Purchase this article with an account.
Jiaying Zhao, Liat Goldfarb, Nicholas B. Turk-Browne; When numbers and statistics collide: Competition between numerosity perception and statistical learning. Journal of Vision 2013;13(9):1087. doi: 10.1167/13.9.1087.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The visual system can extract statistical information from a single experience (summary perception) and across multiple experiences (statistical learning). We recently discovered that summary perception and statistical learning cause mutual interference: extracting the average orientation of a set of objects impeded statistical learning about spatial regularities, and the presence of regularities to be learned reduced the accuracy of average orientation estimates (Zhao et al., 2011). Here we extend these findings to the related domain of numerosity perception — our ability to rapidly estimate the number of objects currently visible in the environment. Across three between-subjects experiments, observers were familiarized with arrays of varying numerosity constructed from four spatial pairs and two singles of colored circles. In Experiment 1, we demonstrate that numerosity estimation prevents statistical learning of the color pairs, which was robust after passive viewing and in a control dual task. In Experiment 2, we show that this interference is bidirectional by observing improved numerosity estimation accuracy for arrays that did not contain spatial regularities over time. In Experiment 3, we clarify that reduced numerosity accuracy for arrays with regularities results from learning of the regularities, not the presence of regularities per se. These findings are notable for at least three reasons. First, they replicate the surprising pattern of mutual interference that we observed previously for summary perception of average orientation. Second, these results generalize our previous findings of interference by statistical learning to a different kind of summary statistic, and one that has been studied extensively and is thought to be computed automatically. Third, numerosity perception represents an even stronger test of the theorized overlap between summary perception and statistical learning: Unlike the estimation of average feature values, estimating numerosity does not require representing the identities of objects beyond their mere existence as discrete entities in space.
Meeting abstract presented at VSS 2013
This PDF is available to Subscribers Only