Towards a theory of language design
More recently Chomsky (1991) has suggested that the theory of grammar developed over the past two decades can be interpreted in terms of more general guidelines for language design. In essence, these guidelines legislate against ‘superfluous steps’ in derivations and ‘superfluous elements’ in representations. For example, the principle of Full Interpretation (Chomsky 1986) requires that every element in Phonetic Form (PF) and Logical Form (LF), the representations interfacing with systems of language use, must receive an appropriate interpretation. From Full Interpretation it follows that LF representations for natural languages may not contain vacuous quantifiers or arguments that are not functionally related to some predicate in the representation. Therefore the functional relatedness requirement of the θ-Criterion is subsumed under the principle of Full Interpretation, a more general requirement that representations be minimal in some sense. In this way Full Interpretation functions as a principle which determines the economy of representations.
With respect to the economy of derivations, Chomsky (1991) proposes a ‘last resort’ condition on movement operations to the effect that movement cannot apply to a constituent unless the nonapplication results in the violation of some principle of grammar. For example in the derivation of sentential passive constructions the last resort condition predicts that the object of the by-phrase is exempt from movement since it is both Case-marked and θ-marked in its underlying position, in contrast to the object of the passive predicate which is Caseless in its underlying position. Notice that the addition of this condition to UG makes a new prediction—namely that there are no languages in which a θ-marked and Case-marked NP can move to a Caseless position which receives no θ-marking. Without the last resort condition such movements are possible under the principles of UG. The last resort condition on movements comes under the more general heading of a ‘least effort’ guideline for derivations.
The last resort condition on movements also subsumes some of the empirical effects of other UG conditions—in particular, the functional uniqueness requirement of the θ Criterion and the TSC and SSC. Consider the standard case of a Functional Uniqueness violation ((1) above, repeated here):
(1)
Example (1) violates the last resort condition on movements because the underlying object Adam is both Case-marked and θ-marked in that position and therefore the movement is presumably not required by any principle of UG. The same sort of analysis applies in the standard cases of TSC and SSC violations. In each case the last resort condition on movement is violated because the NP moved is both Case-marked and θ-marked in its underlying position and therefore is not required to move by any UG condition. Although it remains to be demonstrated, it seems highly probable that the last resort condition on movements may subsume virtually all the empirical effects of the TSC and SSC as well as Functional Uniqueness.
In his most recent work (1993) Chomsky develops the idea that linguistic derivations and representations should be minimal in a certain sense into a more ambitious programme for linguistic theory. In this paper he proposes a minimalist programme for linguistic theory based on a number of plausible, though at this point speculative, assumptions. The primary assumption that motivates much of this programme is that only PF and LF count as levels of representation with respect to UG.
UG must determine the class of possible languages. It must specify the properties of SDs and of the symbolic representations that enter into them. In particular, it must specify the interface levels (A-P [articulatory perceptual], C-I [conceptual-intentional]), the elements that constitute these levels, and the computations by which they are constructed. A particularly simple design for language would take the (conceptually necessary) interface levels to be the only levels. That assumption will be part of the ‘minimalist’ program I would like to explore here. (p. 3)
The assumption is motivated by the fact that most principles of UG apply to either LF or PF representations. Part of the minimalist programme involves a demonstration that virtually all principles of UG apply at PF or LF and that apparent counterexamples of principles that hold at S-structure or D-structure can be reanalysed so that only PF and LF are the relevant levels. ‘Conditions on representations—those of Binding Theory, Case Theory, Theta Theory, etc.—hold only at the interface, and are motivated by properties of the interface, perhaps properly understood as modes of interpretation by performance systems’ (Chomsky 1993:6). To the extent that conditions on the economy of representations and derivations restrict the kinds of representations that occur at the interface levels, these conditions are crucial to carrying out the minimalist programme.
The discussion of the role of economy in grammatical description dates back to the advent of generative grammar—i.e., Chomsky’s MMH. There Chomsky identifies two kinds of criteria of adequacy for grammars—one concerning the correct description of the structure of the language under analysis, and the other concerning requirements imposed by its special purposes, ‘or, in the case of a linguistic grammar having no such special purposes, requirements of simplicity, economy, compactness, etc.’ (p. 1). In a footnote, Chomsky adds the following clarification:
Such considerations are in general not trivial or ‘merely esthetic’. It has been recognized of philosophical systems, and it is, I think, no less true of grammatical systems, that the motives behind the demand for economy are in many ways the same as those behind the demand that there be a system at all. Cf. Goodman (1943).
In other words, a grammar is not merely a description of a language; it is moreover an explanatory theory about the structure of a language—i.e., why a language has the properties it does. It is in this context that considerations of economy, etc. come into play.
In the earliest work on generative grammar, notions of economy and simplicity were considered as part of an evaluation measure, itself part of the theory of grammar, which ranks the potential grammars available (assuming, of course, that the theory allows for more than one grammar that is consistent with the linguistic data). Under the psychological interpretation of grammar, where linguistic theory provides an explanation for language acquisition, an evaluation procedure was thought to be an indispensable element of the theory.
To acquire language, a child must devise a hypothesis compatible with presented data—he must select from the store of potential grammars a specific one that is appropriate to the data available to him. It is logically possible that the data might be sufficiently rich and the class of potential grammars sufficiently limited so that no more than a single permitted grammar will be compatible with the available data at the moment of successful language acquisition, in our idealized ‘instantaneous’ model (cf. notes 19 and 20). In this case, no evaluation procedure will be necessary as a part of linguistic theory—that is, as an innate property of an organism or a device capable of language acquisition. It is rather difficult to imagine how in detail this logical possibility might be realized, and all concrete attempts to formulate an empirically adequate linguistic theory certainly leave ample room for mutually inconsistent grammars, all compatible with primary data of any conceivable sort. All such theories therefore require supplementation by an evaluation measure if language acquisition is to be accounted for and selection of specific grammars is to be justified; and I shall continue to assume tentatively, as heretofore, that this is an empirical fact about the innate human faculté de langage and consequently about general linguistic theory as well. (Aspects, pp. 36–7)
From this perspective the relationship between linguistic theory and the particular grammars it provides is that of an evaluation procedure. Given a corpus (e.g., the primary linguistic data a child is exposed to) and a set of possible grammars, linguistic theory ranks the grammars and selects the most highly valued one.
As noted in SS, ‘the strongest requirement that could be placed on the relation between a theory of linguistic structure and particular grammars is that the theory must provide a practical and mechanical method for actually constructing the grammar, given a corpus of utterances’ (pp. 50–1). A theory of this sort constitutes a discovery procedure for grammars. Early work on generative grammar assumed that linguistic theory would not be able to meet this requirement, and that only the weaker requirement of providing an evaluation procedure for grammars was attainable.
However, with the development of the principles and parameters framework during the past decade and a half an evaluation measure for grammars has become essentially superfluous. With the reduction of the transformational component to a set of elementary operations that are part of UG and with the elimination of phrase structure rules , it appears that linguistic theory allows only a severely limited number of grammars—perhaps only one for each given language. Under the minimalist programme of Chomsky (1993) an even stronger assumption is proposed: ‘there is only one computational system and one lexicon’, apart from a limited kind of variety found in the PF component and in the lexicon (including association of concepts with phonological matrices (what Chomsky refers to as Saussurean arbitrariness), properties of grammatical formatives like inflection, and detectable lexical properties like the linear orientation of heads and complements (the head parameter)). If this is correct, then current linguistic theory now provides a discovery procedure for grammars and has thereby achieved, in Chomsky’s own words, ‘a scientific advance of the highest importance’.
