Selective and delay adaptation of human saccades

Abstract

The consistently triggered step back of a target during primary saccades of a human subject induced a gradual change in gain, the ratio of the saccade amplitude to the target eccentricity. After a few hundred trials, subjects were able to foveate the displaced target in a single saccade. Presentation of a displaced target showed that human memory guided saccades have gain adaptation just like the well-established adaptation of visually guided saccades. Examining the transfer of adaptation between the memory guided saccade and two other types of visually guided saccades showed that each saccade transferred a 10–25% adapted gain change to the other saccades. However, any pair of the three saccades acquired different gains by adaptation in the same horizontal direction simultaneously, hence each saccade had adaptive capability independent of the others. Adaptation took place even when the appearance of a displaced target was delayed by 400–600 ms from the end of a primary saccade. These findings have important implications about the adaptation, particularly the location and temporal property of the adaptive mechanism in saccade generation.

Introduction

Primates move their eyes rapidly to a target of interest to acquire it on their retinal fovea. This fast eye movement is called a saccade. Visual feedback is too slow to control a saccade, because it takes at least 20 ms for the processing within the retina, while small saccades last only 20–30 ms. From the computational viewpoint, the local feedback theory [23] proposed that the saccade control system uses the signal from the neural integrator in the final common oculomotor pathway for the current eye position and compares it with that of the desired eye position.

A computed saccade command may not always be consistent with the target eccentricity. Patients with abducens palsy [18] and a monkey tenectomized in one eye [22] foveated the target with a primary saccade followed by several corrective saccades. In the monkey experiment, a gradual growth in the amplitude of the primary saccade occurred, and 5 days after the tenectomy operation, the monkey was able to foveate a target with a single saccade. The monkey’s saccadic gain (the ratio of the amplitude of a primary saccade to the eccentricity of the target) increased adaptively from 0.3 to its normal value.

In non-invasive adaptive experiments on human and monkey saccades [1], [2], [5], [9], [19], [20], [25], [29], a target was moved back during a primary saccade, and then corrective saccades followed to capture the displaced target. Repetitive and systematic application of the intrasaccadic target-step trial caused a gradual reduction in the gain of the primary saccades. If the target was moved forward during the primary saccade, a gradual increase in the gain of the primary saccades occurred. Saccades also showed an adaptive change in direction when the target changed its direction systematically during primary saccades. Precise examination indicated that adaptation by the tenectomy paradigm and adaptation by the intrasaccadic target-step paradigm occurred at similar rates in monkey saccades [24].

Human saccades undergo selective adaptation using the intrasaccadic target-step paradigm. After the adaptive gain reduction of saccades directed intentionally to a stationary target at a peripheral position (I-saccade), the gain of another type of saccade that quickly followed a jumping target (E-saccade), showed a far smaller change [10]. Deubel [6], [7], [8] demonstrated that when saccades scanning serial dots on a screen (a scanning saccade) experienced a behavioral gain reduction, that reduction showed little transfer to a reactive saccade (E-saccade) and vice versa.

A memory guided saccade (M-saccade) shares a property with an I-saccade, i.e. it is triggered intentionally, but differs in that it is directed to a remembered location that the target once occupied, and is not followed by a corrective saccade. In the adaptive procedures, corrective saccades are always led by a displaced target and executed promptly as a visually guided saccade, not as an M-saccade. We first investigated the adaptiveness of the M-saccade and whether its adaptation was transferred to two other types of saccades: slightly modified versions of E-saccade and I-saccade.

We then investigated whether these saccades could be adapted simultaneously, but to different gain conditions. One saccade type might undergo gain-reduction adaptation while another might undergo gain-increasing adaptation, both in the same horizontal direction.

Finally, we examined the effect on adaptation of the timing of the adapting target step relative to the time of the primary saccade. For example, the time delay extended to over 1 s was still effective in the trace-conditioning of a rabbit [28]. This time delay might correspond to the delay in the appearance of a displaced target in the adaptive procedures. We examined delayed adaptation for the M-saccade and the visually guided saccade. Preliminary results are present elsewhere [11], [12], [13], [14], [15], [16].