Abstract
Event-related brain potentials elicited by lexically associated and unassociated word pairs embedded in normal or semantically anomalous sentences were recorded in order to compare the influences of lexical and sentential context. The design of the experiment was such that second words of associated pairs in anomalous sentences could be subject to lexical context alone, while the second words of unassociated pairs in normal sentences could draw on both types of context, while unassociated words in anomalous sentences were included as a control condition wherein no context effects were expected. N400 amplitude was reduced by both lexical and sentential contexts, and the onset latencies of the two effects were similar. The sentential context effect proved to be longer in duration, and exhibited greater variability across subjects. The amplitude of the purely sentential context effect was predictive of subsequent recognition accuracy for other words occurring in the same sentence. The amplitude of the lexical context effect was unrelated to subsequent recognition performance.