Complex attention filters for dot contrast derived from a centroid judgment task
26.541, Saturday, May 11, 2:45 - 6:45 pm, Vista Ballroom
Howard Yang1, Peng Sun1, Charles Chubb1, George Sperling1; 1Department of Cognitive Sciences, School of Social Sciences, University of California, Irvine
Purpose: In judging the centroid (center of gravity) of a random cloud of dots, it has previously been determined that subjects can give approximately equal attentional weight only to dots lighter than the background (ignoring dots darker than the background), only to dots darker than the background (ignoring light dots) or to all dots (Drew, Chubb, & Sperling, JOV, 2010). Here we determine whether subjects also can allocate attention to dots in direct proportion to dot contrast. Method: On each trial, either 8 or 16 dots with locations drawn from a bivariate Gaussian distribution with a roved mean were shown for 200ms. Subjects moved a cursor to the centroid of the dot cloud. There were two attention instructions (never intermixed within a day): (1) give equal attentional weight to each dot (ignoring contrast), (2) weight each dot in proportion to its contrast magnitude. On every trial, after the response the true centroid and the subject's response were displayed. Blocks (of 150 trials) with 8 dots of only one polarity (either darker or lighter than the background) were interleaved with 16 dot mixed-polarity blocks. Results: Attention filters were derived from a linear regression model for each condition. When performance stabilized, 8 of 9 subjects showed very clear equal-vs-gradient differences in single polarity conditions. Attention filters derived from dual-polarity tasks show smaller, but still significant equal-vs-gradient differences. For dots of both single and dual-polarity, differences between equal-versus-gradient were larger for dark dots than for light dots. Conclusion: In a task in which far more dots are presented than can be individually processed at a high level, subjects are able to produce attention filters that selectively weight dots either in proportion to contrast magnitude or that ignore contrast magnitude.