A restrictive theory of agrammatic comprehension.

In this paper I propose a new, restrictive theory of Trace-Deletion in agrammatism. This theory subsumes the Trace-Deletion Hypothesis (TDH; Grodzinsky, 1984a,b, 1986, 1990), which maintains that traces are deleted from agrammatic representations and that a cognitive strategy augments the patients' performance. This claim accounts for the pattern of loss and sparing observed in these patients' comprehension of a wide variety of syntactic constructions and is thus important for our understanding of the neural representation of syntax. Yet there are reasons for revising the account and making it more precise, stemming from both recent empirical findings and new developments in the theory of syntax. The original TDH was based on observations of agrammatic comprehension of structures containing traces resulting from either NP- or Wh-movement. Nevertheless, heads (as opposed to phrasal projections) also move and leave traces behind. Head movement (of verbs, for instance) has come to play a central role in linguistic theory (which currently postulates a wider variety of empty categories than any previous theoretical framework). Recent findings suggest that verb movement is retained in agrammatism, indicating that a sweeping claim regarding the deletion of all empty categories is too strong. This motivates the first restrictive move, resulting in a theory that picks out a restricted set of traces--only those for which deficient performance is indeed observed. All other empty categories are left intact. Trace-Deletion is tied to theta-positions. The second restrictive move is motivated by two types of surprising asymmetries that have recently been discovered for agrammatic comprehenders: First, agrammatic comprehension on passives of psychological predicates provides an error pattern that distinguishes this construction from agentive passive, indicating that the deficit is tied to the thematic properties of the predicate: Second, asymmetries have been observed in agrammatic comprehension of questions and quantifiers. These findings motivate a modification of the augmentative strategy, whose domain of application is restricted to referential NPs. Thus, the new account amounts to the claim that only traces in theta-positions are deleted, and that the strategy applies to referential NPs alone. This, I argue, not only derives all the data precisely but is also conceptually superior to any previous account of agrammatism. Finally, I discuss the consequences of this account to linguistic theory, and to theories of brain/language relations.