Абстрактный

Visual change detection has a memory limit for ensemble statistics

Neha Kouser

Accounts of remembering supported freelance item representations could overlook a potential contribution of ensemble statistics, higher-order regularities of a scene like the mean or variance of a visible attribute. Here we tend to used amendment observe on tasks to research the hypothesis that observers store ensemble statistics in remembering and use them to detect changes within the visual surroundings. We tend to controlled changes to the ensemble mean or variance between memory and check displays across six experiments. We tend to created specific predictions of observers’ sensitivity victimization associate degree optimum summation model that integrates proof across separate things however doesn't observe changes in ensemble statistics. We tend to found sturdy proof that observers outperformed this model, however only if task issue was high, and just for changes in stimulant variance. Below these conditions, we tend to calculable that the variance of things contributed to vary detection sensitivity additional powerfully than anyone item during this case. In distinction, however, we tend to found sturdy proof against the hypothesis that the common feature price is keep in operating memory: once the mean of memoranda modified, sensitivity didn't disagree from the optimum summation model, that was blind to the ensemble mean, in 5 out of six experiments. Our results reveal that amendment detection is primarily restricted by uncertainty within the memory of individual options; however that memory for the variance of things will facilitate detection below a restricted set of conditions that involve comparatively high remembering demands.