Contra Bentham (And Chalmers) on Physicalism

Metametaphysical issues with arguments from conceivability

Bentham of Bentham’s Newsletter wrote an interesting article a while back where he sketches a series of arguments against physicalism, citing Chalmers extensively. I will argue, in this essay, that they are wrong, but very understandably wrong. The arguments are certainly valid; no one can impeach the logical structure of Bentham’s or Chalmers’ arguments.

I’ll take as a prototypical example of the sort of argument presented the P-zombie argument in order to demonstrate where the argument fails, but all the arguments presented against physicalism are the same in the relevant ways.

On the surface, it seems perfectly legitimate. The argument goes like this:

1. A being could be physically identical to me but could not be conscious

2. Two beings that are physically identical must have all physical properties in common

3. Therefore, consciousness is not a physical property.

Setting aside our intuitions about (1) for the moment—I’ve written before about why I don’t find p-zombies possible—fine. The argument is logically valid. But there are some other assumptions being smuggled in here. In reality, if we unpack (1), it looks something like this:

1*. A being that is physically identical to me but could not be conscious is conceivable.

2*. Anything that is conceivable is logically possible.

3*. Therefore, it is logically possible that there could exist a being that is physically identical to me but could not be conscious.

Now that we’ve spelled that out, let’s dissect the first two premises to see how they hold up under scrutiny.

In particular, I want to think about what it means for something to be conceivable, and I want to determine if it’s true that anything conceivable is logically possible.

We’ll start by thinking about (1*). I claim that the notion of some object X being conceivable is closely related to vagueness. If I can vaguely imagine something but can’t mentally render it in suitable detail, we don’t want to say I can conceive of it. I can sort of, kind of imagine a world where p & ~p is true. I can imagine the possibility, but I can’t tell you in any detail what such a world would be like, and we don’t want to say I can conceive of it. On the other hand, I can conceive of a purple elephant. I can render it in non-trivial detail in my mind. I can visualize it, I can imagine it having brown eyes and a wrinkly trunk, I can tell you what it might eat—in short, I can (in principle) answer probative questions about my purple elephant, and then no one should doubt that I have truly conceived of it.

This points us to the following definition: we’ll say X is conceivable if there exists a person P who can render a mental model of X in suitable non-trivial detail as to, in principle, be able to answer probative questions about X.

And that, too, points us in the direction of two new questions about conceivability: can a person P reliably report if an object X is conceivable, and is whether P believes X is conceivable related to their logical priors?

In order for the anti-physicalism argument to go through, we need it to be the case that people can reliably perceive and report accurately if an object X is conceivable, and we also need it to be the case that whether P perceives X to be conceivable is independent of P’s logical priors (at least on a rhetorical level, if not logically; you won’t have much luck convincing another person P’ that X is conceivable if you can only conceive of it because you already believe in some other proposition Q that P’ disbelieves).

I claim that these conditions are mutually exclusive.

Proof: suppose for all objects X and all people P, whether P will perceive X as conceivable is independent of P’s logical priors. Then the fact that Chalmers and I disagree on the conceivability of p-zombies is proof that at least one of us is (and maybe both are), in principle, not capable of reliably reporting if X is conceivable. Chalmers would, in order to salvage the anti-physicalism argument, have to further furnish an argument showing that he can reliably and accurately report whether an object X is conceivable¹.

And, we note again, if it’s the case that the conceivability of O by X does depend on X’s logical priors in a causal way, then arguing against physicalism on the basis of the conceivability of p-zombies seems to be some kind of circular reasoning.

I believe I’ve at least done enough to cast doubt on the premise (1*). Now let’s turn to (2*), the premise that says “If X is conceivable, then X is logically possible.” Again, it seems on the surface to be reasonable enough; I don’t think anyone wants to admit that we can conceive of, for example, worlds where p & ~p are true or square circles. And yet, there is more depth to the question of whether 2* is true than one might expect.

Note that 2* is equivalent to the proposition “If X is logically impossible, then X is inconceivable.”

I don’t think this is true. At the very least, it doesn’t seem obviously so, and Chalmers and Bentham would have to furnish an additional argument for it. I claim that I can conceive, for example, of finding a counterexample to Fermat’s Last Theorem. I can imagine checking a huge amount of quintuplets of integers in some systematic way, more than we’ve ever checked with computers or by hand, and eventually stumbling upon an x, y, and a z, and an n, such that x^n+y^n=z^n and n>2. I can even imagine it in some detail. I can imagine that they’re all large numbers, they’re all primes, I can imagine a number of the sorts of characteristics they would have to have if they existed—and yet, they don’t. They can’t, as Andrew Wiles proved. The existence of such a quintuplet is logically impossible given standard assumptions about math, but it seems to me I could plausibly conceive of it.

Maybe, we might argue, that’s because I’m just some lowly schlub and not a number theory expert. Because I don’t understand Wiles’ proof of Fermat’s Last Theorem, maybe there’s an explanatory gap in my mind where, if I understood exactly why Fermat’s Last Theorem is true, then I wouldn’t think such a quintuplet is conceivable. If we accept that conceivability is bound by logical priors, then maybe if I don’t really understand why something is true, it doesn’t count as a logical prior even if I believe it is true.

And, sure, that strikes me as a plausible story. But then it seems we must admit the same explanation could in principle apply to p-zombies and all the rest of the arguments against physicalism. Maybe the reason someone would find p-zombies conceivable is because we understand little enough about consciousness to create a similar explanatory gap.

Absent a better explanation, I’m sticking with reductive physicalism, and I don’t think I’ll be convinced otherwise by arguments from conceivability. Or, at the very least, a successful argument would have to be substantially more complicated than the standard conceivability account of p-zombies.

1. I conjecture that might be impossible to prove. We can imagine showing Chalmers extremely complicated mathematical theorems and falsehoods and asking him which ones are conceivable, and I don’t think his intuitions, or anyone else’s, will be particularly reliable.