(A) Table 1A. Subject information. All participants were asked to report their subjective level of hunger (from 1 = ‘not hungry at all’ to 8 = ‘very hungry’) and vigilance (from 1 = ‘very vigilant, not sleepy at all’ to 7 = ‘very sleepy, taking great efforts to keep awake’) before and after the verbal stimulation. Parentheses contain standard deviations. p-Values are calculated using two-tailed Student T test for comparing the difference in mean of each variable between the sleep and wake groups. Parentheses contain standard deviations. (B) Table 1B. Snack items included in the study together with their English translations. Snack items were selected based on a pilot experiment in which 49 subjects were recruited to assess the familiarity, valence, and subjective arousal (Self-Assessment-Manikin scale) of a pool of candidate snacks. Based on those ratings, we selected 60 items with median familiarity (mean ±SD: 3.51 ± 0.84), positive valence (mean ±SD: 5.08 ± 1.26), and median arousal level (mean ±SD: 4.65 ± 1.33) (associated with Figure 1). (C) Table 1C. Linear regressions on the effect of verbal cueing for sleep and wake groups after controlling for differences in age, gender, BMI, as well as self-reported familiarity and differences in vigilance and hunger before and after the cueing session. The dependent variable in the first regression is equal to the average difference of ΔWTP between cued and uncued items for each subject. The dependent variable in the second regression is equal to the likelihood of choosing cued over uncued item in the binary decision task for each subject. *p < 0.05; **p < 0.01; ***p < 0.001, two-tailed (associated with Figure 2). (D) Table 1D. Durations of sleep stages (in minutes) for sleep group subjects with or without ERP. There is no significant difference in the duration of sleep stages. Parentheses contain standard errors. p-Values were calculated using two-tailed Student T test for between-group comparisons (associated with Figure 4). (E) Table 1E. Effects of cueing on preferences and choices in sleep group subjects with or without ERP. There is no significant difference in the effect of verbal stimulation on either the ΔWTP or choice behavior in the binary decision task. Parentheses contain standard errors. p-Values are calculated using two-tailed Student T test for between-group comparisons (associated with Figure 4). (F) Table 1F. Average EEG amplitudes of positive/negative components of the K-complex-like evoked responses (KCs) at each electrode. We averaged the EEG amplitudes measured over the interval from 200 to 600 ms (for KC+) and from 600 to 1200 ms (for KC-) for each subject at each electrode. We found significant KC+/KC- at the frontal electrodes F3 and F4, and the central electrodes C3 and C4 (Student T-test, N = 23, all p values are Bonferroni corrected). For electrodes with significant KCs (F3, F4, C3, and C4), repeated measures ANOVA showed no significant difference among these four electrodes in both KC+ (F3,84 = 0.528, p = 0.664) and in KC- (F3,84 = 0.528, p = 0.664). *p < 0.05; **p < 0.01; ***p < 0.001, two-tailed (associated with Figure 4A). (G) Table 1G. Pearson correlations between ΔWTP and averaged cue-induced power for each frequency band at either subject- or item-levels. For across subject analysis, we examined the correlation between the ΔWTP and the power change of cue-induced band that were averaged over all cued items within each subject. For across item analysis, we examined the correlation between the ΔWTP and the power change of cue-induced band that were averaged over all subjects given the same cued item. All p-values are Bonferroni corrected. *p < 0.05; **p < 0.01; ***p < 0.001, two-tailed (associated with Figure 4B–D). (H) Table 1H. Priors used in the hierarchical Bayesian model estimation for LBA. We performed model estimation under the assumption that individual parameters are drawn from distributions for the sleep and wake group separately, with group level parameters sampled from joint prior distributions (associated with Figure 4—figure supplements 1–4). (I) Table 1I. Percentiles of RT distributions in panel B-C. These data show that a high drift rate bias (therefore high reactivation during N2) is associated with a more prominent change in the tail of the RT distribution and a smaller change in the leading ledge (associated with Figure 4—figure supplements 1–4).