Outlined in these terms, the tool reduction models of Rolland and Dibble are coherent, explicit and in certain respects highly plausible. The basic elements of the models are in good accord with expectations derived from ethnographic observations and with many features of the archaeological record itself. Ethnographically, the extensive resharpen-ing of stone tools has been documented in many different contexts and in some cases (most notably the Australian tula adzes) led to radical changes in the form and dimensions of discarded tools (e.g. Hayden 1979). Similarly, there is now evidence from several archaeological contexts not only that many totally unretouched flakes were employed for a range of functions (Keeley 1980; Beyries 1987; 1988a: 216-8; Beyries & Boeda 1983) but also that several tool types (including typical racloir forms) were systematically resharp-ened at one if not several points in their total use-life (Figs 4.3, 4.4: Cornford 1986; Lenoir 1986; Meignen 1988; Schlanger 1989). These observations leave no doubt that both the systematic use of unretouched flakes and resharpening of tools in the course of use was a regular part of stone tool production apparently during all stages of the Palaeolithic sequence.
The question here, therefore, is not whether the tool reduction models of Dibble and Rolland have some significant role to play in explaining the total range of variation in Middle Palaeolithic assemblages, but simply the scale on which the models can be applied. Some of the basic problems raised by the reduction models concerning the production of different racloir types in Middle Palaeolithic industries have been discussed in Chapter 4 and need not be repeated here. The critical question in the present context is how far the Dibble and Rolland models can be invoked to account for some of the much more radical contrasts in lithic assemblages which have traditionally been used to differentiate between the major Mousterian variants - notably the large variations in racloir frequencies which distinguish principally the Denticulate, Typical and Charentian variants of Bordes' taxonomy, and the occurrence of other taxonomically diagnostic features such as hand axes or backed knives (Bordes 1953a, 1961b, 1968a, 1981, 1984). Clearly, these are fundamental issues to any assessment of industrial variation in the Middle Palaeolithic and deserve close examination.
1. The first and most important point to emphasize is that the tool reduction models have never been seen as a way of accounting for all the differences which define the main industrial variants of the Bordes taxonomy. As discussed above, these models have been focused on two specific aspects: first, the varying production of racloirs in different assemblages, as a result of systematic reshaping of unretouched flakes; and second, on some of the finer variations in the specific forms of racloirs (i.e. the varying percentages of lateral, transverse, double and convergent forms). While these variations could in principle be used to explain many of the documented contrasts between industries of the Denticulate, Typical and Charentian groups, alone, they have no bearing whatever on the basic typological and technological features which have always been used to differentiate between the three most distinctive and sharply characterized variants of the Bordes taxonomy - i.e. the Ferrassie, Quina and MTA variants. As discussed earlier, the fundamental distinction between the Ferrassie and Quina variants is based not on the overall typological composition of the tool assemblages but on the basic primary flaking techniques by which the tools were produced - i.e. on the predominance of Levallois techniques in the Ferrassie assemblages and simpler, non-Levallois techniques in the Quina industries (Bordes 1953a, 1961b, 1981, 1984). Clearly, the tool reduction models of Dibble and Rolland have no bearing whatsoever on these underlying technological contrasts between the Ferrassie and Quina industries and can in no way be used to explain this particular dichotomy within the Bordes system. The same is equally true of the specific hallmarks which have always been used to define the MTA variant - i.e. the presence of characteristic cordiform hand axes and distinctive backed knife forms (Peyrony 1920, 1930; Bourgon 1957; Bordes 1953a, 1961b, 1981, 1984). Again, these features are totally beyond the range of the current tool reduction models and stand as unique, idiosyncratic features of the MTA industries as a whole. All of these points have been recognized by Dibble and Rolland, and have never been presented as a part of their general tool reduction models. Nevertheless, it is important to recognize that whatever credence may eventually be given to the tool reduction models, they leave some of the most basic and striking aspects of industrial variation in the French Mousterian totally unexplained.
2. The suggestion that tool reduction models could account for the varying frequencies of racloirs and notches/denticulates which distinguish the Denticulate, Typical and Char-entian industries is in principle much more plausible and this no doubt stands as the main potential contribution of the Dibble and Rolland models. There is little doubt that tool reduction patterns have some significant role to play in accounting for some of the documented industrial variations within these three groups, at least when seen from a broad geographical and chronological perspective. The question once again is on what scale? In their most recent publications Dibble and Rolland have argued that even the most dramatic contrasts between racloir and notch/ denticulate frequencies - i.e. those of the Ferrassie and Quina industries versus the classic Denticulate-type industries - can be explained entirely in these terms (Rolland & Dibble 1990: 485-6; Dibble & Rolland 1992; see also Rolland 1981: 28). As a specific example they quote the contrast between the later Quina-type Mousterian levels and the immediately overlying Denticulate horizons recorded in the later Würm II levels (layers 26—11) at Combe Grenal. They argue that a shift in the intensity of occupation patterns in these levels could potentially account for all the documented contrasts between the racloir versus notch/denticulate frequencies in these levels - i.e. for the shift from racloir-to-denticulate ratios of around 10 to 1 in the late Quina levels to less than 0.2 to 1 in the immediately overlying Denticulate levels (Dibble & Rolland 1992: Fig. 1.3; see also Rolland & Dibble 1990: 487-8, Table 2). As noted above, their suggestion is that during episodes of intensive, semi-permanent occupation, frequencies of various racloir forms would increase more rapidly in the assemblages as a result of progressive retouching of unretouched flakes than those of the notch/ denticulate group, eventually leading to a massive shift in the relative frequencies of these two tool forms in the tool assemblages (Fig. 10.4). Exactly why these major shifts in the intensity or duration of occupation patterns should have taken place between the Quina and Denticulate levels has not yet been explained. Nevertheless the implication that such changes could have led to the dramatic contrasts between the tool-type composition of the Denticulate and Quina assemblages in this part of the Combe Grenal sequence is clearly set out in Dibble and Rolland's paper.
However elegant and coherent these models may appear in a theoretical sense, there are a number of problems in attempting to reconcile them with some of the hard archaeological data in this particular case. Three observations are especially relevant:
(a) First, the basic mathematics underlying the Dibble and Rolland models lead to some surprising conclusions. If we assume, as they suggest, that all the observed contrasts between frequencies of racloirs versus notches and denticulates recorded in these levels can be attributed to the varying intensity of occupation patterns (i.e. the length of time during which each occupation horizon was occupied) then we must assume that the occupation intensity factor was at least 50 times higher during occupation of the various Quina Mousterian levels than during that of the adjacent Denticulate levels. This conclusion follows directly from their own mathematics (as illustrated in Table 2 of Rolland & Dibble 1990; see also Dibble & Rolland 1992: Fig. 1.3) and makes the minimal assumption that whilst the production (and therefore relative frequencies) of racloirs increased progressively with the duration of occupational episodes, the production of notched and denticulated forms showed no comparable increase over these intervals. If we assume that both racloir and denticulate production would have increased progrès-
sively - though at different rates - over these intervals, as Dibble and Rolland concede is a more likely scenario (see Fig. 10.4), then the contrasts in the occupational-intensity index needed to produce the observed dichotomy between the Denticulate and Quina assemblages becomes even more dramatic - probably in the region of 100:1. Exactly why there should be such massive contrasts between the intensity or duration of occupation patterns in these particular levels of the Combe Grenal sequence remains as yet unexplained.
(b) Presumably, if one were to envisage such massive contrasts between the relative intensity and duration of occupation episodes in the Denticulate and Quina levels, one would reasonably expect to see some reflection of this in certain other features of the archaeological material - for example in the total quantities of faunal remains introduced into the site, or in evidence for intensive use and re-use of hearths and associated hearth refuse in the different occupation levels. In fact, the available evidence from Combe Grenal yields few indications of such contrasts. As discussed in the preceding section, the ratios of identifiable faunal remains to stone tools documented in the various levels of Denticulate Mousterian at Combe Grenal cover broadly similar ranges to those recorded in the adjacent levels of Quina Mousterian and, when calculated as average values, show effectively no discernible contrasts between the Quina and Denticulate levels (Fig. 10.1). If the ratios are calculated to include the unre-touched flake components of the assemblages (on the assumption that these could have functioned as simple, non-reduced tool forms) then admittedly the figures do shift in favour of rather higher overall faunal ratios in the Quina than in the Denticulate levels -potentially by a factor of up to X5. Even this, however, is far removed from the contrast in occupation-intensity ratios implied by the basic mathematics of the Dibble and Rolland models.
Similar observations could be made for the evidence of hearths and other indications of burning in the Denticulate and Quina levels. Clearly, these features are difficult to quantify objectively, but from data provided by Bordes (1972) and from additional observations of Binford (1992, and personal communication) and Laquay (1981), it seems evident that evidence for the use and reuse of hearths is if anything even more conspicuous in the Denticulate levels at Combe Grenal than in the adjacent Quina levels. Binford (1992, and personal communication), for example, has pointed out that many Denticulate levels seem to include not only relatively large, frequently re-used hearths but also often associated evidence for large-scale 'sheet burning' affecting the occupation levels as a whole. Rolland himself (1988b: 168) has drawn attention to the evidence for extensive hearths and ash bands in the single Denticulate level at Pech de l'Aze II. Similar observations are suggested by the frequencies of burnt animal bones and teeth recorded by Binford (1984) and Laquay (1981: 420-2) in the occupation levels, which again suggest significantly higher frequencies of burning in the Denticulate than in the Quina levels. As in the case of faunal remains therefore, there is little evidence to suggest that the intensity or duration of occupation patterns was significantly greater in the Quina levels than in the Denticulate Mousterian at Combe Grenal - and certainly not on the dramatic scale which the current tool-reduction models imply.
(c) Finally, Rolland has quoted some interesting data on the dimensions of tools and unretouched flakes in the various Denticulate and Quina Mousterian levels at Combe Grenal, which appear to conflict directly with the stated implications of the tool reduction models (Rolland 1988b: 173, Table 9.4B). As shown in Table 10.1, these reveal that the average lengths of both racloirs and unretouched flakes are appreciably smaller in the
Temperate very wet
Figure 10.6 Frequencies with which different types of Mousterian assemblage are associated with different climatic regimes in western Europe, according to Dibble & Rolland 1992 (see also Rolland 1990). Note that with the exception of the 'temperate, very wet' category characteristic of the west Mediterranean coastal zone, all types of industry are found in association with all climatic regimes.
majority of Denticulate Mousterian levels at Combe Grenal than in those of the adjacent Quina levels (see also Fig. 4.5). Both of these observations would seem to be in direct conflict with the patterns of tool and flake reduction envisaged for the Denticulate and Quina industries. Clearly, if the largest available flakes were being deliberately selected for retouching into racloirs in the Quina assemblages (as Rolland and Dibble imply: 1990: 485; also Dibble 1988a: 193) then the remaining unretouched flakes in these assemblages should presumably be significantly smaller than in the typologically non-reduced Denticulate assemblages. By the same token it should follow that various racloir forms documented in the Quina-type assemblages should be significantly smaller than in the Denticulate industries, since these tools are (by definition, in the Dibble and Rolland models) more heavily reduced than in the Denticulate industries. In both respects, the recorded artefact dimensions seem to reveal the reverse of the pattern that would be predicted from the tool reduction models. Seemingly the most contradictory feature is the much smaller average size of the racloirs in the Denticulate than in the Quina industries, since this would imply that intensive reduction of the tools (in the Quina assemblages) somehow made the tools larger!
Similar conflicts are apparent in some of the other data cited by Dibble and Rolland for the Denticulate and Quina levels at Combe Grenal. Two features referred to repeatedly in earlier publications as general criteria for identifying intensive reduction patterns in Middle Palaeolithic industries are sharp reductions in the sizes of discarded cores (to achieve maximum economy in the use of available raw materials) and a progressive increase in the frequencies of denticulates compared with notches in tool assemblages, reflecting repeated resharpening of single notches into more complex, multiple-notched forms (Dibble 1988a: 190; Rolland & Dibble 1990: 485-7). Again, the available data on these two features from the Denticulate and Quina levels at Combe Grenal seem to show the reverse of the patterns predicted by the lithic reduction models. Rolland has pointed out (1988b: 196) that the average sizes of cores recovered from the Quina levels at Combe Grenal are substantially larger than those from the adjacent Denticulate levels (hardly an index of heavy reduction or economizing on raw materials in the Quina levels), while according to data provided by Dibble (1988b; Fig. 10.6) ratios of denticulates to notches are in general somewhat higher in the hypothetically unreduced Denticulate levels than in those of the hypothetically heavily reduced Quina horizons (Fig. 4.11). Again, these apparently contradictory observations are passed without comment or explanation in the recent tool-reduction explanations for the Quina/Denticulate Mousterian dichotomy at Combe Grenal.
3. Potentially the most interesting correlations proposed by Dibble and Rolland are those which relate the varying intensity of reduction of particular industries to certain features of the environmental or ecological contexts of the sites. They suggest that heavily reduced industries tend to be encountered primarily either in areas where lithic raw material supplies are in short supply, or where climatic or other environmental factors would have reduced the mobility of the human groups and therefore the frequency of access to available raw material sources (Rolland & Dibble 1990: 484-90; Dibble & Rolland 1992). In both cases there is assumed to be a close correlation between the degree of reduction and the ecological or environmental context of the sites.
Some of the correlations proposed by Dibble and Rolland - such as the tendency for more lightly reduced industries to occur predominantly during mild climatic episodes (Fig. 10.6), or the occurrence of heavily reduced industries in areas far removed from raw material supplies, such as the Zagros region of southwest Asia - are unquestionably interesting and worth pursuing further (Rolland 1981: 29-34; Dibble 1984b, 1991a,b; Dibble & Holdaway 1989; Rolland & Dibble 1990: 486-90). However there are again problems in reconciling these patterns with the totality of the available data from the French sites. As Rolland has pointed out (1988b) there is no very simple correlation between the intensity of tool reduction and the availability of raw materials at specific site locations. At many sites in southwestern France (most notably at Combe Grenal) the complete spectrum of variation from heavily reduced (i.e. Ferrassie and Quina) to almost totally unreduced (i.e. Denticulate) assemblages can be seen to occur in a single location, where the availability of local raw materials must presumably be seen as an effective constant throughout the different occupation episodes (see Bordes 1984: 169). Equally significant is the occurrence of a number of heavily reduced industries in contexts where abundant and high quality flint supplies were apparently readily available on, or closely adjacent to, the occupation sites. Striking illustrations of this are provided, for example, at the site of Combe-Capelle Bas in the Couze valley (Peyrony 1943; Bourgon 1957), or at sites such as Biache-Saint-Vaast, Champvoisy and Riencourt-les-Bapaume in the flint-rich regions of northern France (Tuf-freau & Sommé 1988; Tuffreau 1988b, 1993). Why it should have been necessary to employ heavily economizing tool-resharpening strategies in these and similar contexts remains unclear. Conversely, other non-reduced (Denticulate) industries have been recorded from contexts (such as Arcy-sur-Cure in the Yonne and Roc-en-Pail in Maine-et-Loire)
where local flint supplies were apparently scarce (Girard 1978; Gruet 1976).
There are similar problems in attempting to pin-point specific associations between the degree of reduction and particular combinations of climatic and ecological conditions (Fig. 10.6). As Dibble and Rolland suggest, it may well be significant that many hypothet-ically reduced industries from the southwestern French sites seem to correlate primarily with the extremely cold climatic conditions of the Würm II (i.e. isotope stage 4) phase (as reflected by the various occurrences of Quina and Ferrassie industries), while the bulk of the industries from the milder, Würm I phase (= isotope stages 5a-d) belong to either the Typical or Denticulate variants (Rolland & Dibble 1990: 488-90; Rolland 1981: 29-34, 1988a: 177-8; Jelinek 1988a: 20-8). However, these correlations are by no means exact. Thus several of the Denticulate Mousterian horizons in the later Würm II levels at Combe Grenal demonstrably date from a period of very cold climate (Bordes et al. 1966; Laville et al. 1980), while in other contexts heavily reduced, Charentian-like industries have been recorded from periods of relatively mild climate - for example the Ferrassie-type assemblages from Rescoundudou (Aveyron) and Biache-Saint-Vaast (Jaubert 1983, 1989; Tuffreau & Sommé 1988). Some of the proposed correlations between the relatively unreduced industries and the occupation of open-air sites as opposed to cave and rock-shelter sites are equally ambiguous. The discovery of a number of characteristic Char-entian industries at a range of open-air sites in southern France (Chinchon, Puycelsi, Rescoundudou, Plateau Baillard, Champlost etc.) is particularly significant in this regard and shows clearly that the production of heavily reduced industries was by no means confined either to enclosed cave/rock-shelter sites, or locations where local flint sources were in short supply (Sireix & Bordes 1972; Tavoso 1987a; Jaubert 1983,1989; Le Tensorer 1973; Farizy 1985). The pattern of correlation between lithic reduction intensity and the associated environmental context of the assemblages is therefore at best rather vague and hardly shows the kind of close, direct correlations which the tool reduction models would seem to require.
Was this article helpful?