To predict or not to predict: The role of context constraint and truth-value in negation processing

Studies on negation processing frequently report a polarity-by-truth interaction: False affirmative sentences usually show longer response times and larger N400 amplitudes compared to true affirmative sentences, whereas for negative sentences the effect of truth-value is typically reversed. This interaction has repeatedly been linked to factors such as lexical associations, predictability, or to the need of constructing two subsequent mental representations during the comprehension of negative sentences. In a series of ERP experiments using a picture-sentence verification paradigm, we investigated how sentence polarity, truth-value and predictability interact during sentence processing. Predictability was manipulated by varying the number of alternative sentence continuations provided by the context, similarly for both sentence polarities. For both affirmative and negative sentences, true sentences elicited reduced N400 amplitudes in strongly constraining contexts-where a specific continuation was highly predictable-compared to weakly constraining contexts, where no clear prediction could be made. For false sentences, the effect of context was reversed for both sentence polarities. Crucially, the effect of Truth was dependent on predictability rather than sentence polarity: Both affirmative and negative sentences showed the same direction of the effect of Truth, namely, larger N400s for false rather than true sentences in the strongly constraining context, and the opposite pattern in the weakly constraining context, although the size of these effects differed across the two polarities. In addition, we observe a long-lasting positivity effect for negation, in both context conditions, for both truth-values and across all five experiments. We interpret this effect as reflecting inhibitory mechanisms recruited during negation processing.