Semantic constraints in reading An eye-tracking study of highly restrictive verbs in Italian

Predictability plays a major role in reading research as it’s thought to strongly influence word recognition in cognitive models of eye movement. While it is broadly defined by the probability of a word appearing in a given context, experimental designs in text-based eye-tracking studies are usually centred around lexical combinations, thus conflating a rather wide range of linguistic phenomena, from statistical co-occurrence (e.g. collocations) to intrinsic semiotic properties of words (e.g. light verb constructions). Therefore, the question arises whether changes in processing costs can be explained solely by context. The present study investigated a specific class of verbs in Italian whose properties entail a restriction on the semantic field of their referent (e.g. mangiare, ‘to eat’) since previous studies within the visual world paradigm have shown anticipatory effects in similar cases. Forty -six native speakers of Italian took part in a text-based eye-tracking experiment, where 18 pairs of highly and weakly restrictive verbs were embedded in sentences with no prior informative context. Three areas of interest (Verb, Noun, Spillover) were inspected for early and late measures: weakly restrictive verbs and their referents showed increased total reading time, while no effect was found on the duration of first fixation. We interpret these results as an integration of the subsequent context, thus indicating that verbs behave differently depending on their intrinsic properties. In line with current approaches to categorization, we think that there is scope for future research on reading and language processing to explore gradience within traditional lexical categories.