Training sentence production in agrammatism: implications for normal and disordered language.

This paper presents an overview of our work concerned with treatment of sentence production deficits seen in agrammatic (Broca's) aphasic individuals. Using a single-subject experimental research paradigm, we examined emergent sentence production patterns in a subset of Broca's subjects who evinced sentence production (and comprehension) deficits involving "complex" sentences in which noun phrases (NPs) have been moved out of their canonical positions. We used aspects of Chomsky's Principles and Parameters approach of Government Binding (GB) Theory (Chomsky, 1986, Chomsky & Lasnik, 1991), as well as findings from the psycholinguistic and neurolinguistic literature as a basis for selecting sentences entered into our experiments and for designing our intervention approach, in general. Subjects were trained to produce sentences which rely on NP-movement (i.e., passives) or WH-movement (i.e., wh-questions, object clefts)--sentences formed by applying the rule "Move-alpha" in which moved constituents leave behind a "trace" or "gap" of their movement. Training emphasized the lexical and syntactic properties (e.g., thematic role assignment, NP-movement) of target sentences. Throughout this training we carefully measured generalization to untrained sentences relying on similar movement operations and error patterns were examined as they evolved over time. Results of our work indicated not only improved sentence production abilities in all subjects under study, but also--in many cases--generalization of sentence production across linguistic lines. That is, training WH-movement structures (e.g., object clefts) improved production of untrained WH-movement constructions (e.g., wh-questions) that are very different in their s-structure representation; however, no effect of this training on NP-movement structures occurred. In addition, within the class of wh-questions, generative production across questions relying on argument (direct object NP) movement (i.e., what- and who-questions) occurred in the absence of generalization to wh-questions requiring movement from adjunct position (i.e., where- and when-questions) for some subjects. For others, generalized production occurred only when wh-morphemes were the focus of treatment, indicating that at least two processes must be completed for successful wh-question production to take place: movement of the wh-item itself and control of sublexical features that determine wh-morpheme selection. These data are discussed in terms of the contribution that detailed recovery data, controlled for lexical and syntactic properties of sentence production, can make both to understanding the nature of sentence production deficits and to issues regarding normal sentence production.