Differential Visual Behavior to Human and Humanoid Faces in Early Infancy.

do not differ reliably from each other. Taking each week separately, the simple effect of stimulus materials is significant at weeks two (p < .01), three (p < .05), five (p < .01), and six (p < .05) (Winer, 1962). The overall effect of least attention to mother held true for 16 of the 18 subjects. This proportion of subjects is significantly above chance ( ρ < .002) as deter mined by the binomial test (Siegel, 1956). Thus, the live human face of the mother, (who was also the most familiar stimulus of the experiment), was distinguished from the humanoid "faces" by two weeks of age. The main effect of week of age is significant (F = 17.66, df = 7/105, ρ < .005) and is shown in Figure 4. Within the main effect of weeks, all weekly mean attention scores in month two are significantly higher (p < .01) than all weekly means in month one as determined by the Newman Keuls procedure. Within the first month there were no significant mean differences; in month two, only the comparison between weeks five and eight was statistically significant (p<.05). This main effect of increased attention with age is what one might expect from the rapidly developing organism whose visual system is maturing and who is spending more time awake and in visual contact with the environment. The stimulus materials X week of age interaction falls short of the .05 level of significance (F = 1.70, df = 14/210, ρ < .10). However (Figure 3), the attention curves for manikin and abstract had a similar configuration in contrast with the more clearly linear curve for mother. The increase in attention to manikin and abstract from week four to week five and the related increase from month one to month two may mark a point of pro nounced change in the perceptual process. 4 Since all babies did not attend each week nor complete all stimuli within a session, 17% of the 1,728 data points were missing. Only 10% of the data points were missing using the first six stimuli of each session. Plots of the first six vs. second six stimuli showed virtually the same configuration of means. Since the first six stimuli were representative of the entire session and contained less missing data, the findings reported here were based on the first six stimulus conditions. Data for still and nodding conditions have been combined. Missing data points were estimated by obtaining the average of: a) the baby's overall mean for the given stimulus and; b) the mean of the baby's group for that stimulus at the week in which the datum was missing. 5 The Newman-Keuls procedure is a multiple comparison technique designed to minimize chance findings resulting from more than two statistical tests on the same data. They are two-tailed tests. This content downloaded from 157.55.39.35 on Sat, 03 Sep 2016 05:42:32 UTC All use subject to http://about.jstor.org/terms MERRILL-PALMER QUARTERLY