Physicalism
The relationship between the mind and the brain is hotly contested among philosophers, but there is an unspoken consensus among neurologists that we should always make clinical decisions on the ward under the assumption that the mind is solely due to the physical actions of the brain. As an everyday working hypothesis, it is useful.
In my experience, neurologists who doubt this hypothesis keep it to themselves, and so the hypothesis does not even need a name. Over many years on the neurology wards, I have not heard the Hard Problem mentioned once on hospital grounds. I have not been asked my philosophical stance on consciousness in any clinical setting —except, very rarely, by patients.
Among philosophers, the view taken for granted by neurologists is usually known as physicalism, and although it is the majority view among professional philosophers, opinions vary widely.
The precise definition of “physicalism” can be debated, but, in the context of consciousness, it is simply the mainstream scientific view that there are no special, fundamental properties of reality specifically responsible for mentality; everything we find on introspection is ultimately a natural part of the physical world, which is more-or-less accurately described by the physical sciences. Or accurately enough. The project of physics is clearly ongoing, but, even in its current, incomplete form, it already provides an adequate underpinning for biochemistry, which supports cellular biology, and neurophysiology, and so on. Cognitive neuroscientists don’t need to dig down to those fine layers to understand the function of neurons, and the functions of neurons are all we need to account for consciousness, which, for most physicalists, is seen as an evolved property of brains.
It is commonly stated by anti-physicalists that physicalism is a metaphysical hypothesis, and it certainly can be, but physicalism with respect to consciousness should not be confused with more fundamental questions about the nature of reality itself.
All we are doing, when we adopt a physicalist stance with respect to consciousness, is excluding consciousness from the list of the base ingredients of reality; we are not specifying what the other ingredients are and how they came to exist. The question of what puts reality into the laws of physics is a substantial mystery, of course, but it is not one that needs to be addressed to adopt a physicalist view of consciousness, and it is not one I will attempt to answer anywhere in this blog.
My answer to the deeper metaphysical backing of physicalism is a simple one: I don’t know why anything exists. (For what it’s worth, ontic structural realism looks like a plausible stance for now, but I could be convinced to change my views. If you don’t know what structural realism is, it doesn’t matter for the purposes of the current discussion.)
If you ask a gastroenterologist what puts reality into the fundamental particles of the intestine, they don’t know either, but we don’t often accuse them of not having valid opinions on the gut. That’s because no one is suggesting that digestion is a base ingredient of the universe.
And, of course, we don’t model out metaphysics with our gut.
If I meet someone who believes that ChatGPT is conscious, and I argue that its actions and its opinions are all just aspects of some physical machine, then they can’t refute my position by pointing out that I don’t know what’s inside the electrons that make ChatGPT work, so I must remain agnostic. We might not know exactly what makes up the physical stuff of ChatGPT, but we do know that ChatGPT is not generating its answers within a ghostly non-physical mind, and that’s what counts. There are plenty of mysteries below the engineering level of ChatGPT, but they are out of scope for questions about whether ChatGPT is conscious — right up until some anti-physicalists drag in the idea of some deeper fundamental consciousness lurking inside the machine.
Note that physicalism, defined this way, does not necessarily entail a simple identity claim that equates consciousness with soggy grey matter (and even if there are physicalists who might accept this view, I would not support such a position).
In the series of posts ahead, I will argue for a variant of physicalism that I call virtualism. It is closely allied to illusionism, which is often presented as the claim that the consciousness we find on introspection does not exist, but virtualism manages to avoid some of the contentious language that turns people away from illusionism. In some ways, virtualism could be considered a subspecies of illusionism; I will consider the differences later.
Consciousness exists, I will argue — but not in the form we might have expected from our initial impression of consciousness; reality does not directly match the impression we formed from our own privileged, subjective perspective, which resides within a representational system.
That means, between physical reality and how consciousness seems to us, there is a striking contrast, and we face difficulty reconciling the two sides of that contrast.
The Explanatory Gap
According to virtualism and related physicalist positions, everything we find on introspection can be attributed to cognitive representations that are sustained by natural physical processes within the brain.
It will take some work in the posts ahead to show that virtualism is plausible, but for now we can consider what sort of cognitive fallout we might expect if a virtualist situation pertained.
A natural consequence of seeing the mind in representational terms is that the indirect, interpreted relationship between mind and brain sets us up for confusion. Nothing we find in our minds necessarily has the properties it seems to have. In fact, the representational relationships that make thought possible guarantee that there will be conflict between our neural substrate and our cognitive engagement with what that substrate represents.
For instance, if we think about what red looks like, our neural structures represent the property common to the fire-trucks and red apples of our childhood, but the neural structures doing the representing don’t themselves have that redness property. (Neither do fire-trucks or apples, as far as we can tell, but that’s a discussion for another day.)
If we think about what consciousness itself seems like, via a process of introspection, I think we find something that doesn’t readily map to anything outside our heads. It is entirely virtual. Even what we seem to find in our heads is hard to pinpoint. Consciousness is often depicted in images as an inner light or as a silvery fog, but all pictorial representations are hopelessly inadequate.
I am confident, though, that you, as a conscious reader, know what consciousness is like for you; you’re familiar with how it seems from inside the representational system of your cognition. It’s not like a dark silent void, for instance. It’s not much like a silvery cloud. It’s definitely not like anything you might initially expect in the world described by physicists. (Personally, I think consciousness itself is an attention schema, but — you guessed it — that’s a discussion for another day.)
When we introspect, our cognitive representations are usually taken at face value, but this is not necessarily a stance that science should adopt. If we accept our own internal representations as necessarily accurate in how they portray reality, but we impose different standards for scientists, the mind will seem to possess all sorts of properties that are in conflict with a straightforward physical account of the contents of our skull. We are using different rules in the asking and the answering of our questions.
Accordingly, it will be difficult to make sense of the mind-brain relationship.
That means, if we think of what we find on introspection as “non-physical”, we will end up being half right — or, at least, we will end up wrong in a way that is defensible, given the inadequacies of language. Moreover, the contrast between the brain’s representations and the representational substrate is expected.
But then there’s the Explanatory Gap, which involves more than simple acknowledgement of a mind-brain contrast.
A major problem with physicalism, it is often said, is that we encounter a mysterious shortfall whenever we try to explain phenomenal properties: functional theories of the physical brain seem destined to leave out all the subjective properties of the mind. The stubborn persistence of the Gap — including the anticipated failure of science to close the Gap as neuroscience progresses in future — is sometimes taken as confirmation that physicalism is false.
“There is an explanatory gap (a term due to Levine, 1983) between the functions and experience, and we need an explanatory bridge to cross it. A mere account of the functions stays on one side of the gap, so the materials for the bridge must be found elsewhere.” (Chalmers, 1995)
Conversely, people convinced of the truth of physicalism sometimes counter that the Gap is only temporary — or at least it is too early to tell if we will ever close it. Life itself once seemed incompatible with an entirely physical perspective, and we faced an explanatory gap for Life, but Life is now largely explicable in terms of underlying physical processes. Maybe we can be similarly optimistic about the Gap for subjective properties, such as qualia.
I have some sympathy for this view. If I were forced to choose, it seems more likely to me that we simply don’t yet know how to explain qualia (or that we have muddled concepts) than that science has got it wrong and we need materials for a bridge. Other gaps have indeed been closed. If we had to bet among these two opposing choices, we might be wise to put our money on science closing the Qualia-Neuron Gap, just like it closed the Life-Matter Gap and so many others.
In the next few posts, though, I will be arguing that we don’t have to choose between these two positions; we can take a more nuanced middle view that falls between them.
We can be Gap Compatibilists.
For a start, if the properties we find on introspection are merely represented, as proposed by virtualism, then we wouldn’t necessarily expect science to take the same view as our own cognition. We would expect that views from within a cognitive system and views from outside that cognitive system would be in stark contrast.
Consider, for instance, a less mysterious representational system: the hardware of a computer and the dragon being represented in some game running on that computer. The hardware and the dragon could not be more different, but we usually accept that they are actually reflecting the same patch of reality; behind the different modes of interpretation, they are the very same thing, provided that we remember to think of the dragon as merely represented.
But here the anti-physicalist can point to the Explanatory Gap and ask:
Why can’t science, in the pursuit of understanding, simply adopt the same representational perspective that the brain is adopting? Why does science, even when it is armed with representational theories of cognition, seem destined to leave out the subjective flavours of cognition?
Here is where a Two-Brain Approach can be useful.
The Subject and the Scientist
The Explanatory Gap is based on the idea that functional theories of cognition leave out qualia, and the Hard Problem is based on the idea that qualia might plausibly have gone missing from physical brains entirely.
Many people roll these two issues together, and this conflation is a key component of hardism. (For a brief introduction to hardism, see my first post, The Zombie’s Delusion). The Hard Problem is the challenge of trying to cross the Gap, says the hardist. It can’t be done. Functional accounts will inevitably fail. Experience must be non-functional.
This sequence of conclusions is a mistake, I believe, because there are two main ways in which functional theories of the human brain might be considered to “leave out” subjective properties. We need to distinguish between them, because they have radically different implications.
Firstly, as noted in a previous post (Your Twin) some folk think that, in some hypothetical universe with different natural laws, a functional or physical duplicate of a human brain might conceivably lack qualia, and hence might be a zombie. The functional process might take place in the physical domain and literally leave out the experiential elements. I think this is likely to be impossible for reasons that were briefly mention in my first post (The Zombie’s Delusion), but these issues are complex, and the possibility of zombies will need to be explored in later posts.
Secondly, functional theories might leave out subjective properties in the quite different sense that we could study neural circuitry forever and not know what that circuitry was like, from the inside.
If qualia are lacking, in the first sense, this entails that we won’t be able to derive them; we can’t find what is not there.
The reverse is not true. The second sense of leaving out qualia does not necessarily entail the first, more radical sense. We could experience the frustration of not deriving qualia while remaining agnostic about whether the perceptual state was like anything for that circuitry.
If I studied your brain, for instance, I might be unable to confirm what some subjective experience was like for you, while nonetheless accepting that it probably is like something for you — I am happy to accept that you are not a zombie.
I could be colour-blind, for instance, while you were not, such that both of us had fundamentally different colour concepts. In that setting, I could be quite convinced that your red is very different from your green, but I might be unable to understand this difference as you do, because I can’t know what red is like for you. I could study your circuits and fail to derive any sense of what redness was like, for you, because I lack cognitive access to the relevant concepts.
In this example, your brain and mine have fundamentally different access to your perspective on colour. As far as you are concerned, you seem to get your insights into your own mind directly, but I am forced to study your neural circuits from the outside, and they can’t give me the insight I’m seeking.
If you were a zombie, that would a be a you-problem.
If I can’t derive qualia from your circuits, though, that’s only a me-problem.
Suppose that you and I sit across from each other in the laboratory; in this scenario, you are the notional Subject, and I am the Scientist. We can consider the possibility of neural circuits “leaving out” qualia in Subject-focussed terms (considering your subjective access to the sought-after concept) or in Scientist-focussed terms (considering my objective access).
These are fundamentally different types of “leaving out”, because your circuits are literally in your head, but they are only represented in mine. They represent redness in your head, but, as mere representations, they don’t achieve much in mine.
Subject-Focussed Leaving-Out = Zombiehood
If your neural circuits leave out subjective properties for you, then you are a zombie. (If you behave exactly the same without your qualia, because your brain maintains behaviour, then this implies that qualia in non-zombies like me are epiphenomenal, meaning that they are outside the causal nexus of cognition and have no effects on the physical world). This state of affairs would break physicalism.
Scientist-Focussed Leaving-Out = Explanatory Frustration
If your neural circuits, studied by me, do not allow me to derive what your qualia are like, in my own head, then I have an Explanatory Gap (and I would describe this by saying that your qualia seem irreducible to me.) This state of affairs would not break physicalism; it would just be frustrating for me, making it harder for me to understand the nature of qualia. You still have your qualia; I just can’t reproduce them in my own head to know what they are like.
Whether qualia can ever go missing in the Subject-focussed sense is a philosophical puzzle that cannot be directly addressed with the methods of science. Zombiehood is undetectable by any empirical test, by definition.
Whether my brain can read your circuit diagrams and reach the subjective state of knowing what red looks like also includes several philosophical issues, but among them, there is a well-defined empirical question.
If I am blind, or colour-blind, or locked in a black-and-white laboratory, can my brain get into a state that is functionally comparable to yours, through any deductive process, simply by considering your neural circuits?
I will have quite a bit to say about the Subject-focussed form of “leaving out” qualia later on, when I discuss zombies, but for the remainder of this post I just want to consider the explanatory frustration at the Scientist end of this puzzle, starting with the most famous qualia scientist of them all.
First there was Mary
Most people interested in these issues are familiar with Mary the Colour Scientist.
Mary is a character invented by the philosopher Frank Jackson; she stars in an anti-physicalist thought experiment known as the Knowledge Argument, which was initially intended to discredit physicalism by showing that it is subject to an Explanatory Gap.
Here is the usual brief summary of her tale, which regular readers should skip:
Mary is a brilliant scientist who is, for whatever reason, forced to investigate the world from a black and white room via a black and white television monitor. She specializes in the neurophysiology of vision and acquires, let us suppose, all the physical information there is to obtain about what goes on when we see ripe tomatoes, or the sky, and use terms like “red”, “blue”, and so on. She discovers, for example, just which wavelength combinations from the sky stimulate the retina, and exactly how this produces via the central nervous system the contraction of the vocal cords and expulsion of air from the lungs that results in the uttering of the sentence “The sky is blue”. [...]
What will happen when Mary is released from her black and white room or is given a color television monitor? Will she learn anything or not? It seems just obvious that she will learn something about the world and our visual experience of it. But then it is inescapable that her previous knowledge was incomplete. But she had all the physical information. Ergo there is more to have than that, and Physicalism is false. (Jackson, 1982)
As a physicalist, I don’t agree with Jackson’s conclusion — and, as it turns out, neither does Jackson. He received so many rebuttals to his argument in the years following its publication that he was forced to think through the issues more carefully, and he eventually changed his mind, becoming a physicalist.
Like me and many other physicalists, Jackson now sometimes explains Mary’s journey in representational terms. While in her lab, Mary could not represent redness to herself in the usual way available to most people; then she was released, and she could. The reasons for the change are physical.
In later posts, I will consider some of the rebuttals and counter-arguments that have been raised in response to Jackson’s provocative tale, as part of a broader exploration of physicalist theories of consciousness. I will be arguing that physicalism can accommodate gaps in knowledge of just the sort that Mary demonstrates, and my account of Mary’s frustration is not far from the representational view that Jackson eventually adopted.
Indeed, I’ll have quite a bit to say about Mary in later posts, because I think she forces us to address an important question:
What can a Scientist reasonably be expected to achieve, armed with black-and-white circuit diagrams?
In other words: What should we expect from a theory of consciousness?
I believe that we have been expecting too much, and it has led to unnecessary confusion.
Mary’s story embodies the plausible intuition that subjective mental properties, like colours, tastes and pains cannot be derived from neural circuit diagrams.
Not everyone accepts that this intuition has any merit, but I am one of those who does accept it. Some of the disagreement relates to confusion about what it really means for something to be non-derivable. It is not even clear that subjective redness is a true property of any real thing, so it is not clear what it would take to know anything certain about that elusive property. Deriving the quale for redness is often thought to entail tapping in to some non-physical aspect about which we could have no grounds for any empirical conclusion one way or the other. We can’t even distinguish zombies from non-zombies, for instance.
All our thoughts on this issue are a muddle, and that muddle can get in the way of objective questions.
In many discussions, the philosophical question of what we should make of the non-derivability of qualia and whether zombies are possible gets mixed up with the completely empirical question of which brain states can be reached with which inputs.
Just because the first sort of question is murky should not stop us from considering the second sort of question.
In other words, I think it is useful to take Jackson’s lead and operationalise the question of what it means to derive qualia, specifying a defined cognitive journey in the brain of a Scientist.
In future posts, I will talk about Jacksonian derivation, which I briefly introduced in an earlier post, as follows:
To perform a Jacksonian derivation of a quale, one must start off not knowing it, sit in a room with access to all the relevant discursive facts, think about them very hard, and then acquire the cognitive state that we usually think of as knowing what that quale is like.
Also, one must not use some cheating method that relies on meditation, dreaming, inducing migraines or seizures, or optical illusions that achieve glimpses of colour by having fine gratings of black and white at different spatial frequencies. The process must be a derivation that is essentially analytical in nature.
Jacksonian derivation is a deductive process capable of reaching a good, natural sense of the property of redness, or the taste of cinnamon, or the feel of pain, without ever having had the experience — just by studying the physical substrate of those experiences.
If this is possible, then it is not necessary to have the relevant trained networks on-hand to reach the desired cognitive state; it is enough to read about those networks. If so, the same analytical process can be shared with others who have not had the sensory deprivation, and they can follow it and acquire the cognitive state by a new, deductive route. At that point, they can announce that they have an illuminating understanding of the quale in question. This process of derivation without prior deprivation would also constitute Jacksonian derivation, but in this second context, the prior existence of the relevant trained perceptual networks would make it harder to know if the analytical process had reached its target independently.
The barriers to Jacksonian derivation are complex, so I will approach them in stages — and, within this post, I won’t consider Mary’s barriers in any detail at all.
Instead, I will merely introduce Mary’s lesser known cousin, Rhonda, who faces a somewhat analogous frustration that is not puzzling at all.
(Clearly, because Rhonda’s frustration is not puzzling, her situation is necessarily disanalogous to Mary’s in many important ways, but I believe there are enough factors in common that we can gain insights by studying Rhonda, first, and then switching to Mary later.)
Then there was Rhonda
As a brief teaser for the sort of explanation that might be employed in the face of the non‑derivability of colour qualia, consider a parallel explanatory challenge in a much simpler part of our neural circuitry: the human retina.
Suppose we were asked to create a theory of colour vision in the retina and then apply that theory to the retina.
The general outline of such a theory is already available. Colour in the human retina is detected and subsequently represented via the activation of three different types of retinal neurons, known as cone cells. Red‑sensitive pigments absorb photons of the appropriate wavelength more effectively than other pigments, individual cone cells differ in their dominant pigment, and this difference determines the relative firing rate of different retinal cone cells, launching the entire neural cascade of recognising redness. From there, the colour information travels down the optic nerve and is encoded in neural firing patterns in the dark of the skull.
Suppose that we took this theory of colour-encoding in the retina and we elaborated it extensively with everything that could ever be known about the physical function of the retina. We could then tell a Jacksonian story of Rhonda the Retinal Researcher, who memorises the entire body of physical facts about the retina. Could Rhonda apply this theory to recreate the described firing ratios in her retinas? Could she “think of redness” by activating her retinas in just the right way — not with coloured light or physical probes, but from thought alone — and then observe the result?
If not, why not?
Of course, no one would even attempt this. We intuitively know it would not be possible to command our retinas to fire a certain way, no matter how well we understood their functions. Retinas don’t work like that.
But what would someone have to believe to think that this was a reasonable challenge in the first place? What conceptual errors would they have to commit?
How would you explain to someone why reading about a retinal activation ratio in a black‑and‑white textbook did not provide any means for reproducing the described ratio? Would you argue that this inability reflected poorly on the accuracy of the description in the textbook? Has the textbook left something out?
Because this, in essence, is what the Knowledge Argument attempts to do — albeit in a different region of the nervous system.
One way of explaining Rhonda’s frustration would be to appeal to the fact that the optic nerve is essentially a one‑way street, that the retina is causally upstream from — and not controlled by — the parts of the brain that extract meaning from text.
Another way of explaining the same situation would be to note that trying to force anyone’s retinas to behave in a way that matched the textbook description was a doomed exercise because it completely ignored representational factors. Rhonda would be trying to get her real retinal neurons to copy the merely represented neurons in the textbook, generating a physical example from theory alone. Even if she memorized the textbook, getting all of the relevant information into her nervous system, she would still only be working with merely represented retinal neurons. The faux represented neurons would be represented by real neurons in her occipital and parietal regions, but they are still not retinal neurons; she would have no way to get the faux retinal neurons to influence her real retinal neurons at an anatomically distant location.
Unsurprisingly, the brain doesn’t allow this sort of unconstrained representational level‑leaping, though it does allow some level of top-down control in some systems (so we are already encountering important disanalogies.)
This will be a recurring theme: when considering these issues, we always need to track what we are asking of the cognitive machinery in the Scientist’s Brain. What seems plausible in a non‑physical mind imagined as an ethereal fog of sentience turns out not to be plausible in an actual neural machine with specialised parts and unidirectional pathways. Philosophers love to invoke idealised omniscient demons when considering tricky puzzles; they sometimes refer to this as a priori reasoning. In the study of consciousness, this is ignoring the cognitive modularity of the organ at the centre of the inquiry.
Yet another way of describing Rhonda’s difficulties would be to note that textbook descriptions of retinal activity (and Rhonda’s engagement with those descriptions) provide a clearly inadequate proxy for actual retinal simulation with photons. It is not appropriate to assess the abilities of real retinal neurons by seeing what can be achieved with the proxy.
Rhonda can engage with the description for as long as she wants, and the description could be as detailed and accurate as we might care to specify, and all that accuracy will make no difference. Textbook photons can stimulate textbook retinal neurons, and a sufficiently comprehensive textbook could follow the entire cognitive story into the labyrinthine depths of human cognition, until the textbook brain was claiming to see red and pondering the mysteries of qualia, but none of those pathways would be Rhonda’s; without invasive access to her retinal neurophysiology, she will not achieve what the textbook cognitive system achieves, and we would never expect the information to leap from one cognitive system to another. Moving the entire textbook description into Rhonda’s brain creates a new proxy, but it is equally invalid – she still doesn’t have retinal neurons firing in the sought-after ratios, and the cognitive contemplation of those ratios is not the same thing.
You can see what I will say about Mary: very similar considerations apply in the parts of the brain that usually contemplate redness during idle philosophical speculation. The factors preventing the derivation of redness in the manner explored by Jackson’s Knowledge Argument are much less obvious than the factors preventing cognitive control of retinal cone‑cell activation ratios, but, until the relevant real‑world neural training is in place, the neuro‑anatomical barriers are of the same essential nature.
There are one-way streets, there are barriers to representational level-leaping, and a description is no substitute for the real thing.
Considering the cognitive landscape in which our questions are posed is important whenever we reach a conceptual impasse in any domain, but it is particularly important when it comes to consciousness because hardism — the framework around the Hard Problem — rests entirely on intuitions, not on objective evidence.
By definition, a non-functional experiential property can never have any empirical support; all empirical tests yield the same results on a notional zombie world. What is less often noted is that such a property cannot influence our intuitions, either, or at least not any intuition we could ever report in the physical world. That means something else is driving our intuitions, but, paradoxically, those intuitions are pushing us toward a concept of a functionless entity.
If we have hit an intellectual impasse, and our concepts are paradoxical, then it is at least plausible that our guiding intuitions are flawed and we have taken a misstep, and I will be arguing that this is indeed the case.
In the comments section, feel free to point out the many disanalogies between Rhonda and Mary; they are just as important as the commonalities.
But the most important is this one: whereas Rhonda can never get her retinal neurons to fire in a way that matches her textbook, we can often get our colour-representing neurons to represent colour, through what feels like thought alone. We read the word red, and (unless we have aphantasia) we picture redness. Once we have the appropriate history of colour exposures, we can effectively order up re-enactments from the relevant circuits, generating a representation of redness on demand without requiring coloured retinal inputs, achieving the relevant brain-state from abstract semantic considerations alone.
You’ve probably been doing that as you read.
It almost seems plausible, then, that we should be able to reach the cognitive state of representing redness from black-and-white inputs.
What we’re trying to do is tantalisingly close. And that’s the problem.
In another post, I spoke of Fiona the Font Fanatic, and I made the general point that some fact-like statements rely heavily on examples, such as: “Italics look like this.”
I suggested that statements of this sort do not necessarily express unique truths that cannot be expressed in more circuitous, less satisfactory ways.
I will make a similar claim here. What we are trying to do when we attempt Jacksonian derivation of redness is conjure up an example to use in a self-referential thought.
Redness is like this.
If producing an example of this nature is impossible for physical reasons, or cannot be achieved by a deductive process, but we need those examples to approach the most natural way of expressing what perceptions are like to ourselves, during our attempts at understanding, then we will find it difficult and frustrating to think about these issues. Whereas Fiona could not write certain facts in the most natural way, we will not be able to think certain facts in the most natural way.
If we assign the source of our problem to the Subject’s Brain, we will start inventing fictitious functionless entities.
Post Script
A later post will bring Fiona, Rhonda and Mary together to consider the many ways in which Jackson’s Knowledge Argument fails.
Feel free to comment down below on how that might work — or not.
