Mazzarella, D. 1 , Trouche, E. 1 , Mercier, H. . 1, 2 & Noveck, I. 1
1 Institut des Sciences Cognitive 'Marc Jeannerod', Université de Lyon / CNRS
2 University of Neuchâtel
Recent behavioural studies in experimental pragmatics investigate the effect of politeness on the derivation of scalar inferences, e.g. 'Not all X-ed' from 'Some X-ed'. Bonnefon and colleagues (2009, 2011) claim that when the scalar utterance is face-threatening ('Some people hated your speech') (i) the scalar inference is blocked, (ii) the semantic interpretation of 'some' (at least some) is arrived at slowly and effortfully. This contradicts much evidence showing that the derivation of the scalar inference comes with extra cost (since Bott & Noveck, 2004). Drawing on the distinction between comprehension and acceptance of communicated information, we hypothesized that in face-threatening contexts scalar inferences are derived but less likely to be accepted as true, and that the extra processing effort is associated with rejecting the communicated content. We tested our hypothesis through a series of MTurk studies inspired by Bonnefon et al.'s task. Our main innovation is that we separated the presentation of the scalar utterance (comprehension stage) from participants' evaluation of it (epistemic assessment stage) so that Reaction Times to each part could be measured separately. In line with our predictions, we found that the latencies in the face-threatening condition were due to longer RT at the epistemic assessment stage.