While we have used the Shoemakeresque view to set up the preceding account of mental causation, there is reason to think that the account does not require the Shoemakeresque view to work. What it requires is just that mental and physical properties and events share a common structure while remaining distinct, and more specifically it seems to require that mental events and the physical events that realize them share token causal powers. A philosopher who accepts the Humean view according to which natural properties are related only contingently to their causal profiles, just as much as a reductive causal structuralist, could try to capture this idea of shared structure by identifying physical properties with conjunctive properties having mental properties as conjuncts. Or perhaps she can capture the idea in some other (perhaps more plausible) way. The main reason we have been using reductive causal structuralism to set up the account of mental causation provided is just that reductive causal structuralism makes it especially easy to see how this idea of shared structure might work. Reductive causal structuralism in effect renders natural properties, including both mental and physical properties, as molecular rather than atomic things. Once this structure is in place, the subset view of realization helps us see how structure might be shared without there being the sort of reductive relations that nonreductive physicalists hope to avoid.
I would like for last section’s account of mental causation to be as neutral as possible with respect to underlying metaphysical framework, that is, with respect to views about the nature of natural properties, their connection to their causal profiles, the nature of causation itself, realization, etc. To the extent that some metaphysical framework is incompatible with the preceding account of mental causation though, it seems to me that this could be viewed as a potential objection against that framework. After all, if the considerations raised in Chapter 7 are correct, then one of the central tasks for any view of natural properties is to explain how nonreductive mental and special science causation works. The present account of mental causation does just that, and it seems to do it in a more satisfying way than does any significantly different account that I know of.
In this last section, I hope to show how my own antireductionism about belief properties can be reconciled with the present account of mental causation. In doing this, I will be showing what I vowed to show back in Chapter 5, that my antireductionism about belief is compatible with a physicalistically acceptable account of mental causation.
9.3.1 INCLUSION AND OVERLAP
Both Yablo and Shoemaker take part of the key to solving the causal exclusion problem to be the idea that mental phenomena are in some sense included within physical phenomena. The Shoemakeresque view captures this inclusion idea in an especially literal-minded way, treating mental events as parts of physical events and mental properties as conjuncts of conjunctive physical properties. If inclusion is what is crucial to solve the exclusion problem, though, then my antireductionism about belief properties would appear to be in trouble: I take belief properties and events to have normative features which in some sense are not included within the physical properties and events that realize them.
As a first step toward addressing this prima facie problem my for view, let me observe that though the idea of inclusion is captured by the Shoemakeresque view which we used to set up the present account of mental causation, inclusion per se seems to play no essential role on the account. Given the part/whole causation thesis, a mental event and a physical event will cause a common effect if they share a part (a causal power token) that causes that effect. Now, one way that wholes can share parts is if one is the part of the other. This is not the only way though. Consider two wholes, one composed of a b-token, a c-token, and a d-token, and the other composed of that very c-token, that very d-token, and an e-token. Though they share parts, neither of these wholes is a part of the other – each has a part the other lacks. Suppose next that one of their shared parts, the c-token say, causes some effect. Given the part/whole causation thesis, it then follows that each of these wholes causes that effect.
Two composite events that are related in precisely this way are the striking of a dry match – that is, the match’s instantiation of (being struck & being dry) – and the striking of a match in the presence of oxygen – that is, that very match’s instantiation of (being struck & being in the presence of oxygen). If there is no causal competition between the striking of a match and the striking of a dry match, then surely there is likewise no causal competition between the striking of a dry match and the striking of that match in the presence of oxygen.
Inclusion, then, does not seem to be required to kill the appearance of causal competition. And this would appear to be a good thing for Yablo’s and Shoemaker’s sakes. For, even aside from my own normativity-based antireductionism, there is reason to think that Yablo and Shoemaker will need to give up on the idea of inclusion in favor of the sort of partial overlap we have been describing. For something that we have completely left out of the picture thus far are natural properties’ causal liabilities. That is, the propensities that natural property instantiations have to be caused by certain events. Causal liabilities are what Shoemaker calls “backward-looking” causal powers, although to minimize confusion I will not refer to them as powers at all. Let us now turn to some of the complications that arise once causal liabilities are introduced.
9.3.2 CAUSAL LIABILITIES
To have an example of a causal liability, consider pain. Not only does pain tend to cause wincing, sobbing, teeth-gnashing, etc., but it tends to be caused by tissue damage. For a causal-functionalist account of the mind to have any hope of being viable, it will need to take into account such causal inputs of pain as well as its causal outputs. Similarly, for a form of causal structuralism (reductive or not) to have any hope of being viable, it will need to include causal liabilities just as much as causal powers in its account of the relation between natural properties and their causal profiles.
The reason causal liabilities pose a problem for the idea of inclusion is that realized properties seem to have more causal liabilities than do the physical properties that realize them, not less. As Shoemaker himself recognizes, it seems that the set of causal liabilities of a multiply realizable property will be a superset of each of the sets of causal liabilities of its physical realizers.314 To see why this is, imagine a firing C-fiber stimulator, a devise that causes C-fibers to fire when it is pointed at them. Since firing C-fibers realizes pain, anything that causes C-fibers to fire will also cause pain to be instantiated. So then, one of pain’s causal liabilities will be that instantiations of it tend to be caused by firing C-fiber stimulators. But now, consider inflating D-tubes. A firing C-fiber stimulator cannot cause D-tubes to inflate, and so this particular causal liability will not belong to the causal profile of inflating D-tubes. Thus, this causal liability will belong to pain’s causal profile without belonging to the causal profile of one of its physical realizers. And the point generalizes.
Shoemaker thinks that causal liabilities can be set aside for the purpose of setting out his subset view of realization.315 This is incorrect though. Consider a physical property PZ whose instantiations typically cause wincing, sobbing, teeth-gnashing, etc., but whose instantiations never, as a matter of nomological impossibility, are caused by tissue damage. Given a broadly causal-functionalist view, PZ should then not qualify as a physical realizer of pain. It lacks the right sorts of causal inputs. On Shoemaker’s subset view of realization, though, with its focus on causal powers to the exclusion of causal liabilities, PZ will qualify as a pain realizer: pain’s set of causal powers (excluding its liabilities) will be a subset of PZ’s set of causal powers (excluding its liabilities). To block this consequence the subset view will need to be amended somehow, it will need to take liabilities into account as well. I suspect there are ways this can be done which retain the general spirit of Shoemaker’s original account, although carrying this out does not seem to be completely unproblematic, and the resulting view cannot be called a literal subset view, it seems to me.
Let’s set aside the fate of the subset view of realization, though, while pretending at this point that the notion of realization is in no way problematic. For our purposes, the most important consequence of the introduction of causal liabilities is that it now seems impossible to maintain that mental properties are fully included within the multiple physical properties that realize them. For each physical realizer property will lack certain features possessed by the mental property it realizes. And a similar conclusion seems to apply to events as well. For the same sorts of reasons that came up in connection with Kim’s argument against natural and multiply realizable properties, unless we want to deny that mental properties are natural, it seems we will need to hold that they bestow the same causal liabilities on each of their various instantiations. But then it follows that every mental event will possess certain features which the physical event realizing it lacks. If this is right, then it is impossible to maintain that mental events are fully included within the physical events that realize them.
There are different ways we could try to introduce causal liabilities into the metaphysical framework we have constructed in this chapter. Perhaps the most straightforward way would be to treat them on analogy with the way we have treated causal powers – that is, as properties and their instantiations. We could then continue to treat natural properties as conjunctions, but with conjuncts that include both causal power types and causal liability types. Pain and firing C-fibers will both be conjunctive properties that have certain conjuncts in common, but each property will have conjuncts that the other does not. Similarly, we could continue to treat natural property instantiations as composite events, but with parts that include both causal power tokens and causal liability tokens. Pains and firing C-fibers events will then have parts in common, but each will have parts the other does not.
On this view, mental events will not be parts of the physical events that realize them. Again though, this is not something that matters from the standpoint of the account of mental causation set out in the previous section. What matters is just that mental and physical events have parts in common and that those parts cause effects. And this condition is satisfied on the present view. Every causal power token that is a part of a mental event will also be a part of the physical event which realizes that mental event. Because of this, whenever such a causal power token causes some effect, that effect will be caused by both the mental event and the physical event of which that causal power token is a part, given the part/whole causation thesis.
9.3.3 PURE NORMATIVE TYPES AND TOKENS
Given this treatment of causal liability types and tokens, I now want to introduce what I will be calling pure normative types and tokens, and treat them in a similar fashion. What makes these normative types and tokens pure is something that will become clearer below, but for now we simply can ignore this part of the story. On the view described in the preceding subsection, natural properties are conjunctive but their conjuncts include both causal power types and causal liability types. When we try to fit my antireductionism about belief properties into this sort of framework, we can think of it as follows: belief properties are conjunctive, but their conjuncts include pure normative types in addition to causal power and liability types. Given my antireductionism about normativity, these pure normative types are to be regarded as irreducible to causal power or liability types. Their introduction thus marks a departure from causal structuralism to be sure, but not a completely radical one. Let’s call the resulting metaphysical view causal/normative structuralism.
According to this view, belief events will be composites whose parts include causal power tokens, causal liability tokens, and pure normative tokens. Just as we saw above in connection with causal liabilities, the present view is compatible with holding that every causal power token that is a part of a belief event is also a part of the physical event that realizes it.316 Given the part/whole causation thesis, it then follows that whenever such a causal power token causes some effect, that effect will be caused by both the mental event and the physical event of which that causal power token is a part. Since this is but a single causal power token, albeit one that belongs to two distinct wholes, there is no room for causal competition between the mental event and the physical event in question.
That’s it. That is my account of how nonreductive mental causation works. While I have formulated the account using the framework provided by causal/normative structuralism, again I would prefer to avoid committing myself to this or any similarly specific view. The key idea according to the present account is not that of inclusion but rather that of partial overlap, both between mental and physical properties and between mental and physical events. I claim that such partial overlap kills the appearance of causal competition, which is the key to solving the causal exclusion problem. If alternative metaphysical frameworks can capture this idea of partially shared structure, then I would be willing to consider adopting those frameworks instead.
9.3.4 PURE NORMATIVE EPIPHENOMENALISM?
This account of nonreductive mental causation is extremely attractive, it seems to me. However, the view is not completely without its potential problems. Appreciating these problems will help us get a better handle on the account. To set up its problems, let’s begin by getting clear on what are not problems for the view. Satisfying both the letter and spirit of (Closure) is not a problem for the view, I claim. If every causal power token is a part of some physical event, which is fully compatible with the present view, then it follows that whenever any casual power token at all is exercised, the resulting effects are caused by physical events. And these physical causes will be complete, we can suppose, in the sense that every causally relevant power token will be a causal power token of a physical event. There is nothing causal here which is non-physical.
If there is a problem with the present view, it comes from the opposite direction: the view seems to render the normative epiphenomenal. Some care must be taken in stating the supposed problem here, for not all normative properties and events are epiphenomenal according to the view. In particular, belief properties and events are not epiphenomenal, and yet they are normative. What appears to be epiphenomenal are the pure normative types and tokens: if, as we were just supposing, all causes are either causal power tokens or wholes that have causal power tokens as parts, then it would seem to follow that pure normative tokens will not be causes.
Let me try to state the matter a little more clearly. I want to take the notion of pure normative types as primitive. Below I will offer a few examples of potential candidates for pure normative types, but for now such examples would be distracting. Let’s then say that a normative type is impure if it is a conjunctive property having as conjuncts pure normative types, causal power types, and causal liability types. Similarly, a normative token will be impure if it is a composite event having as parts pure normative tokens, causal power tokens, and causal liability tokens. According to the view defended in this work, belief properties will then count as impure normative types while belief events will count as impure normative tokens. The causal power tokens that are parts of belief events will sometimes cause effects. When they do, the belief events of which those causal power tokens are parts will themselves cause those effects, given the part/whole causation thesis. However, the pure normative tokens that are parts of belief events will cause no effects. Thus, the component of belief that makes belief normative is epiphenomenal. Or so it seems.
Beliefs appear to have epiphenomenal parts then. This thought might tempt us to want to rethink the part/whole causation thesis. Or, abstracting way from the details of the metaphysical framework I have adopted in this chapter, it might tempt us to think that the present account of mental causation has a flaw similar to the one Davidson’s original account is thought to have: if a state qualifies as a belief only by virtue of having certain (pure) normative features, while no events are ever causes by virtue of their pure normative features, then it might seem that no events are ever causes by virtue of being beliefs. In the subsection that follows I will be addressing the objection at this higher level of abstraction, though what I say will in effect constitute a further defense of the part/whole causation thesis.
9.3.5 “BY VIRTUE OF” RUN AMOK
Setting aside normativity and regardless of how one understands the relation between natural properties and their causal powers, it seems that one will have to admit that for any given effect caused by a natural property instantiation, not all of the features of the instantiation will be relevant to the effect. Suppose that electricity is conducted through a wire made of copper. Is it by virtue of the wire’s being made of copper that this effect is caused? Well, copper has many features that appear causally irrelevant to the conduction. For instance, that copper reflects reddish light seems causally irrelevant: had the laws of nature been somewhat different and copper reflected a different colored light, the conduction of electricity still would have occurred, provided that copper still had the electrical resistivity that it does. Analogous considerations seem to suggest that copper’s having the boiling point it does is also causally irrelevant. And this conclusion appears to generalize fairly widely.
Many of copper’s features, then, appear to be causally irrelevant to the conduction of electricity through the wire. Should we then deny that the wire conducts electricity by virtue of being made of copper? If we do, we will be headed down a path which most likely will have us denying the causal efficacy of all natural properties and their instantiations, since the situation we are encountering with copper seems to be the rule rather than the exception. Now, belief differs from copper in that a belief’s pure normative features appear to be causally irrelevant to all effects, while the copper wire’s ability to reflect reddish light, say, is causally relevant to some effects but not others. But it is unclear why this difference should be relevant to the present point. If a belief’s having epiphenomenal features disqualifies it from being a cause of any effect, then presumably copper’s having features that are causally irrelevant with respect to the conduction of electricity should disqualify the wire’s being made of copper from being a cause of the conduction of electricity. But this latter conclusion should clearly be resisted, and so the former one should be too.317
We can accept this result even as we hold that Davidson’s critics had a point in pushing their by-virtue-of objections against him. If mental events are not causally efficacious by virtue of their mental properties, this would seriously undermine mental causation. According to the present view, though, mental events are causally efficacious by virtue of their mental properties. While states qualify as beliefs partly by virtue of their normative features, they also qualify as beliefs partly by virtue of their causal features. Thus, when some effect is caused through the exercise of such a causal feature, that effect is caused by virtue of part of what makes a state a belief. The issue here is that mental events are not causally efficacious by virtue of their pure normative components. What is threatened here is not mental causation but pure normative causation.
9.3.6 EMBRACING PURE NORMATIVE EPIPHENOMENALISM?
What I promised back in Chapter 5 was an account of mental causation, and this I have now provided. No promises about pure normative causation were ever made. Still, let me set describe what strikes me as the two most promising ways that the view defended here could be further developed. I am undecided between which of these two options I prefer.
The first option involves simply embracing pure normative epiphenomenalism. This embrace will be most plausible if we combine with the claim that many of the normative types that most interest us are pure normative types, just as belief properties are. Perhaps we can riff off Bernard Williams’s distinction between thick and thin normative concepts at this point.318 Just as physical properties and events are “thicker” than mental properties and events are on the Shoemakeresque view, impure normative types and tokens will be thicker than pure normative types and tokens on the view defended in this section. Every impure normative token will be a whole composed of pure normative tokens, causal power tokens, and causal liability tokens, but no pure normative token will have causal power tokens or (thus) impure normative tokens as parts. Similarly, every impure normative type will be a conjunctive property having pure normative types, causal power types, and causal liability types as conjuncts, but no pure normative type will have causal power types or (thus) impure normative types as conjuncts.
The invocation of Williams is meant to be partly suggestive of which sorts of normative types and tokens will qualify as impure (thick) and which will qualify as pure (thin). There is room for disagreement about individual cases, but impure normative types presumably will include things like being courageous, being treacherous (these first two are taken from Williams), being justified, being rational, and being a believer (according to the view defended in this work). Pure normative types might then include things like being right and being as one ought to be (again, both examples are taken from Williams). The rough idea guiding the distinction here is that impure normative types and tokens have pure normative components to them. So, for instance, a component of being courageous, say, is being as one ought to be. However, impure normative types and tokens also have non-normative components to them. So, for instance, a component of being courageous is having certain thoroughly non-normative causal dispositions, dispositions which conceivably could be possessed even if there were no normativity – even if normative eliminativism were true. By contrast, pure normative types and tokens do not have such non-normative components to them. What is the non-normative component of being as one ought to be?
Just as belief properties and events are causally efficacious on the present account, so too will be all impure normative types and tokens. Thus, we will be able to say truly that a soldier acted as he did because he is courageous – this can be a correct causal explanation. Similarly, we will be able to say truly that a student stayed in and studied the night before the final rather than going out and partying because she is rational.319 Still, pure normative types and tokens will not be causally efficacious. In connection, what makes impure normative types and tokens normative will be epiphenomenal, just as in the case of belief.
There is plenty of normative causation then, the question is just whether there is enough of it – that is, enough for both the metaphysics and epistemology of normativity. Offhand, I do not find it obvious that pure normative epiphenomenalism would bring with it the end, in the way that it does seem obvious that mental epiphenomenalism would be world-ending, or perhaps even that general normative epiphenomenalism would be world-ending. There is at least something to be said for the idea that pure normative types and tokens of the sort that seem to be in question do not even purport to be causal. However, much more work would need to be done before it could be shown that this option is fully acceptable.
9.3.7 A DOUBLE ASPECT THEORY?
Here would be a happy thought, perhaps: the pure normative epiphenomenalism that seems to be forced on my view is really just an artifact of the casual/normative structuralism I have been using to formulate my view as clearly as possible, but which is not really essential to my position. One thing which suggests that this might be so is that the assumption of physicalism, and more specifically of (Closure), plays absolutely no role in pressuring my view toward pure normative epiphenomenalism. Suppose physicalism and (Closure) were false. That is, on the causal/normative structuralist view, suppose that some causal power tokens are not parts of physical events. It still seems that since pure normative types and tokens are things of one sort while causal power types and tokens are things of another sort, the threat of pure normative epiphenomenalism looms just as large.
Causal powers and liabilities are often contrasted with qualities, the standard examples of which are often taken to be qualia. Now, reductive causal structuralists attempt to do without qualities. They attempt to account for all features of properties completely in terms of causal powers and liabilities. In introducing pure normative types as distinct from causal power and liability types, what I am in effect claiming is that this cannot be done. In making this move, my position effectively assigns pure normative types a role analogous to that played by qualities on certain other metaphysical views that are opposed to causal structuralism: roughly, I claim that pure normative types are property-like entities over and above causal powers and liabilities, while other philosophers make similar claims about qualities.
Just as my pure normative types seem to run into causal problems, qualities, once they are set alongside causal powers and liabilities, seem to run into causal problems. John Heil expresses the point like this: “Purely qualitative properties appear epiphenomenal. A world containing purely qualitative and purely dispositional properties looks like a combination of two bad ideas: a world of pure powers and a world of inert undetectable qualities.”320 Heil thinks that the way to solve this apparent problem is not to do away with qualities, as causal structuralists do, or to drain causal powers of their power, as Humeans like Lewis do. Rather, the way to solve it is to adopt a kind of double aspect theory, according to which some or all of a natural property’s causal powers and liabilities are essential to it, but they do not exhaust that property’s essence. According to Heil, we are to think of
power-bestowing properties as qualitative: the possession of a property is the possession of a powerful quality. Differently put: properties (intrinsic properties of concrete objects) are both qualitative and dispositional; every such property is a quality and is a power. Although a position of this kind produces its share of incredulous stares, I think it both utterly plausible and perfectly natural.321
If such a double aspect theory for qualities and powers/liabilities could be made to work, then perhaps I could appropriate it to the view being defended here. The idea is not necessarily that pure normative types are qualities, but rather that the same sorts of moves that can be made to endow qualities with causal efficacy will be available for normative types as well. I cannot say that I have much confidence at this point that this really can be done. I do not fully understand how double aspect theories like Heil’s are supposed to work – I am one of the incredulous starers, at least for now.
Still, perhaps solace ca be taken in the thought that the causal problem that pure normative types seem to be running up against on my view is just a special case (or at least something like a special case) of a causal problem that faces the qualitative component or aspect of natural properties of any sort – physical, mental, normative, etc. If so, then the view defended in this work solves certain causal problems while leaving us with no causal problem that we – that is, all philosophers, regardless of whether or not they are antireductionists either about the mental or normativity – do not have to solve anyway. If so, this would constitute genuine philosophical progress.