

But this does not constitute a thin substantival theory, for two reasons. First, it does not even imply thin substantivalism: as noted in the last section any substantivalistincluding the thick substantialist who thinks that diffeomorphic shifts are possiblecan interpret her models in this way since models are our tools not our masters. But second (and more importantly), talking of invariances does nothing to tackle task (1). In the case of rational choice theory this does not really matter, for the analogue of task (1) is trivial. We have a firm grasp of the kinds of facts that might be psychologically basicperhaps they are facts to the effect that an agent prefers x over y, or perhaps that she prefers x twice as much as yand so it is harmless to use talk of invariances to communicate which of these are indeed psychologically basic. But in the case of thin substantivalism we have not yet been told what kinds of qualitative facts might be taken to be basicthat is precisely what task (1) is. It is no good simply saying they are the sorts of facts that M and d (M ) agree on, for one kind of fact they agree on are existentially quantified facts and as I emphasized earlier it is not clear what it could mean for these to be basic. Finally, at other times structuralists talk of discernibility. Call two objects absolutely discernible iff there is a qualitative formula with one free variable that is satisfied by one but not the other. The spheres in a Max Black world, for example, are not absolutely discernible. It is widely agreed that the traditional bundle theory implies that distinct objects are absolutely discernible and that this is a serious problem for the view. However, it was then noticed that there is a weaker notion of discernibility: two objects are weakly discernible iff there is an irreflexive qualitative formula with two free variables that they satisfy.45 The spheres in the Max Black world, notice, are weakly discernible, and the thesis that distinct individuals must be weakly discernible is considered by some to be far less problematic than the thesis that they must be absolutely discernible. Thus, some theorists who sound like they have thin substantival leanings have claimed, as an alternative to the bundle theory, that, necessarily, distinct objects (including regions of the manifold) are weakly discernible.46 But I find all this talk of discernibility a red herring. For the thesis that, necessarily, distinct regions are weakly discernible does not even imply thin substantivalism, since the former is a purely modal thesis that a thick substantivalist may accept as a bare modal claim. It might be objected that the claim of interest is not this modal claim but rather the claim that the distinctness between two regions a and b is grounded in the fact that they are weakly discernible. But the fact that a and b are weakly discernible is an individualistic fact, concerning (as it does) a and b respectively, so this does not constitute a view on which the distinctness of a and b is grounded qualitatively. And in any case, there is still the question of what grounds all the other individualistic facts about them, facts that seem to go undiscussed in the literature on weak discernibility. But the most important problem is that none of this talk of weak discernibility does anything to tackle task (1): nowhere in the literature on weak discernibility have I seen an attempt to clearly articulate what the

The Bare Necessities / 135

underlying qualitative facts might be that are taken to ground individualistic facts about the manifold. Moreover, in Section 11 I will outline an approach to grounding individualistic facts qualitatively that has no implications about discernibility at all, and in particular allows that distinct objects can be weakly indiscernible. So I worry that the recent focus on discernibility is doubly misplaced: not only do claims about discernibility fail to imply thin substantivalism, they are not implied by thin substantivalism either. I suspect that the intense focus on discernibility recently is the result of conflating two traditional claims about the Max Black world: first, that it shows that the bundle theory is false; and second, that it shows that distinct objects can be indiscernible. It is correct to point out that this second claim is false given that the two spheres are weakly discernible. But it is hard to see how this is relevant to the thin substantival project, for it remains the case that the only live theory about how individualistic facts might be qualitatively grounded (i.e. the traditional bundle theory) cannot make sense of symmetric systems like the Max Black world. What the thin substantivalist needs to do is present a clear theory about what the underlying qualitative facts are like in a Max Black world, but this talk of weak discernibility makes no progress on that. I have argued that formulating a thin substantival theory requires work that has not been done. Thus insofar as we can extract clear, fully formulated theses from the literature that deny (H1*), they are all theses that endorse thick substantivalism plus some kind of bare modal claim such as anti-haecceitism. Now these modal theses are (in my opinion) perfectly legitimate responses to the Hole Argument. But the trouble is that there are other arguments which purport to show that thick substantivalism should be rejected in favor of thin substantivalism that do not make modal assumptions in the first place. So the problem with the modal theses is not that they are wrong, but rather that they do not constitute a serious defense of substantivalism.

6. Two Occamist Arguments

Philosophers of physics should find these other arguments compelling, for they are analogues of widely accepted arguments that are used to reject Newtonianism in the context of classical physics. Roughly speaking, these arguments go like this. First, note that if Newtonianism is true then there is a fact of the matter as to the velocity of a given body that is independent of its velocity relative to other material bodies, namely its velocity through Newtonian space. Call this its absolute velocity. The first argument is then that a bodys absolute velocity is redundant to classical physics, and the second argument is that absolute velocity is undetectable according to classical physics. In both arguments an Occamist principle is then wielded stating that (all else being equal) we should prefer theories about the structure of the material world that dispense with

136 / Shamik Dasgupta

redundant or undetectable structure; thus, (all else being equal) we should reject the Newtonian view of space. I personally prefer the second argument, but I will discuss them both here for the benefit of those that prefer the first.47 These arguments are normally used to motivate what is known as Galilean substantivalism. On this view, facts about the spatio-temporal relations between material bodies and events are grounded in facts about where those bodies and events are situated in Galilean spacetime, a 4 dimensional array of regions organized as an ordered set of 3 dimensional simultaneity slices. Importantly, there is no fact of the matter as to the spatial distance between two nonsimultaneous points, so no fact of the matter as to the spatial distance traversed by a body over an interval of time, and so no fact of the matter as to its absolute velocity. There are facts about a bodys velocity relative to other bodies just like there are according to Newtonianism. But while the Newtonian grounds these in facts about the absolute velocities of each body, the Galilean substantivalist grounds does not. Thus, Galilean substantivalism dispenses with precisely those facts that were shown to be redundant and/or undetectable by the Occamist arguments. I believe that precisely analogous arguments militate against thick substantivalism. Call an individualistic fact thick iff it does not have a qualitative ground. In the next three sections I will argue (first) that analogous arguments show that thick individualistic facts are redundant to GTR and undetectable according to GTR; and (second) that these Occamist arguments make no essential appeal to assumptions about modality or determinism that are denied by the substantivalists considered earlier. Thus, just as the Occamist arguments motivate us to dispense with absolute velocity and reject Newtonianism in favor of the Galilean view, they should also motivate us to dispense with thick individualistic facts and reject thick in favor of thin substantivalism. At the very least, on pain of inconsistency the contemporary substantivalist faces a dilemma: either reject thick substantivalism in GTR, or else reject the kinds of arguments used to motivate Galilean substantivalism over Newtonianism in the classical case. I suspect that few would be willing to embrace the second horn. The Occamist arguments in the classical case use a notion of invariance analogous to the notion of diffeormorphism invariance in GTR. To set this up, consider the laws of Newtonian Gravitation Theory (NGT), by which I mean his three laws of motion and the inverse-square gravitational force law. Let us assume that there is some way of expressing these laws that is independent of the question of Newtonian substantivalism.48 The models of NGT will therefore be many and varied: some will contain a manifold of regions isomorphic to Newtonian space, others will contain a manifold of regions isomorphic to Galilean space-time, and so on. Now, let a uniform boost v be a function that takes as input any model and outputs a structure that differs only by adding a uniform absolute velocity to each body at each time. Such a function will be identity of all models with a manifold isomorphic to Galilean spacetime and will only map models with a notion of absolute velocity to distinct models. It then turns out that given any

The Bare Necessities / 137

model M of NGT and any uniform boost v , v (M ) is also a model of NGT. Let us call this property of NGT velocity invariance.

7. The Argument from Redundancy

So much for the set-up; we must now examine the Occamist arguments in more detail. As I said, the idea behind first argument is that absolute velocity is physically redundant in NGT. But redundant to what? It is easy to make a false accusation here. Sometimes the accusation is that facts about the particular absolute velocities of things make no difference to how a system evolves according to NGT.49 Now it is true that the absolute velocities of bodies at a given time make no difference to, say, the future inter-particle distances between things. After all, a uniformly boosted world is a world in which current absolute velocities are different but future inter-particle distances are the same; and since the laws of NGT would obtain in this boosted world (by velocity invariance) this shows that according to NGT the specific absolute velocities of things make no difference to future inter-particle distances. But while true, the point does not have the required generality. For according to NGT, the current absolute velocities of things make a difference to some aspects of how a system evolves such as the future absolute velocities of things. After all, if everything were uniformly boosted, the laws of NGT imply that things will have different future velocities than they actually will.50 It would be question begging to ignore these ways in which absolute velocity makes a difference to how a system evolves. A more accurate accusation is that facts about the particular absolute velocities of things make no difference to whether NGT obtains. I will discuss what this amounts to in some detail, but roughly speaking the idea is this. Consider an arbitrary model M of NGT and an arbitrary uniform boost v . Since NGT is velocity invariant, we know that v (M ) is a model of NGT. But in v (M ) the absolute velocities of things are different than they are in M (except for the trivial case in which v is identity). So whether material bodies have the velocities that are assigned to them in M or in v (M ) makes no difference to whether NGT is true. Since M and v were arbitrarily chosen, this shows that the particular absolute velocities that things have make no difference to whether NGT is true. Roughly speaking, this is what I mean when I say that absolute velocity is redundant to NGT. The same is not true, note, about relative velocity. For uniformly changing the relative velocities between things will not, in general, preserve the truth of NGT. This then is the argument from redundancy: (R1) Absolute velocity is redundant to NGT. (By the fact that NGT is velocity invariant)

138 / Shamik Dasgupta

(R2) If a theory about the metaphysics of the material world implies that it contains features that are redundant to the truth of the true and complete physical laws, that is reason to think it is false. (Premise) (R3) Therefore, if the laws of NGT are true and complete, there is reason to think that Newtonianism is false. I take this to be one of the standard arguments that motivate many people to reject Newtonianism in favor of the Galilean view in a classical context.51 I will argue (first) that if you find it compelling in the case of NGT, a precisely analogous argument should motivate you to reject thick substantivalism in favor of thin substantivalism, and (second) that it makes no essential appeal to assumptions about determinism or modality that substantivalists described in Sections 4 and 5 deny. The analogous argument in the case of GTR appeals to (R1*) Thick individualistic facts about the manifold are redundant to the truth of GTR. (By the fact that GTR is diffeomorphism invariant) Along with (R2), this implies that if the laws of GTR are true and complete, there is reason to think that thick substantivalism is false. But what is the motivation for (R1*)? Following the rough idea described above, it goes like this. Consider an arbitrary model M = (M , g, T ) of GTR and an arbitrary diffeomorphism d on M . By diffeomorphism invariance, d (M ) is a model of GTR. But d (M ) differs from M with respect to thick individualistic facts concerning which particular regions play which particular qualitative role. Since M and d were arbitrarily chosen, this shows that those individualistic facts make no difference to whether GTR is true in the same sense as facts about the absolute velocities of things make no difference to whether NGT is true. It should be uncontroversial that some kind of argument like this can be run against thick substantivalism, and that the argument makes no controversial assumptions about determinism. The question is whether it makes controversial assumptions about modality. For the thick substantivalist might object that when I talked of M and d (M ) in the rough description of the argument above, I implicitly assumed that they represent distinct possible worlds that differ only in a diffeomorphic shift. And as we know, the thick substantivalist may endorse anti-haecceitism as a bare modal claim and thereby deny that possible worlds can differ in this way. But I will now argue that, properly understood, the argument from redundancy makes no contentious modal assumptions. If this is right, then even though the thick substantivalist can legitimately respond to the Hole Argument by endorsing anti-haecceitism, she cannot respond to this argument similarly. To this end, it helps to return to the case of NGT and the redundancy argument against the Newtonian. I gave a quick argument for (R1) earlier in terms of models, but according to the objection this was just shorthand for a modal argument that goes like this. We first note that given any NGT world W

The Bare Necessities / 139

in which Newtonianism is true and any absolute velocity v , there is a distinct possible world v (W) that differs from W by a uniform boost of v . We then show that v (W) is an NGT world by the velocity invariance of NGT. But the absolute velocities of things in v (W) are all different than they are in W. Since this is true for any NGT world in which Newtonianism is true and any boost v , it shows that the particular absolute velocities that things have make no difference to whether NGT is true. Grant that (R1) can be motivated in these modal terms. The important point is it does not need to be. To see this, imagine a Spinozist who has recently become a Newtonian. Suppose we try to convince her to give up Newtonianism in favor of the Galilean view of spacetime by presenting her with the argument from redundancy. And suppose we motivate (R1) with the modal argument just described. She could then reasonably reply Wait, you assume that there are distinct possible worlds W and v (W), but I reject the assumption. Therefore your argument for (R1) is unconvincing and I retain my belief in Newtonianism. As a response to our modal argument for (R1) in the last paragraph, this is perfectly fair. But does it suffice to put the argument from redundancy to bed? It seems to me that it does not. For in denying that there are distinct worlds like W and v (W) our Spinozist makes a bare modal claim, a claim that follows not from her views about he metaphysics of space per se but rather from independent views about modality. Thus, as a Newtonian she still insists that there are facts about the absolute velocities of things that are over and above the relationships between bodies in the sense of not being grounded in those relationships. And we would not take the Spinozists curious modal views to imply thatsurprise surprise! those facts about the absolute velocities of things are suddenly relevant to the truth of NGT in just the same way as, say, facts about relative velocities are! So regardless of whether uniform boosts are genuinely possible our Spinozist should agree that, given the structure of the laws of NGT, the particular absolute velocities of things make no difference to whether those laws obtain. Intuitively, then, the Spinozists bare modal claim simply has no bearing on whether absolute velocity is redundant to NGT. If this is right, then even a Newtonian who denies that there are distinct worlds like W and v (W) should accept (R1). This is not to deny that (R1) can be motivated in modal terms. Indeed, the sense in which absolute velocity is redundant to NGT is particularly vivid when presented in modal terms; and since most of us do not share the Spinozists quirky views it is not surprising that one often hears the argument presented in that way. But still, the modal argument is optional and not essential to the argument from redundancy. How else might one motivate (R1)? The natural answer is to use models, just like I did originally. There I considered an arbitrary model M of NGT and an arbitrary boost v and noted (by velocity invariance) that v (M ) is also a model of NGT. But the question then is what the models represent, for if they are taken to represent possible worlds the Spinozist may object that I am sneaking in modal assumptions by the back door.

140 / Shamik Dasgupta

There are a few options here. One is to interpret the models as representing counterfactual truths. Thus, we might take the fact that v (M ) is a model of NGT whenever M is to represent the fact that if everythings velocity were to differ in a certain uniform way, NGT would still be true. The Spinozist thinks that the antecedent of this counterfactual is necessarily false, but she might nonetheless agree that the counterfactual is non-trivially true. If so, then the sense in which absolute velocity is redundant to NGT is that the truth of NGT is not counterfactually dependent on the particular absolute velocities that things have. Alternatively, one might interpret the models as representing explanatory patterns. The idea is this. The equations of NGT are universal generalizations, so it is plausible that their truth is grounded in (i.e. explained by, in the metaphysical sense) their instances, in this case facts about the trajectories of particular particles.52 But what facts about their trajectories are relevant to that explanation? That v (M ) is a model of NGT whenever M is might be taken to show that the absolute velocities of things are not relevant to the explanation in the sense that they do not make a difference to whether or not the laws obtain. If so, then the sense in which absolute velocity is redundant to NGT is that the particular absolute velocities of things are not relevant an explanation of why NGT is true. What then is the correct argument for (R1)? This is a bad question, for we should not expect there to be an argument that all Newtonians will accept. The modal argument will be rejected by a Newtonian who is also a Spinozist; the counterfactual argument will be rejected by Newtonians with quirky views about counterfactuals; and so on. Indeed, attempting to shoehorn the argument into a pre-specified template is likely to make it look weaker than it is, for there will inevitably be some quirky Newtonian who rejects that templates presuppositions. Instead, we should tailor an argument for (R1) to the particular Newtonian we are faced with. The point of the discussion, of course, is that the lesson carries over to the case of GTR when we use the argument from redundancy to argue against the thick substantivalist. To be sure, we can motivate (R1*) using a modal argument. If we choose to do so, we will assume that for any GTR world W and any diffeomorphism d there is a distinct possible world d (W) in which the fields are shifted over the manifold in accordance with d . We will point out that d (W) is a GTR world (by diffeomorphism invariance), and then conclude that the thick individualistic facts that distinguish W and d (W) are redundant to the truth of GTR. But what if the thick substantivalist also happens be a Spinozist? Or what if she happens to be an anti-haecceitist like Brighouse? She will then deny that there are distinct worlds like W and d (W) and therefore object to our modal argument for (R1*). But does that mean that the argument from redundancy must be abandoned? I think not. For when the thick substantivalist denies that there are worlds like W and d (W) because of her Spinozism or her antihaecceitism, she makes a bare modal claim: a claim that follows not from her

The Bare Necessities / 141

views about the metaphysics of the manifold but rather from her independent views about modality in general. As a thick substanativalist, she still insists that the individualistic facts about the manifold are over and above the qualitative facts in the sense that the former are not grounded in the latter. And we would not take the Spinozists quirky modal views to imply thatsurprise surprise! those thick individualistic facts are suddenly relevant to the truth of GTR in just the same way as (say) qualitative facts are! Similarly, we should not take our anti-haecceitists bare modal views to imply that those thick individualistic facts are suddenly relevant to the truth of GTR either. So regardless of whether diffeomorphic shifts are possible our Spinozist and anti-haecceitist should agree that, given the structure of the laws of GTR, the thick individualistic facts detailing which particular regions of the manifold play which qualitative role do not make a difference to whether the laws obtain. The bare modal claims made by our Spinozist and anti-haecceitist therefore have no bearing on whether those thick individualistic facts are redundant to GTR. Thus, if faced with a thick substantivalist who makes either of these bare modal claims, we should simply run the argument by appealing to models as illustrated above. We could take the models to represent counterfactual truths, or perhaps patterns of explanatory relevance, or perhaps something else. Whatever works for the thick substantivalist in question. Admittedly, the interpretation of models as representations of possible worlds is deeply entrenched. Indeed, one sometimes hears that the very content of GTR is that a certain set of possible worlds are physically possible.53 But the fact that a Spinozist could be found working in a GTR laboratory suggests to me that this interpretation of models is entirely optional and that they could be taken by a proponent of GTR to represent counterfactual truths or explanatory patterns instead. In any event, I conclude that the argument from redundancy makes no essential appeal to assumptions about modality that anti-haecceitists deny. It also (clearly) makes no assumptions about determinism. The standard ways of defending thick substantivalism from the Hole Argument therefore do nothing to defend it from the argument from redundancy.

8. The Argument from Undetectability

That is the first Occamist argument against thick substantivalism. The second is an argument from undetectability. Again, it will again be useful to discuss it in the context of classical physics in which it is used to argue against Newtonianism. In that case the argument is this: (U1) If the laws of NGT are true and complete, then if there is any such thing as absolute velocity it is undetectable.

142 / Shamik Dasgupta

(U2) If a theory about the metaphysics of the material world implies that it contains features that are undetectable, that is reason to think it is false. (U3) Therefore, if the laws of NGT are true and complete, there is reason to think that Newtonian substantivalism is false. It is important to guard against misconceptions of the Occamist principle stated in premise (U2). It is not a verificationist principle since it does not state that it is meaningless to talk of undetectable structure. The thought is just that it is a virtue of ones theory of the material world if it does not posit undetectable structure. Moreover, it uses the term undetectable in a very specific way. If the term were used to apply to anything that we cannot see with the naked eye, (U2) would recommend that we become radical scientific anti-realists and dispense with so-called theoretical entities such as electrons. But that is not the intended meaning of the term. As it is used in (U2), something is undetectable if, roughly speaking, it follows from the laws of motion governing our world that it is physically impossible for it to have an impact upon our senses. Electrons are therefore detectable in this sense because there are physically possible processes, such as those that occur in particle accelerators, by which the presence of an electron can be made to have an impact on our senses via its impact on (say) the movement of a dial. So (U2) cannot be used to motivate scientific anti-realism of an objectionable sort. I therefore take (U2) to be reasonably weak and acceptable by most. This then is the other standard argument that motivates many people to reject Newtonianism in favor of the Galilean view in a classical context. As before, I will argue (first) that if you find it compelling, a precisely analogous argument should motivate you to reject thick substantivalism in favor of thin substantivalism, and (second) that it makes no essential appeal to assumptions about determinism or modality that the thick substantivalists described in Sections 4 and 5 reject. The crucial premise is (U1). I should say at the outset that there is widespread agreement that this is true. For example, Earman writes that because Newtons laws of motion and gravitation have (Gal) as their dynamic symmetries [in particular, because they are velocity invariant], no feature of the lawlike behavior of gravitating bodies can be used to distinguish an absolute frame: in that sense, absolute space is unobservable, and it is reasonably clear from the context that by absolute space is unobservable he means what I mean when I say that absolute velocity is undetectable.54 Earman offered no argument for his claim, but there turns out to be a reasonably well known argument for it. It is most vivid when presented modally, so I will first present it in those terms and then show that it can be run without recourse to modality.55 We first note that given any NGT world W in which Newtonianism is true and any absolute velocity v , there is a distinct possible world v (W) that

The Bare Necessities / 143

differs from W by a uniform boost of v . We then note two things: (i) that v (W) is an NGT world (by the velocity invariance of NGT), and (ii) that v (W) is indiscernible from W. Precisely what indiscernible means is an delicate issue that I will discuss a little later, but the intuitive idea is just that everything looks and feels and tastes and smells exactly the same in v (W) as it does in W. Now since W was arbitrarily chosen, we can imagine that it contains organisms like us performing an experiment designed to reveal the absolute velocity of a given body. v (W) then contains those organisms performing the same experiment too. Now, the absolute velocity of that body is different in v (W) than it is in W (except in the trivial case that v is the 0 velocity boost). But we know by (ii) that the experiments outcome in v (W) is indiscernible from its outcome in W. Finally, we know by (i) that the outcome in v (W) is the outcome predicted by NGT in the situation in which everythings velocity is uniformly boosted at the beginning of the experiment. Thus, if NGT is true and complete, there is no physically possible experiment we could run that would reveal the bodys absolute velocity to us. I find this argument compelling, but I will not defend it here. Instead, I want to show that a precisely analogous argument can be given in support of (U1*) If the laws of GTR are true and complete, then if there are thick individualistic facts they are undetectable. Along with (U2), this implies that if the laws of GTR are true and complete then there is reason to think that thick substantivalism is false.56 What is the argument for (U1*)? If run analogously to the above, we would assume that given any GTR world W and any diffeomorphism d there is a distinct possible world d (W) in which the fields are dragged over the manifold in accordance with d . We would then note (i) that d (W) is a GTR world (by diffeomorphism invariance), and (ii) that W and d (W) are indiscernible in the sense that things would look and feel and taste exactly the same the latter as they do in the former. Then by exactly the same reasoning as above, it would follow that if GTR were true and complete then there is no physically possible experiment we could run that would reveal the thick individualistic facts concerning which particular region of the manifold plays which qualitative role. Now there is a subtle issue about my claim that W and d (W) are indiscernible that I will discuss in the next section. But putting that aside, it should be uncontroversial that some kind of argument from undetectability can be run against thick substantivalism. The question is whether the argument makes essential appeal to the assumption that there are distinct worlds like W and d (W), for thick substantivalists who are anti-haecceitists will reject the assumption. The answer is no: like the argument from redundancy, the argument from undetectability is not essentially a modal argument and makes no contentious modal assumptions. To see this, return to the argument from undetectability in the case of NGT. Imagine you were using it to try to convince a Spinozist who

144 / Shamik Dasgupta

had recently endorsed Newtonianism to give up the latter view. And suppose you motivated (U1) with the modal argument described above, in which you assumed that there are distinct worlds like W and v (W). She would of course deny the modal assumption. But does that mean that the argument from undetectability has no force against her? Surely not. For in denying that there are distinct worlds W and v (W), the Spinozist makes a bare modal claim, a claim that follows not from her views about the metaphysics of space per se but rather from independent views about modality. As a Newtonian, she still believes that there are facts about the absolute velocities of things over and above the relationships between bodies in the sense of not being grounded in those relationships. And surely we would not take her quirky modal views to imply thatsurprise surprise!those facts about the absolute velocities of things are revealed to us in experience after all! So the Spinozist should agree that absolute velocity is undetectable, even if the modal argument fails to establish the point. Thus, the Spinozists curious bare modal views intuitively have nothing to do with the question of whether absolute velocity is detectable or not. How then can (U1) be established without the modal assumption? The obvious move is to use models M and v (M ) rather than worlds W and v (W). But what would those models be taken to represent? As before, I do not want to recommend any fixed template. One approach would be to take parts of the models to represent observable features, and then say that two models are indiscernible if they agree on what they represent vis a vis the observables. One could then take the models themselves to represent counterfactual truths. On this approach, the fact that v (M ) is indiscernible from M and is a model of NGT whenever M is would be taken to represent the fact that if the individualistic facts were to differ in a certain way, any experiment devised to reveal them to us would yield an indiscernible outcome. Alternatively, one could take the models to represent patterns of causal dependence, in which case these facts about models would represent the fact that the observable outcomes of experiments do not causally depend on individualistic facts concerning which regions of the manifold play which qualitative role. This is not to deny that there is a certain vividness to running the argument for (U1) in modal terms, describing the boosted possible worlds and seeing in ones mind that they are indistinguishable. But this modal description is not essential to the argument. The point, of course, is that what goes for (U1) goes for (U1*). True, if the thick substantivalist is also a Spinozist or an anti-haecceitist, she will deny the assumption I made in my modal presentation of the argument, namely that there are distinct worlds like W and d (W) that differ only in a diffeomorphic shift. But in denying this the thick substantivalist makes a bare modal claim that follows not from her view about the metaphysics of the manifold but rather from independent modal views. As a thick substantivalist, the Spinozist agrees that there are individualistic facts concerning which particular regions of the manifold play which qualitative role that are over and above the qualitative

The Bare Necessities / 145

facts in the sense that the former are not grounded in the latter. And surely we would not take her quirky modal views to show thatsurprise surprisethose individualistic facts are revealed to us in experience after all! The same goes for the anti-haecceitist. Thus, the bare modal views endorsed by the Spinozist and the anti-haecceitist have nothing to do with the question of whether thick individualistic facts are detectable or not. Even the anti-haecceitists discussed in the last section, then, should agree that the thick individualistic facts are undetectable. Or, more precisely, if they do not, this should not be because of their bare modal views.

9. Easy Detection?
It is worth saying a little more about the assumption that M and d (M ) are indiscernible models. When the argument is presented modally this is the assumption that W and d (W) are indiscernible. Even if she grants that there are such worlds the thick substantivalist might object to the claim that they are indiscernible. The issue here is analogous to one that arises in the case of NGT. Just as absolute velocity would be undetectable if the laws of NGT were true and complete, so it has been traditionally thought that our position in Newtonian space would be undetectable too. The argument has traditionally been thought to be precisely analogous: if run modally, one would note that possible worlds which differ only in a rigid spatial shift of all matter, say, 3 feet to the right are indiscernible, and also that the shifted world is an NGT world if the original world is. By the same reasoning as was rehearsed above, one would then argue that there is no physically possible experiment which could be used to reveal our position in Newtonian space. That is the traditional idea. But Maudlin has argued that while the argument in the case of velocity is sound, its analogue in the case of position is not:
Various positional states of the universe as a whole are possible: It could be created so my desk is here, or three meters north of here, or 888 meters from here in the direction from Earth to Betelgeuse, and so on. Which is the actual state of the world? . . . the answer is easy: In its actual state, my desk is here, not three meters north of here or anywhere else . . . To even formulate the appropriate question . . . one must indexically pick out a spatiotemporal location, and it is then no great trick to observe what material body that location actually contains.57

Maudlins idea is this. In the case of velocity, we can ask the question Am I in a state of absolute rest or a state of uniform motion? and the argument outlined in the last section purports to show that according to NGT it is physically impossible to give a reliable answer. Maudlins observation is that in the case of location there is no analogous question: given the resources we have by which

146 / Shamik Dasgupta

to ask the question of where we are in Newtonian space, the only questions we can ask are questions we can readily settle. The point might be put like this. Call a question about our absolute location or velocity open if it cannot be reliably answered by verifying facts about the relative positions or relative velocities of material bodies. Then Maudlins point is that we have the conceptual resources to ask an open question about or velocity through Newtonian space but not about our location in Newtonian space. It is clear how this carries over the case of GTR, for we lack the conceptual resources to ask open questions about which individualistic facts about the manifold obtain in just the same way that we cannot ask open questions about our position in Newtonian space. Let us grant Maudlin all this. What follows? Maudlin concludes that our location in Newtonian space is detectable after all. After all, if there are no unanswerable questions about our position in space, in what sense could our position be said to be undetectable? On this view, our epistemic situation vis a vis our location in space is much better that our epistemic position vis a vis our velocity through space. But Maudlins conclusion does not follow from what we are granting him. For another conclusion one might draw is that our position in space is undetectable, and moreover that we cannot even express what it is that we cannot detect. On this view, our epistemic situation vis a vis our location in space is much worse than our epistemic position vis a vis our velocity through space, since in the former case we cannot even formulate questions about what it is we cannot detect! Both conclusions are entirely consistent with Maudlins insight, which I am granting for the sake of argument. But which is correct? Surprisingly, Maudlin himself did not offer any reasons for drawing his conclusion. And there are reasons to reject it. For one thing, when one remembers that there are uncountably many shifted worlds that would all look and feel and taste exactly the same as the actual world, there is a clear feeling that our location in space is therefore in some sense beyond our epistemic grasp. And (speaking for myself) this feeling is not dissipated by being told that one cannot formulate an unanswerable question about ones position in space. Indeed, the mere fact that one can answer questions such as Am I here or 3 feet to the right of here seems like a cheap conceptual trick which does not speak to the intuitive question of where in absolute space one is. Moreover, Maudlins view implies that whether something is detectable depends on factors that are, intuitively, entirely irrelevant to the matter. For on his view, whether our location in Newtonian space is detectable depends on whether or not we can ask open questions. So, for example, suppose that God had a favorite point in space. Then we would be able to ask open questions about our location in space, for example Am I 3 feet or 6 feet from Gods favorite point? On Maudlins view, it would then turn out that location is undetectable after all. So on Maudlins view, whether absolute position is detectable or not depends on whether God had a favorite point. And that seems clearly false: whether or not God has a favorite point is surely irrelevant to my epistemic situation vis a vis our position in space!58

The Bare Necessities / 147

Maudlin is right that there is an interesting disanalogy between absolute velocity and absolute position: we can ask open questions about the former but not the latter. But I believe that he was wrong to conclude that our location in absolute space is detectable. Similarly, even if we cannot ask open questions about which regions of the manifold play which qualitative role, it would be wrong to conclude that those facts are detectable.59

10. Towards a New Metaphysics of Manifolds

I have argued that the traditional Occamist arguments used to motivate rejecting Newtonianism in favor of the Galilean view of spacetime also motivate rejecting thick in favor of thin substantivalism. I also tried to show that those arguments make no essential appeal to modal assumptions that a thick substantivalist might deny by endorsing anti-haecceitism. But in Section 5 I argued that a thin substantivalist theory has not yet been properly formulated. That is my reason for claiming that substantivalisms status as the received view is premature. I have focused on substantivalism in GTR, but the same point applies to substantivalism in classical contexts. For example, similar Occamist arguments should motivate our grounding individualistic facts about regions in Newtonian space or Galilean space-time qualitatively. The idea is that if individualistic facts about which particular region a given body is located at in Newtonian space or Galilean space-time were ungrounded, they would be redundant and undetectable. Rather than appealing to diffeomorphic shifts, these arguments would appeal to the rigid spatial shifts described earlier. Now it is often thought that these Occamist arguments assume that rigid spatial shifts are metaphysically possible and that they can therefore be refuted by endorsing anti-haecceitism.60 But if I am right this is a mistake, as the arguments make no essential appeal to modal assumptions after all. How then might we develop a thin substantival theory? I will end by suggesting how the task might be approached. For reasons that will become apparent I do not yet have a completed theory to offer, but I can offer a toy model that might be a helpful guide to the sort of theory we should be looking for. The thin substantivalist claims that individualistic facts about the manifold are grounded qualitatively. As I said in Section 5, developing this idea requires that we (1) clearly articulate what the underlying qualitative facts are like, and (2) show that they are sufficient to explain (in the metaphysical sense) individualistic facts about the manifold. We saw that one theory that at least attempts task (1) is the traditional bundle theory on which all facts about the world are ultimately grounded in facts of the form F, G, . . . are compresent

148 / Shamik Dasgupta

where F, G, . . . are intrinsic, monadic, qualitative properties and compresent is a primitive, plural predicate. Thus, consider a given individual a and suppose we would ordinarily say that it has qualitative properties F, G, . . . . The bundle theorist says that the underlying qualitative fact about this situation is that F, G, . . . are compresent. She is usually said to then identify the individual a with the set of compresent properties, but for now I am just concerned with her account of what the underlying qualitative facts are and not how she makes sense of individuals. So the thin substantivalist might try being a bundle theorist about the regions of the manifold. But the traditional bundle theory is not well suited to her needs. For one thing, it is highly unclear what the intrinsic, monadic properties of regions of the manifold could be.61 But even if this were clarified, the proposal cannot make sense of symmetric physical systems. In order to make progress, let us be clear on why it cannot and why this is a problem. To see why it cannot make sense of such systems, consider the simple Max Black world, a world containing just two spheres of iron 5 feet from each other which share all their intrinsic, monadic, qualitative properties (they are exactly the same mass, color, shape, etc). How can the bundle theorist characterize this world? Suppose that the Fs are the intrinsic, monadic qualitative properties of each sphere. Then she can only say that the Fs are compresent (to account for the first sphere), and then say that the Fs are compresent (to account for the second). But saying something twice over does not make it true twice over! Her description of this world is therefore the same as her description of a world with just one of the spheres. Returning to the case of GTR, if we tried to be a bundle theorist about regions of the manifold we would not be able to make sense of symmetric GTR worlds in which distinct regions of the manifold play the same qualitative role.62 Why is this a problem? It might be said that it is intuitively possible for there to be such worlds, in which case the charge is that bundle theory runs up against intuition. But a better charge is that ruling out such worlds was never motivated by the Occamist arguments. The lesson of those arguments, remember, is that we should ground facts about the manifold qualitatively. But symmetric worlds can be characterized in purely qualitative terms. For example, the Max Black world can be characterized as a world in which (1) (x) (y) (Fx & Fy & Rxy & Ryx & x = y) where F expresses the collection of intrinsic, monadic properties of each sphere and R expresses the distance between them.63 Thus, by failing to make sense of symmetric worlds the bundle theory implies that certain qualitative states of affairs are impossible, and this goes beyond what the Occamist arguments purport to show. Indeed, for all we know on the basis of those arguments, we live in a symmetric world ourselves! Endorsing the bundle theory on the basis of those Occamist arguments is therefore akin to responding to the Occamist argument

The Bare Necessities / 149

in NGT by rejecting Newtonianism in favor of a metaphysics of motion that can make no sense of certain patterns of relative velocities. Such a metaphysics might succeed in dispensing with facts about absolute velocities, but it would be a perverse over-reaction to the Occamist arguments. There is a general moral here. Consider a language PL of predicate logic with identity but no constants, in which every predicate expresses a qualitative property or relation.64 Every fact that can be expressed in PL is a purely qualitative fact. So our new metaphysics of the manifold should be capable of making sense of any physical system that we can express in PL; if it cannot, it is an over-reaction.65 In other work I have described an approach that tries to do exactly this.66 The idea is that the bundle theorist went wrong by trying to construct the underlying qualitative description of the world by considering each individual one by one. It is this localist strategy that means she cannot account for symmetrical systems like the Max Black world in which more than one individual plays the same qualitative role. But this localist strategy was not enforced. Starting with an ontology of qualitative properties, we could simply try to construct the qualitative nature of the entire world all at once, as it were, rather than considering each individual in turn. It turns out that this holistic approach can then account for any physical system described in PL, including symmetric ones.

11. An Algebraic Approach

This globalist approach can be developed in a number of ways, but elsewhere I presented one toy model.67 We start with a domain D of simple qualitative properties and relations each of a given adicity greater or equal to 1, denoted with terms of the form Pn (I will call them all properties, for short). We then construct more complex properties out of this basis. To describe these more complex properties we introduce six term functors: &, , c, p, , and . For example, if R1 is the 1-place property of being red and S1 is the 1-place property of being square, then R1 is the 1-place property of being not red and (R1 & S1 ) is the 1-place property of being red and square. The functors and are permutative functors. For example, if L2 is the 2-place property of loving, which we ordinarily think of as being instantiated by x and y iff x loves y, then L2 is the 2-place property of being loved, which we ordinarily think of as being instantiated by x and y iff y loves x. (L2 & L2 ) is then the 2-place property of loving unrequitedly. We also let our domain D include the simple property I of identity, i.e. the 2-place property we ordinarily understand as holding between x and y if and only if x is identical to y. The functor c is known as a cropping functor. If L2 is the 2-place property of loving, then cL2 is the 1-place property of being loved by someone, the property we ordinarily think of as being instantiated by an individual x iff someone loves x. Roughly speaking, c takes an n-place property Pn and yields the

150 / Shamik Dasgupta

(n-1)-place property cPn we would ordinarily think of as holding of x 2 , . . . x n iff there is something x 1 such that Pn holds of x 1 , x 2 , . . . x n . I leave a more complete description of the term functors to a footnote.68 Now, applying c again to cL2 gives us a 0-place property, i.e. a state of affairs that we would ordinarily describe as the state of someones loving someone. To say what the world is like we say which states of affairs obtain. So on this view the underlying qualitative facts are of the form P0 obtains where P0 is a 0-place property constructed out of the simple, qualitative properties by successive applications of the term functors. Now let AG be the language consisting of terms that denoting properties, the six term-functors, the predicate x obtains, and the normal sentential connectives. One can define a consequence relation on the sentences of the language AG. Now, let AG be the set of equivalence classes of sentences of AG (under the relation of logical equivalence in AG). And let PL be the set of equivalence classes of sentences of PL (under the relation of logical equivalence in PL). Then one can prove the central theorem of interest, namely that there exists an isomorphism between AG and PL that preserves logical equivalence.69 This means that we can make sense of any physical system that we can express in PL, as desired. For example, if the Max Black world is characterized in PL to be a world in which (1) (x) (y) (Fx & Fy & Rxy & Ryx & x = y) then we can take the basic qualitative fact about this world, which will ground the individualistic facts, to be the fact that (1 ) cc(F1 & pF1 & R2 & R2 & I2 ) obtains. By the central theorem just mentioned, (1*) is logically inequivalent to a sentence of AG that would characterize a world in which there is just one sphere. Thus, unlike the Bundle Theory we can make sense of the Max Black world. Now the spheres in a Max Black world are weakly discernible (in the sense defined in Section 5) since they stand in the irreflexive relation R. But the current approach can also make sense of symmetric systems with weakly indiscernible spheres. For suppose that R were reflexive. The resulting system can be described in PL just by adding a couple of conjuncts to the matrix in (1), so by the central theorem mentioned above there is a corresponding statement of the qualitative nature of the system in AG. As I said, this is one reason why I find the recent literature on weak discernibility a red herring. For if I am right we can ground individualistic facts qualitatively regardless of whether objects are weakly discernible.

The Bare Necessities / 151

The key to all this is our holistic approach. For in constructing the state of affairs that is said to obtain in (1*), there was no attempt to first construct the qualitative nature of one sphere and then construct the qualitative nature of the second, as the Bundle Theorist would. Rather, we constructed a worldstate, a state of affairs that completely characterizes the entire world all at once; in this case a state that we would ordinarily describe as the state of there being exactly two iron spheres.70 The fact (1*), that that world-state obtains, is therefore the most fundamental fact about that world. Other qualitative facts, such as that there is at least one iron sphere, are then grounded in the fact that this world-state obtains. So far we have only described an attempt at task (1), i.e. an attempt to state what the underlying qualitative facts are like. What about task (2), the task of saying how they are to explain the individualistic facts? It is not obvious how to do this. For note that if a given individualistic fact has any qualitative ground, it follows from the holistic approach just outlined and the transitivity of ground that it is ultimately grounded in the fact that the entire world-state obtains. I argue elsewhere that that is an undesirable consequence.71 Very briefly, the worry is that it is implausible that the qualitative nature of the entire cosmos must be cited in order to explain, say, my own meager existence. The solution, I argue, is to recognize that ground is irreducibly plural, in the sense that sometimes a collection of facts Y are grounded in another collection X even though no Y taken on its own is grounded in anything.72 We may then say that the individualistic facts together are grounded in the fact that the qualitative world-state obtains even though no individualistic fact has a qualitative ground when taken on its own. On this approach we avoid the undesirable consequence that my existence is grounded in the qualitative nature of the entire cosmos since we deny that my existence (taken alone) has a qualitative ground in the first place! Still, we remain faithful to the idea that the world is fundamentally qualitative since the individualistic facts are together grounded qualitatively. The result is a radically structuralist view of individuals. On this view an individualistic fact such as my existence is not fundamental, but nor is there any qualitative matter of fact that accounts for itnot even a highly relational or extrinsic one. My existence only comes into view, as it were, when considered alongside all individualistic facts about all individuals, for it is only when those facts are taken together that they can be explained in qualitative terms. That at least is the view; in other work I have tried to argue in its favor.73 So far I have just described a general framework for grounding individualistic facts qualitatively. How might the thin substantivalist make use of it? The natural idea would be to let the domain D of simple properties be the properties and relations that the thick substantivalist believes to hold of the regions of the manifold. They might be topological properties, space-time intervals, the properties of points that are taken to constitute the physical mass-energy field, or whatever. The idea would then be to use the term functors described above to construct a world-state W0 that characterizes the entire qualitative nature of

152 / Shamik Dasgupta

the manifold without mentioning any specific individuals. And she may then use the fact that W0 obtains to ground all other facts about the system. In the first instance, she will say that the plurality of all individualistic facts about the manifold of the form (r 1 , r 2 , . . .) are, together, grounded in the fact that W0 obtains even though no one of them has a ground on its own.74 By the principle of Necessitation (modified slightly given our new pluralist logical form of ground) it then follows that any two worlds in which W0 obtains agree on which individualistic facts obtain. So this view about what grounds individualistic facts about the manifold implies, given the principle of Necessitation, that diffeomorphic shifts are impossible. That is why it counts a thin substantivalist view, as opposed to a thick substantivalist view on which the impossibility of diffeomorphic shifts is a bare modal claim. Note that we do not deny that there are regions of the manifold, we just (plurally) ground facts about them in qualitative terms. Since there are such regions, and since we are used to thinking about the world in terms of it containing a domain of individuals propertied and related in various ways, it is no surprise that we tend to represent GTR systems with models M = (M , g, T ), where M is a manifold of individuals. This method of representation is not incorrect, so long as it makes no claim to represent GTR systems in their most fundamental respects. But since the fundamental facts about these systems are qualitative, it makes sense to ignore differences between M and d (M ), for any diffeomorphism d . Thus, if one takes the space of models to represent the space of physically possible worlds, it makes sense to then interpret M and d (M ) as representing the same possible world.

12. Conclusion
The above is just a toy model, not a completed thin substantivalist theory. To take just one issue, the equivalence theorem stated above only says that AG is equivalent to PL. But PL is a finitary language, and the thin substantivalist will likely need to construct a more sophisticated algebra of states of affairs for which there is an equivalence to an infinitary first-order language. Second-order languages might also be necessary. Still, my point here is just that insofar as the Occamist arguments motivate a rejection of thick in favor of thin substantivalism, it is precisely this sort of work that the substantivalist needs to do. The result will be a substantival view that stands to thick substantivalism just as the Galilean view of spacetime stands to Newtonianism: it will dispense with the kinds of facts that were shown by the Occamist arguments to be redundant and/or undetectable.

The Bare Necessities / 153

One might wonder whether a thin substantivalist view along the above lines ultimately deserves the title substantivalism. After all, any reference to the manifold on this approach will ultimately be grounded in facts that do not concern the manifold at allat least, not when it is conceived of as an individual bearer of properties. This is to some extent a verbal issue, and the meaning we pre-theoretically attach to the word substantivalim might not be clear enough to yield a determinate answer. I am inclined to count it as substantivalism because it is constructed by putting thick substantivalism through a general recipe that removes reference to underlying individuals from any theory, but I am not attached to the term. In any event, if this is substantivalism, it is substantivalism exorcized of redundant structure and stripped to its bare necessities.

Notes
Many thanks to David Baker, John Morrison, Jill North and David Plunkett for their extremely helpful feedback on earlier drafts of this paper. 1. The way this debate is characterized in the literatureand the way I characterize it hereis not obviously faithful to the views of these historical figures. But our concern here is not with historical accuracy so I will not elaborate on this point. 2. Sklar 1974, p. 161; my emphasis. 3. Earman and Norton claim that affirming this sort of possibility is the acid test of substantivalism (see their 1987, p. 521). Belot agrees in his 2000. 4. I am told that this is something like Spinozas actual view. But I am not qualified to defend this interpretive point, so I use the term Spinozist as a label for a view which may or may not have been Spinozas. 5. For discussions of ground see Fine 2001 and MS, and Rosen 2010. For some recent literature on ontological dependence see Fine 1995, Lowe, 2006 and Koslicki MS. For a notion of truth-making that is understood non-modally see Cameron MS. And see Sider 2011 for a development of metaphysical semantics. 6. I choose ground because I believe that it does a better job at capturing many issues in metaphysics than the other notions, but I will not argue this here. 7. This is how I interpret Fine as using the term in his influential papers 2001 and MS. Not all recent philosophers use the term in the same way; see Schaffer 2009 for a different idea under the same name. 8. This is how Fine characterizes the logical form in his 2001 and MS. is taken to be a list and not a conjunction so as to allow one to make sense of the plausible view that a conjunction is grounded in its conjuncts without it collapsing into the view that the conjunction grounds itself. 9. This is how Rosen characterizes the logical form in his 2010. 10. Transitivity: if the Xs ground Y* and the Ys along with Y* ground Z, then the Xs along with the Ys ground Z. Irreflexivity: for any X, there are no Ys such that X along with the Ys ground X. Strictly speaking I consider these principles to be open for dispute, but it will smooth the presentation that follows to presume them.

154 / Shamik Dasgupta 11. This statement of the principle quantifies over facts, so it breaks down in modal frameworks in which the notion of a facts obtaining in a world is not well defined (e.g. Lewisian counterpart theory). So it is sometimes better to revert to our official way of talking of ground as a sentential operator and state Necessitation as the scheme If S because , then it is metaphysically necessary that if then S. 12. 13. 14. where is the conjunction of the sentences in . But for our purposes we can largely ignore this complication. This follows from my original statement of the view by the transitivity and irreflexivity of ground. Though this allows that the fact that m 1 is 3 meters from m 2 is grounded in different facts about space in different possible worlds. Moreover, it follows from what I take to be the correct view about what grounds the grounding facts that the corresponding contingent claims are false (see Dasgupta MSa). But there is no space to discuss this here. This terminology is perhaps a little misleading since the second theorists modal view follows from something, namely her views about religion. But our topic here is the nature of space, and the point is that her modal view does not follow from her views about that. Precisely how these notions are defined will not matter to us here. See Geroch 1978 for details. Hoefer 1996, p. 5; my emphasis. One might have a different account of each field, for example that g represents space-time intervals between regions while T represents an extended object with parts. And of course this does not exhaust the options. Maudlin 2007 intriguingly suggests that the correct metaphysics of fields will place them in some new category as yet undreamt of by armchair metaphysicians. By placing so few constraints on I take the last paragraph to be consistent with Maudlins suggestion. Again, how one thinks about these facts depends on ones metaphysics of the mass-energy field. If one thinks of it as an extended object with parts, these facts can be taken to consist of space-time relationships between its parts. Alternatively, if one thinks of it as an instantiation of properties by the manifold, these facts will consist of facts about the space-time intervals between different instantiations of those properties, for example that property F and property G are exmplified a certain space-time interval apart. Again, this is implied by my original statement of the view given the transitivity and irreflexivity of ground. It is tempting to interpret the debate between so-called manifold substantivalists and metric substantivalists as a disagreement on precisely this question. For a defense of the former view, see Earman 1989 chapter 9. For a defense of the latter view see Maudlin 1988 and Hoefer 1996. (I would add parenthetically that I doubt the question of what counts as spacetime in GTR has a determinate answer, since I doubt that the meaning we attach to spacetime is determinate enough to withstand such close scrutiny.) See Earman 1986 for counterexamples involving so-called space-invaders. See Norton 2008 for another kind of counterexample.

15.

16. 17. 18.

19.

20. 21.

22.

The Bare Necessities / 155 23. This property of GTR is sometimes known as General Covariance. However, there is some controversy over how General Covariance should be understood, so to remain neutral on that issue I use the term diffeomorphism invariance stipulatively to label the property I am interested in. See Rynasiewicz 1999 for a good review of the literature on General Covariance. 24. In their 1987, their premise that substantivalists must deny a principle they call Leibnizian Equivalence is equivalent to my statement of the premise above. 25. I should add that I do not count the property of Socratizing as qualitative. I do not count the property of having a particularized trope as qualitative either. Thus, even if the property of being green is qualitative, the particular greenness of this apple is a particularized trope and the property of having this particular greenness is then not qualitative since it concerns a particular individual, namely this particular greenness. 26. Earman and Norton 1987, p. 524. 27. Well, (H2) is not justified by diffeomorphism invariance alone. For diffeomophism invariance was defined to be a property of GTR that concerns its models, while (H2) is a statement about worlds. To bridge the gap, we need the assumption that if M is a model of GTR and represents a possible world W, then W is a world in which GTR is true. But this is granted by all parties to the debate so I will not discuss it further. 28. See Melia 1999 and Skow 2005 for a motivation and defense of this approach. 29. Thus I take it that Caulton and Butterfield were speaking loosely when they said the correct response is to consider diffeomorphically related histories [i.e. models] as representatives of the same physical state of affairs (MSb, p. 9). What they surely meant to say is that, in addition, the correct response denies that there are distinct possible worlds like W and d (W). Without the addition, their claim about the representational properties of models is consistent with the claim that GTR is indeterministic. 30. I point this out because there was a lively debate in the literature as to whether the substantivalist should, when denying (H1), say that M and d (M ) represent the same possible world or that just one of them represents a possible world (see Brighouse 1994 and Maidens 1992 for a defense of the first view, and Butterfield 1988 and Maudlin 1988 for a defense of the latter view). But this issue is largely irrelevant to the matter, since it is (H1*) rather than (H1) that does all the work in the argument and (H1*) says nothing of the representational properties of models at all. 31. Pooley 2006, p. 101. Theorists who respond in this way include Brighouse 1994, Butterfield 1988, Caulton and Butterfield MSb, Hoefer 1996, Maidens 1992, Maudlin 1988 and Pooley 2006. This response constitutes a position sometimes known as sophisticated substantivalism, though that term is also used to describe the conjunction of denying that there are such worlds as W and d (W) and asserting that M and d (M ) represent the same possible world. For reasons just mentioned I do not consider the second conjunct particularly important. 32. Brighouse 1994. 33. Finer distinctions are of course possible. One might wish to replace the iff in the definition of an individuating qualitative profile with only if, since that is enough to imply that diffeomorphic shifts are impossible. Maudlin 1988 endorses a view similar to that which would result by replacing iff with if,

156 / Shamik Dasgupta though the details of his view are subtle and there is no space to discuss them here. Brighouse 1994, p. 122. I have in mind Lewis idea that counterpart theory can often mimick the sort of modal claims that the haecceitist makes. See Lewis 1986, Chapter 4. Thus the language I have in mind does not contain substitutes for constants such as x Socratizes. It is tempting to read some structuralists as implicitly having this sort of thesis in mind, though it is hard to say for sure since they are typically wary of using the notion of ground in the way I do. Specifically: that if (x)Fx, then for any individual a that is F, the fact that (x)Fx is grounded in the fact that a is F. Hoeffer 1996, p. 4 (his emphasis). For example, Hoefer sometimes sounds like he is proposing the purely modal thesis that some modal questions are unintelligible. Baker 2010, p. 1163. Of course, I am being a little unfair here since both these authors go on to expand on what I quoted above. I do not believe that what they say amounts to an attempt at task (1), but I leave this for the reader to judge for herself. Similar remarks apply to Caulton and Butterfield MSa when they express their view QII as the denial that there are haecceitistic properties, for this is a purely negative thesis and does not tell us what the underlying qualitative facts are like. Ladyman and Ross 2007, pp. 143 and 144 respectively. Steven French once suggested this line to me in conversation, though I do not know if he would defend it. This might be what Caulton and Butterfield had in mind when they wrote that they consider diffeomorphically related histories [i.e. models] as representatives of the same physical state of affairs (MSb, p. 9). See Saunders 2003b and 2003a. Ladyman and Ross explore this approach in their 2007, as does Ladyman in his 2007. The two arguments here correspond to two different approaches to thinking about the how symmetry in general bears on metaphysics. I discuss the two approaches at length in Dasgupta MSc. We can perhaps see these arguments as distant descendents of Leibnizs arguments that he described in the LeibnizClark correspondence, but I will not trace out this historical point here. For more on this see Pooley MS. This is analogous to our implicit assumption in the case of GTR that both the substantivalist and the anti-substantivalist use GTR to refer to the same set of laws. In both cases the discussion should really be made without this assumption, but it is somewhat cumbersome to do so. See, for example, Baker 2010, p. 4. Sklar made a similar point in his 1974, p. 180. For example, (R2) is implied by Earmans more general principle of using symmetry considerations to derive metaphysical conclusions. As he puts it on p. 46 of his 1989,

34. 35. 36. 37.

38. 39. 40. 41. 42.

43. 44.

45. 46. 47.

48.

49. 50. 51.

The Bare Necessities / 157 The motivation [for dispensing with features that are altered by the symmetries of our laws] derives from combining a particular conception of the main function of laws of motion with an argument that makes use of Occams razor. Laws of motion, at least in so far as they relate to particles, serve to pick out a class of allowable or dynamically possible trajectories. If [there are features that are altered by the symmetries of our laws], the same set of trajectories can be picked out by the laws working in the setting of a weaker space-time structure. The theory that [posits these features] is thus using more space-time structure than is needed to support the laws. More recently, North defended a principle like (R2) when she advised that we should not posit structure beyond that which is indicated by the fundamental dynamical laws (North 2009, p. 9). This is not to presuppose a Humean conception of laws. The question of Humeanism vs Anti-Humeanism is the question of what grounds the fact that those equations are laws, whereas here I am just asking what grounds the fact that those equations are true. Thanks to David Baker for explaining this worry to me. Earman 1989, p. 48. Here I will just present the argument in barest outline. Roberts presents perhaps the most developed version of it in his 2008. I have also heard oral presentations of it in seminars given by David Albert and Tim Maudlin. I have presented versions of it in Dasgupta 2009 and 2013. Earman and Norton presented a similar argument in their 1987 that they called The Verificationist Dilemma, but the argument was largely ignored in the subsequent literature. Perhaps one explanation for that is the misleading title they gave the argument; as emphasized above, the argument as I understand it here does not appeal to verificationist principles at all. Maudlin 1993, p. 190. Thanks to Gideon Rosen for a helpful conversation on this point. While I do not think that Maudlins insight has the consequences he thinks it does, I do believe it has profound consequences about the nature of inquiry and ignorance that have not yet been fully appreciated. For example, according to the conclusion I draw from Maudlins insight, the sort of epistemic state we attempt to attain when engaging in an inquiry is not merely the ability to knowledgeably answer questions. Moreover, I think that Maudins insight shows that it is extremely difficult to see what sort of epistemic state we do hope to attain as a result of our inquiries, at least when we try to analyze the state in the terms normally used in the contemporary epistemology literature such as truth, justified belief, knowledge, reliability, and so on. I hope to develop these consequences of Maudlins insight in other work. Pooley explores this approach in his MS. See Maudlin 2007 for more on this issue. Parsons and McGivern (2001) offer a bundle theory about regions of the manifold according to which the properties that are said to be compresent are particularized tropes rather than repeatable properties. They advertise that

52.

53. 54. 55.

56.

57. 58. 59.

60. 61. 62.

158 / Shamik Dasgupta this approach can make sense of symmetrical systems. But even if they are right, it does not count as a thin substantival view since facts about particularized tropes are not qualitative facts (see footnote 25). The point here is not just terminological, for the kinds of Occamist arguments that motivate rejecting fundamental individualistic facts about the manifold also motivate rejecting fundamental facts about particularized tropes. One may insist that a totality clause be added to this characterization of the world, to the effect that nothing else exists and that the spheres instantiate no other properties or relations. But I ignore this complication for ease of prose. Note then that PL does not contain substitutes for constants such as x Socratizes. At this point the thin substantivalist might suggest that we ground individualistic facts about the manifold in facts expressible in PL, such as that expressed by (1). But as I argued in Section 5, it is unclear what this could mean. See Dasgupta 2009. The following few paragraphs are based on Dasgupta 2009. The approach is based on work in formal logic by Quine 1976, Kuhn 1983, and others. If F is the n-place property we ordinarily understand as holding of individuals x 1 . . . x n just in case (x 1 . . . x n ), and G is the m-place property we ordinarily understand as holding of individuals y 1 . . . y m just in case (y 1 . . . y m ), then 1. F is the n-place property we ordinarily understand as holding of x 1 . . . x n just in case it is not the case that (x 1 . . . x n ); 2. (F &G) is the max(n, m)-place property that we ordinarily understand as holding of x 1 . . . x k just in case (x 1 . . . x n ) and (x 1 . . . x m ), where k = max(n, m); 3. F is the n-place property we ordinarily understand as holding of x 1 . . . x n if and only if (x n x 1 x 2 . . . x n1 ); 4. F is the n-place property we ordinarily understand as holding of x 1 . . . x n if and only if (x 2 x 1 x 3 . . . x n ); 5. pF is the (n+1)-place property that we ordinarily understand as holding of x 1 . . . x n+1 if and only if (x 2 . . . x n x n+1 ). 6. If n is greater than or equal to 1, then cF is the (n-1)-place property that we ordinarily understand as holding of x 2 . . . x n if and only if there is something x 1 such that (x 1 . . . x n ); otherwise cF is the 0-place property F. See Kuhn 1983 for these results. Here again I am slurring over the issue that one might have wished to add a totality fact to (1), stating that there are no other objects and that the two spheres have no other properties. See Dasgupta MSb. Here I am modifying the proposed logical form of a grounding claim that I stated in Section 1. See Dasgupta MSb. She may then ground the facts expressed by PL, such as existentially quantified facts, in those individualistic facts just as we would normally do. I was not clear about this in my 2009. For given that AG is difficult to read (for most of us), and given that it is isomorphic to PL, it was convenient when discussing AG there to write in PL instead. Unfortunately, it gave some readers the impression that

63.

64. 65.

66. 67. 68.

69. 70.

71. 72. 73. 74.

The Bare Necessities / 159 my view was that those facts expressed by PL were themselves fundamental. As I hope is clear here, that is not my view.

References
Baker, D. (2010). Symmetry and the metaphysics of physics. Philosophy Compass 5, 11571166. Belot, G. (2000). Geometry and motion. British Journal for the Philosophy of Science 51, 561 595. Brighouse, C. (1994). Spacetime and holes. PSA: The Proceedings of the Biennial Meeting of the Philosophy of Science Association 1, 117125. Butterfield, J. (1988). Albert Einstein meets David Lewis. PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association 2, 6581. Cameron, R. (MS). Truthmakers. In M. Glanzberg (Ed.), The Oxford Handbook of Truth. Forthcoming. Caulton, A. and J. Butterfield. (MSa). On kinds of indiscernibility in logic and metaphysics. British Journal for the Philosophy of Science. Forthcoming. Caulton, A. and J. Butterfield (MSb). Symmetries and paraparticles as a motivation for structuralism. British Journal for the Philosophy of Science. Forthcoming. Dasgupta, S. (2009). Individuals: An essay in revisionary metaphysics. Philosophical Studies 145(1), 3567. Dasgupta, S. (2013). Absolutism vs comparativism about quantities. In Oxford Studies in Metaphysics, Volume 8. Dasgupta, S. (MSa). The status of ground. In preparation. Draft available at www.shamik.net. Dasgupta, S. (MSb). On the plurality of grounds. In preparation. Draft available at www.shamik.net. Dasgupta, S. (MSc). Symmetry as an epistemic notion. In preparation. Draft available at www.shamik.net. Earman, J. (1986). A Primer on Determinism. Springer. Earman, J. (1989). World Enough And Space-Time. Cambridge, MA: MIT Press. Earman, J. and J. Norton (1987). What price substantivalism? The hole story. British Journal for the Philosophy of Science 38, 515525. Fine, K. (1995). Ontological dependence. Proceedings of the Aristotelian Society 95, 269290. Fine, K. (2001). The question of realism. Philosophers Imprint 1, 130. Fine, K. (MS). Guide to ground. Geroch, R. (1978). General Relativity from A to B. Chicago: The University of Chicago Press. Hoefer, C. (1996). The metaphysics of space-time substantivalism. The Journal of Philosophy 93, 527. Koslicki, K. (MS). The varieties of dependence. Cambridge: CUP. Forthcoming. Kuhn, S. (1983). An axiomatization of predicate functor logic. Notre Dame Journal of Formal Logic 24(2), 23341. Ladyman, J. (2007). On the identity and diversity of objects in a structure. Proceedings of the Aristotelian Society Supplementary Volume 81, 2343. Ladyman, J. and D. Ross. (2007). Every Thing Must Go: Metaphysics Naturalized . Oxford: OUP. Lewis, D. (1986). On the Plurality of Worlds. Cambridge: Blackwell. Lowe, E. J. (2006). The Four-Category Ontology: A Metaphysical Foundation for Natural Science. Oxford: Clarendon Press. Maidens, A. (1992). Review of John Earmans World Enough and Space-Time. British Journal for the Philosophy of Science 43, 129136. Maudlin, T. (1988). The essence of space-time. PSA: The Proceedings of the Biennial Meeting of the Philosophy of Science Association, 8291.

160 / Shamik Dasgupta

Maudlin, T. (1993). Buckets of water and waves of space: Why spacetime is probably a substance. Philosophy of Science 60(2), 183203. Maudlin, T. (2007). Suggestions from physics for deep metaphysics. In The Metaphysics Within Physics. Oxford: OUP. Melia, J. (1999). Holes, haecceitism and two conceptions of determinism. British Journal for the Philosophy of Science 50, 639664. North, J. (2009). The structure of physics: A case study. Journal of Philosophy 106, 57 88. Norton, J. (2008). The dome: An unexpectedly simply failure of determinism. Philosophy of Science 75(5), 786798. Parsons, G. and P. McGivern,(2001). Can the bundle theory save substantivalism from the hole argument? Philosophy of Science 68, 358370. Pooley, O. (2006). Points, particles and structural realism. In Rickles, French, and Saatsi, (Eds.), The Structural Foundations of Quantum Gravity. Oxford: Oxford University Press. Pooley, O. (MS). Substantivalism and haecceitism. Quine, W. (1976). Algebraic logic and predicate functors. In The Ways of Paradox and Other Essays (cond Edition ed.)., pp. 283307. Cambridge MA: Harvard University Press. Roberts, J. (2008). A puzzle about laws, symmetries and measurable quantities. British Journal for the Philosophy of Science 59, 14368. Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In B. Hale and A. Hoffmann (Eds.), Modality: Metaphysics, Logic, and Epistemology. Oxford: OUP. Rynasiewicz, R. (1999). Kretschmanns analysis of covariance and relativity principles. In H. Goenner, J. Renn, J. Ritter, and T. Sauer (Eds.), The Expanding Worlds of General Relativity, pp. 431462. The Center for Einstein Studies. Saunders, S. (2003a). Indiscernibles, general covariance, and other symmetries: The case for non-reductive relationalism. In J. Renn, (Ed.), Revisiting the Foundations of Relativistic Physics. Kluwer. Saunders, S. (2003b). Physics and Leibnizs principles. In K. Brading and E. Castellani (Eds.), Symmetries in Physics: Philosophical Reflections. Cambridge: CUP. Schaffer, J. (2009). On what grounds what. In D. Manley, D. J. Chalmers, and R. Wasserman (Eds.), Metametaphysics: New Essays on the Foundations of Ontology, pp. 34784. Oxford: OUP. Sider, T. (2011). Writing the Book of the World . Oxford: OUP. Sklar, L. (1974). Space, Time and Space-Time. Los Angeles, CA: University of California Press. Skow, B. (2005). Once Upon a Space-Time. Ph.D. thesis, New York University. Teller, P. (2001). The ins and outs of counterfactual switching. Nous 35(3), 365393.

The Possibility of Physicalism

Shamik Dasgupta
Draft of July 2013

It has recently been suggested that many core philosophical theses theses like physicalism, normative naturalism, phenomenalism, and so onshould be understood in terms of ground.1 The rough idea is that physicalism, for example, is (at least in part) the thesis that everything is grounded in purely physical matters of fact. But Ted Sider disagrees. He claims that if one tries to formulate physicalism as a grounding thesis then it fails at the get-go. For (so formulated) it describes connections of ground between the physical and the non-physical, and he thinks that those connections will not have a purely physical ground, contra physicalism.2 Here I develop this problem and suggest a response. I agree that the connections have no purely physical ground. But this is not in conict with physicalism, I argue, because physicalism only requires the restricted claim that (roughly speaking) everything except those connections has a physical ground. I claim that the restriction is reasonable and not ad hoc because the connections have a special status: they are not apt for being grounded. It is not that the question of what grounds them is well taken and the answer is nothing, it is rather that the question of what grounds them does not legitimately arise in the rst place. The connections fall outside the proper scope of ground, so to speak. Pictorially, one might think of physicalism as a multi-story building, with physical facts on the 1st oor, chemical facts on the 2nd oor, and so on. Then my view is that the ungrounded connections between facts on the 1st oor and those on higher oors are not part of the building itself but are rather the scaffolding around which the building is built. That at least is the picturein what follows I develop it into a thesis.

Physicalism as a grounding thesis

Let us start by clarifying the notion of ground. As I use the term, ground is a purely explanatory notion: to say that some facts ground another is
1 See 2 See

Fine [9], Rosen [18], and Schaffer [19]. Sider [23], chapter 8. This is not quite how he raises the problem; see note 8.

just to say that the former explain the latter, in a particular sense of explain.3 To illustrate, imagine that you are attending a philosophy conference, and imagine that you are asked why there is a conference occurring. A causal explanation might describe a sequence of events over the previous year that led up to the conference: someone thought that a meeting of minds would be valuable, sent invitations, the recipients accepted, and so on. But another explanation would try to say what it is about the event that makes it count as a conference in the rst place. Regardless of what caused the conference, someone in search of this second explanation recognizes that conferences are not sui generis and that there must be something about the event in virtue of which it counts as being a conference. Presumably the answer is that it is a conference in virtue of how its participants are acting, for example that some are giving talks, others are asking questions, and so on. Following Fine I refer to this as a grounding explanation, or a statement of what grounds there being a conference. Or imagine asking why there is a table in the room. A causal explanation might be that someone recently carried it in. But another explanation recognizes that tables are not sui generis items and that if there is a table in the room then that must be in virtue of various other facts, for example that some particles are arranged table-wise. This would again be a statement of what grounds the fact that there is a table in the room. We should distinguish between full and partial explanations. One conversation might be part of a (grounding) explanation of why the event was a conference, but it does not fully explain it. When I say that some facts ground another, I mean that the former fully explain the latter. Why might one think that theses like physicalism are grounding theses? Well, what is physicalism? Not just physicalism about the mind, but physicalism period: what kind of a thesis is it? A rough physicalist picture is that at some basic level the world is constituted wholly out of physical stuff, and that everything elsefootball matches, string quartets, consciousness, valuesomehow arises out of that physical stuff, or (to use other locutions) that everything else is xed by or determined by or is nothing over and above that physical stuff. As the metaphor goes, all God had to do when making the world is make the physical stuff, and then Her job was done. It may be that physicalism also requires more than this. But something like this picture is a necessary condition on physicalism.4 But what does this talk of determining or xing amount to? What does it mean to say that a football match arises out of various physical goings-on? The recent interest in ground stems, I think, from the idea that we should understand this picture in terms of ground, as the thesis that
I follow Fines use of the term in his [9] and [10]. may dispute exactly what kind of physical stuff constitutes the basic level: space-time? matter? elds? laws?. But I bracket those issues here.
4 Physicalists 3 Here

facts about football matches and string quartets and consciousness and so on are grounded in purely physical facts. The same goes for other theses such as normative naturalism, phenomenalism, and so on. For the exciting idea behind all theses is that at some basic level the world is constituted by a relatively sparse basisnatural or phenomenal factsand that this sparse basis gives rise to phenomena such as normativity or the external world (respectively). In this sense they are all ampliative theses, theses to the effect that a relatively sparse basis is sufcient to give rise to these striking phenomena. But again we must ask what giving rise to means, and one answer is that it is best understood in terms of ground: that natural facts ground normative ones, and that the external world is grounded in patterns of sense data (respectively). There are of course alternative answers. One might try formulating physicalism as an identity thesis, the thesis that every thing (event, fact) is identical to a physical thing (event, fact). Or as a theses of analysis, that all truths are logically implied by purely physical truths and analytic denitions of non-physical terms. Or as a supervenience thesis, that all properties supervene on physical properties. No doubt there are other formulations too. So the interest in ground stems largely, I think, from the idea that the grounding formulation is better than these alternatives.5 I will not rehearse the reasons to think that the grounding formulation is superior.6 Instead I will discuss a reason to think that, whatever its virtues, it is ultimately unusable.

The impossibility of physicalism

One might argue that it is unusable because the notion of ground itself is unintelligible. Hofweber [11] has recently argued along these lines. But this is not the problem I want to discuss here, so for the sake of argument I will assume that the notion of ground is intelligible.7 The problem I want to discuss instead was raised by Ted Sider [23], but I will develop it a little differently here. In rough outline the problem is this. Suppose for reductio that the physicalist formulates her view in terms of ground. Then her view is that purely physical facts about (say) particle positions or eld values are sufcient to ground all else. So, consider the
be clear, the idea is not that a so-called physicalist who chooses an alternative formulations would not thereby count as a physicalist. Suppose someone defends the general physicalist picture but formulates it as a supervenience thesis. And suppose all her arguments would carry over to the grounding formulation. Then it would be perverse to accuse her of not being a physicalist. If the grounding formulation really is better, the charge should only be that she has formulated her view inadequately. 6 For some reasons see Horgan [12], Fine [9], Schaffer [19] and Rosen [18]. 7 Though I believe that Hofwebers arguments are pressing and deserve signicant discussion. See Raven [17] for one line of response.
5 To

fact that I am conscious. The physicalist must say that this has a purely physical ground, for example in my being in a certain physical state P. Suppose that she is right. And consider the resulting grounding fact: (C) The fact that I am in physical state P grounds the fact that I am conscious. (C) is not a purely physical fact since it is not just about particle positions or eld values or the like. So the physicalist must say that (C) is also grounded in purely physical facts. And the problem (I will argue) is that this is implausible. If it is grounded in anything, it is grounded in facts about consciousness: it is because of something about consciousness that my being in state P grounds my being conscious. So if physicalism is formulated in terms of ground, it follows that it is false. (I am talking freely of facts, and of some facts being physical and others not. I will clarify this talk below in section 3for now I beg your charity.) The same goes for other ampliative theses like normative naturalism and phenomenalism. If formulated in terms of ground, they say that the world is constituted at the bottom level by a relatively sparse basis of natural or phenomenal facts (respectively), and that this sparse basis grounds all facts about some higher level phenomena, i.e. normativity or the external world. But it turns out that the grounding factsfacts about how those higher-level phenomena are groundedare not themselves grounded in those lower-level facts. So if these ampliative theses are formulated in terms of ground, it follows that they are false. Importantly, the reason why the grounding facts are not themselves grounded in the lower-level phenomena has nothing to do with the particulars of each ampliative thesis. It has nothing to do with consciousness or normativity or the external world per se. Rather, it has to do with ground: it is because of how ground works that facts about how (say) the normative is grounded in the natural are not themselves grounded in the natural. So the problem is that when formulated in terms of ground these ampliative theses fail at the get-go, prior to any rst-order considerations about consciousness or normativity in particular; an unacceptable result. Just to be clear, the conclusion of this argument is not that physicalism or these other ampliative theses are false. The conclusion is that the grounding formulation is inadequate. Let us state argument explicitly in the case of physicalism, though any ampliative thesis would do. The initial idea is that physicalism requires that everything arise out of physical facts. Here a fact is physical iff it concerns only physical matters: the conjunction of a physical fact and the fact that I am conscious is not a physical fact. Then the initial idea is (1) Physicalism is true only if all non-physical facts arise out of physical facts. 4

And the suggestion that we understand this in terms of ground then amounts to this: (2) Y arises out of the Xs iff Y is grounded in the Xs. These two propositions imply that physicalism is true only if all nonphysical facts are grounded in physical facts. But the grounding facts themselves, like (C), are not so grounded. And not because of anything to do with the nature of consciousness, but because of the nature of ground: (3) General considerations about the nature of ground suggest that some non-physical facts are not grounded in physical facts. If (1), (2) and (3) are true then general considerations about the nature of ground suggest that physicalism is false. And this (I claim) is unacceptable. To be clear, physicalism may be false. It might be that thinking about the nature of consciousnesse.g. in conceivability arguments, knowledge arguments, and so onreveals that physicalism is false. But this should not be revealed just by thinking about ground. That is: (4) If physicalism is false, this is because of rst-order considerations (e.g. about the nature of consciousness or value), not general considerations about the nature of ground itself. This contradicts (1), (2), and (3). So the idea that physicalism is understood in terms of ground leads to a contradiction.8 One of the propositions must be rejected, but which? (2) and (4) are working hypotheses. For my question is whether we can formulate physicalism as a grounding thesis in such a way that it remains an interesting thesis of rst-order metaphysics. (2) and (4) express the hypothesis that we can.
8 The core idea behind this argument can be found in Sider [23], but my presentation here differs from Siders in some respects. Sider rst develops the idea in chapter 7 with the principle of Completeness, which says every nonfundamental truth holds in virtue of some fundamental truth. This could be used instead of (4) to make trouble for the grounding theorist. For if we assume that only physical facts are fundamental, then Completeness implies that (C) must hold in virtue of (i.e. be grounded in) in some physical facts, contra (3). But Siders notion of a fundamental truth is not dened in terms of ground. Rather, a truth is fundamental iff it is a truth involving only fundamental terms. And the notion of a fundamental term is, for Sider, a primitive. So to use Completeness instead of (4) one needs an extra notion that a grounding theorist may nd unintelligible. A virtue of my formulation of the argument is that it gets at the problem without using the extra notion. In chapter 8 (section 2) Sider gave a version of the argument without introducing the extra notion. But his argument there was only directed at the view that grounding facts like (C) are ungrounded. So he did not need a premise like (3), and so he did not need to discuss the question of whether grounding facts like (C) can be given physical grounds. In my formulation of the problem (3) is crucial and I will defend it below. Also, in that argument, instead of (4) Sider used the premise that there are no ungrounded facts about cities. In my formulation of the issue we remain neutral on specic grounding claims and instead focus on the idea that the truth or falsity of ampliative theses like physicalism should not depend merely on considerations about the nature of ground.

That leaves (1) and (3). So far I have given no reasons for (3), but I will offer some in sections 36. So to defend the idea that physicalism is a grounding thesis, I claim that we should deny (1). We must show that the initial physicalist picture does not require that all non-physical facts arise out of physical ones; in particular it does not require that (C) does. The idea is has some initial plausibility. For even if it turns out that grounding facts like (C) have no physical ground, a natural thought is that physicalism was never supposed to be concerned with that kind of fact. Still, this approach faces a signicant challenge. For it is not enough to just say that the physicalist need not give (C) a physical ground and leave it at that. Some principled reason must be given as to why (C) is special. After all, a physicalist cannot say that the fact that I am conscious has no physical ground, so why is (C) different? And it is not enough just to say that it is a grounding fact, or that it is a fact of metaphysics, for (putting aside the question of what metaphysics means) the question remains as to why this means that the physicalist can legitimately ignore it. The point might be put as follows. Physicalism is sometimes expressed as the view that the fundamental facts are all physical; and so if facts about the connection between mind and body (like (C)) are ungrounded this might be thought consistent with physicalism so long as the connections are not fundamental. But what does it mean to be fundamental? If one works in terms of ground the natural idea is to dene a fact to be fundamental iff it is ungrounded. But then if the connections are ungrounded they are fundamental after all: their status to the physicalist would then be the same as the status of ungrounded facts about (say) particle positions. If the physicalist wants to say that (C) is ungrounded but not fundamental then she must mean something else by fundamental. And so the question is what this other notion of fundamentality could be. So if we deny (1) and say that physicalism does not require that (C) has a physical ground, we must identify some property of (C)a property that distinguishes it from (say) facts about particle positionsthat explains why it is special in this way. I will do this in sections 79. But before that it is worth asking whether this approach is enforced or whether the grounding theorist would be better off denying (3) instead.

Physical facts
(3) General considerations about the nature of ground suggest that some non-physical facts are not grounded in physical facts.

Let us turn, then, to

I believe that (3) is true. As indicated, my argument for (3) (given in sections 36) is that general considerations about the nature of ground 6

suggest that (C) has no physical ground; therefore, (3) is true. To be clear about the dialectic, I do not know whether my being conscious has a physical ground (indeed I suspect that it does not). But if it does not then it follows immediately that there is a non-physical fact without a physical ground. So the argument is that even if my being conscious has a physical ground, there is then another non-physical fact, (C), that does not (for reasons pertaining to the nature of ground). So I assume that my being conscious has a physical ground for the sake of argument. As stated, the argument assumes that there is such a thing as the fact (C). Strictly speaking this goes beyond the assumption that my being conscious has a physical ground. For one might agree that it is true that my being conscious has a physical ground but deny that this truth corresponds to a fact, and so deny that the fact (C) exists. But this is not germane to the point, for my reference to facts not essential to the argument. To see this let us be explicit about the logical form of ground. My ofcial view, which I assume here, is that a grounding claim is of the form S because where S is a sentence, is a list of sentences, and because is read in the metaphysical rather than causal sense.9 Note that on this regimentation there is no explicit reference to facts: S and are sentence positions, not variables ranging over facts. The argument can then be formulated without reference to facts as follows. We start by assuming that my being conscious is grounded in my physical state. This is to assume that there is a list of sentences expressed in purely physical vocabulary such that (Cs ) I am conscious because is true (the subscript s is for sentence, to remind us that (Cs ) is a sentence and not a fact). And our question is then whether there is a list of sentences , expressed in purely physical vocabulary, such that (I am conscious because ) because is true. And what I will argue is that the answer is no. This statement of the argument makes no reference to facts. Still, it is convenient for ease of prose to pretend that the logical form of a grounding claim is the Xs ground Y where Y is a singular variable and the Xs is a plural variable, both ranging over facts.10 On this approach we assume for convenience that (Cs )
10 This 9 This

is how Fine characterizes the logical form in his [9] and [10]. is how Rosen characterizes the logical form in his [18].

corresponds to a fact, which we call (C), and our question is then whether it has a physical ground. But the assumption that (C) exists is a mere convenience, eliminable from the ofcial formulation of the argument. This claried, I will largely talk in terms of facts for ease of prose.11 The argument also assumes that (C) is not a physical fact (by which I mean it does not concern only physical matters). One might object that while I expressed (C) with some non-physical vocabulary (e.g. conscious), (C) is nonetheless identical to a physical fact. But this objection depends on a thesis about the individuation of facts. So, given that my ofcial formulation the issue makes no reference to facts, the objection is not on point. In our ofcial statement of the issue, the assumption that (C) is not a physical fact amounts to the assumption that (Cs ), the sentence, contains non-physical vocabulary. And this is clearly true: it contains conscious.12

Connective explanations

So let us argue that general considerations about the nature of ground suggest that (C) The fact that I am in physical state P grounds the fact that I am conscious. has no physical ground. I emphasize that the argument is only suggestive. The general question facing us is what grounds the grounding facts: if the Xs ground Y, what (if anything) grounds the fact that the Xs ground Y? This is a difcult question that I cannot hope to settle here. My aim is just to give some reason to think that (C) has no physical ground, in order to motivate interest in my argument (in sections 79) that physicalism does not require it to have a physical ground. So, what might explain a grounding fact like (C)? One might think that (C) is the result of chaining together a number of intermediary grounding facts: perhaps my being conscious is grounded in the rst instance in biological facts, which are then grounded in chemical facts, which are in turn grounded in the physical facts. If so, one might say that (C) is grounded in those intermediary grounding facts.13 But this would not constitute a
my ofcial logical form and the one I use for convenience are controversial. Schaffer has argued that the logical form of a grounding claim is contrastive: the fact that rather than * grounds the fact that rather than * (Schaffer [21], p. 130). And Dasgupta [3] has argued that ground is irreducibly plural: the Xs ground the Ys. I have no objection to either of these views, but I put them aside for simplicity. 12 To be sure, it is not at all clear how to dene physical vocabulary (see Ney [16] for a discussion of this issue). But we do not need a denition to recognize that conscious is not physical vocabulary in the intended sense of the term. 13 In the terminology of Fine [10], this is to say that facts about mediated ground are grounded in facts about immediate ground.
11 Both

physical ground, since in the course of explaining (C) one would talk about biological and chemical matters. And in any case, this explanation is not available if the grounding fact at issue is not the result of chaining together intermediary grounding facts.14 So, are there other ways to explain the grounding facts? It is important that we hear candidate answers in the esh rather than consider them in the abstract. We want to hear whether a candidate ground really is a good explanation of a grounding fact. But (C) is (at best) controversial, so we should not expect to have a good sense of what might explain it. So let us work with a more mundane example. Imagine again that you recently attended a philosophy conference. It was an event with various properties, but focus on the fact that it was a conference. What grounds this? Presumably it was a conference because it contained people acting in various familiar ways, for example that some gave papers, others listened, and so on. Call this way of acting W. Then we have: (F) The fact that the event included people acting in way W grounds the fact that it was a conference. Our question is: what (if anything) grounds (F)? Why did the fact that the event included people acting in way W give rise to its being a conference? The question is particularly vivid in the contrastive form: why did those people acting in way W make the event a conference, rather than (say) a football match? For Schaffer [21], who takes the basic logical form of a grounding claim to be contrastive (see note 11), this contrastive question is the one to ask. For us, assuming (as we are) that the basic logical form of a grounding claim is not contrastive, our ofcial question can only be the non-contrastive one in the last paragraph. Still, since the contrastive locution helps make the question come alive, I will sometimes engage in loose talk and speak contrastively in what follows. So: why did the fact that the event included people acting in way W make the event a conference (rather than, say, a football match)? A natural thought is that the answer has to do with what conferences are. A conference is the kind of thing that you get when people act in way W; that is why, when those specic people at your favorite conference acted like that, the result was a conference. So it is natural to ground (F) in a general connection between conferences and ways of acting. What kind of general connection? There are a number of options. One is an essential connection. The idea is that it is in the essence or nature of being a conference that an event is a conference if it consists in people acting in way W. And the idea is then that this is (at least in part) why the fact that those specic people at your favorite conference acted like
(C) is functioning as an arbitrary example of a grounding fact, so we cannot assume without loss of generality that it results from a chain.
14 Remember,

that gave rise to a conference (rather than a football match). Kit Fine [10] tentatively endorses this kind of view, as does Rosen [18]. Now suppose that the essentialist fact here is groundless. Then the resulting view is what I will call brute essentialism, the idea that grounding facts like (F) are grounded (at least partly) in an essential connection and that there is no further explanation of the essentialist fact. Brute essentialists may disagree over the details. One question is whether the full explanation of (F) also appeals to the fact that those specic people acted in way W. If it does, then the result is that (F) is grounded in (F.i) Those people acted in way W. (F.ii) It is essential to being a conference that if an event includes people acting in way W then it is a conference. Another point of dispute is the precise form of the essentialist fact, for one might argue that what (partly) explains (F) is not (F.ii) but (F.ii*) It is essential to being a conference that if an event includes people acting in way W then it is a conference and moreover its being a conference is grounded in the people acting in that way. Rosen [18] develops this view. But there is no need to settle this issue. Here the basic idea will do, namely that (F) is partly grounded in an essential connection between actions and conferences that is itself groundless. The idea that (F) is grounded in a general connection between actions and conferences need not be cashed out in terms of essence. One might instead suggest that the general connection is a necessary truth, or a conceptual truth, or perhaps even a metaphysical law. Thus one might replace (F.ii) with one of these: It is necessary that if an event includes people acting in way W then it is a conference. It is a conceptual truth that if an event includes people acting in way W then it is a conference. It is a metaphysical law that if an event includes people acting in way W then it is a conference. Suppose one then takes ones favored connection to be groundless. Then we can call the resulting views brute necessitarianism, brute conceptualism, and brute nomicism, respectively. For now the differences between these views do not matter. They are all examples of what I call brute connectivism, the view that (F) is explained by one of these general connections between actions and conferences that is itself ungrounded. The view bears some analogy to the familiar idea 10

that a specic causal event, say of an F causing a G, is underwritten by the general causal law that every F causes a G.15 One virtue of brute connectivism is that it predicts the following: Necessitation: If some Xs ground Y, then necessarily if the Xs obtain then Y obtains. Applied to (F), this says that it is necessary that if those people act in way W then the event is a conference. Brute essentialism predicts this because it says that if (F) obtains then this is (partly) because it is an essential or conceptual or necessary or metaphysically nomic truth that any event in which people act like that is a conference. And it is independently plausible that essential truths, conceptual truths, and metaphysical laws are necessary. It follows that it is necessary that if those people at your favorite conference act in way W then the event is a conference, as required.16 Brute connectivism has its attractions. But all brute connectivist views imply that (C) is grounded in some ungrounded connection between brain states and consciousness. In which case it does not have a physical ground.

Simple reductionism

Is there an alternative to brute connectivism on which (C) does have a physical ground? One alternative is to say that the general connections have a physical ground. I will consider this soon, but let us rst consider an idea that Bennett [1] and deRosset [5] have each (independently) proposed. The view they converged on is that if the Xs ground Y, then the fact that the Xs ground Y is grounded in the Xs. Call this simple reductionism. Simple reductionism implies that (C) has a physical ground: it is the same as what grounds my being conscious in the rst place, i.e. my being in physical state P. Indeed this is one of the virtues that Bennett and deRosset advertise of the view: it implies that there is no problem formulating theses like physicalism in terms of ground. But there are two problems with simple reductionism. The rst problem is that it is false, and the second problem is that even if it were true it does not have this virtue that Bennett and DeRosset attribute to it. Start with the rst problem. To assess the truth of simple reductionism it is important that we not confuse it with the following principle:
15 Though there are differences. Few would say that causal laws cause the specic causal event, but the brute connectivist says that the general connection grounds (F). Still, the similarity is that a general connection underwrites the specic cases of ground and causation. 16 Admittedly, Necessitation is controversial. Fine [10], Rosen [18], and Trogdon [24] endorse it, while Leuenberger [15], Schaffer [20] and Skiles [22] reject it. So the claim here is that if you endorse Necessitarianism then this is a virtue of brute connectivism.

11

Internality: If some Xs ground Y, then necessarily if the Xs obtain then Y obtains and the Xs ground Y. Suppose that some Xs ground Y. Internality says that these Xs necessitate the fact that the Xs ground Y. Simple reductionism goes further and says that the Xs ground the fact that the Xs ground Y. I have no objection to Internality, just to this further claim. For remember, ground is an explanatory notion, so the question is whether the Xs explain the fact that the Xs ground Y, and I do not think they do.17 For one thing, the proposed explanations sound bad. To see this let us consider one in the esh. First question: why is this event a conference (rather than, say, a football match)? Answer: because various people are acting in way W. So far so good. Second question: why is it that those people acting in way W make the event count as a conference (rather than, say, a football match)? The simple reductionist says: because those people were acting in way W. This is not a good explanation. Compare this to brute connectivism. To the question why is it that those people acting in way W make the event count as a conference, rather than a football match?, the brute essentialist (for example) answers: because it is lies in the nature of being a conference that an event with people acting like that is a conference. It is clear which is the better explanation.18 Moreover simple reductionism implies that facts that should get different explanations get the same explanation. For example, suppose (as is customary) that if P obtains then P grounds PQ. And suppose (as is also customary) that P grounds P. Then simple reductionism implies that what grounds the fact that P grounds PQ is exactly the same as what grounds the fact that P grounds P: namely, P. This is wrong: the grounds are surely different and involve something about disjunction in the rst case and negation in the second. It is because of the way disjunction works that P is a sufcient explanation of why PQ, while it is because of how negation works that P is a sufcient explanation of why P. The point is emphasized by noting that even if P&Q obtains, P does not (on
between explanation and (mere) necessitation is not splitting hairs. It is a well-known phenomenon: the fact that I exist necessitates the fact that 2+2=4 but does not explain it. Indeed the raison detre of ground is to mark this distinction. 18 I am appealing to our capacity to tell good explanations from bad. There is of course a deep question as to how we manage to do this. It is analogous to the Humean question about how we can ever come to know the causal structure of the world: in our case the question is how we can ever come to know the grounding structure of the world. We have not yet answered the Humean question, but this is no reason to stop investigating causes as best we can. Similarly, we do not yet have an epistemology of ground, but this does not mean that we cannot reasonably assess claims of ground. If you ask what grounds the fact that the event is a conference and I answer that it is because 2+2=4, you should complain that my proposed explanation is terrible. Somehoweven if we do not yet know howour grasp of the fact that it is a conference puts us in a position to know that my proposed explanation is bad. My objection to simple reductionism is similar.
17 Distinguishing

12

its own) ground P&Q.19 So why is PQ different? What is it about PQ in virtue of which it is sufciently explained by P but P&Q is not? Surely the answer has something to do with disjunction (perhaps its truth-table). Brute connectivism respects these truisms, simple reductionism does not. My objection then is that brute connectivism offers better explanations than simple reductionism. To be sure, highly theoretical considerations might lead one to think that simple reductionism must be correct, on pain of it being impossible to formulate physicalism as a grounding thesis otherwise. This is one reason that Bennet and deRosset give for endorsing the view. But we must not be blinded by theory. The fact is that when we hear the explanations in the esh, we nd them wanting.20 There is much more to be said about this, but it would be distracting to say everything here.21 As I said, my aim here is just to give some reason to think that (C) has no physical ground, in order to motivate interest in the strategy (that I develop later) of arguing that physicalism does not require that it has a physical ground. So let me turn to the second problem with simple reductionism. Suppose that simple reductionism is correct. Would it follow that there is no problem formulating physicalism in terms of ground? Bennett and deRosset think soindeed that is one of the principle virtues of the view, they say. But I think that this is a mistake. To see why, return to the example of your favorite conference. Suppose the fact that it contains people acting in way W grounds the fact that it was a conference. And now suppose that there is another event in which some other people are acting in the very same way W. If we investigate this other event we will nd (I submit) that it is also a conference, and moreover is a conference in virtue of those people acting in way W. And if we investigate yet other events that include people acting in way W we will nd that the pattern continues. What explains this striking pattern? Why, whenever we nd an event that containing people acting in way W, is it a conference in virtue of their acting like that? Why do we never nd an event where people act like that but where their actions do not ground a conference? The pattern is strikingwhat explains it? To be clear, I am using explains broadly here: the answer need not be a grounding explanation. Indeed at least one grounding explanation does not answer the question. This grounding explanation understands the
19 Remember, I use ground to mean a full explanation: P&Q is grounded by P and Q together but not by either one alone. 20 It may be that Bennett and deRosset mean something different by ground than I do. In particular, if ground in their mouth is not constitutively tied to explanation then my remarks here do not engage with their view. But then, by the same token, nor would their view engage with the issue I am discussing in this paper (specically, the truth of (3), coming (as it is) from my mouth). 21 deRosset has (in personal communication) offered a reply to this objection. I hope to discuss his reply elsewhere.

13

pattern in need of explanation to be a (mere) universal generalization i.e. that for any event x, if x contains people acting in way W then x is a conference in virtue of those people acting in way Wand it then says that (like all universal generalizations) it is grounded in its instances. This is clearly no answer to our question. We want to know why all those instances turned out alikejust repeating the instances is no answer. So there is an issue of how to properly formulate the question of what explains the pattern. Perhaps the question is what grounds the necessity of the universal generalization. Or perhaps it is what explains the universal generalization in some non-grounding sense of explains. But it would be distracting to settle exactly how to formulate the question: we understand it well enough to see that merely repeating the instances is not an answer. Now a brute connectivist has a ready answer. On her view, each instance of the pattern has a common ground in some general connection between acting in way W and being a conference. Noting that each instance has a common explanation is one way to explain a pattern. We might explain why New York and Boston and Washington are all covered in snow by saying that a storm swept up the east coast: the pattern is explained by the common cause. Similarly, the brute connectivist can explain the pattern by citing a common ground. Indeed, some brute connectivists have another explanation available too. Suppose that the brute essentialist grounds (F) partly in (F.ii*) It is essential to being a conference that any event including people acting in way W is a conference and moreover is a conference in virtue of including people acting like that. Then each instance of the pattern follows from this essentialist truth as a matter of logic. This is an explanation of sorts, regardless of whether (F.ii*) also (partly) grounds each instance. So a brute connectivist has a ready answer to the question of why this striking pattern obtains. But what can the simple reductionist say? The rst explanation is not available to her, for there is (on her view) no general principle that is a common ground of each instance. And no alternative explanation is forthcoming. For simple reductionism just says, of each fact Y and facts the Xs, that if the Xs ground Y then the Xs also ground the fact that the Xs ground Y. But this gives us no indication of why, whenever there is a fact to the effect that some people act in way W, it grounds a fact to the effect that there is a conference. For all that the simple reductionist has said the pattern is a brute coincidence, a massive accident. This is unacceptable. There must be some explanation of the pattern, even if it is not an explanation that points to a common ground. The simple reductionist might now try adding to her theory in order to explain the pattern.22 For example she might endorse (F.ii*) and say
22 Though

this move is not open to deRosset, who endorses a deationary view of

14

that each instance of the pattern follows from it by logic (even though no instance is grounded in it). Fine. But what is the status of (F.ii*)? Is it ungrounded or not? If not, then the simple reductionist now has an ungrounded fact about conferences. And now the principle virtue of simple reductionism is in doubt. The supposed virtue is that it implies that nothing about the nature of ground precludes formulating physicalism in its terms. Bennett and deRosset think that it has this virtue because it implies that particular grounding facts like (C) have a physical ground. But it is not just particular grounding facts that we must account for: we must also account for patterns in the grounding facts. And simple reductionism says nothing about that. To explain the patterns she might appeal to a general connection between physics and consciousness, like (F.ii*). But until it has been shown how to ground this connection in physical terms, it remains that something about ground (i.e. the patterns) precludes formulating physicalism in its terms. To be clear, the claim is not that (F.ii*) has no physical ground. The claim is that, if it does, this does not follow from simple reductionism. So simple reductionism does not have the principle virtue attributed to it.23

Reductionism

I have argued that we should prefer brute connectivist views over simple reductionism: they provide a better grounding explanation of particular grounding facts like (F) and (C), and also a ready explanation of patterns in the grounding facts. But why the focus on brute connectivism, the view that the general connections between (say) mind and body are themselves ungrounded? Might one endorse the explanation of (C) in terms of the general connection, but then say that the general connection has a physical ground? And if so, might one then chain the explanations together and say that (C) has a purely physical ground? One might then endorse this physicalist explanation of (C) regardless of whether it goes via the general connection. Either way, the result is what I call reductionism: (C) may not have the physical ground that the simple reductionist claims, but it has some physical ground nonetheless. As far as I know, no one has developed a reductionist view (other than simple reductionism). And it is difcult to argue against a view that remains unformulated. But there is reason to be skeptical in advance. For
ground on which the simple reductionist formula is all there is to say about ground. 23 A similar point can be made with regards to other principles governing ground, such as Necessitation and Internality. If the simple reductionist endorses those principles (as I believe she should) then she must somehow explain them (in the broad sense of explains). Merely producing the instances will not do. And it is unclear how the simple reductionist could explain them without appealing to general truths like (F.ii*).

15

the same reasons to doubt simple reductionism look like they generalize to any reductionist view. To see this, return again to the example of your favorite conference. The rst question is: why is this event a conference, rather than (say) a football match? Answer: because various people are acting in way W. So far so good. Second question: why is it that those actions made the event count as a conference, rather than (say) a football match? The simple reductionist said: because those people were acting in way W. That was a bad explanation. But so too, it seems, is any explanation that does not mention conferences. Talk about other actions if you like, talk about particle positions or cabbages or kings. None of that (on the face of it) explains why those actions made the event a conference rather than a football match. An analogy with causal explanation might help. Suppose that the initial state of the world caused its current state. What explains this causal connection? Why did that initial state cause this state, rather than some other? A natural answer is to appeal to the dynamical laws. Never mind whether the specics of this answer is correct, the point is that the reason why the initial state causally produced the current state (rather than some other state) is clearly not to be found just in the initial state itself. It is rather some general, law-like connection between states that explains why one state causally produced another.24 This suggests that only a general connection between actions and conferences can explain why those particular actions produced a conference. Mutatis mutandis, that only a general connection between physical states and consciousness could explain (C). Again, this is just suggestive and does not prove that reductionism is false. For now I leave the burden on the reductionist to produce and defend her theory. The last two sections just serve to illustrate the difculties for this project. Which lends credence to (3) General considerations about the nature of ground suggest that some non-physical facts are not grounded in physical facts.

Autonomy
(1) Physicalism is true only if all non-physical facts arise out of physical facts. (2) Y arises out of the Xs iff Y is grounded in the Xs.

Now recall that (3) and the following two claims

24 Of course the analogy is not perfect, since the idea is not that the law causes the causal connection between initial and current states. Still, the law explains the causal connection in some sense of explains, and the point of the analogy is to give insight into how these explanations might work.

16

imply that general considerations about the nature of ground suggest that physicalism is false; an unacceptable result. So if we wish to understand physicalism in terms of ground, we should explore denying (1). On this approach we argue that physicalism does not require that grounding facts like (C) arise out of physical facts. And if we endorse brute connectivism, then we must show that physicalism does not require that the general (essential or necessary or conceptual or metaphysically nomic) connections have a physical ground either. But as I emphasized in section 2, it is not enough to just say that the physicalist need not give (C), or these general connections, a physical ground. We must also say why they are special. After all, a physicalist clearly cannot say that the fact that I am conscious has no physical ground. Why then are (C) and these general connections any different? Nor is it enough to say that our intuitive understanding of physicalism does not require them to have a physical ground. For there are (in logical space) indenitely many grounding theses, some requiring them to have a physical ground and some not. We want to know why the latter are important and deserving of the attention they receive. To this end, I suggest that we introduce a distinction between substantive and autonomous facts. Roughly, a fact is substantive if it is apt for being grounded, if the question of what grounds it can legitimately be raised and admits of a sensible answer (an answer that either states its ground or else states that it has none). In contrast, a fact is autonomous if it is not apt for being grounded in the rst place, if the question of why it obtains does not legitimately arise. The rough idea is then that physicalism does not require the autonomous facts to have a physical ground, and that the general connections are autonomous. This will be rened as we go along, but that is the basic idea. I will not argue that this is the best (or only) way to deny (1). But I will argue that it is a reasonable way. My approach hangs on this notion of autonomy. What does it mean? I will not try to dene it (in part because I suspect that it has no denition). But something can be said to clarify the notion I have in mind, starting with a general characterization and then turning to examples.25 The basic idea is reasonably familiar in the case of causal explanation. Consider facts about how various particles are arranged. Some particle arrangements (we may assume) have a causal explanation. The arrangement that constitutes my laptop in its current state might be an example. Other particle arrangements may not. If I ask what causally explains why a certain group of particles were arranged in a certain way and if the denoted arrangement happens to have been the initial condition of our universe then the appropriate answer would be that there is no explanation, that it
25 The

rest of this section overlaps signicantly with section 6 of Dasgupta [2].

17

is a causally brute fact about the worlds initial state. But all particle arrangements are apt for causal explanation: the question of what causally explains how they came to be arranged like that can legitimately be raised even if the answer is in some cases nothing. Consider by contrast the fact that 1+2=3. Like the initial condition, this lacks a causal explanation. But there is a difference. If someone asked you what causally explains why 1 and 2 came to equal 3 you would not just say nothing, you would start talking about the notion of causation and the nature of abstract objects in an attempt to show that the question should not have been raised in the rst place. The particle arrangement that happened to be the initial condition lacks a causal explanation even though it is a good question why they came to be arranged like that. The mathematical fact by contrast lacks one because it is not apt for being causally explained in the rst place. The result is a three-fold distinction. There are those facts that are apt for causal explanations but lack onee.g. the initial conditions. And there are those that are apt for causal explanations and have onee.g. the current arrangement of particles that constitute my laptop. And then there are those that are not apt for causal explanation in the rst placee.g. facts of pure arithmetic. My thought is that an analogous three-fold distinction can be drawn in the case of grounding explanations too. There are those that are apt for having a ground but lack one. These are the so-called fundamental or brute facts. And there are those that are apt for having a ground and have one. These are the so-called derivative facts. And nally there are those that are not apt for having a ground in the rst place. These are what I call autonomous. That is one route into the notion of autonomy. But it might help to triangulate onto the notion with a second analogy, an analogy with proof (this will also help us to nd examples of autonomy). Suppose one is doing standard, axiomatic set theory. One starts with a language containing the binary predicate of set-membership, states some axioms governing it, and proves some theorems. It is then useful to expand the language and introduce by stipulation a new word: Denition: x is a subset of y d f any member of x is a member of y. One can then express various claims about subsets, for example Proposition: There is an x and a y such that x is a subset of y but not vice-versa. and attempt to prove them from the axioms and the denitions. Now the question of whether the Proposition has a proof is clearly a good question regardless of whether it actually has one. Indeed this kind of question is the bread and butter of set theory. The same applies to the axioms of the 18

system: the question of whether one of them is provable from the rest is a good question that can lead to remarkable advances even if the answer is nojust think of the attempt to prove Euclids 5th postulate! But the same is not true of the denition itself: the question of whether we can prove the denition from the axioms is clearly not a good question. If your homework buddy asked whether you can prove the above denition you would take her to be confused about its status as a denition. You would not just say no, you would start talking about the role of the denitionthat its function is to introduce by stipulation a new term to the language, and so onso that your buddy comes to see that the request for a proof is in some sense illegitimate. As we might put it, the denition is not apt for being proved. There may be other good questions to ask of the denition, such as whether it is useful or corresponds to some prior concept we had in mind. But how it might be proved from the set-theoretic axioms is not one of them.26 So as in the case of causation we recognize a three-fold distinction. There are those truths that are apt for proof but lack one (if your derivation system is complete then these are just the axioms). Then there are those truths that are apt for proof and have one. And nally there are the denitions that are not apt for proof in the rst place. Autonomous facts are those that stand to ground as denitions stand to proof. Indeed this analogy with proof suggests some candidate examples of autonomous facts. For a stipulative denition like the above states what a term means. And the worldly analogue is a statement of what something isa statement (that is) of its nature or essence. When we ask in the philosophy seminar room What is knowledge? and consider the answer that knowledge is true and justied belief, we are considering a claim about what knowledge is, about the essence of knowledge. And when we ask in a chemistry lab What is water? and consider the answer that water is a substance composed of H2 0 we are again considering a claim about what water is, about its essence. Regardless of whether these claims are correct we seem to understand them reasonably well. Let us follow Fine and take the logical form of these claims to be It is essential to x that where x is an item of any ontological category and is a sentence.27 I will take this to be synonymous to saying that it is in the nature of x that or
course in another axiomatic system the word subset might be dened differently, in which case the material equivalence x is a subset of y iff any member of x is a member of y might have a (non-trivial) proof from those other axioms and denitions. Moreover one might construct a logic of denition, i.e. a logic of the d f operator, and one might conceivably then show that the denition in the text is provable from the axioms governing d f . But the point is just that when doing standard axiomatic set-theory, the question of whether the denition in the text can be proved is intuitively a bad question. 27 This is how Fine regiments talk of essence in his [7].
26 Of

19

that it is part of what x is that .28 So essentialist truths are the worldly analogue of denitions. And it seems to me that they do indeed stand to ground as denitions do to proof: the question of what grounds them strikes us as illegitimate in something like the way that the question of how one might prove a denition does. For suppose that it is essential to knowledge that someone knows only if she truly and justiably believes. And suppose someone asks what explains this (in the metaphysical sense). In virtue of what (the question is) is it part of what knowledge is that someone knows only if she truly and justiably believes? It is difcult to know how to respond. One is tempted to say that this is just what knowledge is. . . but of course this is what we were asked to explain! In saying this one is most naturally heard not as trying to explain this fact about knowledge in any serious sense but rather as deecting the demand for explanation. Or suppose that it is essential of water that it is a substance composed of H2 O. And suppose someone asks what explains this (in the metaphysical sense). The question is not why we should believe it or how water came to refer to what it does, there are good answers to both those questions. The question is rather: in virtue of what is it part of what water is that it is composed of H2 O? It is again hard to know what to say other than that is just what water is! And in saying this one is again naturally heard as sidestepping the question rather than giving it a serious answer. Compare these essentialist facts to (say) the fact that some particles are arranged thus-and-so. This might (for all we know) be grounded in the undulations of a quantum wave-function in a massively high-dimensional Hilbert space.29 Or it might be a brute, ungrounded fact about the world. But either way we consider it a perfectly good and legitimate question why they are so arranged even if the answer turns out to be no reason. So we appear to distinguish the brute facts from the essentialist facts. Both are groundless, but there is a difference. The former are (by denition) apt for being grounded, it is just that they lack a ground. By contrast, the latter are not apt for being grounded in the rst place, in roughly the same sense that arithmetic facts are not apt for causal explanation and that denitions are not apt for proof. I am using autonomous to label the latter category, and brute or fundamental to mark the former. Thus the brute facts play the same role vis a vis metaphysical explanation as the initial conditions play vis a vis causal explanation and the axioms play vis a vis proof. The autonomous facts play a different role, one more analogous
28 It may only be part of what x is that , so we should not presume that uniquely species x. Nor should we presume that there is a collection of such that does uniquely specify x. It might be that the only answer to the question of what canary yellow is, is that it is a determinate shade. 29 See Albert [?] for a view like this.

20

to the role that denitions play in proof.30 The claim (to be clear) is not that it is autonomous that someone knows only if she truly and justiably believes. For that may be grounded in the fact that that is part of what knowledge is. More generally one might endorse what Rosen [18] called Essential Grounding: If it is essential to x that , then because it is essential to x that . So the essentialist truth that is claimed to be autonomous is not that but rather that it is essential to x that . Essentialist truths may not be the only examples of autonomous truths. Conceptual truths are also an obvious candidate, though if Fine [6] is right that they are essentialist truths about concepts then this would not be an example of a new kind. Identities might be another example. Clearly, the question of what kinds of facts exhaust the category of autonomy depends on all sorts of issues, and so is not one that I can settle here (though I will defend the claim that essentialist truths are autonomous later on). In each caseof ground, causation, and proofI suggested that we nd intelligible the idea that certain facts are not apt for ground or causal explanation or proof (respectively). I motivated this by pointing out that we respond to certain questions in a distinctive way (we respond not by answering them but by pointing out that they should not have been asked in the rst place). This led to me glossing the notion in epistemic or cognitive terms, for example in terms of certain questions not making sense or being illegitimate. But the notion of autonomy is not dened in epistemic or cognitive terms. Our distinctive response to these questions is evidence that we nd the notion intelligible, but is not necessarily that in terms of which the notion is to be dened.31 To be sure, this evidence is not conclusive and one might still refuse to nd the notion of autonomy intelligible. And even if one nds it intelligible it is reasonable to hope that more can ultimately be said about what it amounts to. But that is a large project and I do not want to get sidetracked here. So for the sake of argument I will assume that we have enough independent grip on the notion to proceed.
30 I should say that the analogy with causation is (like any analogy) not perfect. I said that arithmetical facts are not apt for causal explanation but they also (arguably) do not causally explain anything. In contrast, even if essentialist facts are autonomous they certainly ground other facts. A conjunction of two essentialist facts will (like any conjunction) be grounded in its conjuncts. Still, the point of the analogy is to bring out the similarity between arithmetic facts and essentialist facts vis a vis the question of what explains them regardless of what (if anything) they explain. 31 Though one might of course propose a link between autonomy and some epistemic notion, for example that a fact is autonomous iff it is apriori or conceptual. Della Rocca (on one reading) defends something like the view that a fact is autonomous iff it is conceptual in his [4]. I am inclined to think that the case of water discussed above is a counterexample to this thesis but I will not discuss this further here.

21

Physicalism regained

I argued earlier that the best explanation of (C) The fact that my brain is in physical state P grounds the fact that I am conscious. is a brute connectivist explanation on which (C) is in part grounded in a general (essential or necessary or conceptual or nomic) connection between brain states and consciousness, a connection that is itself ungrounded. We must now show that this is consistent with physicalism, contrary to appearances. I will argue in this section that if the connection is autonomous then this is consistent with physicalism. And I just motivated the idea that essentialist truths are autonomous. If that is right, then at least one brute connectivist view, brute essentialism, is consistent with physicalism. (I leave open whether the other connections are autonomous, and so whether other brute connectivist views are consistent with physicalism.) So, let us argue that if the connection is autonomous then there is no conict with physicalism. The argument applies equally to any connection but for clarity focus on brute essentialism, on which (C) is grounded in (C.i) I am in physical state P, and (C.ii) It is essential to being conscious that if something is in physical state P then it is conscious. Suppose that (C.ii) is autonomous. Our question is then whether this is consistent with physicalism. But this is not a question of logic: we do not have a clear thesis marked physicalism in hand, such that we can then ask whether it is logically consistent with brute essentialism. Indeed this paper is an investigation into what physicalism is. All we have is a rough idea, a picture, so our question is whether brute essentialism ts with that picture. So, to proceed, let us identify two theses that are logically consistent with brute essentialism and ask to what extent they approximate the physicalist picture. To the extent that they approximate it, brute essentialism will have been shown to t with that picture. The thesis that best approximates the physicalist picture is perhaps Strong Physicalism (SP): All non-physical facts are grounded in physical facts. This is not logically consistent with brute essentialism, for according to brute essentialism both (C) and (C.ii) are non-physical facts without physical grounds. But we can identify two weaker theses that are consistent with brute essentialism. One is Weak Physicalism (WP): All substantive non-physical facts are grounded in facts that are either physical or autonomous. 22

This is consistent with brute essentialism, if essentialist truths are autonomous. For in that case (C.ii), being autonomous, lies outside the scope of WP. And while brute essentialism implies that (C) is substantive, it also implies that it is grounded in facts that are either physical (i.e. (C.i)) or autonomous (i.e. (C.ii)), and so the situation is consistent with WP. The question then is: to what extent does WP approximate the physicalist picture? And the answer is: to some extent, but only some. Let me describe the extent to which it does, and then the extent to which it does not, and then state a modication that does better. The quick argument that WP ts the physicalist picture to some extent is that WP gives explanatory pride of place to physical facts, rather than (say) mental or normative ones. Admittedly, WP does not state that physical facts are enough to ground everything. But it does state that once the autonomous facts are in place the only other facts needed to ground all else are physical facts. Insofar as the physicalist picture has physical facts playing a basic and special explanatory role, WP ts well with this picture. An objection to this quick argument is that WP also gives autonomous facts a special explanatory role. Indeed, two special roles: rst, WP allows that they are ungrounded; and second, WP allows that they help (along with physical facts) ground all else. Why are autonomous facts special? It would certainly not t with the physicalist picture to say that the fact that I am conscious is ungrounded. And yet WP allows that (C.ii), a fact about consciousness, is ungrounded. Why then is (C.ii) special? Why does it t with the physicalist picture that it plays these special roles? In response, consider each role in turn. To see why it ts with the physicalist picture that (C.ii) is ungrounded, remember that our working hypothesis is that physicalism is an explanatory thesis, a thesis about what grounds what. But autonomous facts are not apt for being grounded. So in asking what grounds what, the meaningful question is what grounds those facts that are apt for being grounded, i.e. what grounds the substantive facts. In asking this we bracket the autonomous factsthey are simply not under investigation. Insofar as physicalism is an answer to this question, it is consistent with physicalism that autonomous facts are ungrounded. The point is supported by our analogy with causal explanation. Imagine that a cosmologist says that everything is causally explained by the initial state of the universe. If it turned out that the formation of our solar system had no causal explanation, this would be a counterexample. But it is no counterexample that there is no causal explanation of the fact that 2 is even, for this is not apt for causal explanation in the rst place. Just imagine how absurd it would be for someone to raise this counterexample at a conference! Her thesis is (when interpreted charitably) a thesis to the effect that everything that is apt for causal explanation is ultimately explicable in terms of that initial state. Likewise, the physicalist idea that everything is grounded in the physical is, charitably interpreted, the claim 23

that everything substantive is grounded in the physical. That is why it is consistent with autonomous facts like (C.ii) being ungrounded. Now according to WP, autonomous facts play a second special role too: they (along with physical facts) help ground all else. Does it t with the physicalist picture that they play this role? It does, to some extent. For autonomous facts are autonomous not thanks to their content (i.e. whether they are physical, normative, etc) but thanks to their role in grounding explanations: they are characterized as those facts that are not apt for such explanation. So when we ask what grounds what, one natural and non-ad hoc explanatory project is to take autonomous facts for granted and ask what else one needs to ground the rest. WP answers: just physical facts. This gives a natural and non-ad hoc explanatory pride of place to the physical facts. Insofar as the physicalist picture gives physical facts an explanatory pride of place, WP ts with that picture. Think of it (metaphorically) like this. Substantive facts come in two kinds: grounded and ungrounded. The grounded ones arise for free, if you like, once their grounds are in place. And the ungrounded ones had to be placed there by at for the world to get going, as it were. But autonomous facts are different: they do not have to be placed there by at, yet nor do they arise for free out of others. If you like, they were there anyway. So one principled and natural explanatory project is to take them for granted and ask what needs to be placed there by at in order to ground everything. Since WP says just the physical facts, it gives the physical an special explanatory pride of place.32 That is the extent to which WP ts the physicalist picture. Still, there are respects in which it does not. The existence of a Christian God does not t with that picture regardless of whether its existence is autonomous; and yet its existence is consistent with WP if its existence is autonomous. Or suppose that facts about the natural numbers (including their existence) have no physical ground and are instead grounded in some autonomous facts about (say) zero and the successor function. Then this would be consistent with WP, but arguably jars with the physicalist picture. What then is the difference between (C.ii) and the existence of a Christian God? Why does the former t with the physicalist picture so long as it is autonomous, whereas the latter jars with the physicalist picture regardless of whether it is autonomous? Well, (C.ii) states a physical sufcient condition for my being conscious, and (according to brute essentialism) underwrites a physical explanation of my being conscious. But the exis32 This is a little wooly, no doubt. But it had to be. For our question is not whether WP is logically consistent with physicalism but whether WP ts with the physicalist picture. This is a wooly question, concerning the connection between a clear thesis and a wooly picture. So the answer had to get wooly at some point, and this is that point. Still, the claim is just that WP expresses a principled and natural and non-ad hoc respect in which the physical facts are explanatorily special, and this is tolerably clear.

24

tence of God does neither. This suggests that the following thesis better approximates the physicalist picture: Moderate Physicalism (MP): WP, and all autonomous facts help underwrite the kind of explanations required by WP. What does it mean to help underwrite the kind of physicalist explanations required by WP? Helping to ground an explanation of consciousness in physical terms, as (C.ii) does, certainly counts. But more counts too. For one thing, an explanation of consciousness in physical terms might proceed in stages, rst explaining it in biological terms and then in chemical terms and nally in physical terms. A fact that helps ground any of these intermediary stages would help underwrite a physicalist explanation, as I use the phrase. And for another thing, if my brain were in a different physical state then my being conscious might have a different physical ground; and this different grounding fact might itself be grounded in an essentialist fact other than (C.ii). This other essentialist fact would help ground a physical explanation of consciousness were the physical facts different, even if it actually does not. Still, it helps underwrite the kind of explanations required by WP as I intend the phrase. This notion of underwriting should at some point be dened more precisely, but it would distract from the main thread to do so here. The idea is clear enough to see that brute essentialism is consistent with MP. For we know that it is consistent with WP, and the essentialist truth (C.ii) clearly underwrites the explanations required by WP in the intended sense. Moreover MP assigns physics more of a privileged explanatory role than WP and so ts the physicalist picture better than WP. For MP agrees with WP that, along with the autonomous facts, the physical facts are sufcient to ground all else. But MP goes further and says that the autonomous facts are entirely at the service of those physicalist explanations. This rules out the possibility of a Christian God whos existence is autonomous. Indeed if the autonomous truths that underwrite physicalist explanations of some phenomena are essentialist truths about it (as brute essentialism says), then what MP implies is that there can be nothing more to the essence of that phenomena than what underwrites physicalist explanations of it. This is a strong claim. It rules out, for example, the view that part of the essence of goodness involves a link to motivation, for this essential link would not underwrite a physicalist explanation of goodness.33 Any link between goodness and motivation must, according to MP, be grounded in physics (and, perhaps, autonomous facts of the right kind). So, when coupled with brute essentialism, MP gives a very strong explanatory pride of place to physics indeed: it requires that there be nothing more to what something is than would underwrite physicalist explanations of it.
correct formulation of this link is a notoriously delicate matter. But nothing hangs on how it is formulated so feel free to use your own favorite formulation.
33 The

25

One might now think that MP goes too far and that physicalism is something weaker. But there is no need to settle the issue here, for I am not offering a conceptual analysis of physicalism (it is not clear to me that the English word physicalism has a unique analysis).34 The argument is just that (i) the physicalist picture gives physics an explanatory pride of place, (ii) brute essentialism is consistent with views (like MP and WP) that give physics an explanatory pride of place; therefore brute essentialism approximates the physicalist picture well. Against this, one might object that the physicalist picture nds its natural expression in various slogans such as All God had to do when making the world was determine the purely physical facts. But, the objection goes, this slogan expresses the idea that the only ungrounded facts are physical facts, contra brute essentialism. In response, I think that a reasonable interpretation of the slogan is consistent with brute essentialism. For what does the general slogan all God had to do when making the world was mean? One interpretation is that is whatever would be sufcient to causally explain everything, but that is not what we have in mind when we use the slogan to express physicalism. Another interpretation is that is whatever would be sufcient to ground all else. This is closer to what we might have in mind, and the objection is right that brute essentialism contradicts this interpretation. But there is a third interpretation of the slogan. The slogan derives from the theistic view that there really is a God that ultimately explains things. Since we are concerned with grounding explanations, the theistic view at issue is that everything about the world is ultimately grounded in facts about God (e.g. Her will, etc). But even if this theistic view were true, God would not ground the autonomous facts since they are not apt for being grounded in the rst place. Nor need She ground what is fully grounded in those autonomous facts, since they are explained without Her help. What then does help God ground, on this theistic view? All else, i.e. those facts that are not autonomous or fully grounded in autonomous facts. So on the interpretation of all God had to do when making the world was that ts with this theological view, consists in determining the substantive facts that, along with the autonomous facts, ground all else. So interpreted, brute essentialism is consistent with the physicalist slogan after all. For once God determined that I am in physical state P, brute essentialism implies that this was indeed sufcient, along with the autonomous fact (C.ii), to ground (C).
34 It is because I am not offering a conceptual analysis that it is unnecessary (and would be distracting) to dene the notion of underwriting in MP in more detail.

26

Ampliativity regained

I just argued that if essentialist facts are autonomous then brute essentialism is consistent with physicalism. And I suggested earlier that essentialist facts are autonomous. If that is right then brute essentialism is consistent with physicalism. And it then follows that we should reject (1) Physicalism is true only if all non-physical facts arise out of physical facts. For physicalism does not require that (C) and (C.ii) arise out of physical facts, if (C.ii) is autonomous. This is my response to the Siderian argument that ground is unsuitable for formulating physicalism. The argument from the last section generalizes to the other brute connectivist views like brute necessitarianism, brute conceptualism, and brute nomicism. The argument shows that if the general connection that they each use to ground (C) is autonomous, then these views are all consistent with physicalism and that therefore (1) is false. (I leave the question of whether those other connections are autonomous for another time.) The picture of ampliative theses like physicalism and normative naturalism that emerges is this. The exciting idea behind them is that at some lower level the world is constituted by a relatively sparse substantive basis (e.g. of physical or natural facts), and that this sparse basis grounds higher level facts concerning something new (e.g. consciousness or normativity, respectively). But this does not require that the links between the levelsthe specic grounding facts, and the general (essential or necessary or conceptual or nomic) connectionsare themselves grounded in the lower level, so long as those general connections are autonomous. On this approach ampliative theses are both stronger and weaker than one might have thought. They are weaker insofar as (say) physicalism does not require that every non-physical fact has a physical ground. But they are also stronger, for physicalism is now seen to require autonomous connections between (say) mind and body. Indeed if brute essentialism is correct then physicalism requires that there are essential connections between mind and body. This will be disappointing to physicalists who hoped that formulating physicalism in terms of ground would recuse her from having to offer tight connections of essence or analysis between mind and body. On the current approach, this hope is dashed.35 For the remainder let us focus on one implementation of this approach to ampliativity, the brute essentialist one. It consists of two claims: (i) Brute essentialism, and (ii) the claim that essentialist facts are autonomous.
35 I should emphasize that this is a view about what ampliative theses are, not about how they might be established. How one might demonstrate or refute (say) an essential connection between mind and body is a delicate matter in the epistemology of metaphysics, but not one that can be addressed here.

27

I have already motivated each, but let me end by discussing objections to (ii) (section 10), and an objection to (i) (section 11).

10

Are essentialist facts autonomous?

I claimed earlier that they are. As I said then, suppose that it is essential to knowledge that someone knows only if she truly and justiably believes. And suppose someone asks what grounds this. Remember that an essentialist fact about x is a fact about what x is. So the question is: in virtue of what is it part of what knowledge is that someone knows only if she truly and justiably believes? One is tempted to say that this is just what knowledge is. . . but of course this is what we were asked to explain! I suggested that one is most naturally heard not as offering an explanation but rather as expressing the idea that it is not apt for explanation in the rst place. By an essentialist fact I mean a statement of what something is in its most core respects, what Fine calls a statement of constitutive essence. I take this notion of constitutive essence to be primitive. A number of extensions can then be dened. We can (for example) close it under logical consequence, so that (for instance) if it is essential to x that then it is essential to x that , for any . Fine calls this the notion of consequential essence. Or we can chain essences together to get what Fine calls a mediated essence. If it is constitutively essential to knowledge that knowledge is true and justied belief, and constitutively essential to truth that truth corresponds to the facts, then it is mediately essential to knowledge that knowledge is justied belief that corresponds to the facts. Now this mediated essence of knowledge is plausibly grounded in the two constitutive essences of knowledge and truth, in which case meditated essences are not autonomous. Similarly, if it is constitutively essential to x that , then its being consequentially essential to x that is plausibly grounded in the constitutive essence. So my claim is just that constitutive essences are autonomous (though I drop the qualication for brevity).36 Let me now defend this claim from objections. There are a number of possible objections to consider; here I consider just ve. First, one might point out that ingenious philosophers have proposed general theories about essentialist truths, for example that if something is an essentialist truth then that is in virtue of its being a necessary truth
36 For more on the notion of constitutive and mediated essence see Fine [7]. In his papers on this topic Fine took the notion of consequential essence as basic, and he then denes the notion of constitutive essence by generalizing away (see Fine [7]). But statements of consequential essence would only be recognizable as statements of what things are by a logician, and so it strikes me as a curious choice of a primitive. The notion of constitutive essence seems a better choice. Koslicki [14] also distinguishes a number of other notions of essence. It is a good question whether one of them can be taken to ground the others but I cannot discuss it here.

28

of some sort. If essentialist truths are autonomous then the attempt to give this kind of theory is deeply confused in something like the way that attempting to prove a denition is. And even these attempts are mistaken (the objection is) surely they are not confused in that way.37 In response I think that they are confused, but that the confusion is explained as arising from a natural but misleading picture. On this picture one starts with the idea that there are two ways to have a propertyan essential way and an accidental wayand one then takes the essentialist facts about something to be facts about which properties it has in the essential way. On this picture the essentialist facts are just facts about which properties are had in that way by a given domain of things. And such facts then look to be substantive if anything is: surely we can ask why (in the metaphysical sense) the given object has that property in the essential way. But this picture is misleading: on my conception the essentialist facts concern what those things are in the rst place. It is not that there is some independently given domain and the essentialist facts concern what properties they have, it is rather that the essentialist facts specify what the domain is in the rst place. It is those facts that strike me as autonomous.38 I suspect that those tempted to provide grounds for essentialist facts are in the grip of the picture just described. Insofar as we all mean the same thing by essence I claim that the picture is deeply misleading, albeit understandable. But it is also possible that the word essence has been used by others to mean something different in which case the above picture may be accurate of their notion and there may be no disagreement. A second objection to the idea that essentialist facts are autonomous stems from patterns in essences. Suppose one holds that it is essential to {Socrates} that it contain Socrates as a member. Presumably one will also think that it is essential to {Aristotle} to contain Aristotle as a member. And so on for other singletons. Why this striking pattern? Perhaps it is because each essentialist fact has a common ground: each is grounded (at least in part) in the fact that it is essential to the set-membership relation that it is essential to any set that it has the members it has. I must disagree. But I need not deny that the pattern exists or say that it is entirely random. For the above fact about the essence of setmembership, along with the fact that {Socrates} and {Aristotle} are sets, imply that {Socrates} essentially contains Socrates and that {Aristotle} essentially contains Aristotle. So there is a relation of implication here, though
to Tobias Wilsch for pressing this objection on me. my notion of an essentialist fact one may dene the notion of having a property essentially: x essentially has the property of being F d f it is part of what x is that x is F. So I can make good sense of the idea that there are two ways to have a property and that the essentialist truths are those regarding which properties are had in the essential way. But we are misled if we forget that the primary notion is that of a specication of what something is. This is why I call the above picture misleading and not false.
38 Given 37 Thanks

29

not one of ground. The resulting picture is not unreasonable. Suppose I want to tell you what something is. Once I tell you that it is a set it follows (from the nature of set-membership) that I must then tell you what its members are. But each statement of what the thing isthat it is a set, that it contains these membersis autonomous. A third objection is that while there is something special about essentialist facts, it is not that they are autonomous but rather that they are self-grounding. In response one might say that ground is irreexive and therefore that no fact can ground itself. But I do not assume irreexivity: if ground is transitive, the assumption of irreexivity precludes the possibility of closed circles of ground in which a fact grounds itself via a number of chains in a loop, and I see no reason to preclude such possibilities at the outset. Such cases are the analogue of closed causal loops: odd, perhaps, but a way that the world could turn out. Still, I nd it hard to understand how a fact could immediately explain itself in the very same sense that (say) peoples actions can can ground the existence of a conference, which is what the objector is asking me to believe.39 A fourth objection is that while there is something special about essentialist facts, it is not that they are autonomous but that they are zero grounded. Fine introduces this notion by analogy with sets: Any nonempty set {a, b, . . . } is generated (via the set-builder) from its members a, b, . . . The empty set {} is also generated from its members, though in this case there is a zero number of members from which it is generated.40 So Fine distinguishes between the null-set, which is zero-generated, from things that are not generated by the set-builder operation at all, like me. Similarly, thinks Fine, we should distinguish facts that are ungrounded from facts that are zero grounded. And so one might claim that what is special about essentialist facts is that they are zero grounded.41 In response, it is not clear that the notion of being zero grounded is intelligible. But even if it is, the claim that essentialist facts are all zero grounded strikes me as false because it has the implausible consequence that all essentialist facts have the same ground. Suppose that it is essential to water that it is a compound, and essential to {Socrates} that it contain Socrates. If we say that essentialist facts are zero-grounded then we are conceding that both these essences are substantive (i.e. apt for being explained), and moreover that they both have an explanation; and we are then saying that their explanation is exactly the same (i.e. the zero explanation). And this is hard to believe: if the facts are substantive and if they have an explanation, surely the explanation is different in each case.42
for a defense of the claim that a fact can immediately ground itself, see Jenkins [13]. 40 Fine [10], p. XX. 41 Kit Fine and John Litland have both suggested this to me in conversation. 42 To be sure, one might think for independent reasons that everything has a common
39 Though

30

Fifth (and nally) one might argue that any essentialist fact is grounded in an iterated essential fact. For consider the following principle: (E4) If it is essential to x that , then it is essential to x that it is essential to x that .43 And recall Rosens principle of Essential Grounding, which states that if it is essential to x that , then because it is essential to x that . Now consider any true sentence S of the form it is essential to x that . By (E4), it is essential to x that S. And then Essential Grounding implies: S because it is essential to x that S. Therefore, any essentialist fact is grounded in an iterated essentialist fact; so no essentialist fact is autonomous.44 In response, I deny (E4). Remember, a statement of constitutive essence is a statement of what something is in its most core respects. It is the essential core of the thing from which extended essentialist claims about (say) mediated essence can be derived (and perhaps grounded). To state the constitutive essence of Socrates might require stating that he is human. But it is odd to think that his essential core also includes the fact that it is part of the essential core that he is human. This latter, iterated claim of essence is something that follows from (or is grounded in) his essential core and not part of the essential core itself. Admittedly, (E4) is a theorem of the system presented in Fine [8]. But that system was developed to govern consequential essence, the notion that is closed under logical consequence, and I do not believe that it plausibly governs constitutive essence. To see this, consider the axiom scheme that Fine uses to prove (E4), an instance of which is (*) If it is not essential to Socrates that he drank the hemlock, then it is essential to Socrates that it is not essential to him that he drank the hemlock.45 Our question is whether this is plausible when understood as governing constitutive essence, and I think it is not. Stating the constitutive essence of Socrates might require stating that he is human, that he had certain parents, and so on. But it is odd to add to this description of what he is in the most direct and immediate sense that it is not essential of him that he drank the hemlock. So I claim that (*) is false when understood to govern constitutive essence. Still, there is a truth in the vicinity. Say that it is derivatively essential to x that iff the fact that is grounded in (i) facts of the form
ground, e.g. in the mind of God. But that thesis is in need of independent support. What I object to here is the stock idea that essentialist facts all have the same ground. 43 I have used the label (E4) out of respect for the axiom of modal logic that it resembles. 44 I am extremely grateful to Jon Litland for bringing this objection to my attention and indeed for many enlightening conversations on the topic of this paper. 45 This is an instance of the axiom scheme (II)(iii) in Fine [8], p. 247.

31

it is constitutively essential to x that , and (ii) a totality fact to the effect that those are all the true claims about the constitutive essence of x. Then it is tempting to say that if it is not constitutively essential to x that , then it is derivatively essential to x that it is not constitutively essential to x that . Roughly speaking: the fact that is not a member of the set of all true claims about the constitutive essence of x is what makes it the case that it is not constitutively essential of x that .46 Then (*) is true when its second occurrence of essential is understood to express derivative essence.

11

Regress?

Let us nish by returning to brute essentialism and considering the objection that it leads to regress. We started with the fact that I am conscious, and we supposed that it has a physical ground: (C) The fact that I am in physical state P grounds the fact that I am conscious. We then asked whether (C) has a ground, and the brute essentialist says: (C) (C) is grounded in the following two facts (C.i) I am in physical state P, and (C.ii) It is essential to being conscious that if I am in physical state P then I am conscious. But what about (C)? Does it have a ground? If not, then it is a counterexample to Weak Physicalism and so the possibility of physicalism is lost. To satisfy even just Weak Physicalism we must say that (C) is grounded in some Xs, and that the fact that those Xs ground (C) also has a ground, and so on. But does this not lead to regress?47 Yes, a regress of sorts, but not a problematic one. To see this, let us identify the canonical form of a brute essentialist explanation of ground, so that we can apply it to (C). Compare (C) with: (F) The fact that the event included people acting in way W grounds the fact that it was a conference. The brute essentialist says: (F) (F) is grounded in (F.i) The fact that those people acted in way W.
is analogous to the popular view that negative facts in general are grounded in the positive facts plus a totality fact. 47 Bennet [1] considers a similar regress. Her solution is somewhat different than mine, though, since she endorses simple reductionism.
46 This

32

(F.ii) The fact that it is essential to being a conference that if an event includes people acting in way W then it is a conference. These explanations of (F) and (C) exhibit a common form: some Xs ground Y, and this is explained partly by the Xs and partly by a fact about the essence of a constituent of Y that implies that the Xs are materially sufcient for Y. The brute essentialist then claims that all grounding facts admit of a canonical explanation of this same form. This description of the canonical form slurs over some details, not least over the notion of implication involved. And it does not determine which constituent of Y is the thing whose essence implies that the Xs are sufcient for Y. But the basic idea is clear enough for current purposes. Applying this canonical form to (C), we get that (C) is grounded in (C.i), (C.ii), and a fact about the essence of ground that implies that (C.i) and (C.ii) are sufcient for (C). Following the above examples, this essentialist fact might be something like (G) It is essential to ground that for any Xs and any Y, if the Xs obtain and if a fact about the essence of a constituent of Y implies that the Xs are materially sufcient for Y, then the Xs ground Y. For (G) implies that if (C.i) and (C.ii) obtain then (C) obtains. (To see this substitute in (C.i) for the Xs and the fact that I am conscious for Y). It may be hard to parse, but the proposed explanation is plausible. The question is: If it is essential to consciousness that I am conscious if I am in physical state P, and if I am in state P, why does this ground the fact that my being in state P grounds my being conscious? And the proposed answer is: Because that follows from the essence of ground. Given that I am in state P and given the essence of consciousness, it follows from what ground is that my being in state P grounds my being conscious. And on the next iteration, the question is why (C.i), (C.ii) and (G) ground (C), and the answer is because of (C.i), (C.ii), (G), and (G). Here (G) performs double duty: in the above canonical form it is one of the Xs, and it is also the fact about the essence of a constituent of Y. The pattern is clear. As we keep iterating, the (iterated) grounding fact is grounded in some combination of physical factsi.e. (C.i)and essentialist facts i.e.(C.ii) and (G). Which is consistent with Weak and Moderate Physcalism. We started with the one grounding fact (C). We now have innitely many grounding facts: that some Xs ground (C), that some Ys ground the fact that the Xs ground (C), that some Zs ground the fact that the Ys ground the fact that the Xs ground (C), and so on. Is this problematic? I think not. For one thing, it is not an innite descending chain of ground. It would be an innite descending chain if (C) is grounded in the Xs, and

33

the Xs are grounded in the Ys, and so on.48 But that is not the situation: it is not the Xs that are grounded in the Ys but rather the fact that the Xs ground (C) that is grounded in the Ys. So it remains open that the Xs are all ungrounded and that there are no innitely descending chains. One might instead object that on this view one grounding fact implies the existence of innitely many grounding facts. But I do not see why this is a problem. For one thing, any fact A implies the existence of innitely many facts A&A, A&(A&A), and so on. For another thing, notice that all the implied grounding facts use some combination of (C.i), (C.ii), and (G) as their grounds, so it is not that new facts are needed in each case. Finally, remember that I ofcially treat ground as a sentential operator and remain neutral on the existence of facts. So on my ofcial view the situation is just that one true because claim implies the truth of innitely many because claims, and I do not see what is problematic about that.49

12

Conclusion

I said at the beginning that one can think of physicalism pictorially as a multi-story building, with physical facts on the 1st oor, chemical facts on the 2nd oor, and so on. My view, I said, is that the ungrounded connections between the physical and the non-physical are not part of the building itself but are the scaffolding around which the building is built. I have tried to turn this picture into a theory. The facts in the building are substantive facts, with the ungrounded ones on the ground oor and grounded ones further up. The scaffolding are the autonomous facts (in particular, essentialist facts). They are not apt for grounding explanations and so do not appear on any particular oor of the building. Physicalism requires that the facts on the 1st oor of the building are all physical, but it allows that the scaffolding contain non-physical facts without grounds. In this way the possibility of physicalism as a grounding thesis is secured.50
sense in which each one of the Xs has a ground amongst the Ys. 49 Thanks to Tom Dougherty for an enlightening conversation about this section. 50 This material was presented in the spring of 2011 at a colloquium on Fundamentality at the APA Pacic Division meeting and in my graduate seminar at Princeton. It was also presented in January 2013 at Dartmouth University and a conference on The Philosophy of Kit Fine at NYU. Thanks to those present at all these events for their invaluable feedback. The paper has also benetted from enlightening conversations with Ralf Bader, Louis deRosset, Tom Dougherty, Kenny Easwaran, Branden Fitelson, Elizabeth Harman, Thomas Hofweber, AJ Julius, Shieva Kleinschmidt, Boris Kment, Kathrin Koslicki, Jon Litland, Eliot Michaelson, John Morrison, David Plunkett, Jonathan Schaffer, Susanna Schellenberg, Ted Sider, Will Starr, and Bruno Whittle.
48 When I say that the Xs are grounded in the Ys this can be understood in the distributive

34

References
[1] Bennett, K. By our Bootstraps. Forthcoming in Philosophical Perspectives. [2] Dasgupta, S. Metaphysical Rationalism. Manuscript. [3] Dasgupta, S. On the Plurality of Grounds. Manuscript. [4] Della Rocca, M. 2003. A Rationalist Manifesto: Spinoza and the Principle of Sufcient Reason. Philosophical Topics 31 (1): 7593. [5] deRosset, L. Grounding Explanations. Manuscript. [6] Fine, K. 1994. Essence and Modality: The Second Philosophical Perspectives Lecture. Philosophical Perspectives 8: 116. [7] Fine, K. 1995. Ontological Dependence. Proceedings of the Aristotelian Society (95): 269290. [8] Fine, K. 1995. The Logic of Essence. Journal of Philosophical Logic 24 (3): 241273. [9] Fine, K. 2001. The Question of Realism. Philosophers Imprint 1 (1): 130. [10] Fine, K. Guide to Ground. Manuscript. [11] Hofweber, T. 2009. Ambitious, yet Modest, Metaphysics. In Metametaphysics, edited by D. Chalmers, D. Manley and R. Wasserman. Oxford, UK: OUP. [12] Horgan, T. 1993. From Supervenience to Superdupervenience: Meeting the Demands of a Material World. Mind 102: 555586. [13] Jenkins, C. Forthcoming. Is Metaphysical Dependence Irreexive? The Monist 94. [14] Koslicki, K. Essence, Necessity and Explanation. Manuscript. [15] Leuenberger, S. Grounding and Necessity. Manuscript. [16] Ney, A. 2008. Dening Physicalism. Philosophy Compass 3: 10331048. [17] Raven, M. In Defense of Ground. Forthcoming in Australasian Journal of Philosophy. [18] Rosen, G. 2010. Metaphysical Dependence: Grounding and Reduction.

35

[19] Schaffer, J. 2009. On What Grounds What. In Metametaphysics, edited by D. Chalmers, D. Manley and R. Wasserman. Oxford, UK: OUP. [20] Schaffer, J. 2010. The least discerning and most promiscuous truthmaker. The Philosophical Quarterly 60: 307324. [21] Schaffer, J. 2012. Grounding, transitivity, and contrastivity. In F. Correia and B. Schnieder (eds), Metaphysical Grounding: Understanding the Structure of Reality. Cambridge: Cambridge University Press. [22] Skiles, A. Against Grounding Necessitation. Manuscript. [23] Sider, T. Writing the Book of the World, manuscript. [24] Trogdon, K. Grounding: Necessary or Contingent? Forthcoming in Pacic Philosophical Quarterly

36