Comparing averaging limits for social cues over space and time.

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.