| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1Visual Sciences, Research School of Biological Sciences, Australian National University, Canberra, Australian Capital Territory; and 2School of Medical Sciences and Institute for Biomedical Research, University of Sydney, Sydney, New South Wales, Australia
Submitted 17 April 2007; accepted in final form 26 May 2007
 |
ABSTRACT |
|---|
|
|
|
INTRODUCTION |
|---|
|
, 1962, 1968) identified two types of neurons in the primary visual cortices of cats and monkeys: simple and complex. Subsequently, these cell types have been described in the cortices of many mammalian species (e.g., eutherians: Van Hooser et al. 2005; marsupials: Ibbotson and Mark 2003; Rocha-Miranda et al. 1976). Simple cells have separate ON and OFF subregions in their receptive fields, whereas complex cells have uniform receptive fields where ON and OFF stimuli generate excitatory responses regardless of where in the receptive field they are presented. These two cell types provide the main feedforward input to all subsequent levels of cortical visual processing outside the primary visual cortex, so it is important to understand the mechanisms that generate their spiking responses.
It is common practice to distinguish between simple and complex cells quantitatively by analyzing the spiking responses to moving sinusoidal gratings (cat: Movshon et al. 1978a,b; cat and monkey: Skottun et al. 1991; ferret: Baker et al. 1998; rat: Girman et al. 1999; wallaby: Ibbotson et al. 2005). When a grating of preferred spatial and temporal frequency is moved across the receptive field of a cell in the primary visual cortex there is a component of the response that is phase sensitive (i.e., the spike rate oscillates at the temporal frequency of the grating) and a component that is phase invariant. The F1/F0 ratio is then calculated by dividing the first Fourier coefficient of the response (F1) by the mean time-averaged response (F0). The response of a linearly summating simple cell should oscillate sinusoidally about the mean spontaneous activity of the cell. However, because most simple cells have little or no spontaneous activity the response appears as a half-wave–rectified sinusoid, thus producing F1/F0 ratios >1 (Fig. 1A). If a simple cell behaves as a linear spatiotemporal filter followed by half-wave rectification the F1/F0 ratio would be 1.57 (Tolhurst and Dean 1990). For complex cells, the phase-invariant component of the spiking response dominates such that F1/F0 ratios are <1 (Fig. 1B). The choice of an F1/F0 ratio of one as the dividing line between simple and complex cells is based on large data sets (n >1,500) (Skottun et al. 1991). Cluster analysis has also confirmed that the dividing line between simple and complex cell populations is close to one (Ibbotson et al. 2005). Importantly, classification based on this analysis agrees well with classification based on the spatial distribution of ON and OFF subregions within cells' receptive fields (Bardy et al. 2006; Dean and Tolhurst 1983; Mata and Ringach 2005; Priebe et al. 2004).
|
). The phase-invariant depolarization and the amplitude of the oscillatory component both increase with increasing contrast. Furthermore, in complex cells the depolarization associated with the phase-invariant response component has a higher contrast gain than the voltage modulation (Anderson et al. 2000; Carandini and Ferster 1997, 2000). Contrast adaptation generates a phase-invariant hyperpolarization accompanied by a smaller reduction in the amplitude of the voltage modulation at the stimulus frequency (Carandini and Ferster 1997, 2000; Sanchez-Vives et al. 2000). However, neither stimulus contrast nor adaptation appears to affect the absolute spike threshold (Anderson et al. 2000; Carandini and Ferster 2000). Mechler and Ringach (2002) demonstrated that even if the V1/V0 ratio is unimodal, the F1/F0 ratio can be bimodal because of the highly nonlinear relationship between the membrane potential and spiking output. Priebe et al. (2004) tested this hypothesis by comparing V1/V0 and F1/F0 ratios from intracellular recordings in cat cortex. They confirmed that the V1/V0 ratio was unimodal, whereas the F1/F0 ratio was bimodal. The differences between the intracellular and spiking distributions could be fully explained by the nonlinear relationship between membrane potential and firing rate.
In the present paper we test some predictions of the spike-threshold hypothesis by measuring the F1/F0 ratios of cells in the cat primary visual cortex during two experimental manipulations. First, we altered the contrast of the grating stimulus and, second, we adapted the cells using prolonged high-contrast motion stimulation. Considering both previously published intracellular recording data and the spike threshold hypothesis we predict that decreasing stimulus contrast or adapting to high contrast will act to pull the membrane potential down toward the spiking threshold (Fig. 1B, bottom row). Due to the nonlinear effect of the spike threshold, both stimulus manipulations are expected to generate an increase in the F1/F0 ratio of the spiking responses of complex cells. A far smaller effect would be expected in simple cells because the intracellular membrane potential is in a compressive range where changes in the effective spike threshold should have minimal effects (Mechler and Ringach 2002). Tolhurst and Dean (1990) measured F1/F0 ratios for five simple cells in area 17 of the cat that were stimulated by moving gratings of different contrasts and reported that F1/F0 ratios changed very little with contrast, although they did not present any complex cell data. We recorded from 205 neurons located in areas 17 and 18 of cat primary visual cortex and show that cells defined as complex at 100% contrast became more phase sensitive as contrast declined, but the phase sensitivity of cells defined as simple at 100% contrast remained unchanged across contrasts. Furthermore, based on the preceding definitions, complex but not simple cells became more phase sensitive following con trast adaptation.
|
|
METHODS |
|---|
|
Nine cats of both sexes were prepared as described previously (Crowder et al. 2006; Price et al. 2006). Briefly, single-unit recordings were made from areas 17 and 18 in anesthetized and paralyzed cats. Initially, the trachea and right cephalic vein were cannulated under ketamine HCl (20 mg/kg, administered intramuscularly) anesthesia. Further surgery was performed with high halothane doses (1–2%). Anesthesia during unit recordings was provided by inhaled halothane (0.5%) and a 2:1 ratio of N2O and O2. The level of anesthesia was monitored by recording electrocardiograms and electroencephalograms. All methods regarding the animal preparation were approved by the animal ethics committee of the Australian National University.
Extracellular recordings and visual stimuli
Extracellular recordings were made with lacquer-coated tungsten microelectrodes (FHC, Bowdoinham, ME) driven by a piezoelectric drive (Burleigh inchworm and 6000 controller, Burleigh Instruments, Victor, NY). Extracellular signals from individual units were isolated, amplified, and filtered, then acquired with a CED1401 interface and Spike2 software sampled at 40 kHz (Cambridge Electronic Designs, Cambridge, UK).
After the dominant eye and receptive field location of each neuron had been determined, the nondominant eye was covered and testing was performed with visual stimuli produced by a VSG Series 2/5 stimulus generator (Cambridge Research Systems, Cambridge, UK) and presented on a calibrated monitor (Eizo T662-T, 100-Hz refresh, 1,024 x 768 pixels) at a viewing distance of 57 cm. Sine-wave gratings were presented in a circular aperture surrounded by a gray background of mean luminance (Lum; 50 cd/m2). Sine-wave contrast is defined as: Michelson contrast = [(Lummax – Lummin)/(Lummax + Lummin)] x 100 where Lummax and Lummin are the maximum and minimum luminance of the grating. For all cells, the location and size of the classical receptive field (RF), as well as the preferred orientation/direction, spatial frequency (SF), and temporal frequency (TF) were determined by calculating on-line tuning functions using 100% contrast gratings. Off-line, SF and TF tuning curves were fit using a least-squares algorithm to a skewed-Gaussian function
 |
Receptive field size was determined in two ways: 1) a circular patch of moving grating centered on the middle of the RF was expanded in size to find the diameter that produced the largest response and 2) an annulus of moving grating was centered on the RF and the diameter of the empty center was decreased until spiking responses could just be elicited, thus indicating the edge of the excitatory RF. The first measure found the diameter at which the response saturated, whereas the second method found the outer boundary of the excitatory RF. Both measures usually agreed but, when they differed, the result from method 2 was used to set the stimulus size.
The experimental stimuli were sine-wave gratings of optimal orientation, SF, and TF presented in a circular aperture the size of the classical receptive field. Nonadapted contrast response functions were collected by presenting each contrast (4, 8, 12, 16, 24, 32, 48, 64, 82, 100%) in random order for 1- or 0.5-s tests (10 repetitions) interleaved with 4 s of mean luminance. A 3-min recovery period was allowed between collecting the data for every contrast response function. For adapted contrast response functions, an adapting stimulus was presented for 60 s followed by 0.5-s tests (aforementioned contrast levels for 10 repetitions) interleaved with 4-s-adaptation top-ups. The default contrast for the adapting stimulus was 32%, but lower (16%) or higher (100%) contrasts were used if the contrast response function produced after 32% adaptation was very compressed or changed very little, respectively.
Data analysis
F1/F0 ratios were calculated from the responses produced by the moving sine-wave grating during the 1- or 0.5-s test interval. F0 values were calculated as the mean firing rate (averaged across repeats and time) above spontaneous and F1 values were the first Fourier coefficient of the mean response (averaged across repeats) obtained using the FFT function in MATLAB. Due to the short test window, our analysis included only cells with optimum temporal frequencies of 
2 Hz. We were concerned that this short sampling time may have placed too much emphasis on the onset transient. Therefore we examined TF tuning curves that always used 1-s sampling windows and found that F1/F0 ratios that were calculated with a start time 100 ms after stimulus onset to exclude the onset transient were strongly correlated with F1/F0 ratios that were calculated with a start time at the stimulus onset (r2 = 0.85). Based on this correlation, it is clear that including the onset transient did not systematically alter the F1/F0 ratios. Therefore to include cells with optimum TFs as low as 2 Hz in our contrast adaptation analysis, while still calculating Fourier transforms based on at least one full cycle, we chose to calculate F1/F0 ratios with the start time at the stimulus onset. For 98% of our data the presentation time was an integer multiple of the stimulus period. For the remaining 2% of cells, F1/F0 analysis was performed on a section of the recorded response that was an integer multiple of the stimulus period.
We used two independent criteria to ensure the accuracy of the F1/F0 analysis at low contrasts. The first criterion addressed whether mean stimulus-evoked spike rates statistically exceeded the spontaneous firing rate. First, spontaneous firing rates were calculated from the response to the 4 s of mean luminance interleaved between tests in the nonadapted contrast stimulus. A Poisson distribution was then calculated from this spontaneous spiking. We calculated a response threshold based on the 99% cutoff from the Poisson fit to the spontaneous rate (Fig. 2, A and C, dotted horizontal lines). F0 responses that did not exceed this Poisson threshold for 20 ms were not included in the main F1/F0 analysis (Figs. 3–7).
|
|
|
Control experiments
Previous research has shown that the RF size and SF tuning of macaque V1 neurons can change with contrast (Sceniak et al. 1999, 2002). In our main experiment the aperture size of the stimulus was fixed across contrasts. Therefore we also wanted to determine how F1/F0 ratios change with contrast when the stimulus size as well as SF and TF of the test grating were optimized at each contrast. For a subgroup of cells, RF size and both SF and TF tuning were tested at 8, 16, and 100% contrasts. For each cell, F1/F0 ratios were then calculated from responses to two tests. The first test presented gratings in a circular aperture at 8, 16, and 100% contrasts with grating parameters (aperture size, SF, and TF) obtained using 100% contrast gratings. This stimulus had parameters identical to those of the normal experimental procedure. The second test again presented gratings in a circular aperture at 8, 16, and 100% contrasts, but used the optimal grating parameters for each contrast. In a separate control experiment the grating's contrast, SF, and TF were kept constant, but the stimulus size was varied in three steps between the largest and smallest RF sizes obtained for each cell. This experiment was repeated using 4, 8, and 16% contrasts. Each stimulus in these control experiments was presented for 32 repetitions.
Histology
At the end of recording sessions electrolytic lesions were placed (40 µA, 10 s, electrode positive) 2, 4, and 6 mm below the brain surface at a location displaced 3 mm medially from the recording site. This procedure generated clear lesions that could be correlated for depth with the recording tracks, which were clearly visible in all preparations. Small physical lesions were produced at the bottom of each track using the method of Crowder et al. (2006) to allow accurate localization of the deepest cell location. Lesions were not made at every recording site because this was found to greatly reduce the recording quality of the electrode. Animals were given a lethal dose of pentobarbitone sodium (100 mg/kg) and immediately perfused transcardially with saline (0.9%) followed by 10% formol saline. Brains were extracted and processed in the standard fashion (Crowder et al. 2006). Sectioned tissue was examined using light microscopy to confirm the locations of electrode tracks. Drawings of the sectioned brains were made and orientations of the electrode penetrations reconstructed (taking careful note of the deepest recording site). Cell locations were then determined by working backward from the deepest point toward the shallowest recording using shrinkage-adjusted measurements based on the recording depths noted on the piezoelectric drive. The borders of the cortical layers were identified by visual inspection. Lamination data are presented in Fig. 9. Each cell location was first converted to a percentage of the total thickness of the lamina in which it resides. This made the cells comparable from the same layer but different cats and thus different layer thicknesses. Layers were then plotted as a fraction of the total cortical thickness, also called the normalized cortical depth (see Hawken et al. 1988; Ringach et al. 2002).
|
|
|
|
RESULTS |
|---|
|
Simple-cell responses depend on the spatial phase (or location) of the stimulus, whereas complex cell responses are spatial-phase invariant (Movshon et al. 1978a,b). Here we classify simple and complex cells based on the ratio between the first Fourier coefficient (F1) and mean spiking responses (F0) to moving sine-wave gratings (Skottun et al. 1991). F1/F0 ratios >1 or <1 indicate simple or complex cells, respectively. We classified a cell as simple or complex based on its response to 100% contrast gratings.
F1/F0 ratio and stimulus contrast
Spike density functions plotting the responses of a typical simple cell to stimulation at four contrasts are shown in Fig. 2A. At all tested contrasts the cell shows highly oscillatory responses with bursts of spikes that coincide in time with the fundamental temporal frequency of the stimulus. Figure 2B shows that there is no consistent change in the F1/F0 ratio between contrasts, suggesting that contrast has no influence on this cell's classification. The response of a complex cell to four contrasts is shown in Fig. 2C. This cell also shows oscillatory responses at the F1 frequency, but at high contrasts they are small relative to the mean response (F0). Note that as contrast decreases, the amplitude of the oscillations remains quite steady, although the relative amplitude of the F0 component decreases. Consequently, the F1/F0 ratio increases with decreasing contrast, such that this particular cell had an F1/F0 ratio of 0.05 at maximum contrast but a ratio of 0.76 when the contrast was 8% (Fig. 2, C and D).
In this section, we assessed the responses of 22 simple and 159 complex cells using sine-wave gratings of different contrasts. The mean F1/F0 ratios across simple cells are shown by the solid triangles in Fig. 3A. Changes in contrast did not significantly alter the mean F1/F0 ratios of the simple cell population (two-way ANOVA, P > 0.92). On the contrary, the complex cell population showed a clear trend toward increased F1/F0 ratios at lower stimulus contrasts (Fig. 3A, solid circles). For complex cells, there was a significant relationship between contrast and F1/F0 ratio (two-way ANOVA, P << 0.001). F1 and F0 values were also examined separately for both complex and simple cells, as shown in Fig. 3, B and C, respectively. For complex cells, the average normalized F0 values (Fig. 3B, solid squares) increased with increasing contrast in a sigmoid-like fashion, typically observed in studies of contrast adaptation (e.g., Crowder et al. 2006; Ohzawa et al. 1982, 1985). The F1 values (open squares) for the complex cell population showed fairly small changes in amplitude with increasing contrast. The large increases in F0 amplitude seen in complex cells, accompanied by relatively small increases in F1 amplitude, are consistent with decreasing F1/F0 ratios with increasing contrast. For the simple cell population, both F1 and F0 values increased proportionally with increasing contrast (Fig. 3C), which is consistent with the relatively steady F1/F0 ratios across contrasts.
Scatterplots showing changes in the F1/F0 ratios of individual cells are shown in Fig. 4. This population data were used to generate the plots in Fig. 3. For complex cells, there is a clear trend toward higher F1/F0 ratios at lower contrasts (population migrates above the line of equality as contrast declines). The increase in F1/F0 ratios for the complex cell population compared with values at 100% contrast was significant for all test contrasts 
32% (paired t-test, P < 0.01), with the remaining points being insignificant (paired t-test; 48%, P > 0.3, 64%, P > 0.21, and 82%, P > 0.87). The changes in the F1/F0 ratios for the simple cell population were not significant for any test contrasts (paired t-test, P > 0.3). Note that at some test contrasts a few cells that had been classified as simple at 100% contrast had an F1/F0 ratio <1. For all but one of these points, the reduced ratio did not fall outside the SE error bars of the other nine contrasts tested and therefore was not a significant trend.
|
Prolonged exposure to moving gratings generates contrast adaptation in the great majority of neurons in the primary visual cortex (Crowder et al. 2006; Ohzawa et al. 1982, 1985). This is manifest as a reduction in firing rate to most test contrasts, particularly those at low values. Perceptually, the impact of adaptation is to reduce the apparent contrast of the stimulus (Blakemore et al. 1973; Hammett et al. 1994). Given that our previous data showed that low contrasts lead to higher F1/F0 ratios, we examined whether contrast adaptation could increase F1/F0 ratios. That is, does contrast adaptation increase the phase sensitivity of cells in a similar way as lowering stimulus contrast?
Adapted contrast response functions were obtained using a top-up contrast adaptation protocol (see METHODS). The F1/F0 ratios of simple and complex cells following contrast adaptation are shown, respectively, as open triangles and circles in Fig. 3A. When in the adapted state, simple cells did not significantly change their F1/F0 ratios for any test contrast (Fig. 3A; two-way ANOVA, P > 0.22). However, complex cells showed significant increases in their F1/F0 ratios following adaptation for most test contrasts (Fig. 3A; two-way ANOVA, P << 0.001). Note that for the lowest test contrast (4%) there is no increase in F1/F0 ratio after adaptation, suggesting a saturation effect.
F1 and F0 values after adaptation were plotted separately for complex and simple cells classified at 100% contrast (Fig. 3, D and E, respectively). For complex cells, both F1 and F0 values undergo response gain control (compression) and contrast gain control (rightward shift of the contrast response function) after contrast adaptation (Fig. 3D). Similar to the nonadapted condition, F0 values show larger changes with contrast than F1 values. F1 and F0 values are more similar in the adapted state than in the unadapted state, which matches the observed upward shift in the F1/F0 ratios after adaptation (Fig. 3A). For the simple cell population, both F1 and F0 values also undergo changes in response gain and contrast gain following contrast adaptation (Fig. 3E). However, as in the nonadapted condition F1 and F0 values decreased proportionately, thereby producing no change in F1/F0 ratios following contrast adaptation.
Data from areas 17 and 18 were pooled because neurons in both brain areas show similar changes in F1/F0 ratios with changing contrast and adaptation. A four-way ANOVA using brain area, simple or complex classification at 100% contrast, contrast, and adaptation as variables revealed that neither the effect of contrast (P > 0.99) nor that of adaptation (P > 0.54) on F1/F0 ratio was statistically different between areas 17 and 18. The only significant difference between brain areas was that complex cells in area 17 had slightly higher F1/F0 ratios than those in area 18 (P < 0.03).
In the preceding results, simple and complex cells were classified by their F1/F0 ratios to 100% contrast. However, it is evident that the population distribution of simple and complex cells can change dramatically depending on the conditions under which the F1/F0 ratio is calculated. Figure 5A shows histograms of cell count against the F1/F0 ratio obtained using 100% contrast gratings for the 205 cells in our sample. Twenty-two cells (11%) had F1/F0 ratios >1. Of these cells, 91 also gave significant responses at 8% contrast and these are highlighted in Fig. 5A as gray histograms. It is evident that most cells in this subgroup were classified as complex when stimulated at 100% contrast (i.e., 81 of 91). When tested at 8% contrast, 27 of the 91 cells (30%) had F1/F0 ratios >1 (Fig. 5B). It is clear that stimulus contrast has a significant influence on the F1/F0 ratios of cells. How does adaptation influence the population data? In Fig. 6, A–C we plot histograms of cell count against F1/F0 ratio for 77 neurons from which F1/F0 ratios were calculated under three conditions: unadapted with a 100% test contrast stimulus (Fig. 6A), unadapted using a 12% test contrast stimulus (Fig. 6B), and using a test contrast of 12% after adaptation to 32% contrast (Fig. 6C). At 100% test contrast there is a clear oversampling of complex cells, but eight (10%) simple cells are evident at F1/F0 ratios >1. When the F1/F0 ratios were calculated from responses collected with a 12% test contrast the overall distribution shifts toward higher ratios and the number of cells with F1/F0 ratios >1 increases to 13 (17%; Fig. 6B). Finally, when the F1/F0 ratio is calculated at the same low test contrast but with neurons in an adapted state, we see a large shift toward F1/F0 ratios >1 (34 cells, 44%). Thus when tested at relatively low contrast and in an adapted state, 26 neurons previously categorized as complex cells now have F1/F0 ratios >1. A similar, albeit weaker, trend was evident when a 32% test contrast was used (Fig. 6, D–F, n = 126). The population shift toward more cells with F1/F0 ratios >1 with lower contrast and with adaptation to 32% contrast is evident, even though the effect at 32% contrast is smaller than that with a 12% test contrast (compare Figs. 3 and 6).
|
|
Sceniak et al. (1999) demonstrated that the extent of spatial summation in macaque V1 neurons shrinks at high contrasts. Furthermore, although the preferred SF does not change with contrast, SF tuning bandwidth is narrower at low contrasts (Sceniak et al. 2002). In our main experiment the stimulus aperture size was fixed across contrasts. Therefore it was reasonable to ask whether F1/F0 ratios would still increase with decreasing contrasts if parameters for each test grating (aperture size, SF, TF) were optimized for that contrast.
We analyzed data from 24 area 17 neurons whose RF size, SF tuning, and TF tuning were tested at 8, 16, and 100% contrast. Consistent with the findings of Sceniak et al. (1999, 2002), the size of the classical receptive fields decreased with increasing contrast (P < 0.021, one-way ANOVA), and the preferred SF (xpref, see METHODS) did not vary with contrast (P > 0.49, one-way ANOVA). Figure 7A plots F1/F0 ratios generated by testing with 8% (filled circles) and 16% (open circles) contrast as a function of F1/F0 ratios generated using 100% contrast. For this graph, the grating parameters of aperture size, SF, and TF were constant across contrasts and were set by values obtained from 100% contrast test stimuli. Most points (65%) are above the line of equality, indicating that, similar to the main data set, F1/F0 ratios tend to increase with decreasing contrasts. The data plotted in Fig. 7B were obtained using optimal aperture sizes, SFs, and TFs for each contrast. Again, most points (71%) are above the line of equality, indicating that F1/F0 ratios tend to increase with decreasing contrasts. A two-way ANOVA was used to look at the effects of contrast and grating parameter optimization on F1/F0 ratios that had been normalized to the 100% test. This analysis revealed that F1/F0 ratios increase significantly with decreasing contrasts (P < 0.022), the effect of which did not depend on whether grating parameters were optimized for each contrast (P > 0.26).
To further examine the effect of stimulus size on F1/F0 ratio we presented each cell with gratings of constant contrast, SF, and TF, but at three different aperture sizes. Only neurons whose classical receptive field size increased monotonically with decreasing contrast were included in this analysis. Large medium, and small aperture sizes were the RF sizes obtained with 8, 16, and 100% contrasts, respectively. Figure 7C plots F1/F0 ratios calculated from firing in response to the large aperture size on the abscissa and F1/F0 ratios calculated from firing in response to the medium (solid symbols) and small (open symbols) aperture sizes on the ordinate. This experiment was repeated using three contrasts (circles: 4%; squares: 8%; triangles: 16%). Stars represent centroids of the distributions for the medium (solid) and small stimulus sizes (open), respectively. A two-way ANOVA using aperture size and contrast as variables revealed that changes in aperture size (P < 0.93) did not produce consistent changes in F1/F0 ratios at any contrast (P < 0.76). Also note that at a given contrast the increased aperture size generated increased firing rates. A two-way ANOVA using size and contrast as variables revealed that firing rate increased with both contrast (P < 0.0003) and aperture size (P << 0.01). Therefore for a particular contrast the F1/F0 ratio was stable, even if the firing rate was adjusted by changing stimulus size.
Spike count analysis
Figure 8 summarizes the results of the second independent analysis, which compared measured F1/F0 ratios with those produced by chance alone (see METHODS). Figure 8, A and B shows the spike count plotted against complex cell F1/F0 ratios when the stimulus gratings had contrasts of 4 and 100%, respectively. As expected, across the population, 100% contrast stimulation elicited a greater number of spikes during the stimulus period than did 4% contrast. The shaded area represents the middle 95% of the distribution of F1/F0 ratios expected from different numbers of randomly timed spikes during the stimulus interval. Interestingly, the mean value of the F1/F0 ratio is predicted to increase as the spike count decreases, even though the same random process is used to assign spikes. Thus even a random process could produce increased F1/F0 ratios as contrast (and thus the spike count) is decreased. For a real effect to be present, increases in F1/F0 at low contrasts must be larger than predicted by our stochastic simulations (we will therefore term these ratios significant). In fact, at 4% contrast 67% of cells had significant F1/F0 ratios, indicating that for most cells the increase in F1/F0 at low contrasts could not be accounted for by chance alone. At 100% contrast 54% of the total population of cells had significant F1/F0 ratios. Most (76%) cells that had significant F1/F0 ratios at 100% contrast (Fig. 8, A and B, squares) remained outside the shaded region at 4% contrast.
We then reconsidered the effects of our stimulus manipulations if we include only the significant F1/F0 ratios. Figure 8C plots the average F1/F0 ratio at different contrasts using only complex cells that were shown to have significant F1/F0 ratios. As in Fig. 3A, solid circles show average nonadapted F1/F0 ratios at different contrasts and open circles show average F1/F0 ratios following contrast adaptation. The shape of this curve is qualitatively similar to the one plotted in Fig. 3A, although it is shifted upward toward higher F1/F0 ratios. Interpretation of this result must be tempered with common sense. The responses of complex cells to high contrasts (which elicit high spike counts) are expected to be phase invariant and therefore many bona fide complex cells with insignificant F1 components at high contrasts have been removed using the spike count criterion. Because of the restrictive nature of the spike count criterion we believe that it serves as a complement to, rather than a replacement for, the previous analysis.
To further show that low spike counts could not explain the higher F1/F0 ratios observed at low contrasts we reexamined our control data in which increased spike counts were produced by larger-size stimuli with the same contrast. Figure 8D plots spike count against complex cell F1/F0 ratios for contrasts of 4% (white), 8% (gray), and 16% (black) with small (circles), medium (diamonds), and large (squares) aperture sizes. Spike count was poorly correlated with F1/F0 ratio for 4% (R2 = 0.010), 8% (R2 = 0.012), and 16% (R2 = 0.15) contrasts, which demonstrates that F1/F0 ratios have only slight dependence on spike count for a fixed contrast.
Laminar organization
We have reconstructed the locations of the recording sites for all 205 cells relative to the laminar organization of the cortex. Figure 9 plots the locations of recording sites relative to the normalized cortical depths for each layer, as derived from nine cats (for methodology, see Hawken et al. 1988; Ringach et al. 2002). When using 100% contrast gratings as stimuli 22 cells had F1/F0 ratios >1 and were therefore classified as simple cells (Fig. 9A). One of the simple cells (5% of the population) was located in layer 2, three in layer 3 (14%), eight in layer 4 (36%), and five each in layers 5 and 6 (23% each). It is also noteworthy that 17 of 21 complex cells with F1/F0 ratios >0.6 were clustered in or very close to the border of layer 4. Thus cells that are not formally classified as simple, yet have clear phase-sensitive response oscillations, are also primarily located in or very close to layer 4. Most simple cells in our population (59%) were located in layers 4 and 6.
Figure 9, B and C plots the layer distributions of a population of cells (n = 62) whose F1/F0 ratios were measured at 100% (Fig. 9B) and 4% (Fig. 9C) contrasts, respectively. For stimulation at 100% contrast, seven simple cells (11%) are evident. All of these cells are in layers 3–6 (Fig. 9B). When stimulated with 4% contrast the cell population shows a dramatic shift to higher F1/F0 ratios. That is, at low contrast 33 cells (53%) have F1/F0 ratios >1 and they reside in layers 2–6. The percentage of cells with F1/F0 ratios >1 increases by 400–500% in layers 3–6 when stimulated with a 4% contrast grating relative to a 100% contrast grating.
|
|
DISCUSSION |
|---|
|
), these latter neurons are classified as simple cells regardless of the contrast used (Tolhurst and Dean 1990). Our results suggest a strict new classification system for cortical cells. To be classified as a simple cell a neuron must have an F1/F0 ratio >1 for all stimulus contrasts that generate significant responses, when stimulated at its optimum spatial frequency. Conversely, to be classified as a complex cell a neuron must have an F1/F0 ratio <1 at 100% contrast but can have any F1/F0 ratio at other contrasts. It will be interesting in the future to cross-reference this strict definition with other accepted classification methods (e.g., Skottun et al. 1991).
The contrast and adaptation effects we observe can be predicted from previous intracellular recording data and the spike threshold model (Anderson et al. 2000; Carandini and Ferster 1997, 2000; Mechler and Ringach 2002; Priebe et al. 2004). Intracellular recordings have shown that the modulation ratio of the membrane potential (V1/V0) is unimodal, but skewed toward low V1/V0 ratios (Priebe et al. 2004). Around 40% of cells defined as simple using extracellular F1/F0 ratios also have V1/V0 ratios >1. Furthermore, another 25% of cells defined as simple using F1/F0 ratios have V1/V0 ratios >0.6, so 65% of simple cells have highly phase sensitive membrane oscillations (Priebe et al. 2004). Conversely, from the same data set only around 5% of cells defined extracellularly as complex have V1/V0 ratios >0.6. Thus there is a strong bias toward complex cells having only small-amplitude membrane oscillations in response to moving gratings. Consequently, the bimodal distribution of spiking modulation ratios (F1/F0) arises from two sources: 1) the degree to which the membrane potential oscillates in response to a moving grating, which can be manipulated by the amount of converging, spatially offset inputs (Chance et al. 1999; Hubel and Wiesel 1962; Tao et al. 2002), and 2) the bimodal distribution can arise from the nonlinear relationship between membrane potential and firing rate produced by the spike threshold (Mechler and Ringach 2002; Priebe et al. 2004). The latter is of particular importance for cells that have very low V1/V0 ratios but have F1/F0 ratios >1.
Importantly, the V0 values of complex cells show stronger contrast gain than the V1 and contrast adaptation preferentially decreases the V0, whereas the V1 is barely affected (Carandini and Ferster 1997, 2000; Sanchez-Vives et al. 2000). The spiking data for the complex cell in Fig. 2C (see also Fig. 3B) also indicate that decreasing contrast affects the F0 much more than the F1. Thus from both intracellular and extracellular data, it is clear that the nonoscillating component of the response (the V0, F0, or DC components) is much more sensitive to changes in contrast and adaptation than the phase-locked oscillating component (the V1 or F1). Other stimulus parameters also change V1/V0 and F1/F0 ratios, notably the spatial frequency of the gratings (Movshon et al. 1978a,b; Priebe et al. 2004; Skottun et al. 1991). Complex cells have modulated responses at low spatial frequencies because bar widths are similar to the receptive field widths. On the contrary, simple cells can have phase-invariant responses at high spatial frequencies because their ON and OFF zones are covered by many relatively narrow bright and dark bars, thus removing strongly modulated input. Thus it is important to remember that the discussion of the contrast dependence of V1/V0 and F1/F0 ratios is based strictly on stimulation at the optimum spatial frequency of the stimulated cell, which does not change with contrast (Sceniak et al. 1999, 2002).
The division of neurons in the primary visual cortex into simple and complex types has prompted numerous models to explain the spiking behavior of these two cell classes. In the hierarchical scheme, simple cells receive a restricted range of spatial-phase signals from the dorsal lateral geniculate nucleus (dLGN) and complex cells receive a wider range of spatial-phase information either directly from the dLGN or by simple cells (Hubel and Wiesel 1962). This differing input structure could lead to the phase dependence of simple cell responses and the spatial-phase invariance of complex cells. More recently, recurrent models have been proposed where simple and complex cells share the same basic cortical circuit, but have low and high levels of phase-invariant recurrent amplification, respectively (Chance et al. 1999; Tao et al. 2004). However, circuitry is not the primary concern of the present paper because both models have the net effect of pooling signals with different spatial phases. Furthermore, both models must include a transformation from membrane potential to spike rate. Instead, here we show that complex cell classification must be refined because complex F1/F0 ratios <1 are guaranteed only at high contrasts (the contrast-specific phase sensitivity of complex cells has functional implications, which we subsequently discuss). Moreover, it has been shown recently that phase sensitivity of complex cells in cat primary visual cortex is influenced by feedback signals from the posterotemporal visual cortex and by the stimulus encroaching into the suppressive extraclassical receptive field surround (Bardy et al. 2006). Indeed, increasing F1/F0 ratios are positively correlated with the relative strength of the silent surround suppression of responses to stimulation within the classical excitatory receptive field. Collectively, these recent findings suggest that the classification system for cells in the primary visual cortex must be modernized, as suggested earlier at the beginning of this discussion. Moreover, models of cortical function need to incorporate these new findings.
From our finding that F1/F0 ratios of complex cells increase at low contrasts and after adaptation we predict that the spatially homogeneous receptive fields (i.e., spatially overlapping ON–OFF subregions) of complex cells may develop more distinct ON and OFF subregions at low contrast and following adaptation. That is, the spatial structure of the classical receptive field of complex cells at low contrasts might more closely resemble the spatial structure of simple cells. Using sparse noise stimuli, Priebe et al. (2004) showed that cortical cells are more simple-like in their spiking responses than in their membrane potential responses. Manipulations such as lowering the contrast of the stimulus or contrast adaptation, which lower the mean membrane potential during stimulation so it is closer to the spike threshold, are expected to accentuate this effect. It will be interesting to test this prediction using reverse-correlation stimuli with low contrasts or when the cells are in an adapted state. These techniques provide a direct measure of the spatial structure of cortical cell receptive fields (e.g., Mata and Ringach 2005; Priebe et al. 2004). However, one pitfall of reverse-correlation methods such as white-noise analysis is that the low spatial correlation of the stimulus usually produces low firing rates. Lowering the contrast of these types of stimuli would lower the firing rates even further, perhaps to levels where no spikes can be recorded. It will be a challenge in the future to develop reverse-correlation stimuli that elicit responses robust enough that they can be used at low contrasts (but see Touryan et al. 2005). Intracellular recording techniques could also be used to monitor changes in the synaptic input at low contrasts and after adaptation directly (e.g., Priebe et al. 2004).
Implications for cortical architecture
The finding that the F1/F0 ratios of the units classified as simple cells at 100% contrast did not show any contrast or adaptation dependence suggests that these cells might form a distinct cell class. Recent data from cat cortex combining receptive field mapping and intracellular labeling techniques have conclusively shown that simple cells defined as having separate ON and OFF subregions reside in layer 4 and the upper part of layer 6 (Martinez et al. 2005). Layer 4 and its borders, along with the upper region of layer 6, receive direct thalamic input (for review see Payne and Peters 2005). Martinez et al. (2005) conclude that "simple receptive fields represent an exclusive feature of the first stages of cortical processing." We have identified the recording sites for all cells in our population and correlated them with cortical layers and depths. Using F1/F0 ratios to classify simple and complex cells at 100% contrast, we found that around 60% of simple cells were located in layers 4 and 6. Moreover, a high percentage of simple cells in layer 5 were close to the border of layer 4 and the great majority of complex cells with high F1/F0 ratios (>0.6) were also found in layer 4. Thus from our data layer 4 has an overrepresentation of highly phase sensitive cells when tested using 100% contrast gratings. When stimulated at low contrasts, it is also layer 4 that shows by far the largest increase in the number of cells with high F1/F0 ratios. We see in Fig. 9 that the proportion of cells that have F1/F0 ratios >0.6 increases from 22% when the stimulus is at 100% contrast to 77% when the stimulus is at 4% contrast. Our results are strongly supportive of the concept that simple receptive fields and highly phase sensitive complex cells are prominent in geniculorecipient layers and represent early stages of cortical processing.
How do the preceding observations correlate with the spike-threshold hypothesis, which suggests that the division of cells into two types arises from the nonlinear interaction of spike threshold with membrane potential responses? The spike-threshold hypothesis suggests a continuum of cell types at the intracellular level (Mechler and Ringach 2002). That is, V1/V0 ratios have a unimodal distribution (Priebe et al. 2004), yet this unimodal distribution becomes bimodal at the extracellular level (i.e., F1/F0 ratios) arising from the biophysical properties of the spike threshold. However, the degree of phasicity of the membrane potential responses does play an important role. In fact, Priebe et al. (2004) showed that no cells that are classified as complex based on F1/F0 ratios have V1/V0 ratios >1. Therefore all complex cells already have relatively low oscillatory responses in their membrane potentials, even before the nonlinearities of the spike threshold take effect. This fact suggests to us that these cells have undergone phase averaging through some mechanism and represent a "higher" level of processing than simple cells. In contrast, most simple cells have quite large V1/V0 ratios (i.e., 65% have V1/V0 ratios >0.6; Priebe et al. 2004). We interpret this result to mean simple cells are at a lower level of processing than complex cells. Whether this is because they receive exclusive input from the LGN (as in the classical hierarchical model; Hubel and Wiesel 1962) or because they receive weaker recurrent excitation from their neighboring cortical cells (as in the recurrent model; Chance et al. 1999), cells that have F1/F0 ratios >1 at all contrasts appear to form an anatomically segregated cell class whose phase sensitivity is not as malleable as complex cells. By combining these concepts with the observation that simple cells reside in the thalamorecipient layers of the cortex, our results provide support for the concept that there is an anatomical hierarchy, with cells in layers receiving direct thalamic input exhibiting more phasic response properties.
Functional considerations
The distribution of contrast levels in natural images peaks at low contrasts: 0–25% (Balboa and Grzywacz 2000; Chirimuuta et al. 2003; Ruderman and Bialek 1994; Tadmor and Tolhurst 2000; Vu et al. 1997). Our present results show that the phase sensitivity of complex cells increased substantially at these contrasts. Moreover, cortical cells rapidly adapt to the prevailing contrast, so under natural viewing conditions it can be expected that most neurons will be in an adapted state, which we show increases phase sensitivity even further. Therefore we predict that the primary visual cortex is dominated by cells with phase-sensitive response properties when viewing scenes containing naturalistic contrast distributions. This shift to simple-like responses may be advantageous because under certain circumstances simple cells may transmit information at faster rates and carry more edge-like or line-like information than complex cells (Mechler et al. 2002; Reich et al. 2001). However, there is also evidence that simple and complex cells do not differ greatly in their ability to discriminate spatial features (Mechler et al. 2002). Finally, it makes the most sense to consider the processing of natural visual scenes in an awake and freely behaving animal. Attention and behavioral state could strongly influence the activity of early visual cortex, thereby altering the contrast- and adaptation-dependent effects observed in the present study (Kagan et al. 2002; McAdams and Reid 2005).
|
|
GRANTS |
|---|
|
|
|
ACKNOWLEDGMENTS |
|---|
|
|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: M. R. Ibbotson, Visual Sciences, Research School of Biological Sciences, Australian National University, Canberra, ACT, Australia 2601 (E-mail: Michael.Ibbotson{at}anu.edu.au)
|
|
REFERENCES |
|---|
|
Baker GE, Thompson ID, Krug K, Smyth D, Tolhurst DJ. Spatial-frequency tuning and geniculocortical projections in the visual cortex (areas 17 and 18) of the pigmented ferret. Eur J Neurosci 10: 2657–2668, 1998.[CrossRef][Web of Science][Medline]
Balboa RM, Grzywacz NM. Occlusions and their relationship with the distribution of contrasts in natural images. Vision Res 40: 2661–2669, 2000.[CrossRef][Web of Science][Medline]
Bardy C, Huang JY, Wang C, Fitzgibbon T, Dreher B. "Simplification" of responses of complex cells in cat striate cortex: suppressive surrounds and "feedback" inactivation. J Physiol 574: 731–750, 2006.
Blakemore C, Muncey JP, Ridley RM. Stimulus specificity in the human visual system. Vision Res 13: 1915–1931, 1973.[CrossRef][Web of Science][Medline]
Carandini M, Ferster D. A tonic hyperpolarization underlying contrast adaptation in cat visual cortex. Science 276: 949–952, 1997.
Carandini M, Ferster D. Membrane potential and firing rate in cat primary visual cortex. J Neurosci 20: 470–484, 2000.
Chance FS, Nelson SB, Abbott LF. Complex cells as cortically amplified simple cells. Nat Neurosci 2: 277–282, 1999.[CrossRef][Web of Science][Medline]
Chirimuuta M, Clatworthy PL, Tolhurst DJ. Coding of the contrasts in natural images by visual cortex (V1) neurons: a Bayesian approach. J Opt Soc Am A 20: 1253–1260, 2003.[CrossRef][Web of Science]
Crowder NA, Price NS, Hietanen MA, Dreher B, Clifford CW, Ibbotson MR. Relationship between contrast adaptation and orientation tuning in V1 and V2 of cat visual cortex. J Neurophysiol 95: 271–283, 2006.
Dean AF, Tolhurst DJ. On the distinctness of simple and complex cells in the visual cortex of the cat. J Physiol 344: 305–325, 1983.
Girman SV, Sauvé Y, Lund RD. Receptive field properties of single neurons in rat primary visual cortex. J Neurophysiol 82: 301–311, 1999.
Hammett ST, Snowden RJ, Smith AT. Perceived contrast as a function of adaptation duration. Vision Res 34: 31–40, 1994.[CrossRef][Web of Science][Medline]
Hawken MJ, Parker AJ, Lund JS. Laminar organization and contrast sensitivity of directionally-selective cells in the striate cortex of the old world monkey. J Neurosci 8: 3541–3548, 1988.[Abstract]
Hubel DH, Wiesel TN. Receptive fields of single neurones in the cat's striate cortex. J Physiol 148: 574–591, 1959.
Hubel DH, Wiesel TN. Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. J Physiol 160: 106–154, 1962.
Hubel DH, Wiesel TN. Receptive fields and functional architecture of monkey striate cortex. J Physiol 195: 215–243, 1968.
Ibbotson MR, Mark RF. Orientation and spatiotemporal tuning of cells in the primary visual cortex of an Australian marsupial, the wallaby Macropus eugenii. J Comp Physiol A Sens Neural Behav Physiol 189: 115–123, 2003.[Web of Science][Medline]
Ibbotson MR, Price NS, Crowder NA. On the division of cortical cells into simple and complex types: a comparative viewpoint. J Neurophysiol 93: 3699–3702, 2005.
Kagan I, Gur M, Snodderly DM. Spatial organization of receptive fields of V1 neurons of alert monkeys: comparison with responses to gratings. J Neurophysiol 88: 2557–2574, 2002.
Martinez LM, Wang Q, Reid RC, Pillai C, Alonso JM, Sommer FT, Hirsch JA. Receptive field structure varies with layer in the primary visual cortex. Nat Neurosci 8: 372–379, 2005.[CrossRef][Web of Science][Medline]
Mata ML, Ringach DL. Spatial overlap of ON and OFF subregions and its relation to response modulation ratio in macaque primary visual cortex. J Neurophysiol 93: 919–928, 2005.
McAdams CJ, Reid RC. Attention modulates the responses of simple cells in monkey primary visual cortex. J Neurosci 25: 11023–11033, 2005.
Mechler F, Reich DS, Victor JD. Detection and discrimination of relative spatial phase by V1 neurons. J Neurosci 22: 6129–6157, 2002.
Mechler F, Ringach DL. On the classification of simple and complex cells. Vision Res 42: 1017–1033, 2002.[CrossRef][Web of Science][Medline]
Movshon JA, Thompson ID, Tolhurst DJ. Receptive field organization of complex cells in the cat's striate cortex. J Physiol 283: 79–99, 1978a.
Movshon JA, Thompson ID, Tolhurst DJ. Spatial summation in the receptive fields of simple cells in the cat's striate cortex. J Physiol 283: 53–77, 1978b.
Ohzawa I, Sclar G, Freeman RD. Contrast gain control in the cat visual cortex. Nature 298: 266–268, 1982.[CrossRef][Medline]
Ohzawa I, Sclar G, Freeman RD. Contrast gain control in the cat's visual system. J Neurophysiol 54: 651–667, 1985.
Payne BR, Peters A. The concept of cat primary visual cortex. In: The Cat Primary Visual Cortex, edited by Payne BR, Peters A. San Diego, CA: Academic Press, 2002, p. 1–129.
Price NSC, Crowder NA, Hietanen MA, Ibbotson MR. Neurons in V1, V2, and PMLS of cat cortex are speed tuned but not acceleration tuned: the influence of motion adaptation. J Neurophysiol 95: 660–673, 2006.
Priebe NJ, Mechler F, Carandini M, Ferster D. The contribution of spike threshold to the dichotomy of cortical simple and complex cells. Nat Neurosci 7: 1113–1122, 2004.[CrossRef][Web of Science][Medline]
Reich DS, Mechler F, Victor JD. Formal and attribute-specific information in primary visual cortex. J Neurophysiol 85: 305–318, 2001.
Ringach DL, Shapley RM, Hawken MJ. Orientation selectivity in macaque V1: diversity and laminar dependence. J Neurosci 22: 5639–5651, 2002.
Rocha-Miranda CE, Linden R, Volchan E, Lent R, Bombardieri RA Jr. Receptive field properties of single units in the opossum striate cortex. Brain Res 104: 197–219, 1976.[CrossRef][Web of Science][Medline]
Ruderman DL, Bialek W. Statistics of natural images: scaling in the woods. Phys Rev Lett 73: 814–817, 1994.[CrossRef][Web of Science][Medline]
Sanchez-Vives MV, Nowak LG, McCormick DA. Membrane mechanisms underlying contrast adaptation in cat area 17 in vivo. J Neurosci 20: 4267–4285, 2000.
Sceniak MP, Hawken MJ, Shapley R. Contrast-dependent changes in spatial frequency tuning of macaque V1 neurons: effects of a changing receptive field size. J Neurophysiol 88: 1363–1373, 2002.
Sceniak MP, Ringach DL, Hawken MJ, Shapley R. Contrast's effect on spatial summation by macaque V1 neurons. Nat Neurosci 2: 733–739, 1999.[CrossRef][Web of Science][Medline]
Skottun BC, De Valois RL, Grosof DH, Movshon JA, Albrecht DG, Bonds AB. Classifying simple and complex cells on the basis of response modulation. Vision Res 31: 1079–1086, 1991.[CrossRef][Web of Science][Medline]
Tadmor Y, Tolhurst DJ. Calculating the contrasts that retinal ganglion cells and LGN neurones encounter in natural scenes. Vision Res 40: 3145–3157, 2000.[CrossRef][Web of Science][Medline]
Tao L, Shelley M, McLaughlin D, Shapley R. An egalitarian network model for the emergence of simple and complex cells in visual cortex. Proc Natl Acad Sci USA 101: 366–371, 2004.
Tolhurst DJ, Dean AF. The effects of contrast on the linearity of spatial summation of simple cells in the cat's striate cortex. Exp Brain Res 79: 582–588, 1990.[Web of Science][Medline]
Touryan J, Felson G, Dan Y. Spatial structure of complex cell receptive fields measured with natural images. Neuron 45: 781–791, 2005.[CrossRef][Web of Science][Medline]
Van Hooser SD, Heimel JA, Nelson SB. Functional cell classes and functional architecture in the early visual system of a highly visual rodent. Prog Brain Res 149: 127–145, 2005.[Web of Science][Medline]
Vu TQ, McCarthy ST, Owen WG. Linear transduction of natural stimuli by dark-adapted and light-adapted rods of the salamander, Ambystoma tigrinum. J Physiol 505: 193–204, 1997.
This article has been cited by other articles:
|
|
 |
|
  N. A. Crowder, M. A. Hietanen, N. S. C. Price, C. W. G. Clifford, and M. R. Ibbotson Dynamic contrast change produces rapid gain control in visual cortex J. Physiol., September 1, 2008; 586(17): 4107 - 4119. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. van Kleef, R. Berry, and G. Stange Directional Selectivity in the Simple Eye of an Insect J. Neurosci., March 12, 2008; 28(11): 2845 - 2855. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A. Hietanen, N. A. Crowder, N. S. C. Price, and M. R. Ibbotson Influence of adapting speed on speed and contrast coding in the primary visual cortex of the cat J. Physiol., October 15, 2007; 584(2): 451 - 462. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
TOP
ABSTRACT
INTRODUCTION
