Abstract
In three experiments, we investigate whether speakers tend to perseverate in the assignment of emphasis to concepts with particular thematic roles across utterances. Participants matched prime sentences involving clefts (e.g., Het is de cowboy die hij slaat, “It is the cowboy that he is hitting”) to pictures and then described unrelated transitive events. Participants were more likely to produce a passive after a cleft that emphasised the patient than after a cleft that emphasised the agent. Because prime and target sentences are syntactically unrelated, our study demonstrated nonsyntactic structural priming. We propose that speakers use such priming to facilitate the construction of coherent discourse.
Notes
1We also conducted linear mixed effect models on proportions of all patient-first structures (i.e., including short passives, impersonal structures, and reflexives). In all experiments, emphasis contributed (though marginally in Experiment 3) to the fit of the model and significantly predicted the tendency to produce passive descriptions in targets [Experiment 1: χ2(1) = 5.73, p < .02; N = 970, log-likelihood = –495.9, Wald Z = 2.465, p < .01; Experiment 2: χ2(1) = 6.13, p < .01; N = 953, log-likelihood = –479.8, Wald Z = 2.54, p < .01; Experiment 3: χ2(1) = 3.41, p < .06; N = 1072, log-likelihood = –550.2, Wald Z = 1.90, p < .05].
2To be consistent with a traditional approach in psycholinguistic research (which is, however, highly problematic for categorical data, Jaeger, 2008), we also submitted the data of all three experiments to ANOVAs on aggregated proportions with subjects (F1) and items (F2) as random effects. Results were consistent with those of mixed effect models.