The interplay of syntax and form in sentence production: A cross-linguistic study of form effects on agreement

Abstract

We report four cross-linguistic experiments (in Spanish, Italian and French) testing the influence of morphophonological gender marking in the subject noun phrase on the production of gender agreement. Agreement errors are elicited using a methodology in which participants are required to complete, with a predicative adjective, a sentence preamble. Results confirm a role for morphophonological gender marking in agreement. More precisely, we show that this role varies with two factors of different nature. The first factor is structural and has to do with the position of the marker in the noun phrase (article vs. noun). The second factor is distributional and has to do with the validity of the marker in the language. A model of agreement production is proposed in which two functionally distinct processes are identified: Feature selection, the locus of the morphophonological influences, and Feature copy, operating under strict syntactic guidance.

Acknowledgements

The research reported here was supported by grant number 1114-068250.02 of the Fonds National de la Recherche Suisse to Julie Franck and Uli Frauenfelder, and by the NSF grant SBR 9729118 to Gabriella Vigliocco. We would like to thank Kay Bock, Fernanda Ferreira, Matt Goldrick, Rob Hartsuiker, Luigi Rizzi and Dave Vinson for insightful discussions and comments on previous versions of the manuscript.

Notes

¹Throughout the text, P refers to plural, S refers to singular, M refers to masculine, F refers to feminine, while 0 refers to the absence of a morphological marker.

²At the linguistic level, the three languages differ as to the morphological status of the gender marker on the noun: whereas in Italian and in Spanish it is inflectional in nature, the status of nouns’ endings in French is strictly phonological.

³As Vigliocco and Hartsuiker (2002) discussed it, a strictly feedforward model can also account for the data if it includes the assumption of an internal monitoring component (e.g., Levelt, 1983, 1989). However, the empirical criteria recently proposed by Hartsuiker (2006) to distinguish feedback from monitoring rather support an interpretation of morphophonological effects on agreement in terms of feedback.

⁴Hupet, Fayol, and Schelstraete (1998) and Thornton & MacDonald (2003) reported significant effects on agreement of the plausibility of the intervener. That is, whether the intervener could be plausibly used as subject of the sentence affects agreement. This effect may be taken to suggest that agreement is influenced by the conceptual properties of the intervener. However, these effects may straightforwardly be accounted for in terms of occurring during functional assignment processes, rather than agreement processes: the more plausible as a subject the intervener, the more likely it could be erroneously assigned the subject function.

⁵The output of feature selection is an abstract, syntactic feature that is not available directly and has to be inferred. What is the syntactic feature when the conceptual and morphophonological representations differ (e.g., ‘label’ in ‘the label on the bottles’ or ‘eggs’ in ‘the scrambled eggs in the corner’)? In our view, the syntactic feature is the one that shows up on the agreement target, by definition of agreement. If the speaker says ‘the label on the bottles is green’, the singular feature has been selected in line with its morphological realisation; if the speaker says ‘the label on the bottles are green’ or ‘the scrambled eggs in the corner wants more coffee’, the feature has been selected in line with its conceptual representation. That is, the lack of a mapping between the different levels of representation is at the lexical level of feature selection, not at the syntactic level of feature copy which blindly transmits the selected feature to the agreement target.