4

D. Lewis, "Reductionof Mind"

I am a realist and a reductive materialist about mind. I hold that mental states
are contingently identical to physical - in particular, neural states. My posi-
tion is very like the 'Australian materialism' of place, smart, and especially
Armstrong. Like Smart and Armstrong, I am an ex-Rylean, and L.ruir, ,o-.
part of the Rylean legacy.In view of how the term is contested.I do not know
whether I am a'functionalist'.

Supervenience and analysis

My reductionism about mind begins as part of an a priori reductionism about
everything. This world, or any possible world, consistsof things which instan-
tiate fundamental properties and which, in pairs or triples o;..., instantiate
fundamental relations. Few properties are fundamental: the property of being
a club or a tub or a pub, for instance, is an unnatural gerrymandeq a condi-
tion satisfied by miscellaneousthings in miscellaneousways. A fundamental,
or 'perfectly natural', property is the extreme opposite. Its instances share
exactly some aspect of their intrinsic nature. Likewise for relations.l I hold,
as an a priori pinciple, that every contingent truth must be made true,
somehow, by the pattern of coinstantiation of fundamental properties and rela-
tions. The whole truth about the world, including the mental part of the world,
superveneson this pattern. If two possible worlds were exactly isomorphic in
their patterns of coinstantiation of fundamental properties and relatiorrr. th.v
would thereby be exactly allke simpliciter.2
It is a task of physics to provide an inventory of all the fundamental prop-
erties and relations that occur in the world. (That's becauseit is also a task of
physics to discover the fundamental laws of nature, and only the fundamental
properties and relations may appear in the fundamental laws.3) 'we have no a
priori guarantee of it, but we may reasonably think that present-day physics
akeady goes a long way toward a complete and correct inventory. Remember
that the physical nature of ordinary matter under mild conditions is very well
understood.a And we may reasonably hope that future physics can finish the
job in the same distinctive style. we may think, for insiance, that
mass and

51
 D . L EW IS

charge are among the fundamental properties; and that whatever fundamental d
properties remain as yet undiscovered are likewise instantiated by very small h
'We o
things that come in very large classesof exact duplicates. may further think
that the very same fundamental properties and relations, governed by the very @
@

same laws, occur in the living and the dead parts of the world, and in the m
sentient and the insentient parts, and in the clever and the stupid parts. In
short: if we optimistically extrapolate the triumph of physics hitherto, we may ff
provisionally accept that all fundamental properties and relations that actu- ill

ally occur are physical. This is the thesis of materialism. u

(It was so named when the best physics of the day was the physics of matter d
alone. Now our best physics acknowledgesother bearersof fundamental prop- f,

erties: parts of pervasivefields, parts of causally active spacetime.But it would m

be pedantry to change the name on that account, and diso'"r'nour intellectual I
ancestors.Or worse, it would be a tacky marketing ploy, akin to British Rail's s
decree that second class passengers shall now be called 'standard class
c us t om er s ' . ) il
If materialism is true, as I believe it is, then the a priori supervenience ft
of everything upon the pattern of coinstantiation of fundamental properties d
and relations yields an a posterlori supervenience of everything upon the m
pattern of coinstantiation of fundamental physical properties and relations. m
Materialist supervenienceshould be a contingent matter. To make it so, we t
supply a restriction that makes referenceto actuality. Thus: if two worlds were
physically isomorphic; and if no fundamental properties or relations alien to n
actuality occurred in either world, then these worlds would be exactly alike 5
simpliciter. Disregarding alien worlds, the whole truth supervenesupon the
physical truth. In particular, the whole mental truth supervenes.So here we 5
have the common core of all materialist theories of the mind.s d
A materialist who stops here has already said enough to come under formi- rT
i:,*.11"'i dable attack. An especially well-focused version of the attack comes from t
lJ.';-. rank Jackson.6Mary, confined in a room where all she can see is black or n
.,n
,,", white, studies the physics of colour and colour vision and colour experience
\,!.,1g ".

i-!:*,.t': (and any other physics you might think relevant) until she knows it all. Then
she herself seescolour for the first time, and at last she knows what it's like {
to seecolour. What is this knowledge that Mary has gained? It may seemthat li
she has eliminated some possibilities left open by all her previous knowledge; N
she has distinguished the actual world from other possible worlds that are t
exactly like it in all relevant physical respects.But if materialist supervenience
is true, this cannot be what has happened. I
Materialists have said many things about what does happen in such a case.
I myself, following Nemirow, call it a case of know-how: Mary gains new (
imaginative abilities.T Others have said that Mary gains new relations of !

acquaintance, or new means of mental representation; or that the change in I

her is just that she has now seen colour. These suggestionsneed not be taken il

as rival alternatives. And much ink has been spent on the question whether I

52
 " R ED U C T IO N OF MIN D "

these various happenings could in any sense be called the gaining of .new
knowledge', 'new belief, or 'new information'. But for a materialist. the heart
of the matter is not what does happen but what doesn't: Mary does not distin-
guish the actual world from o;her worlds that are its physical duplicates but
not its duplicatessimpliciter.#
Imagine a grid of a million tiny spots - pixels - each of which can be made
light or dark. when some are light and some are dark, they form a picrure,
replete with interesting intrinsic gestalt properties. The case evokes reduc-
tionist comments. Yes, the picture really does exist. yes, it really does have
those gestalt properties. However, the picture and the properties reduce to the
arrangement of light and dark pixels. They are nothing over and above the
pixels. They make nothing true that is not made true already by the pixels.
They could go unmentioned in an invenrory of what there is withogt fhereby
rendering that inventory incomplete. And so on. |""it&i' ct t,* r*, "
,l',><+/,
suchiommentsseemto obviouslyright. The picffie 'Wt{
pixels. And that is becausethe -. picture superveneson the pixels: there could
be no difference in the picture and its properties without some difference"nts**
in
the arrangement of light and dark pixels. Furrher, the supervenienceis asym-
metric: not just any difference in the pixels would matter to the gestalt
properties of the picture. And it is supervenienceof the large upon the small
and many. In such a case, say I, supervenienceis reduction. And the materi-
alist supervenienceof mind and all else upon the arrangement of atoms in
the void - or whatever replacesatoms in the void in true physics - is another
s uc h c as e.
Yet thousands say that what's good about stating materialism in terms of
supervenienceis that this avoids reductionism! There's no hope of settling this
disagreement by appeal to some uncontested definition of the term .reduc-
tionism'. Becausethe term ls contested, and the aim of some contestants is to
see to it that whatever position they may hold, 'reductionism' shall be the
name for something else.
At any rate, materialist superveniencemeans that for anything mental, there
are physical conditions that would be sufficient for its presence,and physical
conditions that would be sufficient for its absence. (These conditions will
include conditions saying that certain inventories are complete: an electron
has only so-and-so quantum numbers, for instance, and it responds only to
such-and-suchforces. But it's fair to call such a condition 'physical', since it
answers a kind of question that physics does indeed address.)And no matter
how the world may be, provided it is free of fundamental properties or
relations alien to actuality, a condition of the one sort or the other will
obtain. For all we know so far, the conditions associatedwith a given mental
item might be complicated and miscellaneous- even infinitely complicated and
miscellaneous.But so long as we limit ourselves just to the question of how
this mental item can find a place in the world of fundamental physics, it is
irrelevant how complicated and miscellaneousthe conditions might be.

53
 D . LE S TIS
It may *r*
that physical conditions should
T,ri-'rsfactory
whether the mentar always settre
item is present o. #r.rrr. For
a vague question with no determinate mightn't that sometimes be
f.o;u,i;;;J ifitdid;i;; #',il:;i#l T:::il:rllffi
ar

',,r,",
mentalphysicso.:fl* supervenienc.,lt*oold r*.*ir.
,'j,,:;ff
irreduciblero fundamentalphysics rrro* *at boiling was
- i, ubrurd. ro, i, i, . u"go" ma*er
just wheresimmering

1^s two
ror
1."u., #
A longerreply has threeparts.
anrhing dgesr.we still have
".ra
-rri.r, begins.
Aoit*g
rrtiiirt. prrylicalsettlesthe mentarinsofar
*ut..i"trri
phvsicallvisomorphicworlds
supervenience.
to be just
part of what ir
means
t
il
mental indererminacyin one it is that any
.";;;iy matchedby "fiil;;;;y
the other.(2) Sfheneverir.i: in
uue.r.qir.rrio' whetlre, -.;-;;eterminacy
,o_. ,i_ptirtic menral
classificationapplies,. 1
it will be a".,.r,oirr"r. that some
tion applies.what's determinate more subtlecrassifica_
don't' but rather that you're in -;;;;;., that you d;il ,',* or that you
a certainequivocalstateof mind
easydescription.(3) If alr indeterm*".y'r, that defies
then thereis no indeterminacy a matter or r.-"*i. indecision,s
in tt thi.rg, th.*r.ru.r. ul*.""ra
up some irreducible mental conjure
item "just iy failing to d..idl".xa*ly -. which
reducibleitem we're referring
to?
rt may seemthat *h"n ,ip.rvenience
guaranteesthat there are physicar
forthep'.'.n."o,lb',.,,..,f
::ftlffi;':?:T:, ";;;;;';;;tal item,
the
youmigh,;;;:::i;HJlliTl*.iiT":;T,.,:il:'?Hr;z:;:
new knowledgewhen she.first
r"*;;r;;"ttt ough;;;rh. iJ";ned foilowed
necessarilyfrom ail the physicssheknew'beforeh"and,
rant becauseit didn't follow priori.- ' ' il;ained igno_
a "h.
A short repry to this objection
fro- necessitya posterioriis that
show that the mental if it did
irr.a.r.irt.lo'rurra"-ental physics,
it wourd like_
-"r*u, i,,.au.;i;;, fo,d"_..,t1"i _ which
or,rJ,.,
;;ilf'il,,H: ,oo:lt*, is
r""a"-*i"r;r,r."rfl *:6.1*#r"^;:;,;:;;X*mi:,d;-
(A longerreplS followi"g
Ju.r.*",,', iou.ra.a upon the .twJ-dimensional,
posteriori'e
Two-dimensionalirm
il$tili :'r':':t,' ,"; ;;; thereis no
sentencedmay*",,ill1il',,rri^if
I

tions,oneof themnecessary
{,ito,.liT"lil,.i,},T:n::;,T.# I
and the otherone
onecan beknown onr.y posteriori.suff"r. contingent;and the contingent
of meaningunderwhich .a ,". .hooi. ,.'"Jrpr'" conception a

;rr.;^;";;;ffi hrsu"s; ,;;#:;i*

only as a function of matrcrs _."ning,
of contingent fact -for example, fl
on which the meaningof ,boils, a conception
r11.irl".,n.1qent on ,f,l.f, n
enonturns outto o..ylv the,bo'ins-*r.. ,ifrn.al phenom_
usingthemeanings rrt., ri *;l;#;Jr; senrenceg tl
of its wordsasfiled in world Wr, d
usingthemeanings ,. g.i prolositionHr;
asfixed wr, _. s. i'rtr; I
srtronshorizontallyexpressed ciiiri.lf th. p.opo_
- ly 6 ut"rt.'u".io*
^ndr. "". h
worlds;and let H be the
fit
54
 "REDUCTIONOF MIND"
rmle propositionhorizontallyexpressed by s at the actualworld. The proposition
s be d-iagonallyexpressed by g_isthe propositionD that holds at any world
rion wiff
the propositionhorizontallyexpressed by 0 at w is true at w. So if we know
sda- D, y. know that $ horizontallyexpresses sometruth or other, but we rnay
s'as not know which truth. sentence ii necessary
Ntter s d posterioriiff H is necessary
but D is knowable onry a posteriori.Likewise,a propositio p
n necessariry
implies that g iff p implies H; but p a priori impries
pfar th"t o iff t, implies D.
our worry was that when $ was aboutih. and p was p."-ir"
eans true by fundamentalphysics,p might imply that "
-ittd, .r...rr".iif ro',
an._v
I 4 -"d.
,ro, a priori.
But if so, and if you think it matters,just-takeanother
proporioo' e: let e
ti n b,etrue at exactlythoseworlds where horizontally.*pr.rr.,
ntal $ th. ,"-. propo-
sition H that it actuallydoes.e is true. Given the materialist
6ca- supervenience
of everything,Q as well as p is madetrue by-fondam.ntai;;yri.r.
J'ou
t, and e
together imply a priori th.at s. so the gap between prryrii"i
Efies f,remis.s and
mentalconclusionis closed.Anyonewho wants to reopen
on.8 it - for instance,in
order to squarematerialistsupervenience with M"ry,J ,"pf*.J ignorance-
i*. must somehowshow that the two-dimensionalanaiysis
of'n.c.ssity a poste-
hich riori is inadequate.)

dcal If we limit ourselvesto the questionhow mind finds a place

the in the world of
physics,our work is done.Materialistsupervenience
bri. offersa full answer.But
if we expand our interestsa little, *.'ir ,." that among
th. ,op.ru.rri.rrt
Prn featuresof the world, mind must be very exceptionar.
red Thereare countlesssuch
features'In our little toy-exampleof the picture and the
pixels,the superve-
Fo- nient propertiesnumber2 to thepower: i to themilionti;-;;
In rhe case
of materialist supervenience,the number will be f^, grrutrr.
drd ih. irrfirri,.
cardinal beth-3is a conservativeestimate.The u"rt
ike- ,.rp.rrr..rr.rr,
featuresof the world are givenonry by miscelraneourry -"yJritf-of
his infirrit. JiJiunctionsof
infinitely complexphysicalconditi,ons.Therefore,rr.y
lin u.ynrrl o.r, power
to detect,to name,or to think about one at a time. "r.
Mental featuresof the
world' however,are not at all beyondour ken. Finite assemblies
oal' of particles
- us - can track them.Thereforetheremust be somesort of
no simplicityto them.
Mavbe it will be a subtle
qie 9f simplicity,visibleonly if y." ;;k in just the
_rort_
right way. (Think of the Mandelbrot set:its overwhel-irg .o-pr.*ir)! its short
mi- and simplerecipe.)But somehowit must be there.neueitirrftiris
simpricityis
Fnr a job for conceptualanalysis.
ion Arbiters of fashionprocraim_ that analysisis out of date. yet without it, I
ngs seeno possibleway ro establishthat any featurebf the
ion world doesor does
not deservea namedrawn from our tradiiional menral
xn- should
repudiatenot analysisitself, but only somesimplisticgoals "r.";;;;we
ed for ii. w. should
allow for semanticindecision:any interestinganalysa=ndum
Hi is likely to turn
uugyr and ambiguous.often the bestthat any one analysis
po- 9yl can do is ro
fall safelywithin the rangeof indecision.And we should
the alltw for semantic
satisficing:analysismay revealwhat it would take to deserve
,r"-. perfectly,
"
55
 D . L E IT I S

but imperfect deservers of the name may yet deserve it well enough. (And *Mo
sometimes the perfect case may be impossible.) If so, there is bound to be pmT
semantic indecision about how well is well enough. :fm
I offer not analyses, but a recipe for analyses. We have a very extensive :rotrl
shared understanding of how we work mentally. Think of it as a theory: tri,&w
folk psychology. It is common knowledge among us; but it is tacit, as our -fri$fi[
grammatical knowledge is. I7e can tell which particular predictions and expla- rdam
nations conform to its principles, but we cannot expound those principles I
systematically.loFolk psychology is a powerful instrument of prediction. We .-rrnm|

are capable of all sorts of behaviour that would seem bizarre and unintelli- 5.er
gible, and this is exactly the behaviour that folk psychology predicts, rightly, &roil
will seldom occur. (But we take a special interest in questions that lie beyond 'mfll

the predictive power of folk psychology; wherefore ingrates may fairly |rGS
complain of a lack of interesting predictions!) Folk psychology has evolved :rct
through thousands of years of close observation of one another. It is not the trdal
last word in psychology, but we should be confident that so far as it goes - -W
and it does go far - it is largely right. rrur
Folk psychology concerns the causal relations of mental states, perceptual n
stimuli, and behavioural responses.It says how mental states, singly or in m 8'
combination ) are apt for causing behaviour; and it says how mental statesare rrui
apt to change under the impact of perceptual stimuli and other mental states. Ffril
Thus it associateswith each mental state a typical causal role. Now we have n]iid
our recipe for analyses.Supposewe've managed to elicit all the tacitly known rsitr
general principles of folk psychology. l7henever M is a folk-psychological s
name for a mental state, folk psychology will say that state M typically occu- ::nrl
pies a certain causal role: call this the M-role. Then we analyse M as meaning .m.I
'the state that typically occupiesthe M-role'. Folk psychology implicitly defines et'fr
the term M, and we have only to make that definition explicit. rEl
Sincethe causal roles of mental statesinvolve other mental states,we might :dfl!
fear circularity. 'The remedy is due in its essentialsto Ramsey.11Suppose,for i[ sl
instance, that folk psychology had only three names for mental states: L, M, chu
'We
N. associatewith this triplet of names a complex causal role for a triplet :rE I
of states, including causal relations within the triplet: call this the LMN-role. :cd
Folk psychology says that the states L, M, N jointly occupy the LMN-role. !: q
That implies that M occupiesthe derivative role: coming second in a triplet of 3{rl!
statesthat jointly occupy the LMN-role. Taking this as our M-role, we proceed il
as before. Say that the names L, M, N are interdefined, The defining of all r[}rul
three via the LMN-role is a package deal. 3fi'Af
\We might fear circularity for another reason. The causal roles of mental I
states involve responsesto perceptual stimuli. But the relevant feature of the ]mil
'We
stimulus will often be some secondary quality - for instance, a colour. :O T
cannot replace the secondary quality with a specification of the stimulus in ,lrlq
purely physical terms, on pain of going beyond what is known to folk :i€tl
psychology. But if we analyse the secondary quality in terms of the distinctive ,-ilrrcr
mental statesits presenceis apt to evoke, we close a definitional circle. So we :gfo

56
 "REDUCTIONOF MIND'
should take interdefinitionfurther.Let
fork psychologyincludefolk psycho-
physics.This will say,for instance, 'pio
tr,uiirr. of J rr.r"-1"r"", and the
correspondingsensationiointry-occupy
a comprexcausar.rl. *", consistsin
part, but only in part, o{ the'formei
being apt to causethe latter. Now
have a derivativerole associa*i ;irh we
;. name of rhe colour, and anorher
associated with the nameof the sensation:the role
second,respectivelS "il;;;;;?st or coming
in a pair that jointly occupiesthi, .ffi." .o1..
'we might worry arso about the behaviourthat mental
causing.often we describebehaviour states are apt for
in a mentallyl."dJ;;;;as action.To
say that you kicked.the balr to your teammat€
is to describeyour tylo, behaviour.
But such a descriptionpr€supposes
;;i;;;;;
was meantto serveyour desiresaccording " ;r;"; u.t
"',,io.,,
to your beriefs; about the
presenceof the ball and the pl"y;;;r;'.Ae
and the other"rri "tro
player,and about
the socialfactsthat ,-,niteplayersir,[ t."-r.
More threat of iirlur"lry? More
needfor interdefinition?I-don't know
ho* suchfurther interdefinrtionwould
work; and anvwav,it would be well
to calr a halt b.f;;;-i;il psychology
expandsinto a folk theory of the entireLebensweh!
Describingthe behaviourin purely physical
*r-, - the angleof
the velocityof the foot - wouli g., tid of ,hor. presuppositions. the knee,
But, iusr as
in the case of the stimuri, it would go beyond
what is known to folk
psychology.Furrher,thesedescriptionswrould
neverfit tt. l.rr*lo,rr of space
aliensnor of humanoidshape; y.t we shourd,rot air-i., o.ri of hand the
speculationthat folk psychology ".rd
might apply to aliensu, *.ir ,o ou.r.trr.r.
Fortunatelythere is a third *"y1o a.oiiu. behaviour. ",
rrh;; y." kicked
the-ball,your body m.ovedin sucha way that if you
hadbeenon a flar surface
in Farth-normalgravity with_asuitablyplaced'bail
r" frorrt of yo., a suit_
ably placedteammatesom.edisranc.- then thei-p".i;;;;;, "rra
^*:^y, foot upon
the ball would havepropelledthe ball otrto a trajectorybringingit within the
teammare's reach.That descriptionis availabletothe folk. Tfi.f'*ourant
it spontaneousl,but they can recognizeit as correct. give
It presupposes nothing
about your menralsrates, eyenih"t yo., haveany;
r19r
the.ball and the playing fierd whether
and.the gruiiry and the teammatJ ".aG;i;ut
arereaty there;
nothing about your humanoid.h"p.l exceptthat you
h"rr. ,orrlrort of foot.
It could just as well describethe behaviourof a mindl.r,
contrap_
tion_'in the shapeof a spacearien(with a foot), thrashing -..iru.ri.ali" free falr.
(I don't saythat we shourdreally usethese'ii - "b;;;
. ttr.rr, a."r.rifrio", or behav_
iour. Ratheq my_point is that their availability
shows how to unload the
presuppositions from our ordinary descriptions.)
If M means'the statethat_typiiallyoc.upie, the M-rore,
and if that role is
only imperfectly-occupied, whal are we to iol - satisfice:r.tit. rr"-. ll go
to a state that deservesit imperfectly.And if nothing
.";;; ,r.",
occupyingthe M-role? - Then the name M has t
o i.r.r*i. "rr1.*ir.r.
ili. bo,rrrd"ry
betweenthe casesis vague.To take an examplefrom
a differentterm-rntro-
ducingtheorS I supposeit to be indeterminatewherher'dephlogisticared
refersro oxygenor to nothing.But folk psychol"g/it-ir air,
r".'r.,r3r"rrrapethan

57
 D . L EW IS
phlogiston theorS despite scare stories to
the contrary. we can happily grant
that there are no perfect deservers of folk-psycttoiogi."i q{

shouldn't doubt that there are states that d.seive "rru-., ,rumes, but we r
thor. weil enough.
what to do if the M-role, or the LMN-role, turns
out to be doubly occu-
pied? I used to think that in this case too
the name M hadno referent.l2 Bur
now I think it might be better, sometimesor always,
to ,uy th"t th. n"-e t,rrrrs
out to be ambiguous in reference. That follows
ih" t."a .i ei.ra, and it is
consistent with, though not required by, the
treatment of CarnapJ3 Note that
we face rhe same choice with phrasesriiie .the moon
of M;;';-""d in rhat case dr
too I'd now lean toward ambiguiry of reference
rather than lack of it. @t
My recipe for analyses, like Rylean anarytic behaviourism,
for*, anarytic ,d
truths that consfrain the causal relations oi mental
rt"t., to b'.huuiour. iw.
have no necessary connections between distinct existences,
of course; the thd
necessityis verbal. The state itself could have failed
,o o..upy its causar role, iMt
but would thereby have failed to deserveits mental name.) g.rt
th. constraints ufu
are weak enough to be credible. Becauserhe state that
typically occupiesa role h0
need not occupy it invariablg and also because
a state may deserve a name #r
well enough in virtue of a role that it occupies imperfectry,
we are safe from llm
the behaviourist'sbugbears.\7e have a pra.e for the.esolote
deceiver,disposed
come what may to behave as if his mentar states were
other than they realry W
w. have a place for the total and incurable paralytic
r..
life and no behavioural dispositions whatever. we
with a rich mental w
even have a prace for a ffi
madman whose mental statesare causally related to behaviour iliil
and stimuli and
one another in a totally haywire fashion.la And yet not JIE
anything goes.At some
point - and just where that point comes is a mater -!r.r!
of ,.-unii. indecision -
weird tales of mental statesthat habitually offend against the principles rod
of folk
psychology stop making senselbecauseat some point the I
off*di.rg stateslose
all claim to their folk-psychological names. To that exrent, sn
arraiyti. behav-
iourism was right. To quote my crosest aly in these matters, .. flr@,
. . outward
physical behaviour and tendenciesto behavl do in some flmt
way enter into our
ordinary concept of mind. l7hatever theory of mind is true,
it h"s debt to I
pay, and a peace to be made, with behaviourism.,ls " t@
when we describe mental state M as the occupant of the wd[
M-rore, that is
what Smart calls a topic-neutral description.16It saysnothing rhd
about what sort
of state it is rhat occupies the role. It might be a non-phyrt"l o, nrm
a physical
state, and if it is physical it might be a stare of neorai aitivity in the brain, $rc
or a pattern of currents and charges on a silicon chip, or the jangling rml
oi
an enormous assemblageof beer cans. 'what state occupies the M-role and roh
thereby deservesthe name M is an a posteriori matter. Bui if materialist super- flm
Trm
venienceis true, and every feature of the world supervenesupon fundamental
physics,then the occupant of the role is some physical state 0ffi,
or other - because rtufr
there's nothing else for it to be. we know enough to rule out the chip
and croi
the cans' and to support the hypothesis that what occupies the role is
some t
pattern of neural activity. \7hen we know more, we shall know what pattern
flu[l

58
 " R ED U C T IO N OF MIN D "

of neural activity it is. Then we shall have the premises of an argument for
psychophysical identification:17

mental state M - the occupant of the M-role (by analysis),

physical state P = the occupant of the M-role (by science),
therefore M = P.

That's how conceptual analysis can reveal the simple formula - or anyway,
the much less than infinitely complicated formula - whereby, when we know
enough, we can pick out a mental feature of the world from all the countless
other features of the world that likewise superveneon fundamental physics.
The causal-role analyseswould still hold even if materialist supervenience
failed. They might even still yield psychophysical identifications. Even if we
lived in a spook-infested world, it might be physical states that occupied the
causal rules (in us, if not in the spooks) and thereby deservedthe folk-psycho-
logical names. Or it might be non-physical statesthat occupied the roles. Then,
if we knew enough parapsychology,we would have the premises of an argu-
ment for psycho-nonphysical identifi cation.

'!7hen
our argument delivers an identification M = P, the identity is contin-
gent. How so? - All identity is self-identity, and nothing could possibly have
failed to be self-identical. But that is not required. It's contingent, and it can
only be known a posteriori, which physical (or other) states occupy which
causal roles. So if M means 'the occupant of the M-role' it's contingent which
state is the referent of M; it's contingent whether some one state is the common
referent of M and P; so it's contingent whether M = P is true.
Kripke vigorously intuits that some names for mental states, in particular
'pain', are rigid designators: that is, it's not contingent what their referents
are.18I myself intuit no such thing, so the non-rigidity imputed by causal-role
analysestroubles me not at all.
Here is an argument that 'pain' is not a rigid designator. Think of some
occasion when you were in severe pain, unmistakable and unignorable. All
will agree,except for some philosophers and faith healers,that there is a state
that actually occupies the pain role (or near enough); that it is called 'pain';
and that you were in it on that occasion. For now, I assume nothing about
the nature of this state, or about how it deservesits name. Now consider an
unactualized situation in which it is some different state that occupiesthe pain
role in place of the actual occupant; and in which you were in that different
state; and which is otherwise as much like the actual situation as possible. Can
you distinguish the actual situation from this unactualized alternative? I say
not, or not without laborious investigation. But if 'pain' is a rigid designator,
then the alternative situation is one in which you were not in pain, so you
could distinguish the two very easily. So 'pain' is not a rigid designator.
Philosophical arguments are never incontrovertible - well hardly ever.Their
purpose is to help expound a position, not to coerce agreement. In this case,

59
 D . L E !rIS

the controverter might say that if the actual occupant of the pain role is not to r al
a physical state, but rather is a special sort of non-physical state, then indeed is. for
you can distinguish the two situations. He might join me in saying that this m ail
would not be so if the actual occupant of the role were a physical state - else Token
neurophysiology would be easier than it is - and take this together with intu- trom {
itions of rigidity to yield a reductio against materialism. Myself, I don't see of Se,
how the physical or non-physical nature of the actual occupant of the role has But
anything to do with whether the two situations can be distinguished. Talk of recasf,I
'phenomenal character' and the like doesn't help. Either it is loaded with ques- then yi
tion-begging philosophical doctrine, or else it just reiterates the undisputed a caufr
fact that pain is a kind of experience.le are rk
If there is variation across worlds with respect to which states occupy thartr
the folk-psychological roles and deservethe folk-psychological names (and if reladq
this variation doesn't always require differences in the laws of narure, as folL-p
presumably it doesn't) then also there can be variations within a single world. anead
For possibility obeys a principle of recombination: roughlS any possible a !-auff
kind of thing can coexist with any other.2oFor all we know, there may be rn rhir
variation even within this world. Maybe there are space aliens, and maybe Ita
there will soon be artificial intelligences,in whom the folk-psychological roles sub'qa
are occupied (or near enough) by states very different from any states of and br
a human nervous system. Presumably, at least some folk-psychological roles identl
are occupied in at least some animals, and maybe there is variation across tLring;
species.There might even be variation within humanity. It depends on the occur i
extent to which we are hard-wired, and on the extent of genetic variation in not PI
our wiring. Ns
\7e should beware, however, of finding spurious variation by overlooking an,rn t
common descriptions. Imagine two mechanical calculators that are just alike don'r r
in design. When they add columns of numbers, the amount carried goes into en] - S
a register, and the register used for this purpose is selected by throwing a h'e sall
switch. Don't say that the carry-seventeenrole is occupied in one machine by r ._ xl
a state of register A and in the other by a state of register B. Say instead that a r . Ys
in both machines alike the role is occupied by a state of the register selected dis"ml
by the switch. (Equivalently,by a state of a part of the calculator large enough is not
to include the switch and both registers.) If there is a kind of thinking that role h
some of us do in the left side of the brain and others do in the right side, that haring
might be a parallel case. cound
If M means 'the occupant of the M-role' and there is variation in what occu- c,l.ckq
pies the M-role, then our psychophysical identities need to be restricted: not in Bul
plain M = P, but M-in-K = P where K is a kind within which P occupies the more I
M-role. Human pain might be one thing, Martian pain might be something Sin
else.21As with contingency, which is variation across worlds, so likewise with \t-rol
variation in a single world: the variability in no way infects the identity rela- il rt
"M
tion, but rather concerns the referenceof the mental name. are ut
The threat of variation has led many to retreat from 'type-type' to 'token- er y i l
token' identity. They will not say that M = P, where M and P are names Tney I

60
 " R E D U C TION OF MIN D "

role is not for a state that can be common to different things at different times - that
fren indeed is, for a property had by things at times. But they will say rhat ffi = p, where
rg that this m and p are mental and physical names for a particular, unrepeatable event.
stare - else Token-token identities are aII very well, in their derivative b.rt the flight
r n- rh intu- from type-type identities was quite unnecessary.For our restr'icted
-"y, identities.
I don't see of the form M-in-K = P, are still type-type.
he role has But don't we at least have a choice? couldn't our causal role analysesbe
od. Talk of recast in terms of the causal roles of tokens, and if they were, would they not
rn-irhques- then yield token-token identities? After all, the only way for a rype to occupy
undisputed a causal role is through the causesand effects of its tokens. The eifects of pain
a_rethe effects of pain-events- I think, following Jackson, pargetter, and prior,
Bs oL-cupY that this recasting of the analyseswould not be easy.zzThere are more causal
nes and if relations than one. Besidescausing, there is preventing. It too may figure in
[arure. as folk-psychological causal roles; for instance, pain tends to prevent ,r.rJiuid.d
ryle rvorld. attention to anything else. Prevention cannot straightforwrrdly b. treated as
ry pc,ssible a causal relation of tokens, becausethe prevented tokens do not exrst - not
re mar- be in this world, anyway. It is better taken as a relation of types.
rnti naar-be rf a retreat had been needed, a better retreat would harr. be.r, to .subtype-
rylcal roles subtype'identity. Let MK name the conjunctive property of being ir,,trie u
r siares of and being of kind K; and likewise for PK. Do we .."lly *".rt psychophysical
grc-alroles identities of the form MK = PK? - close, but I think .rot quit. ,ight. For one
ioqi across thing, M-in-K is not the same thing as MK. The former bui not the latter can
ids on the occur also in something that isn't of kind K. For another thing, it is p itself,
rnaron in not PK, that occupies the M-role in things of kind K.
Non-rigidity means that M is different statesin different possiblecases:van-
rer"lxrking ation would mean that M was different states in different actual .ur.r. Bua
I iurr alike don't we think that there is one propefty of being in the state M - one prop-
lg.:es lnto erty that is common to all, actual or possible, of whatever kind, who cartruly
hrclr-r:lq a be said to be in state M? - There is. It is the property such rhat, for any possible
lnchrne L'r- X, X has it just in caseX is in the state that occupies the M-role for X,s kind
uread rhar x's world'23 The gerund 'being in M' can be taken, at least on one good
1t
fn *-:;ted disambiguation, as a rigid designator of this property. However, this properry
Eg .r:'ruqh is not the occupant of the M-role. It cannot occupy that or any other causal
di::: lhar role becauseit is excessivelydisjunctive, and therefore no eventsare essentially
: sadt. rhat havings of it.2aTo admit it as causally efficaciouswould lead to absurd double-
counting of causes.It would be like saying that the meat fried in Footscray
lfra: ,:rccu- cooked becauseit had the property of being either fried in Footscray or boiled
icted: not in Bundoora - only worse, becausethe disjunction would be much ionger and
ftrrFresthe more miscellaneous.
mrne-hlns Since the highly disjunctive properry of being in M does not occupy the
mr,se-nr:h M-role, I say it cannot be the referenr of M. Many disagree.They wouid like
ffilt-r ::-a- it if M turned out to be a rigid designator of a property .o--on to all who
are in M. So the property I call 'being in M', they call simply M; andthe prop_
*f
D : !\-1- erty that r call M, the occupant of the M-role, they call 'the realisation of M'.
f B:li= rs They have made the wrong choice, since it is absurd ro deny that M itself

61
 D. LE\rIS
-: 1infry
is causally efficacious. Still, their mistake is superficial. They have the right Prr@
properties in mind, even if they give them the wrong names.
-4 laruridI
It is unfortunate that this superficial question has sometimes been taken to
rlmg4 t
mark the boundary of 'functionalism'. Sometimesso and sometimesnot - and :rlwsm
that's why I have no idea whether I am a functionalist. ia : -I
Those who take 'pain' to be a rigid designator of the highly disjunctive : 5fr-
property will need to conrroveft my argument that 'pain' is not rigid, and they t6 - r-t
will not wish to claim that one can distinguish situations in which the pain- :35m
role is differently occupied. Instead, they should controvert the first step, and -; S'ge m
deny that the actual occupant of the pain-role is called 'pain'. I call that denial l;'6fl
a reductio. _ ,]fi
'3Effilffi
:emffi
Notes : Sfrirrm
1"8 SauNt
1 See David Lewis,'New Workfor a Theoryof Universals' , AustralasianJournalof 1l rlhe m
Philosophy6I (i983), pp.343-377; and DavidLewis,On the Pluralityof Worlds
a'rd t
( B lac k we l l1, 9 8 6 ),p p . 5 9 -6 9 .
aEnseI
2 See David Lewis, 'Critical Notice of D. M. Armstrong,A CombinationalTheory 2i lsffiEb
of Possibility',AustralasianJournalof Philosophy7O (1992), pp. 2II-224. 2; .-gtrE,
3 Lewis,'NewWorkfor a Theoryof Universals', pp. 365-370. =laldl
22
4 GeraldFeinberg,'Physicsand the ThalesProblem',Journalof Philosophy66
lFe{
( 1966) ,pp . 5 -1 3 .
2r - 'frd
5 Lewis,'NewWorkfor a Theoryof Universals', pp. 361-365. 24 :aurd l
6 FrankJackson,'Epiphenomenal Quarterly32(1982), pp.
Qualia',Philosophical tusq'
rzt*r50.
7 LaurenceNemirow,'Physicalism and the CognitiveRoleof Acquaintance' and
DavidLewis,'What ExperienceTeaches',both in Mind and Cognition:A Reader,
ed. by W. G. Lycan(Blackwell,1990).
8 Lewis,On the Pluralityof Worlds,pp.2I2-2I3.
9 FrankJackson,'Armchair Metaphysics', in Philosophy in Mind,ed. by J. O'Leary
Hawthorne and M. Michael(Kluwer,1994); RobertStalnaker,'Assertion', Syntax
and Semantics9 (1978), pp 315-332; M. K. Daviesand L L. Humberstone,
'TwoNotionsof Necessity',PhilosophicalStudr'es 38 (1980), pp. 1-30; and Pavel
Tichli, 'Kripke on NecessityA Posteriori',PhilosophicalStudies 43 (1983),
pp. 225- 2 4 I.
10 PaceDavidLewis,'Psychophysical and Theoreticalldentifications', Australasian
Journalof Philosophy50 t.972), pp. 249-258, eliciting the generalprinciples
of folk psychology is no merematterof gatheringplatitudes.
11 F. P. Ramsey,'Theories'in Ramsey,TheFoundations of Mathemafics. (Routledge
& K eganP a u l , 1 9 3 1 ), p p . 2 1 2 -2 3 6 ; R u d ol fC arnap,' R epl i es
and xposi ti ons'
E
in The Philosophyof Rudolf Carnap,ed. by P. A. Schilpp (CambridgeUniversity
Press,1963), pp. 958-966. See also DavidLewis,'How to DefineTheoretical
Terms',Journalof Philosophy67 (197O), pp. 427-446, reprintedin Lewis,Philo-
sophicalPapers,Vol. 1 (OxfordUniversityPress,1983); and Lewis,'Psychophys-
ical and Theoreticalldentifications'.
12 Dav id Le w i s ,' H o w to D e fi n eT h e o re ti calTerms' and' P sychophysi cal and
Theoreticalldentifications'.

62
 " R E D U C TION OF MIN D "

ight l3 HartryField, 'Theorychangeand the Indeterminacy of Reference', Journalof

PhilosophyTO (1973), pp. 462-48I; Carnaptoc. cit.
nto 14 DavidLewis,'Mad Painand MartianPain',in Readingsin philosophyof psych-
and
ology,Vol. 1, ed. by N. Block (HarvardUniversitypress,19g0), reprintedwith
postscriptin Lewis,PhilosophicalPapers,Vol. 1 (OxfordUniversitypress,l9g3).
15 D. M. Armstrong, A Materialistrheoryof Mlnd (Rouiledge & Keganpaul, 196g),
tive
o. 68.
fi.1- 16 J. J. c. Smart, 'Sensationsand Brain processes',philosophicalReview,6g
ein- (1 9 5 9 ),p p . 1 4 1 -1 5 6 .
and 17 S e e D a v i d L e w i s ,' A n Arg u mentforthel denti tyTheory' , of phi l osophy63
Journal
nial (1966), pp.17-25, reprintedwith additionsin Lewis,phitosophical papers,vol.
I (oxfordUniversityPress,1983); and Lewis,'psychophysical and rheoretical
ldentifications'.See Armstrong,A Materialistrheory of the Mind, for an inde-
pendentand simultaneous presentation of the sameposition,with a much fuller
discussion of whatthe definitivecausalrolesmight be.
daf 18 Saul Kripke,Namingand Necessity (Blackwell,l9S0), pp. I47*I49.
ttds 19 Thecontroverter just imaginedwouldagreewith the discussion in Kripke,Naming
and Necessity, pp. 144-155. But I don't mean to suggestthat Kripkewould
wy agreewith him. At any rate,the wordsI haveput in his mouthare not Krioke's.
20 Lewis,On the Pluralityof Worlds,pp. 86-92.
2I L e w i s ,' Ma d Pa i na n d Ma rti a nP ai n' .
'66
zz F ra n k J a c k s o n ,R o b e rt Pa rgetter,and E l i zabethpri or,' Functi onal i smano
Type-typeldentity Theories',PhilosophicatStudies42 (1982), pp. 209-225.
23 In ' H o wto D e fi n eT h e o re ti c arerms'
l I cal l edi t the' di agonal i zed
sense' ofM.
pp. 24 David Lewis,'Events' in Lewis, Philosophicalpapers,Vol. 2 (oxford Universitv
Pre s s ,1 9 8 6 ).
anc
&r.

!ary
fta/.
ne.
re,
B).

la,r.r
ies

@
n5'
$b
ta
rilr-
D'5-

63