INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

It was proposed by Gibson (1950) that our visual system utilizes various kinds of depth cues to perceive the three-dimensional (3D) shape of an object from two-dimensional (2D) retinal images. Visual depth is perceived not only with binocular cues of disparity, but also with monocular cues of linear perspective, texture gradient, and shading. Similarly, Marr (1982) postulated in his computational theory of vision that the description of the geometry of a visible surface by integrating various depth cues is a critical step in visual information processing for forming a representation of a 3D shape.

Many psychophysical studies have shown that in addition to binocular disparity, monocular depth cues, such as the linear perspective of contours (Clark et al. 1955; Freeman 1966; Olson 1974; Stevens 1983) and texture gradients (Clark et al. 1956; Cutting and Millard 1984; Flock and Moscatelli 1964; Goodenough and Gillam 1997; Gruber and Clark 1956), are also important for the perception of the surface orientation. It was reported that such monocular depth cues interact with binocular depth cues during the perceptual process (Poom and Borjesson 1999).

In our previous studies, we identified surface-orientation-selective (SOS) neurons in the caudal part of the lateral intraparietal sulcus (area CIP) of the monkey and found that they are sensitive to binocular disparity (Shikata et al. 1996). It was found that these neurons are sensitive to the gradients of disparity across the surface and/or those along the contour, and that they integrate these disparity signals to represent 3D surface orientation (Taira et al. 2000). It was shown by further study that some of these SOS neurons are also sensitive to monocular cues for depth such as linear perspective and texture gradient (Tsutsui et al. 1999); however, it is still unclear how the neural signals of monocular depth cues are integrated with those of disparity cues to generate a representation of the 3D surface orientation in these neurons.

In the present study, we compared the effects of monocular depth cues of linear perspective with those of binocular depth cues on the response of SOS neurons in area CIP by single-unit recording experiments. We also examined the critical role of SOS neurons in the perception of 3D surface orientation by muscimol microinjection into the recording sites of these neurons.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Two male Japanese monkeys (Macaca fuscata) were used in the present study. Throughout the experiments, the monkeys were treated in accordance with the National Institutes of Health Guide for Care and Use of Laboratory Animals. This project was approved by the Ethical Committee of Nihon University School of Medicine.

Apparatus

All stimuli used in the present study were generated by a graphics computer (SGI, Indigo2) and presented on a display (1,240 × 1,024 pixels, 21 in.) with a liquid crystal polarized filter (NUvision). The display was placed 44 cm in front of the monkey at eye level. The filter was switched at 120 Hz, whereby 60 frames/s of stimulus were presented to each eye. The monkey wore polarized glasses to view these stimuli stereoscopically.

Stimulus

Figure 1A shows samples of the binocularly presented stimuli used in the single-unit recording and the muscimol injection experiments. 1) Solid-figure stereogram (SFS) of a square plate with perspective cues (D+P condition, top row): this type of stimulus had perspective cues as well as disparity cues; therefore the 2D shapes of fused images were dependent on their orientation in depth. 2) Solid figure (SF) with perspective cues alone (P-only condition, middle row): these were presented binocularly in the task. This type of stimulus had perspective cues, but no binocular disparity cues (i.e., 2D shape stimuli). 3) SFS with disparity cues alone (D-only condition, bottom row): this type of stimulus had binocular cues, but no perspective cues; therefore the 2D shape of a fused image was the same (square) in all orientations. We used nine different orientations in the task: a plate in the frontoparallel plane and plates of eight different orientations that were slanted 45° against the frontoparallel plane and rotated every 45° around the sagittal (Z) axis (see insets in Fig. 2). We defined the direction of slant as "tilt" following Stevens (1983), so that the tilt of the slanted plate ranged from 0 to 315° at 45° intervals. In the muscimol injection experiments, 2D shape stimuli (circle, triangle, square, and hexagon) were used for 2D shape discrimination, which was the control task (2D-shape condition). The 2D shape stimuli were presented binocularly in the task.

View larger version (29K): [in this window] [in a new window]   Fig. 1. A: stimulus sets used in the single-unit recording experiments. Solid figure stereogram (SFS) of a plate in 0° tilt (left column) and in 270° tilt (right column) under the D+P (top row), P-only (middle row), and D-only (bottom row) conditions. The figure on the left is for the left eye, and that on the right is for the right eye (see METHODS for details). B: schematic illustration indicating the location of the caudal part of the lateral bank of the intraparietal sulcus (area CIP). The intraparietal sulcus (ips), lunate sulcus (lu), and the parietooccipital sulcus (po) are unfolded. Area CIP is located between areas LIP and V3A. C: single-unit recording sites were superimposed on the trace of the frontoparallel section of the left hemisphere of one monkey. Small dots indicate the points where SOS neurons were recorded.

View larger version (32K): [in this window] [in a new window]   Fig. 2. Responses of a typical surface-orientation-selective (SOS) neuron with sensitivity to perspective cues. A: responses to 9 different orientations under the D+P condition. The two-dimensional (2D) shape of the fused image changed in accordance with the linear perspective. B: responses to 9 different orientations under the P-only condition in which all stimuli were presented with zero disparity so that the linear perspective was the only cue for surface orientation. Insets indicate the stimuli presented: solid lines represent the orthographic projection of the simulated plate onto the frontoparallel plane, dashed lines schematically represent the orientation of the surface in depth caused by binocular disparity, and arrow represents the surface normal. The numbers at the top left of the rasters indicate the tilt (the orientation in the frontoparallel plane of the axis around which the surface was rotated). In the present study, the slant of the surface (the angle between the line of sight and the surface normal) was fixed at 45°. Cor. indicates the plate in the frontoparallel screen. Rasters and histograms are aligned to the stimulus onset (dashed line). Bold underline denotes the stimulus presentation period (750 ms). Note that preferred orientations of this neuron are similar under both conditions although the response magnitude is much higher under the D+P condition. This neuron also responded selectively under the D-only condition (not shown) and was classified as a DP neuron.

The size of these stimuli was 6.3 × 6.3° when they were in the frontoparallel orientation and was of minimum thickness (1 dot on the display, 0.0385 × 0.0385°). The fixation point and the center of the stimulus were presented at a distance of 44 cm from the monkey at eye level. The simulated distance for the perspective cues of the stimuli was the same as the optical distance. Pure red was chosen as the color of the stimuli to prevent ghost stimuli from appearing inappropriately in the eye when the filter was switched. All stimuli were rendered without shading or texture, and the background of the stimulus was black.

Behavioral task

We used a GO/NO-GO-type delayed-matching-to-sample (DMTS) task, or successive same/different discrimination, in which the monkey had to judge whether the surface orientations of successively presented sample and test stimuli were the same or different. The difference in tilt angles between a pair of sample and test stimuli in a trial ranged from 0 to 180° at 45° intervals, because the sample and test stimuli were chosen from a stimulus set consisting of nine orientations. The time sequence of the task is as follows. When a small fixation spot (0.2° diam) appeared, the monkey pressed the key and fixated on the spot. The monkey had to fixate on the spot until it released the key at the end of the trial. The presentation time of sample and test stimuli was 1 s or 750 ms, and the delay period (time interval between sample stimulus offset and test stimulus onset) was 2 s. If the surface orientation of the sample stimulus was the same as that of the test stimulus, the monkey had to release the key as soon as possible after the color of the fixation spot changed (GO trial); however, if the surface orientations were different, the monkey had to withhold key release for 1.5 s after the color change of the fixation spot until the fixation spot was turned off (NO-GO trial). The monkey was rewarded for the appropriate key release in both GO and NO-GO trials (symmetrical reinforcement). In the muscimol injection experiments, a 2D shape discrimination task was performed as the control task in addition to the surface orientation discrimination tasks. The sample and test stimuli were chosen from a 2D shape stimulus set, which consisted of a circle, triangle, square, and hexagon. The monkey had to discriminate whether the shapes of the sample and test stimuli were the same or different.

Eye movement recording

The movement of the right eye was monitored using an infrared eye movement recording system (RMS). The trial was canceled immediately when the eye position exceeded the limit of 1° from the fixation spot. The eye movement during each trial was also monitored off-line to confirm that small saccades or vergence eye movements did not occur. Please refer to our recent paper (Taira et al. 2000) for a detailed description of eye movement recording.

Single-unit recording

Before the single-unit recording, an atlas of the stereotaxic magnetic resonance image (MRI) of the brain of each monkey was constructed. For head fixation, a halo-like metal ring was implanted in each monkey's skull, and a microelectrode recording chamber was stereotaxically implanted in the opening of the skull over the parietal cortex under pentobarbital sodium anesthesia. After recovery from the surgery, extracellular single-unit recordings were carried out in the lateral bank of the intraparietal sulcus using tungsten microelectrodes. Since the recording chamber was implanted stereotaxically, the penetration track of the electrode could be accurately superimposed on the stereotaxic MRI brain map.

Single-unit activity was recorded during the performance of the DMTS task under three conditions (D+P, P-only, and D-only) in a blocked manner. Each block consisted of 45 trials, in which each of 9 different surface orientations was presented as a sample stimulus for every 9 trials in a random order. Test stimuli were selected so that 22 or 23 trials became GO trials and the remaining ones became NO-GO trials in one block. The recording procedure was as follows. We first tested the response of the neurons to a flat wooden plate held by the experimenter. If a neuron appeared to be selective to the surface orientation, it was further tested using the computer-generated stimuli during the monkey's performance of the DMST task, first under the D+P condition. If the neuron did not clearly exhibit selectivity as determined by the experimenter's visual inspection of rasters displayed on-line, it was discarded before completing the entire sequence (45 successful trials) of the D+P condition. All neurons recorded under the D+P condition were further recorded under the P-only condition. If the single-unit activity could be maintained in good isolation, the activity was further recorded under the D-only condition.

Analysis of neuronal activity

In the present study, neuronal responses to the sample stimulus in successful trials only were analyzed. To examine whether a neuron was visually responsive, neuronal activities in the prestimulus (500 ms) and stimulus (1 s or 750 ms) periods under the D+P condition were compared using the Student's t-test (P < 0.05) for each orientation. If the activities during the prestimulus and stimulus periods were different at least for one orientation, the neuron was classified as visually responsive. For visually responsive neurons, their selectivity to surface orientation under each condition was tested by comparing the activities during the stimulus period of eight orientations, excluding frontoparallel orientation, using the Rayleigh test (Mardia 1972). If a neuron responded selectively under the D+P condition, it was classified as an SOS neuron.

To estimate the preferred orientation and tuning sharpness of each SOS neuron under each condition, we transferred the response frequency for eight different orientations into vectors (_{0, 1 ... 7} for 0, 45 ... 315° tilt) so that the vector angle corresponded to the surface tilt and the vector length corresponded to the response frequency. The preferred orientation (tilt) was obtained by calculating the direction of the sum vector ( ). This method of calculating the preferred orientation is mathematically the same as the method of fitting a sinusoidal function to a discharge rate, as described by Georgopoulos et al. (1982). For the index of the tuning sharpness, we used the angular deviation S, which is equivalent to the standard deviation in linear statistics. The angular deviation S was calculated using

where | | is the length of the sum vector,  || is the sum of each vector's length, and S is the angular deviation in degrees.

Neurons recorded under all three (D+P, P-only, and D-only) conditions were classified according to their selectivity under the three conditions (Table 1). If a neuron responded selectively under the D+P and P-only conditions but not under the D-only condition, it was classified as a P neuron, that is, it is sensitive to perspective cues alone. If a neuron responded selectively under the D+P and D-only conditions but not under the P-only condition, it was classified as a D neuron, that is, sensitive to disparity cues alone. If a neuron responded selectively under all three conditions, it was classified as a DP neuron, that is, it is sensitive to both disparity and perspective cues.

Muscimol microinjection

After the single-unit recording experiments, the muscimol microinjection experiments were performed on the same monkeys. We microinjected muscimol using an injection-recording device (a stainless steel microinjection cannula containing a Teflon-coated tungsten wire that serves as a unit recording electrode; Crist). The tip of the microinjection cannula and that of the electrode were about 1 mm apart. The cannula was attached to the electrode manipulator and inserted into the recording site. To confirm that the tip of the cannula was in area CIP, extracellular single-unit recordings were carried out through an attached electrode. A total of 15-20 µg of muscimol (5 µg/µl) was injected over a period of 10 min. Immediately after the completion of the microinjection, the monkeys were required to perform the task.

In the muscimol microinjection experiments, the monkeys were required to perform the DMTS task under three conditions (D+P, D-only, and 2D-shape). The time course of the trial was the same as that in the unit recording experiments (1-s or 750-ms stimulus presentation and 2-s delay). One session consisted of three blocks of different conditions in a random order. In three of six cases of microinjection experiments, a block consisted of 18 trials. In the other three cases, a block lasted until the number of correct trials reached 18 trials, and then it was switched to another block. The duration of one session was approximately 15 min. Between blocks, five practice trials of the stimulus condition of the following block were inserted, to avoid the effect of changing the stimulus condition on the monkeys' performance. Before the microinjection, the monkeys were required to perform the task of two sessions as the control. After the microinjection, they were required to perform one session every 30 min during the first 2 h, one session after 3 h, and two sessions after 24 h.

Histology

After completion of the muscimol microinjection experiments, the monkeys were sacrificed and the entire brain was removed from the skull and soaked in 20% formaline. After the formaline fixation, 50-µm-thick sections were cut along the frontal plane in both hemispheres of two monkey brains. One in every two sections was stained with thionine to trace the penetrations. The unit recording sites and muscimol injection sites were determined indirectly from the relative positions of the penetrations to the anatomical landmarks, as well as from the stereotaxic lesions made after performing all the experimental sessions. The recording and injection sites were superimposed at 2-mm intervals on the trace of the frontal section of the left hemisphere of one monkey (Figs. 1C and 6A, see details in the figure legend). The plotting by this histological method and those by stereotaxic MRI method did not differ by more than 1 mm.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We recorded 211 visually responsive neurons in area CIP (Fig. 1B), and 66 of these were tested for orientation selectivity with a complete test routine (9 orientations × 5 repetitions) using a SFS of a square plate with both disparity and perspective cues (D+P condition). [We discarded neurons when they appeared to be nonselective to surface orientation in the initial examination (see METHODS for details concerning the recording procedure). Therefore there might be a sampling bias in selecting 66 neurons to be examined with a complete test routine.] Of these 66 neurons, 62 were identified as SOS neurons, as they responded selectively (Rayleigh test, P < 0.05) to the surface orientation. The mean angular deviation S (index for tuning sharpness; see METHODS for details) for SOS neurons (n = 62) was 68.5°, and that for non-SOS neurons (n = 4) was 79.4°. For each SOS neuron, the preferred orientation under the D+P condition was calculated by transferring the response frequency for eight different orientations into vectors and calculating the direction of the sum vector. These preferred orientations of SOS neurons were distributed randomly (Rayleigh test, P > 0.10).

To examine the effect of linear perspective cues on the response of SOS neurons, we tested their response to the solid figure with perspective cues alone (P-only condition). Figure 2 shows responses of a typical SOS neuron sensitive to linear perspective cues. Under the D+P condition (Fig. 2A), this neuron responded most vigorously to the surface tilted 315° and also responded with a high discharge rate to those tilted 0 and 45° (preferred orientation: 358.7°). Under the P-only condition (Fig. 2B), this neuron showed almost the same orientation selectivity (preferred orientation: 5.7°) as that under the D+P condition, but its response was much weaker. Thus this neuron appeared to be sensitive to both perspective and disparity cues. Of all the SOS neurons, 58% (36/62) showed orientation selectivity under the P-only condition, suggesting their sensitivity to perspective cues.

To examine the effect of binocular disparity cues on the response of SOS neurons, 35 of 62 SOS neurons were further tested using stimuli with disparity cues alone (D-only condition) in addition to the D+P and P-only conditions. Figure 3 shows responses of a typical SOS neuron with sensitivity to binocular disparity cues. Under the D+P condition (Fig. 3A), this neuron responded vigorously to surfaces tilted 0, 270, and 315° (preferred orientation: 323.3°). Under the D-only condition (Fig. 3B), this neuron showed almost the same orientation selectivity (preferred orientation: 313.6°) as that under the D+P condition, and its magnitude of response was almost the same. Thus this neuron did not appear to be sensitive to perspective cues. Of all the SOS neurons tested, 86% (30/35) showed orientation selectivity under the D-only condition, suggesting their sensitivity to disparity cues. This population was significantly higher (² test, P < 0.01) than that of neurons selective under the P-only condition (58%, 36/62).

View larger version (30K): [in this window] [in a new window]   Fig. 3. Responses of a typical SOS neuron with sensitivity to disparity cues. A: responses to 9 different orientations under the D+P condition. B: responses to 9 different orientations under the D-only condition in which all stimuli were presented without linear perspective so that the disparity was the only cue for surface orientation. The 2D fused image of each stimulus was always square under the D-only condition. Conventions are the same as those in Fig. 2. This neuron showed nonselective response under the P-only condition (not shown) and was classified as a D neuron. Note that preferred orientations and discharge rates of this neuron are similar under both conditions.

Classification of SOS neurons

Thirty-five SOS neurons tested under three conditions were classified into four groups according to their selectivity under each condition (Table 1). Two major types are shown in Fig. 4.

View larger version (36K): [in this window] [in a new window]   Fig. 4. Two major types of SOS neuron. A: DP (disparity and perspective) neuron. B: D (disparity) neuron. Responses under the D+P, P-only, and D-only conditions are shown in the top, middle, and bottom rows, respectively. For each condition, only responses to the most preferred and the diagonal orientations are shown, although responses to nine orientations were recorded. The 2D shape of the fused image changed in accordance with the linear perspective under the D+P and P-only conditions, but it was always the same (square) under the D-only condition. Other conventions are the same as those in Fig. 2.

The largest group (19/35, 54%) consists of neurons responding selectively under all three conditions, suggesting their sensitivity to both disparity and perspective cues (DP neuron). The neuron whose responses are shown in Fig. 4A is an example of this type. This typical DP neuron responded selectively to the surface tilted 225° under the D+P condition (top row), P-only condition (middle row), and D-only condition (bottom row). The figure shows that the response was greatly enhanced under the D+P condition in the same orientation. This may be due to the summation of responses to disparity and perspective cues. The average response of all DP neurons to the preferred orientations was significantly higher under the D+P condition (50.7 ± 5.2 spikes/s, mean ± SE) than under the D-only condition (37.9 ± 4.7, Student's t-test, P < 0.01) and the P-only condition (31.2 ± 3.5; Student's t-test, P < 0.001).

The second largest group (11/35, 31%) consists of neurons responding selectively under the D+P and D-only conditions but not under the P-only condition, suggesting their sensitivity to disparity cues alone (D neuron). The neuron whose responses are shown in Fig. 4B is an example of this type. This typical D neuron responded to the surface tilted 270° under the D+P condition (top row) and the D-only condition (bottom row) with almost the same response magnitude. However, it showed only a weak and nonselective response under the P-only condition (middle row). The average responses of all D neurons to the preferred orientations under the D+P condition (45.7 ± 6.7 spikes/s) and the D-only condition (34.6 ± 5.6) were not significantly different (Student's t-test, P > 0.10).

In contrast to DP and D neurons, only a few (6%, 2/35) neurons were classified as P neurons for responding selectively under the D+P and P-only conditions but not under the D-only condition. Furthermore, in these neurons, the selectivity under the P-only condition was relatively low (just above the significant level of P = 0.05). The remaining three neurons (9%, 3/35) did not show selectivity either under the P-only or D-only condition, showing response selectivity only under the D+P condition.

Concerning the location of neurons, these four types of neuron were uniformly distributed in area CIP, and there was no difference in the distribution according to the type of neuron.

Interaction between linear perspective and binocular disparity cues

To examine the interaction between linear perspective and binocular disparity cues in SOS neurons, relationships of preferred orientations under the three different conditions were analyzed.

For SOS neurons sensitive to perspective cues (DP and P neurons, 21/35), we plotted the preferred orientation under the P-only condition versus that under the D+P condition (Fig. 5A). The correlation of preferred orientations under the two conditions was positive but not sufficiently strong to reach statistical significance (circular-circular regression, r = 0.343, P > 0.10). However, the distribution of the preferred orientation difference was significantly concentrated to <45° (² test, P < 0.05; Fig. 5B), showing that preferred orientations under these two conditions tend to coincide to some extent. Furthermore, when we restrict the sample to DP neurons (n = 19), the preferred orientations under the two conditions were significantly correlated (circular-circular regression, r = 0.432, P < 0.05).

View larger version (26K): [in this window] [in a new window]   Fig. 5. A: plot of the preferred orientation of perspective-sensitive (DP and P) neurons under the P-only condition (ordinate) against that under the D+P condition (abscissa). , P neurons; , DP neurons. B: the distribution of the preferred orientation difference between the 2 conditions in these neurons. , P neurons; , DP neurons. C: plot of the preferred orientation of disparity-sensitive (DP and D) neurons under the D-only condition (ordinate) against that under the D+P condition (abscissa). , D neurons; , DP neurons. D: the distribution of the preferred orientation difference between the 2 conditions in these neurons. , D neurons; , DP neurons. E: plot of the preferred orientation of DP neurons under the P-only condition (ordinate) against D-only condition (abscissa). F: the distribution of the preferred orientation difference between the 2 conditions in these neurons.

For SOS neurons sensitive to disparity cues (DP and D neurons, 30/35), we plotted the preferred orientation under the D-only condition against that under the D+P condition (Fig. 5C). The preferred orientations under the two conditions were well correlated (circular-circular regression, r = 0.760, P < 0.01), and the distribution of the preferred orientation difference was significantly concentrated to <45° (² test, P < 0.01; Fig. 5D), showing that the preferred orientation under the D-only condition had a strong tendency to coincide with that under the D+P condition. For DP neurons (n = 19), the circular-circular regression coefficient was r = 0.888 (P < 0.001), whereas that of D neurons (n = 11) was r = 0.596 (P < 0.05).

For SOS neurons sensitive to both disparity and perspective cues (DP neuron, 19/35), we plotted the preferred orientation under the P-only condition versus that under the D-only condition (Fig. 5E). The correlation of the preferred orientations under the two conditions was positive but did not reach statistical significance (circular-circular regression, r = 0.273, P > 0.10). However, the population of neurons with the preferred orientation difference of <90° (14/19) was significantly larger than that with the preferred orientation difference of more than 90° (² test, P < 0.05; Fig. 5F), indicating that the preferred orientations under these two conditions tend to coincide to some extent.

To evaluate and compare the orientation tuning sharpness of SOS neurons under three different conditions, the angular deviation S, which corresponds to the standard deviation in normal distribution, was calculated (see METHODS for details concerning the calculation of the S value). For the perspective-cue-sensitive neurons (DP and P neurons), the mean S value under the P-only condition (76.1°) was significantly higher (Student's t-test, P < 0.01) than that under the D+P condition (68.4°), suggesting a broader orientation tuning under the P-only condition. For the disparity-cue-sensitive neurons (DP and D neurons), the mean S value under the D-only condition (68.5°) and that under the D+P condition (69.3°) did not differ significantly (Student's t-test, P > 0.10). For the neurons sensitive to both disparity and perspective cues (DP neuron), the mean S value under the P-only condition (76.3°) was significantly higher (Student's t-test, P < 0.01) than that under the D-only condition (68.0°) or that under the D+P condition (68.6°), suggesting a broader orientation tuning under the P-only condition in this group. These data suggest that the orientation selectivity in many SOS neurons tends to be relatively weak when only perspective cues are available.

Muscimol microinjection

After the completion of the unit recording experiments, we microinjected muscimol into the caudal part of the lateral bank of the intraparietal sulcus (area CIP), where we recorded the responses of SOS neurons (Fig. 6A), for the reversible functional block of this area.

View larger version (28K): [in this window] [in a new window]   Fig. 6. A: muscimol injection sites were superimposed on the trace of the frontoparallel section of the left hemisphere of one monkey. , , and +, the injection points in 3 effective injection cases; , those in 3 ineffective cases. In one case (+), muscimol was injected at 3 points along the lateral bank of IPS. Muscimol was injected at a single point in all other cases. B: time course of the success rate after muscimol injection in case 1. Only the performance level under the D-only condition decreased immediately after the injection and remained low for at least 3 h. It returned to the normal level on the next day (24 h). C: success rates before () and after () the injection in 3 positive cases. Graphs 1, 2, and 3 represent the success rates of cases indicated in the plot above by , , and +, respectively. * P < 0.05.

Out of six inactivation experiments, the performance of 3D surface orientation discrimination in the DMTS task was impaired in three cases (Fig. 6C). In one case (case 3), in which muscimol was microinjected into three spots rostrocaudally along area CIP, the performance of 3D surface orientation discrimination under both D-only and D+P conditions became significantly worse than that before microinjection, although 2D shape discrimination (see METHODS) was not affected. In two cases (cases 1 and 2), in which muscimol was microinjected into only one spot in area CIP, only the performance under the D-only condition was impaired after microinjection.

Figure 6B shows the time course of the level of task performance after microinjection in case 1. The performance level under the D-only condition decreased immediately after microinjection and remained impaired for at least 3 h, then returned to the normal level after 24 h. However, the ability to discriminate the 2D shape was not impaired at all.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study demonstrates that linear perspective is effective as cues for 3D surface orientation in many SOS neurons of area CIP, although they are less effective compared with binocular disparity cues. It was also demonstrated that perspective cues are integrated with disparity cues in many SOS neurons to generate the representation of 3D surface orientation. Furthermore, the performance of 3D surface orientation discrimination was impaired following the inactivation of SOS neurons. These results strongly support the view that the activity of SOS neurons is a neural correlate of the perception of 3D surface orientation.

Representation of 3D surface orientation in SOS neurons of area CIP

In the single-unit recording experiment, more than one-half (58%, 36/62) of all SOS neurons responded selectively to surface orientation under the P-only condition. This result suggests that many SOS neurons can encode the surface orientation with perspective cues alone. However, perspective cues seem to be less important than disparity cues for SOS neurons, because the proportion of perspective-sensitive neurons (58%, 36/62) was smaller than that of disparity-sensitive neurons (86%, 30/35), and the proportion of neurons sensitive to perspective cues alone (P neuron, 2/35) was much smaller than that of neurons sensitive to disparity cues alone (D neuron, 11/35). The finding that disparity cues are of great importance is in line with the finding in our previous studies that the majority of SOS neurons are sensitive to binocular disparity cues (Shikata et al. 1996; Taira et al. 2000). On the other hand, it appears that monocular cues of linear perspective are incorporated with binocular disparity cues in area CIP to represent surface orientation, because the majority (19/35) of SOS neurons were sensitive to both disparity and perspective cues (DP neuron). In this type of neuron, summation of the effect of disparity and perspective cues was observed in the response magnitude; the response to the preferred orientation was stronger under the D+P condition than under the D-only and P-only conditions (Fig. 4A). Since it is a general phenomenon in visual perception that the presence of a larger number of cues enables clearer perception of visual features, the activity of DP neurons may be strongly related to the perception and judgment concerning 3D surface orientation.

Results of the analysis of correlation of the preferred orientations provide strong support for the notion that SOS neurons utilize linear perspective cues to represent 3D surface orientation. Since the preferred orientations under the P-only condition tended to coincide with those under the D+P condition in many of perspective-sensitive neurons (Fig. 5, A and B), it was suggested that they do not simply respond to a 2D shape (trapezoid) in different orientations, but that they extract the signal of 3D surface orientation from a 2D contour viewed in a linear perspective. In other words, these neurons interpreted the stimuli as the silhouette of a square plate slanted in depth, rather than that of a frontoparallel trapezoid. The notion was also supported by the coincidence of the preferred orientations under the D-only and P-only conditions in neurons that are sensitive to both disparity and perspective cues (DP neurons, Fig. 5, E and F). Thus the nature of the perspective-sensitive SOS neurons should be distinguished from that of neurons in other cortical areas, such as the inferior temporal cortex, selective to 2D shapes reported previously (Desimone et al. 1984; Fujita et al. 1992; Gross et al. 1972; Kobatake and Tanaka 1994; Rolls et al. 1977; Tanaka et al. 1991), in the sense that the perspective-sensitive SOS neurons extract the signal of 3D surface orientation from a 2D image.

Results of the analysis of tuning sharpness also suggested that perspective cues have a weaker effect than disparity cues on the response of SOS neurons. In perspective-sensitive neurons, tuning was broader under the P-only condition than under the D+P condition (mean S value 76.1 and 68.4°, respectively), while in disparity-sensitive neurons, tuning sharpness under the D-only condition and that under the D+P condition were almost the same (mean S value 68.5 and 69.3°, respectively). Furthermore, in neurons sensitive to both disparity and perspective cues (DP neurons), tuning was broader under the P-only condition than under the D-only condition (mean S value 76.3 and 68.0°, respectively). These data suggest that the orientation selectivity in many SOS neurons tends to be relatively weak when only perspective cues are available. This relatively weak effect of perspective cues compared with that of disparity cues may be partly responsible for the weaker coincidence of orientation preference between the P-only and D+P conditions in perspective-sensitive neurons (Fig. 5, A and B) than that between the D-only and D+P conditions in disparity-sensitive neurons (Fig. 5, C and D), and the relatively weak coincidence of orientation preference between the D-only and P-only conditions in DP neurons (Fig. 5, E and F).

The dominance of disparity cues over perspective cues in the response of SOS neurons is in good correspondence with psychophysical findings. Since the perspective cues are somewhat ambiguous compared with disparity cues, the judgment of a surface slant is often unstable when based only on perspective cues, and the judgment is much more reliable when perspective cues are coupled with binocular disparity cues (Clark et al. 1955; Olson 1974). However, it is possible that we have underestimated the effect of perspective cues in the present study. Since we binocularly presented the SF with perspective cues alone instead of monocularly presenting them under the P-only condition, the power of perspective cues may have been weakened to some extent by conflicting with those of disparity cues, which provide information that the figure is in the frontoparallel orientation. This kind of cue conflict always occurs when we look at a painting or a photograph with both eyes open. The sensation of a 3D space becomes slightly stronger when we look at them monocularly (Koenderink et al. 1994).

Area CIP and the perception of 3D surface orientation

In muscimol microinjection experiments, the ability to discriminate a 3D surface orientation was selectively impaired following the inactivation of SOS neurons in area CIP, while the ability to discriminate a 2D shape remained intact. On the basis of this observation, it was suggested that a 3D surface orientation is a discrete perceptual entity and that its perception is directly achieved by the activity of SOS neurons in area CIP.

There were three effective cases out of six microinjection experiments (Fig. 6C). In the case of muscimol being microinjected into a relatively large area (case 3), discrimination of a 3D surface orientation was impaired under both the D+P and D-only conditions. However, in two of three effective cases in which muscimol was microinjected into only one spot in area CIP (cases 1 and 2), discrimination of a 3D surface orientation was impaired only under the D-only condition but not under the D+P condition. One possible interpretation of these results is that the muscimol microinjection into a single spot may have selectively inactivated D neurons leading to a selective effect under the D-only condition. However, this interpretation is unlikely because the distributions of DP and D neurons did not differ in area CIP. The response property of neurons in area CIP revealed by the single-unit recording experiments suggests another possibility. The activity of the entire population of SOS neurons may have been higher under the D+P condition than under the D-only condition, because the activity of DP neurons was enhanced under the D+P condition. It is a general phenomenon in visual perception that the presence of a larger number of cues enables easier perception of visual features. Under the D+P condition, in which whole population activity was relatively high, inactivation of a large number of neurons in area CIP with a relatively large amount of muscimol may have been necessary to affect the performance of discrimination.

There were effective and ineffective cases of muscimol microinjection experiments, although the injection sites were almost the same in all injection cases. One possible explanation for this is that the behavioral task was not sufficiently sensitive to detect a subtle change in perceptive ability caused by such small amounts of muscimol. In the present study, the difference in tilt angles between a pair of sample and test stimuli in NO-GO trials in the DMTS task ranged from 45 to 180° at 45 intervals (see METHODS). If the difference in tilt angles had been sufficiently small to be close to the threshold level of the monkey's orientation discrimination, we might have found some impairment in the three "ineffective" cases.

The main findings in muscimol microinjection experiments that a 3D surface orientation is a discrete perceptual entity and that its perception is directly achieved by the activity of neurons in area CIP are in good agreement with those in human studies. In positron emission tomography (PET) and fMRI studies with human subjects, the caudal part of the intraparietal area, which seems to correspond to the monkey area CIP, was activated while the subjects discriminated a 3D surface orientation based on texture gradients (Shikata et al. 2001), or a 3D surface geometry (convex-concave) based on monocular cues of shading (Taira et al. 2001) or binocular cues of disparity (Taira et al. 1997). Neurological studies have shown that 3D constructional apraxia occurs in patients with right parietal lobe lesions (Critchley 1953; De Renzi 1982). These patients showed abnormalities in assembling blocks according to a 3D model and drew a characteristic line drawing with no linear perspective. Furthermore, patients with parietal lesions were reported to exhibit severe impairment in 3D perception (Holmes and Horrax 1919; Riddoch 1917; Rothstein and Sacks 1972). It is notable that these patients seem to have problems not only in stereopsis (use of binocular disparity) but also in the use of monocular cues such as linear perspective and shadings.

Relations of area CIP to other cortical areas

The intraparietal area including area CIP is a part of the dorsal visual pathway and receives strong input from the V3-V3A complex, where columns based on binocular disparity tuning are found (Adams 1997). Area CIP appears to receive binocular disparity signals from this area. However, it is not yet determined from which region area CIP receives information regarding monocular cues. The most plausible candidates are certain areas in the ventral visual pathway. A group of neurons that selectively responded to 2D shape, texture, or shading was identified in the inferotemporal cortex (Desimone et al. 1984; Fujita et al. 1992; Gross et al. 1972; Kobatake and Tanaka 1994; Rolls et al. 1977; Tanaka et al. 1991). The intraparietal area including area CIP appears to receive input from areas in the ventral visual pathway (Baizer et al. 1991; Webster et al. 1994).

Recently, neurons representing 3D curvature of a surface were found in the lower bank of the superior temporal sulcus (STS) in the rostral temporal cortex (area TEs) (Janssen et al. 1999, 2000a,b). This area appears to be interconnected with area CIP (Baizer et al. 1991). It is one of the objectives of future research to reveal how area CIP and area TEs, one in the dorsal pathway and the other in the ventral pathway, cooperate with each other to represent 3D structures of objects. One characteristic of area CIP may be its anatomically close location to the area related to hand manipulation movements. In the anterior part of the lateral bank of the intraparietal sulcus (area AIP), a group of neurons was found to be related to hand manipulation movements, and many of these neurons were visually sensitive to the axis and surface orientations of objects as well as to their shape (Murata et al. 2000; Sakata et al. 1995; Taira et al. 1990). These neurons are likely to receive information regarding axis and surface orientations from area CIP neurons. As proposed in our previous studies (Sakata et al. 1997, 1998; Taira et al. 2000), area CIP seems to provide a viewer-centered representation of 3D surfaces for the manipulation of objects. On the other hand, area TE seems to be more closely related to object recognition, because a lesion in the inferior temporal cortex causes severe impairment of object recognition (Dean 1976; Iwai and Mishkin 1969; Weiskrantz and Saunders 1984), and many neurons in area TE respond selectively to a complex 2D shape and texture (Desimone et al. 1984; Fujita et al. 1992; Gross et al. 1972; Kobatake and Tanaka 1994; Rolls et al. 1977; Tanaka et al. 1991).