Abstract
To understand the limits of visual performance, it is critical to determine the causes underlying response variability. One possibility is that response variability is caused by strategies that are based on one's response history, as opposed to the specific stimulus at hand (Green, 1964). We tested this idea using a variation of the double-pass method described by Burgess & Colborne (1988). In the double-pass method, observers are presented with identical stimulus-plus-noise patterns at different times. The relationship between response agreement and response accuracy is an index an observer's internal-to-external noise ratio (I:E).
We measured performance at 5 contrast levels in a 2AFC orientation discrimination task in which observers judged whether a sine wave grating was oriented 1 deg to the left or right of vertical. Gratings were embedded in white Gaussian noise, and identical stimuli were separated by a trial lag of either 1, 2, 4, 8 or 16 trials. We reasoned that if response history strongly influences internal noise, then I:E should vary across different trial lags — an observer's response on any given trial would likely be influenced by particular trials in the past. Initially, we found differences in I:E across different trial lags. However, these differences diminished by the third session. Our results suggest that, with practice, response history plays only a minor role in producing response variability. We are currently examining the effects of spatial frequency on I:E to determine the locus of contrast-variant noise, as well as the effects of luminance on I:E to determine the influence of photon noise on contrast-invariant noise.