This is the draft of a paper that will be published in its fully developed form in my book Lines of Thought: Rethinking Philosophical Assumptions (Cambridge: Cambridge Scholars Publishing, 2014)
ANSWERING THE TWO MAIN OBJECTIONS AGAINST IDENTITY THEORY: MULTIPLE REALIZABILITY AND IRREDUCTIBILITY OF QUALIA
There are two main objections against the identity theory of mind: the objection of multiple realizability and the objection of the irreductibility of qualia. In this paper I will answer both of them, beginning with the objection of multiple realizability.
The concept of type, as distinct from token, may require a greater or lesser degree of similarity among things that are tokens of the same type. For example: two similar round patches of pure ultramarine blue can be regarded as two tokens of the same type, distinguished only by the different spaces they occupy. A round patch of ultramarine blue and a round patch of sky blue can likewise be distinguished by differences in their shades of colour, although they have the same blue colour type. We can clarify this point by saying that types can be more abstract relative to their tokens in ways that tend to be directly proportional to the number of properties that are abstracted, namely, those that are left out of consideration because they aren’t necessarily common to all their tokens.
According to the well-known multiple realisability objection to type-type identity theory, there is no way to identify a mental state with a neurophysiological type, because a mental state of a mental type is in fact realizable in tokens of very distinct physical types. In searching for a reason to dismiss the multiple realisability objection, some philosophers have considered the property of types that they are more or less abstract. What these philosophers have discovered is that the supposed identity between mental and neurophysiological types can be much more abstract than was initially supposed by the advocates of the multiple realisability objection, which allows realisations in many different types of tokens belonging to the domain of the intended more abstract type.
I agree with this position, but I also agree with those who say that this would not be enough. Consequently, my first aim here is to call attention to an insight about the nature of the properties that identify a token as a token of the same type. In the case of the colour blue, we are usually speaking about a certain phenomenally identifiable range of the visible spectrum. In this case, as in the case of a stone, a mountain, a cloud, etc., the identifying properties can be regarded as qualitative and intrinsic, what we could call material properties. However, not all the properties that identify types are material properties in this sense, since as a rule many kinds can only be identified by means of what in contrast may be called functional properties, that is, by what they are able to do. In what follows, I will give three very common examples of functional definitions of types:
(i) Consider the natural kind concept of mammal. There are many different sorts of mammals: human beings, whales and bats are all mammals. How can the concept include such different kinds of living creatures? The reason is that its definition draws much more on functional characteristics than on material ones. Typically, mammals are warm-blooded vertebrate animals that maintain a constant body temperature. They have hair, and females have milk-producing glands used to nourish their infants. There are several other specifically mammalian features, such as a lower jaw directly hinged to the skull, etc. ... However, this definition doesn’t include a precise material determination, for example, of the chemical composition of milk. It is functions (combined with some general material determinations) that make a token a token of the mammalian type.
(ii) Artefacts are good examples of functionally defined things. Consider the concept of knife. One can define a (real) knife as ‘an instrument with a handle and a sharp blade designed for cutting things’. This is essentially a functional definition. In addition to this, a minimal material specification can also be made. When one says that a knife is a solid, hard, very sturdy object, one is delimiting, in a generous way, a domain of possible materials. Real knives can be made of hard materials such as steel, flint, silver, bronze, aluminium, maybe fibreglass, plastic, wood… but not of soft, easily deformed materials such as paper, marzipan, butter, liquids and gases in general. Of course, one could complete the definition with a list of hard-and-not-easily-broken materials that can be used to make knives, a list that would unavoidably remain open-ended.
(iii) Consider now the much-discussed concept of chair – another artefact. In general, this concept can be adequately defined as ‘a seat with a back, designed for one person to sit on’. A stool isn’t a chair because it has no back; a park bench isn’t a chair, because it is wider and designed to seat more than one person. Although mainly functional, the definition of a chair cannot be entirely functional because a chair must be made of materials that are rigid and stable. One can easily imagine a chair made of wood, steel, fibreglass, plastic, etc. But one cannot easily imagine a chair made of chocolate, butter, sea-water (at room temperature) or air (except for inflatable plastic furniture…).
What these cases first demand is a sufficiently precise specification of their function; to this we can add an indeterminate qualitative identification of the material, but not the other way around.
Although there is no hard and fast distinction between material and functional definitions, they can easily be distinguished. One could object that each material property can be further defined in a functional way. The colour blue, for example, is a qualitatively perceptible property, but usually it is scientifically defined in terms of its wave-length, functionally determined by means of instrumentally measured diffraction, etc. Nevertheless, the point is that material properties like the colour blue aren’t in fact accessed in this way, but rather in a way that for us is more direct, by the perception of their qualitative aspects, while the reverse isn’t the case: Functional properties, like ‘being designed to cut things’, cannot be materially defined.
Now, I will show that the neurophysiological side of the identities foreseen by identity theory requires a functional definition, and that this fact has implications for the multiple realisability objection. Restricting our considerations to phenomenal states, my point is that qualia must be seen as constituted not of precise material elements, but rather of more specific neurofunctional elements added to indeterminate, vague or open-ended material conditions.
We can try to defend this proposal by pointing out the deficiencies of hypothetical counterexamples. Consider the identification of the neurophysiological type corresponding to the sensation of pain. This identification has similarities with those presented above. It should be done using a sufficiently precise functional identification added to an indeterminate material identification that refers to biophysical, neurochemical, histological and anatomical commonalities. We already know something about them. We can say something about the relevant centres in the limbic system. We can also say something about the neuronal processes that produce what we call pain. This is the beginning of a neurofunctional identification, but it also brings us to the borders of the unknown. Neuroscience is still in its infancy, and our ignorance of the real workings of the brain tends to be much greater than we are able to imagine. Up to now all that we have really accomplished is to describe the neurophysiology of pain in a manner similar to naïve scientific thinking: A child may try to understand how a steam locomotive works after seeing one, indigenous peoples may try to understand the flight of an airplane that has crashed in their jungle. An inquisitive child might conclude that the forward movement of a locomotive has something to do with men shovelling coal into a firebox and water heated in a boiler, and perhaps even guess correctly that in some way steam pressure causes the wheels to turn... The indigenous peoples could rummage through the smouldering wreckage of the airplane and perhaps guess that it was able to fly because of its engine and wings. In the same way, we can already offer vague, hypothetical and superficial neurophysiological proto-definitions of mental states like pain. One of these could be the following:
Pain sensation (P-Df.) = a kind of pre-cortical response, resulting from the stimulation of peripheral sensory structures called ‘nociceptors’ (which are neuronal cells activated by noxious stimuli and inhibited by analgesic drugs), usually leading to some sorts of cortical activation and appropriate avoidance reactions…
As a definition, this would be plainly insufficient. But it already serves to expose some of the confusion that philosophers have caused by applying the multiple realisability objection to the case of pain. Consider the ‘definition’ of pain as the activation of C fibres, once seriously proposed by philosophers when discussing identity theory. It is true that this can be intended as an as if definition. The problem, however, is that this as if definition is also misleading, because it is based on the already named material elements of the definition, tempting us to think that an adequate definition of pain must identify something materially in a much more specific way than is needed or expected. An analogous case would be that of someone insisting that nothing can be a knife that isn’t made of steel. And just as maintaining that a knife must be made of steel excludes many things usually called ‘knives’, it must be similarly misleading to claim that as peripheral nociceptors, C fibres are a necessary condition for experiencing pain, so that animals lacking C fibres could not possibly feel pain. This would lead all too easily to the species-chauvinistic conclusion denounced by Hilary Putnam, according to which only human beings can experience pain. Now, it is this kind of mistake that pervades Putnam’s multiple realization argument, at least insofar as it concerns phenomenal qualities. To quote a well-known passage by Putnam:
Consider what the brain-state theorist has to do to make good his claims. He has to specify a physical-chemical state such that any organism (not just a mammal) is in pain if and only if (a) it possesses a brain of a suitable physical-chemical structure; and (b) its brain is in that physical-chemical state. This means that the physical-chemical state in question must be a possible state of a mammalian brain, a reptilian brain, a mollusk’s brain (octopuses are mollusca, and certainly feel pain), etc. At the same time, it must not be a possible (physically possible) state of the brain of any physically possible creature that cannot feel pain. Even if such a state can be found, it must be nomologically certain that it will also be a state of the brain of any extra-terrestrial life that may be found that will be capable of feeling pain before we can even entertain the supposition that it may be pain.
These comments, intended to reduce to absurdity the hypothesis of an identity between pain and certain physical states of the brain, can easily be shown to be misleading by considering our proto-definition of pain. For the material specifications in it are restricted to a minimum, consisting in a reference to (non-specified) cortical and pre-cortical centres, along with peripheral cells called nociceptors, which are only functionally defined as cells activated by noxious stimuli and inhibited by analgesic drugs. And even the functioning of these cells is not a necessary condition, since some animals (including human beings) can feel pain simply through direct neuronal activation of the reticular formation and the thalamus. Indeed, once our proto-definition of pain is accepted, the objection of inter-specific multiple realisability loses its relevance. It is very probable, for example, that an octopus can experience pain, and that this involves the activation of paleo-encephalic cells, which usually results from the stimulation of nociceptors… Indeed, the anatomical and histological elements of the proto-definition are vague enough to make this possible. Even octopuses have neuronal cells, they must have pre-cortical centres, and it is obvious that they also have nociceptors, defined as cells excitable by noxious stimuli like heat that can be inhibited by at least some analgesic blockers. (If it should be found that their nociceptors were not blocked by any known analgesic substance, physiologists would want to broaden the concept of nociceptor, since we are convinced that octopuses also feel pain.)
We conclude that, at least concerning inter-specific cases, the multiple realisability objection to identity theory is misleading, because it relies on a mistaken notion of how we specify types, assuming that we identify types on the basis of material properties, ignoring neurofunctional properties. Here the multiple realisability objection is to blame for what we could call a fallacy of false specification. We commit this fallacy when we define something using an unsuitable parameter of specification that sets the boundaries of the concept either too broadly or too narrowly. Often it is committed when people do not sufficiently understand the functional nature of things and consider them only or primarily in terms of their material aspects. For example, we find this fallacy when it is assumed that since insects lack retinal cells, they cannot have eyes. The obvious mistake is to define the concept of eye not based on the functional property of having vision, but instead on the material property of having retinal cells. The thesis that the neurophysiological type of pain in other species is different from that in humans exemplifies the fallacy of false specification. It is possible for different types of pain to arise from the same type of neurophysiology, because pain must be defined mainly in terms of its neurophysiological function, and only secondarily and vaguely in material (anatomical and histological) terms.
If the inter-specific argument of multiple realisability is equivocal when applied to pain, and pain is a phenomenal quality, we may ask if we cannot say the same of other phenomenal qualities. Consider the case of anxiety. One can give a proto-definition of anxiety by saying that it is a cortical effect resulting from the reverberation of neuronal firing in circuits belonging to the limbic system (in this case, we exclude the possibility that insects could experience anxiety – because they don’t have a limbic system – but we are more cautious when considering the possibility of anxiety in fish and birds.) What about visual images? We can say that these are a sort of effect of the isomorphic activation of brain cells, normally caused by the stimulation of the visual organs by light (this case is very inclusive: even insects have some kinds of visual images). It seems that it is always possible to justify general attributions of phenomenal qualities (like pain, anxiety and visual images) by means of a sufficiently broad neurophysiological characterization matching our phenomenally and behaviourally based use of the concept.
The upshot of our considerations is that, regarding phenomenal qualities, we have sufficient reason to believe that the proposal of inter-specific multiple realisability concerning phenomenal qualities is a philosophical fantasy made possible by our lack of knowledge of how the brain really works.
However, there are other philosophical fantasies about the multiple realisability of phenomenal states. Consider the case of phenomenal states in hypothetical extra-terrestrial beings with a non-carbon based physiology, as Putnam suggests at the end of the cited passage. In my view, the question: ‘Why are sophisticated non-carbon based brains not physio-chemically possible?’ is naively motivated. Anyone who has studied biochemistry realizes this. It is like asking: ‘Why is there no life on the other planets of our solar system?’ The obvious answer is that life is an extremely complex and rare phenomenon. The earth is the one place in our solar system where all the myriad necessary conditions were present that made possible and led to the emergence and evolution of an order of reality we call life. Furthermore, the evolutionary knowledge that led biologists to formulate the fundamental principle of convergence between different species suggests that life on other planets in other solar systems would probably not be very different from life on ours. But if life is a complex phenomenon, conscious life is a much more complex and demanding phenomenon. Probably extra-terrestrial beings, in order to be conscious and have phenomenal states, would need to be so biochemically similar to us that it would be possible, if not to use the same neurofunctional definitions for their phenomenal states, at least to extend our definitions in a natural way in order to include them.
Finally, what about the future possibilities of phenomenal states in advanced silicon-based computers? Here the answer is that the belief that phenomenal states such as pain could occur in computers or robots is an even more naïve fantasy that became popular at the beginning of the computer era. Robots can replicate the external manifestations of psychological states like sensations and feelings, but cannot intrinsically experience real sensations or real feelings. For one thing, they don’t belong to the same order of physical reality that makes such phenomena possible, but instead to a far less complex physical order. Computers and robots with human feelings such as pain are the stuff of science fiction films by Hollywood producers like Steven Spielberg or Stanley Kubrick (e.g., A.I.: Artificial Intelligence; 2001: A Space Odyssey). Influenced by such fanciful ideas but lacking an understanding of the real dimensions of the biological nature responsible for the incredible sophistication of real brains, people simply project human consciousness and mental faculties into supercomputers, with their enormous databases and capability to make extremely rapid calculations involving vast numbers of variables. However, this is in fact not so very different from the anthropomorphic projection of human feelings and thoughts into the animal kingdom that has been found in totemic societies.
We conclude that at least concerning phenomenal qualities the neurofunctional interpretation of the constitution of mental states immunizes the identity thesis against the multiple realisability objection.
Now, what about the multiple realisability objection on the level of cognitive states like thoughts, beliefs, intentions, etc., which from Aristotle to Aquinas were seen as belonging to the more abstract and general dimension of the mental? Here I have real doubts as to whether the same kinds of answers are suitable. It remains implausible that a particular thought, for example, that Kauai is the oldest Hawaiian island, should have similar neurofunctional paths in both my brain and yours. Moreover, a chain of reasoning, understood as a logical sequence of thoughts, seems to be at least structurally reproducible in the form of a mechanical manipulation of symbols according to an algorithm. Indeed, there must be something that allows us to say that computers can think in some derivative sense of the word, since they are able to solve complicated mathematical problems or play chess. Even if the biological process of human thinking is very different, it could be maintained that there is no reason why the symbolic manipulation of a computer couldn’t at least in some respects resemble what occurs in the mind of a mathematician or a chess player.
Nevertheless, the idea that a computer would be able to think or reason in the strict and proper sense of the word remains intuitively implausible. Why is this the case? – Mere prejudice? I don’t think so. It seems to me that when we speak about thinking and reasoning we are not restricting ourselves to the generation of sequences of symbols in accordance with algorithms. We are assuming that these are integrated into a whole which includes in its foundations the lower levels of the thinker’s emotions, desires, sensory experiences and memories, that is, things that must involve phenomenal states depending on our biology and life form.
The necessary involvement of our emotions and desires in reasoning was famously pointed out by David Hume, who maintained that reason is the slave of passion, for our reasoning must have changeable aims originally grounded in the emotional and volitional nature moulded in our life form. But these are not the sorts of phenomena that a computer could duplicate. Moreover, the necessary engagement of sensations as grounding elements for our reasoning processes was already pointed out by Kant, who concluded that thoughts without any association to sensible intuitions are empty, that is, lacking a semantic basis. Indeed, if artificial reasoning and thinking were possible, then from a wider perspective they would be seen as irrational and in the end thoughtless, because they would not be adequately grounded in the phenomenal states constitutive of our biologically-rooted and socially-moulded nature, the only things able to make full sense of them. It seems, consequently, that in order to reason or think in the full sense of the word, a mind would need to possess appropriate phenomenal states that could give content to its reasoning and thinking. However, as already explained, a true phenomenal state for consciousness and thinking must necessarily be a biological, neurophysiological state.
The result of these considerations seems to be that even if the multiple realisability objection were applicable to cognitive processes, forcing us to accept only a token-token identity on that level, this would not necessarily lead us to the fanciful conclusion that some minds could be very different from our own, perhaps transcending the biological realm. It is true that a token-token identity is the identity of something with something else, which can be simply anything, insofar as the kind of identity always remains open. However, the reason to reject the conclusion that at the cognitive level the multiple realisability hypothesis trounces all its rivals is that the cognitive states of conscious minds would always have to be grounded in phenomenal states. Since in this way cognitive states would depend on something identifiable with definite functional neurophysiological states by means of a type-type identity theory, this would allow us to restrict the objection of multiple realisability for cognitive states in a satisfactory way, limiting them to biological brains.
With this conclusion, we reach the point where we can make a speculative wager about the right form of an identity theory of mind. It would be a double-edged solution that could be motivated by and against the medieval doctrine that only the intellective soul – but not the sensitive soul – could survive the destruction of the body. Mind has indeed both a cognitive and a phenomenal dimension. The phenomenal dimension, as we have seen, is connected to the brain by type-type identities, because of the concrete singularity of its proper neurofunctional states. However, because of its abstract nature the cognitive dimension is structurally linked to the physical realm only by a token-token identity. Nevertheless, contrary to the medieval doctrine, the cognitive mind continues to depend on the living brain. And the reason for this is that the cognitive mind is rooted in the phenomenal dimension of the mind. This in turn is inextricably dependent upon our biological brain and life form. In short: the phenomenal mind is the glue that cements the cognitive mind in the brain.
After the multiple realisability objection, the second main objection to type-type identity theory consists in the claim that this theory is unable to reduce phenomenal qualities in the sense of identifying them with neurophysiological properties. The sentence ‘I have a visual experience of red’ seems to say something different from ‘Such and such brain-cells and centres are being activated in such and such ways’. This is made clear by the fact that a blind person could know everything about how these brain cells work to produce the experience of the colour red, without knowing, as we do, what the colour red looks like, since she has no access to the qualia of red.
Since we have answered the first main objection, it makes sense to finish this paper with a tentative answer to the second. In my view, the second objection arises from a confusion as old as that between sense and reference. The sentence
(1) Having the visual experience of the colour red is the same as having such and such neurofunctional phenomena.
is in some sense analogous to
(2) The roundness of this tennis ball that I experience by seeing it is the same as the roundness of this tennis ball that I experience by touching it.
The roundness of the tennis ball is the object of knowledge, which remains one and the same, although differently experienced. However, we have here two different senses, or, as I prefer to call them, two different perspectives of access: the visual and the tactile experiences of roundness. Obviously, we could also take these perspectives of access as objects of our knowledge. Since these two perspectives are different, there is no question about their numerical identity. The sentence
(3) The visual experience that I have when I speak of the roundness of a tennis ball is the same as the tactile experience that I have when I speak of the roundness of the tennis ball.
is obviously false, since here the objects of knowledge are the different perspectives of access.
Now, my suggestion is that the objection that phenomenal qualities cannot be reduced to neurophysiological properties arises from our perception that, even if the object is the same, there is an irreducible remainder in the fact that the associated perspectives of access remain different.
My answer to the objection is that it is misleading, for the perspectives of access are also reducible, so that there is no need for a non-identical remainder. To show this, all we need to do is to choose a perspective of access as our object of knowledge and consider attentively what occurs in experiencing something. In this case, we can show, for example, that the chosen psychological perspective of access is identical with a further physical-physiological object of description that has its own proper public perspective of access. For example:
(4) The visual experience that I have of a tennis ball is the same as a computing process integrating spatial extension, perspective, form and colour in certain centres of my brain.
Unlike (3), (4) is a true sentence. If the subjective psychological perspective of access can also be described as a new public physical-neurophysiological phenomenon, we see that this perspective, taken as an object of knowledge, could also be reduced with no remainder. We can, I am convinced, also detect different perspectives of access to a given perspective of access taken as an object… But it is plausible that we could give the same answer regarding any higher order perspective of access. The conclusion of this argument is that when we make ourselves aware that any supposed psychological remainder responsible for what an experience is like can always be equated with some public physical-physiological state of affairs, the objection breaks down that phenomenal qualities are non-reducible.
 See William Bechtel & Jennifer Mundale, ‘Multiple Realizability Revisited: Linking Cognitive and Neural States’, Philosophy of Science 66 (1999), 175-207. See also chapter 1 of T. W. Polger’s book, Natural Minds (Cambridge: MIT Press, 2004). The basic tenet of these arguments is the idea that neurophysiological types must be more coarse grained (that is, more abstractly understood). By adding the suggestion that neurophysiological types must also be seen as essentially neurofunctionally defined, as I do in this paper, I hope to show something about how they must be more abstractly defined and to what extent.
 Philosophers have conceived of bizarre circumstances that seem to contradict this definition. One can imagine, for example, a possible world where people are extremely overweight and sedentary, so that chairs must resemble benches on which two or three persons of our world could sit side-by-side… One can also imagine a situation in which benches function like chairs, because no more than one person is permitted to sit on them at one time… See Paul Elbourne: Meaning: A Slim Guide to Semantics (Oxford: Oxford University Press, 2012), chap. 1. Nonetheless, in the world where people are extremely overweight, the chairs are not benches, because they are still designed to be chairs, that is, to be used by only one person at a time, and in the second situation, the benches are still designed as benches, even if their use is limited to just one person at a time. As can be expected, there are borderline cases, like that of a trunk that happens to have the form of a chair, and because of this, with some minor changes, is used as a chair and is also called a chair. And there are extended (non-proper) senses, like the chair of rock in Victor Hugo’s Les Travailleurs de la mer, not to mention metaphorical senses. Still, I don’t see why we should conclude that concepts are generally indefinable merely because there are borderline cases. What is important is that definitions usually work.
 The idea of a neurofunctional state does not commit us to functionalism, at least insofar as functionalism tends to confuse our forms of epistemic access to a mental state with the mental state as such.
 Hilary Putnam, ‘The Nature of Mental States’, in Mind, Language and Reality: Philosophical Papers (Cambridge: Cambridge University Press, 1975), Vol. 2, p. 436 (my italics).
 For a converging approach, see P. M. S. Hacker’s interpretation of Wittgenstein in his book, Wittgenstein: Meaning and Mind, part I: Essays (Oxford: Blackwell, 2001), chap. IV.