The "Mental" and the "Physical"

Herbert Feigl, The "Mental" and the "Physical": The Essay and a Postscript (1967).

B. Unitary or Dual Language Reconstruction?
In most of the crucial parts of the present essay I have taken a unitary language to be the ideal medium of epistemological reconstruction. By this I mean the following: Both the phenomenal terms (designating raw feel data) and the illata terms (designating unobservables) occur in the language of com-monsense or of science, and they are connected by strands in the nomo-logical net. I believe that if this sort of unitary language is constructed with care, category mistakes can be avoided. This reconstruction differs essentially from the dual language reconstruction pursued by Carnap and W. Sellars (cf. their essays 73, 315). Purely phenomenal terms are_ there excluded, presumably owing to their conviction that category mistakes as well as solipsism would be unavoidable if we chose a phenomenal basis of reconstruction. But with the reinstatement of realism, i.e., with the insistence on the synthetic character of the strands in the nomological net, solipsism is no longer a consequence, and category mistakes can be avoided if we dismiss pictorial appeals as cognitively irrelevant, and if we take care to distinguish sharply between universals and particulars, among phenomenal as well as among non-phenomenal terms.
I admit, of course, that there are certain distinct advantages in the dual language reconstruction. All evidential statements are there couched in terms of the observation language; and the observation language is conceived as inteisubjectively meaningful right from the beginning. The connections between the observation language and the theoretical language are formulated with the help of correspondence rules. This type of reconstruction is very illuminating in the analysis of the meaning and the confirmation of scientific theories. But, as I have pointed out, it does not do full justice to statements about the data of direct experience, whether they are one's own or someone else's. In our unitary language the "partition" between the data and the illata is located very differently. The correspondence rules in the unitary language would ultimately be statements of f-Q correlations, i.e., of the raw-feel denotations of neurophysiological terms. Since precise knowledge of these correlations is only a matter of hope for a future psychophysiology, the unitary language is largely in the "promissory note" stage. It is therefore not very illuminating if our epistemological reconstruction is to reflect the progress of knowledge in our very unfinished and ongoing scientific enterprise. For this purpose, the dual language reconstruction is much more adequate.
But if we are satisfied with relatively low probabilities for the strands in the nomological net, the unitary reconstruction might do the job too. As a sketch for a reconstruction of an ideally finished science, however, the unitary language approach is preferable. What this would amount to can at present be indicated only by some sort of "science-fiction" illustration: Suppose that we had a complete knowledge of neurophysi-ology and that we could order all possible human brain states (if not metrically, then at least topologically) in a phase space of n dimensions. Every point in this phase space would then represent a fully specific type of brain state. And, taking isomorphism for granted, a subset of these points would also represent the total set of possible mental states.
Suppose further that we could teach children the vocabulary of the language of brain states. If this requires n-tuples of numbers, then simple expressions like "17-9-6-53-12" (or even abbreviatory symbols for these) might be inculcated in the child's language. If we took care that these expressions take the place of all introspective labels for mental states, the child would immediately learn to speak about his own mental states in the language of neurophysiology. Of course, the child would not know this at first, because it would use the expression, e.g., "17-9-6-53-12" as we would "tense-impatient-apprehensive-yet hopefully-expectant." But having acquired this vocabulary, the child, when growing up and becoming a scientist, would later have no trouble in making this terminology coherent with, and part of, the conceptual system of neurophysiology, and ultimately perhaps with that of theoretical physics. Of course, I not only admit, but I would stress, that in this transformation there is a considerable change in the meaning of the original terms. But this change may be regarded essentially as an enormous enrichment, rather than as a radical shift or a "crossing of ontological barriers." In other words, introspection may be regarded as an approach to neurophysiological knowledge, although by itself it yields only extremely crude and sketchy information about cerebral processes. This sort of information may concern certain Gestalt patterns, certain qualitative and semiquantitative distinctions and gradations; but it would not, by itself, contain any indication of the cerebral connections, let alone localizations.