Chomsky论形式语法理论

In the traditional study of language, an important distinction was made between"particular grammar" and" universal grammar", the first being concerned with the idiosyncratic properties of specific languages, the second, with the general features of any human language. A concern for this distinction has reappeared in the foreground of research in the past few years, and this resurgence of interest has coincided with the initiation of serious research in a branch of mathematical linguistics that is sometimes called" algebraic linguistics", to distinguish it from statistical investigations of language or the study of probabilistic models for performance. The coincidence of these developments is quite natural, in several respects. Algebraic linguistics is the study of formal properties of natural language, abstracted from the specific realization in particular languages. So defined, it is really quite indistinguishable from universal grammar, although in practice there has been a division of labor between the more empirically oriented studies of universal grammar and the strictly mathematical studies of formal structures suggested by linguistic research. The existence of this division is in part a sign of the immaturity of the field, in part a reflection of the differing motivations and research interests of particular investigators. If the division can be substantially narrowed, a true theory of mathematical linguistics may emerge which studies in a purely abstract way the class of systems defined by the principles of universal grammar - the class of "possible human languages".
A second reason for the coincidence of the rise of algebraic linguistics and the revived interest in universal grammar can be found in the development of the study of language in the modern period. Over a century ago, the study of “grammaire générale et raisonnée“ entered into a decline and its value and significance were seriously questioned as scholars turned to historical and comparative study, and, later, to study of language within the very fruitful "structuralist" or "descriptivist" traditions. The older studies of “grammaire générale et raisonnée” were held to be too speculative, not firmly enough grounded in linguistic fact. It was felt that they seriously underestimated the possible variety of language structures and that they imposed, arbitrarily, a model of language structure that distorted the actual phenomena of particular languages. Some scholars went so far as to maintain that languages can differ without limit and that there are no general conditions of any significance that
restrict and constrain the form of any possible human language. At the same time, attention shifted to sound structure, which had by no means
been neglected in the classical linguistic tradition but which had been regarded as ancillary to the investigation of the deeper syntactic properties of language. A major concern of the classical tradition was a property of language that might be called its "creative aspect", that is, the fact that a language provides recursive mechanisms that permit the expression of infinitely many thoughts, feelings, intentions, and so on, independently of any control by external stimuli or identifiable internal states. The "creative aspect" of language was in effect denied by many modern students of a behaviorist orientation, who came to view a language as a system of habits, a network of stimulus-response connections, or something of the sort; by others, for example, de Saussure, it was not denied, but was assigned to “parole”, and regarded as either peripheral, or else definitely beyond the scope of the study of language, and not subject to strict linguistic rule.
Certainly at least one reason, perhaps the major reason, for this shift of interest and modification of doctrine lay in the fact that available technique and understanding of formal processes did not make it possible to study the recursive mechanisms of syntax in any clear and productive way. In fact, it was not until a generation ago that the proper concepts were developed and sharpened, in the course of investigation of foundations of mathematics. With the new understanding of recursive mechanisms and the nature of algorithms that has developed in the past thirty years, it becomes possible to return to the study of the creative aspect of language use, and to attempt to formulate, in a precise way, the mechanisms that each language makes available for the free and unbounded use of language. The study of these mechanisms is now generally referred to as "generative grammar"; the generative grammar of a language is understood as the system of rules and processes that characterize the potentially infinite class of sentences of a natural language and that assign to each of these objects a structural description representing its significant phonetic, syntactic, and semantic properties. Thus the study of generative grammar became feasible as a result of developments in mathematics, and it is therefore not surprising that interest in formal properties of grammars, in algebraic linguistics, was a natural outgrowth of this new approach to the study of language.
There is much to be said about these matters, but I think it is fairly clear that the skepticism of structural and descriptive linguistics with respect to the possibilities for a “grammaire générale et raisonnée” was unwarranted, and that, in fact, the classical tradition was probably much too conservative in the extent to which it postulated restrictive conditions and constraints that govern the form of any human language. Speculating, I think it is not unreasonable to assume that the linguistics of the next generation will reveal that each natural language is a specific realization of a highly restrictive schema that permits grammatical processes and structures of a very limited variety, and that there are innumerable "imaginable" languages that violate these restrictions and that are, therefore, not possible human languages in a psychologically important sense, even though they are quite able, in principle, to express the entire content of any possible human language. If this is true, then a mathematical study of universal grammar, in the sense described earlier, may well become the central domain of linguistic theory. It is too early to know whether these hopes can be realized, but they seem not unreasonable, given what we know and are coming to know today.
I would like to stress again that there is still a significant gap between the mathematical and the empirical investigations that fall within the domain of what ultimately may become a mathematical theory of universal grammar. The schema for grammatical description that seems empirically motivated by the facts of particular languages specifies a class of systems that are for the moment, much too complex for fruitful and far-reaching mathematical investigation; furthermore, it must be borne in mind that any proposals that can be made today concerning this universal schema are both highly tentative and also somewhat loose in important respects. At the same time, there have been interesting and suggestive studies of much more restricted schemata for grammatical description - the theory of so-called "context-free languages" is the primary example - but these systems are surely empirically inadequate. A mathematical theory of universal grammar is therefore a hope for the future rather than a present reality. The most that one can say is that present research appears to be tending in the direction of such a theory. To me it seems that this is one of the most exciting areas of study today, and that if it proves successful, it may place the study of language on an entirely new footing in coming years.
引自——《Notions sur les Gramaires formelles》by M. Gross, A. Lentin, M. Salkoff
英译本《Introduction to Formal Grammars》translated by M. Salkoff 1970
此文是N. Chomsky为此书英译本写的序言
【注】grammaire générale et raisonnée：中译是：普遍唯理语法，波尔·罗瓦雅尔语法学派的代表作。