The IBE Argument against Reductionism

Each of the forms of reductionism examined in Chapter 6 denies the existence of natural and multiply realizable mental properties. This denial can be cast as the rejection of a certain realist thesis. The views in question each deny that all pained beings, for instance, share a metaphysically significant feature (i.e., a natural property) – something completely independent of our ways of grouping such beings together with concepts, or of semantic facts about the meaning of our word ‘pain.’ A common way to defend realist theses in general is by making some sort of inference to the best explanation (IBE) argument: antirealist views may be consistent with certain data, but only realist views can adequately explain that data. This, in fact, is the sort of argument I will be making in this chapter. I will attempt to show that the reductionists we have been considering are not able to adequately explain certain empirical findings, while those nonreductive physicalists who are realists about natural and multiply realizable properties are able to do so.

Exposition will go smoothest if we start by focusing on just one of the forms of reductionism we’ve been considering, realizer-state functionalism say, and then later generalize the conclusions we reach to cover alternative reductionisms. Assuming realizer-state functionalism, then, suppose that each of the following causal generalizations is true.

(1): Human pains, when of a certain duration and intensity, typically cause hair loss.

(2): Both human pains and Martian pains typically cause wincing.

(3): Both human pains and Martian pains, when of a certain duration and intensity, typically cause hair loss.
I want to examine each of these generalizations in turn and consider how a realizer-state functionalist could try to account for its truth.

7.1.1 EXPLAINING (1)

The feature of (1) that is of special interest to us is that the causal power it describes – the power to cause hair loss in the specified circumstances – is not one that the realizer-state functionalist we have been imagining takes to be built into pain’s causal role. Now, as long as this feature is taken by itself, it poses no real threat to the realizer-state functionalist. She can fully accept that there might be empirical discoveries about the causal powers possessed by a property like firing C-fibers (a.k.a. P_H, a.k.a. pain-in-humans), that is, a property that occupies a given causal role.

In fact, it would be disastrous for realizer-state functionalists if they could not allow for this sort of thing. For surely, firing C-fibers (or, for that matter, any other physical occupant of pain’s causal role) will possess many, many such causal powers. For instance, firing C-fibers will possess a certain mass, and so they will exert a certain gravitational force on the planet Neptune. No realizer-state functionalist will want to take this power to be part of pain’s defining causal role though – if she did, she would need to deny that anything more massive or less massive than firing C-fibers could occupy pain’s causal role. Thus, the realizer-state functionalist needs to allow that there inevitably will be true generalizations relevantly like (1) out there to be discovered. Perhaps they will be discovered through psychology, perhaps through neuroscience, and perhaps through yet other scientific disciplines.

Something close to the present point can also be expressed if we put things in terms of causal laws rather than causal powers. Realizer-state functionalists can consistently allow that (1) is an empirically discovered causal law.²³⁰ This sort of translation from powers talk to laws talk will be helpful at certain points in the discussion that follows, so I will make free use of it. I want to emphasize in advance, though, that I won’t be try to get any real metaphysical mileage out of it. For the sake of the argument I want to make, I can grant that (1) is not really a law. For instance, I can grant that the implicit ceteris paribus clause in (1) – marked by the word “typically” – disqualifies it from properly counting as a law.

7.1.2 EXPLAINING (2)

Moving on, the feature of (2) that is of special interest to us is that it ranges over events that are physically dissimilar, events that are the instantiations of two wholly distinct natural physical properties, P_H and P_M. Again though, as long as this feature is taken by itself, it poses no real threat to realizer-state functionalism. If the realizer-state functionalist is right about the meanings of our mental terms, than (2) is something like an analytic or conceptual truth: ‘pain’ just means something like the property whose instantiations are typically caused by tissue damage and typically cause wincing.²³¹ According to the realizer-state functionalist, if instantiations of P_H and instantiations of P_M didn’t typically cause wincing, then the term ‘pain’ wouldn’t properly apply to them in the first place.

Think about if from an epistemological angle. Imagine that the way we first learn of Martians is by receiving a radio transmission from an astronaut of ours who is exploring their planet. The transmission tells us two things: Martians exist and they sometimes suffer pains. If realizer-state functionalism is correct, then even with this minimal empirical knowledge we have just acquired there will be excellent reason to expect that Martian pains will typically cause wincing. It is not on the basis of induction that that this expectation would be justified. It is not that our past observations of various human pains that cause wincing constitute empirical evidence for (2) (which entails that Martian pains typically cause wincing). Rather, our expectation is justified on the conceptual grounds that if whatever physical property occupying pain’s causal role in Martians didn’t have instantiations that typically cause wincing, then ipso facto it would not count as an occupant of pain’s causal role.

7.1.3 EXPLAINING (3)

Finally consider (3), which is the sort of generalization that will interest us most in this chapter. Like (1) but unlike (2), (3) is no analytic or conceptual truth by the realizer-state functionalists’ lights. Like (2) but unlike (1), (3) ranges over physically dissimilar events. When these two features are combined together, a problem arises for the realizer-state functionalist. Realizer-state functionalists, it seems, are forced to regard the truth of (3) as a certain kind of coincidence. To clarify the sort of coincidence involved, let me first say what sort of coincidence is not involved.

Given the metaphysics of realizer-state functionalism, (3) can be true only if the physical properties P_H and P_M both contain within their causal profiles the power to cause hair loss. Translating this powers talk into laws talk, (3) can be true only if there is a law that P_H instantiations cause hair loss and also a law that P_M instantiations cause hair loss. These laws taken jointly entail that (3) is nomologically necessary. Given this result, whatever the sense is in which (3) would be a coincidence if realizer-state functionalism were true, it must be importantly different from the sense in which we might call accidentally true generalizations “coincidences.” The accidentally true generalization that all the objects in Nelson Goodman’s pocket are made of silver is importantly unlike (3) in that there are nomologically possible worlds where Goodman’s pocket contains pennies (or other objects not made of silver), but there are no such worlds where (3) is false.

Rather, the sense in which (3) would be a coincidence if realizer-state functionalism is true is like the sense in which it’s a coincidence that both emeralds and alligators are green. Being an emerald and being an alligator are two wholly distinct natural properties (suppose they are genuinely natural), and so the fact that all alligators are green is completely metaphysically independent from the fact that all emeralds are green. What’s more (further suppose), it’s no analytic truth that either emeralds or alligators are green. If realizer-state functionalism is true, then the truth of (3) is relevantly like this. There is no underlying unified explanation of why both P_H and P_M typically cause hair loss. Rather, the explanation is just that there is one fact that P_H instantiations do so, and another completely independent fact that P_M instantiations do so.

Again, think of it from an epistemological perspective. Imagine that before receiving any further radio transmissions from our astronaut on Mars, a psychological study is released here on Earth stating that (1) is true: human pains do cause hair loss. Do we then have good reason to expect that Martian pains will cause hair loss as well? Given that there is no analytic truth in play this time, the question here is whether our newly gained empirical knowledge that human pains cause hair loss gives us good inductive grounds for expecting that Martian pains will cause hair loss too. That is, does (1) inductively support (3)?

It seems that if realizer-state functionalism is correct, the answer to this question must be no. The truth of (1) tells us something about the causal profile of P_H, and this may give us reason to infer something about the behavior of all unobserved P_H instantiations. But unless P_H is also the occupant of pain’s causal role in Martians, this would give us no reason to infer anything about the behavior of pains in Martians. If pain’s causal role is occupied by some other physical property in Martians – like P_M, as we’ve been supposing – then the truth of (1) will tell us nothing about this physical property’s causal profile. In short, taking (1) to support (3) would be like comparing emeralds to alligators. Even if we discover that every last emerald in the world is green, this gives us no compelling inductive reason for thinking that all alligators are green. Of course, as it turns, out all alligators are green. And similarly, we can suppose, as it turns out, Martian pains typically cause hair loss. In each case, though, that things turn out in this way is just a kind of coincidence.

This is not an inevitable consequence for a metaphysics of mind. Reconsider the case while supposing that, pace realizer-state functionalism, pain is a natural and multiply realizable mental property. If so, then both human pains and Martian pains will be instantiations of that single natural property. But then, discovering that human pains cause hair loss in the specified circumstances really would seem to give us good inductive reason to expect that Martian pains will too: (1) really would inductively support (3). For, (1)’s truth would show that one of the powers belonging to pain’s causal profile is that of causing hair loss. Since it is a general metaphysical principle – accepted by many philosophers, including each of the reductionists we have been considering – that natural properties bestow the same causal powers on their various instantiations, it would then follow that the power to cause hair loss will be bestowed on those instantiations of the (natural and multiply realizable) property pain that occur in Martians. And so, we would have reason to expect that Martian pains will typically cause hair loss in the specified circumstances.

The key issue here is the link between a property’s being natural and its being projectible – that is, its ability to figure in justifiable inductive inferences.²³² Nonreductive physicalists who take pain to be a natural and multiply realizable property regard it as projectible, meaning that it can figure in justifiable inductive inferences of the sort we’ve been considering. Realizer-state functionalist who deny the existence of natural and multiply realizable mental properties must deny that there is a projectible property common to both human pains and Martian pains. Because of this, they are committed to denying the sort of inductive inferences we’ve been considering. The IBE argument against reductionism turns on this difference in projectibility between reductive and nonreductive views.