A comparison of two lateral inhibitory models of metacontrast

Metacontrast, an apparent reduction in brightness of a target that is follo wed by a non-overlapping mask, has been modeled with simulated neural nets incorporating either recurrent lateral inhibition or forward and backward i nhibition with lateral components. A one-layer lateral inhibitory model (B. Bridgeman, 1971, Psychological Review 78, 528-539) and a six-layer model ( G. Francis, 1997, Psychological Review 104, 572-594) both simulate the basi c metacontrast effect, showing that stimulus-dependent activity that reverb erates for some time in the model after stimulus offset is essential to sim ulate metacontrast. The six-layer model does not simulate monotonic masking with low response criterion, an essential property of metacontrast; the la teral inhibitory model uses duration of reverberation to simulate the crite rion. Each model simulates several variations of masking, such as changing the relative energy of target and mask, but neither can handle effects of p ractice or attention that apparently engage higher processing levels. (C) 2 001 Academic Press.