Spatial frequency summation in visual noise

Kersten [Vision Res.27, 1029 (1987)] reported that absolute efficiency for the detection of static, one-dimensional bandpass noise was high and approximately constant for stimulus bandwidths ranging from 1 to 6 octaves. This result implies that human observers integrated information efficiently across a wide range of spatial frequency. One interpretation of this result—and similar results obtained with auditory stimuli [J. Acoust. Soc. Am.32, 121 (1960)]—is that human observers, like ideal observers, can detect stimuli using an internal filter that has an adjustable bandwidth. The current experiments replicate Kersten's findings, extend them to the case where observers are uncertain about stimulus bandwidth, and use the classification image technique to estimate the filter used to detect noise stimuli that differ in bandwidth. Our results suggest that observers do not adjust channel bandwidth to match the stimulus and that detection thresholds are consistent with the predictions of a multiple-channel model.