We studied the dynamics of voluntary, horizontal, binocular gaze-shifts between pairs of continuously visible, real three-dimensional targets. Subjects were stabilized on a biteboard to allow full control of target angles, which were made to differ only in distance (pure vergence), only in direction (pure version; conjugate saccades) or in both distance and direction (disjunctive saccades). A wide range of changes in vergence (0-25 deg) and version (0-65 deg) was recorded to study the dynamics of disjunctive saccades, described until now for limited ranges, throughout the horizontal oculomotor range within manual working space, and to study the velocity-duration-amplitude relations (''main sequence'') of disjunctive vs conjugate saccades. Pure vergence was almost never observed; divergence, especially, was always associated with saccades. Likewise, horizontal saccades were never strictly conjugate, they always contained a transient divergence-convergence sequence. The amplitude and velocity of these transient components varied systematically with saccadic size. In combined version-vergence movements, vergence was, in general, accelerated and shortened as a function of increasing version. This effect was fairly uniform for divergence, which appeared to increase in velocity by about as much as the transient peak divergent velocity of the version saccade. The intrasaccadic fraction of divergence increased from about 50% to close to 100% as a function of increasing version. For convergence, saccades up to about 20 deg were also accelerating; in this case it appeared as if the transient peak convergent velocity of the version saccade was added to the basic convergence velocity. For larger saccades this effect was partly counteracted by the penetration of an initial divergence associated with the saccade. This initial divergence delayed and slowed down convergence. The intrasaccadic fraction of convergence varied between about 40% and 70%. In disjunctive saccades the individual eyes did not follow the main-sequence parameters of conjugate saccades of comparable sizes, except for the eye that moved with the combination ''abduction and divergence''. For all other combinations of vergence and version, disjunctive saccades had lower peak velocities and longer durations than conjugate saccades. As a consequence, disjunctive version was also slower than conjugate version. Thus, while version accelerates vergence, vergence slows down version: in the generalized case of three-dimensional gaze-shifts, peak velocities and durations are in between those of the limiting cases of pure version and pure vergence. We conclude that, within manual working space, binocular gaze-shifts are effected by the highly integrated action of conjugate and disjunctive mechanisms, both of which are expressed preferentially in fast, saccadic movements.
