Abstract
Studies of spatial vision and attention seek to measure sensitivity to visual stimuli at multiple locations within a single experimental session. A key challenge for behavioral psychophysics is to decouple the contributions of perceptual sensitivity from those of response bias at each location. Signal detection theory (SDT) is a powerful approach for decoupling sensitivity from bias in two alternative forced-choice (discrimination) and Yes-No (detection) tasks. However, conventional SDT cannot be readily applied to tasks with more than two response alternatives. Here we introduce a theoretical framework that extends SDT to a detection task with multiple (more than two) alternatives. In multiple-alternative detection tasks, targets are presented at one among several potential locations, and the subject is rewarded for detecting and indicating the location of the target (when it occurred), or for not responding when no target was presented ("catch" trials). Based on a structural model and decision rule, our framework permits principled estimation of sensitivity and bias at each location from observed response probabilities. We applied this framework to estimate sensitivity and bias for chickens (Gallus domesticus) and monkeys (Macaca mulatta) performing four-alternative visual detection tasks. In each task, the target could occur at one of four locations (one in each visual quadrant) or not at all (catch), generating a 5x5 contingency table of response probabilities. For both species, the theory demonstrated remarkable predictive ability: with parameters estimated from only 8 contingencies (the 4 miss and 4 false-alarm rates) response probabilities for all remaining 17 contingencies could be predicted with less than 3% deviation. Our results demonstrate that a common model accounts for spatial detection behaviors in these diverse species. This framework will find important application in assessing the differential effects of neural manipulations (microstimulation or inactivation) on sensitivity and bias in multialternative detection tasks of spatial vision and attention.
Meeting abstract presented at VSS 2013