Our tally of social referencing did not include instances of the child turning to the parent/experimenter during a display change, or if the parent or the experimenter initiated

spoken communication to the child, both of which elicited the child’s attention. We hypothesized that if infants detected the perceptual anomaly in the picture of the impossible cube, it might elicit an increased frequency of vocalizations and/or social referencing to the parent accompanying the child during the study. Infants’ responses were analyzed using a repeated-measures 2 (Sex) × 2 (Order: Possible versus Impossible First) × 3 (Display) analysis of variance (ANOVA). Preliminary analyses revealed no reliable differences in the extent of reaching, social referencing, vocalizations, or mouthing behaviors based on sex or stimulus order,

Metformin mouse F(1, 10) = n.s., Proteasome inhibitor all p-values > .25, and no interactions, so these between-subjects factors were omitted from further analyses. Data points from the perceptual control displays (tree bark, gray patches, and brown lines) were collapsed into one within-subjects variable for comparison with the possible and impossible cube displays. In order to assure reliability of the experimenter’s judgments, an independent observer who was blind to the hypotheses also coded manual gestures offline for 100% of the final sample. Pearson correlations between the experimenter’s and the coder’s judgments indicated strong interrater reliability for all measures (manual gestures r = .90, p < .01; sequential gestures r = .92, p < .001; social referencing r = .89, p < .01; vocal utterances r = .80, Amino acid p < .01). All

tests of statistical significance used an alpha level of .05, and all t-tests were two-tailed. Results of a within-subjects ANOVA yielded a main effect of display, F(2, 26) = 8.76, p < .001, due to differences in mean quantity of categorical types of manual gestures across displays. Pairwise comparisons (with least squares differences [LSD]) revealed that the infants engaged in a greater number of different types of manual exploration toward the impossible cube relative to the possible cube display, t(13) = 2.74, p < .001, and the perceptual controls, t(13) = 4.25, p < .02, as shown in Figure 2a. The mean impossible preference score was .63, which differed significantly from chance, t(13) = 2.48, p < .03. Infants attempted an average of one additional different type of manual gesture toward the impossible cube display above that of the possible cube display and the perceptual controls. The pattern of increased manual exploration toward the impossible cube display was observed in nine of the 14 infants, with four infants responding equally to the two displays, and one with more reaching to the possible cube, Z = 2.13, p = .03.