Mary's Room does not refute physicalism

Contra Bentham's Newsletter

This is a response to Bentham’s Newsletter’s recent post, Mary’s Room Refutes Physicalism.

I don’t think people should take Mary’s room or similar thought experiments to tell us much, if anything, about what the world is like. Mary’s room is supposed to somehow threaten physicalism: we’re supposed to imagine that Mary knows all the physical facts about the color red while having never seen red. Then we’re supposed to imagine she sees red, and we’re asked: does she learn something new? If the answer is “yes,” then, simply put, she’s learned some nonphysical fact, and therefore physicalism is false.

Physicalists may respond by arguing that she doesn’t learn anything new, or they may argue that whatever she learns is consistent with physicalism. I don’t think physicalists (or anyone else) should take either approach. I think we should just reject the notion that this thought experiment is a useful way of learning about whether physicalism is true or not. Simply put: how we’re disposed to respond to the thought experiment may tell us something about our psychology (our beliefs, dispositions, the way we’re inclined to use sentences, etc.), but I see little reason to think it tells us anything about what the world is actually like.

We’re tasked with imagining a person knowing all the physical facts about seeing red. Well, what are those facts? What would a person be like if they knew all those facts? I don’t know, and I don’t think anyone else does. So how are people supposed to imagine this scenario, if they don’t even know what it would entail? Are we supposed to imagine something like “well, whatever those facts are, they wouldn’t be enough to allow her to know what it’s like to see red without seeing it”? If so, why think we’re in a position to make reliable inferences of this kind?

Simply put, whether Mary learns anything when seeing red isn’t something we’re in a position to reach any reasonable conclusions about simply by thinking about it. What are we supposed to do? Simulate the situation she’s in and then “observe” the outcome? We can’t do that. Whether one thinks she learns something or not tells us something about your inclinations to reach one or another conclusion. It’s not at all clear it tells us anything about what the world is like.

Recently, Matthew over at Bentham’s Newsletter has suggested that the Mary’s Room scenario somehow disproves physicalism. It does not. And it does not because the thought experiment isn’t a legitimate method for acquiring information about the status of physicalism in the first place. Some people may report having “intuitions” in response to the thought experiment that, if they tracked the truth, would track truths inconsistent with physicalism. But there is little reason suppose that people have such truth-tracking faculties and that those who have anti-physicalist intuitions in particular are tracking the truth in this particular case. The conclusions one reaches in responding to it can and do vary: merely because some people think Mary would learn something new, and that what she learns would be non-physical, on consideration of the thought experiment, it doesn’t follow that these people are correct in thinking this. This conclusion is based on those people’s intuitions on the matter. And other people, myself included, simply don’t share those intuitions. Why should we privilege their intuitions over our own? I don’t think we should. At best, Mary’s Room shows that some people are inclined on considering the scenario to think that Mary learns new non-physical facts, and others aren’t. This tells us something about what those people think, but why should we think how people react to this scenario tells us about reality?

Matthew says:

When you present this argument to normal physicalists, you almost never get anyone [sic?]. I have, on at least 1 billion occasions, at various Rationalist meet-ups given the argument, only for my interlocutor to flail around soliloquizing rather than clearly enumerate [sic] a premise they reject. 

Maybe Matthew shouldn’t expect people at rationalist meet-ups to be prepared to have discussions in analytic philosophical terms. Do physicalist philosophers flail around and soliloquize? Matthew should talk to some of them and report back. In any case, here’s Matthew’s argument for non-physicalism:

“Mary in her black and white room can, merely by reading textbooks, learn everything physical.

Mary in her black and white room cannot, merely by reading textbooks, learn everything about consciousness (she cannot learn what red looks like).

Therefore, some things about consciousness are non-physicali [sic].”

This is trivially easy for a physicalist to respond to: They can simply reject (2). As for myself, I’m not a physicalist, and my goal isn’t to defend physicalism but to object to the methods Matthew uses to arrive at conclusions: thinking that one’s personal intuitions about thought experiments is somehow dispositive about grand questions of metaphysics. 

In this case, we’re supposed to just affirm or deny premises. But premises are often the tip of a much deeper iceberg of potential meaning that is concealed when presented with the premise alone. Absent further specification, premises are often simply too underdescribed for it to be reasonable to demand an affirmation or denial. What is involved in knowing everything physical? What is involved in knowing everything about consciousness? Everything? Can human brains even store that information? Is Mary a human? What kind of person are we supposed to be imagining here?

Matthew says:

I think 1 is on extremely firm footing, and it’s very widely accepted.

I don’t accept it. I’d need to know what is involved in knowing “everything physical.”

Matthew continues:

Mary could learn every fact about chemistry, biology, physics, and so on simply by reading the textbooks.

I don’t think she could. Normal human brains probably don’t have perfect retention capacities. Is Mary a normal human or am I supposed to imagine some kind of cognitive superbeing with perfect recall? It’d be one thing to imagine a normal human discovering something new when seeing red for the first time, but such a person would perhaps not have known all the physical facts. If knowing all the physical facts requires being utterly unlike anyone I’ve met, some kind of galaxy-brained posthuman cyborg comprised of city-scapes of supercomputers,  how am I supposed to reliably imagine what things would be like for such a person? I don’t think I can, and I don’t think anyone else can.

Physical facts can be exhaustively described in terms of physics and in terms of behavior—they’re the sorts of facts that you can learn about just from a textbook.

No, I don’t think they are. Could a person learn all the physical facts about the precise motion and function of every living organism on earth? No. Human brains don’t have the cognitive bandwidth to do so. Textbooks don’t and probably can’t practically convey all the physical facts. They may present models or theories about color perception, the electromagnetic spectrum, and so on, but such information is often general. I’m not sure physical facts could, practically speaking, be “exhaustively described.” What would exhaustively describing the physical facts of a single person walking to the mailbox look like? I don’t know. What if it involved equations that could fill a thousand encyclopedias? To model “seeing red,” we’d need to know a whole bunch of information about brain processes, and how the brain processes visual stimuli, and a whole host of information that would presumably be vast, technical, and intricate, and perhaps require a scale of knowledge retention that is unavailable to human brains.

Am I overstating how much knowledge Mary would need to have to know “everything”? I don’t know! That’s an empirical question! How are random philosophers who don’t study cognition or visual perception supposed to just simulate these facts in their minds if they don’t know the relevant details of the facts themselves?

Regarding the second premise: I see no reason why a physicalist, or anyone else, should affirm this. Whether Mary could learn “everything about consciousness” from reading books is an open empirical question I don’t know the answer to. And just what is it she’s supposed to be learning about? Qualia? Phenomenal states? Clearly she can’t learn from a textbook about phenomena that by stipulation are such that their content couldn’t be fully conveyed in a textbook. So what, exactly, are we talking about when talking about consciousness in this premise? What does “consciousness” mean here? Phenomenal consciousness? If so, then the premise presupposes the very thing a physicalist would be denying. Do we mean something else? If so, what? If it’s “consciousness” of the form Dennett endorses, then physicalism is no barrier to knowing all about it (even if our cognitive limitations were).

Matthew asks:

If Mary could learn every other physical fact just by reading from textbooks, why couldn’t she learn this one?

Supposing she could learn all the other facts from reading textbooks, I don’t see any reason to think she couldn’t also learn facts about consciousness.

Everything in the above paragraph is correct. Physicalism doesn’t entail that Mary would ever see red while in the room. It implies something much stranger: that Mary, while in the room, would be able to learn what it’s like to see red.

“What it’s like” is the problem here. What does that mean? This may be a Trojan conceptual horse smuggling in precisely those preconceptions physicalism couldn’t handle by stipulation, then declaring victory when it inevitably fails. If the notion of “what it’s like” to see red just is the notion of what-its-likeness central to phenomenal consciousness, then Mary might very well be unable to know what it’s like to see red before she leaves the room, but that’s because there isn’t anything it’s like to see red at all, and so she wouldn’t learn what it’s like to see red in this respect after she leaves the room, either.  

In other words, physicalists don’t have to grant that Mary would learn what it’s like to see red at any point, even after seeing red. We shouldn’t grant that “what it’s like” conveys any clear and meaningful concept in the first place. It needs to be disambiguated. If “what it’s like” is cashed out in effable, relational terms, Mary could perhaps come to know what it’s like to see red without seeing it, or perhaps she couldn’t for reasons that are no threat to physicalism. But if the notion of “what it’s like” smuggles in the mysterious, ineffable, intrinsic, private experiential qualities phenomenal states are supposed to have: well, of course Mary doesn’t learn what it’s like before she leaves the room…but she also doesn’t learn what it’s like once she does leave the room, because there isn’t anything it’s like in this respect to see red at all. What is the critic of physicalism going to do? Scoff and say “So you’re saying she doesn’t know what it’s like to see red even after she sees it?”

Yes. That’s exactly what I think a physicalist should say, and they should respond with more than a little anger if the anti-physicalist scoffs at this as “obviously absurd” or something equivalent. If one bakes anti-physicalist notions into the very questions one is posing, frames them in colloquial language, and then asks a person to deny the colloquial phrasing, this is just another instance of equivocating between the technical formulation of “what it’s like” and colloquial notions that carry pragmatic connotations the physicalist isn’t denying. Exploiting an audience’s conflation between the two can make it look like the physicalist is denying things they aren’t denying. This technical/colloquial exploitation is a common feature in the analytic philosophy toolkit: it’s rhetoric that employs false and misleading implications about rival positions under the guise of simply pumping intuitions and evaluating plausibility. As always, analytic philosophy’s primary arsenal relies on confusions about language.

Matthew goes on to talk about standard types of physicalist responses. Roughly, these involve claiming that various sorts of notable things happen when Mary sees red for the first time other than what’s at stake: learning what it’s like to see red. This is mocked with the following meme:

For instance, once Mary is acquainted with red objects, she can sort them from other sorts of objects. As such, she acquires a new skill. And we may say she’s learned something or acquired knowledge of new facts. But, allegedly, something is missing from this picture:

Yes Mary is acquainted with a new fact. But it seems that on top of that, she learns something new.

Once again, Matthew makes the same move as in almost every other argument: he simply makes an assertion about how “it seems.” It seems to who? To Matthew? I cannot overstate how much I don’t care how things seem to Matthew. It seems to some people that they were abducted by aliens, or that they can see the future, or that auras exist, or that Drake’s music is good. People believe all kinds of things and report all kinds of experiences. I simply don’t care (aside some near-negligible Bayesian amount) how things seem to Matthew. I don’t consider Matthew any sort of authority or reliable guide on this or pretty much any other topic of interest. 

As far as it seeming this way to a bunch of philosophers, I care a bit more, but still not very much. Historically, part of the problem with the way academic philosophers have responded to these sorts of claims is to just go along with them, conceding and perhaps sincerely reporting that it does “seem like” whatever the proponent of the view in question says; in this case, that Mary “learns something new.” Physicalists are perhaps too quick to grant this. It doesn’t seem to me that Mary “learns something new.” Physicalists should reconsider whether it really “seems” this way to them. And if it seems this way to Matthew and others, well, why should I care? It doesn’t seem that way to me.

At least part of the problem here is a language issue: we may find ourselves wishing to affirm our confidence that if Mary left the room and saw something red, she’d react in some way, e.g., perhaps exclaiming “Oh, so that’s what it looks like!” We may then wish to affirm the sentence “she learns something new” to account for this anticipated behavior. But, firstly, why think our fantasies of how Mary would react tell us anything other than facts about our own psychology? And second, even if she would say this, there’d still be an open question as to what this means. Why affirm that she’d “learn something new,” i.e., assign “TRUE” to this sentence, then scramble to explain what that amounts to? Why are you assigning truth to sentences prior to knowing what it would mean to do so? The scenario we’re in once again seems like one where philosophers have a bizarre view about the relation between language (or, more aptly, decontextualized toy sentences, not even actual instances of language use) and reality.

Next, Matthew says:

It’s common for physicalists to simply bite the bullet and say that Mary could know everything it’s like just by being in her black and white room. I find that view utterly crazy. If this is right, then if we had a good enough memory, just by reading about octopus brains, eventually we’d know what it’s like to be them and to see the exotic colors that they see but we cannot.  That Mary learns when she leaves the room is about as obvious as anything could be.

I really dislike the way philosophers speak of others “biting the bullet.” This frames someone taking a particular stance as though it’s some kind of concession. But one can not only accept a position but also deny that by doing so that one is “biting a bullet” by denying that accepting that position is a concession. Physicalists are more than entitled to do so in this case. Sure, maybe Mary could learn everything it’s like to see red prior to seeing it. Why is this “biting the bullet”? Why should I grant that it is? Matthew says “I find that view utterly crazy.” Well, so what? I think Matthew’s views are ridiculous. Does that mean Matthew is biting the bullet when holding those views? Wow, look at the  bullets Matthew is biting when he denies physicalism! Talk of “biting bullets” only works against a backdrop of intersubjective agreement about which positions enjoy greater initial plausibility. Since I don’t grant that there’s anything even remotely plausible about Matthew’s views or methods on this many other closely-related topics, I might see concessions towards Matthew’s position as “biting bullets,” in that one is accepting as true various positions that seem really implausible to me. But this perspective isn’t shared, so my interlocutors have no reason to grant that in denying physicalism or affirming weird views about qualia or phenomenal states, that they’re “biting the bullet.” 

Regarding the latter remark about knowing what it’s like to see colors that an Octopus could see: supposing one really knew everything about octopus brains, what’s so ridiculous about this? What that would involve turns on precisely what it would be like to know everything. Perhaps that would require textbooks longer than the diameter of our solar system. Perhaps it’d require a brain the size of the sun. If we don’t actually know or have a means of estimating how much information knowing “everything” would involve, how are we supposed to accurately imagine Mary knowing “everything”? This could very well turn out to require such wildly science-fictionesque considerations that I see little reason to think our intuitions are up to the task of properly modeling these scenarios. If knowing “everything” would require a superhuman megamind of vast proportions, am I to believe Matthew can accurately simulate what seeing red is like for such a being and generate a truth-tracking output about whether this superhuman megamind would be able to know what it’s like to see as an octopus? Why should I think Matthew or anyone else has such powerful imaginative capacities?

Even if the scenario were far more modest than this, and some advanced supercomputer of terrestrial origin can do it a decade from now, that still doesn’t vindicate the imaginative capacities of Matthew or any living philosopher. I suspect Matthew and these philosophers are imagining a human (though a very impressive one) reading books about color vision and octopuses and going “Aha! So that’s what it’s like!” This is implausible. But it’s implausible because textbooks and human brains aren’t good candidates for modeling octopus visual systems: we may simply lack the raw hardware or adequate software configurations to do so. Most people aren’t even that good at engaging in detailed visual imagery of a table or a horse. Some have aphantasia. I couldn’t even come close to drawing my own face from memory. For comparison: a very old computer couldn’t adequately run a modded up version of Cyberpunk 2077. I’d hardly take this to be evidence that physicalism is false. Just so, that textbooks and human brains can’t model visual systems of other organisms may just tell us something about the limits of textbooks and human brains, rather than reveal that some things aren’t physical.

We are only familiar with normal human minds and normal means of information conveyance. There is no good reason to think either is adequate for simulating the visual systems of other entities or whatever else would be required, intellectually, to learn “everything” there is to know about what “it’s like” for those beings to see, whatever that’s even supposed to mean in the first place. Such thought experiments must be run on our impoverished human minds. Our minds may simply lack the software and hardware capacity to properly “run” the thought experiment in question in a useful way. So we have a double-layer of imaginative barriers: first, we don’t know the requirements for knowing “everything” about seeing red: they may require capacities we aren’t familiar with and couldn’t readily imagine. Second, even if we knew what those criteria were, we may still face personal limitations on our ability to accurately model them and form reliable inferences about them. For comparison, most people can’t do very advanced mental math or play chess in their minds. Yet we’re supposed to believe we can accurately model Mary knowing everything about the color red from reading textbooks? This is pure methodological fancy on the part of philosophers.

Finally, note the last remark in this paragraph: “That Mary learns when she leaves the room is about as obvious as anything could be.”

It’s not obvious to me, and I don’t grant that it’s as obvious as anything could be. Perhaps the problem with these sorts of arguments is that they turn so heavily on the diagnosticity of thought experiments like Mary’s Room…, when such thought experiments are simply not useful tools for determining what reality is like. The approach I’d recommend for dealing with Mary’s Room is to question whether the thought experiment can tell us much of anything about whether physicalism is true or false. I doubt how philosophers are disposed to respond to Mary’s Room tells us much of anything about metaphysics or consciousness. At most, I think it tells us something about our own preexisting commitments and inclinations. Perhaps it's useful as a tool of self-discovery, but I seriously doubt it has any value as a tool for discovering anything about the world outside our minds.

Note that I am not the first or only person to raise similar objections. In the comments to Matthew’s post, people raise similar points, often putting them quite succinctly. Here’s one from Anlam Kuyusu:

"Mary in her black and white room can, merely by reading textbooks, learn everything physical. "

Can Mary also learn how to ride a bike merely by reading textbooks? If she can't learn how to bike, does that mean biking is not physical? (I mean regular biking - nothing like ET-style in the sky, magical biking.)

What if Mary could devise an apparatus like those used in Matrix (see the way Neo learns Kung-fu) and download the color red into her brain? Would that work?

I keep coming back to this premise: "Mary in her black and white room can, merely by reading textbooks, learn everything physical." It seems so demanding that it makes physicalism impossible. Because of course there are always things you can't just learn by reading textbooks.

This is a fantastic comment. It highlights that much of what’s wrong with Matthew’s appeal to the thought experiment is that it sets up impossible and absurd standards, setting the bar a physicalist much allegedly pass so high that they couldn’t reasonably pass it. The problem here isn’t physicalism. It’s the thought experiment. Matthew seems enamored with the a prioristic armchair tools of the analytic philosopher. This is unfortunate, because these tools are, at best, extremely limited. And, at worst, they are actively misleading, prompting philosophers to devise theories built on the phantasmagoria of imagined worlds. Philosophy would make far more progress if it were more grounded and more integrated with the empirical sciences.