Purchase this article with an account.
Jun Saiki, Alex Holcombe; Perception of global statistics of color-motion correlation requires surface-based attention to a single motion. Journal of Vision 2009;9(8):136. doi: 10.1167/9.8.136.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
A transparent motion display comprised half of dots moving leftward, half moving rightward, half red, and half green. 100% consistency: all leftward-moving dots were the same color, and all rightward-moving dots were the other color. 0% consistency: half of dots of each motion were red, and half were green. When the color-motion pairing of all the dots is suddenly reversed by swapping dot colors, humans are remarkably poor at detecting this change when consistency is low (Saiki & Holcombe, VSS 2008). This indicates lack of representation of any particular dot's pairing. Here we explore availability of global statistics. If feature attention is completely effective, then attention to a color (say, red) should make detection of the pairing swap easy under high consistency, as the marginal mean (proportion of red dots with a particular direction) changes dramatically with the swap. The same applies to motion attention. We suspected that motion attention would be more effective than color attention because the display is subjectively organized as two surfaces moving in different directions, which attention can select. Observers monitored for the swapping of the colors of all dots, with a concurrent task of localizing a transient size change of a dot. In the ‘diffuse attention’ condition, any dot might change size. In the ‘motion attention’ and ‘color attention’ conditions, observers knew the size change would occur in a dot of the specified direction and of the specified color, respectively. With low consistency displays (0–20%) attention had no effect, suggesting it did not access individual feature pairings. Only under high consistency (60%), motion attention significantly improved detection performance (relative to diffuse attention), whereas color attention provided negligible improvement. The marginal mean provides some knowledge of the overall feature pairing proportion, but apparently this can only be computed via surface-based (here, motion) attention.
This PDF is available to Subscribers Only