Abstract
The human visual system is adept at encoding the global velocity of image motion by integrating local estimates of image speed and direction across space and over time. Yet the precise computations governing this integration process, particularly in the case of global speed perception, are still unclear. Psychophysical studies using random_dot-kinematograms (RDKs) have previously suggested that global speed discrimination is sensitive to the average, but not the mode, of the physical speeds present in the stimulus (Watamaniuk & Duchon, 1992; Vision Research 32: 931–941). However, perceived global speed was not directly measured and other statistical measures of central tendency (e.g. median and geometric mean) were not investigated. To address this issue, in the current study observers judged which of two RDKs had the faster global speed. In the standard RDK all dots moved with the same speed. In the comparison RDK each dot undertook a random-walk in speed drawn (with replacement) from an asymmetric distribution (either a skewed Gaussian or rectangular distribution) with distinct measures of central tendency. Results showed that, in general, perceived global speed tended to coincide with the median of the local physical speeds present in the stimulus and not the other statistics. However, under some conditions (when the comparison speeds were drawn from a rectangular distribution and were much slower than the standard) observers reported transparency and this adversely affected their ability to make reliable global speed judgements. When transparency was minimised by using RDKs with relatively high dot densities or brief durations, perceived global speed was still best characterised by the median physical dot speed. Consequently, global speed discrimination might only be sensitive to the average physical image speed when this inadvertently co-varies with the median speed.