Plausibility and grammatical agreement

Three experiments examined plausibility effects on the production and comprehension of subject–verb agreement. In a production task, participants were given a verb and sentence preamble and asked to create a complete passive sentence. The preambles contained two nouns (e.g., the album by the classical composers). The plausibility of the verb was manipulated so that either (a) both nouns could be plausible passive subjects (e.g., praised, as both albums and composers can plausibly be praised) or (b) only the head noun could be a plausible subject (e.g., played, as only albums can plausibly be played). The comprehension task was self-paced reading with the same materials. The results from both methodologies demonstrated robust plausibility effects. There were higher agreement error rates in production and longer RTs at the verb in comprehension when both nouns were plausible subjects than when only the head was plausible. Implications for current production models are considered and an alternative account is presented that is motivated by current comprehension models and other recent production data. 2003 Elsevier Science (USA). All rights reserved. Many of the world s languages exhibit agreement phenomena, meaning that the forms of two or more words in a sentence match in some way. For example, in English, finite verbs typically agree with their subjects, so that in the sentence my best friend is here, both the noun and verb are singular, whereas in the sentence my best friends are here, both are plural. Because agreement is a fundamental aspect of language processing, studies of grammatical agreement have played an increasingly central role in language production research in recent years. To examine how speakers produce agreement, production researchers have used a sentence completion task to elicit subject–verb agreement errors and thereby shed light on the agreement production processes (e.g., Bock, Eberhard, Cutting, Meyer, & Schriefers, 2001; Bock & Miller, 1991; Eberhard, 1997; Vigliocco, Butterworth, & Garrett, 1996). In this task, participants are given a sentence subject and asked to use it in producing a complete sentence. On critical trials, the subject contains two noun phrases (e.g., the key to the cabinets), one of which is the head NP (the key), because it contains the head noun of the subject, and the other is referred to as the local NP. Several studies have found that when the number marking on the local NP mismatched that of the head, participants were more likely to produce an agreement error than when the number of both nouns matched (e.g., Bock & Miller, 1991). These findings, along with other studies showing that non-syntactic factors do not seem to modulate agreement error rates (Bock & Miller, 1991; Bock & Eberhard, 1993), led many production researchers to assume that the mechanism underlying agreement errors is some sort of syntactic feature processing during a strictly grammatical stage of the production process (Bock & Eberhard, 1993; Eberhard, 1997). On this view, which we will term the Journal of Memory and Language 48 (2003) 740–759 Journal of Memory and Language www.elsevier.com/locate/jml Corresponding author. Present address: Department of Linguistics and Cognitive Science, 550 Harvard Ave., Pomona College, Claremont, CA 91711, USA. Fax: 1-909-621-8065. E-mail address: [email protected] (R. Thorn-