Attention neglects a stare-in-the-crowd: Unanticipated consequences of prediction-error coding.

Area of Science:

• Cognitive Neuroscience
• Visual Attention
• Predictive Coding

Background:

• Direct gaze is a salient social stimulus, typically thought to capture attention strongly.
• Predictive Coding (PC) models suggest attention is drawn to prediction-violating stimuli.
• This implies PC models might predict attention away from direct gaze towards averted gaze.

Purpose of the Study:

• To investigate attentional biases towards direct versus averted gaze within a Predictive Coding framework.
• To determine if attentional bias aligns with the consensus view or PC model predictions.
• To explore the role of contextual factors, like gaze uniqueness, in modulating these biases.

Main Methods:

• Participants searched visual displays containing salient, 'odd-one-out' gazes.
• Eye-tracking measured rapid, initial saccades to assess attentional bias.
• Experimental conditions manipulated the distinctiveness or uniqueness of gaze directions.

Main Results:

• An attentional bias towards averted gaze was observed, consistent with PC models.
• This bias was evident specifically when gaze uniqueness was highlighted by the experimental conditions.
• No consistent bias towards direct gaze was found under these specific conditions.

Conclusions:

• Attentional responses to gaze direction are context-dependent.
• Predictive Coding principles can explain attentional biases, particularly when stimuli violate expectations (e.g., unique averted gaze).
• Task demands and the salience of prediction errors significantly influence how gaze direction affects perception and attention.