Complement coercion in Polish and the role of selectional restrictions revealed in a self-paced reading study

Abstract

So-called ‘complement coercion’ (‘begin a book’), understood as a combinatorial conflict, is mainly analysed as a repair operation in composition. Experimental data has shown that there is an extra cognitive effort in the processing of event-selecting verbs with entitydenoting arguments. These results support the formal analysis of ‘complement coercion’ as an enriched form of semantic composition (Pustejovsky, 1995; Egg, 2003; de Swart, 2011; Asher, 2015). Recently, an alternative view has been proposed by Piñango and Deo (2015) arguing that there is no mismatch between the verb and its complement, but an ambiguity resulting from the different dimensions along which an aspectual verb in composition with its argument can be interpreted (e.g. temporal, spatial, etc.). This approach has been supported by experiments showing that aspectual verbs like begin incur greater processing cost in coercing contexts than psychological verbs like enjoy (Lai et al., 2014). We designed a self-paced reading experiment to compare the different predictions that the two approaches make for the processing of matching and mismatching verb-noun combinations. Our results are compatible with enriched composition, but not with dimensional ambiguity. We find facilitation in conditions where the selectional restrictions of the verbs are satisfied (‘begin a fight’, ‘see a book’), and longer processing times in the conditions ‘begin a book’, ‘see a fight’, which does not support the dimensional ambiguity approach. Our experiment thus provides evidence that selectional restrictions are a fundamental property of a predicate and that they need to be understood as a graded continuum of combinatorial preferences, as also argued in Spalek (2014).