Multiple Realization,
Reduction and Mental
Properties
in International Studies in the
Philosophy of Science 13 (1999), No. 2, pp. 135-149.
Abstract.
The paper tries to
remove some obstacles standing in the way of considering mental properties as
both genuine natural kinds and causally efficacious rather than epiphenomena.
As the case of temperature shows, it isn't justified to conclude from a
property's being multiply realizable to its being irreducible. Yet Kim's (1992)
argument to the effect that if a property is multiply realizable with a
heterogeneous reduction base then it can't be a natural kind and possesses only
derivative "epiphenomenal" causal efficacy isn't conclusive either.
The fact that temperature is, but jade is not, a natural kind cannot be
established by comparing the heterogeneity of their respective reduction bases,
but rather by the fact that the former is and the latter isn't embedded in laws
of nature.
Multiple Realization, Reduction and Mental Properties
There has been a broad consensus in the philosophy of mind regarding the
functionalist conception of the nature of mental states. According to
functionalism, a token mental state belongs to a given type because it is
causally related, or potentially causally related, to other mental states and
to sensory input and behavioral output. When this idea was first proposed by
Putnam (1967), the major competing theoretical hypothesis about the mind was that
mental state types should turn out to be reducible to brain state types.
Functionalism won the assent of the philosophical community mainly by Putnam's
now famous argument that such a reduction is impossible because mental states
of the same type can be biologically or mechanically realized in many ways.
According to Putnam himself, this is an empirical possibility; as a result -
what came later to be called - the "argument from Multiple Realizability
(MR)[1]" appeared to stand on an empirical premise. Since Lewis (1969),
however, this argument has been reinterpreted: it is now considered to be
purely conceptual. Our concept of a mental state implies that its physical
realization does not condition its identity. Materialism requires each mental
state to be physically realized in some way or another, but a priori the possible realizers form an
open disjunction.
This way of putting things was responsible for some intriguing features
of the functionalist consensus. On the one hand, the argument from MR linked functionalism
with antireductionism. If the possible realizers of a given mental state type
are many, one can't reduce it to any single type of realizer. But on the other
hand, the concept of realization deprives the (realized) mental state not only
of independent existence (this much, every materialist will concede anyway),
but also of independent causal powers. To speak of a mental state M as being realized, in a given cognitive agent and
in a particular situation, by a token of a physical state P, means that P is
where we have to look if we want to know about M's causal powers.
The aim of this paper is to examine whether the fact that mental
properties are multiply realizable warrants the conclusion that they are not natural
kinds. Such a result would have far-reaching consequences for if mental
properties are not natural kinds then they cannot constitute the subject matter
of a genuine science. Kim (1992) has argued along these lines against the
scientific character of general psychology. Considering his arguments will
constitute my starting point for tackling the issue of the status of multiply
realizable properties. It will soon become clear that assessing them
presupposes a critical examination of the role of theoretical reduction of
higher-level by lower-level theories. Eventually both the traditional argument
from MR to the irreducibility of the mental and Kim's argument from MR to the
thesis that the mental doesn't constitute the subject matter of a genuine science
will be found wanting.
1. Functionalism,
realization, and causal powers
The conception according to which a mental state is "realized"
by a neural state in the brain contains the idea that the "realizer"
monopolizes causal power (Cf. Kim 1984; 1993a, p. 366) at the expense of the
realized mental state. Now, there is a considerable tension between this
implication of the concept of realization and the original functionalist
conception of a mental state (Cf. Block 1990). The question of where the causal
powers lie is essential for functionalism: it is based on the postulate that
the identity of a mental state M is determined by and only by its causal
interactions (including its potential causal interactions) with other mental
states (and with sensory input and behavioral output), but not by the
properties of the brain in which these mental states occur. Functionalism thus
crucially depends on the causal efficacy of the mental states themselves as
opposed to the brain states that realize them[2]. But it is difficult to see how functionalism could resist the
following argument leading to the opposite conclusion: if a mental state has a
realizer which is a neural or other biologically or physically specified state,
then this realizing state has causal powers of its own. Now attributing causal
power both to realizer and realized mental state makes it look as if there were
systematically causal overdetermination each time a mental state acts as a
cause of the same effect as the realizer causes. Kim proposes to avoid this
apparent overdetermination by appealing to a principle of explanatory exclusion
(Kim 1988; 1989b): of a given effect there cannot be two complete but
independent causal explanations. But where a mental property and the neural
property realizing it compete for causal power, the neural property wins. Kim
reconciles this with the apparent causal efficacy of mental states by way of
the thesis that mental causation - as indeed all other higher-level causation -
is "epiphenomenal" or "supervenient" causation (Kim 1984).
Kim explicates this relation as follows:
"The general
schema for reducing a macrocausal relation between two events, x's having F and
y's having G, where F and G are macroproperties, is this: x's having F
supervenes on x's having m(F), y's having G supervenes on y's having m(G),
where m(F) and m(G) are microproperties relative to F and G, and there is an
appropriate causal connection between x's having m(F) and y's having m(G). Any
causal relation conforming to the pattern set forth above will be called a
'supervenient causal relation'." (Kim 1984, p. 262).
In (Kim 1992), Kim does not argue for his theory of supervenient
causation, but simply reminds us of its conclusion, namely the "Causal
Inheritance Principle" which says that "if mental property M is realized in a system at t in virtue of physical realization base
P, the causal powers of this instance of M are identical with the causal powers of P." (Kim 1992, p. 18). From this he draws the conclusion that
functionalism is wrong in thinking that it is possible to identify mental
states on the basis of their causal interactions. In trying to do so,
functionalists in fact just apply a generally valid methodological principle
for the identification of natural kinds, a principle which Kim calls the
Principle of Causal Individuation of Kinds. It says that "kinds in science
are individuated on the basis of causal powers; that is, objects and events
fall under a kind, or share in a property, insofar as they have similar causal
powers." (Kim 1992, p. 17).
Not the least of Kim's merits is the fact that he has spelled out more
explicitly than others that the idea that mental states are (multiply)
"realized" by physical states leads to an irrealist position with
respect to mental states. The reason is that causation is the best indicator we
have to know whether a property is real. But if the Causal Inheritance
Principle is correct, the functionalist's attempt will not lead to the
specification of any well-determined natural kind. Kim concludes that "mental
kinds cannot satisfy the Causal Individuation Principle, and this effectively
rules out mental kinds as scientific kinds." (Kim 1992, p. 18)[3].
2. Theoretical reduction
and property identity
Kim (1992) attacks antireductionism on two grounds. We have seen his
first argument: mental kinds are realized by physical kinds; kinds are
individuated causally and the causal powers of the realizing property are
ontologically prior to the causal powers of the realized state. From these
premisses he concludes that "mental kinds are not causal kinds, and hence
are disqualified as proper scientific kinds" (Kim 1992, p. 18)[4].
His second argument begins with a careful spelling out of his opponents'
- the antireductionists - line of reasoning according to which MR entails the
irreducibility of the higher-order multiply realized state. Why is it that MR
is said to lead to irreducibility? If M can be realized by a large number of
different physical states N1, N2,..., then the reduction
base would have to be a "disjunctive property" N1 Ú N2 Ú N3 ... with many or indeed infinitely numerous disjuncts[5]. But why is a mental state type irreducible if the only candidate for
such a reduction is a disjunction? According to Kim,
"for genuine
reduction, the bridge laws must be construed as property identities, not mere property
correlations [...] This of course requires that each T2-term [i.e. each term of the reduced theory, M.K.]
have a nomic (or otherwise suitably modalized) coextension in the vocabulary of
the reduction base. To put it in another way, ontologically significant
reduction requires the reduction of higher-level properties, and this in turn requires (unless one takes an
eliminativist stance) that they be identified with complexes of lower level properties.
Identity of properties of course requires, at a minimum, an appropriately
modalized coextensivity." (Kim 1992, p. 9; italics are Kim's).
This requirement is the first premiss of the argument from MR to
irreducibility. It amounts to a very restrictive interpretation of the concept
of theoretical reduction. Kim cites Hellman and Thompson's (1975) formulation
according to which reductionism is the thesis "that all scientific terms
can be given explicit definitions in physical terms" (Hellman and Thompson
1975, p. 551; Kim 1992, p. 2). But instead of talking in terms of the definition of terms, Kim himself then goes on to formulate his discussion of
reduction in terms of laws linking
higher-level to lower-level properties.
In the latter - realist - terminology, reduction requires bridge laws linking
reduced to reducing properties. This in turn requires that both the reducing
and the reduced properties are natural kinds.
Now, the argument from MR precisely denies that this latter requirement
can be fulfilled in the case of multiply realizable properties. The denial
depends on a second premiss which Kim sets out to explicate and to defend. The
potential realizers form a heterogeneous disjunction, but "a disjunction
of heterogeneous kinds is not itself a kind." (Kim 1992, p. 9). Kim's main
argument for this claim is that predicates expressing a property with a
disjunctive realization basis, like "being jade", are not
projectible: they are not confirmed by their instances[6].
With these premises the irreducibility of M follows from its being
multiply realizable: If M is "realized" in many different ways, the
only plausible candidate for reduction is a "non-kind", namely the
disjunction of all these actual and potential realizers. But Kim pursues the
entailments of the argument further than its antireductionist proponents:
"If M is
identified with non-kind Q (or M is reduced via a biconditional bridge
principle 'M «
Q', where Q is a non-kind), M could no longer figure in special science laws;
e.g., the law, 'M ® R', would in effect reduce to 'Q ®
R', and therefore loses its status as a law on account of containing Q, a
non-kind." (Kim 1992, p. 10).
This consideration leads to turning the arguments' conclusion against
the antireductionists. If being nomically equivalent to a non-kind means to be itself a non-kind, then if the only
plausible reduction base for M is disjunctive, we can't have both: hold that M
can be reduced to a disjunctive property and
hold that M continues to be a natural kind, figuring in laws on the level of
mental properties. Faced with this alternative, instead of choosing to maintain
M as a natural kind and abandoning the perspective of its reduction, Kim
proposes to make the opposite choice: M is reducible though only locally, i.e.
there is a local reduction of M whose validity is restricted to precisely one
type of realizer. On the other hand, Kim is prepared to accept the conclusion
that M itself can't be considered as a natural kind any more. He is thus led to
"a form of mental eliminativism" (1992, p. 25)[7].
The above argument defended both by the antireductionists and their
opponents relies crucially on the identity
of reducing and reduced property. Indeed, only if M and Q are identical, do we
get the crucial conclusion that "'M ®
R' would in effect reduce to 'Q ® R'" (Kim 1992, p. 10). But
it seems to me that it is wrong to impose such a strong requirement on the
possibility of reduction. If I can show that this conception of what
theoretical reduction comes to is misguided, I shall have shed doubt on the
soundness both of the argument from MR to irreducibility and of Kim's argument
that local reduction leads to the elimination of the reduced kind.
The relevant condition Nagel (1961) imposes on the reduction of a
predicate P1
in a higher level theory T1 by a predicate P2 in a lower-level theory T2 is weaker. Indeed, in order to accomplish a successful reduction of T1, T2 must be supplemented with linking postulates establishing a connection
between those predicates of T1 which do not appear in T2, and T2.
In general T2
will not contain any predicate synonymous
with P1, e.g., neither
"mean molecular energy" nor any other term of statistical mechanics
is synonymous with "temperature". Once this possibility is excluded,
the linkage could in principle either be the result of a convention which
stipulates a coordinating definition of P1 by P2,
or the object of a factual hypothesis[8]. What is crucial for our present argument is that neither of the
remaining two interpretations - convention or factual hypothesis - implies that
a successful reduction of P1 by P2
shows that both predicates designate one and the same property.
It order not to beg any questions, it may be useful to compare reduction
with a non-reductive systematic link between two families of properties where
those properties are clearly not identical with each other. Thus, basic colors
are systematically correlated to light of definite wavelengths. This link
presumably has the status of an empirical psychophysical hypothesis, but it
could also be interpreted as a set of coordinating definitions. Yet neither
interpretation would imply that, e.g., the secondary property of being yellow
is identical to the property of
possessing a wavelength of 589 nm.
Other than electromagnetic waves may have that wavelength without thereby being
yellow.
Linking postulates are necessary but not sufficient for successful
reduction. Rather, the decisive criterion is the theoretical fruitfulness of
the connection achieved by those postulates. One is justified in considering a
reduction as successful if it suggests new and interesting generalizations,
both in the reducing and in the reduced theory. There are indeed good reasons
for not requiring the linking postulates either to be "bridge laws"
or to express property identities. On looking closer to paradigmatic examples of
successful property reductions in physics, it turns out that they don't have
the properties philosophers often think they have. Why should the reduction of
the mental to the neural fulfil requirements so rigourous that even a
paradigmatic case of successful reduction in physics, such as that of
temperature to mean molecular energy, fails to satisfy them?
First, temperature is multiply realizable, but its being multiply
realizable is not an obstacle to its reduction. Second, even if one takes into
account only the main reduction base of temperature, mean molecular energy, it
turns out that there are objects which possess the property in the reduction
base - mean molecular energy - without possessing the reduced property -
temperature. Coextensivity[9], and a fortiori identity of
reduced and reducing property, fail[10].
In order not to bring in any controversial issues which are inessential
for the point presently at stake, let us focus on gases: these are the
paradigmatic substances for which the reduction is valid. Consider an object
constituted by only two molecules of
some gas. This object definitely has mean molecular energy - the mean being
taken both over time and the two molecules - but it doesn't have any
temperature. The concept of temperature can only be applied to macroscopic
objects[11]. Or consider the artificial object constituted by the sum of a great
number of gas molecules, but where these molecules are isolated from each
other. For example, take the mereological sum of one molecule out of each of a
series of boxes filled with some gas. Again, the resulting object will possess
mean molecular energy, but no temperature. This is because the number of
elements is not the only restriction placed on the application of thermodynamic
concepts like temperature to a given object. The elements constituting the
object in question must furthermore form an interacting whole. But in virtue of
the constitution of our mereological sum, its elements do not interact with
each other; thus the object doesn't have a temperature. Nevertheless, it has
mean molecular energy, for the application of this concept requires neither a minimal number of elements nor
interaction between these elements. This should be sufficient to show that in
the paradigmatic case of the reduction of temperature to mean molecular energy,
this reduction does not result in the identity
of the reduced with the reducing property. In fact, these properties aren't
even coextensive[12], let alone nomically or necessarily coextensive.
Secondly, the reduction base of temperature is multiple[13]. The concept of temperature has recently been applied both to describe
the state of the matter forming the atomic nucleus and to empty space. In the
first case, there are no molecules which could possess mean molecular energy;
in the second case there aren't any particles at all. In view of the case of
nuclear temperature, we can drop the mention of molecules which now appears to
reflect the narrow application of the temperature concept before the discovery
of the inner structure of the nucleus. If we redefine the reduction base to be
simply mean kinetic energy, this property can reduce both nuclear temperature
and the temperature of solids, liquids and gases. However, the reduction base
so redefined still doesn't cover the application of the concept of temperature
to empty space. Indeed, this application isn't based on kinetic energy but on
the fact that black bodies[14] have a spectrum of radiation which is specific for each temperature.
This perfect mapping allows one to extend the concept of temperature to objects
deprived of matter but containing radiation - which is the case for empty
space. This means that temperature is, after all, neither reducible to mean
molecular energy nor to mean kinetic energy but to mean molecular energy or to the black body radiation spectrum.
How does this help us with our problem about knowing whether a mental
property with a multiple reduction base can nevertheless be considered as a
natural kind, and even an irreducible natural kind - as antireductionists argue
- or whether "as a property or kind [it] must go", for as Kim argues,
"to be reduced is to be eliminated as an independent entity" (Kim
1992, p. 24)? I propose to infer from the analogy with temperature that neither
of these conclusions is correct.
1. The argument from MR to irreducibility is wrong: Temperature has been
reduced[15], pain and other mental states are potentially reducible, all with a
multiple reduction base. This means that a disjunctive predicate is on one end
of the translation hypothesis which is the core of theoretical reduction. The
argument from MR to irreducibility is wrong, because on the basis of the
multiplicity of their (actual or potential) reduction base, temperature and
mental states are in the same boat. Parity of reasoning prevents us from taking
MR as a ground for denying reducibility to mental states in so far as the
application of the same reasoning to the case of temperature leads to the
manifestly wrong result that temperature is irreducible.
2. Kim is wrong, too: Reduction with a multiple reduction base doesn't
entail elimination "as an independent entity". Kim's thesis can be
interpreted as a thesis about elimination as
a causally efficacious property. The extension of the concept of temperature
from matter to empty space shows that reduction is not the basis for
considering a property as real. As Kim himself points out, to be considered as
real - where "real" always means "(potentially) causally
efficacious" - a property has to figure in genuine laws[16]. Kim (1992, p. 12) proposes to distinguish the latter by the fact that
they are confirmed by their instances, whereas other generalizations are not[17]. This is not the place to enter into the difficult subject of criteria
for lawfulness. But it is certainly the case that the success of macroscopic
thermodynamics combined with the fact that if that theory were true there would
be many objective laws of nature involving temperature, are responsible for our
considering temperature as an independent property. "Independent" of
the issue of its reducibility, and a
fortiori of whether it is multiply or simply reducible. We should consider
both the reality of a property and its reducibility as empirical issues: to
find out about the former, we have to look at scientific inquiry and see if
there are any lawful connections between the property in question and other
properties on the same ontological level. This criterion of when some property
is "real", i.e. constitutive of a natural kind, and when on the contrary
it "must go" as a natural kind property, also proves successful in
more controversial cases. Kim elaborates the case of jade - which is a kind for
common sense and for gemology but which has a dual reduction base in
mineralogy, consisting of jadeite and nephrite. Presumably, neither common
sense nor gemology are sciences. They don't identify laws of nature. Being jade
is not a natural kind property because no law contains this property, whereas
being jadeite and being nephrite are kinds because they are lawfully connected
to other properties identified by mineralogy.
Questions about reducibility must be postponed to a time when both the
property itself is embedded in a scientific theory and there are candidate
theories describing reality on a lower ontological level. Pain (and other
mental states) in general - as opposed to pain in a given species - will turn
out to be a natural kind if there turns out to be a general psychology
providing evidence that there exist laws linking pain (and other mental states)
to other mental or non-mental properties which are instantiated by more than
one species. This is an empirical issue, and it is independent of the question
about the reducibility of these kinds - multiple or not.
3. Homogeneous and
heterogeneous reduction bases
One might object to this reasoning that it doesn't do justice to the
important question as to how
heterogeneous the set of properties in the reduction base is. In this
sense, Kim says that "there is nothing wrong with disjunctive predicates
as such; the trouble arises when the kinds denoted by the disjoined predicates
are heterogeneous, 'wildly disjunctive', so that instances falling under them
do not show the kind of 'similarity', or unity, that we expect of instances
falling under a single kind." (Kim 1992, p. 13). There is, it might be
argued, a crucial difference between temperature on the one hand and jade and
pain (and other mental states) on the other. Both the functionalists' argument
from MR to irreducibility and Kim's argument from MR to the elimination of the
multiply realized property as a natural kind are motivated by, if not
explicitly dependent upon, the supposed heterogeneity of the lower level
properties[18]. Thus, the objection might go, temperature is reducible, although
reducible by a multiple reduction base, because the properties contained in the
reduction base are relatively homogeneous. This is why the functionalist
needn't consider temperature as irreducible and why Kim isn't forced to
conclude that temperature is not, after all, a natural kind. But, according to
the objection, the situation is crucially different with pain and jade: Their reduction bases are heterogeneous
to a degree higher than some treshold.
In order for this objection to be successful and thus to be able to
rescue either the functionalists' or Kim's argument, it would have to be
supplemented by a clear-cut criterion for distinguishing homogeneity from
heterogeneity in the relevant sense. A plausible criterion seems to be that two
properties are homogeneous if and only if they have (important) properties in
common, and (totally) heterogeneous, if and only if they have few (or no), or
only unimportant, properties in common. It should be uncontroversial that such
a distinction must allow for degrees. But how can it be made precise? -
sufficiently precise to justify the judgment that the pair jadeite/nephrite is
significantly more heterogeneous than the pair mean kinetic energy/specific
distribution of radiation (obeying the law of the black-body), and - what is
certainly even more difficult - to justify the further judgment that pain is
more similar to jade than to temperature in this respect. It is clear that
neither pair is totally
heterogeneous: Jadeite and nephrite have sufficiently many properties in common
to make for the very similar macroscopic properties of the crystals of those
types. Similarly, mean kinetic energy and the specific distribution of
radiation obeying the black body law have something in common which enables
objects possessing either one or the other of these properties to reach a
common state of thermodynamic equilibrium[19].
Plausibly, though, neither of the common properties shared by the pair
in question could be used decisively in the issue at hand, namely that of
distinguishing homegeneous from heterogeneous in the context of different types
of MR. For both the properties common to jadeite and nephrite underlying the
similar superficial properties of the respective crystals, and the property
common to mean kinetic energy and the specific distribution of radiation
obeying the black body law, enabling the objects possessing them to reach a
state of thermodynamic equilibrium, are properties which are characterised by
reference to a property which belongs to the level of the reduced property. But the fact that they have
something in common on that level is
trivial for we precisely undertook to consider those pairs of properties
because they are reducing the same property! It means nothing else than that
they share the property of belonging to the same reduction base.
Kim proposes a criterion of confirmability: A general statement with an
antecedent denoting a "wildly disjunctive" property is not in general
confirmed by its instances. Thus, according to Kim, we would not be justified
in considering
(1) "Jade is green"
as confirmed by its positive instances if all samples ever considered
turned out to have been jadeite (Cf. Kim 1992, p. 12). In that case, he argues,
all those jadeite samples do not licence any conclusion about the properties of
nephrite and thus don't in the end confirm (1) either.
I think one should resist that conclusion. The samples of green jadeite do confirm the statement (1), for the
green appearance is precisely one of the properties which are responsible for
the grouping together of jadeite and nephrite under the higher-level property
of being jade. To make his point, Kim would have to find a property belonging
exclusively to jadeite and not to nephrite, say containing atoms of aluminium
(Al). The jadeite samples still confirm the general statement
(2) "Jade contains
Al",
but this time, the confirmation is deceptive. (2) is refuted by the
first observed sample of nephrite which does not contain Al. However, its being
deceptive is perfectly compatible with its being a confirmation. We know that
induction is logically invalid; confirmation is inductive and can therefore
never provide a proof of the confirmed general statement. Many of the best
confirmed statements of laws of nature have turned out to be false.
Nevertheless they had been confirmed, until cases refuting them were
discovered. Nothing else happens when the (hypothetical) generalization (2) is
refuted by a sample of nephrite, i.e. a sample of jade not containing Al. The
reasonable move is then to modify statement (2) and choose as a substitute a
more modest conjecture:
(3) "Jadeite
contains Al",
as long as it is not
discovered to be false in turn. The case of temperature is analogous. Before
discovering that empty space has temperature as well, one might have
conjectured that
(4) "Everything which has temperature has moving parts".
Such a putative law[20] would have been regularly confirmed by all material samples[21]. However, it was refuted by the discovery that regions of space devoid
of matter also had temperature. As with jade, the correct move is to modify the
statement by reducing the width of its scope:
(5) "Everything material which has temperature has moving
parts".
The important point to note is that the refutation of a general
statement like (2) or (4), i.e. the discovery that the preceding confirmation
was deceptive, does not automatically entail that the property in question was
not after all a natural kind. Temperature is and remains a natural kind even
though there are generalizations involving temperature - like (4) - which have
been refuted. By parity of reasoning, Kim would have to say that it then turned
out that (4) was "not confirmed by its instances", for the
conclusions one can draw about the properties of empty space on the basis of
the observation of samples of matter with temperature are just as limited as
the conclusions one can draw about samples of nephrite on the basis of the
observation of samples of jadeite. But unlike jade, temperature's having a
disjunctive reduction base doesn't threaten its being a natural kind. The
reason is that there are other laws
involving temperature which we are well justified to believe, such as the 0th
law of thermodynamics, or the Boyle-Mariotte law for ideal gases: pV = nRT. It
suffices that there exists one law implying a property in order for this
property to be a natural kind. If, on the other hand, we don't take jade to be
a natural kind, this is not because, as Kim suggests, there are general
statements - like (2) - involving the predicate "jade" and whose
confirmation by its instances was deceptive, but because there is no law at all
which we have good reasons to believe and which involves the property of being
jade.
I conclude that the capacity of a general statement to be confirmed by
its instances does not provide a criterion for whether the predicates linked by
that statement denote natural kind properties or not. In the end, the criterion
Kim proposes doesn't allow one to make precise the intuitive distinction
between predicates denoting "wildly disjunctive" properties and
predicates denoting genuine natural kinds. In the form Kim gives it, parity of
reasoning forces him once more to rule out temperature together with jade from
the realm of genuine natural kinds. This consequence suffices to undermine the
credibility of Kim's criterion.
It seems to me that the burden of proof lies with the advocate of the
homogenous/heterogeneous distinction: In order to block my argument by
invalidating the analogy with temperature, both the functionalist and Kim must
either spell out a clear-cut criterion of what counts as being sufficiently
heterogeneous for membership in the same category as jade and what counts as
being sufficiently homogeneous for membership in the same category as
temperature, or at least show that such a criterion could be found in
principle. In the absence of such a criterion, the analogy stands and with it
my argument against the functionalists' irreducibility thesis and Kim's elimination
thesis.
4. Local reduction and
eliminativism
Before finishing I would like to point to an unwelcome consequence of
Kim's argument for the "conclusion that pain as a property or kind must
go" (Kim, 1992, p. 24). If it were correct, it would lead to radical
eliminativism, Kim's explicit denial nonwithstanding. According to Kim,
"the present approach is not, in its ontological implications, a form of
the standard mental eliminativism currently on the scene" (Kim 1992, p.
25). I shall briefly try to show that Kim's argument, if it were correct,
couldn't both show that general pain is
not a natural kind and that its
species-specific substituents are
natural kinds. By parity of reasoning, all the local species-specific
restrictions of our usual general psychological properties would fall prey to
analogous arguments from reduction by a disjunctive reduction base.
According to Kim, whenever a property has a disjunctive reduction base,
neither the reducing disjunctive property nor the reduced property are genuine
kinds. "Local reduction" permits one to show where the genuine kinds
are. In the case of mental properties like pain, the genuine properties are
species-specific. "The present view allows, and in fact encourages,
'species-specific psychologies', but the standard eliminativism would do away
with all things psychological - species-specific psychologies as well as global
psychology." (Kim 1992, p. 26). In an earlier text, Kim (1989a) himselfs
hints at a general difficulty for the intended substitution of species-specific
psychologies for general psychology. He considers the possibility that science
may discover species-specific bridge laws reducing mental properties separately
for each biological species. These laws would have the form "Si ® (M « Pi)" (Kim 1989a, p. 273). For each particular species Si, there is
some physical property Pi whose instantiation by an individual of this species
is both necessary and sufficient for its instantiation of the mental property
M. Kim pauses to note that
"in order to
generate laws of this kind, biological species may turn out to be too wide;
individual differences in the localization of psychological functions in the
brain are well known. Moreover, given the phenomena of learning and maturation,
injuries to the brain etc., the neural structure that subserves a psychological
state or function may change for an individual over its lifetime" (Kim
1989a, p. 273)[22].
Kim doesn't seem to appreciate the fact that there is an enormous
difference between a linking proposition - I hesitate to call it a law -
establishing a nomic link between a mental state type and the neural state of one individual at one moment of his lifetime on the one hand, and on the other hand a
genuine bridge law between that mental state type and a general neural state
type, i.e. a neural state which could be instantiated in all (or at least many)
individuals of the species, and more than only once in their lifetime. On the
contrary, Kim immediately continues by playing down this difference. "What
is important, he says, is that these laws are relative to physical-biological
structure-types, although for simplicity I will continue to put the matter in
terms of species" (Kim, 1989a, p. 273). But the thesis that for a given
species, there exists a non-disjunctive reduction base for mental properties,
is far from being a negligible detail in Kim's argument which could be set
aside for simplicity's sake. On the contrary, it is the centerpiece of his
response to Putnam's (1967) argument according to which MR blocks all hopes for
reduction of mental state types. If it really turned out to be the case that
the neural reduction base for a given mental state M is different not only for
each species, but for each individual and even for each moment in the lifetime
of this individual, this would according to the notion of reduction used by Kim
indeed make even "local", i.e. species-specific, reduction
impossible.
We have good reasons to expect that the properties which are at the
basis of a species-specific reduction turn out to be themselves multiply
reducible. Let M be reducible, in humans, to the neural property P. Then due to
individual differences and to the evolution of each individual, P is likely to
be itself reducible to a disjunction of complex chemical and physical
properties of the brain. If Kim's argument concerning M were sound, he couldn't
help applying it to P as well. The neural state P would be eliminated as a
natural kind for exactly the same reason for which M has been eliminated as a
natural kind. His argument according to which either both the reduced property
and the reduction base are natural kinds or neither of them is, naturally
drives him toward the eliminativist position according to which only
fundamental physics individuates objective natural kinds. Kim succeeds in
avoiding eliminativism only by drawing the opposite conclusion in two arguments
whose logical structure is strictly the same. Pain in general is not a respectable natural kind because it is multiply realizable on the
neural level, but human pain is a
respectable natural kind although it
is multiply realizable on the chemical and physical level. Kim's conclusion that
general psychological properties are eliminated by local reduction to
species-specific psychological properties, where these latter conserve their
status as natural kinds, is based on an ad
hoc distinction.
5. Conclusion
The aim of this paper has been to remove some important obstacles
standing in the way of considering mental (and other higher-level) properties
as constitutive of genuine natural kinds. I argued for the thesis that the
causal efficacy of macroscopic properties (which goes hand in hand with their
being natural kinds and their being embedded in laws of nature) is compatible
with, though logically independent of, their reducibility to underlying
properties. In order to show this, I analyzed an argument of Kim's (1992)
intending to show that multiply realizable macroproperties are not genuine
natural kinds - and thus possess only derivative "epiphenomenal"
causal efficacy. My strategy for refuting this claim was to show that
temperature is in relevant respects similar to more controversial
macroproperties. In particular, it is multiply realizable and it is neither
identical nor coextensive with any one of the properties in its supervenience
base, e.g. mean kinetic energy. Nevertheless temperature is reducible if anything
is.
Against antireductionism, this shows that - by parity of reasoning - the
fact that a macroproperty is multiply realizable and neither identical nor even
coextensive with any of the properties in its supervenience base is not an
obstacle to its reduction. But against Kim, the example of temperature shows
that its being in that sense a typical macroproperty isn't an obstacle to its
being a genuine natural kind either, and therefore to its being genuinely
causally efficacious. A property's being involved in laws of nature is a better
criterion for both naturalness and causal efficacy. The fact that temperature
is part of the Boyle-Mariotte law suffices to establish its being a natural
property and causally efficacious.
There is an important objection against the legitimacy of the analogy
between temperature and mental states which is crucial to my argument. It might
be said that this analogy overlooks the distinction between multiply realizable
properties whose supervenience bases are respectively homogeneous and
heterogeneous. This is not a clear-cut distinction, but if the objection is to
go through, it should at least be plausible that a mental state like pain comes
closer with respect to the homogeneity of its supervenience base to a property like
jade which is not a natural kind than
to a property like temperature which is one. I tried to show, however, that
neither the homogeneity or heterogeneity of the supervenience base of a
property P nor the confirmability of general statements containing a predicate
denoting that property can serve as a criterion for deciding whether P is a
natural kind or not. My argument was based on a comparison of jade with
temperature with respect to these criteria. Contrary to what the objection
says, these properties turn out to be similar with respect to the heterogeneity
of their respective supervenience bases, and this similarity is also brought
out by applying the criterion of confirmability. Therefore, these are not
satisfactory criteria for establishing that temperature is a natural kind
property whereas jade is not. But if these criteria fail for jade and
temperature, we cannot expect them to tell us whether pain is a natural kind property or not.
I ended by pointing out an additional problem for Kim's position. Let us
assume that there are, or will be, species-specific reductions of mental
properties to properties of the brain. As Kim recognizes, we can also assume
that the neurophysiological reducing property will in general itself be
instantiated in different ways, depending on the individual and the moment in
his lifetime. If one accepts his thesis that a property whose reduction base is
heterogeneous can't be a natural kind, and if this criterion is sufficient for
deciding about naturalness, then one isn't justified either in considering the
neurophysiological property as a natural kind in its turn. Kim's line of
reasoning leads to eliminativism, recognizing only the properties of
fundamental physics as genuinely natural and causally efficacious[23].
References
Blackburn, S. (1993) Losing
Your Mind: Physics, Identity, and Folk Burglar Prevention, in: Essays
in Quasi-Realism (Oxford, Oxford University Press).
Block N. (ed.) (1980), Readings in the Philosophy of Psychology,
Vol. 1 (Cambridge, MA, Harvard
University Press).
Block, N. (1990) Can the
Mind Change the World?, in: G. Boolos (ed.), Meaning and Method, Essays in
Honour of Hilary Putnam (Cambridge, Cambridge University Press).
Block, N. (1997)
Anti-Reductionism Slaps Back, in: James E. Tomberlin (ed.), Philosophical Perspectives,
11: Mind, Causation, and World, pp. 107-132.
Churchland, P.M. (1988)
Matter and Consciousness: A Contemporary
Introduction to the Philosophy
of Mind. Revised Edition (Cambridge, MA, MIT Press).
Churchland, P.S.
(1982) Review of: Austen Clark, Psychological Models and Neural Mechanisms, Journal of Philosophy, 79, pp. 98-111.
Churchland, P.S.
(1986) Neurophilosophy (Cambridge,
MA, MIT Press).
Crane, T. (1992) Mental
Causation and Mental Reality. Proceedings
of the Aristotelian Society, 92, pp.
185-202.
Hellman, G.P. and Thompson, F.W. (1975) Physicalism: Ontology, Determination, and Reduction. Journal of Philosophy, 72, pp. 551-564.
Jackson, F. and Pettit, P.
(1988) Functionalism and Broad Content. Mind,
97, pp. 381-400.
Jackson, F. and Pettit, P.
(1990) Causation in the Philosophy of Mind. Philosophy
and Phenomenological Research,
50, Supplement, pp. 195-214.
Kim, J. (1984) Epiphenomenal
and Supervenient Causation. Midwest
Studies in
Philosophy, IX, pp. 257-270;
repr. in: Kim (1993b), chap. 6.
Kim, J. (1988) Explanatory
Realism, Causal Realism, and Explanatory Exclusion. Midwest Studies in Philosophy,
XII, pp. 225-239.
Kim, J. (1989a) The myth of
nonreductive materialism, in: J. Kim (1993b), chap. 14.
Kim, J. (1989b) Mechanism,
Purpose and Explanatory Exclusion, in: James E. Tomberlin (ed.), Philosophical
Perspectives, 3: Philosophy of Mind and Action Theory, pp. 78-108; repr.
in: Kim (1993b), chap. 13.
Kim, J. (1992) Multiple
Realization and the Metaphysics of Reduction. Philosophy and Phenomenological
Research, 52, pp. 1-26; repr. in: Kim (1993b), chap. 16.
Kim, J. (1993a) Postscripts on
mental causation, in: Kim J. (1993b), chap. 18.
Kim, J. (1993b) Supervenience and mind (Cambridge, Cambridge
University Press).
Kim, J. (1997) The Mind-Body
Problem: Taking Stock after Forty Years, in: James E. Tomberlin (ed.),
Philosophical Perspectives, 11: Mind, Causation, and World, pp. 185-207.
Kistler, M. (forthcoming),
Causes as events and facts, Dialectica.
Lewis, D. (1969) Review of
Putnam, in: Block (1980).
Lewis, D. (1972)
Psychophysical and Theoretical Identifications, repr. in: Rosenthal D. (ed.), The Nature
of Mind (Oxford, Oxford University Press, 1991); also repr. in: Block
(1980).
Nagel, E. (1961) The Structure of Science (London,
Routledge and Kegan Paul).
Pacherie,
E. (1993) Naturaliser l'intentionnalité, (Paris,
Presses Universitaires de France).
Pereboom, D. and Kornblith, H.
(1991) The Metaphysics of Irreducibility, Philosophical
Studies, 63, pp. 125-145.
Proust, J. (1995)
Functionalism and Multirealizability. On Interaction Between Structure and Function, in: K. Gavroglu et al. (eds.),
Science, Mind and Art (Dordrecht,
Kluwer), pp. 169- 185.
Proust,
J. (1997) Comment l'esprit vient aux bętes (Paris,
Gallimard).
Putnam, H. (1967) The Nature
of Mental States, in: Rosenthal D. (ed.), Materialism
and the Mind-Body-Problem
(Englewood Cliffs, NJ, Prentice Hall, 1971); also in: Philosophical Papers, Vol. 2
(Cambridge, Cambridge University Press, 1975).
Note
on contributor
Max Kistler is Lecturer in the Department of
Philosophy, Université Blaise Pascal of Clermont-Ferrand, France. He earned his
Ph.D. in philosophy from the Ecole des Hautes Etudes en Sciences Sociales
(E.H.E.S.S.) in Paris, with a thesis on "Causality, law, and
representation" in 1995.
Correspondence: Département de
philosophie, Université Blaise Pascal, 29 Boulevard Gergovia, 63037
Clermont-Ferrand, France
[1]I
shall use "MR" to refer the phenomenon
of multiple realizability, not - as for example Kim (1992) does - to the thesis that mental states have the
property of being multiply realizable.
[2]It
should be mentioned that the opposition between a mental state and its realizer
is not inevitable for functionalism as such. David Lewis' (1972), e.g.,
formulates functionalism as a version of the identity theory which identifies
brain states and mental states rather than opposing them. The opposition
between a mental state and its realizer exists only for those who take it that
MR forbids their identification insofar as the different realizers are not
identical to each other.
[3]Kim's
way in (1984a) of analyzing mental causation as "supervenient
causation" or "epiphenomenal causation" is misleading in this
respect : while attributing causal efficacy exclusively to the physical
realizers of mental states, and thus denying independent causal efficacy to the
mental states themselves, he continues to call
them causes and effects. More recently, Kim seems to have admitted this fact :
he now wonders whether "'supervenient causation' [...] might [...] not be
'causation' in name only" (Kim 1993a, p. 359).
[4]Since
1992, Kim has modified his judgment about the scientific character of
properties which are, like mental properties, multiply realizable. In (Kim
1997), he argues for the compatibility of functionalism with the reduction of
the mental to the physical, and even for the thesis "that the
functionalist conception of mental properties is required for mind-body reduction" (Kim 1997, p. 203). But this
modification doesn't touch the crucial point of the present paper which
concerns the reality of the multiply realized property. In the sense intended
here, a property is "real", or is constitutive of a natural kind, if
it is capable of being causally efficacious. According to nomological
conceptions of causation, this capacity in turn depends on the property's being
embedded in laws of nature. But Kim still holds - in line with the position
expressed in (Kim 1992) - that if a (mental or other) property M is reducible
with a multiple reduction base, then "there is no need [...] to think of M
itself as a property in its own right" (Kim 1997, p. 201). The reason is
that the scientific respectability of properties which are multiply reducible
goes hand in hand with their functional definition as "second
order-properties" (ibid., p. 194). But Kim makes it clear that such talk
of "second-order properties" is misleading, for contrary to what it
suggests, "by existential quantification over a given set of properties,
we do not literally bring into being a new set of properties" (ibid., p.
200). What remains in the place of properties which have been reduced (and
thus, as Kim might still say, "eliminated as independent entities"),
are "second-order descriptions or
designators of properties, or
second-order concepts, [rather] than
second-order properties" (ibid., p. 201).
[5]The
former is the case if multiple realization is an empirical fact : the reduction
base is the disjunction of the actually existing realizers. The latter is the
case if MR is conceptual : there is an infinite number of possible realizers
for any given mental state, thus the potential reduction base is an infinite
disjunction.
[6]I
postpone the examination of this argument until section 3 below. In this
section I shall rather concentrate on the more general question whether
reduction between properties implies (or presupposes) their identification.
[7]I
shall come back to this conclusion later, in section 4.
[8]In the
present context, we can leave aside the question which of the remaining
interpretations of the character of the nature of the linking postulates is
preferable. As Nagel (1961, pp. 356f.) points out, they are not, appearences to
the contrary nonwithstanding, incompatible alternatives. It may depend on the
context of exposition of the theoretical reduction as a whole whether the link
between P1 and P2 appears as a coordinating definition or
as a factual hypothesis.
[9]I
follow a common contemporary usage - which is also Kim's own - in talking not
only of a predicate's extension which
is the class of objects satisfying that predicate, but also of a property's extension which is the class
of objects possessing (or instantiating) that property.
[10]If
this is true, the reduction can be accomplished neither by nomic equivalence
nor by a law - at least on accounts which consider laws to imply necessary, or
at least universal coinstantiation of the lawfully linked properties.
[11]"Macroscopic"
is a term whose conditions of application are context-dependent. The inferior
limit below which temperature and other thermodynamic concepts can't be applied
is theoretically fixed by an object's having at least three constituent parts capable of independent movement. Objects
with only two parts and "atomic" objects without any parts at all
can't be objects of thermodynamics. In particular, it is impossible to ascribe
temperature to them.
[12]It is
interesting to note that Kim himself notes in passing that Nagel's model of
intertheoretic reduction "does not in general require hat each T2-term
be correlated with a coextensive T1-term"
(Kim 1992, p. 9; his emphasis). However, as I argue in the text, his subsequent
argument is vitiated by his failure of taking this important fact about
reduction seriously. If he did, as I think one should, the resulting picture of
the relation between reduced and reducing properties would turn out very
different.
[13]This
has been noted by a number of authors. Cf. P.S. Churchland (1982, pp. 101f.;
1986, pp. 356ff.), P.M. Churchland (1988, pp. 41f.), Pereboom and Kornblith
(1991, p. 138), Crane (1992, p. 195), Pacherie (1993, pp. 31f.).
[14]An
object is a black body if it absorbs all incident radiation of all wavelenghts,
without reflecting any of it.
[15]Given
what I have said about the multiple realizability of temperature, one might
rather prefer to say that is has not. But that would amount to interpreting the
meaning of the word "reduction" as exclusively defined by its role in
the Nagelian theory of intertheoretical reduction. But it is one of the aims of
this paper to show that in some cases at least, a property may be reduced
without there being a unique bridge law in Nagel's sense. In a way it comes to
a terminological decision to keep the name "reduction" for the
relation between a macroproperty like temperature and its microphysical basis
even when one questions the universal applicability of Nagel's theory to such
relations. This decision seems to me to be justified by the fact that the
application of the word "reduction" to whatever relation exists
between temperature and its microphysical basis, as well as between other
paradigmatical cases of reduction, seems to be at least partially constitutive
of the meaning of that word.
[16]On the
link between the causal efficacy of a property and the laws of nature in which
it is embedded, cf. Kistler (forthcoming).
[17]Cf.
section 3 below.
[18]It is
precisely upon examination of the issue of the heterogeneity which the potential
reducers of a given mental property M can be expected to have across the large
variety of (natural or artificial) cognitive systems able to possess it, that
Block (1997) sheds doubt on the justification of Kim's radical conclusion.
Indeed, as Block argues, there are "forces of design or selection"
(Block 1997, p. 122) putting heavy constraints on which physical structure
could possibly underlie a given mental property. (On this point, see also
Proust 1995; 1997). These contraints impose similarities on those underlying
properties. If Block is right, both the functionalists' argument from MR to
irreducibility and Kim's argument from MR to elimination as a natural kind are weakened insofar as they both rely on
the assumption of (strong) heterogeneity.
[19]Blackburn
(1993) uses the fact that the different realizations of temperature share the
capacity to bring bodies possessing them in a common state of thermodynamic
equilibrium to argue for temperature's being a perfectly respectable natural
kind property and against P.S. Churchland's (1986) contention that its multiple
realizability eliminates temperature as a candidate for being a natural kind
property sanctioned by science. I would rather argue that temperature is a
natural kind not because all of its realizations sustain thermodynamic
equilibrium, but because there are laws linking it - not the lower-level properties of its particular instances,
like mean kinetic energy - to other natural kind properties. The 0th law of
thermodynamics (It says that if A and B are in thermal equilibrium, and B and C
are as well, then so are A and C.) is such a law, and equilibrium is such a
property, but it belongs to the same (macroscopic) ontological level as
temperature itself.
[20]I do
not claim to have a criterion of lawhood, enabling me to justify that I treat
general statements involving "jadeite", "nephrite" and
"temperature" as hypothetical laws, but not general statements
bearing on "jade". I simply rely on the generally shared conviction
that the first three terms designate natural kinds of mineralogy or
thermodynamics, whereas the latter designates a kind of gemology which is not a
science and whose kinds are not natural
kinds.
[21]I
abstract here from the further difference between moving molecules/atoms of gases and vibrating atoms/molecules of solids. For the purposes of the
present argument, vibration can be considered as a kind of motion.
[22]In
(Kim, 1992, pp. 5 and 7/8), Kim repeats the observation that MR might
"occur 'locally' as well." He goes on to note that "we may want
to avail ourselves of the flexibility of allowing a psychological state, or
function, to be instantiated by alternative mechanisms within a single system.
This means that Pi can be a disjunction of physical properties" (Kim,
1992, p. 7). He does not, however, take into account the implications of this
possibility that I point out in the text.
[23]Parts
of an earlier version of this paper have been read at the 5th Annual Meeting of
the European Society of Philosophy and Psychology, Barcelona, July 1996. I
would like to thank my auditors in Barcelona and Ned Block, Joan Cullen, Marcel
Lieberman, Joëlle Proust and an anonymous referee for this journal for their
helpful criticisms.