Since the pioneering work by Wind (1984), Conroy and Vannier (1984), and Zonneveld and Wind (1985), two-dimensional (2D) and three-dimensional (3D) computed tomography (CT) have opened a new window to the study of fossil hominids. Using CT-based computer imaging techniques, endocranial and intracranial features often hidden or covered by stone matrix have been analyzed on “virtual” specimens. Among the major features studied have been endocranial volume, the bony labyrinth, the paranasal sinuses, the location of the cerebral lobes, and basicranial flexion (Zonneveld et al., 1989; Conroy et al., 1990; Kalvin et al., 1992; Spoor and Zonneveld, 1995, 1999; Spoor et al., 1994; Hublin et al., 1996; Seidler et al., 1997; Conroy et al., 1998; Zollikofer et al., 1998; Bookstein et al., 1999; Conroy et al., 2000). Some of these studies have claimed specific morphological conditions (e.g., the orientation of the anterior cranial fossa or the massive pneumatization of sinuses) as significant for the phylogenetic interpretation of particular hominids. In addition, conventional (non-CT-based) studies have suggested features such as the groove for the superior petrosal sinus and the domelike arcuate eminence in modern Homo sapiens, but generally missing in Neanderthals (Schwartz and Tattersall, 1996), or the Breschet sinus found only in European archaic H. sapiens (Arsuaga et al., 1997). The precise identification of endocranial features of phylogenetic potential is of great importance to placing fossils in paleoanthropological perspective. Thus, the main aims of the present study are to provide additional data on endocranial and intracranial morphology of African archaic H. sapiens, and to evaluate the possibilities and problems of such virtual analysis.
For the first time, the well preserved matrix-filled cranium KNM-ES 11693 from Eliye Springs, Kenya, has been studied by computed tomography and virtual-3D reconstruction. The Eliye Springs cranium (Fig. 1) was found in beach deposits at the western shore of Lake Turkana. Its heavy degree of mineralization, which is also shown by associated faunal remains, strongly suggests that the specimens derived from later deposits of the Koobi Fora Formation, which underlie the Holocene Galana Boi Beds (Bräuer and Leakey, 1986). Analysis and comparisons of the cranium have shown close affinities to later Middle Pleistocene archaic H. sapiens (Bräuer and Leakey, 1986; Bräuer, 1989), which, according to recent dating, might have existed between 300,000 and 150,000 years B.P. (Bräuer et al., 1997). Ectocranially, the Eliye Springs hominid deviates from modern anatomy in a number of features, which include indications to a well-projecting supraorbital morphology, massive temporal crests, upward converging parietal walls, strong occipital curvature, and heavily pneumatized maxillary sinuses. Phylogenetically, there can be little doubt that the hominid belongs to the pre-modern or late-archaic range of variation in East Africa. Recent analysis of the Eliye Springs cranium has revealed a thickening of the vault bones (Fig. 2) which was probably caused by chronic anemia in the childhood or youth of this individual (Bräuer et al., 2003). To assess the morphological conditions seen in the Eliye Springs hominid, a sample of modern crania from the Mumba and Strauss Caves in Tanzania, which date from the final late Pleistocene and Holocene (Bräuer, 1980; Bräuer and Mehlman, 1988), was also studied by CT-based 3D reconstruction.
MATERIALS AND METHODS
Scanning of the matrix-filled cranium from Eliye Springs was carried out by Fred Spoor at the Diagnostic Centre in Nairobi. Using a Siemens Somatom AR. SP scanner with a tube voltage of 130 kV and an exposure of 249 mAs, 201 coronal scans were acquired. Slice thickness and distance were set to 1 mm. The modern comparative sample consists of three crania from the Mumba Rockshelter (Mumba 4, 6, and 10) and two specimens from the Strauss Rockshelter (Strauss 2 and 3), which were preserved well enough for this study. Scanning of this sample was performed by C.G. at the Department of Neuroradiology, University Hospital Hamburg-Eppendorf. Since nearly no matrix had to be x-rayed, a much lower exposure could be used in these cases. The scanning was done on a Siemens Somatom Plus 4 using a slice thickness of 1 mm and a slice distance of 0.5 mm.
By linear interpolation, the datasets were scaled in axial direction to obtain isotropic voxels. All further processing was performed using VOXEL-MAN, a high-resolution volume visualization system originally developed for medical applications (Höhne et al., 1995; Tiede et al., 1998; Pommert et al., 2001). As a first step, an interactive segmentation was done to identify and label the objects of interest. In cases with a sufficient contrast between object and background, such as fossil bone and air, an intensity threshold was used. This technique could be applied for the modern sample and many parts of the Eliye Springs cranium, such as its outer surface. The next step for this study of the Eliye Springs cranium was the virtual removal of the sandstone matrix from the endocranium, the sinuses, and labyrinths. Since the fossil bone and the matrix exhibit close similarities in density and fall into the same intensity range, the different objects often had to be separated by manual editing on orthogonal (transverse, sagittal, and coronal) CT slices. Marked lines of different grey levels helped in separating bone and matrix. In case of an interrupted borderline, as with the frontal squama, a “mental interpolation” of the line was necessary (Fig. 2). Difficulties in separating bone from matrix occurred especially with regard to the right orbital roof, parts of the pituitary fossa, and the sphenoidal sinus. In a subsequent step, the manually edited contour was smoothed using a Gaussian filter. Results of segmentation were visually inspected on orthogonal cross-sectional images and perspective views, and stored as object membership labels assigned to every voxel. All segmentation work was performed by K.K.
VOXEL-MAN is a system for direct volume visualization; i.e., 3D views are created directly from the segmented volume data. Compared to other approaches, where an intermediate surface description is created using triangles or other surface patches, the major advantage is that all image information originally acquired is kept during the rendering process and may be accessed (e.g., by placing arbitrary cuts). This makes it an ideal technique for interactive data exploration (Pommert et al., 2002a). For rendering, an algorithm was developed that considers both object labels and local object intensities; it can display isointensity surfaces defined by a threshold with subvoxel resolution (Tiede et al., 1998).
As could be shown, the accuracy of the thus defined surface position is indeed in the subvoxel range, provided that a suitable threshold value is used (Pommert et al., 2002b). With a poor threshold, the deviation can be as large as the width of the 3D point-spread function, which is determined by scanner filter kernel, slice thickness, and spiral pitch, and may be up to several millimeters (Prevrhal et al., 1999). However, since a poor threshold selection leads to marked artifacts on the surface, such as stair steps or increased noise, choice of a suitable threshold can be assumed. For manually edited borders, the accuracy is limited by the voxel size. For the Eliye Springs cranium, with a voxel size of 0.41 mm, an accuracy of about 1 mm seems reasonable, except possibly in the particularly difficult regions as described above. VOXEL-MAN also provides a set of tools for measurements of distances, angles, and volumes. Since the distances are measured between points marked on a visible surface, accuracy is in the same range as described above. Volumes are measured by counting the voxels belonging to a segmented object.
In their first study, Bräuer and Leakey (1986) could only roughly determine the endocranial capacity to between 1,300 and 1,450 cc, based on correlations between major ectocranial dimensions and the capacity as suggested by Olivier and Tissier (1975). Now, a more precise estimation of the endocranial volume is possible. Direct measurement of the endocast yielded a volume of 1,243 cc. In order to correct some postmortem deformations on the right side of Eliye Springs's cranial vault, the left side was mirror-imaged at a midsagittal plane. For this purpose, two slightly different planes were interactively defined by setting three points on the outer surface of the skull, corresponding to some anatomical landmarks (nasion, inion, opisthion, and nasion, lambda, opisthion, respectively). Accuracy of the thus obtained midsagittal planes was visually controlled using an automatically drawn line, indicating the intersection of plane and skull, and an overlay of distorted right and mirror-imaged left halves of the skull. Measurements of the thus corrected endocast yielded a volume of 1,222 cc and 1,170 cc, respectively. The average of all three measurements is 1,212 cc. Thus, the new estimations of the volume of Eliye Springs's virtual endocast are considerably smaller than the previous rough estimate based on ectocranial measurements. With a likely volume of about 1,210 cc (and a possible range between ∼1,170 and 1,245 cc), Eliye Springs is close to the capacities of the other two archaic H. sapiens specimens from Bodo (Ethiopia) and Kabwe (Zambia), for which new CT-based studies yielded average estimations of 1,245 cc and 1,270 cc, respectively (Conroy et al., 2000; Seidler et al., 1997). For Bodo, however, a range between 1,200 and 1,325 cc has been suggested, depending on different reconstructions of the missing parts of the basicranial region. Apart from frequent uncertainties in exactly determining the capacity, conventional data point to a great variation among African archaic H. sapiens from about 1,000 to 1,400 cc (Bräuer, 1984). Nevertheless, Eliye Springs's volume, which is also close to that of Saldanha (South Africa), is smaller than those of the late archaics Laetoli H 18 (Tanzania), Omo Kibish 2 (Ethiopia), and Jebel Irhoud 1 and 2 (Morocco).
Anterior Cranial Fossa
According to Aiello and Dean (1990) the full expansion of the frontal lobes occurred very late in human evolution, and only in late Middle Pleistocene hominids is the anterior cranial fossa comparable in size to that of modern humans. Following Haas's (1980) definition, the medial length of the anterior cranial fossa (the distance between the endocranial opening of the optic canal and the rounded transition from the anterior cranial fossa to the frontal squama, measured on a parasagittal section through the endocranial opening of the optic canal; Fig. 3) could be measured on the 3D reconstruction of the intact left side of Eliye Springs and on some Holocene crania (Table 1). To also determine the relative medial length of the anterior cranial fossa, the ratio (index) between this measurement and the medial length of the endocast (between the anterior point of the medial length and the most posterior point of the endocast in the same parasagittal plane) was calculated. In addition, the breadth of the fossa was determined by modifying Haas's (1980) definition in using the projective distance between the midpoint of the lateral endofrontal fovea and the midsagittal plane. As Table 1 shows, Eliye Springs's anterior cranial fossa is only slightly shorter both absolutely and relatively than the ones in the small modern sample and falls into its range with regard to the breadth.
Table 1. Measurements (mm) of the anterior cranial fossa of the Eliye Springs hominid and the comparative modern specimens
Medial length endocast
(Medial length/medial length endocast) × 100
Eliye Springs (l)
Mumba 4 (l)
Mumba 6 (r)
Strauss 2 (l)
The borders of the cribriform plate of the ethmoid bone are not definable precisely in Eliye Springs due to the limits of resolution, but a coronal CT scan at the level of the crista galli shows the position of the plate in relation to the orbital roofs (Fig. 4). This relation has been previously examined and commented upon other fossil hominids. For example, Weidenreich (1943) described a gradual decline of the floor of the anterior cranial fossa from lateral to the cribriform plate in modern humans, whereas he found a rather abrupt slope toward the plate from a generally even floor in the Zhoukoudian H. erectus sample. In OH 9 the cribriform plate is also described as deeply wedged in between the orbits (Maier and Nkini, 1984), and according to Nkini (1986) its lowest level is situated about 25 mm below the orbital roof, which is protruding well toward the endocranium. According to Aiello and Dean (1990), it can be argued that since the late Middle Pleistocene, the expansion of the forebrain has finally extended downward, so that the orbital cavities have been restrained from above during growth and remain at the level of the cribriform plate. Coronal sections of four crania from the Mumba and Strauss Caves at the level of crista galli (Fig. 4) reveal a remarkable variability showing height differences between the cribriform plate and the orbital roof at the level of the endofrontal eminence (Lang, 1981) of 16.7, 18.1, 20.3 and 22.5 mm for Strauss 2, Mumba 4, Mumba 6, and Mumba 10, respectively. Interestingly, Eliye Springs exhibits the smallest height difference (15.1 mm) and a weak decline toward the cribriform plate (Fig. 4b), whereas the modern specimens include similarly weak slopes, but also steeper declines, as in Mumba 6 and 10, even reaching height differences closely approaching that given for the Olduvai H. erectus cranium. Based upon only these few specimens, it appears evident that this feature is rather variable. Further study is necessary to evaluate its phylogenetic relevance.
Based on coronal (Fig. 4a) and midsagittal CT scans (Fig. 5a), it appears likely that the crista galli in Eliye Springs is not defective. The crest is only slightly projecting and seems to be pneumatized. Although the crista galli exhibits great variation in size and shape among recent humans (Heringhaus, 1959; De Villiers, 1968), a functional connection to increased dural tensions resulting from rotations of the neurocranium and falx cerebri (Moss, 1963) can be assumed. Whereas Weidenreich (1943, 1951) could not find any crista galli in both the Zhoukoudian and Ngandong samples, and its existence in OH 9 is unclear (Nkini, 1986), the feature occurs in African archaic H. sapiens from Bodo, Ndutu, and Kabwe. Seidler et al. (1997) have even observed that in Kabwe and Petralona, the apex of the crista galli is rotated in an occipital direction, whereas in Arago 21 and modern humans the tip is directed more or less vertically. These authors suggested a connection between the orientation of the crista galli and the more steeply inclined orientation of the frontal lobes. Arsuaga et al. (1997) saw no significant differences in size and shape of the crista galli between the pre-Neanderthals from the Sima de los Huesos (Atapuerca) and modern humans. In our comparative sample from Tanzania, the apex is generally vertically oriented with a possible slight posterior inclination in Mumba 4 (Fig. 5c), whereas it appears more difficult to decide on the orientation of the very low apex in Eliye Springs (Fig. 5a). The relevance of this feature as well can be assessed only on a larger sample of fossil and recent specimens.
Another attachment for the falx cerebri is formed by the frontal crest, which is also a variable structure. In Eliye Springs it is only moderately projecting, with a maximum height of about 9 mm. Its base measures about 42 mm. The lower end of the frontal crest is well defined, and the foramen caecum is visible on sagittal CT scans (Fig. 5a). The length of Eliye Springs' frontal crest falls close to the lower end of the range of the five modern specimens (42–58 mm), and its height falls well into the range (4–11 mm). Concerning other fossil hominids heights of at least 5 mm for OH 9 (Nkini, 1986) and 5 mm for La Ferrassie 1 (Heim, 1976) have been reported. In the Zhoukoudian sample, the frontal crest has been described as a thin, high, bladelike elevation (Weidenreich, 1943), and as a broad, high crest (Wu and Poirier, 1995) in the Lantian specimen. The late Pleistocene Ngandong crania exhibit very pronounced and somewhat rounded frontal crests (Weidenreich, 1951).
Weidenreich (1951) also noted another feature, the possible bifurcation at the transition from the frontal crest to the sagittal sinus common in modern humans. In the Ngandong skulls he could see no trace of such a division into two “lips.” According to Wu and Poirier (1995) a bifurcation is also missing in the Lantian hominid, but a low division is reported for one of the Zhoukoudian crania (Skull III) (Weidenreich, 1943). Due to the difficult virtual removal of the matrix, the condition in Eliye Springs is not clearly determinable, although a separation into two lips such as that present in Strauss 2 is unlikely. Interestingly, this specimen is the only one in the Mumba/Strauss sample showing a bifurcation.
Seidler et al. (1997) found that in Kabwe and Petralona, the angle between the course of the anterior cranial fossa and the Frankfurt Horizontal (FH) measured 24° and 25° respectively, whereas for modern humans (n = 20) they found a range of 5° to 12°. Based on these measurements the authors suggested that in these archaic hominids the frontal lobes are inclined more steeply than those in modern humans. However, according to Seidler et al. (1997), Arago 21 falls into the modern range, indicating a greater complexity of this morphology. Following Seidler et al.'s definition of the angle (between the FH and a parasagittal section through the medial edge of the orbit), we determined the values for Eliye Springs on the intact left side as well as on the modern sample. The results (Table 2) indicate that there is considerable variation among the three modern Tanzanians (8.2–20.8°), and Eliye Springs's value of 1.7° (Fig. 6b) is even smaller than the smallest figures found by Seidler et al. and us in modern humans (Fig. 7). It is also remarkable that Mumba 4 exhibits 20.8°, a value close to that of Kabwe. In order to evaluate the robusticity of this angle (hereinafter called “S1), we determined two additional angles on parasagittal planes: one 5 mm lateral to the plane of S1, called “S2 and another 10 mm lateral to the plane of S1, called “S3.” The results presented in Table 2 show that such minor deviations or uncertainties in placing the plane can lead to considerable differences with regard to the angles that might mainly be due to the uneven surface of the anterior cranial fossa. For example, the S1 angle of Mumba 6 increases considerably when moving the plane by only 5 mm (S2). On the other hand, in Mumba 4 the angles decrease from S1 to S3. Our results underscore the need for broad and detailed analysis on the orientation of the frontal lobes, including the variation along the anterior cranial fossa, before any conclusions about the phylogenetic significance in later hominid evolution can be drawn.
Table 2. Angle of orientation of the frontal lobes: “S1”,a “S2” determined 5 mm more laterally; “S3” determined 10 mm more laterally than “S1” (see text)
Seidler et al. (1997) emphasized not only the unusual orientation of the frontal lobes in Kabwe and Petralona but also their location. In these specimens the superior surface of the orbits is formed by the lower part of the frontal sinus, also deviating from the modern condition in which the frontal lobes are positioned directly above the orbits. In Eliye Springs, in which the supraorbital torus is mostly broken off, it is clearly evident (Fig. 8) that the frontal lobes are placed above the orbital cavities, very similar to the placement in modern humans. Arago 21 shows this modern pattern as well (Seidler et al., 1997), in spite of its well-developed supraorbital torus. Thus, the spatial relationships between the orbital roofs and the position of the frontal lobes also deserve broader study.
Middle Cranial Fossa
This part of the endocranium, which supports the temporal lobes and exhibits a small depression for the pituitary gland, is also rather well preserved in Eliye Springs, and most of the sediment matrix could be virtually removed from the bony surface. Only the pituitary fossa caused some difficulties, with its defective condition and nearly completely missing dorsum sellae.
Using Lang's (1981) measurements for the middle cranial fossa (Fig. 3) we determined the lateral length (oblique distance between the transition from the crest of the lesser wing to the calvaria and the most lateral point of the petrous superior margin) to be 61.4 mm and the medial length (distance between the posterior edge of the lesser wing and the petrous superior margin determined on a parasagittal plane through the lateral edge of the oval foramen) to be 45 mm on the left undistorted side of the Eliye Springs cranium. The points of measurement were checked on CT scans. The values fall close to the lower end of the ranges of the modern sample (n = 4; lateral length: 61.8–71 mm; medial length: 46.6–49.8 mm). There are no data on other fossil hominids available for these measurements.
Due to the lateral expansion of the brain in modern humans, the glenoid fossa lies entirely below the middle cranial fossa, whereas in apes and most fossil hominids the glenoid fossa lies more lateral to the brain under the root of the zygoma. H. erectus specimens (e.g., Zhoukoudian, OH 9) exhibit a somewhat intermediate position in this respect (Weidenreich, 1943; Nkini, 1986; Aiello and Dean, 1990). Coronal sections of the Eliye Springs virtual cranium reveal a modern condition for the location of the glenoid fossa below the temporal lobe.
Although the pituitary fossa is very similar in early and later hominids, with its well-developed dorsum sellae and the prominent anterior clinoid processes (Aiello and Dean, 1990), Weidenreich (1951) found that length and breadth of the fossa in the Ngandong 11 specimen (22 mm/22 mm) are double the average length and breadth of the fossa in modern humans (mean= 10.7 mm/10 mm), while the depth is about the same. Since the pituitary fossa could not be completely cleaned from the matrix, this observation on Ngandong has been doubted (Washburn and Howell, 1952) and needs to be verified by CT-based studies. Thus, speculations on the size of the pituitary gland remain so far hypothetical. Due to the fragmentary condition of this region in Eliye Springs, only a rough estimate of the breadth of the pituitary fossa is possible following Weidenreich's (1951) definition (distance between the medial borders of the sulci carotici). To determine the points, CT scans were also useful. By doubling the right half, an estimate of 14 mm was yielded. This figure falls into the range of a modern sample (7–17.5 mm) as cited by Weidenreich (1951) as well as into the range of the three comparative specimens, on which we could determine this measurement (12–17 mm).
In spite of existing uncertainties in the diagnosis of the pituitary fossa in the Ngandong crania (Washburn and Howell, 1952), Weidenreich (1951) arrived at the conclusion that in these hominids the fossa is located further back, compared to the location in modern humans, so that its center falls behind a coronal plane through the two oval foramina. A coronal section of the Eliye Springs cranium through the middle of the oval foramina indicates that the center of the pituitary fossa might lie slightly behind this reference plane (Fig. 9). In the two Tanzanian crania in which the feature could be assessed, the coronal reference plane cuts the dorsum sellae (Strauss 2, Fig. 9c) or the posterior part of the fossa (Mumba 4), which might possibly point to a difference with regard to Eliye Springs. But from the data currently available, it is difficult to assess the significance of these minor diversities.
Lang (1981) defined another measurement that could also be of interest in the study of fossil hominids. This is the vertical distance between the floor of the pituitary fossa and the Frankfurt Horizontal. In extant humans there appears to be great variability. Lang (1981) found that on average the floor is situated 13.2 mm above the Frankfurt Plane, with a range of –2.3 to 21.3 mm. For Eliye Springs a value of 21.6 mm could be determined, which falls just outside the upper limit of the recent sample. The two Tanzanian skulls that could be measured (Strauss 2 and Mumba 6) yielded values of 11.3 and 15 mm, respectively. Detailed studies on the vertical position of the floor of the pituitary fossa might be of interest, also by using different horizontal planes.
Another important region of the middle cranial fossa is formed by the petrous part of the temporal bone, which projects medially across the cranial base. The petrous crest on top, to which the tentorium cerebelli attaches, forms the border between the medial and posterior cranial fossae. The orientation of the petrous parts varies and changes in hominid evolution mainly due to the growth of the cerebellum and the size of the posterior cranial base. According to Dean (1988; p. 110), “The short posterior part of the cranial base of modern humans forces the cerebellum to expand laterally under the tentorium cerebelli so orientating the petrous temporal bones coronally.” On fossil hominids the angle formed by the right and left petrous pyramids has generally been determined ectocranially. Weidenreich (1943; p. 57) defined the ectocranial pyramid angle (“degree of deviation of the pyramid axis from the mid-sagittal plane”) and determined a value of 40° for Skull III of the Zhoukoudian H. erectus and 63° for a modern European specimen, suggesting that the petrous pyramid has moved to a more coronal orientation in modern humans. In a more recent worldwide study, Mathews (1987) found a variation of 35–68° in a sample of 200 crania with regional means of around 50°. Putz (1974; p. 260) defined an endocranial pyramid angle (“anatomical angle”) between the petrous crests of both sides and found an average of 102° for a modern human sample (n = 46). If one only considers the endocranial pyramid angle on one side, i.e. the angle between the petrous crest and the mid-sagittal plane (as usually done for the ectocranial angle), the average for Putz's extant human sample would be 51°. In the present study both the endocranial and ectocranial petrous angles have been determined on the virtual Eliye Springs cranium. On the undistorted left side, an endocranial angle of 52.3° and an ectocranial angle of 39.3° could be measured (Fig. 10a). For comparison, the endocranial and ectocranial angles were also determined on the virtual images of the Tanzanian crania, yielding values between 41.7 and 52.3° (n = 4) endocranially and 45.1 and 54.9° (n = 3) ectocranially. Our data support a modern condition for Eliye Springs' endocranial angle, which is identical to that of Mumba 4. On the other hand, the ectocranial angle is very similar to that of Zhoukoudian Skull III, as reported by Weidenreich () and lies well below all geographic means found by Mathews (1987), but not outside the range of variation as determined in that study.
The petrous temporal part exhibits a number of additional features of phylogenetic relevance in the middle cranial fossa. Several researchers have described deviations between modern and non-modern humans with regard to the shape and level of the petrous crest. For example, Weidenreich (1943) found that in modern humans the superior crest mostly forms a sharp edge that partly overhangs the posterior surface, whereas in Zhoukoudian H. erectus the crest forms an obtuse or completely rounded corner. Wu and Poirier (1995) emphasized that in the Lantian hominid both the anterior and posterior surfaces of the pyramid are steeper and the upper border of the pyramid is less obtuse than in the Zhoukoudian sample, indicating closer affinities to Trinil than to Zhoukoudian. However, there is a relatively obtuse angle in Sangiran 4, with the posterior surface sloping posteriorly. Weidenreich (1951) also sees differences in the Ngandong crania. The posterior wall of the pyramid rises to a much lower level than in modern humans, and its superior surface does not sink down to the bottom of the fossa as abruptly as in modern crania but has a more gradual, gentle incline. In modern humans the pyramid should rise to a much higher level, and the anterior and posterior surfaces would meet at an obtuse angle in contrast to the right angle and the lacking sharp edge seen in Ngandong. For the Neanderthal from Amud, Suzuki (1970) also described a fairly low pyramid, which is said to be characteristic of H. erectus and Neanderthals. In Amud the anterior and posterior surfaces form an obtuse angle of 97° (measured at the lateral margin of the internal auditory meatus), whereas that angle in modern humans is more acute (a value of c. 78° is given by Suzuki (1970)).
Analysis of these features in Eliye Springs revealed that the superior margin is relatively sharp-edged (Fig. 11a), similar to modern conditions, and that the angle between the anterior and posterior surfaces following Suzuki's definition measures about 93° (Fig. 11b). The latter angle was determined by cutting the pyramid at the lateral edge of the internal auditory meatus, perpendicular to the superior margin. The comparative specimens (Fig. 12) show values of 74° (Mumba 4), 71° (Mumba 10), and 75.3° (Strauss 3). The angle in Eliye Springs could thus indicate some nonmodern tendency, but the variability of this feature is still less explored and standardized. To describe the course of the superior surface with regard to Weidenreich's (1951) observation in Ngandong, we cut the virtual pyramid oriented to the Frankfurt Horizontal parasagittally at the middle of the superior margin (Figs. 11c, 12). In contrast to Weidenreich's observation, the superior surface follows a rather flat course toward the bottom of the fossa in all three Mumba specimens and in Strauss 2; a slightly stronger decline might be present in Strauss 3. In view of the conditions and variation among these modern specimens, no clear deviation can be assumed for Eliye Springs. But again, as with most of the other traits dealt with here, more basic reproducible studies are needed to adequately describe the extant variation of this complex morphology.
Recently, Schwartz and Tattersall (1996) have drawn attention to the arcuate eminence by arguing that adult and juvenile modern humans and juvenile Neanderthals share a prominent domelike arcuate eminence, whereas in adult Neanderthals (e.g., La Quina 5, Spy 1 and 2, La Ferrassie 1 and 2), with the exception of Gibraltar 1, this region is only minimally convex, following the curvature of the superior surface of the petrosal, or flat. Schwartz and Tattersall (1996) regard this flattening as the result of ontogenetic remodeling leading to a derived condition in Neanderthals. In spite of this intriguing hypothesis, it cannot be overlooked that other researchers described the arcuate eminence in La Ferrassie 1 as “très saillant,” or very prominent (Heim, 1976; p. 218), and in Amud 1 as a circular eminence of 15 mm in diameter with greatly depressed areas anterior and medial to the eminence (Suzuki, 1970). There seems to be variation not only among Neanderthals but also among Asian and African H. erectus. Weidenreich (1943; p. 67) described the arcuate eminence in the Zhoukoudian specimens as ”much less pronounced“ than in modern humans, whereas Wu and Poirier (1995) regard the condition in the Lantian H. erectus as more similar to modern humans than to Zhoukoudian; and Plhak (1983) characterized the eminence of Sangiran 4 as well developed. The conditions of the arcuate eminence of three specimens of African H. erectus are described as marked by a clear transverse ridge in OH9 (Nkini, 1986) and clearly defined in KNM-ER 3891 (Wood, 1991) to very low in the Turkana Boy (Walker and Leakey, 1993). Whatever the evolutionary significance of this feature might be, in Eliye Springs the eminence (Fig. 13a) is well developed by a clearly defined mount. Figure 13b shows the location of the anterior semicircular canal. All three modern specimens in which the region could be assessed (Mumba 4, Strauss 2 and 3) show prominent eminences as well. Another complex feature, which has revealed to be of great phylogenetic relevance, is the bony labyrinth with its semicircular canals (Spoor et al., 1994; Hublin et al., 1996; Spoor and Zonneveld, 1998), but the condition of this feature will be treated separately.
Posterior Cranial Fossa
The endocranial surface of the posterior fossa can also be well studied in Eliye Springs. Following Lang's (1991) definition of the length of the posterior cranial fossa (lateral border of the trigeminal impression to the lower edge of the transverse sinus, parallel to midsagittal plane), a value of approximately 73 mm could be determined for Eliye Springs, though slight uncertainties remain with regard to the exact placement of the points. Nevertheless, the value falls into the range of variation of the modern Tanzanians (n = 3; 67.4–74.5).
Among the phylogenetically interesting features is the orientation of the posterior petrous surface (Aiello and Dean, 1990), which is nearly vertical or even “undercut” in modern humans and some early Homo (Clarke, 1977), whereas in H. erectus and some australopithecines (such as KNM-WT 17000), the posterior aspect of the petrous bone slopes posteriorly and is much less vertical in its profile. However, it is obvious that the orientation of the posterior surface depends on the orientation of the whole cranium and thus needs clear standardization. Lang (1991) determined the angle between the posterior portion of the petrous part and the Frankfurt Horizontal at the level of the internal auditory meatus and found an average of 79.2° for a recent sample. Parasagittal cuts of the virtual 3D reconstruction of Eliye Springs and the modern Tanzanian specimens were done at the level of the lateral edge of the foramen as with the angulation of the posterior and anterior surfaces (see above). For Eliye Springs a value of 78.3° was found (Fig. 14), which is close to Lang's average. In only two of the modern specimens a reliable determination of the angle was possible: Mumba 4 yielded a value of 80.2° and Strauss 2 68°. More detailed studies on the orientation of the posterior petrous surface would be useful for other Homo species as well, also considering the mediolateral variation along the petrous part.
The internal auditory meatus is clearly visible in the 3D reconstruction of the Eliye Springs endocranium (Fig. 15a). Although size and shape of this opening are variable even in the same specimen (e.g., in La Ferrassie 1; Heim, 1976), some differences in its location have been described among fossil hominids. McCown and Keith (1939) mentioned a more superior location of the opening in the Gibraltar Neanderthal, and Suzuki (1970) saw a more transverse direction of the porus in Amud than in modern humans. In contrast, Weidenreich (1943) found no difference in the internal auditory meatus between H. erectus and modern humans. Walker and Leakey (1993) describe the shape of the internal porus of the Turkana Boy as oval as is usually found in modern temporals (see also Lang, 1991). In Eliye Springs the porus exhibits a superior location and is oval in shape. The small modern sample shows considerable variation, with Strauss 2 (Fig. 15c) having a superior location similar to that in Eliye Springs and Mumba 4 having a more inferior location. The condition in Mumba 10 could be described as intermediate (Fig. 15g), whereas the location in Strauss 3 is difficult to determine due to the oblique orientation of the foramen (Fig. 15e).
Other morphological features of the petrous temporal bone, for which variation among fossil hominids and extant humans has been described, include the subarcuate fossa and the external aperture for the vestibular aqueduct (Schwartz and Tattersall, 1996; Hofmann, 1987; Heim, 1976; Plhak, 1983). In contrast to most primates, humans lack a subarcuate fossa which extends through the arc of the anterior semicircular canal and houses the petrosal lobule of the cerebellar paraflocculus. Instead, in humans and the great apes a similarly shaped fossa containing only vascular tissue is present in fetal stages and closes after birth, becoming the petromastoid canal to transmit the subarcuate artery to the periotic bone (Gannon et al., 1988; Spoor and Leakey, 1996; Scheuer and Black, 2000). Although homology and terminology of these “subarcuate fossae” have long been discussed (see also Mojà-Solà and Köhler, 1993), size and shape of the “closed subarcuate fossa” have been considered with fossils and recent humans (Lang et al., 1981). For example, Schwartz and Tattersall (1996) mentioned the existence of a well-defined, but not deeply concave, subarcuate fossa in Gibraltar 1 and Le Ferrassie 1. On the better-preserved petrosal of Eliye Springs (Fig. 15a), no trace of a fossa can be seen. However, it cannot be excluded that a weak depression exists on the original that is not visible on the virtual bone due to the limits of resolution. Small but more clearly defined depressions are present on Strauss 2 and Mumba 10 (Fig. 15c, g). Somewhat medial to the sulcus for the sigmoid sinus there is a horizontal slit, which is the external opening for the vestibular aqueduct. Weidenreich (1943) and Suzuki (1970) mentioned that in Peking Man and in Amud Man, the cover of the aperture appears as a distinct eminence, whereas in modern humans the slit is covered by a plate. The parasagittal scan through the respective area in Eliye Springs (Fig. 15b) shows the aqueduct, or canal, of the vestibule, the opening of which appears to be overhung by a well-developed eminence. In Strauss 2 (Fig. 15c, d) the aqueduct is clearly visible, but an eminence as in Eliye Springs is missing. However, such an overhanging eminence is present in Strauss 3 (Fig. 15e, f). In Mumba 10 the feature is visible, but CT scans show no posteriorly protruding eminence (Fig. 15g, h).
The relatively deep cerebellar fossae in Eliye Springs (Fig. 16) are separated by a projecting but rounded, toruslike internal occipital crest, which is about 8 mm in breadth and exhibits general similarities to that of Kabwe, as illustrated and described by Seidler et al. (1997). For the Turkana Boy, a wide (>13 mm), low internal occipital crest (Walker and Leakey, 1993) and for OH9 a thick and rounded crest (Phlak, 1983) have been described, whereas Weidenreich (1943) stated that because of the smallness of the area of the cerebellar fossae in the Zhoukoudian H. erectus, such a crest is practically missing.
The internal occipital protuberance in Eliye Springs is not located close to the foramen magnum but about 36 mm apart from basion and at about the same level as the external occipital protuberance. Although the phylogenetic relevance of this feature combining endocranial and ectocranial aspects has been questioned (Bräuer and Mbua, 1992), no deviation from modern conditions can be seen in Eliye Springs's morphology or the location of the internal protuberance. The toruslike internal occipital crest divides into two ridges, which pass on either side of the foramen magnum. However, a so-called Vermian fossa, or depression, is not present in this hominid (Fig. 16). It is also lacking in the modern Tanzanian sample.
The intracranial venous sinus system is only partly visible in Eliye Springs (Fig. 17a, 16). The sulci for the sigmoid sinuses are well marked on both sides, whereas those for the transverse sinuses become continuously weaker toward the endinion region, where they are hardly traceable. This reduction in depth of the sulcus is also visible in the CT scans. Seidler et al. (1997) report on faint traces of the transverse sinuses in Kabwe and Petralona. According to Arsuaga et al. (1991) a so-called lateral sulcus, a deep groove near the asterion (where the sulcus runs from the occipital to the temporal, crossing the asterion region of the parietal) is missing in H. erectus from Zhoukoudian and often lacking in Neanderthals, whereas it is generally present in modern humans and also occurs on the early Middle Pleistocene parietal from Tighenif. Such a deep sulcus in continuation of the sigmoid sulcus is well developed on both sides in Eliye Springs (see Fig. 17a). In Strauss 2 and 3 as well as in Mumba 4 and 10 there are lateral sinuses crossing the asterion region (Fig. 17b–e).
Numerous researchers have dealt with interspecific differences in the degree of basicranial flexion and its relations to the position of the vocal tract, relative brain size, facial projection, locomotion, and other factors (Laitman et al., 1978; Ross and Ravosa, 1993; Ross and Henneberg, 1995; Spoor, 1997; Lieberman and McCarthy, 1999; Lieberman et al., 2000). Both ectocranial and endocranial landmarks were used. Studies of the ectocranial angulation between the hard palate and foramen magnum pointed to an apelike basicranial morphology in Plio-Pleistocene hominids, whereas a more flexed condition was found in H. erectus (Laitman and Heimbuch, 1982; Laitman, 1985). Quite a number of different endocranial angles have been suggested for studying basicranial angulation. Ross and Henneberg (1995) used a cranial base angle (CBA), which describes the angle between the clival plane and the planum sphenoideum. These authors found that, although hominids have more flexed basicrania than other primates, Australopithecus africanus,H. erectus and archaic H. sapiens do not differ significantly from modern humans. According to Spoor (1997), however, the planum sphenoideum (which does not include the cribriform plate) represents only a very short segment of the anterior cranial base. Thus, Spoor (1997) used the line from sella to foramen caecum for defining the orientation of the anterior fossa. In contrast to Ross and Henneberg (1995), Spoor found that A. africanus (Sts 5 = 147°) is significantly different from modern humans (mean= 137°), whereas the H. erectus cranium Sangiran 17 (129°) is not. With regard to this angle Lieberman et al. (2000) pointed out that Neanderthals and archaic H. sapiens fossils like Kabwe have considerably more extended basicranial angles than modern humans (about 15°). This also indicates that no single explanation, such as relative brain size, can account for the total variation of this angulation. It is beyond the scope of this paper to discuss in detail the various possible factors involved in the complex developmental processes of basicranial flexion. Nevertheless, and although there is probably “no best or most useful measure of the angle of the cranial base” (Lieberman and McCarthy, 1999; p. 512), the two endocranial angles discussed here were determined for Eliye Springs and the comparative specimens.
Although in Eliye Springs the dorsum sellae is largely broken off, the CT scans allow a reliable placement of the superior point for the clival plane (from basion to the point on the clivus before the dorsum sellae curves posteriorly). The planum sphenoideum is defined by the superior-most point on the sloping surface of the pit in which the cribriform plate is set and the most posterior and superior midline point on the tuberculum sellae (Ross and Ravosa, 1993; Lieberman and McCarthy, 1999). The CBA for Eliye Springs measures about 87° (Fig. 18b). Lieberman and McCarthy (1999) gave a mean of 112.6° ± 6.15° for an adult modern sample (n = 17), which is similar to the figure of 111.8° ± 7.43° (n = 93) provided by Ross and Henneberg (1995). The latter authors found a wide variation of 92–135° for their modern sample, and provided a value of 128° for Kabwe and estimations between 92.7 and 104° for OH9. For three Tanzanian crania we found values of 108.9° (Mumba 6), 115.4° (Mumba 4), and 129.6° (Strauss 2) (Fig. 18e and h). Eliye Springs' low figure for the CBA reveals a strong basicranial flexion close to the lower end of extant humans. On the other hand, however, this value also appears close to that of OH 9. Using the basicranial angle suggested by Spoor (1997), which is defined as the angle between the lines sella to basion and sella to foramen caecum, Eliye Springs yielded a value of 122.6° (Fig. 18c). The three modern Tanzanian specimens provided figures of 130.8° (Mumba 6), 139.3° (Mumba 4), and 156.5° (Strauss 2) (Fig. 18f and i). Lieberman and McCarthy (1999) gave a mean of 134.2° ± 3.1° for an adult modern sample (n = 16), which falls very close to the figure of 137° ± 4.9° (n = 48) provided by Spoor (1997). Thus, for this angle as well, Eliye Springs exhibits a small value lying below recent means and the figures of the modern sample from Tanzania, underscoring its strong basicranial flexion. Nevertheless, the value for Sangiran 17 determined by Spoor (1997) on a cast lies remarkably close to the recent data as well.
The frontal sinuses of Eliye Springs are partly defective due to the broken supraorbital torus, so the volume cannot be calculated with any certainty (Spoor and Zonneveld, 1999). Nevertheless, the lateral and superior extension can be assessed (Fig. 19a and b). Size and shape of the frontal sinuses show great variation in recent and fossil humans (Szilvássy, 1986; Spoor and Zonneveld, 1999; Seidler et al., 1997). Arsuaga et al. (1991) assume that frontal sinus development cannot be used to estimate phylogenetic affinities, at least in European human evolution. Eliye Springs exhibits well-developed frontal sinuses, which might have been restricted to the supraorbital torus region not invading into the squama in contrast to Kabwe or Petralona (Seidler et al., 1997). On the better-preserved left side, the sinus extends up to the medial third of the orbit, whereas on the right it extends roughly to midorbit. An extreme variation is also present among the five Tanzanian specimens reaching from tiny interorbital sinuses (Mumba 6) to very broad sinuses leading superiorly up to about midsquama (Strauss 2) (Fig. 19c and d).
In Eliye Springs the ethmoidal and sphenoidal sinuses could largely be cleaned from the matrix, but due to some destruction their full extent or volume cannot be determined. The sphenoidal sinus (Fig. 20a) is relatively large and extends on the left side into the greater wing of the sphenoid bone up to the level of the oval foramen, and posteriorly up to the superior third of the clivus (Fig. 20b). Analysis of the modern sample revealed that in Strauss 2 the sphenoidal sinus is also large extending laterally into the greater wing up to the same level as in Eliye Springs (Fig. 20c) and posteriorly well into the upper third of the clivus (Fig. 20d). In contrast, the sphenoidal sinus of Mumba 4 is smaller and does not extend into the wing or clivus (Fig. 20e and f). The sphenoidal sinuses of Petralona and in Kabwe (less strongly) continue from the body into the greater wings. The heavy pneumatization in Petralona is also characterized by an extension of the sinus into the medial part of the clivus and the temporal squama (Seidler et al., 1997).
Because of damage to the bony surface of the better-preserved left maxilla, virtual 3D reconstruction and stereolithographic modeling will be necessary before shape and volume of the maxillary sinuses can be estimated (in preparation). Although a larger portion of the alveolar process is missing, preliminary reconstruction points to the presence of a canine fossa.
Endocranial Affinities of the Eliye Springs Hominid
This study examined 35 endocranial features or morphological aspects in both the Eliye Springs cranium and anatomically modern specimens from Tanzania. The Eliye Springs hominid was found to lie well within, or close to, the modern ranges for nearly all of the traits considered. In only a few features did Eliye Springs indicate deviations from a more modern appearance. The endocranial volume of 1210 cc is small and even close to that of early archaic H. sapiens specimens, such as Bodo and Kabwe. The anterior cranial fossa also tends to be relatively small. No other feature of the anterior fossa exhibits a condition approximating that found in H. erectus or other nonmodern. The middle cranial fossa is also relatively short and close to the lower end of the range of variation of the modern sample. There are a few other conditions noted here that could represent nonmodern tendencies. These include: the anterior-posterior position of the center of the pituitary fossa, which seems to be located slightly further back than in modern samples; the relatively large vertical distance between the floor of the pituitary fossa and the Frankfurt Horizontal; and the somewhat greater angle between the superior and posterior surfaces of the petrous temporal bone. None of the features of the posterior cranial fossa shows any tendency of being marginal to or outside of the modern range of variation. The cranial base angulation is remarkably strong in Eliye Springs, even stronger than in those of the modern comparative sample. This holds true for the two different angles measured, the CBA (Ross and Henneberg, 1995) and the basicranial angle (Spoor, 1997). Although this might support the modern morphology, similarly small values have also been reported for H. erectus, but this has been based on limited fossil data. The extension of the paranasal sinuses shows no clear deviation from the large range found in modern samples. Finally, in spite of the likely occurrence of a canine fossa, as indicated by preliminary reconstruction, the maxillary sinuses appear to be rather well developed.
Since the few comparative modern African specimens revealed great ranges of variation, it can be expected that broader studies on larger samples will lead to still greater ranges of variation. Therefore, with the possible exception of endocranial capacity, the current list of features in which Eliye Springs shows tendencies to deviation from modern conditions can be regarded only as hypothetical, based on very limited material. Nevertheless, this first endocranial description of the archaic H. sapiens specimen from West Turkana points to a basically modern morphology.
Issues and Future Directions of Virtual Studies of the Endocranium
The present work has also revealed problems with regard to many of the features examined. Among those of the anterior cranial fossa, the vertical distance between the cribriform plate and the orbital roof and the variation regarding the slope from the orbital roof toward the cribriform plate has led to somewhat confusing results. Based on the small recent sample studied, Eliye Springs exhibits conditions that not only diverge from the suggested ancestral ones but also are even more divergent than those found in the recent modern sample. Another feature that turned out to be difficult to assess is the orientation of the crista galli. The extant variation of this variable crest cannot be adequately described by using only two categories, such as vertical or posterior orientation. The present study has also shown that the angle suggested by Seidler et al. (1997) to describe the inclination of the frontal lobes is problematic. The angle appears to vary considerably and can increase or decrease by just moving the parasagittal plane a few millimeters. According to the present study, Eliye Springs exhibits a very small value for this angle lying below the recent modern range of variation given by Seidler et al. (1997). On the other hand, one of the modern Mumba specimens has a much larger value of about 21°, close to those in Petralona and Kabwe. Lack of clarity can also be seen with regard to the location of the frontal lobes relative to the orbital cavities. No phylogenetic significance can be seen in the occurrence of a modernlike condition in Eliye Springs and the Ante-Neanderthal specimen Arago 21, or of an archaic pattern in another Ante-Neanderthal from Petralona and the African archaic specimen from Kabwe.
Turning to the middle cranial fossa, the relative location of the pituitary fossa could be of phylogenetic interest. Not only detailed metrical data on the size of the fossa but also its anteroposterior position can be well determined on virtual specimens. With regard to the latter feature, Eliye Springs showed an intermediate position between that described for Ngandong and those seen in the modern sample studied here, but this still remains dubious in view of the sparse data and the uncertainties in determining the feature on Ngandong by Weidenreich (1951). In addition, it would be interesting to study the vertical position of the floor of the pituitary fossa relative to different horizontal planes. Here, Eliye Springs shows a large vertical distance to the Frankfurt Horizontal lying above the range of a large modern sample studied by Lang (1981).
A number of potentially important but poorly understood features are related to the petrous temporal bone. Whereas the ectocranial angle of the petrous pyramid is an often-used trait, the endocranial angle of the petrous crest has rarely been considered. It is most surprising that Eliye Springs should have a H. erectus–like ectocranial angle combined with a fully modern endocranial angle. Considerable diversity also occurs with regard to the angle between the superior and posterior surfaces of the petrous bone. However, an exact definition of this angle is necessary before a reliable determination on virtual reconstructions can be carried out. Until such data are available, it remains unclear whether Eliye Springs's value of 93° approaches nonmodern conditions or just falls into the modern range of variation. Another feature that can be well examined only by CT-based 3D reconstruction is the incline from the superior petrous surface toward the bottom of the cranial fossa. Eliye Springs's condition is similar to those found in modern specimens, and it would be valuable to check Weidenreich's (1951) observation of a different course in Ngandong by using CT-based images. Another interesting feature of the petrous bone is the development of the arcuate eminence, which has been suggested as phylogenetically relevant. Although this trait is a small feature, CT-based data appear to be an adequate way to explore the occurrence and variation of it.
Further relevant features of the petrous bone belong to the posterior cranial fossa. Among these is the orientation of the posterior surface, which, of course, depends on the orientation of the whole cranium. Moreover, the level at which to measure the angle along the petrous crest has to be defined, as has the plane at which the angle is determined. It would be of interest to study the change of this angle at different planes along the petrous crest to determine the robusticity of this feature. All this can be done only on the virtual bone. A further interesting feature is the external opening of the vestibular aqueduct, which is said to be covered by either a plate or a distinct eminence. Eliye Springs exhibits a clear eminence as has also been described for the Zhoukoudian H. erectus and the Amud hominid (see above). However, our study indicates that such an eminence also occurs among the few modern crania examined. Thus, CT scans might help to assess this feature and also identify whether there are more than two distinguishable character states. A last important feature of the posterior fossa considered here is the lateral sulcus. It is present in Eliye Springs as well as in the modern Tanzanian sample. Indications point to its lack in Chinese H. erectus and most Neanderthals, but much unclarity still remains as to the occurrence of this trait in the different hominid taxa.
The present analysis and the points raised in this section clearly show that there are many questions regarding variation and phylogenetic relevance of endocranial features. Clear definitions and reproducible scoring systems are needed for many of these traits, especially if they can now be assessed by using virtual tools. Much of the earlier endocranial descriptions had to be based on fragmentary crania or fragments and, accordingly, need to be reassessed in light of current methodological possibilities. Large worldwide samples of extant humans, as well as samples of different fossil hominid taxa or morphs, have to be examined to provide a more robust database for the occurrence and variation of endocranial features. Only then we will be able to better understand the relevance of endocranial morphology to the assessment of hominid phylogenetic analyses.
We thank Meave Leakey and the Kenya National Museums for permission to do a CT analysis of the Eliye Springs hominid as well as Nicholas Conard for extending the loan of the Mumba/Strauss material. We are grateful to Fred Spoor for doing the scanning of the specimen for us and for his valuable comments on an earlier version of this paper. We thank Karl-Heinz Höhne and Hermann Zeumer for supporting this research, as well as Jean-Jacques Hublin for his kind cooperation. We also thank Maximilian von Harling, Hermann Müller, and Julia Sandberger for their assistance. Finally, we appreciate the useful comments on our manuscript as well as the advice and help by Jeff Laitman.