Craniometry This file is one of six making available supplementary material to the printed text of:

Mary Jackes, David Lubell & Christopher Meiklejohn/ Healthy but mortal: human biology and the first farmers of Western Europe
ANTIQUITY volume 71 number 273 (1997): 639-658.

The files are entitled:
Craniometry
Dates
Metric data
Non-metric data
References

A further file reproduces the full text as printed in the paper publication: Printed

Craniometry
Lalueza Fox (1996) states that skull measurements indicate complete discontinuity between Mesolithic and Neolithic populations in Iberia. We have discussed this opinion briefly elsewhere (Jackes et al. 1997), but provide additional background here.

Our full data set is given in CRANIOMETRY TABLE 1. The sexes are pooled to increase sample sizes, the variables have been chosen primarily on the basis of available data, and missing data has been filled with mean values. Information on Cova da Moura from Gallay & Spindler (1970) was too incomplete to use and, just as Garralda & Mesa (1984) question whether the separation of the Portuguese Bronze Age site Eira Pedrinha from other Iberian samples may not be caused by the limited sample, the placement of our Neolithic samples is uncertain.

CRANIOMETRY TABLE 1. Means of Martin measurements
(Martin 1957-66) for males and females combined, used in classificatory analyses (CRANIOMETRY FIGURES 1 & 2)

M1 glabello-occipital length.
M5 basion-nasion length.
M8 maximum cranial breadth.
M9 minimum frontal breadth.
M48 nasion-prosthion height.
M52 orbital height.
M54 nasal breadth.

a Missing data (in brackets) derived from means of Neolithic Portuguese skulls.
b Missing data (in brackets) derived from mean of Moita, Arruda, Téviec and Hoëdic (Meiklejohn unpub. data).
c Individuals aged 30-39 years not dying of pulmonary TB; missing data derived from multiple regression.
d Not used in all analyses.

ID sample code M1 M5 M8 M9 M48 M52 M54 source
1 medieval Granada Muslims MUS 180.50 98.81 134.26 94.59 67.51 32.85 24.17 Fox et al. 1996
2 19th century Lisbon c
180.55 98.60 135.05 94.69 (65.84) 32.66 24.50 Olivier & Almeida 1972
3 Eira Pedrinha
181.02 99.24 137.30 95.13 65.36 30.92 24.01 Mendes Corrêa & Teixeira 1949
4 Melides
180.75 96.50 137.00 91.25 64.33 31.50 23.75 Meiklejohn & Schentag nd
5 Escoural
181.70 97.60 133.40 92.00 62.80 31.95 24.20 Isidoro 1981
6 Casa da Mouraa
180.00 (95.32) 137.00 95.77 62.96 30.27 (23.54) Jackes unpublished
7 Cabeço da Arruda
180.60 96.75 136.00 93.50 71.00 28.60 23.10 Meiklejohn unpublished
8 Moita do Sebastião
179.95 92.60 131.75 91.60 68.05 29.70 24.65 Meiklejohn & Schentag nd
9 Fontainhasa
177.00 90.60 135.00 94.43 (64.50) 30.87 23.17 Jackes unpublished
10 Bronze Age central Catalonia CAT1 183.86 97.87 138.70 94.88 66.01 31.55 23.35 Fox et al. 1996
11 Bronze Age upper Andalusia GRA 183.08 97.18 138.62 93.93 69.11 31.28 23.65 Fox et al. 1996
12 Guipuzcoan Basques
182.81 98.77 140.96 95.85 68.90 33.02 22.79 Aranzadi in Morant 1929
13 Tarragona TAR 183.33 99.55 139.88 96.30 69.58 33.55 24.46 Fox et al. 1996
14 Basques in de la Rua's sample BAS 184.07 97.53 141.19 96.13 70.98 34.51 22.66 de la Rua 1992
15 medieval Barcelona Jews JEW 183.85 100.00 137.99 97.05 67.99 33.42 23.38 Fox et al. 1996
16 medieval central Catalonia CAT2 185.36 99.83 137.96 95.27 68.19 33.27 23.45 Fox et al. 1996
17 Visigothic North Meseta VIS 186.19 100.95 137.92 97.42 70.22 32.85 24.76 Fox et al. 1996
18 Gruta da Feteira
180.40 97.00 138.33 97.56 70.50 32.50 22.50 Jackes unpublished
19 medieval Christian Cantabria CAN 188.90 99.92 140.14 96.13 72.24 33.50 23.77 Fox et al. 1996
20 Carenque a,d
182.79 94.00 135.55 94.33 57.50 30.75 21.50 Bubner 1986
21 Cascaisd
188.88 103.08 142.50 97.67 63.96 32.92 25.08 Riquet 1972
22 El Colladob, d
186.00 (98.27) 143.60 95.80 78.00 28.00 (24.59) Pérez-Pérez et al. 1995

In short, the data are extremely poor, and when taken from published material, congruity of methods is not assured. We have, however, avoided using measurements such as Martin 17, 51or 51a for which methodology may be uncertain when not specified. The values for M55 (nasal height) are so variable that there may be differences among researchers on techniques of measurement. As a result, this variable was removed from the analysis. We have multiple measurements for several of the Neolithic samples, permitting checking for inter-observer error (Jackes unpub. data; Meiklejohn & Schentag n.d.; Paula e Oliveira 1889; Riquet 1972). In addition, a number of papers provide information on Basque craniometry (Morant 1929; de la Rua 1992; Lalueza Fox et al. 1996). In order to include all possible samples, the variables were limited to the Martin measurements 1, 5, 8, 9, 52 and 54; measurement 48 was then included when it was clear that use of the mean for the Lisbon sample would not alter placement.

A large number of analyses, using either pooled sexes or males only, showed that sites with small samples sizes may produce unstable results. As a result the challenge became one of not allowing preconceptions to dominate the choice. We performed a number of cluster analyses on some or all of the samples and tabulated the results for those sites that were unstable.

Moita and Arruda separated completely in around 44% of our initial 30-40 analyses. Cascais and Feteira often group loosely with Basques and Cantabrians. This is almost invariable for Cascais, which grouped with Cantabria and other northern samples in 87% of 34 cluster analyses. Feteira occasionally grouped with Casa da Moura, but in 68% of the analyses it joined the northern Basque/Catalonia/Cantabria grouping. The more southerly of the Portuguese sites often grouped together, separate from those north of the Tagus; however in 50% of analyses Escoural joins Casa da Moura. The sites of the Portuguese Estremadura were very unstable, drifting between the Basque pole and the more southern Portuguese group. Mesolithic Spain, as represented by El Collado, either fell closer to the Portuguese Mesolithic or to the Basque/Catalonia/Cantabria grouping.

While the explanation for the instability may lie in the weakness of the data set, Lalueza Fox et al. (1996) ascribe a similar lack of robusticity in their groupings to population homogeneity.

After our initial analyses, Eira Pedrinha, a northern Portuguese cave/shelter site which has a relatively large sample of crania (Mendes Corrêa; & Teixeira 1949) spanning a long period of time from Early Neolithic to Chalcolithic (Arnaud 1982), was brought into the data- set.

CRANIOMETRY TABLE 2 shows to a very limited extent how cluster membership changes according to the method of analysis. While not proof, this instability suggests that variability in Iberian populations from Mesolithic to modern times indeed shows no temporal discontinuity, having an underlying homogeneity modified by clinal variation, with the Basques at one pole together with some of the more northern Spanish and Portuguese sites, very slightly differentiated from the more southern and western sites.

CRANIOMETRY TABLE 2. Shifting cluster membership for samples using two different methods of analysis and three different measures of distance (using SPSS for Windows 6.1.3) with partial and full data sets.
Average linkage combines clusters to minimize average distance between all pairs of items in which one member of the pair is from each of the clusters.
Ward's method calculates the means for each variable within each cluster, then calculates the squared Euclidean distance to the cluster means and these distances are summed for all of the cases.
SE Squared Euclidean Distance (sum of the squared differences in values for each variable.)
Ch Chebychev (maximum absolute difference in values for any variable.)
CB City Block Measure (sum of the absolute differences in values for each variable.)



average linkage Ward's method average linkage Ward's method
ID sample SE Ch CB SE Ch CB SE Ch CB SE Ch CB
1 medieval Granada Muslims 1 1 1 1 1 1 1 1 1 1 1 1
2 19th century Lisbon 1 1 1 1 1 1 1 1 1 1 1 1
3 Eira Pedrinha 1 1 1 1 1 1 1 1 1 1 1 1
4 Melides 1 1 1 2 1 2 1 1 1 1 1 1
5 Escoural 1 2 1 2 1 2 1 1 1 1 1 1
6 Casa da Moura 1 2 1 2 1 2 1 1 1 1 1 1
7 Cabeço da Arruda 2 3 2 3 2 3 1 2 1 2 2 2
8 Moita do Sebastião 3 4 3 4 3 4 2 3 2 3 3 3
9 Fontainhas 4 4 4 4 3 5 2 3 2 3 3 3
10 Bronze Age central Catalonia 5 5 5 5 4 3 3 2 3 2 2 2
11 Bronze Age upper Andalusia 5 5 5 5 2 3 3 2 3 2 2 2
12 Guipuzcoan Basques 5 5 5 5 5 6 3 2 3 2 2 4
13 Tarragona 5 5 5 5 5 6 3 2 3 2 2 4
14 Basques in de la Rua's sample 5 5 5 5 5 6 3 2 3 2 2 4
15 medieval Barcelona Jews 5 5 5 6 4 6 3 2 3 2 2 4
16 medieval Central Catalonia 5 5 5 6 4 6 3 2 3 2 2 4
17 Visigothic north Meseta 5 6 5 6 6 6 3 2 3 2 2 4
18 Gruta da Feteira 5 5 2 5 5 3 3 2 1 2 2 2
19 medieval Christian Cantabria 6 6 6 6 6 6 3 2 3 2 2 4
20 Carenque





4 4 4 4 5 3
21 Cascais





5 5 5 5 6 5
22 El Collado





6 6 6 6 4 6

The separation of Moita and Arruda, highlighted in CRANIOMETRY TABLE 2, is demonstrated by CRANIOMETRY FIGURE 1 where we have used Ward's method and squared Euclidean distance. Here there is a clear differentiation into two clusters, one containing five of the Portuguese Neolithic and Chalcolithic sites plus Moita, and the other containing one of the Portuguese Neolithic sites plus Arruda.

CRANIOMETRY FIGURE 1. Dendrogram produced by hierarchical cluster analysis of the data for samples 1 to 19 of CRANIOMETRY TABLE 1, using Ward's Method and the Squared Euclidean Distance measure. Two distinct clusters are shown. The upper cluster contains Neolithic to nineteenth century Portuguese samples, slightly separated from one Mesolithic (Moita) and one Neolithic sample (Fontainhas). What is noteworthy is that our second Portuguese Mesolithic cranial sample (Arruda) separates out completely from Moita. This occurs in about 50 % of analyses. Arruda and Feteira cluster loosely with Basque and Bronze Age Spanish samples. The linkage of Feteira with northern samples is almost invariable.

Figure 1

Lalueza Fox (1996: Figure 1) and our own analyses suggested that our data set might reasonably be assigned to six clusters and we used hierarchical cluster analysis (Ward's method, squared Euclidean distance measure) to assign group membership for each sample to one of six clusters. We planned to refine this by iterations in discriminant function analysis, but found the clustering completely stable. We used the direct method, rather than stepwise discriminant function, having determined that the discrimination is basically spread over all seven variables (indicating that we are, in fact using total conformation of the skull, rather than crude size as indicated by length and breadth).

CRANIOMETRY FIGURE 2 plots the first discriminant function (explaining 78.08% of the variance) and the second discriminant function (explaining 10.21% of the variance) for the full data set, after discriminant function analysis determined the highest probability of cluster membership for the three unassigned samples 20, 21 and 22. We can immediately see that, although outliers, these three samples can be accommodated within the Iberian data. As with Arruda, these skull samples are discriminated on the basis of the second, rather than the first function. Mesolithic Moita and Neolithic Fontainhas remain together as outliers on function 1 which is the most significant one.

CRANIOMETRY FIGURE 2. Scatter plot of the first two of three significant discriminant functions generated by direct method analysis of the craniometric data in CRANIOMETRY TABLE 1. The grouping variable for the analysis is cluster membership predicted by Ward's Method hierarchical cluster analysis (six cluster solution) on the first 19 samples. With forced cluster membership prediction, sample 20 (Carenque - Chalcolithic) is classified into group B, sample 21 (Cascais - Neolithic) into group F, and sample 22 (El Collado - Mesolithic) into group C. Moita and Fontainhas separate out most distinctively from all other skull samples, and Moita and Arruda are not identical.

Figure 2

Cascais is obviously an outlier, and it may be relevant that a comparison of the figures provided by Riquet (1972) for male skulls from Melides with those of Meiklejohn indicate clear differences in methods of measurement. However, re-analysis excluding Cascais does not alter the relationships shown in CRANIOMETRY FIGURE 2. Similarly, removal of El Collado and Carenque makes no difference to the scatter relationships.

The samples of complete Portuguese Neolithic crania are inadequate to do more than suggest that simple explanations will not provide all the answers. We cannot assume that the Basques are remnant Mesolithic people whose forebears were replaced in the rest of the Iberian Peninsula by newcomers from the east. We cannot assume that a homogeneous Mesolithic population was replaced by heterogeneous Neolithic groups. CRANIOMETRY FIGURE 2, in so far as it tells us anything, suggests that heterogeneous Mesolithic populations blend into succeeding Neolithic populations. CRANIOMETRY FIGURE 3 supports this. Here grouping for discriminant analysis is specified by geography, resulting in a clustering of all central Portuguese samples, apart from Cascais and Feteira. Mesolithic samples are not differentiated.

In Jackes et al. (1997) we provide our comment on Lalueza Fox's opinion using measurements of male skulls only in order to duplicate the method of analysis used by Lalueza Fox. CRANIOMETRY TABLE 3 provides the data we used to produce Figure 1 in Jackes et al. (1997). Our analyses provide no support for an hypothesis of discontinuity between Mesolithic and Neolithic populations in Iberia.

CRANIOMETRY FIGURE 3. Direct method discriminant function analysis of the craniometric data in CRANIOMETRY TABLE 1 with geographical location coded as the grouping variable. Skull length (GOL, Martin measurement 1) correlates at .54 with Function 1, but all other variables contribute most significantly to Functions 3 and 4. Basque samples, and historic Spanish samples from the northern Meseta and Cantabria are outliers. Portuguese samples in general cluster together indicating that geography, rather than chronology, mediates skull form.

Figure 3

CRANIOMETRY TABLE 3: Means of Martin measurements (Martin 1957-66) for males only used in classificatory analyses for Figure 1 in Jackes et al. (1997).
Key

M1..........................glabello-occipital length.
M5..........................basion-nasion length.
M8..........................maximum cranial breadth.
M9..........................minimum frontal breadth.
M40..........................basion-prosthion length.
M45..........................bizygomatic
M52..........................orbital height.
M54..........................nasal breadth.

ID sample code M1 M5 M8 M9 M4 0 M45 M52 M54 source
1 Talayotic Majorca MAL2 188.56 104.52 140.29 97.31 95. 72 133.22 32.11 23.50 Lalueza Fox et al. 1996
2 Talayotic Majorca MAL1 186.82 101.22 139.59 97.61 98. 32 134.00 30.95 23.60 Lalueza Fox et al. 1996
3 Tarragona TAR 186.93 102.47 142.07 97.18 95.56 134.84 33.68 24.93 Lalueza Fox et al. 1996
4 Basques
185.80 99.60 143.50 97.10 90.70 128.80 33.60 23.20 Morant 1929
5 Basques BAS 188.39 100.02 143.53 97.29 92.27 129.08 35.09 22.74 de la Rua 1992
6 medieval Christian Cantabria CAN 192.68 101.40 142.12 97.82 94.0 0 130.69 33.91 24.40 Lalueza Fox et al. 1996
7 medieval Central Catalonia CAT2 189.28 102.47 139.52 96.54 94. 26 131.45 33.59 23.96 Lalueza Fox et al. 1996
8 medieval Barcelona Jews JEW 187.88 103.00 139.60 98.87 95.57 131.00 35.78 24.17 Lalueza Fox et al. 1996
9 medieval Christian Burgos BUR1 187.90 100.31 140.02 95.28 92.95 130.36 32.84 24.95 Lalueza Fox et al. 1996
10 Visigothic North Meseta VIS 190.87 102.90 140.20 97.42 97.70 129.51 32.87 25.49 Lalueza Fox et al. 1996
11 Bronze Age central Catalonia CAT1 188.07 100.97 141.04 95.74 95. 57 128.49 31.87 23.70 Lalueza Fox et al. 1996
12 Bronze Age upper Andalusia GRA 186.45 100.29 140.07 95.14 95.69 128.61 32.09 24.26 Lalueza Fox et al. 1996
13 medieval Christian Burgos BUR2 186.68 101.76 140.65 94.63 96.87 129.69 32.10 22.39 Lalueza Fox et al. 1996
14 Eira Pedrinha
185.70 102.00 138.40 97.20 95.60 125.30 (33.07) 23.53 Mendes Corrêa & Teixeira 1949
15 medieval Granada Muslims MUS 184.74 101.46 135.69 96.30 97.3 4 128.47 33.38 24.86 Lalueza Fox et al. 1996
16 Escoural
183.50 101.50 134.38 92.70 96.60 (128.93) 32.60 25.00 Isidoro 1981
17 Visigothic Portuguese
187.60 101.73 136.13 95.07 96.17 128.88 34.94 25.19 Cunha & Neto 1953
18 Melides
181.00 99.00 134.20 93.00 95.00 126.00 31.50 22.50 Meiklejohn & Schentag n.d.
19 Cabeço da Arruda
186.00 95.50 135.80 94.00 100.00 124.50 28.70 23.70 Meiklejohn unpublished
20 Moita do Sebastião
181.90 93.80 134.40 94.80 93.50 121.50 30.40 24.90 Meiklejohn and Schentag n.d.

Dates
Metric data
Non-metric data
References
Printed
[Home]
© Antiquity Publications Ltd 1997