All child participants passed the selection measures. The three responses, ‘small’, ‘big’ and ‘huge strawberry’ are coded as response 1, 2 and 3. The adults invariably produced the 3-, 2- and 1-response for the optimal, underinformative and false utterances respectively. The results from the child group are presented in Table
1. A series of between-group comparisons using Mann–Whitney U tests for each cell reveal that children did BEZ235 research buy not perform significantly different than adults in any condition (all U < 2.1, p > .05). Within the child group, there were significant differences in the responses to every type of utterance (optimal, underinformative, false) both for both scalar and non-scalar expressions (all six Friedman’s ANOVA χ2(2) > 20.45, p < .001). The preferred responses in the false, underinformative and optimal conditions were 1, 2 and 3 respectively for both expressions (all 12 Wilcoxon Signed Ranks tests W > 3.1, p < .001, r > .73). There was no significant difference between the preferred responses for scalar and non-scalar expressions given the same utterance CB-839 ic50 type (all three W < 1.3, p > .1). Critically, 2-responses were more frequent in the underinformative than in the false condition, but less frequent than in the optimal condition; 3-responses were more frequent in the optimal than
in the other two conditions; and 1-responses were more frequent in the false than in the other two conditions (all W > 3.3; p < .001, r > .77). Thus, at the group level, children were sensitive to informativeness (rating it lower than optimal) but also tolerant (rating it higher than false). Furthermore, an
analysis of individual performance reveals that 16 out of 18 children consistently gave the middle reward to the underinformative utterances (at least 5 out of 6 cases for each expression), with the remaining two children giving underinformative utterances the lowest reward in at least four cases for each expression. Moreover, the children consistently awarded the top reward to the optimal condition and consistently gave the lowest reward to the false condition for each expression (with the exception of one child who did not consistently award the top reward to the optimal medroxyprogesterone condition for scalar expressions). Thus, given a ternary judgment task, each and every individual child participant revealed consistent sensitivity to underinformativeness (lower reward than optimal) and 16 out of 18 also revealed tolerance (higher reward than false). Every adult participant demonstrated both sensitivity to informativeness and tolerance of pragmatic infelicity. This has implications for the interpretation of experiment 1, where the majority of children consistently accepted underinformative utterances (13/20 and 12/20 children for scalars and non-scalars respectively).