2011.01.23 — Language & Responsibility continued

Language & Responsibility, begun 2011.01.06, continues to be excellent. I seem to be stumbling into writers who with great intelligence and perception are clearing identifying empirical failures of the 'soft' sciences — sociology, linguistics, psychology — because of irrational exclusions and limitations to their scope of study. It was funny, even almost fushigi-like, to read Chomsky castigating psychologists and linguists, so soon after having posted Jung's similar lament a few weeks ago about psychologists.

Noam Chomsky. Language and Responsibility: Based on conversations with Mitsou Ronat. (Out of Print, but you might like the book behind this hyperlink.) New York: Pantheon Books (KnopfDoubleday), 1979. Translated by John Viertel. ISBN0394736192

So, from the mid-1970s, here's Chomsky's observation that...

... many psychologists have a curious definition of their discipline. A definition that is destructive, suicidal. A dead end. They want to confine themselves solely to the study of performance—behavior—yet, as I've said, it makes no sense to construct a discipline that studies the manner in which a system is acquired or utilized, but refuses to consider the nature of this system.

Okay, that was just a short clip from the extended text! But interesting. Here's the extended version that includes the above:

Mitsou Ronat: The linguistic model is a model of what is termed competence. You have just mentioned process models or models of performance. This opposition, competence-performance, was first clearly stated around 1964-5. You defined linguistic competence as that knowledge internalized by a speaker of a language, which, once learned and possessed, unconsciously permits him to understand and produce an infinite number of new sentences. Generative grammar is the explicit theory proposed to account for that competence. In performance, other cognitive systems, aside from competence (memory, etc.), intervene.

In Language and Mind you indicate that the other branches of psychology—dealing with vision, memory, and so on—must, in order to become scientific, define an equivalent concept of competence. Now it is evident that most psychologists oppose just that concept.

Noam Chomsy: In my opinion, many psychologists have a curious definition of their discipline. A definition that is destructive, suicidal. A dead end. They want to confine themselves solely to the study of performance—behavior—yet, as I've said, it makes no sense to construct a discipline that studies the manner in which a system is acquired or utilized, but refuses to consider the nature of this system.

In my opinion, in order to do good psychology one must start by identifying a cognitive domain—vision, for example—that is to say, a domain which can be considered as a system, or a mental organ, that is more or less integrated. Once that system is identified, one can try to determine its nature, to investigate theories concerning its structure. To the extent that such a theory can be formulated, it is possible to ask on what basis the system is acquired, what are the analogues in it to universal grammar, its biologically given principles. Similarly, study of performance presupposes an understanding of the nature of the cognitive system that is put to use. Given some level of theoretical understanding of some cognitive system, we may hope to study in a productive way how the cognitive system is used, and how it enters into interaction with other cognitive systems. Something like that should be the paradigm for psychology, I think. Of course, this is an oversimplification. One cannot legislate the "order of discovery." But this paradigm seems to me basically correct.

M.R.: That is the approach which you have followed in linguistics. You have identified the system: the competence—and you have proposed a theory, that of generative grammar. Universal grammar is the set of hypotheses that bear on the acquisition of the system and so on. But such is not the customary path of psychology.

N.C.: No, because until fairly recently psychologists have tried to leap over the initial stages; and going directly to the subsequent stages, they have been unable to accomplish as much as they could. Because you cannot study the acquisition or use of language in an intelligent manner without having some idea about this language which is acquired or utilized. If all you know of language is that it consists of words, or if you have a theory of the Saussurean type that tells you: "Here is a sequence of signs, each having a sound and a meaning," that limits very greatly the type of process model you can investigate. You must work with performance models, which produce word-by-word sequences, with no higher structure. You can only work with acquisition models, which acquire a system of concepts and sounds, and with the relations between these systems. That would be a primitive psychology, limited by the conception of language that was the point of departure. The same holds quite generally.

Psychologists often say that they don't presuppose a model of competence, that is to say, a theory of language. But that is not true; they could not do anything without having a conception of the nature of language. Every psychologist presupposes at least that language is a system of words: that is a model of competence. A very bad model of competence, but a model just the same. If they want to do better psychology, they must choose a better model of competence.

Why are many psychologists reluctant to consider richer and more abstract models of competence? Many linguists too, for the matter. In my opinion, because they are still under the influence of empiricist doctrines that are restricted in principle to quite elementary models of competence. These doctrines maintain that all learning, including language acquisition, proceeds by the accumulation of specific items, by the development of associations, by generalization along certain stimulus dimensions, by abstracting certain properties from a complex of properties. If this is the case, the models of competence are so trivial that it is possible to ignore them.

M.R.: When looked at this way, the Saussurean system of signs, conceived as a store slowly deposited in memory, corresponds very well, in effect, with the trivial empiricist model.

Do you know Gregory's experiments on vision? They prove that vision is produced by an interaction between an innate system and experience.

N.C.: Gregory is one of those who are trying to construct a model of competence for vision. That is interesting work, and it seems a logical way to treat these questions. Apparently, the visual cortex of mammals is predetermined in part, with a certain margin of indeterminacy. There exist, for example, cells of the visual cortex which are designed to perceive lines at a certain angle, and others at another angle; but the development of these receptors, their density, in particular, or their precise orientation within a predetermined range of potential orientation, all this depends on the visual environment, so it appears.

M.R.: Vision is thus a construction, like grammar?

N.C.: It seems that the general structure of the visual system is fixed, but the particular detailed realization remains open. For example, it is supposedly virtually impossible to determine precise binocular coordination genetically. It seems that visual experience is required to solve this engineering problem in a precise way, though binocular vision is genetically determined.

In general, serious psychology will be concerned primarily with domains in which human beings excel, where their capacities are exceptional. Language is one such case. There one is sure to find rich structures to study. In the domain of visual perception, for example, one of the most extraordinary abilities is to identify faces. How can one, after having seen a face from a certain angle, recognize it from another angle? That involves a remarkable geometric transformation. And to distinguish two faces! It would be no small task to design a device to match human performance in these respects.

It is possible that the theory of face perception resembles a generative grammar. Just as in language, if you suppose that there are base structures and transformed structures, then one might imagine a model which would generate the possible human faces, and the transformations which would tell you what each face would look like from all angles. To be sure, the formal theories would be very different from those of language ...

M.R.: ... because we are passing from linear sequence to volume.

N.C.: There has also been very interesting work recently on the perceptual system of infants. During the past few years experimental methods have been devised that permit one to work with very young infants, even just a few days old, or a few weeks, and to determine some aspects of their perceptual systems, which exist, evidently, prior to relevant experience. It has been reported, for example, that infants distinguish the phonetic categories P, T, and K, which acoustically form a continuum: there is no line of demarcation between these categories, and no physical necessity to divide the acoustic continuum just this way. But perceptually they do not form a continuum. Particular stimuli along this dimension will be perceived as P or T or K. It seems that infants already make this categorial distinction, which indicates that it must reflect part of the human perceptual system that is not learned, but is rather an innate capacity, perhaps specifically related to language, though this is debated (48-52).