Abstract
It is well established that there is a close relationship between smooth pursuit eye movements and MT visual motion signals (Lisberger & Movshon, J.Neurosci.,1999). However, existing models of this visual-motor transformation contain components that are inconsistent with known physiology or lack the millisecond-by-millisecond response outputs needed to adequately capture the temporal dynamics of pursuit. We sought to overcome these deficits by using a model of MT neuron responses based directly on V1 neuron inputs (Perrone & Thiele, ARVO., 2000). The model incorporates two V1-like units based on spatio-temporal energy filters: the first has sustained low-pass temporal frequency tuning (S), whereas the second has transient band-pass temporal tuning (T). The response of the model MT sensor is given by: log(T+S+a)/(|logT−logS|+d). The additional terms control the spatial (a) and temporal (d) frequency tuning bandwidths of the MT sensor. This model provides an excellent description of the spatio-temporal frequency response properties of MT neurons (Perrone & Thiele, Nature Neurosci., 2001). Here, we examined the time course of the model sensor responses to a variety of moving dot pattern stimuli, including the step-ramp stimulus motions (Rashbass, J. Physiol., 1961) typically used to elicit smooth pursuit. The model inputs were 16-frame movie sequences composed of 256 × 256 pixel images. To compare the performance of the model to the properties of MT neurons, we measured the latencies and the transient/sustained ratios from the responses of the model MT sensor. The model replicated the transient overshoots in firing rate often found in the activity of MT neurons when the speed of the visual stimulus changes (Lisberger & Movshon). Our results show that a model of MT responses based on VI neuron properties can account for some of the distinguishing features of the visual motion inputs that drive pursuit.
This work was supported by NIH grant EY12212 and a McKnight Scholar Award to R.J.K.