Horizontal Propagation of Visual Activity in the Synaptic Integration Field of Area 17 Neurons

See allHide authors and affiliations

Science  29 Jan 1999:
Vol. 283, Issue 5402, pp. 695-699
DOI: 10.1126/science.283.5402.695


The receptive field of a visual neuron is classically defined as the region of space (or retina) where a visual stimulus evokes a change in its firing activity. At the cortical level, a challenging issue concerns the roles of feedforward, local recurrent, intracortical, and cortico-cortical feedback connectivity in receptive field properties. Intracellular recordings in cat area 17 showed that the visually evoked synaptic integration field extends over a much larger area than that established on the basis of spike activity. Synaptic depolarizing responses to stimuli flashed at increasing distances from the center of the receptive field decreased in strength, whereas their onset latency increased. These findings suggest that subthreshold responses in the unresponsive region surrounding the classical discharge field result from the integration of visual activation waves spread by slowly conducting horizontal axons within primary visual cortex.

The average size of the minimal discharge field (MDF) in area 17 neurons is ∼2° of visual angle (for the representation near the area centralis) when it is mapped with a small spot or slit of light (1, 2). The strength of the spiking response results from the amplification of the feedforward thalamo-cortical drive by a local recurrent intracortical loop that preserves the retinotopic mapping of visual input onto cortex (1, 3). However, firing responses to stimuli presented within the MDF can also be modulated by the concomitant stimulation of its surround, over a region up to 10° of relative eccentricity (4, 5). These center-surround interactions are thought to be mediated by long-range connections, which originate within area 17 (6), or from higher cortical areas (7). In vivo intracellular recording techniques allow the analysis of subthreshold synaptic inputs and their retinal origin (8, 9). They were applied here to dissect out the different synaptic contributions potentially involved in binding information originating from distant parts of the visual field.

We performed intracellular sharp and patch recordings in the primary visual cortex of cat (10), using three mapping protocols, which evoked different degrees of spatial and temporal summation (11). First, brief punctate stimuli of positive or negative contrast were sequentially flashed in random positions in the visual field [two-dimensional (2D) impulse-like input]. Second, optimally oriented bars whose length matched that of the MDF length were flashed for longer duration in random positions across the receptive field width. Third, sinusoidal luminance gratings were presented through large co-centered annular windows, resulting in a high level of input summation. The stimulus-locked analysis of the membrane potential responses of 66 cells showed that visual cortical neurons were synaptically activated by stimuli well beyond the boundaries of their discharge field. Figure 1A illustrates a typical 2D impulse response field, with a compact MDF (2° by 1.6°), surrounded by a homogeneous subthreshold depolarizing field (D-field) (3.2° by 3.2°). Figure 1B illustrates, for another cell, the MDF (1.8°) and the subthreshold field (9°; D-field alone: 6.3°) when stimulating with long bars. Figure 1C illustrates the case of subthreshold depolarizing responses evoked by annular gratings even when separated from the MDF center by more than 11.3°.

Figure 1

Mapping the synaptic integration field. The different protocols (A toC) are symbolized in the upper row, with the preferred orientation of the recorded cell being set to vertical by convention. For each graph, synaptic responses eliciting spikes and defining the MDF extent are illustrated in bold. Shaded background delineates the region (or regions) where the significant subthreshold depolarizing responses are found. The stimulus onset is indicated by a black triangle in (B) and (C). Scale bars are 5 mV (vertical) and 100 ms (horizontal). (A) Two-dimensional impulse-like input: ON responses (averaged over 36 trials) to light patches (0.4° by 0.4°, 50 ms duration) in a unimodal simple cell. (B) Optimally oriented bars: averaged responses (15 trials) to dark bars (0.9° by 3.9°, 1 s duration) in a complex cell in 10 contiguous positions. (C) Sinusoidal gratings: averaged ON responses (30 trials) to a stationary optimally oriented grating in a simple cell. The grating is counterphased at 2 Hz in the MDF (region 1), or regions co-centered around it (2: NEAR periphery, 3: FAR periphery). Cells (A) and (B) were recorded in adult cats, and cell (C) in a 15-week-old kitten. Resting potential [dotted line in (B) and (C)] was −51 mV (A), −65 mV (B), and −62 mV (C). Retinal eccentricities of the discharge field centers were 1.3° (A), 3.9° (B), and 5.7° (C).

Two parameters characterizing the spatial sensitivity profile of the synaptic integration field were extracted at the population level: mean spatial attenuation gradient and mean spatial extent of the subthreshold responses. Because in most cells at rest impulse-like input evoked mainly depolarizations, the strength of the significant depolarizing responses for the three mapping protocols is plotted inFig. 2 as a function of the distance between the actual stimulus position and the location in the visual field eliciting the maximal spike discharge (12). The spatial sensitivity envelope of the D-field was generally co-centered with the MDF and declined with relative eccentricity. D-fields mapped with random 2D impulse-like inputs were characterized for each cell by a linear (mean correlation coefficient 〈r〉 = −0.83), steep (−24.3 ± 15.9% per degree) decrease of the postsynaptic response strength, at increasing distances from the MDF center (Fig. 2B). The same analysis for long bars gave a similar gradient estimate (−23.7 ± 16.7% per degree; mean 〈r〉 = −0.82, Fig. 2C). In contrast, the slope for gratings was significantly attenuated when compared with the 2D impulse-like input (−16.8 ± 13.4% per degree in Fig. 2D;P < 0.05).

Figure 2

Spatial profile of the depolarizing field. The visually evoked strength of the depolarizing response is normalized relative to that observed at the location of the MDF (dotted contour) eliciting the maximal discharge (circle). (A). Color-coded map of the response strength, obtained for the cell shown in Figs. 1A,3A, and 4A. Black pixels indicate the absence of significant responses. Horizontal scale bar, 2°. (B to D) Each individual profile represents for a given cell the change in response strength across the width or length of the RF, expressed as a function of the eccentricity of the stimulus from the MDF center. These different profiles have been superimposed together on the same graph, each corresponding to a particular mapping protocol [(B) 2D impulse-like input (n = 37); (C) flashed bars (n = 21); (D) flashed (n = 19) or moving (n = 2) sinusoidal luminance gratings]. The average MDF extent and its SD are indicated by thick and thin horizontal line segments, respectively. In the case of annular stimuli, the abscissa corresponds to the distance between the center of the MDF and the inner radius of the annulus.

We estimated the boundary of the D-field by the angular distance between the extreme locations where the linear fits of each flank of the spatial sensitivity hill (measured along the MDF width or length axis, or both) intercepts the threshold level of significant response. The mean D-field outer diameter was 6.6 ± 4.3° (n = 26) for 2D impulse-like input, slightly less than that observed with long bars (8.5 ± 4.3°; n = 20). When gratings were used, because of the internal symmetry of the annular stimulus, only a lower bound estimate of the D-field boundary could be extracted (half width of the inner rectangle); this value was 14.6 ± 8.6° (n = 19). Thus, independently of its precise shape, the extent of the D-field increased beyond that of the MDF (2.74 ± 2.3° for impulse-like input and 2.5 ± 1.9° for long bars) by a factor dependent on the mapping stimulus (×2.5 for impulse-like input, ×3.3 for long bars, ×5.6 for gratings). These results show the existence of a large field of depolarizing responses whose detection is favored by spatiotemporal synaptic summation. This conclusion differs from the classical suppressive role attributed to surrounding regions of the cortical discharge field (5, 13). It may also explain the additional excitatory drive provided by surrounding regions for bipartite center and surround stimulus configurations (14).

To measure quantitatively the topographic arrangement of the responsive areas, while distinguishing between spikes, excitatory, and inhibitory events (15), we devised a method adapted from the reverse correlation (RC) technique (16). This method allows us to assess which pixels in the visual field elicited a significant response at a preset confidence level (99%) (17). The occurrence times of several types of voltage events were detected in the filtered membrane potential signal, and each of the corresponding time-series was correlated with the random stimulus sequence. Three different receptive field maps were established for each neuron, based on the detection of spikes (MDF), of subthreshold depolarizing events (D-fields), and of hyperpolarizing events (H-fields) (Fig. 3). This analysis estimated the mean equivalent diameter of MDFs as 3.3 ± 1.3° and that of the D-field as 4.4 ± 2.4° (by compacting all responsive domains into a single disk of the same total area). In 30% of cases, the internal structure of the D-field was patchy, with one to four additional islands of subthreshold depolarizing responses, detached from the one co-centered with the MDF (Fig. 3C). Patchiness was even more apparent in the H-fields whose extent was more limited (equivalent diameter: 2 ± 1°). Thus, the topological union of the MDF, D-field, and H-field, that is, the synaptic integration field, on average covered an area (equivalent diameter: 5 ± 2.4°) that was four times as large as that of the MDF. This result, obtained when the cell is at rest, can only give a lower bound estimate of the maximal extent of recruitable synaptic input, the functional activation of which also depends on the voltage at which the cell is held.

Figure 3

Reverse correlation maps of the synaptic integration field. All color-coded maps (red: ON; blue: OFF) show, for all statistically significant pixels (P < 0.01), the z score of the optimized reverse correlation count (17). The zones of overlap between ON and OFF responses are illustrated by generating a transparency color map mixing linearly ON and OFF color codes. Black pixels indicate the absence of significant responses. The three columns illustrate the MDF (left), the D-Field (center), and the H-field (right) for three different cells recorded in adult cortex. The z max scores (MDF, D-Field, H-Field) are, respectively, 17.5, 14, and 6.5 (A); 32, 6.5, and 3.5 (B); and 10, 8.5, and 4 (C). (A) The same cell as in Figs. 1A, 2A, and 4A, for which the RC method reveals a much larger integration field than stimulus-locked averaging. (B) Three-zone simple cell (OFF-ON-OFF ). (C) Unimodal ON simple cell. Resting potential was −51 mV (A), −55 mV (B), and −59 mV (C). The stimulus sizes were 0.4° by 0.4° (A), 1.6° by 0.6° (B), and 1.89° by 1.39° (C). Retinal eccentricities of the discharge field centers were 1.3° (A), 5.1° (B), and 3.4° (C).

A second source of information that could help to trace the origin of subthreshold responses is their onset latency (18). A strong correlation was observed in most cells between the latencies of the depolarizing response and the eccentricity of the flashed stimulus relative to the center of the MDF, giving an isotropic latency basin shape. The center of this basin corresponding to the shortest latencies was typically near the region of maximum discharge (19). We explored the hypothesis that the peripheral subthreshold depolarizing responses are relayed by horizontal connections within area 17 (6, 20). According to this framework, the linear relation observed between the latency of the postsynaptic evoked depolarization and the stimulus eccentricity relative to the MDF center results from a constant propagation velocity of action potentials along intracortical axons. To test this prediction, we converted the distance separating two loci of activation in the visual field into a distance in the cortex. We used an average cortical magnification factor of 1 mm in the cortex for 1° in retinal space (21). The apparent speed of horizontal propagation (ASHP) of the hypothetical cortical wave of activity (distance in the cortical projection map/difference in visual latencies of the synaptic responses) can be derived directly from the inverse of the slope estimates of the latency basins. Distributions of the ASHP values (Fig. 4D) for both 2D impulse-like stimuli and long-bar stimulations peak at 0.1 m/s and are indistinguishable (unpaired t test, P > 0.55). The cortical dynamics extrapolated here match the slow velocity of activation waves monitored in vivo with optical imaging techniques along the supragranular layer plane [0.1 to 0.25 m/s (22)]. Moreover, the intracellular subthreshold responses, evoked when stimulating the “silent” surround of the discharge field with annular gratings to the exclusion of the discharge field center and its immediate periphery, were selective to orientation (23), which strongly suggests a cortical origin. Our electrophysiological estimates also match in vitro measurements of horizontal action potential propagation in rat and cat visual cortex at physiological temperature (24). Thus, our data suggest the dominant participation of a single type of connectivity responsible for the subthreshold peripheral responses, namely, that subserved by intracortical horizontal axons.

Figure 4

Basin of latencies of subthreshold depolarizing synaptic responses. (A) Color-coded map of the absolute latency of the depolarizing response, obtained for the same cell and protocol as in Figs. 1A, 2A, and 3A. Black pixels indicate the absence of significant change in the onset slope of the postsynaptic response. In (B) and (C), the graphs represent changes in latency measured across the width or length of the RF and expressed as a function of the eccentricity of the stimulus from the latency basin center (18), for two mapping protocols [(B) 2D-impulse-like input (n = 37); (C) flashed bars (n = 27)]. (D) Distributions of the ASHP values calculated from the slopes of the basins of latency illustrated in (B) (green) and (C) (blue).

Other pathways could potentially be invoked to explain lagged peripheral responses. Diverging thalamocortical fibers are unlikely candidates because of their limited lateral spread and their very high conduction velocity (25). Feedback projections from extrastriate cortical areas, or from precortical structures where MDFs are large, could explain large synaptic integration fields in area 17 (7, 26). But these pathways do not account in a simple way for the linear dependency of latency on stimulus position with respect to the MDF center. Polysynaptic horizontal propagation within area 17, or in subcortical structures, would require firing at every step of synaptic integration and therefore the unsubstantiated existence of very large MDFs (27).

The main findings of this study are threefold: (i) Spatial sensitivity at the subthreshold level decreases linearly from the center to the periphery of the discharge field. (ii) The extent of the synaptic integration fields of primary visual cortical neurons spans a much larger window of visual space than the MDF. (iii) The latency of depolarizing responses increases linearly from the MDF center toward its surround. Cortical spatiotemporal dynamics may be viewed in two complementary ways: A focal visual stimulus evokes in cortex a radial wave of activity that spreads in the plane of cortical layers at a constant velocity over a radius of more than 10 mm. This is in agreement with recent optical imaging of the cortical point spread function (22, 28). Reciprocally, functional integration realized by the cortical cell may be described by the summation of propagated inputs originating from distant intracortical sources in a spatiotemporal coordinate system centered on its minimal discharge field. A straightforward prediction of this model is that synaptic summation may be optimized by adjusting the relative latencies of activation evoked through the feedforward pathway and the peripheral intracortical connectivity. This could explain the boosting of cortical responses by bipartite stimuli composed by low-contrast center and high-contrast surround features (14). This also suggests that the processing of new visual information arising in the center of the classical discharge field could be highly influenced by contextual information that was present a few tens of milliseconds earlier in the periphery.

  • * Present address: Neurobiology Division, Medical Research Council LMB, 2 Hills Road, Cambridge CB22QH, UK.

  • To whom correspondence should be addressed. E-mail: yves.fregnac{at}


View Abstract

Navigate This Article