Initial directions and landing positions of binocular saccades.

Existing models for the generation of saccades predict fixed trajectories between start and landing positions of saccades. Experimental data show that saccades have rather variable trajectories. The objective of the present research is to quantify the variability in trajectories of binocular saccades and to test in how far spatial variability can be described by adding noise to components of existing models. We studied the trajectories of self-paced saccades. Saccades were made between a number of stationary, visual targets lying in the frontal plane. More than 75 saccades were made to each target. Horizontal and vertical movements of both eyes were measured with a scleral coil technique. We defined the direction from starting position to end position of each primary saccade as the effective direction. We defined the direction from starting position to the eye position when the saccade had covered a distance of 2.5 deg as the initial direction of the saccade. We find that variability is two to seven times larger in initial directions than in effective directions. Effective directions are more accurate and more precise than initial directions. For each eye, initial and effective directions of saccades made to a particular target are negatively correlated, although in most cases rather weakly (0.1 < r2 < 0.5). Contrastingly, initial directions are always highly correlated (r2 > 0.8) of associated binocular saccades. High correlations are also found between effective directions. We conclude that curvedness of saccades is the result of a purposeful control strategy. The saccadic trajectories show that, initially, the eyes are accelerated roughly in the direction of the target and subsequently they are guided to the target. Analysis of possible models suggests that variability predominantly enters the saccadic system at a central stage of neuronal saccade generation. We conclude from simulations, in which we used different models of saccade generation, that the major sources of directional variability are part of a feedback loop. This conclusion provides indirect evidence for the presence of a feedback loop in the saccadic system.