The activity of V1 neurons evoked by stimuli within the classical receptive field (CRF) is known to be modulated by stimuli in the surrounding field, the extra-receptive field (ERF). By varying the relative spatial phase (RSP) between a central grating presented in the CRF and a surround grating in the ERF, we studied the contextual modulation in V1 neurons of alert monkeys (Macaca mulata). Results from two monkeys show that most of the V1 neurons with suppressive ERF are sensitive to the RSP, and the degree of sensitivity is strongly dependent on the strength of ERF suppression. This sensitivity is maximal when the RSP is generated at or near CRF/ERF boundary, but is observed over the entire ERF. Interestingly, the suppressive effect of the surround grating can be largely abolished by inserting a narrow gap between the center and surround gratings or by a phase displacement between them corresponding to <10% of the CRF diameter. These properties of V1 neurons may serve important perceptual functions.
Previous studies have shown that the responses of V1 neurons to stimuli in the classical receptive field (CRF) are influenced by stimuli in the extra-receptive field (ERF) (Blakemore and Tobin, 1972; Maffei and Fiorentini, 1976; Nelson and Frost, 1978; Allman et al., 1985; Gilbert and Wiesel, 1990; Knierim and van Essen, 1992; DeAngelis et al., 1994; Li and Li, 1994; Sillito et al., 1995; Lamme et al., 1998a,b; Nothdurft et al., 1999; Jones et al., 2001). Such influence depends on various parameters of the ERF stimuli, such as the orientation. When an optimally oriented bar or grating extends beyond the CRF into the surround regions, the response of the V1 cell is often suppressed. This surround suppression was demonstrated by either lengthening the stimulus beyond the ends of the CRF (‘end inhibition’; Hubel and Wiesel, 1965; Rose, 1977; Orban et al., 1979; Li and Li, 1994) or broadening the stimulus beyond the sides of the CRF (‘side inhibition’; Bishop et al., 1973; De Valois et al., 1985; Li and Li, 1994). The surround suppression is maximal when the stimuli within the CRF and ERF share the same orientation (iso-orientation suppression) (Allman et al., 1985; Knierim and van Essen, 1992; Li and Li, 1994; Sillito et al., 1995; Levitt and Lund, 1997). Besides orientation, other parameters of the ERF stimuli, such as contrast (Levitt and Lund, 1997) and motion direction and speed (Li et al., 1999), can also affect the neuronal response.
In the present study, we used an iso-orientation paradigm to stimulate the CRF and ERF of V1 neurons in alert monkeys. By varying the relative spatial phase (RSP) between the central (CRF) and the surround (ERF) gratings, we investigated whether neurons in primate V1 are sensitive to the RSP. In a similar experiment, DeAngelis et al. (1994) found that most cells in V1 of anesthetized cat were not sensitive to the RSP between the CRF and ERF. In the present study, however, we have found significant neuronal sensitivity to RSP in V1 of the alert monkey. We further examined the dependence of the RSP sensitivity on the ERF property and on the spatial configuration of the center and surround stimuli.
Materials and Methods
Behavioral Training and Surgery
Experiments were performed on two female Macaca mulatta weighing 4.5–5.5 kg during the experimental period. All experimental procedures were performed in accordance with the National Institutes of Health guidelines. The monkey sat in a restraining chair that allowed movement of the head. Juice was given as a reward for performing the task. The animal was trained to fixate at a small spot (0.3° in diameter) located at the center of a monitor placed at a distance of 57 cm from the eyes. When the monkey pressed a lever, a trial began with the fixation point appearing as a white spot. After a variable delay of 0.5–5 s, the spot changed color to light yellow and the lever had to be released within 500 ms for a successful trial. Then, the fixation spot was extinguished until the next trial began. During this interval, the monkey received a drop of juice if the trial was successful. Once the monkey learned the task, a head-restraining implant, as well as a stainless steel recording chamber overlaying areas V1 and V2, were surgically attached to the skull. Surgery was conducted under aseptic conditions while the monkey was under deep pentobarbital anesthesia. A remote, infrared eye tracker (ASL Model 504, resolution 0.25° visual angle, system accuracy 0.5° visual angle) monitored the fixation and eye movements during post-operative training and throughout the experiments.
Activity from single neurons or clusters of neurons was recorded extracellularly with glass-coated tungsten electrodes prepared according to the method of Li et al. (1995). The electrode was inserted through the intact dura by means of a hydraulic micro-drive mounted on the chamber. Responses were recorded from neurons in the striate cortex whose CRFs were at eccentricities in the range of 3–6°. Nerve impulses, after being converted to standard pulses by a window discriminator, were fed into a computer, along with eye position data for real-time monitoring and analysis. Both the behavioral and physiological data were processed using software written in our laboratory. During fixation trials, eye position was continuously sampled at 60 Hz. A fixation window (1°) was set and centered on the fixation position, so that deviations of the eye position from the fixation point resulted in cancellation of the trial. To confirm that the recording sites were in V1, the CRF positions of the neurons in each penetration were traced according to the topographic mapping of V1, and the recording depth was determined with reference to the reading on the electrode manipulator and the response properties characteristic of layer 4 cells (little or no orientation selectivity, synchronized with light on and off). When more than one neuron was recorded simultaneously, the spikes from different neurons were differentiated on the basis of both amplitude and slope of the spike waveform.
A computer (Pentium III, 800) with a graphics card (Gforce GTS) was used to generate visual stimuli and the fixation point on the monitor (frame rate, 85 Hz). The screen was 40 × 30 cm. This visual stimulator could generate multiple patches of sinusoidal or line-grating stimuli of various sizes, spatial frequencies, orientations, velocities and contrasts. The stimulus patterns consisted of two gratings with a mean luminance of 10 cd/m2 and a contrast of 0.95. The small rectangle grating was set to have the same size and location as the CRF of the recorded neuron and the large rectangle grating in the ERF covered both the inhibitory end and side regions. Various RSP (0–330°) was introduced between the CRF and ERF grating stimuli. To test the spatial phase tuning of the end and side regions separately, elongated gratings extending along either the length or the width of the CRF were used. The center grating stimulated the CRF, while the bilaterally elongated grating stimulated either the end or the side regions.
Data collection was synchronized with the stimulus presentation, which began at the on-set of the stimulus and ended at the termination of the stimulus. To quantify the neuronal responses, we computed the mean firing rate over the entire period of the stimulus presentation (2–3 cycles of drifting gratings, 500–1000 ms), but the results were similar if the period immediately following stimulus onset (50–100 ms) was excluded. For each stimulus configuration, the responses to 5–20 repeats were averaged.
Determination of the CRF and ERF
The procedures were identical to those described in a previous paper (Li and Li, 1994). Briefly, to locate the center of CRF, a small rectangular gating patch (typically 0.1 × 1°) was moved at successive positions along axes perpendicular or parallel to the optimal orientation of the cell as the monkey fixated. The peak in the response profile along the length and width axes was defined as the center of the CRF, and the length and width of the CRF were determined by the distance between the bases of the rising and falling sections of the curves. The ERF properties of the neuron were assessed with reference to the spatial summation properties. To determine a neuron's length-summation curve, patch width was held constant (equivalent to the CRF's width) and the grating length was varied systematically. A similar method was used to determine a neuron's width summation curve. In V1 of the alert monkey, the majority of cells show suppression at both the end and side regions. Consequently, both the length and width summation curves decline when the stimulus length or width extends into the end or side regions. The length (or width) at which the response decreases to the spontaneous level reflects the dimensions of the end (or side) regions.
We analyzed the responses of 103 units in V1 of two monkeys. Most of the cells (77/103, 74.8%) exhibited strong surround suppression, with response reduction >70% when a uniform grating stimulus covered a large region beyond the CRF.
Relative Spatial Phase Tuning of V1 Cells
We presented the center–surround sinusoidal grating stimuli at the optimal orientation and spatial frequency drifting in the optimal direction, with the central grating confined to the CRF and the surround grating to the ERF. The two gratings drifted at the same speed, but their RSP varied randomly from trial to trial with an interval of 30° phase shift. Within each trial, the RSP between the CRF and ERF gratings was held constant, and the responses were recorded for 2–3 cycles of the drifting grating; for each RSP the data from 5–20 trials were averaged (Fig. 1). Since no substantial difference was found between the simple and complex cells in their RSP tuning properties, we grouped them together in subsequent analyses.
The data illustrated in Figure 2a are from a cell exhibiting strong inhibition at both the side and the end regions. The cell was strongly activated when the CRF was stimulated by the central grating alone (solid horizontal line in Fig. 2a), but was almost silent (at the level of spontaneous firing, dashed line) when the ERF grating was presented in phase with the CRF grating (0° RSP). The strength of the ERF inhibition, however, decreased dramatically (i.e. the response increased), with a slight displacement between the center and the surround gratings. A complete release of ERF inhibition occurred when the center–surround gratings were presented at ∼180° RSP, with the response comparable to that elicited by the CRF grating alone. A similar result is shown in Figure 2b for another cell. In this test, the RSP tuning curve was obtained using stimulus patterns consisting of abutting thin lines (inset in Fig. 2b). In the following analyses, data obtained using gratings and abutting lines are combined.
We next tested the contribution of end- and side-inhibition to RSP sensitivity separately. Figure 3a,b shows the results from a cell exhibiting suppression at both the end and the side regions. By varying the RSP between the CRF grating and the ERF grating at either the end or the side regions, we observed RSP sensitivity in both regions. In contrast, for a cell exhibiting suppression exclusively at the side region, RSP sensitivity was observed only at the side region (Fig. 3d) but not at the end region (Fig. 3c). The converse situation was observed for neurons showing suppression only at the end region.
Relationship between Spatial Phase Sensitivity and Strength of Surround Suppression
We further analyzed the relationship between spatial phase sensitivity and the strength of ERF suppression. The neuronal RSP sensitivity was quantified by the index of spatial phase sensitivity (SPS), which was defined as:
We analyzed the relationship between SPS and the suppression index for a total of 86 cells from two monkeys (55 in monkey A and 31 in monkey B). Figure 4a shows the relationship when both the end and the side regions of the ERF were stimulated. A significant correlation was found (r2 = 0.54, P < 0.01, regression slope = 0.87). Figure 4b and 4c depict the relationship between SPS and the suppression index for end-suppression (r2 = 0.78, P < 0.01, regression slope = 0.87) and side-suppression (r2 = 0.46, P < 0.01, regression slope = 0.69), respectively. Thus, SPS is strongly correlated with the strength of ERF suppression, suggesting a causal relationship between them.
Spatial Extent of SPS
In the experiments described thus far, SPS was measured when the center grating was confined to the CRF and the RSP was placed at the boundary between the CRF and the ERF. To explore the spatial extent of the modulation, we systematically changed the location of the RSP within the ERF. Figure 5a shows the result from a cell with a 0.5° CRF and a 3° ERF with extremely strong surround suppression (suppression index = 0.97). The RSP–response curves of the cell were measured at various sizes of the center grating (0.5, 1.0, 1.5, 2.0 and 2.5°). We found that RSP modulation occurred across the entire ERF, although the amplitude decreased with the size of the center grating. As shown in Figure 5b, SPS was highest at the CRF border (0.5 × 0.5°) and at 0.25° away from the border (1.0 × 1.0°) and diminished with increasing distance, disappearing at the border of the ERF (2.5 × 2.5°). The data presented in Figure 5c were obtained from another cell with a 0.6° CRF and 2.3° suppressive end region (suppression index = 0.42). The SPS also exhibited a gradual decay from the CRF border to the ERF border. Thus, although SPS is strongest at or near the CRF/ERF boundary, the entire ERF contributes to the modulatory effect.
Effect of Gap between Center and Surround Gratings
A prerequisite for V1 cells to detect spatial phase difference between the CRF and ERF stimuli is that the terminations of constituent lines at the boundary must be conjunctive. To demonstrate this, we conducted an experiment in which narrow gray gaps were introduced at the boundary between the center and the surround gratings. The RSP–response curves measured with the gap were compared with those without the gap. An example is shown in Figure 6 for a cell with a 1.5° CRF and a 4° suppressive ERF. Without the gap, we observed a normal RSP modulation, with complete suppression at 0° RSP and a maximum response at 240° RSP (SPS = 0.75) (Fig. 6a). When small (0.2°) gray gaps were introduced, the SPS decreased substantially, although it was still detectable (Fig. 6b). As the gap size increased to 0.5°, the SPS was completely abolished (Fig. 6c). This effect indicates that the spatial conjunction of displaced gratings at the interface between CRF and ERF is crucial for the SPS of V1 cells.
Spatial Resolution of RSP Modulation
For most cells, a small RSP at the interfaces can produce a substantial release of end-suppression. This implies that V1 cells are capable of detecting small displacements in straight lines at the interfaces. In order to obtain a more precise measurement of this capacity of V1 cells, we determined the minimum RSP that caused 50% reduction of ERF suppression (half-height release) at various spatial frequencies. The visual angle corresponding to the RSP was calculated as
Figure 7a–c shows the RSP–response curve of an example cell tested at three spatial frequencies. At 0.5 c/deg, Dmin was 30° RSP (Fig. 7a). When the spatial frequency increased to 0.8 c/deg (Fig. 7b), Dmin shifted to 45°. Further increasing the spatial frequency to 1.5 c/deg caused a shift of Dmin to 75° RSP (Fig. 7c). At these spatial frequencies, SR was found to be 0.167, 0.156 and 0.139°, respectively, which was approximately constant. Such spatial-frequency invariance allows assessment of the spatial resolution at which the cortical cell can detect the displacement of edges or line segments at the CRF/ERF border. For the cell shown in Figure 7a–c, the mean SR was 0.154 ± 0.015°, corresponding to the displacement of line segments shown in Figure 7d, which was 6% of the CRF size. For a population of 42 cells analyzed, the mean SR was 0.14 ± 0.07°, which was 6.9 ± 5.4% of the CRF size (Fig. 7e). Such a high spatial resolution of the striate cortical cells in detecting line displacement at the interface between CRF and ERF may serve useful perceptual functions.
In the present study, we found that V1 neurons in awake monkeys are sensitive to the RSP between the CRF and ERF gratings, and the sensitivity is strongly correlated with the strength of ERF suppression. This finding is different from two previous studies (DeAngelis et al., 1994; Akasaki et al., 2002), which reported that most cells had weak phase sensitivity and that the phase sensitivity did not depend on the strength of end- and side-inhibition. The discrepancy between the previous and the present studies may be due to the difference in the animal species and in the state of consciousness (awake monkey versus anesthetized cat). Since RSP sensitivity is correlated to the strength of ERF suppression, if there were more cells with strongly suppressive ERF in the awake monkey than in the anesthetized cat, there would be more cells sensitive to RSP. This in fact seems to be the case. Using a stimulus with a contrast of 0.36, Jones et al. (2000) found the average strength of surround suppression to be 46% in V1 of anesthetized cat, similar to that observed by Akasaki et al. (2002). On the other hand, the same group (Jones et al., 2001) reported that most cells (94%) through all layers of primate V1 exhibited strong suppression (mean reduction of 67%) to uniform stimuli exceeding the CRF. Thus, the difference in the percentage of cells with strong suppressive ERF may largely account for the difference between the awake monkey and the anesthetized cat.
We found significant SPS over the entire ERF, but the effect is strongest at the CRF/ERF border, declining monotonically from the CRF border to the ERF border (see Fig. 5b,c). This result is consistent with the findings of Rossi et al. (2001), in which the magnitude of the response to an orientation-defined figure was dependent on the distance between the boundary of the figure and the border of the CRF, with a maximum at, or in close proximity to, the CRF border. Rossi et al. suggested that this effect is due to the response of the neuron to the orientation-defined boundary within the RFs. Here we propose an alternative interpretation. Our data showed that for most neurons with small CRFs (≤1° in diameter) and strongly suppressive ERFs (suppression index > 0.9), the suppressive zone is often confined to a narrow space (typically 4° in diameter or smaller). When we expanded the grating stimulus from CRF into ERF, the response of the neuron decreased gradually and reached 70% reduction when the stimulus was 3° in diameter. Under such a condition, the surrounding grating (with 0° RSP) will exert little suppressive effect even at 0° RSP, thus changing the RSP does not significantly affect the neuronal response. This is why many neurons show obvious sensitivity to RSP at small central grating sizes (2° or less in diameter), but little or no RSP sensitivity when the central gratings reach 3° or more in diameter.
As shown in Figures 6 and 7, the suppression at 0° RSP can be largely eliminated by a small shift of the surround grating in either the horizontal (thus leaving a small gap between the center and surround gratings, Fig. 6) or vertical (thus creating a small RSP, Fig. 7) direction. The spatial resolution at which the cells can detect the shifts corresponds to a small fraction of the CRF size. The effect of the gap between the center and surround gratings has been studied previously in anaesthetized cat (Akasaki et al., 2002), which was found to be spatially co-extensive with the suppressive effect of the surround (i.e. the suppression is abolished only when the gap covers the entire suppressive surround). This is different from our finding that the suppression at 0° RSP was greatly reduced by including a small gap with a width of 0.5° (a small fraction of the width of the suppressive surround), leaving a small component of the suppression that is not phase sensitive. It is possible that this phase-insensitive component is analogous to the suppressive effect found in the cat, which is spatially co-extensive with the surround (Akasaki et al., 2002), where as the phase-sensitive component that can be abolished by the small gap represents an additional effect in the awake primate. As discussed above, this difference may be due to the difference in the animal preparation.
Another important difference between the studies in anesthetized and awake, behaving animals is the presence of eye movement. In the present study, we used a 1° fixation window to determine whether a trial is included for analysis. Due to the variability of eye position in different fixation trials within this window, the boundary between the center and the surround stimuli may fall in different locations relative to the neuronal RF, leading to inaccuracy in the measurement of the SPS extent (Fig. 5). In addition, if the boundary falls within the CRF in some trials, the edge created by RSP may contribute to the response through the CRF. Note, however, that the variability in fixation position similarly affects our measurement of the CRF border, causing an overestimate of the size of the CRF. The mean CRF size among our cells was 1.27 ± 0.6° (n = 26) in monkey A and 1.17 ± 0.34° (n = 31) in monkey B, larger than that reported by others at similar eccentricities (0.52° in Lamme et al., 1999; 0.69° in Ito and Gilbert, 1999). Thus, the boundary between the center and surround stimuli that we set at the estimated CRF border is likely to be outside of the real CRF, which should reduce the probability that the stimulus boundary falls within the CRF.
In conclusion, the RSP modulation we have found in awake monkeys was different from that observed in anesthetized cats. The exquisite sensitivity of V1 neurons to the center/surround stimulus configuration may serve important perceptual functions.
We thank Drs David Tigwell and Yang Dan for comments on the manuscript. We thank Mrs X.Z. Xu for technical assistance. This work was supported by the Major State Basic Research Program (G2000077804), the Natural Science Foundation of China (90208006), The Program of Brain and Mind of CAS, the Laboratory of Visual Information Processing of CAS and the Laboratory of Mental Health of CAS.
1Institute of Neuroscience, Shanghai Institutes of Biological Sciences, Chinese Academy of Sciences, 320 Yue-Yang Road, Shanghai, 200031, China, 2Graduate School of the Chinese Academy of Sciences