Binocular phase interactions in area 21a of the cat

Abstract

1. Binocular interactions related to retinal disparity were investigated in single neurons in area 21a of extrastriate cortex in the anaesthetized cat using sinusoidal luminance gratings.

2. The responses of approximately two-thirds of neurons were profoundly modulated by a relative phase difference between identical drifting gratings presented to each eye. This modulation included both facilitatory and inhibitory interocular interactions. The selectivity for binocular disparity was about twice as sharp as the selectivity for monocular spatial position.

3. Significant phase modulation was retained in many neurons at interocular orientation differences exceeding 45 deg. The response suppression associated with stimulation at a phase shift 180 deg from the optimum was stronger than the response suppression to an interocular orientation difference of 90 deg.

4. The proportion of phase modulated neurons and the potency of modulation in area 21a neurons exceed that reported for phase-selective complex cells in area 17. Neurons in area 21a show sharp disparity tuning that is relatively insensitive to changes in orientation and monocular position, which suggests that this extrastriate region has a role in stereoscopic depth perception.

Binocular disparity between the image of an object on the retina of each eye is a sufficient cue to elicit the perception of relative depth. The first site at which binocular disparity is explicitly encoded is the primary visual cortex (Barlow et al. 1967), where simple cells are tuned to binocular disparity based on differences between both the structure and position of their receptive field in each eye (Anzai et al. 1997). At least 40 % of complex cells in the cat primary visual cortex also appear to encode binocular disparity, as indicated by the modulation of their responses to a phase difference between dichoptically presented drifting sinusoidal gratings (Ohzawa & Freeman, 1986b;Hammond, 1991). This property of complex cells is remarkable, as their responses are not strongly modulated by the phase of a monocularly presented grating (Movshon et al. 1978). Complex cells code binocular disparity in a manner that is relatively independent of stimulus position, which may be an important level of abstraction for encoding object depth as a distinct parameter (Ohzawa et al. 1997).

Area 21a is a region of extrastriate cortex in the cat that receives its principal input from areas 17 and 18, and also from the lateral posterior nucleus of the thalamus (Symonds & Rosenquist, 1984; Wimborne et al. 1993; Morley et al. 1997). Most neurons in area 21a have receptive field centres within 15 deg of the area centralis and have a receptive field structure similar to that of complex cells in area 17. The neurons are predominantly binocular, strongly orientation selective, and show little direction preference, and about half display low-pass tuning properties for temporal frequency (Wimborne & Henry, 1992; Dreher et al. 1993; Morley & Vickery, 1997). A recent study by Wang & Dreher (1996) in which single bars were presented to each eye found that almost 70 % of area 21a neurons showed significant response modulation to binocular retinal disparities. This evidence, together with the fairly sharp spatial frequency selectivity demonstrated by area 21a neurons, has led to the suggestion that this extrastriate area may play a role in form discrimination and binocular depth perception (Dreher et al. 1993; Morley & Vickery, 1997).

In the present study, extracellular recordings were made from single area 21a neurons in the anaesthetized cat. We used dichoptically presented drifting sinusoidal gratings that differed in relative phase to investigate further the disparity selectivity of area 21a neurons and its dependence on interocular orientation difference.

METHODS

Animal preparation

The data presented here were obtained from experiments on 20 adult cats weighing from 2 to 5 kg. Anaesthesia was induced by intramuscular injection of ketamine hydrochloride (20 mg kg⁻¹) and xylazine hydrochloride (1 mg kg⁻¹) and was supplemented by i.m. injections of 4 mg kg⁻¹ ketamine/0.2 mg kg⁻¹ xylazine as needed. During preparatory surgery, the trachea, femoral vein and femoral artery were cannulated. A small craniotomy was made at Horsley-Clark co-ordinates 1-7 mm anterior-posterior and 9-15 mm medial-lateral, and the dura removed to expose the posterior part of the middle suprasylvian gyrus. Throughout the rest of the experiment, the animal was artificially ventilated and anaesthesia was maintained with intravenous infusion of alphaxalone and alphadolone (12 mg kg⁻¹ h⁻¹; Pitman-Moore, Sydney, Australia). Expired CO₂ was monitored and maintained within 3.5-4.0 % and body temperature was monitored and maintained at 37.5°C by a feedback-controlled heating pad. Blood pressure was continuously monitored via the arterial cannula, and screws positioned in the skull overlying the frontal lobes allowed monitoring of the electroencephalogram (EEG). The level of anaesthesia was judged sufficient when the blood pressure, heart rate, EEG and CO₂ records were stable and when mild noxious stimuli produced no change in these parameters. The animal's level of anaesthesia was stable for at least 1 h prior to inducing muscle paralysis by gallamine triethiodide (5 mg kg⁻¹ h⁻¹i.v. in 4 % glucose-0.18 % NaCl) to minimize eye movements, and depth of anaesthesia was closely monitored throughout the period of paralysis. The procedures for animal care, anaesthesia, surgery and recording complied with the guidelines of the National Health & Medical Research Council on the care and use of animals for scientific purposes and with those of the Animal Care & Ethics Committee of the University of New South Wales. The animals were given injections of atropine sulphate (s.c.) to reduce salivation and 4 mg dexamethasone phosphate (i.m.; Decadron: Merck, Whitehouse Station, NJ, USA) to minimize cerebral oedema. The exposed cortex was covered with warm agar (4 %) in physiological saline. Pupils were dilated with topical application of atropine sulphate and nictitating membranes retracted with topical phenylephrine hydrochloride. Zero power contact lenses protected the corneas, a 3 mm diameter pupil improved optical quality and where necessary spherical spectacle lenses focused the eye on a tangent screen or display monitor.

Visual stimulation and recording

The extracellular activity of single cells was recorded using glass-coated tungsten microelectrodes. The electrode was advanced into the cortex using a hydraulic micromanipulator with the electrode set at an angle of 20-30 deg to the vertical so that penetrations were approximately perpendicular to the surface of the cortex. Extracellular action potentials from isolated cells were conventionally amplified, displayed on an oscilloscope and passed to an audio monitor. Recording of the responses of single cells involved the action potential being passed through a window discriminator and fed to a computer for on-line analysis and data storage.

The optic disc and area centralis of each eye were plotted on a tangent screen using the method of Pettigrew et al. (1979). Receptive fields of isolated cells were plotted on the tangent screen using an ophthalmoscope to project a bar of light. All subsequent visual stimuli consisted of drifting sinusoidal gratings at a Michelson contrast of 0.8 generated using a Cambridge Research Systems video board (Rochester, Kent, UK) and presented on two computer monitors (NEC; model XV15) viewed through half-silvered mirrors, which resulted in a mean luminance of 10 cd m⁻². Spatial and temporal frequency and orientation tuning curves were determined monocularly for each eye. Binocular stimulation employed drifting gratings with parameters optimal for the dominant eye, presented simultaneously to each eye for 1.1 s. The relative phase offset between the gratings was varied between stimulus periods in a pseudo-random sequence of 45 deg increments that covered the full 360 deg. Each phase offset was tested at least 5 times in a pseudo-random interleaved series, with each presentation consisting of at least two temporal cycles of the drifting grating. Stimuli were presented every 5 s, and in the intervals between stimulus periods the screens were blank with the same mean luminance as the grating stimulus.

In some neurons we also investigated the relationship between relative phase offset and a change in the interocular grating orientation. Drifting gratings of optimum spatial and temporal frequency were presented simultaneously to each eye for 1.1 s. The grating presented to the dominant eye was set at the optimum orientation, while the grating presented to the non-dominant eye was changed from iso-oriented to orthogonal in steps of 15 deg. At each interocular orientation the relative phase offset was varied between stimulus periods in a pseudo-random sequence of 45 deg increments over the full 360 deg.

RESULTS

Monocular response properties

The results are based on extracellular recordings from 46 neurons in area 21a of the anaesthetized cat in which we obtained quantitative data on the spatial and temporal frequency tuning, orientation tuning and selectivity to relative binocular phase offset. Most neurons were similar to complex cells in the primary visual cortex in that they displayed little phase modulation to monocular grating stimulation, while a small proportion of neurons (19 %) resembled simple cells, in agreement with previous studies (Tardif et al. 1996; Morley & Vickery, 1997). The monocular response properties (spatial and temporal frequency tuning and orientation tuning) of the area 21a neurons were consistent with the findings of our previous study (Morley & Vickery, 1997) and were usually well-matched between the two eyes.

Response modulation by binocular phase difference

The response of the majority of neurons was significantly affected by the relative phase offset between otherwise identical drifting sinusoidal gratings presented dichoptically. Figure 1 shows the responses of two neurons to binocular disparity resulting from relative phase offset. The neuron in Fig. 1A responded vigorously at a phase offset of 45 deg, while at a phase offset of 225 deg (180 deg from optimum) there was scarcely any response. The binocular response at the optimum phase offset was about double the response to monocular stimulation of the dominant eye (thick dashed line on the graph in Fig. 1A), while the response at the least effective phase offset was less than half the response to stimulation of either the dominant or non-dominant eye. The peristimulus time histograms (PSTHs) at the left of Fig. 1A illustrate the dramatic variation in response over the 360 deg range of relative phase offsets. The neuron illustrated in Fig. 1B was less modulated by relative phase offset than the neuron shown in Fig. 1A, but is more representative of the degree of modulation in the majority of neurons.

Figure 1. Response modulation of two area 21a neurons to a relative phase offset between dichoptically presented drifting sinusoidal gratings.

A, neuron showing strong facilitation and suppression. PSTHs accumulated over five runs are shown on the left; the mean ±s.d. response is plotted on the right, together with the monocular responses to optimum stimulation (dashed lines: C = contralateral eye, I = ipsilateral eye). The fitted curve is a sine function constrained to a period of one cycle. B, neuron with relative phase modulation typical of our sample.

Phase modulation index

The extent of response modulation to binocular disparity was determined by calculating a phase modulation index, which is defined as the difference between responses at the optimum and least effective phases, divided by the sum of these two responses. This measure ranges from zero for no phase modulation, to 100 % which corresponds to total suppression of the response at the least effective phase offset. The phase modulation index ranged from 13 % to 97 % for our sample of 46 cells (Fig. 2A), with a mean ±s.e.m. of 53 ± 3 %, which is equivalent to a response reduction from peak to trough of about two-thirds. The neuron illustrated in Fig. 1B had a phase modulation of 49 %, and so was typical of our sample. The phase modulation data were tested using one-factor ANOVA to ascertain which cells displayed significant response variation due to the binocular phase difference, and 78 % of cells were significantly modulated (P < 0.05). The distribution of the phase modulation index in Fig. 2A indicates those cells that did not show significant phase modulation by the hatched segments of the columns. An alternative criterion for binocular disparity selectivity of area 21a neurons used by Wang & Dreher (1996) requires a doubling of neuronal response level with a change in binocular disparity, which corresponds to a phase modulation index > 33 %. In our sample of 46 neurons, 35 (76 %) had phase modulation values greater than 33 % and were classified as selective for binocular disparity. The proportion of area 21a neurons classified as disparity selective is substantially larger than the proportion found for area 17 complex cells, and the difference in the median values of the phase modulation index, 53 % compared with 35 % in area 17, is highly significant (P < 0.01, Mann-Whitney U test; Hammond, 1991).

Figure 2. Measures of disparity tuning in the sample of area 21a neurons.

A, distribution of the phase modulation index for all neurons. ▪, neurons whose phase modulation was statistically significant (ANOVA, P < 0.05). , neurons whose phase modulation was not statistically significant. B, the BII is plotted against the phase modulation index for 46 neurons. ○, values with an S/N ratio less than 2 (see text).

Phase modulation calculated from fitted sine functions

We also quantified phase modulation by computing a binocular interaction index (BII), which is mathematically equivalent to the phase modulation index but calculated from a sine curve fitted to the phase response function. The fitted sine curve was not constrained to be positive and hence the minimum was occasionally less than zero, yielding BII values that were greater than one. We used this measure to allow comparison with the series of studies reported by Ohzawa & Freeman (1986a,b) on the effect of binocular phase offset on neurons in area 17. Sine curves were fitted to the binocular phase offset responses of our sample of neurons and found to be an effective description of the data. The sine curves fitted to the response functions from the two neurons in Fig. 1A and B had correlation coefficients (r²) of 0.99 and 0.83, respectively. An additional measure of the goodness of fit, used by Ohzawa & Freeman (1986a), is the S/N ratio, determined by dividing the amplitude of the sine function by the root-mean-square of the residual error of the fit. A value of the S/N ratio greater than 2 was used by Ohzawa & Freeman (1986a) to indicate that the data were well fitted by a sine curve. The two neurons illustrated in Fig. 1A and B have S/N ratios of 8.4 and 7.0, respectively. Using a BII value > 0.3 together with an S/N ratio > 2 as the criteria for significance, again as adopted by Ohzawa & Freeman (1986a) for area 17 cells, 63 % of our area 21a neurons were disparity selective. The relation between BII and phase modulation index is plotted in Fig. 2B and shows that there is good agreement between the two values for most cells (dashed line is 45 deg). The open symbols in Fig. 2B are those with an S/N ratio less than 2, which are therefore not well fitted by a sine wave, and so do not produce reliable BII estimates.

Relation between binocular phase modulation and monocular parameters

One aim of this study was to assess the extent to which area 21a neurons were responsive to disparity as a global stimulus, independent of changes in stimulus orientation. We first determined the relationship between binocular disparity selectivity and optimum monocular orientation and spatial frequency for the dominant eye. The optimum orientation for dominant eye input to each neuron when plotted against their BII showed no correlation for our sample of 46 neurons (r < 0.03; Fig. 3A). Even some cells tuned to horizontal orientation, and therefore vertical disparities, showed marked response modulation by binocular vertical phase offset (Fig. 3A). There was also no relationship between optimum monocular spatial frequency of area 21a neurons and BII, as shown in Fig. 3B for all 46 cells (r= -0.16). The eccentricity of the receptive field centre of each neuron and the BII showed no correlation (r < 0.1).

Figure 3. Relation between binocular phase modulation and the optimum orientation and spatial frequency of area 21a neurons.

A, distribution of BII plotted against the optimum orientation determined with stimulation of the dominant eye for all 46 cells. B, distribution of BII plotted against the optimum spatial frequency for the dominant eye. Two of the 46 cells were classified as spatial low-pass, and are plotted at 0.02 cycles (c) deg⁻¹.

We also determined whether a balanced binocular excitatory drive was required for effective binocular disparity modulation of the response of area 21a neurons. We calculated an ocular dominance index for each neuron by dividing the ipsilateral response by the sum of the contralateral and ipsilateral responses (Tardif et al. 1996). Two-thirds of area 21a neurons lay within the range 35-65 % for ocular dominance. There was no correlation between the ocular dominance index for a cell and its BII (r= 0.01), but the two neurons that had values outside the range 10-90 % for the ocular dominance index were among the non-modulated neurons.

To elucidate the interactions between inputs from each eye that give rise to phase-modulated responses, we compared the maximum and minimum binocular response level with that evoked by monocular stimulation. For most cells, phase modulation produced response rates both above (binocular peak) and below (binocular trough) the dominant eye monocular response rate, as shown in Fig. 4. We interpret this to indicate that most of the area 21a neurons responsive to binocular phase difference exhibit both facilitatory and inhibitory interactions between the inputs from each eye. There were some cells that showed modulation that was predominantly facilitatory or inhibitory, and in some cases the binocular response was entirely greater than, or less than, the monocular response (those cells outside the bottom right quadrant of Fig. 4). In the sample of 34 cells, 82 % had a binocular peak greater than the dominant eye monocular response level, and 32 % of all cells had a binocular peak that exceeded the sum of the monocular response levels from each eye. The remaining 18 % included cells that showed little phase modulation, as well as cells that were strongly modulated, but showed an overall binocular suppression below the monocular response level. Nearly all cells (91 %) had a minimum binocular response level lower than the monocular dominant eye response. The mean decrease at the binocular trough was 46 ± 5 % which was very similar to the mean increase at the peak of 45 ± 9 %.

Figure 4. Facilitation and inhibition of area 21a neuron responses by binocular phase interaction.

The maximum binocular response (abscissa) and the minimum binocular response (ordinate) are expressed as the change from the monocular dominant eye response level.

Relation between phase modulation and interocular orientation-dependent suppression

We have previously described the immediate response suppression in area 21a that occurs by presenting a grating to the non-dominant eye that is oriented orthogonally to the optimally oriented grating presented to the dominant eye (Vickery & Morley, 1997). We confirmed our earlier assumption that there would no response modulation by binocular relative phase offset to the presentation of orthogonal gratings. We have now extended this study to compare the mechanisms of interocular orientation-dependent suppression and interocular phase-dependent suppression. Gratings presented to the dominant eye were at a fixed phase and orientation while gratings shown to the non-dominant eye had their phase and orientation systematically altered. Both interocular phase difference and orientation difference played a role in shaping the response of area 21a neurons. Phase modulation was greatest for iso-oriented gratings, as described in the preceding sections, and became progressively less effective as the orientation of the grating presented to the non-dominant eye was altered towards orthogonal.

We determined the extent of response modulation to relative phase and interocular orientation for each neuron. The relative binocular phase offset that yielded the highest response level at each interocular orientation difference was considered to be the optimum phase. In general the response at the optimum phase difference decreased with larger interocular orientation differences, as illustrated for a typical neuron in Fig. 5A. As we have reported previously, the suppression by an interocular orientation difference was a mean reduction of 35 ± 10 % (mean ±s.e.m.; Vickery & Morley, 1997). The responses at a phase shift of 180 deg from optimum for iso-oriented gratings were usually less than the response at a 180 deg phase shift for orthogonal gratings (Fig. 5A). This suggests that the response inhibition produced by an interocular phase difference is stronger than that due to interocular orientation difference. The phase modulation for 36 cells is plotted as the difference between the response at the optimum phase and at a phase shift of 180 deg against the difference between the response to the iso-oriented and orthogonal gratings at the optimum phase in Fig. 5B. There was a correlation between the extent of phase modulation and orientation modulation (r= 0.69), but response modulation due to a shift in relative phase was significantly larger than modulation due to a difference in interocular orientation (P < 0.005, Student's paired t test).

Figure 5. Relation between phase modulation and interocular orientation.

A, responses of an area 21a neuron plotted against the difference in interocular orientation. •, responses at the optimum relative phase offset at each interocular orientation; ○, responses at a 180 deg phase shift from the optimum. The dotted line shows the monocular dominant eye response at the optimum orientation. B, phase modulation for 36 neurons is plotted as the difference between the response at the optimum phase and at a phase shift of 180 deg from optimum against the difference between the response to the iso-oriented and to the orthogonal gratings at the optimum phase for 36 neurons. C, mean ±s.e.m. BII of nine neurons at interocular orientation differences ranging from 0 to 90 deg. The dashed line indicates a BII level of 0.3, which is the criterion level adopted to indicate disparity selectivity.

In the ten neurons that were studied over a range of interocular orientation differences, modulation by relative binocular phase offset decreased to non-significant levels for orthogonal gratings. In Fig. 5C the mean BII is plotted against interocular orientation difference for the nine neurons that were disparity selective for iso-oriented gratings. It is apparent that disparity selectivity is maintained over a wide range of interocular orientation differences as the BII at 45 deg interocular orientation difference still exceeds the criterion level of 0.3.

Comparison of monocular and binocular phase modulation

Monocular response modulation by a drifting sinusoidal grating can be measured as the relative modulation, which is the ratio of the response component at the stimulus temporal frequency (f₁) divided by the mean response (f₀). A binocular equivalent of relative modulation can be determined from the phase offset tuning curves and, for data that are well fitted by a sine wave, the binocular relative modulation is twice the BII. The monocular and binocular relative modulation values are plotted in Fig. 6 for 38 cells. Monocular and binocular modulation indices are significantly correlated (P < 0.01, Kendall's rank correlation, τ= 0.30). However, this correlation across all the cells is due to the strong contribution of only a few neurons. Those neurons with monocular relative modulation values < 1 and therefore classified as complex-like, showed no correlation between binocular and monocular relative modulation values (P > 0.5, τ= 0.08, n= 31). The binocular response modulation of these complex-like neurons was almost double their monocular modulation (mean ±s.e.m. monocular modulation, 0.46 ± 0.04; binocular modulation, 0.85 ± 0.09). The small group of simple-like cells had much greater monocular and binocular response modulation, but there was little difference between the two values (mean ±s.e.m. monocular modulation, 1.53 ± 0.10; binocular modulation, 1.70 ± 0.17). The simple-like cells also showed no correlation between their monocular and binocular modulation indices (P > 0.5, τ= 0.14, n= 7).

Figure 6. Relation between monocular and binocular relative modulation for the 28 neurons that had binocular phase responses well described by a sine function, and a monocular response exceeding 5 impulses s⁻¹.

The binocular relative modulation is twice the BII; the monocular relative modulation is the f₁/f₀ ratio (see Results). Cells with a monocular relative modulation greater than 1 were classified as simple-like cells (shaded area). The 45 deg line indicates expected performance for well-matched monocular and binocular modulation: this holds true for simple-like cells, but not for complex-like cells.

DISCUSSION

Position and phase in disparity selectivity

The selectivity of area 21a neurons to binocular disparity was determined using relative binocular phase offset between drifting gratings. We were not able to address directly whether disparity tuning at this level of visual processing involves signalling of receptive field position disparity or phase disparity. Although the animal was paralysed, we did not control for slow drift in eye position and so we did not attempt to determine the absolute position disparity between a cell's receptive field in each eye. Even with no eye movement, determining absolute position disparity for complex-like cells is difficult because their monocular receptive fields show little structural detail. The binocular receptive fields of complex cells in area 17 plotted by dichoptic stimulation indicate that complex cells make use of receptive field phase-based coding, although position-based coding cannot be discounted (Ohzawa et al. 1997). Simple cells in area 17 employ both mechanisms and so it is likely that both mechanisms contribute to the final disparity selectivity (Anzai et al. 1997). A drifting sinusoidal grating as used in the present study should, like most natural stimuli, effectively activate both systems and so our measured disparity response should correspond to that used by the behaving animal, irrespective of whether a phase and/or a position-based disparity coding mechanism is employed.

Measurement of disparity selectivity and comparison with previous studies

We assessed the extent of modulation in the response of area 21a neurons to a difference in relative phase between gratings presented to each eye by calculating a phase modulation index that uses only the maximum and minimum responses in the phase response function and a binocular interaction index (BII), which involved fitting a sine function to the phase response function. The BII has the advantage of weighting the responses of a neuron over the 360 deg range of phase differences used in the determination of the phase modulation. Although a sine function is a remarkably good fit, it is not the theoretical best fit (Ohzawa & Freeman, 1986a) and so the true extent of phase modulation probably lies somewhere between the values computed by the two methods. Regardless of the method chosen to determine the extent of phase modulation, approximately two-thirds of our sample of area 21a neurons were strongly modulated by a relative phase difference between otherwise identical gratings presented to each eye. Our findings are in marked contrast with the report of only 5 % of area 21a neurons displaying disparity selectivity (Wieniawa-Narkiewicz et al. 1992), but are consistent with those of Wang & Dreher (1996) who reported that almost 70 % of their sample of area 21a neurons showed significant response modulation to a disparity between single bars presented to each eye.

For the majority of area 21a neurons, the response modulation consisted of a response peak where the response was greater than the monocular dominant eye response, and a response trough where the response was lower than the monocular dominant eye response. This is similar to the findings of Ohzawa & Freeman (1986b) and Hammond (1991) for area 17 complex cells, but in area 21a the incidence of phase selective cells was almost double that of area 17. Eleven (32 %) of our sample of area 21a neurons displayed binocular facilitation, in which the response was greater than the sum of both the left and right eye monocular responses. This proportion of area 21a neurons displaying binocular facilitation is similar to a previous study where we reported that seven of 25 (28 %) neurons displayed binocular facilitation (Vickery & Morley, 1997). In addition to the binocular enhancement, we have shown that in area 21a neurons when the binocular phase offset was shifted 180 deg from optimum, almost all neurons had a response level that was less than the monocular dominant eye response. These findings indicate that area 21a neurons receive both excitatory and inhibitory input, and the relative balance between the excitation and inhibition depends on the nature of the stimulus in each eye. It remains unclear whether it is binocular enhancement or binocular inhibition that plays the more important role in signalling information about a particular phase disparity, and presumably therefore information about relative depth.

Orientation tuning and binocular disparity selectivity

The horizontal offset of the two eyes gives rise to binocular parallax that results in spatial disparities of an object's image on each retina. Retinal image disparity is considered a sufficient cue to the perception of relative depth, and it has been suggested that the visual system could be specialized to detect horizontal disparities, as horizontal disparities are of greater significance for stereopsis than vertical disparities (Barlow et al. 1967; LeVay & Voight, 1988). Indeed, Barlow et al. (1967) reported that neurons in cat area 17 responded over a much greater range of horizontal disparities than vertical disparities (6.6 deg vs. 2.2 deg, respectively).

More recently, experiments by Ohzawa & Freeman (1986a,b), using a stimulus protocol similar to that employed in the present study, and by LeVay & Voight (1988) found that neither simple nor complex cells in area 17 show a correlation between disparity selectivity and preferred orientation (see Fig. 9 in Ohzawa & Freeman, 1986b). In the present study we found no correlation between orientation preference and disparity selectivity for area 21a cells, in agreement with the findings of Wang & Dreher (1996). The role of neurons that are tuned to horizontal orientations and sensitive to vertical disparities is not known, but it has been suggested for area 17 neurons that they could play a role in detecting the vertical image disparities that result when an object is closer to one eye than the other (Barlow et al. 1967; Maske et al. 1986; LeVay & Voight, 1988). Area 21a neurons may also be involved in processing these vertical disparities.

Binocular orientation interactions are important in explaining phenomena such as binocular rivalry and tilt perception. In the present study we found that the disparity selectivity of neurons was maintained in the face of quite considerable changes between the grating orientations presented to each eye (Fig. 5C). The BII for our sample of cells indicated that disparity selectivity is retained for interocular orientation differences up to approximately 50 deg. A similar retention of disparity tuning at substantial interocular orientation differences has been reported for complex cells in area 17 (Nelson et al. 1977).

Mechanisms of disparity selectivity and interaction with interocular suppression

There appears to be a substantial interaction in area 21a between the binocular systems responsible for disparity detection and interocular suppression. There was strong correlation between the maximum phase offset modulation and the maximum interocular orientation-dependant suppression (Fig. 5B). Both of these measures used a common high point at the optimum phase and orientation, but the low points were due to a difference in interocular phase offset or a difference in interocular orientation (Fig. 5A). As the phase-dependant modulation is significantly greater than the orientation-dependent suppression, this may indicate that more potent inhibition is brought into play in enforcing disparity selectivity. It is interesting to note that both binocular orientation and phase-dependent suppression result from stimuli that generally show a net excitatory effect when presented monocularly. This implies that the net suppression observed in the binocular case is due to either presynaptic inhibition or postsynaptic inhibition that causes a marked shunting of current from synaptic inputs of the other eye.

Comparison of disparity selectivity in area 21a and primary visual cortex

The disparity tuning curves for neurons in area 21a are very similar to those reported for neurons in areas 17 and 18 (LeVay & Voight, 1988). The average optimum spatial frequency of area 21a neurons is substantially less than that of area 17 neurons (Movshon et al. 1978; Tardif et al. 1996; Morley & Vickery, 1997), which makes the retention of sharp disparity tuning all the more remarkable. As most area 21a cells have receptive fields and response properties that resemble complex cells in the primary visual cortex, we have restricted our comparisons to these cells. A substantially larger proportion of area 21a neurons display disparity-modulated responses than do complex cells in area 17 (∼70 % in area 21a compared with 40 % in area 17; Ohzawa & Freeman, 1986b). Furthermore, in addition to a greater proportion of cells that show significant disparity modulation, the extent of the modulation is also greater in area 21a than in primary visual cortex. The binocular phase modulation index observed in area 21a neurons had a median of 53 % (Fig. 2A), which was significantly greater than the median value of 35 % determined for complex cells in area 17 (Hammond, 1991). Binocular phase disparity modulation is twice as great as monocular phase selectivity for most area 21a cells, similar to complex cells in area 17 (Ohzawa et al. 1997). The association of these factors suggests that area 21a may play an important role in stereopsis.