Introduction. When our eyes independently view dissimilar images, we perceive an ongoing alternation between the images: binocular rivalry. Rivalry’s alternation cycle comprises periods of exclusive dominance and transition periods during which both images are partly visible. Two key driving forces are self-adaptation of, and cross-inhibition between the two images’ neural representations. Models based on these factors correctly reproduce dominance durations, but their predictions on transitions are unknown. We studied rivalry’s temporal dynamics in relation to stimulus contrast, considering dominance durations, transition durations and the occurrence of return transitions: Occasions on which an eye loses and regains dominance without an intervening dominance period of the other eye. In simulations we investigated if existing models could reproduce our findings.

Results. We find that dominance durations and the incidence of return transitions depend similarly on contrast; transition durations show a different dependence. We also demonstrate that previous descriptions of contrast dependence of dominance durations are incomplete. Our data refute existing models, particularly the large transition durations and high frequency of return transitions. We argue that deterministic factors are insufficient to explain rivalry, and show that noise is a crucial driving force. Relevant models should cover this interplay between deterministic and stochastic forces.