Climate Signatures in the Morphological Differentiation of Worldwide Modern Human Populations


  • Mark Hubbe,

    1. Instituto de Investigaciones Arqueológicas y Museo, Universidad Católica del Norte, San Pedro de Atacama, Chile
    Search for more papers by this author
  • Tsunehiko Hanihara,

    1. Department of Anatomy, Kitasato University School of Medicine, Sagamihara, Japan
    Search for more papers by this author
  • Katerina Harvati

    Corresponding author
    1. Department of Human Evolution, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
    2. Department of Early Prehistory and Quaternary Ecology, University of Tübingen, Tübingen, Germany
    • Department of Early Prehistory and Quaternary Ecology, University of Tübingen, Rümelinstrasse 19-23, 72070 Tübingen, Germany
    Search for more papers by this author
    • Fax: +49-7071-295717


The ability of cranial morphology to reflect population/phylogenetic history, and the degree to which it might be influenced by environmental factors and selection pressures have been widely discussed. Recent consensus views cranial morphology as largely indicative of population history in humans, with some anatomical cranial regions/measurements being more informative on population history, while others being under selection pressure. We test earlier findings using the largest and most diverse cranial dataset available as yet: 7,423 male specimens from 135 geographic human population samples represented by 33 standard craniometric linear measurements. We calculated Mahalanobis D2 for three datasets: complete cranial dataset; facial measurement dataset; and neurocranial measurement dataset; these morphological distance matrices were then compared to matrices of geographic distances as well as of several climatic variables. Additionally, we calculated Fst values for our cranial measurements and compared the results to the expected Fst values for neutral genetic loci. Our findings support the hypothesis that cranial, and especially neurocranial morphology, is phylogenetically informative, and that aspects of the face and cranium are subject to selection related to climatic factors. The Fst analysis suggest that selection to climate is largely restricted to groups living in extremely cold environments, including Northeast Asia, North America, and Northern Europe, though each of these regions appears to have arrived at their morphology through distinct adaptive pathways. Anat Rec, 2009. © 2009 Wiley-Liss, Inc.

The degree to which human cranial morphology reflects population history or adaptive and developmental changes related to environmental conditions is the subject of ongoing scientific discussion in anthropology. Our understanding of the evolutionary processes affecting morphological variation directly impacts both the study of modern human geographic diversity as well as the interpretation of the human fossil record. The relevance of cranial morphology to phylogenetic reconstruction has been questioned in the past (see e.g., Collard and Wood, 2000), and convergence, parallelism, reversals, and environmental plasticity are commonly cited as obscuring any phylogenetic signal that it may preserve. The effects of mastication and climatic adaptation on the cranium are thought to be particularly extensive, with some anatomical regions or measurements believed to be affected differentially by these processes (e.g., Hylander, 1977; Carey and Steegmann, 1981; Olson, 1981; Beals et al., 1983; Skelton and McHenry, 1992; Larsen, 1999; Wood and Lieberman, 2001; Lieberman et al., 2004; Roseman, 2004; Roseman and Weaver, 2004; González-José et al., 2005; Harvati and Weaver, 2006a, b; Sardi et al., 2006).

On the other hand, several studies have demonstrated a geographic structure in modern human craniometric diversity on a global level (e.g., Howells, 1973, 1989; Hanihara, 1996). Craniometric data have been found to follow a common geographic pattern with genetic markers, including both classical and microsatellite DNA markers (Relethford, 1994, 2004a, 2009; Manica et al., 2007; Betti et al., 2009). These findings have been interpreted as resulting from an isolation-by-distance model of evolutionary diversification. Furthermore, population relationships inferred from cranial morphology (as reflected both by traditional linear measurements and by 3D geometric morphometric data) have been shown to match those inferred from genetic data (Roseman, 2004; Harvati and Weaver, 2006a, b; Smith, 2009).

Taken together, these results suggest that human cranial morphology preserves a relatively strong population history signal, in addition to a climatic and possibly also dietary/masticatory signal (e.g., Relethford, 2004a). In view of the prevalence of cranial morphology in interpreting the human fossil record, however, it is imperative to elucidate the effects of these processes on the cranium and to determine as much as possible which anatomical regions or measurements reflect adaptive or developmental changes and which can be reliably used to reconstruct phylogeny. Here, we follow up on previous work on the climatic and population history signature in the morphological diversity of modern human population using a modern human craniometric dataset (Hanihara, 1996, 1997) several times larger than the Howells dataset (1973, 1989) and other datasets used in most previous analyses (e.g., Harvati and Weaver, 2006a, b; Smith, 2009). The goals of our study are as follows: 1) to explore this unique dataset for patterns of correlation with geography and climate such as those found in analyses of smaller datasets (e.g., Howells, 1996; Roseman, 2004; Harvati and Weaver, 2006a, b; Smith, 2009); 2) to determine which of the linear measurements analyzed present higher inter-regional diversity than would be expected due to purely stochastic microevolutionary processes; 3) whether these measurements are correlated to climatic variables; and finally 4) to identify the geographic regions responsible for the differentiation seen in each of these variables.


Our sample comprises 7,422 male modern human crania representing 135 geographic populations. A list of the populations used, sample size, geographic coordinates, and climate parameters obtained for each is presented in the Appendix. Populations were selected from the larger dataset collected by one of us (TH) according to sample size: only samples larger than 15 individuals were included in the analysis. Females were excluded because far fewer population samples were measured for females relative to males, resulting in much smaller female total sample. In a few cases, closely related populations were pooled to maximize sample size (see Appendix for details). All crania were measured by one observer (TH) and each individual is represented in this study by 33 linear measurements (Table 1). For each population, geographic coordinates were established to the closest reference point possible and climatic variables were obtained from BIOCLIM database using the DIVA-GIS software (Hijmans et al., 2004). Climatic variables collected include temperature-related as well as humidity and rainfall measurements (Appendix).

Table 1. Linear measurements included in this study
  • a

    Variables included in the neurocranium analysis.

  • b

    Variables included in the facial analysis.

Maximum cranial length (GOL)aBrauer, 1988
Nasion-opisthocranion (NOL)aBrauer, 1988
Cranial base length (BNL)aBrauer, 1988
Maximum cranial breadth (XCB)aBrauer, 1988
Minimum frontal breadth (M9)aMartin and Saller, 1957
Maximum frontal breadth (XFB)aBrauer, 1988
Biauricular breadth (AUB)aBrauer, 1988
Biasterionic breadth (ASB)aBrauer, 1988
Basion–bregma height (BBH)aBrauer, 1988
Sagittal frontal arc (M26)aMartin and Saller, 1957
Saggital parietal arc (M27)aMartin and Saller, 1957
Saggital occipital arc (M28)aMartin and Saller, 1957
Nasion-bregma chord (FRC)aBrauer, 1988
Bregma-lambda chord (PAC)aBrauer, 1988
Lambda-opisthion chord (OCC)aBrauer, 1988
Basion prosthion length (BPL)Brauer, 1988
Breadth between frontomalare temporale (M43)Martin and Saller, 1957
Bizygomatic breadth (ZYB)Brauer, 1988
Nasion prosthion height (NPH)bBrauer, 1988
Interorbital breadth (DKB)bBrauer, 1988
Orbital breadth (M51)bMartin and Saller, 1957
Orbital height (OBH)bBrauer, 1988
Nasal breadth (NLB)bBrauer, 1988
Nasal height (NLH)bBrauer, 1988
Palate breadth (MAB)bBrauer, 1988
Mastoid height (MDH)Brauer, 1988
Mastoid width (MDB)Brauer, 1988
Frontal chord (M43(1))bMartin and Saller, 1957
Frontal subtense (No 43c)bBräuer, 1988
Simotic chord (WNB)bBrauer, 1988
Simotic subtense (SIS)bBrauer, 1988
Zygomaxillary chord (ZMB)bBrauer, 1988
Zygomaxillary subtense (SSS)bBrauer, 1988

Before analysis, size correction was performed by dividing each measurement by the geometric mean of the individual (Darroch and Mosimann, 1985). Missing values were replaced by multiple regression estimations. The estimated values in this case correspond to the expected value of the missing variable (used as the dependent variable in the multiple regression) based on the remaining observed variables of that individual (which are used as the independent parameters of the regression). One multiple regression is performed for every variable that has missing values in a dataset that has all missing values replaced by the global mean of each variable and the predicted values are subsequently used to replace the missing one. This procedure is preferable to simple mean replacement because of its relatively limited effect on the observed variance of a variable and on the covariances among variables. Multiple regressions were performed in Statistica 7 (Statsoft Inc., 1984–2007).

To test for patterns of correlation among morphology, geography, and climate, we calculated morphological, geographic, and climate distance matrices among all 135 groups (see Relethford, 2004a; Harvati and Weaver, 2006a, b). Morphological distances were calculated as Mahalanobis squared distances (Mahalanobis, 1936), which represent the morphological distance among groups, scaled by the inverse of the pooled within-group covariance matrix. In order to test the findings of previous studies showing that different cranial regions preserve population history and climatic signatures differentially (Harvati and Weaver, 2006a, b; Smith, 2009), three morphological distance matrices were calculated: one for the whole skull, one considering only facial variables, and the last one considering only neurocranial variables (Table 1). As genetic data are not available for all the populations included in this study, we used geographic distances as a proxy for genetic relationships, assuming a pattern of diversification through isolation by distance in accordance with the findings of Relethford and others (e.g., Relethford, 2004a; Manica et al., 2007; Betti et al., 2009). Geographic distance consisted of the linear distance between groups in kilometers, using Cairo, Bangkok, Bering and Panamá as check-points to limit the distances to terrestrial routes; and climate matrices are simply the differences among series for each climate variable. Distance matrices were compared through Mantel Correlation tests (Mantel, 1967), using NTSYSpc, version 2.10t (Rohlf, 1986–2000).

Our second objective was to determine which linear measurements show greater inter-regional diversity than would be expected due to stochastic microevolutionary processes alone. For this purpose, we calculated minimum Fst estimates (Relethford and Blangero, 1990; Relethford, 1994) for the whole set of variables and for each variable independently, using the 135 populations as the units of analysis. Fst offers a measure of the amount of variation found between the units of analysis in relation to the total variation of the sample expected under neutral evolution (Relethford, 1994). In other words, Fst gives an estimate of the amount of the variation that is a result of the differences observed between units of analysis (e.g. groups, populations). As such, it allows the assessment of the general influence of the subdivision of the data into groups in the overall variance observed. The Fst estimate has been originally devised for genetic data; however, it can also be derived from phenotypic data such as craniometric dimensions (Relethford and Blangero, 1990; Relethford, 1994) when the variance matrix for these data is proportional to the variance matrix of the genetic information underlying the phenotypic trait. As heritability (i.e., proportionality between phenotypic and genetic data) of craniometric measurements is usually relatively high (Devor, 1987; Carson, 2006), Fst estimates of such data offers a good parameter to compare the apportionment of intergroup variation seen for craniometric variables. Fst values were calculated in Excel, through a Macro written in Visual Basic by André Strauss (Instituto de Biociências, Universidade de São Paulo, Brazil), who allowed its usage for this study. We chose to analyze individual variables, rather than Principal Components, because we consider their interpretation to be more straightforward. Following the work of Relethford (1994, 2002) and the suggestion of Roseman and Weaver (2004), we accepted an Fst value of 0.3 as the threshold of variability that can be accepted as the product of stochastic microevolutionary processes alone. All Fst estimates considered here were based on an equal heritability value of 0.55, following previous studies (Devor, 1987; Relethford, 1994; Roseman and Weaver, 2004). Heritability values are used to correct Fst estimates so that it better reflects the proportionality between the phenotypic and the genetic variance matrices, and would present better estimates of Fst for those traits were the true heritability is known. Although Carson (2006) and Martínez-Abadías et al. (2009) published individual heritability estimates (h2) for some of the variables used here, the use of Carson's results for those variables found to have high Fst values with h2 = 0.55 did not change significantly the results (data not shown). Since both these studies found most h2 values to be smaller than the one considered here (0.55), Fst estimates using their values would be much higher than the ones presented here. Thus, an h2 of 0.55 must be seen as a conservative value in this case.

As a next step, we correlated those variables with high Fst values with the climatic variables, to see if the high intergroup diversity observed in these measurements can be explained by some of the environmental factors considered here. Because of the large sample size in this study, a very conservative alpha of 0.001 was adopted and so the climate correlations found are also considered conservative.

For the last objective of identifying the geographic regions responsible for this differentiation, the 135 populations were grouped in 15 geographic regions (Table 2; Fig. 1). For each region, Fst estimates were calculated for the groups in it, to see if our subdivisions into regional groupings could be considered coherent units from the viewpoint of internal variability. We then performed a jackknife procedure, removing all populations of one region at a time and recalculated the Fst for those variables with values above the 0.3 threshold. This procedure allowed us to evaluate the effect of each region in the apportionment of between-group variation.

Figure 1.

Geographic regions in which the series included in the study were grouped. Polynesia includes all dark dots of the Pacific Ocean not included in a gray ellipsis.

Table 2. List of geographic regions, number of populations included, and Fst estimates for each
RegionNumber of populationsFst (h2 = 0.55)
East Africa40.0843
East Asia80.1132
North Africa30.0368
North and Central America140.0997
Northeast Asia30.1042
North Europe180.0673
Extreme North America120.1681
South Africa40.0638
South America40.0873
South Asia170.101
West Africa60.0486


Geographic and Climate Correlations

The results of the Mantel correlations tests comparing morphological distances with geographic and climate variables (Table 3) support previous findings (Harvati and Weaver, 2006a, b; Smith, 2009) that, when considered as a whole, cranial morphology shows a strong correlation with geographic distance (r = 0.38; P = 0.0001). Significant correlations were also observed with some climate variables, especially those reflecting temperature (r ranging from 0.27 to 0.39, all significant at the 0.0001 level). When only neurocranial and only facial measurements were considered, the former showed a higher correlation coefficient with geographic distances (r = 0.40) and a lower one with temperature variables (r from 0.20 to 0.31), while the latter showed the inverse pattern (r = 0.29 for geography and ranging from 0.25 to 0.40 for climate; Table 3).

Table 3. Mantel correlations results between morphological distances and geographic and climate distances
Geographic and climate variablesAnatomical region
Whole skullNeurocraniumFace
  • a

    Relative humidity correlations are based on 132 populations, since data for some Polynesian islands could not be retrieved.

Geographical distancer = 0.3769r = 0.4033r = 0.2925
p = 0.0001p = 0.0001p = 0.0001
 Annual minimum temperaturer = 0.3509r = 0.2502r = 0.3736
p = 0.0001p = 0.0001p = 0.0001
 Annual maximum temperaturer = 0.3943r = 0.3099r = 0.3938
p = 0.0001p = 0.0001p = 0.0001
 Annual average temperaturer = 0.3863r = 0.2917r = 0.3957
p = 0.0001p = 0.0001p = 0.0001
 Maximum temperature of the warmest monthr = 0.2738r = 0.2549r = 0.2573
p = 0.0001p = 0.0001p = 0.0001
 Minimum temperature of the coldest monthr = 0.3083r = 0.1967r = 0.3421
p = 0.0001p = 0.0001p = 0.0001
 Temperature ranger = 0.1466r = 0.0764r = 0.1830
p = 0.0002p = 0.0245p = 0.0001
 Annual rainfallr = −0.0142r = −0.0211r = −0. 436
p = 0.4125p = 0.3613p = 0.1323
 Rainfall of the wettest monthr = 0.0117r = −0.0295r = 0.0149
p = 0.3636p = 0.2404p = 0.3176
 Rainfall of the driest monthr = −0.0155r = 0.0161r = −0.0890
p = 0.4196p = 0.6471p = 0.0085
 Relative humidityar = −0.0174r = 0.0180r = −0.0186
p = 0.3856p = 0.3321p = 0.3485

Apportionment of the Variation of Craniometric Dimensions

Table 4 lists the Fst values calculated for the entire set of craniometric measurements and for each of the variables separately. The value obtained (0.21 for h2 = 0.55) for all measurements considered together, representing overall cranial morphology, is similar (if a little higher) to Fst values published previously for cranial morphology at a worldwide level (Relethford, 1994, 2002). This result shows that increasing the number of population samples, as well as the total number of specimens considered in the analysis, does not affect the observed pattern of apportionment of craniometric variation. Of the Fst values calculated for each measurement separately, 10 of the 33 were relatively high (>0.3). The variables with high Fst values are concentrated in the facial region, and reflect facial height (NPH), breadth (ZYB, ZMB) and projection (BPL), and the dimensions of the nasal aperture (NLB, NLH, and SIS). Only three neurocranial measurements showed high Fst values, and all reflect cranial breadth (XCB, XFB, and AUB). From these variables, those associated with cranial breadth and with nasal height have previously been correlated with temperature by Roseman (2004) and thus suggested to be influenced by selection to cold climatic conditions.

Table 4. Fst estimates for the variables included in the study
VariablesFst (h2 = 0.55)Fst (h2 = 1.00)
  1. Variables with Fst estimates higher than 0.3 are shown in bold.

All Variables0.21170.1287
Maximum cranial length (GOL)0.27710.1741
Nasion-opisthocranion (NOL)0.28430.1793
Cranial base length (BNL)0.23350.1435
Maximum cranial breadth (XCB)0.43780.2999
Minimum frontal breadth (M9)0.18010.1078
Maximum frontal breadth (XFB)0.38120.2531
Biauricular breadth (AUB)0.49340.3488
Biasterionic breadth (ASB)0.27300.1712
Basion –bregma height (BBH)0.27650.1737
Sagittal frontal arc (M26)0.18200.1090
Saggital parietal arc (M27)0.19750.1172
Saggital occipital arc (M28)0.14640.0862
Nasion-bregma chord (FRC)0.15820.0937
Bregma-lambda chord (PAC)0.22210.1357
Lambda-opisthion chord (OCC)0.14940.0881
Basion prosthion length (BPL)0.31920.2050
Breadth between Frontomalare temporale (M43)0.28320.1785
Bizygomatic breadth (ZYB)0.43680.2990
Nasion prosthion height (NPH)0.40170.2697
Interorbital breadth (DKB)0.25170.1599
Orbital breadth (M51)0.26530.1657
Orbital height (OBH)0.28010.1703
Nasal breadth (NLB)0.35000.2285
Nasal height (NLH)0.36360.2391
Palate breadth (MAB)0.23650.1456
Mastoid height (MDH)0.20380.1234
Mastoid width (MDB)0.18300.1097
Frontal chord (M43(1))0.23630.1454
Frontal subtense (No 43c)0.27750.1744
Simotic chord (WNB)0.26570.1660
Simotic subtense (SIS)0.34810.2270
Zygomaxillary chord (ZMB)0.36080.2369
Zygomaxillary subtense (SSS)0.25080.1555

Climate Correlations of Variables with High Amounts of Intergroup Variation

The correlation coefficients and probability values obtained between the cranial measurements with high Fst values and the climate variables compiled for this study are reported in Table 5. The strongest relationship was found with temperature variables, which were strongly correlated with (in order of greatest r) facial height (NPH; r ranging from −0.57 to −0.71), nasal height (NLH; r ranging from −0.53 to −0.64) and biauricular breadth (AUB; r ranging from −0.55 to −0.65), as well as with facial breath (ZYB; r ranging from −0.45 to −0.56), nasal breadth (NLB; r ranging from −0.38 to −0.48), and the other two neurocranial breadth measurements (XCB, XFB; r ranging from −0.38 to −0.47 and from −0.33 to −0.41, respectively). Relative humidity was also correlated with breadth dimensions, both in the face (ZYB; r = 0.40) and in the neurocranium (XCB and AUB; r = 0.33 and 0.38, respectively). Finally, precipitation and temperature range appear correlated to nasal breadth, as well as, more weakly, to some of the other dimensions, mostly in the face (BPL, NPH, NLB, and SIS for precipitation; BPL, NPH, NLH, and AUB for temperature range). The only variable with high Fst which does not show a significant correlation with climate variables was one reflecting mid-facial breadth (ZMB).

Table 5. Correlations coefficients and associated probabilities between the variables with high Fst values and the climate parameters included in this study
Craniometric variablesTemperatureTemperature rangePrecipitationRelative humidity
Annual minimum temperatureAnnual maximum temperatureAnnual average temperatureMaximum temperature of warmest monthMinimum temperature of coldest monthAnnual RainfallRainfall of wettest monthRainfall of driest month
  1. Significant correlation at α = 0.001 are shown in bold.

p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p = 0.015p = 0.122p = 0.187p = 0.166p < 0.001
p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p = 0.012p = 0.012p = 0.061p = 0.006p = 0.003
p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p = 0.195p = 0.128p = 0.861p < 0.001
p = 0.109p = 0.325p = 0.192p = 0.268p = 0.030p = 0.001p < 0.001p < 0.001p = 0.002p = 0.109
p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p = 0.023p = 0.886p = 0.829p = 0.359p < 0.001
p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p = 0.002p = 0.001p = 0.182p = 0.013
p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p = 0.008p < 0.001p < 0.001p = 0.364p = 0.075
p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p = 0.256p = 0.175p = 0.917p = 0.003
p = 0.685p = 0.980p = 0.848p = 0.124p = 0.607p = 0.103p < 0.001p < 0.001p = 0.183p = 0.017
p = 0.075p = 0.064p = 0.066p = 0.026p = 0.079p = 0.466p = 0.034p = 0.002p = 0.492p = 0.017

Influence of Each Region on the Apportionment of Variation

With the exception of Extreme North America, all regions showed fairly low Fsts (Table 2), lower than the Fst estimate obtained for the entire dataset, justifying the geographic criteria adopted for the grouping of the populations into regions. The high Fst found for Extreme North America may be due to the large geographic distances observed among the samples it included. However, it was kept as a region due to the impossibility of further subdivision.

Figure 2 shows the results of the jackknifing procedure of the regions in calculating Fsts values for each of the 10 variables with a high proportion of intergroup variation (i.e., the variables listed in bold in Table 4). The columns depicted in the figure present the value of Fst obtained after removing the geographic region indicated under each column, as well as the Fst obtained originally with all groups included. To facilitate the reading of the graphs, the Fst values have been sorted in increasing order. As a way to assess the significance of the values obtained after removing each region, the graphs also show in darker gray those values that fall bellow the 95% confidence range of the original Fst. The confidence limit in this case was calculated as the value of the Fst for all region −1.96 times the standard deviation of the Fst values obtained with the jackknifing procedure. As can be observed, in all cases at least the lowest Fst value obtained is below this confidence limit. In no case was the highest Fst obtained in these analyses above the 95% confidence limit (original Fst + 1.96 times the standard deviation). This suggests that the reduction in Fst observed is more accentuated than its increase.

Figure 2.

Effects of the removal of each of the geographic regions in the Fst estimates of the variables presenting high between-group variation apportionment (Fst > 0.3). The black columns show the Fst obtained with all regions included. Dark gray columns represent those regions whose Fst value falls bellow the 95% confidence of the jackknife procedures (refer to text for more details).

The pattern that emerges from these analyses regards the impact of the northernmost regions on the Fst values of these measurements. As can be observed, for eight of the variables, the regions that contribute most for the intergroup diversification (i.e., the regions that by being removed reduce the Fst values most) are the northernmost (Northern Europe, Northeast Asia, or Extreme North America). However, while the removal of Northeast Asia and Extreme North America always appear to affect the Fst of the same variables, the removal of Northern Europe mostly affects the Fst of different measurements. Since Northeast Asia and Extreme North America behave in a similar manner, the effect of their removal together is also plotted in Fig. 2. These two regions contribute most to the intergroup differentiation seen in facial and nasal height (NPH, NLH), facial breadth (ZYG), and biauricular breadth (AUB), (Fig. 2f,h,c,d, respectively). The removal of Northern Europe, on the other hand, affected most the Fst values for frontal breadth (XFB), facial projection (BPL), and nasal breadth (NLB; Fig. 1b,e,g, respectively). These three geographic regions together also affected the Fst observed for midfacial breadth (ZMB; Fig. 1j), the only variable that did not show a strong correlation with the climate parameters included in this study.

The two remaining variables (XCB and SIS) show a strong contribution of regions other than the three northernmost. Maximum cranial breadth is mostly influenced by Australia and Melanesia (although only this first is below the 95% confidence range) and simotic chord is influenced mostly by South Africa and North Europe (both below the 95% confidence range).

To summarize, the jackknife analyses performed suggest that most of the variables with a high amount of intergroup variation which are strongly correlated with climate variables have a significant part of their differentiation driven by populations living in the northernmost regions of the planet. Also, it appears that morphological differentiation of these northernmost populations was not uniform, because the measurements under the influence of Northern Europe are a distinct set from those influenced by Northeast Asia and Extreme North America.


The results presented here indicate a strong relationship between cranial morphology and geographic distance, similar to that found by Relethford (2004a) in a much smaller population and total sample. Our findings further suggest a climatic signature on the morphological diversity of modern human groups, evidenced by both the morphological distances among group centroids and by the apportionment of between-group variation. Our study agrees well with previous work on limited geographic and total samples indicating that both isolation by distance and climate adaptation explain cranial morphological variation in modern humans (Relethford, 2004b; Roseman 2004; Harvati and Weaver, 2006a, b; Smith, 2009).

Our results also confirm that different anatomical regions are affected differentially by population history, here assumed to be closely related to geographic distance, and by environmental conditions, as previously suggested (Roseman, 2004, Harvati and Weaver, 2006; Smith, 2009). Although the nature of the linear measurements we used here did not allow as fine a subdivision of the cranium as is possible with geometric morphometric data (see Harvati and Weaver, 2006a, b; Smith, 2009), we were able to partition our data into neurocranial and facial measurements, and to consider these in our analysis in addition to the complete cranial dataset. We found that, while a larger part of neurocranial variability can be explained by geographic, rather than climatic, distances, facial morphology exhibits the opposite pattern. If geographic distances can be used as a proxy to genetic distances and population history in situations where isolation by distance is an important microevolutionary force, as has been suggested for modern humans (Relethford, 2004b), this result implies that neurocranial variability reflects population history to a greater degree than climate, in agreement with Harvati and Weaver (2006a, b; but contra Smith, 2009). Nonetheless, the facial measurements in the present study were found to be correlated, albeit more weakly than neurocranial measurements, with geographic distances as well as with climate. This result supports the recent finding by Smith (2009), who reported that facial morphology was correlated with neutral genetic distances (contra Harvati and Weaver, 2006b).

In agreement with these findings, our Fst analysis of individual measurements suggests that roughly one third of the craniofacial measurements studied are associated with higher Fst values than would be expected as result of stochastic microevolutionary forces alone. Most of these variables are concentrated on the face and are correlated with some aspects of climate. However, the high Fst values observed in these variables were mostly influenced by populations living in the northernmost regions of the planet, i.e., extreme North America, Northeast Asia, and Northern Europe. This result agrees with previous studies that found no correlation between climate and cranial morphology once Arctic samples were removed from the analyses and that suggested that climatic adaptation was observed in populations living under extreme cold conditions (Roseman, 2004; Harvati and Weaver, 2006b).

Most of the high Fst variables showing a correlation with climate were located on the face, with a few also reflecting neurocranial breadth, again agreeing with previous work (Coon et al., 1950; Carey and Steegmann, 1981; Beals et al., 1983; Franciscus and Long, 1991; Roseman, 2004; Roseman and Weaver, 2004; Harvati and Weaver, 2006a, b). They comprise measurements of facial breadth and height, as well as the nasal dimensions and simotic subtense. The cranial breadth measurements, facial height and breadth, and nasal dimensions were correlated with temperature variables; while nasal breadth and, to a lesser extent, nasal height, simotic subtense, and facial projection were correlated with rainfall variables. Two of the three cranial breadth measurements and facial breadth were correlated with relative humidity. These measurements were also found by Roseman (2004) to be related to climate.

The results of the jackknife procedure of the regions suggest that, although the northernmost populations are those that are maximally affecting the Fst values of the variables with high degrees of interpopulation variability, these regions influence different sets of measurements, and therefore different aspects of cranial morphology. While Northeast Asia and extreme North America affect mostly facial height and breadth, nasal height and inferior cranial breadth, Northern European populations influence facial projection, nasal breadth, and frontal breadth. These results suggest that populations from these regions followed distinct adaptive pathways to their cold environment. These different pathways are evident in two sets of variables: the nasal dimensions and the neurocranium breadth. In all three extreme cold regions included here, adaptation to cold climate in nasal shape seems to have favored small breadth/height relations of the nasal cavity: extreme North America, Northeast Asia, and Northern Europe show the lowest nasal index of all 15 regions. Yet, the path by which small nasal indexes have been achieved is different for each region. While Northern Europe populations exhibit the smallest mean nasal breadth, Northeast Asia and extreme North America show the largest mean nasal height. Both strategies result in smaller nasal indexes. The same pattern is observed for the facial height, which suggests that this measure might be reflecting the nasal height that is included in it.

Similarly, when cranial breadth is considered, all three northernmost regions (together with East Asian series) are characterized by wide neurocrania, although the regional differentiation is associated with distinct measurements: while Northern European groups are mainly affecting the Fst of maximum frontal breadth, Northeast Asians and Extreme North Americans are affecting the Fst observed for biauricular breadth. These differences can be interpreted again as distinct pathways toward wider braincases, which would be advantageous for populations in extremely cold climates under the assumption of Bergman's and Allen's rules (Beals et al., 1983, 1984): By increasing neurocranium breadth, globularity of the braincase is increased and the ratio surface/volume reduced, thus reducing heat loss through the surface of the skull. These changes toward a wider skull could be related to changes in basicranial morphology, which seems to be influencing vault morphology (Lieberman et al., 2000), as well as to the increase of general brain size (Beals et al., 1984). However, the lack of basicranial measurements in our dataset limits at this moment our capacity to further explore these hypotheses.

The distinct adaptive pathways suggested here could be the result of the constraints on the morphology of each population derived from its unique population history; or, alternatively, as a result of differences in the complex interplay among the various climatic parameters (e.g., temperature, humidity, rainfall, seasonality, etc.), only some of which could be included in this study. It is also possible that some of the measurements with greater-than-expected Fst values, particularly those on the face, are additionally influenced by other types of adaptation, most importantly related to dietary or masticatory practices. While dietary factors were not included in our study, the extreme masticatory and paramasticatory behavior of some extreme North American populations is well documented (e.g., Applet et al., 2000), and could be contributing to the observed pattern of high population diversity. This could be particularly relevant for those variables related to the masticatory complex, such as bi-zygomatic breadth, which is influenced by the size of the masticatory muscles (temporalis and masseter).

Our results also have some implications for the interpretation of the Neanderthal distinctive morphology, which has in the past been interpreted as resulting from climatic adaptation (e.g., Coon et al., 1950; Sergi, 1958). However, it has been observed that Neanderthal facial and especially nasal morphology does not necessarily match expected climatic patterns seen among modern human populations, and is sometimes described as a “paradox” (e.g., Holton and Franciscus, 2008; though see Márquez and Laitman, 2008 for an example of wide nasal apertures in cold climates in macaques). Alternative explanations for the evolution of the Neanderthal face have focused on functional demands associated with paramasticatory activities (e.g., Rak, 1986; Trinkaus, 1987) or on random genetic drift (e.g., Howell, 1952; Hublin, 1998) as driving forces. A recent study by Weaver et al. (2007) could not reject genetic drift as responsible for the evolution of Neanderthal cranial morphology, though selection was not excluded particularly for some aspects of facial anatomy. Here we have shown that different cold-adapted modern human populations do not show identical morphological patterns in response to their environments. Instead they appear to have followed distinct trajectories probably constrained by their ancestral morphology as well as by subtle differences in their environments. In this context, one might expect a possible Neanderthal cold adaptation to share some elements with the observed pattern in modern humans, but to be limited by the ancestral morphology of that species. Given the constrains of such morphology (i.e., elongated, low crania, tall faces, wide nasal apertures), the greatly increased height of the Neanderthal nasal aperture, as well as their increased cranial breadth, relative to the ancestral morphology are consistent with an adaptation to extremely cold climates.

Finally, since our assessment of climate signature in the morphological differentiation of modern human population is based mostly on Fst estimation for individual craniometric measurements, some notes of caution are in order. First, the fact that we chose to consider as significant only those variables with Fst values higher than 0.3 means that our conclusions are centered on diversifying evolutionary forces. As stated by Roseman (2004) and Roseman and Weaver (2004), stabilizing selection could also be affecting general morphological trends in modern humans. Such stabilizing selection would result in a reduction of Fst estimates for those variables or anatomical regions affected. However, no apparent reduction of Fst estimates was observed in our analysis. The variable with smallest Fst estimate (Occipital arc; Fst = 0.1464) is relatively close to the Fst estimate for the whole cranium (0.2117), and well within the range of Fst estimates published for neutral molecular data (Lewontin, 1972) and other craniometric datasets (Relethford, 1994, 2002).

Second, Fst estimates for craniometric traits depend a great deal on reliable estimates of heritability (Relethford and Blangero, 1990; Relethford, 1994). Carson (2006) and Martínez-Abadías et al. (2009) presented estimates for heritability of craniometric measurements that are in most cases smaller than the usually accepted heritability values of 0.55 (Devor, 1987; Sparks and Jantz, 2002). The consequence of assuming a moderately high and uniform heritability means that we are probably underestimating the number of variables that have real Fst values above the threshold of 0.3. Thus, although the heritability value assumed here does not affect the variables presented as having high degrees of between population diversity, it probably excluded a number of variables that would have been included in this category, had we used the more recent heritability estimates.

In conclusion, both geographic distances and climate differences seem to contribute to the morphological diversity seen among modern human populations. Assuming that geographic distances reflect isolation by distance processes of morphological differentiation due to stochastic microevolutionary forces (Relethford, 2004a; Harvati and Weaver, 2006a, b) and that climate signature might be a result of diversifying selection (Relethford, 2004b; Roseman, 2004; Harvati and Weaver, 2006a, b), our analysis corroborates that the between-groups cranial morphological variation has been shaped both by neutral evolutionary processes and natural selection.

Diversifying selection to cold environments mainly affects dimensions in the face and neurocranial breadth. High Fst values, pointing to diversifying selection, are apparent in the northernmost populations of Europe, Asia, and America. This pattern has been previously observed for North American groups (Roseman, 2004; Harvati and Weaver, 2006b) in much smaller samples (10–13 geographic population samples, comprising one Arctic sample each). Although natural selection to cold climate seems to have been a considerable force of morphological differentiation among modern humans, it appears limited to northern populations and has not eliminated the geographic, population history, signal observed in the overall pattern of modern human craniometric differentiation. Furthermore, diversifying selection to similar environmental conditions was found here to follow distinct pathways of morphological differentiation among the northern populations of Europe, East Asia, and North America. Some aspects of Neanderthal morphology are consistent with an adaptation to extreme cold conditions, given the constraints of their ancestral morphology.


The authors thank Danilo Vicensotto Bernardo for his help with the original database, and André Strauss for sharing with us his Visual Basic code which allowed us to perform the Fst calculations. Two anonymous reviewers provided helpful comments, which improved this manuscript.


Appendix Table

Table Appendix Table. Sample size, geographic and climatic information for the series included in the study
SeriesRegionSampleLongitudeLatitudeReference point for the geographic coordinates of the seriesTemperature (°C)Temperature range (°C)Rainfall (mm)Relative humidity
Annual minimum temperatureAnnual maximum temperatureAnnual average temperatureMaximum temperature of warmest monthMinimum temperature of coldest monthAnnual rainfallRainfall of wettest monthRainfall of driest month
New South WalesAustralia66151.24−33.89Sydney13.12217.526.17.11912931496162
QueenslandAustralia22144.53−22.59Approx. center of Queensland16.331.223.7377.929.148199940
South AustraliaAustralia147138.61−34.92Adelaide11.221.616.428.66.921.7543771556
West AustraliaAustralia28122.18−25.49Approx. Center of Western Australia15.130.222.738.75.832.921243432
KenyaEast Africa9536.81−1.27Nairobi12.724.918.827.41116.49042001567
MalawiEast Africa2633.77−13.98Lilongwe1426.420.229.98.521.4862224070
SomaliaEast Africa6445.342.03Muqdisho23.130.726.932.922.110.848081168
TanzaniaEast Africa8935.73−6.17Dodomi16.12922.631.612.718.9572129066
AinuEast Asia35141.3443.05Sapporo4.312.28.225.8−8.634.411651405874
KoreaEast Asia33126.9537.53Seoul7.316.411.929.7−7.937.613193762070
Han NorthEast Asia62123.3843.8Sheniang0.−20.549.3457136263
Han SouthEast Asia67113.6434.75Zhengzhou920.114.532.3−4.737636146765
Jomon JapanEast Asia50139.8235.66Tokyo11.819.515.730.70.929.814321854768
ManchuriaEast Asia56111.4349.58Manzhouli−95.3−1.922.6−28.150.7406119363
Tohoku JapanEast Asia107140.8938.25Sendai8.−2.730.912681835172
Tokyo JapanEast Asia43139.8235.66Tokyo11.819.515.730.70.929.814321854768
Ancient ItalyMediterranean4012.4841.89Rome10.620.715.730.53.726.87981141873
Badari EgyptMediterranean4132.6725.18El Sibaiya16.233.724.940.86.734.100035
Gyzeh, EgyptMediterranean10031.2430.06Cairo14.
Iron Age IsraelMediterranean7835.2231.77Jerusalem11.823.217.530.75.325.4511119064
Iron and Bronze Age IsraelMediterranean9135.2231.77Jerusalem11.823.217.530.75.325.4511119064
Naqada EgyptMediterranean8032.7225.89Naqada15.732.924.340.55.634.900040
Recent GreeceMediterranean3423.7137.97Athens1321.717.332.55.52743975562
Recent ItalyMediterranean10112.4841.89Rome10.620.715.730.53.726.87981141873
Spain + PortugalMediterranean14−3.740.41Madrid9.3518.914.127.853.324.55821.5111.51358
BorneoSoutheast Asia79113.310.05Approx. Center of Borneo19.527.923.728.219.29329337516685
Caroline IslandsMicronesia22158.186.91Palikir222925.529.421.38.1511349727784
Easter PapuaMelanesia21151.1−8.49Trobriand Island22.930.326.631.622.59.1373640019480
East Sepik, PNGMelanesia29143.61−3.56Wewak22.630.926.731.322.19.2205220614581
Gulf Province, PNGMelanesia33145.83−7.92Kerema22.630.326.431.721.89.9337642219180
JavaSoutheast Asia62106.84−6.21Jakarta22.83227.433.222.111.119883766779
Madang, PNGMelanesia33145.78−5.23Madang22.630.326.530.522.28.3319238812981
Mariana IslandsMicronesia26144.7413.76Hagatña23.729.326.53022.97.123473519185
MoluccaSoutheast Asia22129.45−3.21Maluku Island14.420.817.622.413.88.6276031913682
Negritos PhilippinesSoutheast Asia22120.9814.6Manila23.431.327.433.821.612.22101456876
New BritainMelanesia74150.7−5.78Approx. Center of New Britain21.628.925.329.9218.9408144023680
New CaledoniaMelanesia51166.43−22.29Nouméa19.526.422.929.316.113.210291354276
New HebridesMelanesia52168.32−17.34Port-Vila20.32723.729.318.410.9241338111081
New IrelandMelanesia35152−3.39Approx. Center of New Ireland17.224.120.72516.88.2287430418581
PhilippinesSoutheast Asia122120.9814.6Manila23.431.327.433.821.612.22101456876
SumatraSoutheast Asia27100.35−0.95Padang22.731.32731.922.59.4404149320883
Torres StraitMelanesia61142.26−10.18Moa Island23.429.826.631.522.19.41561339676
West Sepik, PNGMelanesia24141.3−2.68Vanimo2229.725.830.421.49255830415480
Early NubiaNorth Africa7830.4719.16Dunqulah18.536.527.542.99.333.6128025
Kerma, NubiaNorth Africa8630.4719.16Dunqulah18.536.527.542.99.333.6128025
NubiaNorth Africa4630.4719.16Dunqulah18.536.527.542.99.333.6128025
ArizonaNorth America41−112.0733.44Phoenix12.930.121.540.83.537.320428337
ArkansasNorth America29−92.2834.74Little Rock10.522.616.633.6−1.234.812561338070
FloridaNorth America96−84.2830.43Tallahassee12.92619.532.9428.914991937375
IllinoisNorth America118−89.6439.8Springfield5.717.111.430.6−8.639.2912964171
KentuckyNorth America37−84.8734.2Frankfort8.821.915.331.5−1.833.313341547970
MarylandNorth America54−76.4938.97Annapolis8.118.513.330.4−3.433.810691147166
MéxicoNorth America73−99.1319.41Mexico city7.524.415.927.22.924.3644129762
New MexicoNorth America63−105.9335.68Santa Fé1.117.49.328.8−937.8395741946
New YorkNorth America27−73.7542.65Albany2.814.88.828.8−10.839.6946925770
North CaliforniaNorth America61−121.4938.58Sacramento9.123.116.1343.330.7458100166
South CaliforniaNorth America163−118.2434.05Los Angeles12.523.918.
South DakotaNorth America119−100.3544.36Pierre1.715.58.632.2−1446.242779864
UtahNorth America58−111.8940.76Salt Lake City3.517.510.533.1−7.440.5447572055
VirginiaNorth America58−77.4637.55Richmond7.520.313.931.2−3.734.910971117770
BuriatNortheast Asia27104.2452.31Irkutsk−−24.248.4481113973
ChuckchisNortheast Asia21170.6865Approx. Center of Chuktchi Peninsula−14.7−3.4−920.4−3252.4351561486
MongolsNortheast Asia151106.947.92Ulaanbaatar−7.15.6−0.822.9−26.949.824866263
AustriaNorth Europe8016.3748.2Wien514.69.826−3.829.8631763573
CzechoNorth Europe9517.2549.52Charvarty4.113.18.624−5.129.1586852677
EnsayNorth Europe67−7.0857.76Ensay Island5.310.57.915.70.814.911951376488
FinlandNorth Europe2524.9560.16Helsinski28.35.120.8−8.829.6649783380
GermanyNorth Europe8213.452.52Berlin513923.7−3.427.1574703479
HungaryNorth Europe9319.0847.49Budapest6.715.41127−3.330.3562683270
LappsNorth Europe3426.8967.68Approx. Center of Lappi−63.2−1.418.6−2038.6492682578
NetherlandsNorth Europe334.8952.37Amsterdam612.59.220.7−0.120.8802874383
Poundbury, UKNorth Europe102−2.4350.71Dorchester5.9139.520.70.919.8827945186
Recent FranceNorth Europe632.3548.85Paris7.115.511.324.70.823.9636584474
Repton, UKNorth Europe43−1.4752.92Derby5.913.39.621.10.420.7721715184
RussiaNorth Europe4435.1348.42Dnepropetrovsk4.913.49.127.8−7.435.2501593373
SerbiaNorth Europe3420.4644.8Belgrade7.617.112.328.6−1.530.1663853869
Spittafields UKNorth Europe57−0.1251.5London6.914.810.823.21.321.9641624181
Spittafields UKNorth Europe86−0.1251.5London6.914.810.823.21.321.9641624181
SwedenNorth Europe3118.0759.29Stockholm3.810.3722.2−4.927.1526662580
SwitzerlandNorth Europe377.4446.94Bern4.312.78.522.7−3.225.99441156076
UkraineNorth Europe2634.144.96Kiev6.115.510.827.5−3.531516593474
AlaskaExtreme North America36−134.5758.36Juneau1.58.5518−6.724.715732287980
British ColumbiaNorth America78−123.3448.44Victoria6.213.59.920.11.918.26941161681
East AleutsExtreme North America45−169.7152.99Kagamil Island26.34.112.7−3.215.99801195684
GreenlandExtreme North America107−51.7264.18Nuuk−41.7−1.210.8−11.522.3668783981
Iroques OntarioNorth America40−79.3843.67Toronto3.712.48.126.5−9.235.7790825275
N AlaskaExtreme North America70−156.7871.29Barrow−15.3−9.2−12.27.6−30.938.511325481
AmurNortheast Asia22−170.5365.84Pouten−10.5−4.6−7.58.3−24.532.8335561692
NE CanadaExtreme North America30−68.5263.74Inqaluit−13.4−5.8−9.611.5−31.142.6406611875
NW AlaskaExtreme North America107−16465.4Approx. Center of Seward Peninsula−8.2−0.8−4.514.5−21.636.1357781476
SW AlaskaExtreme North America135−160.4561Approx. Southwest Alaska−5.92.7−1.617.6−19.336.94651021977
Tlingit AlaskaExtreme North America35−134.5758.36Juneau1.58.5518−6.724.715732287980
West AleutsExtreme North America82−166.6953.7Aleutian West1.25.63.312.9−4.617.59471115385
Chatam IslandsPolynesia76−176.53−43.91Chatam islands8.714.411.618.15.212.990411753 
Easter IslandPolynesia81−109.36−27.11Easter Island16.223.319.826.61412.611601377779
Marquese IslandsPolynesia66−139.56−8.9Ua Huka Island21.928.625.329.521.58127618164 
New ZealandPolynesia120174.78−41.31Wellington9.116.512.821.35.415.911661325881
Society IslandsPolynesia49−149.56−17.53Papeete22.729.225.930.621.29.4171531450 
BushmanSouth Africa3218.42−33.92Cape town11.821.916.826.27.918.38581422574
South Africa, BantuSouth Africa4328.21−25.23Pretoria11.327.319.330.52.827.7582102558
South AfricaSouth Africa2428.21−25.23Pretoria11.327.319.330.52.827.7582102558
ZuluSouth Africa3028.21−25.23Pretoria11.327.319.330.52.827.7582102558
ChileSouth America20−70.64−33.46Santiago7.122.114.629.62.62736387172
Fuego-PatagoniaSouth America57−68.37−54.09Approx. Center of Tierra del Fuego1.495.213.8−1.815.6437452576
PeruSouth America223−77.62−11.11Huacho15.423.119.22713.213.881086
Venezuela + ColombiaSouth America12−74.084.61Bogotá11.722.0516.92310.112.9901.512226.581
AfghanistanSouth Asia4069.1734.52Kabul4.319.311.832.2−8.640.834082150
Andaman IslandsSouth Asia5392.812.47Andaman Islands23.929.526.731.722.49.33088545679
BangladeshSouth Asia2590.423.7Dacca21.330.625.934.512.322.21990367775
Bengal, IndiaSouth Asia5188.3622.57Kolkata21.331.426.336.112.623.51730366771
Bihar, IndiaSouth Asia2185.1325.61Patna20.831.3263910.928.11033289363
LaosSoutheast Asia32102.6117.96Viang Chan20.430.725.533.813.919.91635339375
Madras, IndiaSouth Asia10080.2413.06Chennai23.833.428.638.519.419.11176311271
MalaysiaSoutheast Asia56101.73.15Kuala Lumpur22.331.827.132.721.711247528512882
MyanmarSoutheast Asia10396.1516.78Yangon22.53227.336.717.918.82348507373
NepalSouth Asia3685.3127.7Katmandu12.12418.1283.224.81548390876
NW IndiaSouth Asia2977.228.63New Delhi18.631.625.
Punjab, IndiaSouth Asia7373.5336.73Approx. Center of Jammu & Kashmir−8.90−4.414.1−22.836.9267411051
SikkimSouth Asia2588.4627.6Approx. Center of Sikkim District−0.611.75.516.6−9.726.3833175574
SingaporeSoutheast Asia43103.821.29Singapore22.830.826.831.422.19.3238529715682
ThailandSoutheast Asia47100.5213.73Bangkok23.632.528.134.920.214.71414315675
TibetEast Asia6491.1329.65Lhasa−−12.133.2379116050
VietnamSoutheast Asia24105.8521.02Hà Noi20.327.423.83313.319.716703231683
CameroonWest Africa3311.513.87Yaoundé19.128.223.629.618.411.216593002279
GabonWest Africa769.450.39Libreville2329.226.130.621.88.82883501284
GhanaWest Africa51−0.195.55Accra23.230.226.73221.610.48032071482
Ibo, NigeriaWest Africa797.489.05Abuja19.631.225.434.716.618.11401287162
Ivory CoastWest Africa27−5.286.81Yamoussoucro20.931.12633.819.214.611551741475
NigeriaWest Africa457.489.05Abuja19.631.225.434.716.618.11401287162