Phase-disparity coding in extrastriate area 19 of the cat

Abstract

Binocular interactions were investigated in area 19 of the anaesthetized cat using dichoptically presented phase-shifted static spatial frequency gratings that flickered at a fixed temporal rate. More than two-thirds of the binocular cells showed phase specificity to static phase disparities leading to either summation or facilitation interactions. This proportion of spatial disparity selectivity was higher than that shown for the same area (one-third of the units) when drifting light bars or drifting spatial frequencies were used to create disparities. The range of phase disparities encoded by binocular cells in area 19 is inversely related to the optimal spatial frequency of the dominant eye. Thus, cells in this area are tuned to coarse spatial disparities which, as supported by behavioural studies, could reflect its involvement in the analysis of stereoscopic pattern having gross disparities but devoid of motion cues. Because of the nature of its interconnections with numerous visual cortical areas, area 19 could serve as a way station where stereoscopic information could be first analysed and sent to other higher order areas for a complete representation of three-dimensional objects.

The early work of Hubel & Wiesel (1962) showed that the signals from each eye converge upon single neurons in striate cortex. The great majority of these binocular neurons are highly selective to interocular disparities between image features and are known to form the neural substrate of depth perception (Barlow et al. 1967; Nikara et al. 1968). Thus, the neural interactions required for two of the fundamental properties of binocular vision, namely, stereopsis and fusion, are initiated at this first site of binocular convergence.

Several studies have shown that binocular cells in several visual areas of both the cat and monkey are able to code interocular disparities as in the striate area (for a review see Cumming & DeAngelis, 2001). Indeed, in the monkey, disparity-sensitive neurons were identified in most of its visual areas. However, based on behavioural studies and on the proportion of disparity detectors in the various areas, it has long been advanced that disparity processing is mainly carried out by areas located along the so-called dorsal stream (Maunsell & VanEssen, 1983; Livingstone & Hubel, 1987). Thus, studies have shown that a topographical map of disparity-selective cells is present in the thick stripes of monkey's area V2 (Ts'o et al. 2001) and recent evidence reported a columnar organization of disparity-selective units in the higher order extrastriate middle temporal (MT) area (DeAngelis & Newsome, 1999). The latter area receives massive inputs from the thick stripes of V2 (De Yoe & Van Essen, 1985; Shipp & Zeki, 1985). Areas along the dorsal stream are commonly known to process visual information related to motion and stereomotion and their ablations greatly affect perception of depth. Recent studies have, however, shown that cells in area V4 and in the inferotemporal area (IT), both located along the ventral stream, can also code binocular disparities (Uka et al. 2000; Hinkle & Connor, 2001). The former provides the major source of input to IT (Felleman & VanEssen, 1987). Although disparity-selective neurons in area IT are locally clustered, no clear columnar organization has been shown (Uka et al. 2000). A common property of cells located along the ventral stream is that they tend to respond best to complex forms, independently of the motion component of the stimulation. Inversely, cells in area MT and those along the dorsal stream contain only few units responding to binocular disparities devoid of motion cues (DeAngelis & Newsome, 1999).

In the cat visual system, there is no clear-cut segregation of information as in the monkey. However, parallel inputs from the retina (the X-, Y- and W-classes of retinal ganglion cells) to different visual areas have led Pettigrew & Dreher (1987) to propose a model for the functional processing of disparity information. The cat's visual system is provided with a large proportion of disparity-selective detectors. Indeed, numerous areas along with the striate (area 17) area (Lepore et al. 1992; Ohzawa et al. 1996, 1997) contain binocular cells responding selectively to interocular disparities created by drifting light bars or drifting sinusoidal gratings (area 19: Guillemot et al. 1993; Mimeault et al. 2002b; area 21a: Wang & Dreher, 1996; Vickery & Morley, 1999; posteromedial lateral suprasylvian area (PMLS): Mimeault et al. 2002a). The proportion of disparity-selective units varies from one area to another. However, it is clear that a larger number of these units (close to 80 %) are found in area PMLS (an homologue of area MT, see Payne, 1993) while the smallest proportion (less than 40 %) are found in area 19 (Bacon et al. 2000; Mimeault et al. 2002a,b). It appears that area PMLS of the cat, like its presumed primate homologue area MT, plays an important role in binocular processing and stereoperception (DeAngelis & Newsome, 1999; Mimeault et al. 2002a).

No study has attempted to verify whether cells in cortical areas of the cat respond selectively to binocular disparities without motion (drifting) components in the stimulation, which are not necessarily present in the normal viewing of a three-dimensional scene. A good candidate for this sort of coding, as suggested by behavioural and anatomical studies, could be area 19. This extrastriate area receives direct projections from the C-laminae of the dorsal lateral geniculate nucleus, the extrageniculate structures such as the medial part of the lateral posterior complex and from the medial interlaminar nucleus (Holländer & Vanegas, 1977; Berson & Graybiel, 1978). As these subcortical regions receive both Y and W inputs from the retina, area 19 is likely to be innervated by these two kinds of inputs. However, functional anatomy studies suggest that the principal visual input to area 19 is provided by the W channel (Dreher et al. 1980; Dreher, 1986). Thus, it is likely that poor direction selectivity and relatively poor orientation selectivity of area 19 neurons combined with their fairly sharp spatial selectivity (Duysens et al. 1982a,b) are related to paucity of X-type input and predominance of W-type input to this area. Moreover, behavioural studies indicate that area 19 is involved in form discrimination (Doty, 1971; Sprague et al. 1977; Hughes & Sprague, 1986), in the detection of stationary or moving figures on either stationary or moving noisy backgrounds (Krüger et al. 1988; Dinse & Krüger, 1990) and in texture segregation (De Weerd et al. 1994). In a recent study (Khayat et al. 2000), we also suggested that area 19 is implicated in form perception and texture segregation.

Experimental results suggest that neurons in area 19 have all the properties needed to code binocular disparities in large proportions. Indeed, these cells are mostly binocularly driven with small receptive field sizes, occupying the central part of the visual field (Hubel & Wiesel, 1969; Pettigrew & Dreher, 1987; Bergeron et al. 1998). The spatiotemporal properties of the left and right receptive fields are quite similar and highly matched (Bergeron et al. 1998) and stimulation of both eyes evokes strong interactive effects (Pettigrew & Dreher, 1987; Guillemot et al. 1993; Mimeault et al. 2002b). Thus, it is possible that the reason why only a few (a third) of the binocular neurons in area 19 are disparity-selective is due to the inclusion of the motion component in the stimulation (Guillemot et al. 1993; Mimeault et al. 2002b). The aim of the present study was to verify whether the receptive fields of binocular cells in area 19 are able to code selectively and show specific interactions to binocular presentations of phase-shifted spatial frequency gratings that are devoid of motion cues.

Methods

Animal preparation

The experiment was carried out on 10 adult cats weighting between 2 and 4 kg, which came from a Université de Montréal approved supplier. All manipulations were carried out in accordance with the guidelines proposed by the Canadian Council on Animal Care and with those of the National Institutes of Health (NIH). The university animal care committee approved all experimental protocols. All efforts were made to ensure the humane treatment of the animals and to minimize the number used.

The techniques of animal care and preparation, anaesthesia, surgery, recording, data analysis and eye movement control have been fully described in previous papers and will only be summarized herein (Bergeron et al. 1998; Mimeault et al. 2002a). On the day prior to the experiment, the animal was injected (i.m.) with 2 ml kg⁻¹ of dexamethasone sodium phosphate (5 mg ml⁻¹; Vetoquinol Canada Inc., Joliette, Canada) to limit inflammation during surgery. Before the induction of the anaesthesia the cat received an i.m. injection of atropine (Atro-Sol, 0.2 mg kg⁻¹; Ormond Veterinary Supply Ltd, Lancaster, Canada). The induction of the anaesthesia was performed using a facemask with a gaseous mixture of nitrous oxyde, oxygen (N₂O:O₂, 70:30) and isoflurane (5 %; Bimeda-MTC Animal Health Inc., Cambridge, Canada). The animal was intubated with an endotracheal tube connected to a respiratory pump and the rate was controlled so as to maintain a constant level of expired CO₂ (4 ± 0.5 %). Throughout the surgery, the anaesthesia level was maintained constant (1 to 2 %) to keep the animal deeply anaesthetized. A small trepanation was performed over the cortex representing area 19 (AP: 5 to -5; L: 4 to 8). A tungsten microelectrode that had an impedance of 3-6 MΩ (measured at 1000 Hz) was lowered into area 19 where the centre of the visual field is represented (Tusa et al. 1979).

All pressure points and incision sites were routinely infiltrated with local anaesthetic (xylocaine, Astra Pharma Inc., Mississauga, Canada). At the end of the surgery, the anaesthesia level was reduced progressively (0.5 % per 15 min) to a final level of 0.5 %, which was then kept constant during all the recording session. Body temperature was maintained constant (38 °C) with the help of a heated water pad. The absence of reflexes and a stable heart rate established that the level of anaesthesia was sufficient. During the recording sessions, the EEG and heart rate were monitored intermittently yet regularly, the former showing slow-wave activity and the latter a stable rate. From that point on, neuromuscular blockade was established with a mixture of gallamine triethiodide (Flaxedil: 200 mg; Rhone-Poulenc, Montréal, Canada) and d-tubocurarine (Tubarine: 15 mg; Sigma Chemicals, St Louis, MO, USA) dissolved in a 30 ml solution of lactated Ringer solution with dextrose (5 %). This mixture was continuously infused during the experiment through the saphenous vein (5.6 ml h⁻¹) to maintain neuromuscular blockade of the extraocular muscles. During neuromuscular blockade and throughout the recording sessions, artificial respiration was given to the animal using a respiratory pump. A full 2 h period of stabilization was allowed before the beginning of the recording sessions.

To prevent dehydration of the eyes, and improve image resolution, neutral contact lenses with an artificial pupil (3 mm) were placed on both eyes. The optic disks and major blood vessels were projected onto a tangent screen located at 57 cm from the nodal point of each eye. Then, appropriate dioptric lenses, as dictated by direct ophthalmoscopy, were placed in front of the eyes of the animal. The relative position of the areae centrales were considered to be located 16 deg medially and 7.5 deg below the iso-elevation of the centre of each optic disk (Bishop et al. 1962). The optic quality of the eyes and the position of the optic disks and major blood vessels were routinely checked before and after each quantitative protocol.

Visual stimulation and recording

Upon isolating a cell, the positions and limits of the receptive fields were first mapped on a translucent screen, located 57 cm from the animal's eyes, with light and dark bars using a manually controlled projector. The optical axis of the dominant eye was deviated, with the help of a prism, onto the tangent screen. This allowed both simultaneous and independent stimulation of each receptive field. The optimal stimulus parameters (orientation, width and length) were determined with sinusoidal spatial frequency gratings generated by a G3 Macintosh computer using Vpixx software (Vpixx Technologies Inc., Longueuil, Canada). The stimulation field (70 deg × 52deg) was back-projected with a LCD projector (Mitsubishi LVP-X100A) and the mean luminance of this field was 40 cd m². The resolution of the image was 11.9 pixels deg⁻¹ and the refresh rate was 75 Hz.

After determining carefully the limits of the receptive fields, the optimal orientation, temporal frequency and monocular spatial frequency tuning function were assessed for each cell. Thus, stimulation consisted of sinusoidal gratings whose contrast was modulated over time but not across space. That is, for each presentation, a grating appeared from a blank luminous field (contrast = 0 %), and the contrast of the pattern was increased sinusoidally to attain a maximum value that ranged between 20 and 50 %. The contrast of the spatial frequency grating flickered in each eye (i.e. increased from 0 % to maximum and decreased to 0 %) at a temporal rate of 1 to 6 Hz without ever changing its position in space. These static sinusoidal gratings were modulated in spatial frequency by steps of 0.25 octaves varying from 0.04 to 2.4 cycles deg⁻¹ and presented in each receptive field in a pseudo-random fashion. Each spatial frequency was presented 10 times at the best estimated orientation and temporal frequency (1 to 6 Hz).

A trial started with a gradual increase in contrast of the spatial frequency over a period of 500 ms, until it reached the appropriate contrast. The gradual increase was carried out in order to avoid a transient cell response to the sudden appearance of the stimulus. Then the spatial frequency grating was presented for 1000 to 2000 ms and an interval of 10 to 15 s was introduced between trials to minimize cell adaptation. While stimulating one receptive field, a blank field (mean luminance: 40 cd m²) was presented in the other. Special care was taken to adjust the size (width and length) of the gratings to the size of each receptive field in order to obtain the highest response rate. The receptive fields were classified following the procedures and criteria proposed by Skottun et al. (1991) and DeValois et al. (1982) in terms of simple, complex and end-stopped categories. The principal inclusion criteria for the latter class was that the cell preferred an oriented grating of optimal length, whereby extending the stimulus out of the boundaries of the receptive field caused an important decrease in response (end-stopping). This category therefore included both simple and complex receptive fields.

The phase disparity tuning function was next assessed by stimulating both eyes simultaneously with the optimal spatial frequency of the dominant eye having the same contrast and temporal frequency used to determine the spatial frequency tuning function.

The spatial phase of the optimal spatial frequency was varied in either eye, which created phase disparities. Phase disparities varied by steps of 22.5 deg and ranged from 0 to 337.5 deg. While the optimal spatial frequency of one eye flickered at its specific temporal rate and spatial position, the other flickered at the same temporal rate, appearing at the same spatial position (phase = 0 deg) or at a disparate spatial position (phase = 22.5 to 337.5 deg). For control purposes, monocular stimulation of each eye and a null condition (blank field) were interleaved with the disparity conditions in the stimulation protocol.

Histology

Electrolytic lesions were made in each recording track. At the end of the experiment, the cat was deeply anaesthetized with 5 % isoflurane (Bimeda-MTC Animal Health Inc., Cambridge, Canada), and perfused through the heart with isotonic saline followed with formalin (10 %). The brain was removed, placed in formalin and prepared for histology. Blocks of tissue containing the electrode tracks were coronally sectioned (40 μm) using a freezing microtome and stained with cresyl violet. All the recorded cells were in area 19.

Results

A total of 62 binocular cells were recorded and studied in detail. The receptive fields of the majority of the recorded cells were classified in either complex (58 %) or end-stopped complex categories (31 %). The remaining cells were classified into simple (3 %) and end-stopped simple (8 %) categories. The receptive fields of the cells were usually small in size and their spatiotemporal properties were comparable with those shown by others (Tanaka et al. 1987; Bergeron et al. 1998). However, since we used static spatial frequency gratings instead of drifting gratings, the optimal spatial frequency and spatial bandwidth distribution are therefore worth examining more fully. Since these spatial properties for the left and right receptive fields were closely matched, as appears to be a common feature for cells in this area (Bergeron et al. 1998), only the properties derived from the dominant eye will be briefly exposed. A typical cell, whose dominant receptive field was the ipsilateral one, is shown in Fig. 1A. As was the case for all the other units recorded, the cell had a band-pass tuning function and was optimally excited by a spatial frequency of 0.32 cycles deg⁻¹. The spatial bandwidth was measured at the half-height of the spatial frequency tuning curve. This measure is considered to be a good indicator of a cell's selectivity to spatial frequencies. The cell shown in Fig. 1A was relatively selective to spatial frequencies having a spatial bandwidth of 1.3 octaves. The cell shown in Fig. 1B was optimally excited by stimulation coming from the controlateral eye. The optimal spatial frequency of the latter was relatively low (0.16 cycles deg⁻¹). This cell was highly selective to spatial frequencies with a narrow bandwidth of 0.85 octaves.

Figure 1. Spatial frequency properties of all cells recorded in area 19.

Representative examples of spatial frequency tuning functions of the dominant eye for two complex cells (A and B). The stimuli used to derive the spatial frequency tuning functions were sinusoidal gratings (contrast: 30 %) flickered at a temporal frequency of 2 Hz (A) and 4 Hz (B). The distribution of the optimal spatial frequency (C) for the dominant eye of 62 cells shows that most units are sensitive to low spatial frequencies (mean = 0.18 cycles deg⁻¹). The spatial bandwidth distribution (D) of the dominant eye reveals that the cells are highly selective to static spatial frequency gratings (mean = 1.6 octaves).

The distribution of the optimal spatial frequencies for the dominant eye for 62 cells is shown in Fig. 1C. Most of the cells were sensitive to low spatial frequencies (mean = 0.18 cycles deg⁻¹; σ = 0.13 cycles deg⁻¹) ranging from 0.04 to 0.56 cycles deg⁻¹, although some (8 %) units were sensitive to higher spatial frequencies (≥ 0.41 cycles deg⁻¹). A closer examination of Fig. 1C reveals that a large proportion of cells (60 %) were selective for optimal spatial frequencies ≤ 0.16 cycles deg⁻¹ and some (32 %) for even lower optimal spatial frequencies (≤ 0.08 cycles deg⁻¹). The overall distribution of the optimal spatial frequency was not significantly different from what we had previously shown (Bergeron et al. 1998; t test, P ≥ 0.05). The distribution of the spatial frequency bandwidths for the dominant eye of all recorded units is shown in Fig. 1D. The spatial bandwidth ranged from 0.45 to 3.6 octaves (mean = 1.6 octaves; σ = 0.82 octaves). Most of the cells (71 %) were highly selective to spatial frequencies having bandwidths ≤ 1.8 octaves, although a few units (13.3 %) were tuned to coarse spatial frequencies (≥ 3.1 octaves). Our results suggest that cells in area 19 act as fine spatial frequency analysers, a property that has often been attributed to cells in area 17 (Movshon et al. 1978). These results are not significantly different from those obtained using drifting gratings (results compared to Bergeron et al. 1998; t test, P ≥ 0.05).

Figure 2 shows the tuning curve to phase disparity of a typical end-stopped simple cell. The peristimulus time histograms (PSTHs) obtained from stimulation of each eye are shown on either side of the tuning response profile. The PSTHs at the lower left and right of Fig. 2 show the modulated monocular response to the optimal spatial frequencies. It is clear that the maximum increment in response amplitude for this cell occurred at a phase disparity of 67.5 deg. The binocular response obtained at the worst phase disparity condition was situated between that of the contralateral and the ipsilateral monocular responses, as revealed by a decrease in amplitude shown by the PSTHs. The summed values of each PSTH at a particular disparity were used to define a point on the tuning response curve.

Figure 2. Peristimulus time histograms and phase disparity tuning function for a typical phase-sensitive cell.

The cell shows strong facilitation and an optimally modulated response at a phase disparity of 67.5 deg. Horizontal lines represent the monocular responses of the contralateral (contra) and the ipsilateral (ipsi) eye to the optimal spatial frequency. The phase disparity tuning functions were derived using optimal sinusoidal gratings (0.08 cycles deg⁻¹) at a temporal frequency of 4 Hz and a contrast of 30 %.

Figure 3 shows the phase disparity tuning curves of six binocular cells. In order to classify the phase disparity interactions, the tuning response to phase disparity must show a modulation profile. Thus, a modulation index was calculated using the formula: ((OR - WR/OR + WR) × 100), where OR represents the optimal response at one phase disparity and WR the worst response at another disparity (Hammond, 1991; Mimeault et al. 2002a). For example, a cell having an index of 100 % must have a null response (complete inhibition) at a specific phase disparity. A cell must have an index ≥ 30 % to be considered phase disparity-sensitive (Hammond, 1991; Mimeault et al. 2002a). These phase-sensitive cells often show binocular interactions of either facilitatory or summative types. The former is present when responses to binocular stimulations are higher than the sum of both monocular responses and the latter, when binocular responses are higher than the best monocular response (i.e. the dominant receptive field).

Figure 3. Examples of phase disparity tuning profiles for six binocular cells.

The cell in A shows a phase insensitive profile while the others are phase sensitive. Cells in B, D, E and F show binocular facilitation while the cell in C shows binocular summation to phase disparity. Horizontal lines represent the monocular response of the ipsilateral (I) and the contralateral (C) eye at the optimal spatial frequency. The stimuli used to derive the phase tuning functions were optimal sinusoidal grating for the dominant eye (cell in A: 0.28 cycles deg⁻¹; cell in B: 0.14 cycles deg⁻¹; cell in C: 0.12 cycles deg⁻¹; cell in D: 0.56 cycles deg⁻¹; cell in E: 0.1 cycles deg⁻¹; cell in F: 0.1 cycles deg⁻¹) at a temporal frequency of 4 Hz; except for the cell in F, which was tested at 2 Hz. The contrast was 30 % (cells in A to D) or 50 % (cells in E and F). Cells in A, B and E were classified as complex while cells in C and F were classified as end-stopped complex. The cell in D was classed as an end-stopped simple.

Of all the cells (n = 62) studied, a total of 29 % exhibited binocular interactions that were not phase specific. The tuning curve of the cell shown in Fig. 3A shows some binocular summation interactions at its optimal phase disparity but has a modulation index of only 23 %. Phase disparity between the gratings, therefore, does not significantly influence the responses of the cell. Thus, two classes of profiles can be easily distinguished from the phase disparity tuning curves: phase-sensitive and phase-insensitive categories

A total of 71 % (44/62) of the recorded neurons fulfil the modulation classification criteria (≥ 30 %) mentioned above and are thus classified as phase sensitive. Of these, 37 (84 %) showed facilitation and the remaining demonstrated summation interactions. A mirror image of these types of interactions could also be seen for the worst response: the cell showed either strong inhibition at some disparity values, so that its response was even below that of the non-dominant eye or an occlusive response to binocular stimulation such that the response was weaker than that of the dominant eye but stronger than the response of the non-dominant eye.

Typical phase disparity tuning profiles for sensitive cells are presented in Fig. 3. The cell whose binocular phase disparity tuning profile is illustrated in B showed facilitation interactions around the optimal (180 deg) phase disparity and a moderate modulation index of 66 %. The binocular response at the worst phase disparity was at the level of the non-dominant monocular response. The phase-sensitive neuron shown in C was optimally responsive to a phase disparity of 135 deg, which represents a summation interaction. However, it had a high (85 %) modulation index. This is due to the fact that the binocular responses at other phase disparities (202.5 to 67.5 deg) show inhibitory effects, where responses are lower than that of the non-dominant eye. The phase-sensitive cell whose tuning is shown in Fig. 3D had a facilitation interaction at a phase disparity of 90 deg while binocular responses at about 180 deg away from this optimal value were lower than that of the non-dominant eye. The latter cell had a modulation index of 69 %. The tuning profile illustrated in Fig. 3E shows a cell optimally excited at a phase disparity of 292.5 deg while its binocular response was nearly abolished at a phase disparity of 135 deg resulting in a strong modulation index of 96 %. A summation interaction was present at an optimal phase disparity of 225 deg for the cell whose phase disparity profile is shown in Fig. 3F. The binocular response at the worst phase disparity was between the levels of the monocular contralateral and ipsilateral responses. This cell was characterized by a weak (57 %) modulation index.

The distribution of the modulation indices of 44 phase-sensitive cells is shown in Fig. 4A. Cells having a modulation index < 30 % (phase insensitive) were not included in the distribution. In general, cells in area 19 had high modulation indices (mean 74 %) ranging from 43 to 97 %. Indeed, an important proportion (66 %) of the cells had strong modulation indices situated between 75 and 90 % Moreover, a non-negligible proportion (12 %) had even stronger modulation indices ≥ 91 %. Except for the cells we classified as phase insensitive using a strict phase sensitivity classification criteria of 30 %, all phase-sensitive neurons had modulation indices ≥ 43 %.

Figure 4. Modulation index derived from the phase disparity tuning profiles.

A, distribution of the modulation indices of 44 phase-sensitive cells. Cells having an index ≥ 30 are classified as phase sensitive. Most of the phase-sensitive cells have high modulation indices (mean = 74 %) ranging from 43 to 97 %. B, relationship between facilitation and the modulation indices of phase-sensitive neurons. The relationship between the facilitation index and the modulation index shows a positive correlation. Values located at and above the dashed line represent binocular facilitation interactions at the optimal phase disparity.

The amount of interaction of the phase disparity-sensitive neurons varied quite extensively, from simple summation to strong facilitation. In order to quantify the extent of the interactions across the complete sample of sensitive neurons, a facilitation index was calculated using the formula: (OR/I + C)) where OR represents the (best) optimal response to binocular phase stimulation and I and C the monocular ipsilateral and contralateral responses, respectively. Cells having a facilitation index ≥ 1 are those whose binocular responses are higher than the sum of both monocular responses (facilitation interaction) whereas those having an index < 1 have responses to binocular stimulation which are only greater than those of the dominant monocular response (summation interaction). Figure 4B shows the relationship between the facilitation index and the modulation index. Values that are located below the dashed line represent summation whereas those above this line represent facilitation interactions. Most of the phase disparity-sensitive cells (37/44) showed facilitation while the remaining showed summation interactions. These interactions seem to be closely related with the modulation profiles, as revealed in Fig. 4B. Indeed, a significant correlation between these two indices was found (r = 0.4; P ≤ 0.001). Cells that exhibited strong facilitation interactions tended also to have strong modulation tuning profiles while those showing simple summation interactions had more moderate modulation tuning profiles.

Since phase disparity corresponds to a change of position of the spatial frequency gratings inside the left and right receptive fields, cells having low optimal spatial frequencies should have larger spatial displacements than cells having higher optimal spatial frequencies for identical phase disparity values. Figure 5A shows the distribution of spatial displacement in deg (mean = 1.62 deg; σ = 1.60 deg) of visual angle. Results show that more than half of the phase-sensitive cells (52 %) were able to code small (< 1.5 deg) spatial displacements between the left and right receptive fields although a few (less than 15 %) coded values ≥ 3.5 deg. The cell whose phase disparity profile is presented in Fig. 3E had an optimal spatial displacement of 1.86 deg which is close to the average spatial displacement of our sample. However, the cells shown in Fig. 3B and D coded larger and smaller spatial displacements, respectively. Indeed, the former had an optimal spatial displacement of 3.57 deg while the latter, 0.45 deg. The remaining cells coded large spatial displacements (cell shown in C = 3.13 deg; cell shown in F = 3.11 deg). Figure 5B presents the relation between the spatial displacement and the optimal spatial frequency of the dominant eye. Cells responding to low optimal spatial frequencies coded larger spatial displacements whereas those having higher optimal spatial frequencies generally coded smaller spatial displacements. Indeed, a highly significant negative correlation was found (r = -0.4; P ≤ 0.001). For example, the cell shown in Fig. 3D had an optimal spatial frequency of 0.56 cycles deg⁻¹ and coded optimally a small spatial displacement of 0.45 deg, while the cell shown in Fig. 3F, whose optimal spatial frequency was 0.1 cycles deg⁻¹, coded a larger spatial displacement of 3.11 deg.

Figure 5. Relationship between displacement and spatial frequency of phase disparity-sensitive cells.

A, distribution of the optimal displacement in visual angle of phase-sensitive cells. B, relationship between optimal displacement and optimal spatial frequency of the dominant eye. The relationship is demonstrated by the regression line and is negatively correlated. C, distribution of the spatial phase bandwidth of phase-sensitive cells. D, relationship between the spatial phase bandwidth and the optimal spatial frequency of the dominant eye. The relationship is negatively correlated, as shown by the regression line.

In order to quantify the selectivity for phase disparity, a spatial phase bandwidth in terms of visual angle was calculated at half-height of the phase disparity tuning curves. The distribution of this spatial phase bandwidth is shown in Fig. 5C. A homogeneous distribution was found: some cells were finely tuned while others were coarsely tuned to phase disparity (mean = 3.05 deg; σ = 2.27 deg). Indeed, a total of 12 % of cells were highly selective to phase disparities (≤ 0.5 deg) while some 19 % of the units had larger phase bandwidths ≥ 5.1 deg. Cells shown in Fig. 3B and F were coarsely selective to phase disparity having phase bandwidths of 4.6 and 4.28 deg, respectively, contrasting with cells shown in D and E, which were sharply tuned. In effect, the latter had a narrow phase bandwidth of 0.72 deg and the former, 0.5 deg. Finally, the cell shown in C had a phase bandwidth of 2.05 deg. The spatial phase bandwidth is plotted against the optimal spatial frequency of the dominant eye for each phase-sensitive cell in Fig. 5D. The results indicate that cells code a larger range of phase disparities when they are sensitive to low optimal spatial frequencies whereas they seem to code a smaller range of phase disparities when they are sensitive to higher optimal spatial frequencies. This highly significant negative correlation (r = -0.74; ≤ 0.001) is represented by the regression line in Fig. 5D. These results clearly show that phase disparity sensitivity is dependent on optimal spatial frequency. Indeed, large phase disparities are coded by cells having low spatial frequencies and inversely for small phase disparities.

Discussion

The main objective of this study was to determine if binocular neurons could encode a static form of disparity for stimuli that are devoid of motion cues and if they do, to quantify the interactions. Our study shows that cells in area 19 do in fact selectively code static phase disparities. Indeed, a total of 71 % of the recorded cells showed modulated tuning response profiles and facilitation (84 %) or summation interactions (16 %).

Other studies have also demonstrated that cells in area 19 exhibit strong interactive effects when both eyes are stimulated simultaneously at disparate retinal loci (Pettigrew & Dreher, 1987; Guillemot et al. 1993). From our laboratory, Mimeault et al. (2002b) in fact found that 38 % of cells manifested interaction effects when drifting gratings were used to create the disparities. The major difference between the present study and those mentioned above is that we did not include a motion component in the stimulation. Indeed, a drifting grating is a periodical stimulus that constantly changes its spatial position over time, while a static grating is also a periodical stimulus in that it is modulated in contrast, but its spatial position and thus its spatial disparity in a binocular presentation is constant over time. The results, therefore, clearly indicate that for this area, stationary gratings presented with different spatial offsets to each eye are more efficient in driving the cells than moving gratings.

The disparity tuning curves of the cells reveal that area 19 may be implicated in the process of analysing the coarse spatial aspects of a three-dimensional scene. Indeed, our results show that cells in area 19 are coarsely selective to spatial disparities having relatively large spatial phase bandwidths (mean = 3.05 deg). This large selectivity can be explained by inputs to area 19. One of the major projections to area 19 is from the lateral-posterior pulvinar complex (Dreher, 1986) where cells have large receptive fields and coarse spatial tuning properties (Casanova et al. 1989). These make them inappropriate to signal fine spatial characteristics of disparity-tuned neurons. A similar argument could be made for its other major subcortical sources of inputs, namely, the Y and W inputs from the C-lamina of the lateral geniculate nucleus and the medial interlaminar nucleus. On the other hand, in the present study, we found receptive fields of cells that were small in size and some were highly selective to spatial frequencies (mean = 1.6 octaves; see Fig. 1D). These properties are generally attributed to simple cells in area 17 (Movshon et al. 1978). Cells in area 19, although mainly coarsely selective to phase disparities, do in some cases code small spatial displacements (< 0.5 deg; see Fig. 5A) and have very narrow (< 0.5 deg; see Fig. 5C) phase bandwidths. An example of this fine selectivity is evident for the cell shown in Fig. 3D. This fine spatial disparity selectivity could be explained by its direct inputs from areas 17-18 (Rosenquist, 1985; Scannell et al. 1995). These latter areas are known to code fine spatial selectivity (Movshon et al. 1978; Lepore et al. 1992).

Behavioural studies showed that the integrity of the X system is not required for the discrimination of form (Doty, 1971). However, its disturbance produces a significant reduction of visual acuity (Berkeley & Sprague, 1979). This suggests that the Y and W systems, which are present in area 19 of the cat, could play a role in the analysis of form and coarse spatial details of the visual scene. Binocular cells in area 19 that are sensitive to static phase disparities could be implicated in the analysis of aspects of a three-dimensional scene that does not require high spatial resolution. The analysis of the texture of objects is one such possibility and is supported by a study from our laboratory that showed that an important proportion of cells in area 19 responded to textured bars on textured backgrounds (Khayat et al. 2000).

A parallel can be made between the processing of disparity information for the cat and monkey. Since many neurons in the dorsal visual pathway of the monkey are disparity-selective, it has been thought that this pathway was the most important for stereopsis (Maunsell & Van Essen, 1983; Livingstone & Hubel, 1987). However, Uka et al. (2000) have shown that a large proportion of cells in area IT located along the ventral visual pathway are selective to binocular disparities, which suggest that there is more than one pathway for stereopsis. Area V4, also having a large proportion of disparity-selective units, is a probable source of disparity information for IT (Hinkle & Connor, 2001). In the cat, it has been shown that the largest proportion of disparity-selective cells are found in areas 21a and PMLS (Vickery & Morley, 1999; Bacon et al. 2000; Mimeault et al. 2002a). The latter is known, as is its homologue MT, to process motion information related to the spatial dimension of the visual scene (Payne, 1993). Studies of Lomber et al. (1996a,b), have suggested that the ventral-posterior suprasylvian gyrus (mostly area 20) of the cat could be the homologue of area IT of the monkey. Area 20 receives a major source of inputs from area 19 (Rosenquist, 1985; Scannell et al. 1995). Area 19 of the cat, therefore, which has a large proportion of static disparity-selective cells, as shown in the present study, could feed disparity information to area 20 as area V4 does to IT in the monkey (Hinkle & Connor, 2001).

Considering the numerous connections of the visual cortical areas with area 19 (Rosenquist, 1985; Scannell et al. 1995), it is likely that the disparity processing in area 19 could be forwarded to other visual areas. It is thus possible that this area could serve as a way station to higher order areas. The final analysis of disparity with or without motion cues could be achieved by higher order areas PMLS/21a or 20a, respectively, areas that are strongly connected with area 19 (Rosenquist, 1985; Scannell et al. 1995). Rather than having an independent pathway for disparity processing, the visual system of the cat as in the monkey seems to process disparity information in parallel. Indeed, the multiple visual areas could each play different roles in stereoperception.