Evolution of directional preferences in the supplementary eye field during acquisition of conditional oculomotor associations

Abstract

We assessed the preferred directions (PDs) of supplementary eye field (SEF) neurons during conditional visuomotor learning. Monkeys learned to select one of four saccadic eye movements in response to a foveal instruction stimulus (IS). ISs were either familiar or novel. Each familiar IS reliably evoked one saccade: 7 degrees left, right, up, or down form the central fixation point. Novel ISs initially triggered virtually random responses among those four possibilities, but the monkeys ultimately learned to select the instructed saccade. As reported previously, activity rates on novel IS trials significantly changed during learning. Some of these cells (learning-dependent) also have significant modulation on familiar IS trials, but others (learning-selective) lack such activity. Of the former, the familiar IS activity can be either directionally selective or omnidirectional. For most neurons, PDs were apparent during all phases of learning, but they were rarely constant. Only infrequently did a neuron's PD for novel ISs closely match that for familiar ISs throughout the learning process. In directional learning-dependent cells, the PD usually reoriented near the end of learning to resemble that for familiar IS trials. In omnidirectional cells, initially evident PDs dissipated with learning, even as the cell became more strongly modulated. Learning-selective cells typically began with significant PDs, but became unmodulated as learning progressed. Our findings show a pervasive lability in SEF PDs that may reflect a flexible and rapid remapping between inputs and responses within the premotor cortical network.

Fig. 1.

Conditional oculomotor learning task. Schematic drawing of the video monitor and the monkey’s oculomotor responses.A, The monkey maintained gaze at the central fixation point (not shown) for 500 (or occasionally 600 or 800) msec. A visual instruction stimulus was presented for 500 (rarely 600) msec, followed by 1.5–3.0 sec delay period, while the fixation point remained on (left). At the trigger stimulus (fixation point off), the monkey made a saccade to one of the four targets and maintained fixation at the target for 600 msec (right). B, The task periods and durations (arrows, top) and a schematic of horizontal eye position (Eh,bottom).

Fig. 4.

A, B, Three-point moving average of performance (open squares) and activity (filled circles) plotted for novel IS trials, separated for the downward (A) and rightward (B) saccades. From the same cell as Figure 3. Dashed horizontal lines indicate ±1 SD of the mean activity for the familiar IS trials. The criterion performance trial is marked as trial 0. C, Activity for familiar IS trials during different learning phases, and a polar plot (right) showing the mean activity and ±1 SD for each saccade direction. In the polar plot, the circle represents the average activity during the reference period. D, Activity modulation for different saccade directions and the novel ISs associated with those saccades for each phase of learning. At thebottom of C and D, the mean vector for each learning phase is shown (bold arrow), as well as the familiar IS mean (solid radius) and 99% confidence limit (dashed radius) for the entire block. The vector length is scaled with reference to that of the novel PD during the established phase. Open circles, Down; filled squares, right;open squares, left; filled circles, up. Each value represents the average activity in a given learning phase.fml, Mean activity on familiar IS trials; ref, mean activity ±1 SD in reference period.

Fig. 2.

Mean reaction times of the second monkey for leftward (A), rightward (B), upward (C), and downward (D) saccades. Data are taken from a representative sample of behavior during 15 recording sessions, balanced over early, middle, and late data collection sessions. Correct response trials are aligned on the trial of criterion performance (trial 0, dashed line), which is the first instance of three consecutive correct responses. Error bars indicate means ± 1 SD. fml, Familiar IS trials (asterisk).

Fig. 3.

SEF cell with learning-dependent activity in the instruction period. A–D, Histogram and rasters showing the activity evolution when four saccade directions were instructed by four different novel ISs. Only data from correctly executed trials are shown, in the order of their occurrence (for each movement direction) from top (first) to bottom (last). Thin arrows mark the trial on which criterion performance was achieved (criterion). IS, Instruction stimulus onset;x, IS offset; TS, trigger stimulus;acq, target acquisition; rew, reward. Note that the familiar IS trials and these novel ISs were interleaved pseudorandomly in the block with several other novel and familiar ISs. Activity scale in impulses/sec, the same for all plots.

Fig. 5.

Directional tuning of two learning-dependent, directional SEF cells. A, Polar plots of the directional tuning of a cell, instructed by the familiar IS (top left) and the directional tuning in each learning phase for novel IS trials, plotted from lower left to upper right in a row along an arrow. Data were obtained in postsaccadic period. The bold arrow in each learning phase indicates the length and angle of the mean vector. The vector length, ranging from 0 to 1, is scaled relative to the outer circle of each plot. The activity scale is identical for all parts of A, and vectors with a length <0.2 are not plotted. Note that the mean vectors in the late and established learning phases point in the direction similar to that of the familiar IS trials. B, Directional tuning of a different learning-dependent, directional SEF cell. Note that this cell decreases its modulation for the nonpreferred directions rather than increasing modulation for PDs. Data were obtained in the instructed delay period. Format as in A.

Fig. 6.

PD evolution during different learning phases. Data shown are from learning-dependent cases with a significant PD for familiar IS trials. A, Distribution, for each case, of the absolute value of the angular difference between the PD for familiar IS trials and that for novel IS trials in each learning phase. Thehatched bars show the cases with absolute angular differences of ≤30°. B, DI, in each learning phase, for the direction of saccade that is associated with the strongest modulation in familiar IS trials (familiar-maximum direction).C, DI, in each learning phase, for the direction of saccade associated with the least activity in familiar IS trials (familiar-minimum direction). fml, Familiar; nvl, novel; PD, preferred direction.

Fig. 7.

Correlation between the directional tuning on familiar IS trials and that of early, middle, late, and established phases for novel IS trials. Box plots show the median (solid line) and the mean (dashed line) for each phase, confined by the 25th and 75th percentile. The capped lines indicate the 10th and 90th percentiles.

Fig. 10.

Magnitude of directional bias based on mean vector length (A and C) and proportion of cases with directional biases during different learning phases (Band D). Data from learning-dependent cases lacking PDs for familiar IS trials (A and B) and those from learning-dependent cases with PDs on familiar IS trials (Cand D) are displayed separately. Box plots show the median (solid line) and the mean (dashed line) for each phase, confined by the 25th and 75th percentile. The capped lines indicate the 10th and 90th percentiles. Bar charts (B and D) illustrate the proportion and number of cases showing a significant PD during each learning phase based on a cutoff criterion of 0.2, a value that separated directional cases from omnidirectional ones for the familiar IS trials.mid, Middle phase of learning to respond to initially novel ISs; fml, familiar IS trials.

Fig. 8.

Activity modulation of an omnidirectional, learning-dependent SEF cell. Format as in Figure 4, except partC shows only a polar plot of block averages. Data obtained from the postsaccadic period.

Fig. 9.

PD evolution of three SEF cells with learning-dependent, omnidirectional activity. Data for each cell is from a different task period: the postsaccadic (A), target-hold (B), and presaccade (C) periods, respectively. Note that, regardless of the strength of modulation, transitory PDs emerged during the early phases of learning. Format as in Figure 5.

Fig. 11.

Activity modulation of a learning-selective SEF cell. Format as in Figure 4. For the novel IS associated with rightward saccades, the monkey had two distinct phases of learning, as shown inC. Asterisks in C and Dindicate the trials that are normalized to the second attainment of criterion performance for rightward saccades. Dashed horizontal lines indicate +1 SD of the mean activity for the familiar IS trials, for the same saccade direction. −1 SD lines, not shown, are <0 impulses/sec. Data were obtained from the instructed delay period.

Fig. 12.

Polar plots for a learning-selective SEF cell, with activity modulation during instruction period (A), instructed delay period (B), and postsaccadic target hold period (C), respectively. Note that the PD can be dramatically different in different task periods. Format as in Figure5.

Fig. 13.

A, Lateral view of the cortical surface from the second monkey examined in this study. B, Proportional distribution of directional (open circles) and omnidirectional (plus signs) cases of learning-dependent activity. C, Distribution of learning-selective cases (open circles). In B and C, the size of the symbol is proportional to the number of cases in each class.Ar, Arcuate sulcus; Pr, principal sulcus.