Abstract

The ambiguous Necker cube stimulus appears to switch between two alternative perspectives, 'cube-up' and 'cube-down', such that the inter-switch interval gives percept duration. Percept frequency, plotted as a function of duration, is modelled by a gamma distribution, and the model fit is improved by discarding arbitrarily durations greater than three standard deviations above the mean (Borsellino et al, 1972 Kybernetik 10 139 - 144). Alternatively, good fits can be achieved by discarding the 'dead period' at the beginning of the distribution that represents the observer's decision time between a switch and the subsequent response with a key press that signals the switch (Martin, 1967 Papers in Psychology 1 2).

Twenty-nine naïve observers recorded switches between cube-up and cube-down percepts. During a total observation period of 10 min, observers recorded up to 300 switches. We found that the best fits were generated with the gamma distribution when the 'dead period' was discarded. Arbitrary trimming of the tail, three standard deviations above the mean, had little impact on goodness-of-fit. This suggests that (i) the perceptual changes can be adequately modelled as a simple stochastic process, and (ii) the modelling is improved by accounting for the decision time between switch and response.