Experimental paradigms. A, Speed-pulse experiment. The goal of this task was to detect a brief speed change. The monkey fixated on a colored fixation spot and two diametrically opposite colored annuli appeared on the screen. After 500 ms, random dot patches with 100% coherent motion appeared inside the annuli. The monkey's task was to release a lever in response to a transient (53 ms) increase in speed at either of the dot patches (the speed pulse). The color of the fixation point matched one of the annuli, indicating the likely location of the speed pulse. On 40% of trials the fixation point color would change midtrial indicating that the likely speed-pulse location had changed (data not shown). The monkey had to release a lever within 200–600 ms of the speed pulse to obtain a reward. B, Motion-step experiment. The goal of this task was to detect the occurrence of a coherent motion step. The monkey fixated on a central point and a static random dot patch appeared indicating the likely position of the coherent motion step. Afterward, two diametrically opposite random dot patches began moving at 0% coherence, with one patch at the location of the static cue. Zero percent coherent motion was shown for a random amount of time between 500 and 8000 ms (flat hazard function), followed by a coherent motion step lasting 750 ms. The monkey had to release a lever during the coherent motion step to obtain a reward. On 20% of the trials the coherent motion step occurred in the uncued patch. C, Motion-pulse experiment 1. The goal was to detect a brief coherent motion pulse. After the monkey fixated, 0% coherent motion began in two random dot patches located in the same hemifield, each one matched to the RFs and preferences of two neurons simultaneously recorded. After a random amount of time between 500 and 10,000 ms (flat hazard function), a 50 ms pulse was shown in one or both patches. Afterward, 0% coherent motion would resume. The monkey had to release a lever from 200 to 800 ms after the start of the motion pulse to obtain a reward. D, Motion-pulse experiment 2. The monkey initiated fixation, followed by 0% coherent motion for a random time between 500 and 10,000 ms in a single random dot patch. The monkey had to release a lever from 150 to 650 ms after a 33 ms pulse of coherent motion.
Main sequence analysis of the microsaccades collected from the three experiments. Most microsaccades appeared within the boundaries of the graphs (>80% for the 50 ms motion-pulse task, >97% for all others). The two-dimensional histogram of the peak velocity versus amplitude for each experiment is shown, with the number of microsaccades per bin indicated by grayscale values. Bins with zero frequency are plotted in white and the bin with the greatest frequency is shown in black. The two-dimensional histogram of the peak velocity (x-axis) versus the amplitude (y-axis) of each microsaccade collected during the speed-pulse experiment (A), motion-step experiment (B), 50 ms motion-pulse experiment (C), and 33 ms motion-pulse experiment (D). The value for the maximum bin is 100, 166, 176 and 241 for A, B, C, and D, respectively.
Three example trials from the same speed-pulse experiment. A, An example trial with three microsaccades (indicated by the asterisks). The black curve shows eye velocity that sharply peaks at the time of the microsaccades, while the neural activity is shown using both the raster (black dots) and the average spike rate (gray curve). Neural activity is suppressed after all three microsaccades. B, An example correct trial that contained no microsaccades. Neural activity increased after the speed pulse (vertical dashed line). C, An example failed trial from the same experimental session. A microsaccade occurred just after the speed pulse and resulted in a large reduction in neural activity. This example trial suggests a mechanism by which microsaccades might simultaneously affect both the neural response to a stimulus and the monkey's ability to perceive the stimulus.
The effect of microsaccades on the monkeys' perception. A, In the left panel, the microsaccade rates for correct trials (blue curve), failed trials (red curve) and all trials (black curve) relative to stimulus onset are shown for the speed-pulse experiment. A greater number of microsaccades occurred around stimulus onset during failed trials. In the right column, we plot the monkey's ability to detect the stimulus as a function of microsaccade time relative to stimulus onset. The panel shows that when a microsaccade occurs near stimulus onset, the ability to correctly detect the stimulus is reduced. B, Same as above, but for the motion-step experiment. Microsaccades after stimulus onset are also associated with a reduction in performance. C, Same as above, but for data pooled from both the 50 and 33 ms motion-pulse experiments. All graphs are averages across experimental sessions. Dashed lines show ±1 SE.
The effect of microsaccades on neural activity. Averages were computed for each brain area and experiment type. The left column shows the average neural activity aligned on microsaccades for microsaccades completed before test stimulus onset. Only the responses that occurred before the test stimuli (speed or coherence change) were used. The neural response for each neuron was normalized so that its baseline rate was equal to one. The right column shows the average neural activity in response to the test stimulus separated by whether or not a microsaccade occurred near stimulus onset. The test stimulus was either a speed change (A, B) or a coherence change (C–E). The gray curve shows the average activity given a microsaccade occurred from 200 ms before to 100 ms after stimulus onset for the pulse tasks, and from 100 ms before to 200 ms after stimulus onset for the step task. The black curve shows the average activity for trials with no microsaccades in these windows. Only neurons with at least 10 trials with microsaccades completed in these windows and 10 trials without were included. A, The average response across LIP for the speed-pulse task; B, the average response for area MT for the speed-pulse task; C, the average response for area MT for the motion-step task; D, the average response for area VIP for the motion-step task; E, the average response for area MT for the 50 ms motion-pulse task. Dashed lines show ±1 SE.
The partial correlations between microsaccades, neural activity, and perception. The partial correlation analysis was performed at 10 ms intervals relative to stimulus onset and averaged across all neurons for each condition. Perception was measured as either the behavioral response (correct = 1, failed = 0, shown in the left column), or the reaction time for correct trials (right column). The blue curve shows the partial correlation between neural activity and perception with the effects of microsaccades removed. The red curve shows the partial correlation between neural activity and microsaccades with the effect of perception removed. The green curve shows the partial correlation between microsaccades and perception with the effect of neural activity removed. The neural activity was the number of spikes in the previous 100 ms and the number of microsaccades from 100 to 200 ms before each point was used in the calculation. The horizontal black and gray bars give the windows used to count spikes and microsaccades, respectively, for Figure 7. A, Speed-pulse in LIP; B, speed-pulse in MT; C, motion-step in MT; D, motion-step in VIP; E, 50 ms motion-pulse in MT. Dashed lines show ±1 SE. μsac, Microsaccades.
The detect probability with and without the effect of microsaccades. The histograms show the difference between the detect probabilities calculated using all trials and using trials with microsaccades near stimulus onset excluded. Detect probability was computed either using all trials or removing trials where a microsaccade occurred near stimulus onset. Only neurons with 15 correct and failed trials were included in this analysis. For the speed-pulse and motion-pulse experiment, trials were removed if a microsaccade occurred from 200 ms before to 100 ms after the stimulus onset. For the motion-step experiment, the window was from 100 ms before to 200 ms after for C and D and from 200 ms before to 50 ms after for E. A, Speed-pulse in LIP; B, speed-pulse in MT; C, motion-step in MT; D, motion-step in VIP, late window; E, motion-step in VIP, early window; F, 50 ms motion-pulse in MT.