Neuronal responses to orientation and motion ... - Semantic Scholar

Visual Neuroscience (1999), 16, 587–600. Printed in the USA. Copyright © 1999 Cambridge University Press 0952-5238099 $12.50

Neuronal responses to orientation and motion contrast in cat striate cortex

SABINE KASTNER, HANS-CHRISTOPH NOTHDURFT, and IVAN N. PIGAREV AG Neurobiology, Max Planck Institute for Biophysical Chemistry, 37070 Göttingen, Germany (Received September 1, 1998; Accepted December 14, 1998)

Abstract Responses of striate neurons to line textures were investigated in anesthetized and paralyzed adult cats. Light bars centered over the excitatory receptive field (RF) were presented with different texture surrounds composed of many similar bars. In two test series, responses of 169 neurons to textures with orientation contrast (surrounding bars orthogonal to the center bar) or motion contrast (surrounding bars moving opposite to the center bar) were compared to the responses to the corresponding uniform texture conditions (all lines parallel, coherent motion) and to the center bar alone. In the majority of neurons center bar responses were suppressed by the texture surrounds. Two main effects were found. Some neurons were generally suppressed by either texture surround. Other neurons were less suppressed by texture displaying orientation or motion (i.e. feature) contrast than by the respective uniform texture, so that their responses to orientation or motion contrast appeared to be relatively enhanced (preference for feature contrast). General suppression was obtained in 33% of neurons tested for orientation and in 19% of neurons tested for motion. Preference for orientation or motion contrast was obtained in 22% and 34% of the neurons, respectively, and was also seen in the mean response of the population. One hundred nineteen neurons were studied in both orientation and motion tests. General suppression was correlated across the orientation and motion dimension, but not preference for feature contrast. We also distinguished modulatory effects from end-zones and flanks using butterfly-configured texture patterns. Both regions contributed to the generally suppressive effects. Preference for orientation or motion contrast was not generated from either end-zones or flanks exclusively. Neurons with preference for feature contrast may form the physiological basis of the perceptual saliency of pop-out elements in line textures. If so, pop-out of motion and pop-out of orientation would be encoded in different pools of neurons at the level of striate cortex. Keywords: Cat striate cortex, Single-cell recordings, Classical receptive field, Contextual response modulation, Pop-out

oriented line elements (Nothdurft & Li, 1984, 1985; Gilbert & Wiesel, 1990; Knierim & Van Essen, 1992; Lamme, 1995; Zipser et al., 1996; Kastner et al., 1997; Nothdurft et al., 1999), or random dot patterns (Allman et al., 1990). Knierim and Van Essen (1992) studied responses to bars presented in a neuron’s RF and surrounded by an array of bars at the same or orthogonal orientation. They found that a substantial number of neurons responded more strongly to the bar in the RF when it was surrounded by orthogonal bars. Perceptually, the central bar was highly salient in that condition, “popping out” from the texture background. Similarly, responses at the border of dissecting segments (“texture segmentation”) were found to be stronger than responses to homogeneous texture fields in monkey V1 (Nothdurft et al., 1992; Gallant et al., 1995). Other investigators have found stronger responses to texture elements belonging to a figure than to those belonging to the background (Lamme, 1995; Zipser et al., 1996; but see Rossi et al., 1998). These findings indicate that contextual response modulation in the monkey may contribute to fundamental visual processes such as detecting salient stimuli or segregating figures from ground at early processing stages.

Introduction The modulation of neural responses by stimuli presented outside the classical receptive field (RF) has been demonstrated in various species and at various levels of the visual system (see Allman et al., 1985). In striate cortex of cats and monkeys, contextual modulation has been demonstrated using various stimuli such as sinusoidal gratings (Blakemore & Tobin, 1972; Maffei & Fiorentini, 1976; Fries et al., 1978; Nelson & Frost, 1978; DeAngelis et al., 1992; Li & Li, 1994; Sillito et al., 1995; Sengpiel et al., 1997), noise patterns (Hammond & MacKay, 1981; Hammond & Smith, 1982; Gulyas et al., 1987; Orban et al., 1987), arrays of

Correspondence and reprint requests to: Sabine Kastner, Laboratory of Brain and Cognition, National Institute of Mental Health, Bldg. 49, Room 1B80, Bethesda, MD 20892, USA. Present address of Sabine Kastner: Laboratory of Brain and Cognition, National Institute of Mental Health, National Institutes of Health, Bldg. 49, Room 1B80, Bethesda, MD 20892, USA. Present address of Ivan N. Pigarev: Division of Psychology, Faculty of Science, The Australian National University, ACT 0200, Australia.

587

588 Pop-out and texture segmentation have been studied extensively in human psychophysics. The saliency of target elements in these tasks is seen only with certain feature properties of the target such as line orientation (Treisman & Gelade, 1980; Treisman, 1986). It also depends on the local feature contrast between the target and surrounding distractors (Nothdurft, 1991, 1992, 1993). This dependence has been shown for several feature properties such as line orientation, direction of motion, color, luminance, and stereopsis (Nothdurft 1992, 1993, 1995). It has been hypothesized that the perception of pop-out and texture segmentation in humans is based on neurons with preference for feature contrast similar to those found in monkey striate cortex (Knierim & Van Essen, 1992; Nothdurft, 1994; Sillito et al., 1995; Nothdurft et al., 1999). We have investigated neural responses to pop-out of motion and orientation in cat striate cortex. Cats are able to detect texture boundaries defined by local discontinuities in line orientation (Wilkinson, 1986). Moreover, their performance in texture segmentation tasks is similar to human performance (De Weerd et al., 1992). Therefore, the neural mechanisms underlying texture segregation and related perceptual phenomena such as pop-out can be studied in cats and the results can be related to human texture vision. In a previous study, we have reported that several striate neurons responded more strongly to salient elements in texture patterns with motion or orientation contrast than to nonsalient elements in uniform textures (Kastner et al., 1997). In the present study, we will provide a more detailed analysis of response modulation induced by texture surrounds. We used texture patterns that were similar to those used in human psychophysics (e.g. Nothdurft, 1985) and similar to those previously studied in monkeys (Knierim & Van Essen, 1992). Stimuli were composed of a central bar presented in a neuron’s RF and similar bars in the surround that either had the same orientation or direction of motion as the center bar, or displayed orientation or motion contrast. We sought to address three major questions. First, we wanted to compare contextual effects in the orientation and in the motion dimension. Second, we wanted to compare the effects in cats with those obtained in monkeys. And third, we wanted to investigate the contribution of end-zones and flanks to contextual response modulation in order to yield insights into the underlying mechanisms. Abstracts of this work have been published (Kastner et al., 1995a,b). Methods Animal preparation Experiments were carried out on 13 adult cats in acute preparations. The animals were initially anesthetized with an intramuscular injection of xylazine (Rompunt; 2 mg0kg) and ketamine hydrochloride (Ketanestt; 5–15 mg0kg). All standard surgical procedures (venae sectio, tracheotomy, head fixation in a stereotaxic frame, craniotomy) were done under this medication, and, when necessary, supplemented by pentobarbital (Nembutalt; 3– 4 mg0kg i.v.) after venae sectio. Subsequently, anesthesia was maintained with a 70%030% mixture of N2O0O2 , with the addition of 0.5–1.0 mg0kg0h pentobarbital to a continuous intravenous infusion of glucose in Ringer’s solution. To prevent brain edema prednisolone (Solu-Decortint; 1 mg0h) was administered. When a stable level of anesthesia was reached with elimination of reflexes and autonomous responses to painful stimuli, the animal was paralyzed with a loading dose of gallamine triethiodide (Flaxedilt, 20–30 mg0kg) followed by a continuous infusion of Flaxedilt (10 mg0kg0h). The

S. Kastner, H.-C. Nothdurft, and I.N. Pigarev anesthesia was maintained until the end of the experiment, which could last up to 120 h. End-tidal CO 2 , EKG, EEG, and body temperature were continuously monitored during the experiment. CO2 was kept around 4% and body temperature was held around 37.58C. The animal’s blood pressure was stabilized with an intravenous infusion of plasmaexpander (Hemohaest 6%; 3 ml0h) if necessary. Pupils were dilated by topical application of atropine and nictitating membranes retracted with neosynephrine. The eyes were covered with contact lenses (0 diopter) and artificial pupils of 4-mm diameter were fixed close to the lenses. Eye background was backprojected on a screen and blind spots were marked for analysis of RF eccentricity. Refraction was measured with a Rodenstock refractometer and corrected to monitor distance with appropriate lenses in front of the eyes. The clarity of the eye media was checked every few hours. Single-unit activity was recorded with varnished tungsten electrodes inserted vertically through small bone holes (,2 mm in diameter) and through the dura at Horsley-Clarke coordinates P 2-4 and L 1-2. Neural activity was monitored on oscilloscopes and loudspeakers, and action potentials were transformed into unitary signals and fed into the interface of a computer. Visual stimulation Visual stimuli were presented on a 17-inch raster monitor (60-Hz frame rate) at a distance of 57 cm from the animal’s eyes. Texture stimuli used in this study were composed of line arrays as shown in Fig. 1. A bar was centered in the “classical” receptive field (indicated by the rectangle) and surrounded by similar bars outside the RF. In the motion test (Figs. 1A and 1B), bars in the surround had the same orientation as the bar in the RF, and moved in the opposite direction (motion contrast) or the same direction (uniform motion). All bars moved at the same velocity. In the orientation test (Figs. 1C and 1D), bars in the surround were oriented either orthogonal to the center bar (orientation contrast) or parallel to it (uniform texture). The role of end-zones and flanks was investigated using “butterfly”-configured texture stimuli (Figs. 1E and 1F) similar to those used by Knierim and Van Essen (1992). The spacing between elements in the texture raster was generally adjusted to be 1.5-fold of the length of the individual bars. Only if texture surrounds alone evoked a strong excitatory response, the spacing (and hence the distance of the surround bars from the RF center) was increased. This was the case for less than 10% of the neurons. The position of individual bars within a pattern was jittered up to 20% of the distance between bars except for the center line, which was always presented at the same position; the jitter was refreshed for each new stimulus presentation. Textures contained bright elements of 55 cd0m 2 on a dark background of 3.7 cd0m 2 . Stimuli were presented monocularly. Experimental procedure and data analysis After a single unit had been isolated using a window-discriminator, the length, width, preferred orientation, and direction of motion of the optimal bar were roughly determined, and the borders of the excitatory RF were plotted under mouse control. Then, the RF center was determined more exactly in test series with varying bar positions along the two axes of the RF. Also the preferred orientation and the optimal bar length and width were quantitatively estimated: 15-deg or 30-deg steps were used for orientation estimates, and length and width were varied between 0.2 and 4.0 times

Orientation and motion contrast in cat striate cortex

589

Fig. 1. Texture patterns used as visual stimuli. Patterns were composed of bright bars (55 cd0m 2 ) on dark background (3.7 cd0m 2 ) on a 17-inch monitor. The “center bar” was placed inside the “classical” RF (rectangle) at fixed position, “surround bars” (same length and width as the center bar) were arranged outside the RF with a positional jitter (20% of element spacing) refreshed for each trial. A, B: Stimuli for the motion test. Center and surround bars moved either in opposite directions (A, motion contrast) or in the same direction (B, uniform motion) as indicated by the arrows. All texture elements had the same orientation, and moved at the same velocity. C, D: Stimuli for the orientation test. Surround bars were either orthogonal (C, orientation contrast) or parallel to the center bar (D, uniform texture). All bars were presented stationarily. E, F: Butterfly stimuli for selective stimulation of end-zones or flanks of the RF. Only the two conditions with orientation contrast are shown. Analogous butterflies were constructed for the uniform condition and for both motion textures.

of the preliminary estimates. Movement of bar stimuli was always orthogonal to bar orientation. In both the motion and the orientation test, responses to center bars alone, to uniform textures, to contrast textures, and to surrounds alone were recorded. Test series included stimuli with optimal center bars (C; preferred orientation in the orientation test; preferred orientation and direction of motion in the motion test) as well as nonoptimal center bars (C9; orthogonal orientation in the orientation test; preferred orientation, but nonpreferred direction of motion in the motion test). The following briefings will be further used for the different conditions (cf. sketches in Figs. 2 and 3): center bars presented alone (C, C9), uniform textures (CS, C9S9), contrast textures (CS9, C9S), and texture surrounds presented alone (S, S9). In our nomenclature, primes always refer to the nonoptimal stimulus conditions. All patterns were presented in pseudorandomized order and each condition was repeated 10–20 times. Stimuli of the orientation test were presented for 600 ms. In the motion test, stimuli were first presented stationarily for 500 ms, then the pattern moved for 230– 600 ms depending on the velocity

Fig. 2. Examples of response modulation in the motion test. A–C: Mean discharge rates and standard errors of the mean (S.E.M.) for three neurons to stimulus conditions outlined below (surrounds contained many more elements than shown; center and surround motion as indicated by arrows). A: General suppression (GS). Responses to the optimal center bar were similarly suppressed by both surrounds. B: Preference for motion contrast (MC). A larger response was obtained if the surround moved in opposite direction to the center bar than when it moved in the same (preferred) direction. C: Preference for uniform motion (UF). The response to coherent motion was larger than that to motion contrast. The texture surrounds alone did not evoke a response.

590

S. Kastner, H.-C. Nothdurft, and I.N. Pigarev preference of the cell (as determined qualitatively); amplitudes reached from 0.5 deg for neurons with small RFs (length , 1.5 deg) to 1.0 deg for neurons with larger RFs (1.5–3.5 deg). The movement of the center bar usually started and ended outside the CRF. Activity before stimulus onset (spontaneous firing) and after stimulus offset (off-activity) was also recorded. Neurons with RFs larger than 3.5 deg were not studied to exclude tests with too few texture elements. Data analysis was based on the mean discharge rates during stimulus presentation in the orientation test, or the movement interval in the motion test, shifted by a 50-ms delay for cortical response latency. Mean spontaneous firing rates were subtracted. Various indices were used for the quantitative analysis of contextual effects, which will be introduced below. An end-stopping index (ESI ) was derived for each neuron by calculating the difference between the response to the optimal bar and the response to the longest bar tested (usually four times the optimal length), normalized to the response to the optimal bar. Orientational and directional selectivity indices (OSI, DSI ) were derived from the responses (R) to the respective optimal (C ) and nonoptimal (C9) center bar conditions in these tests. OSI 5 ~R C 2 R C 9 !0R C , DSI 5 ~R C 2 R C 9 !0R C . Results We have recorded from 169 striate neurons with RFs up to 12 deg of eccentricity. Responses to uniform texture and to texture displaying orientation or motion contrast were compared (motion or orientation test). The motion test was performed in 135 neurons, the orientation test in 153 neurons. In 119 neurons, both tests were completed thus allowing for direct comparison of data in the two dimensions. Types of response modulation

Fig. 3. Examples of response modulation in the orientation test. A–C: Mean discharge rates and S.E.M. of three neurons to the stimulus conditions outlined at the bottom. A: General suppression (GS). The response to the center bar was suppressed by surrounds at either orientation. B: Preference for orientation contrast (OC). Responses to textures with orientation contrast were stronger than those to uniform textures. C: General enhancement (GE). Both texture surrounds enhanced the response to the center bar. Texture surrounds alone did not evoke a response.

In the following paragraph, we will describe the major types of contextual response modulation qualitatively. Figs. 2 and 3 show examples of contextual response modulation. The histograms in each figure show mean discharge rates of individual neurons tested in the stimulus conditions outlined at the bottom. It should be noticed that many more bars were presented in the surround than shown in these outlines (cf. Fig. 1). Stimulation conditions were considered reliably distinct if responses differed by at least 2 S.E.M. With the neuron in Fig. 2A, responses to a center bar moving in the preferred direction (C ) were moderately suppressed by both texture surrounds (CS and CS9). Responses to nonoptimal center bars were not modified by the surrounds. The neuron in Fig. 3A showed complete suppression of responses to optimal and nonoptimal center bars induced by both texture surrounds (CS, CS9, C9S9, C9S ). With the neuron in Fig. 2B, responses to the optimal center bar were suppressed by uniform texture motion (CS ), but enhanced by motion contrast (CS9), resulting in a considerable response difference between the two conditions. The neuron in Fig. 3B was suppressed by both texture surrounds; however, the uniform condition induced stronger suppression, so that the response to the contrast condition appeared to be relatively enhanced. In both neurons (Figs. 2B and 3B), the responses to nonoptimal center bars were suppressed by both texture surrounds. In some neurons, responses to uniform texture were stronger than those to texture

Orientation and motion contrast in cat striate cortex displaying feature contrast. With the neuron in Fig. 2C, uniform motion enhanced the responses to the center bar while motion contrast induced suppression. In very few neurons responses were enhanced in both surround conditions (cf. Fig. 3C). Despite the strong modulatory effects illustrated in Figs. 2 and 3, texture surrounds presented alone did not evoke strong responses in these neurons. In general, the types of response modulation obtained in the motion and in the orientation test were similar and can be summarized as follows (exact classification criteria are given below): (1) general suppression (GS): Both uniform and contrast texture conditions suppressed the responses to the center bar to a similar amount (e.g. in Figs. 2A and 3A); (2) preference for motion contrast (MC) or orientation contrast (OC): Contrast patterns evoked stronger responses than uniform patterns (e.g. in Figs. 2B and 3B); (3) preference for uniform patterns (UF): Uniform patterns evoked stronger responses than contrast patterns (e.g. in Fig. 2C); (4) general enhancement (GE): Both uniform and contrast conditions enhanced the responses to the center bar to a similar amount (e.g. in Fig. 3C). Various combinations of these types were found in conditions with optimal and nonoptimal center bars. It was generally not possible to predict from the surround effects obtained for one center line the effect to be obtained for the other. The following analysis is focused on responses to optimal center bars. We classified responses using criteria of Nothdurft et al. (1999) that were similar to those of Knierim and Van Essen (1992). The obtained distributions of response categories are shown in Fig. 4. In both the motion and the orientation test, about the same proportion of neurons was modified by texture surrounds (Motion: 860135, 64%; Orientation: 970153, 63%). Differential effects (MC, OC, UF) were obtained more often in motion (44%) than in orientation tests (28%). In contrast, nondifferential effects (GS, GE) were seen more frequently in orientation (35%) than in motion tests (19%). MC cells were found more often than OC cells (34% vs. 22%), whereas GS cells were more frequent in orientation than in motion tests (33% vs. 19%). Only three neurons showed general enhancement (GE cells), and only in the orientation test. The differences in both distributions were greater than expected by chance (x 2 ; motion: P , 0.001, orientation: P , 0.001). Averaged suppression In most neurons, texture surrounds suppressed the center bar responses. To quantify these effects, we computed the averaged suppression index (ASI; Knierim & Van Essen, 1992) for patterns with optimal center bars. This measure estimates the averaged amount of suppression induced by both texture surrounds normalized to the response to the center bar alone: ASI 5 1 2 ~~R CS 1 R CS9 !02!0R C ,

R 5 mean responses.

Positive values indicate suppression, negative values enhancement (21 5 100% response enhancement). For example, the ASI of the neuron in Fig. 3A was 1.00, indicating complete suppression of the center bar response by both texture surrounds, whereas the ASI for the neuron in Fig. 3C was 20.55, indicating a mean enhancement of 55% compared to the response to the center bar. It should be noticed that opposite effects such as enhancement induced by one texture condition and suppression induced by the other may be canceled out by this measure, such as for the neurons in Figs. 2B

591

Fig. 4. Frequency distribution of response categories for contextual modulation. Distributions of contextual effects in the motion (A) and in the orientation test (B). Neurons were grouped into five classes according to the response modulation obtained for optimal center bars. Numbers give percentages and the count of neurons in each class. Responses of about two-thirds of all neurons were modified by texture surrounds. The main effects were preference for motion or orientation contrast (MC, OC) and general suppression (GS). UF: preference for uniform texture; GE: general enhancement; and NO: no modulatory effect.

and 2C. The ASIs were 0.02 for each, indicating no averaged suppression effect. The distributions of ASIs for neurons in the motion test and in the orientation test are shown in Fig. 5. Responses to center bars were suppressed to a similar degree by texture surrounds in motion and orientation tests. The means of the distributions were 0.31 6 0.03 S.E.M. for motion and 0.37 6 0.03 for orientation. These averages deviated statistically significant from 0 (one-sample t-tests; P , 0.0001); means for motion and for orientation were not significantly different (t-test on independent samples, P 5 0.15). Averaged suppression was strongest in GS cells (ASI means: 0.70 6 0.05 for motion, 0.68 6 0.04 for orientation), but was also seen in cells with preference for motion or orientation contrast (ASI means: 0.37 6 0.04 for MC cells, 0.45 6 0.05 for OC cells). Differential firing The differences in responses to uniform and to contrast patterns were quantified by the differential firing index: DFI 5 ~R CS9 2 R CS !0R C ,

R 5 mean responses.

592

Fig. 5. Distribution of averaged suppression indices (ASI ) for neurons in the motion (A) and in the orientation test (B). Calculations were based on the responses to optimal center bars; suppression is indicated by positive, enhancement by negative values. In both tests, a similar amount of averaged suppression was obtained; the means were 0.31 (motion) and 0.37 (orientation).

(cf. Knierim & Van Essen, 1992). Positive values indicate preference for orientation or motion contrast, negative values preference for uniform patterns. For example, the DFIs for the neurons in Figs. 2B and 2C were 0.67 and 20.67, respectively, indicating response differences of 67% between the surround conditions compared to the responses to the center bar alone. The distributions of DFIs obtained in the motion and in the orientation test are shown in Figs. 6 and 7. In each figure, distributions are given for the total sample of neurons (Figs. 6A and 7A) and for two subsamples, neurons with differential effects (Figs. 6B and 7B) and neurons with nondifferential effects (Figs. 6C and 7C). One-sample t-tests were used to assess a significant deviation of mean values from 0. DFI means were 0.52 6 0.04 S.E.M. for MC cells (Fig. 6B) and 0.37 6 0.03 for OC cells (Fig. 7B) (P , 0.0001). For UF neurons, means were 20.66 6 0.09 for motion (Fig. 6B; P , 0.0001) and

S. Kastner, H.-C. Nothdurft, and I.N. Pigarev

Fig. 6. Distribution of differential firing indices (DFI ) in the motion test. Calculations were based on test conditions with center bars moving in the preferred direction. Positive values indicate preference for motion contrast, negative values preference for uniform motion. Distributions are shown for the total cell sample (A) and different subgroups of cells (B, C). In MC cells and in the total sample, but not in the group of GS cells, DFIs were significantly shifted toward preference for motion contrast.

20.43 6 0.05 for orientation (Fig. 7B; P , 0.0001). Thus, differential effects turned out to be, on average, stronger in the motion than in the orientation test (t-test on independent samples, P , 0.01). Means of GS cells were distributed around 0 (Figs. 6C and 7C). DFI means of the total samples were 0.13 6 0.04 for motion (Fig. 6A; P , 0.001) and 0.06 6 0.02 for orientation (Fig. 7A; P , 0.05). Comparison of modulation effects across dimensions Fig. 8 shows the ASI (A) and DFI values (B) for 119 neurons tested in both the orientation and the motion test. The vast majority of


Fig. 7. Distribution of differential firing indices (DFI ) in the orientation test. Calculations were based on test conditions with optimally oriented center bars. Positive values indicate preference for orientation contrast, negative values preference for uniform texture. Distributions are shown for the total cell sample (A), and for neurons with (B) or without differential effects (C). In OC cells and in the total sample, but not in the group of GS and GE cells, DFIs were significantly shifted toward preference for orientation contrast.

neurons showed suppression with both tests (Fig. 8A; upper-right quadrant); and in most neurons, this suppression was similarly strong (correlation coefficient r 5 0.62, P , 0.0001). Only few neurons showed mean enhancement in one test and suppression in the other (Fig. 8A; upper-left and lower-right quadrants). In contrast, differential effects were not related across dimensions (Fig. 8B; r 5 0.19, P 5 0.08), although several cells showed preference for the contrast conditions in both tests (Fig. 8B; upperright quadrant). Of the 119 neurons, 26 (22%) were classified as OC cells and 43 (36%) as MC cells. Thirteen neurons were classified as both

593

Fig. 8. Comparison of contextual effects in the motion and in the orientation tests. ASI (A) and DFI values (B) for all neurons studied in both tests. A: The average amount of suppression induced by texture surrounds was similar for many neurons. B: Differential effects obtained in the motion and in the orientation tests were not related across neurons.

OC and MC. To exclude the possibility that neurons without preferences for feature contrast distorted the picture outlined in Fig. 8, data were replotted for the subpopulations of OC and MC cells in Fig. 9. For both subpopulations, the results were remarkably similar to those for the entire population tested across dimensions: ASI values showed good correlation (orientation: r 5 0.58, P , 0.01; motion: r 5 0.55, P , 0.001), whereas DFI values showed no correlation (orientation: r 5 0.20, P 5 0.33; motion: r 5 0.11, P 5 0.45). Normalized population responses To get an estimate of the mean responses of striate cells, we analyzed the population responses of our sample. For each neuron, responses were normalized to the response to the optimal center

594


Fig. 9. Comparison of contextual effects for OC cells and MC cells. As for the entire population tested across dimensions, ASI values (A) showed good correlation, whereas DFI values (B) were not related.

bar presented alone. The normalized responses were then averaged for all cells studied in the motion (Fig. 10) or orientation test (Fig. 11). Mean values were computed for two subgroups of the population, for the neurons showing preference for orientation or motion contrast (MC, OC cells; Figs. 10A and 11A) and for all other cells (Figs. 10B and 11B), and also for the total cell samples (Figs. 10C and 11C). MC and OC cells responded, on average, 41% and 36% more strongly to contrast than to uniform texture for optimal center bars (Figs. 10A and 11A). Responses of MC cells to center bars moving in the nonpreferred direction were not significantly modified. OC cells were generally suppressed when tested with orthogonal center bars. The responses of all other cells tested for motion showed general suppression, with a slightly, but not significantly stronger response to uniform texture, when tested with optimal center bars. Responses to nonoptimal center bars were not affected (Fig. 10B). All other neurons in the orientation test revealed general suppression for both center stimulus conditions (Fig. 11B). The total sample of neurons showed preferences for the contrast conditions: In the motion test, these responses were 11% stronger than to the uniform motion texture for the preferred center bar movement, and 7% stronger for nonpreferred direction (Fig. 10C). In the orientation test, responses were 7% stronger than to uniform textures for optimal center bars (Fig. 11C). No difference was seen for the

orthogonal center bars. The preference for contrast conditions was statistically significant (paired t-test; motion: t 5 22.99, P , 0.05; orientation: t 5 23.14, P , 0.05). Contribution of end-zones and flanks To investigate whether contextual response modulation originates from specific regions surrounding the excitatory RF such as inhibitory end-zones or inhibitory flanks, several neurons were tested with “butterfly” stimuli (Figs. 1E and 1F; cf. Knierim & Van Essen, 1992). With these stimuli, texture elements were presented predominantly to end-zones (Fig. 1E) or flanks (Fig. 1F) of the RF. Responses to butterfly stimuli were compared with the responses to “full” texture stimuli. Since the mechanisms underlying preference for feature contrast are not yet clear, the contribution of end-zones and flanks to these effects was of particular interest. Of 76 neurons tested with butterfly stimuli, 20 were MC cells (20051 5 39%) and 13 were OC cells (13055 5 24%). The mean responses of these cells are shown in Fig. 12. Responses were normalized to the response to the optimal center bar presented alone and then averaged across cells. With full texture surrounds, the MC cells in this sample responded 54% more strongly to motion contrast than to coher-


Fig. 10. Normalized population responses: motion (N 5 84). Responses were normalized to each neuron’s response to the optimal center bar; these values were then averaged over all cells. Histogram plot means and S.E.M. A: MC cells; B: all other neurons; C: total sample. A preference for motion contrast was strong for MC cells but also present in the responses of the total sample.

595 ent motion compared to the responses to the center bar alone (Fig. 12A). When texture elements were presented only to endzones or flanks, the preference for motion contrast was still present, but the effects were smaller (response difference between contrast and uniform condition: 27% (end-zones), 18% (flanks); two-tailed paired t-test, P , 0.05). The OC cells responded 44% more strongly to texture contrast than to uniform texture with full texture surrounds (Fig. 12B). Texture elements presented to the end-zones induced general suppression, but elements presented to the flanks produced a marginally significant preference for orientation (twotailed paired t-test, P 5 0.05) with a reduction of the effect compared to that seen with full texture surrounds. It should be pointed out that neither texture elements presented to end-zones nor texture elements presented to flanks induced differential effects as strong as those induced by full texture surrounds. In MC cells, preference for motion contrast was seen in both subregions, but response differences were much smaller than those seen with the full texture surrounds. In OC cells, small effects were seen from the flanks. However, because of the small OC cell sample tested with butterfly stimuli, it cannot be concluded that differential effects were predominantly related to flanks in orientation tests. In summary, these findings do not suggest that preference for motion or orientation contrast originates exclusively from either end-zones or flanks. This is confirmed by the comparison of ASI and DFI values for full texture and for butterfly conditions (Figs. 13 and 14). In most cells, texture elements presented to end-zones and flanks both induced suppression; ASI values from these regions were moderately correlated (Fig. 13A; correlation coefficients— orientation: 0.38 and motion: 0.46; P , 0.001). Compared to the ASI seen with full texture, neither the stimulation of end-zones alone nor the stimulation of flanks induced an equally strong effect (Figs. 13B and 13C). The differential effects obtained with butterfly stimuli varied considerably and were generally not related in individual cells (Fig. 14A; correlation coefficients—orientation: 0.13 and motion: 0.19; P 5 0.15). In general, the effects seen with full texture could not be predicted from the effects seen with butterfly stimuli (Figs. 14B and 14C). In some cells, the differential effects with full texture surrounds were stronger than those obtained with butterfly stimuli; in other neurons, however, the reverse was seen. In summary, presenting texture elements only to end-zones or flanks induced substantial general suppression. However, these suppressive effects were usually smaller than those obtained with full texture suggesting that surround effects from different regions around the RF partially sum up. Differential effects obtained with butterfly stimuli presented to flanks or end-zones were not related and could not predict the differential effects obtained with full texture. Neither of these effects originated exclusively from endzones or flanks. In 69 neurons, ASIs and DFIs obtained with full texture were compared with the cells’ degree of end-inhibition as quantified by the end-stopping index (ESI, cf. Methods). Values close to 1 indicate strong end-inhibition, while values close to 0 indicate no endinhibition. The correlation between the degree of end-inhibition and averaged suppression was moderate (Fig. 15A; correlation coefficient r 5 0.42, P 0.001). Neurons with strong end-inhibition (e.g. ESI . 0.7) showed some degree of averaged suppression, and the few neurons with negative ASI, on the other hand, were not strongly end-inhibited. End-inhibition was not correlated with the differential firing index (Fig. 15B; r 5 20.02, P 5 0.91) suggest-

596

Fig. 11. Normalized population responses: orientation (N 5 99). Normalized responses (cf. Fig. 10) of OC cells (A), all other neurons (B), and the total sample of neurons (C). Preference for orientation contrast was seen in the response of the OC cells and of the total sample.

ing no special role of end-inhibition in differential response modulation. Differential effects and averaged suppression were also compared with the cells’ orientation and direction selectivity, as mea-


Fig. 12. Responses of MC and OC cells to butterfly stimuli. Responses to center bars surrounded by full texture surround (black columns), or surrounded by texture elements presented exclusively to end-zone regions (crossed columns), or flanks (hatched columns). Responses were normalized to those to the center bar alone. A: MC cells. Preference for motion contrast was strong for full texture surrounds and reduced for texture elements presented to end-zones and flanks alone. B: OC cells. Preference for orientation contrast was seen for full texture surrounds and for texture elements presented to the flank regions. No differential effects were seen for the stimulation of end-zones.

sured by the respective indices (OSI, DSI; cf. Methods). There was a moderate anticorrelation between orientation selectivity and averaged suppression; responses of neurons without much orientation tuning were more strongly suppressed by the texture surrounds (ASI vs. OSI: r 5 20.44, P , 0.001). No correlations were found with the other comparisons (ASI vs. DSI: r 5 0.07, P 5 0.57; DFI vs. DSI: r 5 0.07, P 5 0.61; DFI vs. OSI: r 5 20.21, P 5 0.43).


Fig. 13. Averaged suppression with butterfly stimuli and full texture surrounds. A: ASI values obtained with butterfly stimuli presented to endzones (abscissa) or flanks (ordinate). In many neurons, both regions induced suppression. B: ASI values obtained with full texture surrounds and butterfly stimuli presented to the flanks. C: ASI values obtained with full texture surrounds and butterfly stimuli presented to the end-zones. End-zones or flanks evoked less suppression than full texture surrounds.

Discussion Neural correlates of pop-out in the cat The present study adds to the growing list of investigations demonstrating contextual response modulation in cat striate cortex, that depends on the orientation or direction of motion of patterns presented outside the classical RF (Blakemore & Tobin, 1972; Maffei & Fiorentini, 1976; Fries et al. 1977; Nelson & Frost, 1978; Hammond & MacKay, 1981; Hammond & Smith, 1982, 1984; Nothdurft & Li, 1984, 1985; Gulyas et al., 1987; Orban et al., 1987; Gilbert

597

Fig. 14. Differential firing with butterfly stimuli and full texture surrounds. A: DFI values derived from responses to end-zone or flank stimulation for individual neurons. B, C: DFIs obtained for full texture surrounds and either flanks or end-zones. The DFI values for these different stimuli were generally not related.

& Wiesel, 1990; DeAngelis et al., 1992; Li & Li, 1994; Sillito et al., 1995; Kastner et al., 1997). These studies and studies in the monkey (Allman et al., 1985, 1990; Knierim & Van Essen, 1992; Kapadia et al., 1995; Lamme, 1995; Sillito et al., 1995; Zipser et al., 1996; Nothdurft et al., 1999) have fundamentally altered the view of primary visual cortex as a pure processor of simple feature properties. However, the functional relevance of contextual modulation is not yet clear. We sought to relate response modulation induced by texture patterns to the perception of saliency by using a stimulation paradigm adopted from human psychophysics. Behavioral studies in the cat have shown that this species can detect differences in texture orientation (Wilkinson, 1986). We have shown in the present and a previous report (Kastner et al., 1997) that a

598

S. Kastner, H.-C. Nothdurft, and I.N. Pigarev in the anesthetized monkey in order to compare contextual modulation in the two species. The same major types of response modulation were found in these studies: preference for orientation contrast, preference for uniform texture, and general suppression evoked by any texture surround. General enhancement effects, also observed in the studies by Knierim and Van Essen (1992) and Nothdurft et al. (1999), were obtained only in few neurons and have been described in several other studies in both species (Maffei & Fiorentini, 1976; Nelson & Frost, 1985; Gilbert & Wiesel, 1990; Sillito et al., 1995; Kapadia et al., 1995). Knierim and Van Essen (1992) grouped types of response modulation according to the effects obtained with optimal and nonoptimal center bars, whereas we only considered effects on optimal center bars, which are sufficient to describe functional response properties (Nothdurft et al., 1999). Despite these minor differences in response classification, the distributions of response classes were similar across these different studies. OC cells were slightly less frequent in the anesthetized cat (22%) than in the awake monkey (32%; Knierim & Van Essen, 1992), whereas GS cells were seen more often (33% compared to 27%). However, it is unlikely that these differences reflect species differences rather than differences in animal preparation, since similar deviations are seen in the anesthetized monkey (24% OC cells; 40% GS cells; Nothdurft et al., 1999). In all three studies, texture surrounds suppressed the center bar responses by about one-third (37% in the present study compared to 34% [Knierim & Van Essen, 1992] and 33% [Nothdurft et al., 1999]). Also, a general preference for contrast in the population response was seen in all these studies. This suggests that processing of orientation contrast, at the level of striate cortex, is organized in a similar way in cat and monkey. In both species, neurons responding better to orientation contrast than to uniform texture may provide the basis for the perception of saliency.

Fig. 15. Relation of end-inhibition to averaged suppression and differential firing. ASI (A) and DFI values (B) were compared to the degree of endinhibition as quantified in 69 neurons. Values close to 1 indicate strong, values near 0 weak end-inhibition. End-inhibition was moderately related to averaged supression, but not to differential firing.

considerable number of neurons responded more strongly to elements embedded in a texture displaying orientation or motion contrast, thus appearing to be salient, than to nonsalient elements in uniform texture. The response properties of these neurons are well suited to process local discontinuities in pop-out texture patterns. For line textures, such filters were demanded (e.g. De Weerd et al., 1992) but not yet found in the cat visual system. Although the various types of response modulation were obtained in cat striate cortex, a preference for orientation and motion contrast was predominant and even apparent in the population response of striate neurons. Processing of orientation and motion contrast in cat and monkey Orientation contrast We used patterns similar to those used by Knierim and Van Essen (1992) in the awake monkey and by Nothdurft et al. (1999)

Motion contrast Neurons responding more strongly to motion contrast than to coherent motion in line patterns have not yet been demonstrated in monkey striate cortex. However, Allman et al. (1990) found such effects with dot-pattern surrounds. Out of 21 striate cells investigated, 33% were less suppressed if the surround moved opposite to the center bar than when it moved in the same direction. Almost all other neurons were generally suppressed by background motion independent of its direction. In the cat, modulatory effects by relative motion have been reported with moving noise patterns (Hammond & Smith, 1982, 1984; Gulyas et al., 1987; Orban et al., 1987). These effects are partly similar to the modulation effects observed with moving line patterns in the present study. Thus, there is evidence that not only pop-out of orientation but also pop-out of motion might be processed similarly in the two species. Contextual modulation for different visual dimensions We have compared contextual modulation in two dimensions, orientation and motion, in single neurons from the same population. The amount of averaged suppression seen in the different tests was similar in most neurons, whereas differential effects were not related across dimensions. In general, it was not possible to predict from the effect obtained in one visual dimension the effect obtained in the other. This suggests that differential effects in the orientation and motion dimension are generated by different pools of neurons and that averaged suppression and differential response modulation may perhaps underlie different mechanisms. Common suppressive effects may be generated independently of the partic-

Orientation and motion contrast in cat striate cortex ular features of the stimuli in the surround. In contrast, differential effects may be based on more specific mechanisms depending on certain stimulus features such as orientation or direction of motion. However, our data did not show a correlation of differential effects with a neuron’s selectivity for these properties. End-zones, flanks, and end-inhibition The mechanisms underlying contextual response modulation are not yet understood. Anatomical substrates subserving these mechanisms may include local circuits within striate cortex via longrange horizontal connections and feedback connections from higher cortical areas with larger RFs (cf. discussions in Knierim & Van Essen, 1992; Lamme, 1995; Nothdurft et al., 1999). We were particularly interested in the modulatory effects from specific regions surrounding the excitatory RF such as inhibitory end-zones or flanks. For orientation and motion stimuli, both regions contributed to averaged suppression, in many cells to a similar amount, in agreement with DeAngelis et al. (1994), who found a high correlation between end-zone and flank inhibition in most neurons of cat striate cortex. We have found a moderate correlation of averaged suppression with the degree of a neuron’s end-inhibition, in agreement with the results of Knierim and Van Essen (1992) for a smaller cell sample (Fig. 15A). Differential effects elicited with end-zone or flank stimuli were not related to each other and were also not related to the differential effects obtained with full texture surrounds. These effects were also not related to the degree of end-inhibition. This does not exclude that inhibitory flanks and end-zones do contribute to the differential effects in some neurons, but there seems to be no exclusive role for one or the other of these subregions in generating differential effects. End-inhibition has been related to several tasks such as the detection of corners or borders (Hubel & Wiesel, 1965), the representation of curvature (Dobbins et al., 1987), and the perception of subjective contours (von der Heydt & Peterhans, 1987). We conclude from our results that end-stopping does not play a major role in generating response preference for orientation or motion contrast. Implications for human texture vision The relationship of our results to psychophysical data has been discussed elsewhere (Nothdurft 1993, 1994; Kastner et al., 1997). Here, we will focus on human studies obtained with electrophysiological and functional imaging methods. Given the similarity of the cat’s behavioral performance in texture segmentation tasks with that of humans (Wilkinson, 1986; De Weerd et al., 1992), one should assume that cells with preference for orientation or motion contrast similar to those described in this study may also form the neural basis of pop-out of orientation or motion in the human visual system. This implies that processing of pop-out starts as early as in striate cortex. Several evoked potential studies have recorded response differences between checkerboard patterns defined by orientation or motion contrast and homogeneous texture patterns (Bach & Meigen, 1992; Lamme et al., 1992, 1994; Meigen et al., 1997). However, it is not yet clear where the source of this activity is localized. In a functional MRI study, Karni et al. (1993) found stronger hemodynamic responses in the calcarine to a texture segment embedded in a background of orthogonal line elements than when the same segment was embedded in a background of line elements at the same orientation, suggesting that texture segmentation occurs as early as in area V1 (but see De Weerd

599 et al., 1998). The signals obtained with these different methods may be generated by large populations of striate neurons responding more strongly to feature contrast in texture patterns than to homogeneous texture, similar to the OC and MC cells described in this study. Psychophysical studies have shown that the saliency of target elements in textures depends on the local feature contrast between the target and surrounding distractors. This dependence on local feature contrast appears to be a general phenomenon; it was shown for feature properties such as line orientation, direction of motion, color, luminance, and stereopsis (Nothdurft, 1991, 1992, 1993, 1995). In accordance with these results, it was recently found that texture borders defined by orientation, direction of motion, luminance, or stereopsis generated similar visually evoked potentials (VEPs; Bach & Meigen, 1997), suggesting that the neural processes underlying these signals are largely similar across visual dimensions. Our results suggest that populations of neurons responding preferentially to feature contrast may generate these signals. However, these populations may differ across the visual dimensions. In the present study, for example, preferences for motion and orientation contrast were represented by different pools of neurons. Neurons responding better to feature contrast have so far been found with pop-out (Knierim & Van Essen, 1992; Kastner et al., 1997 and present study; Nothdurft et al., 1999) and texture border stimuli (Nothdurft & Li, 1985; Nothdurft et al., 1992; Gallant et al., 1995; Lamme, 1995; Zipser et al., 1996), and for different visual dimensions such as orientation (Nothdurft & Li, 1985; Knierim & Van Essen, 1992; Li & Li, 1994; Gallant et al., 1995; Kastner et al., 1997; Lamme, 1995; Sillito et al., 1995; Zipser et al., 1996; Nothdurft et al., 1999), motion (Zipser et al., 1996; Kastner et al., 1997) and luminance (Nothdurft & Li, 1984, 1985; Zipser et al., 1996). However, further investigations are needed to establish the neural mechanisms underlying these effects, in particular the role of feedback from higher cortical areas or intrinsic circuits within striate cortex.

Acknowledgments We thank our colleagues B. Jagadeesh, J. Reynolds, and U. Ziemann for valuable discussions. The study was supported by a grant from the Deutsche Forschungsgemeinschaft (No 18008-1).

References Allman, J., Miezin, F. & McGuinness, E. (1985). Stimulus specific responses from beyond the classical receptive field: Neurophysiological mechanisms for local–global comparisons in visual neurons. Annual Review of Neuroscience 8, 407– 429. Allman, J., Miezin, F. & McGuinness, E. (1990). Effects of background motion on the responses of neurons in the first and second cortical visual areas. In Signal and Sense: Local and Global Order in Perceptual Maps, ed. Edelman, G.M., Gall, W.E. & Cowan, M.W., pp. 131– 142. New York: Wiley. Bach, M. & Meigen, T. (1992). Electrophysiological correlates of texture segregation in the human visual evoked potential. Vision Research 32, 417– 424. Bach, M. & Meigen, T. (1997). Similar electrophysiological correlates of texture segregation induced by luminance, orientation, motion and stereo. Vision Research 37, 1409–1414. Blakemore, C. & Tobin, E.A. (1972). Lateral inhibition between orientation detectors in the cat’s visual cortex. Experimental Brain Research 15, 439– 440. DeAngelis, G.C., Robson, J.G., Ohzawa, I. & Freeman, R.D. (1992). Organization of suppression in receptive fields of neurons in cat striate cortex. Journal of Neurophysiology 68, 144–163.

600 DeAngelis, G.C., Ohzawa, I. & Freeman, R.D. (1994). Length and width tuning of neuron’s in the cat’s primary visual cortex. Journal of Neurophysiology 71, 347–374. De Weerd, P., Vandenbussche, E. & Orban, G.A. (1992). Texture segregation in the cat: A parametric study. Vision Research 32, 305–322. De Weerd, P., Kastner, S., Desimone, R. & Ungerleider, L.G. (1998). Texture segregation in human extrastriate cortex: A f MRI study. Society for Neuroscience Abstracts 24, 1396. Dobbins, A., Zucker, S.W. & Cynader, M.S. (1987). Endstopped neurons in the visual cortex as a substrate for calculating curvature. Nature 329, 438– 441. Fries, W., Albus, K. & Creutzfeldt, O.D. (1977). Effects of interacting visual patterns on single cell responses in cat’s striate cortex. Vision Research 17, 1001–1008. Gallant, J.L., Van Essen, D.C. & Nothdurft, H.C. (1995). Twodimensional and three-dimensional texture processing visual cortex of the macaque monkey. In Early Vision and Beyond, ed. Papathomas, T.V., Chubb, C., Gorea, A. & Kowler, E., pp. 89–98, Cambridge, Massachusetts: MIT Press. Gilbert, C.D. & Wiesel, T.N. (1990). The influence of contextual stimuli on the orientation selectivity of cells in primary visual cortex of the cat. Vision Research 30, 1689–1701. Gulyas, B., Orban, G.A., Duysens, J. & Maes, H. (1987). The suppressive influences of moving textured backgrounds on responses of cat striate neurons to moving bars. Journal of Neurophysiology 57, 1767– 1791. Hammond, P. & MacKay, D.M. (1981). Modulatory influences of moving textured backgrounds on responsiveness of simple cells in feline striate cortex. Journal of Physiology 319, 431– 442. Hammond, P. & Smith, A.T. (1982). On the sensitivity of complex cells in feline striate cortex to relative motion. Experimental Brain Research 47, 457– 460. Hammond, P. & Smith, A.T. (1984). Sensitivity of complex cells in cat striate cortex to relative motion. Brain Research 301, 287–298. Hubel, D.H. & Wiesel, T.N. (1965). Receptive fields and functional architecture in two nonstriate visual areas (18 and 19) of the cat. Journal of Neurophysiology 28, 229–289. Kapadia, M.K., Ito, M., Gilbert, C.D. & Westheimer, G. (1995). Improvement in visual sensitivity by changes in local context: Parallel studies in human observers and in V1 of alert monkeys. Neuron 15, 843–856. Karni, A., Ungerleider, L.G., Haxby, J., Jezzard, P., Pannier, L., Cuenod, C.A., Turner, R. & LeBihan, D. (1993). Stimulus dependent MRI signals evoked by oriented line-element textures in human visual cortex. Society for Neuroscience Abstracts 19, 1501. Kastner, S., Nothdurft, H.C. & Pigarev, I.N. (1995a). Neuronal responses to orientation and motion contrast in feline striate cortex. Society for Neuroscience Abstracts 21, 770. Kastner, S., Nothdurft, H.C. & Pigarev, I.N. (1995b). Neuronal correlates of pop-out in cat striate cortex. Perception 26, S128. Kastner, S., Nothdurft, H.C. & Pigarev, I.N. (1997). Neuronal correlates of pop-out in cat striate cortex. Vision Research 37, 371–376. Knierim, J.J. & Van Essen, D.C. (1992). Neuronal responses to static texture patterns in area V1 of the alert macaque monkey. Journal of Neurophysiology 67, 961–980. Lamme, V.A.F., Van Dijk, B.W. & Spekreijse, H. (1992). Texture segregation is processed by primary visual cortex in man and monkey. Evidence from VEP experiments. Vision Research 32, 797–807. Lamme, V.A.F., Van Dijk, B.W. & Spekreijse, H. (1994). Organization of contour from motion processing in primate visual cortex. Vision Research 34, 721–735.

S. Kastner, H.-C. Nothdurft, and I.N. Pigarev Lamme, V.A.F. (1995). The neurophysiology of figure-ground segregation in primary visual cortex. Journal of Neuroscience 15, 1605–1615. Li, C.Y. & Li, W. (1994). Extensive integration field beyond the classical receptive field of cat’s striate cortical neurons—classification and tuning properties. Vision Research 34, 2337–2355. Maffei, L. & Fiorentini, A. (1976). The unresponsive regions of visual cortical receptive fields. Vision Research 16, 1131–1139. Meigen, T., Kastner, S. & Bach, M. (1997). A comparison of psychophysical and electrophysiological threshold detection in texture segregation. Perception 26, S128. Nelson, J.I. & Frost, B.J. (1978). Orientation selective inhibition from beyond the classical receptive field. Brain Research 139, 359–365. Nothdurft, H.C. (1985). Sensitivity for structure gradient in texture discrimination tasks. Vision Research 25, 1957–1968. Nothdurft, H.C. (1991). Texture segmentation and pop-out from orientation contrast. Vision Research 31, 1073–1078. Nothdurft, H.C. (1992). Feature analysis and the role of similarity in pre-attentive vision. Perception and Psychophysics 52, 355–375. Nothdurft, H.C. (1993).The role of features in preattentive vision: Comparison of orientation, motion and colour cues. Vision Research 33, 1937–1958. Nothdurft, H.C. (1994). Common properties of visual segmentation. In Higher Order Processing in the Visual System, ed. Bock, G.R. & Goode, J.A., pp. 245–259. New York: Wiley. Nothdurft, H.C. (1995). Generalized feature contrast in preattentive vision. Perception 24, S22. Nothdurft, H.C. & Li, C.Y. (1984). Representation of spatial details in textured patterns by cells of the cat striate cortex. Experimental Brain Research 57, 9–21. Nothdurft, H.C. & Li, C.Y. (1985). Texture discrimination: Representation of orientation and luminance differences in cells of the cat striate cortex. Vision Research 25, 99–113. Nothdurft, H.C., Gallant, J. & Van Essen, D.C. (1992). Neural responses to texture borders in macaque area V1. Society for Neuroscience Abstracts 18, 1275. Nothdurft, H.C., Gallant, J. & Van Essen, D.C. (1999). Response modulation by texture surround in primate area V1: Correlates of “popout” under anesthesia. Visual Neuroscience 16, 15–34. Orban, G.A., Gulyas, B. & Vogels, R. (1987). Influence of a moving textured background on direction selectivity of cat striate neurons. Journal of Neurophysiology 57, 1792–1812. Rossi, A., Desimone, R. & Ungerleider, L.G. (1998). Late onset responses to extra-receptive field stimulation in V1. Society for Neuroscience Abstracts 24, 1979. Sengpiel, F., Sen, A. & Blakemore, C. (1997). Characteristics of surround inhibition in cat area 17. Experimental Brain Research 116, 216–228. Sillito, A.M., Grieve, K.L., Jones, H.E., Cudeiro, J. & Davis, J. (1995). Visual cortical mechanisms detecting focal orientation discontinuities. Nature 378, 492– 496. Treisman, A. & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology 12, 97–136. Treisman, A. (1986). Features and objects in visual processing. Scientific American 255, 106–115. von der Heydt, R. & Peterhans, E. (1987). Mechanisms of contour perception in monkey visual cortex. I. Lines of pattern discontinuity. Journal of Neuroscience 9, 1731–1748. Wilkinson, F. (1986). Visual texture segmentation in cats. Behavioral Brain Research 19, 71–82. Zipser, K., Lamme, V.A.F. & Schiller, P.H. (1996). Contextual modulation in primary visual cortex. Journal of Neuroscience 16, 7376–7389.