Computational and Neural Studies of Eye Movement

Eye movements provide many advantages for investigating neural mechanisms of both reflexive and voluntary behavior. Eye movements exhibit properties such as visual-motor coordination, learning, and memory that can be generalized to other systems. Yet, the neural machinery of eye movement is accessible for quantitative study and modeling. Our experiments are conducted in awake, behaviorally trained rhesus monkeys and entail quantitative assessment of motor performance, extracellular recording from single brain cells during eye movement, and computational analyses of models that are based on biological observations.

The VOR responds to head turns with compensatory eye movements that keep the eyes stable in space. Even in total darkness, the performance of the VOR is superb -- a head turn in one direction produces an equal amplitude smooth eye rotation in the opposite direction. If a monkey views the stationary environment through telescope spectacles, there are gradual "learned" changes in the eye movements evoked by head turns in darkness. After several days, the eye movements of the VOR in the dark are nearly appropriate for the visual conditions seen through the spectacles, demonstrating "memory" of the learned changes. In 1994, we published the results of a long series of experiments leading to the hypothesis that there are two sites of memory: one in the cerebellar cortex and one in the brainstem (Lisberger et al 1994a,b,c; Lisberger 1994). In collaboration with Mike Mauk, we then argued that the circuits for motor learning in the VOR have much in common with those for classical conditioning of the eyelid response (Raymond et al 1996). In 1998, we published the analysis of possible error signals to control modifications at both sites (Raymond & Lisberger 1998). This analysis revealed that different signals might be important for vestibular stimuli over different frequency ranges, and showed that a comparison of vestibular inputs and climbing fiber responses would be an effective error signal only if each climbing fiber response was compared with the vestibular inputs from 100 ms earlier. Ongoing projects include:

Elaboration of previous "lumped" models of the VOR into distributed models in which the simulated units have both cellular and discharge properties of neurons in the circuit.

• Ablation of the flocculus and ventral paraflocculus to determine whether these two cerebellar structures have separate functions during eye movements.