Abstract
According to the old idea of the “hard threshold”, if the stimulus is too weak, it has no effect on the nervous system. Many experiments motivated by signal detection theory in the 1960s cast serious doubt on the hard threshold concept. However we have collected data from human observers that seem consistent with the presence of a hard threshold in the detection of a drifting grating, in detecting the sum of two oppositely drifting gratings, and in discriminating the direction of drifting gratings. The observers viewed 2 c/deg drifting or flickering patterns embedded in dynamic Gaussian white noise. The contrast was varied and d′ was measured. For an ideal observer in each task, the psychometric function should be a line through the origin. The same is true for a nonideal observer with suboptimal efficiency and internal noise. In fact, for all tasks the psychometric function was linear but with a substantial rightward shift away from the origin. Thus once the contrast declined below some hard threshold, d′ was 0. The shape of the functions was not consistent with observer uncertainty. When the temporal frequency was varied, the psychometric functions did not change slope but shifted sideways. Thus a major factor determining the shape of the temporal contrast sensitivity function is the position of the hard threshold rather than internal noise or sampling efficiency as conventionally defined.