*3.2. Audiovisual Task*

*Brain Sci.* **2021**, *11*, x FOR PEER REVIEW 6 of 12

Infants' looking times to the screen were recorded with an eye-tracker. We defined dynamic areas of interest (AOI) covering the background, the arm, the face, the eyes, and the mouth (Figure 1). For each trial, we calculated the proportion of looking time to the AOIs in comparison to the whole screen. Next, for each infant we averaged proportions across all trials. The total number of trials differed across infants, depending on how long they were interested in the task (between 1 and 8 blocks). A Wilcoxon-Mann-Whitney test revealed that groups did not differ in the number of completed trials (MTD = 3.91, range 2–6; MDS = 3.4, range 2–5; Z = −0.71, *p* = 0.47, r = 0.13). The looking patterns for the two groups of infants are depicted in Figure 3. **Figure 2.** Mean orientation latency (in seconds) in the visual attention task across the Down syndrome (DS) and typically developing (TD) groups. The values above the bars refer to the mean latency value for each group. Error bars represent 1 (+/−) standard error of mean.

**Figure 3.** Mean proportion of looking time in the audiovisual task across the Down syndrome (DS) and typically developing (TD) groups. Error bars represent 1 (+/−) standard error of mean. Significant differences are signaled: 0.05 = \*, 0.01 = \*\*, 0.001 = \*\*\*. **Figure 3.** Mean proportion of looking time in the audiovisual task across the Down syndrome (DS) and typically developing (TD) groups. Error bars represent 1 (+/−) standard error of mean. Significant differences are signaled: 0.05 = \*, 0.01 = \*\*, 0.001 = \*\*\*.

The looking pattern was analyzed for each group separately. First, we compared the 4 levels of AOIs (the eyes, the mouth, the arm, and the background) separately for each group. The results revealed that AOIs significantly differed in both groups (a Kruskal-Wallis test for the DS, H(3) = 17.7, *p* = 0.004, η2 = 0.61; a one-way-ANOVA for the TD The looking pattern was analyzed for each group separately. First, we compared the 4 levels of AOIs (the eyes, the mouth, the arm, and the background) separately for each group. The results revealed that AOIs significantly differed in both groups (a Kruskal-Wallis test for the DS, H(3) = 17.7, *p* = 0.004, η <sup>2</sup> = 0.61; a one-way-ANOVA for the TD F(3, 92) = 8.82, *p* < 0.001, η <sup>2</sup> = 0.22). Pairwise comparisons (Bonferroni controlled) showed that regarding the background, both groups looked longer at the background than the arm (both *p*s = 0.001, dDS = 1.3, dTD = 0.35), and longer to the background than the mouth (*p*TD = 0.0012, dTD = 1.8, *p*DS = 0.0017, dDS = 0.66). However, DS infants looked longer at the background than the eyes (*p* = 0.04, d = 0.78), but not TD infants (*p* = 0.86, d = 0.03). Regarding the arm, TD, but not DS infants, looked more at the eyes than the arm (*p*TD = 0.001, dTD = 0.38; *p*DS = 0.2, dDS = 0.47), whereas DS, but not the TD, looked more at the arm than the mouth (*p*TD = 0.86, dTD = 0.01; *p*DS = 0.039, dDS = 0.88). Finally, both groups looked longer at the eyes than the mouth (*p*TD= 0.001, dTD = 0.4; *p*DS = 0.035, dDS = 0.89). To further understand this complex looking pattern across the groups we reduced the number of AOIs, thus we compared the face (including the eyes and the mouth), the arm, and the background separately for each group. We observed that both groups looked more at the background than the arm (*p*TD = 0.01 dTD = 0.32; *p*DS = 0.0017, dDS = 1.25) and more at the face than the arm (*p*TD < 0.001, dTD = 1.02; *p*DS = 0.0017, dDS = 1.22). However, only TD, but not DS infants, looked more at the face than the background (*p*TD < 0.001, dTD = 0.71; *p*DS = 0.6, dDS = 0.18). To directly compare the two groups, we computed a linear-mixed analysis on proportion of

looking time with AOI (face, arm, background) and group (TD and DS) as fixed factors (with the interaction term), and by-subject as a random intercept. This analysis confirmed that the AOIs differed (F = 26.94, *p* < 0.001, η 2 <sup>p</sup> = 0.38, 95% CI = 0.22–0.51), and more importantly that there was an interaction between AOI and Group (F = 3.21, *p* = 0.04, η 2 <sup>p</sup> = 0.07, 95% CI = 0.01–0.18). Further pairwise comparisons (Bonferroni controlled) revealed that groups did not differ in their looking time to the arm (t = −0.03, *p* = 0.97, d = 0.06). However, results suggest a trend of TD looking more to the face than DS (MTD = 0.65, MDS = 0.48, t = 1.8, *p* = 0.074, d = 0.38), and a trend for TD looking less to the background than DS (MTD = 0.26, MDS = 0.42, t = −1.77, *p* = 0.079, d = 0.38).

Finally, considering that DS infants were slower in the visual orientation task, we tested whether individual latency in the visual orientation task modulated performance in the audiovisual task. To the previous mixed model analysis, we added the average latency for each subject as a fixed effect, while other parameters maintained the same. The results were similar as in the previous analysis, with a main effect of AOI (F = 26.64, *p* < 0.001, η 2 <sup>p</sup> = 0.38, 95% CI = 0.22–0.51) and an interaction between AOI and Group (F = 3.17, *p* = 0.04, η 2 <sup>p</sup> = 0.03, 95% CI = 0.01–0.18). The pairwise comparisons revealed the same pattern: TD and DS do not differ in their proportional looks to the arm (t = −0.02, *p* = 0.97, d= 0.006), while we observed a trend of TD looking more to the face than DS (t = 1.7, *p* = 0.09, d = 0.37), and TD looking less to the background than DS (t = −1.68, *p* = 0.09, d = 0.36). This suggests that even when average latency in the visual orientation task is taken into account the same pattern of findings across groups holds in the audiovisual task.
