Fig. 4: A more biomimetic model of MT and the downstream pursuit circuit.

a Overview of the biomimetic circuit model of pursuit. Activation of model middle temporal (MT) neurons (triangles within square, left) depends on patch size (colored circles). Each neuron drives (1) the motion reliability pathway, which computes a vector sum of MT activity (top) and (2) the speed estimation pathway, which computes a vector average (bottom). Application of the output of the motion reliability pathway to the output of the estimation pathway generates simulated pursuit (right). b Direction tuning (top) and speed tuning (bottom) of a random sample of model MT neurons. Purple highlights one neuron. c Noise correlations between model neurons as a function of the difference in direction preference (Δ Direction preference; top), distance between receptive field positions (Δ Position; middle), and difference in speed preference (Δ Speed preference; bottom). Each dot plots the measured noise correlation for a different pair of model neurons sampled from the population. d Receptive field centers (gray points) of the population of MT neurons used to represent motion for pursuit. Green, red, and black circles correspond to the locations of the 2 deg, 6 deg, and 20 deg targets. Purple and cyan circles plot the classical receptive field of two example neurons. e Response of each model neuron for an example trial (spikes/s; color). The center of each point corresponds to the neuron’s preferred direction and speed. Large purple symbol indicates the example neuron from panels b and d. f Mean gain of the population as a function of target speed and size. g Mean simulated eye speed as a function of target speed and size (colors as in panel f). Source data are provided as a Source Data file.