The co-ordination of horizontal memory-guided saccades is similar to that of visually-guided saccades, albeit more variable. When the remembered target is made disparate, horizontal memory guided-saccades become disjunctive; disjunctivity persists even when the remembered target is no longer disparate, suggesting learning (Kapoula et al., 1998). The purpose of this study is, first, to examine the natural binocular co-ordination of vertical memory-guided saccades. Second, to test if such binocular co-ordination is modifiable by a learning mechanism. Subjects wore an afocal magnifier (8%) in front of their dominant eye. The following paradigm was used to elicit memory-guided saccades. The subjects fixated a central spot. Another target point was flashed, for only 100 ms, up or down at a randomly chosen position ± 5°, 10°, 15°. Subjects were instructed to continue fixate the central dot. After a memory delay of 1 s, the central dot was switched off and the subject was instructed to saccade in complete darkness, to the remembered target location. This paradigm was performed under the following three conditions:

• Before training (3 min) : Viewing was monocular with the eye wearing the magnifier; thus the target to be remembered was non-disparate. This control condition was used to determine the natural conjugacy of memory-guided saccades.
• Training (15 min) : Viewing was binocular. The size of the target and its eccentricity were 8% larger in the eye wearing the magnifier, thereby disparate.
• After training (3 min): to test the persistence disjunctivity, subjects performed memory-guided saccades under monocular viewing (non-disparate targets).

Nine experiments were run on 7 subjects. For six runs, eye movements were recorded with the IRIS SKALAR device; for the other 3 runs, the search coil-magnetic method was used (Robinson, 1963). The results show that the natural binocular co-ordination of vertical memory-guided saccades is good comparable to that of vertical visually-guided saccades : the group mean disjunctivity was 0.37 ±0.19° n = 9 subjects. During training with disparate targets, in 6 of the 18 individual cases (9 runs up or down saccades), saccades developed disjunctivity. The group mean difference from the before value was 0.40 ±0.12°, n = 6 cases. After training, for five from the 18 cases, the disjunctivity persisted even though the remembered target was no longer disparate: the mean difference between, the before and after training condition was 0.53 ±0.21°, n = 5 cases. These results indicate that vertical disparity can be memorised and can modify the conjugacy of vertical saccades. It should be emphasised, however, that the disjunctive oculomotor learning based on the memorisation of vertical disparity is more subject-dependent and smaller in amplitude than that found for horizontal memory-guided saccades (Kapoula et al., 1998). Perhaps, memory based disjunctive oculomotor learning can be reinforced, in conditions where vertical disparities occur naturally (e.g. proximal and tertiary gaze ).