Epiphenomenalism Defended

Consciousness doesn't do anything

The brief case for epiphenomenalism

For a while I described myself as a hardcore Chalmersian about consciousness. I pretty much thought Chalmers was right across the board about consciousness—the Chalmers party line was consistently correct. The one place I diverged was that Chalmers is very sympathetic to epiphenomenalism, and I thought epiphenomenalism was crazy.

But now the small island of things I disagree with Chalmers about related to consciousness has disappeared in the sea of things I agree with Chalmers about related to consciousness, for I recently changed my mind about epiphenomenalism. Epiphenomenalism is the doctrine that consciousness is causally inert—physical things, in conjunction with the psychophysical laws, cause consciousness, but consciousness itself doesn’t cause anything. For a while I was a very staunch critic of epiphenomenalism, thinking that it was super implausible. While it’s strictly more parsimonious than interactionism—which is otherwise the same as epiphenomenalism but says that consciousness causes stuff—I thought that epiphenomenalism suffers from fatal objections.

But I’ve recently changed my mind about that. Reading Chalmers’ Tour de Force The Conscious Mind, I think epiphenomenalism is now very plausible—such that I’m now undecided between it and interactionism. Here, I’ll explain why I’m now on team epiphenomenalism.

The basic case for epiphenomenalism is relatively simple. It starts with the assumption that physicalism is wrong. I’ve argued for that elsewhere, so I won’t rehash the various arguments. But once we conclude that physicalism is false, it seems like epiphenomenalism is the default. It’s hard to make interactionism consistent with the conservation of energy and charge, for example. So unless we’re given a good reason to posit extra causal laws that conflict with our best science, it seems like we shouldn’t do so.

There’s also some solid empirical evidence for epiphenomenalism, summarized here. The basic facts are as follows:

1. Libet showed that for many decisions that we believe to be under our conscious control, one can predict whether we’ll take them before we come to a decision about them. But this provides evidence that there’s some nonconscious brain process that causes the movements—if it were the byproduct of consciousness, then one wouldn’t be able to predict it in advance just by looking at the physical. Of course, this isn’t a proof—it could very well be that consciousness plays some role in potentially vetoing the actions of our unconscious minds, as Libet admits, but it’s some evidence. Others have found similar results, which provides evidence that there are plausibly physical causes of our behavior.

2. Various people have found that we have the illusion of control. When some physical change is the result of some purely mechanistic processes outside of our conscious control, we mistakenly believe that we’re in control. But this means that we can have the illusion of control and at least some cases that we believe to be under our conscious control are the result of purely mechanistic physical processes.

I think most people agree that epiphenomenalism is in some sense the default—it has something going for it, so unless there’s a good reason to reject it, it’s worth accepting. They simply think that epiphenomenalism is totally unacceptable for one reason or another. I used to think that there was one pretty damning objection to epiphenomenalism, but I’ve recently shed that belief. Here I’ll explain why none of the objections to epiphenomenalism are fatal, and few are convincing at all. I’ll begin by arguing for the conclusion that there is a fundamental distinction between views that are epiphenomenalist at the level of the brain and views that are interactionalist at the level of the brain. I will argue that this is the fundamental distinction involved in dualism and that various views like idealism that appear to preserve mental causation devolve into something like epiphenomenalism unless they posit strongly emergent physical forces. Thus, the fundamental question is whether at the level of the brain—at the level of unified consciousness—there are new causal laws.

There are only two choices (also, why I’m not a panpsychist, idealist, or panprotopsychist, or at least why they don’t have much of an advantage over traditional epiphenomenalism)

This section is sort of technical so skip it if you’re not interested in hearing about why all views have to either inherit the problems of interactionism or epiphenomenalism.

There are lots of views about philosophy of mind. Many of them are physicalist—but such views are just complete nonstarters; Mary in her room saw to that (I actually have a housemate named Mary, and so have been in Mary’s room, though it sadly contains red). But even among nonphysicalists, there are roughly 5 views:

1. Epiphenomenal dualism: consciousness is caused by psychophysical laws that apply to brains. It is causally inert.

2. Interactionist dualism: consciousness is caused by psychophysical laws that apply to brains. It is causally effective.

3. Panpsychism: consciousness is fundamental and is had by atoms.

4. Idealism: consciousness is all that exists.

5. Panprotopsychism: there is fundamental stuff that is neither mental nor physical but from which the mental and physical emerge.

I’m going to mostly ignore idealism and panprotopsychism partly because they are less unified and hard to pin down, partly because the considerations raised in relation to panpsychism apply to them as well, and partly because it takes far too long to type out the word panprotopsychism.

Even under the category panpsychism, there are three different versions:

1. Consciousness is fundamental to atoms. It weakly emerges at the higher levels. In other words, there are no other forces that cause the consciousness in the atoms to combine. Similarly, the physical description of the world is identical to that proposed by physicalists.

2. Consciousness is fundamental to atoms. It strongly emerges from the lower levels. There are fundamental laws about how consciousness combines. However, this combined higher-level consciousness doesn’t generate any new physical effects. The physical description of the world is identical to that proposed by physicalists.

3. Consciousness is fundamental to atoms. It strongly emerges from the lower levels. There are fundamental laws about how consciousness combines. This combined higher-level consciousness does generate any new physical effects. Thus, the physical description of the world isn’t identical to that proposed by physicalists.

1 is totally implausible. My consciousness is coherent and unified. Proposing that it is nothing over and above a bunch of simple atoms being conscious is obviously false (of course, it has some smart defenders, but I find it nuts!).

View 2 is epiphenomenal at the level of the brain. It holds that consciousness at the level of the brain doesn’t cause anything, though it is strongly emergent. It’s only the fundamental consciousnesses at the level of atoms that causes things. View 3 in contrast is interactionist at the level of the brain—it holds that the consciousness in the brain causes things, not just the consciousness at the fundamental level.

But view 3 inherits all the problems of interactionism. It violates causal closure of the physical and potentially conflicts with conservation of energy. View 2 inherits the problems of epiphenomenalism because it holds that the consciousness at the level of the brain is causally impotent.

This problem generalizes; either consciousness is causally efficacious at the level of the brain or it isn’t. If it isn’t, then the brain’s consciousness is epiphenomenal—if it is, then all the problems for interactionism reappear.

Bearing this in mind, panpsychism, idealism, and panprotopsychism lose much of their advantage. They either inherit all the problems for interactionism or for epiphenomenalism. And it’s not clear that they’re simpler, because they have to posit complex laws about how different consciousnesses combine which are strictly less parsimonious than the epiphenomenalist or interactionists law. For instance, a panpsychist might think the fundamental laws are:

1. Atoms are conscious.

2. This consciousness has X physical effects when observed from the outside.

3. When such and such physical things happen, the consciousness combines.

If they’re a panpsychist interactionist holding the third view described above, they’ll also think:

4. Consciousness causes such and such things when it combines.

In contrast, the dualist thinks the fundamental laws are:

1. There are X physical effects from atoms.

2. When such and such physical things happen, the consciousness combines.

If they’re an interactionist dualist, they’ll also think the fundamental laws include:

3. Consciousness causes such and such things.

So the panpsychist picture is necessarily more complex. Of course, they can have a simpler model by positing that law 3 is unnecessary, but then it’s still only as simple as dualism, and requires a deeply implausible form of reductionism, according to which macro experiences reduce to micro experiences. It also involves positing things that are very implausible, namely, that atoms are conscious. Even if the simplest view says atoms are conscious—which it doesn’t—you shouldn’t believe that unless you’re given a good reason!

For this reason, I think that while panpsychism isn’t a crazy view, it isn’t super appealing. It has pretty much the same problems as dualism, is pretty unintuitive and unnatural (I’ll talk more about that virtue later), and is a bit less parsimonious.

The core error

A lot of responses to epiphenomenalists forget that epiphenomenalists believe in neural correlates of consciousness. Epiphenomenalists think that there are some brain states that correlate with consciousness. In fact, we generally think that those neural states that correlate with consciousness represent similar information to that represented by consciousness. For instance, your brain processes pain as a negative signal that it tries to avoid, just as you consciously experience pain as something negative that you want to avoid.

As a result, it looks like consciousness is causing things. This is because consciousness represents the same information as is represented in the brain, and the information represented in the brain generates consciousness. So while it looks like pain causes you to pull your hand to move away, instead the purely physical brain state that represents pain causes one to pull away, and in combination with the psychophysical laws also causes the conscious states.

Once we have this understanding in hand, a lot of objections to epiphenomenalism disappear quickly. These epiphenomenal objections are impotent; they don’t do anything to move me (pun intended). For instance, Emerson Green lists a lot of objections to epiphenomenalism in this article. Several of Emerson’s arguments seem easily refuted by this consideration. Emerson first lists Popper’s argument against epiphenomenalism, that consciousness requires enormous energy expenditure, so it would be selected against by evolution. But what’s selected for, according to the epiphenomenalist, isn’t consciousness itself but the neural correlates of consciousness, which the epiphenomenalist says are perhaps a kind of information processing. I’ll justify later why this isn’t an ad hoc assumption.

Once we have that conception, then the epiphenomenalist has a perfectly good explanation of why consciousness evolved. It evolved because the neural correlates of consciousness are adaptive.

Emerson next argues that the placebo effect is best explained by consciousness having causal effects. According to the placebo effect, our beliefs affect how healthy we are. But the epiphenomenalist has a perfectly good explanation—there are neural correlates of consciousness that produce exactly the same types of physical effects that the interactionist supposes. Saying “consciousness causes physical placebo effects,” is no better of an explanation than “neural correlates of consciousness cause physical placebo effects.”

A final argument that is quite popular—see here for one version—is just the brute Moorean shift. Opponents of epiphenomenalism suggest that epiphenomenalism is unintuitive—it sure seems like consciousness causes things. And we shouldn’t deviate from commonsense absent a good reason. But the epiphenomenalist has a perfectly good explanation of why it seems that consciousness causes things. They think that the physical states that cause consciousness cause things, so it looks like consciousness causes things.

It turns out that when a person snaps their fingers, the movement of their fingers isn’t what causes the noise. The thing that causes the noise is their fingers hitting the side of their palm (you can test this by blocking your palm). But it’s no surprise that people think that the snap itself causes the noise, because they observe a systematic correlation. It seems that the epiphenomenalist can give the same explanation of our intuitions. And just as it would be silly to Moorean shift in the snapping case (it’s more intuitive that snapping causes noise than any argument against it), so too is it silly to Moorean shift here.

In fact, the epiphenomenalist can give a very natural-sounding explanation of conscious causation. The epiphenomenalist can accept, for instance, that the pain in my hands when I touched the hot stove explains why I pulled my hand away. They just think that pain has both an experiential component and a physical component which both overlap in information content, and that it is the physical component that causes me to pull my hand away. This doesn’t strike me as crazy! The SEP page on epiphenomenalism notes:

Epiphenomenalists, however, can make the following reply. First, it can never be obvious what causes what. Animated cartoons are full of causal illusions. Falling barometers are regularly followed by storms, but do not cause them. More generally, a regularity is causal only if it is not explained as a consequence of underlying regularities. It is part of epiphenomenalist theory, however, that the regularities that we observe to hold between mental events and actions can be explained by underlying regularities.

Richard Chappell has a pretty natural-sounding explanation of epiphenomenalism:

We may think of mental states as having both physical and experiential components: their physical effects are due entirely to the physical aspects of our thoughts. The non-physical (experiential) component, on the other hand, constitutes what it feels like to be in that state. There's then an obvious sense in which our mental states have causal effects, insofar as their physical aspects do. That doesn't require that the causal 'oomph' come from the experiential aspect -- indeed, how could it? Experiential feels aren't the kinds of things that push atoms around. You need other particles to accomplish that!

On this view, we may (speaking loosely) say that I pulled my hand away from the hot stove "because it hurt", and this can be perfectly informative, without implying that the hurty feel itself provided the causal force that moved my hand...

These considerations are enough to undermine the evidence the placebo effect, Moorean shift, and the simpler version of the evolution objection.

I’m not quite sure if these are no evidence or just a tiny amount of evidence, but either way, none are convincing. Certainly not convincing enough to justify deviating from our well-understood physics by positing new interaction laws, almost all of which would violate energy conservation and conflict with natural interpretations of experimental results.

Epiphenomenalists can accept that other people are conscious

Michael Huemer recently objected to epiphenomenalism on the grounds that it can’t explain why we are rational to infer that other people are conscious. He claimed that this argument made epiphenomenalism the “next most insane view,” after the view that consciousness doesn’t exist. His argument is paraphrased as the following:

1. We infer that other people have minds through inference to the best explanation.

2. If epiphenomenalism is true, other people’s minds don’t cause anything.

3. If something doesn’t cause anything, it doesn’t feature in our best explanation of human behavior.

So epiphenomenalists have no reason to think other people have minds.

I think that this argument is not at all convincing. Huemer is very smart, so maybe I’m missing something, but it strikes me as a genuinely terrible argument! Before I get into the details, it seems totally crazy to regard epiphenomenalism as the second craziest view after eliminativism about consciousness based on this. The assumptions, I shall argue, aren’t plausible, but even if they are, they aren’t obvious. Other views like physicalism have many many arguments that appeal to extremely obvious premises that disprove them conclusively. This argument seems way too controversial to be sufficient to label epiphenomenalism crazy.

On to the details of the argument. 1 and 2 are clearly true, but 3 is false. Huemer would presumably agree that 3 is false—he’s a moral realist, but he thinks the moral facts are causally inert. So he thinks that moral facts are the best explanation of certain obvious intuitions, even though they don’t cause them.

The epiphenomenalist has a pretty straightforward account of why we infer that other people are conscious. We know that we are conscious—more on that later—and so we infer that certain brain processes lead to consciousness. In fact, we know that we’re conscious only under certain conditions—e.g. when we’re not in deep sleep. So we come up with a theory that has various virtues that explains the circumstances under which we’re conscious. We observe that our level of consciousness correlates systematically with various brain states and certain verbal reports, and so we infer that those are the neural correlates of consciousness. We then see that other people have those things—and more broadly, brains very similar to ourselves, and similar reports—and so we infer that they’re conscious.

Imagine you had direct access to the following fact: each time your right arm is raised, a star moves in a different galaxy. Even if you can’t observe that other people cause galaxies to move, if other people report similar things, you should infer that their arms also cause faraway stars to move, because the best explanation of your arms causing such things is that there are general laws that apply to other people.

Of course, this gets practically difficult. It is super difficult to know whether animals, for instance, are conscious. But this just seems like an unfortunate fact about the world, not a reason to reject the theory. It is the case that we don’t have a good way to know whether animals are conscious, at least, at the margins. We have some pretty good ways of telling, but it’s far from perfect.

It might even be impossible in principle to figure out the right theory of consciousness. Maybe we’ll figure out two theories of consciousness that explain the data equally well. But while this would be unfortunate, it’s not impossible in principle. So we shouldn’t reject epiphenomenalism just because it might bring some bad news.

Evolutionary arguments (the good kind)

Another argument, original to William James, yet reproduced by Emerson Green and various other people is that epiphenomenalism can’t explain the systematic correlation between physical harm and harmful stimuli. On epiphenomenalism, the mental state of feeling pain doesn’t cause anything, so why would it be associated with harm and damage?

The interactionist has a perfectly good explanation of such facts. They suppose that pain causes one to want to pull their hand away, so pain feels bad. But if pain doesn’t cause anything, why would the mental state of pain produce an aversive feeling?

Now, it’s no challenge explaining this once we set the psychophysical laws. The epiphenomenalist thinks that there is a law of nature that results in the experience produced by bodily harm being painful. Specifically, the epiphenomenalist thinks that there is a law of nature according to which conscious experience has a similar content to the information represented by physical brains—so just as physical brains represent the causes of pain as “bad thing—try to stop immediately,” our conscious experience is the same.

But you might think that this is a cop-out. Yes, epiphenomenalists can give an explanation when they brutely stipulate that the laws of nature force the internal states to produce feelings like those represented by the physical systems. This is no more satisfying than solving fine-tuning by brutely asserting a law of nature that finely tunes the universe. I think there are three considerations that defang this objection quite thoroughly.

First, I think this is broadly everyone’s problem. The interactionist says that consciousness causes particular desires to form in the mind which cause physical actions which cause aversive behavior. But there are many different psychophysical laws, some of which involve pleasure forming, a desire not forming, and aversive behavior being caused. So even for the interactionist, the psychophysical laws can be scrambled such that their experiences are radically out of accordance with what is physically going on.

To illustrate this more, the interactionists explanation is physical state A (some brain state, maybe C fibers firing) causes mental state B (say pain) which causes physical state C (some brain stuff that results in one pulling their hand away). The epiphenomenalist says that physical state A causes physical state C and mental state B, but mental state B doesn’t cause anything.

The interactionist objects that mental state B could be any mental state—mental state D, C, E, F, G, etc. So the fact that mental state B pairs with an analogous physical state is a vast coincidence. But the interactionist has the same problem—mental state B could be swapped out with mental state D, E, F, G, etc, and as long as the causal regularities stay the same, so that, for instance, E causes aversive behavior, there would be no physical difference. So for the interactionist, there’s the same problem—just as you can scramble consciousness without changing the physical, interactionism says that you can scramble consciousness while keeping the physical facts the same. In fact, this same problem avails panpsychism, idealism, and all types of physicalism except type A—as Crummett and Cutter show.

I don’t even know if type A physicalism avoids the problem. There doesn’t seem to be any natural entailment from the physical facts to the phenomenal facts feeling aversive. Why couldn’t the illusion of consciousness differ greatly in feel from what’s going on in the brain?

But even if this is everyone’s problem, you might wonder what the solution is. Can epiphenomenalists give a solid account of the pairing between our mental states and physical states. I think we can.

This takes us to the second consideration. It seems very plausible to me that psychophysical laws of the kind that exist in our universe are much more intrinsically probable than unscrambled psychophysical laws. It could be that every conscious being has the conscious experience of lying in cool bathtub water. That would be simpler! But it would do much worse on the theoretical virtue of naturalness. Some explanation is natural if it’s harmonious and seems like the way you’d expect things to be. Huemer explains it well:

Be that as it may, I want to defend a theoretical virtue related to but distinct from simplicity and elegance. This is the virtue of theoretical naturalness, where “naturalness” is understood in the sense intended when one says, for example, “p is a natural suggestion” or “it is natural to wonder whether p.” If, when abstractly considering some philosophical problem, it is natural to suppose that p, then, other things being equal, it is a virtue of a philosophical theory that it be able to accommodate p. Naturalness is closely related to but distinct from the virtue of intuitiveness. A natural suggestion is not necessarily one that seems true, but it will be a suggestion for which we feel that we have no trouble seeing how it would be true. The contrast to a natural suggestion is a suggestion that seems ad hoc, puzzling, or in need of theoretical explanation, that is, in need of an explanation for why the suggestion should be true. The most natural theories are frequently simple and elegant, but an unnatural theory can be more simple than a natural theory. For example, the idea that mental states are nothing over and above bodily behavior is simpler but less natural than the idea that mental states are nonphysical states.

If a computational system that internally represented some stimuli as harmful made the stimuli feel good, that would seem odd. It would seem like the kind of thing that cries out for explanation. In contrast, if the experiences are roughly similar to what is represented computationally, that’s very natural and is thus more probable.

Thus, the pairing between the mental and the physical stops being so mysterious. It’s no more puzzling than the fact that the laws of physics are of roughly equal strength. Would it be weird if one law of physics had 100^100^100 times the force of the others, such that the others were almost totally irrelevant? Yes—it would be a disharmony that has no explanation. It would be similarly weird if there was a radical disharmony between consciousness and the physical state represented.

You might doubt that naturalness is a virtue. But naturalness as a virtue is quite easy to motivate. Holding that naturalness is a virtue accounts for this otherwise puzzling psychophysical harmony, and also explains why the laws of physics are so elegant. Because elegant laws are intrinsically more likely.

Similarly—and I’m personally less sympathetic to this way to go, but worth mentioning as its not totally crazy—you might think that skeptical worlds where the things going on in our minds are radically out of accord with what’s actually happening are intrinsically less probable. As Richard says:

Well, given that I'm at least aware of my present qualia, we can rule out most weird worlds as incompatible with my present evidence. So then we're getting into "Should I believe that the world just came into existence one second ago" territory, and I just can't take that hypothesis seriously, which commits me to thinking that it must be a priori less likely than the "normal" hypothesis. Is there an explanation for this? Not sure. It'd be nice to have one, but it doesn't seem essential to me. If forced to choose, I'd sooner just take sensible priors to be brute and unexplained than to have silly skepticism-inducing priors.

This has the nice advantage of avoiding skepticism in one fell swoop. I find it a bit unintuitive, but it’s not totally crazy and is worth accepting if you think that other solutions to psychophysical harmony fail.

The final response to this argument is that, if one is really very worried by it, there are many explanations they can posit. If one holds that the fundamental laws are not random, but instead are molded to be conducive to value either by a deity or natural force or intrinsic purpose then they can explain this easily. I’m not at all sympathetic to the belief that there’s a perfect God, but a limited God can explain this, as can axiarchism and natural teleology. Similarly, we can explain the results by an amoral God who just likes harmony or some natural tendency towards harmony between conflicting forces. I don’t think these are super likely, but they at least offer a plausible explanation that doesn’t require accepting type-A physicalism.

Consciousness talk

For a while, the main reason I wasn’t an epiphenomenalist was that I thought epiphenomenalism had a huge problem with explaining our talk about consciousness. What are the odds that we’d be conscious and our talk about consciousness would precisely line up with our conscious experience if consciousness wasn’t the cause of it? If A makes no difference to B, then why would A and B line up? This concern is pressed well by Eliezer Yudkowsky and Emerson makes it as well. Several replies are worth making:

1. Even if the epiphenomenalist doesn’t have a great exact theory of why we talk about consciousness, it isn’t hard to imagine that there would be some explanation. The epiphenomenalist doesn’t need a fully worked-out picture, just a reason to think that such a solution isn’t terribly implausible.

2. The epiphenomenalist has a pretty good explanation of roughly the following form. A system like the brain will have lots of internal parts going on. Only some of them are worth reporting on—it takes too much energy for a computational system to report all the details of the computation. So it makes sense for it to have judgments that it reports being directly acquainted with—if you ask it how it knows it’s in pain it would say “I just know it.” But once you have a physical system that talks about experiences that it’s directly acquainted with, you have consciousness talk.

3. You might think that this still seems a bit unlikely. But 2 seems super plausible—you don’t build a robot to report on all its internal goings on! But even if you’re skeptical, the epiphenomenalist has a further explanation if they believe in the multiverse—which I do for fine-tuning reasons. As we saw in response to the better kind of evolutionary worry, the most probable sets of psychophysical laws involve what’s going on in the brain being roughly represented in conscious experiences. But that means that the only beings that have metacognition—that think about their thinking or are conscious of their consciousness—in most universes have physical representations of consciousness. So almost all those who talk about consciousness and are conscious throughout the multiverse would have accurate beliefs.

4. The same problem applies to interactionism. Of all the ways that consciousness could have a physical effect, very few are significant. Even of the physically significant psychophysical laws, few make one’s desires be implemented in action. Of those that do, few result in beings having desires to talk about consciousness—there could be other mental states that don’t desire talking about consciousness. So this is everyone’s problem!

5. The naturalism+ solutions—natural teleology, axiarchism, and limited God—also can explain this perfectly adequately.

Consciousness knowledge

One objection to epiphenomenalism that even its proponents seem to regard as the biggest problem for the view is that it makes it hard to see how we could know about consciousness. If consciousness doesn’t cause anything in the brain, how could we come to know about it?

Lots of people make a big deal about this objection. Chalmers dedicated an entire chapter of his book almost exclusively to this argument. But I’ve never been very moved by it, and I think Helen Yetter-Chappell develops a pretty knock-down objection to it. The argument is well summarized in that paper by Yetter-Chappell:

1. If epiphenomenalism is true, I have a zombie twin (z-twin) in another possible world.

2. As she scratches at her leg, the very same things go on inside of her brain as inside of my brain. These brain processes cause the very same sounds “Itchiness feels like this” to come out of her mouth as come out of my mouth.

3. So my z-twin has the very same phenomenal belief as me—“Itchiness feels like this”—formed by the very same mechanism.

4. But my z-twin’s belief is not only false, it’s not justified.

5. If her belief is not justified, and my belief was formed by the same mechanism, then my beliefs can’t be justified either.

6. So—if epiphenomenalism is true—my phenomenal belief isn’t justified.

7. But it clearly is justified.

8. So epiphenomenalism is false.

Yetter-Chappell goes on to argue that we should reject 3—the zombies don’t have the same beliefs. If we’re dualists, we should think that part of what determines your belief is your conscious state—you might have a belief if you have a particular brain state combined with a particular kind of conscious experience. But if that’s true, then zombies don’t have the same beliefs about consciousness as I do. She notes:

(i): If consciousness is essential to the phenomenal beliefs I have, then my z-twin does not have the same phenomenal beliefs as me, since she doesn’t have any such experiences. My z-twin’s brain and vocal cords may do the same things as mine, but unless we embrace a physicalistic conception of phenomenal beliefs that renders them incapable of coming apart from these physical processes, there is no reason to think that my z-twin’s phenomenal beliefs must also be the same as mine. And the dualist has no reason to embrace such a physicalistic conception of phenomenal beliefs. (Chalmers, 2003) (More on this to follow in Sect. 3.)

(ii): There is likewise no reason for dualists to accept the claim that our beliefs were formed by the same mechanisms. If the sole mechanism at work in forming my beliefs was the physical workings of my brain, it would follow that my z-twin’s beliefs are formed by the very same mechanism. But—insofar as the dualist holds that (non-physical) phenomenal experiences are essential constituents of phenomenal beliefs—my phenomenal beliefs are formed not only by the physical workings of my brain, but by the bridging laws responsible for generating these phenomenal experiences. No such mechanism is at work for my z-twin.

The original paper develops this view in greater detail. But the basic point is that if we’re dualists, we should adopt a dualist conception of belief, according to which beliefs depend in some way on your experiences. If so, we should think zombies don’t have the same beliefs as we do even if they behave identically.

There are various other responses that can be given, summarized well here. I think collectively these are enough to almost entirely eliminate the force of the objection. Lots of epiphenomenalists think this is the best objection to epiphenomenalism, but I find it totally unconvincing. In contrast, I think explaining consciousness talk is more challenging.

Conclusion

Epiphenomenalism sounds like a totally crazy view. When I first heard it, I thought it was wholly unreasonable. But the more you think about it, the easier it is to accept. Once we’re nonphysicalists we already accept that consciousness is not necessarily causal. And once we reach that conclusion, we have no reason to think consciousness is causal in the actual world.

There are various lines of converging empirical evidence for epiphenomenalism. Nothing in neuroscience suggests violations of causal closure of the physical, as we’d expect if epiphenomenalism were false. And interactive laws aren’t necessary to explain anything. I think if I had to defend some view about consciousness in a public debate, I’d probably pick epiphenomenalism.

That said, even after I’ve said all this, the view seems a bit crazy. It’s super hard to believe consciousness does nothing. The moves that one makes defending it, while internally consistent, strike me as something like cheating. It reminds me somewhat of David Lewis’s modal realism—it’s not obvious why it’s wrong, but it feels quite like you’re being duped. However, the more I think about it, and try to imagine things from an epiphenomenalists perspective, the more plausible it starts to sound.

I end up not fully on team epiphenomenalism, but pretty sympathetic.