Describing simple events: The dynamics of incremental grammatical encoding

Abstract

Past research on grammatical encoding in speech production mostly involved either determining the proportions of choices between alternative syntactic structures as a function of the conceptual availability of referents, or comparisons of utterance onset latencies for given syntactic structures. We will argue that this approach does not do full justice to the dynamics of the incrementality of production. In five experiments we asked participants to describe animations involving two moving objects. We manipulated the temporal availability of pieces of the conceptual input by withholding for short periods of time either (1) the information about the movement direction of one of the objects, (2) its identity, or (3) both types of information. The results show that both the temporal availability of different types of conceptual information and the latency of utterance onset affect the syntactic structure of descriptions. The results are discussed in the context of theories of incremental syntactic encoding in language production.

Keywords:

• Descriptions
• Grammatical encoding
• Incrementality
• Language production

Notes

¹Translation by the authors. The original quotation in German is as follows: “Ich glaube, dass mancher grosse Redner, in dem Augenblick da er den Mund aufmachte, noch nicht wusste, was er sagen würde. Aber die Überzeugung, dass er die ihm nötige Gedankenfülle schon aus den Umständen, und der daraus resultierenden Erregung seines Gemüts schöfen würde, machte ihn dreist genug, den Anfang, auf gutes Glück hin, zu setzen.”

²In the remaining experiments of this paper, we applied the same exclusion criteria. In this way, we ensured that the participants whose data are entering the eventual analysis are comparable across experiments in not using any specific strategies.

³From a strict statistical perspective, the use of linear contrasts could be seen as not being fully adequate because the levels of the factor SOA are not spaced equally. However, additional standard ANOVAs for all trend analyses reported in the present paper confirmed the results from the trend analysis for all but one comparison: The interaction effect between SOA and speed in Experiment 2 was not significant in the standard ANOVA.