Language

This work grew out of presentations at a 1999 LSA Institute workshop, which marked the twenty-fifth anniversary of David Perlmutter and Paul Postal's 1974 LSA Institute course (where the theory of relational grammar (RG) was first presented). [End Page 468]

While the papers in this volume all bear on the issue of GRAMMATICAL FUNCTIONS (or GRAMMATICAL RELATIONS, henceforth GRs) in contemporary syntax, they also give a snapshot of the state of syntax around the turn of the millennium. There is a lot to like about this work: there are interesting data, large theoretical issues, subtle argumentation, and clever analyses. Why, then, do I come away feeling that the field is lacking something? Perhaps it is because the machinery of modern syntax has become quite elaborate; that is, the delta between the minimal structure demanded by the data and those proposed for largely conceptual reasons has exceeded my pain threshold. The result is a heavy burden on engineering devices whose primary task appears to be to derive more concrete structures from the abstract ones.

Relational grammar departed from standard transformational grammar by giving theoretical primacy to GRs, such as SUBJECT, OBJECT, and so on. While these notions were derived from constituent structure in Chomsky 1965, RG treated them as theoretical primitives.

As transformational grammar evolved, eventually into the minimalist program, the constituent structures became more abstract, less tied to constituency, and more transparently tied to thematic role and Case assignment (or checking). Given a rough correlation between thematic roles and initial GRs on the one hand, and abstract Case and final GRs on the other, coupled with the de-emphasis of constituency as a determining factor behind phrase structure, one is naturally led to the central question of the volume, which might be restated, somewhat uncharitably, as 'has transformational grammar evolved into a "Rube Goldberg" version of RG?'.

The papers in Part 1 most directly focus on the central question of the volume. Davies and Dubinsky's introduction discusses the fate of final GRs in current minimalist approaches. They note that the increase in functional projections has created analogs for relation changes (RG REVALUATIONS) for both subjects and objects: in both cases there have been arguments for movement from a VP-internal position to the specifier of a functional projection. Thus, they conclude that GRs are largely epiphenomenal in this current framework. Nevertheless, they point out that subjects (involving the TP and Agr_SP projections) exhibit behaviors that are distinct from objects (involving the Agr_OP projection).

Davies and Dubinsky note that there is a good deal of crosslinguistic (not to mention cross-analysis) variation with respect to whether and how movement occurs. For example, in the case of subjects, movement to Spec of TP may be motivated by the EPP. However, as has been proposed by Alexiadou and Anagnostopoulou (1998), among others, the EPP may be satisfied by head movement of the verb, leaving the VP-internal subject in situ. Davies and Dubinsky go on to suggest that the several subject properties (e.g. nominative case, verb agreement, and controllability) correlate with different structural syntactic properties—these happen to coincide in languages like English, but are dissociated in other languages. Similarly, evidence for object movement varies crosslinguistically and cross-construction types. Davies and Dubinsky propose a set of four binary parameters that defines an empirical space that is consistent with this crosslinguistic variation. While these four parameters define sixteen language types, only four of these sixteen are discussed. This may be alternatively seen as the beginning of a research program or an indication that the tools are too blunt to accurately model attested behaviors.

Alex Alsina's paper examines the correlation between government binding (GB) Case and LFG grammatical functions (similar to RG final GRs). He correctly points out that GB, as exemplified by Baker (1988a,b), cannot claim to have derived GRs, as the two objects of double object constructions are distinguished by an...