Figure 6.13 Multi-dimensional-scaling analysis of 33 Mousterian assemblages from southwestern France, from Mellars 1967. The assemblages represented are: nos 1-4 Pech de l'Aze I (layers 4, B, 6, 7); nos 5-6 Hauteroche (lower, upper layers); nos 7-10 Abri Chadourne (layers A, B, C, D); no. 11 L'Ermitage; nos 12-14 Roc de Marsal (layers 6,10,11); no. 15 Mas-Viel; nos 16-18 Petit-Puymoyen (cave, C2, breccia); nos 19-33 Combe Grenal (layers 11,14,17, 20, 22, 23, 26, 27, 29, 33,35,38, 50, 52.

MTA Type B) are separated clearly, on the right-hand side, from those with higher racloir frequencies on the left-hand side (those of the Ferrassie, Quina, Typical Mous-terian and MTA Type A), Within these two broad divisions, however, the groupings of individual assemblages correspond closely with Bordes' basic taxonomic divisions. Thus the 15 assemblages attributed to the Quina Mousterian are grouped together closely in the lower, left-hand side of the diagram and in this area no other non-Quina industries occur. In keeping with Bordes' notion of the wider 'Charentian' grouping it is interesting that the group of five Ferrassie-type assemblages are placed immediately adjacent to the Quina group but consistently separate from the latter group and located towards the upper, left-hand zone of the distribution. (It is particularly interesting to see that although Bordes7 distinction between the Ferrassie and Quina variants is based on the technical features of the industries - which are not incorporated directly in the present analysis - the separation between the two forms is equally apparent on purely typological grounds). The smaller grouping of five Typical Mousterian assemblages is similarly placed consistently to the upper right-hand-side of the main Quina-Mousterian distribution. The most widely dispersed industries are those of the MTA group. In this case there is a fairly sharp separation of the different assemblages based on the overall frequencies of racloirs, with the single industry of the Type A group on the left of the diagram and those of the Type B on the right. Even so, these industries are separated from those of the other industrial groupings and significantly reveal an overall linear arrangement which corresponds exactly with the documented stratigraphic sequence of the respective assemblages in the different levels at Pech de 1'Aze site I.

A second approach to multivariate analysis was published by Callow and Webb (1977, 1981), employing the technique known as Canonical Variates analysis. The aim of this technique is not so much to scan a body of data for its inherent structure but rather to test how far a preconceived perception of clustering in the material can be substantiated in terms of the overall range of the analytical data. In other words, given an a priori classification of the material, the technique is designed to assess how far this classification has succeeded in identifying genuine patterns of clustering, or clear discontinuities, in the analytical data. In the same way the technique can assess whether individual units have been ascribed to their taxonomically correct grouping, or whether they would be better assigned to another group.

Callow and Webb's analysis was applied to many more assemblages than that of the multidimensional scaling study discussed above (a total of 96 assemblages in all) but was again restricted entirely to industries from the Peri-gord and immediately adjacent areas of southwestern France. The analytical data also differ in several respects from those of the earlier study and rely mainly on the principal typological and technological indices employed by Bordes to summarize the major quantitative features of the assemblages. In essence this analysis uses a more restricted range of purely 'typological7 features of the assemblages but expands on this data base by incorporating all the major 'technical7 features of the industries (i.e. frequencies of Levallois flakes, facetted striking platforms, blades etc.).

The main results of Callow and Webb's analysis are illustrated in Fig. 6.14, which again compresses a multi-dimensional pattern into two main visual dimensions. As they point out, the pattern provides strong support for the major components of Bordes' taxonomy. All the major variants, with the possible exception of the Typical Mousterian, occupy separate zones on the diagram with no significant areas of interchange or overlap between the groups. The Typical Mousterian emerges as the least isolated group and

Figure 6.14 Canonical variates analysis of 96 Mousterian assemblages from southwestern France according to Callow & Webb 1981, with the taxonomic classification of the assemblages according to Bordes.

Figure 6.14 Canonical variates analysis of 96 Mousterian assemblages from southwestern France according to Callow & Webb 1981, with the taxonomic classification of the assemblages according to Bordes.

Was this article helpful?

0 0

Post a comment