Understanding the Middle Palaeolithic assemblage typology.

MOYER, COLIN CAMPBELL ; ROLLAND, NICOLAS

Recent research on Middle Palaeolithic stone-tool assemblages has focused narrowly on individual flint-knapping sequences or on single assemblages, neglecting the relationship between individual behaviour and population-scale questions. Furthermore, the description of countless chaines operatoires makes inter-assemblage comparisons difficult at best. Without a shared quantitative reference, using the presence or absence of a technology to compare sites could ultimately result in replacing the fossile directeur with a chaine directrice. At present, Francois Bordes' assemblage and artefact typologies represent the only means of comparison. Many authors, either implicitly (Grayson & Cole 1998; Turq 1989) or explicitly (Mellars 1996) uphold the validity of Bordes' assemblage categories. One cannot, however, dispense with Bordes' artefact typology and retain the assemblage typology, because together they create a unified interpretative and analytical scheme.

Historical background

Bourlon (1906) first recognized Middle Palaeolithic variability during his excavations at the abri superieurat Le Moustier. He observed (1906: 317-18) the presence of handaxes in the middle and upper layers and their absence in the base layers. Peyrony, on the basis of his own work at the abri inferieur and Bourlon's observations, divided the Mousterian into two traditions, Mousterien Typique (MT) and Mousterien de Tradition Acheulien (MTA) (containing handaxes). However, Peyrony incorrectly asserted that the base layers at the abri superieur contained handaxes and concluded that these traditions were interstratified and not chronologically patterned, i.e., they were the results of parallel toolmaking traditions. Nonetheless, Peyrony's (1930: 171-2) model and his interpretation of these different traditions as the product of distinct tribes became the basis for Bordes' assemblage typology and his theoretical position. Recent absolute dating methods support a late date for the MTA in southwestern France and at Le Moustier, thus supporting Bourlon's initial observations (Mellars 1988; Mellars & Grun 1991). Additionally, the repeated occurrence of MTA assemblages after Quina and Ferrassie type Mousterian assemblages in southwestern France supports their relatively late chronological position regardless of the absolute dating (Mellars 1969; 1988).

Bordes' assemblage typology

Bordes' early writing (1958: 180; 1960: 101; 1969: 2) clearly reveals that his artefact and assemblage typologies were created in part to explain a pre-existing interpretative framework. With few notable differences, his interpretations closely follow those of Peyrony. Although he accepted a model of stylistic differences between tribes, Bordes established a more rigorous system to classify the diverse Mousterian assemblages into formally defined types.

Bordes' system incorporates two typological systems, an artefact typology and an assemblage typology. His system of classifying the different assemblages is typological rather than taxonomic because the different assemblage types are not hierarchically ordered groupings of the tool types themselves (Adams & Adams 1991: 47, 202). The assemblage categories are defined primarily by the varying proportions of tools and secondarily by the presence or absence of specific type-fossils.

The assemblage typology uses two processes to discriminate between different assemblages:

1 cumulative graphs of artefact types (all retouched tools and flakes made using Levallois technology); and

2 `secondary determinants' (Bordes & de Sonneville-Bordes 1970).

The cumulative graphs contrast the proportions of notches and denticulates to scrapers, whereas the secondary determinants include the proportion of Levallois flakes and the presence or absence of diagnostic tool types (e.g. backed knives and handaxes; see TABLE 1). The result is five types, the MTA being divided into two subtypes. Bordes & de Sonneville-Bordes (1970) provided a chart showing a multimodal distribution of the proportion of scrapers relative to the total tool count. They used this distribution to support the statistical discreteness of the categories.

[TABULAR DATA 1 NOT REPRODUCIBLE IN ASCII]

Although the tribal element is no longer accepted, the discreteness of these categories is upheld through their continued use. Mellars (1996) cites Callow & Webb (1981: 137) in support of their discreteness.

Callow and Webb's (1981) study suffers from a number of statistical flaws. Most notably, it uses discriminant analysis. This statistic assumes the existence of the categories at the outset and describes the differences between them. Discriminant analysis is not well suited to testing the existence of categories employed at the outset. Callow & Webb also used Bordes' technological and typological indices and groups instead of his type-list and thus over-emphasized certain tool classes; several scraper types were counted several times in their analysis. This overrepresentation of a single tool type was likely the cause of the apparent discreteness of the assemblages.

Causes of assemblage variability

The purpose of the present study is first, using Bordes' defined analytical framework, to assess his claim that the assemblage categories he defined are real and meaningful analytical units. If the categories are not meaningful within his own framework, why are they used in an uncritical fashion? Secondly, we describe the nature of stone tool assemblage variability in Western Europe.

A principal components analysis (PCA) of 103 assemblages identified the primary sources of variation between these assemblages (Moyer 1998). PCA searches for a few uncorrelated linear equations of the original variables (e.g. proportions of artefact types) used to describe each item (e.g. archaeological assemblages) to summarize the information contained in the original variables. Each principal component is ranked in order of the amount of variance (based on the correlations between the original variables) it explains. Only assemblages reported using Bordes' artefact typology and containing 300 or more artefacts from Western Europe were included (see Moyer 1998 for a list of assemblages and variables) and Fischer's arcsine transformation was used (Baxter 1994). The artefact types chosen for analysis included the proportions of Bordes' artefact types in addition to unretouched, non-Levallois blanks and hand-axes. Three components, which explain nearly half of the variance between assemblages, were chosen as significant. The interpretation of these principal components is presented in TABLE 2 and a plot of the first three principal components is given in FIGURE 1.

TABLE 2. Interpretation of principal components.

component interpretation proportion of
 variance

1 Scrapers vs. unretouched non-Levallois 24%
 flakes and blades
2 Notches, denticulates and backed knives 12%
 vs unretouched non-Levallois flakes
 and blades
3 Levallois end-products and Mousterian 10%
 points

[Figure 1 ILLUSTRATION OMITTED]

The first component suggests that the bulk of Bordes' scraper types are correlated and that there is an inverse linear relationship between all scrapers and unretouched non-Levallois flakes and blades. This equation explains approximately 24% of the variance in the sample and explains the largest proportion of variance, confirming Rolland's (1977; 1981) observation that differential scraper manufacture explains the most significant proportion of the variability in Mousterian assemblages.

The second component suggests that notches, denticulates and backed knives are correlated and that there is an inverse relationship between the proportions of these types and unretouched non-Levallois flakes and blades. This equation explains approximately 12% of the variance in the sample. In the case of notches and denticulates, they are inversely proportionate to both the proportion of flakes and blades and the proportion of scrapers (Rolland 1981: 26-7; Rolland & Dibble 1990: 485-6).

The third component shows that the various forms of Levallois artefacts (types 1 to 4 in Bordes' type-list) are correlated and Levallois technology explains approximately 10% of the variance in the sample. Interestingly, Mousterian points were most strongly correlated with Levallois technology, not with scrapers, thus questioning whether they were intended tool forms or a by-product of scraper manufacture (Dibble 1987; Mellars 1996: 110-17). Many Mousterian points are made on Levallois points. When the first three principal component scores for each assemblage are plotted, there is no clear separation of assemblages into clusters corresponding to Bordes' assemblage types (FIGURE 1). Each assemblage type overlaps with at least one other.

In short, the PCA suggests three things:

1 Several classes (taxa) of tool types are mutually correlated (scrapers; notches and denticulates; backed knives; and Levallois artefacts). These taxa represent more `natural' categories than the subtle variations suggested in Bordes' artefact typology.

2 Certain tool classes explain the bulk of the variability in the assemblages (TABLE 2). This study suggests the importance of these types and provides a quantitative expression of variability in Western European Middle Palaeolithic assemblages. The importance of scrapers and unretouched non-Levallois flakes and blades supports the attention given them by Rolland & Dibble (1990). It also highlights the importance of other tool classes and the importance of Levallois technology in explaining significant proportions of variability.

3 There is little evidence of discrete categories of assemblages. The assemblage types proposed by Bordes represent grades of a continuously distributed, but internally heterogeneous group in Western Europe (FIGURE 2). However, the types of tools Bordes used to discriminate between assemblage types correspond to the same classes that the principal components analysis suggests. Other statistics further support the lack of discrete clusters corresponding to the defined assemblage types (FIGURE 3; Moyer 1998).

[Figures 2-3 ILLUSTRATION OMITTED]

Discussion

Bordes' assemblage types are not statistically discrete and this should change the way we deal with them. Rather than equating an assemblage type with an essential character, we should recognize that the bulk of Middle Palaeolithic sites are situated within a single, internally heterogeneous category. The lack of discrete categories and the fact that the differences between Mousterian assemblages can be explained by differences in the proportions of tools suggests that factors other than style, function or cognitive `templates' underlie the most fundamental differences in assemblage composition. Many of the documented differences between assemblages in terms of Bordes' artefact types are likely the product of simple economizing behaviour. More tools (particularly sidescrapers) are produced in some assemblages than others, and correspondingly there is a proportional decrease in the total numbers of unretouched flakes. Two different, but related, behavioural patterns have been documented which may help to explain this pattern. The first is the differential transformation of unretouched flakes into tools (Principal Components 1 and 2; Rolland 1977; 1981). The other is a change in core-reduction strategy that is particularly evident in the transition to Quina assemblages from Ferrassie assemblages. The core reduction strategies demonstrated by Turq (1989) appear to result in a greater proportion of flakes suitable for manufacturing large scrapers than for Levallois technology (P.C. 1 and 3; see TABLE 2). Both of these strategies are economizing behaviour.

The causes of this economizing behaviour are more difficult to determine. The shapes of the distributions of the various features are unimodal and continuous (FIGURE 2; Rolland 1981), suggesting multiple underlying causes combining to produce a continuous range of variation. The mathematical basis of the expression of multicausal phenomena is found in the binomial and other probability distributions that are the combined effect of multiple events. Different combinations of events (possibly including differences in access to raw materials, social density, longevity of occupation, mobility, faunal exploitation patterns, etc.) could lead to similar results: the economizing of raw material. Palaeoclimatic differences remain the most notable set of evidence supporting this pattern. Certain assemblage types correspond to specific environments (Rolland 1981; Rolland & Dibble 1990). The relationship between climate and geography is complex, however, and different climatic events in different geographic locations may yield similar archaeological remains. If economizing behaviour is a socially held rule, how is this reflected in individual acts of tool production (Giddens 1984)? The lack of standardization in describing and comparing these technological sequences is problematic. Population-scale comparative analyses must remain a fundamental task in the development of theories explaining socio-economic patterns. Taxonomy constitutes a fundamental scientific operation for ordering complex phenomena (Gould & Purcell 1994). Bordes' artefact and assemblage typologies remain the only practical system for comprehensively comparing assemblage structure in the Middle Palaeolithic using quantitative methods.

Acknowledgements. We gratefully acknowledge the support of a University of Victoria Fellowship granted to Colin C. Moyer.

References

ADAMS, W.Y. & E.W. ADAMS. 1991. Archaeological typology and practical reality. A dialectial approach to artifact classification and sorting. Cambridge: Cambridge University Press.

BAXTER, M. 1994. Exploratory multivariate analysis in archaeology. Edinburgh: Edinburgh University Press.

BORDES, F. 1958. Le passage du Paleolithique Moyen au Paleolithique Superieur, in G.H.R. von Koenigswald (ed.), Hundert Jahre Neanderthaler: 175-81. Utrecht: Kemink en Zoon.

1960. Evolution in Paleolithic cultures, in S. Tax (ed.), Evolution after Darwin: 99-110. Chicago (IL): University of Chicago. The University of Chicago Centennial.

1969. Reflections on typology and techniques in the Palaeolithic, Arctic Anthropology 7: 1-29.

BORDES, F. & D. DE SONNEVILLE-BORDES. 1970. The significance of variability in Palaeolithic assemblages, World Archaeology 2: 61-73.

BOURLON, M. 1906. Les industries Mousteriennes au Moustier. Paper presented to the Congas International d'Anthropologie et Archeologie Prehistorique, Monaco.

CALLOW, P. & R.E. WEBB. 1981. The application of multivariate statistical techniques to Middle Palaeolithic assemblages from southwestern France, Revue d'Archeometrie 5:129-38.

DIBBLE, H.L. 1987. The interpretation of Middle Paleolithic scraper morphology, American Antiquity 52: 109-17.

GIDDENS, A. 1984. The constitution of society: Outline of the theory of structuration. Cambridge: Polity Press.

GOULD, S.J. & R. PURCELL. 1994. Pride of place. Science without taxonomy is blind, The Sciences 34: 38-9.

GRAYSON, D.K. & S.C. COLE. 1998. Stone tool assemblage richness during the Middle and early Upper Palaeolithic in France, Journal of Archaeological Science 25: 927-38.

MELLARS, P. 1996. The Neanderthal legacy. An archaeological perspective from Western Europe. Princeton (NJ): Princeton University Press.

1988. The chronology of the southwest French Mousterian. A review of the current debate, in M. Otte (ed.), La Technique. L'Homme de Neandertal: 97-120. Liege: Universite de Liege. ERAUL.

MELLARS, P. & R. GRON. 1991. A comparison of the electron spin resonance and thermoluminescence dating methods. The results of ESR dating at le Moustier (France), Cambridge Archaeological Journal 1: 269-76.

MOYER, C.C. 1998. The structure of Middle Palaeolithic variability. A multivariate assessment. Unpublished MA thesis, Department of Anthropology, University of Victoria.

PEYRONY, D. 1930. Le Moustier, ses gisements, ses industries, ses couches geologiques, Revue d'Anthropologie 40: 48-76, 155-76.

ROLLAND, N. 1977. New aspects of Middle Palaeolithic variability in Western Europe, Nature 266: 251-2.

1981. The interpretation of Middle Palaeolithic variability, Man (n.s.) 16: 15-42.

ROLLAND, N. & H.L. DIBBLE. 1990. A new synthesis of Middle Palaeolithic variability, American Antiquity 55: 480-99.

TURQ, A. 1989. Approche technologique et economique du facies Mousterien de type Quina: etude preliminaire, Bulletin de la Societe Prehistorique Francaise 86: 244-56.

COLIN CAMPBELL MOYER & NICOLAS ROLLAND(*)

(*) Moyer, Department of Archaeology, University of Cambridge, Downing Street, Cambridge CB2 3DZ, England. ccm24@cam.ac.uk Rolland, Department of Anthropology, University of Victoria, PO Box 3050, Victoria BC V8W 3P5, Canada. nrolland@uvic.ca

Received 30 January 2000, accepted 23 June 2000, revised 29 September 2000