Why Thoughts Cannot be Chemical Events
The usual kinds of theoretical identity exhibit a common feature: we can give a compositional analysis of the thing being identified in terms of what it is being identified with. Thus in the case of “water is H2O” we can analyze water as composed of two parts hydrogen to one part oxygen: these are the constituents of water. This is chemical compositional analysis. If water could not be analyzed in this way, it would be false to say that water is H2O. Similarly for “heat is molecular motion”: heat can be analyzed as composed of rapidly moving molecules. Likewise for “light is a stream of photons/pattern of electromagnetic waves”: light is composed of such constituents. In all these cases there is a kind of hidden structure to the thing in question and it reduces to that structure by compositional analysis. The structure may not be obvious, the constituents may be remote from common sense, but the thing in question is in fact made up of these constituents in that structure. It certainly does not have a compositional analysis that differs from that with which it is identified. The analysis is what grounds the claim of identity, so the identity cannot hold without the analysis holding. Often the analysis reveals the phenomenon to be more complex than we might have supposed; but there is no doubt that the phenomenon is to be analyzed in that way. It has turned out that water, heat, and light have such-and-such a compositional analysis. If someone tried to claim that heat is light, say, it would suffice to point out that these two phenomena have quite different sorts of composition: heat is composed of molecules in motion, light is composed of photons or electromagnetic waves. Since these are not the same, nor can heat and light be the same.
It is the same story with other sorts of identity, such as those that apply to physiological processes. Suppose we identify digestion with a complex sequence of stages of food breakdown and absorption: “Digestion is the process that begins with food being taken into the mouth, then chewed, then swallowed, then entering the stomach, and so on.” We have analyzed digestion into its component parts and explained how those parts aid the process. That process was largely hidden, but science has revealed it, and we can now state the appropriate identity, which is underpinned by physiological compositional analysis. Similarly for breathing and gestation: we can say what breathing is by talking about the lungs and their action, oxygen and the blood, thus providing a compositional analysis; and we can say what gestation is by describing the processes of embryological development. We spell out the underlying complexity that constitutes respiration and gestation: this is what those processes are composed of. We suspect a hidden structure to something for which we have a mere label, and science tells us just what that hidden structure is. If someone were to try to claim that respiration is identical to gestation, it would suffice to point out that these two processes are composed quite differently—with the lungs involved in one case and the womb in the other.
But now how does it work in the case of thoughts and the brain? The identity theorist claims that a thought is identical to a chemical event in the brain: my thinking that it is about to rain is identical to a batch of chemicals crossing a synaptic cleft or some such. Thoughts are identical to their chemical correlates—actual molecules moving and combining. When we observe brain chemistry we are observing thoughts: this is what thoughts have turned out to be. Science has discovered the hidden structure of thoughts, and it is the structure of chemicals in the brain. The correct compositional analysis of thoughts is in terms of chemical constituents and their interactions. The trouble with this familiar picture is that thoughts do not appear to have any such chemical compositional analysis. They do have a compositional analysis, in terms of their conceptual constituents, but this is not a chemicalanalysis: the constituents of thoughts comprise concepts of different kinds standing in the relation of predication and other logical relations, not chemicals that form compounds and travel across ion channels. So it is hard to see how the former can be identified with the latter. It is like trying to identify heat with light or respiration with gestation: wrong compositional analysis.
Suppose we had an identity theory of sentences: when someone hears a sentence in his mind this event is identical to a chemical event in his brain. The trouble with this theory is that while the sentence does have a compositional analysis, in terms of nouns, verbs, prepositions, modifiers, and so on, this structure is not the same as a structure of chemicals in the brain. We can’t map the sentence structure onto a structure of chemicals shuttling around between neurons. Consider that molecules break down into atoms and that atoms break down into protons and electrons: that is the compositional analysis of a molecular event. But nouns and verbs don’t break down into atoms and elementary particles—they don’t have any such compositional analysis. So sentences (uttering or hearing them) cannot be identified with chemical events. Chemical events may—and no doubt must—correlate with sentences and thoughts, but we can’t claim that the latter can be compositionally analyzed by reference to the former.  By contrast, the standard examples of theoretical identity do involve clear instances of compositional analysis. Water consists of H2O in a way that the thought that it’s about to rain does not consist of a collection of chemical events. The structure of thought is not molecular structure.
It is, however, very plausible to suppose that physiology reduces to chemistry. The physiological processes of the body—digestion, respiration, gestation, and so on—do consist of molecular processes. The body is composed of chemicals, and these chemicals provide a compositional analysis of all the processes that occur in the body. This is something we have discovered; it is certainly not obvious (there might have been a “vital force”). Molecules have the power, by joining and separating, and by energy exchanges, to produce the full range of physiological processes. Chemistry may not be alchemy, but it can achieve some remarkable transformations. We can take an activity of the body and dissect it into its chemical components, thereby achieving full explanation. There is no “explanatory gap” between physiology and chemistry, no “hard problem”, no “mysterious link”. The body really does have a chemical compositional analysis (that, ultimately, is how it is possible). Thus we can endorse “chemicalism” about the body. And that includes the brain: it too is an organ consisting of a collocation of molecules. Nerve transmission reduces to activities of molecules. The body really is a complicated molecular machine, a vast chemical factory, because molecules are versatile and productive things, capable of elaborate feats of construction and energy transfer. The body is composed of cells and the cells are composed of complex molecular structures—so it all comes down to chemistry. Thus it is reasonable to propound various kinds of theoretical identity with respect to physiology, which are entirely analogous to the identities propounded in physics and chemistry itself (such as the identification of the elements with atoms of varying atomic number). Given that chemistry is reducible to atomic physics, we then reach the conclusion that the body is indeed a physical thing—by no means a trivial result.
But psychology does not reduce to chemistry—which is interesting because it surely depends on chemistry. Thoughts don’t have a chemical compositional analysis, unlike events of digestion or respiration—even though they are, in some sense, grounded in chemicals. We cannot be “chemicalists” about thoughts. We thus reach the conclusion that the mind is not the body: for the body is subject to “chemicalism” but the mind is not—therefore the mind is not the body. If the mind were the body, the mind would have a chemical compositional analysis; but as it doesn’t, it isn’t. Nevertheless, chemicals are, in some sense, the machinery of thought—what make thought possible, give it efficacy, embody it. We have given various technical-sounding names to this vital relation, none of them very illuminating: “supervenience”, “realization”, “emergence”, “grounding”, “hardware”, “substrate”, and so on. The connection is real enough—there is no thought without a chemical basis—but it is completely opaque. My aim has been to articulate one reason why thoughts cannot be chemical events. The fundamental point is the asymmetry between standard theoretical identifications and putative psychochemical identifications, in respect of compositional analysis.
No doubt there are ways in which one might try to wriggle out of the argument, which are quite familiar. It might be claimed that thoughts really do have a chemical analysis, contrary to initial impressions, because we can assign their conceptual analysis to the appearances and insist that chemistry tells us their hidden real essence—as we can assign the superficial properties of water to their appearances to the human senses. The obvious reply to this is that the conceptual structure of thought is not just a dispensable superficial appearance but is the very essence of thought. Or we could dispense with thoughts altogether and just talk chemistry. Or we could protest that it’s all just a matter of “intuition” and hence methodologically suspect. Or we could question the whole idea of compositional analysis. What is clear is that there is a significant difference between standard scientific identities and the claim that thoughts are identical to chemical events, concerning the question of compositional analysis. The difference is at its most telling in the contrast between physiology and psychology in respect of chemistry. Physiology really is a branch of chemistry, but psychology is apparently not. Accordingly, thoughts are not physiological processes. 
 If we adopt the language of thought theory, we can easily convert the argument from language to thought: given that inner sentences are not chemical events, thoughts cannot be either. Thoughts cannot have a chemical compositional analysis if the sentences that constitute them do not.
 We can claim that thoughts are biological processes in the sense that they are aspects of organisms that evolved by natural selection, but it doesn’t follow that they are physiological processes, like respiration and digestion. No physiological process in the brain can be a thought, on pain of reducing thinking to chemistry.