Color-Change Detection Activity in the Primate Superior Colliculus

Abstract

The primate superior colliculus (SC) is a midbrain structure that participates in the control of spatial attention. Previous studies examining the role of the SC in attention have mostly used luminance-based visual features (e.g., motion, contrast) as the stimuli and saccadic eye movements as the behavioral response, both of which are known to modulate the activity of SC neurons. To explore the limits of the SC's involvement in the control of spatial attention, we recorded SC neuronal activity during a task using color, a visual feature dimension not traditionally associated with the SC, and required monkeys to detect threshold-level changes in the saturation of a cued stimulus by releasing a joystick during maintained fixation. Using this color-based spatial attention task, we found substantial cue-related modulation in all categories of visually responsive neurons in the intermediate layers of the SC. Notably, near-threshold changes in color saturation, both increases and decreases, evoked phasic bursts of activity with magnitudes as large as those evoked by stimulus onset. This change-detection activity had two distinctive features: activity for hits was larger than for misses, and the timing of change-detection activity accounted for 67% of joystick release latency, even though it preceded the release by at least 200 ms. We conclude that during attention tasks, SC activity denotes the behavioral relevance of the stimulus regardless of feature dimension and that phasic event-related SC activity is suitable to guide the selection of manual responses as well as saccadic eye movements.

Keywords: change detection; color; spatial attention; superior colliculus.

Figure 1.

Attention task procedure, stimuli, and behavior. A, Procedure: the monkey was required to hold down the joystick and maintain central fixation throughout the trial, releasing his hold only for cued stimulus changes. The cued stimulus patch was presented in the location previously occupied by the flashed cue-ring (the bottom right of the screen, in this example), and the foil was presented at an equally eccentric opposing location (top left). Only one stimulus changed per trial (see Materials and Methods). B, Hit rates (HRs) and false-alarm rates (FARs) for monkey 1 and 2 in each session; percentages are across-session binomial parameter estimates and horizontal lines are 95% confidence intervals. C, Trial timeline illustrates the time window in which a color change was possible relative to stimulus presentation; stimuli were presented for 1 s beyond the time of the color change. D, Stimulus colors were drawn from an isoluminant plane of the DKL color space. Axes were scaled to the [−1, +1] interval with limits representing the maximum possible contrast for the display (see Materials and Methods). Numbered arrows (not to scale) represent color changes in saturation increase (1, 2, 3, 4) and saturation decrease (5, 6) trials; saturation decreases were approximately two to four times larger than saturation increases and were arranged so that the mean saturation was the same following either an increase or a decrease. E, Example stimulus frames before (left) and after (right) saturation increases (1, 2, 3, 4) and decreases (5, 6). Note that after a blue increase (1) or a blue decrease (5), the saturation was the same, as mentioned above; the same was true for yellow increase (2) and decrease (6). F, Example saturation distributions for stimuli before and after a green saturation increase; before increase, Gaussian distribution has (mean, SD) of (0.1, 0.05) and (0.15, 0.05) after increase.

Figure 2.

Example unit and population-level color preferences between color-pairs. A, Example unit average firing rate traces aligned on stimulus onset (left) and color change (right), for yellow and blue stimuli. Significant preferences for one of the two colors in each bin is indicated by the row of colored boxes below; significance was determined using bootstrapped ROC areas comparing spike counts from nonoverlapping 20 ms bins (see Materials and Methods). Right panel inset shows a time-expanded view of the outlined rising portion of change-related phasic activity. B, Population-level percentage yellow/blue preference (n = 139): the percentage of individual units with a significant preference in each bin; the row of colored boxes below mark bins in which the proportion of units preferring one of the colors was significantly greater than the proportion preferring the other as measured by χ²-proportion tests. C, Example unit average firing rate traces for red and green stimuli (conventions as in A). D, Population-level percentage red/green preference (n = 31): conventions as in B.

Figure 3.

The strength of cue-related modulation varies by unit-type category in distinct ways from color-related modulation. A, Population average firing rates for cued stimulus (blue) and foil (yellow) aligned on stimulus onset (left) and change (right), for V (n = 42; far left), VM (n = 45; middle), and VMp (n = 46; far right) categories. B, Population average AMI by category: for each unit, average spike count in nonoverlapping 20 ms bins for the cued stimulus (c) and the foil stimulus (f) was used to calculate AMI = (c − f)/(c + f); individual AMIs were then averaged to obtain population AMI. Gray shaded regions mark time windows used for ANOVA (see Results). C, Population-level percentage cue/foil preference by category: the percentage of units in each bin with significantly larger spike counts for the cue (blue) or the foil (yellow); significance was determined with bootstrapped ROC area. Shaded regions indicate time windows used for ANCOVA. D, Population-level percentage yellow/blue preference by category: the percentage of units in each bin with significantly larger spike counts for the yellow or the blue stimulus determined with bootstrapped ROC area. Because of the similarity of individual neuron responses to increases and decreases, data from increase trials and decrease trials were pooled.

Figure 4.

Peak change-evoked activity is comparable to onset-evoked activity. Data from V, VM, and VMp neurons only (n = 113). A, Change versus onset peak activity when the cued stimulus was in the RF. Onset peak is the maximum average firing rate during the 50–200 ms after stimulus onset; change peak is the maximum during the 100–300 ms after stimulus change. Inset histogram is the distribution of distances from the dashed identity line. Data from increase trials and decrease trials were pooled.

Figure 5.

Change-detection activity is comparable for saturation increases and decreases. A, Population (n = 22) average firing rates for saturation increases (blue) and decreases (gold). Inset is a scatter plot of maximum change-detection activity (in the 100–300 ms after change). B, Population percentage preference for increases/decreases: the percentage of units with significantly larger spike counts for increases or decreases in each 20 ms bin; significant individual unit preferences were determined by bootstrapped ROC area. C, Individual unit preferences for increases or decreases sorted by the total number of bins with significantly greater spike counts for saturation increases.

Figure 6.

Change-detection activity is predictive of behavioral choice. A, Population average firing rate traces for hits (blue) and misses (gold). After averaging spike trains convolved with a waveform resembling a postsynaptic potential (see Materials and Methods), activity in nonoverlapping 20 ms bins was averaged for comparison with preferences in B and ROC area in C. B, Population percentage hits preference: the percentage of units in each bin with significantly larger spike counts in hit trials versus miss trials; significance was determined by bootstrapped ROC area. C, Population average ROC area or detect probability: average ROC area from the bin-by-bin comparison of spike counts for hits to those for misses. Data from increase trials and decrease trials were pooled.

Figure 7.

Timing of change-detection activity predicts jRT. A, Example unit average firing rate for hits falling into each of three jRT quantiles: slow (yellow), medium (green), and fast (blue). Colored arrows indicate the median in-quantile jRT. B, Example hinge model fits to the data in A (solid colored lines; see Materials and Methods). Colored Xs near hinge points indicate +2 SD from baseline, while those near the plateau of activity indicate −1 SD from the peak; the yellow X and connected gray line are the criterion firing rate value (minimum of the three −1 SD points) used to calculate threshold times, which are indicated by larger dashed lines. C, To selectively determine the effect of differences in hinge model slope on threshold time, hinge models were aligned at hinge points and the “slope time” time at which each crossed the criterion firing was found (dashed lines). D, To determine the effect of differences in baseline firing rate on threshold time, the “baseline time” required to go from each quantile’s baseline offset to an across-quantiles baseline mean was determined (horizontal double-arrowhead lines). E, The relationship between threshold time and median jRT for each unit; this analysis follows the method of Thompson et al. (1996). Data from the example unit shown in A is plotted in darker gray. F, Histogram of the proportion of jRT accounted for by threshold time. Each unit’s proportion is defined as the average slope of the two line segments belonging to that unit in E, so that each neuron contributes one value to the distribution in F; the red diamond on the abscissa is the proportion for the example unit in A. G–I, Proportion of threshold time accounted for by start-of-rise time (G), slope time (H), and baseline time (I): each unit’s proportion is the average slope of the relationship between the quoted time variable and median jRT; the red diamond is the example unit’s value. Data were pooled across increase trials and decrease trials.