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

C. One-one Correspondence and the "Riddle of the Universe."

The isomorphism of the mental and the physical consists, according to our interpretation, in a one-one correspondence of elements and relations among the phenomenal data with the elements and relations among the referents of certain neurophysiological terms. And we proposed to explain this isomorphism in the simplest way possible by the assumption of the identity of phenomenal data with the referents of (some) neurophysiological terms. The question arises whether the identity view could be held if we were, for empirical reasons, forced to abandon ψ-φ-one-one correspondence and to replace it by a doctrine of one-many correspondence. As was pointed out previously, the physicalistic predictability of the occurrence of mental states would in principle still be unique, if one-many correspondence holds true. Comparison with an example of the identification of purely physical concepts may shed some light on this issue. Macro-temperature, as thermometrically ascertained, corresponds in one-many fashion to a multitude of micro-conditions, viz., a very large set of molecular states. Strictly speaking, this correspondence holds between one state description on the macro-level with a specifiable infinite disjunction of state descriptions pertaining to the micro-level. Since, as we have also pointed out, this correspondence is empirically ascertained, there is here as little reason to speak of logical identity as in the ψ-φ case. Nevertheless, we have seen that it makes sense, and what sense it makes, to regard the relation of temperature to mean molecular kinetic energy as an example of a theoretical identity.

In the mind-body case, just as in the temperature case, prediction of the ensuing micro- (and ultimately even macro-) constellations on the basis of information about, respectively, the mental state, or the macro-temperature state, could not be unique under the supposition of one-many correspondence. This is obvious for the temperature example in the light of the principles of statistical mechanics. Analogously, the precise behavior subsequent to the occurrence of a specified mental state would not be predictable either. This is not too disturbing by itself. After all, even if one-one correspondence held true, the neural correlates of a mental state would form only a very insignificant part of the relevant total initial conditions. Talk of identity in the case of one-many correspondence, however, would seem unjustified, because here we are (ex hypothesi) acquainted with the phenomenal datum, and the corresponding disjunction of cerebral states could not plausibly be identified with that individual datum.

Even if one-one correspondence is assumed, there is an intriguing objection1 against the identity view. According to the view presented in section V, there is no empirically testable difference between the identity and the parallelism doctrines. We said that the step toward the identity view is a matter of philosophical interpretation. But, so the objection maintains, if identity is assumed, it would be logically impossible to have a stream of direct experience (a "disembodied mind") survive bodily death and decay. It is further asserted that this would not be a logically entailed consequence of parallelism. For it could well be maintained that the one-one correspondence holds only during the life of the person, but that as drastic an event as bodily death marks the limits of this correspondence. Mental states could then occur independently of physical correlates.

Thus it would seem as if our philosophical identity theory implied consequences which are testably different from those of parallelism. This is quite paradoxical. My tentative reply to this argument is twofold. First,ψ-φ identification being empirical, it could of course be mistaken. But if the identity does hold, then survival is indeed logically impossible. This is logically quite analogous to the conditional: If the law of the conservation of energy holds, then a perpefuum mobile (of the "first kind") is thereby logically excluded. But, of course, the energy law has only empirical validity and might some day be refuted by cogent empirical evidence. Second, and perhaps more important, the parallelism doctrine, as I understand it, holds that there is a ψ-φ-oneone correspondence and that this correspondence is a matter of universal and irreducible law. This seems to me to exclude disembodied minds just as much as does the identity thesis. I therefore think that the identity thesis is a matter of epistemological and semantic interpretation, and does not differ in empirical consequences from a carefully formulated parallelism.

Another perplexity was formulated in Leibniz's monadology, and in different form presented by E. Du Bois-Reymond as one of his famous unsolvable "riddles of the universe." If I may put the core of the puzzle in modern form, it concerns the irreducible (synthetic) character of the ψ-φ correlations. Wherever we find co-existential or correlational regularities in nature, we hope to find a unitary explanation for them, and in many cases scientific theories have provided fruitful and well-confirmed explanations of this sort. But in the case of the ψ-φ correlation we seem to be confronted with a fundamentally different situation. There is no plausible scientific theory anywhere in sight which would explain just why phenomenal states are associated with brain states. Many philosophers have resigned themselves to regard the ψ-φ correlations as "ultimate," "irreducible," "brute facts." Since any explanation presupposes explanatory premises which at least in the context of the given explanation must be accepted, and since even the introduction of higher explanatory levels usually reaches its limit after three or four "steps up," one might as well reconcile oneself to the situation, and say that "the world is what it is, and that's the end of the matter." Now, I think that it is precisely one of the advantages of the identity theory that it removes the duality of two sets of correlated events, and replaces it by the much less puzzling duality of two ways of knowing the same event—one direct, the other indirect.

Nevertheless, there are some "brute facts" also according to the identity theory. But they are located differently. Besides the basic physical laws and initial conditions, there are according to our view the only empirically certifiable identities of denotation of phenomenal and of physical terms. But this identity cannot be formulated in laws or law-like sentences or formulas. The identity amounts, merely to the common reference of acquaintance terms on the one hand and unique physical descriptions on the other. Any other way of phrasing the relation creates gratuitous puzzles and avoidable perplexities. For example, it is misleading to ask, "Why does a mental state 'appear' as a brain state to the physiologist?" The brain-state-as-it-appears-to-the-physiologist2 is of course analyzable into phenomenal data forming part of the direct experience of the physiologist. The "brute fact" simply consists in this, that the phenomenal qualities known by acquaintance to one person are known (indirectly) by description to another person on the basis of phenomenal (evidential) data which, in the vast majority of cases, are qualitatively quite different from the data had by, or ascribed to, the first person. I see nothing paradoxical or especially puzzling in this account of the matter.

A little reflection upon the autocerebroscopic situation shows clearly that the correspondence between, e.g., musical-tones-as-directly-experienced and certain excitation patterns in the temporal lobes of one's brain as represented by visual patterns (perceived on the screen) is simply a correlation between patterns in two phenomenal fields. The conceptual neurophysiological account of the visual data in this case consists in explanatory hypotheses about cerebral processes which are causally responsible for the production of the image on the screen, and these are in turn causally responsible for the emergence of certain patterns in the visual field. Strictly speaking, and in the light of physical laws, there must even be a minute time lag between the moment of the occurrence of a neural event in the temporal lobe and its "representation" via the autocerebroscope in one's own visual field. The experienced patterns in the visual field are in this situation the causal consequences of (among other things) the auditory data. Disregarding the small time lag we could here speak of a parallelism indeed. But this is a parallelism between the data (or patterns) in different sense modalities; or, in the case of visual experience autocerebroscopically "represented" by other visual data, within one and the same modality. (May I leave it to the reader to think this through and to find out for himself that this special case of autocerebroscopy does not involve any paradoxical consequences.)

Another puzzle that may be raised is the question as to whether the proposed identity theory does not involve the undesirable consequences of epiphenomenalism. It should be obvious by now that our solution of the mind-body problem differs quite fundamentally from materialistic epiphenomenalism in that: (1) it is monistic, whereas epiphenomenalism is a form of dualistic parallelism; (2| the "physical" is interpreted as a conceptual system (or as the realities described by it), but not as the primary kind of existence, to which the mental is appended as a causally inefficacious luxury, or "shadowy" secondary kind of existence; (3) quite to the contrary, mental states experienced and/or knowable by acquaintance are interpreted as the very realities which are also denoted by a (very small) subset of physical concepts. The efficacy of pleasure, pain, emotion, deliberation, volitions, etc. is therefore quite definitely affirmed. In this respect monism shares the tenable and defensible tenets, without admitting the objectionable ones, of interactionism.

Speaking "ontologically" for the moment, the identity theory regards sentience (qualities experienced, and in human beings knowable by acquaintance) and other qualities (unexperienced and knowable only by description) the basic reality. In avoiding the unwarranted panpsychistic generalization, it steers clear of a highly dubious sort of inductive metaphysics. It shares with certain forms of idealistic metaphysics, in a very limited and (I hope) purified way, a conception of reality and combines with it the tenable component of materialism, viz., the conviction that the basic laws of the universe are "physical." This means especially, that the teleology of organic processes, the goal directedness or purposiveness of behavior are macro-features, and that their explanation can be given in terms of non-teleological concepts and laws which hold for the underlying micro-levels. In other words, the monistic theory here proposed does not require irreducibly teleological concepts in its explanatory premises.

In this connection there is, however, a perplexity which may give us pause. Inasmuch as we consider it a matter of empirical fact and hence of logical contingency just which physical (neurophysiological) concepts denote data of direct experience (raw feels), one may wonder whether the causal efficacy of raw feels is satisfactorily accounted for. There are countless teleological processes in organic life which, unless we be pan-psychists or psychovitalists, must be regarded as occurring without the benefit of sentience. For examples, consider the extremely "ingenious" processes of reproduction, growth, adaptation, restitution, and regeneration, which occur in lower organisms as well as in many parts of human organisms. On the other hand, the causal efficacy of attention, awareness, vigilance, pleasure, pain, etc. on the human level is so striking that one is tempted, with the panpsychists, to assume some unknown-by-acquaintance qualities quite cognate with those actually experienced.

The new puzzle of epiphenomenalism would seem to come down to this: An evolutionary, physiological, and possibly physical explanation of adaptation, learning, abient or adient, goal-directed behavior can be given without any reference whatever to raw feels. The distribution of raw feels over the various possible neural states could be entirely different from what in fact it is. For example, raw feels might be associated with the peristaltic movements of the stomach or with coronary self-repair, and not with cortical processes. But, I repeat, such different distribution of raw feels or even their complete absence would still not prevent an adequate explanation of teleological behavior. Of course if we accept the actual distribution, i.e., the total set of ψ-φ-correlation rules as ultimate parallel laws, and interpret these according to the identity theory, then we can quite legitimately speak of the efficacy of raw feels. This is so, because the raw-feel terms are then precisely in those loci of the nomological net where science puts (what dualistic parallelism regards as) their neural correlates. But if the biopsychological explanations offered by the theories of evolution and of learning can thus incorporate the efficacy of raw feels, those theories presuppose, but do not by themselves explain, the ψ-φ correlations.

That pleasure or satisfaction reinforces certain forms of adient behavior can be formulated in the manner of the law of effect (cf. Meehl, 220). But in the ultimate neurophysiological derivation of this empirical law of behavior, the correlation of pleasure or gratification with certain cerebral states is not required. Behaviorists, especially "logical behaviorists," have taken too easy a way out here in simply defining the pleasurable as the behaviorally attractive and the painful as the behaviorally repellent. The "illumination" of certain physically described processes by raw feels is plainly something a radical behaviorist cannot even begin to discuss. But if the synthetic element in the ψ-φ relations that we have stressed throughout is admitted, then there is something which purely physical theory does not and cannot account for. Is there then a kind of "brute fact" which our monistic theory has to accept but for which there is possibly no explanation, in the same sense as there can be (within a naturalistic empiricism) no explanation for the fact that our world is what it is in its basic laws and conditions? Possibly, however, I see a riddle here only because I have fallen victim to one of the very confusions which I am eager to eliminate from the mind-body problem. Frankly, I suspect some sort of "regression" rather than "repression" has engendered my bafflement. If so, I should be most grateful for "therapeutic" suggestions which would help in clearing up the issue. Possibly, the solution may be found in a direction which appears plausible at least for the somewhat related puzzle of the "inverted spectrum."

This ancient conundrum, we have seen, is not satisfactorily "dissolved" by Logical Behaviorism. A "captive mind" is logically conceivable, and might know by acquaintance that his sense qualia do not stand in one-one correspondence to his autocerebroscopically ascertained neural states. If physical determinism is assumed, then it is true that such knowledge would have to remain forever private and uncommunicable. But under these conditions a systematic interchange of the qualia for one person at different times and as between different persons is logically conceivable. It would of course ex hypothesi not be intersubjectively confirmable, and thus never be a possible knowledge claim of science. But the logical conceivability of the inverted spectrum situation demonstrates again the empirical character of the ψ-φ correspondence. This empirical character is, however, (as we have also emphasized) extremely fundamental in that it is closely bound up with the basic principle of causality or of "sufficient reason." Systematic interchange of qualia for the same sort of neural states would be something for which, ex hypothesi, we could not state any good reasons whatever.

Furthermore, there is a grave difficulty involved in the assumption that a captive mind could even "privately" know about the interchange. Normal recollection by memory presumably involves (at least) quasi-deterministic neural processes. The captive mind could be aware of the inverted spectrum type of interchange of qualia only if we assume some peculiar breach in normal causality. If the captive mind is to know that today the correlation of raw feels with neural states differs from what it was yesterday, he would have to remember yesterday's correlations. But how could this be possible if memory depends upon modifications in the neural structures of the cortex? These considerations show clearly that under the supposition of normal physical causality the systematic interchange would remain unknowable even to the private captive mind. (Converse, but otherwise analogous, puzzles arise for the assumption of the survival of a private stream of experience beyond bodily death. How could such a private mind have knowledge about the continuance of his "physical" environment?)

All these reflections seem to me to indicate that in our world at least, there is nothing that is in principle inaccessible by "triangulation" on an intersubjective (sensory) basis. The having of raw feels is not knowledge at all, and knowledge by acquaintance does not furnish any truths which could not in principle also be confirmed indirectly by persons other than the one who verifies them directly. The ψ-φ-identity theory as I understand it, makes explicit this "ontological" feature of our world. The criterion of scientific meaningfulness formulated in terms of intersubjective confirmability, far from being an arbitrary decree or conventional stipulation, may thus be viewed as having ontological significance—but "ontological" in the harmless sense of reflecting an inductively plausible, basic characteristic of our world.

Empirical identity, as I conceive it, is "weaker" than logical identity but "stronger" than accidental empirical identity, and like theoretical identity stronger than nomological identity in the physical sciences (just as causal necessity is weaker than logical necessity, but stronger than mere empirical regularity). If the coreference of a phenomenal term with a neurophysiological term is conceived as something more than mere extensional equivalence, if it is conceived as characteristic of the basic nature of our world (just as the basic natural laws characterize our kind of world and differentiate it from other kinds), then perhaps the inference from a neural state to its ("correlated") raw feel is at least as "necessary" (though of course not purely deductive) as is the inference from, e.g., the atomic structure of a chemical compound to its macro-physical and chemical properties.

I hope that readers sympathetic to my admittedly speculative gropings will try to formulate in logically more precise and lucid form what I have been able to adumbrate only so vaguely. Such readers should in any case keep in mind one of the ideas which seem to me indispensable for an adequate solution of the phenomenalism-realism as well as the mind-body problems: The paradigm of symbolic designation and denotation is to be seen in the relation of a token of a phenomenal term to its raw-feel referent. All non-phenomenal descriptive terms of our language, i.e., all physical terms (no matter on which level of the explanatory hierarchy) designate (or denote) entities which -- within the frame of physical knowledge -- are unknown by acquaintance. But if our "hypercritical" realism is accepted, we must ascribe denotata to all those physical terms which designate individuals, properties, relations, structures, fields, etc., i.e., entities which can justifiably be said to be described (i.e., uniquely characterized) on the basis of evidential data by Russellian descriptions on one or the other level in the hierarchy of logical types. "To exist" means simply to be the object of a true, uniquely descriptive statement. But since such descriptive knowledge (on a sensory evidential basis) by itself never enables us deductively to infer the acquaintance qualities of its objects, there is always a possibility for some sort of modal identification of a datum with a specifiable descriptum. This is the central contention of the present essay.


Notes

1. Raised in Minnesota Center for Philosophy of Science discussion by Mr. H. Gavin Alexander.

2. No matter whether the physiologist observes someone else's brain, or -- autocerebroscopically -- his own.