Stimulus and observer characteristics jointly determine the relevance of threatening facial expressions and their interaction with attention

Abstract

Most emotional stimuli, including facial expressions, are judged not only by their intrinsic characteristics, but also by the context in which they appear. Gaze direction, for example, modifies the salience of explicitly presented facial displays. Yet, it is unknown whether this effect persists when facial displays are no longer task-relevant. Here, we first varied the salience of fearful, angry or neutral displays using gaze direction, while participants performed a gender (attended faces) or a scene discrimination task (unattended faces). Best performance occurred when faces were unattended and emotional expressions were highly salient (direct anger and averted fear), suggesting that these combinations are sufficiently important to capture attention and enhance visual processing. In a second experiment, we transiently changed participants’ individual characteristics by instructing them to hold either expansive or constrictive postures. Best performance occurred for direct anger and averted fear following expansive and constrictive postures, respectively, demonstrating that stimulus and observer characteristics jointly determine the attribution of relevance of threatening facial expressions and their interaction with attention.

Appendix

Experiment 1: supplementary results

The data were cleaned so that only responses with a reaction time superior to 200 ms were included in analyses and for reaction time data analyses, only correct responses were included.

Analyses on reaction times

Reaction times collapsed across tasks reached 620.14 ± 24.97 ms (SEM). The repeated-measures ANOVA across tasks revealed a main effect of task (F(1,39) = 133.76, p < .001, $\eta _{p}^{2}$ = 0.745), indicating that participants were slower during the scene task (675.49 ms) as compared to gender task (564.80 ms). The ANOVA also showed a main effect of emotion (F(1,39) = 5.399, p = .006, $\eta _{p}^{2}$ = 0.122). Planned comparisons of this effect found that participants were significantly faster in the presence of fearful, as opposed to angry faces (t(40) = 2.942, p = .005, d = 0.16) and for fearful as opposed to neutral faces (t(40) = 2.616, p = .012, d = 0.15). Importantly, the interaction of interest between task, emotion and gaze was significant (F(2,78) = 6.371, p = .003, $\eta _{p}^{2}$ = 0.140), driven by a significant emotion × gaze interaction in the scene task (F(2,78) = 6.130, p = .003, $\eta _{p}^{2}$ = 0.136) in which participants were significantly faster at discriminating scenes in the presence of a fearful face with an averted, as opposed to a direct gaze (t(40) = 3.630, p = .001, d = 0.31). However, the difference in reaction times between direct and averted anger conditions was not significant (t(40) = 1.024, p = .312, d = 0.08) (see Table 1).

Experiment 2: supplementary results

Analyses on reaction times

The data were cleaned so that only responses with a reaction time superior to 200 ms were included in analyses and only correct responses were included. Reaction times collapsed across sessions reached 628.68 ± 24.77 ms (SEM).

To account for the within-subject design, we first ran a repeated-measures ANOVA with both order of pose and sex of subject (expansive–constrictive, constrictive–expansive) as between-subject factors, and pose (expansive, constrictive), task (gender, scene), emotion (neutral, fear, anger) and gaze (direct, averted) as within-subjects factors. Significant interactions between order of pose × sex × pose × emotion × gaze (F(2,80) = 3.874, p = .025, $\eta _{p}^{2}$ = 0.088), order of pose × pose × task (F(1,40) = 14.500, p < .001, $\eta _{p}^{2}$ = 0.266), order of pose × pose (F(1,40) = 31.643, p < .001, $\eta _{p}^{2}$ = 0.442) suggested that the impact of pose may be different in Session 1 and Session 2, akin to accuracy results. We therefore analyzed Session 1 and 2 separately, by running two independent repeated measures ANOVAs for each session, with both posture and sex of subject as between-subject factors and task (gender, scene), emotion (neutral, fear, anger) and gaze (direct, averted) as within-subjects factors.

In session 1, neither the interaction between pose × task × emotion × gaze (F(2,80) = 0.745, p = .478, $\eta _{p}^{2}$ = 0.018) nor the one between task × emotion × gaze (F(2,80) = 0.845, p = .433, $\eta _{p}^{2}$ = 0.021) were significant (see Table 2). Similarly, in Session 2, neither the interaction between pose × task × emotion × gaze (F(2,80) = 0.096, p = .909, $\eta _{p}^{2}$ = 0.002) nor the one between task × emotion × gaze (F(2,80) = 0.249, p = .707, $\eta _{p}^{2}$ = 0.009) were significant (see Table 3).

In both sessions, the ANOVA did reveal a main effect of task (Session 1: F(1,40) = 86.502, p < .001, $\eta _{p}^{2}$ = 0.684; Session 2: F(1,40) = 80.421, p < .001, $\eta _{p}^{2}$ = 0.668), indicating that participants were significantly faster for the gender task (Session 1: 585.07 ± 24.09 ms (SEM); Session 2: 536.27 ± 21.94 ms (SEM)) as opposed to the Scene task (Session 1: 731.52 ± 31.16 ms (SEM); Session 2: 656.64 ± 28.01 ms (SEM)). In summary, reaction time analyses for Experiment 2 demonstrate no impact of pose.