From the outset, modern generative grammar has been concerned with the construction of precise models for linguistic structure. As Chomsky noted in the preface to Syntactic Structures, such models
can play an important role, both negative and positive, in the process of discovery itself. By pushing a precise but inadequate formulation to an unacceptable conclusion, we can often expose the exact source of this inadequacy and, consequently, gain a deeper a deeper understanding of the linguistic data. More positively, a formalized theory may automatically provide solutions for many problems other than those for which it was explicitly designed. Obscure and intuition-bound notions can neither lead to absurd conclusions nor provide new and correct ones, and hence they fail to be useful in two important respects. (Chomsky Reference Chomsky1957, p. 5)
One can evaluate competing proposals in linguistics to the extent that they are formulated precisely. In the case of the target article, this is not possible because there are no such proposals.
In general, the target article eschews precise analysis and offers instead a number of general descriptive statements about a range of languages that are taken to be refutations of proposed universals. Often these are offered without citation of data or, more importantly, the explicit analysis that would support the claims. Such refutations are based on unstated and unverified assumptions about grammatical analysis. In general, such discussion is not useful.
Consider the discussion of the Subjacency Condition in section 3, one of only two UG principles in generative grammar discussed.Footnote 1 Evans & Levinson (E&L) give no formulation of this condition; nor do they give explicit syntactic analyses of the examples cited in (6–8). Example (6) is actually ambiguous (i.e., ‘where-said’ versus ‘where-got off’) and Example (8), which is marked deviant, is grammatical under the interpretation of ‘where-said’. The inadequate analysis of (8) offered in Example (9) does not distinguish between the possible and impossible interpretations because a place adverbial modifying the main clause VP occurs in the same linear position as one modifying the subordinate clause VP. Only the analysis of hierarchical structure disambiguates the two interpretations. Either where starts out as a constituent of the subordinate clause (the impossible interpretation) or of only the main clause (the possible interpretation). Moreover, if whether is analyzed as a complementizer on a par with that in Example (6) rather than a wh-phrase, then the Subjacency Condition does not account for the deviance of (8) on the ‘where-get off’ interpretation.
To understand E&L's criticism, one needs to know that the Subjacency Condition is a locality constraint that restricts the scope of syntactic movement, thereby contributing to an account of so-called island phenomena in natural language, a central topic with an extensive literature.Footnote 2 E&L attempt to dismiss the Subjacency proposal on the grounds that it “does not work in Italian or Russian in the same way, and theorists have had to assume that children learn the specifics of the constraint after all” (sect. 3, para. 6). This is simply wrong, as illustrated for Italian in Rizzi's famous 1978 paper, which demonstrates that Subjacency holds in Italian but is parameterized in terms of which clause boundary type counts as a bounding domain for the constraint (CP for Italian, but TP for English)Footnote 3 (Rizzi 1978). The parametric analysis (including Russian, which apparently has no interclausal wh-movement) has been spelled out in detail in this journal; see Freidin and Quicoli (Reference Freidin and Quicoli1989), which analyzes the poverty of stimulus issues for parameters, arguing that the choice of parameter settings for Subjacency cannot be made solely from presentation of data.Footnote 4 Thus, some parameters that account for variation across languages are also part of UG.
E&L conclude their discussion of Subjacency by saying, “Each constraint in UG needs to be taken as no more than a working hypothesis, hopefully sufficiently clearly articulated that it could be falsified by cross-linguistic data” (sect. 3, para. 6). The first part is a truism. However, the assumption in the second part that data alone can falsify any working hypothesis (especially one that works for a significant range of data) is clearly false. As Chomsky noted about Subjacency more than thirty years ago,
It is easy enough to find phenomena that appear to violate the subjacency condition. Consider, e.g., the sentences (9), (10), where there is a relation between the phrase in bold face and the position marked by t, “violating” subjacency under the assumption that the rule in question is a movement rule:
(9) Johnseems [S′to be certain [S′t to win] ]
(10) whodid Mary hope [S′that Tom would tell Bill [S′that he should visit t] ]
Putting the matter more carefully, a proposed condition on rules, such as subjacency, cannot be confirmed or refuted directly by phenomena of this (or any other) sort. A condition on rules can be confirmed or refuted only by rules, which observe or violate it, respectively. If the rule of NP movement that yields (9) applies successive cyclically, as often assumed, then the rule will observe subjacency. If, as I have argued in the references cited, the rule of wh-movement applies successive cyclically, then it too will observe subjacency, giving (10). To find evidence to support or to refute a proposed condition on rules, it does not suffice to list unexplained phenomena; rather, it is necessary to present rules, i.e., to present a fragment of a grammar. The confirmation or refutation will be as convincing as the fragment of grammar presented. This is a simple point of logic, occasionally overlooked in the literature. The status of conditions on rules is empirical, but evidence can only be indirect and the argument, one way or another, is necessarily rather abstract and “theory bound.” (Chomsky Reference Chomsky and Culicover1977, p. 74)
Data alone cannot speak to the validity of explicit proposals about the content of UG. What is required is an explicit analysis of data that follows from a precisely formulated fragment of a grammar. This is a comment about methodology, independent of any particular linguistic theory. In science there is no alternative to providing explicit analysis of data. The discussion of UG in this article misses the mark entirely.
