Time series of the durations each eye was dominant during binocular rivalry were obtained psychophysically. The oscillations showed an adaptation effect with mean and standard deviations of rivalry dominance durations increasing as a square root function of time over the course of a trial. The data were corrected for this non-stationarity. Dominance durations had a log-normal probability distribution and the autocorrelation function revealed no short term correlations in the time series. In an attempt to distinguish whether the variability of durations was due to a deterministic, low-dimensional chaotic attractor or to a stochastic process, the data were subjected to two tests. The first was calculation of correlation dimensions and the second was nonlinear forecasting of the time series. Both tests included comparisons with randomized `surrogate data' as controls. In neither case was there a large difference between test results for actual data and surrogate data. We conclude that chaos is not a major factor underlying variability in binocular rivalry.