Retinotopy and Functional Subdivision of Human Areas MT and MST

Identifying MT+

MT+ was identified, separately for each subject, based on a combination of anatomical and functional criteria. Specifically, a contiguous region was marked by hand to include voxels on the lateral surface of the occipital lobe, where the fMRI time series correlated strongly with the moving/stationary stimulus alternations (r > ∼0.5, chosen separately for each subject). Figure 2A shows a flat-map representation of MT+ localizer responses in one hemisphere.

MT+ was similarly localized bilaterally in all hemispheres of all subjects (Figs. 3A, 4A). Its location was anterior to and distinct from the retinotopically defined areas V1, V2, V3, V3A, and V4v, whose locations had been identified previously in all subjects. It fell mostly or entirely within a single sulcus. We occasionally noticed (three hemispheres) a swath of activity slightly posterior and/or ventral to MT+ on the flat maps. Despite its close proximity to MT+, we excluded this patch of activity from MT+ for two reasons. First, this activity was often found in a different sulcus (or sulci), with MT+ clearly on the other side of an intervening gyrus [a fact somewhat obscured on the flat-map representation but more evident when the data are viewed in sagittal slices of the three-dimensional (3D) brain volume]. Second, the application of a high correlation threshold (higher than that used in the figures) to the MT+ localizer responses revealed a clear distinction between MT+ and this posterior–ventral activity. In fact, the MT+ localizer stimulus elicited activity throughout much of the occipital lobe; the responses were simply stronger (i.e., withstanding a higher correlation threshold) in MT+.

Identifying MT: angular component of retinotopy

Area MT was defined, separately for each subject, to include a contiguous subregion of MT+ that exhibited strong response modulations during the retinotopy scans. The same correlation threshold was applied to the MT+ localizer (Fig. 2A) and the retinotopy data (Fig. 2B). Because we collected equal numbers of repeats of both of these conditions, and because the duty cycles of both of these stimuli were the same (see Materials and Methods, visual stimuli), applying the same correlation threshold allowed for a fair comparison of the spatial extent of the responses to these two stimulus conditions.

A retinotopic subregion of MT+ is clearly visible in Figure2B and is marked by the green curve drawn on the flat map. The fact that this subregion of MT+ responded strongly to the rotating-wedge stimulus suggests that neurons within this area have relatively small receptive fields. In addition, the phase map varies smoothly from magenta/red (upper-left quadrant of visual field) through purple [horizontal meridian (HM)] through blue/cyan (lower-left quadrant), suggesting orderly retinotopic organization.

Area MT was discernable based on strong responses to the rotating-wedge stimulus in both hemispheres of all subjects (Figs.3B, 4B). We were also able to discern a qualitatively clear and orderly retinotopic phase map in 5 of the 10 hemispheres. In all hemispheres for which the angular retinotopic map was easily discernable, the representation of the upper vertical meridian (UVM) was anterior to the representation of the lower vertical meridian (LVM). Even when the retinotopic map within the area did not include a smooth progression of phases, we chose to define the area whenever there was a contiguous subregion of MT+ that responded strongly to the rotating-wedge stimulus. A strong response to the rotating-wedge stimulus implies that neurons within each fMRI voxel had relatively small receptive fields at the same or nearby locations in the visual field. Thus, even when visual inspection did not reveal an orderly retinotopic phase map, the presence of a strong response modulation provided evidence of relatively small receptive fields and local retinotopic organization.

Identifying MST: ipsilateral stimulation

Area MST was defined, separately for each subject, to include a contiguous subregion of MT+, distinct from retinotopically defined MT, that responded strongly to peripheral, ipsilateral stimulation. Figure 2C shows the ipsilateral responses in the right hemisphere of one subject. Although ipsilateral responses were relatively weak compared with contralateral responses, a subregion of ipsilateral activity was clearly identifiable, marked by the cyan curve drawn on the flat map. This same subregion did not respond strongly to the retinotopy stimulus; this double dissociation is evident by contrasting Figure 2B,C.

Area MST, as defined by the dual criteria of a response to ipsilateral stimulation and lack of a strong response modulation to the retinotopy stimulus, was evident in both hemispheres of four of the five subjects (Figs. 3C, 4C). In defining MST, we first noted the subregion of MT+ that did not exhibit a strong modulation of response to the retinotopy stimulus and then defined area MST as a nonretinotopic region that did respond strongly to the ipsilateral stimulus. Furthermore, we chose to identify MST only if a strong ipsilateral response was not also present in the retinotopic region. We note that these conservative criteria sometimes left some parts of MT+ unclassified (neither MT nor MST). MST, in the eight hemispheres in which it was identified, was always anterior and often dorsal to MT, although there was some degree of variability across subjects. In these eight hemispheres, MST typically abutted MT; when some degree of separation was apparent, the areas were still within ∼5 mm of one another along the gray-matter surface.

However, in the remaining subject (R.F.D., left and right hemispheres) we did not observe a clear double dissociation between two subregions of MT+. Although we were able to identify a retinotopic MT subregion in both hemispheres of this subject, responses to ipsilateral stimulation were either too weak or too diffuse to confidently identify a distinct MST region. Ipsilateral responses in both hemispheres of this subject were notably weaker than those observed in the other subjects. Also, in both hemispheres of this subject, the anatomical location of MT+ was less distinct and did not fall primarily within a single sulcus, as it did in most subjects. Because the local cortical anatomy of this region is quite variable across individuals (Watson et al., 1993; Tootell et al., 1995; Dumoulin et al., 2000), our sample is too small to determine where subject R.F.D.'s organization lies with respect to the normal range. Critically, the failure to identify MST in this subject demonstrates that our procedure (first identifying a retinotopic subregion and then looking for a distinct subregion that responded to ipsilateral stimuli) did not logically guarantee that we would observe the desired double dissociation.

Central and peripheral retinotopic representations

We also measured the cortical representations of the central and peripheral portions of the visual field by alternating moving dots within central (4° outer radius) and peripheral (4° inner radius; 16.5° outer radius) parts of the visual field. The response to the central versus peripheral stimulus for one subject is shown in Figure2D. A large central representation is evident in the ventral portions of both MT and MST, and a smaller peripheral representation is evident in the more dorsal portions, particularly within area MT.

The subregion of MT corresponding to the central representation of the visual field typically (in 8 of 10 hemispheres) covered the ventral (and sometimes posterior) extreme of MT. Responses to peripheral stimulation were typically found at the dorsal and/or anterior borders of MT, although responses to the peripheral stimulus covered much less cortical area than responses to the central stimulus. Responses in this experiment were rather noisy, particularly in MST, consistent with larger receptive fields that might be expected to cover both the central and peripheral stimuli. Despite the noise, we did observe a clear response to the central stimulus in MST in five of the eight hemispheres in which we were able to identify MST. The representation of the central part of the visual field in both regions provides additional evidence that MST reflects a distinct cortical area and not simply the peripheral retinotopic representation of a single, larger area.

Position and size of MT+ subregions

To better evaluate the relative positions of these areas in the 3D cortical volume, we transformed the regions corresponding to MT and MST from the flat map to the corresponding gray matter in the high-resolution anatomy images of each subject's brain. In all of the eight hemispheres in which we were able to define both MT and MST, we observed that MT fell primarily on the posterior (-ventral) bank of a sulcus, whereas MST fell on an anterior (-dorsal) bank. This sulcus could usually be identified as a dorsal/posterior limb of the ITS (Dumoulin et al., 2000). Although this dorsal/posterior continuation of the ITS was the clearest anatomical landmark, we also observed that MT+ sometimes continued posteriorly into the lateral occipital sulcus and/or onto the lateral occipital gyrus.

Viewing the fMRI responses in the high-resolution volume anatomy also reveals the relative positions of MT and MST and confirms that the geometrical distortion inherent in transforming the functional data to the flat maps did not introduce any systematic artifacts. Figure5A shows the responses to the MT+ localizer, the retinotopy stimulus, and the ipsilateral stimulus on an axial slice in one subject. In the left panel, a strong response to the MT+ localizer is evident on both sides of the sulcus (the center of the sulcus is indicated by the arrow). In the center panel, a strong response to the retinotopy stimulus is evident only on the posterior bank of the sulcus. Conversely, in the right panel, the region responding to ipsilateral stimulation lies on the anterior bank of the sulcus. Similar organization is evident in Figure 5B,C, which shows coronal and sagittal slices, respectively, in two additional subjects.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Spatial separation of retinotopy and ipsilateral responses in the cortical volume. A, fMRI responses shown in axial slices through the cortical volume (subject A.R.W., right hemisphere). The arrow indicates the center of the sulcus. Note that localizer responses (MT+) fall on both sides of the sulcus (left panel), retinotopy responses (MT) fall primarily on the posterior bank (middle panel), and ipsilateral responses (MST) are primarily restricted to the anterior bank (right panel).B, Coronal slices (subject A.A.B., left hemisphere, same format as in A). C, Sagittal slices (subject D.J.H., left hemisphere, same format as in A).A, Anterior; L, lateral;D, dorsal; M, medial; P, posterior.

Table 1 reports the gray-matter surface area for MT, MST, and MT+. On average, MT subsumed ∼243 mm² and MST subsumed ∼83 mm². Sizes for MST are likely to be underestimates, because of the conservative criteria used in defining this area (see Results, Identifying MST: ipsilateral stimulation). Size can also be estimated by visual inspection of the figures, because all flat maps had a 35 mm radius. Postmortem anatomical studies in a similar part of human cortex have identified a region of dense myelination, believed to correspond to MT, covering ∼200 mm² (Tootell and Taylor, 1995).

Inferences about neuronal receptive field sizes

We interpret the retinotopy and the ipsilateral measurements as evidence that receptive fields are larger in MST than MT. MT responses modulated strongly to the rotating-wedge stimulus but were weak or absent to stimuli presented at least 10° into the ipsilateral visual field, implying relatively small receptive fields. MST responded to the ipsilateral stimulus but exhibited weak or absent response modulations to the retinotopy stimulus, implying larger receptive fields.

Our measurements of ipsilateral responses are in general agreement with those of Tootell et al. (1998), who measured activity in human visual cortex to ipsilateral stimulation and reported ipsilateral responses in a broad region of extrastriate visual cortex including MT+. However, Tootell et al. (1998) excluded only the central 0.5° of the visual field, in contrast to our exclusion of the central 10° of the visual field. Given that monkey MT neurons representing the fovea have receptive fields several degrees in diameter and that many of them cross into the ipsilateral visual field, it is likely that their stimuli, unlike ours, would have evoked activity in both monkey MT and MST. Thus, our results are consistent with those of Tootell et al. (1998), but our use of farther-peripheral stimuli permit more reliable inferences concerning the relative receptive field sizes within MT+.

Our results are also in agreement with those of Dukelow et al. (2001), who reported ipsilateral responses in the anterior portion of MT+. Our results extend their observations by demonstrating that the more posterior region in MT+ (which did not exhibit ipsilateral responses) often exhibits clear retinotopic maps. In addition, Dukelow et al. (2001) only presented data from the right hemispheres of subjects (because of their use of a specialized head coil). We acquired fMRI responses from both hemispheres and confirmed that the organization of human MT and MST is similar in the two hemispheres.

Inferences about retinotopic organization

Our results suggest that a subregion of MT+ exhibits retinotopic organization that can be assessed using fMRI. We observed a strong response to the rotating-wedge retinotopy stimulus in all hemispheres. Furthermore, in five hemispheres we observed a smooth map of the visual field, with the LVM represented more ventral and/or posterior and the UVM represented more dorsal and/or anterior.

We believe that the lack of a clear retinotopic map in some hemispheres may reflect methodological failures. The cortical surface in the vicinity of MT+ is one of the most highly gyrified areas of the human cortex (Zilles et al., 1989; Dumoulin et al., 2000). Consequently, fMRI measurements in this region may sometimes be blurred across opposite sides of a sulcus, mixing responses from the posterior bank (putative MT) and the anterior bank (putative MST). Precise segmentation and unfolding of the gray matter in this convoluted part of the cortex is difficult, which could introduce additional errors. In particular, we noted that manually refining the gray-matter segmentation in this region often yielded clearer retinotopic maps (see Materials and Methods), suggesting that at least some of the variability in our data could result from these technical difficulties. We also note that the retinotopically organized area that we have identified as MT is approximately an order of magnitude smaller than V1 (∼240 vs ∼2400 mm²), and corresponded to only ∼5 × 12 fMRI voxels. In light of the relatively small size of this area, the ability to discern any retinotopic organization is notable.

In addition to these technical issues, electrophysiological measurements in the macaque have revealed that the MT retinotopic map is not as orderly as in earlier visual areas like V1. Maunsell and Van Essen (1987) observed that: (1) the lower visual field was often over-represented in comparison with the upper visual field; (2) the angular component exhibited local discontinuities; (3) the same angular position could be represented in multiple distinct regions; and (4) retinotopic maps were variable across hemispheres. Given this variability and local representational disarray observed in macaque MT, one might not expect to observe an orderly retinotopic map in every human MT.

Maps of visual field eccentricity in the macaque yield a clearer picture, with more lateral/ventral neurons exhibiting central receptive fields and more medial/dorsal neurons exhibiting more peripheral receptive fields. We observed a similar organization in the responses to our central versus peripheral stimuli, with the response to central stimulation often lying at the ventral edge of MT or lying ventral to a region exhibiting a clear peripheral response.

The maps of visual field angle that we observed were consistent and reproducible within subjects. We observed similar retinotopic subregions in subjects A.R.W., A.C.H., and R.F.D. in scanning sessions performed >1 year before the data reported in this paper (Dougherty et al., 1999). The locations, orientations, sizes, and shapes of the retinotopic regions were similar within subjects across the two data sets, and in the hemispheres in which the retinotopic map was most orderly in both of the data sets (A.R.W., right and left), the precise organization of the angular map was also found to be in close correspondence. In these preliminary sessions, we used a conventional retinotopic mapping stimulus, a flickering checkerboard wedge, instead of the motion-defined wedge used in this study. The similarity of the maps we observed when using such different stimuli also demonstrates the robustness of the MT retinotopic maps we identified.

Possible homologies between human and macaque motion-sensitive areas

Our results are consistent with the existence of areas MT and MST in the human that are homologous to those in the macaque. This homology is supported by three main observations. First, macaque MT and MST are adjacent to one another, with MST lying anterior to MT on the opposite bank of the dorsal STS. We found human MT and MST to be immediately adjacent in seven of the eight hemispheres in which we confidently identified both areas. MT and MST typically lay on opposite sides of the same sulcus (the ITS), with MST anterior and/or dorsal to MT. Second, macaque MT exhibits a clearer retinotopic organization than macaque MST. In human MT, we observed clear maps of the angular component of the visual in field in 5 of 10 hemispheres, and we observed evidence for a distinction between central and peripheral parts of the visual field in 7 of 10 hemispheres. Third, neurons in macaque MST have larger receptive fields than corresponding neurons in MT representing the same eccentricity. We inferred that human MST has larger receptive fields than MT, based on responses to the rotating wedge (retinotopy) and ipsilateral stimuli.

Although our data are consistent with a proposed homology between human and macaque MT and MST, they are not conclusive. There are four adjacent, motion-sensitive areas in macaque STS (MT, MSTl, MSTd, and FST), whereas our measurements discerned only two areas within human MT+. We chose to concentrate on distinguishing two regions based on differences in retinotopy and receptive field size, because electrophysiologists often use receptive field size as a rule of thumb to distinguish macaque MT and MST. In addition, fMRI measurements of retinotopic organization are well established as a technique for subdividing larger regions of visual cortex (Engel et al., 1994, 1997;Sereno et al., 1995; DeYoe et al., 1996), including dorsal motion-responsive regions V3A, V3B, and V7 (Tootell et al., 1997; Smith et al., 1998; Press et al., 2001; Tootell and Hadjikhani, 2001).

Additional measurements of function within the macaque STS and the human ITS will shed further light on the proposed homologies between the monkey and human motion-sensitive areas. For example, most neurons in macaque MT respond only according to the local direction of translation of a moving stimulus (Dubner and Zeki, 1971; Maunsell and Van Essen, 1983; Albright, 1984), whereas many neurons in macaque MST also exhibit selectivity for particular components of optic flow (e.g., expansion/contraction or rotation) (Saito et al., 1986; Duffy and Wurtz, 1991b). In a human fMRI experiment, Morrone et al. (2000)compared MT+ responses to translation, expansion/contraction, and rotation. They observed stronger responses to translation in a dorsal and/or posterior portion of MT+ and stronger responses to expansion/contraction and rotation in a ventral and/or anterior part (although in two subjects this layout was reversed). However, their results do not appear to clearly align with the areas that we have identified as MT and MST.

In addition, neurons in macaque MST, but not in macaque MT, receive extraretinal eye movement signals, so that some MST neurons respond during smooth-pursuit eye movements in the absence of retinal motion (Newsome et al., 1988). Dukelow et al. (2001) reported activity in the most anterior portion of human MT+ when subjects performed “nonvisual” pursuit of a self-generated somatosensory target (their own finger moving back and forth) in darkness, and suggested that this region corresponded to the human homolog of MSTl. However, it is not known whether self-guided pursuit is mediated by the same cortical mechanisms that control normal pursuit eye movements.

The visual response properties of neurons in macaque FST have not been well studied. FST neurons are typically thought to exhibit weak and erratic visual responses (Komatsu and Wurtz, 1988a), although one study did observe direction-selective responses in some FST neurons (Erickson and Dow, 1989). If a human homolog of FST exists, it is unclear as to whether this region would respond strongly enough to visual motion to fall within our original definition of MT+. fMRI measurements in monkeys (Dubowitz et al., 1998; Stefanacci et al., 1998; Disbrow et al., 1999; Logothetis et al., 1999, 2001) could provide a standard against which to evaluate the measurements from human brains to further test the proposed homologies.

Another possibility, of course, is that not all of the macaque motion-sensitive areas are preserved in humans. For example, in the owl monkey, homologs to macaque MST and FST have not been unambiguously defined (Rosa et al., 1993). Regardless, because the subregions we identified exhibit differences in retinotopic organization and receptive field sizes, it would be prudent to analyze data from future fMRI experiments separately for each of these subregions rather than treating them as one larger area.