You are on page 1of 22

The Journal of Political Philosophy: Volume 17, Number 4, 2009, pp.

418439

Liberal and Republican Freedom*


Boudewijn de Bruin
Philosophy, University of Groningen

I. INTRODUCTION HE revival of republican thought has prompted several liberal theorists to compare their notion of freedom as non-interference with the republican ideal of non-domination. Liberals have tried to accommodate the latter ideal, contending that the liberal notion of freedom can describe all that republican freedom can, except that whichin their viewno notion of freedom should be able to describe anyway.1 Republicans object to this argument, defending the cogency and irreducible originality of their concept of freedom (also called antipower, non-domination, or neo-Roman freedom).2 The ensuing exchange of ideasstill ongoing and very livelyshows increasing analytic clarity in both camps. This article adds to the debate by presenting a new counterargument against republican freedom. I start by investigating three relations between liberal and republican freedom: (i) Logical Equivalence, or the question whether republican freedom entails liberal freedom (and vice versa); (ii) Degree Supervenience, or
jopp_334 418..439

*Warmest thanks are due to Matthew Braham, Govert den Hartogh, Martin van Hees, Frank Hindriks, Jan-Willem van der Rijt and Bruno Verbeek, to the anonymous referees of this Journal and its editor, Robert Goodin, and to the audiences in Amsterdam and Groningen to whom I presented this paper in 2006 and 2007. I have beneted enormously from the gracious and detailed written comments by Ian Carter, Matthew Kramer and Philip Pettit which have helped to shape the views I defend in this essay. 1 Ian Carter, A Measure of Freedom (Oxford: Oxford University Press, 1999) and How are power and unfreedom related? Republicanism and Political Theory, ed. C. Laborde and J. Maynor (Oxford: Blackwell, 2008), pp. 5882. Robert Goodin, Folie rpublicaine, Annual Review of Political Science, 6 (2003), 5576. Matthew Kramer, The Quality of Freedom (Oxford: Oxford University Press, 2003) and Liberty and domination, Republicanism and Political Theory, ed. C. Laborde and J. Maynor (Oxford: Blackwell, 2008), pp. 3157. Kristjan Kristjnsson, Is there something wrong with free action? Journal of Theoretical Politics, 10 (1998), 25973. 2 Philip Pettit, Republicanism: A Theory of Freedom and Government (Oxford: Oxford University Press, 1997) and Republican freedom: three axioms, four theorems, Republicanism and Political Theory, ed. C. Laborde and J. Maynor (Oxford: Blackwell, 2008), pp. 10230. Iseult Honohan, Civic Republicanism (London: Routledge, 2002). Christian List, Republican freedom and the rule of law, Politics, Philosophy and Economics, 5 (2006), 20120. Quentin Skinner, Liberty Before Liberalism (Cambridge: Cambridge University Press, 1998) and Freedom as absence of arbitrary power, Republicanism and Political Theory, ed. C. Laborde and J. Maynor (Oxford: Blackwell, 2008), pp. 83101. Jean-Fabien Spitz, The concept of liberty in a theory of justice and its republican version, Ratio Juris, 7 (1993), 33147. Maurizio Viroli, Republicanism (New York: Hill and Wang, 2001). 2009 The Author. Journal compilation 2009 Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.
doi: 10.1111/j.1467-9760.2009.00334.x

LIBERAL AND REPUBLICAN FREEDOM

419

whether changes in the degree (amount, quantity) of republican freedom are mirrored by changes in the degree of liberal freedom (and vice versa); and (iii) Epistemological Priority, that is, whether knowledge about arrangements of republican freedom presupposes knowledge about arrangements of liberal freedom. If Logical Equivalence holds, liberals are right to claim that republicans have not introduced a new concept of freedom. It is easy to see, though, that Logical Equivalence does not hold (Section III). If Degree Supervenience holds, liberals can maintain that while republican freedom is a new concept of freedom, fostering or promoting it is not very different from fostering or promoting liberal freedom. Degree Supervenience does not hold either though (Section IV), and as a result two straightforward counterarguments against republican freedom fail. In this article I present a new counterargument. First, I argue that the relation of Epistemological Priority holds: knowing something about arrangements of republican freedom presupposes that you know something about arrangements of liberal freedom (Section V). Using Epistemological Priority, I show that the benets claimed for republican freedom over and above liberal freedomit minimizes the need for strategic deference, it minimizes uncertainty, and it minimizes subordinationcan be accounted for in purely liberal terms (Section VI).3 While close in spirit to critical evaluations of republicanism by authors such as Ian Carter and Matthew Kramer, this essay is concerned with arguments that differ from theirs in a variety of ways. Ignoring subtle yet signicant dissimilarities between their respective views, Carter and Kramer both present an equivalent-judgments thesis to the effect that while republicans and liberals disagree about the precise analysis of freedom, they give very similar answers to questions about degrees and distributions of freedom.4 Liberals maintain that in the few cases where there is disagreement, the republicans misuse the term freedom.5 While there is much to recommend this line of argumentation, it might lead some republicans to respond that it reduces the debate to one of terminology, not substance. In contrast to Carter and Kramer, I accept the republican terminology and develop an argument that cannot be countered by playing the terminology card.6 II. PRELIMINARIES Philip Pettit denes republican freedom as the negation of domination: non-domination. Agent T dominates agent S if and only if T has the capacity
These supposed benets are discussed in Pettit, Republicanism, pp. 8590. The term equivalent-judgments thesis is due to Carter, How are power and unfreedom related, p. 59. 5 Carter, A Measure of Freedom, pp. 2434. 6 Carter, How are power and unfreedom related, p. 59 observes that republicans would have to show that the equivalent-judgments thesis is false, but that this would force them to adopt a positive concept of freedom as autonomy.
4 3

420

BOUDEWIJN DE BRUIN

to interfere, on an arbitrary basis, in certain choices S is in a position to make.7 Interference, for Pettit, is a form of more or less intentional obstruction, caused by human actions, which makes the situation of the dominated party worse.8 What matters is that T could interfere with S if she wanted to, not that T is or is not disposed to interfereshe need not even have the slightest inclination to do so. The capacity for interference, moreover, has to be a capacity to interfere arbitrarily, which means that it is subject just to the arbitrium, the decision or judgment, of the [dominating] agent. It is chosen at the agents pleasure . . . without reference to the interests, or opinions, of those affected.9 This does not mean that T dominates S whenever T can perform certain actions that go against the interests of S. Even though, according to Pettit, interference as such already constitutes a worsening of someones situation by denition, not all capacities to perform actions that worsen someones situation are capacities for arbitrary interference. If T s worsening of Ss situation is to count as an arbitrary worsening, it has to be the case that T was not forced to track the relevant interests of S in her decision to interfere.10 To belabor the point, it is instructive to consider what it means to act against someones interests in a non-arbitrary fashion. The standard case in point is imprisonment. It typically goes against a persons interests to put her in jail; it worsens her situation. Yet, Pettit suggests, if a criminal is put in jail this is not an arbitrary form of intervention. First, it is not arbitrary because the judge who sentenced the prisoner to jail does not incarcerate the prisoner at her own pleasure, as her decision is backed by a well-functioning, well-enforced legal system affording the possibility of review. Even a legal system that works properly may be tyrannical, though, so the laws underlying the judges verdict must in addition be grounded in decisions made in the shared or common interests of the citizens. No denition of republican freedom will work until one has specied criteria governing the interests that are relevant here, as well as criteria about the acceptability of the collective decision procedures determining such interestsand it is the business of a defender of republican freedom to spell this out in detail.11 For our current purposes, however, we need not delve deeper. Before proceeding, I turn briey to an epistemic issue concerning republican freedom that plays an essential role in a large part of the essay. Republicans believe that as a matter of empirical truth the occurrence of relations of
Pettit, Republicanism, p. 52. Pettit, Republicanism, p. 53. Kramer, The Quality of Freedom, pp. 1335 observes that requiring the worsening to be intentional is problematic. The argument I present in this article also extends to a denition of republican freedom that drops this requirement. 9 Pettit, Republicanism, p. 55. 10 Pettit, Republicanism, p. 52 (emphasis mine). Cf. Pettit, A Theory of Freedom: From the Psychology to the Politics of Agency (Oxford: Oxford University Press, 2001), pp. 1389 for a denition in terms of avowable interests. 11 See Pettit, Republicanism, passim, and Pettit, The determinacy of republican policy: a reply to McMahon, Philosophy and Public Affairs, 34 (2006), 27583. An earlier article is Pettit, Freedom as antipower, Ethics, 106 (1996), 576604.
8 7

LIBERAL AND REPUBLICAN FREEDOM

421

domination and non-domination is completely open and transparent to the agents involved and to many of the bystanders.12 If T dominates S, then S and T and all other relevant agents know that. Not only that, they also know that they know, and know that they know that they know, and so on. It is common knowledge among S, T and the bystanders that T has the capacity arbitrarily to interfere with S. Moreover, whenever such capacities are absent, the fact that they are absent is common knowledge, too. In other words, republican unfreedoms and republican freedoms are common knowledge. I call this condition Common Knowledge of Republican Freedom (CKR).13 III. LOGICAL EQUIVALENCE Gerald MacCallum analyzed freedom as a relation between an agent S who is free from constraints c to perform an action a, and thus showed that the apparent dichotomy of negative and positive freedom, famously defended by Isaiah Berlin, is not one of logical form.14 Rather it should be cast in terms of the kind of objects that one considers to be agents, constraints and actions. A similar lesson can be drawn to enlighten the debate between liberals and republicans. As non-interference and non-domination both involve agents, constraints and actions, MacCallums triadic analysis also applies to their concepts of freedom.15 Since both concepts are concerned with the same sort of agents and the same sort of actions, the disagreement centers round what counts as constraints. Republican freedom takes capacities for arbitrary interference as constraints; liberal freedom, by contrast, takes physical barriers established by humans (or dispositions to establish such barriers) as constraints.16 As a result, liberal and republican freedom are clearly different within MacCallums schema:
Pettit, Republicanism, pp. 5861. CKR is not part of the denition of republican freedom, because it arises in most cases, but not in all. If T has the capacity to modify Ss preference for a particular brand of canned soup by subliminal advertising during a break in a baseball game, T dominates S according to Pettits denition, but only few people may know that. 14 Gerald MacCallum, Negative and positive freedom, Philosophical Review, 76 (1967), 31234. Isaiah Berlin, Two Concepts of Liberty (Oxford: Clarendon Press, 1958). MacCallum also included becomings in the third relatum. 15 Pettit suggests (personal communication) that as a fourth relatum in the relation one could adopt the standing protection against others that republican free agents may receive. This may be more appropriate analytically, but it would make a claim against Logical Equivalence trivial (three- and four-place relations can never be equivalent). A stronger claim results therefore if we assume that republican freedom ts in the MacCallum scheme, and then argue against Logical Equivalence. 16 Ian Carter, Positive and negative liberty, Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/entries/liberty-positive-negative, posted Oct. 8, 2007, notes that liberals may disagree about the types and sources of constraints of freedom. With respect to type I follow the pure negative theorists, taking only physical obstructions as constraints, excluding psychological barriers in the agents themselves, but allowing for dispositions (on the part of interfering agents) to establish physical barriers. With respect to source, my argument works for a denition of freedom that sees as constraints only those obstacles for which an agent can be held causally responsible, but also for a denition taking intentionally erected actions as sources of constraints. Roughly, then, the denition I use here is a standard negative liberal one according to which an agent S is liberally free to perform an action a whenever there is no agent T who physically blocks S from performing a or
13 12

422

BOUDEWIJN DE BRUIN

The two propositions S is liberally free to perform a and S is republican free to perform a are not logically equivalent.17

The precise argument is straightforward yet instructive. If Sue cannot reach home because the police are blocking the road to search a house linked to criminal activities, she is physically obstructed from reaching home and the obstructions are created by human actions. Liberally unfree, she is nevertheless free in the republican sense of the word, because the police do not interfere arbitrarily.18 The conclusion is that liberal freedom is not a necessary condition for republican freedom. To show that it is not a sufcient condition, consider a case in which Tania can arbitrarily bar Sue from going home, but in fact does not do so. Sue does not face any physical obstructions, is liberally free to go home, and will reach home whenever she attempts to, but all the same she is not republican free to go home.19 IV. DEGREE SUPERVENIENCE Logical Equivalence does not hold, and hence one straightforward liberal argument against republican freedom breaks down. In this section I investigate a second such argument. If Degree Supervenience holds, liberals can maintain that while republican freedom is a new concept of freedom, any change in the distribution of republican freedom is reected by a change in the distribution of liberal freedom.20 But Degree Supervenience does not hold, and as a result a second straightforward counterargument against republican freedom fails. A property P supervenes on a property Q if and only if it is necessarily the case that if two worlds differ qua P they also differ qua Q, or equivalently, if it is not
who possesses the disposition so to block S. The arguments presented here apply, then, not only to the denitions of Carter and Kramer, but also to those of: Berlin, Two Concepts of Liberty; Friedrich von Hayek, The Constitution of Liberty (London: Routledge, 1960); David Miller, Constraints on freedom, Ethics, 94 (1983), 6686; Felix Oppenheim, Political Concepts: A Reconstruction (Oxford: Blackwell, 1981); and Hillel Steiner, An Essay on Rights (Oxford: Blackwell, 1994). 17 More precisely, these two propositions are not logically equivalent without extra assumptions. Boudewijn de Bruin, A note on Lists modal logic of republican freedom, Politics, Philosophy and Economics, 7 (2008), 3419 proves an equivalence result adopting two extra modal axioms (the plausibility of which may be questioned, of course). 18 The use of human intervention is not essential. The argument still runs if we assume that Sue is constrained by heavy rainfall. Pettit, Republicanism, p. 56n would say that the rainfall does not compromise Sues freedom, but only conditions it in the sense of narrowing the available choice options. Kramer, The Quality of Freedom, p. 3 denes liberal freedom to perform an action as being able to perform it, and liberal unfreedom as, roughly, not being able to do it because of human intervention. Cf. Steven Wall, Freedom, interference and domination, Political Studies, 49 (2001), 21630. 19 Pettit, Negative liberty, liberal and republican, European Journal of Philosophy, 1 (1993), 1538 contains a denition of republican freedom as resiliently realized non-interference that has fallen into desuetude. It is clear that with this denition republican freedom entails liberal freedom, but not the other way round. 20 The presupposition that degrees of overall freedom can be measured is warranted given an emerging literature on measures of freedom. See, e.g., Carter, A Measure of Freedom, and Kramer, The Quality of Freedom. Neither of them argues for full Degree Supervenience, though.

LIBERAL AND REPUBLICAN FREEDOM

423

possible that there is a P-difference without a Q-difference.21 Against Degree Supervenience I argue, then, that it is possible that two worlds are different qua republican freedom without being different qua liberal freedom.22 Consider two worlds v and w. In both worlds Sue can push a button to call the elevator of her apartment building, which is the only way for her to get to her apartment. In world v Tania stands in front of the elevator, blocking the button, but as soon as Sue approaches the button Tania gives way. In contrast to what this phrasing suggests, though, Tania is not a subservient assistant, for she has all the physical and psychological power to bar Sue from taking the elevator, and the current state of anarchy in the country does not impose any legal restrictions on her dealings with Sue. But while Tania has the capacity arbitrarily to interfere with Sues going home, she never actually bars her from pressing the button. World w, in turn, is different only in that Tania does not inhabit it, and as result of this Sue has a strictly greater degree of republican freedom in w than in v. While in w she has any republican freedom she has in v, she has the extra republican freedom to push the button and go to her apartment.23 But although the degree of Sues republican freedom is different in the two worlds, she has the same degree of liberal freedom; there is no difference with respect to pushing elevator buttons, nor is there any difference with respect to other actions.24 To see this, note that while in w Sue cannot (of course) talk to Taniain the sense that no one can talk to someone who is absentthat does not constitute an unfreedom according to a negative, liberal concept of freedom,
21 Note that (P-difference Q-difference) is logically equivalent to (P-difference Q-difference). 22 A number of arguments put forward in the literature come close to arguing against Degree Supervenience. See, e.g.: Pettit, Negative liberty, liberal and republican, p. 19; Pettit, Republicanism, pp. 738; Pettit, Republican freedom: three axioms, four theorems, pp. 1116; Skinner, Liberty Before Liberalism, p. 70; Skinner, Freedom as the absence of arbitrary power, pp. 96100. For arguments in favor of it, see: Carter, A Measure of Freedom, p. 240; Carter, How are power and unfreedom related? pp. 6971; Kramer, The Quality of Freedom, pp. 10524; Kramer, Liberty and domination, pp. 4150. The republican arguments show no sensitivity to the fact that changes in republican freedom often entail changes in liberal freedom. The liberal arguments show no awareness of the fact that changes in republican freedom do not necessarily entail changes in liberal freedom. The equivalent-judgments thesis is to the effect that republicans and liberals agree quite extensively in their judgments concerning freedom, but this thesis allows for exceptional cases in which differences in republican freedom are not mirrored by changes in liberal freedom. Unlike Degree Supervenience, the equivalent-judgments thesis concerns judgments about freedom rather than degrees of freedom, and it is not implied by my arguments. Yet, to the extent that failures of Degree Supervenience are rare, the equivalent-judgments thesis is compatible with Degree Supervenience. Determining the exact extent, however, is an empirical issue. 23 The kind of supervenience used here is global supervenience to the effect that for any two worlds v and w, if v and w have the same distribution of P-properties, they also have the same distribution of Q-properties. This assumes that there is a bijection between the domains of worlds v and w. See Robert Stalnaker, Varieties of supervenience, Philosophical Perspectives, 10 (1996), 199206. The present example can be easily (though slightly articially) modied to t this format. Both worlds have two apartment buildings. World v remains as in the argument. World w does not remove Tania, but places her in front of the elevator of the second apartment building. (Just moving Tania to a place far away from Sue does not work as she will lose the republican freedom to push an elevator button.) 24 The truth of this statement is independent of whether degrees of freedom involve ordinal or cardinal measurements.

424

BOUDEWIJN DE BRUIN

because in w she can still perform the action (utter the words) that in v would constitute talking to Tania (no one interferes with Sues doing that). Nor is Sues freedom decreased in that she can no longer carry her piano upstairs with Tania, for this concerns group freedom, not individual freedom (and anyway, no one interferes with the group doing that).25 This establishes the following claim:
The degree of republican freedom does not supervene on the degree of liberal freedom.

The converse relation of Degree Supervenience does not hold either; liberal freedom does not supervene on republican freedom. To defend this claim I show that it is possible that two worlds are different qua liberal freedom without being different qua republican freedom. The argument can be given quickly. Imagine two worlds v and w where Tania has the capacity arbitrarily to interfere with Sues going home. Being Sues master in v, Tania is genuinely disposed to exercise this capacity. In w, by contrast, Tania is Sues friend and will not interfere. By stipulation, Sue has the same degree of republican freedom in v and w, for even though Tania likes Sue in w, Tania still dominates her. Yet Sue has different degrees of liberal freedom. In v Tania cannot go home; in w she can.26 Thus we nd:
The degree of liberal freedom does not supervene on the degree of republican freedom.

V. EPISTEMOLOGICAL PRIORITY Neither Logical Equivalence nor Degree Supervenience hold, and this means that two straightforward ways to criticize republican freedom are closed. I now turn to an investigation of a third relation between liberal and republican freedom, that of Epistemological Priority.27 I show that if someone possesses knowledge about arrangements of republican freedom she also possesses knowledge about arrangements of liberal freedom. In the next section I show how Epistemological Priority can be used in an argument against republican freedom. A disclaimer rst. The claim that knowledge about republican freedom presupposes knowledge about liberal freedom does not apply to knowledge obtained by testimony. If a reliable source tells me that Tania dominates Sue, and if I do not have more information concerning Sue and Tania, then this entails nothing at all about my knowledge of arrangements of liberal freedom. So the
Carter, A Measure of Freedom, p. 230 considers a related case in terms of group freedom. If one is unwilling to grant probability one interference in world v, the claim is still valid, but now only for cardinal measures of freedom taking care of probability: the degree of liberal freedom in v is less than the degree of liberal freedom in w, because there is a higher probability of interference in v than in w. Carter, A Measure of Freedom and Kramer, The Quality of Freedom incorporate probabilistic judgments in their measures of freedom. 27 Martin Davies, The philosophy of language, The Blackwell Companion to Philosophy, 2nd edn, ed. N. Bunnin and E. P. Tsui-James (Oxford: Blackwell Publishers, 2003), pp. 90146.
26 25

LIBERAL AND REPUBLICAN FREEDOM

425

scope of the claim is rather knowledge based on perception, inference and statistical reasoning: primary knowledge.28 This is no serious restriction, though. First, much of our knowledge about domination is based on our own observations and is consequently primary. Moreover, even if we obtain knowledge about domination by testimony, this ultimately has to be grounded in primary knowledge, because a chain of epistemic agents testimonially transferring knowledge about a proposition always needs a prime mover with primary knowledge about the proposition. The claim, then, is that primary knowledge about arrangements of republican freedom presupposes primary knowledge about arrangements of liberal freedom. The rst step in defending this claim is to show that primary knowledge about domination presupposes primary knowledge about unequal distribution of resources. Correlating resource inequality and domination is, in fact, part and parcel of the republican literature:
The resources in virtue of which one person may have power over another are extraordinarily various: they range over resources of physical strength, technical advantage, nancial clout, political authority, social connections, communal standing, informational access, ideological position, cultural legitimation, and the like. They also include the resource of being someonesay, the only doctor or police ofcer aroundwhose help and goodwill the other may need in various possible emergencies. They even include the resource of perceived intractabilityat the limit, perceived irrationalitythat enables someone to drive a hard bargain. . . . Such resources tend to be prominent and detectable by those to whose disadvantage they may be deployed, and . . . this helps ensure that where one person has any dominating power over another, in virtue of an inequality in such resources, it is a matter of common knowledge that that is so.29

To obtain knowledge about Tanias capacity for arbitrary interference with Sue, I have to know that there is resource inequality between Sue and Tania, and I have to know that the fact that there is resource inequality entails that there is domination. The rst element is clear:
(1) There is a situation x of which I know that x is a particular instance of resource inequality affording more to Tania and less to Sue.

The second element postulates knowledge about the republican tenet from the above quotation, that is, knowledge about a certain empirical regularity correlating resource inequality and domination:
(2) I know that for all situations x and all agents S and T, x is an instance of resource inequality affording more to T and less to S whenever x is an instance of domination of S by T.30
28 Robert Audi, The sources of knowledge, The Oxford Handbook of Epistemology, ed. P. Moser (Oxford: Oxford University Press, 2002), pp. 7194. 29 Pettit, Republicanism, pp. 5960. 30 A probabilistically more sophisticated rendering of (2) is cast in terms of the statistical correlation between the two random variables measuring resource inequality and domination, but

426

BOUDEWIJN DE BRUIN

If these two conditions are met, it follows that I have knowledge of a particular instance of republican unfreedom:
(3) There is a situation x of which I know that x is a particular instance of domination of Sue by Tania.

Where can we nd knowledge about liberal freedom? Statement (1) clearly presupposes no knowledge about liberal freedom. As it is assumed that I do not obtain knowledge by testimony, I obtain the knowledge described in (1) by perceiving, for instance, Sues frail build in comparison with Tanias, the difference in nancial clout between them, or their different social status, but not their respective freedoms. This does not mean that resource inequality can always be read off so easily. Evidence about resource inequality between Tania and Sue due to Tanias perceived intractability, as Pettit mentions in the earlier quotation, has to refer to Sues way of perceiving Tania. While Tania may seem intractable to me, this appearance does not form a resource inequality between Sue and Tania as long as Sue is perfectly able to make sense of Tania; Sue, too, would have to nd Tania intractable. But even so, the knowledge embodied in (1) is no knowledge about interference.31 Turning to (2), we rst have to understand the specic epistemic access human beings have to capacities; for while resource inequality can be established without further ado, primary knowledge about capacities requires a detour. A perfectly general example illustrates this. Suppose I want to know whether zebra nches can sing at lower frequencies if their testosterone levels increase. While I have direct perceptual access to the intake of hormones, my epistemic access to a nchs capacity to sing certain frequencies goes via observations: recordings of birds actually singing. As a result, to make a claim about the correlation between testosterone and capacity for singing lower frequencies is to make an inductive generalization that is based on a nite set of observations of increased testosterone levels in certain zebra nches and the actual singing of particular frequencies.32

this does not increase analytical precision in the present case. Republicans hold that (almost) all cases of resource equality go together with domination. Since, moreover, both random variables only take two values (yes or no), the correlation coefcient will be (almost) equal to one, thus entailing that my rendering of (2) is valid in (almost) all cases. An advantage of my formalization is that it brings out the logical dependence of the propositions more clearly. 31 The logical form of (1) is $x Know RI(x, Tania, Sue) featuring de re knowledge of a particular instance of resource inequality x to stress that I have primary knowledge about x. De dicto knowledge as in Know $x RI(x, Tania, Sue) would be appropriate if my knowledge were based merely on someone telling me that there is resource inequality between Tania and Sue, or my concluding it from certain general facts in a way that does not conserve primariness of knowledge. 32 Jeffrey Cynx, N. Jay Bean and Ian Rossman, Testosterone implants alter the frequency range of zebra nch songs, Hormones and Behavior, 47 (2005), 44651. Plenty of other examples could be given relating, for instance, air temperature to athletic achievements, IQ to academic potential, or music to memory capabilities.

LIBERAL AND REPUBLICAN FREEDOM

427

Similarly, (2) represents an inductive generalization based on a nite set of observations about instances of resource inequality that were signicantly often accompanied by instances of actual arbitrary interference.33 As an intermediate claim we thus nd:
Primary knowledge about republican unfreedom to perform a-like actions entails primary knowledge about actual instances of resource inequality that were signicantly often accompanied by arbitrary interference with a-like actions.34

While this goes some way to arguing for the main claim of this section, it is not enough for a full argument. The reason is the notion of arbitrariness. To progress, it is instructive to look at the differences between interference under resource inequality and interference under resource equality. It is plain that under resource equality interference by no means becomes impossible. If Sue and Tania are of equal physical strength, Tania can still push Sue from the sidewalk by surprise, and even if Tania is one of many doctors in town, Tania can still decide to help Sue only if Sue pays double the normal fee. But there is a difference. In situations of resource equality, Sue can counter. Countering sometimes means striking back, as in the case of two pedestrians of equal strength; sometimes it means sidestepping Tania, as in the case of Sue turning to another doctor in town; and it is precisely due to resource equality that Sue can counter Tanias interfering actions.35 The claim now is that arbitrary interference is interference that cannot be countered. To summarize in a second intermediate claim:

33 This is reected in the logical form of (2), Know "x"S"T (RI(x, T, S) Dom(x, T, S)). This statement ranges over all instances x and individuals S and T (it is an inductive generalization), but it is backed by observations of only nitely many such x, S and T. (If I had observed all, then the de re statement "x"S"T Know (RI(x, T, S) Dom(x, T, S)) would be appropriate.) In particular, the observations will probably not include resource inequality between Sue and Tania. To compare, I have seen hundreds of black ravens (observation of correlation) and conclude that all ravens are black (inductive generalization). I now hear a raven croak and conclude it is black, even though I have never even seen it. 34 The point is subtle as (3) embodies knowledge about a capacity not based on an observation of an instantiation of that capacity. Rather, knowledge about the existence of the capacity is derived from (1) and (2), and it might seem that a similar derivation could yield knowledge about the existence of capacities in (2) itself. This is true, but then (2) would itself require a second-order inductive generalization. This generalization would in turn either have to be based on observations of instantiations of capacities, or require a third-order inductive generalization, etc. This process of invoking higher-order inductive generalizations cannot continue innitely, because primary knowledge ultimately has to be based on observation. Without loss of generality I assume that the rst-order inductive generalization (2) subsumes all observations on which these subsequent higher-order generalizations are based. 35 One denition of arbitrariness is to the effect that arbitrary actions are not forced to track the relevant interests of the people concerned. See, e.g., Pettit, Republicanism, p. 55. If relevance can be defended independently of the concepts of resource, capacity and interference, then a liberal can grasp it without being committed to republican freedom. In that case I can safely leave it out of my argument. If, on the other hand, it cannot be dened independently, then the concept of republican freedom is circular. I take it not to be circular, though, and conceptualize it in terms of counterability. Also see Pettit, Republican freedom: three axioms, four theorems.

428

BOUDEWIJN DE BRUIN

Primary knowledge about republican unfreedom to perform a-like actions entails primary knowledge about actual instances of resource inequality that were signicantly often accompanied by non-counterable interference with a-like actions.

As much as primary knowledge about capacities rests on knowledge about actual instantiations of capacities, knowledge about non-counterable interference is ultimately grounded in knowledge about non-countered interference. The knowledge captured by (2) depends, then, on an inductive generalization over a nite number of observations of cases in which resource inequality co-occurred signicantly often with non-countered interference. If in a signicant number of cases of resource inequality the better endowed agent interferes with the less well endowed agent, and the latter agent does not counter, then this is evidence that resource inequality is correlated with capacity for non-counterable, arbitrary interference. This, of course, involves knowledge about arrangements of liberal freedom. It presupposes, to be precise, knowledge about the liberal unfreedom to perform the a-like action that the interferer blocks, and knowledge about the liberal unfreedom to perform another action that would counter the interferers blocking of the performance of an a-like action. It is, for instance, knowledge about the fact that a doctor can interfere with a patient obtaining treatment for a standard fee, and knowledge about the fact that she cannot visit any other doctor who would give her the treatment for the standard fee. All in all, the claim defended in this section is:
Primary knowledge about republican unfreedom to perform a-like actions entails primary knowledge about actual instances of resource inequality that were signicantly often accompanied by (i) interference with a-like actions and (ii) interference with actions countering interference of a-like actions.

This establishes that primary knowledge about arrangements of republican unfreedom presupposes primary knowledge about liberal unfreedom. What about knowledge about freedom? There is an interesting interpretative issue here that complicates matters a little. To see this, let p be the proposition that no one at all can interfere with Sues going home, and let q be the proposition that, even though no one can arbitrarily interfere with Sues going home, someone can interfere non-arbitrarily. Now, if I know that Sue is republican free to go home, does this mean that I have knowledge of a disjunction of two propositions, so that Know (p q) fully describes my epistemic state? Or does it mean that I have knowledge (more detailed than the rst reading) about one of two non-disjunctive propositions, so that either Know p, or Know q fully describes my epistemic state?36 There seems to be no reason in principle to prefer one reading over another, so I consider both.
36 Know(p q) does not entail (Know p Know q). That you know that either McCain or Obama will win the US presidential elections does not necessarily entail either that you know that McCain will win, or that you know that Obama will. This interpretative issue does not arise with respect to

LIBERAL AND REPUBLICAN FREEDOM

429

If we take it that knowledge about republican freedom is knowledge about a disjunction of two propositions, then deleting the ins and uns from (1), (2) and (3) yields a representation of how I arrive at knowledge about republican freedom. Clearly the inductive generalization involving resource inequality can no longer be used, so a new inductive generalization has to be made on the basis of observations of resource equality signicantly often accompanied by non-interference (see proposition p) or countered interference (see proposition q), and this presupposes knowledge about liberal freedom and about liberal unfreedom to perform some action together with liberal freedom to perform relevant countering actions. If, alternatively, knowledge about republican freedom is knowledge about one of two non-disjunctive propositions, then this inductive generalization does not contain enough information. It does not make a difference between the kind of resource equality that entails non-interference (proposition p) and the kind that entails non-arbitrary interference (proposition q). Two inductive generalizations are needed in that case, one relating resource equality of kind P and non-interference, the other relating resource equality of kind Q and non-arbitrary interference. Furthermore, to make this effective, (1) has to be modied not only to characterize the present situation as one of resource equality, but also to specify whether it is resource equality of kind P or Q. This sounds rather technical, but it is of course nothing other than distinguishing between such situations as when someone is sitting in her living room (kind P), and situations such as when someone is shoplifting (kind Q), and while again (the modied version of) statement (1) does not presuppose knowledge about liberal freedom, the inductive generalization cannot get off the ground without knowledge about non-interference. What is true for the earlier inductive generalization applies here, too: to relate resource equality of kind P to non-interference presupposes knowledge about liberal freedom, and to relate resource equality of kind Q to nonarbitrary interference presupposes knowledge about liberal unfreedom to perform some action together with liberal freedom to perform relevant countering actions. Thus we nd:
Primary knowledge about republican freedom to perform a-like actions entails primary knowledge about actual instances of resource equality that were signicantly often accompanied by (i) non-interference with a-like actions and/or (ii) interference with a-like actions and non-interference with countering actions.

knowledge about republican unfreedom. Knowledge that Tania has the capacity to interfere with Sue does not plausibly entail knowledge about whether she will interfere or not. (If it did, it would directly constitute knowledge about liberal freedom, and the present argument would be much shorter.) See also the third reply to minimization of uncertainty.

430

BOUDEWIJN DE BRUIN

VI. SUPPOSED BENEFITS OF REPUBLICAN FREEDOM OVER LIBERAL FREEDOM Republicans maintain that republican freedom has certain benets that liberal freedom lacks; it leads, they claim, to a minimal need to display strategic deference, minimal uncertainty, and minimal subordination. In this section I show that these benets are not uniquely republican as they can be understood in purely liberal terms. A. MINIMIZATION OF THE NEED FOR STRATEGIC DEFERENCE The rst supposed benet of republican freedom is that it minimizes the need to display strategic deference. The inhabitants of republican worlds have more rarely to resort to placating the powerful and pleasing their superiors than those of liberal worlds, it is claimed, and they can accomplish their goals signicantly more often without the toadying and currying characteristically loathed by republicans. In a maximally liberal world l, the republican argument goes, various third agents may possess capacities for arbitrary interference, and it is exactly the exercise of capacities for arbitrary interference that one can often successfully allay by displaying strategic deference. In a maximally republican world r, however, there will be far fewer such agents and far fewer such capacities, and therefore one will not as often be forced to keep a weather eye on the powerful, as Pettit puts it, and turn to displays of deference.37 Although the liberal retort to this claim does not appeal to Epistemological Priority, I mention it as it informs the replies to the next two supposed benets of republican freedom, and also because it sheds more light on Degree Supervenience. Suppose that in some world v the number of cases where individuals have to show strategic deference is rather extensive, and that, for instance, Sue has to write three elaborate letters to her superiors in order to be able to go on vacation. Then, the liberal argues, world v is not the maximally liberal world, for we can easily extend Sues freedom by giving her the option of going on vacation after calling her superiors only once. This is a signicant increase in Sues freedom (and if it represents any decrease of her superiors freedoms at all, it is only minor), and is therefore fully justied from the liberal point of view. The maximally liberal world, that is, contains far fewer possibilities for arbitrary interference than republicans tend to believe.38
Pettit, Republicanism, p. 86. This does not contradict the argument against Degree Supervenience. While changes in republican freedom are not in all cases reected by changes in liberal freedom (or vice versa), I am concerned here with a very specic kind of case, namely, where someone is perfectly capable of reaching her goals, but only via the detour of strategic deference, and removing the detour not only increases republican, but also liberal freedom.
38 37

LIBERAL AND REPUBLICAN FREEDOM

431

B. MINIMIZATION OF UNCERTAINTY The idea underlying the second benet of republican freedom, minimization of uncertainty, is that inhabitants of republican worlds will face signicantly less uncertainty with respect to the actions from which they can choose than do inhabitants of liberal worlds. A maximally liberal world l contains more actions about which the inhabitants do not know whether they can perform them than a maximally republican world r, because supposedly in l numerous third agents may possess capacities for arbitrary interference, of which one can never exactly predict whether they will be exercised or not. In r, by contrast, there are far fewer such agents and far fewer such capacities, and therefore one does not face such uncertainty as frequently. Or so republicans argue. There are four liberal replies. The rst copies the reply to the previous supposed benet of republican freedom. The second considers a maximally liberal world l and a maximally republican world r under the assumption that there is Common Knowledge of Republican Freedom (CKR) in both worlds; it is shown that republican freedom only minimizes uncertainty if CKR obtains. The third reply considers l and r under the assumption that in r there is CKR, while in l the analogous assumption of Common Knowledge of Liberal Freedom (CKL) obtains, thus arriving at the remarkable conclusion that, under these assumptions, liberal freedom minimizes uncertainty. The fourth reply, nally, uses Epistemological Priority to show that CKR (which the second reply showed to be necessary so that the argument from minimization of uncertainty can be run) is grounded in knowledge about liberal freedom. 1) The rst reply just repeats the reply to the rst supposed benet of republican freedom, casting doubt on the claim that the number of capacities for arbitrary interference in l is as high as the republicans suggest. So let us turn to the second reply. 2) The second reply grants republicans the view of the number of capacities for arbitrary interference in worlds l and r, but points out that the minimization of uncertainty is a consequence not of the specic concept of freedom that is maximized in r (non-domination), but of the specic epistemic assumption (CKR). Without CKR, that is, l and r score equally on uncertainty. To see how CKR fullls its function, consider Sue and Tania living in world l. Suppose that Tania has the capacity arbitrarily to interfere with Sues going home and assume that in world r this capacity does not exist. If CKR obtains in both worlds, Sue knows in l about the presence of Tanias capacity arbitrarily to interfere with her going home, and she knows in r about the absence of that capacity. Indeed, in l the assumption of CKR does not give Sue much knowledge. Knowing that Tania has a capacity for arbitrary interference does not as such mean knowing whether she will exercise that capacity, and as a result Sue remains completely uncertain in l about whether she will face any interference from Tania in her going home.

432

BOUDEWIJN DE BRUIN

In r, however, the assumption of CKR does give Sue some extra knowledgeprovided we interpret knowledge about republican freedom in the non-disjunctive wayfor Sues knowing that Tania cannot arbitrarily interfere with her going home then entails that either:
Sue knows that p, that is, she cannot be interfered with at all,

or
Sue knows that q, that is, she can be interfered with non-arbitrarily, but not arbitrarily.39

If in r Sue knows that p, then she denitely knows more than she knows in l (under the assumption that there is CKR in l), for as we saw, although she knows in l that Tania possesses a capacity for arbitrary interference, she does not know whether Tania will exercise it. Sue does not know in l whether she will succeed in going home if she tries, but in r she knows she will succeed. This is clearly an increase in certainty. If, on the other hand, Sues knowledge about republican freedom entails that Sue knows that q, this constitutes more certainty than she has in l only if Sue knows how to estimate the likelihood that non-arbitrary interference takes place. She may have such knowledge, for instance, when she observed that police ofcers block roads when they are engaged in crime scene investigation or when there are accidents (and not when they are checking incorrectly parked cars, say). But such knowledge does not directly ensue from CKR.40 Furthermore, full certainty is rare here, due to a systematic asymmetry between the information of ordinary citizens and, say, the police. Nonetheless, it is true that in a maximally republican world there is less uncertainty than in a maximally liberal worldprovided, of course, that there is CKR in both worlds and that knowledge about republican freedom is interpreted as a relatively detailed epistemic state involving knowledge about one nondisjunctive proposition rather than about a disjunction of two propositions. Liberals may remain unconvinced, however, as long as this benet is due to additional epistemic assumptions rather than to the concept of republican freedom itself. 3) Yet even if liberals license the introduction of epistemic assumptions, they could reasonably object to the precise assumption of CKR. While it is fair to assume CKR in a republican world, it is unnatural, the third reply states, to assume common knowledge of republican freedom in a liberal world. The

39 The conjunction of Know(p q) and Know(p , q) entails that Know q. Here and elsewhere I assume that Sue does not face any natural obstacles, and that she knows this. 40 Such knowledge is grounded in primary knowledge correlating interference to certain situations (accidents, etc.), and non-interference to others, and thus on knowledge about liberal freedom and unfreedom.

LIBERAL AND REPUBLICAN FREEDOM

433

comparison ought to be between a maximally republican world r where CKR obtains and a maximally liberal world l where there is common knowledge of liberal freedom (CKL). Such an alternative comparison does not support the benet of minimization of uncertainty, though. Clearly, Sues knowledge in r (with CKR) is what it was before; either she knows that p (she cannot be interfered with at all), or she knows that q (she can be non-arbitrarily interfered with, but not arbitrarily). So let us turn to l. As Sue is either liberally free to go home or not in that world, CKL in l entails that either:
Sue knows that s, that is, she will not be interfered with,

or
Sue knows that t, that is, she will be interfered with.41

But with these assumptions Sue does not have less certainty in l than in r; she possibly has even more certainty. To see this, rst observe that knowledge about p and knowledge about s yield the same level of certainty. In both cases Sue knows she can go home.42 Moreover, if she knows that t, then she also has complete certainty, because she then knows for sure that she cannot go home. Only knowledge about q gives less than full certainty. While Sue probably has the expertise to estimate the likelihood of non-arbitrary interference to some extent, knowledge that non-arbitrary interference is possible does not reach the degree of full certainty. In fact, if CKR entailed that each agent is always completely certain about the likelihood of non-arbitrary interference, common knowledge of republican freedom would be entirely equivalent to common knowledge of liberal freedom. This is a conclusion republicans ought to resist, for they have to allow, for example, for the possibility that Sue, speeding on a highway at night and seeing a police ofcer engaged in crime scene investigation, does not know for sure whether the ofcer will exercise her capacity for non-arbitrary interference. As a result, with analogous epistemic assumptions, inhabitants of a maximally liberal world have no less certainty than those of a maximally republican world. A republican might object that unlike CKR, the assumption of CKLentailing in the example that Sue knows for sure what the police ofcer will dois unrealistic. But that would miss the arguments point. The second reply to the supposed benet of minimization of uncertainty is that this benet follows from the epistemic assumption of CKR, not from the very concept of republican freedom. Granting the acceptability of introducing additional epistemic assumptions, the third reply objected that, as much as it makes sense to

41 Knowledge of a disjunction does not give any knowledge in the case of liberal freedom, so the non-disjunctive reading is the only plausible one. 42 The subtle modal difference between knowledge of p (will not) and knowledge of s (cannot) is immaterial here.

434

BOUDEWIJN DE BRUIN

investigate CKR in republican ideal worlds, it makes sense to investigate CKLthe analog for liberal freedom of CKRin liberal ideal worlds. 4) With non-analogous epistemic conditions, republican worlds may have more certainty than liberal worlds; with analogous epistemic conditions, it is the other way round. But the liberal can even grant non-analogous epistemic conditions, and turn to a fourth reply. This is to the effect that the knowledge that arises under CKR in a maximally republican world can be accounted for in purely liberal terms. Given the previous work on Epistemological Priority and on the logical form of knowledge about republican freedom, the argument can be brief. As we saw, Sues knowledge about republican freedom under the non-disjunctive interpretation is either knowledge about p or it is knowledge about q. Such knowledge, now, is grounded in two inductive generalizations, one correlating resource equality of kind P to non-interference, and one correlating resource equality of kind Q to countered interference. But that is knowledge about liberal freedom, and similar argumentation shows that this also holds for the disjunctive reading of knowledge of republican freedom. If, to conclude, inhabitants of republican worlds have more certainty than those of liberal worlds, then the knowledge that affords these higher degrees of certainty is just knowledge about arrangements of liberal freedom. C. MINIMIZATION OF SUBORDINATION The third benet of republican freedom, minimization of subordination, is supposed to be that inhabitants of a maximally republican world r signicantly more often feel that they are of equal status than inhabitants of a maximally liberal world l. While most inhabitants of r can supposedly look each other in the eye without any need to bow and scrape, in l a signicant number of individuals feel subordinated.43 There are two ways for the liberal to reply. 1) The rst is again an analog of the earlier observation that the number of capacities for arbitrary interference in l may not be as high as the republicans suggest, and that as a result subordination in l and r may be comparably extensive. If Tania can arbitrarily interfere with Sue and if there is some chance that Tania will exercise this capacity, then liberals too can describe the situation as less than optimal and as one in which Sues freedom is restricted. Subordination, in other words, is liberal unfreedom. 2) Yet republicans can take issue with this response on the ground that it leaves untouched cases where there is zerogenuinely nilprobability of interference. Republicans hold the view that Tania subordinates Sue even when
43 Pettit, Republicanism, p. 87. Whether or not this kind of effect can indeed be demonstrated is a matter for empirical psychology. For the sake of argument I grant the republican that this is in fact the case. See also Jan-Willem van der Rijt, Republican dignity: the importance of taking offence, Law and Philosophy (forthcoming).

LIBERAL AND REPUBLICAN FREEDOM

435

there is a genuine zero probability that she actually interferes. Since liberals evaluate such a situation as one of freedom (Sue will not encounter any interference), maximizing liberal freedom does not necessarily minimize subordination. Or so the republicans demur. One way for the liberals to respond to this objection is to argue that the cases the republicans invoke do not actually involve genuine zero probability. As such argumentation has to remain dependent on the specic details of the cases, I here offer a completely general reply to the effect that zero probability cases can never arise in genuinely republican contexts.44 To start with, it is important to see that, as before, the republican argument for the supposed benet of minimization of subordination requires that arrangements of republican freedom and unfreedom are common knowledge (CKR). To see this, suppose that Tania subordinates Sue. If Sue does not know that Tania subordinates her, Sue will not have any problem looking Tania in the eye. Tania may of course nd Sues looks audacious or presumptuous, but that again depends on an additional level of CKR, namely, whether Tania herself knows that she dominates Sue. So for there to be genuine subordination, Sue has to know about Tanias capacity for arbitrary interference, Sue has to know that Tania knows that Sue knows about the capacity, and so on, showing that all levels of CKR are required.45 A second thing to note is that there is nothing conceptually or empirically amiss with capacities that have a zero probability of being exercised, and in fact something of this sort was exploited in the arguments about Logical Equivalence and Degree Supervenience.46 To borrow an example from Isaac Levi, that it is physically possible for someone to run a four-minute mile is completely consistent with there being zero probability that she will actually accomplish such a feat.47 As a result, an argument that zero probability cannot arise in cases that are relevant to the issue of minimization of subordination has to refer to specic characteristics of these cases. And indeed it does, for the argument is precisely that zero probabilities are impossible when CKR comes into play. The argument about Epistemological Priority revealed that knowledge about republican unfreedom rests on knowledge about an actual instance of resource inequality and knowledge about an inductive generalization correlating resource inequality and non-countered interference. Now, clearly, not all observations about resource inequality are observations of non-countered interference, but if a sufciently large number of cases show non-countered interference, this is evidence to the effect that in all

Carter, A Measure of Freedom, pp. 2434. Kramer, The Quality of Freedom, pp. 13843. Pettit, Republicanism, pp. 878 explicitly acknowledges as much. 46 Primary knowledge of that capacity is then impossible, of course. 47 Isaac Levi, The Enterprise of Knowledge: An Essay on Knowledge, Credal Probability, and Chance (Cambridge: The MIT Press, 1980), pp. 2412.
45

44

436

BOUDEWIJN DE BRUIN

observed cases there was a capacity for non-counterable interference.48 (This line of reasoning is embodied in the two intermediate claims from the previous section.) Evidence underlying the claim that resource inequality entails capacity for arbitrary interference is thus a set of observations of resource inequality of which a sufciently large number show actual non-countered interference. To start the argument against zero probability subordination, assume that in 75% of the cases I observed actual non-countered interference. Then I am justied in holding two claims. First, there is a claim about resource inequality and republican unfreedom:
(1) There is resource inequality whenever there is a capacity for arbitrary interference.

This is a claim about a capacity. Second, there is a claim about the likelihood of actual interference:
(2) If there is resource inequality, then there is a 75% chance of actual non-countered interference.

This reveals that while the knowledge embodied in CKR has to be grounded in observations supporting an inductive generalization about resource inequality and actual non-countered interference, these observations are incompatible with assigning zero probability to actual interference in cases of resource inequality. If I have seen non-countered interference in 75% of the cases of resource inequality, I can conclude that there is a capacity for arbitrary interference, but I cannot simultaneously hold the view that there is zero probability of interference in this particular case. The evidence underlying knowledge about domination is inconsistent with assigning zero probability to actual interference. It may be objected that in some cases we do have additional evidence that there is zero probability of interference, but an investigation of how to deal with such a case of conicting evidence reveals that this does not refute my argument. Suppose I have adopted the belief that there is resource inequality between Sue and Tania, that I have made an inductive generalization about resource inequality and non-countered interference, but that I also have some evidence to the effect that there is zero probability of Tania interfering with Sue. What should I do? There are (at least) three ways to stay consistent. First, I can give up the inductive generalization. But for this to be rational the additional evidence underlying the zero probability claim has to be genuinely stronger than the evidence underlying the inductive generalization, and since the inductive generalization is a central idea of republican thought, this move is unattractive in the context of the present
48 That not all cases of resource inequality are cases of interference is not due to the fact that there is no strict correlation (such as when the administration of certain medication goes together in many, but not in all cases, with curing a certain disease). Rather it is due to the fact that the capacity for arbitrary interference is not in all cases exercised (such as when the administration of testosterone to zebra nches goes together with a capacity to sing low frequencies, even though nches with increased testosterone levels will not actually sing in all cases).

LIBERAL AND REPUBLICAN FREEDOM

437

argument. Conversely, I can give up my belief in the additional evidence precisely on the grounds that the inductive generalization about resource inequality and republican unfreedom is backed by a large number of observations, but here liberals may object that such belief revision would be biased toward republicanism. The simplest and least damaging thing to do, then, is just to give up the belief that the present situation is one of resource inequality; statement (1), that is. In fact, this way to revise beliefs is arguably the most rational one. The evidence underlying the inductive generalization involves a large number of observations, and if the additional evidence about zero probability is any good, it also has to be based on a signicant number of observations. One just cannot reasonably form beliefs about probability (zero and non-zero probability alike) if the sample space is small. By contrast, the observation that the present case is one of resource inequality is a one-shot judgment, not backed by a large reservoir of evidence. As a result, if there is an indication that the probability of actual interference is zero in a certain case, then a rational epistemic agent holding on to the inductive generalization gives up on the one-shot judgment. She sticks, that is, to the inductive generalization; she sticks to the additional evidence that there is zero probability of actual interference, but she discards the belief that there is resource inequalityand this is how she remains consistent. A second objection could be raised, however, to the effect that the argument only involves subjective probabilitiesreecting beliefsand that nothing is said about whether in reality there may be any zero probability cases of subordination. But the argument applies equally well to objective probabilities, because it can be shown that in situations where CKR obtains there cannot be objective zero probability of interference. The idea is simply that if the data on which the beliefs are formed are primary (not involving hallucinations, and so on), observing that there is non-countered interference in 75% of the cases of resource inequality just suggests that the objective probability of non-countered interference in situations of resource inequality is 75%. In other words, the subjective probability estimates are just equal to the objective probabilities. True, the objective probability could only be ascertained with full condence if we had access to (nitely or even innitely many) instances of resource inequality; and if from now on resource inequality were no longer accompanied by non-countered interference, the objective probability would turn from 75% to zero.49 But this would lead us to sever the connection between resource inequality and domination that, as we saw, republicans are unwilling to give up, so the present argument can safely ignore such possibility. Furthermore, additional evidence could certainly support zero probability, but as in the case of subjective probabilities, this would indicate that there is, in fact, no resource inequality.
49 This is true for frequentist interpretations of probability, but not necessarily for propensity and Bayesian interpretations. See Donald Gillies, Philosophical Theories of Probability (London: Routledge, 2000).

438

BOUDEWIJN DE BRUIN

Granting the republicans what can be granted, no objective zero probability of interference is possible in a situation in which CKR is satised.50
Primary knowledge about republican unfreedom is inconsistent with assigning subjective zero probability to actual interference, and cannot be satised in situations where there is objective zero probability of interference.

All in all, if there is non-zero probability of subordination, both liberals and republicans have reasons to increase freedom. If there is zero probability, then CKR does not hold. As I have argued, however, without CKR the third supposed benet of republican freedom does not arise. This leads to the conclusion that the benet of minimization of subordination is as much a benet of liberal as of republican freedom. VII. CONCLUSION In this article I started by looking at three relations between liberal and republican freedom. I argued against two of them (Logical Equivalence and Degree Supervenience), and in favor of one of them (Epistemological Priority): to know something about arrangements of republican freedom presupposes that you know something about arrangements of liberal freedom. Turning to a critique of the republican position, I considered the three benets that adherents of republican freedom claim their concept possesses over and above those of liberal freedom: that it minimizes the need for strategic deference; that it minimizes uncertainty; and that it minimizes subordination. I showed that republican freedom only possesses the latter two benets if there is common knowledge of arrangements of republican freedom, and using the reasoning from the argument about Epistemological Priority I showed that these benets are not uniquely republican. In fact, their necessary and sufcient conditions can be phrased in purely liberal terms. An independent argument revealed that the rst benet can be understood in liberal terms. While I did not examine the relation of Normative Priority between liberal and republican freedom, I believe that the work presented in this article may be of help here. Ian Carter has suggested that liberal freedom is normatively prior to republican freedom in that liberal freedom gives normative reasons for favoring republican freedom, but not vice versa.51 Liberals, the argument goes, can defend republican institutions simply to the extent that they favor liberal freedom, and Carter suggests that this extent could be rather broad. While the plausibility of
50 This does not rule out that there are situations in which there is zero probability of exercising a capacity for arbitrary interference. In fact, such cases gure in my attacks on Logical Equivalence and Degree Supervenience. What the present argument shows is that these cases cannot at the same time satisfy CKR based on primary knowledge. 51 Carter, How are power and unfreedom related? p. 81. Cf. Lena Halldenius, Liberty and its circumstances: a functional approach, New Waves in Political Philosophy, ed. B. de Bruin and C. Zurn (Basingstoke: Palgrave Macmillan, 2009), pp. 1939.

LIBERAL AND REPUBLICAN FREEDOM

439

this suggestion depends on empirical claims about how good actual republican policies are at increasing liberal freedom, the argument presented in this article to the effect that the benets of republican freedom can be understood in liberal terms underscores the prima facie plausibility of Carters view.

You might also like