Figure-ground activity in primary visual cortex is suppressed by anesthesia

Abstract

By means of their small receptive fields (RFs), neurons in primary visual cortex perform highly localized analyses of the visual scene, far removed from our normal unified experience of vision. Local image elements coded by the RF are put into more global context, however, by means of modulation of the responses of the V1 neurons. Contextual modulation has been shown to follow closely the perceptual interpretation of the scene as a whole. This would suggest that some aspects of contextual modulation can be recorded only in awake and perceiving animals. In this study, multi-unit activity was recorded with implanted electrodes from primary visual cortex of awake, fixating monkeys viewing textured displays in which figure and ground regions were segregated by differences in either orientation or motion. Contextual modulation was isolated from local RF processing, by keeping RF stimulation identical across trials while sampling responses for various positions of the RF relative to figure and ground. Contextual modulation was observed to unfold spatially and temporally in a way that closely resembles the figure-ground percept. When recording was repeated, but with the animals anesthetized, the figure-ground related modulatory activity was selectively suppressed. RF tuning properties, however, remained unaffected. The results show that the modulatory activity is functionally distinct from the RF properties. V1 thus hosts distinct regimes of activity that are mediated by separate mechanisms and that depend differentially on the animal being awake or anesthetized.

Figure 1

Sequences of visual stimulation and resulting percepts for two varieties of figure-ground displays. (Upper) Structure from motion defines a square figure region of 4° width. Initially, the display is covered with a random dot pattern; 300 ms after fixation, a brief period of motion (30 ms) occurs in which dots inside and outside of a square region move in opposite directions, each dot moving 0.06°. After the motion, the random dot pattern remains stationary and contains no physical trace of the figure. The percept obtained from this sequence of stimulation is of a square figure that persists against the static random dot pattern for approximately half a second after the motion, if fixation is maintained. (Lower) Orientation contrast defines a square figure region. Initially, the display is covered with randomly oriented line segments; 300 ms after fixation onset, these are replaced by oriented texture in which line segments inside and outside of the 4° square region are orthogonal; this remains for 500 ms. The percept obtained from this sequence of stimulation is of a square figure that persists for the duration of the presentation. In the actual displays used in our experiments, background textures subtended 28° × 21° of visual angle, and thus were in fact more extensive than the ground texture shown in these examples.

Figure 2

Sampling of V1 neural responses with the RF located at various positions in the figure-ground displays. To avoid having results depend on local RF features such as texture orientation or direction of motion, we always used complementary pairs of stimuli with the same figure-ground relationships but opposing local features (Left); for a given position of the RF relative to the figure, the responses to these complementary stimuli were averaged. The results for structure from motion (Middle) and oriented texture (Right) are shown. For purpose of comparison, the response for the RF on ground is depicted in composite (thin lines) with edge and figure responses. The difference in response is indicated by gray shading. SEM of each response is given by the dotted lines above each plot. When the RF appeared on the edge or within the motion-defined figure, an elevation in response can be seen, relative to the response to ground, that faded approximately 500 ms after motion onset. When the RF appeared on the edge or within the orientation-defined figure, a persistent elevation in response, relative to the ground response, is observed for as long as the figure remains. The responses shown are the average from all 32 micro-electrodes implanted in area V1 of two awake, behaving macaque monkeys trained to maintain fixation during stimulus presentation.

Figure 3

Spatio-temporal representation of neural activity in response to figure-ground displays in awake monkeys. The figure-ground display was sampled at 15 positions (including the three shown in Fig. 2); samples were taken at 0.5° increments across the figure-ground display. For both the structure-from-motion display (Upper) and for the oriented-texture display (Lower), a uniform level of elevated activation within the figure region of the display evolves after the initial response. Before being combined in the population average, the responses from each electrode were normalized by the maximum of the responses evoked at the various positions of RF relative to figure. In this way, each electrode contributed equally to the average, independent of the absolute response level at the particular electrode.

Figure 4

Spatio-temporal representation of neural activity in response to figure-ground displays in anesthetized monkeys. The data are represented in the same format as in Fig. 3. The elevation in activity for figure versus ground, as observed in the awake animals, is no longer present for either structure-from-motion or oriented-texture displays.

Figure 5

Control experiments. (A) After the recordings in anesthetized animals, figure-ground modulation still could be recorded in the awake condition. Shown are average responses for the two types of stimuli with RFs either within the boundaries of the square figure (thick lines) or overlying background (thin lines). The gray shading indicates the difference between figure and ground responses, i.e., the figure-ground specific contextual modulation. (B) Figure-ground modulation in the awake animals is not caused by eye movements. Shown are the average eye velocities during fixation (Top) (eye movements were recorded during about 40% of the data collection resulting in Figs. 2 and 3), showing that the stimulus does not evoke any eye movements beyond the already present low level of micro-saccades within the fixation window. The second and third rows show the figure-ground responses either before or after correcting for eye movements. Correction consisted of removing those 50% of the trials where eye movements within the fixation window were largest. Figure-ground modulation does not appear to depend quantitatively on eye movements. The fourth row shows figure-ground responses for flashed stimuli (see Results), indicating that figure-ground modulation could not have been caused by eye movements over the textures generating responses. (C) Figure-ground modulation in the awake is not an effect of focal attention. Quantitatively similar figure-ground modulation is observed when two figures are presented (Lower) as when one figure is presented (Upper), whereas an effect of focal attention would be expected to split in about equal halves when two figures are presented, thereby diminishing figure-ground modulation for each figure. Format of all figures is: figure responses, thick lines; ground responses, thin lines; SEM of figure responses, dotted lines; prestimulus level of activity, short horizontal dashes; difference between figure and ground responses, gray shading. Not all of these control experiments were performed in both animals.