Kim's Argument

articles

Jaegwon Kim’s Supervenience Argument

Introduction

Jaegwon Kim in Physicalism or Something Near Enough argues for a physicalist approach to the mind-body problem and concludes that mental causation and what we most value about consciousness, our agency and cognitive ability, are consistent with a physicalist world view. This leaves the qualitative states of consciousness to form ‘[a] “mental residue” that cannot be accommodated within the physical domain’ (Kim, 2008: 170). In Kim’s account, the qualia, which constitute this residue, are not ‘functionalizable’, are not functionally reducible to the physical. Kim’s argument is well constructed and very clearly expressed: there is much to admire in it, even if I disagree with this part of his conclusion.

Since Gilbert Ryle (1973 [1949]) persuasively argued that the two-worlds view of mind and body prevalent since Descartes arose from a family of category mistakes, one might have thought that dualism would not play a part in contemporary philosophy of mind. Although Cartesian substance dualism has been abandoned, its essence remains, transmuted into an initially plausible and attractive property dualism as well as a more subtle and insidious dualism based on the distinction between the phenomenal and the intentional.

Property dualism is founded on the idea that if mental properties are not reducible to physical properties then the distinctiveness of the mental over the physical is not an illusion and our everyday experience of mental causation might not be mistaken after all. Kim reveals the serious difficulties in this viewpoint, concluding

Causal efficacy of mental properties is inconsistent with the joint acceptance of the following four claims: (i) physical causal closure, (ii) causal exclusion, (iii) mind-body supervenience and (iv) mental/physical property dualism (2008: 21-22).

Kim says that physical causal closure and mind-body supervenience should be shared commitments of all physicalists, while exclusion principles are general metaphysical constraints without credible challenge, which leaves property dualism as the only negotiable claim. But if property dualism is questionable, then a reductionist position must be taken seriously, which “will cause a chill in those physicalists who want […] both the irreducibility and causal efficacy of the mental” (ibid: 22).

This article summarises Kim’s line of reasoning, which he calls the “supervenience argument”, and also covers his response to three objections: the possibility of over-determination, the “generalisation argument” and the spectre of causal drainage. The supervenience argument is only one part of Kim’s wider case for physicalism: these other questions are not dealt with here, nor is the concession in his final conclusion that there a mental residue insusceptible to physical reduction or explanation.

The Supervenience Argument

This is grounded in a non-reductive physicalism, which has the following tenets.

Supervenience: mental properties strongly supervene on biological / physical properties. This is an ontological thesis of dependence: a mental property is instantiated because of or in virtue of the instantiation of a physical base property.
Irreducibility: mental properties are not reducible to and not identical with physical properties.
Causal efficacy: mental properties have this, that is their instantiations can and do cause the instantiation of other mental and physical properties.

Edwards’ dictum. There is a tension between “vertical” determination and “horizontal” causation. In fact, vertical determination excludes horizontal causation. So we have two competing determinants of properties: one across levels (vertical) and another, causal one from past to future (horizontal). The vertical determination occurs irrespective of what happened in the past. This view was based on Edwards’ belief that God was a sustaining cause at every instant. Kim translates this idea to mind-body supervenience, where vertical determination is the hierarchical dependence of supervenience and horizontal causation is mental causation or any other higher-level causation.

To clarify the supervenience argument, restate it as follows.

Stage I.

Let M and M* be mental properties. M causes M* (strictly, an instance of M causes M* to instantiate).
For some physical property P*, M* has P* as its supervenience base.

A tension arises from (1) and (2). Has an instance of M* been created because M caused it to be or because P*, a supervenience base of M*, came into being? As P* is present on this occasion, so must M* be present, no matter whether M had previously existed or had a causal role. Unless of course M had brought about the creation of P*. This suggests a way of relieving the tension:

M caused M* by causing its supervenience base, P*.

This result shows that if supervenience is accepted, then mental-mental causation entails mental-physical causation, or more generally same-level causation entails downward causation. “In short, level-bound causal autonomy is inconsistent with supervenience or dependence between the levels” (Kim, 2008: 40).

Stage II

There are two ways of completing the argument.

Completion 1

M has a physical supervenience base, P.

There are reasons for thinking P is a cause of P* (see Closure below). Note also that P (by supervenience) is sufficient for M and, since ex hypothesi M is a cause of P*, P qualifies as a cause too.

M causes P* and P causes P*.
From irreducibility, M ≠ P.

Again, a tension arises from (5) and (6). The causal exclusion principle seems applicable here.
Exclusion. No single event can have more than one sufficient cause occurring at any given time – unless it is a genuine case of causal overdetermination.

(Assumption) P* is not causally overdetermined by M and P.

By Exclusion, either M or P must be eliminated as a cause of P*. To decide, appeal to the causal closure of the physical domain.
Closure. If a physical event has a cause at t, it has a physical cause that occurs at t.

The putative mental cause, M, is excluded by the physical cause, P. That is, P, not M, is a cause of P*.

This result can be displayed graphically as

      M              M*

      ↑              ↑

 supervenes     supervenes

      ↑              ↑

      P →  causes →  P*

Figure 1.

Completion 2

M is a cause of P* (from (3) in Stage I).
By Closure, P* has a physical cause – call it P – occurring at the time that M occurs.
From irreducibility, M ≠ P.
Hence P* has two distinct causes, M and P, and this is not a case of causal over-determination.
By Exclusion, either M or P must go.
By Closure, P stays, so from (8) M goes.

This is simpler than Completion 1, in that supervenience is not required as a premise, so there is no claim that P is a cause of P* on account of P being the supervenience base of M. However, Completion 1 is more intuitive. Whichever completion is used, Stage 2 demonstrates that the principles of causal exclusion and physical closure disallow “vertical” causation.

Objection I: is P* over-determined?

As a challenge to Stage II, is the assumption that P* is not over-determined justified? If P and M are to be each a sufficient cause of P*, we need to consider a possible situation where M occurs without P and ask if in this situation P* would follow. However, if this is not in fact possible, because the following counterfactual is true: “if P had not occurred, M would not have occurred”, then what can be made of the claim that M as well as P is a sufficient cause of P*, so that M and P together over-determine P*? Consider a world W in which M occurs without P: Supervenience does not make such a world impossible, but it does require that M has an alternative physical base; let this base be P´. In this case, the earlier argument applies: M supervenes on P´, M causes P*, so P´ is also causally sufficient for P*. Thus P* has a physical cause in W and Closure applies there. As Kim says:

As long as Supervenience is held constant, there is no world in which M by itself, independently of a physical base, brings about P*; wherever M claims to be a cause of P*, there is some physical property waiting to claim at least an equal causal status (ibid: 47).

So in the P-M world. it would be incoherent to assert that there is a causal chain from M to P*, independent of the one from P to P*; indeed, the former, if it exists, must incorporate the latter. In exemplary cases of over-determination, such as two bullets entering the heart at the same time, or a short-circuit and overturned oil lamp both causing a fire, each cause is sufficient and distinct. In the Supervenience case, such distinct causes are not evident.

A simpler way of ruling out physical over-determination is to adopt a stronger version of physical causal closure.

Strong closure. Any cause of a physical event is itself a physical event – that is, no nonphysical event can be a cause of a physical event (ibid: 50).

Revisiting completion 1 with Strong closure gives, as before:

M causes M* by causing P*.
M has a physical supervenience base, P.
M causes P* and P causes P*.

From here, the argument is as follows:

For every physical property P, M ≠ P Irreducibility.
M does not cause P* (from (6) and Strong closure).
M does not cause M* (from (3) and (7) - (3) implies causing P* is the only way M can cause M*).
P causes P* (from (5)).

The result is as for the original Completion 1, as illustrated by Figure 1. By replacing Closure and Exclusion with Strong closure, the argument is simpler and avoids any difficulty some might have with the principle of causal exclusion. Most physicalists would see no objection to Strong closure as a principle. On the other hand, by adopting Strong closure, one starts the argument with mind-body causation already ruled out. Better, philosophically, to base the conclusion of an argument on the weakest possible premises rather than fewer but stronger premises. The latter approach is apt to beg the question and is vulnerable to a more concentrated attack.

Objection II: the generalisation argument

Block, in discussing Kim’s supervenience argument, asserts

First, it is hard to believe that there is no mental causation, no physiological causation, no molecular causation, no atomic causation but only bottom level physical causation (Kim, 2008: 52).

Kim infers from this that Block subscribes to the generalisation argument, in which the supervenience argument generalises beyond mind-body causation, so that causation at any level gives way to the next lower level. Other writers endorse this argument, saying that if mental causation is a problem then there is a parallel problem for the other special sciences, leaving genuine causation to be found only at the deepest physical level. Such an argument presupposes a layered model of the domains of science, where the layers are arranged in a hierarchy of levels, with the domain of physics at the bottom, the domain of chemistry above it , the domain of biology above the chemical domain and, finally, the domain of psychology above the biological domain. This hierarchical model is not new: Steven Rose (1978: 28-33) described it as a operational construct, as one useful to handling problems and providing explanations at different levels. Rose, a biochemist and neuroscientist, is careful to avoid bold ontological statements and is well aware of the pitfalls around multi-level causation: he displays a caution many metaphysicians would do well to adopt.

The generalisation argument presupposes something else: an irreducibility between layers, at least at the higher levels. Block construes Kim’s supervenience argument as a reductio against mental causation, but Kim says his target is more against the idea of irreducibility. Kim says one must choose between causal impotence or reducibility. If then, one can reduce (find reductive identities) from a level to the one below, causal impotence disappears: there is mental causation, biological causation and so on since they are reductively identical with physical causation. This position is as illustrated below:

      M →  causes →  M*

      ↑              ↑

 reductively identical with

      ↑              ↑

      P →  causes →  P*

Figure 2.

However, although such reductive identities would make causation possible at all levels, such a neat outcome does not establish that these identities exist. Beyond the case for reducibility made by Kim’s supervenience argument, I think that this can only be done empirically. As the discussion in the next section mentions, there is evidence to support the view that identities exist from physical to biological levels and arguably to the mental level as well. These identities may be compositional, where an entity at one level is composed of an aggregate of entities at a lower level, or representational, where an entity at the mental level is represented by entities at lower levels. If the identities form an unbroken chain from a base physical level up to some higher level then there is causation at the higher level and all intermediate levels; they are just more organised forms of the base level causation.

Objection III: the causal drainage argument

Block continues:

Second, it is hard to believe that there is no causation at all if there is no bottom level of physics (Kim, 2008: 52).

Kim counters by using the idea of micro-based properties. A micro-based property of an object is one which characterises the object’s microstructure, that is, which specifies the micro-constituents of which the object is composed and the structural relations which configure these constituents into a stable whole. Kim says such a property is one of the macro-properties of the object, belonging to the whole object, and it does not supervene on the properties of the micro-constituents. As such micro-based properties are not susceptible to the supervenience argument and this prevents causation draining downwards from macro to micro. Recognising this strategy, Block uses a multiple composition argument to prevent a macro property being identified with a micro-based property:

But why can’t micro-based properties be micro-based in alternative ways? Why isn’t jade an example of a micro-based property, micro-based in both calcium magnesium silicate (nephrite) and sodium aluminum silicate (jadeite)? (Kim, 2008: 57)

There is a parallel with multiple realisation, which is used as an argument against against reduction. Here, multiple composition is being used to argue against identifying a macro-property, being jade, with its micro-based properties. The counter-argument is as for multiple realisation: the problem vanishes if one is dealing with instances, any piece of jade will be an instance of nephrite or jadeite. As Kim says:

All we need is identity at the level of instances, not necessarily at the level of kinds and properties; causation after all is a relation between property or kind-instances, not between property or kinds as such (ibid: 58).

Alternatively, if dealing with types, one can say jade is identified with a disjunctive type, the type nephrite or the type jadeite. Kim concludes:

To disarm Block’s multiple composition argument, adopting either disjunctive property/kind identities or instance (or token) identities seems sufficient (ibid: 59).

However, Kim recognises this does not completely defeat the causal drainage argument. Micro-based properties exist at each level in the hierarchy, so although a micro-based property may not be supervenient on the micro-constituents it is supervenient on the equivalent micro-based property at a lower level. The general form of the supervenience argument implies:

Seepage. If a property Q supervenes on a property Q* at a lower level without being reducible to it, Q’s causal powers are pre-empted by those of Q* (ibid: 60).

If matter is infinitely divisible (despite, conjecturally, its smallest independent existence being in the form of discrete quanta), then there is no physical bottom level. According to Block, the supervenience argument in this case must produce the conclusion that there is no causation anywhere.

Can the spectre of causal drainage be avoided? Kim says there are alternatives to Seepage when considering the causal implications for a property Q that supervences on a property Q* at a lower level without being reducible to it. These alternatives do not seem susceptible to causal drainage:

Explanation. The causal powers of Q are explained in terms of the causal powers of Q*.
Constitution. The causal powers of Q are constituted by those of Q*.
Derivation / determination. The causal powers of Q are derived from or determined by those of Q*.

However, as mentioned previously, it is hard to conceive of an adequate explanation that does not involve reduction. The same is true for constitution. Derivation or determination need closer definition to avoid the same trap, and that may not be possible. Using Derivation, one might say that M “superveniently causes” M*, meaning M can be considered as having a causal effect derived from supervenience. Thus we would have:

      M →  superv.causes →  M*

      ↑                     ↑

 supervenes              supervenes

      ↑                     ↑

      P   →    causes  →    P*

Figure 3.

But this adds nothing and gives rise to the misconception that M has a real causal effect in addition to that of P.

Block provides a further alternative, popular amongst non-reductive physicalists. He posits non-competing causal efficacy at multiple levels. If correct, this would avoid the problem of causal drainage. But Kim says this is precisely what he is challenging. What evidence is there for non-competing causal processes or relations? If they compete, the tension mentioned earlier must be addressed. So, the problem of causal drainage must be faced; Kim has two counter arguments.

First, notwithstanding Seepage, Kim makes clear that he is not talking about causation at level L giving way to causation at level L-1 in the manner of collapsing rungs of a ladder. The supervenience argument relies on Closure, where the lower level in the argument is causally closed, Thus the mental rung does not collapse onto the biological rung, because the biological domain is not causally closed. The same is true for the levels of chemistry and macro-physics; it is only at the level of micro-physics that we have a causally closed domain. It has been observed by Bohm (Kim, 2008: 67) that even here, if there is an explanatory need to descend to a lower micro-physical level, the current level is not causally closed. On this basis, it could well be that no level in the infinite hierarchy of levels can be identified as closed. However, as Kim argues, the (infinite) union of the lowest micro-physical levels will be causally closed and this is enough to stop causal drainage.

Second, for the drainage argument to work, irreducibility must hold all the way down. Block’s multiple composition argument does not succeed in precluding reducibility. Empirical research and current scientific theories suggest that most (if not all) levels are indeed reducible. Biological phenomena can be reduced to biochemical and hence chemical ones (philosophers who still think biological facts are irreducible have perhaps not understood twentieth century discoveries and theoretical advances, particularly in cell biology and biochemistry). Chemical and macro-physical phenomena can be reduced to micro-physical ones. I contend that mental phenomena are implemented by physiological processes within biological structures (support for this view is outside the scope of this article), but even if they are not, causal drainage would be stopped by the reducibility of biological, chemical and physical levels.

Kim notes that the causal drainage argument has similarities with other well known arguments involving infinite series, such as Aristotle’s argument for the existence of a prime mover or Aquinas’ argument for the existence of God. This suggests to me yet another approach, one that is simpler and to the point. The fundamental flaw in the causal drainage argument is that is applies finite reasoning to an infinite sequence: one cannot assume that what obtains for an interval in an infinite sequence, where the interval contains a finite number of levels, will also obtain for the entire sequence. In mathematics, there are countless examples of infinite sequences where the terms converge to a finite limit. Even if matter is infinitely divisible, the differences between levels become smaller, causal pre-emption becomes a smaller move, leading to causal convergence rather than annihilation.

Conclusion

The supervenience argument rests on three assumptions: the supervenience of the mental on the physical, mental irreducibility to the physical and mental causal efficacy. The argument consists of two stages. In the first stage, mental state M causes mental state M* by causing the instantiation of the latter’s supervenience base P*. In the second stage, by physical closure, P* has a physical cause P (which may be identified as M’s supervenience base). So P* has two possible causes, M and P, and they are distinct since M is irreducible to P. Assuming P* is not over-determined, the exclusion principle dictates that only one cause can remain; closure favours P. Thus, the supervenience argument seems to indicate that one of its initial assumptions is wrong: mental states are causally impotent. Alternatively, another of the initial assumptions is incorrect. Kim identifies irreducibility as the most suspect, in which case reductive identities preserve mental causal efficacy. Looking at the objections in turn, Kim provides a strong case for P* not being over-determined, since, given supervenience, M is not independent of some base physical property P (or P´). Generalising the supervenience argument does not result in causation at only the bottom physical level, provided there are reductive identities between levels; empirical evidence suggests such identities. Kim uses micro-based properties and other arguments, including the causal closure of a union of the lowest micro-physical levels, to counter the prospect of causal drainage; my approach, which avoids extrapolating from a finite number of levels to an infinite number, relies on causal convergence.

Overall, Kim’s argument makes a case for mental causation, not as something independent of physical causation but linked to it by reductive identities. Starting with a dualistic assumption, the supervenience argument proves to be a rejection of dualism. Kim’s approach is very different to that taken by Ryle, for whom mental causation is not a counterpart to physical causation, but of a different type altogether. Strangely, I think a similar conclusion can be obtained.

Bibliography

Kim, J. (2008) Physicalism, or Something Near Enough, Princeton, Princeton University Press.
Ryle, G. (1973 [1949]) The Concept of Mind, Harmondsworth, Penguin Books.