Abstract
Perception is inherently ambiguous. Binocular and perceptual rivalry model such ambiguity by presenting constant stimuli that evoke alternating perceptual interpretations. We modeled key phenomena that are common to nearly all forms of rivalry: i) Dominance durations, the times in which a single percept is perceived, follow a heavy-tailed distribution. ii) Changes in stimulus strength (e.g., contrast of one or both stimuli) have well defined effects on dominance durations (Levelt's propositions). iii) Sufficiently long periodic stimulus removal ("blanking") stabilizes the percept, while short blanking destabilizes it. Our model consisted of three coupled winner-take-all (WTA) circuits. Each WTA circuit contained two excitatory units that projected to one inhibitory unit, which in turn projected to both excitatory units. To test the model's quantitative predictions, we performed two binocular rivalry experiments for obtaining fine-grained measurements about the relation between stimulus strength and dominance durations. Small (0.6° visual angle) sinusoidal gratings (oriented +45° in one eye, -45° in the other) were used at 6 contrast levels per eye (36 in total). Contrast levels were adjusted individually. In experiment 1, stimuli were continuously presented (blank duration 0); in experiment 2 the number of contrast levels was reduced to 3, but blank duration and inter-blank presentation duration were varied. We found that the network exhibited all three aforementioned hallmarks of rivalry. Furthermore, the model made novel predictions on the functional dependence of dominance durations on stimulus strength, blank duration and inter-blank duration, which were well compatible with experimental data. Beyond predicting all hallmarks of rivalry, our model is well founded in neuronal circuitry. It is a generic model of competitive processes rather than tailored to explain specific aspects of rivalry. Hence our model provides a natural link from rivalry to other forms of perceptual ambiguity and to other competitive processes, such as attention and decision-making.
Meeting abstract presented at VSS 2013