We proposed a theoretical framework to distinguish three mechanisms underlying performance improvements in visual attention (1-5) and perceptual learning (6,7): stimulus enhancement, external noise exclusion via template retuning, and reduction of contrast gain-control or multiplicative noise. Measuring TVCs (threshold vs external noise contrast) at multiple performance levels under joint manipulations of external noise and attention/training can characterize the nonlinearities in the perceptual system and distinguish mechanisms and their mixtures (7,8). In visual attention, several pure cases of template retuning (2-4) and stimulus enhancement (1,4,5 have been reported. In perceptual learning in visual periphery, we found a mixture of stimulus enhancement and template retuning (6). Gold et al.(9) replicated our results with other stimuli; but instead concluded that perceptual learning only changed the signal not the noise. Our conclusions were based on a multiple-TVC constraint on nonlinearity, while Gold et al’s were based on double-pass response consistency (10). However, the double pass method only assesses the internal to external noise ratio. We show that response consistency is a function only of the ratio. Mathematically, neither the multiple-TVC nor the double-pass method can distinguish stimulus enhancement from internal additive noise reduction, nor a mixture of stimulus enhancement + external noise exclusion from the changes only in signal claimed by Gold et al. We discuss empirical results, the mathematical properties of multiple-TVCs and measures of response consistency, their relation to different classes of observer models, and to performance signatures of attention and perceptual learning.

1. Lu & Dosher, VR’98. 2. Dosher & Lu, PsychSci’99. 3. Dosher & Lu, VR’00. 4. Lu & Dosher, JEPHPP’00. 5. Lu, et al, VR’00. 6. Dosher & Lu, PNAS’98. 7. Dosher & Lu, VR’99. 8. Lu & Dosher, JOSA’99. 9. Gold, et al, Nature’99. 10.Burgess & Colborne, JOSA’88.

Supported by AFOSR, NIMH & NSF.