Macaque V1 activity during natural vision: effects of natural scenes and saccades

Abstract

In the present study, we examined the way that scene complexity and saccades combine to sculpt the temporal response patterns of V1 neurons. To bridge the gap between conventional and free viewing experiments, we compared responses of neurons across four paradigms ranging from less to more natural. An optimal bar stimulus was either flashed into a receptive field (RF) or brought into it via saccade and was embedded in either a natural scene or a uniform gray background. Responses to a flashed bar tended to be higher with a uniform rather than natural background. The most novel result reported here is that responses evoked by stimuli brought into the RF via saccades were enhanced compared with the same stimuli flashed during steady fixation. No single factor appears to account entirely for this surprising effect, but there were small contributions from fixational saccades and residual activity carried over from the previous fixation. We also found a negative correlation with cells' response "history" in that a larger response on one fixation was associated with a lower response on the subsequent fixation. The effects of the natural background and saccades exhibited a significant nonlinear interaction with the suppressive effects of the natural background less for stimuli entering RFs with saccades. Together, these results suggest that even responses to standard optimal stimuli are difficult to predict under conditions similar to natural vision, and further demonstrate the importance of naturalistic experimental paradigms to the study of visual processing in V1.

Figure 1

Stimuli and behavioral task. (A) A small optimally-oriented bar was placed within the classical RF and superimposed upon either a uniform gray (top) or natural image (bottom) background. For natural background conditions, the bar was surrounded by a gray buffer patch. Stimulus elements are not shown to scale. In the experiments, the stimulus bar was roughly the size of a small tree branch in the image. (B) Timing comparison between flash (top) and saccade trials (bottom). In both conditions the animal was first required to fixate an initial peripheral fixation point (FP), and then saccade to the central FP when it appeared. In flash conditions the bar stimulus appeared in the RF during fixation; in saccade conditions it was brought into the RF with the saccade to the central FP.

Figure 2

Statistics of single-unit firing during free-viewing and optimal bar presentation with fixation. (A) Distributions of firing rates for one cell. Firing rates were computed in 50 ms windows placed at 10 ms increments across the response. The hatched histogram shows the response distribution for free-viewing and the plain histogram shows responses to flashed optimal bars. (B) Firing rates in flashed-bar and free-viewing conditions for all cells, averaged across all window positions. Error bars are standard error. (C) Comparison of peak firing rates across the cell population in flashed-bar and free-viewing conditions. Peak rates are the highest firing rates observed in any 50 ms window across all trials, for each cell. Symbol sizes are proportional to number of cells plotted at each location. (D) Comparison of average free-viewing activity to the peak response to a flashed bar. In more natural visual situations, V1 activity is well below the response rates commonly reported in experiments with optimal stimuli. In figures B -D, solid lines are diagonals.

Figure 3

Comparison of responses to bar stimulus on natural image and uniform gray backgrounds. (A) Single-unit responses to a bar flashed on the gray and natural image backgrounds during fixation (top) and to an identical bar appearing in the RF via saccade (bottom). Dark lines are mean responses across 15 trials; light lines are mean ± s.e.m. (B) Population responses (normalized for 59 neurons) to a bar presented on natural and gray backgrounds during fixation (top) and with a saccade (bottom). With either flash or saccade presentation modes, responses were generally less on a natural background.

Figure 4

Comparison of responses elicited by a flashed bar with a small buffer (dashed trace), large buffer (solid trace), and full screen gray background (bold trace). Data are averaged normalized responses for 19 neurons. The small buffer was twice the RF diameter and the large buffer was four times the RF diameter. Much of the surround suppression comes from a distance greater than four times the RF size.

Figure 5

Comparison of responses to a bar flashed in the RF or entering via saccade. (A) Single unit saccade and flash responses to a bar presented on the natural image background (top) and gray background (bottom). (B) Population responses to flash and saccade stimuli on natural image (top) and gray background (bottom). Traces represent normalized responses for 59 neurons. Responses to the identical stimulus were generally higher in saccade conditions.

Figure 6

Interaction between surround suppression and presentation mode. The bold curve shows the average response to a bar flashed on a uniform gray background. The other two curves contrast the average responses to the same bar on a natural scene background with a small (2×RF) buffer presented either via flash (dashed curve) or saccade (intermediate solid curve). In both flash and saccade conditions, the natural background beyond the buffer suppresses the response. In the saccade condition, the response with the small buffer patch is greater than the response with the same buffer in the flash condition. The small-buffer saccade response is comparable to the flash response with a much larger (4×RF) buffer (compare Figure 4). These data suggest that there is less surround suppression near the RF in the saccade condition.

Figure 7

Contribution of fixation saccades to bar responses. (A) Absolute value of eye velocity for saccade and flash conditions with natural background, averaged across all trials in all recording sessions. Absolute values were taken before averaging. The vertical scale is expanded and the saccade curve is cropped so that details can be seen at later times. (B) Raw single cell responses for saccade (dashed trace) and flash conditions. (C) The same cell response as in (B) after discarding spikes following fixation saccades by less than 100 ms. Eliminating spikes possibly resulting from fixation saccades does not abolish the larger response in the saccade condition.

Figure 8

Preferred orientations of recorded cells. The dashed distribution shows preferred orientations for cells with a significant difference between saccade and flash conditions. The solid distribution represents cells without a significant mode difference. There is no apparent correlation between preferred orientation and significant mode effects, suggesting that the difference between flash and saccade responses does not result from movement of the bar across the RF during saccades.

Figure 9

Relationship between residual activity and responses in saccade conditions. (A) Flash and saccade conditions were identical in the light-shaded time interval except that the bar stimulus was not yet presented in flash conditions. Spike counts in the dark-shaded interval were used to quantify residual activity possibly carried over from the previous fixation and the saccade that acquired the final fixation point. The dashed lines indicate when the bar stimulus appeared in the RF. (B) Each data point represents the saccade condition response for one cell, normalized to the cell’s bar response in the flash condition. Two points are plotted for each cell, one corresponding to rightward saccades, and the other to leftward saccades. Open symbols denote cells with significant differences between saccade and flash bar responses. The dashed line shows the best linear fit to all data points. There is a tendency for neurons with higher residual responses to have higher responses to the bar in the saccade condition.

Figure 10

Contribution of response history to bar-evoked responses. (A) Relationship between response history and left/right response differences. Each data point relates the response margin between left and right bar responses (x-axis) with left and right pre-bar responses (y-axis). The upper and lower insets show examples of RF contents during penultimate fixations to the left or right, respectively, of the final fixation point. Two points are plotted for each cell in the population, one each for saccade and flash conditions. The dashed line shows the best linear fit through all data points. Cells with higher responses on the penultimate fixation tend to have lower responses to the final bar stimulus. The gray boxes highlight the disproportionate number of points in the upper left and lower right quadrants. (B) Relationship between response history and saccade/flash response differences. Each data point relates the RM between saccade and flash bar responses (x-axis) with saccade and flash pre-bar responses (y-axis), for conditions with the natural image background. Two points are plotted for each cell in the population, one each for trials started with saccades from the left and right. Open symbols are plotted when saccade and flash bar responses were significantly different. The dashed line shows the best linear fit through all data points. There is a mild but highly significant correlation such that greater response differences on penultimate fixations in flash and saccade conditions are associated with smaller response differences to the bar stimulus on the final fixation.

Figure 11