Abstract
Many commonplace activities (e.g. waiting for a gap in traffic) involve deciding to act upon one of a series of events. In such circumstances, the probability of acting successfully must be estimated and weighed against the probability of success for subsequent events. Here, we examined whether observers can maximise gain in a task requiring the selection of a single entry in a finite series, for use in a subsequent motion estimation task. In a series of training trials, observers were presented with a dot moving from the centre to the edge of a circle. Dot movement was randomly selected from a von Mises distribution, with spread K. The dot disappeared before reaching the edge of the circle, and participants were asked to ‘catch’ the dot by estimating the point at which they believed it would hit the circle’s edge. The probability of catching a dot was measured at a range of K-values. To examine selection strategies, participants were required to attempt to catch a single dot from a set of seven. The seven dots were presented in sequence, and observers were unaware of the K-values used in each set. For each dot, observers had the option to attempt a catch, or to skip and be presented with the next item in the set. Dots could not be returned to once skipped. Results were compared to a Maximum Expected Gain model, which uses observer performance in training to define the optimal selection strategy. Contrary to the strategy adopted by observers, this model predicts that observers should typically wait until later in the set to act, skipping earlier items. Observers instead chose lower than predicted K-values, selecting to act earlier in the set. These results suggest that the ability to maximise gain in perceptual tasks may be limited by task complexity.
Meeting abstract presented at VSS 2013