Inactivation of primate superior colliculus impairs covert selection of signals for perceptual judgments

Abstract

Primates base perceptual judgments on some sensory inputs while ignoring others. The covert selection of sensory information for perception is often thought to be accomplished mostly by the cerebral cortex, whereas the overt orienting toward relevant stimuli involves various additional structures such as the superior colliculus, a subcortical region involved in the control of eye movements. Contrary to this view, we show that the superior colliculus is necessary for determining which stimuli will inform perceptual judgments, even in the absence of orienting movements. Reversible inactivation of the superior colliculus in monkeys performing a motion discrimination task caused profound inattention for stimuli in the affected visual field, but only when distracters containing counterinformative signals appeared in the unaffected field. When distracting stimuli contained no information, discrimination performance was largely unaffected. Thus, the superior colliculus is a bottleneck in the covert selection of signals for perceptual judgments.

Figure 1

Selective attention task design. (a) Task sequence. After a brief fixation period, colored cue rings were presented. Stochastic motion patches appeared next, and then the cues disappeared. Following a delay, brief coherent motion pulses occurred in both the cued location (red arrow) and the diametrically opposite location (yellow arrow). When responding by saccade, monkeys reported the direction of the cued motion signal by making an eye movement to a response dot in the same direction; when responding by button push, monkeys pressed a button corresponding to the motion direction. (b) Normal behavioral task performance for both subjects in the saccade response version (top) and the button press response version of the task (bottom). Red dots represent proportion of correct choices (based on cued signal) in each session. Errors could be either driven by the foil signal (yellow dots) or by neither signal (gray dots). Scatter indicates variability across control sessions collected over several months before the inactivation experiments. Black lines, population averages; error bars, 95% multinomial confidence intervals; dashed line, proportion of responses consistent with the foil that would be expected by chance.

Figure 2

Map of inactivation effects. (a) Single session data from subject F. Black dots, saccade end points; interpolated color map, changes in peak velocity after muscimol injection. Cooler colors in the lower left quadrant indicate the decrease in peak velocity caused by SC inactivation; white contour delineates the affected region. Gray circles indicate the positions of the four stochastic motion stimuli, which were at fixed locations throughout the set of experiments. (b) Summary of SC injections for the saccade response task. Crosses, average end points of saccades evoked by microstimulation at the injection sites; shaded regions, extent of the visual field affected by each injection experiment.

Figure 3

Sample data from one inactivation session. (a) Behavior of subject F in a representative stimulus condition before inactivation. Schematic of stimulus indicates cued signal position with red ring and directions of motion in the cued signal (red arrow) and foil signal (yellow arrow). Red dots indicate end-points of saccade trajectories. A majority of responses were correctly directed by the cued signal. (b) Behavior in the same condition after inactivation of the SC. Schematic of stimulus now indicates the affected portion of visual space as a blue shaded region. Only a minority of responses were correctly guided by the cued signal; instead, the majority of decisions were based on the foil.

Figure 4

Summary results from inactivation sessions in saccade-response version of the task. The proportion of choices after injection is plotted against the proportion before injection. Red circles, correct choices matching the cued signal; yellow and gray symbols, errors driven by either the foil signal or neither signal. Error bars, 95% multinomial confidence intervals for each of the sessions, which included 176–264 trials per session. (a,b) Data for subject F. (c,d) Data for subject M. When the cued signal was in the affected region (a,c), subjects ignored this signal and instead based their choices on the foil. Conversely, when the foil signal appeared in the affected region (b,d), subjects tended to base their choices on the cued signal and ignore the foil.

Figure 5

Summary results from inactivation sessions in button-press version of the task. Same conventions as in Figure 4. (a,b) Data for subject F. (c,d) Data for subject M. When the cued signal was in the affected region (a,c), subjects ignored this signal and instead based their choices on the foil, much as in the results from the saccade-response version. The injection of saline during a control experiment (large, open symbols in c,d) produced no significant changes in performance.

Figure 6

The effects of SC inactivation on local motion discrimination. (a) Task sequence. After a brief fixation period, a stochastic motion patch appeared in one quadrant of the visual field. After a random delay (480 ms plus a geometrically distributed interval with mean 480 ms), a brief (160 ms) coherent motion pulse occurred. The direction of motion was drawn at random from any of the four diagonal directions. Monkeys reported the direction of motion by making a saccade to a response dot in the same direction. Data were pooled across sessions on the basis of the direction of motion with respect to the affected quadrant: ipsilateral, other quadrant on the same side; opposite, diagonally opposite quadrant on the other side; contralateral, directly opposite quadrant on the other side. (b–e) Performance on motion discrimination task for subject F for each of the four motion directions. Black circles, correct task performance in control sessions; red circles, correct performance after injection. Error bars, 95% multinomial confidence intervals. Solid lines, fits by multinomial logistic regression; dashed lines, 95% confidence intervals on the fits. Insets: bias (control, white; after injection, gray) and sensitivity (control, black; after injection, red) relative to choices into the quadrant containing the stimulus. Error bars, s.e.m. of the fitted parameters. After inactivation, bias significantly increased away from the injection site and sensitivity significantly decreased for all directions of motion. (f–i) Performance on motion discrimination task for subject M. Same conventions as in b–e.

Figure 7

Impairments in selective attention after SC inactivation required the presence of a foil signal. (a,b) Data for subject F. (c,d) Data for subject M. (a,c) After injection, monkeys tended to base judgments on the near-threshold foil signal in the unaffected region even when the cued signal in the affected region was set to maximal coherence. Individual symbols without error bars indicate performance on individual sessions; symbols with error bars indicate pooled performance. (b,d) When no foil signal appeared in the unaffected region, however, monkeys successfully ignored the three distracter stimuli with 0% motion coherence and based their choices on the cued signal.