Monthly
288 pp. per issue
6 x 9, illustrated
ISSN
0899-7667
E-ISSN
1530-888X
2014 Impact factor:
2.21

Neural Computation

December 1, 2003, Vol. 15, No. 12, Pages 2863-2882
(doi: 10.1162/089976603322518777)
© 2003 Massachusetts Institute of Technology
Lateral Neural Model of Binocular Rivalry
Article PDF (714.72 KB)
Abstract

This article introduces a two-dimensionally extended, neuron-based model for binocular rivalry. The basic block of the model is a certain type of astable multivibrator comprising excitatory and inhibitory neurons. Many of these blocks are laterally coupled on a medium range to provide a two-dimensional layer. Our model, like others, needs noise to reproduce typical stochastic oscillations. Due to its spatial extension, the noise has to be laterally correlated. When the contrast ratio of the pictures varies, their share of the perception time changes in a way that is known from comparable experimental data (Levelt, 1965; Mueller & Blake, 1989). This is a result of the lateral coupling and not a property of the single model block. The presentation of simple and suitable inhomogeneous stimuli leads to an easily describable perception of periodically moving pictures like propagating fronts or breathing spots. This suggests new experiments. Under certain conditions, a bifurcation from static to moving perceptions is predicted and may be checked and employed by future experiments. Recent “paradox” (Logothetis, 1999) observations of two different neuron classes in cortical areas MT (Logothetis & Schall, 1989) and V4 (Leopold & Logothetis, 1996), one that behaves alike under rivaling and nonrivaling conditions and another that drastically changes its behavior, are interpreted as being related to separate inhibitor neurons.