 |
"Discrimination of Complex Motion
Patterns is Consistent with an Interconnected Population
|
 |
|
|
"Discrimination of Complex Motion
Patterns is Consistent with an Interconnected Population
|
|
Scott A. Beardsley, Colin W.G. Clifford & Lucia M. Vaina
Brain & Vision Research Laboratory
Department of Biomedical Engineering
Boston University
Boston MA 02215
USA
Purpose. To use a complex
motion discrimination task to examine the structural and functional role
of radial and circular motion detectors and characterize the neural architecture
necessary to obtain comparble performance in a simulated population of
neurons from the medial superior temporal area (MST) of nonhuman primates.
Methods.
Complex motion stimuli were presented in a 24 degree diameter annular region
for 440±40 ms. Observers performed a 2TAFC task on interleaved pairs
of complex motion patterns at two dot speeds (8.4 and 30 deg/s). Discrimination
thresholds were obtained as a function of the type of complex motion using
stimulus pairs generated by symmetrically perturbing the patterns of radial
and circular motions. Human performance was compared with a population
of 100-500 MST-like units under conditions simulating (1) an "unconnected"
population code of convergent stimulus information and (2) a "fully connected"
population code incorporating excitatory/inhibitory lateral connections.
Results.
At 8.4 deg/s observer thresholds were lowest for radial motions with a
significant asymmetric increase in circular motion thresholds. At 30 deg/s,
thresholds decreased for circular motions and remained consistent across
observers In simulations with simple "unconnected" populations, performance
varied with population size and the distribution of preferred motions.
In simulations with "fully connected" populations, discrimination was independent
of these parameters and was consistent with human performance. Analysis
of the lateral interactions showed a strong dependence on the strength
and spread of inhibitory activity. Conclusion. Perceptual discrimination
for the patterns of complex motion varied significantly as a function of
the type of motion, suggesting perceptual correlates to the preferred motion
bias reported in MST. Robust psychophysical performance was achieved in
a fully connected population code which excited preferred units and inhibited
non-preferred units making a highly interconnected architecture appealing
as a method for encoding information consistently in high-level motion
areas such as MST.