No scalar inferences under embedding

Bart Geurts and Nausicaa Pouscoulous

We present three experimental studies showing that localist theories of scalar inference are wrong and the Gricean theory is right. The first two experiments used an inference task to explore upper-bounding inferences associated with scalar expressions occurring in the scope of various operators. The main finding of these experiments was that scalar inferences plummeted in embedding conditions, though the drop was less steep with embedding under 'think' than in the other embedding conditions (i.e., 'must', 'all', and 'want'), where positive response rates fell by as much as 70% on average. One issue that the first two experiments gave rise to is that, in some cases, we still observed substantial rates of local scalar inferences. We hypothesised that this was due at least in part to the experimental paradigm, and the results of our third experiment confirm this diagnosis. Taken together, our results confirm Gricean and disconfirm localist predictions.

Click here for the full paper.