Supervenience, Necessary Coextension, and Reducibility

Author(s): John Bacon

Source: Philosophical Studies: An International Journal for Philosophy in the Analytic
Tradition , Mar., 1986, Vol. 49, No. 2 (Mar., 1986), pp. 163-176
Published by: Springer

Stable URL: https://www.jstor.org/stable/4319818

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Springer is collaborating with JSTOR to digitize, preserve and extend access to Philosophical
Studies: An International Journal for Philosophy in the Analytic Tradition

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 JOHN BACON

SUPERVENIENCE, NECESSARY COEXTENSION, AND

REDUCIBILITY

(Received 7 January, 1985)

In Keith Campbell's clear definition,

One feature is supervenient upon others if, while not following deductively from those
others, it nevertheless cannot vary unless they do. (1981, p. 15)

Supervenience has been popularized by Hare in ethics (1952, pp. 131ff)l and
by Davidson in his treatment of the mind-body problem (1970, p. 88). In
the words of Kim, who has led the way in clarifying it, supervenience is to be
a "determinative relationship between properties without requiring correla-
tions between them" (1978, p. 154). It is supposed to provide a happy
medium between dualism and reductionism, for example Yet, as Kim has
pointed out, a variant of supervenience entails necessary coextension: each
supervenient property is necessarily coextensive with some base property
(1983, p. 49 (4); 1985, pp. 23f). To be sure, necessarily coextensive proper-
ties need not be identical or interdeducible. But in the context of would-be
physical theory, nomically necessary coextension does establish a presump-
tion of reducibility. Thus if the 'cannot' in Campbell's definition expresses
nomic impossibility, supervenience threatens to collapse into reducibility.
My purpose here is to work out the logical relationships among several
closely related variants of supervenience. Among other things, I shall prove
a stronger form of Kim's result of necessary coextension. This narrows still
further the options for supervenience theorists.

I. CONCEPTS OF SUPERVENIENCE

I.O. Logical Tools

In order to establish the logical relations among different concepts of super-

venience, some formal tools will be handy. Besides intuitive set theory, I
shall make some informal use of second-order modal logic. [1 will be a place-

Philosophical Studies 49 (1986) 163-176.

?) 1986 by D. Reidel Publishing Company

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 164 JOHN BACON

holder for varying grades of necessity according to context. As a minimal

assumption, let necessity be closed under logical consequence:

Conditional Necessitation. If A 1- B then C1A 1- OB.

Taking transworld identification for granted, I assume D as the single common
domain (of existent or nonexistent individuals) for all possible worlds. (This
assumption is natural if the possible worlds are thought of as a heuristic aid
to speculating about the actual world, rather than as theoretical entities in
their own right.) This means that the converse Barcan principle will hold for
individual quantification:2

Barcan(x). CJVxA I1 VxODA.

Following Kripke, I take a property to be a function from possible worlds

to classes of individuals. Thus logically equivalent properties are identical.
(This assumption, while ultimately questionable, seems refined enough for
present purposes.) The predicate variables <p, 4 have properties as their sense-
range and classes as their denotation-range (in the sense of Church). Thus
in extensional contexts they behave just like class variables, with epx for
x E <p, etc. The conjunction of two properties is the function assigning to each
world the intersection of the properties' extensions (classes) there; and
similarly for disjunction, negation, infinite conjunction, etc. For quantifica-
tion over properties so construed, a second-order analogue of the Barcan
equivalence holds:

Barcan(sp). Y1pZA AS ZVpA.

Thus the same properties are found (although not necessarily instantiated)
in all words. Finally, I define

p C 4 for Vz(pz D 4'z)

Q C R for VxVy(xQy D xRy)
<p - for Vz(lpz4 z)
X 8- Y for O(X C Y)3

1.1. Base Properties and Supervenient Properties

As I use the term, the base B of a supervenience relation4 will be a set of

properties that is closed under (infinite) disjunction V, conjunction A, nega-
tion, etc. It seems natural also to assume the following closure property,

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 SUPERVENIENCE AND REDUCIBILITY 165

which I call diagonal closure. Where fp is

Bw = {<p: p E B}, B is also to contain an
for each world w. (This might also be thought of as the closure of B under
"resplicing".) The set S of supervenient properties is similarly closed. (As a
consequence of closure, S and B cannot be empty: they will always contain
the universal property and the empty property.) Since properties are func-
tions defined over all worlds, S and B will each be the same sets respectively
in all worlds, so that

(1) Sp-1FOSp, Bip A EOBep.

The restricted variables s and b will range over S and B respectively. In view
of (1) and diagonal closure, we have the surprising

(2) ? 3bA 1- 3bOA,

where A is free of modal operators and second-order predicates. For if in

every world (of some kind) some base property's extension satisfies A, then
diagonal closure enables us to splice those extensions together into a new
base property with extensions satisfying A in each world (of the relevant
kind). (In general, principles enunciated here for b and B are to hold for s and
S too.)

1.2. Equivalence-Classes and Natures

B-equivalence (likewise S-equivalence) is defined as

x- yforYb(bxDby).

[Given closure, this is equivalent to Vb(bx by).] As an equivalence relation,

- partitions D at each world into equivalence-classes whose members are in-
B
discernible by B properties. An individual x's B-equivalence-class { y: x B y}
will be the extension of x's B-nature, the conjunction of all its B properties:

[XIB for A{b : bx}.

As a defimed predicate, IXIB expresses a property, but at any particular

world it denotes x's B-equivalence-class there. By conjunctive closure, B-
natures are B properties:

(3) If-B[xIB.

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 166 JOHN BACON

Individuals share a B-nature iff they are B-equivalent:

(4) BX]BY 4 X-
For y has the conjunction of x's B properties iff it has each of them. Since
B
- is reflexive, we get in particular

(5) Iv- [XIBX.

(This paragraph holds also with s and S in place of b and B.)

1.3. Concepts Compared

We are now ready to compare the following variants of supervenience. The

leading formulation simply contraposes Campbell's definition from which we
set out. (Campbell's proviso 'while not following deductively' is taken as a
criterion of adequacy rather than as part of the definition.) Kim calls it
"weak supervenience":

Supervenience. B 8 - (Kim, 1978, p. 152)

Consequentiality. VXx([XlB C [xis) (Ross, 1930, p. 122)
Determination. E]VsVx{ sx D 3 b [bx & (b C s)] }
(Kim, 1983, p. 48)
Strong supervenience1. VsVx{ sx -8 3b [bx & (b 8- s)] }

(Kim, 1978, p. 155)

Strong supervenience2 F]VsVx{ sx D 3b [bx & (b 8- s)] }
(Kim, 1983, p. 49)5
Coextension. OVs 3b(s b) (Kim, 1983, p. 49(3))
Necessary coextension. Vs 3b L(s b) (Kim, 1983, p. 49(4))
Strong necessary coextension. OVs 3b n(s b)
Propositional coextension. VsVx 3b 3 yO(sx =by)

'Determination' is my term; compare Teller (1983b) and Post (1984), who

use it in slightly different senses. I credit "consequentiality" to Ross on the
basis of his characterization of value as a "consequential" property: "value
follows from the whole intrinsic nature of its possessors" (1930, pp. 121f).

1.4. Entailment Relations

It turns out that supervenience, consequentiality, determination, and coex-

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 SUPERVENIENCE AND REDUCIBILITY 167

tension are all equivalent. If the modality O is at le

von Wright's system T, then the remaining conce
Fig. 1 (the arrows indicate entailment).

strong necessary coextension

strong supervenience2

supervenience strong supervenience

necessary coextension

propositional coextension

Fig. 1

In S4, supervenience becomes equivalent to strong supervenience2 and strong

necessary coextension, yielding the simpler picture shown in Fig. 2.

supervenience

strong suprvenience,

necessary coextension

propositional coextension

Fig. 2

In S5, finally, all these notions collapse into equivalence except propositional
coextension.

1.5. Proofs

Here I sketch proofs of the entailments just claimed. It simplifies things that
nonmodal formulas can be treated set-theoretically, as only the denotations
of predicates matter in that case.
TI. Supervenience entails coextension. From the lemma that - C = entails

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 168 JOHN BACON

S- V{ [XiB: SX} for each s E S. For suppose sy: by (5) [YIBY. Thus [YIB E
{ [X]B: sx}, and [YIB C V{ [XIB: sx} by generalized disjunctive addition.
Since [Y1BY, it follows that y has V{ [XJB: SX}. Conversely, if y has this dis-
junction, then it has one of the disjuncts [XIB such that sx. By (4) we get
B S
X B Y, whence x
establishes the l
ties. By disjunct
tially generalize the conclusion of the lemma to 3b(s b) for each s. Ti
results by conditional necessitation.
T2. Coextension entails determination. Suppose Vs 3b(s b) and sx.
Then x has some b such that b C s. Thus demodalized determination follows
from demodalized coextension. T2 results by conditional necessitation.
T3. Determination entails consequentiality. By (5) x has [x]s, which by
(3) is in S. It follows by demodalized determination that x has some b such
that b C [xis. Since [XlB C b by generalized conjunctive simplification, we
get [XIB C [XiS. Conditional generalization yields Vx([xIB C [x]s), and
conditional necessitation yields T3.
B
T4. Consequentiality entails supervenience. Suppose tnat x = y. By (4)
[xJBy. Demodalized consequentiality then yields [x]sy, or x - y by (4). Thus
demodalized consequentiality entails B C_, so EVx([XiB C [X]S) entails
B S
=8- S.
T5. Supervenience -4 F- coextension - F determination - I- consequen-
tiality. From T1 -T4.
T6. Strong necessary coextension entails strong supervenience2. By a
proof similar to that of T2, necessary coextension entails VsVx{ sx D 3b [bx &
(b 8- s)] }. Conditional necessitation then gives T6.
T7. Strong supervenience2 entails supervenience. By denecessitation the
b 8- s component becomes b C s, yielding determination. Supervenience
follows by T3 and T4.
T8. Supervenience entails necessary coextension. By TI, supervenience
entails LVs 3b(s b), i.e. DVp[Sp D 3 b(up b)] . By Barcan(p) and necessi-
ty distribution, we get Vp[LlS<p D l 3b(p-b)]. (1) Let us erase the left
box. The right box permutes with the existential quantifier by (2), yielding
Vp[S D 3bO(po b)], which is necessary coextension.
T9. Strong supervenience2 entails strong superveniencel. By methods
similar to those of T8. [This is the only place where Barcan(x) is needed.]
TI0. Strong supervenience, entails necessary coextension. From the

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 SUPERVENIENCE AND REDUCIBILITY 169

lemma that strong supervenience1 entails LO(s V{b: 3 x[bx & (b 8- s)]})
for each s eS. I abbreviate {b: 3x[bx & (b 8- s)] } as B(s). Right-to-left:
by concretion, every b E B(s) is such that b 8- s. Hence, by the Boolean
analogue of constructive dilemma, VB(s) 8- s. Left-to-right: suppose sx.
Then it necessarily follows by strong supervenience1 that there is a b such
that bx & (b 8- s). Existential generalization gives 3 x [bx & (b 8- s)], whence
b E B(s) by abstraction. As x has b, by disjunctive addition x has VB(s). That
establishes s 8- VB(s), the other half of the consequent of the lemma. By
disjunctive closure, VB(s) is a B property. Thus T1O follows from the lemma
by existential and universal generalization of the consequent.
TI 1. Necessary coextension entails propositional coextension. By univer-
sal instantiation, necessary coextension yields a b such that O(sx bx) for
arbitrary x. Existential generalization gives 3 b 3yL(sx by). Finally, we
get VsVx 3b 3yEZl(sx by) by universal generalization.
That establishes the entailment relations in Fig. 1 of ? 14. Next we add the
NN-thesis of S4.
T1 2. In S4, supervenience entails strong necessary coextension. By condi-
tional necessitation, T8 gives us strong necessary coextension from OI(- C
S
_). T12 follows by the NN-thesis.
T13. In S4, supervenience strong supervenience2 - strong neces-
sary coextension. From T1 2, T6, and T7.
Finally, we move to S5 with the principle

(6) DADOC F-DA -8DC.

T14. In S5, necessary coextension entails strong necessary coextension.

From necessary coextension we get Vp{ Sp D 3 4 [O B 4 & 0U(p 4p)]} by
(1) and the NN-thesis. Let R(4p, 4) abbreviate [B4 & L(p 4)]. Converse
distribution gives Yep[Sp D 3 4,UR(p, 4,)], whence VFp[Sep D U 34 R(ip
by quantified modal logic. By denecessitation this becomes Vip[DSp D
O 3 D R(p, 4)], and by (6) VFp[USp D O 3 R(p, 4)]. Barcan(p), (1),
and denecessitation finally yield OVp[S&p D 3 R(p, 4)], i.e. strong neces-
sary coextension.

T15. In S5, supervenience strong supervenience, -IL necessary co-

extension. By T14, T13, T9, TIO.
T15 and T13, along with T5, equate all the variants of supervenience
except propositional coextension.

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 170 JOHN BACON

1.6. Propositional Supervenience

Before proceeding to evaluate the above results, we should take note of a

couple more varieties of supervenience that don't fit so straightforwardly into
the list of ? 13. First, Horgan (1976, 1982) and Currie (1982) have proposed
replacing individuals by possible worlds in the definition of supervenience.
The simplest such scheme, intended only as provisional, is Currie's

(7) VwVu[VxVb(bwx b ux) D VyVs(swy suy)]

(1982, Definition 2) (my notation: u, w range over possible worlds). We have

taken properties thus far as functions from possible worlds to classes of indi-
viduals, but functions from individuals to propositions (sets of possible
worlds) come to the same thing. Thus we may understand b as assigning a
proposition {u : bux} to each thing x in D. {u : bux} is true at w, i.e. b,x, if
w E {u: bux}. Now consider each such proposition { u: pux}, {w: pwy}, etc.
to be in a set -W iff p E B, and similarly for Y. Let ,B range over propositions
in M% and a over Y .
Now we can collapse quantifications like VxVb into quantifications over
propositions, and (7) becomes

(8) VwVu [Vf3(O,B -3u) D Va(u au)].

But this has the same form as demodalized supervenience, with the domain D
specialized to the set of possible worlds. Thus (8) is equivalent by (demodal-
ized) T5 to

Va 3, VYw(aw - )
i.e.,

(9) Va3 0(a =3),or Y C ,

where 0 expresses logical necessity. Since c, ,B range over propositions of the

form { u: sux}, { w: bwy}, we may re-expand (9) to

VsVx 3b3yyl(sx by),

which we recognize as propositional coextension, provided we identify the
D's of ? ? 12 and 16. It is accordingly appropriate to call the equivalent (7)
propositional supervenience.6 Similar observations apply to Horgan's proposals
(S-3), (s-4) (1982). Kim calls (7) "global supervenience" and asserts mis-
takenly that it entails strong supervenience2 (1985, pp. 22f).7

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 SUPERVENIENCE AND REDUCIBILITY 171

1.7. Supervenient Languages

Haugeland and Post put forward a notion very like (8), except that wand
Y are replaced by "languages". A language is a set of sentences with assigned
truth-values for each possible world in W. In other words, each sentence is
assigned a proposition (subset of W). Now Haugeland defines

Two worlds in W are discernible with language L just in case there is a sentence of L
which is true at one, and not at the other....

K weakly supervenes on L (relative to W) just in case any two worlds in W discernible

with K are discernible with L. (1982, p. 97).

Post's definition of "determination" is essentially the same (1984, p. 318

(DD)). The definiens could be translated into (8) if we took a to range over
K, , over L, and read 1w as "S3 is true at w". For an even closer approximation
to (8), note that each Haugeland language determines a set of propositions.
Let Y be the set of propositions expressed by K, and W those expressed
by L. Then K weakly supervenes on L (in Haugeland's sense) iff (8).
This still involves a slight reinterpretation of (8), as Haugeland's Y and
Mpropositions, in contrast to those of ? 16, are not confined to monadic
predications over a common domain. Nor is there any guarantee that Y and
M(K and L) are closed under truth-functions or their infinitary analogues.
To be sure, this makes no difference to (8), as two worlds are discernible
with L iff they are discernible with L's (infinitary) truth-functional closure.
But if we assume such closure of K and L, propositional coextension in the
form of (9) follows once again as in ? 16. Haugeland's weak supervenience is
thus very closely related to propositional supervenience.

1.8. New Results

I have now exhibited and demonstrated the logical relations among several
different concepts of supervenience. Some of these are fairly obvious, and
some were known before. So far as I know, T8, T10, and T12 are new. By
T8, (weak) supervenience entails necessary coextension, and not just co-
extension as in Kim (1983, p. 49(3)), provided we assume diagonal closure.
Barring that, strong supervenience, still entails necessary coextension (T10)
According to Kim, only strong supervenience2 entails necessary coexten-
sion. And by T12, in S4 (weak) supervenience entails strong supervenience2,
notwithstanding Kim's assertion, "Strong supervenience entails weak super-

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 172 JOHN BACON

venience; but the conve

pace Kim, global or propositional supervenience does not entail strong super-
venience2-

II. IMPLICATIONS

11.0. Necessary Coextension and Reducibility

The upshot of part I is that most of the concepts of supervenience considered

entail necessary coextension. The remainder entail the similar propositional
coextension, with the partial exception of Haugeland's linguistic supervenience.
What is the import of necessary coextension? According to Kim, it is distinct
from reducibility but "points to the possibility of reducing" (1983, p. 50).
Nevertheless, he holds, necessary coextension undercuts the autonomy of
S from B which supervenience theorists have maintained (ibid.).
As Kim points out, reducibility is partly an epistemological notion. Reduc-
tion may elude us because the particular predicates needed are not known, or
are too numerous and patternless for us to formulate. Also, reduction relates
entire theories, not just sets of properties. We can nevertheless capture the
ontological core of reducibility thus: S is reducible to B in principle iff every
S property is a B property; i.e.,

Reducibility in principle. S C B

Leaving to methodologists of science the question of when reducibility in

principle supports actual reduction, let us investigate the relation between
necessary coextension and reducibility in principle.

II1.. The Modality [

Following Kim, I have not fixed upon a specific interpretation of O. If it

were logical necessity, then necessary coextension (and strong necessary
coextension: T14) would be the same as reducibility in principle. But the
O in supervenience is not normally taken to be that strong. Only in mathe-
matical definitions by abstraction (e.g. of the rationals in terms of the natural
numbers) would the O approach logical necessity.
If S and B comprise properties expressible in two theories, then O will be
nomic necessity. Nomically necessary coextension is of course weaker than

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 SUPERVENIENCE AND REDUCIBILITY 173

reducibility in principle, but in many cases it will provide a good reason

for property identification. For example, the nomic co-ocurrence of schizo-
phrenia and a certain chemical in the brain may be grounds for identifying
schizophrenia with the presence of that chemical. Thus nomically necessary
coextension establishes a presumption of reducibility in principle. The burden
would be on the dualist to show theoretical cause for refusing reduction.
Nomic necessity has the structure of Feys and von Wright's modal system
T, not S4. For the latter, we should have to show that whatever is a law of
nature in a given possible world is likewise a law in every nomic alternative
to that world. Yet it seems clear that the same generalization might be a law
of nature in our world but a mere constant conjunction in an alternative
world.8 Thus nomic necessity does not satisfy the S4-model condition of
transitivity of accessibility. It accordingly fits Fig. 1 (? 14). For O as nomic
necessity, supervenience is weaker than strong supervenience2, as Kim has
observed.

11.2. Mind-Body Reduction

If an adequate psychology may be expected to come forth as science, then

the thesis of psychophysical supervenience will involve nomic necessity. By
T8, it will entail nomically necessary coextension. Thus we have a presump-
tion of psychophysical reducibility in principle. Is the supervenience theorist
obliged to accept this, or is there reason to refuse the reduction? I see no
reason, but defenders of a contingent-identity theory do. They insist that
materialism is a contingent thesis, false in some possible worlds (see Teller,
1983b, p. 149). On this question hinges the tenability of psychophysical
supervenience as a distinctive middle ground between reductive materialism
and dualism.

11.3. Supervenient Values

Moore, the father of modern supervenience (though not by name), argued

that in value theory, a is an "unconditional must" far stronger than nomic
but not quite so strong as logical necessity (1922, pp. 271-275). In light of
the "naturalistic fallacy", it is clear that Moore did not mean this to lead to
property identification and reducibility in principle. Similarly, Hare argues
at length that good-making characteristics do not entail a thing's goodness

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 174 JOHN BACON

(1952, p. 80 et seq.), although elsewhere he identifies the modality in super-

venience as logical!9. Kim suggests that in moral contexts Ol is "perhaps
best taken to signal logical or metaphysical necessity" (1985, p. 18). Since
logical and metaphysical necessity entail nomic necessity, if these philosophers
are right, moral supervenience would be in the same boat with psychophysical
supervenience, skirting reducibility. But to me it seems likely that O here
may be as weak as or even weaker than nomic necessity. If so, supervenience
might well be sustained as a distinctive nonreductive position in ethics.
This seems even more likely in aesthetics. 'No aesthetic difference without
a physical difference' is not even necessary across species, if aesthetic qualities
like beauty are taken as monadic. If beauty is a relation and materialism holds,
then the L in aesthetic supervenience might conceivably be nomic. The same
goes for monadic beauty supervening on a phenomenal rather than a physical
base. In value theory, then, supervenience may be more tenable as a happy
medium between reductive naturalism and extreme nonnaturalism.

11.4. Infinite Disjunctions

Kim's results (1983, 1985) as extended above might seem to narrow the use-
fulness of supervenience, particularly where potential theories are concerned.
Yet I have found most supervenience theorists unmoved. David Lewis has
suggested to me that they probably get off the boat right at the lemma of TI:
the formation of a possibly infinite disjunction of possibly infinite conjunc-
tions of base properties. (The disjunction will be infinite if the supervenient
property has infinitely many instances. A conjunction will be infinite if the
corresponding instance of the supervenient property has infinitely many
simple base properties.) Teller puts the point forcefully:

Properties which in this way do not show up similarities I call properties in an extended
sense only, an epithet which clearly applies to the disjunctive properties which follow
from strong supervenience as described by Necessary Coextension.
Next, we must keep in mind that these disjunctive properties are going to be awfully
fat....
Now what, I want to ask, makes it appropriate to call a property 'physical' when it
is such a disparate and infinitely long disjunction of disjuncts each of which is prob-
ably already an uncountable [?] conjunction of fimite physical descriptions? Yes, it is
a Boolean combination of physical properties, but I feel that to call it "physical" threat-
ens to be misleading. (1983a, p. 58f)

My feeling is the opposite, but the essential point is conceded: a Boolean

combination of physical properties. To stick at calling such combinations

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 SUPERVENIENCE AND REDUCIBILITY 175

physical seems a desperate verbal ploy to save the "mental". If that's all
it is for a property to be mental, then ontologically materialism has carried
the day. In any event, the results of part I can all be reinstated by simply
replacing the word 'property' by 'Boolean combination of properties' through-
out.

11.5. Conclusion

Supervenience in most of its guises entails necessary coextension. Thus

theoretical supervenience entails nomically necessary coextension. Kim's
result, thus strengthened, has yet to hit home. I suspect that many super-
venience enthusiasts would cool at necessary coextension: they didn't mean
to be saying anything quite so strong. Furthermore, nomically necessary
coextension can be a good reason for property identification, leading to
reducibility in principle. This again is more than many supervenience theorists
bargained for. They wanted supervenience without reducibility. It is not
always available for this mediating role.

ACKNOWLEDGEMENT

Stimulating discussions with William G. Lycan helped fo

supervenience. I am grateful to David Benfield for usefu
John L. Pollock for improvements.

NOTES

'Supervenient' in this sense seems to appear first in Hare (1952, p. 80), but Hare
denies originating this use of the term. For further history, see Kim (1982).
2 A single comprehensive domain D is compatible with varying inner domains Eu, EW,...
of "existents" for different words u, w,... "Inner" quantifiers, restricted to existents,
would not satisfy the Barcan equivalence, for existence is a contingent matter. Here I
use "outer" individual quantifiers, ranging over all of D.
3 The symbol for strict inclusion, or property-entailment, is due to Barcan Marcus
(1946), viz. an inverted fishhook, here approximated as '8-'.
4 Sometimes 'base' is used for a more limited set of properties whose closure is B, and
sometimes for the particular property upon which another supervenes.
' In (1982, p. 134) Kim defined strong supervenience as VsVx{sx D 3 b(bx & (b 8- s)] }
(my notation), but this is clearly a slip of the pen for strong supervenience1 or 2
6 Currie himself does not rest with (7), but develops a more complicated notion of
supervenience involving quantification over times and permutations of D. He derives a
corresponding version of coextension (1982, Claim 3).

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms
 176 JOHN BACON

7 The assertion is refuted by the following countermodel. Let D = {x, y } and let w be
the sole possible world. Let the accessibility relation be universal over worlds. Let s be
true of x and false of y in w. Let b be the universal property. S is the closure of {s} and
B of {b}. [XIB or B* (x's B-maximal property) is then the same as b. In this model,
(7) is trivially true, while b I s.
I am indebted to D. M. Armstrong for this insight. See my (1980).
"Suppose that we say 'St. Francis was a good man'. It is logically impossible to say
this and to maintain at the same time that there might have been another man placed
in precisely the same circumstances as St. Francis, and who behaved in exactly the same
way, but who differed from St. Francis in this respect only, that he was not a good man."
(1952, p. 145, my italics)

BIBLIOGRAPHY

Bacon, John: 1980, 'What is physical necessity?', Proceedings of the Russellian Society
(Sydney) 5, pp. 12-26.
Barcan Marcus, Ruth C.: 1946, Dissertation (Yale, New Haven).
Campbell, Keith: 1981, 'Naturalism in moral philosophy',Proceedings of the Russellian
Society (Sydney) 6, pp. 12-22.
Currie, Gregory: 1982, 'Individualism and global supervenience', Australasian Associa-
tion of Philosophy (La Trobe University, Melbourne, 26 August).
Davidson, Donald: 1970, 'Mental events', in Lawrence Foster and J. W. Swanson (eds.):
Experience and Theory (University of Massachusetts Press, Amherst), pp. 79-101.
Hare, R. M.: 1952, 7he Language of Morals (Clarendon Press, Oxford).
Haugeland, John: 1982, 'Weak supervenience', American Philosophical Quarterly 19, pp.
93-103.
Horgan, Terence E.: 1976, 'The supervenience of the mental' [abstract], Journal of
Philosophy 73, p. 762 (Main defmition not in abstract, but quoted in Kim (1982,
p. 133n)).
Horgan, Terence E.: 1982, 'Supervenience and microphysics', Pacific Philosophical
Quarterly 63, pp. 29-43.
Kim, Jaegwon: 1978, 'Supervenience and nomological incommensurables', American
Philosophical Quarterly 15, pp. 149-156.
Kim, Jaegwon: 1982, 'Psychophysical supervenience as a mind-body theory', Cognition
and Brain Theory 5, pp. 129-147.
Kim, Jaegwon: 1983, 'Supervenience and supervenient causation', Southern Journal of
Philosophy, supp. 22 (1984), pp. 45-56.
Kim, Jaegwon: 1985, 'Concepts of supervenience', Philosophy and Phenomenological
Research, forthcoming. (MS pages cited.)
Moore, G. E.: 1922, 'The conception of intrinsic value', in his Philosophical Studies
(Routledge and Kegan Paul, London), pp. 253-275.
Post, John F.: 1984, 'On the determinacy of valuation', Philosophical Studies 45, pp.
315-333.
Ross, W. D.: 1930, The Right and the Good (Clarendon Press, Oxford).
Teller, Paul: 1983a, 'Comments on Kim's paper', Southern Journal of Philosophy,
supp. 22 (1984), pp. 57-6 1.
Teller, Paul: 1983b, 'A poor man's guide to supervenience and determination', ibid.,
pp. 137-162.

Department of Philosophy, University of Sydney,

Sydney, NSW2006, Australia.

This content downloaded from

130.251.200.3 on Sat, 09 Jan 2021 16:38:03 UTC
All use subject to https://about.jstor.org/terms