Huanping Dai
Psychoacoust. Lab., Psychol. Dept., Univ. of Florida, Gainesville, FL 32611-2065
In typical profile-detection tasks there is little difference, about 3 dB in signal-to-standard ratio, between signal threshold when the signal frequency is fixed (signal known) and when the signal frequency is randomly selected on each trial (signal unknown). For an ideal observer, this difference should be 8 dB when using 21 components. One hypothesis about the discrepancy between the human and ideal observer is that, for the human observer, the central noise (e.g., fluctuations of the decision criterion) is much greater than the peripheral noise. This assumption can account for both the slope of the psychometric function obtained in conditions when the signal frequency is known and the relatively small threshold difference between the signal-known and signal-unknown conditions. [Work supported by NIH.]