Spatial attention and the latency of neuronal responses in macaque area V4

Abstract

The effects of attention on neuronal responses in visual cortex have been likened to a change in stimulus contrast. Attention and stimulus contrast both modulate the magnitude of neuronal responses. However, changes in stimulus contrast also affect the latency of visual responses. Although many neurophysiological studies have examined how attention affects the strength of neuronal responses, few have considered whether attention affects neuronal latencies. To compare directly the effects of stimulus contrast and attention, we recorded responses from individual neurons in area V4 of macaque monkeys while they performed a task that independently controlled spatial attention and stimulus contrast. As expected, changes in stimulus contrast affected both the magnitude and latency of neuronal responses. Although attention had the expected effects on the magnitudes of neuronal responses, we did not detect statistically reliable changes in neuronal latency. A direct comparison of the effects of contrast and attention revealed a reliable difference. When a shift in spatial attention decreased response magnitude, response latency increased much less than when the same magnitude change was caused by reducing stimulus contrast. Thus, attention is distinct from contrast in the way it affects the relationship between neuronal response magnitude and latency.

Figure 1.

Task and stimulus timing. One series of Gabor stimuli was centered in the receptive field of a V4 neuron, and a second was located at the same eccentricity directly across the fixation point. The location to be attended was signaled to the monkey by instruction trials. A series of Gabor stimuli was presented in synchrony at both sites (94 ms duration, 141–294 ms interstimulus period) and the monkey's task was to detect an orientation change in the series at cued location while ignoring orientation changes at uncued location. The contrast of each Gabor was selected in a random manner and different orientations were presented on different trials.

Figure 2.

Responses of two example V4 neurons. A, D, PSTHs for the responses of two cells to different stimulus contrasts and different attention states. The thick black line at the bottom of each histogram marks the stimulus duration. Each histogram was Gaussian filtered (8 ms SD). The dark gray area shows the response in the attended condition and the light gray area shows the response in the unattended condition. Horizontal solid lines show the mean response magnitudes (firing rate) during the sampling period in the attended condition, and dashed lines show mean response magnitudes in the unattended condition. Vertical lines show the corresponding response latencies in each condition. B, E, Plots of the relationship between response magnitude and latency for the same cells, parameterized by stimulus contrast and attention. Filled circles show response magnitude and latency at different contrasts in attended condition, and open circles show response magnitude and latency for the same stimuli in the unattended condition. Error bars are 95% confidence intervals from bootstrap. C, F, Some of the same data from B and E, plotted on an expanded time scale.

Figure 3.

Population contrast responses of V4 neurons (n = 78). The same format as in Figure 2, A and D, is shown. Histograms were smoothed with a Gaussian filter (4 ms SD). A, Population average responses of V4 neurons. The black arrow in the top panel points to elevated spontaneous activity caused by attention. Latency values for the attended and unattended conditions were as follows: 100% contrast, 62 and 63 ms; 25% contrast, 73 and 74 ms; 6.2% contrast 96 and 85 ms; 1.6% contrast, 209 and 144 ms. B, Normalized average responses of V4 neurons. The response of each neuron was normalized to its maximum response magnitude over all trials. Horizontal lines show normalized response magnitudes of the population. Latency values for the attended and unattended conditions were as follows: 100% contrast, 65 and 64 ms; 25% contrast, 76 and 77 ms; 6.2% contrast 96 and 86 ms; 1.6% contrast, 192 and 139 ms.

Figure 4.

Average latency–magnitude relationships for V4 neurons. Cells that satisfied bootstrap criterion were included in this plot (100% contrast, n = 62; 25% contrast, n = 54; 6.2% contrast, n = 16). The latencies for each cell were normalized to their minimum value and magnitudes were normalized to their maximum value. Each data point represents the sample average of normalized response latencies and magnitudes. Filled circles are data for the attended condition and open circles are data for the unattended condition. Error bars are SEs of the means. Gray regions represent the 95% confidence interval for regression lines from the bootstrap in each condition and the lines are the means of the bootstraps. The means and 95% confidence intervals for the regressions were as follows: attended slope, −1.15 (−0.89 to −1.44), intercept, 2.11 (1.82 to 2.42); unattended slope, −1.00 (−0.79 to −1.24), intercept, 1.83 (1.60 to 2.09).