Contemporary theories of subject-verb agreement production generally assume a process whereby features of the subject are conveyed or copied to the verb. Based on several problem cases, it is argued in this thesis that the identity of the subject cannot be taken as a given. An alternative approach is proposed in which identifying the subject is seen as a non-discrete, probabilistic process that is an important component of computing agreement. This view contrasts with the traditional view, in which identifying the subject is logically prior to and separate from agreement itself. This thesis seeks to distinguish between these views by testing whether distributional cues can influence which noun a verb agrees with. Distributional cues only become available during the middle stages of production, when sentence structure is determined and word order is set, and after grammatical functions are assigned. Thus, under the traditional view, distributional cues should play no role in identifying the subject.; Eight studies examined two different types of distributional cues. The first four studies examined disjunctions such as "the boy and the girls", and yielded strong evidence that simple linear proximity between elements influences which noun a verb agrees with. The final four studies examined more subtle cues (definiteness and choice of preposition), and yielded preliminary evidence for distributional influences. Collectively, the studies establish the viability of the new approach, and point the way to future research to further distinguish between the alternatives. The implications of the new approach for theories of agreement production are discussed.
展开▼
