Purchase this article with an account.
Ben S. Webb, Tim Ledgeway, Paul V. McGraw; Adaptive spatial integration of orientation signals over time. Journal of Vision 2008;8(6):353. doi: 10.1167/8.6.353.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Psychophysical studies often claim, or implicitly assume, that the process of spatial integration is synonymous with averaging (or other statistical combinations) of local information contained within a visual image. We have recently shown that this is not the case for the spatial integration of local motion directions; physiologically plausible neuronal population decoders (maximum likelihood, winner takes-all) rather than image-based statistics accurately predicted human observers' perceived direction of global motion at extended stimulus durations. Here we ask which of these processes (image-based statistical estimates or neuronal population decoding) underpin the spatial integration of local orientation signals over different time frames. In a temporal two-alternative forced choice task, observers discriminated which of two sequentially presented texture patterns had a more clockwise surface (global) orientation. Texture patterns were composed of 500 oriented Gaussian lines (envelope SD, 0.16 x 0.33 deg) randomly positioned within a circular aperture (diameter, 10 deg), presented at a range of stimulus durations (0.05-3.33 sec). Lines in the standard texture had a common orientation, randomly assigned on each trial; the orientations in the comparison texture were chosen independently from a skewed (asymmetric) probability distribution with distinct measures of central tendency. We simulated observers' performance on this task on a trial-by-trial basis with a bank of orientation tuned neurons that respond to the stimulus distributions with a Gaussian sensitivity profile corrupted by Poisson noise. The perceived surface orientation of texture patterns was accurately predicted by an image-based statistic (vector average orientation) at short stimulus durations and by algorithms (winner takes-all, maximum likelihood) that decode the orientation tuned activity of the simulated population of neurons at longer stimulus durations. Our results suggest that the spatial integration of local orientation information is an adaptive process that uses different strategies as more sensory evidence is accumulated over time.
This PDF is available to Subscribers Only