ABSTRACT
In work on grammatical agreement in sentence production, there are accounts of verb number formulation that emphasise the role of whole-structure properties and accounts that emphasise the role of word-driven properties. To evaluate these alternatives, we carried out two experiments that examined a referential (wholistic) contributor to agreement along with two lexical-semantic (local) factors. Both experiments gauged the accuracy and latency of inflected-verb production in order to assess how variations in grammatical number interacted with the other factors. The accuracy of verb production was modulated both by the referential effect of notional number and by the lexical-semantic effects of relatedness and category membership. As an index of agreement difficulty, latencies were little affected by either factor. The findings suggest that agreement is sensitive to referential as well as lexical forces and highlight the importance of lexical-structural integration in the process of sentence production.
Acknowledgements
We would like to thank Ben Carter for collecting, transcribing, and coding the Experiment 2 data, Taylor Eighmy for providing voice recordings, transcription, and coding, and Megan Krull, Alexa Mazza, and Stacey Cohen for data collection and transcription. We are particularly grateful for advice from anonymous reviewers and from Gary Dell, Maureen Gillespie, Maryellen MacDonald, and Darren Tanner.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. Margin of error is defined as the half-width of the 95% confidence interval for differences between condition means. For these experiments, we calculated this confidence interval from the mean-squared error (MSE) of the highest-level interaction in a repeated measures ANOVA by items, using a Scheffé correction and the type III sum of squares.
2. The numbers of items going into each of these cells are extremely unbalanced. Standard error of the mean (SEM) gives a sense of the reliability of each mean estimate while taking the sample size into account.