The direction of social attention from groups provides stronger cueing than from an individual. It has previously been shown that both basic visual features such as size or orientation and more complex features such as face emotion and identity can be averaged across multiple elements. Here we used an equivalent noise procedure to compare observers’ ability to average social cues with their averaging of a non-social cue. Estimates of observers’ internal noise (uncertainty associated with processing any individual) and sample-size (the effective number of gaze-directions pooled) were derived by fitting equivalent noise functions to discrimination thresholds. We also used reverse correlation analysis to estimate the spatial distribution of samples used by participants. Averaging of head-rotation and cone-rotation was less noisy and more efficient than averaging of gaze direction, though presenting only the eye region of faces at a larger size improved gaze averaging performance. The reverse correlation analysis revealed greater sampling areas for head rotation compared to gaze. We attribute these differences in averaging between gaze and head cues to poorer visual processing of faces in the periphery. The similarity between head and cone averaging are examined within the framework of a general mechanism for averaging of object rotation.

When looking at someone, we combine information about their head orientation and eye deviation to judge their direction of gaze. What remains unknown, however, is how these cues combine when we are not looking directly at the person, but rather are using our peripheral vision. Given that peripheral vision helps direct future attention, understanding how we perceive other people's gaze is key to determining their future actions. To examine this, we asked participants to categorize gaze direction in faces whose heads were turned in different directions, and which were viewed using either central or peripheral vision. We report that the weight given to head orientation increases in the periphery, in which forward-facing heads were categorized as "direct" over a wider range of eye deviations than when viewed centrally. When peripheral heads were turned, the number of "direct" responses fell for all gaze deviations with no consistent shift in left-right responses toward the head rotation. For centrally presented heads, head orientation typically repulsed the perceived direction of gaze, and our finding of no consistent shift in responses indicates that such effects are reduced in the periphery. This is not simply the result of poorer spatial resolution in the periphery-other influences, such as crowding and priors for gaze or head direction, may play a role. (PsycINFO Database Record

Observers are able to extract summary statistics from groups of faces, such as their mean emotion or identity. This can be done for faces presented simultaneously and also from sequences of faces presented at a fixed location. Equivalent noise analysis, which estimates an observer's internal noise (the uncertainty in judging a single element) and effective sample size (ESS; the effective number of elements being used to judge the average), reveals what limits an observer's averaging performance. It has recently been shown that observers have lower ESSs and higher internal noise for judging the mean gaze direction of a group of spatially distributed faces compared to the mean head direction of the same faces. In this study, we use the equivalent noise technique to compare limits on these two cues to social attention under two presentation conditions: spatially distributed and sequentially presented. We find that the differences in ESS are replicated in spatial arrays but disappear when both cue types are averaged over time, suggesting that limited peripheral gaze perception prevents accurate averaging performance. Correlation analysis across participants revealed generic limits for internal noise that may act across stimulus and presentation types, but no clear shared limits for ESS. This result supports the idea of some shared neural mechanisms b in early stages of visual processing.

No abstract