The evolution and development of cranial form in Homo sapiens

The majority of the previously proposed diagnostic cranial characters of AMHS listed in Table 1 and several other variables (see below) were measured by using external landmarks from several samples: 100 recent H. sapiens (50 of each sex) from five craniofacially diverse populations (from Australia, China, Egypt, Italy, and West Africa; for details, see ref. 17); 10 relatively complete Late Pleistocene fossils commonly classified as early AMHS (Cro Magnon 1, Jebel Irhoud 1, Liujiang, Minatogawa 1, Obercassel 1, Predmosti 4, Qafzeh 6, Qafzeh 9, Skhul V, and Zhoukoudian 101); and nine relatively complete crania typically assigned to AH (but not H. erectus) comprising five Neanderthals (Gibraltar 1, Guattari, La Chapelle aux Saints, La Ferrassie 1, and Shanidar 1), and four crania usually attributed to H. heidelbergensis or H. rhodesiensis (Bodo, Dali, Broken Hill, and Petralona). Fossil crania lacking the upper face were not included. All external measurements were taken from casts at the American Museum of Natural History (New York), with the exception of the recent human sample and Skhul V, which were taken from original specimens. Geometric morphometric comparisons of cranial differences between AH and AMHS were computed from two- and three-dimensional landmarks digitized from computed tomography (CT) scans of four adult fossil crania (Bodo, Broken Hill, Gibraltar 1, and Guattari) and four fairly robust recent adult male H. sapiens (two Australian, two Native American) from the National Museum of Natural History (Smithsonian Institution, Washington, DC). Two-dimensional landmarks were digitized from lateral radiographs of a longitudinal study of six male and six female recent H. sapiens from the Denver Growth Study (details in ref. 18) and from lateral radiographs of a cross-sectional sample of Pan troglodytes (details in ref. 18). All radiographs were compared at three ontogenetic stages: stage I, 50% through the neurocranial growth phase (≈3 years in H. sapiens and 1.5 years in P. troglodytes); stage II, at the end of the neurocranial growth phase (≈6 years in H. sapiens and 3 years in P. troglodytes); and stage III, adult (based on third molar eruption).

Factor analysis identifies combinations of variables that account for morphometric covariation among a given sample (19, 20). Although identified factors need to be further tested against a priori developmental models by using methods such as confirmatory factor analysis (21), exploratory factor analysis is a useful initial test of the hypothesis that a few structural modifications underlie much of the taxonomically important cranial variation in recent Homo. Factors were extracted from the AMHS as well as the combined AMHS and AH samples described above by using principal components analysis; both the initial factor solution and a varimax transformation were examined (20). Included variables quantify most of the previously proposed diagnostic characters of AMHS in Table 1: frontal angle, parietal angle, and occipital angle were measured following Howells (22); vault height relative to length was measured as basion-vertex/nasion-opisthocranion; vault width relative to height (measured as euryon–euryon/bregma–vertex) was substituted for bregma-asterion chord/biasterionic breadth (from ref. 1) because it better quantifies vault curvature in the coronal plane; canine fossa depth was measured as the maximum subtense between zygomaxillare and alare; supraorbital torus size/shape was quantified by using Lahr's system of grades (ref. 6, pp. 344–346); browridge length was calculated as the midsagittal distance from glabella to the bifrontomaxillare chord. Mandibular characters such as the chin and dental size measurements were not included in the analysis (see ref. 23).

ANOVA and comparison of sample ranges were used to test the hypothesis that structural changes identified by the factor analysis discriminate between AH and AMHS. Because of unequal sample sizes, ANOVA significance was determined conservatively by using Scheffé's F (19). Variables compared include the previously proposed diagnostic cranial characters of AMHS (Table 1) and three additional variables included on the basis of the factor analysis results (see below) that have recently been proposed as structural determinants of AMHS cranial form (5–7, 14, 24–27): neurocranial globularity, defined as the roundedness of the cranial vault in the sagittal, coronal, and transverse planes; facial retraction, defined as the anteroposterior position of the face relative to the anterior cranial base and neurocranium; and facial prognathism, defined as the orientation of the lower face relative to the upper face. Table 2 provides details of how these variables were measured and standardized. Facial retraction could not be estimated reliably from external measurements, and was measured from radiographs and/or computed tomography scans of the comparative AMHS samples and those fossils for which nasion–foramen cecum can be measured: Bodo, Broken Hill, Cro Magnon I, Gibraltar I, Guattari, La Chapelle aux Saints, Obercassel I, Petralona, and Skhul V.

To examine the structural and ontogenetic bases of differences in facial form between AH and AMHS, two morphometric analyses were calculated by using subsets of 17 landmarks digitized directly from computed tomography scans of fossil hominids measured ETDIPS (www.cc.nih.gov/cip/software/etdips/) and from radiographs of the ontogenetic samples of H. sapiens and P. troglodytes. Landmarks used were anterior nasal spine, basion, bregma, foramen cecum, glabella, lambda, nasion, opisthocranion, orbitale, pituitary point, posterior maxillary (PM) point, prosthion, sella, sphenoidale, the most inferoposterior midline point on frontal squama above glabella (frontex), and the midline points of greatest elevation between nasion and bregma (metopion), and bregma and lambda (see refs. 22 and 28 for landmark definitions). After Procrustes superimposition (29, 30), Thin Plate Spline (TPS) analysis (www.usm.maine.edu/%7Ewalker/; ref. 31) was used to visualize major differences in projected lateral view between taxa. Euclidean distance matrix analysis (EDMA) was also used to quantify significant differences in three-dimensional shape, by dividing all interlandmark lengths by a global geometric mean, and by using nonparametric bootstrapping (n = 100) to determine confidence intervals of 0.90 (α = 0.10) for each size-corrected linear distance (32, 33).

Orthogonal and varimax solutions of both the AMHS and combined AMHS and AH samples yield virtually identical results, indicating similar, statistically robust patterns of covariation among the diagnostic features of AMHS listed in Table 1. Fig. 1 summarizes the initial (untransformed) factor solution of the AMHS sample in which factors 1–3 explain 61% of the sample variance. Variables that contribute substantially to factor 1 (factor scores > 0.50) are parietal angle, occipital angle, vault height relative to length, and vault height relative to width. These moderately correlated variables (mean r = 0.40 for the combined sample) all quantify cranial vault curvature in the coronal, sagittal, and transverse planes. As noted above, neurocranial globularity has previously been proposed to be diagnostic of AMHS (5, 6, 17, 24). In contrast, browridge size and frontal angle contribute to most of the variation in factor 2, and canine fossa depth explains most of the variation in factor 3. Of these patterns of covariation, the association between browridge size and frontal angle (factor 2) is related structurally to facial retraction, another key proposed structural autapomorphy of AMHS (24, 25). It has been well established that primates with more retracted faces have smaller, shorter browridges with steeper frontal squamae, reflecting the supraorbital region's role to integrate spatially the upper face and the neurocranium (see ref. 14); linear measurements of facial retraction explain ≈80% of browridge length and frontal angle variation across primates and in ontogenetic samples of humans and chimpanzees (25, 34). The structural basis for canine fossa depth, which contributes most of the variation in factor 3, is less clear and requires further study. Variation in this feature may be a function of maxillary arch retraction relative to the zygomatic, but could also reflect maxillary sinus expansion into the infraorbital region.

Table 2 compares ranges and degrees of cranial variation for a number of features to test whether the two structural variables identified above, neurocranial globularity and facial retraction, discriminate between AH and AMHS better than the features traditionally thought to be diagnostic of AMHS. Although the mean values for all features in Table 1 differ significantly (P < 0.05) between the two samples, they do not completely separate AH and AMHS (see also refs. 7 and 8). Ranges overlap considerably for these variables, especially browridge size/shape and facial prognathism. However, measurements of facial retraction and vault globularity completely discriminate between the two taxa with no overlap (Table 2). Thus, as characters, neurocranial globularity and facial retraction appear to represent AMHS autapomorphies.

Variations in facial retraction are thought to be a function of interactions between several cranial components including facial size, cranial base angle, cranial base length, and brain size (14, 25–28). Likewise, variations in neurocranial globularity presumably derive from multiple interactions between portions of the brain and the size and shape of the cranial base (17, 28, 35). Thus, to better understand the origin of AMHS cranial form it is useful to identify more proximate variables that interact during growth to generate variations in facial retraction and neurocranial globularity among recent humans. As a preliminary effort, we first used TPS and EDMA analyses of landmarks that include major loci of cranial growth to compare the pattern of shape differences between adult AMHS and two taxa of AH: Neanderthals and African archaic Homo. The results, summarized in Fig. 2, not only highlight the above described differences in facial retraction and neurocranial globularity, but also reveal several important differences in facial and cranial base shape that provide clues about their structural and developmental causes. The most obvious difference is that the AMHS face is much smaller relative to overall cranial size than in either group of AH. According to the landmarks used here, facial reduction in AMHS appears to be concentrated in the upper face, with 10–15% decreases in both supero-inferior height and antero-posterior length relative to overall cranial size. AMHS also have smaller midfaces than Neanderthals but not the archaic Africans because of autapomorphic midfacial prognathism in Neanderthals (2).

At least three important differences in shape between the AMHS and AH samples are also evident in the cranial base. First, anterior cranial base length (e.g., from sella to foramen cecum) is ≈15–20% longer relative to overall cranial size in AMHS than in either taxon of AH. Second, the anterior cranial base (and with it the face) is more flexed relative to the posterior cranial base in AMHS (as indicated by arrows in Fig. 2; see also refs. 25 and 28). Average cranial base angle in AMHS is 134° (ref. 18), but is ≈15° more extended in Guattari and Broken Hill. Third, the EDMA analyses indicate that the middle cranial fossa in AMHS is ≈20% wider relative to overall cranial size, as shown by the distance between the midline of the sphenoid body and the poles of the temporal lobes (the PM points). These differences in relative cranial fossae dimensions suggest that the temporal (and possibly the frontal) lobes are proportionately larger in AMHS than AH.

Although the above analyses suggest that a few variables may underlie major cranial shape differences between AH and AMHS, further analyses are necessary to test whether and how growth differences explain these contrasting patterns, especially in terms of facial retraction and neurocranial globularity. Recent geometric morphometric comparisons (36) show that Neanderthal and AMHS crania have distinctive, ontogenetically early growth patterns that may result from shifts in basicranial and facial development. However, the available sample of infant AH crania is too small and insufficiently complete, particularly in the basicranium, to test directly the effects of facial size, cranial base flexion, anterior cranial base length, and middle and anterior cranial fossae size on cranial ontogeny. In addition, there are no well-preserved fossil Neanderthal crania with undistorted or complete cranial bases, and none younger than 2.2 postnatal years, by which time most cranial base growth (e.g., flexion) is complete (18).

An alternative, preliminary way to test the effects of facial diminution, cranial base flexion, anterior cranial base elongation, and expansion of the middle and anterior cranial fossae on facial retraction and neurocranial globularity in H. sapiens is to compare ontogenetic samples of human and nonhuman primates to test whether the same variables contribute to homologous structural differences. This hypothesis is supported by the analyses summarized in Fig. 3, which compare cranial ontogeny in humans and chimpanzees by using just the cranial base and facial landmarks from the analysis presented in Fig. 2. Fig. 3 shows that, in contrast to humans, facial retraction decreases during Pan ontogeny. During early postnatal ontogeny (between stages I and II), facial projection is associated with a decrease in the relative length of the anterior and middle cranial fossae and with an increase in relative facial length and height (Fig. 3A). After neural growth is complete in Pan (between stages II and III), the relative lengths of the cranial fossae continue to shorten, facial height and length continue to increase, and the cranial base extends, rotating the face dorsally relative to the neurocranium (Fig. 3B; refs. 18 and 28). In contrast, development of relative cranial base length, relative facial size, and cranial base angulation is different in H. sapiens ontogeny, as the neurocranium remains highly globular and the face stays retracted under the anterior cranial base. Between stages I and II (while the brain is still growing but cranial base flexion is complete; ref. 18), the posterior cranial fossa becomes relatively shorter as facial size remains constant relative to overall cranial size (Fig. 3C). As the human face increases in relative size (mostly inferiorly) between stages II and III, facial retraction decreases slightly, the anterior cranial fossa becomes relatively shorter, and the cranial base remains flexed rather than extending as it does in Pan (Fig. 3D). Basicranial flexion is important because it positions most of the face beneath the anterior cranial fossa. In conclusion, the major variables that apparently underlie differences in facial retraction and neurocranial globularity between AH and AMHS are the same ones that contribute to similar differences evident in human and chimpanzee cranial ontogeny: cranial base angle, the relative length and width of the cranial fossae, and relative facial height and length (Fig. 3E).