MP vs. GB
As Chomsky and others have pointed out (see Chomsky 1995d; Freidin, 1997a; Lasnik, 1999), the MP is grounded in the principles and parameters framework, just like the GB theory it has to a large extent superseded. It shares at least three basic assumptions with virtually all of its predecessors within modern generative grammar. First and foremost, the mind/brain contains a language faculty, a component that interacts with other cognitive systems. Next, the cognitive system of language connects with performance systems via levels of linguistic representation, perhaps limited to only two external systems, one involving articulation and perception of the physical linguistic signal and the other involving the encoding or decoding of the meaning of the signal. Under this limitation there are only two interface levels, phonetic form (PF) and logical form (LF). Finally, performance systems do not differ across languages. Even at this fundamental level Chomsky takes nothing for granted, noting that these assumptions are ‘not at all obvious’ (1995d: 3).
As both the GB theory and the MP are grounded in the principles and parameters model, they share further specific assumptions [cf. Chomsky, 1995d: 170]:
(i) regarding computational system for human language, CHL, the initial state S0 contains invariant principles and parameters (i.e., options restricted to functional elements)
(ii) selection Σ of parameters determines a language (iii)a language determines an infinite set of linguistic expressions (SDs), each pair (π, γ) obtained from the interface levels, PF and LF
(iv) language acquisition involves fixing Σ
(v) the grammar of language states just Σ, lexical arbitrariness and the PF component aside
These shared assumptions have been standard for nearly two decades.
There is one further assumption articulated for the first time in Chomsky (1993), which could easily have been proposed prior to the MP. Chomsky offers as a narrow conjecture the suggestion that there is no variation in the overt syntax or the LF component. Ignoring PF options and lexical arbitrariness, Chomsky suggests that Variation is limited to nonsubstantive parts of the lexicon and general properties of lexical items,’ in which case ‘there is only one computational system and one lexicon’, apart from the variation mentioned (1995d: 170).1 Thus viewed at the appropriate level of abstraction, there is only one human language. Of course the truth or falsity of this bold conjecture remains to be established.
We come at last to the three assumptions that are unique to the MP: (1) that the interface levels LF and PF are the only relevant linguistic levels, in spite of apparent empirical evidence to the contrary (p. 169), (2) that all conditions are interface conditions (p. 194), and (3) that a linguistic expression is the optimal realization of these conditions (p. 194). These three assumptions constitute what Chomsky (2001) calls the strong minimalist thesis. In contrast to the six basic assumptions that the MP shares with GB theory, the three assumptions unique to the MP do not add up to ‘a major paradigm change in the theory of grammar’ as will become clear from the following discussion.
For LLJ the contrast between GB theory and the MP is that ‘the MP adds economy principles’ ‘in addition to local constraints on operations and the structures they produce.’ While it is correct to highlight the role of economy in the development from GB to the MP, it is inaccurate to claim that economy conditions are an innovation unique to the MP.
The discussion of the role of economy in grammatical analysis begins with the advent of modern generative grammar—i.e., Chomsky’s MMH (1951). There Chomsky identifies two kinds of criteria of adequacy for grammars—one for the correct description of the structure of the language under analysis, and the other for 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.’ (1951:1). In a footnote, Chomsky supplies 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 rather than other conceivable properties. It is in this context that considerations of economy, etc. first came into play.
Applying economy conditions to the selection of derivations within each grammar represents a significant leap from applying them to the selection of grammars. Although economy conditions like Full Interpretation (FI) and Last Resort (LR) were not articulated in the earliest discussions, both were introduced as part of GB, the former in Chomsky (1986a) and the latter in Chomsky 1991 (which first appeared in MITWPL #10 in 1989, three years prior to the advent of the MP). What happened in the 1990s, in a nutshell, was that certain economy conditions (e.g., FI) were interpreted in a way that made it natural for them to supplant a significant portion of the older GB principles.
Consider for example the Case Filter. It prohibits any phonetically realized nominal expression that is not marked for Case. Within GB, why this should be is just a matter of stipulation that appears to accord with the facts. From the point of view of economy, Case features are extraneous to interpretation at the LF interface at least and therefore should be eliminated from the derivation before LF. Otherwise, these uninterpretable features at the LF interface will violate FI, causing the derivation to crash. Thus legibility conditions imposed by the cognitive system that interfaces with CHL determines how derivations proceed with respect to Case features. In this way FI and the Case Filter overlap in a way that FI subsumes the empirical effects of the Case Filter, but not conversely. The preference for FI over the Case Filter is just the standard preference for the more general constraint, all things being equal.
In short, the heuristic of eliminating overlapping conditions, which has resulted in much fruitful research over several decades, is one of the central motivations for switching from the analysis of GB to that of the MP.
The same logic provides one of the strongest motivations for eliminating the relation of government from current discussion. Consider the standard case of an ECP violation given in (1), where T stands for the functional category Tense and t is the trace of John.
Under MP analysis, (1) can be interpreted as a violation of FI in the following way. Suppose that the nominative Case feature of John is checked in the embedded clause by T, as is the nominative Case feature of T. When John moves to the subject position of the matrix clause, Spec-T, only the D-feature of the nominal is available to check the D feature of the matrix T. The nominative Case feature of John has been checked and thereby eliminated, so that the nominative Case feature of the matrix T remains, causing the derivation to crash at LF because of FI. Thus the phenomena that fall separately under the Case Filter and the ECP within GB, are captured under a single principle within the MP analysis—moreover, one that functions as a natural legibility condition that regulates the interface between CHL and other cognitive systems.
Given the greater generality of FI, we would prima facie want to eliminate the ECP in favor of the more general principle. Furthermore, this analysis, which explains deviance on the basis of legibility conditions imposed by cognitive systems that interface with CHL, strikes us as a more promising explanatory account than the postulation of various constraints internal to CHL that basically reflect the complexity of the phenomena in an essentially descriptive fashion. Note incidentally that this approach is highly reminiscent of the one followed in Chomsky and Lasnik (1977). However, even if we discount impressions of what might be a more promising explanatory account, the methodological requirement to eliminate overlapping conditions whenever possible motivates the abandonment of much of the machinery of GB in favor of the MP analysis. Thus the change from GB to the MP is motivated by the same methodology that has always motivated changes in syntactic theory.
Another motivation for exploring the MP rather than continuing with GB concerns the central topic of phrase structure. The MP introduces bare phrase structure theory, which eliminates the ambivalent top-down and bottom-up view of phrase structure that has been characteristic of the field since the earliest formulations of X-bar theory in the late 1960s. With bare phrase structure there is only bottom up analysis of a specific kind. The theory of bare phrase structure provides a derivational mechanism for syntactic representations, which has been missing from GB since the elimination of phrase structure rules (circa 1980) on the grounds that they are redundant given X-bar theory, the general principles of GB, and the specific properties of lexical items. Furthermore, bare phrase structure theory as incorporated within the MP cannot produce canonical D-structures, hence the elimination of D-structure as a level of representation follows from the nature of CHL rather than a methodological stipulation.
Bare phrase structure eliminates in principle categorial distinctions for levels of phrasal projection, thereby conforming to the Inclusiveness Condition, which restricts computations solely to the elements (features) contained in lexical items. Given this condition, computations cannot introduce new elements such as bar levels, indices, or syntactic categories that are not already part of the lexical items computed.
Chomsky (1995d) is very clear that the Inclusiveness Condition provides one criterion for the ‘perfection’ of CHL. Although precisely what Chomsky intends by talking about language as a perfect system may not be totally clear in Chomsky (1995d), this discussion is clarified considerably in Chomsky (2000), written in 1998. Here the issue is represented in terms of the language faculty (FL) as a solution to legibility conditions imposed by the cognitive systems that interface with it. An optimal solution would encompass a CHL restricted to just the properties of lexical items involved in computations plus just those few operations required for derivations that connect LF with PF (perhaps just the elementary operations of adjunction/concatenation (for Merge and Move) and deletion (for feature checking)). If the behavior of derivations is controlled solely by legibility conditions imposed by other cognitive systems at the interfaces, then CHL can be reduced to these bare necessities, excluding additional internal machinery like the Case Filter and the EGP. Thus in terms of simplicity, economy, and non-redundancy the MP is clearly preferable to GB.
