OUP user menu

Variability and asymmetry in the human precentral motor system
A cytoarchitectonic and myeloarchitectonic brain mapping study

J. Rademacher, U. Bürgel, S. Geyer, T. Schormann, A. Schleicher, H.-J. Freund, K. Zilles
DOI: http://dx.doi.org/10.1093/brain/124.11.2232 2232-2258 First published online: 1 November 2001


The morphology of the region of the primary motor cortex in the human brain is variable, and putative asymmetries between the hemispheres have been noted since the beginning of last century. Such variability may confound the results of clinical lesion or functional activation studies. We measured Brodmann area (BA) 4 and the identifiable precentral component of the pyramidal tract (PRPT) in 11 human post-mortem brains using techniques of quantitative cytoarchitectonic and myeloarchitectonic image analysis. Topography and variability in the localization of architectonic borders were analysed and mapped to a computerized spatial reference system, which consists of an individual in vivoMRI brain. All maps were superimposed to produce probabilistic maps of BA 4 and PRPT which can be co-registered with any image of brain structure or function that has also been transformed to Talairach coordinates. These maps can be readily applied to future brain mapping studies. We observed a considerable degree of variability between hemispheres (intra-individual) and between brains (inter-individual). The variation zones of BA 4 and PRPT differ from the templates of the Talairach atlas. Voxel-based morphometry shows significant side differences with larger volumes of PRPT in the left hemisphere than in the right hemisphere. This larger volume of the descending cortical motor fibres may be related to the known left-hemisphere dominance for handedness in >90% of the population. In contrast, BA 4 was symmetrically organized. The lack of a significant correlation between the size of BA 4 and the size of PRPT may relate to the fact that additional non-primary motor and sensory cortices contribute to the origins and size of the pyramidal tract proper.

  • human cerebral cortex
  • motor cortex
  • pyramidal tract
  • image analysis
  • brain atlas
  • AC = anterior commissure
  • BA = Brodmann area
  • GLI = grey level index
  • PC = posterior commissure
  • PRG = precentral gyrus
  • PRPT = precentral portion of the pyramidal tract
  • PT = pyramidal tract


The aim of the work reported here was to map and measure the extent of the cytoarchitectonically defined `primary motor cortex' [Brodmann area (BA) 4] and the myeloarchitectonically defined precentral portion of the pyramidal tract (PRPT), as well as their relationship to each other in the same series of post-mortem human brains. Application of the myeloarchitectonic method was confined to the precentral component of pyramidal tract (PT) proper, i.e. PRPT originating in the precentral gyrus (PRG), because reliable analysis of the other non-precentral components of PT proper cannot be achieved in human post-mortem brains even with the presented `modified' myeloarchitectonic technique. We applied modern architectonic image analysis and brain mapping techniques to the study of anatomical variations rather than standard qualitative methods or classical textbook descriptions that have been used in previous studies.

BA 4 and its main descending pathway, i.e. PT, represent important parts of the primary motor system (Foerster, 1936; Penfield and Boldrey, 1937). Cytoarchitectonically defined BA 4 is located in PRG laterally and in the paracentral lobule medially. A large portion of BA 4 occupies the posterior wall of PRG in the depth of the central sulcus (Brodmann, 1903, 1909). From there, the descending axons of PT pass through the corona radiata and the posterior limb of the internal capsule towards the spinal cord (Ross, 1980). BA 4 has also been identified as area 42 of Vogt (1910) and Sanides (1962), area FAγ (Economo and Koskinas, 1925) and the frontal ganglionic core (Braak, 1980). It is widely accepted that BA 4 is the anatomical correlate of functionally defined primary motor cortex. In clinical practice, primary motor cortex has been traditionally considered an executive locus for simple voluntary movements which sends via PT commands to individual muscles or even motor neurones (Fetz and Finocchio, 1972). Recent results from animal experiments and brain mapping studies in humans, however, suggest that BA 4 participates in processes of greater complexity such as complex finger movement sequences (Gerloff et al., 1998), spatiotemporal patterning of muscle activity (Kakei et al., 1999), movement planning (Alexander and Crutcher, 1990), manual skill learning (Karni et al., 1998) and mental rotation of objects (Ganis et al., 2000).

Knowledge about the anatomical topography and variability of BA 4 and PT is becoming increasingly important, because modern neuroimaging techniques such as functional MRI provide the means to analyse the organization of human motor functions with high spatial resolution (Kim et al., 1993; Wildgruber et al., 1996; Dassonville et al., 1997; Lee et al., 1998; Stephan et al., 1999; Toni et al., 1999). For analysis, the resulting foci of activation are usually related to a standard stereotaxic system (Talairach and Tournoux, 1988) or to visible macroanatomical landmarks such as the central sulcus or the internal capsule. Even the most advanced imaging protocols, however, do not permit direct visualization of the laminar heterogeneity which defines the cytoarchitectonic pattern of BA 4. It is the implicit assumption of many brain-mapping studies that the macroanatomical gyral and sulcal landmarks coincide with the borders of architectonic areas. That this hypothesis is not tenable in general has been shown previously (Rademacher et al., 1993; Geyer et al., 1996; White et al., 1997a, b).

With regard to cerebral cortex, it is known that the presence of striking topographical variations may obscure and divert structural–functional relationships, if the range of individual macroanatomical (Cunningham, 1892; Steinmetz et al., 1989; Ono et al., 1990; Thompson et al., 1996) and microanatomical patterns (Rademacher et al., 1993; Amunts et al., 1999) is neglected. Volumetric variations (of up to a factor of 10) exist even at the early hierarchical level of the primary cortices (Amunts et al., 2000). Thus, variability in size and geometry of the motor cortex may lead to relevant structural–functional mismatching in the order of centimetres. The classical cytoarchitectonic reports and templates have mostly neglected anatomical variability and they lack stereotaxic data. With respect to the description of location, shape and size of BA 4, there are considerable discrepancies in the classical literature (Zilles, 1990). Taken together, these limitations could explain, in part, why there is still debate about both precise localization and asymmetries of human motor responses (Kim et al., 1993; Netz et al., 1995; Iacoboni et al., 1997; Triggs et al., 1997; Leocani et al., 2000).

With regard to fibre tract topography, anatomical evidence on topographical variations of PT is scarce (Nyberg-Hansen and Rinvik, 1963; Kertesz and Geschwind, 1971; Nathan et al., 1990). Studies of fibre tracts in nonhuman primates use invasive tracer injections that cannot be performed in humans. Some information on human fibre tracts can be obtained by imaging fibre myelination or degeneration patterns but only in pathologic cases. Most importantly, these approaches do not permit the systematic and quantitative 3D mapping of PT for its total extent (Barkovich et al., 1988; Kuhn et al., 1989; Fries et al., 1993). Advances in the identification of PT with diffusion-weighted MRI techniques (Virta et al., 1999; Karibe et al., 2000) hold promise for future brain mapping studies, but they are still not readily available to most clinical centres. As such, new neuroimaging techniques are introduced, but have not yet been validated themselves; a re-evaluation of anatomical measures is needed.

The first goal of this study was to acquire anatomical data that will help to clarify these issues related to brain mapping. To do so, we generated probability maps, which contain superimpositions of cytoarchitectonic BA 4 and myeloarchitectonic PRPT fibre topography in a common reference system. In a first step, we applied recently developed cytoarchitectonic and myeloarchitectonic techniques that permit the reliable identification and delineation of BA 4 and PRPT to a series of 11 normal adult post-mortem brains. For cytoarchitectonic analysis of BA 4, we used a modified silver method which stains cell bodies and delineated the areal borders of BA 4 with an observer-independent procedure (Schleicher et al., 1999, 2000). Although antero- or retrograde transport studies cannot be applied in the human, myeloarchitectonic analysis of PRPT by using a modified staining procedure for myelin provides a clear delineation of single fibre tracts in the white matter (Bürgel et al., 1997, 1999). In a second step, we used a 3D probabilistic mapping strategy (Schormann and Zilles, 1998), which is based on the standard reference brain of the European Computerized Human Brain Database, to define the degree and direction of cyto- and myeloarchitectonic variations in stereotaxic space (Roland et al., 1994; Zilles et al., 1995). BA 4 and PRPT from each brain were individually warped to the 3D stereotaxic reference system (Schormann and Zilles, 1998). Probability maps which reflect the degree of variability of both structures across the sample of brains were generated. In these maps, the degree of overlap in each stereotaxic position was quantified and transferred to Talairach coordinates for better comparability with functional or anatomical data from other studies.

The second goal of this study was to examine possible interhemispheric asymmetries of BA 4 and/or PRPT. Comparative primate data for BA 4 from the squirrel monkey have demonstrated that in the dominant hemisphere (i.e. the hemisphere opposite the preferred hand), the distal forelimb representations were larger than in the non-dominant hemisphere (Nudo et al., 1992) showing that movement representations in BA 4 are use-dependent (Nudo et al., 1996). This close correspondence between behavioural performance and electrophysiologically defined motor representations may also hold true for the human brain. Since side differences of functional activation in human motor cortex appear to reflect direction and degree of handedness (Dassonville et al., 1997; Volkmann et al., 1998), one may speculate that an anatomical asymmetry is the underlying morphological substrate. The question whether there is a correlation between handedness or other functional measures of hemispheric lateralization and anatomical asymmetries is not only of fundamental importance for our understanding of the human motor system (Darian-Smith et al., 1999) but it also relates to a more general model of cortical organization, which assumes that superior performance is based on greater allocation of neural circuitry related to a distinct function (Geschwind and Galaburda, 1985). The greater amount of cortical tissue in BA 4 devoted to the hand and fingers compared with the leg and foot (Penfield and Rasmussen, 1950) appears to be in support of this hypothesis. BA 4 is a perfect candidate to test this hypothesis given that handedness is the clearest example of behavioural lateralization. Until today, only a few studies have examined interhemispheric differences in BA 4 reporting contradictory results (Beck, 1950; Rademacher et al., 1993; Amunts et al., 1996; White et al., 1994, 1997a, White et al., b).

The aims of the present study were to (i) analyse whether there is an association between the precise location of architectonic borders and macroanatomical landmarks, (ii) map BA 4 and PRPT to a stereotaxic reference system in order to provide probabilistic maps which can be applied to brain mapping studies and (iii) analyse intersubject variability and interhemispheric (a)symmetry in the positions and volumes of BA 4 and PRPT.

Material and methods

The cytoarchitectonic mapping of BA 4, myeloarchitectonic mapping of PRPT, spatial normalization of all structural maps, and generation of probability maps are summarized as a flowchart in Table 1.

View this table:
  Table 1


Eleven human post-mortem brains (five female and six male brains; age range 37–85 years) were obtained from the body donor program of the Anatomical Institute of the University of Düsseldorf in accordance with the legal requirements. There was no history of chronic neurological or psychiatric disease. An inherent limitation in post-mortem studies is the lack of information about handedness. On the basis of a 90% incidence of right-handedness in the population, we assumed that most of our cases will have been right-handed (Gilbert and Wysocki, 1992).

MRI and histological processing of the brains

The histological processing and the cytoarchitectonic and myeloarchitectonic analysis have been described in detail in previous articles (Geyer et al., 1996, 1999, 2000; Bürgel et al., 1997, 1999; Schormann and Zilles, 1998; Amunts et al., 1999, 2000). All brains were fixed in 4% formalin or in Bodian's solution for several months. In order to compensate for histological deformations (i.e. shrinkage and cutting artefacts), an MRI scan was acquired of each brain prior to further histological analysis. In brief, the immersion-fixed undissected brains were scanned (Siemens Magnetom 1.5 T) with a T1-weighted 3D FLASH (fast low angle shot) gradient echo sequence which had a resulting voxel size of 1 × 1 × 1.17 mm, providing sufficient resolution for 3D computer reformations. For all MRI examinations, the technical factors were 40° flip angle, 40-ms repetition time, 5-ms echo time, 1 excitation, 25-cm field of view, 15 cm thickness of the excited volume, 128 partitions, and a 256 × 256-image matrix. After MRI, all brains were dehydrated in graded alcohols, embedded in paraffin and serially sectioned (20 μm thick coronal whole brain sections). The brains were embedded in a standard canonical way, ensuring a coronal plane of sectioning which approximates Talairach's coronal axis (Talairach and Tournoux, 1988). Obviously, the cortical ribbon is not always cut perpendicularly to the course of PRG on whole brain coronal sections. Those cortical regions which did not show all laminae of the cortex by visual inspection (i.e. tangentially cut surfaces) were excluded from further analysis. Images of the paraffin blockface were obtained after every 60th section with a CCD (charged coupled device) digital camera. Every 60th section was mounted on a gelatin-coated slide and stained for cell bodies with a modified silver method (Merker, 1983). Adjacent sections were mounted and stained for myelin with a modified Heidenhain–Woelcke technique (Bürgel et al., 1997). The mounted histological sections were then digitized with a CCD camera for the subsequent 3D reconstruction of the histological volume (see below).

Definition of macroanatomical landmarks

Several landmarks have been proposed for identifying PRG directly (Kido et al., 1980; Freund and Hummelsheim, 1985; Sastre-Janer et al., 1998) and cytoarchitectonic BA 4 indirectly (Rademacher et al., 1993; Amunts et al., 1996; White et al., 1997a). PRG lies between the precentral and central sulci where it takes an oblique course from the interhemispheric fissure medially to the edge of the sylvian fissure laterally. On the medial hemispheric surface, the paracentral lobule extends from the ascending and descending paracentral sulci anteriorly to the terminal up-swing of the cingulate sulcus posteriorly. Although these gyral and sulcal structures are present in the whole population (Ono et al., 1990), locating PRG in MRI may be unreliable (Sobel et al., 1993). Regarding PT topography, the borders of the internal capsule are the most reliable macroanatomical landmarks for defining its location (Ebeling and Reulen, 1992). Our analysis included the proximal fibre extent of PRPT down to the diencephalon. Crus cerebri, pons and medulla were not part of the present study. The classical stereotaxic approach of Talairach and Tournoux (Talairach and Tournoux, 1988) to map BA 4 and PT is shown in Fig. 1.

Fig. 1

Standard stereotaxic localization of the human primary motor system (modified from Talairach and Tournoux, 1988). The upper row shows images of the lateral (left) and medial (right) brain convexity in Talairach space. Solid black lines indicate the defining Talairach planes passing through the anterior and posterior commissures either horizontally (AC–PC) or vertically (VAC and VPC). BA 4 (black areas) is shown on the precentral gyrus laterally and on the paracentral lobule medially. Broken lines localize the origins of one coronal (left; y = –20) and one axial (right; z = 45) brain slice, which are shown in the lower row. There, the topography of PRPT (dotted areas) is visualized from its dorsal origin in the precentral gyrus down to cerebral peduncles ventrally. The superimposed grids represent stereotaxic Talairach space.

Cytoarchitectonic delineation of BA 4

The present study is based on the quantitative cytoarchitectonic analysis of BA 4 which has been described in detail in a previous article (Geyer et al., 1996). Only major points will be summarized here. Every 60th histological section was stained for cell bodies with a modified silver method (Merker, 1983), which yields high contrast between cells and neuropil for cytoarchitectonic analysis and observer-independent delineation of BA 4 in both hemispheres. Because of insufficient staining, one brain had to be excluded from cytoarchitectonic analysis (Brain 1; Tables 2, 3 and 7). BA 4 is characterized by an absent layer IV, giant pyramidal cells in layer V, and low cell packing density across all layers without pronounced laminar organization (Brodmann, 1909; Economo and Koskinas, 1925; Sanides, 1962; Zilles, 1990; Rademacher et al., 1993; Amunts et al., 1996; Geyer et al., 1996; White et al., 1997a). At the anterior border of BA 4, scattered or absent giant pyramidal cells and large, slender pyramidal neurones in layer III characterize BA 6 (Vogt's area 6aα laterally and SMA medially). The cytoarchitectonic differences at the caudal border between primary motor cortex (i.e. BA 4) and primary somatosensory cortex (i.e. BA 3a) are: a wide band of grey matter, low cell density, a pronounced columnar arrangement of the cells, and a blurred border between grey and white matter in BA 4 versus a narrow band of grey matter, high cell density, an incipient layer IV, and a sharp border between grey and white matter in BA 3a. The anterior border of BA 4 (towards area 6aα; Vogt, 1910) and its posterior border (towards primary somatosensory BA 3a) were mapped with an observer-independent cytoarchitectonic technique (Schleicher et al., 1999, 2000) which detects statistically significant changes in the cytoarchitectonic pattern. This method has already been applied successfully to the analysis of the following cortical regions: primary motor and somatosensory cortices (Geyer et al., 1996, 1999, 2000), Broca's region (Amunts et al., 1999), primary and secondary visual cortices (Amunts et al., 2000), and primary auditory cortex (Rademacher et al., 2001). The level of distortion that is tolerated by the observer-independent cytoarchitectonic grey level index (GLI) method has recently been analysed by means of a cortical model which simulates cortical layering (Schleicher et al., 1999). GLI, as an estimate of volume density, is the only stereological parameter which does not depend on the orientation of the plane of sectioning relative to the orientation of microstructural elements of the cortex (i.e. cells or cell columns). Therefore, GLI differs from the estimation of numerical densities (number of objects per volume unit of reference space) or surface density, which, for anisotropic structures like the neocortex, requires a well-designed strategy for sectioning and sampling. Nevertheless, cell shape and layer width are affected by cortical folding (Economo and Koskinas, 1925). Schleicher and colleagues have observed that cortical folding induces gradual changes in GLI profile shapes (Schleicher et al., 1999). By analysing their cortical model, they demonstrated that the GLI method is characterized by a pronounced sensitivity to abrupt changes in the cortical laminar pattern and an insensitivity to the aforementioned gradual changes in cytoarchitecture. In general, the results obtained in the standard coronal plane could be fully confirmed by analysis of sagittal sections through the regions of interest. Therefore, the GLI method offers a new level of reproducibility and observer independence.

View this table:
Table 2

Stereotaxic coordinates of the maximum extent of BA 4 and PRPT in the three standard planes of each brain

Brain no.Sagittal plane (x-axis)Coronal plane (y-axis)Axial plane (z-axis
Coordinates x, y and z as in Talairach and Tournoux (1988). Dashes indicate missing values because of insufficient staining.
BA 4
2–57–2620 1–42–1–3720692169
3–50–4490–5–39–4–44–1065 –962
4–48–3533–5–394–47 –364 –564
7–4914948–455–33 –359 067
View this table:
Table 3

Stereotaxic coordinates of the maximum extent of BA 4 in the paracentral lobule of each brain

Brain no.Coronal plane (y-axis)Axial plane (z-axis
The coordinates of BA 4 in the sagittal plane are not shown. The sagittal extent of BA 4 is equivalent to the cortical depth at the interhemispheric fissure. Dashes indicate missing values because of insufficient staining.

Myeloarchitectonic delineation of PRPT

PRPT was identified by myeloarchitectonic criteria: (i) fibre staining and (ii) fibre origin and direction. For increased contrast and optimal visualization of PRPT fibres, sections adjacent to those stained for cell bodies were mounted on glass slides and stained with a modified Heidenhain–Woelcke technique for myelin (Bürgel et al., 1997). This procedure enhances the staining contrast between heavily and less-myelinated fibre tracts, thereby permitting the identification of PRPT at microscopic resolution. Heavily myelinated fibre tracts can then be distinguished from the neighbouring less densely stained white matter. The method is restricted to fibre tracts with a differential myelin staining compared with the surrounding white matter. The identification of a circumscribed region in the white matter as an anatomically distinct part of PRPT was achieved by analysing its histological texture and its localization. The myelin sheaths of PRPT appear intensely stained compared with less densely stained surrounding regions and poorly myelinated association pathways. The spatial orientation of PRPT fibres can be recognized along the whole dorsal-to-ventral extent. Dorsally, their origin was identified along the cortex of PRG. Ventrally, the rostral tip of the cerebral peduncles was taken as the inferior border. Based on these criteria, the borders of PRPT were defined in each myelin-stained section. The observer-dependent intra- and inter-rater variability with respect to the tracing of the borders did not exceed ±1 mm for each direction (Bürgel et al., 1999). The quality of the differential labelling of axons obtained using this myeloarchitectonic method has been demonstrated in previous studies by comparison with other myeloarchitectonic techniques (Bürgel et al., 1997, 1999). Because of insufficient staining, one brain had to be excluded from myeloarchitectonic analysis (Brain 11; Tables 2 and 7).

Reconstruction of the histological volume

The histological volume of each brain was then reconstructed in 3D from (i) the images of the paraffin blockface, (ii) the digitized histological sections, and (iii) the MRI volume of the same brain with linear and non-linear transformations (Schormann et al., 1993, 1995; Schormann and Zilles, 1997, 1998). Since the MRI volumes were obtained after fixation, but prior to further histological processing and then spatially normalized to the reference brain of a computerized brain atlas (see below), histological artefacts (e.g. shrinkage of the brain due to postfixation histological processing or compression of the sections due to cutting) could be minimized by this procedure which has been applied already to a series of cyto- and myeloarchitectonic studies (Amunts et al., 1996, 1999, 2000; Geyer et al., 1996, 1999, 2000; Bürgel et al., 1999; Rademacher et al., 2001). This issue deserves attention, since posthumous histological processing of the brain can affect and distort the brain tissues. Uncontrolled histological artefacts have often limited the reliability of architectonic studies (Uylings et al., 1986). Without correction for shrinkage, this will cause serious problems when groups are compared in which the tissue has been shrunken differently. For example, it has been shown for the rat frontal cortex that volumetric shrinkage may vary (between 54 and 67%) as an effect of ontogenetic age in the developing brain (van Eden and Uylings, 1985). For practical reasons, the effects of the initial fixation procedure (i.e. before MRI) have not been considered. Smaller shrinkage effects but also swelling effects have been discussed in the literature. Related issues and other standard techniques for the correction of shrinkage have been reported by Uylings and colleagues (Uylings et al., 1986).

The statistically significant borders of BA 4 were marked on the cell-stained sections as were the borders of PRPT on the sections stained for myelin. An image analysis software package with an interactive voxel-painting program (KS 400; Zeiss, Germany) was used to label the extent of BA 4 and PRPT in corresponding sections of the 3D reconstructed histological volume. With this approach, microstructurally defined representations of BA 4 and PRPT were obtained in the 3D reconstructed histological volume of each brain.

Spatial normalization and probabilistic mapping of BA 4 and PRPT

Each reconstructed histological volume (with the volume representations of BA 4 (n = 10) and PRPT (n = 10) was then spatially normalized to the reference brain of the computerized atlas (Roland et al., 1994) with a new algorithm based on an extended principal axes theory (Schormann and Zilles, 1997) and a fast automated multi-resolution fullmulti-grid movement model (Schormann and Zilles, 1998). To evaluate the positional variability of BA 4 and PRPT, probability maps were created by superimposing BA 4 and PRPT of all normalized brains in 3D space. For each voxel of these maps, a grey value codes for how many brains have a representation of BA 4 or PRPT in this particular voxel.

Stereotaxic position of BA 4 and PRPT

The extent of BA 4 and PRPT in the sagittal, coronal and horizontal planes was calculated in Talairach coordinates. Our topographical data can therefore be compared with the templates of the Talairach atlas (Talairach and Tournoux, 1988) and the results of other studies of structure or function. In Talairach space, which has the anterior and posterior commissures (AC and PC, respectively) as points of origin, the x-axis defines the mediolateral direction, with negative values indicating the left hemisphere. The y-axis defines the anterior–posterior direction, with positive values anterior to the orthogonal plane through AC and negative values posterior to this landmark. The z-axis is oriented parallel to the AC–PC line passing through both commissures, with positive values dorsal to it and negative values ventral to it. The maximal areal extent in stereotaxic space was outlined separately for both hemispheres by defining the bounding box (Penhune et al., 1996) which contains the total volume of BA 4 or PRPT from all 10 brains.

Volumetric measurements of BA 4 and PRPT

The 3D data of the histological volume of each brain (before transformation to the 3D reference brain; see above) were used to calculate the volumes of BA 4 and PRPT according to the formula: volume of the structure (V) = N × A, where N is the number of voxels attributed to BA 4 or PRPT; and A is the volume of one voxel in the histological volume (mm3) (image analysis software KS 400; Zeiss, Germany). Interhemispheric asymmetry of BA 4 or PRPT was defined as any side difference which is >10% and its degree was determined as the asymmetry index of the region-of-interest, δROI = {δVRVL)/(0.5(VR + VL)}, with VR and VL representing each structure's volume in the right and left hemisphere, respectively. Negative values of δROI indicate leftward asymmetries and positive values indicate rightward asymmetries. BA 4 or PRPT were considered to be asymmetrical when δROI > 0.10. Volumes were compared between structures (BA 4 and PRPT) and between hemispheres (left and right) [two-way ANOVA (analysis of variance) with repeated measurements design; P = 0.05]. For analysis of the correlation between the volumes of BA 4 and PRPT, Pearson correlation coefficients were calculated separately for each hemisphere. Left–right comparisons of the volumes of BA 4 and PRPT were performed by a paired (left versus right hemisphere) t-test (P = 0.05).


Topography of BA 4

In contrast to the high variability of gyral and sulcal patterns in other primary cortices (Rademacher et al., 1993; Penhune et al., 1996; Amunts et al., 2000), the gyral and sulcal landmarks of BA 4, i.e. PRG and the central sulcus, showed more regular patterns. As in previous studies, we distinguished two classes of architectonic variations, those which can be predicted from macroanatomical landmarks (Class I) and those which cannot (Class II) (Rademacher et al., 1993). Class I variability includes the finding that BA 4 always occupies the posterior bank of PRG in the central sulcus including the region of the `hand knob' (Yousry et al., 1997). Its presence on the exposed surface of the lateral convexity is restricted to the dorsal portion of PRG. Precentral and central sulci represented maximum anterior and posterior borders of BA 4. Medially, BA 4 is located in the central portion of the paracentral lobule, anterior to the termination of the central sulcus. Class II variability relates to the finding that the precise cytoarchitectonic borders of BA 4 differed between hemispheres and did not coincide with the sulcal fundi. The extent of BA 4 on the anterior lip of the central sulcus showed individual variations of up to 1 cm. At the paracentral lobule, the borders of BA 4 had no limiting sulcal landmarks.

Topography of PRPT

PRPT, as defined by myeloarchitectonic criteria, was characterized as an intensely stained large fibre tract, which originates from PRG and takes a longitudinal course in the ventral direction. The myeloarchitectonic method did not permit the identification of other non-primary premotor and parietal contributions to PRPT. It therefore provides a simplified concept of PRPT, which is classically defined as all fibres which course longitudinally in the pyramid of the medulla oblongata, regardless of their site of origin. This caveat also holds true for previous morphological studies and the Talairach atlas. The course of PRPT through the hemispheric white matter, as identified in the present study, was defined by the topographical relationship between PRG and the internal capsule. PRPT fibres converge within the corona radiata where they show a characteristic fan-like arrangement, which opens against the precentral cortex and closes in proximity to the internal capsule (Ebeling and Reulen, 1992). The medial edge of the fan is bordered by the body of the corpus callosum and the lateral ventricle. The lateral border of PRPT is represented by the insula, putamen and globus pallidus. PRPT was always localized in the posterior limb of the internal capsule (Class I variability), but its precise position varied between horizontal planes and between individuals (Class II variability). The intermingling of PRPT and callosal fibres at the medial edge of PRPT was analysed at high-power magnification, which provided the means to distinguish fibre identity by analysis of fibre orientation.

Probability maps of BA 4 and PRPT

Figures 2 and 3 exemplify the two sorts of topographical data, which have been acquired with our architectonic brain mapping method. The left-hand columns of Figs 2 and 3 show sagittal, coronal and axial sections through the reconstructed histological volume of one brain. The extent of BA 4 in this brain has been shown in Fig. 2(A–C) and the extent of PRPT in the same brain in Fig. 3(A–C). In contrast, the right-hand columns of Figs 2 and 3 show sections through the reference brain. The probability map of BA 4 has been superimposed in Fig. 2(D–F) and the probability map of PRPT in Fig. 3(D–F). Both maps are based on data from five brains and describe, for each structure and each voxel, how many brains have a representation of this structure in this particular voxel. This representation is coded in grey values from black (one brain) to white (five brains). High variability is represented by a low degree of overlap (20% = one brain) at distinct spatial locations. Figures 4 (BA 4) and 5 (PRPT) represent a stereotaxic atlas (based on 10 brains for each structure) which can be readily applied to the localization of BA 4 and PRPT by other study groups.

Fig. 2

The left column shows images of one brain of our sample after reslicing of the reconstructed histological volume in the (A) sagittal (x = +33), (B) coronal (y = –28) and (C) axial (z = +54) Talairach planes with superimposed BA 4 as shown in white. The right column shows images of the MRI volume of the reference brain in the same (D) sagittal, (E) coronal and (F) axial planes onto which the grey value-coded probability map of BA 4 has been superimposed. The grey value code indicates the absolute frequency of voxels containing BA 4 from 1 (black) to 5 (white) in individual brains warped to the format of the reference brain.

Fig. 3

The left column shows images of one brain after reslicing of the reconstructed histological volume in the (A) sagittal (x = –28), (B) coronal (y = –18) and (C) axial (z = +28) Talairach planes with superimposed PRPT in white. The right column shows images of the MRI volume of the reference brain in the same (D) sagittal, (E) coronal and (F) axial planes onto which the grey value-coded probability map of PRPT has been superimposed. The grey value code indicates the absolute frequency of voxels containing PRPT from 1 (black) to 5 (white) in individual brains warped to the format of the reference brain.

Fig. 4

Probability maps of BA 4 at 20% steps in eight coronal slices overlaid on a grid in standardized space (z separation = 4 mm; range: z = –4 to –32). In all slices, the outer contour (black) represents 20% overlap. The left hemisphere is on the left side. (A) Coronal slices from y = –4 to –16; (B) Coronal slices from y = –20 to –32 (grid scale in millimetres; modified from Talairach and Tournoux, 1988).

One important aspect of the presented probability maps needs to be clarified. The topographical relationship between the central sulcus and the probability map of BA 4, for example, relates to the macroanatomy of our computerized atlas brain. This reference brain is the in vivo MRI scan of one individual human brain. Thus, although the central sulcus represents the maximum posterior border of BA 4, in Fig. 4 it is shown going beyond this boundary, extending into postcentral areas and into white matter. The main reason for this discrepancy lies in the co-registration with a single reference brain. This is a critical feature of any probabilistic mapping technique. While probabilistic contour maps can be optimally used to localize a region of interest within BA 4 for a given range of probability, they do not represent the ideal tool to study the relationship between individual gyral/sulcal patterns and cytoarchitectonic topography (Rademacher et al., 2001). In the present study the reference brain is represented by the reference brain of the Human Brain Atlas (Roland et al., 1994) and the sulci depicted in Figs 4 and 5 are the individual sulci of this brain. It has been selected as the anatomically least deviating (in brain shape) specimen from a larger sample of brains (Roland et al., 1994). In contrast, the probabilistic maps of BA 4 relate to the findings from 10 different brains which do not have their central sulci in the same stereotaxic position of the reference brain. Thus, the Human Brain Atlas provides Talairach coordinates which allow comparison with other atlases, but it does not permit a high-resolution study of microscopic–macroscopic relationships (Rademacher et al., 2001). Individual maps of BA 4 represent the ideal tool to visualize the latter (Rademacher et al., 1993). Alternative probabilistic mapping procedures which are based on a `mean' brain constructed by averaging numerous individual MRI brains (Collins et al., 1994) share the same limitations. In this case, blurred sulcal contours of the reference brain are the result of the variations in sulcal topography.

Fig. 5

Probability maps of PRPT at 20% steps in eight coronal slices overlaid on a grid in standardized space (z separation = 4 mm; range: z = –4 to –32). In all slices, the outer contour (black) represents 20% overlap. The left hemisphere is on the left side. (A) Coronal slices from y = –4 to –16; (B) Coronal slices from y = –20 to –32 (grid scale in millimetres; modified from Talairach and Tournoux, 1988).

Stereotaxic position of BA 4

Individual Talairach coordinates of BA 4 (bounding box) in each brain are summarized for its dorsolateral extent in PRG in Table 2 and for its medial extent in the paracentral lobule in Table 3 (dashes = missing values due to insufficient staining). For comparison, the respective coordinates of the Talairach atlas (1988) are shown in Table 4. Figure 6 shows the discrepancies between our results and the Talairach atlas for two slices in the standard coronal plane by superimposing the total extent of BA 4 (i.e. outer contour of the probability map of 10 brains) to the respective Talairach templates. The location of primary motor cortex as defined by cytoarchitectonic BA 4 differed in all planes from its location as defined in the Talairach atlas. The anterior border was shifted caudally by 5 mm on the left side and the posterior border was shifted caudally by 9 mm (right side) to 11 mm (left side). The lateral border was shifted medially by 3 mm (left side). The superior and inferior borders were located more dorsally by 6 mm (right side) to 7 mm (left side) and by 6 mm (left side) to 7 mm (right side), respectively. The maximum differences between the stereotaxic coordinates of BA 4 in individual hemispheres and the location in the Talairach atlas (i.e. the highest degree of mismatch) were 19 mm in the coronal plane, 5 mm in the sagittal plane, and 12 mm in the axial plane. At the paracentral lobule, the dorsal extent of BA 4 was identical with the values from Table 2 and the medial-to-lateral extent was defined by the depth of the paracentral cerebral cortex (not shown in Table 3). Its ventral extent was larger than shown in the atlas of Talairach and Tournoux (1988) where it is depicted from z = +60 to +65 mm. In our series, the mean values for the ventral extent were bilaterally +43 mm (SD ± 2.7 mm). Side differences for the spatial position of BA 4 at PRG were observed for its anterior–posterior extent with right-sided BA 4 being located 6 mm anteriorly, compared with left-sided BA 4. Maximum asymmetries between hemispheres ranged from 7 mm in the sagittal and 8 mm in the axial planes to 12 mm in the coronal plane. At the paracentral lobule the maximum difference between any pair of hemispheres was 6 mm in the axial plane and 18 mm in the coronal plane.

View this table:
Table 4

Stereotaxic coordinates of BA 4 from the atlas of Talairach and Tournoux (1988)

SideSagittal plane (x-axis)Coronal Plane (y-axisAxial plane (z-axis
*In the Talairach atlas, the right side of the brain is supposed to be the mirror equivalent of the left side.
BA 4 (total)
    Right* 59 20–351665
BA 4 (paracentral)
    Right* 5 20–356065
Fig. 6

Discrepancies between the probability maps and the Talairach templates for the extent of BA 4 and PRPT. (A) The left column shows images of the maximal variation zone of BA 4 as defined cytoarchitectonically in 10 brains, superimposed to the respective coronal Talairach templates (y = –8 and –28). White areas, BA 4 as shown in the Talairach atlas and overlapping with the probability map of BA 4; hatched areas, BA 4 as shown in the Talairach atlas but not overlapping with the probability map of BA 4; grey areas, zone of the probability map of BA 4 which does not overlap with BA 4 as depicted in the Talairach atlas. (B) The right column shows images of the maximal variation zone of PRPT as defined myeloarchitectonically in 10 brains, superimposed on the respective coronal Talairach templates (y = –8 and –28). White, hatched and grey areas are used by analogy to the left column.

Stereotaxic position of PRPT

Individual Talairach coordinates of PRPT (bounding box) in each brain are summarized in Table 2. Figure 6 shows the discrepancies between our results and the Talairach atlas for two slices in the standard coronal plane by superimposing the total extent of PRPT (i.e. outer contour of the probability map of 10 brains) on the respective Talairach templates. The location of PRPT as defined myeloarchitectonically differed in all planes from the location of PT as defined in the Talairach atlas (Table 6; also the Talairach atlas does not show PT proper). The anterior border was shifted rostrally by 2 mm on the right side and caudally by 3 mm on the left side. The posterior border was shifted caudally by 6 mm (right side) and by 7 mm (left side). The lateral border was shifted medially by 7 (right side) to 10 mm (left side). The inferior border was shifted ventrally by 7 mm bilaterally. The maximum differences between the stereotaxic coordinates of PRPT in the present study and PRPT in the Talairach atlas were 14 mm in the coronal plane and 15 mm in the sagittal and axial planes. The largest side differences in spatial location between any pair of hemispheres were found in the coronal plane where right-sided PRPT was shifted up to 9 mm anteriorly, compared with the left side. The stereotaxic coordinates of PRPT at the level of the internal capsule were documented for four different z planes because of the special clinical interest in this region (Table 5). In comparison with the Talairach atlas (Table 6), PRPT was shifted more posteriorly bilaterally (by 3–6 mm) and more medially on the left side (by 4–7 mm).

View this table:
Table 5

Stereotaxic coordinates of PRPT for their average extension in the internal capsule

SideSagittal plane (x-axis)Coronal plane (y-axis)
Lateral(mean ± SD)Medial(mean ± SD)Anterior(mean ± SD)Posterior(mean ± SD)
PRPT (z = 0 to +12), PRPT in the internal capsule at four different z-planes.
PRPT (z = +12)
    Left–24 ± 2–12 ± 4–6 ± 6–25 ± 4
    Right27 ± 218 ± 2–5 ± 5–23 ± 4
PRPT (z = +8)
    Left–23 ± 2–12 ± 4–6 ± 6–25 ± 4
    Right27 ± 218 ± 3–7 ± 5–22 ± 5
PRPT (z = +4)
    Left–23 ± 4–12 ± 3–8 ± 6–23 ± 4
    Right27 ± 218 ± 3–8 ± 5–21 ± 5
PRPT (z = 0)
    Left–21 ± 3–11 ± 3–9 ± 5–22 ± 4
    Right26 ± 317 ± 2–8 ± 5–19 ± 4
View this table:
Table 6

Stereotaxic coordinates of PRPT from the atlas of Talairach and Tournoux (1988)

SideSagittal plane (x-axis)Coronal plane (y-axisAxial plane (z-axis
*In the Talairach atlas, the right side of the brain is supposed to be the mirror equivalent of the left side. PRPT (z = 0 to +12), PRPT in the internal capsule in four different z planes.
PRPT (total)
PRPT (z = +12)
PRPT (z = +8)
PRPT (z = +4)
PRPT (z = 0)

Volumetric measurements of BA 4 and PRPT

Table 7 shows the individual volumes and asymmetry coefficients of BA 4 and PRPT (dashes = missing values due to insufficient staining) and Fig. 7 shows the distribution of asymmetry coefficients in the present series. ANOVA for repeated measures (regions: BA 4 and PRPT) showed a significant effect of region [F(1,8) = 61.8; P < 0.05). In most of the hemispheres, the volumes of BA 4 were larger than the ipsilateral volumes of PRPT. There was no significant correlation between the volumes of BA 4 and PRPT (P > 0.1). BA 4 varied between 8.62 and 13.08 cm3 (mean ± SD: 10.69 ± 1.30 cm3) on the left side, and between 8.99 and 13.14 cm3 (11.06 ± 1.18 cm3) on the right side. Side differences of BA 4 were not statistically significant. Direction and degree of BA 4 asymmetry coefficients varied between individual brains. Two brains (20%) showed a leftward asymmetry and three brains (30%) had a rightward asymmetry. Most of the brains (50%) were symmetrical for the volume of BA 4. The morphometric results for individual PRPT volumes are also listed in Table 7. On the left side, PRPT varied between 5.36 and 13.46 cm3 (9.32 ± 2.58 cm3), and on the right side it varied between 5.24 and 11.83 cm3 (7.94 ± 2.34 cm3). A paired t-test showed a significant side difference of PRPT between the hemispheres with a left > right asymmetry (P < 0.01). A leftward asymmetry of PRPT was present in seven brains (70%), while a rightward asymmetry was present in only one brain (10%). Two brains (20%) were symmetrical.

View this table:
Table 7

Left and right volumes and asymmetry coefficients of BA 4 and PRPT

HVolume (cm3)δBA 4Volume (cm 3)δPRPT
Insufficient staining (dashes): BA 4 in Brain 1 and PRPT in Brain 11. L = left > right asymmetry; R = right > left asymmetry; S = symmetry.
111.2910.20–0.07 S
28.6211.110.16 R12.0611.21–0.05 S
39.7010.880.08 S7.265.24–0.23 L
413.0811.12–0.11 L7.686.79–0.08 S
59.9511.230.08 S6.715.69–0.11 L
611.8613.140.07 S10.567.72–0.22 L
710.4510.940.03 S9.066.12–0.28 L
811.0112.020.06 S5.366.770.15 R
911.8011.65–0.01 S9.747.85–0.15 L
1010.738.99–0.12 L13.4611.83–0.09 S
119.749.47–0.02 S
Fig. 7

Volume differences of BA 4 and PRPT between left and right hemispheres. Bars represent asymmetry coefficients, calculated for each brain separately. Volumes that are greater in the left hemisphere appear to the left of each vertical line (negative values), while volumes that are greater in the right hemisphere appear to the right of the vertical line (positive values).


The main results of our study can be summarized as follows. The topographical extent and the volumes of BA 4 and PRPT differed considerably among individuals. The cytoarchitectonic and myeloarchitectonic borders did not coincide with sulcal patterns or other MRI-visible macroanatomical landmarks. We generated probability maps for the spatial localization of BA 4 and PRPT, which can potentially increase the validity of structure–function brain mapping in the human motor system. They can be readily applied to in vivo data sets from various neuroimaging techniques. Comparison of these maps with functional data from the literature showed that the anatomical in vivo localization of BA 4 may vary considerably between studies and that the borders between two cytoarchitectonic areas may have specific functional roles. The volumes of BA 4 were symmetrically distributed between left and right hemispheres. In contrast, the volumes of PRPT were asymmetrical with a larger left side in approximately two-thirds of the brains. This anatomical asymmetry may be related to the functional lateralization of handedness.

The present study appears to be the first quantitative structural analysis of the spatial variability of the human primary motor system combining cytoarchitectonic data of BA 4 with myeloarchitectonic measurements of PRPT in the same brains. Knowledge of the link between functional cortical regions and fibre connections is essential to an integrated understanding of the structural organization of the motor system. Previous reports on the primary motor cortex were restricted to measurements of gyral or sulcal landmarks in the central region (Freund and Hummelsheim, 1985; Amunts et al., 1996; Yousry et al., 1997) or they were focused exclusively on BA 4 morphometry (Rademacher et al., 1993; White et al., 1997a, b). Previous studies of PT were mostly based on the qualitative pattern description of Wallerian degeneration in patients with stroke (Kuhn et al., 1989; Miyai et al., 1998) or on myelogenetic observations in immature brains (Yakovlev and Lecours, 1967; Barkovich et al., 1988). These studies have provided important insights into the pathological mechanisms of fibre tract degeneration after cerebral lesions and into the time course of myelination. However, they cannot contribute to the issue of topographical variability in stereotaxic space which is of increasing importance for neuroimaging studies relating structure to function. Such information was also not provided by previous anatomical studies of PT (Nathan et al., 1990; Ebeling and Reulen, 1992). In our study, the advantages of a modified white matter staining procedure (clear delineation of PRPT) (Bürgel et al., 1997, 1999) are combined with those of the GLI method for cytoarchitectonic analysis (observer-independent quantitative analysis of BA 4) (Zilles et al., 1995; Schleicher et al., 1999, 2000) in order to define the range of variations and intrinsic structural constraints of the primary motor system.

Architectonic borders do not coincide with macroanatomical landmarks

Early on, neuroanatomical and electrophysiological studies supported the hypothesis that cytoarchitectonically defined cortical areas may represent distinct functional units and that the cerebral sulci may coincide with the areal borders (Brodmann, 1909; Vogt and Vogt, 1919; Sanides, 1962). However, striking variations in distinct sulcal landmarks pose serious obstacles to the attempt to map behavioural function onto the brain (Rademacher et al., 1992; Leonard et al., 1998). Topography and size of BA 4 vary between the classical cytoarchitectonic parcellations as well as between individual brains (Brodmann, 1903, 1909; Economo and Koskinas, 1925; Sarkisov et al., 1949; Braak, 1980; Rademacher et al., 1993; Amunts et al., 1996; Geyer et al., 1996; White et al., 1997a, b). On the dorsal portion of PRG, BA 4 correlates well with Sanides' motor area 42 (Sanides, 1962). In contrast, the ventral portions of their homologues in the maps of Brodmann (1909) and Economo and Koskinas (1925) have been depicted with a much larger surface extent towards the sylvian fissure. Medially, BA 4 may be relatively large and extend down to the cingulate sulcus (Brodmann, 1909) or it may be much smaller with only a triangular tongue reaching the sulcus (Economo and Koskinas, 1925). Unfortunately, neither the 2D Brodmann map nor its 3D successors (Talairach and Tournoux, 1988; Damasio and Damasio, 1989), which are based on the assumption that the sulcal fundi represent reliable borders, provide the database which is necessary to map motor function to BA 4 or to the surrounding non-primary motor areas.

The present findings confirm the observation that BA 4 covers only small portions of the exposed surface of PRG, typically in the dorsomedial region. Thus, Brodmann's classical visualization of BA 4 covering most of the PRG surface (Brodmann, 1909) does not represent a useful anatomical convention. Also, the superior frontal sulcus did not generally coincide with the inferior level at which BA 4 may be found on the convexity surface (Rademacher et al., 1993). There was also no consistent association between individual sulci and the cytoarchitectonic borders of BA 4 at the paracentral lobule. Neither PRG nor the central sulcus are identical with primary motor cortex. A similar lack of a precise correspondence between the sulcal landmarks and distinct cytoarchitectonic areas has been reported in the visual cortex for BA 17 and BA 18 (Amunts et al., 2000).

Little evidence is available in modern literature about the white matter course of PT in the human forebrain (Davidoff, 1990; Ebeling and Reulen, 1992). The localization of PT in the internal capsule has always been a matter of discussion. The classic view of the organization of PT fibres at the level of the internal capsule is of a single motor homunculus with the head presented in the anterior limb, the mouth in the genu, the upper extremity in the anterior part and the lower extremity in the posterior part of the posterior capsular limb (Penfield and Boldrey, 1937). Another influential concept, which was proposed by Dejerine ~100 years ago (Dejerine, 1901), localized PT in the anterior two-thirds of the posterior limb. Others reported that PT projects through the middle third of the posterior capsular limb (Fries et al., 1993). A rostrocaudal shift with a more anterior position dorsally and a more posterior position ventrally has also been observed (Ross, 1980). Our data show individual variability in the spatial position of PRPT, especially at the corona radiata. At the internal capsule, most of the fibres of PRPT were located in the posterior half of its posterior limb. The compact arrangement of PRPT fibres at the internal capsule explains why there is a high liability to severe clinical deficits even with relatively small lesions. Our data for PRPT have to be interpreted under the caveat that the presented myeloarchitectonic method does not identify all pyramidal fibres as classically defined by the pyramid of the medulla oblongata. Also with other anatomical (Ebeling and Reulen, 1992) and clinical techniques (Kuhn et al., 1989) analysis of human PT was restricted to the central (and largest) portion of PT which originates in PRG. In comparison, up to nine discrete corticospinal projections were identified in the macaque monkey with origins in the frontal, parietal and insular cortex (Galea and Darian-Smith, 1994). How these multiple sources relate to a coherent model of the cortical control of movement is unclear.

Probability maps of BA 4 and PRPT

While modern imaging techniques have increased our understanding of the organization of motor cortex, they also set new standards for precise and reliable anatomical maps as a prerequisite for the topographical interpretation of activation clusters. The decision whether a distinct motor activation is located within a specific architectonic area depends on the precise identification of areal borders. This information has not been available up until now. In a first effort to overcome such limitations, several studies have quantified the stereotaxic variations of major macroscopic (sulcal) landmarks, which may relate to the underlying microarchitecture. Steinmetz et al. (1989) have described, both qualitatively and quantitatively, the topographical variations of the central sulcus in the stereotaxic Talairach frame. Recently, Royackkers and colleagues (Royackkers et al., 1999) have provided a set of quantitative parameters (for various brain sulci, including precentral, central and postcentral sulci) that describes the variability of sulcal geometry and topology and compares the results with those of the Ono atlas (Ono et al., 1990). Information about the variability of human brain shape including the sensorimotor region has also been quantified by Zilles and colleagues (Zilles et al., 2001), based on the same 3D transformation technique as in the present study. With the possible exception of the hand area (Sastre-Janer et al., 1998), however, indirect identification by sulcal landmarks is unreliable because of macro- and microanatomical variability. In this context, architectonic probabilistic maps provide a reliable and valid basis for structure–function mapping (Penhune et al., 1996). Probabilistic information about the spatial topography of BA 4 is also valuable for studies of the premotor or supplementary motor cortices, because the definition of their borders with BA 4 has been difficult (Freund and Hummelsheim, 1985). In the present sample there was a considerable amount of topographical variability of BA 4 (Figs 2 and 4). The degree of variability was inversely related to the outer borders of BA 4, i.e. the further away from the border the lower the range of variations. Also PRPT showed topographical variability and these spatial variations were characterized by a dorsal-to-ventral gradient (Figs 3 and 5). The highest degree of spatial variability was located in the proximal portions of the corona radiata. But also at the level of the internal capsule there was considerable spatial variability (Table 5).

Beyond the macroscopic view, our maps can be readily used to localize BA 4 for a given range of probability and discuss models of motor function. For example, a fundamental question raised by recent evidence on the localization of oculomotor and somatomotor space coding is whether the convergence of space information from both sources in the right-sided PRG is accomplished in premotor cortex or BA 4 (Iacoboni et al., 1997). This question could not be answered on the basis of that functional study. Thus, mapping of the activity foci to the anatomy of PRG and the underlying architectonic areas is pivotal for these studies on brain–behaviour relationships. Comparison of the stereotaxic coordinates for the functional activation foci in motor cortex (Iacoboni et al., 1997) with our probability maps for BA 4 shows that the functional focus was located at the border between BA 4 and premotor cortex with a 50% probability for one or the other. While this could mean also that anatomical probability maps cannot solve this question, an alternative view may be proposed. The term `border' usually implies that there is a strict separation between two regions and a decrease in local `communication'. In contrast, architectonic border zones do not interrupt the continuity of the cortical ribbon and may represent regions where integration takes place. Thus, a focus at the border between BA 4 and premotor cortex may indicate that the task is performed at this distinct location because it requires resources from both regions. Further comparison of our probability maps and spatial coordinates of BA 4 with stereotaxic functional data from the literature showed that the anatomical localization of BA 4 may vary considerably between in vivo brain mapping studies (Shibasaki et al., 1993; Iacoboni et al., 1997; Mima et al., 1999; Toni et al., 1999; Zald and Pardo, 1999).

The comparison of our anatomical maps with those obtained recently by cortical electrical stimulation of the hand areas also yielded interesting results. Nii and colleagues have challenged one of the major axioms in human brain structure–function relationships by providing evidence that the motor and sensory hand cortices are not divided in a simple manner by the central sulcus (Nii et al., 1996). Hand motor responses occurred not only in PRG but also in the postcentral gyrus. In addition to the common caveat that regards patients with chronic epilepsy, who may have altered structural–functional brain topography, there are at least two further explanations for this unexpected observation. First, if the stimulation findings are valid, this would demonstrate the limitations of any anatomical approach because functional variability may then be greater than the underlying microanatomical constraints. Meta-analysis of the present as well as previous cytoarchitectonic parcellation studies, including a total of 100 hemispheres, shows that BA 4 is never located posterior to the fundus of the central sulcus (Rademacher et al., 1993; Geyer et al., 1996; White et al., 1997a, b). Secondly, if the cytoarchitectonic parcellation defines areas of functional relevance (Roland and Zilles, 1998), then the results of the stimulation study could only be reconciled with our cytoarchitectonic data if one assumes that for certain tasks the sensory and motor cortex may form a strong functional unit, with both sensory and motor functions being located in each of the postcentral and precentral gyri (Sobel et al., 1993). The short-range white matter fibres just beneath the `hand knob' region (Yousry et al., 1997; Sastre-Janer et al., 1998) could then be interpreted as an anatomical marker of an intense connectivity. Manual dexterity, of great evolutionary significance to all primates, has been shown to depend on a sustained and rapid transfer of sensorimotor information between the cerebral cortex and the cervical spinal cord (Darian-Smith et al., 1996; Galea and Darian-Smith, 1997). The critical importance of primary somatosensory cortex for motor skill learning in adult primates has been reaffirmed by the recent study of Xerri and colleagues using intracortical microstimulation (Xerri et al., 1999). These authors postulated that the observed changes in the BA 3b representation of digit tips as defined by a behavioural training task could be expected to substantially influence the shaping of learning-induced changes in the major 3b projection targets that feed cortical BA 4 (i.e. areas 3a, 1, 2, 6aα, secondary somatosensory cortex and supplementary motor cortex).

In contrast to the proportional grid of Talairach, our probability maps also give insight into the range of individual variations of PRPT, i.e. connectivity. This easy identification of PRPT is a major improvement over current (indirect) 2D identification of PT in clinical–-topographical studies at the level of the internal capsule.

Stereotaxic position of BA 4 and PRPT

The analysis of the spatial position of BA 4 and PRPT in the Talairach coordinate system is of interest for several reasons. First, the Talairach atlas (Talairach and Tournoux, 1988) is excellent and was conceived for deep brain structures close to the AC–PC line but it is of limited value for cortical mapping purposes because it provides only crude information about the extent of BA 4 and PRPT. In addition, its validity is compromised by a lack of explicit cytoarchitectonic and myeloarchitectonic criteria in the Talairach atlas for defining BA 4 and PRPT. Compared with its location in the Talairach atlas, the dorsolateral portion of the central sulcus may be shifted posteriorly by ~1 cm when studied with in vivo MRI morphometry (Steinmetz et al., 1989). This variable medial half of the central sulcus is especially relevant for brain mapping studies, as it includes the representations of the hand and fingers (Penfield and Rasmussen, 1950; Woolsey et al., 1979). Secondly, it is difficult to differentiate between the activations at BA 4, primary somatosensory cortex and the surrounding non-primary motor areas in PET studies, because of the limited spatial resolution (Mima et al., 1999). Thirdly, Talairach coordinates help to define the range of variations in spatial extent between hemispheres and between individuals. They allow direct reference to the results of functional neuroimaging studies.

How closely did BA 4 match the location of primary motor cortex in the stereotaxic reference frame of the Talairach atlas? The average location of BA 4 in the present study differed in all dimensions from that indicated in the Talairach atlas, which underestimated the range of spatial variability of BA 4 (Fig. 6). At the precentral gyrus this affects the differentiation between BA 4 and premotor cortex and at the paracentral lobule this affects the differentiation between BA 4 and supplementary motor cortex. The most striking differences were observed in the coronal plane with up to 1.9 cm. Such topographical uncertainty is relevant, because distinct motor centres are located only a few millimetres apart (Woolsey et al., 1979). The spatial extent of the cortical area representing the fingers of one hand is <2 cm (Cheyne et al., 1991). The discrepancies between our results and the Talairach atlas are probably caused by the fact that the latter is based on only one brain and the assumption that one hemisphere is the mirror image of its contralateral counterpart. We found anatomical side differences for the stereotaxic position of BA 4 with right-sided BA 4 being located more anteriorly than left-sided BA 4. Thus, the Talairach atlas underestimates bilaterally the true amount of topographical variability in the population. This has also made it difficult to interpret asymmetries of functional activation (Triggs et al., 1994; Netz et al., 1995; Triggs et al., 1997; Ohtomo et al., 1998). They may reflect different cognitive strategies between the hemispheres if bilateral, but topographically, discrepant foci map onto different cytoarchitectonic areas. However, they may also represent identical functional units if they simply follow asymmetries in cytoarchitectonic topography and map onto the same areas.

How closely did PRPT match the location of pyramidal fibres as indicated in the templates of the Talairach atlas? The average location of PRPT in this study showed discrepancies to the Talairach atlas in all dimensions. The most striking individual differences measured ~1.5 cm in each of the three standard planes, underestimating the amount of topographical variability (Fig. 6). Such topographical uncertainty can be expected to be relevant for studies of fibre tract function. We also found anatomical side differences for the stereotaxic position of PRPT with right-sided PRPT being located more anteriorly than left-sided PRPT. The extent and direction of these asymmetries were similar to those of BA 4 thereby implying that the spatial relationship between BA 4 and PRPT may follow fixed anatomical constraints, at least to a certain degree. Detailed analysis of PRPT topography at the internal capsule revealed discrepancies to the Talairach atlas which may be especially relevant to lesion mapping studies because of the relatively small extent of the capsular portion of PT.

BA 4 is organized symmetrically

The size and putative side differences of BA 4 are of considerable interest since they relate to a set of hypotheses about the biological mechanisms of cerebral lateralization leading to an asymmetrical nervous system (Geschwind and Galaburda, 1985). The volume differences of BA 4 between individuals may reflect the inter-individual variability in the size of functional motor maps (Mortifee et al., 1994). It is well known that disproportionately large regions of BA 4 represent the most agile body parts, i.e. the hand and the face (Penfield and Boldrey, 1937). However, comparison of the volumes of specific movement representations and our anatomical maps shows that the relationship between structure and function cannot be described by a simple equation. Although the functional MRI activation cluster for thumb movements has been estimated to measure ~1.6 cm3 (Sanes et al., 1995) as compared with our mean total volume of 10.9 cm3 for BA 4 it has to be taken into account that the cytoarchitectonic map of BA 4 does not represent strictly segregated somatotopic representations but rather intermingled clusters of body parts which overlap (Schieber and Hibbard, 1993; Sanes et al., 1995; Schieber, 1999) and may show use-dependent variations of size (Sanes et al., 1995; Nudo et al., 1996). Accordingly, Nudo and colleagues (Nudo et al., 1992) have found larger representations of the hand area in the squirrel monkey as defined by intracortical microstimulation for the monkey's preferred hand. Inter-individual variation in the total size of BA 4 in our study (Table 7) was considerably smaller than inter-individual differences in the size of functional motor maps for distinct forelimb movements as described for the squirrel monkey (e.g. factor < 2 versus factor 14; Nudo et al., 1992).

Beck (Beck, 1950) was the first to study systematically inter-hemispheric differences of microanatomically defined BA 4. Since then, asymmetries in BA 4 have been repeatedly discussed, but in contrast to the clear evidence on macroanatomical (Geschwind and Levitsky, 1968; Galaburda et al., 1987) and architectonic (Galaburda et al., 1978) leftward asymmetries in the size of auditory association cortex, there are contradicting results for BA 4. Amunts et al. (1996) provided in vivo evidence of left–right differences by measuring the depth of the central sulcus bilaterally on MRI. In right-handers, the left central sulcus was deeper than the right, and vice versa in left-handers. However, the depth of the central sulcus as measured in MRI may not be a perfect marker for the total volume of BA 4. Our volume measurements of BA 4 demonstrated no significant hemispheric asymmetries, a finding which is in support of similar results in the large morphometric study by White and colleagues (White et al., 1997b). Thus, the preferred use of the right hand in the majority of subjects seems to occur without a volumetric asymmetry of BA 4. Neither the cytoarchitectonic borders of BA 4, nor the cortical Talairach coordinates appear to be reliable indicators of determining structural–functional allocations with respect to handedness.

We suggest that the definition of a `standard' asymmetrical left > right pattern does not seem to be a useful anatomical convention to study how motor function maps onto the cerebral cortex. However, the architectonic organization of left and right BA 4 may well be different between the hemispheres. Cytoarchitectonic evidence of a left > right asymmetry in neuropil has been interpreted as an increased connectivity in BA 4 contralateral to the preferred hand (Amunts et al., 1996). Hence, the left-sided and right-sided processors for primary motor processing may be built up differently in relation to handedness, while their bilateral volumes are symmetrical. Further, our data for the whole of BA 4 would still allow for significant regional side differences, if asymmetry in the cortical hand region was concealed by total volume symmetry for the whole body map of BA 4. But White and colleagues (White et al., 1997b) reported symmetry of BA 4 also on the paracentral lobule. Alternatively, lack of structural asymmetry may result from the multiplicity of functional exigencies of different lateralization. For example, tongue movements produced left hemispheric functional lateralization at the lower BA 4 during automatic speech, symmetrical activation during non-speech tongue movements, and right-sided lateralization during singing (Wildgruber et al., 1996).

PRPT is organized asymmetrically

Phylogenetically, the anatomical development and functional role of PT increase with the ability of mammalian species to perform complex hand movements and culminate in human hand control (Kuypers, 1981). Studies in the macaque monkey have established that PT plays an important role in mediating manual dexterity in the primate (Darian-Smith et al., 1999). Nevertheless, the only fibre tract in the human brain that has received considerable attention in morphometric studies has been the corpus callosum. The finding of an association between corpus callosum characteristics, sex and handedness may have caused this bias (Witelson, 1989; Bishop and Wahlsten, 1997). Unfortunately, the observation that variations in the form of PT could be related to digital dexterity in mammals did not cause a comparable increase in scientific interest (Heffner and Masterton, 1975). This may change after the pioneering study of Bürgel and colleagues (Bürgel et al., 1999), who have recently shown asymmetries in the position and size of the human optic radiation.

In contrast with our measurements of BA 4, we have detected a significant leftward asymmetry in the volume of PRPT. Fibre tract asymmetries have already been discussed by Flechsig (Flechsig, 1876) >100 years ago. He described that the uncrossed ventral pathway of PT is often larger on the right side of the spinal cord than the left. Similar findings and a higher pyramidal decussation on the left side were reported subsequently (Yakovlev and Rakic, 1966; Kertesz and Geschwind, 1971). Nathan and colleagues (Nathan et al., 1990) described a greater number of pyramidal fibres in the right side of the spinal cord in 75% of cases. Because of the higher incidence of right-handedness in the population reaching ~90%, Nathan et al. (1990) rejected the hypothesis that PT asymmetry may be a structural correlate for the lateralization of handedness. Nevertheless, it has been speculated that such PT asymmetry is related at least to some degree to handedness or other asymmetries of motor function (Tan, 1989; Macdonnell et al., 1991; Kim et al., 1993; Triggs et al., 1994, 1997; Chen et al., 1997). In general, lower motor thresholds were reported after left-sided cortical stimulation compared with right-sided stimulation, especially in right-handed subjects (conflicting evidence has been discussed by Civardi et al., 2000). The fact that similar asymmetries were also obtained by stimulation of PT suggests that lower thresholds are probably due to differences in PT anatomy (Beric et al., 1997). This explanation would be in accordance with the absence of cortical map hemispheric differences. Fibre tract asymmetries could also explain the finding of a greater influence of the left descending motor fibres on contralateral limb movement than of the right (Tohgi et al., 1996). Obviously, neither the finding of a macroscopic left > right asymmetry in the `hand area' of the central sulcus, which was present in 60% of right-handers (Foundas et al., 1998), nor the left > right asymmetry of PRPT in 70% of the brains in our series can explain the distribution of left hemisphere dominance for right-handedness, i.e. nearly 100% (Milner, 1974). Our findings relate to that portion of PRPT which originates from PRG. Taking into account the fact that large regions of this gyrus are occupied by premotor cortex (area 6aα; Vogt, 1910), the asymmetry of PRPT may reflect side differences between premotor cortices rather than of primary motor cortex. While not solving these important issues, our data help to define the structural constraints of the human primary motor system which may, in one way or another, be related to the direction and/or degree of handedness.

Relationship between BA 4 and PRPT in the same brains

There was no correlation between the volumes of BA 4 and the volumes of PRPT in the present study. This may be surprising if one assumes that the volume of a cortical area reflects the number of neurones in this area which in turn give rise to the descending fibre tracts. However, BA 4 cannot be assumed to be the only cortical source for PT fibres. Non-primary motor and primary somatosensory cortices also give origin to PT fibres (Minckler et al., 1944; Zilles, 1990). Especially fibres which originate from PRG outside BA 4 in the premotor cortex and in the supplementary motor area may have contributed to the observed asymmetrical pattern of PRPT in the present study. It has been estimated that approximately two-thirds of PT fibres have their cortical cells of origin in PRG and the paracentral lobule, being related to BA 4 and BA 6 (Jane et al., 1967; Schoenen and Grant, 1990). Comparative studies in primates have demonstrated that up to one-third of PT axons may arise from the parietal lobe (Russell and DeMyer, 1961; Galea and Darian-Smith, 1994). Each of these separate populations of corticospinal neurones has continuous access to all spinal motor neurone populations, which underlines their functional importance (Galea and Darian-Smith, 1994). Therefore, PT fibres from the non-primary motor areas and the parietal lobe may confound a putative correlation between the size of PT and the size of the primary motor cortex. In the present study, however, those fibres, which originated posterior to the central sulcus in the parietal lobe were not included in the stereotaxic analysis and volumetric measurement. But the method defines the output of primary motor cortex (BA 4) and area 6aα as defined by the Vogts (Vogt and Vogt, 1919). The latter can be regarded as the likely homologue of the postarcuate premotor cortex in the monkey. On the medial side it includes the fibres originating in the supplementary motor area (SMA proper).

The observed discrepancy in the identification of pyramidal fibres may reflect either variability of the staining procedure or differences in the anatomy of the underlying microarchitecture. A relevant variability of the staining procedure itself (i.e. lack of reproducibility) can be excluded on the basis of previous studies (Bürgel et al., 1997, 1999). Therefore, one may speculate that differences in the level of myelination and/or packing density within the same tract (i.e. total volume of all pyramidal fibres) may have caused this discrepancy. Such data are not readily available for the human brain. Flechsig (1876) did not report significant differences in the degree of myelination within the pyramidal tract. There is evidence from primate data, however, that differences in packing density may be the decisive factor. Thus, the density of retrogradely labelled pyramidal neurones after injections of the cervical grey matter (horseradish peroxidase as a tracer) was higher in BA 4 compared with other non-primary motor cortices (Dum and Strick, 1991). Similar differences (i.e. highest density of pyramidal neurones in BA 4) were also reported between BA 4 and the primary somatosensory cortex (Murray and Coulter, 1981). Since the density of retrogradely labelled pyramidal neurones can be expected to correlate with the density of descending pyramidal axons (Porter and Lemon, 1993), there is good reason to assume that the resulting disproportionately high packing density of pyramidal fibres from BA 4 causes the observed discrepancies in staining patterns. Further studies are needed to see whether the non-correlation between BA 4 and PRPT persists when primary somatosensory and premotor cortices are also included in the calculation. Such data are not available in the literature.

In conclusion, we describe new data for the mapping and measurement of the primary motor system in the human brain. For the first time, quantitative measurements of spatial location and volumes of both BA 4 and PRPT have been performed in the same brains in order to define the anatomical constraints which relate to primary motor action. The presented data predict that lack of understanding of intra- and inter-individual variability of size, shape and location of BA 4 and PRPT may lead to significant structural–functional mismatch. The results provide a rationale for thinking about individual variability and hemispheric asymmetry when designing morphometric and functional studies of the primary motor system. Mapping BA 4 and PRPT to a 3D human brain has provided a stereotaxic atlas-based tool that can quantify the probability of localizing specific functional properties inside or outside the primary motor system. The probability maps of BA 4 and PRPT can be used to clarify the role of variations in both anatomical patterns and cognitive strategy.


We thank U. Blohm and C. Opfermann-Rüngeler for their excellent technical assistance. This `Human Brain Project' research was funded jointly by the National Institute of Mental Health, National Institute of Neurological Disorders and Stroke, National Institute on Drug Abuse and the National Cancer Institute. This work was also supported by grants from the Deutsche Forschungsgemeinschaft, SFB 194/A6 (to K.Z.) and SFB 194/A9 (to H.-J.F.).


View Abstract