Noise location and the slope of the psychometric function for simple motion stimuli.

We measured subject performance as a function of luminance for both detection and discrimination of increment stimuli; some were static, and some were arranged to give two-step apparent motion. Our aim was to examine a prediction for the shape of the psychometric function for motion: an accelerating function due to the presence of a multiplicative nonlinearity contained in many low-level motion models. For the tasks with static stimuli we found psychometric function slopes (of log d' versus log luminance plots) between 1.9 and 2.4 in two subjects, as previously reported. For the tasks with apparent motion stimuli in the same range of detectability, however, the slopes are between 1.2 and 1.7. The lower slopes indicate that many low-level motion models are either incorrect or incomplete as currently specified, and changes in nonlinearities and noise placement are discussed.