Abstract
Faces are frequently used to investigate ensemble processing of complex objects, because visual impressions of faces in a crowd can be highly relevant for social interactions. Most of these studies rely on relatively few stimuli created by morphing among a small number of faces. We asked how the repetition of objects over trials influences performance on ensemble judgments. We hypothesized that when provided such repetition, people can learn to process multiple stimuli as an ensemble, regardless of category or adaptive value. We measured participants’ ability to extract ensemble information from arrays of car images. On each trial, participants viewed two successive, briefly presented arrays of 6 cars and judged which array contained the more diverse set of cars. The more diverse of the two arrays comprised 6 different cars, and the less diverse array comprised 3 or 4 identical cars plus 3 or 2 different cars. The identical cars in the less diverse array were shown either at neighboring or scattered locations. Half of the participants viewed the same set of 6 cars throughout the experiment, and thus had opportunity to familiarize themselves with that ensemble of cars. The other participants viewed arrays of new cars on each trial, precluding stimulus-specific learning. All participants could perform the diversity judgment task, but those who viewed non-repeating cars performed better when identical cars were neighboring than scattered, suggesting reliance on local information. In contrast, the spatial distribution of identical cars had less influence for participants who viewed repeating cars, suggesting that repetition of cars throughout the task promoted learning of stimulus-specific, ensemble-relevant information. Whether or not stimulus-specific learning facilitates ensemble processing of objects for which people have experience of forming ensembles, such as faces, needs to be explored.
Acknowledgement: This work was supported by the NSF (SBE-0542013 and SMA-1640681) and by the Centennial Research Fund (Vanderbilt University).