A 'no' with a trace of 'yes': A mouse-tracking study of negative sentence processing

There is strong evidence that comprehenders can parse sentences in an incremental fashion. However, when the sentence contains a negation, the evidence is less clear. Previous work has shown that increasing the pragmatic felicity of a negative sentence reduces or eliminates any processing overhead relative to affirmative sentences. However, in previous work felicity has gone hand-in-hand with the predictability of critical material in a sentence. In three experiments reported here, we presented equally felicitous sentences with critical material of varying predictability (operationalised as the number of possible completions) to test whether this might be a critical factor determining the ease with which partial sentences containing a negation are interpreted. Participants completed a truth-value judgement task (Experiment 1) or a sentence completion task (Experiments 2 and 3) after viewing a visual environment that provided the context for a test sentence, which could differ in truth value (in Experiment 1 only), polarity (affirmative or negative), and number of possible completions (one, two, or three). In all three experiments, we recorded response times and accuracy, but also response dynamics via participants' computer mouse trajectories, allowing us to test specific hypotheses about the time course of comprehension. Across all experiments, in conditions with one or two possible targets, we observed consistent detrimental effects of negative polarity, suggesting that the difficulty in processing negation cannot be reduced to effects relating to predictability or pragmatic felicity. We discuss this finding in relation to incremental and two-stage models of processing and outline a new account of the processing difficulty arising from negation in terms of a conflict between what is locally activated on the basis of individual words and phrases and the global meaning of a negative sentence.