Geometric morphometrics and paleoneurology: brain shape evolution in the genus Homo

Transcription

1 Journal of Human Evolution 47 (2004) 279e303 Geometric morphometrics and paleoneurology: brain shape evolution in the genus Homo Emiliano Bruner* Dipartimento di Biologia Animale e dell Uomo, Universita` La Sapienza, P.le A. Moro 5, Roma, Italy Istituto Italiano di Paleontologia Umana, P.za Mincio 2, Roma, Italia Received 13 November 2003; accepted 18 March 2004 Abstract Paleoneurology concerns the study and analysis of fossil endocasts. Together with cranial capacity and discrete anatomical features, shape can be analysed to consider the spatial relationships between structures and to investigate the endocranial structural system. A sample of endocasts from fossil specimens of the genus Homo has been analysed using traditional metrics and 2D geometric morphometrics based on lateral projections of endocranial shape. The maximum and frontal widths show a size-related pattern of variation shared by all the taxa considered. Furthermore, as cranial capacity increases in the non-modern morphotypes there is a general endocranial vertical stretching (mainly centred at the anterior ascending circumvolution) with flattening and relative shortening of the parietal areas. This pattern could have involved some structural stress between brain development and vault bones at the parietal midsagittal profile in the heavy encephalised Neandertals. In contrast, modern humans show a species-specific neomorphic hypertrophy of the parietal volumes, leading to a dorsal growth and ventral flexion (convolution) and consequent globularity of the whole structure. Brain tensors such as the falx cerebri have been hypothesised to represent one of the main physical constraints on morphogenetic trajectories, with additional influences from cranial base structures. The neurofunctional inferences discussed here stress the role of the parietal areas in the visuo-spatial coordination and integration, which can be involved in higher cerebral functions and related to conceptual thinking. Ó 2004 Elsevier Ltd. All rights reserved. Keywords: human evolution; endocranial morphology; fossil endocasts; brain shape * Dipartimento di Biologia Animale e dell Uomo, Universitá La Sapienza, P.le A. Moro 5, Roma, Italy. Tel. C ; Fax: C address: emiliano.bruner@uniroma1.it /$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi: /j.jhevol

2 280 E. Bruner / Journal of Human Evolution 47 (2004) 279e303 Introduction Considering the tight ontogenetic relationship between brain shape and skull bones (Moss and Young, 1960; Enlow, 1990), the endocasts from fossil crania are a useful source of information on hominid cerebral morphology and anatomy. This principle is the basis of paleoneurology, namely the examination and analysis of natural or artificial endocasts that reproduce details of the external morphology of the brain (Holloway, 1978; Falk, 1986, 1987; Bruner, 2003a). Although brain size has been the principal issue considered in the paleoanthropological literature, paleoneurology also focuses on specific anatomical traits, like vascular patterns (e.g. Kimbel, 1984; Falk, 1993; Saban, 1995; Grimaud-Herve, 1997), cerebral asymmetries (Le May, 1976; Holloway and De La CosteLareymondie, 1982), or specific circumvolutions (Grimaud-Herve, 1997). Brain shape variation has rarely been analysed, mainly because of the fragmentary nature of the fossil record. Furthermore, there are some difficulties in coding and quantifying endocranial morphology, because of the smooth geometry of the brain itself. A pioneering approach to endocranial shape variation in fossil hominids was performed using a polar coordinate stereoplotting technique to compare extant and extinct Hominoidea (Holloway, 1978, 1981a). This research suggested that a large amount of variation can be localised in the lower parietal areas in the extant taxa, and in the upper parietal districts considering the extinct groups. It has been hypothesised that early changes in the frontal areas in Australopithecus were followed by a late and gradual evolution of posterior districts in early human representatives (Falk et al., 2000). In earlier hominids the principal differences from apes may have been a reduction of the primary visual striate cortex, a reorganisation of the frontal lobe (mostly at the third frontal circumvolutions), and the expression of hemispheric specialisation (Holloway, 1995). Subsequently, the development of the posterior parietal cortex may have been related to an increase in visuospatial integration, sensory reception, and social communication. One of the most studied area in fossil endocasts is represented by the frontal lobes, because of their presumed role in higher cognitive functions and language. Generally, the frontal lobes are narrower in the most archaic Homo taxa, showing a clear encephalic rostrum (Grimaud-Herve, 1997). The lateral development of these structures in more encephalised hominids leads to a more pronounced expression of the Broca s cap, which is involved in speech potentialities (e.g. Aboitiz and Garcia, 1997; Cantalupo and Hopkins, 2001). The frontal midsagittal profile is less variable, without marked shape differences between Middle Pleistocene and modern humans (Bookstein et al., 1999). The position of the frontal lobes can be strongly influenced by a marked pneumatisation, which pushes backward the entire orbital plate in some robust Middle Pleistocene specimens such as Petralona and Kabwe (Seidler et al., 1997). At the opposite extreme, the occipital lobes project backward behind the parietal profile in more archaic brains, shifting under the parietal areas as brain size increases. Accordingly, the cerebellar structures are located under the occipital poles in archaic Homo erectus, under the parietals in more derived taxa and almost under the temporal areas in modern humans (Grimaud-Herve, 1997). The brain s maximum width is located at the temporal base in Homo erectus and other Middle Pleistocene groups, between the temporal and parietal areas in Neandertals, and at the parietals in anatomically modern humans (Holloway, 1980; Grimaud- Herve, 1997; Seidler et al., 1997). KNM-WT (dated to about 1.5 Ma) and the other specimens included in the Homo ergaster hypodigm show an endocast comparable to those of Asian Homo erectus, but with a less developed frontal diameter and without a marked occipital projection (Begun and Walker, 1993). A recent study on the shape variation of the endocast in the genus Homo showed two major patterns of variation (Bruner et al., 2003): an archaic structural trajectory shared by nonmodern taxa and characterised by an allometric vertical development, frontal enlargement, and parietal relative shortening, versus a modern pattern characterised by parietal development leading to brain globularity.

3 E. Bruner / Journal of Human Evolution 47 (2004) 279e In this paper, the endocranial shape of some fossil specimens included in the hypodigm of the genus Homo is investigated using traditional metric comparisons and two-dimensional geometric morphometrics performed on projected lateral views of the endocast. The aim is to consider in detail some specific variables of the human fossil endocasts, and to characterise the endocranial morphology in norma lateralis. This approach is used to check and improve previous results from three-dimensional analyses. The lateral view (and particularly the midsagittal vault profile) has been frequently considered in evolutionary studies because of its recognised availability, usefulness, and evolutionary meaning (e.g. Manzi et al., 2000b; Lieberman et al., 2002; Bruner et al., 2004). Also, projections and 2D data allow the use of type III landmarks (Bookstein, 1991), geometrically defined by chords or fractions of curvature. Type III landmarks have no biological meaning in terms of homology, but can be used to define structures or areas that lack clear anatomical references. These points can be extremely useful when considering endocranial shape, for which type I (i.e. homologous) and type II (structurally corresponding) landmarks are rare. The null hypothesis of this analysis is that human brain evolution was based on quantitative encephalisation, i.e., endocranial enlargement without change of the structural model. This hypothesis suggests that endocranial variation within the genus Homo was characterised by scaled (i.e. allometric) versions of a plesiomorphic morphological plan, described by a shared structural trajectory. Conversely, the recognition of group-specific patterns would suggest the presence of neomorphic processes. Materials and methods Comparative samples Data were collected from high-quality endocasts at the Museum of Anthropology Giuseppe Sergi (Dipartimento di Biologia Animale e dell Uomo, Universita` La Sapienza, Roma), the Istituto Italiano di Paleontologia Umana (IsIPU, Roma), and the Institut de Pale ontologie Humaine (IPH, Paris). As in previous analyses (Bruner, 2003b; Bruner et al., 2003), the specimens were included in three main groups in order to identify general patterns: 1. Archaic morphotypes (ARC): specimens referred to Homo erectus sensu lato (s.l.), namely the Asian Early and Middle Pleistocene remains (Homo erectus sensu stricto e s.s.) plus the specimens from Sale (Africa) and Arago (Europe). It is assumed that, even if the Asian specimens show some apomorphic or autapomorphic traits, they represent a less derived model with respect to Neandertals and modern humans. The Sale skull is not entirely interpreted, but it is assumed to represent a plesiomorphic pattern compared to more encephalised taxa of the genus Homo. Asymmetries at the nuchal ectocranial structure in Sale suggest a pathological congenital torticollis, related to atrophy of some nuchal muscles, and asymmetrical reduction of the surface insertions (see Hublin, 2002). In contrast, no marked asymmetry is shown at the internal surface, except for minor differences related to a common left occipital petalia tilting the midsagittal axis and the underlying cerebellar poles. Furthermore, the relative independence between inner and outer tables of the skull (involving epigenetic musclerelated traits) suggests that endocranial morphology is not strongly affected by the hypothesised ectocranial pathology. The Arago endocast is the result of a hybrid reconstruction composed of the Arago frontal and parietal (respectively Arago 21 and Arago 47) and the occipital from Swanscombe and the temporals from Sangiran 17. Therefore, it represents a heterogeneous but in any case less derived system with respect to Neandertals and modern humans. Clearly, all of these assumptions are based on a set of inferences about the phylogenetic status of these specimens, but they are necessary to produce a reference variability 2. Neandertals (NDR): European Wurmian Neandertals, here recognised as Homo neanderthalensis. A stereolithographic model of the virtual endocast of Saccopastore 1 (Bruner et al., 2002) has been included in this group. Although the

4 282 E. Bruner / Journal of Human Evolution 47 (2004) 279e303 fossils from Saccopastore are dated to about 120 Ka, they show Neandertal derived traits both for the ectocranial (Sergi, 1944; Condemi, 1992) and endocranial (Bruner, 2003b) morphology. The skull from Teshik-Tash represents a juvenile individual, but its ontogenetic stage is assumed not to affect the endocranial comparison. Brain growth is almost complete at this age, with the brain weight of modern human populations reaching 96% and 99.7% of the adult values at the puberty and adolescence respectively (Pen a- Melian, 2000). Furthermore, Neandertal cranial shape shows rate-hypermorphosis compared to modern humans, and the taxonomic differences are expressed early during ontogeny (Ponce de Leo` n and Zollikofer, 2001). According to its cranial development, Teshik-Tash, about 8.5 years old, is comparable to Qafzeh 11, approximately 13.5 years old (Ponce de Leo` n and Zollikofer, 2001). Therefore, the cranial shape must be assumed sufficiently comparable with the adult specimens. 3. Modern Humans (MOD): anatomically modern humans, from Late Pleistocene to recent populations. This group includes the endocast of Vatte di Zambana, recovered near Trento (Italy) in and dated to 8 ka (Corrain et al., 1976; Newell et al., 1979). A modern endocast from a recent Javan individual is also included in this group. It shows an extreme brachycephalic expression of the modern cerebral packing, presenting an almost globular structure that may be useful to understand the trends and variability of the Homo sapiens model. In Table 1 a list of the entire sample is provided, with information about the repository source, the reference group, and the labels used in the text. Univariate and bivariate metrics Fig. 1a shows the diameters employed in this analysis, and Table 2 provides a list of the endocranial variables, together with the respective labels and definitions. Lengths and widths were measured with a spreading caliper directly on the endocast. Chords and heights were measured by projecting the respective landmarks onto the left hemisphere in lateral view through a dioptograph. Table 1 List of the whole sample, with labels used in the text, reference group, and repository site Specimen Label Group Repository* Trinil 2 TRN2 ARC IsIPU Sangiran 2 SNG2 ARC IPH Salé SAL ARC BAU Sinanthropus III ZKD3 ARC BAU Sinanthropus X ZKD10 ARC IPH Sinanthropus XII ZKD12 ARC IsIPU Arago (rec) ARA ARC IsIPU Saccopastore 1 SCP1 NDR BAU La Chapelle-aux-Saints CHP NDR IsIPU La Ferrassie 1 FRS NDR IPH Teshik-Tash TST NDR IPH Guattari 1 GTT NDR IsIPU Feldhofer Grotto FLD NDR BAU Predmostì 3 PRD3 MOD IPH Predmostì 4 PRD4 MOD IPH Predmostì 9 PRD9 MOD IPH Predmostì 10 PRD10 MOD IPH Combe-Capelle CCP MOD IsIPU Vestonice 2 VST2 MOD IPH Vatte di Zambana VTT MOD IsIPU Recent Human Endocast RHE MOD IPH *BAU: Dipartimento di Biologia Animale e dell Uomo. IsIPU: Istituto Italiano di Paleontologia Umana. IPH: Institut de Pale ontologie Humaine. Endocasts were aligned using the plane passing through the frontal crest, the internal occipital protuberance, and the endovertex. Absolute values have been considered as well as relative values, which were divided by the maximum length of the endocast (averaged hemispheres). Differences between groups were analysed by Kruskal-Wallis analysis of variance. Regressions were analysed by the least-squares procedure and both Pearson s and Spearman s correlation coefficient. Statistical significance was set at p! A Major Axis regression has been computed by Model II (Legendre, biol/legendre). Repeated measures of diameters on the endocast sample revealed an average error of 0.6 mm, due mostly to uncertainty in localisation of type II (Bookstein, 1991) and fuzzy landmarks (Valeri et al., 1998). The maximum average error for repeated measures was 0.9 mm (a discrepancy less than 5%). Considering these limits in the metric resolution, the diameters were taken to the nearest millimetre.

5 E. Bruner / Journal of Human Evolution 47 (2004) 279e Fig. 1. a) Interlandmark distances sampled on the endocasts, in left lateral and upper view. The diameters have been measured directly on the endocasts, while chords and projections have been obtained using a dioptograph (see Table 2 for labels and definitions). b) landmark configuration used in the 2D geometric morphometric analysis (left hemisphere - see Table 3 for labels and definitions); for the PCA a reduced configuration has been used, considering only the vault morphology (landmarks between FP and OP). Geometric morphometrics Endocranial shape was also analysed by a landmark-based approach, according to the procedures of geometric morphometrics (Bookstein, 1989; Rohlf and Bookstein, 1990; Bookstein, 1991; Marcus et al., 1993; Rohlf and Marcus, 1993; Marcus et al., 1996). In geometric morphometrics, systems of coordinates are superimposed to minimise the size differences between specimens, with the aim of identifying the shape component of the total form (see Richtsmeier et al., 2002; Rohlf, 2003). A two-dimensional configuration of nineteen landmarks was selected to describe the endocranial lateral profile. Landmarks were chosen based on an operational homology definition (see Smith, 1990). Table 3 shows the list of landmarks, together with their labels and definitions. Fig. 1b shows the entire configuration. The left hemisphere was used, except for Combe Capelle that shows excessive damage on that side. A Generalised Procrustes Analysis (GPA) was performed, except in the 2D pairwise comparison in which a Bookstein superimposition was preferred. The first procedure superimposes the configurations through translation of the centroids, scaling to unitary size, and least-square rotation, while the second uses a reference baseline between two landmarks (Bookstein, 1991). For the Bookstein superimpositon the frontal and occipital poles were used as references for the baseline. This superimposition was used in the pairwise comparisons considering the hemispheric length a good index of quantitative encephalisation, useful to display differences in shape at the same endocranial antero-posterior development. Furthermore,

6 284 E. Bruner / Journal of Human Evolution 47 (2004) 279e303 Table 2 List of the interlandmark distances used in the metric analysis, with labels and definitions Label Measure Definition ML maximum length maximum hemispheric length from frontal to occipital poles - averaged hemispheres MW maximum width maximum width of the endocast, orthogonal to the midsagittal cerebral plane BW Broca width maximum width of the endocast at the base of the third frontal circumvolution H1 anterior height vault height at the anterior quarter - 25% - of the maximum length chord H2 middle height vault height at the middle - 50% - of the maximum length chord H3 posterior height vault height at the posterior quarter - 75% - of the maximum length chord VM vault module mean value between maximum length, maximum width and middle height FC frontal chord chord between the most anterior point of the frontal pole and the rolandic scissure at the midsagittal plane PC parietal chord chord between the Rolandic scissure and the perpendicular scissure at the midsagittal plane OC occipital chord chord between the perpendicular scissure and the torcular herophili at the midsagittal plane this configuration is mainly based on the anterior and posterior extremes. Anyway, in this analysis superimpositions using the baseline or the procrustes procedure gave the same results, more evidenced in the former approach. Centroid size is used as size index, defined as the square root of the sum of squared distances of a set of landmarks from their centroid (Marcus et al., 1996). Data were sampled using a dioptograph and digitised by tpsdig 1.20 (Rohlf, 1998a). The Principal Component Analysis (PCA) required a minimum statistical balance between specimens and landmarks, and the use of missing data was unsuitable. Therefore, to keep a sufficient fossil sample, only the landmarks from the vault (between FP and OP) were used (9 landmarks, 21 specimens). The vault configuration was used in a Cluster Analysis of individual specimens based on the Procrustes distance matrix and the unweighted pair-group method using arithmetic averages (UPGMA). As stressed before, this analysis is aimed at characterising phenotypes only in terms of shape, having landmarks a geometrical and structural meaning. The perpendicular scissure (i.e. the crossing point in lateral view between this scissure and the interhemispheric one) has not been used in the 2D configuration because of incompatibility between fixed homologous landmarks and shifting geometric landmarks. The sequence along the midline presents the perpendicular scissure situated between the 2 nd and 3 rd posterior chord projection (thus, between P2 and P3) in Homo erectus and Homo sapiens specimens. In contrast, in Neandertals and the specimen from Sale the perpendicular scissure lies before the midpoint of the posterior projection (thus, anterior to P2). This is probably related to the occipital projection in Homo erectus, and to the parietal development in Homo sapiens. This inversion leads to some visualisation difficulties in spatial deformation, but it is still a useful geometric indication that must be considered in future studies. Considering the landmarks used in these analyses, it must be stated that the terms parietal is used here to describe a surface that should overlap with the underlying cortical structure. When the Rolandic and perpendicular scissures are involved (as in the interlandmark metrics), this relationship is quite tight, and it is used to describe the presumed boundaries of the parietal lobes. Conversely, when the general outlines are considered (as in the geometric morphometric approach), parietal

7 E. Bruner / Journal of Human Evolution 47 (2004) 279e Table 3 List of the landmark used in the geometric morphometric analysis, with labels and definitions Label Landmarks Definition FP frontal pole the most anterior point of the endocast according to the maximum length RS rolandic scissure meeting point between the rolandic and interhemispheric scissures F1-F3 frontal projections meeting points between the endocranial outline and orthogonal chords drawn at fractions of the frontal chord (respectively at 0.25, 0.50, 0.75) OP occipital pole the most posterior point of the endocast according to the maximum length P1-P3 posterior projections meeting points between the endocranial outline and orthogonal chords drawn at fractions of the posterior chord (respectively at 0.25, 0.50, 0.75) TH torcular herophili point of maximum depression at the internal occipital protuberance TOC temporo-occipito-cerebellar meeting point between the cerebro-cerebellar scissure and the preoccipital scissure, between the 3 rd temporal and 3 rd occipital circumvolutions TV temporal valley posterior end of the temporal valley, between the temporal and cerebellar areas, at the angle between the temporal and cerebellar lobes ACC anterior cerebellar anterior point of the cerebellar outline, at the meeting with the temporal lobe PCC posterior cerebellar posterior point of the cerebellar outline, at the meeting with the occipital lobe C1-C3 cerebellar projections meeting points between the cerebellar outline and orthogonal chords drawn at fractions of the cerebellar chord ACC-PCC (respectively at 0.25, 0.50, 0.75) TP temporal pole point of maximum curvature of the temporal lobe BA Broca area point of maximum curvature between the pars opercularis and pars triangularis at the 3 rd frontal circumvolutions areas are used to describe surfaces that are not necessarily related to specific circumvolutions. Other functionally interesting structures - like the supramarginal gyrus or other specific circumvolutions - have not been included in this analysis because of the excessive uncertainty with respect to the available variation. For the supramarginal gyrus, problems arose in the exact localisation of the boundaries of this area, which is smooth and irregular. This gyrus is included in Wernicke s area and it represents one of the major sources of asymmetry in the human brain, generally being more strongly expressed on the left hemisphere. Therefore, cerebral dominance leads to a different location of the homologous counterpart on the right side. Once the gyrus has been localised, the choice of the landmarks to represent this area is not unequivocal, and often the centroid of the boss does not overlap with the point of maximum curvature of the circumvolutions. Structures like the supramarginal or the angular gyrus would be better considered only in group-averaged and hemisphere-averaged data. As cautionary note, it must be noted the bidimensional projections of volumes can be misleading if the geometric properties of the model are not considered carefully. In this 2D configuration, the results will be influenced by the different orientation of specific areas, as the cerebellar lobes. The more angled they are relative to the midsagittal plane, the shorter they will appear to be in a 2D projection. In any case, such an approach is necessary to consider geometrically derived landmarks from chords and projections and to characterise areas that lack clearly localisable points. It must be stressed that, because of the group-specific limited sample size, it is difficult to conduct between-group tests of significance. Instead, pairwise comparisons, cluster procedures, and multivariate approaches are used to characterise and describe the main phenotypic affinities among specimens. Morpheus et al. (Slice, was used to compute average shapes and pairwise comparisons. Applied Procrustes Software 2.3 (APS; Penin, and tpsrelw 1.18 (Rohlf, 1998b - edu/morph/) were used for the PCA. Procrustes distances were obtained by tpssmall 1.19 (Rohlf, 1998c). Cluster Analysis was performed by

8 286 E. Bruner / Journal of Human Evolution 47 (2004) 279e303 UPGMA using the Phylogeny Inference Package (PHYLIP, 3.57c; Felsenstein, html), while phenograms were computed with TREEVIEW (Page, zoology.gla.ac.uk/rod/treeview.html). Results Univariate and bivariate metrics Univariate distributions are reported for each main group by median and quartile values (Table 4). Considering the absolute diameters, the more derived groups (NDR and MOD) show significantly larger values for the vault module and maximum endocranial length when compared to the archaic average. Neandertals and modern humans show a significant larger value for the frontal chord, with modern humans also showing a larger parietal chord. Considering the relative values, Neandertals display a larger frontal and a shorter parietal chord, while modern humans have a relatively greater development of the parietal value. A bivariate plot of frontal and parietal chords shows that, moving from the archaic specimens (mostly the small individuals) to the Neandertals, an increase of the first value is not associated with an increase of the second. In contrast, the modern sample shows an increase in both variables (Fig. 2a). The vault heights show a clear pattern of increasing values from the archaic group to Neandertals to modern humans. The differences between NDR and MOD are significant for the posterior height (H3 - both absolute and relative values). The anterior and middle heights show significant differences for the smaller values of the archaic group (absolute data) or between archaic and modern humans (relative data), even if the trend is similar to that of posterior height. The heights show the strongest correlation with the vault module (r Z 0.87, 0.91, and 0.85 respectively). Considering this allometric relationship, the modern specimens seem to depart from a trajectory shared by the non-modern sample (Fig. 2b). The differences in the relative height Table 4 Median and quartiles for absolute and relative metrics in the three groups (absolute values are in mm) ML FC PC OC MW BW H1 H2 H3 VM BW/MW FCr PCr OCr H1r H2r H3r MWr BWr ARC med ( ( NDR med ( ( MOD med ( (

9 E. Bruner / Journal of Human Evolution 47 (2004) 279e Fig. 2. Bivariate comparisons. a) frontal chord vs. parietal chord. b) vault module vs. posterior height (H3). c) slopes (alpha) for the regression between vault module and heights (H1, H2, H3) for the non-modern (bold line) and modern (plain line) groups. d) maximum width vs. frontal width at the Broca s cap. The three groups are represented by circles (ARC), squares (NDR) and triangles (MOD). Values are in mm. See Table 1 for labels. values are attenuated but confirm the gradual increase between the main groups, and for the posterior height the differences are again pronounced. A full statistical approach to the speciesspecific regression coefficients has not been performed because of the limited sample size. Nevertheless, if we pool the archaic and Neandertal specimens into a non-modern group, the regressions between vault module and heights show a set of slopes distinctly different from the modern pattern (Fig. 2c). Interestingly, the withingroup scheme seems comparable (with middle height more sensitive to size), but with higher values in the modern group. Among the relative measures, heights are the only variables maintaining a significant correlation with the vault module (r Z 0.48, 0.58, and 0.58 respectively). The maximum width presents significantly greater values for moderns and Neandertals, slightly higher for the latter group. When the maximum width is corrected for the hemispheric length, the differences seem to fade, with a large range of variation in modern humans and slightly higher mean values for Neandertals. The modern range is strongly influenced by the rounded (wide and short) RHE endocast and the lengthened (thin and long) VST2 brain. The width at Broca s cap shows an absolute distribution very similar to the

10 288 E. Bruner / Journal of Human Evolution 47 (2004) 279e303 maximum width. With respect to the hemispheric length, the mean values of the two derived morphotypes are still above that of the archaic group. The ratio between the Broca and maximum widths expresses an archaic pattern, with a limited frontal diameter compared with the total breadth, and two derived morphotypes (Neandertals and moderns) with relatively enlarged frontal width. A Kruskal-Wallis test between archaic and nonarchaic values is significant at 0.08 level (which is indicative, considering the small number of specimens). A bivariate plot of MW vs. BW (Fig. 2d) suggests no clear species-specific departure from a shared trajectory. A correlation between these two diameters and the vault module shows coefficients of 0.84 (MW) and 0.91 (BW). Interestingly, a least square regression shows the frontal width scaling with a slope of 0.95, while the maximum (parieto-temporal) width scaling with a slope of Because of the small sample size, the 95% confidence intervals of these slopes are large, overlapping each other and both including isometry. The major axis computed on the same variables shows slopes respectively of 0.94 (maximum width vs. vault module) and 1.04 (Broca width vs. vault module), with the same considerations on the confidence intervals. In general, this second approach would be more appropriate to analyse the relationship between two variables but, because of the limited sample size and because of the likely non-normal distribution of the variation (which includes multiple taxa), it could be a less robust result (see Martin and Barbour, 1989; Legendre and Legendre, 1998). The regressions are all significant, but the coefficients cannot be properly analysed and should be considered only indicative. Generally, these data may suggest that as size increases the frontal diameter widens slightly faster than the maximum (temporo-parietal) breadth. The occipital chord is larger in Neandertals and modern humans, but the differences are not significant when the relative values are considered. Geometric morphometrics The PCA applied to the vault coordinates shows a polarised morphospace, in which the first two components account for 89% of the total variance (Fig. 3a). The 1 st component (69% of the total variance) is related to parietal development. At lower values the parietal outline flattens and shortens, with consequent relative lengthening of the vault. The frontal area enlarges and bends upward, while the occipital pole undergoes a vertical shortening. At higher values, the parietal area grows almost orthogonal to the midsagittal profile. The vault thus becomes higher and relatively shorter, with reduction of the most frontal surface and vertical development of the posterior districts. This first axis mostly separates modern humans, with a developed parietal outline and shortened frontal and occipital poles, from non-modern specimens, which show endocranial flattening at the parietal outline. The 2 nd component (20% of the total variance) involves a general stretching of the vault vertically at lower values and a vertical flattening on at higher values. The vertical stretching/flattening is particularly localised anterior to the Rolandic scissure, namely at the ascending frontal circumvolutions. This second axis separates Neandertals with relatively high vaults and developed frontal areas, from the archaic samples, which are more platycephalic and have a reduced frontal development. The modern sample is split by this component into more rounded profiles versus high and lengthened ones (mostly VST2 and CCP). A plot of the first two principal components shows a good separation of the three main groups (Fig. 3b), modern humans showing a developed parietal outline, Neandertals displaying high vaults and flattened parietals, and the archaic group with a low vault and flattened parietals. The following axes are related to small percentages of variability (!4%). When the first axis is plotted against the centroid size, two clusters are easily identified (Fig. 3c). The modern sample is separate, with large size and derived parietal growth. The other two groups lie contiguously with Neandertals having larger but slightly more flattened parietals. Therefore, in nonmodern specimens there is a change of size without marked changes of shape, while there is a morphological gap between modern and non-modern specimens in the relationship of shape to size. Considering the whole shape variation, both multiple regression and partial least-square

11 E. Bruner / Journal of Human Evolution 47 (2004) 279e Fig. 3. Vault shape 2D analysis. a) distortion grids showing the two extreme configurations along the PC1 and PC2 vectors (left hemisphere). PC1 is mainly characterised by parietal reduction/development, with consequent bending/convolution of the vault. PC2 describes a vertical stretching/flattening of the vault, localised mainly anteriorly to the Rolandic scissure (frontal ascending circumvolution) b) plotting of PC1 and PC2 values computed on the Procrustes residuals. c) PC1 values are plotted against centroid size (groups and labels like Fig. 2). regression onto centroid size do not find other allometric vectors. Considering just the archaic and Neandertal specimens, a PCA shows a first axis of variation weakly correlated with centroid size (partial least-square regression: r Z 0.64, p Z 0.02), separating archaic specimens from Neandertals, and involving (from smaller to larger configurations) parietal flattening and shortening associated with frontal vertical development (Fig. 4). Clearly, a larger number of specimens would be required for a useful statistical approach to the species-specific allometric pattern. A UPGMA applied to the Procrustes distances between specimens using this endocranial vault configuration shows the main separation to be between modern and non-modern samples (Fig. 5a). Within the non-modern group, Archaic and Neandertal specimens show two principal clusters that, however, do not represent exactly the two respective taxa. The Asian Homo erectus specimens show a close phenetic affinity to each other. Using the complete configuration, the consensus average shapes of the two more derived groups were compared to the mean archaic configuration

12 290 E. Bruner / Journal of Human Evolution 47 (2004) 279e303 Fig. 4. PC1 values from a PCA including only the non-modern specimens (ARCCNDR). The grids show the relative distortion in the two opposite directions of the morphological vector. basal lengths and the development of a more rounded morphology. The relative endocranial length decreases, with relative shortening of the frontal and occipital poles. The cerebellum is pushed downward by this deformation. A further note can be added regarding the position of the Rolandic scissure. Generally, in non-modern specimens the Rolandic scissure lies behind the midpoint of the maximum length, and the middle vault height (H2) crosses the profile approximately at the frontal ascending circumvolution. In contrast, in modern humans the middle height position approaches the Rolandic scissure, or even lies behind it, at the ascending parietal circumvolution. by superimposition of the fronto-occipital baseline (Fig. 5b). ARC vs. NDR. The entire structure stretches vertically in NDR, and the frontal area enlarges. The occipito-cerebellar complex shifts forward. There is increased development of the temporooccipital surface. The grid generally describes a vertical development, with an upward bending of the vault. The endocast undergoes a slight dorsal flexion by means of an upward shifting of the extremities and a retrograde movement of the upper landmarks. The internal occipital protuberance (and the whole posterior district) shows a geometric compression, as a structural meeting point of forces deriving from the dorsal bending and vertical stretching. The confluence of sinuses becomes less deepened under the flattening occipital poles. ARC vs. MOD. The vault enlarges and develops vertically in MOD, mostly because of an extreme enlargement of the parietal contour. The vault development involves an opposite bending with respect to the previous pattern, namely a convolution of the whole structure. The term convolution is used here not to describe an increase of the cerebral gyrification, but to describe a dorsal (parietal) growth that involves a globularisation process of the entire structure. While the parietal enlarges, the temporo-cerebellar areas approach the frontal poles. The extremities turn downward by the action of the parietal pressure, approaching each other with shortening of the Discussion Methodological considerations Paleoneurology generally has limitations because of the poor preservation of fossils, the low prevalence of brain traces on the inner table, a very small sample size, and the heterogeneous training of the few scholars working on this issue (Holloway, 1978). Metric tools in paleoneurology present several additional challenges. First, cerebral landmarks are seldom clearly localisable, because of the fuzzy nature of the whole structure. Fuzzy landmarks have been defined as the position of a biological structure that is precisely delineated, but occupies an area that is larger than a single point in the observer s reference system (Valeri et al., 1998). However, a within-observer error test proved that the uncertainty of fuzzy landmarks is limited and can be reduced by experience; moreover, literature studies report errors comparable with the values reported in this analysis for the more localisable landmarks (Valeri et al., 1998; Free et al., 2001). Type I landmarks (biologically homologous) are rare, and most of the suitable references can be included in type II, such as points of maximum curvature, projections or depths. Many of the endocranial landmarks are often hard Type II, i.e. defined by smooth surfaces or strongly dependent on variables like the orientation of the

13 E. Bruner / Journal of Human Evolution 47 (2004) 279e Fig. 5. a) UPGMA phenogram computed on the Procrustes distance matrix from the geometric morphometric analysis of the vault; b) distortion grids showing the pair-wise comparison between averaged configurations after Bookstein superimposition (baseline: FP- OP): ARC compared to MOD (top), and ARC compared to NDR (bottom). Reference shape (ARC) in bold. specimen or even the handling of the surfaces. Actually, both orientation and perspective play a crucial role during the localisation of some points. For the same reasons, there are few and weak references available to build geometric constructed landmarks, such as chords or projections. Asymmetries play an additional confounding role, making landmarks and diameters extremely dependent upon hemispheres and leading to easy misinterpretation of morphological relationships. One of the consequences of hemispheric asymmetry is the curvature of the interhemispheric scissure, which is the only reference available to localise the midsagittal plane. Therefore, the only plane available to localise landmarks and relative diameters is rather biased by the direct or indirect influence of cerebral dominance. Incompleteness of specimens, damage, subjectivity of the reconstructions, and limits of moulding techniques must be taken into account within this context. Thus, a constant anatomical uncertainty will necessarily be involved in any data set, and any inference will necessarily be limited by some sort of morphological doubt. Even the rare type I landmarks are not immune from subjective decisions. Points such as the Rolandic or the transverse scissure at the midsagittal plane are seldom easily recognised on an endocast, and a certain degree of experience and personal intuition is required. Moreover, they represent the boundaries of some circumvolutions, and therefore are often expressed not as points but as areas, whose geometry shows a high interindividual variation. Experience can limit random error (within-observer), but the results may be influenced by a certain degree of systematic discrepancy (among-observer). Two more sources of biases must be considered. First, the correspondence between endocranial and brain morphology is not complete and univocal, and the endocasts represent only a part of the cerebral structures (Mannu, 1911; Kimbel, 1984; Zollikofer and Ponce de Leo n, 2000). Second, cranial capacity and brain size do not overlap entirely, as cranial capacity includes a subarachnoidal space of about 300 cc (see Pen a-melian, 2000). These biases are obviously related, with the second being the main cause of the first. Therefore, much caution has to be used when inferences about brain morphology are made from endocast

14 292 E. Bruner / Journal of Human Evolution 47 (2004) 279e303 analyses, which mainly describe the variation of endocranial features and not the original cerebral structures. All the noise generated by these variables affects the resolution power of the paleoneurological approach, as well as any landmark-based analysis on cerebral living tissues (Free et al., 2001). The control and consideration of these properties represent the operational bases of paleoneurology. Data must be calibrated on the resolution power available, and results must be discussed only on the basis of a robust approach, taking into account the real numeric meaning of physical differences. It must be remarked that the methodological notes described in this context should be considered not as limits of the approach, but as a priori parameters of the models used in this analysis. They represents limits when ignored, confounding and producing noise in the results. In contrast, they become intrinsic components of the model if taken into account, improving the range of available analysis. Clearly, a more complete and resolute formalisation of brain geometry must be one of the main future targets of paleoneurologists. In this paper, univariate and bivariate metrics and 2D geometric morphometrics generally support and improve the results obtained from a multivariate interlandmark approach and 3D analyses (Bruner et al., 2003). A comparison between the two studies is useful to reinforce the results considering the possible biases of the paleoneurological approach. The data from 3D analysis of the endocasts are largely based on type I landmarks, related to a specific biological meaning, but more scattered on the surface and associated with a larger uncertainty related to the exact localisation of the structures (e.g. the perpendicular scissure). In contrast, the present analysis is based largely on type III landmarks, and a lateral projected bidimensional geometry. Thus, information is not associated with specific functions but simply with overall shape and morphology. Furthermore, in the present paper the complementary metrics from interlandmark distances strengthen the results because they contribute to a synthetic approach, including information from the third dimension (frontal and maximum widths), and considering the vault morphology in terms of frontal and parietal chords (type I-based landmarks) as well as function-independent variables (type III-based vault heights). Considering the convergence of the results from two- and three-dimensional approaches, the use of both geometric and anatomical landmarks are necessary to a full complementary discussion on endocranial shape evolution. Human brain evolution in the Middle Pleistocene Figure 6 synthesises the major differences between four indicative specimens, namely the Sale configuration (African Middle Pleistocene) warped into Zhoukoudian III (Asian Middle Pleistocene), La Chapelle-aux-Saints (European Wurmian Neandertal) and Vatte di Zambana (European Mesolithic modern human). They represent just indicative morphotypes, and the comparisons are not necessarily intended to be phylogenetic representations. Sale is considered as reference because of the general plesiomorph endocranial (Holloway, 1981b) and ectocranial (Hublin, 2002) morphology, associated with its geographical localisation. The patterns of variation described in this analysis will be discussed according to each main endocranial districts. Frontal area. The frontal lobes have been suggested to have a major role in human evolution, because of an interesting development in those morphotypes hypothesised to be the earliest representatives of the genus Homo, and formerly included in the hypodigm of Homo habilis (Holloway, 1995; Tobias, 1995). In the frontal lobes the coronal diameters increase in the more derived taxa, mostly by means of the development of the 3 rd frontal circumvolutions (Grimaud-Herve, 1997). Actually, what seems more interesting is the development of the frontal width compared to the maximum (parieto-temporal) breadth. This ratio increases in larger endocasts, following an allometric pattern shared within the whole genus. In larger hominids it leads to an allometric expression and development of the Broca s cap, while in the opposite direction leads to the typical frontal narrowing of the archaic phenotypes.

15 E. Bruner / Journal of Human Evolution 47 (2004) 279e Fig. 6. Distortion grids showing differences in 2D left lateral configuration between the hypothesised less derived phenotype (Salé) towards three representative specimens (Zhoukoudian III, La Chapelle aux Saints, Vatte di Zambana), with a draw of the respective endocasts. Note that this comparison is aimed at a description of the main morphotypes, and does not represent any phylogenetic hypotheses. Concerning the frontal vertical development, the anterior midsagittal brain profile has been hypothesised to have undergone a long stasis since the Middle Pleistocene, independently from the large variations of the outer bony morphology related to the brow ridge and associated structures (Bookstein et al., 1999). The analyses of the interlandmark and landmark data here show a general size-related vault elevation, particularly at the level of the ascending frontal circumvolution, with the archaic Asian specimens showing more pronounced flattening. This pattern is combined with a backward shifting of the Rolandic scissure, with consequent relative lengthening of the frontal lobe. Comparable results were obtained by superimposition of averaged threedimensional configuration of the endocasts (Bruner et al., 2003). Therefore, even if only some minor variations are recorded in the frontal profile, its shape is nevertheless influenced by size and by a generalised vertical stretching, and absolute stasis is improbable considering the encephalisation processes. Conversely it may be stated that, if the frontal shape differences in extinct human taxa are considered only as the consequence of brain enlargement, it should be possible to hypothesise an allometric stasis, as shape variations along a plesiomorph sizedependent morphological trajectory. The term allometric stasis is used here to indicate

16 294 E. Bruner / Journal of Human Evolution 47 (2004) 279e303 evolutionary changes mainly based on a nonderived structural system, thus shape variations without major neomorphic relationships. The morphology changes, while the biological model does not (at least concerning a large percentage of variability). In any case, in modern humans the generalised endocranial convolution involves a relative shortening of the frontal pole that reduces the effect of the allometric stretching. This result, although secondarily induced by the parietal development, must also be taken into account. It should be emphasised that in this study the Asian variability was included, and the frontal area was entirely considered (until the Rolandic scissure). In Bookstein et al. (1999) the Asian taxa were not considered and only the shape of the anterior cranial fossa was taken into account. Therefore, in the present analysis the Asian specimens may have increased the size-related differences, which moreover may be situated more at the frontal ascending circumvolutions than in the prefrontal cortex housed in the anterior cranial fossa. This volume is further reduced by the postorbital (backward) shifting of the frontal lobes in some robust Middle Pleistocene specimens (Seidler et al., 1997). Moreover, in previous work the inner (endocranial) profile was analysed together with the outer (ectocranial) profile (Bookstein et al., 1999). Extreme variation in the latter profile may have obscured the more limited variation in the endocranial shape. It is worth noting that recent volumetric analyses show a comparable development of the frontal lobes in modern humans and apes, considering their specific brain size (Semendeferi et al., 1997; Semendeferi and Damasio, 2000), suggesting an allometric stasis within the Hominoidea and supporting the hypothesis of a generalised evolutionary inertia of the frontal lobes. This may diminish the importance of these areas during human evolution compared to what has been historically assumed. However, it cannot be excluded that the trespassing of some size boundaries could have involved some discrete (emergent) changes in neuro-functional organisation. Considering the role of the frontal circumvolutions in higher cognitive functions and language, this hypothesis needs to be thoroughly investigated. Parietal area. There is some evidence suggesting that changes in the parietal lobes can be involved in the evolution of early hominids, considering the information available on the Australopithecinae endocasts (Holloway, 1983, 1995; Tobias, 1995). The lower parietal areas are extremely variable in extant Hominoidea, while the upper parietal structures show wide variation in extinct hominids (Holloway, 1981a). It seems than the parietal areas maintained a key role in the subsequent radiation of the genus Homo, particularly when modern humans are considered and compared to the extinct taxa. In general, the whole parietal and temporo-parietal complex widens in different lineages as brain size enlarges, along with an associated decrease of the occipital area (Grimaud- Herve, 1997). All the taxa considered here show a common pattern of size-related parieto-temporal widening. In contrast, as the brain enlarges the parietal outline shows a negative allometric pattern, leading to a relative shortening of the midsagittal profile and flattening of the upperposterior areas. The posterior height shows a sizerelated vertical development of the vault along a shared trajectory in the non-modern sample. The high level of encephalisation of the Neandertals exaggerates these patterns, with parietal areas high and wide, but short and flattened. It must be noted that posterior height (H3) and parietal flattening are not fully comparable variables, and they should not be confused. The parietal development concerns shape, i.e. a more rounded or flattened outline, while its height concerns size, i.e. the absolute elevation above the fronto-occipital chord. Multivariate traditional metrics has been used to characterise two different morphological trajectories based on this inverse relationship between brain size and parietal structure, separating modern from non-modern specimens (Bruner et al., 2003). As a result of relative parietal shortening, the brain undergoes a generalised dorsal bending as it enlarges, and the endocast is flexed upward. It is useful to evaluate the relationship between the inner and outer cranial surfaces, particularly considering the ectocranial lambdoid flattening of Neandertals. The Neandertals ontogenetic trajectory, compared with the pattern expressed in