Page 53 - Emotions through the eyes of our closest living relatives- Exploring attentional and behavioral mechanisms

Emotional attention is modulated by familiarity
Data filtering
We filtered reaction times (RTs) with extreme values (i.e., RT < 250 and RT > 5000
ms). As our dataset contains a large age range, we also filtered reaction times per
age category (0-5, 6-10, 11-15, ..., 56-60, 61-85) and calculated the median absolute 2 deviation for reaction times per age category. Finally, we used the following filter:
[RT < (Median RT + (2.5 * Median Absolute Deviation))]. After applying this filter,
five participants had less than 5% of data points left and were thus excluded from
further analysis, leading to a final N of 444. In total, we excluded 15.25% of the data
for further analysis.
Statistical analyses
Data were analyzed using a generalized linear mixed model in R studio (v1.4.1106, glmmTMB package, α =.05 (Brooks et al., 2017; R Core Team, 2020)). Experimental trials (40) were nested within participants (ID, 444). We used reaction time (ms) as the dependent variable, random intercepts for all IDs (subjects), and used Congruency, Familiarity (both sum coded), and their interaction terms as fixed factors. Moreover, we used the AIC statistic to determine which distribution (gaussian vs. gamma) fit our data best (Lo & Andrews, 2015). Model assumptions were checked by visually inspecting QQ plots and the residuals plotted against fitted values.
Results
A model with a gamma distribution appeared to fit our data best (AICnormal = 200452, AICgamma = 198980). Testing whether humans have an attentional bias towards emotions of familiar and unfamiliar others, we found a significant interaction effect between Congruency and Familiarity (c2(1) = 3.47, p = .047, Figure 4, bottom left). Planned comparisons showed that participants were significantly faster when a probe replaced an emotional stimulus (M = 563.78, SD = 116.89) versus a probe replacing a neutral stimulus (M = 568.89, SD = 121.66) in the Familiar condition (b = -.01, SE = .00, t(16943) = - 2.72, p = .007) but not in the Unfamiliar condition (b = .00, SE = .00, t(16943) = .08, p = .936. See Table S11.1 for further model output. Also see Table S11.2 for an exploratory analysis of sex and familiarity effects on emotion bias). In short, while our control experiment in Experiment 2 showed that humans generally have a bias towards emotions, Experiment 3 shows that this bias is modulated by familiarity such that humans mainly have a bias towards emotional cues from familiar individuals.
51