There is increasing evidence that readers make predictions about upcoming words during sentence comprehension. However, little research has been conducted to understand if this process is automatic (i.e., if readers always activate a set of associated concepts when there is a sufficiently constraining context) or strategic (i.e., readers only anticipate the expected candidates when it is beneficial for sentence processing). The current study examined whether lexical predictions are modulated by word expectancy and prediction validity of a prior set of sentences. In a one-word moving-window self-paced reading task, 100 participants initially read a set of 120 filler sentences to establish an environment with either high or low predictive validity. Then, both groups read 60 experimental sentences with critical words that were either expected or unexpected. The response times following the expected critical words were faster than after reading the unexpected critical words. This effect was significant for participants who had previously read the highly predictive filler sentences, whereas it was not observed for the group with low predictive validity fillers. These findings suggest that the facilitation of lexical predictions in sentence context may depend on the recent success or failure of anticipatory processing. Readers who have experienced many prediction failures may strategically limit their anticipatory processing. This study adds to our understanding of the benefits and costs of lexical predictions and the processes involved in sentence comprehension.

--

sentence comprehension,lexical prediction,strategic processing,context predictive validity,self- paced reading