George A. Miller, the author, after a long period of time in the Cambridge area, associated with both MIT and Harvard University, is now at Rockefeller University. His Ph.D. is from Harvard. He is the president of the American Psychological Association and author of Language and Communication and of Psychology: The Science of Mental Life.

The reviewer, Leon A. James, was born in Romania and his education has spanned two continents, three countries, and five languages. His Ph.D. is from McGill University and has taught briefly at Laval, McGill, and Wisconsin before going to the University of Illinois where he is Associate Professor of Psychology and Communications and Co-Director of the Center for Comparative Psycholinguistics. At present he is involved in cross-cultural investigations of subjective culture, bilingualism, and the development of communicative competence. He is co-editor (with Miron) for Readings in the Psychology of Language.

This book, as the sub-title indicates, consists of seven essays, six of which were published between 1956 and 1964. Included in the selections are four papers on man as an information processing organism ("Information and Memory" "The Magical Number Seven...," "The human Link in communication Systems," Computers, Communication, and Cognition"), a well known article entitled, "The Psycholinguists," an essay "Concerning Psychical Research," and a lengthy report (one-third of the book) entitled "Project Gramarama" that appears in print for the first time. Although most of the content is not new, the arguments are still surprisingly fresh. Part of this is due no doubt to the inherent appeal of the metalanguage of information theory. These essays are written in a style for which Miller is justly famous. Sample: "Each letter of the alphabet carries a log² 26 = 4.70 bits of information. ... There are perhaps 1,000 common monosyllables in English, so a rough estimate of the information value of a monosyllabic word selected at random would be about ten bits" (p. 9). Or, again: "A simple sentence in English an easily run to a length of twenty words, so elementary arithmetic tells us that there must be at least 10²⁰ such sentences that a person who knows English must know how to deal with. . . Putting it differently, it would take 100,000,000,000 centuries (one thousand times the estimated age of the Earth) to utter all the admissible twenty-word sentences of English" (p. 80).

On the Number of Semantic features

Arguments of this kind represent more than cute little number games that astonish even upon second reading. They have certain implications for the nature of the theories designed to account for man's ability to use language. Take for instance that argument that man's limit of "the span of absolute judgment" is "usually somewhere in the neighborhood of seven" (p. 32). One important device we use to augment our "channel capacity" is to increase the number of dimensions along which judgments are made, a technique which can be shown to extend the span of absolute judgment from seven to 150. Miller speculates that the "span of perceptual dimensionality"--the maximum number of simultaneous dimensions that can be differentiated --is "in the neighborhood of ten." It is interesting to note the implication of this limit for semantic theories that define the meaning of words in terms of semantic features--simultaneous dimensions along which word can vary. If Miller's estimate on the limit of perceptual dimensionality is correct, then such semantic theories must limit the number of features they postulate for any one conceptual area to about ten (cf. the work on componential analysis). If we assume that a stimulus can be assigned up to seven values on each dimension, there are then 710, or a little under 3 million, different meanings that could be distinguished by man, --a number far in excess of the dictionary entries for any known natural language. It may be reassuring for lexicographers to know that their task has an upper limit short of infinity!!

The Crisis in Psycholinguistics

It is to be regretted that the only recently written article I the book, "Project Gramarama," fails to come to grips with the crisis in Psycholinguistics today. This is the excessive dependence of psycholinguistics on linguistic theories. Chomsky's early formulation of the "kernel" sentence as the base around on which transformations are made (from active to passive, from question to negative, etc.) was quickly followed up by psychologists in numerous experiments. Later, Chomsky modified his theory in such a way as to alter (or even eliminate) the significance of the "kernel" sentence. As a result, the theoretical relevance and usefulness of those psycholinguistic experiments on transformations is doubtful. Given the highly fluid character of present day linguistic formulations, one must wonder about the wisdom of the psycholinguist's strategy of expanding much research effort in the direction of immediate --and often too superficial--"translation" of specific linguistic speculations into psychological experimentation. Not only does he run the real risk of having the carpet pulled form under him (as happened with the kernel sentence), but even more importantly, such efforts are likely to inhibit the formulation of psychological models of the language user.

In retrospect, Miller's conception of what psycholinguistics is all about, as outline in "The Psycholinguists" suffers form this same weakness. His description of sentence understanding as a hierarchical process involving separate levels (phonological mapping, syntactic analysis, semantic interpretation, implication, pragmatic validity) follows closely the earlier formulations of Chomsky (at least for the first three levels). Not only has this view been challenge by second generation generative linguists (cf. Fillmore, Lakoff, McCawley, Ross) but from the psychological point of view it never really made much sense. We've known for some time that auditory decoding is intimately related to the meaning of speech input, that successful syntactic analysis involves semantic interpretation and, of course, vice versa, that both the implicative meaning of the utterance as well as its communicative context affects its interpretability, and so on. A linguistic model of separate processes functioning hierarchically could not form the basis of an adequate psycholinguistic theory of the language user.

The Concept of "Rule"

Much of what linguists have to say today is of dubious relevance to a psychological theory of language. Take for instance the oft quoted argument made by linguists--which Miller also reiterates--that the grammar of English defines as permissible infinitely recursive sentences, and that therefore there is no longest sentence in language. This argument is at the very core of modern linguistics, which views the structure of sentences as expansion via a set of rewrite rules. Such recursiveness follows logically from the linguistic conception of a "grammatical rule" as a formalized expression of the permissible manipulation of defined symbols. It is difficult to conceive of modern linguistic theory without precisely such a conception. Yet it is clear from psychological evidence, as Miller himself points out, that man must be assigned to "the general category of non-recursive devices (p.116)." So, although man is capable (with difficulty) of handling up to three or four embedding in a sentence, such recursiveness certainly is not his characteristic way of processing information. Hence the conception of the "rule" that is to be psychologically relevant as the basis of psycholinguistic theory of language performance must be something quite different from the linguistic one.

The value of Miller's book lies in the fact that given its emphasis on man's peculiar and limited information processing characteristics it may encourage psychologists to now turn their attention back once again to the problem that preoccupied contributors to the Osgood and Sebeok monograph of 1954, which ushered in the age of "The Psycholinguists," namely theories about man, the language user rather than theories about the language he uses. If this enterprise should turn out to be successful, the dependency relation between the sister disciplines in psycholinguistics, which so far has remained a on-way street--psychologists taking their cues from linguists--might progress toward a more mature relationship in which psychological theories of what human kind can do will impose restrictions and limits on linguistic formulations of the structure of language. This is a lesson that psychology itself had to learn several decades ago when newly discovered neurophysiological facts imposed limits on the nature of psychological speculations about the mind. It will be well to remember Chomsky's classification of linguistics as "a particular branch of cognitive psychology" (Language and Mind, p. 1) or his recent definition of the field: "Linguistics is really a theoretical biology, if you like, a theoretical psychology" (quoted in The Nation, September 9, 1968, p.217).