1. We investigated whether neurons in the supplementary eye field (SEF) of macaque monkeys code saccadic eye movements in oculocentric coordinates (relative to the current direction of fixation) or in craniocentric coordinates (relative to the head). Craniocentric coding in SEF had been previously suggested by the convergent appearance of electrically elicited saccades originating at different orbital positions. 2. We primarily studied SEF neurons that started responding before the beginning of saccades because such presaccadic activity is likely related to saccade generation and metrics. Using a memory-saccade task, we classified the presaccadic activity of each neuron as either purely visual related, purely movement related, or both visual and movement related. 3. We then mapped the response fields (receptive fields and movement fields) of SEF neurons from different orbital positions. When mapped relative to a central fixation point, the strongest responses for a given SEF neuron invariably occurred for a particular polar direction with fairly symmetrical declines for departures from that direction. When tested using other fixation point locations, their strongest responses almost always continued to occur for stimuli having the same polar direction relative to each fixation point tested, and thus they appeared to code both stimulus direction and saccade direction in an oculocentric coordinate system. 4. The effect of eye position on SEF presaccadic activity was quantified in two ways by computing, for each neuron, 1) an ''intersection distance,'' the eccentricity of the point where extensions of the neuron's optimal polar directions measured at two eccentric orbital positions converged, and 2) an ''orbital perturbation index'' such that an index of 0 corresponded to no change in the neuron's optimal polar direction across different orbital positions (i.e., perfectly oculocentric response fields) and an index of 1 corresponded to optimal polar directions that converged to the same craniocentric goal regardless of initial eye position (i.e., perfectly craniocentric response fields). For neurons with both visual and movement responses, these measures were calculated separately for each type of activity using tasks that temporally separated the visual cue presentation and the saccade to it. 5. Almost all of the intersection distances were well beyond the oculomotor range (+/-50 degrees) of the monkey (38/39 for movement activity and 62/66 for visual activity). The median intersection distance for visual activity was very large (274 degrees), and the median for movement activity was slightly divergent (beyond infinity). Thus SEF neurons rarely showed a conspicuous convergence of response field direction. 6. Likewise, the mean orbital perturbation indexes were very small (-0.04 +/- 0.21, mean +/- SD, for movement activity and 0.09 +/- 0.15 for visual activity), also indicating that SEF neurons code stimuli and saccades in an oculocentric manner. 7. For neurons with both visual and movement activities, the orbital perturbation indexes of the two activities were not significantly correlated (r = 0.16), even though their characteristic directions (optimal polar direction estimated from the center of the screen) were almost the same (circular correlation, r = 0.97). The lack of a significant correlation between the visual and movement activity orbital perturbation indexes is consistent with the hypothesis that most of the variation in this index represents statistically independent errors of measurement. Conversely, the strong covariation of visual and movement activity characteristic directions indicates that directional preference is a fundamental functional property of SEF presaccadic activity. 8. The optimal visual target eccentricity and saccade size was also investigated for a smaller sample of SEF neurons. When mapped with visual targets having different eccentricities relative to a central fixation point, a minority of neurons responded maximally for a particular distance and declined for departures from that distance, both larger and smaller. When then remapped from different fixation points, such eccentricity-selective neurons continued to respond maximally for the same eccentricity and thus appeared to be oculocentrically coding stimulus eccentricity and/or saccade amplitude. The majority of SEF neurons tested did not exhibit a clear eccentricity preference, however, examination of their responses on eccentricity tests at different orbital positions indicated that they were not coding for particular craniocentric goals along the path being tested. 9. A smaller sample of neurons from the more laterally located frontal eye field (FEF) in the same monkeys was similarly studied, using the same paradigms and analyses. As in SFF, the initial position of the eye in the orbit generally had little or no effect on the optimal polar direction of FEF presaccadic activity. All of the FEF intersection distances (3/3 visual activity and 17/17 movement activity) were well beyond the monkey's oculomotor range (median 763 degrees), and the mean orbital perturbation index for presaccadic activity in FEF was -0.01 +/- 0.15. 10. Overall, these experiments indicate that SEF codes visual targets and saccades in an oculocentric manner and hence are not consistent with the hypothesis that the convergence of electrically elicited saccades observed in previous studies of SEF constitutes evidence of craniocentric coding. Instead, these data are consistent with the alternative hypothesis that converging elicited saccades are an artifact of the electrical stimulation technique. [We previously argued that converging electrically elicited saccades reflect the inability for punctate electrical stimulation to adequately engage the cerebellar circuitry that normally compensates for the changes in orbital mechanics associated with different directions of fixation, and we also showed that multiple saccade vector representations with orbital-dependent thresholds activated simultaneously through one electrode may also cause convergent perturbations.] Moreover, the similarity of these SEF results with the control experiments on FEF neurons, together with the comparability of SEF and FEF elicited saccade phenomenology that we previously reported, indicate that these reciprocally connected frontal lobe areas both generate saccadic eye movements in a common oculocentric coordinate system, and that the functional specializations that distinguish them must lie in other aspects of oculomotor processing.
