Coordinating Parsing and Grammar

Since the discussion of the competence{performance dichotomy in Chomsky (1965), it has been commonly assumed that the problem of characterizing an individual's grammatical knowledge is distinct from that of characterizing how and whether a speaker/hearer builds grammatical representations during language production/comprehension. A linguist presented with data concerning some facet of language must determine which aspects of this evidence are best explained by appeal to properties of a speaker's competence and which are due to the performance mechanisms. In some cases there has been virtual unanimity as to the locus of the explanation. The degraded acceptability of sentences involving center-embedding has almost always been accounted for in terms of some type of processing limitation (Miller and Chomsky 1963, inter alia). On the other hand, the complex distribution of clitic pronouns in Romance languages is widely agreed to be the result of grammatical properties, be they syntactic or phonological (Kayne 1975, inter alia). There are a number of cases, however, in which the locus of explanation remains the subject of some debate. Witness, for example, the varying accounts of unacceptability of sentences having crossing dependencies (cf. Fodor 1978 vs. Pesetsky 1982), or di ering explanations of the ill-formedness of sentences involving extraction from islands (cf. Ross 1967, inter alia, vs. Marcus 1980). The pervasive phenomenon of coordination has been the focus of linguistic research from the very beginnings of generative grammar (Chomksy 1957, Gleitman 1965). In spite of this history, the peculiar properties of coordination have exerted surprisingly little in uence on the development of theories of syntactic competence, in contrast with, say, passivization and question formation (to name two other phenomena discussed in Chomsky 1957). This curious state of a airs might be explained if there were no non-trivial properties of coordination: any theory could, with equal ease, explain the phenomenon. However, it is well-known that this is not the case. Species of coordination such as gapping, non-constituent coordination, etc. o er a rich array of complicated syntactic patterns. Why then have there been so few attempts to integrate coordination into the fold of syntactic theorizing? Crit Cremers in On Parsing Coordination Categorially (hereafter C) and Marjan Grootveld in Parsing Coordination Generatively (hereafter G) argue that this separation of coordination from the core of grammatical theory is well-motivated. In slightly di erent ways, they make the intriguing proposal that the \interesting" properties of coordination are due to aspects of the parsing mechanism rather than the grammar and that the structure of coordination does not interfere in any way with other grammatical processes. Furthermore, they both conclude that the traditionally vexing issue of the phrasal versus sentential origin of coordination is of no relevance to this crucial parsing process. Both of these suggestions raise a great many empirical and conceptual issues and I will attempt to raise a few of these, highlighting C's and G's perspectives on them, and in some places pointing at alternatives. Naturally, we will be able to draw no de nitive conclusions in the