Grammaticality, acceptability and sentence processing: A psycholinguistic study

Linguistic theory is built on an empirical foundation consisting largely of sentence acceptability judgments, deemed to reflect underlying grammaticality. This thesis focuses on extra-grammatical influences that sometimes obscure such judgments. For example, processing resource limitations may lead a perceiver to reject a grammatical sentence because its constituents cannot easily be recognized. Further, extra-linguistic processes may influence sentence ratings due to the analytic nature of the judgment task. Thus, In order for the researcher to accurately delineate the boundary between grammatical and ungrammatical sentences it is necessary to identify and evaluate non-grammatical influences that muddy acceptability judgments. ^ Three experiments, exploiting the technique of monitoring eye-movements during reading, probe readers\u27 responses to various classes of unacceptable sentences. This technique was chosen because eye-movements serve as a rich source of information about how sentence features are assimilated as they unfold in time. The Referential Model of the human sentence processor (HSP) serves as a theoretical framework for interpreting results, paying careful attention both to computational devices and the resources available to them. ^ Experiments 1 and 2 contrast ungrammatical sentences, containing morpho-syntactic anomalies, with others that are grammatical, yet pragmatically anomalous. The results illustrate that eye-movements are sensitive to constraints invoked by a range of morpho-syntactic and pragmatic anomalies, and that responses to the two classes of anomaly are distinct. These findings are argued to support a model of HSP that engages distinct devices in response to each class of anomalous sentence. ^ Experiment 3 examines the status of constraints on long-distance syntactic movement. Judgment satiation is a phenomenon in which some, but not all, classes of initially unacceptable sentences are judged increasingly acceptable with additional exposure. This study uncovers an on-line counterpart to judgment satiation, where the effort expended in processing (satiable) wh-island violations increases with repeated exposure. No such increase is seen in the case of (non-satiable) adjunct island violations. The initial unacceptability of wh-island violations is argued to stem from a processing resource limitation, while the satiation effect derives from the HSP\u27s ability to dynamically shift resources to meet the moment-by-moment needs of its sub-components.