Gender discrimination task: eye movements
Mean results for each group are described in Table 2.
Whole face area—gaze maintenance. In the emotion × intensity × group ANOVA analyzing gaze maintenance, a significant between-group difference was found irrespective of facial expression and intensity, driven by subjects on citalopram maintaining their gaze over the faces for the whole trial duration more than subjects on placebo (main effect of group: F(1,31)=11.664, p=0.002; Figure 2).
Whole face area—scanpath length. In the emotion × intensity × group ANOVA analyzing the scanpath length, all subjects showed larger degree of exploration of faces with medium emotional intensity (F(2,62)=6.024, p=0.004). A significant two-way interaction was found between group and intensity of facial expression (F(2,62)=3.651, p=0.032), such that within the placebo but not within the citalopram-treated group participants showed larger degree of spatial exploration of medium compared with low and high intensity faces (placebo: F(2,62)=6.372, p=0.005; drug: F(2,62)=2.038, p=0.149). A significant three-way interaction was also present between group, emotion, and intensity of facial expression (F(2,62)=3.692, p=0.031, (Cohen's d=0.64)), indicating that individuals on citalopram tended to spatially explore the fearful faces of medium intensity less compared with individuals on placebo (F(1,31)=3.255, p=0.083) (Figure 3).
Whole face area—scanning time. In the emotion × intensity × group ANOVA analyzing the amount of time spent scanning faces, a significant between-group difference was found irrespective of facial expression and intensity, driven by subjects on citalopram scanning all faces longer compared with subjects on placebo (F(1,31)=9.867, p=0.004) (Supplementary Figure S1).
Eyes vs mouth—gaze maintenance. In the emotion × intensity × ROI × group ANOVA analyzing the proportion of gaze maintenance over different areas of the facial expressions, a significant main effect of group was found driven by subjects on citalopram maintaining their gaze for the whole trial duration over both eyes and mouth areas of all faces more than subjects on placebo (F(1,31)=6.378 p=0.017).
Correlations with behavioral performance and affect ratings. A significant positive correlation was observed between gender discrimination accuracy for fearful faces and scanning time (R=0.374, p=0.027). Instead, a negative correlation was found between gender discrimination accuracy for fearful faces and scanpath length (R=−0.366, p=0.031). No correlations were found with affect ratings.
Correlations with emotion recognition. A significant negative correlation was found between the intensity threshold needed to correctly recognize happy faces and gaze maintenance over the whole area of happy facial expression (R=−0.358, p=0.044) and between the number of misclassifications of happy faces and both gaze maintenance and scanning time over the whole area of happy facial expression (R=−0.518, p=0.002; R=−0.473, p=0.006). Given the significant between-group difference in facial expression recognition, driven by the citalopram-treated group performing better on positive faces, an exploratory regression analysis was also run to test whether emotion recognition was predicted by the pattern of ocular exploration of the faces. A significant predictive model was found with recognition threshold of positive faces as the dependent variable and group, ocular scanning time over happy faces, and the interaction between group and scanning time as independent variables (R=0.503, dF=(3,31), F=3.157, p=0.040). Scanning time over happy faces and the interaction between group and scanning time over happy faces were significant predictors of positive emotion recognition performance (β=−1.366, p=0.018 and β=1.181, p=0.021; Supplementary Figure S2).