Abstract
Observers on a correlation perception task can select a target ensemble among distractor ensembles, even when the two have extremely high color similarity (Elliott & Rensink, VSS 2018). This does not follow predictions about color systems in feature-based attention (Nagy & Sanchez, 1990). To investigate further, we examined another task where attention is used to select a population for which statistical information is gathered—namely, the estimation of total number of items in one of two intermixed populations on the screen. Our number estimation task used intermixed target and distractor populations of squares that were distinguishable by color and number. Observers were asked to select a target ensemble based on its color (ignoring squares of the other color), and to report how many target color squares were in the display. We used the same experimental color space from Elliott & Rensink (VSS 2018) to vary distractors in equal perceptual steps along CIELAB color axes. This allowed us to examine how number and color feature differences influenced number estimation of the target population—a measure of attentional color selection of ensembles. Number estimation showed no interference from opposite-colored distractors, consistent with past work on visual search (Duncan & Humphreys, 1989), but inconsistent with work on correlation perception (Elliott & Rensink, VSS 2016; 2017; 2018). Targets and distractors of similar color led to large estimation errors, even for colors that could be easily separated in the correlation task. Interestingly, similar color distractors also caused lower estimation slopes, with overly high estimates for low numbers of squares and overly low estimates for high numbers. As such, these results show that the color system used in attentional selection depends on the nature of the task involved.
Acknowledgement: UBC 4-Year PhD Fellowship