Implications of interrupted eye-head gaze shifts for resettable integrator reset

The neural circuit responsible for saccadic eye movements is generally thought to resemble a closed loop controller. Several models of the saccadic system assume that the feedback signal of such a controller is an efference copy of "eye displacement", a neural estimate of the distance already travelled by the eyes, provided by the so-called "resettable integrator" (RI). The speed, with which the RI is reset, is thought to be fast or instantaneous by some authors and gradual by others. To examine this issue, psychophysicists have taken advantage of the target-distractor paradigm. Subjects engaged in it, are asked to look to only one of two stimuli (the "target") and not to a distractor presented in the diametrically opposite location and they often generate movement sequences in which a gaze shift towards the "distractor" is followed by a second gaze shift to the "target". The fact that the second movement is not systematically erroneous even when very short time intervals (about 5 ms) separate it from the first movement has been used to question the verisimilitude of gradual RI reset. To explore this matter we used a saccade-generating network that relies on a RI coupled to a head controller and a model of the rotational vestibulo-ocular reflex. An analysis of the activation functions of model units provides disproof by counterexample: "targets" can be accurately acquired even when the RI of the saccadic burst generator is not reset at all after the end of the first, interrupted eye-head gaze shift to the distractor and prior to the second, complete eye-head gaze shift to the "target".