Generative semantics | Статья в журнале «Молодой ученый»


Рубрика: Филология

Опубликовано в Молодой учёный №7 (18) июль 2010 г.

Статья просмотрена: 308 раз

Библиографическое описание:

Назари К. Б. Generative semantics // Молодой ученый. — 2010. — №7. — С. 163-167. — URL (дата обращения: 26.05.2018).

                        Theory of transformational grammar developed from the late 1960s to the mid – 1970s. The original proposal was that a base component of a grammar should directly generate semantic representations of sentences, which would be converted to surface structures with no intervening level of deep structure. This was associated in particular with the view that lexical items were units only at the surface level. But in the 1970s it became clear that the proposed semantic representations could not be assigned by rules of grammar independent of the knowledge, beliefs, etc. of individual speakers. Therefore the break with what was later called the Standard Theory of transformational grammar was in reality far more radical.

Key Words: transformational grammar, surface structure, deep structure, standard theory, interpretive semantics

1. Generative Semantics

The name for the counter position taken by G. Lakoff, J. McCawley, and J. Ross among others in the late 1960s in response to Chomsky's conception of semantics in his 1965 ‘standard theory’ of transformational grammar. Chomsky, Katz and Fodor (1963) argued that the syntactically motivated deep structure presents the only structure applicable to the semantic interpretive components of the grammar. In contrast, the proponents of generative semantics maintained that semantic structures are generated in a form of basic (universal) rules similar to those of predicate logic. The meaning of individual lexemes is described as a syntactically structured complex of basic semantic elements. For example, the verb convince (X convinces Y to do Z) is paraphrased by X does that Y wants that Z, where do and want are atomic predicates which form more complex predicates through transformations. In addition, the number of syntactic categories is reduced to three: S (= proposition), NP (= argument), and V (= predicate). Since the logical-semantic form of the sentence is now seen as the underlying (generative) structure, the otherwise strict division between syntax and semantics collapses, especially between lexical semantics, word formation and the semantics of propositions. Critics of generative semantics pointed out the adhoc nature of the descriptive mechanism and the ‘overpowerful’ generative power of this model, whose apparatus could generate more complex structures than are realized in human languages.

It is widely held view nowadays that the linguist's model of a language system should not only generate all and only the well-formed system sentences of the language but should assign to each system – sentence both a phonological representation (PR) and a semantic representation (SR). The PR is to be thought of as a representation of the way in which the system- sentence would be pronounced and the SR as a representation of its meaning. Looked at from this point of view, the model can be seen as an integrated system of grammatical, phonological and semantic rules relating sound and meaning; and this is how generative grammars are now commonly described.

When Chomsky first put forward his theory of generative grammar, he had little to say about the possibility of integrating phonology, morphology, syntax and semantics within a unified model of a language – system. The illustrative partial description of English that he used in his earliest work did not contain any rules for the semantic interpretation of sentences; and he took the view that the grammatical rules could be established and formalized without making any appeal to sameness and difference of meaning or to any other semantic notions. In this respect, grammar was held to be autonomous and independent of semantics. It was always recognized, however, that there were certain systematic connections between syntax and semantics and that in so far as the choice between two grammatical analyses was otherwise indeterminate, semantic criteria should be used to resolve the indeterminacy. To this extent at least, Chomskyan generative grammar has always taken account of the systematic connections that were held to exist between syntax and semantics. In particular, it has always been concerned with the fact that certain kinds of ambiguity could be regarded as grammatically explicable ; and Harris if not Chomsky, has from the outset emphasized the fact that some part of the meaning of a sentence remains constant under transformation. This point must be stressed in view of the very considerable confusion that now surrounds the thesis that grammar, and, more especially syntax is autonomous. Chomsky, like Harris and other post – Bloomfieldian linguists, has continually professed his methodological commitment to the principles of autonomous syntax. But he has been paying more attention recently, as have other generative grammarians, to the integration of syntax and semantics.

The first explicit proposals for the integration of syntax and semantics within a Chomskyan framework were made by Katz and Fodor (1963). Their proposals were subsequently clarified and extended by Katz and Postal (1964) and taken over by Chomsky (1965) in the construction of what has now come to be called the standard version of Chomskyan transformational –generative grammar. What Katz and Fodor did, in effect, was to add to the grammar a dictionary, providing semantic and syntactic information for each of the lexemes that it contained, and a set of projection – rules, whose function it was to associate with every semantically well – formed sentence at least one semantic representation. The general orientation of the Katz and Fodor approach to the integration of syntax and semantics is evident from their famous slogan: "Linguistic description minus grammar equals semantics" (in which ‘grammar’ is to be understood to cover, not only syntax and morphology, but also phonology). As far as well – formedness was concerned, semantics was residual: "Semantics takes over the explanation of the speaker's ability to produce and understand new sentences at the point where grammar leaves off". Given that a particular string of forms was defined by the grammar to be syntactically ill – formed, the question whether it was semantically well –formed or ill – formed simply did not arise. It was only with respect to grammatical sentences that the projection – rules had any role to play. This view of semantics as purely residual has had the effect that "research has been biased heavily in favor of syntactic solutions to problems" (Jackendoff, 1972: 2).

A further point to be noted about the Katz and Fodor proposals is that they imposed a spurious parallelism upon phonology and semantics. The syntactic part of the integrated model of the language – system, consisting of base – rules and transformational rule, was held to be central, not only in that it came between the phonological and the semantic parts of the model, but also in that it contained all the generative capacity of the whole integrated model. In contrast with the rules of syntax, and more particularly with the base – rules, both the semantic rules (i.e. the projection – rules) and the phonological rules were held to be nongenerative and interpretive. Their function was to take as input syntactically structured strings of forms generated by the syntactic part of the model and to interpret these in terms of allegedly universal elements of meaning and of sound. What is to be noticed here (apart from the alleged universality of the elements of sound and meaning: is the curious use of the term ‘interpret’ , according to which both the pronunciation and the meaning of a sentence constitutes an interpretation of it. Apart from the terminology that is employed, the Katz and Fodor model is strikingly similar, at this point, to the so – called glossematic version of structuralism developed by Hjelmslev and his collaborators some years earlier and it is open to the same objections.

There is an inherent connexion between grammar and semantics which does hold between grammar and phonology; and this fact should be captured in anything that purports to be a model of a language system. It is, to say the least, obscured in the Katz and Fodor model, as it is in any model that treats the phonological and the semantic representations associated with sentences as being comparable theoretical constructs. Henceforth, we will avoid using the terms ‘semantic representation’. On the one hand, and ‘phonological interpretation’ (or ‘phonetic interpretation’), on the other. We will take instead of the phonological representation of a sentence (on the assumption that it is realized in the phonic medium) and of its semantic interpretation (or interpretations). Incorporating these terminological modifications into the Katz and Fodor model, we can formulate the relationship between the several parts of their integrated model of the language –system by means of the diagram in figure 1.

It will be noticed that the base has been distinguished from the other three sets of rules. This is intended to take account of one of the principal changes that Chomsky (1965) made in his formulation of the so – called standard version of transformational grammar: the inclusion of the lexicon as a sub-component of the base. We are not concerned with the reasons for this change or its implications. Another change that has been incorporated in figure 1, in order to bring it into line with the standard version of Chomskyan transformational grammar, is the introduction of the notions of deep structure and surface structure.

Овал: DSБлок-схема: решение: PR


Овал: SS


Блок-схема: решение: SI



Figure. 1 The so-called standard theory

Apart from the base, there are three boxes of rules: transformational rules (T-rules), semantic rules (S-rules) and phonological rules (p-rules). The output of the base is a set of deep structures (DS), to which the S-rules (Katz & Fodor's projection rules) apply, and yield a set of semantic interpretation (SI). The output of the T-rules, on the other hand, is a set of surface structures, to which the phonological rules apply and derive for each sentence its phonological representation (PR).

The general conclusion towards which Katz and Fodor (1963), and more especially Katz and Postal (1964), were working was the thesis that all of the T-rules were obligatory; and this thesis was taken over and made part of the standard version of transformational grammar by Chomsky (1965). It carries as an immediate corollary, by virtue of the semiotic principle that meaningfulness implies choice, the proposition that transformations do not affect meaning. It is only in so far as this proposition is held to be true that one can maintain the principle that all the information that is relevant to the semantic interpretation of a sentence is present in its deep structure. Acceptance of this principle is made explicit in figure 1, it will be observed, by the absence of any path from SS to SI.

It is tempting, having accepted that deep-structure identity is a sufficient condition of semantic identity, to take the further step of making it a necessary condition also. In effect, this is what is done by those calling themselves generative semanticists, who argue that there is no need to postulate any distinction between the deep structure of a sentence and its semantic interpretation. Their approach to the construction of an integrated model of linguistic description is shown in figure 2.





The term ‘interpretive semantics’ and ‘generative semantics’, which have been widely employed to refer to the alternative conceptions of the relationship between semantics and syntax that are diagrammed in figure 1 and figure 2 respectively, are quite inappropriate for this purpose. Any model of a language - system that generates a set of semantically well-formed sentences must rest upon a theory of semantics that is properly described as generative. The difference between the alternative conceptions of the relationship between semantics and syntax is not therefore that one rests upon a theory of generative semantics and the other does not. They both presuppose the existence or possibility of a theory of generative semantics. Indeed, in so far as these two conceptions of the relationship between semantics and syntax have been put forward within the general framework of Chomsky's theory of transformational grammar, they have both taken for granted a very particular kind of generative semantics: they have both accepted that a model of linguistic description should not only generate the set of semantically well-formed sentences, but should also associate with each semantic interpretation in terms of a universal inventory of sense-components. Componential analysis is theoretically suspect of a number of counts. What we are concerned with here is whether any kind of generative semantics is viable, independently of its association with componential analysis, on the one hand, or with various kinds of transformational grammar' on the other.

The difference between the alternatives shown in figures 1 and 2 is that the former draws a distinction, which the latter does not, between the deepest syntactic analysis of a sentence and its semantic interpretation. Another way of expressing this difference is to say that, whereas the former is syntactically based, the latter is semantically based. A syntactically based model operates according to the principle of the antonomy of syntax; a semantically based model does not. Various kinds of syntactically based and semantically based models are conceivable. But most of the discussion of the difference between them that has taken place in recent years has been centered upon the role that is assigned to deep structure in Chomsky's standard theory of transformational grammar.

According to Chomsky 91965) the deep structure of a sentence is a phrase-marker which contains all the lexemes whose forms appear in the surface structure of the same sentence; and it is in terms of the topology of the deep structure phrase-marker that the semantically relevant notions of subject, object and predicate are defined and selection restrictions are accounted for. The so-called generative semanticists take the view that lexicalization is a particular kind of transformational process. For example, the lexeme ‘kill’ might be taken from the lexicon and substituted for an underlying structure containing the sense-components CAUSE, BECOME, NOT and ALIVE; and the operation whereby this substitution is carried out would become, among many, of the transformations involved in the generation of any sentence containing the lexeme ‘kill’. Furthermore, lexicalizing transformations of this kind do not operate in a block, it is argued, prior to the operation of other transformations: they must be interspersed with what would be conventionally regarded as purely syntactic transformations; and it is principally for this reason that the Chomskyan notion of deep structure is rejected. The so – called generative semanticists also reject the treatment of selection – restrictions propose by Chomsky (1965); and like many other linguists, they deny that what the standard theory of transformational grammar defines to be deep-structure subjects, objects, and predicates play any role in the semantic interpretation of sentences.

One of the most striking features of the presentation by Katz and Postal (1964) of the thesis that transformations do not change meaning was the looseness with which the term ‘meaning’ was employed. No account was taken of the fact that the semantic relationship between a declarative sentence and an interrogative sentence, or between a declarative sentence and an imperative sentence, was a different kind of semantic relationship than that which holds, or may hold, between two declarative sentences. Furthermore, no distinction was drawn between the meaning of a sentence and the meaning of an utterance; and what was held to be purely stylistic variation was thereby classified as semantically irrelevant.

There are two classes of sentences which philosophers and linguists have generally treated as anomalous and which are already of concern to the semanticist: sentences which (when they are used to make statements) express tautologies and sentences which express contradictions. Both tautologies and contradictions are, in principle, uninformative: a tautology tells the addressee nothing that he does not know, or could not deduce, by virtue of his knowledge of the language; and a contradiction fails to tell him anything that he can accommodate in his state-description of the world. But to say that tautologies and contradictions are uninformative is not to say that they are meaningless or semantically acceptable. If they were meaningless, they could not have a truth-value; and their status as tautologies or contradictions rests upon their being necessarily true or necessarily false, respectively. There can be no question, therefore, of excluding sentences that express tautologies or contradictory propositions from the set of semantically well-formed sentences.

One of the reasons that is often given for including in a model of a language-system rules that are sensitive to so-called selection-restrictions is that such rules make it possible to account for the native speaker's ability to infer propositions that are presupposed or implied, rather than asserted. For example, from That person is Pregnant, one would normally infer that the person referred to is female; and, on the vast majority of occasions on which this utterance might be produced, the inference would no doubt be correct. But the proposition expressed by the sentence ‘That person is female’ is certainly not entailed by the proposition expressed by ‘that person is pregnant’: the inference is in principle no more than probabilistic, since it is possible to envisage a world in which men could be pregnant. (For the same reason, there is no violation of the rules of the language-system involved in the sentence ‘That man is pregnant’).

Another reason that is given for having rules which are sensitive to the collocational restrictions holding between particular lexemes is that they explicate the alleged fact that a phrase or sentence may be unambiguous even though it contains one or more homonyms or polysemous lexemes in positions that the purely grammatical rules specify as permissible for them. For example, it might be argued that, whereas both ‘ball1’ ("spherical or avoid object use in certain games") and ‘ball2’ ("elegant kind of party featuring social dancing") are permissible in noun-phrases governed by a transitive verb, the sentence ‘The man hit the ball’ cannot be interpreted as containing ‘ball2’, by virtue of the requirement associated with the verb ‘hit’ that its object should refer to a physical object. Granted that The man hit the ball (i.e. some token of the utterance-type that is isomorphic with the sentence ‘The man hit the ball’) will, in all probability, be interpreted by native speakers of English as containing ‘ball1’, rather than ‘ball2’, if the utterance is put to them out of context, it does not follow that the sentence ‘The man hit the ball’ must be treated as non-ambiguous. What has to be demonstrated is that there are no circumstances under which The man hit the ball could be construed as containing ‘ball2’. Even if this could be demonstrated, or safely assumed, it still does not follow that rules must be formulated within the linguist's model of the language – system to exclude the possibility of taking ball as a form of ‘ball2’ in this instance. Other cognitive abilities besides knowledge of the language – system are involved in the recognition and interpretation of utterances.

2. Interpretive vs. Generative Semantics

The controversy surrounding generative semantics stemmed in part from the competition between two fundamentally different approaches to semantics within transformational generative syntax. The first semantic theories designed to be compatible with transformational syntax were interpretive. Syntactic rules enumerated a set of will-formed sentences paired with syntactic structures, each of which was assigned an interpretation by the rules of a separate semantic theory. This left syntax relatively (though by no means entirely) "autonomous" with respect to semantics, and was the approach preferred by Chomsky. In contrast, generative semanticists argued that interpretations were generated directly by the grammar as deep structures, and were subsequently transformed into recognizable sentences by transformations. This approach necessitated more complex deep structures than those proposed by Chomsky, and more complex transformations as a consequence. Despite this additional complexity, the approach was appealing in several respects. First, it offered a powerful mechanism for explaining synonymity. In his initial work in generative syntax, Chomsky motivated transformations using active/passive pairs such as "I hit John" and "john was hit by me", which despite their identical meanings have quite different surface forms. Generative semanticists wanted to account for all cases of synonimity in a similar fashion – an impressively ambitious goal before the advent of more sophisticated interpretive theories in the 1970s. Second, the theory had a pleasingly intuitive structure: the form of a sentence was quite literally derived from its meaning via transformations. To some, interpretive semantics seemed rather "clunky" and adhoc in comparison. This was especially so before the development of trace theory.

3. Conclusion

It is difficult to avoid the conclusion that no convincing case has yet been made for the thesis that a linguist's model of a language-system should be such that it generates all and only the semantically well-formed sentences of the language (as a proper subset of the grammatically well-formed sentences). Nor indeed is it as obvious as it is often assumed to be that a generative model of a language-system should associate with each well-formed sentence one or more interpretations in some appropriate notation semantic interpretations, considered as representations of the meaning of sentences, are theoretical constructs, which must be justified (if they can be justified) in terms of their explanatory value; and there is nothing what every one would agree has to be explained in terms of the structure of the language-system (e.g., synonymy, antonymy, tautology, contradictoriness, entailment and paraphrase) that cannot be explained in terms of relations defined over sentences without the postulation of such intermediate theoretical constructs as semantic interpretations.

Though generative semantics is logically impendent of componential analysis, it undoubtedly derives much of its attraction from the fact that it is commonly presented in association with the assumption that it is possible to analyze the semantic structure of all languages in terms of a set of universal sense – components; and this assumption is, to say at least, questionable. Having made this point, however, it must be emphasized that, independently of the soundness of their underlying assumptions, both generative semantics and componential analysis have been of immeasurable importance in recent years in that they have obliged their practitioners to present their analyses in a precisely specified format; and this has brought them within the range of constructive criticism and emendation which has undoubtedly increased our understanding of the complexity of the issues involved.



  1. Bartsch,R. and Vennemann, T. (1972). Semantic Structures: a study in the relation between semantics and syntax. Frankfurt (2nd ed. 1973).

2.      Bussmann, H. (1996). Routledge Dictionary of Language and Linguistics (translated and edited by Gregory P. Trauth and Kerstn Kazzanza), London: Routledge Publishing Company.

3.      Chomsky, N. (1971). Deep Structure, Surface Structure, and Semantic Interpretation. In D.D. Steinberg and L.A. Jakobovits (eds). Semantics.Cambridge: 183-216.

4.      Chomsky, N. (1972). Studies on Semantics in Generative Grammar. The Hague.

5.      Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press.

6.      Jackendoff, R.S. (1972). Semantic Interpretation In Generative Grammar.Cambridge, Mass.: MIT Press.

7.      Katz, J.J. (1970). Interpretive Semantics vs. Generative Semantics. FL 6. 220-59.

8.      Katz and Fodor, J.A. (1963). The Structure of Semantic Theory. Lg 39. 170-210.

9.      Katz, J.J. and Postal, P.M. (1964). An Integrated Theory of Linguistic Description. Cambridge, Mass.: MIT Press. 16

10.  Lakoff, G. (1970). Linguistics and Logic. Synthese 22. 151-271.

11.  Lakoff, G. (1971). On Generative Semantics. In D.D. Steinberg and L.A. Jakobovits (eds). Semantics. Cambridge. 232-96.

12.  Lyons, J. (1977). Semantics (vol.2). Cambridge: Cambridge University Press.

13.      Matthews, P.H. (2007). The Concise Oxford Dictionary of Linguistics (2nded.). Oxford: Oxford University Press.

14.  Seuren, P.A.M. (1985). Discourse Semantics. Oxford.

15. Semantics

Основные термины (генерируются автоматически): MIT, NOT, CAUSE, BECOME, ALIVE.

Ключевые слова

Трансформационная грамматика, Поверхностная структура, Глубокая структура, Стандартная теория, Интерпретирующая семантика


Социальные комментарии Cackle
Задать вопрос