A. Three stimulus sweeps, presented without interstimulus interval. Circles and arrows indicate the extent and direction of stimulus movement.
B. Simple reverse correlation. Once the latency has been determined, a response window for each stimulus sweep (shaded part of the histogram) is defined as starting at the stimulus onset+latency and extending for the duration of the stimulus sweep. Spikes in this window are then added to the bin corresponding to the location at the center of the stimulus sweep. Each bin keeps separate spike counts for each of the 8 tested directions. A vector sum of the response is computed at the end for each bin.
C. Spatial reverse correlation. The response is subdivided into temporal slices of 16 msec (in the figure, they are wider for graphical purposes), shown in gray. The stimulated region of the visual field is divided into a 10 by 10 grid of 100 locations. A bin is assigned to each location. Spikes are assigned to the bins corresponding to locations covered by the stimulus during a given temporal slice. Thus, for a stimulus moving downwards, spikes generated during the initial part of the stimulus presentation are assigned to bins corresponding to higher locations than the spikes generated towards the end of the stimulus presentation. As before, a vector sum is computed across the entire stimulus set at the end.
D. Temporal reverse correlation. The spikes collected during the stimulus presentation are subdivided in up to 50 partially overlapping temporal slices. Grids at the bottom here represent separate temporal frames that, when seen one after the other, show the changes of the response through time. Spikes are binned into different frames depending on the portion of the stimulus sweep during which they occur. Changes in the within-sweep location of the stimulus are ignored, as with simple reverse correlation.
Back to the cogsci homepage