It takes two to kiss, but does it take three to give a kiss? Categorization based on thematic roles

Language is characterised by broad and predictable mappings between meaning and syntactic form. Transitive sentences typically encode two-participant events while ditransitives typically encode three-participant events. Lightverb constructions, however, systematically violate these mappings; for example, some have ditransitive syntax (‘Romeo is giving Juliet a kiss’) but describe what appear to be agent[1]patient events (Romeo kissing Juliet). We used a conceptual sorting task to explore whether this non-canonical mapping influenced the interpretation of these sentences. Participants were trained to sort events by the number of thematic roles they encoded. After a training phase with only pictures, they sorted a mix of pictures and written sentences, including transitive agent-patient sentences, ditransitive source-theme-goal sentences and ditransitive light-verb constructions. Events described by light-verb constructions were most often grouped with agent-patient events but were sometimes grouped with source-theme-goal events. A control condition using the transitive/intransitive alternation for joint action verbs (e.g., ‘meet’) demonstrates that this is not attributable to misconstruing the task as syntactic sorting. We conclude that non-canonical mappings between meaning and form can affect event construal, but syntactic form does not solely determine the construal that is chosen.