Some Problems for Reductive Physicalism
Pierre Jacob
Institute of Cognitive Science,
CNRS, France.
Jaegwon Kim's book, Mind in a
Physical World, is a dense, sustained and powerful critique of nonreductive
physicalism. Nonreductive physicalists are monist physicalists. Hence, they
reject Cartesian substance dualism. But since they reject the identity between
mental and physical properties, they are property dualists (or pluralists). The
nonreductive physicalists' key word here is 'autonomy': according to all
versions of nonreductive physicalism, the mental is autonomous. This claim,
however, is not unambiguous: it may mean either that mental properties make a
causal difference or that they are not nomic.
Davidson's anomalous monism, for
example, is a version of nonreductive physicalism. It does not purport to meet
the demands of mental realism. It relies instead on mental anomalism, i.e., the
view that there can be neither strict psychological nor strict psychophysical
laws: mental predicates cannot appear in genuine nomic correlations. Hence, the
autonomy of the mental consists in its being anomalous. Davidson's main reason
for embracing mental anomalism is the normativity of the mental: unlike the
application of physical predicates, the application of mental predicates is
governed by normative constraints such as rationality. On this view, the
autonomy of the mental precludes a science of the mental or at the very least
it points towards the limits of a scientific understanding of mental phenomena.
According to orthodox functionalism,
mental properties are higher-order physical properties which can be multiply
realized by lower-order physical properties. A higher-order functional property
in general is defined as the possession of a first-order physical property
meeting a set of causal conditions. If a mental property is
"functionalizable", then its instantiation can both be the effect of
either some sensory input or the instance of some other mental property and the
cause of either the instance of some other mental property or of some
behavioral output. According to Putnam's celebrated multiple realization
argument, since a higher-order physical property can be realized by a
disjunction of distinct lower-order physical properties on different occasions,
it is not identical with any lower-order physical property which happens to
realize it (or on which it supervenes) on a given occasion. So functionalism
too is a version of nonreductive physicalism. Since orthodox functionalists are
mental realists, they are committed to the causal efficacy of mental
properties. Thus functionalism secures the autonomy of mental properties: they
can be both functional and nomic or projectible because psychological causal
generalizations involve mental kinds, not their disjunctive physical realizers.
Hence, the autonomy of the mental is the autonomy of scientific psychological
explanations.
No doubt at the beginning of the
book Kim accepts the motivations of mental realism (e.g., p. 31). This is why
he rejects Davidson's anomalous monism and why on his view, it is a condition
of adequacy on any version of physicalism that mental properties must turn out
to be causally efficacious. Kim's goal in the book is to show that the orthodox
functionalists cannot have it both ways: mental properties cannot both be
irreducible (or autonomous) and causally efficacious. Importantly, Kim's
argument is directed against anti-reductionism, not against the
functionalist conception of mental properties.
Unlike orthodox functionalists, Kim
does not think that if a property can be functionalized, then anti-reductionism
is vindicated. On the contrary, he thinks that functionalism paves the way for
the identification of the functional property with the physical property which
realizes it on a given occasion. He thinks so because if functionalism is true
of e.g., property F*, then there exists a pattern of causal/nomic
relations whereby instances of F* are produced when and only when some
other property F is instantiated and which in turn causes instances of
property F**. The reason why functionalism entails reductionism is that
once such a pattern of causal/nomic relations has been established, it becomes
possible to find underlying physical mechanisms which occupy the various causal
roles.
If so, then the functional property
enjoys the very causal powers of the physical property with which it is
identical and mental realism is saved by reductionism. The principle which
encapsulates Kim's reductionist interpretation of functionalism is his "causal
inheritance principle" which denies the explanatory autonomy of mental
properties (p. 54) and according to which a second-order functional property
can have no more causal powers on a given occasion than its physical realizer
on this occasion. If a mental property is not functionalizable, then (on pain
of dualism) the choice is between eliminativism and epiphenomenalism. Kim calls
"emergentist" the anti-reductionist's claim that a given property is
intrinsic and not open to a functional construal. Kim's own suspicion is that,
unlike intentional (or content) properties, phenomenal properties (or qualia)
are indeed intrinsic properties which will resist functionalization and
therefore reductive physicalism.
Kim's argument against nonreductive
physicalism proceeds in two complementary stages. His first and main argument
is "the supervenience argument". I will examine it in some details.
Then, I'll make some brief comments on the second argument which can be called
"the multiple realization argument revisited". And finally, I'll
raise a question about Kim's fundamental attitude with respect to mental
realism.
The purpose of the supervenience
argument is to show that nonreductive physicalism is doomed by the problem of
causal exclusion. The argument has the general form of a dilemma: either
mind-body supervenience holds or it does not.[1] Kim agrees with nonreductive physicalists that
physicalism would not survive the failure of supervenience. I agree with him.
So let's assume that mind-body supervenience holds.
Consider a case of mental-to-mental
causation: suppose that the instantiation of mental property M causes
the instantiation of another mental property M*. Given the assumption of
supervenience, M* supervenes on some physical property P* which
realizes it. And now Kim asks: what prompted the instantiation of M*?
Given the contribution of P*, what was the contribution of M to
the instantiation of M*? How can the instance of M and the
instance of P* collaborate in the process whereby M* was
exemplified? The only sensible answer open to the nonreductive physicalist
seems to be: M caused the instantiation of M* by causing the
instantiation of P*. Hence, the nonreductive physicalist is committed to
assuming that cases of mental-to-mental causation presuppose what Kim called
"downward causation", i.e., the possibility of mental-to-physical
causation.[2]
I agree with Kim that answering the
question raised in the above paragraph is a challenging task for the
nonreductive physicalist. However, there is, I think, a plausible answer
available to the nonreductive physicalist — an answer which derives its inspiration chiefly from Kim's own critique
of Searle's "biological naturalism" (pp. 47-50). The nonreductive
physicalist ought to deny that the relation between the instance of P*
and the instance of M* is causal. As Kim puts it (p. 44), "in
general, the relation between base properties and supervenient properties is
not happily construed as causal. For one thing, the instantiations of the
related properties are simultaneous, whereas causes are standardly thought to
precede their effects; second, it is difficult, perhaps incoherent, to imagine
a causal chain with intermediate links, between the subvenient and the
supervenient properties". Assuming that M* strongly supervenes on P*,
not only is the instantiation of the former simultaneous with the instantiation
of the latter, but anything which exemplifies the latter at t must
necessarily instantiate the former at t. There is no possible world in
which the latter is exemplified at t and the former is not. Kim, I
think, was wrong when he wrote in an ancestor paper to chapter 2 of this book
that "the situation is essentially identical to the situation we face when
we are given two distinct independent causes for one and the same event, each
claimed to be a sufficient cause".[3] It follows that the nonreductive physicalist must
presumably recognize that the alleged causal relation between M and M*
must be factorized into two sub-relations only one of which is causal: the
instance of M causes P* which in turn realizes M*. So the
causal relation between M and P* does not compete with the causal
relation between P* and M* for there is no causal relation
between P* and M*.
This is not to say that the
nonreductive physicalist is out of the wood. For we now reach the second step
of the argument which involves the principle of the causal closure of the
physical: any instance of a physical property must have a physical cause. Given
the assumption of mind-body supervenience, mental property M itself
supervenes on some physical property P which realizes it. So the
question is: given that an instance of P causes the instantiation of P*,
what causal work is left for M? Unless he is willing to violate the
principle of the causal closure of the physical, the nonreductive physicalist
seems bound to grant that property M is epiphenomenal in the process
whereby P* gets instantiated. The conclusion of the supervenience
argument is that if physicalists want to save the causal efficacy of mental
properties, they should opt for reductionism.[4]
I learnt, I think, two important
things from pondering over Kim's supervenience argument and his reactions (in
chapter 3) to some of the responses offered on behalf of nonreductive
physicalism.
First of all, I agree with Kim that
the problem of causal exclusion is about causation, not about explanation. The
issue is: how do different kinds of properties relate to one another? Any
attempt to overcome this problem by distinguishing different kinds of
explanation will fail to engage with the metaphysical issue generated by the
metaphysical assumption that mental properties are irreducible.[5] This, I think, applies across the board: Jackson and
Pettit's distinction between process and program explanations is explicitly designed
to vindicate the "relevance", not the causal efficacy, of
higher-order physical properties — and
relevance is an epistemic notion. To argue (as
Block and Antony and Levine have) for the causal powers of a higher-order
chemical property such as dormitivity from its involvement in a generalization
such as "dormitive substances provoke traffic accidents" is, I think,
to switch from causation to explanation. Of course, to learn that e.g.,
antihistamines and tricylics have something in common is to learn something
useful. However, if it causes a particular traffic accident, the ingestion of a
substance could only do so in virtue of its chemical property.[6]
Second of all, I think that Kim's
response to "the generalization argument" is quite effective. If correct,
this argument would undermine "the spirit of causal exclusion
reasoning".[7] It relies on the following conditional: if mental
properties are deprived of causal efficacy by their physical realizers, then
all macrolevel special science (chemical, geological or biological) properties
should be similarly deprived of causal efficacy by fundamental microlevel
physical properties. Since nobody in his right mind would accept the
consequent, the antecedent can be safely rejected. As Kim correctly points out,
the conditional confuses the distinction between levels (of reality) and the
distinction between orders (of properties). Macrolevel special scientific
properties belong to entities different from the entities which exemplify
fundamental microlevel physical properties. This is why the problem of causal
exclusion does not apply: the causal efficacy of the former does not compete
with the causal efficacy of the latter. However, a second-order functional
property is exemplified by the very same sort of thing which possesses its
lower-order physical realizer on a given occasion. This is why the problem of
causal exclusion does apply and the causal efficacy of the former is threatened
by the causal efficacy of the latter.
I now turn to Kim's multiple
realization argument revisited. As I said, Kim disagrees with Putnam's
celebrated argument that, if a higher-order functional property is multiply
realizable by a disjunction of different lower-order physical properties on
different occasions, then the former cannot be identical with any of the
physical disjuncts and the scientific autonomy of the former is secured. In an
earlier paper,[8] Kim argued that if the disjunction of lower-order
physical realizers of the higher-order functional property is heterogeneous,
then the following dilemma arises: either both the functional property and the
heterogeneous disjunction are nomic kinds or neither is. He recommended the
latter option on the basis of the following reasoning. Samples of what was once
held to be a single mineral kind, jade, turns out to involve two distinct
minerals with different molecular structures: jadeite and nephrite. Hence, jade
turns out to be a disjunction of two heterogeneous mineral kinds. Hence,
"jade is green" does not express a nomic generalization; rather it
expresses a conditional with a disjunctive antecedent. Being a disjunction of
two heterogeneous mineral kinds, jade is not a projectible property. Not being
projectible, it is not a nomic property. In other words, to say that "if
something is either jadeite or nephrite, it is green" is to accept the
following conjunction: "if something is jadeite, then it's green and if
something is nephrite, then it's green". If all our evidence happened to
confirm only one of the conjuncts, we should not extend our confidence to the
generalization involving the disjunctive antecedent. If so, then psychology
cannot be an autonomous science.[9]
I do think that in combination with
the supervenience argument, Kim's multiple realization argument makes an
impressive case against the autonomy of mental properties. The hard question we
are left with is: does Kim's reductionism really vindicate mental realism? As I
mentioned above, if a mental (e.g., phenomenal) property is not
functionalizable, then the choice faced by a physicalist is: epiphenomenalism
or eliminativism? But what about functional properties? In one section of
chapter 4 (pp. 103-106), Kim entertains the radical view that one ought to
trade higher-order functional properties for higher-order functional concepts.
This proposal, I think, amounts to giving up reductionism and embracing
eliminativism. It strikes me as odd in at least two respects.
First of all, I don't quite I
understand the view that there could be functional concepts and no functional
properties (p. 104). A second-order property is supposed to be the property of
possessing some lower-order physical property meeting some causal conditions. I
fail to see what the relevant notion of a second-order concept might be other
than the concept of a second-order physical property. However, if there are no
second-order physical properties, then what could a second-order concept be?
What would it be a concept of? Presumably, the answer is: it is the
concept of an uninstantiated property. I wonder whether in retrospect Kim does
not himself acknowledge the oddity of his own suggestion when he writes (p.
111): "I also urged that M be viewed as a concept, not a property
in the world; however this is to duck the issue — we need a more straightforward answer".
Finally, it would be ironical if by
the end of the book Kim ended up agreeing with Davidson's anomalous monist view
that we should replace talk of mental properties by talk of mental conceps (or
predicates). Of course, unlike Davidson, Kim is not urging that we replace talk
of physical properties by talk of physical concepts (or predicates). But giving
up mental properties suffices to remove the question: if some physical property
of a cause was efficacious in the process whereby it produced its effect, then
could its mental property be efficacious too? If so, then the problem of causal
exclusion simply does not arise and Kim's supervenience argument is undermined.
[1] As Kim's definition makes clear, the relevant notion is strong supervenience since it involves the claim that "necessarily anything with (subvenient property) P at a time has (supervenient property) M at that time" (p. 39).
[2] "The Nonreductivist's Troubles with Mental Causation", in Kim, Supervenience and Mind, Cambridge: Cambridge University Press, 1993.
[3] Ibid., footnote 30, p. 351.
[4] My earlier comment on the first stage of the argument corroborates, I think, Kim's conclusion: the relation between the instance of M* and the instance of P* is not causal and necessarily if P* is instantiated then so is M*. So it follows that the nomic connection between instances of P* and instances of M* is not causal. Rather it must be some kind of identity such as "water = H2O".
[5] I confess that I earlier missed the point.
[6] Cf. N. Block, "Anti-Reductionism Slaps Back", in J.E. Tomberlin (ed.) Philosophical Perspectives, 11, p. 174, Oxford: Blackwell, 1997, and L. Antony & J. Levine, "Reduction with Autonomy", ibid.
[7] Cf. T. Horgan, "Kim on Mental Causation and Causal Exclusion", in J.E. Tomberlin (ed.) op. cit.
[8] J. Kim, "Multiple Realization and the Metaphysics of Reduction", in Kim, Supervenience and Mind, op. cit. A version of this argument appears in the book in the section "Multiple Realization Again", pp. 106-12.
[9] I fail to see the force of Fodor's distinction between open and closed disjunctions in response to Kim's argument in his "Special Sciences: Still Autonomous After All These Years", in J.E. Tomberlin (ed.) op.cit.