Subject-verb agreement studies have played an increasingly prominent role in the production literature (e.g., Bock et al., in press). Speakers' productions of correctly inflected verbs are generally very fluent but not entirely error-free. Particular difficulty arises when an agreeing verb must be produced for a complex noun phrase (NP) containing both a singular and plural noun, such as "the key to the cabinets". One central finding from agreement studies is that NPs with a singular head noun (key) and a plural local noun (cabinets) result in a relatively large number of agreement errors such as "the key to the cabinets are". In contrast, there are few errors with a plural head and singular local noun, e.g., "the keys to the cabinet is".
This error asymmetry is one of the most replicated results in the agreement literature, but its underlying cause is not well understood. One account has explained the asymmetry in terms of feature markedness (Eberhard, 1997). We, however, propose an alternative account based on a distributional asymmetry in the input. Specifically, we present American English corpus data, from the parsed WSJ and Brown corpora, demonstrating that singular nouns take plural verbs more frequently than plural nouns take singular verbs. This pattern derives from several different constructions. For example, in group phrases such as "a number of analysts are misguided" and "the majority were poor", the singular heads (number, majority) appear with plural verbs. There are also several other constructions in English in which nouns appear with verbs that are homophonous with a plural form. For example, modals appear with infinitives, e.g. "Bill will run", so that a singular head (Bill) appears with a plural-sounding verb (run). Crucially, the converse patterns, plural nouns with singular verbs or singular-sounding verbs, almost never occur.
Thus, we suggest that the asymmetry in agreement error rate reflects distributional patterns of related grammatical utterances. Plural nouns typically agree only with plural verbs, while singular nouns can occur with singular verbs, plural verbs, and plural-sounding verbs. Thus, computing agreement for singular nouns is more difficult and error-prone than for plural nouns. This account makes substantial cross-linguistic predictions: distributional differences in the aforementioned and other constructions should modulate the degree of the error asymmetry. We will discuss these data in the context of other recent studies supporting a constraint-satisfaction account of agreement processes, in which distributional information of various types constrains the verb form to be produced.
References
Bock, J. K., Eberhard, K. M., Cutting, J. C., Meyer, A. S., & Schriefers, H. (in press). Some attractions of verb agreement. Cognitive Psychology.
Eberhard, K. M. (1997). The marked effect of number on subject-verb agreement. Journal of Memory and Language, 36, 147-164.