Purchase this article with an account.
Kelly Chang, Allison Yamanashi Leib, David Whitney; Training Ensemble Perception. Journal of Vision 2016;16(12):52. doi: 10.1167/16.12.52.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
When we glance at a group of people, we can easily discriminate the emotional tenor of the crowd via ensemble perception—a mechanism that allows us to extract gist or summary statistical characteristics from similar stimuli (Haberman & Whitney, 2009). However, some individuals are much more adept at this task than others. Here, we investigated whether it is possible to train participants to improve their sensitivity to overall crowd characteristics. To investigate perceptual learning of summary statistical perception, we trained observers in five sessions of 200 trials, totaling 1000 trials. In each trial, participants viewed a crowd of distinct but similar faces that, collectively, had a randomly determined average expression. Observers adjusted a test face to match the average emotional expression of the crowd. After each response, participants received feedback on their performance, and were awarded points based on accuracy. Across all training sessions, we included a measure to evaluate how many faces in the display participants integrated into their ensemble percept. Additionally, we included a measure to evaluate participants' improvement in single face discrimination. Sensitivity to average crowd expression significantly improved over the five sessions of training. This was accompanied by enhanced integration of information: Participants' accuracy improved as they integrated more and more faces from across the display. Control measures confirmed that participants' improvement cannot be explained by heightened single exemplar recognition or motor improvements associated with the response method. Our results demonstrate that summary statistical perception can be trained and improved. This could yield practical benefits. Individuals who routinely evaluate crowds (e.g. crowd security) can potentially enhance their performance with ensemble coding training. More importantly, these results reveal how ensemble perception improves—through increased integration of multiple exemplars.
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only