Curvilinear, geometric and phylogenetic modeling of basicranial flexion: is it adaptive, is it constrained?

Abstract

Prior work has shown that the degree of basicranial flexion among primates is determined by relative brain size, with anatomically modern humans possibly having a less flexed basicranium than expected for their relative brain size. Basicranial flexion has also been suggested to be adaptive in that it maintains a spheroid brain shape, thereby minimizing connections between different parts of the brain. In addition, it has been argued that the degree of flexion might be constrained such that increases in relative brain size beyond that seen in Australopithecus africanus were accommodated by mechanisms other than basicranial flexion. These hypotheses were evaluated by collating an extensive data set on basicranial flexion and relative brain size in primates and other mammals. The data were analyzed using standard least squares regression, geometric and curvilinear modeling, and phylogenetically independent contrasts (PICs). Geometric modeling does not support the hypothesis that flexion is an adaptation that facilitates enlargement of a spheroid brain. Whether humans have a less flexed basicranium than expected for their relative brain size depends on the phylogenetic vantage point from which it is evaluated. They are as flexed as expected for a descendant of the last common ancestor of the Paranthropus–Homo clade, but their degree of flexion cannot be predicted from the basal hominoid node, even if their relative brain size is specified. Humans undoubtedly occupy an unusual part of morphospace in terms of basicranial flexion and relative brain size, but this does not mean that their degree of flexion is or is not constrained. Curvilinear regression models and standard linear regression models describe the relationship between flexion and relative brain size equally well. Hypotheses that the degree of flexion is or is not constrained cannot be discriminated at present. Consideration of recently published ontogenetic data in the context of the interspecific data for adults suggests that much of the variance in basicranial flexion can still be explained as a mechanical consequence of brain enlargement relative to basicranial length.

Introduction

The degree of flexion of the midline cranial base in primates is of interest because the cranial base is integrated with many structural and functional systems in the head (Lieberman et al., 2000). Changes in the degree of flexion therefore may impact various functional systems. Several factors have been hypothesized to be important in determining the degree of flexion, including head and neck posture, facial orientation, and brain size. The hypothesis gaining the most attention inrecent years is Gould's (1977) suggestion that the degree of basicranial flexion is determined by the size of the brain relative to the length of the basicranium. Ross and Ravosa (1993)corroborated this hypothesis in interspecific comparisons across non-hominid primates. They found that as neurocranial volume relative to basicranial length increases, so the cranial base becomes more flexed (Ross and Ravosa, 1993). Subsequent work on adult primates has supported the general tenor of Gould's hypothesis, but Jeffery and Spoor's investigations reveal that in fetal primates increasing relative brain size is actually associated with decreased basicranial flexion, or flattening of the cranial base (Jeffery, 1999, Jeffery, 2002a, Jeffery, 2003; Jeffery and Spoor, 2002).

Unresolved questions fall into three groups: Is basicranial flexion adaptive or is it a mechanical consequence of brain enlargement? Are there phylogenetic effects on basicranial flexion and how do they impact on tests of the other hypotheses? Is basicranial flexion functionally constrained, as hypothesized by Ross and Henneberg (1995)?

Adaptive explanations for basicranial flexion link it to three functions: reducing the energy required to support the head on an orthograde neck, improving braincase strength by maintaining its spheroid shape in context of brain enlargement, or minimizing wiring length in the brain by maintaining a spheroid brain shape. Strait and Ross (1999)have addressed the postural hypothesis elsewhere. Taking relative brain size and head posture into account, they found no evidence to suggest that variation in flexion is related to variation in head and neck orientation. Instead, relative brain size was the more important determinant of flexion. Thus, although basicranial flexion is often regarded as some kind of postural adaptation, there are no data to support this hypothesis (see also Lieberman et al., 2000).

Is basicranial flexion a mechanism for accommodating increases in relative brain size while retaining a spheroid brain shape (Ross and Henneberg, 1995; Lieberman et al., 2000)? Two possible advantages of a spheroid brain shape have been suggested. First, a spheroid brain might optimize interneuronal communication by minimizing distances between neurons (Ross and Henneberg, 1995). Minimizing of “wiring length” has been suggested to be an important design principle of neural architecture (Allman and Kaas, 1974; Barlow, 1986; Mitchison, 1991; Cherniak, 1995; Van Essen, 1997) and maintenance of a spheroid brain by increased basicranial flexion may be a way to minimize “wiring length” at the level of overall brain shape. Second, a spheroid brain might result from selection for a spheroid braincase. A spheroid braincase provides the strongest and most economical housing for the brain (Ross and Ravosa, 1993) and a spheroid braincase with a flexed base also has the advantage of reducing stresses in the poorly stress resistant region of the basicranium rostral to the occipital condyles (Demes, 1985).

Here we test the hypothesis that basicranial flexion accommodates increasing brain size relative to basicranial length in order to keep the shape of the brain and braincase spheroid. Geometric models are used to generate a variety of predictions regarding relationships between basicranial flexion, basicranial length, and brain shape in the mid-sagittal plane if sphericity in the measured neurocranial volumes is being optimized. The primate data are compared with these models and the closeness of fit between them is used to evaluate the models.

As noted above, it has been suggested that ontogenetic data from Homo sapiens, Macaca nemestrina and Alouatta caraya falsify thehypothesis that increasing basicranial flexion is a mechanical consequence of increasing relative brain size (Jeffery, 1999, Jeffery, 2003; Jeffery and Spoor, 2002). Here we present all the available adult and ontogenetic data and address the issues raised by Jeffery and Spoor.

Recent comparisons of various ways of measuring flexion and relative brain size revealed taxonomic differences in patterns of variation in basicranial flexion measures. The definitions of the different measures are given in Table 1. Analysis of variance components at different taxonomic levels (Smith, 1994) reveals that for one of the measures of relative brain size and one of the measures of basicranial flexion (IRE1 and CBA4) most of the variance is found at the family level, followed by the superfamily and then the genus levels. In contrast, for the other measures of relative brain size and basicranial flexion (IRE5 and CBA1), most of the variance is found at the Infraorder level (Lieberman et al., 2000). Adjusting degrees of freedom to account for these taxonomic effects (Smith, 1994), IRE1 and CBA4 were found to be significantly correlated with each other at P<0.01, as were IRE5 and CBA1. However, correlations between IRE5 and CBA4, and between IRE1 and CBA1 were only significant at P<0.05 (Lieberman et al., 2000). Thus, the highest correlation coefficients and degrees of significance were found between variables that showed maximum variance at the same taxonomic levels. This suggests that phylogenetic effects can affect patterns of correlation between basicranial variables and that phylogenetic relationships among primate taxa need to be accounted for in evaluating hypotheses of relationship.

Following the publication of Felsenstein's seminal 1985 paper, techniques for incorporating patterns of phylogenetic relatedness into comparative analyses have been developed and made available in the form of computer software such as CAIC, COMPARE, and PDAP (Garland et al., 1993, Garland et al., 1999; Garland and Ives, 2000; Blomberg et al., 2003; Rohlf, 2002). In this study the PDTREE component of the PDAP software is used to test hypotheses regarding correlations and regressions between measures of basicranial flexion andrelative brain size using Phylogenetically Independent Contrasts (PICs). To explore the importance of phylogenetic relatedness in analyzing these data, the relationships between these variables estimated using “tip” data (data from the tips of the phylogenetic tree without regard to branching pattern) are compared with the relationships estimated using PICs. In addition, the descriptive statistic, K, is calculated to estimate the strength of the phylogenetic signal in each variable (Blomberg et al., 2003). Various branch length options available in PDTREE are explored to determine the relative importance of tree topology and branch length on the patterns of correlation and regression measured.

Ross and Henneberg (1995)extended the Ross and Ravosa (1993)study to hominins and found Homo sapiens to be less flexed than expected. They interpreted these results as suggesting that the degree of basicranial flexion is constrained by the functional necessities of pharyngeal anatomy. This possibility arises because increased basicranial flexion among anthropoids is associated with increased facial kyphosis (Ross and Ravosa, 1993), so that as the anterior cranial base flexes ventrally relative to the posterior cranial base, so the palate (and orbits) rotate ventrally (i.e., facial kyphosis increases). Ross and Henneberg argued that the functional exigencies of pharyngeal anatomy must limit such facial kyphosis, hence constraining basicranial flexion as well. They also suggested that the brain dorsal to the notochord might preclude retroflexion of the basicranium because this would impinge upon the space available for the brain, leading them to suggest that 180 degrees might be the upper limit for basicranial flexion.

Previous debates regarding these hypotheses of constraint have focused on the methods for quantifying basicranial length and flexion (Spoor, 1997; Lieberman et al., 2000; McCarthy, 2001). McCarthy (2001)showed that basicranial length is most appropriately measured as the sum of the distance from basion to sella to foramen caecum (Basicranial length 2, or BL2 of Lieberman et al., 2000) and relative brain size is best measured using Index of Relative Encephalization 5 (IRE5). When the Ross and Ravosa flexion angle (CBA4) is regressed against IRE5, the hypothesis of constraint is confirmed: Homo sapiens have a less flexed basicranium than predicted (Liebermanet al., 2000; McCarthy, 2001). However, when flexion is measured using the more traditional angle defined by lines joining sella to basion and to foramen caecum, Homo sapiens has the degree of basicranial flexion predicted for its degree of relative encephalization (Spoor, 1997; Lieberman et al., 2000; McCarthy, 2001). Thus, for a primate with its relative brain size, Homo sapiens' planum sphenoideum is less flexed than expected (see above) but its foramen caecum is positioned as predicted. The meaning of these results remains to be fully explored, but it does suggest that different parts of the primate cranial base are either influenced by different factors, or by the same factors but to different degrees (Lieberman et al., 2000).

To test the hypothesis of constraint suggested by Ross and Henneberg (1995)we investigate whether CBA4 of humans is unusual for a primate of their relative brain size. In the past this has been assessed by calculating reduced major axis (RMA) lines calculated for nonhuman primates, and using the position of Homo relative to the estimated 95% confidence limits of those lines to assess whether the mean value for Homo sapiens deviates significantly from the nonhuman primate line (Ross and Henneberg, 1995; Spoor, 1997; Lieberman et al., 2000; McCarthy, 2001). Here we use a method, presented by Garland and Ives (2000), for using PICs to determine whether a species deviates from an allometric relationship. Their method calculates the prediction intervals for traits in a hypothetical species attached to a specific part of the phylogenetic tree by a specific branch length. The prediction intervals are calculated using the standard errors of the least-squares regression (LSR) calculated with PICs. When Homo sapiens falls inside the prediction intervals, this is taken to corroborate the hypothesis that, given their position in the primate phylogenetic tree, their degree of basicranial flexion and relative brain size are not unexpected.

We also evaluate the hypothesis of constraint on flexion by investigating nonlinear models of the relationships between relative brain size and basicranial flexion. Various nonlinear equations describe lines with sigmoidal or logistic shapes, as predicted if the degree of flexion is limited at higher or lower relative brain sizes. We use software (CurveExpert 1.3) to search for the best curve to describe the data and compare these models with the standard linear models traditionally used in studies of basicranial evolution. Because Ross and Henneberg's hypothesis of constraint makes predictions regarding the relationship between flexion and relative brain size at small brain sizes, data on several species of nonprimate mammals are included in these curvilinear analyses.