Abstract
Monkeys were trained to detect 100-msec increments in the intensity of continuous white noise. A response on one of two bars was reinforced with some probability if it conformed to the presence or absence of the increment on that trial. Stimulus parameters of background intensity, increment size, and probability of increment presentation were varied, and response probabilities and latencies were recorded. The task was analogous to the "yes-no" task used in human psychophysics. Data analysis within the context of signal-detection theory revealed response biasing toward one bar or the other to be related to the probability of increment presentation, whereas sensitivity depended on the combination of increment size and background noise intensity. Weber's law was found to hold for a large range of background intensities in that the sensitivity to relative intensity increments varied little. Performance was compared to that of an ideal observer that uses samples of the envelope of the noise waveform on which to base its responses.