The Message in the Bottle: How Queer Man Is, How Queer Language Is, and What One Has to Do with the Other

15. A THEORY OF LANGUAGE

A Martian View of Linguistic Theory,Plus the Discovery That an Explanatory Theory Does

Not Presently Exist, Plus the Offering of a Crude Explanatory Model on the Theory That

Something Is Better than Nothing

ALTHOUGH THE WRITER of this paper is not, strictly speaking, a Martian, his distance from and innocence of standard linguistic disciplines is, if not extraterrestrial, at least extralinguistic. Accordingly, this paper can commend itself to readers more by reason of its ignorance than its knowledge. What virtues it may have are mainly those of its perspective. I do not presume to compare myself to the boy who noticed that the king was naked nor linguists to the king’s subjects. Yet innocence — and distance — may have its uses. Just as a view of the earth from space may reveal patterns in forested areas and deserts which might be missed by the most expert foresters and geographers — because they are too close — so it is that what follows is what might be seen from a Martian perspective, that is to say, a perspective worlds removed from the several admirable disciplines of linguistics.

Imagine, anyhow, a Martian astronaut of average intelligence and an average scientific education. He lands on earth with the assignment of making a brief survey of the status of earth sciences. After taking cram courses in physical, chemical, and biological sciences, he reads up all he can on theoretical linguistics. Upon his return to Mars, and after thinking things over, he fancies he can see a thing or two which might be useful to earth linguists — by virtue of his perspective. It is somewhat as if he had been in orbit a hundred years ago when Sir Richard Burton and Captain Speke were thrashing around East Africa in search of the Nile’s source. From his altitude the astronaut can beam down instructions to the explorers: “No no no! You’re too far west! Head due east and in fifty miles you’ll strike a large body of fresh water. Proceed north along shore until…”

What the Martian sees in the case of earth linguistics is a, to him at least, remarkable bifurcation of theoretical effort of such a nature that the central phenomenon is straddled and, as he sees it, largely missed — as if Speke were on one side of Lake Victoria and Burton on the other. There is, on the one hand, a triumphant tradition in modern linguistics taking several forms and variously named, “descriptive,” “structural,” “transformational,” associated with people like Bloomfield, Harris, Chomsky — varied theoretical approaches, to be sure, yet sharing one important trait in common: that of approaching the phenomenon of language through a formal analysis of the corpora of languages, an analysis which abstracts both from the people who speak the language and the things they talk about. Semantics or the relation between words and things, the Martian notices, is mentioned now and then but is treated by and large like a bastard at a family reunion. Kinship is admitted, yes — after all words are often used to mean things — but the visitor does not exactly fit into the family. He, the Martian, recalls, for example, Bloomfield’s model where “experience” is stuck onto an otherwise neat hierarchy of phonemes, morphemes, and such, somewhat like a sore thumb; or the transformation model (Chomsky) which specifies a “central syntactic component” to which a somewhat mysterious “interpretive semantic component” is added as a kind of afterthought.

This is one branch of the bifurcation then, the structural-descriptive-generative analysis of language as a corpus. The other branch is not so much a working science as it is a shared belief, a faith that human language must surely be of such-and-such an order. Until a few years ago it was set down, again with all the fervor of an article of faith, that human language must not be different in kind from communication in other species. Proposals to the contrary were taken as a rejection of the entire scientific tradition of the West (Hockett). The traditional model was of course that of the behaviorists and learning theorists with one or another refinement, for example, Bloomfield’s notion of language as “secondary response.” The trouble was that this model worked only in carefully chosen cases: Jack getting thirsty and saying “Water” and Jill going up the hill to fetch it, or Malinowski’s example of the Trobriand Island fisherman shouting “Mackerel here!” whereupon other fishermen respond by paddling over. But the theory didn’t seem to work when, the fishing over, the feast eaten, the islanders were sitting around the fire spinning tales about long past or mythical events.

No less astonishing to the Martian is the more recent countervailing view that human language is utterly unlike animal communication (Chomsky), so much so that it was felt necessary to revive the old Rationalist notion of innate ideas to account for it (Chomsky). Maybe Hockett was right after all. Anyhow, what strikes the Martian most about the controversy is the extreme character of the alternatives. If he understands correctly, it appears as if, once the inadequacy of the behaviorist model is admitted, one has no choice but to chuck “empiricism,” rummage in the philosophical attic, and dust off a somewhat decrepit mind-body dualism. Surely, thinks the Martian, empiricism as it applies to science is not a dogma about the nature of the mind — that it is a tabula rasa or whatever — but rather a proposition about the practice of the scientific method, namely, that the scientist relies on data which he obtains through his senses, to account for which he constructs theories and models, and to confirm the latter he must return to sense data.

In what follows, the Martian will revive another idea, not quite so dusty nor so far removed from the practice of science. This is Charles Peirce’s theory of abduction, which is an analysis of scientific hypothesis formation, peculiarly apposite, as the Martian sees it, to linguistic theorizing, and which avoids such ideological extremes as mechanism and mentalism.

From an orbital perspective, it is possible to make other, more or less elementary observations.

It seems to the Martian, to begin with, that the transformationalists’ assault on the learning-theory model was both long overdue and remarkably successful (Fodor and Katz, Chomsky) and that, as a consequence, the latter has been dismantled and can no longer be entertained as a serious explanation of language as phenomenon. The watershed was probably marked by the appearance of Chomsky’s celebrated review of Skinner’s Verbal Behavior. Linguistic theory would never thereafter be the same.

But, unless the Martian is very much mistaken — and it is here that he does resemble somewhat the boy who noticed something wrong with the king — it appears to him that while the prevailing behaviorist theory has been dismantled, no other theory has been advanced to take its place, this in spite of all the talk by transformationalists about “explanatory models.”

It is somewhat as if the Ptolemaic geocentric universe had been dismantled but Copernicus had not yet come along with his heliocentric model.

Accordingly, the assumption will be made in what follows that linguistic theory has not yet reached the level of explanatory adequacy of, say, seventeenth-century biology. It was then that, following the work of Harvey and Malpighi, it became possible to construct crude but accurate models of cardiac and renal function; to suppose, for example, that the heart is like a unidirectional pump or the kidney is like a filter. One may not say as much at the present time about the unique human capacity for language. True, a schema of sorts has been suggested (Chomsky; Katz) to show what happens when a child exposed to fragments of a language acquires a competence in that language: primary linguistic data→ LAD→ Grammar (where LAD is the Language Acquisition Device). What seems fairly obvious, however, is that despite claims to the contrary this schema is in no sense an explanatory model. It is no more than a statement of the problem under investigation. The “LAD” appears to be a black box whose contents are altogether unknown.

Finally, the Martian shall make bold to put forward a crude model not entirely of his own devising — Charles Peirce is its earthly progenitor.

1. Descriptive or structural linguistics cannot be regarded as a theory of language if the word theory is used as it is used in other sciences.

Structural or descriptive linguistics (Harris) deals with regularities in certain features of speech. These regularities are in the distributional relations among the features of speech in question, i.e., the occurrence of these features relative to each other within utterances. The procedure of structural linguistics is “to begin with the raw data of speech and with a statement of grammatical structure…essentially a twice-made application of two major steps: the setting up of elements, and the statement of the distribution of these elements relative to each other. First, the distinct phonological elements are determined and the relations among them investigated. Then the distinct morphological elements are determined and the relations among them investigated.” (Harris)

Such a discipline is undoubtedly beyond reproach — as far as it goes. Indeed one might well agree with Lévi-Strauss in setting up the method of structural linguistics as a model for anthropologists, a distributional method which Lévi-Strauss in fact applies to other cultural phenomena such as art, myth, ritual, religion, even cooking (Lévi-Strauss).

Yet it must not be forgotten that this method, rigorous as it is, does not pretend to be other than descriptive. Thus if one were studying hematology, one could imagine a science called “structural hematology” which consisted of a description of the cellular and chemical components of the blood and of certain “distributional” relations between them, e.g., a high nonprotein nitrogen is regularly associated with a low hemoglobin. But such a discipline, however rigorous, could never serve as a theory of blood formation or as a theory of the function of hemoglobin. Accordingly, the method of structural linguistics has at its core a fundamental ambiguity. This ambiguity can be expressed by two questions which presently go not only unanswered but unasked: (1) Does structuralism make the assumption that both language and culture are by their very nature phenomena of such an order that the search for distributional regularities is a terminus ad quern for both linguists and anthropologists? (2) Or is it rather the case that the current status of both arts is so primitive that we are necessarily at a stage comparable to Linnaean taxonomy, and so it goes without saying that at some future time linguistics shall arrive at a general explanatory theory bearing roughly the same relation to descriptive linguistics that Darwinean theory bears to Linnaean taxonomy?

The answer to either question is not clear, because, for one reason, the questions are not asked. What is clear is that in any case descriptive or structural linguistics is not a theory of language in the ordinary use of the word theory.

2. Behaviorism, both in its early Pavlovian and Watsonian versions, and in the later refinements of modern learning theory, does indeed offer a plenary model of language as phenomenon, which meets all the specifications of explanatory theory except one: It is wrong.

S-R theory, however modified by little s’s and r’s, by “intervening variables,” “dispositions to respond,” “habit structures,” “generalization and analogy,” “stimulus control,” “network of associative connections,” and the like, fails to address itself to, let alone explain, those very features of language behavior which set it apart from other forms of animal communication, e.g., the phenomenon of symbolization or naming, the sentence as the basic unit of language behavior, the learning performance of a child, who, upon exposure to a fragmentary input of a language, is able to utter and understand any number of new sentences in the language — this after a relatively short period and without anyone taking much trouble about it.

3. Transformational generative grammar is not an explanatory theory of language, although it has been advertised as such (Chomsky). That it fails to serve as such is not a consequence of its stated objective, which is in fact correct; namely, to specify the character of the device which mediates the processing of the input, the primary data, and the output, the grammar of the language. Nor does it fail as theory primarily because of the putative and unconfirmed status of so-called “deep structures,” from which surface structures are derived by transformations (Hockett).

Transformational grammar is not an explanatory theory of language as phenomenon but rather a formal description, an algorithm, of the competence of a person who speaks a language. There is no evidence that this algorithm bears a necessary relation to what is happening inside the head of a person who speaks or understands a sentence. There is evidence in fact that it does not.

Transformational grammar also fails as theory because it violates a cardinal rule of scientific explanation, namely, that a theory cannot use as a component of its hypothesis the very phenomenon to be explained. That is to say, if one sets out to explain the appearance of an apple on an apple tree, it will not do to suppose that apple B, which we have in hand, derives from putative apple A, which we hypothesize as its progenitor. An adequate account of the origins of either apples or sentences must contain in the one case only nonapple elements, e.g., pollen, ovary, ovule, etc., and in the other case nonsentential elements. So. it will not do for an explanatory theory of language which must presumably account for the utterance and understanding of a sentence or “surface structure” to hypothesize a “deep structure” as its source when deep structures are themselves described as “kernel sentences” (Chomsky) or “underlying propositions” (Chomsky), when in fact it is the phenomenon of sentence utterance itself in whatever form, kernel sentences or propositions, that is unique among species and therefore, one would think, the major goal of theorizing.

3.1. The main error of a generative grammar considered as a theory of language is that its main component is syntactical with semantic and phonological components considered as “interpretations” thereof (Chomsky). This awarding of the prime role to syntax rules out nonsyntactical elements, for example, semological and phonological components, as primitive generative components of deep structures and in effect posits syntax as an underivable and therefore unexplainable given.

Thus, the phrase-marker or general rule of sentence formation is given as the coupling of syntactical elements:

S → NP-VP

In point of fact, the phrase-marker does not represent the class of all classes of sentences or even the class of all declarative sentences. Consider the one-word sentence in which a child points at a round red inflated object and looks questioningly at his father, whereupon the father says Balloon. Where are the NP and VP of this sentence? What have been coupled here are not two syntactical elements but a class of sounds and a class of experienced objects. The only way this one-word naming sentence can be captured under a syntactical rubric is to consider it as an elliptical version of the NP-VP form That is a balloon, and even this strategy does not give an account of the special semiotic status of the demonstrative that. To parse the one-word naming sentence syntactically is to tailor data to fit theory. It is rather for theory to accommodate data.

It is important to notice, as we shall presently see, that while this one-word sentence does not fit the syntactical phrase-marker NP-VP, it fits very well the more informal definition of a sentence given by the ordinary-language analysts as comprising (1) what one talks about and (2) what one says about it (Strawson). What the father and child are talking about is the experienced balloon itself considered as a member of a class of such objects. What the father says about it is that it is named by a class of sounds balloon.

3.2. The phrase-marker NP-VP describes a subclass of declarative sentences. A general rule of sentence formation must accommodate nonsyntactical as well as syntactical elements such as “noun phrases” and “verb phrases.”

3.3 The central component of the LAD is not syntactical but rather semological-phonological (Chafe). The syntactical component is nothing more nor less than the formal properties which issue from the semological-phonological linkage and later from combinations of semological-phonological complexes or semantically contentive words (Brown’s usage, see Brown and Bellugi).

Thus the formal properties of syntax are already present in two-word combinations of “contentives” in children’s sentences. Later these formal properties are explicitly marked by the addition of “functors” (Brown and Bellugi).

3.4. Accordingly, a rule of sentence formation must be sufficiently general to accommodate both the purely syntactical NP-VP sentence and the naming sentence in which a class of experience (semology) is linked with a class of sounds (phonology) (Chafe).

Some such general rule may be formulated as follows:

(1) Sentence → S (is) P

The “subject” and “predicate” form seems advisable for no other reason than that these terms are sufficiently general to accommodate both naming sentences and NP-VP sentences. Thus “subject” and “predicate” are used not to revive Aristotelian categories but as a shorthand description of a sentence as comprising (1) what one talks about and (2) what one says about it. “Subject” fairly describes both the balloon in the one-word naming sentence Balloon and the NP in The boy hit the dog.

The copula is indispensable because something is asserted in a sentence. The parenthesis is added because such assertion does not always require the presence of a verb phrase, e.g., Balloon.

Subclass (1a), the naming sentence, might be formulated by the semiotic rule

(1a) Sentence → IEc (is) Sc

where I is an index, either an item of behavior, a pointing at or looking at, or some such functor as that in That is a balloon; E is an experience, subscript c added to indicate that it is not such-and-such an experience which is pointed at or looked at as a singular but rather as one of a class of such experiences. Similarly, Sc is not such-and-such a sound but rather an utterance understood as a class of such sounds (phonology). To use Chafe’s terminology, we are dealing not with substances but with forms (Chafe).

Note that it is preferable to use Ec to designate a semological class of experience rather than, say, Oc for a class of objects, e.g., ballons. For in fact a child not only names things (table, doggie, ball) but actions (play, see, drop) and qualities (blue, broke, bad) (Brown and Bellugi). What such words have in common is not a syntactical property but a semantic property. Thus they are not all nouns; some are verbs and adverbs. But they are all semantically contentive.

The second subclass is the standard “syntactical” declarative sentence form:

(lb) Sentence → NP-VP

Sentences of this form, I shall suggest, appear first in children’s speech as those two-word combinations in which contentive words are selected from an inventory of semophones stored up by naming sentences and are paired to form primitive versions of adult NP-VP sentences. Thus: Bobby wet, Doggie fall, Mommy lunch (Mommy had her lunch), etc.

3.5. Four immediate advantages accrue to a linguistic model which proposes a semological-phonological linkage as its genetically prime component:

(1) It is a transsyntactical theory; that is, it is founded on a general semiotic — the science of the relations between people and signs and things — which specifies syntax as but one dimension of sentential theory. Accordingly, it provides theoretical grounds for distinguishing between the two types of declarative sentences, the naming sentence and the NP-VP sentence.

(2) It accords with the data of language acquisition and provides a model for understanding the ontogenesis of speech in children, in particular the stages of word and “phrase” acquisition, a sequence which is presently accounted for by purely descriptive “generative rules of phrase formation.”

(3) It allows the possibility, as we shall see, of looking for a neurophysiological correlate of such a model, a possibility which is disallowed in principle by a generative theory which postulates syntax as a central underived component.

(4) It permits the assimilation of linguistic theory to a more general theory of all symbolic transactions, a theory which must in turn accommodate such nonsyntactical “sentences” as metaphor, a painting, a sculpture, a piece of music. From this larger perspective it will be seen that the division of language into two kinds of sentences is not as arbitrary and unsatisfactory as it might at first appear. Rather is it the case that the standard syntactical sentence of language, the coupling of subject and predicate, is a special case of the more fundamental human capacity to couple any two things at all and through the mirror of the one see the other. Thus, the child’s sudden inkling that the thing ball “is” the sound ball is the progenitor not only of all future sentences about balls but also of his grasp of metaphor, art, and music.

4. The two kinds of sentences formed by (1a) and (1b) can be regarded not only as representing the formal subclasses of constative sentences — anyone at any time can name things with one-word sentences or assert propositions about things and events and relations — but also as delineating major stages in the ontogenesis of language. In the initial naming stage of language acquisition, the first sentences children utter are the linking of semological elements (forms of experience) with phonological elements (forms of sound). As a consequence of this extraordinary naming activity, a repertoire of semological-phonological complexes or “contentive” words is formed. For the sake of convenience I propose to call these semological-phonological complexes “semophones.” Once such a lexicon of semophones is available, it becomes possible by combining any two semophones to form a large number of primitive NP-VP sentences.

4.1. One test of a theory of language is its utility in accounting for the acquisition of language, in particular the ontogenesis of speech as it is observed in intensive studies of individual children.

Judged by the standard of adult syntax, the early manifestations of speech in children appear vagarious and fragmentary. In studies of individual children, such speech forms have been described variously as “single-word utterances,” “phrases,” “holophrases,” “sentence fragments,” “telegraphic sentences,” and so on.

Such studies, however, generally agree in specifying certain stages of language development.

Some time around the end of the first year of life, most children go through a naming stage. As Brown and Bellugi put it, at this period “most children are saying many words and some children go about the house all day long naming things (table, doggie, ball, etc.) and actions (play, see, drop, etc.) and an occasional quality (blue, broke, bad, etc.).

It is interesting to note that the best-known studies of the acquisition of speech in children (Braine, Brown, Ervin, McNeill), while taking note of the naming stage and of “one-word utterances,” skip over it and address themselves to “phrases” of two or more words. The assumption seems to be made both that there is assuredly such a thing as a naming stage and also that there is not much to be said about it. This may be true, but it nevertheless seems curious, considering the fact that no other species on earth ever names anything at all, much less goes about naming everything under the sun or asking its name, that investigators of the genesis of language in children should not have been more intrigued by this apparently unique activity. On the other hand, what is one to say about it? A child names something or hears it named, understands or misunderstands what is named, and that is that. One can only conclude either (1) that the phenomenon of naming is the most transparent of events and therefore there is little to be said about it, or (2) that it is the most mysterious of phenomena and therefore one can’t say much about it. As Fodor said, nobody knows what a name is.

I wish to suggest that one reason for the indifference of these psycholinguistic studies to the naming stage of language acquisition is a necessary consequence of a commitment to structuralism as linguistic theory. That is to say, if one regards grammatical patterns and distributional relations as the primary goal of linguistics, one can’t have much to say when confronted with a single word. For as soon as theory abstracts from behavior and the relation of words to things, and addresses itself only to the relation of words to words, the theorist can only watch the naming child with bemused interest and mark time until he begins to put two or more words together.

The second stage of language acquisition is characterized by two-word utterances, usually described as “pivot-open” constructions. The pivot class has fewer members than the open class. Thus, a child will say my sock, my boat, my fan, or big plane, big shoe, big sock, etc. (Braine); that knee, that coffee, that Adam, or two coat, two stool, two Tinkertoy, etc. (Brown and Bellugi); this arm, this baby, this yellow, or the other, the pretty, the dolly’s, or here baby, here yellow, etc. (Ervin). The following rule therefore holds for both single-word utterances and pivot-open combinations (McNeill):

S→ (P) + O

which would account in a purely descriptive manner for such utterances as ball or my ball or here ball or there yellow or there drop, whatever syntactical or sentential differences may exist.

It has been noted that members of the pivot class are usually functors (a, that, the, here) but not always (pretty, big), while open-class words are nearly always contentives (boy, coffee, sock, knee, wet, yellow).

The third stage is characterized by two developments:

(1) Differentiation of the pivot class.

For example, car, through successive expansions a car, that a car, that a big car, eventually reaches its adult form that is a big car (see Brown and Bellugi; McNeill). Descriptive rules can be inferred:

NP → (P) + N car, a car

NP₁ → Dem + Art + M + N that a my car

NP₂ → Art + M + N a big car

NP₃ → Dem + M + N that big car

NP₄ → Art + N the car

NP₅ → M + N my car

NP₆ → Dem + N that car

These rules allow, for example, that a big horsie but not a big that horsie.

(2) The open-open construction.

Instead of saying here man, here car, here coffee or a bridge, a man, a daddy, the child begins to combine two words of the open class: man car (a man is in the car), car bridge (the car is under the bridge), coffee Daddy (here is coffee for Daddy), etc. (Braine).

These open-open constructions are often uttered in strong contextual situations, for example, where mother and child are looking at the same thing. The mother, in other words, is a reliable interpreter. Indeed, the mother actually repeats and expands the child’s utterance, keeping the order of the contentives but adding functors and inflections, as much as to say, “Isn’t this what you mean?” Some examples of open-open combinations with the mother’s “interpretations” and expansion and, presumably, the child’s approval (from Brown and Bellugi):

Child Mother Baby high chair Baby is in the high chair Mommy eggnog Mommy had her eggnog Eve lunch Eve is having her lunch Mommy sandwich Mommy’ll have a sandwich

Note one consequence of the transition from the pivot-open to the open-open construction. The child’s discovery of the latter makes possible an almost exponential increase in the number even of two-word utterances — a fact of the highest significance in any attempt to account for that unique characteristic of the Language Acquisition Device: the ability to utter and understand any number of new sentences. The number of two-word combinations noted by Braine in one child observed at regular intervals went so: 14, 24, 54, 89, 350, 1400, 2500 + (Braine).

Such then is a too-brief and much-oversimplified summary of the highlights of language acquisition as it is actually observed to occur in children.

Judged by the criterion of adult NP-VP syntax, child’s speech appears somewhat fragmentary and vagarious, an assortment of one-word utterances, phrases, fragments, “pivot-open” and “open-open” combinations, the whole odd lot accounted for by purely descriptive “generative phrase rules.”

The real issue seems to be whether the utterances of children, or of anyone at all for that matter, can be understood by a textual analysis which abstracts from the behavior of people who utter sounds to each other about one thing or another. What if sentences have components other than lexical items? If this is so, then a theory of the ontogenesis of language deriving exclusively from the study of corpora of speech must necessarily issue in a purely descriptive structuralism or a formal deductive calculus which transforms one kind of sentence to another by “rules” which, as matters stand now, cannot even in principle be correlated with anything that happens inside people’s heads.

One wonders therefore: How might the development of child’s speech appear when viewed by a transsyntactical theory of sentences, which sees the conventional NP-VP construction as a subclass of the class of all sentences?

I think it can be shown that if the speech of children is viewed not merely as a corpus from which certain descriptive grammatical rules can be inferred but as a behavior which implicates both syntactical and “nonsyntactical” elements, it is possible to arrive at more general semiotic rules of sentence formation. The various descriptive rules of “phrase formation” will then be seen to be coherent stages in the emergence of sentence utterances.

(1) The “one-word utterance,” so characteristic of the first stage of language acquisition, is nothing more Or less than the earliest appearance of the naming sentence, a complete sentence in semiotic terms albeit lacking some later syntactical and functional elements.

The rule involved is not a phrase rule, such as

NP → N

(Brown and Bellugi)

where NP is “noun phrase” and N a “noun,” or a rule for “pivot-open” combinations

(P) + O

but rather the general behavioral rule of formation of the naming sentence

S→ (I) (E_c) (is) S_c

where I, the index, is in this case an item of behavior (e.g., a pointing at or looking at), E_c is the thing or quality or action experienced by the child and indicated as one of a class of such experiences, (is) is the copula dispensed with until the final adult form, S_c is the contentive word, usually a sound — e.g., noun ball, adjective yellow, verb hop—uttered as a member of a class of such words. (As Charles Peirce would say, a contentive word or symbol is not a single thing but a kind of thing [Peirce]).

(2) A descriptive rule for the formation of pivot-open constructions has been given as

S → P + O (e.g., a knee, a man, a Mommy [McNeill])

A more general rule, which allows for the expansion and differentiation of the pivot class, is given by Brown as a rule for generating noun phrases:

NP → (Dem) + (Art) + (M) + N

This rule accommodates not only one-word utterances, car, boat, yellow, and pivot-open constructions, my car, my boat, my yellow, but also such “phrases” as that a car, that a my car, etc.

What needs to be noted, however, is that none of these expressions is a phrase, save only in that sense decreed by a purely syntactical definition of sentences. Rather, all are complete semiotic sentences as provided by the general rule for naming sentences (1). Where this rule differs from Brown’s descriptive phrase rule is in its specifying (a) that what is formed is not a phrase but a sentence, (b) that the index I is a general semiotic class of which Brown’s demonstratives (that, there) belong to the syntactical subclass, (c) that the copula is added because the final adult form requires it.

Thus it is not a grammatical vagary, to be accounted for by a descriptive rule, that a child may say that a blue flower but never a that blue flower or blue a that flower (Brown and Bellugi). What Brown calls the “privileges peculiar to demonstrative pronouns” is in fact specified by the more general semiotic rule which requires the initial positioning of the index (I), whether it be a demonstrative (there, that) or a behavioral item (looking, pointing). You can’t name something for someone without first pointing at it.

To summarize (1) and (2): Nearly the entire class of pivot-open constructions, plus the entire class of expanded utterances generated by the rules for “differentiation of the pivot class,” are not “phrases” to be accounted for by lists of descriptive generative rules. Rather are all such utterances specified by a general semiotic rule of sentence formation, in this case the naming sentence, in which both behavioral and syntactical components are ordered by presiding semiotic considerations.

(3) The open-open construction, which appears somewhere around the second birthday and accounts for the exponential explosion of language, is semiotically different from the naming sentence. It is in fact nothing more nor less than the adult NP-VP sentence without functors. It always comprises the pairing of contentives, often “nouns,” but also quality words and action words, e.g., baby wet, car go, man car (the man is in the car), etc.

Braine uses a juncture symbol (#) between the two words of an open-open construction to distinguish it from an otherwise similar utterance having a quite different meaning. Thus, again relying on the mother as the best of all interpreters:

baby#chair (The baby is in his chair)

baby chair (There is the baby’s chair)

The juncture symbol # is thus a semiotic mark which might be inserted between the “subject” S and “predicate” P of all sentences, whether naming sentences or NP-VP syntactical sentences. It marks a behavioral pause between what I am talking about and what I say about it.

Accordingly, it will be seen that the conventional syntactical “phrase marker” is in fact a special instance of a more general semiotic structure. Everyone is familiar with the NP-VP phrase-marker for a sentence like The baby is in his chair. A quite different but equally justifiable semiotic phrase-marker could be designed for the naming sentence baby chair, as diagrammed in Figure 10. There is only one lexical-syntactical item in this sentence, the NP baby chair. The other elements are behavioral (pointing, pause) or environmental (chair experienced as a class member).

There is only one lexical-syntactical item in this sentence, the NP baby chair. The other elements are behavioral (pointing, pause) or environmental (chair experienced as a class member).

Contrast the semiotic phrase-marker of the naming sentence baby chair with the conventional phrase-marker of the open-open construction baby#chair (the baby is in his chair), as in Figure 11.

To summarize 4.1: There are, semiotically speaking, two basic classes of linguistic sentences (we will say nothing here about nonlinguistic “sentences,” e.g., van Gogh’s painting “The Chair,” which assuredly uses a symbol to assert something about something else): the naming sentence and the NP-VP sentence.

Both kinds of sentence are acquired, understood, and uttered without the use of functors and other syntactical forms.

The addition of functors to child speech can be understood as the behaviorally necessary substitutes for a diminishing context. Thus a two-year-old child, sitting on his mother’s lap and looking with her out the window and saying boy lawnmower can be reliably understood to mean: The boy is pushing the lawnmower. But as context drops away, until at length the child is twelve and is reporting over the telephone to his mother about the performance of the gardener, the speaker needs his functors and must say Yes, the boy is still pushing the lawnmower.

5. As a genetic theory of language acquisition, we may hypothesize two basic stages, at each of which occurs a coupling of elements and in neither does it seem necessary to postulate a “deep syntactical structure” from which uttered sentences are derived by a series of transformations.

(1) The formation of semophones by the coupling of semological and phonological elements.

Much of the linguistic activity of the first two years of life goes toward the building up of an inventory, or lexicon, of semantically contentive words through which the world of experience is segmented, perceived, abstracted from, and named. So enduring and stable are these semological-phonological combinations that it seems appropriate to regard them as sound-meaning units, perhaps to be designated by some such term as semophone—the “phone” in this case signifying not a phone in the technical linguistic sense but rather the hierarchy of sound units: sound, phoneme, morpheme, word (Chafe).

As the neurophysiological correlate of such a coupling, one can only suppose that there come to be established stable functional interconnections between the visual and auditory cortexes. May we not at this point make bold to reach for the explanatory level of seventeenth-century physiology with its crude but accurate models of body functions — the heart is like a pump, the kidney is like a filter? Accordingly, may we not suggest that the LAD is like a coupler?

The semological and phonological components of the semophone are thoroughly interpenetrated. The resulting configuration is a much more stable and enduring entity than can be expressed by association psychology. Thus it is not so much the case that words like yellow, wet, glass, hop, Elmer, quick “call up” such and such an association or have such and such a “connotation.” Rather is it the case that these sounds are interpenetrated and transformed by the classes of experience to which they refer. The contentive word in a sense contains the thing. Yellow becomes yellow.

(2) The formation of NP-VP sentences by the coupling of semophones.

Semophones are paired in the child’s open-open constructions to form the basic or contentive elements of the adult NP-VP sentence. One of the major tasks of an explanatory linguistic theory is to account for the practically unlimited number of new sentences which can be uttered and understood by a three or four-year-old child, following the input of limited and fragmentary data. If the basic component of the LAD is a coupler, it will be seen that this extraordinary generative capacity can be accounted for in two ways:

(1) The exponential increase in the number of open-open constructions which can be formed once an inventory of semophones is established. Thus, an inventory of n semophones (car, wet, Daddy, sock…) will yield an n2—n number of open-open sentences (car wet, car Daddy, Daddy wet, Daddy sock...). An inventory of 100 semophones will yield a possible 9,900 open-open sentences.

(2) A single open-open sentence is susceptible to as many sentential interpretations as context allows. Thus a child, sitting on his mother’s lap and looking out the window, who utters the sentence car Daddy, can be reliably understood by his mother to be saying Daddy is getting in the car, Daddy is washing the car, Daddy is kicking the car, depending on whether in fact Daddy is getting in the car, washing the car, kicking the car.

Indeed, the number of sentences made possible by (1) the exponential increase of the number of open-open combinations and (2) the contextual application of any one such combination to any number of mutually perceived situations becomes, for all practical purposes, unlimited.

5.1. Two characteristic transformations occur in the two types of coupling or sentence formation.

In (1), the naming sentence, the phonological element is transformed by the semological element. Yellow becomes yellow, wet becomes wet, hollow becomes hollow.

In (2), the open-open coupling, it is the linkage itself which is transformed. Thus the linkage between Daddy and car in Daddy#car becomes is getting into, is washing, is kicking, as the case may be.

Functors or grammatical markers are added to open-open combinations not as a result of overt imitation of adult sentences (Ervin) but the other way around, through the parent’s imitation and expansion of the child’s sentence (Brown and Bellugi).

Presumably the exigencies of communication require that, as context is withdrawn, functors be added. With the child’s increasing mobility and his increasing number of reports of what has happened out of the hearer’s sight, functors come into play. The following conversation occurred between my two-year-old grandson, arriving in some excitement to make a report, and his mother:

Child: Daddy tractor!

Mother: Daddy is driving the tractor?

Child: (Silence)

Mother: Daddy is fixing the tractor?

Child: (Silence with a half nod)

Mother: Daddy is under tractor?

Child: Daddy under tractor!

5.11. The question must be raised about grammatical transformations: Is there any evidence to support the theory that the so-called grammatical transformation, whatever its usefulness to the linguist as an analytical tool, actually operates in the acquisition of language?

Thus, it may be unexceptionable to say with Chomsky that:

If S₁ is a grammatical sentence of the form

NP₁-Aux-V-NP₂

then the corresponding string of the form

NP₂-Aux +be +en-V-by +NP₁

is also a grammatical sentence.

Using this transformation rule, one can obtain Lunch is being eaten by John from John is eating lunch.

But it does not necessarily follow that because a linguist analyzing the corpus of a language can derive one kind of sentence from another kind of sentence by a rule, a formal operation, or because he hypothesizes putative “deep structures” from which “surface structures” are generated by “transformations,” this is what happens when a child learns a language. Indeed, if one follows the principle of parsimony in theorizing, one wonders why such a formal schematism cannot be dispensed with altogether.

For is not the adult passive sentence already implicit in the early open-open construction, later to be filled out by the required functors which the child learns through adult imitation and expansion?

For example, keeping in mind the general S-P form of the sentence, that is, its division into what one is talking about and what one says about it, one can easily imagine some such sequence as follows: Mother and child are watching a dog from the window. Various events occur in which the dog is both subject of attention and subject of sentences about these events. Whether the dog does things or whether things happen to the dog, the dog is what we are talking about.

Child Mother

Dog run Yes, the dog is running.

Dog man Yes, the dog is barking at the man.

Dog car Yes, the dog is chasing the car.

Dog car! Yes, the dog was run over by the car!

Conceivably, then, the child might “verify” the mother’s interpretation of the last sentence by adding one of her passive functors: Dog run by car! (Such, in fact, was my grandson’s first attempt at the passive: his imitation of an adult’s expansion of his original open-open construction.)

The question of course can only be answered by systematic behavioral studies. The point is that it is open to confirmation, or nonconfirmation, by such studies.

6. Is an explanatory theory of language possible?

In this connection, I would like to mention Charles Peirce’s theory of abduction, or explanatory hypothesis, as a valid and possibly useful strategy in approaching language as a phenomenon. My reasons for doing so are two: (1) The present state of theoretical linguistics, considered as a natural science, is so confused, comprising as it does incoherent elements of structuralism and “learning theory” and even Cartesian mentalism, that it might be worthwhile to take a step back, so to speak, in order to view the phenomenon of language from the perspective of perhaps the best-known American theorist of the scientific method. If one should object that Peirce’s theory is almost a hundred years old, I can only reply that since theoretical linguistics is at least three hundred years behind theoretical physics, Peirce can be regarded as being, linguistically at least, ahead of his time. (2) Peirce’s theory of abduction has been revived recently (Chomsky) but in such an odd and what I consider a wrongheaded fashion that there is some danger that its usefulness to linguistics might be permanently impaired.

The assumption will be made then that an explanatory theory of language does not presently exist: that behaviorism does indeed provide an explanatory model but that it is wrong; that structural linguistics and transformational grammar are not explanatory theories (v.s.).

Further, we will accept the following description of both the problem at hand and our ignorance: that every normal human being, and no doubt most abnormal ones as well, are uniquely equipped with what can be characterized abstractly as a Language Acquisition Device (LAD) whose structure and function are unknown but which receives as input primary linguistic data, speech from fluent speakers within hearing range, and has as its output a competence in the language, that is, the ability to utter and understand any number of new sentences (Katz; Chomsky). Now how does Peirce’s theory of abduction relate to the problem at hand, namely, approaching the black box, LAD, toward the end of discovering its workings? Let us reassure ourselves at the outset. Surely the enterprise is worth undertaking, if for no other reason than the depth of our ignorance and the wide divergence of the guesses on the subject. In view of the uniqueness of the human capacity for speech, how different are these workings from the workings of other brains? Are they qualitatively different or quantitatively different? Does the black box hold Cartesian mind-stuff or S-s-r-R neuron circuitry? or both? Certainly it would be a start in the right direction if we had some notion of what to look for, what kind of thing. It is here that abduction or Peirce’s explanatory hypothesis might be of some help.

Let us be clear, first, how Peirce distinguished abduction from induction and deduction, the other two logical components of the scientific method. He believed that neither deduction nor induction could arrive at explanatory theory but only abduction. No new truth can come from deduction or induction. Deduction explores the logical consequences of statements. Induction seeks to establish facts. Abduction starts from facts and seeks an explanatory theory (Peirce). As a classical example of abduction, Peirce cited Kepler’s theory of the elliptical form of Mars’s orbit. Though Kepler had made a large number of observations of the longitudes and latitudes of Mars, and even if he had made a million more, no induction or generalization from these facts could have arrived at the nature of Mars’s orbit. At some stage or other, Kepler had to make a guess, construct a model, then see if the model would (1) fit all the facts at hand and (2) predict new facts which could be verified by observation.

Peirce listed three kinds of abductions or explanatory hypotheses: (1) those which account for observed facts through “natural chance” or statistical methods, e.g., the kinetic theory of gases; (2) those which render the facts necessary through a mathematical demonstration of their truth, e.g., Kepler’s elliptical theory of planetary orbits; and (3) those which account for facts by virtue of the very economy and simplicity of the explanatory model (Peirce).

Presumably we are looking here for 3, at least for the present. Certainly no statistical method or mathematical model known to me has any relevance to what goes on inside a child’s head when he acquires language in the second and third year of life.

Peirce also makes much of the fact, and Chomsky echoes it, that “man’s mind has a natural adaptation to imagining correct theories” (Chomsky). A physicist comes across some new phenomenon in the laboratory. According to Peirce, there are “trillions of trillions of hypotheses” which might be made to account for it, “of which only one is true” (Peirce). Yet as matters usually turn out, the physicist usually hits on the correct hypothesis “after two or three or at the very most a dozen guesses.” This successful guessing or hypothesizing of scientists is not, according to Peirce, a matter of luck. Peirce’s own explanation of the extraordinary success (in the face of such odds) of scientific theorizing is founded in his own allegiance to philosophical realism, the belief that general principles actually operate in nature apart from men’s minds and that men’s minds are nevertheless capable of knowing these principles. But how is this possible? Peirce hazards the guess that, since “the reasoning mind is a product of the universe,” it is natural to suppose that the laws and uniformities that prevail throughout the universe should also be “incorporated in his own being” (Peirce).

Maybe so. This is only speculation, however interesting, about why abduction works. What concerns us here, entirely apart from Peirce’s philosophical realism and his explanation of it, is his theory of abduction itself, which I take to be nothing more nor less than the method of hypothesis formation as it is used in practice by scientists in general, whether one is theorizing about why volcanoes erupt or why people speak and animals don’t. Peirce’s theory of abduction, particularly of the third type, is both sufficiently rigorous that it achieves the level of explanatory adequacy and sufficiently nonspecific that it does not require a commitment to ideology and hence does not fall into the deterministic trap of behaviorism and learning theory. Explanatory theory at the level of the human acquisition of language, it seems fair to paraphrase Peirce, does not require a mechanism, or, in Peircean terms, a “dyadic” model.

I have taken the trouble to review Peirce’s theory of abduction both because of its possible value to linguistic theory and to call attention to the odd use to which Chomsky has put it.

Chomsky has revived Peirce’s theory of abduction, not in order to arrive at an explanatory theory of language, but rather to attribute the capacity for abduction to the child who acquires language. In the same way that, as Peirce speculated, man is somehow attuned to the rest of the universe so that he is able to theorize successfully about it, so it is that “knowledge of a language — a grammar — can be acquired only by an organism that is ‘preset’ with a severe restriction on the form of grammar.”

This innate restriction is a precondition, in the Kantian sense, for linguistic experience, and it appears to be the critical factor in determining the course and result of language learning. The child cannot know at birth which language he is to learn, but he must know that its grammar must be of a predetermined form that excludes many imaginable languages. Having selected a permissible hypothesis, he can use inductive evidence for corrective action, confirming or disconfirming his choice. (Chomsky)

What is odd of course is not Chomsky’s idea that language can only be learned by an organism “preset” with a severe restriction on the form of grammar — this squares very well with the suggestion made in this paper that all sentences in any language must take the form of a coupling made by a coupler — but rather Chomsky’s proposal to shift the burden of explanation from the linguist, the theorist of language as a phenomenon, to the child, the subject under study. Chomsky’s theory of language is that the child is capable of forming a theory of language.

Now Chomsky’s abdication may or may not be justified. Perhaps in the long run it will turn out that it is not possible to arrive at an explanatory theory of language in any ordinary sense of the word and that the only “explanation” available is that the child somehow hits on the grammar of a language after a fragmentary input. If this is so, we must face up to the fact that we have reverted to homunculus biology, explaining human potentialities both in spermatozoa and in children by supposing that each somehow has a little man locked inside. I believe, however, that by the serious use of abduction, hypothesis, not the attributing of it to the child but the figuring out of what goes on inside the head of a child, the theorist can hope to make a start toward the construction of a relatively simple and parsimonious model along Peircean lines.

It is curious to note in passing that if one is seeking philosophical progenitors for Peirce’s theory of abduction and the realism underlying his analysis of the scientific method, one is inevitably led not to Descartes and a mind-body dualism but, according to Peirce, to Duns Scotus!

7. Suppose one were to advance the following tentative hypothesis:

The basic and genetically prime component of the LAD is a semological-phonological device through which semological elements are coupled with phonological elements. Such linkages form a finite inventory of semological-phonological configurations or “semophones,” stable functional entities which correspond to semantically contentive words, e.g., wet, yellow, sock, knee. These semophones in turn become available for couplings to form a large number of “open-open” combinations, which are nothing less than primitive forms of the adult NP-VP sentence.

If this is the case, two questions arise: (1) Does such a model allow the possibility of looking for a neurophysiological correlate of the LAD — a possibility apparently disallowed by a basically syntactical model — and (2) if so, is there presently any evidence of such a correlate?

For some time I had supposed that the basic event which occurs when one utters or understands a sentence must be triadic in nature (Percy). That is to say, sentences comprise two elements which must be coupled by a coupler. This occurs in both the naming sentence, when semological and phonological elements are coupled, and in the standard declarative NP — VP sentence which comprises what one talks about and what one says about it.

Thus, when the father in Peirce’s example points out an object and utters the sound balloon and the son looks and nods, an event of the order of that shown in Figure 12 must occur somewhere inside both father and son.

Later, when the son is older and utters some such sentence as That balloon is loose, a coupling of another sort occurs, as in Figure 13, also a triadic event.

Let us say nothing about the physiological or ontological status of the “coupler.” Suffice it for the present to say that if two elements of a sentence are coupled, we may speak of a coupler. Indeed, the behavioral equivalent of Descartes’s cogito ergo sum may be: If the two elements of a sentence are coupled, there must be a coupler. The latter dictum would seem to be more useful to the behavioral scientist, including transformational linguists, than Descartes’s, because Descartes’s thinking is not observable but his speech is.

Accordingly, I had supposed that what the neurophysiologist and anatomist should look for in the brain is not a neuron circuitry transmitting S-R arcs with little s’s and r’s interposed (what Peirce would call a series of dyads) of the following order:

but rather a structural-functional entity with the following minimal specifications: (1) It must be, considering the unique and highly developed language trait in man, something which is present and recently evolved in the human brain and either absent or rudimentary in the brains of even the highest nonhuman primates. (2) It should be structurally and functionally triadic in character, with the “base” of the triad comprising what must surely be massive interconnections between the auditory and visual cortexes. What else indeed is the child up to for months at a time when it goes around naming everything in sight — or asking its name — than establishing these functional intercortical connections?

It was with no little interest, therefore, that I came across the work of Norman Geschwind, who believes he has identified just such a recently evolved structure, “the human inferior parietal lobule, which includes the angular and supramarginal gyri, to a rough approximation areas 39 and 40 of Brodmann. In keeping with the views of many anatomists, Crosby et al. comment that these areas have not been recognized in the macaque. Critchley, in his review of the anatomy of this region, says that even in the higher apes these areas are present in only rudimentary form” (Geschwind). And further:

In man, with the introduction of the angular gyrus region, intermodal associations become powerful. In a sense the parietal association area frees man to some extent from the limbic system…

The development of language is probably heavily dependent on the emergence of the parietal association areas since at least in what is perhaps its simplest aspect (object naming) language depends on associations between other modalities and audition. Early language experience, at least, most likely depends heavily on the forming of somesthetic-auditory and visual-auditory associations, as well as auditory-auditory associations. (Geschwind)

Such findings are adduced here as a matter of interest only and to show that at least the model here adumbrated gives a hint what to look for. Being fully aware of the strong feelings of many psychologists and psycholinguists against such localization of cognitive functions (e.g., Lenneberg and Premack), and being myself altogether incompetent to evaluate Geschwind’s findings, I suggest only that if a triadic theory of language acquisition is correct, one might expect to find some such structure. If Geschwind is right, what he has uncovered is the cortical “base” of the triadic structure of the typical semological-phonological naming sentence (Figure 14).

The apex of the triangle, the coupler, is a complete mystery. What it is, an “I,” a “self,” or some neurophysiological correlate thereof, I could not begin to say.