Purchase this article with an account.
Émilien Tlapale, Barbara Anne Dosher, Zhong-Lin Lu; Modeling motion perception and perceptual learning in random dot kinematograms. Journal of Vision 2014;14(10):474. doi: https://doi.org/10.1167/14.10.474.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Random dot kinematograms (RDKs) represent a fundamental category of stimuli in the study of visual motion in both neurophysiology and psychophysics. Although models have been proposed to account for certain results from RDK experiments, none offer a quantitative account of the effects induced by systematic manipulations of the parameters of RDKs. Yet, a comprehensive consideration of those stimulus variations could provide significant constraints on models of the visual motion system. Here we propose a detailed dynamical model comprised of motion detection, motion integration and perceptual decision stages, respectively linked to cortical areas V1, MT and LIP. In addition the short term dynamics of the model are influenced by slow connectivity reweighting that leads to long term perceptual learning. We show that the model is sufficiently generic to handle a wide class of moving stimuli. It is also sufficiently precise to quantitatively reproduce the threshold curves of Watamaniuk et al (1989,1992) in which movement direction distribution of random dots, presentation duration and stimulus size are parametrically varied. The richness of those data provides constraints on the dynamics and connectivity of the model. Moreover our model provides a natural explanation for the threshold reductions associated with very broad direction distributions, which are mostly unexplained in the original study. The simple reweighting mechanism (Dosher and Lu, 1998) allows us to replicate the perceptual learning effects observed for RDKs (Ball and Sekuler, 1982,1987). Finally, we consider the impact of several noise sources at the different stages of the model, making testable predictions on their influences on threshold curves. As a whole, we present a precise yet generic model of visual motion processing which is able to quantitatively account for both the effects of parametric stimulus variations and longer term learning effects.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only