Processing light verb constructions

Henry gave Elsa a kiss give kiss give a kiss We test two approaches to light verb constructions: (1) joint predication in light verb constructions is stored as pre-specified, and their high frequency predicts less processing cost. (2) Joint predication in light verb constructions is built in real-time. The entailed extra-syntactic composition predicts greater cost. Results from a cross-modal lexical decision task show delayed, higher reaction times for light verb constructions, supporting (2), which is consistent with a linguistic architecture that has partly autonomous lexico-semantic storage and processing.