Sleep Promotes the Extraction of Grammatical Rules
Figure 4
The effect of brain-state (wake, sleep) on classification performance.
A) In the wake group (pooled 15 min morning, 15 min evening, and 12 h wake groups) participants endorsed almost equal amounts of Grammatical (G) and Non-Grammatical (NG) test items. In the sleep group (pooled 12 h sleep and 24 h sleep groups), however, the difference in the endorsement between G and NG items was more than 0.17. Thus, sleep increased the effect of grammaticality on endorsement (significant interaction between brain state and grammaticality in contrast 1, Table 6). Both in the wake and in the sleep group, participants endorsed more High than Low ACS items. However, sleep did not amplify that effect (no interaction between brain state and ACS in contrast 1, Table 6). B) Sleep especially enhanced the effect the grammaticality status of Low ACS test items had on classification (post hoc T-test, P<0.001, Bonferroni corrected), but also in the High ACS items there is a trend that sleep increased the effect of grammaticality (post hoc T-test, P = 0.052 Bonferroni corrected). Sleep increased the effect of the grammaticality on endorsement more for Low than High ACS items (brain state×grammaticality×ACS 3-Way interaction, P = 0.047 in contrast 1, Table 6). C) The endorsement of each of the four categories of sequences in the Classification Set (Fig. 1D) for the wake and sleep groups. Endorsement is the proportion of sequences classified as Grammatical independent of the actual grammaticality status. Endorsement difference is the difference between those proportions for G and NG sequences, or High ACS and Low ACS sequences. Error bars denote standard errors of the mean.