Using systematic electrophysiological mapping, architectonics and the global pattern of interhemispheric connectivity, we have identified three visual areas in the lateral most part of the posterior suprasylvian gyrus. The most posterior and largest area we call area 20a and anterior to this we defined a smaller area, area 20b. These areas lie lateral to the visual areas 18, 19 and 21 and posterior to a third, but incompletely defined, visual area, area PS. Areas 20a and 20b, emphasize the representation of the upper hemifield. Their interhemispheric connections conform to the so called ‘midline rule’ in that they are abundant in regions representing central portions of the visual field, scarce or absent elsewhere. These areas are probably homologous to the homonymous areas of the cat and might be indicative of a Bauplan from which the temporal areas of primates may have evolved.
The cortical visual areas of primates, including those of man, are organized into dorsal (‘where’) and ventral (‘what’) processing streams (Ungerleider and Mishkin, 1982; Ungerleider and Haxby, 1994). The dorsal stream consists of the parietal visual areas (posterior parietal cortex and the MT complex) and has been implicated in attentional mechanisms, determining the location of objects in space and the preparation and guidance of motor sequences. The ventral stream consists of the temporal visual areas and has been implicated in object recognition. The ‘where’ and ‘what’ pathways have been described in primates, including man, but it is unlikely that they evolved de novo in primates and they might be found in other mammals with complexly organized cortical visual systems.
In a previous paper we detailed the areal organization of the posterior parietal cortex of the ferret and noted several similarities with that of the macaque monkey (Manger et al., 2002b). Despite these similarities, the posterior parietal cortex of the ferret consists of only two areas, whereas that of the macaque monkey appears to contain substantially more. This potential increase in complexity of posterior parietal cortex in the macaque monkey might be causally related to the evolution of a greater range of eye and hand movements in the primates as compared to the carnivores.
In primates, there are at least five temporal visual areas (Kaas, 1997; Rosa, 1997). They are characterized by neurons with large receptive fields, sometimes extending into the ipsilateral visual field, and the precision of the retinotopic maps progressively deteriorates with increasing distance from the occipital visual areas (Desimone and Gross, 1979; Boussaoud et al., 1991). Moreover, very specific neuronal response properties emerge, with neurons tuned to the identification of particular objects and/or faces (Gross et al., 1972; Fujita et al., 1992; Kobateke and Tanaka, 1994; reviewed in Perret et al., 1987; Desimone, 1991). These response properties appear not to be restricted to the temporal visual areas of primates. Electrophysiological investigations of the temporal lobe of sheep have demonstrated neurons specifically tuned to conspecific ‘faces’ (Kendrick et al., 2001); however, it is not known exactly which cortical areas were involved.
Given previous findings in the cat (Tusa and Palmer, 1980; Updyke, 1986; see Discussion), it is likely that the ferret also has several temporal visual areas. Exploratory recordings in this region in the course of another study indicated this to be the case (Manger et al., 2002a). The present study reports the results of our experiments examining retinotopic maps and global interhemispheric connectivity patterns in the temporal region of the ferret visual cortex — part of the putative ventral stream of the ferret.
Materials and Methods
Twelve normally pigmented adult female ferrets (Mustela putorius) weighing between 600 and 1000 g were used. The animals were bought from a local authorized breeder and all experiments were performed according to the Swedish and European Community guidelines for the care and use of animals in scientific experiments.
Four series of experiments were carried out, for electrophysiological mapping (n = 6), architectural analysis (n = 2), interhemispheric connections (n = 2) and combined electrophysiological mapping and interhemispheric connectivity (n = 2). Our histological, hodological, and electrophysiological methods have been described previously (Innocenti et al., 2002; Manger et al., 2002a,b), therefore here we detail mostly what was specific to the present series of experiments.
As in Manger et al. (2002a,b), the animals were initially anesthetized with i.m. doses of ketamine hydrochloride (Ketalar, 10 mg/kg) and medetomidin hydrochloride (Domitor, 0.08 mg/kg), supplemented with atropine sulphate (0.15 mg/kg). An i.v. cannula was secured into the femoral vein and a tracheal tube (5 mm diameter) inserted. The animal was placed in a stereotaxic frame, paralyzed with an initial i.v. bolus of pancuronium bromide (0.15 mg/kg), maintained with a continuous infusion of the same (6 µg/kg/h) and artificially ventilated. Stable conditions were maintained for anaesthesia (1% isoflurane in a mixture of 1:1 nitrous oxide and oxygen), ventilation (3.5–4% expired CO2 ) and body temperature (37–38°C). Heart rate was monitored throughout the experiment.
Multiunit activity was recorded with varnish-isolated tungsten microelectrodes (impedance between 0.95 and 1.3 MΩ) inserted perpendicularly in the exposed temporal cortex. In five cases, only the dorsolateral surface of the temporal cortex was examined (two of which were combined with tracing of interhemispheric connectivity) and in three cases both the dorsolateral and ventral surfaces were explored. The electrophysiological signal was amplified and filtered, visualized on an oscilloscope and played through a loudspeaker. Recordings were made at distances of between 500 and 1000 µm below the pial surface, or when investigating the ventral surface of the temporal lobe, the electrode was lowered to depths of up to 2.5 mm in 250 µm steps. Each penetration in a given antero-posterior row of recordings was separated by between 250 and 500 µm, dependent on the cortical vasculature. Each recording row was separated medio-laterally by 300–1000 µm. Visual receptive fields were mapped with stationary or moving luminous white circles, varying between 1 and 50° diameter and were plotted on a hemispheric screen in front of the animal, on which the optic disc had been ophthalmoscopically projected. Receptive fields were mapped to the contralateral eye, normally the dominant eye, unless stated otherwise. The center of the dome corresponded roughly to the center of the visual hemifield.
At the end of the recording session, the animal was given a lethal dose of pentobarbital (80 mg/kg, i.p.) and perfused intracardially, initially with a rinse of 0.9% saline solution (4°C, 500 ml/kg), followed by fixation with 4% paraformaldehyde in 0.1 M phosphate buffer (4°C, 1000 ml/kg). The brain was removed and post-fixed overnight in the same solution, then transferred into a 30% sucrose solution in 0.1 M phosphate buffer (at 4°C) and allowed to equilibrate. The hemisphere from which we recorded was partially manually flattened (as in Manger et al., 2002a,b) and 50 µm sections were cut on a freezing microtome in a plane parallel to the pial surface. Alternate sections were treated for cytochrome oxidase (CO) reactivity, or myelin. In the cases where the ventral surface of the temporal region was investigated the sections were taken in a coronal plane and stained for Nissl, CO and myelin. For the reconstruction of recording sites, the sections were examined with a stereomicroscope and the outlines of the section, larger blood vessels, tissue artifacts and recording sites marked on a camera-lucida drawing. Architectonic boundaries, particularly the anterior border of area 18 and the medial border of temporal cortex, were determined from the CO and myelin staining at this stage. In the two cases in which connections and mapping were combined, the reconstructed connections were aligned with the architectonic borders at this stage.
For the determination of the physiological boundaries, several criteria were employed simultaneously. These included reversals in receptive field visuotopic progressions, changes in the size and response properties of the receptive fields. Manual reconstruction of the maps (as in Manger et al., 2002a,b) and computer-aided reconstructions were undertaken (as in Manger et al., 2002a). For the computer-aided reconstruction, information from each recording site was entered into Excel (Microsoft) and processed in Matlab v. 5 (MathWorks Inc.) as: azimuth (degrees from zero meridian), elevation (degrees above/below horizontal meridian) and receptive field size (horizontal and vertical diameters in degrees). In this study we used the zero meridian, defined as the most anterior azimuth of the leading edge of the combined receptive fields, in place of the vertical meridian. For the azimuth and elevation, circles of an appropriately scaled size were overlaid on the position of the recording site using a customized program, Image Mapper (written by Laurent Tettoni). For the receptive field sizes, lines representing the width and height of the receptive field were overlaid onto the recording sites. These allowed immediate visualization of all the results of the recording site for each parameter; moreover, the different parameters could be superimposed to examine for any congruency between them and with architectonic borders and patterns of interhemispheric connectivity.
Architectural studies were performed on four hemispheres with material fixed and treated as described above. The hemispheres were sectioned (50 µm thick serial sections) and alternately stained with cresyl violet, reacted for CO (Carroll and Wong-Riley, 1984) and for myelin (Gallyas, 1979). Two hemispheres were sectioned in a coronal plane and two were sectioned in a parasagittal plane (normal to the pial surface of the temporal cortex). Electrophysiologically mapped hemispheres cut parallel to the pial surface and stained for CO or myelin were also used in this analysis. Sections were observed and photographed using low- and high-power light microscopy.
Injections of Tract Tracers
The animals were initially anesthetized as described above and placed in a stereotaxic frame. A mixture of 1% isoflurane in 1:1 nitrous oxide and oxygen was delivered through a mask, while the animal maintained its own respiration. Anesthetic level was monitored using the eye blink and withdrawal reflexes, in combination with measurement of the heart rate. The temporal cortex was exposed under aseptic conditions. A few electrophysiological recordings were made to ensure placement of the tracer within the temporal cortical areas. The vascular pattern was also useful, as large veins were often found to demarcate the border between temporal and occipital cortical areas and the anterior limit of the temporal visual areas. To examine the global pattern of interhemispheric connectivity, between four and six injections of ∼500 nl of 10% wheat germ agglutinin conjugated to horseradish peroxidase (WGA–HRP; Sigma) dissolved in phosphate buffer (0.1 M, pH 7.35) were made throughout the temporal cortex. The spread of tracer resulted in a large, homogenous injection site across all of the temporal cortical areas.
After completion of the injections, a soft contact lens was cut to fit over the exposed cortex, the retracted dura pulled over the contact lens and the excised portion of bone repositioned and held in place with dental acrylic. The temporal muscle was reattached using surgical glue and the midline incision of the skin sutured. Antibiotics were administered in all cases (terramycin, 40 mg/kg/day). The animals were given a 3 day recovery period to allow for tracer transport.
At the end of the recovery period, the animals were given a lethal dose of pentobarbital (80 mg/kg, i.p.) and perfusion-fixed with 3% paraformaldehyde. The cerebral hemispheres were dissected from the remainder of the brain and cut in a semi-flattened plane, parallel to the pial surface as in Manger et al. (2002a,b). Alternate 50 µm thick sections were stained for CO, or reacted to reveal the transported WGA–HRP (Innocenti et al., 2002). In one WGA–HRP case the brain was sectioned in a parasagittal plane and the sections alternately stained for CO, cresyl violet, myelin and reacted for transported WGA–HRP. The labeled cells were plotted using a camera lucida. Architectonic details of the cortical areas were determined and superimposed onto the plots of the labeled cells.
The visually responsive region of the temporal cortex explored in the present study is a small (10–15 mm2 area), wedge shaped portion of cortex occupying the most lateral (ventral) part of the posterior suprasylvian gyrus, between the convexity of the hemisphere and the posterior suprasylvian sulcus. Within this region three visual areas were identified by electrophysiological criteria, containing mirror reversal representations of the contralateral hemifield. Following the nomenclature of similarly located areas in the cat we named these areas, from posterior to anterior, 20a, 20b and PS (posterior suprasylvian). Cytoarchitecture, myeloarchitecture and CO staining, as well as interhemispheric connectivity provided additional criteria for the parcellation. The medial border of these areas runs roughly antero-posteriorly for ∼5 mm. This border is with the visual areas 18 and 21 (Innocenti et al., 2002; Manger et al., 2002a) and further anterior with the, still uncharted, suprasylvian visual areas; it is often marked by a large vein. Laterally, these areas extend ∼0.5 mm beyond the convexity of the suprasylvian gyrus, onto the ventral aspect of the hemisphere, where they are adjacent to a visually unresponsive region.
Areas 20a, 20b and PS contain roughly mirror representations of, essentially, the upper quadrant of the contralateral visual hemi-field extending not further than 30° below the horizontal meridian. In both areas the peripheral visual field is over-represented. Indeed, about half of their surface represents azimuths beyond 45° and up to 100°. These features distinguish the temporal areas from the other visual areas of the ferret charted thus far. In these other areas the central visual field is over-represented and the upper and lower quadrants are either equally represented, as in areas 17, 18, 19 and 21 (Law et al., 1988; Manger et al., 2002a) or the lower quadrant predominates, as in the parietal areas (Manger et al., 2002b). Examples of the results of electrophysiological mapping experiments are shown in Figures 1–3 and a synthetic rendering of the visual field representation in these areas is shown in Figure 8. Figure 4 illustrates the extent of the visual field representation and Figure 5 the progression of receptive field size with eccentricity in the visual areas of the ferret.
This is the largest of the two partitions of area 20, extending ∼3.5 mm antero-posteriorly. The medial border of area 20a is with area 21 anteriorly (and in some cases, in part, with the suprasylvian visual areas) and posteriorly with area 18; it corresponds to high values of elevation. In rows of penetrations oriented from medial to lateral, perpendicular to this border, the receptive field positions move from locations at the extreme periphery of the upper hemifield towards the horizontal meridian. The medial border of area 20a, therefore, corresponds to a mirror reversal of the upper hemifield representation found in the adjacent areas 18 and 21.
The central visual field is found posterior and the representation of the zero meridian extends over 2 mm on the postero-lateral edge of the lateral suprasylvian gyrus. At 10° above the horizontal meridian and ∼5° below it, receptive fields centered at azimuths up to 35° were found. These receptive fields were large and often extended with their periphery up to the zero meridian. In rows of penetrations directed from posterior to anterior, the values of azimuth increase progressively to beyond 70° until the transition between areas 20a and 20b is reached. Therefore the anterior border of area 20a corresponds to a representation of the visual field periphery continuous with that at the anterior border of area 21. At least half of area 20a is devoted to representing the periphery (azimuths >45°). The azimuths close to the zero meridian were compressed close together and movement of the electrode of 250 µm could often result in a shift of azimuth >30°. The largest values of azimuth were found at the anterior-medial corner of the area and therefore the isoazimuth lines tend to run diagonally across the area.
The horizontal meridian is represented close to the lateral edge of area 20a and courses in an approximate antero-posterior direction. The lower visual field representation is found lateral to that of the horizontal meridian and is limited to ∼–30° in elevation and to ∼45°in azimuth. Medial to the horizontal meridian, elevations steadily increase into the upper visual field with isoelevation lines running approximately antero-posteriorly. Receptive fields with upper borders lying 90° above the horizontal meridian were a common occurrence.
The neurons in this region responded vigorously to moving circles of white light, especially when the diameter of the stimulus was 20° or more. The neurons responded best to a stimulus that ‘swept’ across the entire receptive field. The size of the receptive fields showed a strong tendency to increase with increasing azimuth. Interestingly, those receptive fields found in the representation of central vision were only slightly larger than corresponding receptive fields in area 17. However, those fields representing the periphery were very large, similar in size to corresponding receptive fields in area 20b and parietal cortical areas. While this trend may seem unusual, the correlation of azimuth and receptive field size is quite high (R2 = 0.728).
This area spans ∼1.5 mm antero-posteriorly. Its medial border is continuous with that of area 20a and corresponds to high values of azimuth and elevation. Its posterior border corresponds to the representation of the visual field periphery at the anterior border of area 20a. The anterior border of area 20b corresponds, in its lateral two-thirds to the representation of the zero meridian of the visual field. As recordings were made more medial on the anterior border, receptive fields centered at progressively larger azimuths were found, corresponding to elevations >30°. However, these receptive fields often extended with their periphery up to the zero meridian. The azimuths increase progressively moving posterior and medially with the largest values found at the postero-medial corner of the area. Therefore the isoazimuth lines tend to run diagonally across the area and join those of area 20a.
The horizontal meridian is represented laterally and is continuous with that of area 20a. As in area 20a, the elevations proceed smoothly from medial to lateral, with the upper hemifield medially and a restricted representation of the lower hemifield laterally.
Neuronal response properties to moving circles of light were similar to those found in area 20a; however, these neurons did prefer slightly larger stimuli, up to 30°. They responded vigorously to ‘sweeping’ stimuli and did not habituate. In this area there was a trend for increasing receptive field size with increasing azimuth; however, it was not a strong trend. The sizes of the receptive fields in area 20b are similar to those seen in the posterior parietal cortex of the ferret. While differences in neuronal responses were not sufficient to differentiate between areas 20a and 20b, the clear reversal in the topographic progression of receptive fields provided an accurate location of their mutual border.
A full physiological investigation of this region proved to be difficult due to a large overlying blood vessel and because the most anterior part of this area appeared to be within the most lateral portions of the suprasylvian sulcus. However, the observations made indicated that this area was indeed separate from the other two temporal areas. Examples of retinotopic reversals in this area are shown in Figures 1 and 3. As summarized in Figure 8, the retinotopic representation appeared to form a mirror image of that seen in area 20b.1
The posterior border of this region is with area 20b, where a representation of the zero meridian occupied the lateral two-thirds of this border, with increasing azimuths occupying the medial one-third. Often, only two or three recording sites could be made across the extent of this field in a recording row and these recordings showed increasing azimuths towards the anterior border of the field. Thus, along the anterior and medial borders of PS a representation of the periphery was found. The isoelevation lines were continuous with those in more posterior temporal areas and a similar bias in representation of the upper visual field was observed.
This region appears to have different neuronal stimulus preferences to those of areas 20a and 20b (which were quite similar). Very slowly moving circles of white light, with diameters of 30° or more elicited the most vigorous neuronal responses. The response of the neurons was not driven by ‘sweeping’ the stimulus across the field, but, rather, by moving the stimulus around within the receptive field. Turning the stimulus on and off also elicited a strong response.
Architectonics of the Lateral Temporal Areas
The physiological representations described above were located in a cortical territory that is architectonically distinct from the surrounding visual (located medially) and perirhinal areas (located laterally). Within the lateral temporal cortex the architectonic differences between physiological representations are subtle; however, these differences are, for the most part, compatible with the electrophysiological maps.
Areas 20a and 20b are together distinguished from the medial visual areas (see Fig. 3 in Innocenti et al., 2002) by a less cell dense layer 5, containing few or no large pyramidal neurons and by a thick layer 6, but with poorly differentiated cellular palisades (Fig. 6). Layer 4 is difficult to define because of its poor granularity and it contains mainly small pyramids. These features are slightly more marked in 20b than in 20a.
Areas 20a and 20b are lightly myelinated and contain thin radial fascicles traversing layer 6 to layer 3 (Fig. 6). A network of tangential fibers characterizes layers 3, 4 and 5, particularly in 20a, but laminar differences cannot be distinguished. Both these areas are readily distinguished from the more laterally located perirhinal cortex which evidenced extremely light staining for myelin.
A distinguishing anatomical feature of the whole region is its pale reactivity for CO. This feature is most marked in sections taken parallel to the pial surface, where it clearly delineates the temporal cortex from the more medial visual areas, corresponding to the reversal of receptive field progression around the upper periphery of the visual field in the three areas mapped (Fig. 7). In coronal and parasagittal sections, the same border is marked by the disappearance of a strongly CO reactive layer 4, especially so in areas 17 and 18 (Innocenti et al., 2002). The border between lateral temporal areas and perirhinal cortex was not as clear; however, the perirhinal cortex was slightly more reactive to CO than the lateral temporal regions. No further distinction between the three physiological representations of the temporal cortex could be made on the basis of this stain.
At the anterior border of area 20b, area PS (not shown) is characterized by a better-developed layer 4, larger pyramidal neurons in layer 5, a higher density of myelination and weaker CO reactivity.
Interhemispheric Connections to Areal Borders and Visuotopic Representations
Interhemispheric connections, were traced with large, multiple injections of WGA–HRP in the homotopic region of the contralateral hemisphere (Fig. 8). The large injection sites effectively filled the tip of the posterior suprasylvian gyrus. The brains were studied in flattened preparations as well as in parasagittal sections. The latter also allowed determination of the laminar distribution of the interhemispheric connections.
Labeled axons were followed from the injection site to both the corpus callosum and to the anterior commissure. These large injections did not allow us to determine if axons traveling in the different commissures had a different origin and/or destination. However, more localized injections (unpublished results) indicated that the anterior commissure route was followed by axons connecting homotopically the temporal areas, while axons connecting the temporal areas with heterotopic and more dorsal areas of the contralateral hemisphere were traveling in the corpus callosum. As in the occipital visual areas, interhemispherically projecting neurons were found mainly in layers 3 and 5–6 and were of the pyramidal type. As was seen in the occipital areas (Innocenti et al., 2002; Manger et al., 2002a), there was a distinct heterogeneity in the pattern of interhemispheric connectivity in the visual temporal areas of the ferret and this pattern related to both areal borders and the representations of the visual midline (Fig. 8).
In tangential sections through this region, interhemispherically projecting neurons and terminating axons were distributed in two mediolaterally oriented bands, one posterior in area 20a, the other anterior in 20b (Figs 8 and 9). These two bands are, for the most part, separated by a non-connected region, but join laterally so that the whole territory containing interhemispheric connections is U-shaped. The interhemispheric connections of the temporal areas are separated from those of the occipital areas by a band free of commissural connections, continuous with the non-connected band described above (Fig. 8). On the basis of the visuotopic maps and architectonic criteria, it appeared that the bands devoid of interhemispheric connections lie respectively at the medial border of the visual areas of this region and at the 20a/20b border (Fig. 8). Therefore, it seemed probable that, as in the occipital visual areas (Manger et al., 2002a), the interhemispherically connected regions would correspond to the representation of the more central portions of the visual field, including the zero meridian, while the representations of more peripheral portions of the visual field would be devoid of these connections.
This hypothesis was verified in combined anatomical and electrophysiological experiments (Fig. 9). Receptive fields with centers within 40° of the zero meridian were all within the interhemispherically connected region. Receptive fields with centers between 40 and 70° were found both in connected and non-connected regions. More peripheral receptive fields were restricted to non-connected regions. In conclusion, the distribution of interhemispheric connections provides strong supporting evidence for the parcellation of this cortical region determined using electrophysiological mapping and also correlates to architectonic observations. The interhemispherically connected regions correspond to the postero-lateral border of area 20a and also to the area 20b/PS border. The non-connected regions correspond to the remaining borders of these visual areas and provide clear markers of the medial borders of all three areas, as well as corresponding to much of the medial portion of the area 20a/20b border. There was only one region of cortex in which a border between areas did not have a corresponding interhemispheric connections marker; this was the lateral portion of the area 20a/20b border. In this region the representation of the visual field was found to contain smaller azimuths, less than 70° (Figs 8 and 9). This region corresponded to the curvature of the interhemispheric connections that bridged the two major bands of connectivity. Thus, the correspondence between interhemispheric connections and visual midline holds in this region, however, the areal border becomes cryptic in terms of interhemispheric connectivity at this location.
The present study used a variety of techniques to define three temporal visual areas of the ferret — areas 20a, 20b and PS. Topographic reversals in the progression of receptive fields around the upper periphery of the visual field were found along the medial borders of all three areas. This reversal coincided with an increase in the size of receptive fields and with noticeable changes in their response properties. It was matched by changes in architectonic features and corresponded to a strip of cortex that was not connected interhemispherically. Thus, five criteria concur in differentiating these three cortical areas from the remainder of visual cortex explored thus far (Innocenti et al., 2002; Manger et al., 2002a,b). The three areas could be differentiated from each other by topographic reversals in the progression of receptive fields around the zero meridian (the 20b/PS border) and the periphery (the 20a/20b border) of the visual field. These reversals coincide with differences in the interhemispheric connectivity (non-connected at the 20a/20b border, connected at the 20b/PS border) and at the 20b/PS border with architectonic features and neuronal response properties. Since the techniques employed in this study have already been used successfully to parcellate the ferret visual cortex we will not discuss the methodology as this was done previously (Innocenti et al., 2002; Manger et al., 2002a,b).
Comparisons with the Cat and with other Visual Areas of the Ferret
The macroscopic morphology of the temporal lobe of the ferret shows both similarities and differences with that of the cat. In both species the medial/dorsal boundary of the temporal lobe is constituted by the sylvian sulcus, (sometimes called pseudosylvian sulcus). The sylvian sulcus is longer and deeper in the ferret than in the cat. The temporal lobe of the ferret is subdivided longitudinally by the posterior extension of the suprasylvian sulcus. This sulcus exists in the cat, where an additional sulcus, the posterior ectosylvian sulcus is intercalated between the sylvian and the suprasylvian sulci. These are typical differences between the orders Mustelidae and Felidae (Welker, 1990) which may relate to an expansion of the auditory areas in the cat. In both cat and ferret the posterior suprasylvian sulcus corresponds roughly to the border between auditory areas (anterior) and visual areas (posterior) (Brugge and Reale, 1985; Wallace et al., 1997). The auditory areas of the cat, therefore, comprise a wider portion of the temporal lobe, including the posterior ectosylvian sulcus. In both cat and ferret, the temporal visual areas occupy the lower portion of the posterior suprasylvian gyrus, which, in this region, merges with the postero-lateral gyrus. An interesting, similarity is between the posterior suprasylvian sulcus of the carnivores and the anterior portion of the superior temporal sulcus of the rhesus monkey, which also demarcates the border between the visual and the auditory areas (e.g. Pandya and Yeterian, 1985).
As mentioned earlier in this paper, we wish to suggest that the part of the temporal lobe of the ferret explored in this study is homologous to that of the cat. As in the cat (Tusa and Palmer, 1980; Updyke, 1986), the temporal cortex of the ferret contains three separate representations of the visual field, in areas 20a, 20b and PS. The arguments in favor of the homology are the location of the areas, their internal organization, cytoarchitectonics and connectivity. None of these criteria, however, is absolutely stringent and therefore our interpretation must be considered tentative. In both species, areas 20a, 20b and PS are bordered medially by the occipital visual areas (areas 17, 18, 19 and 21) and laterally by perirhinal cortex. However, area 20b is ventral to 20a in the cat and anterior to 20a in the ferret. Both areas extend more ventrally and medially in the cat than in the ferret and, in the cat, area 20b borders with area 17, not with 21, nor with the suprasylvian areas (Tusa and Palmer, 1980; Updyke, 1986). Area 20a of the cat borders areas 18, 19 and 21. This is similar to the ferret except for the absence of a border with area 19, which appears to be foreshortened medio-laterally compared to the cat (Manger et al., 2002a). In both species, all three areas, are strongly biased towards the representation of the upper visual field, with much of the lower visual field having no discernable representation — cf. Tusa and Palmer (1980) and Updyke (1986) with present findings. In both, the periphery of the upper visual field forms the boundary of the temporal areas with the more medial visual areas and the lower periphery is the boundary with the visually unresponsive region. The horizontal meridian runs from posterior-lateral to anterior-medial in the ferret and mainly medio-laterally in the cat.
The major difference between the visual field representations in ferrets and cats is the orientation of the maps. The 20a/20b border corresponds to the vertical meridian in the cat and to the periphery of the visual field in the ferret. In the ferret, the vertical meridian is mapped at the posterior boundary of area 20a and at the anterior border of area 20b, both of which represent peripheral azimuths in the cat. Thus, the retinotopic maps in area 20a and 20b contain mirror representations of the visual hemifield in both species, but there is also a mirror reversal between species. The differences in retinotopic organization of this region in the ferret and the cat are difficult to reconcile; however, the results in the ferret are supported by the pattern of interhemispheric connections, which showed a strong congruency with the mapping studies (see below). Comparable data seem to be lacking for the cat.
Previous anatomical studies of the cat have provided similar descriptions of this cortical region as described here for the ferret. Sanides and Hoffmann (1969), Gurewitsch and Chatschaturian (1928) and Heath and Jones (1971) have all distinguished the lateral temporal cortex from the medially placed occipital and laterally placed perirhinal cortex. A brief architectonic description was provided by Sanides and Hoffmann (1969). At locations on the posterior suprasylvian gyrus corresponding to areas 20a and 20b of the ferret they described two fields, peVlp anteriorly and pL posteriorly. Both areas were less myelinated than the more medial cortex. The anterior field was characterized by a blurred lamination, layer 4 containing mainly small pyramids and was equated to area 19 of Brodmann. The posterior field was considered paralimbic cortex and was characterized by a diminished layer 3 and by prevailing layers 5 and 6. The felid architectonic studies are confirmed in their observations by immunohistochemical (SMI-32) studies of the same region in the cat, where differences between areas 20a and 20b were not obvious (Van der Gucht et al., 2001). A similar set of observations on the cytoarchitecture of the lateral temporal region was made by Brodmann (1909) in another carnivore, the kinkajou. Thus, our architectural observations concur with previous studies and also argue for homology of the lateral temporal cortical region across carnivore species.
What Constitutes the Ventral/‘What’ Processing Stream of Carnivores?
We introduced the idea that the ventral and dorsal visual processing streams of primates (Ungerleider and Mishkin, 1982; Ungerleider and Haxby, 1994) are unlikely to have evolved de novo in the primate lineage and that these processing streams might be found in other species with complexly organized cortical visual systems. In our previous study of posterior parietal cortex in the ferret (Manger et al., 2002b) it was concluded that, as a whole, the posterior parietal cortex of the ferret could be seen as related to a Bauplan, from which that of the macaque monkey evolved and thus formed part of the dorsal/’where’ processing stream.
In primates, all cortical areas lateral to the middle temporal complex of areas (MT, MTc, MST and FST), such as the posterior part of the inferotemporal cortex (posterior IT or TEO), the anterior part of the inferotemporal cortex (anterior IT or TE), other inferotemporal areas, plus occipital visual areas V1, V2 and V4, are considered to form the ventral processing stream (Ungerleider and Mishkin, 1982; Ungerleider and Haxby, 1994). In previous studies (Innocenti et al., 2002; Manger et al., 2002a) homologues of primate V1 and V2, ferret areas 17 and 18, were determined. Moreover, we described an area, area 21, that can be considered either analogous or homologous to area V4 of primates. Thus, in terms of the occipital areas of the ventral processing stream the ferret is comparable to the primates.
Payne (1993) has argued that the temporal visual areas of the cat include not only areas 20a, 20b and PS, but also areas DLS and VLS. Moreover, Payne suggests that these areas have direct homologues in the macaque monkey: 20a being homologous to TF, 20b to TH, VLS to TEO, DLS to TE2, and PS/EPp to TG/TE1. The three areas defined here for the ferret, 20a, 20b and PS, appear to be homologous to the same named areas in the cat (see above), thus we have defined at least three temporal cortical areas that may belong to the ventral processing stream in the ferret. Exploratory recordings made during the course of the present and previous studies (Manger et al., 2002a) have indicated the existence of visually responsive cortex anterior to area 21 and medial to areas PS and 20b, this cortex most likely being the ferret counterparts of cat DLS and VLS. Thus, whatever decision is reached regarding the ventral processing stream of either the cat or the ferret, it most likely applies to both species and possibly to all carnivores.
Comparison of the Ferret with the Macaque Monkey
The comparison of the temporal areas of the carnivores with those of the macaque monkey is more difficult since only the retinotopy of one of the areas, area TEO, seems to be known in detail (Boussaoud et al., 1991; see also Rosa, 1997). This area is bordered posteriorly by V4, which we speculated elsewhere might correspond to area 21 of the carnivore (discussed in Innocenti et al., 2002; Manger et al., 2002a). Area TEO appears to be characterized by a blurred retinotopy that can be best described in terms of rough gradients of retinal representation. Somewhat similar to areas 20a and 20b of the ferret, area TEO contains an exaggerated representation of the upper visual field. However, this is in terms of the proportion of cortical surface devoted to the representation while receptive fields cover similar portions of the upper and lower hemifields. The transition with area V4 in macaques is not sharp in terms of retinotopy and occurs in the periphery of the upper hemifield as in the ferret. From the anatomical point of view area TEO is less myelinated than more medial areas (Boussaoud et al., 1991), a feature which is also in common with the temporal areas of cat and ferret.
In spite of the some similarities between TEO and area 20a, arguments also exist in favor of an homology between areas 20a and 20b of the carnivore with areas TF and TH of the primate (Payne, 1993). Despite these observations, it appears that the visual part of the temporal lobe of primates has undergone a major expansion compared with that of the carnivore. The temporal lobe expansion in primates appears to include an increase in the number of areas and changes in their internal organization. The situation resembles in this respect that of the parietal cortex where despite some common features of global organization, the rhesus monkey appears to have far more cortical areas than the ferret (Manger et al., 2002b). Although the extent of the reorganization is uncertain and it awaits a more detailed definition of the number of areas and connectivity in both species, it seems that the temporal cortex of primates underwent a less dramatic increase in the number of areas than the parietal cortex. This might relate to the fact that while the organization of the temporal cortex is tailored to perceptual processes that are common across species, the organization of the parietal cortex reflects hand–eye coordination mechanisms elaborated in primates. In this light it would be interesting to know if the temporal areas we identified in the ferret, contain neurons selective for complex stimulus features as in the monkey (Gross et al., 1972; Fujita et al., 1992; Kobateke and Tanaka, 1994; reviewed in Perret et al., 1987; Desimone, 1991) as such neurons were described in sheep (Kendrick et al., 2001). It would also be interesting to know if the remarkable developmental plasticity found in the parietal cortex after early lesions of the primary visual areas (Restrepo et al., 2003) applies to this area, as it seems to do in the monkey (Webster et al., 1991a,b). Developmental plasticity might be greater in areas that have undergone extensive evolutionary reorganization.
Retinotopy, Areal Borders and Interhemispheric Connectivity in Higher Order Visual Areas
Interhemispheric connectivity has been touted as a useful criterion in determining cortical area boundaries in the visual system, due to the observation that interhemispheric connections arise mostly from cortical regions representing the zero (or vertical) meridian, corresponding to areal boundaries in occipital visual areas (Innocenti, 1986; Payne and Sivek, 1991). Consistent with this, in the ferret, the representations of the vertical meridian (and of adjacent regions) at the 17/18 and at the 19/21 borders were found to be interhemispherically connected (Manger et al., 2002a). However, the relationship between interhemispheric connectivity and cortical area boundaries is confounded by the existence of complex retinotopic maps in the occipital visual areas resulting in complex patterns of connectivity, and thus must be interpreted carefully and in combination with electrophysiological mapping (Manger et al., 2002a).
The border between the temporal areas and the more medial visual areas is denoted by a stripe devoid of interhemispheric connections, which extends laterally into the border between areas 20a and 20b. This stripe contains receptive fields centered beyond 40° from the zero meridian. The posterior border of 20a and the anterior border of 20b correspond to more central portions of the visual field are heavily interhemispherically connected. From this it can be concluded that the interhemispheric connections of the temporal areas of the ferret, as those of areas 17, 18, 19 and 21, conform to a somewhat stretched ‘midline rule’, as discussed elsewhere (Innocenti et al., 2002; Manger et al., 2002a). Thus far, in the ferret the only visual representations clearly violating the so-called midline rule appear to be the parietal areas (Manger et al., 2002b). Nevertheless, the degree of visual field represented in interhemispherically connected region seems to increase from areas 18 and 19 to the temporal areas. In areas 18 and 19, only receptive fields centered within 10–15° from the zero meridian were invariably in callosally connected regions and few if any were found beyond 40° (Manger et al., 2002a). In the temporal cortex, instead, receptive field centers within 40° from the vertical meridian were found invariably in interhemispherically connected regions and only receptive field beyond 70° of azimuth were invariably in non-connected regions. Taking into account the larger size of receptive fields in the temporal areas, it can be estimated that the portion of the visual field which is connected to the opposite hemisphere is at least double in the temporal areas compared with occipital areas 18 and 19. It should be noticed that a part of the visual space represented in the interhemispherically connected region probably falls in the ipsilateral visual hemifield. Given tthe difficulties in determining ophthalmoscopically the retinal landmarks with sufficient precision and in evaluating the degree of ocular torsion due to paralysis, the estimate of the extent of ipsilateral visual field representation in the ferret awaits systematic recordings of receptive fields in the two hemispheres. Estimated with this method, the nasotemporal overlap is ∼10° at the 17/18 border (Innocenti et al., unpublished), corresponding to previously estimated values (White et al., 1999).
This finding is of interest as it indicates: (i) that a greater range of the visual field is processed between hemispheres in the higher order visual areas and (ii) that the extent of binocular processing might be increased in these regions of cortex through interhemispheric neural mechanisms in animals with lateral eye placement. The extended interhemispheric connectivity in the ‘what’ pathway might provide completeness to images that are limited by the lateral placement of the eyes. Alternatively, these connections might subserve attention switching mechanisms alerting the animal to salient images falling on the opposite hemifield, as has been suggested in studies of the zebra finch (Schmidt and Bischof, 2001).
The work reported in the present study was funded by a grant to G.M.I. from the Swedish Medical Research Foundation (No. 12594) and a Wenner-Gren Foundation fellowship to P.R.M. The authors wish to thank Ms Paola Santunione for her assistance in the computer-aided analysis used in the present study and Linda Danielson for her help with the analysis of the connections.