Production of subject-verb agreement in Slovene and English

Abstract

This thesis explores the mental representation of subject-verb agreement, and the factors that can affect the determination of agreement in language production. It reports nine experiments that used a task in which participants produced sentence completions for visually presented complex subjects such as “The greyhound which two lively rabbits were tempting”. Such completions typically agree with the head noun (greyhound) as in “A greyhound which two lively rabbits were tempting is jumping” but sometimes agree with the local noun (rabbits) as in “A greyhound which two lively rabbits were tempting are jumping”. The first experiments examined the value of the concept of markedness in subject verb number agreement to see whether it has explanatory power for languages like Slovene with more than two number values. Results from two experiments employing complex sentence preambles including a head noun post modified by a prepositional phrase or a relative clause (e.g., “The nudist(s) near the sand dune(s)”) show that Slovene number agreement differs from number agreement in languages with no dual, but that it is not possible to simply state that the singular is the least marked and the dual the most. I argue that using languages with more complex number systems allows greater insight into the processes of correct and erroneous subject-verb agreement, and shows that it is necessary to dissociate susceptibility to agreement from error-causing status. To conclude, the concept of markedness seems unable to explain my results. Semantic effects in agreement are then examined using two comparison experiments in English. Experiment 3 shows that although English has only a two value system, speakers are sensitive to semantic differences in number. Experiment 4 explores the possible influence of speakers’ native language three-value number system on their two-value second language system. It shows that native speakers of English are more sensitive to semantic number differences in English than Slovene speakers of English. Experiment 5 explores gender agreement in Slovene (which has three genders) and shows that there is a complex pattern of agreement. As with number, there is not just one number value which is problematic: neuter and masculine are most confusable, but masculine errors are also common when feminine agreement would be expected, thus suggesting that speakers revert to two different defaults, masculine and neuter. Finally, the results of four experiments examining number and gender agreement in coordinated phrases are presented. Agreement in such phrases may be resolved (i.e. the verb agrees with the whole subject) but may instead agree with one conjunct. Agreement with one conjunct is affected byword order (agreement with the nearest conjunct is most common), coordinator (e.g., single-conjunct agreement is more common after “or” than “and”) and the gender or number of the conjuncts (e.g., dual number is associated with single-conjunct agreement. Taken together, my results suggest that agreement is affected by a complex interplay of semantic and syntactic factors, and that the effects of a three-valued system are quite distinct from those of a two-valued system.