Contextual modulation in primary visual cortex

Abstract

We studied extra-receptive field contextual modulation in area V1 of awake, behaving macaque monkeys. Contextual modulation was studied using texture displays in which texture covering the receptive field (RF) was the same in all trials, but the perceptual context of this texture could vary depending on the configuration of extra-RF texture elements. We found robust contextual modulation when disparity, color, luminance, and orientation cues variously defined a textured figure centered on the RF of V1 neurons. We found contextual modulation to have a spatial extent of approximately 8 to 10 degrees diameter parafoveally. Contextual modulation correlated with perceptual experience of both binocularly rivalrous texture displays and of displays with a simple example of surface occlusion. We found contextual modulation in V1 to have a characteristic latency of 80-100 msec after stimulus onset, potentially allowing feedback from extrastriate areas to underlie to this effect.

Fig. 1.

Example texture displays. a, Illustration of an orientation-defined figure. b, Illustration of a homogeneous texture display. Texture in the center of the orientation-defined figure is identical to texture at the corresponding position in the homogeneous texture display. Typically, the luminance of the gray background was 24 cd/m² and that of the black bars was 6 cd/m², although we saw no evidence that a particular contrast was critical. Furthermore, our results do not seem to depend on the exact texture distribution. Nonetheless, we generally used texture bars 0.5° in length with the pattern illustrated in this figure. The gray between texture bars was the same as the gray that covered the screen in the intertrial period.

Fig. 2.

Extra-RF contextual modulation for orientation-defined texture figures. a, Configuration of the fixation spot, figure, and RF. The RF is completely enclosed by the figure contour. b, Illustration of the response of a multiunit site to stimulation with the homogeneously textured display flashed on a gray background for 267 msec. c, Illustration of the response of the same site when a 3.6° wide orientation-defined figure was flashed on in randomly interleaved trials. The initial response is nearly identical, but the tonic phase of the response is elevated in this condition compared withb. The response profile for the homogeneous texture display is shown in composite for comparison (fine-line waveform), and gray shading highlights the positive difference in response. d, Comparison of the average responses of all 75 recording sites for which we have quantitative data for stimulation both with RF textureand with extra-RF texture alone. Extra-RF texture alone gave at best an extremely weak response. e, Histogram of extra-RF contextual modulation ratios for the orientation display. This ratio is defined by the average response to the figure display divided by average response to the homogeneous texture display. Average response rates were measured in the interval of 100–250 msec after stimulus onset (thereby ignoring the initial transient response). Ratio values > 1.0 indicate larger responses to figure displays. Single- and multiunit sites were qualitatively and quantitatively similar (see text for details).

Fig. 3.

Extra-RF contextual modulation for diverse figure-defining cues. The responses of one isolated parafoveal V1 neuron (cell a) are illustrated in this figure; quantitative description of this cell’s responses appears in Figure 5.a, Illustration of the responses to the homogeneous texture display. In each of the following conditions (b–f), the pattern and disparity of RF texture were identical to that in the corresponding region of the homogeneous texture display. b, Illustration of the response when the RF is centered in a texture figure 3.6° wide, defined by binocular disparity. The disc figure appeared at zero disparity, the background texture at 0.14° far disparity. The neuron’s initial response to this display was nearly the same as to the homogeneous texture display. Yet after the initial response, the disparity-defined disc evoked significantly more vigorous responses.c, Illustration of the response when the disc figure was defined by chrominance cues; in this condition, the space between texture elements in the background was a green [CIE coordinates (x,y) = (0.344, 0.486)] equiluminant to the gray between texture elements in the disc [gray CIE coordinates (x,y) = (0.333, 0.333)], as confirmed with measurement by chrominance and luminance meters. We chose green (as opposed to, say, red or blue) solely to minimize chromatic aberration. The color-defined disc evoked a response very similar to the disparity-defined disc. d, Illustration of the response when the disc figure was defined by luminance cues; luminance of bars outside the disc was 43 cd/m², with the gray background the normal 24 cd/m² and the black bars the normal 6 cd/m². Again, the response of the V1 neuron was very similar to that for the disparity-defined disc. e, Illustration of the response to an orientation-defined disc. The cell here also showed elevated activity after the initial response as compared with the homogeneous texture display. f, Illustration of the response to a disc defined by each of the four preceding cues. The response magnitude for this “combination” display is not significantly different from that for discs defined by the four constituent cues. g, Illustration of the response for the disc-alone condition, in which the region outside the disc remained a constant gray throughout the trial. The response magnitude for this condition was not significantly different from the preceding five conditions.

Fig. 4.

Single-unit extra-RF modulation ratios for diverse cues. Activity measures, as always, are from 100 to 250 msec after stimulus onset. The first five histograms compile modulation ratios for various figure-defining cues for all 64 cells tested with each of the following: the disparity-, color-, luminance-, orientation-, and combination-defined figures. The last histogram compiles modulation ratios for the 43 neurons tested with the disc-alone condition that were also tested with the preceding five disc displays. The form of each distribution is similar (i.e., most cells have ratio values > 1.0). See text for details.

Fig. 5.

Variation of cue receptivity among single units for diverse cues. This figure deals with the 64 isolated V1 neurons tested with the homogeneous texture display (H) and the five common disc displays: disparity (D), color (C), luminance (L), orientation (O), and combination (Cb). We define a cue average variance index (CVI) as the standard deviation of a cell’s responses to disc displays in excess of the response to the homogeneous texture display, normalized by the response to the homogeneous texture display. For cell a (the same cell as in Fig. 3), with activity levels shown as a bar chart in theupper left of the figure, this corresponds to the SD of the heights of the gray portions of the response bars divided by the height of the leftmostbar (the homogeneous display response). We conservatively define a neuron to be cue-invariant in extra-RF contextual modulation if it has aCVI < 0.25 and shows significantly greater response to each of the five disc displays as compared with the homogeneous texture display (p < 0.05 for one-sided t test for each disc). Thecenter of the figure shows a pie chart that divides the cells into three classes: cells not significantly modulated by any of the five disc displays, cells that are cue-invariant, and cells with significant modulation that fall short of the cue-invariant classification. At the top and bottom of the figure are example cells.

Fig. 6.

Spatial extent of extra-RF contextual modulation tested with discs of variable diameter. a, Illustration of some sample responses from one multiunit site tested with variable diameter discs defined by luminance. b, Illustration of the entire diameter-response function for the same multiunit site. Response rates fall off essentially monotonically with disc diameter.c, Illustration of the median extra-RF modulation ratio as a function of disc diameter for all 84 sites tested (n = 65 tested with orientation-defined discs, 14 with luminance-defined discs, and 5 with color-defined discs). Extra-RF modulation declines monotonically with disc diameter, reaching the value 1.0 at ∼10° diameter. d, Illustration of the fraction of sites with significant modulation (p < 0.05, one-sided t test) as a function of disc diameter. The fraction of modulated sites reaches chance level at ∼8° diameter.

Fig. 7.

Extra-RF contextual modulation and binocular rivalry. a, Illustration representing examples of three types of binocularly rivalrous displays (case 1, case 2, and case 3) for each illustrating left- and right-eye images and an approximate representation of the cyclopean percept. See text for complete description. b, Illustration of the response of one multiunit site to these displays. c, Illustration of another multiunit example. d, Histograms of extra-RF modulation ratios for case 2/case 1 and case 3/case 1, with case 1 filling the role of the homogeneously textured display. See text for details.

Fig. 8.

Extra-RF contextual modulation and perceived distal structure. a, Configuration of a texture display in which a band of texture surrounding the RF may vary in apparent depth through binocular disparity cues. The default texture was typically at zero disparity, although we used far background disparity as the default in some experiments. b, Illustration of how this display may be configured to appear as a homogenous texture display, moat display, or frame display. Typically, the disparity offset of moat and frame was ±0.14°, although this value was not critical. Monkey binocular vision is similar to that of man (see Materials and Methods), and we assume that our monkey subjects perceive these displays as do human observers. c, Illustration of the responses of a multiunit site to the displays in b. For this experiment, RF texture was always at zero disparity.d, Illustration of responses of another multiunit site. For the data shown, RF texture was at 0.14° far disparity (with moat and frame moved back accordingly to preserve the relative depth arrangements). The contextual modulation pattern is the same as inc. The site in d also gave the same pattern of response when RF texture was at zero disparity (data not shown). e, Extra-RF contextual modulation ratios for 146 sites for moat (top) and frame (bottom) displays. See text for details.

Fig. 9.

Characteristic delay of extra-RF contextual modulation. In a two-step texture presentation procedure, we initially present the homogeneous texture display and then 150 msec later, change to the moat display by manipulating only extra-RF texture. As the RF is entirely within the moat-defined figure, it receives static RF texture stimulation whether or not the moat appears. In a, we compare average response profiles of 53 sites for the two-step moat presentation (trace M) and simple long-duration homogeneous texture (trace H). The responses are identical until ∼80 msec after the moat appears, after which the neural response rebounds for the moat condition. In b, we show the results of the analogous one-step moat experiment performed on the same 53 sites in randomly interleaved trials. Despite the different time course of RF stimulation, the timing of extra-RF contextual modulation is the same.

Fig. 10.

Comparison of response latencies of various visual areas and extra-RF contextual modulation in area V1. Extra-RF contextual modulation in V1 occurs well after the extrastriate cortical areas are initially activated. If we assume that feedforward and feedback pathways have the same time delay (Domenici et al., 1995), we see that the V1-inferotemporal cortex (IT) latency would permit IT signals to reenter V1 before extra-RF contextual modulation is fully expressed at the poststimulus onset time t = 100 msec.