The role of emotion in learning trustworthiness from eye-gaze: Evidence from facial electromyography

Abstract

Gaze direction can be used to rapidly and reflexively lead or mislead others' attention as to the location of important stimuli. When perception of gaze direction is congruent with the location of a target, responses are faster compared to when incongruent. Faces that consistently gaze congruently are also judged more trustworthy than faces that consistently gaze incongruently. However, it's unclear how gaze-cues elicit changes in trust. We measured facial electromyography (EMG) during an identity-contingent gaze-cueing task to examine whether embodied emotional reactions to gaze-cues mediate trust learning. Gaze-cueing effects were found to be equivalent regardless of whether participants showed learning of trust in the expected direction or did not. In contrast, we found distinctly different patterns of EMG activity in these two populations. In a further experiment we showed the learning effects were specific to viewing faces, as no changes in liking were detected when viewing arrows that evoked similar attentional orienting responses. These findings implicate embodied emotion in learning trust from identity-contingent gaze-cueing, possibly due to the social value of shared attention or deception rather than domain-general attentional orienting.

Keywords: Emotion; Face evaluation; Facial EMG; Gaze-cueing; Trustworthiness.

Figure 1.

Schematic illustration of trial procedure for rating and cueing trials. On rating trials before and after cueing, participants observed each face for 1000 ms after which a visual analog rating scale appeared requiring participants to click the point on the scale that represented how trustworthy they judged the face to be. During cueing trials, participants saw a fixation cross for 1500 ms, followed by a face looking directly for 1500 ms after which it changed its gaze direction and remained for 500 ms when an object appeared to the left- or right-hand side of the face and disappeared when the participant responded which also triggered the face to look back directly at the participant for another 2000 ms. Not drawn to scale. Faces reprinted with permission from the MacArthur Network.

Figure 2.

Mean trustworthiness ratings by validity and face gender, given before (left) and after (middle) cueing and the change in ratings (right), computed by subtracting beginning from end ratings. Error bars show ±1 standard error of the mean.

Figure 3.

Mean reaction times on congruent and incongruent trials across blocks within participants who did (left panel) and did not show trust effects (right panel). Error bars show +/-1 standard error of the mean.

Figure 4.

Mean stimulus-locked corrugator activity on congruent (solid line) and incongruent trials (dashed line) for trust effect (left panels) and no-trust effect (right panels) participants across trial periods 2, 3, 4, and 5 (rows). EMG units on the y-axis represent the ratio of activity relative to baseline (fixation). Error bars show ±1 standard error of the mean.

Figure 5.

Mean stimulus-locked zygomaticus activity on congruent (solid line) and incongruent trials (dashed line) for trust effect (left panels) and no-trust effect participants (right panels) across trial periods 2, 3,4, and 5 (rows). EMG units on the y-axis represent the ratio of activity relative to baseline (fixation). Error bars show ±1 standard error of the mean.

Figure 6.

Mean corrugator activity across blocks for trust effect participants (left panels) and no-trust effect participants (right panels) in trial period 2 (top), trial period 3 (middle) and trial period 4 (bottom). Dashed lines represent incongruent and solid lines congruent. EMG units on the y-axis represent the ratio of activity relative to baseline (fixation). Error bars show ±1 standard error of the mean.

Figure 7.

The 16 arrows in their neutral state used in experiment 2.

Figure 8.

Trial procedure on rating and cueing trials for experiment 2 using arrows. On rating trials before and after cueing, participants observed each arrow/shape for 1000 ms after which a visual analog rating scale appeared requiring participants to click the point on the scale that represented how much they liked the arrow/shape. During cueing trials, participants saw a fixation cross for 1500 ms, followed by an arrow/shape in its neutral state for 1500 ms, after which it changed shape to point to the left or right for 500 ms when an object appeared to the left- or right-hand side and disappeared when the participant responded. This also triggered the arrow to go back to its neutral state for another 2000 ms.

Figure 9.

Liking ratings of congruent and incongruent arrows before (left panel) and after (middle panel) cueing and the change in ratings computed by subtracting beginning from end ratings (right panel). Error bars show +/-1 standard error of the mean.

Figure 10.

Mean arrow-cueing reaction times on congruent and incongruent trials across blocks. Error bars show +/-1 standard error of the mean.

Figure 11.

Mean change in trust ratings (end-beginning) for congruent and incongruent conditions in trust effect (left panels) and no-trust effect participants (right panels). Error bars show +/-1 standard error of the mean.