Cartesian Linguistics

by Noam Chomsky

  you would not understand anything in it except by putting together in your mind the words in their relation to one another, and you can do this only after you have heard the whole sentence.

  (pp. 198–199)

  In Latin, for example, it is the “relative word-endings which makes us consider the words in the completed proposition in accordance with the order of their interrelations, and hence in accordance with the order of the simple, necessary and meaningful construction” (pp. 241–242). This “simple construction” is an “order which is always indicated, but rarely observed in the usual construction of languages whose nouns have cases” (p. 251). Reduction to the “simple construction” is an essential first step in speech perception:

  The words form a whole that has parts. The simple perception of the relations between these parts makes us conceive the whole of them, and comes to us solely from the simple construction. Setting forth the words in accordance with the order of succession of their relations, this presents them in a manner that is best fitted to make us recognize these relations and to make the whole thought arise.

  (pp. 287–288)

  Constructions other than the “simple constructions” (namely, “figurative constructions” [constructions figurées])

  are understood only because the mind corrects their irregularities, with the help of accessory ideas which make us conceive what we read and hear as if the sense were expressed in the order of the simple construction . . . (p. 292)

  In short, in the “simple construction” the relations of “syntax” are represented directly in the associations among successive words, and the undifferentiated thought expressed by the sentence is derived directly from this underlying representation, which is regarded, throughout, as common to all languages (and, typically, as corresponding to the usual order of French – cf., e.g., p. 193).

  The transformations which form a “figurative construction” effect reordering and ellipsis. The “fundamental principle of all syntax” (p. 218) is that reordering and ellipsis must be recoverable by the mind of the hearer (cf. pp. 202, 210ff., 277); that is, they can be applied only when it is possible to recover uniquely “the strict metaphysical order” of the “simple construction.”89

  Many examples of reduction to simple constructions are presented to illustrate this theory.90 Thus the sentence “Who said it?” is reduced to the simple construction “The one who said it is which person?” (Sahlin, p. 93); the sentence “Being loved as much as lovers, you are not forced to shed tears” is reduced to “Since you are loved as much as you are lovers, . . .”; the sentence “It is better to be just than to be rich, to be reasonable than to be wise” is reduced to four underlying propositions, two negative, two positive, in the obvious way (p. 109), etc.

  A rather different sort of example of the distinction between deep and surface structure is provided by Du Marsais in his analysis (pp. 179–180) of such expressions as “I have an idea/fear/doubt,” etc. These, he says, should not be interpreted as analogous to the superficially similar expressions “I have a book/diamond/watch,” in which the nouns are “names of real objects that exist independently of our thought [manière de penser].” In contrast, the verb in “I have an idea” is “a borrowed [empruntée] expression,” produced only “by imitation.” The meaning of “I have an idea” is simply “I am thinking” or “I am conceiving something in such-and-such a way.” Thus the grammar gives no license for supposing that such words as “idea,” “concept,” “image” stand for “real objects,” let alone “perceptible objects.” From this grammatical observation it is only a short step to a criticism of the theory of ideas, in its Cartesian and empiricist forms, as based on a false grammatical analogy. This step is taken by Thomas Reid, shortly after.91

  As Du Marsais indicates with abundant references, his theory of construction and syntax is foreshadowed in scholastic and renaissance grammar (see note 67). But he follows the Port-Royal grammarians in regarding the theory of deep and surface structure as, in essence, a psychological theory, not merely a means for the elucidation of given forms or for analysis of texts. As indicated above, it plays a role in his hypothetical account of the perception and production of speech, just as, in the Port-Royal Grammar, the deep structure is said to be represented “in the mind” as the utterance is heard or produced.

  As a final example of the attempt to discover the hidden regularities underlying surface variety, we may mention the analysis of French indefinite articles in Chapter VII of the Port-Royal Grammar, where it is argued, on grounds of symmetry of patterning, that de and des play the role of the plural of un, as in Un crime si horrible mérite la mort, Des crimes si horribles méritent la mort, De si horribles crimes méritent la mort, etc. To handle the apparent exception, Il est coupable de crimes horribles (d’horribles crimes), they propose the “rule of cacophony” that a de de sequence is replaced by de. They also note the use of des as a realization of the definite article, and other uses of these forms.

  Perhaps these comments and examples are sufficient to suggest something of the range and character of the grammatical theories of the “philosophical grammarians.” As noted above, their theory of deep and surface structure relates directly to the problem of creativity of language use, discussed in the first part of the present work.

  From the standpoint of modern linguistic theory, this attempt to discover and characterize deep structure and to study the transformational rules that relate it to surface form is something of an absurdity;92 it indicates lack of respect for the “real language” (i.e., the surface form) and lack of concern for “linguistic fact.” Such criticism is based on a restriction of the domain of “linguistic fact” to physically identifiable subparts of actual utterances and their formally marked relations.93 Restricted in this way, linguistics studies the use of language for the expression of thought only incidentally, to the quite limited extent to which deep and surface structure coincide; in particular, it studies “sound-meaning correspondences” only in so far as they are representable in terms of surface structure. From this limitation follows the general disparagement of Cartesian and earlier linguistics,94 which attempted to give a full account of deep structure even where it is not correlated in strict point-by-point fashion to observable features of speech. These traditional attempts to deal with the organization of semantic content as well as the organization of sound were defective in many ways, but modern critique generally rejects them more for their scope than for their failures.

  Description and explanation in linguistics

  Within the framework of Cartesian linguistics, a descriptive grammar is concerned with both sound and meaning; in our terminology, it assigns to each sentence an abstract deep structure determining its semantic content and a surface structure determining its phonetic form. A complete grammar, then, would consist of a finite system of rules generating this infinite set of paired structures and thus showing how the speaker-hearer can make infinite use of finite means in expressing his “mental acts” and “mental states.”

  However, Cartesian linguistics was not concerned simply with descriptive grammar, in this sense, but rather with “grammaire générale,” that is, with the universal principles of language structure. At the very outset of the work under review, a distinction was made between general and particular grammar. These are characterized by Du Marsais in the following way:

  Some points [observations] of grammar apply to all languages. These form what we call general grammar – for example, those we made regarding articulated sounds and the letters which are the signs of these sounds, the nature of words and the various ways they must be ordered or terminated in order to have meaning. Apart from these general points, there are some which are peculiar to one particular language, and these form the special grammar of that language.95

  Beauzée elaborates the distinction in the following way:

  GRAMMAR, whose object is the expression of thought with the help of spoken or written words, comprises two sorts of p
rinciples. One sort, being immutably true and universally applicable, derive from the nature of thought itself, following its analysis and being its result. The other sort are only hypothetically true and depend on conventions which, being accidental, arbitrary and changeable, have given rise to different idioms. The first sort of principles constitute general grammar and the second are the object of various particular grammars.

  GENERAL GRAMMAR is therefore the rational science of the immutable and general principles of spoken or written Language [Langage], whatever language [langue] this may be.

  A PARTICULAR GRAMMAR is the art of applying the arbitrary and usual conventions of a particular language to the immutable and general conventions of written or spoken Language.

  General Grammar is a science, because its object is rational speculation on the immutable and general principles of Language.

  A particular Grammar is an art, because it considers the practical application of the arbitrary and usual conventions of a particular language to the general principles of Language.

  The science of grammar is anterior to all languages in so far as its principles presuppose only the possibility of languages and are the same as those which guide human reason in its intellectual operations; in short, because they are eternally true.

  The art of grammar, by contrast, is posterior to languages in so far as linguistic usages must exist before they can stand in an artificial relation to the general principles of Language, and the analogical systems that form this art can be determined only by observations made on these pre-existent usages.96

  In his Eloge de du Marsais, D’Alembert gives this account of “philosophical grammar”:

  Grammar is therefore the work of philosophers. Only a philosophical mind can ascend to the principles on which its rules are based . . . This mind first recognizes, in the grammar of each language, the general principles which are common to all of them, and which form General Grammar. It then distinguishes, among the usages peculiar to each language, those which can be founded on reason from those which are the work of chance or negligence: it observes the reciprocal influences that languages have had on each other and the alterations that this mingling has brought about without entirely destroying their individual character; it weighs their mutual advantages and disadvantages; differences in their construction . . .; the diversity of their genius . . .; their richness and freedom, poverty and servitude. The development of these various factors is the true metaphysics of grammar. Its object . . . is to advance the human mind in the generation of its ideas and in the use it makes of words to transmit thoughts to other men.97

  The discovery of universal principles would provide a partial explanation for the facts of particular languages, in so far as these could be shown to be simply specific instances of the general features of language structure formulated in the “grammaire générale.” Beyond this, the universal features themselves might be explained on the basis of general assumptions about human mental processes or the contingencies of language use (for example, the utility of elliptical transformations). Proceeding in this way, Cartesian linguistics attempts to develop a theory of grammar that is not only “general” but also “explanatory” [raisonnée].

  The linguistics of Port-Royal and its successors developed in part in reaction against the prevailing approaches represented, for example, in such work as Vaugelas’s Remarques sur la langue françoise (1647).98 Vaugelas’s goal is simply to describe usage “which everyone recognizes as the master and ruler of living languages” (Preface). His book is called Remarques . . . rather than Décisions . . . or Loix . . . because he is “a simple observer [tesmoin].” He disclaims any intention of explaining the facts of speech or finding general principles that underlie them, just as he generally suggests no modification or “purification” of usage on rational or esthetic grounds. His grammar, then, is neither “explanatory” nor prescriptive.99 He is quite aware of the problems of determining actual usage and provides an interesting discussion of “elicitation procedures” (pp. 503f.), in which, among other things, he points out the inadequacy of the kinds of “direct question” tests for grammaticalness that have occasionally been proposed and applied by structural linguists, with predictably inconclusive results. He does not restrict his descriptive comments to surface structure.100 For example, he points out that one cannot determine from the form of a word whether it has an “active meaning” [signification] or a “passive meaning” or, ambiguously, both (pp. 562–563). Thus in the sentence “My esteem isn’t something from which you can derive any great advantage,” the phrase “my esteem” has the sense “the esteem which I hold for you,” whereas in the sentence “My esteem does not depend on you,” it means “the esteem in which I am held” or “the esteem in which I may be held;” and the same is true of such words as “aid,” “help,” and “opinion.” There are other examples of a concern for descriptive adequacy on a broad scale. At the same time, Vaugelas’s work foreshadows many of the defects of modern linguistic theory, for example, in his failure to recognize the creative aspect of language use. Thus he regards normal language use as constructed of phrases and sentences that are “authorized by usage,” although new words (e.g., brusqueté, pleurement) can be correctly formed by analogy (pp. 568f). His view of language structure, in this respect, seems not very different from that of Saussure, Jespersen, Bloomfield, and many others who regard innovation as possible only “by analogy,” by substitution of lexical items for items of the same category within fixed frames (cf. p. 65 above).

  The reaction of “philosophical grammar” is not against the descriptivism of Vaugelas and others as such101 but against the restriction to pure descriptivism. The Port-Royal Grammar takes it as a general maxim for anyone working on a living language that “the ways of speaking that are authorised by undisputed general usage must be accepted as good even if they go against the rules and analogy of the language” (p. 83; PRG 113). Lamy, in his rhetoric, echoes Vaugelas in describing usage as “the master and arbitrary ruler of languages” and in holding that “no one may contest this rule which necessity has established and the general agreement of people has confirmed” (op. cit., p. 31). Du Marsais insists that “the philosophical grammarian must consider the particular language he is studying in relation to what this language is in itself and not in relation to another language.”102 Philosophical grammar, then, was not characteristically attempting to refine or improve language, but to discover its underlying principles and to explain the particular phenomena that are observed.103

  The example which, for more than a century, was used to illustrate this difference between descriptive and explanatory grammar was provided by a rule of Vaugelas (pp. 385f.) regarding relative clauses, namely, the rule that a relative clause may not be added to a noun that has no articles or only the “article indefini” de. Thus one cannot say “II a fait cela par avarice, qui est capable de tout” or “II a fait cela par avarice, dont la soif ne se peut esteindre.” Similarly, one cannot say “II a esté blessé d’un coup de fleche, qui estoit empoisonnée” (p. 385), although it is correct to say “II a esté blessé de la fleche, qui estoit empoisonnée” or “II a esté blessé d’une fleche qui estoit empoisonnée.”

  In Chapter IX, the Port-Royal Grammar first notes a variety of exceptions to this rule and then proposes a general explanatory principle to account both for the examples that fall under the rule of Vaugelas and for the exceptions to his rule.104 The explanation is, once again, based on the distinction between meaning and reference. In the case of a “common noun,” the meaning [signification] is fixed (except for ambiguity or metaphor), but the reference [estendue] varies, depending on the noun phrase in which the noun appears. A particular occurrence of a noun is called indeterminate “when there is nothing that indicates whether it must be taken generally or particularly and, if the latter, whether for a determinate or indeterminate particular” (p. 77; PRG 109); otherwise, it is determinate. Vaugelas’s rule is now restated in terms of determinatio
n: “in current French usage one may not put qui after a common noun unless it is determined by an article or some other thing that determines it no less than it would be determined by an article” (p. 77; PRG 109). A detailed analysis follows, attempting to show that the apparent counter-examples involve occurrences of nouns that are “determined” by some feature other than the article. In part, the analysis is based on assumptions about deep structure that are not without interest in themselves. The rule is also discussed by Du Marsais, Beauzée, and others at some length. We need not go into the details here. The point, in the present context, is that this was taken as a paradigm example of the necessity for supplementing descriptive statements with a rational explanation, if linguistics was to go beyond compilation of facts to true “science” – in the terminology of the day, if grammar was to become “philosophical.”

  In connection with the rule of Vaugelas and several other cases, the explanations that are proposed, in universal grammar, have some substance and linguistic content. All too often, however, they are quite empty, and invoke assumptions about underlying mental reality in a quite mechanical and unrevealing way. In fact, it seems to me that in general the modern critique of “philosophical grammar” is quite misplaced. The error of this position is generally taken to be its excessive rationality and a priorism and its disregard for linguistic fact. But a more cogent criticism is that the tradition of philosophical grammar is too limited to mere description of fact – that it is insufficiently “raisonnée”; that is, it seems to me that the faults (or limitations) of this work are just the opposite of those which have been attributed to it by modern critics. The philosophical grammarians considered a wide realm of particular examples; they tried to show, for each example, what was the deep structure that underlies its surface form and expresses the relations among elements that determine its meaning. To this extent, their work is purely descriptive (just as modern linguistics is purely descriptive in pursuit of its more restricted goal of identifying the units that constitute the surface structure of particular utterances, their arrangement into phrases, and their formally marked relations). Reading this work, one is constantly struck by the ad hoc character of the analysis, even where it seems factually correct. A deep structure is proposed that does convey the semantic content, but the basis for its selection (beyond mere factual correctness) is generally unformulated. What is missing is a theory of linguistic structure that is articulated with sufficient precision and is sufficiently rich to bear the burden of justification. Although the examples of deep structure that are given in abundance often seem quite plausible, they are unsatisfying, just as modern linguistic descriptions, though often quite plausible in their analysis of particular utterances into phonemes, morphemes, words, and phrases, remain unsatisfying, and for the same reason. In neither case do we have an underlying hypothesis as to the general nature of language that is sufficiently strong as to indicate why just these and not other descriptions are selected by the child acquiring the language or the linguist describing it, on the basis of the data available to them.105