Purchase this article with an account.
Jesse Moyer, Timothy Vickery; Interference from an integral feature in visual statistical summary representations. Journal of Vision 2014;14(10):1054. doi: https://doi.org/10.1167/14.10.1054.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Humans are capable of computing an accurate representation of the average value of a given feature within a set, including size (Ariely, 2001), brightness (Bauer, 2009), or orientation (Dakin & Watt, 2009). In the natural world, members of a set may vary across many feature dimensions. Can we selectively attend to a single feature and ignore other features when computing a summary statistic? Prior research (Moyer, Payne, Pitts, & Palomares, VSS 2012) demonstrated that estimations of average length or orientation are unaffected by the presence of variability in the other feature dimension. However, length and orientation are separable features; are feature averages similarly immune to variability of an irrelevant but integral feature? In the current study, we examined whether estimation of average height is influenced by irrelevant variability in width, given that height and width are known to be integral features. On each trial participants were briefly presented with sets of rectangles and judged average height, or sets of lines and judged average line length, using an adjustment procedure. An irrelevant integral or separable feature (rectangle width or line orientation, respectively) was either homogenous (all items were the same width or orientation) or heterogenous (widths or orientations were randomly varied). Consistent with prior work, we found no accuracy cost for computing the average of length while orientation was varied. However, participants were less accurate at judging average height when width was varied than when it was held constant. Our results suggest that statistical summary representations are not immune to variance in non-target feature dimensions if the two dimensions are integral, yet summary representations remain robust to variance in separable feature dimensions.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only