radius) containing the fixation point. The stimulus in Fig. 1A was displaced by 20° of polar angle from the vertical meridian; other stimuli (outlined in Fig. 1B) were displaced by either 0°, 5°, 10°, or 40°. Stimuli were presented either to the left or right hemifield within a given scan, always in alternation (16-sec epochs) with a uniform gray control stimulus including a central fixation point. Subjects were instructed to stare continuously at the fixation point during fMRI acquisition.
The stimulus displacement from the vertical meridian increased with eccentricity, to approximate the well-known and systematic decreases in cortical magnification factor with eccentricity. The rationale for this stimulus configuration was as follows. Because cortical receptive field size generally increases as cortical magnification decreases within a cortical area, a thin vertical line displaced from the vertical meridian would be expected to stimulate a smaller range of cortical polar angles at a large eccentricity than at a small eccentricity. To overcome this bias, for each stimulus, the edge of the occluding aperture was moved further away from the vertical meridian with increasing eccentricity. The goal was to shift the representation of the aperture’s medial borders roughly as a line across cortex, approximately equal in cortical distance from vertical meridian representations, irrespective of stimulus eccentricity. The topography of the retinotopy in the contralateral hemisphere (see below) suggest that the polar coordinate stimuli used here approximately achieved this result.
Consistent with our basic hypothesis, such stimuli produced significant activation in the ipsilateral hemisphere. Fig. 1C–G shows the typical pattern in one subject, produced by the stimulus in Fig. 1A. The ipsilateral activation produced by these stimuli has a distinct topography, consisting of two broad branches (see Fig. 1 and below). One branch extends superiorly toward inferior parietal cortex, and the other one runs antero-posteriorly along the inferior lateral surface. Finding two distinctive branches of ipsilateral activation was not obviously predicted by the anatomical topography of callosal connections in previous animal experiments (3–8).
In most cortical regions, the amplitude of the ipsilateral magnetic resonance (MR) increase was not as high as that in the contralateral hemisphere; this finding is consistent with the generality of crossed visual input. In fact, in cortical regions showing significant contralateral retinotopy (e.g., areas V1, V2, and V3), there were consistent, significant decreases (blue through white) in response (relative to the control stimulus) during presentation of our ipsilateral visual field stimuli. This unusual finding does not appear to be because of “blood stealing” in the fMRI signals, at least in any simple way. However, the results are consistent with existing single-unit electrophysiology in animals demonstrating response inhibition by ipsilateral stimuli. In macaque area V4, at least, it has been reported that inhibitory ipsilateral influences extend much further into the ipsilateral visual field, compared with excitatory influences (14).
In the contralateral (control) hemisphere, these stimuli produced a pattern of activation predictable from the shape of the stimulus relative to previously described retinotopic maps in areas V1, V2, V3, VP, V3A, and V4v. For example, Fig. 2A shows a map of the contralateral retinotopy in one subject, and Fig. 2B shows the contralateral response to our most restricted (40°) stimulus (see Fig. 1B), in the same subject. This unilateral stimulus activates the cortical representations of the contralateral horizontal meridian, but spares the representation of the vertical meridian and the foveal representation—exactly as predicted by the stimulus geometry. These control results confirmed the appropriateness of our polar coordinate stimuli, the fixation stability, and our understanding of the contralateral retinotopy.
Further analysis reveals that the ipsilateral activation is systematically related to other topographical features in the visual cortical map. For instance, the ipsilateral activation produced by the mid-range (20°) stimulus appeared to be concentrated immediately anterior to those areas showing classical (contralateral) retinotopy (e.g., V1, V2, V3, VP, V3A, and V4v). To test this idea directly, we compared the contralateral and ipsilateral representations in the same subjects, in the same hemispheres. Fig. 3 shows such a comparison, produced by stimuli in left and right visual hemifields (activated in different scans). It suggests that the ipsilateral representation indeed “begins” approximately where the contralateral retinotopy “ends.” Although the thresholds in such a comparison are not directly comparable, this same contralateral-to-ipsilateral retinotopic transition is apparent even when other thresholds and visual field extents are chosen (see Fig. 2 and below).
When using this same stimulus, the superior branch of ipsilateral activity borders the anterior portion of V3A. which is the most anterior and most coarsely retinotopic vertical meridian representation revealed by our current tests of contralateral retinotopy (see Fig. 3). That branch then continues superiorly with significant activation across the anterior segment of the transverse occipital sulcus, continuing anterior to and past the superior terminus of the parieto-occipital sulcus. The inferior branch of activity always begins near the foveal representation of V3/VP, extends somewhat inferiorly, and then somewhat superiorly through and beyond the motion-selective area MT (see Figs. 1, 3, and 4).
The above experiments reveal only a single “snapshot” of the ipsilateral representation. However, it is known that ipsilateral influence actually extends continuously but nonuniformly into different cortical regions, dependent on both cortical area boundaries and the retinotopy (if any) within each area (3–8, 16).
This complex spatial relationship was revealed by presenting the full set of ipsilateral stimuli (see Fig. 1B), within the same hemisphere (e.g., Fig. 4). Retinotopic cortical visual areas also were labeled as described elsewhere (23–31). The stimuli closest to the vertical meridian produced thin strips of activation extending along the representations of the vertical meridian, especially along the borders between V1 and V2 (see Fig. 4C and D). The appearance of activation along the vertical meridians of the ipsilateral hemisphere coincides with the complete filling-in of activation in the classic retinotopic areas of the contralateral hemisphere.
These tests also reveal differences in the overall extent of ipsilateral influence between different cortical areas. The most anterior retinotopic areas (V3A and V4v) show distinctively greater interhemispheric activation compared with immediately adjacent areas V3 and VP, across a considerable range of stimulus extents (see Fig. 4B–D). Although the ipsilateral architecture has not been imaged previously, these results are generally consistent with: (i) the retinotopy of these areas in humans (e.g., refs. 23–31) and macaques (e.g., ref. 39), (ii) related differences in receptive field sizes among areas in macaque (40) and human (24) cortex, and (iii) anatomical studies of callosal (interhemispheric) connections in animals (3–8, 16).
The two-branched topography revealed above is not predictable from the anatomical topography of callosal connections in macaques (3–8, 16) or humans (17–19). Of course, the topography of callosal connections is incompletely known in hu-