In real-world vision, both objects in the environment and the observer's eyes move, but these conditions are rarely replicated in psychophysical experiments. Here, observers watched a continuous movie of naturalistic stimuli (a nature documentary) making free eye movements. Localised regions of the image were incremented in contrast within a 1 octave spatial band, and the position of these targets on screen was updated at 120 Hz to remain at the same retinal location as the observers' eyes moved. Observers reported the location of the contrast increment relative to their fovea (4AFC). A contrast response function was fitted from these judgments using simulations (MCMC) to estimate the full Bayesian posterior of a multilevel model incorporating inter-subject variability. There was no evidence of saturation within the range of contrasts occurring in our stimuli, and the fastest response acceleration appeared for target bands around 1.5 cycles per degree, indicating that the contrast sensitivity function peaked at lower spatial frequencies than for static narrowband stimuli. I will also present the results of applying statistical learning techniques to isolate stimulus features that are predictive of performance. Experiments with simple grating stimuli may measure aspects of the human visual system that are atypical of vision in the real world.
Tom Wallis did his undergraduate and PhD studies at the University of Queensland and spent the last three years as a postdoctoral fellow at the Schepens Eye Research Institute, Boston. He is currently an itinerant scientist.