Purchase this article with an account.
Oliver Layton, Ennio Mingolla, N. Andrew Browning; A Model of MT Motion Pooling Explains Human Heading Bias. Journal of Vision 2011;11(11):715. doi: 10.1167/11.11.715.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
In a rigid environment the observer's translational optic flow field has a focus of expansion (FoE) that specifies the direction of self-motion (heading) without eye rotations. Humans accurately judge to within 1 degree their direction of heading, unless the environment has independently moving objects (IMOs) that cross the observer's FoE. Studies have identified that IMOs that move horizontally with respect to the observer's translation and maintain a fixed distance bias human estimates of heading differently than those that approach the observer. While Royden & Hildreth (1996, Perception & Psychophysics 58) investigated non-approaching objects and found biases in the direction of object motion, Warren & Saunders (1995, Perception 24) found biases in the direction of the object's FoE for approaching objects. The motion-pooling model of W & S could not explain the findings of R & H. Royden (2002, Vision Research 42) argued that motion-opponent operators in primate area MT explained both sets of data, concluding that differential motion was critical to human heading estimation. This approach is difficult to reconcile with studies showing that motion pooling cells in MT project directly to dorsal MST, where direction-of-heading sensitive cells are located, but differential motion cells do not (Born & Bradley 2005, Annu. Rev. Neurosci. 28). We present a motion pooling model of MT and MST based on the model of Browning et al. (2009, Cog Psy 59) that demonstrates that differential motion is not necessary to explain human heading judgments. We generate motion sequences that mimic those viewed by W & S and H & R's subjects, using analytically computed V1 representations. Model MT pools over V1, followed by distance-weighted template-matching and competition stages in model MST. The model produces heading biases of the same direction and magnitude as humans (r = 0.84) while maintaining consistency with known primate neurophysiology.
This PDF is available to Subscribers Only