2.7. ERP Analyses

ERP components (target) were identified and quantified across Fz, Cz, Pz, and Oz midline sites. The following components were identified: N1 ERP peak was (M ± SD = 104.0 ± 5.0 ms), quantified as negative values as the baseline-to-peak difference in voltage for the most negative peak within the 90–140 ms window following face stimulus onset; N2 (217.7 ± 9.5 ms, window 170–310 ms); P3 (320 ± 12.4 ms, window 200–390 ms); N4 (382.2 ± 9.2 ms, window 330–500 ms). Finally, the N170 ERP wave was also examined at the lateral posterior-temporal sites T5 and T6 (closest to the occipito-temporal sites P7 and P8). This ERP component was peaking at 180.7 ± 13.1 ms and was measured within the 140–260 ms time window [35,36,58,63]. Peak values were first semi-automatically detected as local minima for negative waves (or maxima, for positive waves) and then, after visual inspection, the position of the peak changed manually if necessary. The N1, N2, and N170 peak values were then multiplied by -1 and expressed as positive values for our convenience. The ERP analysis included both correct and incorrect behavioral responses. The amplitude and latency of each ERP component were quantified by the highest peak value within the chosen latency window. ERP amplitude and latency were analyzed with repeated measures ANCOVAs using AQ scores as a covariate. For these analyses, Emotion (neutral, happy, sad), Condition (subliminal, supraliminal), Facial-Gender (female, male), and Electrode Location (Fz, Cz, Pz, Oz) as within-subject factors were used. For the N170 wave, Electrode Location was replaced with Hemisphere (T5, T6) factor. An alpha criterion level of 0.05 was used unless otherwise noted. Huynh-Feldt adjustments were used when the assumption of sphericity was violated [82].

To report effect size estimates, partial ɳ²_p values (see supplemental material for statistical details) were also calculated. Paired samples t-tests were performed to supplement the behavioral and ERP findings. To control for false-positive errors, significance levels obtained for ERP measures were corrected by applying the false discovery rate correction (FDR) method across all ERP amplitude and latency measures [81]. Only for graphical illustrations, and to understand the direction of changes of significant main and/or interaction effects involving AQ trait, we applied a separate median split on this personality measure (M = 14.9, Md = 14.5; Skewness = 0.278, Kurtosis = −0.984). Participants were considered as belonging to either group Hi-AQ (N = 25, M = 20.6, SD = 4.3, Range = 15–26) or Lo-AQ (N = 25, M = 9.2, SD = 2.8, Range = 3–14) when their AQ scores were above or below the median. None of the AQ scores fell on the median, therefore, none of the participants were excluded.