Different ways of modeling spatial-frequency uncertainty in visual signal detection

Inferior human signal-detection behavior compared with that of ideal observers has been explained by intrinsic uncertainty of the human observer with respect to certain signal parameters. One way to model this uncertainty is to assume that the observer simultaneously monitors multiple channels, corresponding to possible parameters. However, it is also conceivable to assume that an observer, uncertain about which channel to monitor, chooses a suboptimally tuned single filter. Finally, uncertainty may also cause the filter underlying a single channel to broaden. In this paper these different models are investigated with respect to spatial-frequency uncertainty for matched filters detecting Gabor signals. All three mechanisms predict a decrease in detection performance. However, it is shown that the resulting psychometric functions are different. While the slopes increase with uncertainty for the multiple-channel models, they decrease for a randomly chosen single channel. Broadening a single filter leads to parallel psychometric functions.