Hysteresis and slow drift in abducens unit activity.

Two trained monkeys made saccadic eye movements to a small visual target. The activity of 39 isolated abducens units, presumed to be motoneurons or abducens internuclear neurons, was recorded in relation to these eye movements. After a calibration trial, a test trial repeatedly elicited 20 degrees horizontal saccades to primary position from either the left or right. On average, the steady-state firing rate at primary position depended on the direction of the saccade. For saccades where the neuron showed a burst in activity during the saccade (on-saccades) the steady-state firing rates were usually higher than for those saccades that showed a pause in activity during the saccade (off-saccades). For the population of units this hysteresis measured 5.4 spikes/s, which may be compared with an average primary-position rate of 97 spikes/s. The average hysteresis for individual units ranged from -2.1 to 18.5 spikes/s. The steady-state firing rate after equal saccades in the same direction and ending at the same position (primary) varied slowly over time. Across all units the variability (standard deviation) ranged from 0.5 to 11.8 spikes/s with a mean of 4.7 spikes/s. Furthermore, for any one unit the variations following on-saccades generally correlated with the variations following the off-saccades. Hysteresis, doubted by many, does exist. Fortunately, it is small enough, 5.5% of typical primary-position rate, that it can be neglected for many purposes. Nevertheless, it poses the interesting theoretical question of how the oculomotor system compensates for hysteresis. The simplest explanation of slow variations in background rate is cocontractive noise: a slow fluctuation in all abducens neurons so that these variations do not result in fluctuations of eye position.