Anticipating Negation: The Do s and Don't s of Neg Raising

Absract A sentence like (i) can often be used interchangeably with (ii). I don't think it's raining. I think it's not raining. With particular matrix predicates (e.g., think , seem , be probable ) the negation from a complement clause may appear in the matrix clause, according to syntactic accounts, owing to a rule of Neg Raising, or according to semantic accounts, owing to a cancelable presupposition of the Excluded Middle (I believe p or I believe ¬ p ). Six written acceptability judgment studies approached the phenomena of Neg Raising ( NR ) as anticipation. A speaker who intends a negation may produce the negation early when the matrix verb permits the negation to be raised, especially if the polarity of the complement clause is at issue. Experiments 1 and 2, with a strict negative polarity item ( NPI ), showed higher ratings for sentences with Neg Raising (+ NR ) than for sentences without (− NR ), failing to support the complexity predictions of a syntactic account where + NR involves an extra rule or dependency. Experiment 3 compared + NR predicates with − NR predicates and showed a penalty for matrix negation in the latter. Experiments 4 and 5 showed that increasing the distance between the raised negation and the negated predicate reduced the + NR advantage, but they showed that this was only an effect of increased sentence length. Experiment 6 showed an increased advantage for Neg Raising in a response to a polar question about the truth of the complement sentence. The results do not support a syntactic account but rather some version of a semantic or pragmatic account of Neg Raising coupled with the assumption that speakers anticipate salient upcoming material, such as a negative.