Processing morphological information in the parafovea: What role does prediction play?

ilkin, Z. & Sturt, P.

University of Edinburgh, Psychology Department

We will examine to what extend syntactic prediction could influence eye movement behavior in reading. We will report two experiments where the morphology of the upcoming word was predicted from previous sentence context. We found preliminary evidence indicating that if the morphology of the upcoming noun (plural/singular) is predicted in advance then readers are more likely to skip that noun. In the first experiment we will investigate the fixation probabilities and durations on nouns (puppet/puppets) when there is an initial mismatch between the article and the subsequent noun in phrases like ‘these/the green puppet creations’ or ‘these/the green puppets created’. If readers are more likely to access morphological information from the parafovea when this information is predicted, this should reflect it self in skipping rates and following fixation durations on the parafoveal word. The second experiment will investigate whether the readers are more likely benefit from the parafoveal information when the upcoming word had the typical form features of the expected syntactic category of that word. In a MEG study Rabagliati et al. (2010) showed that when a noun is predicted the typical form features that are associated with that category is also activated. They claim that there is sensitivity in the visual cortex for the expected typical information. We will compare processing of typical form features of the parafoveal noun, when there is a strong expectation for a noun then when there is not; in an eye tracking experiment. We will also manipulate the availability of the parafoveal preview information. If there is actually an early sensitivity to the typical features of the predicted syntactic category this should also influence the processing of the morphological information in the parafovea. We will discuss the implications of these results for models of lexical access in sentence processing and for the reading models.