Abstract
One of the interesting aspects of binocular rivalry is its time dynamics, which have been used to evaluate models of binocular rivalry. Blake and Mueller (1989) introduced a novel method to modulate the time dynamics of binocular rivalry using contrast modulation either in synchrony or independent of a percept switch. Those data supported a reciprocal inhibition oscillator as a rivalry mechanism. A laminar cortical model of 3D vision, called 3D LAMINART (Grossberg and Swaminathan, 2004), quantitatively simulates rivalry percepts of real images that obey the duration-contrast relations reported by Blake and Mueller, both the slope sign and slope value of linear regressions. Key model mechanisms are binocular perceptual grouping circuits in layer 2/3 of V2 interacting with orientational competition and habituative synapses. The model clarifies how these mechanisms control visible rivalrous surface percepts in V2 and V4, and modulate cell responses in V1. These mechanisms have elsewhere explained perceptual and brain data about stable cortical development and 3D perceptual grouping, notably data about 3D figure-ground separation, transparency, neon color spreading, perception of slanted and curved surfaces, Necker cube, perceptual learning, and object attention. Thus the model shows how rivalry can emerge from brain mechanisms that carry out other perceptual tasks. The model can also reproduce the gamma distribution of dominant phase duration (Levelt, 1967), mixed phase output (Ngo et al., 2000), why feature-based rivalry (Diaz-Caneja, 1928; Alias et al, 2000; Lee and Blake, 2002) rather than eye-based rivalry can take place, consistent with the fact that the location of rivalry is higher than monocular loci (Leopold and Logothetis, 1996; Sheinberg and Logothetis, 1997). The model also explains the Logothetis, Leopold, and Sheinberg (1996) demonstration that swapping two half-stereopairs between the two eyes with the rate of 3 Hz does not affect the time dynamics of rivalry.
Supported in part by NSF and ONR