Online syntactic storage costs in sentence comprehension

This paper presents three self-paced, word-by-word reading experiments that test for the existence of on-line syntactic storage/expectation costs in English. To investigate this issue, we compared reading times for sentence regions in which syntactic expectation costs varied, keeping other factors constant. Experiment 1 manipulated the number of verbs needed to form a grammatical sentence. It was observed that in the critical region, people read the condition in which zero verbs were predicted fastest, followed by the conditions in which one verb was predicted, with the condition in which two verbs were predicted slowest. Experiments 2 and 3 investigated whether incomplete filler-gap dependencies also incur storage costs. It was observed that people read the critical region in which a wh-filler is pending slower than if no such wh-filler is pending. The results of all three experiments demonstrate the role of online syntactic storage costs in sentence comprehension.