Bottom-up and top-down processes in attention to emotional faces

Emotional faces convey important information on social chances as well as danger. Therefore, automatic attention allocation to emotional facial expressions is a frequently investigated phenomenon. Processing advantages of emotional over neutral stimuli have been shown using various experimental paradigms like Dot-Probe (Bocanegra, Huijding, & Zeelenberg, 2012), Eriksen-Flanker-Task (Fenske & Easwood, 2003) and Visual Search (Fox et al., 2000). Nevertheless, more and more evidence appears showing that these effects are not unconditional (i.e. Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg, & van IJzendoorn, 2007). Person characteristics, like anxiety, as well as task characteristics (Wirth & Wentura, 2017) modulate attentional effects. The talk includes own data from dot-probe studies (Puls & Rothermund, 2017), showing no attentional capture of emotional faces at all, as well as data from the Eriksen-Flanker-Task (Tannert & Rothermund, 2018) showing attentional effects of emotional faces being conditional on relevance of the affect dimension and necessity of flanker processing. One explanation for these and other results, reviewed in this talk, is a certain amount of top-down influences on attention even at this early stage of processing. Theories considering this possibility (e.g. contingent capture account¸ Folk & Remington, 1998) will be outlined and discussed in light of the present as well as previous findings.

Several theories of emotional attention claim that angry faces capture attention because they are highly relevant stimuli for the observer. However, in the dot-probe task – a paradigm commonly used to asses emotional attentional bias – usually only anxious participants show an attentional bias towards angry faces. We conducted three experiments to investigate social processing as a potential moderator for the occurrence of such a bias in non-anxious samples. In Experiment 1, participants performed a dot-probe task with two different target types. In the social target condition, participants had to categorise socially meaningful targets (schematic faces). In the non-social target condition, socially meaningless targets (scrambled schematic faces) were employed. Before the onset of the target display, two photographic face cues – one angry and one neutral – were presented for 100 ms. Participants only showed a significant bias towards angry face cues in the social target condition, but not in the non-social target condition. In Experiment 2, we increased the SOA between cue and target onset to 200 ms. Again, a significant bias towards angry face cues only occurred in the social target condition. In Experiment 3, we additionally observed the N2pc component of the ERP as a measure of shifts in spatial attention. N2pc components elicited by angry face cues were significantly larger in the social target condition than in the non-social target condition. These results suggest that social processing moderates the occurrence of attentional bias towards angry faces in non-anxious samples.

Many studies support the existence of a general bias for the detection of threat in humans. In particular, the threat-capture account claims that attentional resources are captured by threat-related stimuli, mainly based on behavioral evidence. However, these studies do not inform us about how the detection of such stimuli in the visual space influences the spatial orienting of attention. In fact, the capture of attention by a threatening item in the environment should automatically activate information about its spatial location to elicit a proper reaction, giving the localization of threat and an enhanced selection or suppression of a threatening item. Electrophysiological indexes, specifically the N2pc or the Pd component, are well suited for detecting this early attentional selection of threat signals. In this talk, I will show how these components provide interesting evidence of the threat-capture account. I will also discuss their limits, as well as the limits of the threat-capture account, specifically the impact of top-down processing on the capture of attention by threatening items.

Increasing evidence suggests that emotional information can be utilized in an “automatic” way in processes that are inherently related to the emotional meaning: For example, seeing an angry face can trigger subtle facial responses, prepare avoidance reactions, and influence deployment of attention, judgments and decisions in line with or as a reaction to the emotional meaning (e.g., Hess & Fischer, 2013; Stins et al., 2011). In order to provide contextual flexibility in various critical situations, we proposed that emotional information can be utilized fast, efficiently, and based on limited perceptual input also for novel, goal-directed processes. To test this assumption, we used emotional faces as central cues in a series of endogenous cueing experiments, the facial expression signaled the location of targets. Endogenous cuing emerged fast, based on valence and specific negative emotions (e.g., sadness versus fear), and even based on masked emotional information (Folyi, Rohr, & Wentura, 2019). In a further study, we tested whether this cueing effect emerges based on explicit knowledge of cue-target contingencies in line with the prevailing view of endogenous cueing; or, at least partly, based on implicit learning of these contingencies. Cue emotion was again informative to the target location, but participants did not receive any information about this relationship. Overall, there was no indication of cueing without contingency information, and majority of the participants could not report these contingencies. Our results suggest that attentional processes based on implicit probability learning do not explain the efficacy of cueing by emotional faces.