Neural network simulations of the primate oculomotor system. V. Eye–head gaze shifts

Abstract

We examined the performance of a dynamic neural network that replicates much of the psychophysics and neurophysiology of eye-head gaze shifts without relying on gaze feedback control. For example, our model generates gaze shifts with ocular components that do not exceed 35 degrees in amplitude, whatever the size of the gaze shifts (up to 75 degrees in our simulations), without relying on a saturating nonlinearity to accomplish this. It reproduces the natural patterns of eye-head coordination in that head contributions increase and ocular contributions decrease together with the size of gaze shifts and this without compromising the accuracy of gaze realignment. It also accounts for the dependence of the relative contributions of the eyes and the head on the initial positions of the eyes, as well as for the position sensitivity of saccades evoked by electrical stimulation of the superior colliculus. Finally, it shows why units of the saccadic system could appear to carry gaze-related signals even if they do not operate within a gaze control loop and do not receive head-related information.

Human eye-head gaze shifts preserve their accuracy and spatiotemporal trajectory profiles despite long-duration torque perturbations that assist or oppose head motion

“…Furthermore, even if the voluntary command stayed the same in control and perturbation trials-and head trajectory adjustments were ensured by reflex mechanisms not part of the corollary discharge-the voluntary motor command would still not predict the actual head trajectory. Thus, within the context of gaze control models in which the saccade burst generator is attenuated by a corollary discharge of the head motor command (e.g., Freedman 2001; Kardamakis et al 2010), our paradigm ensured that any corollary discharge in torque trials-be it from the voluntary command alone or from the sum of voluntary and reflex mechanisms-could not predict the resulting true head trajectory and therefore could not modulate the eye saccade command appropriately. For example, in opposing trials the head was slowed by the motor and subjects opposed the torque, generating a motor command that would have given a faster head movement had the head not been impeded.…”

“…For example, in opposing trials the head was slowed by the motor and subjects opposed the torque, generating a motor command that would have given a faster head movement had the head not been impeded. In the model structures proposed by Freedman (2001) and Kardamakis et al (2010), a stronger head motor command means the eye saccade burst generator is more attenuated, with the result that the saccade is slower. However, in our results (Fig.…”

“…Can existing gaze control models explain these observations? As mentioned in the introduction, some papers have proposed separate feedback controllers for the eye and head, each driving their platform to a specific end point (Freedman 2001(Freedman , 2008Kardamakis et al 2010;Phillips et al 1995). These approaches incorporated the mechanism that the VOR, via PVP cells, is deficient during saccadic gaze shifts Roy and Cullen 1998).…”

“…In these models there is no updated GPE signal, but, to enable the eye to respond to changes in head motion, a head velocity signal, obtained from either the semicircular canals (Bizzi 1981;Phillips et al 1995) (2001) and Kardamakis et al (2010) were proposed to explain how the eye compensates for natural changes in head motion so as to generate "twin-peak" eye velocity profiles that are very different from those of head-fixed saccades.…”

“…In these models there is no updated GPE signal, but, to enable the eye to respond to changes in head motion, a head velocity signal, obtained from either the semicircular canals (Bizzi 1981;Phillips et al 1995) or a corollary of the head motor command (Freedman 2001;Freedman and Quessy 2004;Freedman and Sparks 2000;Kardamakis and Moschovakis 2009;Kardamakis et al 2010;Vliegen et al 2005) attenuates the saccade burst generator. The models of Freedman (2001) and Kardamakis et al (2010) were proposed to explain how the eye compensates for natural changes in head motion so as to generate "twin-peak" eye velocity profiles that are very different from those of head-fixed saccades.…”