Abstract
The ideal observer approach has been used to reveal the causes of human limitations in detecting simple visual patterns at near threshold contrasts. In everyday situations, however, the patterns to be detected often have very high contrasts. We studied the sources of visual inefficiency in detecting high contrast Gabor patches in dynamic Gaussian noise by measuring simple reaction time (RT) and by comparing the results to predictions from an ideal observer model. The ideal observer in a simple RT task must form an estimate of the time of arrival of the signal and hit the button at that time. For an ideal observer, the variance of this time of arrival estimate increases linearly as the variance of the external Gaussian noise increases. The human observer's RT variance behaves in a similar way, but humans have low sampling efficiency and add internal noise.
Supported by a grant from the Engineering and Physical Sciences Research Council (UK) to WS.