Purchase this article with an account.
Carl Gaspar, Wei Chen; Random-sampling leads to multiplicative noise in crowded displays. Journal of Vision 2014;14(10):774. doi: 10.1167/14.10.774.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Is crowding stochastic? Dakin, Cass, Greenwood and Bex (2010) show that their crowding results are consistent with a model that assigns random weights to multiple flanking elements. If observers are judging the orientation of a peripheral Gabor target, then this random-sampling model predicts that internal noise should increase with the variance of flanker-orientation – a form of multiplicative noise. We tested this hypothesis directly by measuring the effect of flanker-orientation variance on response consistency in a 2AFC orientation discrimination task. 2 observers (Burgess & Colborne, 1988) identified which of 2 peripheral Gabor patches (2.5 degrees eccentricity), surrounded by 8 Gabor flankers, was tilted (clockwise or counterclockwise from horizontal). Target-orientation was varied using method-of-constant-stimulus, and an exact copy of each stimulus was shown in repeated trials (same target-orientation and configuration of flanker orientations). Orientation thresholds and response-consistency were measured separately for various levels of flanker-orientation variance (3 levels in Observer 1, spanning an 8-fold range; and 4 levels in Observer 2, spanning a 16-fold range). Critically, the mean of flanker orientation was always 0 (horizontal) for every stimulus; performance could only be affected by flanker variance. For both observers, orientation thresholds increased with flanker variance by a log-log slope of 1/2. Most importantly, response-consistency data for both observers show that the slope between proportion-correct and proportion-agreement reached an upper asymptote at the highest flanker variance, which is the signature of muliplicative noise and consistent with the model of random-sampling suggested by Dakin et al. (2010). In a current experiment we are varying the number of flankers as well as their orientation-variance, and predict that a random-sampling model can be well fit to all these data.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only