Fig. 3.

Persistent activity bump in the C-ring. a Top: A visual cue (yellow bar in the black visual field) is modeled by the spike input to two corresponding PB regions (see Methods). Bottom left: The visual cue elicited firing activity (represented by graded cyan) in several neurons and the responses persisted after the cue offset. Bottom right: A stable activity bump can be observed in the EB region-based firing rate plot. The red line is the estimated center of the activity bump, determined by the peak position of the Gaussian curve fit to the activity bump in each time step. b When we presented two visual stimuli with different brightnesses (top), the C-ring circuit developed only one activity bump in the EB (bottom) due to the nature of the neural competition. The thick and thin red ticks mark the EB regions corresponding to the stronger and weaker visual cues, respectively. c We investigated the response of the C-ring to a visual cue that moves across the visual field (top). Bottom left: the cue first moved clockwise for 16 s with a speed of 22.5° s⁻¹ and then reversed direction. Bottom right: the cue moved with a double speed for 8 s in each direction and then stopped. The circuit was able to accurately track the moving visual cue at a speed of 22.5° s⁻¹. The quality of the activity bump and its ability to track the moving cue (at 22.5° s⁻¹) are indicated by d the deviation between the bump center and the cue location, and by e the bump width (FWHM). The error bars indicate the within-trial s.d. f The activity bump is stable and does not shift in the dark during simulated body rotation (CW = clockwise, CCW = counterclockwise). g The height (red) and width (FWHM) (green) of the bump following stimulus offset were independent of the strength (top) and duration (bottom) of the stimulus, suggesting that the activity bump only encodes the direction of the stimulus, not its other non-spatial properties. Each data point was averaged over 10 trials and the error bars indicate s.d