Cerebral Cortex Advance Access published online on March 29, 2007
Cerebral Cortex, doi:10.1093/cercor/bhm019
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Cortical Response Field Dynamics in Cat Visual Cortex
Neurobiology Department, Weizmann Institute of Science, Rehovot 76100, Israel, 1 Current address: Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA 02129, USA, 2 Current address: Department of Cognitive Neurobiology, Ruhr-University Bochum, D-44780 Bochum, Germany, 3 Current address: INCM, UMR6193, CNRS—Aix-Marseille Université, 31 Chemin Joseph Aiguier, 13402 Marseille Cedex 20, France
Address correspondence to email: dahlia{at}nmr.mgh.harvard.edu.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Little is known about the "inverse" of the receptive field—the region of cortical space whose spatiotemporal pattern of electrical activity is influenced by a given sensory stimulus. We refer to this activated area as the cortical response field, the properties of which remain unexplored. Here, the dynamics of cortical response fields evoked in visual cortex by small, local drifting-oriented gratings were explored using voltage-sensitive dyes. We found that the cortical response field was often characterized by a plateau of activity, beyond the rim of which activity diminished quickly. Plateau rim location was largely independent of stimulus orientation. However, approximately 20 ms following plateau onset, 1–3 peaks emerged on it and were amplified for 25 ms. Spiking was limited to the peak zones, whose location strongly depended on stimulus orientation. Thus, alongside selective amplification of a spatially restricted suprathreshold response, wider activation to just below threshold encompasses all orientation domains within a well-defined retinotopic vicinity of the current stimulus, priming the cortex for processing of subsequent stimuli.
Key Words: area 18 • local oriented grating • optical imaging • orientation map • voltage-sensitive dye imaging
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Electrophysiological studies have offered a wealth of data about how single neurons respond to various stimuli, frequently exploring the "optimal" stimulus for the examined neuron and determining the degree of selectivity to this preferred stimulus (Hubel and Wiesel 1962
). However, a given sensory stimulus necessarily elicits some activity in many neurons for which it is not optimal, and the complex cortical spatiotemporal response pattern that emerges may not be straightforwardly predicted from the activity of single cells. Therefore, it is ultimately necessary to complement the single-/multiunit approach with high speed imaging studies of patterns of population activity. Population responses to full-screen visual stimuli reveal a dynamic combination of orientation-selective and orientation nonselective components (Sharon and Grinvald 2002). Local visual stimuli evoke waves of suprathreshold and subthreshold activity in visual cortex (Grinvald et al. 1994; Bringuier et al. 1999; Jancke, Chavane, et al. 2004). Thus, the sensory stimulus induces a dynamic field of cortical activity. The term cortical point-spread function refers to the region of cortex activated by a point stimulus and was calculated indirectly in previous studies from responses to nonpoint stimuli (Albus 1975; Cynader et al. 1987). In contrast, here we explore the full spatiotemporal activity pattern of responses to local gratings, which is thus more comprehensively termed the cortical response field.
The myriad intracortical connections, the complex and variable behavior of cortical neurons, and the variability in the retinotopic map make it difficult to predict a stimulus' cortical response field, that is, the spatiotemporal pattern of the population response, from the responses of single cells. The cortical response field is expected to include both spiking and subthreshold responses within partially overlapping cortical positions, but how these combine to represent a given stimulus is not known. It is now possible, using optical imaging of population activity (Grinvald et al. 1982), to directly measure the cortical response field elicited by a visual stimulus. Optical imaging using voltage-sensitive dyes (Grinvald et al. 1984) has shown that quickly after onset, responses spread over millimeters of cortex (Grinvald et al. 1994). Because voltage-sensitive dye imaging (VSDI) reflects neuronal population responses with millisecond temporal resolution, emphasizing dendritic subthreshold activity (Sterkin et al. 1998; Grinvald et al. 1999; Petersen, Grinvald, et al. 2003), it is particularly suited to study cortical response field dynamics. However, the previous publication (Grinvald et al. 1994) did not offer columnar resolution due to technical limitations of VSDI at the time. Recent advances in the methodology (Shoham et al. 1999; Grinvald and Hildesheim 2004) now allow us to report their spatiotemporal structures at the necessary 10-fold higher spatial resolution and at much finer temporal detail than previously possible (Grinvald et al. 1994).
In particular, we were interested in the dynamic representation of local oriented gratings at the population level because this stimulus type has often been used in single-unit electrophysiology of primary visual cortex. Our own previous work has shown that for full-screen gratings, an orientation-selective component exists from response onset, growing in amplitude over several tens of milliseconds but constant in tuning width, in addition to a large orientation nonselective component, which increases in amplitude for a longer duration (Sharon and Grinvald 2002). Studying responses to nonoriented square stimuli, we found low-amplitude activity spreading far beyond the retinotopic representation of the stimulus, as well as a region of high-amplitude activity that does not spread and to which spiking is limited (Jancke, Chavane, et al. 2004). Based on these 2 previous studies, one would expect that the spatial profile of the response to local gratings will demonstrate a smooth gradation from peak to periphery, punctuated perhaps by local peaks corresponding to orientation patches of appropriate preference—the latter depending on the relative amplitudes of the orientation-selective and nonselective components for such stimuli. Existing theoretical studies of population representations in general also predict smooth population profiles, for example of Gaussian shape, based on tuning functions analogous to those obtained from single-cell electrophysiology (for review, see Pouget et al. 2003). Another common assumption is that the peak of the response to a local oriented stimulus coincides with an orientation patch of appropriate preference. Both these predictions cannot be tested with electrophysiological methods at sufficient sampling density and indeed have not been directly examined before. We tested both these predictions and found them to be wrong. Instead, we found that the population response profile includes a plateau beyond which activity amplitude exhibits a sharper decay. Later in the response, a peak (or series of up to 3 peaks) emerged on the plateau. Though the location of the peak strongly depended on the inducing stimulus' orientation, it only partly coincided with it. To the best of our knowledge, these phenomena have not been observed or predicted by previous work, and we propose a simple model as an initial working hypothesis to explain them.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
All surgical and experimental procedures were in accordance with the National Institutes of Health guidelines. Nine imaging sessions from nonoverlapping cortical regions in 8 hemispheres of 7 adult cats were used for the VSDI-only analysis. In 3 of these, local stimuli in 2 different locations, separated by a distance equal to the stimuli's size, were used—totaling 12 different local stimuli used in this study for VSDI-only analysis. Two additional sessions from 2 adult cats were used for the combined VSDI–electrophysiology data.
Surgery
Cats were initially anesthetized with intramuscular ketamine (15 mg kg–1) and xylazine (1 mg kg–1), supplemented by atropine (0.05 mg kg–1). Following tracheotomy, animals were artificially respirated and anesthetized with 1–1.5% (0.6–1% during recording) halothane in a mixture of equal O2 and N2O. The skull was opened above area 18, and the dura was resected. Paralysis was obtained with pancuronium bromide (0.2 mg kg–1 h–1, intravenous) administered starting >3 h before imaging to abolish eye movements (see also below). The position of the area centralis of each eye was projected on the screen using a fundus camera (Nikon, Tokyo, Japan) before and after imaging. Eyes were fit with zero-power contact lenses, and external lenses were used to focus them on the screen. A prism was placed in front of the ipsilateral eye to obtain convergence of the 2 eyes when necessary. Electrocardiogram, electroencephalography, expired CO2, and body temperature were continuously monitored during the experiment. Additional details have been described previously (Bonhoeffer and Grinvald 1993).
VSDI and Electrophysiology
The exposed cortex was stained for 2.5–3 h with the oxonol dye RH-1691. A FUJIX HR Deltaron 1700 camera with an array of 128 x 128 detectors, each monitoring 
64 x 64 mm, was used for data acquisition. A detailed account of the data acquisition setup and procedure has been published (Shoham et al. 1999). Data frames were acquired at a rate of 9.6 ms per frame. For the electrophysiological confirmation experiments, single-/multiunit recordings were carried out using tungsten electrodes and the in-house sliding-top cranial window and electrode positioner microdrive previously described (Arieli and Grinvald 2002). Electrophysiology was performed prior to VSDI, allowing extensive averaging of spiking responses without danger of photodynamic damage. Vertical penetrations under a microscope were guided by a high-resolution image of the cortex taken with a Mavica MVC-FD91 camera (Sony, Tokyo, Japan) under green illumination to emphasize the blood vessel pattern. The signal was filtered between 0.1 and 3 kHz and digitized at 25 kHz via a PCI-6601 National Instruments board. Data acquisition was controlled with in-house programs, and spike sorting was performed using Multi Spike Detector (Alpha-Omega, Nazareth, Israel).
Eye-movement–free data were ensured for the VSDI-only experiments by single-trial analysis where an eye movement is detected as spatial movement of the peak response between trials: all trials after an eye movement, if such occurred, were rejected. During combined VSDI–electrophysiology experiments, the receptive field (RF) of each cell was manually mapped to determine its location in the visual field and preferred orientation. We then performed the electrophysiological recording at that site using the same stimulus location in all penetrations. RF mapping was repeated after recording to verify that no eye movement had occurred in the meantime. During VSDI, the electrode was kept at a constant cortical location, and the RF at that site was mapped from time to time when VSDI was paused in order to verify constant eye position. Furthermore, we performed an additional recording after VSDI at a cortical location near the very first penetration, to verify that the eyes had not moved between the start and end of the whole experiment.
Visual Stimulation
Stimuli were 100% contrast black/white square-wave gratings of 0.2 cycles/degree drifting for 500 ms at 6 Hz. Local stimuli, 4 degrees in diameter, and full-screen stimuli were presented at vertical and horizontal orientations, at 2 drifting directions. Local stimuli were presented at a horizontal eccentricity of 4–5 degrees contralateral to the vertical meridian and a vertical eccentricity of 2–10 degrees below the horizontal meridian. Between stimuli, the screen was blank with luminance of 49 cd m–2, equal to mean luminance during stimulation. Stimuli were pseudorandomly interleaved with recording epochs in which the screen was blank. A recording epoch of 1 s surrounded each stimulus presentation. The stimuli were displayed binocularly using VSG series 3 (Cambridge Research Systems, Rochester, England) on a 38 x 29 cm2, 640 x 480 pixel2 monitor, located 50 or 57 cm from the cat's eyes, at a refresh rate of 150 Hz.
The present experiments were much more demanding than our previous VSDI studies with respect to signal-to-noise considerations and pushed the current technology to its limits. The expected increase in relative noise level when shortening the integration interval of the dye signal 40-fold (from 380 ms in Figs 1–3, when studying the cortical response field spatial profile, to 9.6 ms in Figs 4–7, when studying cortical response field dynamics) is over 6-fold, assuming signal-to-noise ratio increases with the square root of integration interval. Furthermore, signals are smaller with a local stimulus. Therefore, in order to maximize statistical power by increasing the number of samples before photodynamic damage starts affecting the data and before an eye-movement occurred, we presented only 2 orientations and in each cycle of presentations showed the local stimuli twice as many times as the full-screen stimuli. In the experiments reported here, we were typically able to analyze only 40 responses to each local stimulus.
|
F/F, parts in a thousand 
; the brightness of pixels relative to resting fluorescence level). The scale bar, and the spacing between grid lines in the (x, y) plane, here and in all other figures is 1 mm. In the far periphery, net activity was suppressed to below baseline levels (dark blue regions in this case—color bar at left). (B) The same as in (A) for another experiment—the eccentricity here was (4, –2) degrees. (C) The cortical response fields from the 10 additional cases in this study shown for the vertical (3) or horizontal orientations (4–12), in decreasing rim saliency order. Arrows point at plateau rims. Title colors denote different hemispheres (e.g., 8 and 12 are 2 different stimulus locations from one hemisphere).
|
|
|
Data Analysis
Peristimulus time histograms of the electrophysiological recordings were calculated using in-house software. For VSDI data, the evoked response to each stimulus was calculated for each pixel on a frame-by-frame basis in 2 steps. The recorded value at each pixel was first divided by the average value at that pixel before stimulus onset (to remove slow stimulus-independent fluctuations in illumination and background fluorescence levels), and this value was subsequently divided by that obtained for the blank condition. This procedure eliminates most of the noise due to heartbeat and respiration (Grinvald et al. 1984), and the result reflects neuronal activity after subtraction of spontaneous activity. To obtain the evoked response for each orientation, the responses to the 2 directions of motion were averaged. The average cortical response field of a stimulus was defined as the average evoked response over 40 data frames starting at 140 ms after stimulus onset. For all single-condition analysis (steady state and dynamics), raw, unfiltered, data were used. For differential orientation analysis only—Figure 2A,B and Figure 3 inset—the data were mildly high-pass filtered (1250 or 2500 µm, 2-dimensional [2D] Butterworth filter) to remove slow gradients and isolate the local orientation patches.
|
The rim of the plateau was located by smoothing (250 µm low-pass 2D Butterworth filter) the cortical response field averaged over time, and determining as follows an area in which the slope quickly changes from small to large, in the direction from peak to periphery of the cortical response field (see Supplementary Fig. 1). Contour levels at 30 intervals were drawn using Matlab's (The Mathworks, Natick, MA) contour function and a Hausdorff-like distance measure between each level, and the one above it was determined. First, the distance of a single point on the lower level to the entire next contour line was calculated as the minimum distance between that point and each one on the next contour line. Then, the distance between the entire 2 contours was calculated as the mean distance between the points on the lower level and the entire higher level (the Hausdorff distance takes the maximum instead of the mean and is symmetric—but is more sensitive to noise). Local maxima of the distance as a function of level were taken as candidate rims if they were among the 2 or 3 closest to the peak of the cortical response field. The eventual rim was determined as the one for which a histogram of cortical response field values of all pixels within it had a local peak at low values, meaning that indeed within it lay a relatively flat region, in which values change slowly, and a region where the rate of change increases. For 2 cases, we verified that changing the number of intervals between 20 and 40 did not change the determination of the rim.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Cortical Response Fields: Spatial Structure
To determine the cortical response field in response to small oriented gratings, we began by characterizing their spatial structure. One might expect a smooth decline in activity level from the cortical response field peak to periphery, as the overlap between the stimulus and the aggregate RF at each cortical location decreases. However, this was not the case, as shown in Figure 1A for a horizontal drifting grating stimulus, 4 degrees in diameter, presented for a duration of 500 ms. The average cortical response field, during the late time window of 140 to 515 ms after stimulus onset, had a distinctive spatial structure: an initial rapid drop from the peaks is followed by a slower sloped region, the plateau, beyond the rim of which (black contour line) a rapid decline in activity level occurs in the periphery. This spatial pattern has not been reported previously. Figure 1B shows the averaged cortical response field from another hemisphere. The population of late cortical response fields in this study exhibited a spectrum with respect to the strength of the discontinuity at the plateau rim (Fig. 1C). In some, it was more pronounced than in others, but in all, it was possible to determine the plateau rim as explained in the Materials and Methods and shown in Supplementary Figure 1. To clarify, the nonsmooth feature of the cortical response field is the plateau rim, which bounds an area including both the peak and the plateau. There is a smooth transition between peak and plateau. The plateau rim makes a sharp distinction between the plateau area and the activated periphery. We first examine the orientation selectivity of cortical locations within the plateau rim and then study the events of the population response that lead up to these late cortical response fields: the emergence of the peak and the subsequent dynamics of peak and plateau amplitudes.
Relationship to the Orientation Map
Stimuli of different orientations in the same visual location evoke a very similar cortical response field plateau. In contrast, the location of the peak is more strongly related to the orientation map. A comparison of the cortical response field structure with the orientation map of the same cortical region is shown in Figure 2A,B for the same hemispheres shown in Figure 1A,B, respectively. On the full-screen (left) and local (right) differential orientation maps are overlaid several contour lines from the horizontal (yellow) and vertical (purple) cortical response fields. The thinnest corresponds to the level of the rim of the plateau (see Materials and Methods) and the 2 others—to the levels of the 25% and 5% highest pixels. Indeed, the difference between the cortical response fields of the 2 orientations is larger at higher levels. At higher activity levels, correspondence to the orientation maps increases—meaning that most pixels at the peak of a cortical response field have orientation preference equal to the orientation of the inducing stimulus, whereas at plateau level, there is no apparent correspondence between cortical response field and orientation map. The strongest local orientation map occurs in the region near cortical response field peak.
At each cortical location, the response thus has an orientation-selective and an orientation nonselective component. The orientation-selective component accounts for 4.8 ± 2.8% of the signal at cortical response field peak (on average over experiments) for local stimuli, which is about 4 times less than for full-screen stimuli (see Supplementary Fig. 4 of Sharon and Grinvald 2002). At plateau level (and lower), the nonselective component is even more dominant, although an orientation-selective component exists, as demonstrated by the fact that an orientation map does appear in response to the local stimuli (Fig. 2A,B middle).
The location of cortical response field peak depended more strongly on the inducing stimulus' orientation than did the location of cortical response field plateau for the full set of experiments, as quantified and shown in Figure 3. The full-screen orientation map (e.g., Fig. 2A,B, left) was used as a reference map, according to which each pixel was designated as preferring horizontal or vertical. We then asked to what degree a local horizontal stimulus induced activity only in cortical pixels preferring horizontal, for each activity level from plateau rim to peak (likewise for vertical). To do this, we divided the pixels within the rim of the plateau of each orientation's cortical response field into 20 equal bins according to their activity level and calculated the fraction of pixels in each level for which the preferred orientation was equal to the orientation of the inducing stimulus. These are the "correct" pixels at that level, and their fraction is a measure for how exclusive the activation at that level is to pixels whose orientation preference matches the inducing stimulus. We show the fraction of correct pixels averaged over the vertical and horizontal cortical response fields, with color denoting different hemispheres. The thick black trace is the average over all experiments, showing that pixels near plateau rim were equally likely to prefer either of the 2 orientations independently of the inducing stimulus' orientation (fraction of correct pixels near 0.5 for all bins in lower half). In contrast, a sharp increase in the likelihood to prefer the orientation of the inducing stimulus is exhibited as cortical response field peak is approached. Interestingly, among the 5% highest pixels, the average fraction of correct pixels reaches only 0.775. This is in agreement with the visual impression in Figure 2 that the contour of the 5% highest pixels is near but does not fully coincide with the appropriate neighboring orientation patch.
A similar though weaker trend was seen for the amplitude of orientation preference, the modulation depth, defined as the difference between the response to preferred and orthogonal orientations in a previous VSDI full-screen grating study (Sharon and Grinvald 2002), as a function of activation level (Fig. 3, inset). Thus, an orientation map could be detected at least as far as plateau level, but near the peak it is strongest.
To summarize, an orientation map, that is, orientation-selective activity, exists from the peak and outward beyond plateau rim. At the level of plateau rim, there are equal numbers of neurons with orientation preference matching the stimulus and of neurons with an orthogonal preference. That is, at plateau rim, the magnitude of the orientation-selective component is very low relative to the total signal. Even the very peak of the response does not fully coincide with an appropriate orientation patch, meaning that stimulus location dominates the coded signal (see Discussion).
Delayed Emergence of the Peak on the Initial Plateau
We hypothesized that the spatial profile of the cortical response field is a signature of the amplification process of the retinotopic feedforward thalamic input by the local cortical circuit, which, together with modulation from other cortical areas, determines the strength of the spiking response (Douglas and Martin 1991). Hence, the next goal was to investigate the cortical response field dynamics of local gratings. Knowing that the signal-to-noise ratio in this case would necessarily be considerably lower both than when averaging over 380 ms (Figs 1–3) and when observing responses to full-screen stimuli, we did our utmost to increase the quality of the data by extensive signal averaging, so that decisive statistical testing would be possible (see Materials and Methods).
In Figure 4A,B, we show, for the same 2 examples shown in Figure 1A,B, 3 consecutive "snapshots" of the local horizontal grating cortical response field shortly after stimulus onset. An additional prestimulus snapshot is shown for comparison (top). The peak and plateau regions started responding together, initially at the same amplitude. A comparison to the average cortical response field shown in Figure 1 reveals that the characteristic peak on the plateau was absent in the beginning of the response and developed slightly later. Supplementary Movie 1 shows the sequence of events early in the response, clearly showing the flat plateau at 38 ms, whereas in the following frame, the peak has visibly emerged on it.
To quantify these events, we next performed region-of-interest (ROI) analysis to compare the time courses of pixels within different regions (Fig. 5A) of the cortical response field shown in Figure 1B: the highest 25% of the pixels within the rim of the plateau, corresponding to the peak of the cortical response field; the lowest 25% of these pixels, coming from the plateau; and the same number of pixels from the farthest periphery of the cortical response field within the imaged area. The time course of the response to the preferred orientation, averaged over these 3 ROIs, is displayed in Figure 5B. The peak ROI obviously gave the strongest response; however, a zoom in (inset) shows that for the initial response (until 43.2 ms, left asterisk), its activity level was equal to that of the plateau ROI (or even slightly less than that). In the subsequent data frame, at 52.8 ms, the activity level of the peak ROI is already larger than that of the plateau ROI. Note that the minute apparent lead of plateau ROI response amplitude over that of the peak ROI up to and including 43.2 ms is insignificant. Thus, the very earliest response consisted of a uniform rise in activity level within plateau rim of both the peak and plateau regions together, followed by larger/faster activation in the peak relative to the plateau.
|
: periphery. Shading is the standard deviation over pixels in each region. Inset: zoom in on response onset. At the initial response, the activity of peak ROI was not greater than that of plateau ROI. Shading in inset: 99% confidence interval. Asterisks: initial significant separation. Sampling rate: 9.6 ms/frame. (C) Population summary of time of initial peak-from-plateau and plateau-from-periphery separation, obtained from ROI analysis as in (B). Colors are the same as in Is the fact that initially the plateau and peak regions activate together statistically significant? For each data frame, we performed a t-test to assess whether activity of pixels in the peak ROI was stronger than that of pixels in the plateau ROI and whether that in turn was stronger than pixels in the periphery ROI. Asterisks denote the initial time for which significantly stronger activity was consistently observed (one-sided t-test, 0.001). At 43.2 ms, plateau pixels were already activated more strongly than periphery pixels (first asterisk in inset), whereas peak pixels exceeded plateau pixels only at 52.8 ms (second asterisk in inset). Initial activity thus began at equal levels for peak and plateau pixels, confirming the visual impression in Figure 4 that, in contrast to later times, the initial response consisted of a flat plateau-like structure, with no peak on it.
The emergence of the peak on an earlier plateau was confirmed for the full set of experiments. Figure 5C summarizes the results of the analysis shown in Figure 5B for all 12 cortical response fields examined in this study. For each cortical response field, we calculated first the initial time at which significantly stronger activity in peak ROI than in plateau ROI was consistently observed (i.e., for at least 40 uninterrupted ms—but in 11 of 12 cases, stronger activity persisted for the full duration of the response), see, for example, first asterisk in Figure 5B. We then calculated the initial time for plateau–periphery separation—for example, second asterisk in Figure 5B. All cases are plotted in Figure 5C. For all but one case, initial peak–plateau separation occurred later than initial plateau–periphery separation (points above equality line). Initial consistent peak–plateau separation occurred later than plateau–periphery separation at a confidence level of 0.003 (one-sided paired t-test). The average times for initial consistent peak–plateau and plateau–periphery separation were 61.4 ± 17.9 ms (mean ± standard deviation) and 42.2 ± 17.9 ms. For these final population calculations, we were forced to exclude 2 deviant data points—the one for which plateau activity started to be consistently higher than periphery activity already before stimulus onset (negative value on y axis) and the extreme outlier point exhibiting initial peak-plateau separation at 216 ms (though both points conform to the trend of the other data points). Thus, the full data set shows that, similarly to the example shown in Figure 5B, the peak emerged 20 ms after appearance of the plateau.
Plateau Activity Continues to Increase after Peak Activity Relative to the Plateau Is Maximal
In order to further probe the characteristic temporal behavior of the peak and plateau responses, we compare the time course of peak height to that of plateau height. Relative peak height refers to its elevation above the plateau and is calculated as the difference between peak and plateau ROIs (Fig. 6A, trace labeled x; ROIs defined identically to Fig. 5). Although the total response continues to rise until 220 ms (Fig. 5B), relative peak height abruptly reaches a steady state level more than 100 ms earlier. Thus, after this time, the peak on the plateau has a nearly constant size, and the peak and plateau regions continue to rise together at an identical rate. At this stage, however, the plateau itself still rises relative to the periphery, as shown in Figure 6A (trace labeled +). Relative plateau height (above periphery) stops increasing about 50 ms before the total activity reaches its maximum. To find the precise time at which these traces stopped increasing, we performed a t-test on the difference at each data frame to assess whether the difference at the next data frame was larger. Asterisks in Figure 6A denote that at 91.2 ms, peak height stabilized, whereas plateau height stabilized at 168 ms. After this time, there was thus a uniform increase in signal across the imaged cortex until 220 ms after stimulus onset.
|
Figure 6B next summarizes this analysis over all cases. For each cortical response field, the time at which there was no significant increase in peak–plateau amplitude difference for at least 2 consecutive data frames—see, for example, first asterisk in Figure 6A—is plotted as a function of the time at which plateau-periphery amplitude difference stabilized—second asterisk in Figure 6A. In all but one case, data points lie above the equality line, and this trend is significant to a confidence level of 0.001 (one-sided paired t-test). Stabilization of relative peak height occurred on average at 86.4 ± 25.2 ms, whereas the height of the plateau above the periphery stopped increasing only at 128 ± 39.5 ms.
Unique Behavior at the Peak Zone
As shown above, the response at the peak ROI has distinct features in terms of both stimulus selectivity and response dynamics. More evidence that the response at the peak is distinct may be offered by the fact that for 5 out of the 12 cases studied here, the time course of modulation depth was markedly different between peak pixels and plateau pixels (Fig. 7). In these cases, for peak pixels, maximal modulation depth was reached up to 100 ms after response onset followed by decline, reminiscent of the typical modulation depth time course in response to full-screen stimuli (see Fig. 3C of Sharon and Grinvald 2002). In contrast, plateau pixels exhibited a more gradual and asymptotic increase in modulation depth. This is shown in Figure 7 for the 2 examples shown in Figure 1. These properties taken together led us to hypothesize that the peak zone is the locus of spiking activity.
Electrophysiological Confirmation of Spiking Absence Away from the Peak
To test the hypothesis that the peak, with its slightly delayed emergence on the plateau and rapid amplification, is the spiking cortical zone while the rest of the cortical response field is mostly subthreshold, we performed electrophysiological recordings. By definition, spiking is triggered in cortical regions with neurons having RFs centered on the stimulus. Furthermore, it has previously been confirmed that spiking occurs at high levels of the VSDI response (Jancke, Chavane, et al. 2004). Therefore, the critical question was whether away from the peak, spiking activity occurs. Hence, in a separate set of experiments employing both VSDI and extracellular recording, we confirmed that there is no spiking activity away from cortical response field peak.
In these experiments, electrophysiological recordings were performed in a series of penetrations at different cortical locations, using precisely the same stimulus used during VSDI. This allowed us to directly compare spiking responses to the VSDI signal. Responses to full-screen stimuli were also acquired, to verify that in cortical loci where no spiking was evoked in response to the local stimulus, it was not trivially due to problems with the extracellular recording. Eye movements were rigorously monitored to ensure that the retinal location of the stimulus stayed constant (see Materials and Methods).
Figure 8A shows the cortical image (left) and cortical response field (right) with penetration locations superimposed, and Figure 8B presents the cortical response field in surface format. One of these electrode penetrations (1) was targeted to the peak zone and the other (4) to the plateau away from the peak. The extracellular responses are shown in Figure 8C,D as average response time courses. For each recording, responses to full-screen (left) and local (right) stimuli are shown, with the traces in each corresponding to 4 different stimulus orientations. The full-screen responses exhibit normal levels of evoked activity and orientation preference. However, the local stimulus evokes spiking only in penetration 1 (peak zone) and does not evoke spiking above that exhibited during the blank stimulus for any orientation in penetration 4 (plateau, away from peak). The full experiment, with 2 additional peak zone penetrations, is presented in Supplementary Figure 2. An additional experiment, with 3 penetrations on the plateau (away from peak) and 2 in the periphery, is shown in Supplementary Figure 3. There were no spiking responses to the local stimulus in any of these penetrations. Spiking was therefore limited to the peak zone of the cortical response field.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The main results are that at its onset, the cortical response to a local oriented grating is a flat-topped structure (Fig. 4 and supplementary movie). Response onset occurs
40 ms after stimulus onset. Approximately 20 ms later, a peak emerges on this plateau (Fig. 5) and reaches its maximal amplitude in about 25 ms. The height of the plateau continues to increase for 50 ms more (Fig. 6). Spiking exists at the peak zone and is absent elsewhere, both at the plateau level and in the periphery (Fig. 8 and Supplementary Figs. 2 & 3). At the peak zone, most of the pixels have the appropriate preferred orientation, matching that of the inducing stimulus, whereas at plateau level, no such bias exists (Fig. 3 main panel). Although this means that pixels of all orientation preferences are located within the plateau rim, pixels with orientation preference matching the stimulus respond slightly stronger than those with orthogonal orientation preference (Fig. 3 inset). This orientation-selective "ripple" on the plateau allows orientation maps to emerge in response to local stimuli (Fig. 2). The total detected signal continues to slightly increase in the imaged cortical region—but in a much less localized manner—until 200–300 ms after stimulus onset (Fig. 5B). Later than this, there is an overall decrease in signal amplitude. In 3 cortical response fields (Fig. 1C, responses 7, 8, 11), we observed multiple peaks, probably due to the specific relationship between the retinotopic mapping of these specific stimuli and the local orientation map (i.e., the directly stimulated visual area contained more than a single orientation patch).
The voltage-sensitive dye signal reflects the population membrane potential response, and the dendritic compartment's contribution is emphasized because of its large membrane area and lack of myelin. The signal at each imaged pixel will be strongly affected by the dendritic arborization within it, including dendrites of neurons whose cell bodies lie up to 300 µm away, in columns with different selectivity properties. Thus, the VSDI signal is somewhat smeared relative to somata activity. In what way do the current reported cortical response field properties differ at the level of the cell bodies? First, the orientation selectivity at the cell bodies will be somewhat higher. Second, the dendritic contribution to the signal is equivalent to smoothing its spatial structure, assuming that the dendritic arborization is isotropic on average (Buzas et al. 2006 show that the lateral connections of populations of neurons can be predicted by a Gaussian distribution as a function of cortical distance from the site of origin, convolved with an orientation-selective connectivity component). This means that the discontinuity exhibited at the rim of the plateau (the sudden increase in response decline rate) may even be more pronounced at the somatic level than revealed by VSDI, but it is unlikely that its location changes. Likewise, the peak may be sharper at the somatic level, but its location is unaltered. Thus, it is predicted that at the cell body level, the cortical response field characteristics revealed here would be very similar but sharper and less fuzzy. The fact that the plateau structure is present in the earliest response and that it is orientation selective rules out glial activity as an explanation of the present results.
In addition to the different stimuli from those explored previously (Jancke, Chavane, et al. 2004), here we also applied a more detailed data analysis of the spatial profile as well as finer sampling of electrode penetrations near the cortical response field peak. This enabled us to reveal the plateau, in addition to the peak and periphery zones (termed high- and low-amplitude activity in that study). The plateau does not spread (see Fig. 4) and does not exhibit spiking (see Fig. 8) and is therefore distinct from both the peak, which does not spread but exhibits spiking, and the periphery, which does spread but does not exhibit spiking (see Figs 1 and 2 in Jancke, Chavane, et al. 2004).
The relatively low orientation selectivity exhibited overall by VSDI signals probably arises from 3 main reasons. The first is that, due to its dependence on total membrane surface, VSDI emphasizes subthreshold potentials (Sterkin et al. 1998; Grinvald et al. 1999; Petersen, Grinvald, et al. 2003), which are less orientation selective (Carandini and Ferster 2000; Volgushev et al. 2000; Monier et al. 2003), over spiking activity. Furthermore, subthreshold potentials measured intracellularly are measured at the soma, whereas VSDI provides a view of the whole dendritic tree. A third, related reason is that the signal at each imaged pixel receives contribution from membrane potential events occurring in dendrites of neurons whose soma is not within the imaged pixel. Regarding the stronger orientation selectivity in full-screen responses than at the peak of local responses, we speculate that it is due to cooperative contributions from horizontal connections, received at each point in the full-screen response from the entire surrounding cortex. To the best of our knowledge, a systematic single-unit comparison of full-screen and local responses has not been reported. However, using VSDI, we obtained results suggesting that lateral connections do indeed serve to increase orientation selectivity when an extended stimulus is used (Chavane F, Sharon D, Jancke D, Marre O, Frégnac Y, Grinvald A, in preparation).
A Simple Model Qualitatively Accounts for the Data
We propose a conceptual model to account for the results (Fig. 9), which relies on 2 observations. The first is that for a given cortical column, spiking is evoked in a much smaller area of the visual field than the area that evokes synaptic activity, that is, the suprathreshold RF is much smaller than the subthreshold RF. This has been observed in intracellular RF mapping work (Bringuier et al. 1999). The second is that the stimulus size we used was the same order of magnitude as the suprathreshold RF size, because it is the way the stimulus was designed, based on Tusa et al. (1979). See Figure 9A, top. These 2 observations account for the spatial and temporal properties of cortical response fields we measured.
|
Given a set of cortical locations, such as those along a straight line going through the peak zone of the cortical response field (Fig. 9A, bottom), the subthreshold RF and suprathreshold RF for each location will have some degree of overlap with the stimulus (Fig. 9B), and will therefore be activated to a corresponding extent. The magnitude of the subthreshold and suprathreshold response at each cortical location will depend on the degree of overlap between the stimulus and the subthreshold and suprathreshold RF, respectively. Assuming that stimulus size is much smaller than the subthreshold RF (see above), leads to the prediction of a plateau in the subthreshold response (Fig. 9C): As one moves from cortical response field center (red point) outward, there is first a region (purple and blue points) where the stimulus is fully contained within the subthreshold RF (full overlap—though not concentric) and therefore the subthreshold response will remain near maximum, decreasing slowly. As soon as the overlap between stimulus and subthreshold RF is not maximal (orange and green), the subthreshold response starts to drop quickly, thus giving rise to a predicted plateau structure of the subthreshold response.
Assuming that stimulus size is the same order of magnitude as the suprathreshold RF leads to the prediction of a sharper peak in the suprathreshold response (Fig. 9D): As soon as one moves away from the center of the response, spiking decreases due to a decreased overlap between stimulus and suprathreshold RF. On the plateau away from the peak (purple and blue), because subthreshold RF is much larger than suprathreshold RF, there is already no overlap between stimulus and suprathreshold RF and therefore no spiking response.
Because the VSDI signal is a weighted sum of the voltage events occurring in the imaged area, it will reflect both the subthreshold and suprathreshold responses. This accounts for the spatial structure of the cortical response field, that of a plateau with a peak on it (see Fig. 1). These considerations can also explain the measured cortical response field dynamics, as detailed in Figure 9E. At the top, we present a time line summarizing the events exhibited by the present data, with the rows below showing each of the predicted components at each of these events. The first row illustrates the subthreshold response, which is the earliest to appear, with its characteristic plateau structure (40 ms). For simplicity, we did not draw the "ripples" on the plateau that correspond to areas preferring the orientation of the stimulus and those preferring the orthogonal. As the response continues, the height of the plateau above its surroundings increases until 130 ms, while activity spreads horizontally to more distant cortical locations. Due to spike threshold, a small difference in the subthreshold response, such as that between the peak region and the plateau region, can of course lead to a large difference in the suprathreshold response, as discussed further in the last paragraph of this subsection. The suprathreshold response, on the second row, starts with a slightly later onset (60 ms), but quickly amplifies to a maximum at 85 ms. Finally, the VSDI is shown on the bottom row as a sum of the subthreshold and suprathreshold activities, with a larger weight for the contribution of the subthreshold activity due to the larger participating membrane area. The cortical response field measured at each point in time is thus accounted for by this simple model. The temporal development of the plateau is paralleled by psychophysical results which may hint at its role in visual processing, as discussed below.
This simple model suggests that the peak zone is the spiking locus, in agreement with the electrophysiological data (Fig. 8 and Supplementary Figs 1 and 2) and that the plateau is the result of full overlap between the stimulus and the subthreshold RF of neurons within it. Within this model, neurons somewhat further than plateau rim still receive direct thalamic input, though weaker, so it strongly predicts orientation-selective signals beyond plateau rim—as indeed observed (see local maps in Fig. 2).
At present, the relationship between the plateau region described here and the "up" membrane potential state (Wilson 1993) is not known. It is possible that the plateau corresponds to a cortical region where many neurons are pushed into an up membrane potential state in response to the visual stimulus because the present data indicate that neurons within plateau rim are stimulated either to the extent that they exhibit spiking (in the more central peak zone) or to the near spike threshold in the plateau. Indeed Petersen, Hahn, et al. (2003) have shown concordance between an elevated VSDI signal and the up state recorded intracellularly. However, alternation between up and down states in cortical neurons is heavily dependent on the animal's vigilance state and type of anesthetic, being readily observed in cat under ketamine–xylazine anesthesia (Steriade et al. 1993; Destexhe and Pare 1999). To the best of our knowledge, the distribution of intracellular membrane potential states has not been described for halothane anesthesia as used here, so it is difficult to draw conclusions about the relationship between our results and the dichotomy of up and down states. Under barbiturate anesthesia and during wakefulness, the mean membrane potential is near that of the ketamine–xylazine up state (Destexhe and Pare 1999). However, 2 membrane potential states do exist under barbiturates, although the potential difference between them is only 2.6 mV on average (Anderson et al. 2000). We conjecture that such a situation may also exist in the halothane-anesthetized and awake animal, although the latter would be much more difficult to detect. Qualitatively, our results are reminiscent of those reported by Anderson et al. (2000), who showed that visual stimulation can increase the fraction of time a neuron stays in the up state, especially if the stimulus is high contrast or close to the neuron's preferred orientation. If indeed 2 closely spaced but distinct states exist in the awake condition, it could be that the plateau region observed here characterizes the cortical extent of neurons pushed into the up state in response to visual stimulation and that it is involved in the perception of local stimuli (see below).
In addition to (and possibly instead of) the passive, feedforward-driven model suggested above, shunting inhibition could be involved in maintaining the plateau zone below threshold, clamping the depolarized response. Indeed, intracellular recordings have shown cases in which depolarized responses are concomitant with reduction in trial-to-trial variability, a hallmark of shunting inhibition (Monier et al. 2003).
Combined Representation of Position and Orientation
Observing the spatial properties of cortical response fields of local oriented gratings allowed us to probe the combined coding of position and orientation. Previous studies have aimed at the important goal of separating the 2 stimulus attributes and characterizing the retinotopy and orientation maps independently and examined whether they were correlated (Das and Gilbert 1997; Bosking et al. 2002). A further important question is how the representations of the 2 attributes are combined in responses to actual stimuli (Jancke 2000), rather than how the maps interact. The question for a local oriented stimulus is—a pure retinotopic (orientation free) representation will peak in one cortical locus according to the retinotopic map, whereas a pure orientation (retinotopy free) representation will peak in a second cortical locus according to the orientation map (a set of iso-orientation patches)—where will the peak response to the stimulus actually lie? Our results show that the closer an imaged pixel is to the peak of the cortical response field, the stronger its tendency to contain neurons preferring the orientation that matches the stimulus (Fig. 3). Interestingly, however, even the very peak rarely coincides fully with the nearest appropriate iso-orientation patch. This is easily seen in Figure 3A, which shows that even the 5% highest pixels are not all "correct" (rightmost value is 0.775, rather than 0.95–1). See also the contour delineating the 5% highest pixels over the orientation map in Figure 2. Therefore, these results suggest that in the representation of a stimulus, there is a compromise between its "correct" retinotopic and orientation mappings, in effect convolving the 2 representations.
Potential Role of Cortical Response Field Plateau
The coding advantage of having a focal spiking peak with response coming mainly from neurons that prefer the orientation of the stimulus is clear because it provides a suprathreshold response that is selective in both the retinotopic and orientation domains. Going beyond the peak of the response, the rim of the plateau is very similar for the cortical response field of different orientations (Fig. 1A,B; thin lines in Fig. 2). This is due to a large orientation nonselective response component, which is probably just below threshold and arises from large subthreshold RFs. What could be the advantage of population coding employing plateau-like orientation nonselective subthreshold activation? Activating all orientations within a given retinal neighborhood to just below threshold means that the response to any stimulus that will appear next within this vicinity will be greatly accelerated. Different features/objects appearing in the same location, motion of an object to a slightly different retinal location, and rotation of the object in its place, are some stimuli the response to which will be greatly accelerated by the characteristics of the cortical response field and that are likely to be important next stimuli to process.
Indeed, extracellular recordings show that responses to moving stimuli exhibit shorter peak latencies than those to stationary stimuli (Jancke, Erlhagen, et al. 2004). Behaviorally, preceding the onset of a long bar stimulus with a square located at one of its ends gives rise to the line-motion illusion—the bar appears to spread from the location of the preceding square (Hikosaka et al. 1993). This effect is maximal when the square precedes the bar by 100–200 ms (Hikosaka et al. 1993), while the present data show that the plateau reaches its maximal height at 130 ms. (Figs 6 and 9). Previous VSDI results from our group suggest that this illusion may involve faster processing at the location of the preceding square than at the other end of the bar due to the gradient of activity set up by the preceding square (Jancke, Chavane, et al. 2004). Our current results imply that the strongest illusory motion is generated from the rim of the plateau outward because within plateau rim the subthreshold activity gradient is small.
In addition, spatial attention has been reported to decrease latencies of visual evoked potentials to task-irrelevant high-contrast gratings at the attended location (Di Russo and Spinelli 1999). Stimulus-driven spatial attention results in faster and more accurate behavioral processing of visual stimuli (Posner 1980), with the effects of this reflexive orienting peaking at around 100–150 ms (Weichselgartner and Sperling 1987; Nakayama and Mackeben 1989), again paralleling the time plateau height reaches maximum in our data (130 ms). Subthreshold plateau-like activity as described here is likely to dynamically organize information about potential stimuli colocalized with a current salient stimulus and about its potential trajectory.
Perhaps, the strongest support for the behavioral relevance of the current VSDI results comes from a recent psychophysical study of attentional capture using direct uninformative cues, that is, cues appearing in possible target locations but conveying no information about the actual target location, showing that multiple cues each accelerate processing of targets at their respective locations (Wright and Richard 2003). These effects were stronger for 100 ms cue-target asynchrony than for 200, 300, and 400 ms intervals. Our present physiological results are in agreement with their suggestion that bottom-up sensory-based processing, as opposed to only top–down effects, contributes significantly to attentional capture under appropriate circumstances, and the current reported cortical response fields represent direct visualization of the hypothesized "activity distributions" that they adopt from LaBerge (LaBerge and Brown 1989; LaBerge et al. 1997).
An important unresolved question is whether voluntary, goal-directed spatial attention can affect the cortical response field via top–down interactions to facilitate behavioral demands, for example, expand or contract it when the spotlight of attention is diffuse or focal, respectively. Such questions remain to be explored using behaving monkeys. The rich cortical response field dynamics presented here demonstrate an interesting processing strategy; the cortex selectively amplifies a portion of the response, while allowing widespread synaptic interactions and priming the surround for enhanced processing of subsequent stimuli by pushing the activation level of the retinotopic vicinity's representation to just below spiking threshold.
|
Acknowledgments |
|---|
We are grateful to R. Hildesheim for synthesizing RH-1691; D. Ettner, N. Mushcatel, Y. Toldedo, and C. Wijnbergen for expert technical and engineering assistance; and D. Backlash-Omer, T. Fekete, and L. Rom for help during the combined VSDI–electrophysiology experiments. We are grateful to A. Sterkin for generously providing his data acquisition and data analysis softwares for spike recordings. Supported by grant from the BMBF, ISF, Minerva, the Goldsmith foundation, and the Grodetsky center. FC was supported by a Marie Curie European Union grant. Conflict of Interest: None declared.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Albus K. (1975) A quantitative study of the projection area of the central and the paracentral visual field in area 17 of the cat. I. The precision of the topography. Exp Brain Res 24:2159–179.[Web of Science][Medline]
Anderson JS, Lampl I, Reichova I, Carandini M, Ferster D. (2000) Stimulus dependence of two-state fluctuations of membrane potential in cat visual cortex. Nat Neurosci 3:6617–621.[CrossRef][Web of Science][Medline]
Arieli A and Grinvald A. (2002) Optical imaging combined with targeted electrical recordings, microstimulation, or tracer injections. J Neurosci Methods 116:115–28.[CrossRef][Web of Science][Medline]
Bonhoeffer T and Grinvald A. (1993) The layout of iso-orientation domains in area 18 of cat visual cortex: optical imaging reveals a pinwheel-like organization. J Neurosci 13:104157–4180.[Abstract]
Bosking WH, Crowley JC, Fitzpatrick D. (2002) Spatial coding of position and orientation in primary visual cortex. Nat Neurosci 5:9874–882.[CrossRef][Web of Science][Medline]
Bringuier V, Chavane F, Glaeser L, Fregnac Y. (1999) Horizontal propagation of visual activity in the synaptic integration field of area 17 neurons. Science 283:695–699.
Buzas P, Kovacs K, Ferecsko AS, Budd JM, Eysel UT, Kisvarday ZF. (2006) Model-based analysis of excitatory lateral connections in the visual cortex. J Comp Neurol 499:6861–881.[CrossRef][Web of Science][Medline]
Carandini M and Ferster D. (2000) Membrane potential and firing rate in cat primary visual cortex. J Neurosci 20:1470–484.
Cynader MS, Swindale NV, Matsubara JA. (1987) Functional topography in cat area 18. J Neurosci 7:51401–1413.[Abstract]
Das A and Gilbert CD. (1997) Distortions of visuotopic map match orientation singularities in primary visual cortex. Nature 387:594–598.[CrossRef][Medline]
Destexhe A and Pare D. (1999) Impact of network activity on the integrative properties of neocortical pyramidal neurons in vivo. J Neurophysiol 81:41531–1547.
Di Russo F and Spinelli D. (1999) Electrophysiological evidence for an early attentional mechanism in visual processing in humans. Vision Res 39:182975–2985.[CrossRef][Web of Science][Medline]
Douglas RJ and Martin KAC. (1991) A functional microcircuit for cat visual cortex. J Physiol (Lond) 440:735–769.
Grinvald A, Anglister L, Freeman JA, Hildesheim R, Manker A. (1984) Real-time optical imaging of naturally evoked electrical activity in intact frog brain. Nature 308:5962848–850.[CrossRef][Medline]
Grinvald A and Hildesheim R. (2004) VSDI: a new era in functional imaging of cortical dynamics. Nat Rev Neurosci 5:11874–885.[CrossRef][Web of Science][Medline]
Grinvald A, Lieke EE, Frostig RD, Hildesheim R. (1994) Cortical point-spread function and long-range lateral interactions revealed by real-time optical imaging of macaque monkey primary visual cortex. J Neurosci 14:52545–2568.[Abstract]
Grinvald A, Manker A, Segal M. (1982) Visualization of the spread of electrical activity in rat hippocampal slices by voltage-sensitive optical probes. J Physiol 333:269–291.
Grinvald A, Shoham D, Shmuel A, Glaser DE, Vanzetta I, Shtoyerman E, Slovin H, Wijnbergen C, Hildesheim R, Sterkin A, et al. (1999) In-vivo optical imaging of cortical architecture and dynamics. In Windhorst U and Johansson H (Eds.). Modern techniques in neuroscience research(Springer-Verlag, New York) pp. 893–969.
Hikosaka O, Miyauchi S, Shimojo S. (1993) Focal visual attention produces illusory temporal order and motion sensation. Vision Res 33:91219–1240.[CrossRef][Web of Science][Medline]
Hubel DH and Wiesel TN. (1962) Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. J Physiol (Lond) 160:106–154.
Jancke D. (2000) Orientation formed by a spot's trajectory: a two-dimensional population approach in primary visual cortex. J Neurosci 20:14RC86.
Jancke D, Chavane F, Naaman S, Grinvald A. (2004) Imaging cortical correlates of illusion in early visual cortex. Nature 428:6981423–426.[CrossRef][Medline]
Jancke D, Erlhagen W, Schoner G, Dinse HR. (2004) Shorter latencies for motion trajectories than for flashes in population responses of cat primary visual cortex. J Physiol (Lond) 556:3971–982.
LaBerge D and Brown V. (1989) Theory of attentional operations in shape identification. Psychol Rev 96:1101–124.
LaBerge D, Carlson RL, Williams JK, Bunney BG. (1997) Shifting attention in visual space: tests of moving-spotlight models versus an activity-distribution model. J Exp Psychol Hum Percept Perform 23:51380–1392.[CrossRef][Web of Science][Medline]
Monier C, Chavane F, Baudot P, Graham LJ, Fregnac Y. (2003) Orientation and direction selectivity of synaptic inputs in visual cortical neurons: a diversity of combinations produces spike tuning. Neuron 37:4663–680.[CrossRef][Web of Science][Medline]
Nakayama K and Mackeben M. (1989) Sustained and transient components of focal visual attention. Vision Res 29:111631–1647.[CrossRef][Web of Science][Medline]
Petersen CCH, Grinvald A, Sakmann B. (2003) Spatiotemporal dynamics of sensory responses in layer 2/3 of rat barrel cortex measured in vivo by voltage-sensitive dye imaging combined with whole-cell voltage recordings and neuron reconstructions. J Neurosci 23:41298–1309.
Petersen CCH, Hahn TTG, Mehta M, Grinvald A, Sakmann B. (2003) Interaction of sensory responses with spontaneous depolarization in layer 2/3 barrel cortex. Proc Natl Acad Sci USA 100:2313638–13643.
Posner MI. (1980) Orienting of attention. Q J Exp Psychol 32:13–25.[Web of Science][Medline]
Pouget A, Dayan P, Zemel RS. (2003) Inference and computation with population codes. Annu Rev Neurosci 26:381–410.[CrossRef][Web of Science][Medline]
Sharon D and Grinvald A. (2002) Dynamics and constancy in cortical spatiotemporal patterns of orientation processing. Science 295:5554512–515.
Shoham D, Glaser DE, Arieli A, Kenet T, Wijnbergen C, Toledo Y, Hildesheim R, Grinvald A. (1999) Imaging cortical dynamics at high spatial and temporal resolution with novel blue voltage-sensitive dyes. Neuron 24:791–802.[CrossRef][Web of Science][Medline]
Steriade M, Nunez A, Amzica F. (1993) A novel slow (< 1 Hz) oscillation of neocortical neurons in vivo: depolarizing and hyperpolarizing components. J Neurosci 13:83252–3265.[Abstract]
Sterkin A, Arieli A, Ferster D, Glaser DE, Grinvald A, Lampl I. (1998) Real-time optical imaging in cat visual cortex exhibits high similarity to intracellular activity. Neurosci Lett 51, S41.
Tusa R, Rosenquist A, Palmer L. (1979) Retinotopic organization of areas 18 and 19 in the cat. J Comp Neurol 185:4657–678.[CrossRef][Web of Science][Medline]
Volgushev M, Pernberg J, Eysel UT. (2000) Comparison of the selectivity of postsynaptic potentials and spike responses in cat visual cortex. Eur J Neurosci 12:1257–263.[CrossRef][Web of Science][Medline]
Weichselgartner E and Sperling G. (1987) Dynamics of automatic and controlled visual attention. Science 238:4828778–780.
Wilson CJ. (1993) The generation of natural firing patterns in neostriatal neurons. Prog Brain Res 99:277–297.[Web of Science][Medline]
Wright RD and Richard CM. (2003) Sensory mediation of stimulus-driven attentional capture in multiple-cue displays. Percept Psychophys 65:6925–938.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J.-Y. Wu, Xiaoying Huang, and Chuan Zhang Propagating Waves of Activity in the Neocortex: What They Are, What They Do Neuroscientist, October 1, 2008; 14(5): 487 - 502. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||