Purchase this article with an account.
Howard S. Hock, Cynthia L. Park, Gregor Schoner; A model of self-organized motion pattern formation. Journal of Vision 2001;1(3):371. doi: https://doi.org/10.1167/1.3.371.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose. In a variety of physical and biological systems, pattern formation occurs without a pattern-creating agent, without rules or instructions, and without the specification of the pattern by energy put into the system. Such emergent, self-organized patterns also exist in vision; either unidirectional or oscillatory motion patterns are perceived for the directionally ambiguous, counterphase row-of-elements (Hock & Balz, 1994). If pattern formation for this stimulus is indeed self-organized, it would be possible to create a model for which the patterns are not specified, but are nonetheless generated. Method. We determine whether global pattern formation for the counterphase stimulus can be based strictly on the activation of local motion detectors that respond selectively to leftward or rightward motion. Non-linear dynamical equations incorporate: 1) Local stability: Activation values for leftward and rightward detectors have stable fixed-points, 2) Cooperativity: A motion detector contributes to the activation of other detectors with similar directional selectivity through excitatory interactions that depend non-linearly on the detector's activation level, and 3) Competition: Inhibitory interactions among leftward and rightward detectors result in one or the other being activated, not both at the same time. Results. Unidirectional and oscillatory global motion patterns, though unspecified in the model equations, are self-organized on the basis of the interaction-modified activation of local, directionally selective motion detectors. The current model is unique in reproducing global motion patterns that are more than collections of local motions with a common direction (as for RDCs); they have emergent temporal structure.
This PDF is available to Subscribers Only