Rock art dating in the Northern Cape, South Africa:
a review of dating techniques for rock engravings; and a consideration of the archaeological context of Driekopseiland
By David Morris, McGregor Museum, Kimberley
Go to opening page for note on copyright
"Ultimately what is of primary relevance about dating in archaeology
is not the age as such, but the temporal positioning of the
thing dated in relation to other archaeological material."
- Andree Rosenfeld & Claire Smith 1997:409.
Given current debates on the possibly differing cultural contexts of varying bodies of rock engravings and of finger paintings in the Northern Cape, establishing the archaeological associations and dating of engravings and paintings is of central significance in research. Neither the physical landscape (e.g. Butzer et al. 1979) nor the social history (e.g. Humphreys 1988; Beaumont & Morris 1990) of the region was static. Firm temporal linkages are critical to the interpretation of rock art relative to these and other variables (e.g. Morris 2002).
Dating, and the development of a regional chronology - what Trigger (1989:409) refers to as archaeology's "first problem", and one which it "can never outgrow" - is hence a principal objective for longer term rock art research in the Northern Cape. As Trigger asserts, "only insofar as archaeologists understand the order in which cultural factors change, do they have a basis for beginning to understand the causal relations linking them" (ibid.:409 citing Wylie 1985:77-8).
In rock art studies, the teasing out, and reliable documentation, of a relevant chronology has been notoriously difficult. Three decades ago Inskeep (1971:104) observed that South Africa's rock art was "a source of wealth not easily tapped." In the 1930s and 1940s the place of rock art studies at the core of archaeological endeavours here - notwithstanding their empiricist bent - seemed almost secure (e.g. Goodwin 1936; van Riet Lowe 1937; 1945); but, after the war, priorities were shifting. Other questions were being pursued by archaeologists, where new techniques such as radiocarbon dating could be applied; while a distinct pessimism was growing as to the future of rock art studies, in which there appeared to be no similar progress. "We find that at every turn we may take but a few steps," lamented Inskeep (ibid.:103), "before the path becomes too treacherous to tread." The impasse, where the lack of temporal control was a primary obstacle, was summed up in his remark that "on the shelter walls, and in their floors, we have two worlds which cannot yet be brought together" (ibid.:102). The art in its "flat manifestation" (Humphreys 1971) had somehow to be separated out through time if its study was going to proceed beyond a "pre-stratigraphic" stage and be integrated fully into a broader archaeological perspective. This was a matter of basic empirical (not empiricist) concern.
Mazel (1993) was making the same fundamental point a little over two decades later in answer to a critique of "chronocentricism" - the suggestion that "much of the art can be sufficiently dated to allay the misgivings of all but the most incorrigible chronophiles" (Lewis-Williams 1993:49). In fact paintings (and engravings) of domesticates, that potentially linked associated art into limited time-spans, are found only within restricted spatial distributions; while the number of direct dates for art of earlier periods was minimal. Against this, "educated guesses" (Lewis-Williams & Dowson 1992:18) were based on paper chromatography dating of the late 1960s (Denninger 1971), considered unreliable (Mazel 1993). "In essence," suggested Mazel, "we are no nearer to dating the majority of the Natal Drakensberg paintings...we are not yet in a position properly to integrate the rock paintings with the 8000-year-old Natal Drakensberg hunter-gatherer historical record known from surveys and excavations, except perhaps in an uncertain way for the ad 1830-1870 paintings" (ibid.:891). Russell (2000:61) recently commented that the absence of a relevant chronology effectively consigns nearly the entire corpus of rock art to the same status as a heap of unprovenanced artefacts. Indeed, at a Northern Cape site such as Driekopseiland, a sequence of cultural material of Pleistocene to late Holocene age flanks the engravings; and it remains to be proven which assemblage (if indeed any of those directly alongside the site) best matches the engravings (which may themselves be of variable age).
Ways of dating rock engravings
Instead of focussing primarily on the difficulties of rock art dating, the main advancements in rock art research in South Africa since the 1960s have been made along different trajectories, revolutionising thinking beyond the "game of ethnological 'snap'" of which Inskeep had disapproved, and elucidating much concerning the meaning of the art (Vinnicombe 1976; Lewis-Williams 1981; Lewis- Williams & Dowson 1994). There has nevertheless been some progress with dating the art, or with determining its broad archaeological contexts - both in South Africa (e.g. Butzer et al. 1979; Thackeray 1981; 1983; Thackeray et al. 1981; van der Merwe 1982; van der Merwe et al. 1987; Morris 1988; Beaumont 1989; Morris & Beaumont 1994; Yates & Jerardino 1996; Mazel & Watchman 1997; Ouzman & Wadley 1997; Jerardino & Swanepoel 1999) and internationally (see reviews by Rosenfeld & Smith 1997; Rowe 2001). Indeed the number of dates obtained, together with the development of new instrumental dating methods, from the 1980s, has invigorated research (Rowe 2001); even if the initial promise of direct radiocarbon dating techniques (e.g. van der Merwe 1982; van der Merwe et al. 1987; Loy et al. 1990) has proven often to be illusory because of continuing technical difficulties (Nelson 1993; Rosenfeld & Smith 1997; Rowe 2001; Dorn 2001; Whitley & Simon 2002a, 2002b; pers comm David Whitley). Spectacular claims, as were made initially at Jinmium in Australia, have given way to revised estimates (e.g. Owen 1999); while claims for a recent age, based on more than one technique, were refuted in the politically charged case of Foz Coa in Portugal (Zilhao 1995). Tempering his review of "substantial and exciting progress" (2001:5), Rowe suggests that "rock art dates in general should be considered provisional pending further research" (ibid.:10); while Whitley and Simon (2002a:13; cf. Whitley & Simon 2002b) stress, in the context of the AMS petroglyph dating controversy, that "substantial basic research must be conducted to determine when it is working, and why, as well as when it does not work, before it is applied to empirical cases."
Prospects (or present constraints) for dating Northern Cape rock art sites are assessed by way of a review of the major current techniques.
Several dating techniques are based on measuring the time it takes for rock coatings and varnishes - commonly referred to as 'patination' - to form or grow on the surface of an engraving. Indeed one of the oldest relative dating techniques was based on the notion that the deeper the colour of patination, the older the engraving. Today a detailed understanding of the way rock coatings form shows that 'patination' colour is not a reliable indication of relative age, since many factors other than time are involved.1 Apart from technical aspects of rock coating formation itself - the variety of potential such coatings, their chemistry, and their possible interdigitation on a single surface - other variable factors include underlying rock type, water flow and water ponding dynamics, presence or absence of organisms such as lichen, corrosion, surface roughness, and proximity to soil (Dorn 2001:174- 175; Pope 2001; cf. Butzer et al. 1979:1201-1202).
But this understanding has contributed to the development of other potential dating techniques based on the stratigraphic principle that the minimum age of an engraving may be determined by dating that which covers it. Conversely, knowing the time when the support rock became available for making engravings can, in some instances (of which Driekopseiland may well be one), provide a useful maximum possible age. It is critical that the substance (or event) to be dated should relate as closely as possible in time to the engraving event; and, as obviously, that there be a means of calibrating measurements against a time scale.
Numerous methods have been proposed, but none is by any means universally applicable. Biofilms, for instance, formed by organisms such as lichen, may have growth rates that can be calibrated (e.g. Joubert et al. 1983) - as has been shown in Arctic and alpine settings - but such calibration has proven far more difficult in temperate and arid regions, where most engravings occur (Dorn 2001:172-173). Some success has been reported with direct radiocarbon dating of calcium carbonate coatings over engravings (Dorn 2001:173); and Thermoluminescence (TL) dating was applied with informative results to stalagmite-derived samples covering Palaeolithic engravings at Venta de la Perra in Spain (Arias et al. 2000). The greatest snags have emerged in relation to AMS (Accelerater Mass Spectrometer) dating of organic aggregates encapsulated within rock coatings or varnishes (Dorn 2001:173-182; Whitley & Simon 2002; in press), with Dorn's retraction of the technique in 1995 - "an unfortunate set-back to a decade of petroglyph research" - sparking a controversy that continues in the present. Apart from difficulties inherent to radiocarbon dating itself, such as irregularities in the atmospheric production of 14C isotopes and their differential absorption in the biosphere, it was found that the organics within a rock varnish might be "heterogeneous in nature, age and therefore origin" (Whitley & Simon 2002) - some potentially far older than the engraving to be dated, and some far younger. Taken in 'bulk', samples of material in cases where such heterogeneity pertains would produce entirely spurious results - as they did in the Foz Coa dating projects (Zilhao 1995; Dorn 2001). By an odd chance the AMS radiocarbon rock varnish technique was first tested in a situation that inadvertently yielded valid time signals: its trial run was under geological conditions untypical of most empirical settings. One of the lessons, as Whitley and Simon (2002) stress, is that accidental 'false positive' results must be eliminated by running trial applications of new techniques in a wide range of contexts and conditions.
Another dating technique based on changes within a rock coating that can be calibrated against time is cation-ratio dating. The theory behind it is that the mobile elements Potassium and Calcium in rock varnish leach at a steady rate relative to the immobile element Titanium. Applied to geomorphological and rock engraving contexts in Australia, China, Peru, Russia, USA and South Africa, a key to its success is establishment of a reliable local leaching curve. Critics have pointed to factors other than time that may influence leaching rates in some environments. Dorn (2001:175) nevertheless suggests that its low cost and performance in blind tests lends it utility for preselecting samples for radiocarbon or cosmogenic nuclide (see below) analysis; and as an inexpensive cross-check on other methods. Cation- ratio dating was applied to rock engravings at Klipfontein, west of Kimberley. The results of that analysis (Whitley & Annegarn 1994) will be examined in detail presently.
Layering that results from rock varnish deposition under differing palaeoclimatic conditions is the subject of a new technique that distinguishes orange-yellow (Manganese-poor) layers, reflecting drier periods, relative to black ones (Manganese-rich) that signal wetter regimes. Correlated with palaeoenvironmental data for a given region, these microlaminations over engravings - visible in ultra thin sections under a transmission light microscope - have been used to separate out rock art of Pleistocene and Holocene age in the western USA, while the identification of a mid-Holocene wet pulse in Tunisia shows further promise for this low-cost technique (Dorn 2001:175-178; pers.comm. David Whitley). It may well have future applicability in the Northern Cape.
A further technique that may have some utility at a site such as Driekopseiland is one based on changes in a rock surface following its exposure - although the technique may not be able to resolve dates relevant to the engravings per se. The principle behind it is that cosmic ray bombardment produces cosmogenic nuclides in minerals, which can be measured and calibrated to yield maximum possible ages for engravings in certain situations. Its value was demonstrated in the Coa valley debate when it showed that the rock was geomorphologically stable enough to support engravings of Palaeolithic age - in the face of arguments to the contrary (Dorn 2001:169-170). Applications of the technique, comments Granger (1998), "are quickly diversifying to include estimation of erosion rates and soil formation rates, and to tackle more complicated problems of landscape development." At Driekopseiland it has been suggested that the glacial pavements first became available for the engravers at around 2500 years ago (Butzer et al. 1979), which may be too recent an event to register in terms of this method. A gravel unit at the base of the flanking +15 m deposits containing Acheulean artefacts indicates that the pavements were probably previously exposed, and hence cosmogenic nuclide readings on bedrock could possibly reflect evidence pertinent to other aspects of the site's history.
Controversy surrounds the method known as microerosion dating (Bednarik 1992; 2002; cf. Willcox 1984:8-9). The key to the technique is the assumption that minerals such as quartz and feldspar weather and erode at rates that can be measured and calibrated. In its application in the Coa valley - yielding age estimates of not greater than c. 6500 years - Zilhao (1995:891) suggested it was, as a means of dating, "at best...an interesting possibility" that remained to be tested, even at experimental level. Dorn (2001:171-172) identifies several inherent difficulties. The first is that it cannot be applied in any situation where rock coatings are present, or have been present, since these, by virtue of their chemistry, potentially accelerate or retard development of erosion phenomena; a possibility not factored into the method. Secondly, different varieties of quartz and feldspar may weather differently, and hence a locally relevant calibration curve would need also to be based on comparable varieties of these minerals. Thirdly, it is assumed that the minerals exposed by the engraving were not previously weathered, which, Dorn suggests, is almost never the case, unless the engraving was made in a new lava flow. At a macro level, geographers have used microerosion rates to arrive at qualitative assessments from many readings over extensive areas of rock (Dorn 2001; Rosenfeld cited in Zihlao 1995:891); but the type of quantitative information necessary for the dating of a specific rock art image, resolving many variables that may operate at any one spot, continues to elude weathering researchers (Dorn 2001, citing Dove 1995; Pope 2001). Bednarik (2002) continues to experiment with the technique.
Beyond these dating techniques, rock engraving researchers must fall back on the more traditional methods of sequences based on superimposed images, stylistic correlations, and the archaeological or geomorphological associations of particular engravings or sites.
Superimposed images can be a basis for constructing relative dating sequences, which have often then been extrapolated within and between sites in terms of stylistic characteristics. The principle in question is essentially that of stratigraphy - the upper layers of art being younger relative to those below - although it is impossible on the basis of superpositioning alone to say whether the interval separating these events amounts to minutes, years, or centuries (cf. Lewis-Williams 1974). Many writers caution against extrapolation even within a single site, let alone between sites, suggesting that sequences based on superpositioning are reliable only where specific individual images are in direct contact (Dorn 2001). Chippindale and Tacon (1993:35) introduced use of Harris matrices - originally developed for bringing order to complex excavated stratigraphies - to resolve possible sequence in complex painted panels in Arnhem Land. Apparent similarities between images in their analysis, which lacked, however, demonstrable linkages, were expressed as 'affinities' and indicated in the matrix as 'relations of equivalence' - a form of reasoned extrapolation. Harris matrix analysis has since been applied to painting sites in South Africa (e.g. Loubser 1993; 1997; Russell 2000). But its utility on engraving sites - where panels are usually far less complex and images, or restricted image clusters, are usually isolated on separate rocks - might not be as impressive. Linkages across a site would be correspondingly dependent upon 'relations of equivalence' that boil down to stylistic and other subjective judgments. At Driekopseiland, where the engravings do occur across vast sheets of rock, there are remarkably few superimposed images, and other potential criteria for assigning images within a matrix are compromised by variables such as differential exposure to water abrasion (Butzer et al. 1979).
Stylistic dating was recently brought under the spotlight - again, by the Coa valley rock art controversy - and given a degree of renewed respectability (Zilhao 1995). The debate prompted Rosenfeld and Smith's (1997) review of radiometric and stylistic methods of dating, and their argument, anticipated by Chippindale and Tacon (1993), that rock art researchers should recognise the two approaches as being complementary: "radiocarbon to probe and anchor stylistic sequences, and stylistic observations to identify problems and inconsistencies in radiocarbon dates" (Rosenfeld & Smith 1997:409). There is a need to move beyond simple correlations, they add, and to face, critically and determinedly, the problems and complexities inherent in both radiocarbon and stylistic methods of rock art dating.
In an important sense the question of style relates back to issues of reification and the 'culture model' (e.g. Binford & Sabloff 1982; Morris 2002), since the notion of 'style', in some of its classic applications has often been equated uncritically with a distinct 'people' in a particular place and/or time, producing a distinctive form of art. It was partly because of this kind of approach, but also in that 'style' had come to have "so many contradictory meanings" (Chippindale & Tacon 1993), or had lost analytical precision (Davidson 1996 cited in Rosenberg & Smith 1997), that it came to be rejected as a useful concept in rock art studies, and was replaced by what was heralded as a "post-stylistic era", in the early 1990s. Bednarik (1995:91), in most explicit, positivist posture, has dismissed style as "iconographically guided intuition", irreconcilable with objective scientific procedure. But that the need remains for some concept of style, suggest Chippindale and Tacon (1993:39), is apparent from the very re-invention of it by its critics: the difficulty, they contend, lies more in its definition, and chiefly in the way it has tended to bundle together distinct characteristics. "A figure has many traits - of subject, of size and scale, of pigment, of technique, of manner, of convention in perspective. By studying these separately and at the same time noticing which of them covary," they suggest, "we can better explore similarity and difference than under an all-embracing and ill-defined 'style'" (ibid.:39). Indeed, with Comaroff and Comaroff (2001:153), one might suggest, it is in the tendency to conceive of style as noun, rather than using it adjectivally, that an untenable structuring device emerges. For, as Rosenfeld and Smith suggest, 'style' is dynamic, and bound up in the historicity of social practice. Such a characterisation is strongly borne out in the anthropological literature on style in contemporary settings, notably in the Kalahari (e.g. Wiessner 1984) where it is seen inter alia as being part of an on-going negotiation and, potentially, contestation of personal and social identities.
Finally, archaeological excavations may provide clues to the age of rock art by association, such as in instances where engravings or paintings have been buried beneath datable deposit (Thackeray 1981; 1983). In other instances, probable associations between engravings and adjacent occupation/activity debris have been posited on the basis, in some instances, of consistent co-occurrence at many sites (Morris 1988; Beaumont & Vogel 1989). Erosional geomorphological settings may represent, as Dorn (2001:169) remarks, the inverse of excavation into sediments; and where it is possible to date erosion history, it is theoretically possible to arrive at maximum possible ages for engravings on surfaces exposed by that process. These more 'traditional' archaeological scenaria for the dating of rock art remain all- important, Dorn (ibid.) has stressed, as yielding independent test data for assessing other newer dating techniques.
The age of Northern Cape rock art
Rock art in South Africa has been a feature particularly, but not exclusively, of the Later Stone Age, with several secure associations with Oakhurst and, more commonly, Wilton Complex contexts (e.g. Thackeray 1983; Deacon & Deacon 1999). But, whereas perhaps most of the surviving art is of late Holocene age (cf. Jerardino & Swanepoel 1999), and some of the most recent art has been linked to the Iron Age and colonial eras (e.g. Maggs 1995; Ouzman 1999), the origins or rock art can be traced back into at least Middle Stone Age times (Wendt 1976; Thackeray 1983; McBrearty & Brooks 2000; d'Errico et al. 2001; Henshilwood et al. 2002; Beaumont in prep.), from which period there is also indubitable evidence for pigment use, most likely in ritual circumstances (recently reviewed by Watts 2002; cf. Beaumont 1992 for earlier possible instances).
At Driekopseiland the only systematic attempt to date the engravings thus far has been that of Butzer et al. (1979), who posit a late Holocene age for these engravings. On present evidence this seems not implausible; although on the face of it there are other possible archaeological associations, given the spread of Pleistocene to Holocene lithics in sediments immediately abutting the engravings (pers. comm. Peter Beaumont; cf. Whitley & Annegarn 1994). It is unlikely that any direct dating technique could be applied here - in view of the site's situation in an active river bed - but future work at other sites in the region may well provide closer circumstantial or supporting evidence on the age of the engravings.
Such age estimates as exist for the rock engravings of the Northern Cape are based mainly on 'traditional' archaeological and geomorphological associations. The excavations at Wonderwerk Cave near Kuruman yielded five "certain" and six "probable" small engraved stones bearing 'hairline' or 'fineline' images, mainly of a geometric nature, but with some figurative motifs, which came from layers dated to between about 2000 and 10 200 BP (Thackeray 1981; Thackeray et al. 1981). Previously, it had been deduced from superpositioning and differences in patination that 'hairline' engravings (stylistically distinct from a category of recently incised art) were older than 'pecked' engravings (Butzer et al. 1979), a conclusion that was essentially supported by subsequent work on the archaeological associations of rock engraving sites in the Upper Karoo (Beaumont et al. 1985; Morris 1988; Beaumont & Vogel 1989; Beaumont et al. 1995: though Busby et al. 1978, cited by Whitley & Annegarn [1994:190], suggest from replicative experiments that the stratigraphic relationships between fully pecked and fineline engravings may be difficult to ascertain empirically). The Karoo study indicated repeated associations, by contiguity, of mid-Holocene Wilton Complex and, in some cases, Oakhurst lithics with 'hairline' engravings; while relatively younger 'scraped' and 'pecked' engravings appeared by similar associations to be at least partly coeval. There was a suggestion in the art itself of some formal link between 'hairline' and 'pecked' engravings, the latter often made directly on 'hairline' outlines (Morris 1988:117): theoretically, indeed, the two techniques might sometimes have been part of a single engraving event. A major pulse of 'scraped' engraving activity (lightly patinated) was found consistently to appear alongside sites, in the Upper Karoo, of a blade-dominated 'Swartkop' industry, dated to some 250-700 BP; but other 'scraped' engravings looked older than these and could in some instances be associated spatially with earlier 'Swartkop' assemblages (Beaumont & Vogel 1989). What distinguished this study was the large number of radiocarbon dates obtained for surface lithic scatters associated with rock engravings. Many of the readings, which were based on the carbonate fraction of ostrich eggshell samples, require to be adjusted in light of an initial deficit of 2.2% in 14C caused by the large intake of fossil limestone during egg laying (Vogel et al. 2001). On average, the readings appear to be some 180 years too old; and while dates based on ostrich eggshell now seem more reliable than those on bone, since the former material is less prone to contamination, the ostrich eggshell derived datings will always have an uncertainty of about 120 years (ibid.).
Previously, the combined criteria of superpositioning (particularly in relation to technique) and associations (both archaeological and geomorphological) were applied by the Focks and their colleagues (especially Butzer et al. 1979; Fock & Fock 1989) to assess, in a broad multi-disciplinary framework, the age and context of the engravings of South Africa's central interior. As with the Karoo study, elements of the methodology employed could be traced back to the 'typological' work of Burkitt (1928) evident in his writing, inter alia, on the engravings at Keurfontein, Vosburg. Goodwin's (1936) analysis of Keurfontein elaborated somewhat upon Burkitt's 'Series I - III', proposing no less than eleven "styles" grouped into seven "general phases" (ibid.:173). These were linked chronologically with 'Vosburg', 'Smithfield B', and 'Smithfield B-C' "folk", while the most recent styles originated in "the age of metals" (ibid.:208-209). Though "tentative", Goodwin was confident that the relative order of styles based on patination1 was in accord with that obtained from palimpsests (ibid.:168). Such sequences of styles were not readily applied from one area, or even site, to another, however, and rock art 'zones' or 'regions' came to be defined in the literature (e.g. Willcox 1984). Part of the strength of the study by the Focks and Butzer et al. (1979) was the recognition that the causes of spatial and temporal variability should be sought not so much in terms of regional styles and sequences in themselves, but relative to a more holistic framework that saw the art as a complex and dynamic tradition influenced by a number of potential social factors including those relating to environmental resources. Some of these remained to be counted in, or eliminated, by future contextual archaeological research (Butzer et al. 1979:1211).
In general the research programme of Fock and Butzer demonstrated that there was distinctive patterning in rock art distributions both within and between sites (which was not determined by macroenvironmental constraints); and that while the variability could reflect, in some instances "the presence of functionally discrete sites or site segments" and, in others, potentially distinct identity-conscious groups, there were also temporal "trends" and "shifts in emphasis" in engraving techniques and content. Subject matter in some instances - for example, depictions of domesticated cattle and ovicaprines - reflected the appearance of farming in the landscape (not that the 'embedded' meanings of this contact art are by any means self-evident, as Loubser and Laurens  have shown). In other cases, the range of semi-tropical species depicted in what is today a semi-arid environment appeared to be compatible with pollen evidence of palaeoenvironments that would have supported such faunas at certain periods in the past (again, the meanings of motifs were not what Butzer et al. addressed; but, as Manhire et al. [1985:162] have said, "it is possible to refer to what a painting is as well as to what it means, in the sense that many or almost all symbols are objects or activities selected from hunter- gatherer life").
The geoarchaeological contexts of key sites were used to argue that: the "main geometrics phase" most likely belonged to a period of stream down-cutting and drier climate after 1300 BP; that "classical" engravings "most probably coincided" with an era of more abundant water some 3200 to 2500 BP; while animal engravings of a "classical and related" category might fall between these periods. Earlier hairline engravings, it was suggested, could be linked with earlier phases of 'Smithfield' settlement following an apparent mid-Holocene hiatus in parts of the region (cf. Deacon 1974), or even with earlier Oakhurst or terminal Pleistocene occupation (Butzer et al. 1979:1211). In sum, Butzer et al. (ibid.) saw "the several rock art genres of southern Africa [as forming] part of the archaeological record of the flexible cultural system represented by the Southern San. Systematic temporal and spatial variation of the engravings appear to reflect processual change and distinct identity-conscious groups within that system."
Finally, in this review of efforts to establish the age of Northern Cape rock art, the only direct dating attempt on engravings here thus far has been that by Whitley and Annegarn (1994). Their cation-ratio analysis produced some surprisingly early dates, with a temporal spread spanning the entire Holocene for a range of individual engravings in each of the sampled "pecked representational", "outline-pecked representational" and "geometric" types. Forty four engravings at Klipfontein and two at Susana were sampled. Varnish development was found to be only moderate, the coatings "relatively thin", so that the technique favoured selection of engravings that were more heavily varnished and hence probably older (Whitley & Annegarn ibid.). Twenty of the Klipfontein and one of the Susana samples yielded enough varnish for bulk chemical analysis using PIXE; and none had sufficient varnish for replication and assessment of error margins. Apart from these limitations, a further constraint identified by the authors concerned the derivation of a leaching curve. This was based on just two points in time, namely on aeolian deposit on the modern surface at the site, equivalent to approximately 100 BP, and an upthrust varnished cobble embedded in the top of Member IV of the Riverton Formation flanking the Vaal River nearby. For purposes of the study Member IV was estimated to have been laid down by 8000 BP (Helgren cited in Whitley & Annegarn 1994), "with the embedded cobble dating to some time after that point." The resulting age-assignments could not be considered as other than "maximum-limiting ages", Whitley and Annegarn (1994:193) suggest; and they caution further (ibid.) that "the cation-leaching curve (and numerical ages on engravings calculated therefrom) should be considered preliminary."
Two points have been drawn by Whitley and Annegarn from the results of the cation-ratio dating at Klipfontein. The first is that the spread of dates for geometric engravings in particular was consistent with expectations based on Lewis-Williams and Dowson's (1988) neuropsychological model, namely that such geometric motifs, as entoptic phenomena, were indeed 'signs of all times'. The second is that, considering the mean dates for each style, the results were still to some extent in line with the chronological synthesis of Butzer et al. (1979). Possible sampling error (including small sample size) remains as the major constraint limiting the value of these dates, which have been calibrated against a leaching curve of uncertain reliability (the Riverton Member IV estimate of "a little after 8000" [Helgren 1977:306] is based on an unspecified suite of radiocarbon dates known in places to be problematic [Butzer et al. 1973:355]). Moreover, minimal information is given on the particular engravings that were dated, so that where there may be 'stylistic' trends - or indeed significantly distinct cultural subdivisions - within, for example, the 'geometric' corpus (e.g. Smith & Ouzman in press), these cannot be evaluated on the basis of the published results.
1. Goodwin (1936) was not unaware of this, devoting a section of his paper to a review of the diverse processes involved in weathering and patination. He cautioned that "it is not time which produces patina, but the various factors...acting alone or together, and more or less intensely over varying periods" (1936:166). See also Goodwin 1960.
Environment, history and rock art: on the dating of Driekopseiland's rock engravings
Driekopseiland was one of the key sites on which the synthesis of Butzer et al. (1979) was based. The site offers a singular opportunity for geomorphological age estimation for rock engravings, its flanking sediments reflecting recent geological history and human activity in Pleistocene to Holocene times. Its detailed stratigraphy (Butzer et al. 1979) can now be supplemented with new evidence, providing for the following scenario:
Bedrock - and the exposed surface of it that came to be festooned with rock art at some point in its history - consists of glacially scoured Ventersdorp andesite, in places overlain by Dwyka tillite and shales.
Three alluvial sediment Units (1-3), containing fresh Acheulean artefacts, overlie bedrock, forming a terrace 20 m above the Riet River low water. Preserved discontinuously at the base are Unit 1A 'Younger Gravels'/river floor conglomerates (Helgren 1979:274) which contain rolled Acheulean lithics (ibid.; Butzer et al. 1979:1207) and occasional fossil fauna (pers. comm. Juri van den Heever to E.A. Voigt, 19911). Helgren (ibid.) Tentatively equates these conglomerates with Rietputs C along the Vaal. It appears that the glacial pavement here was probably exposed in places in a riverine setting where gravels accumulated in Unit 1A times, at some point in the Middle Pleistocene. The impressive silt sediments of Units 1-3, with Unit 4, indicate subsequent long periods of "aggradation or floodplain stability", contemporary with Acheulean and subsequent Middle Stone Age presence in the landscape (Butzer et al. ibid.). These accumulations undoubtedly represent Pleistocene Interglacials, one of which may well have been the Holsteinian/Oxygen Isotope Stage 11, at around 405 000 years ago - generally regarded as the warmest of the Pleistocene Interglacials (Beaumont et al. 1992:495, citing van Coevering & Kukla) - which may well have promoted the conditions that resulted in the highest +20 m terrace at Driekopseiland.
Unit 4, which is embanked against the older alluvia, forming a terrace at +13-15 m, consists of two generations of flood silts separated by an episode of channel cutting and soil formation (Butzer et al. 1979). The authors (ibid.: note 86) relate this Unit to Member III of the Riverton Formation. The younger of these channel beds has an associated minimum date, based on shell, of 38 500 1150 years BP (Table 3) - which, being close to the limits of radiocarbon is probably tantamount to an infinite 14C reading. It is also to be noted that high degrees of younger carbonate contamination pertain in these settings and were earlier acknowledged by Butzer et al. (1973:355; cf. Humphreys & Thackeray 1983:273) to be "a serious liability in all samples." In fact it may be suggested that the unit in question more likely relates at least to Oxygen Isotope Stage 5e (at around 125 000 BP), as has been proposed for the similar overbank 'Older Silts' at Pniel 6, which are also characterised as Riverton III (Butzer et al. 1973), also contain freshwater shell, and, like the Driekopseiland Unit 4, have associated Middle Stone Age (MSA 1) artefacts (Beaumont 1990a; Beaumont et al. 1992).
There is evidence both at Driekopseiland and at Pniel 6 for a decline, after this, in human occupation; and following MSA 3a times along the Vaal (Beaumont 1990a) there is indeed a total lack of archaeological traces for the remainder of the Late Pleistocene (Beaumont et al. 1992). This drop in archaeological visibility is matched by pollen evidence to indicate that the Pleniglacial period was cooler and drier than at present, with increased desert in the west and grasslands with fewer trees in the central and eastern parts of the subcontinent (Beaumont et al. 1992).
Unit 5A is an alluvial terrace at +10 m, containing no artefacts, which Butzer et al. (ibid.) see as correlating with other alluvial fills of terminal Pleistocene or early Holocene age; with its equivalent in the Riverton Formation being Member IV (Butzer et al. 1979). Given that the datings obtained by Butzer et al. (ibid.) (see Table 3), may well be problematic, further work is needed to interpret the palaeoenvironmental correlations and implications of this unit. It is not clear, for instance, whether there is an equivalent at Driekopseiland of the 'Younger Silts' at Pniel 6, which date most likely from Holocene Climatic Optimum times, c. 8000-5000 years ago.
Unit 5B is a 'Drab Sand' regional marker horizon that is colluvial in origin, having been swept down onto the Unit 5A +10 m and exposed Unit 4 flood plain. It grades locally into gray vertisols. The dating of this unit relies on regional extrapolations (Table 3). Dates based on ostrich eggshell may to be more reliable than those based on soil carbonates (Vogel et al. 2001; cf. Butzer et al. 1973:355; Humphreys & Thackeray 1983:273). Butzer et al. (1979:1207) remark that "Smithfield occupation [at Riverview] is verified midway in the related sediments" (Butzer et al. 1979:1207), while the Weltevreden burial dated to 3360 60 (Humphreys 1974) probably also relates to the LSA occupation contemporary with Unit 5B (Butzer et al. 1979 note 92).
Unit 5C is a +5 m alluvial terrace, apparently lacking artefacts, and estimated by Butzer et al. (ibid.) to be about 2250 to 1300 years old. They equate it with Riverton Formation Member V.
Unit 5D is a 'Pale Cover Sand' which, like the 5B 'Drab Sand', is another widespread regional marker horizon, of aeolian origin that overlies Units 5B and 5C. It dates between 1220 and 800 BP and commonly includes 'Smithfield' occupation debris.
Unit 6 consists of loose sands up to +3.5 m, which represent sub-recent bedload. This unit has been augmented by flood surges of recent decades, and includes secondary deposition of sediments derived from Unit 1B which was significantly cut back in places in the late 1980s-early 1990s, when the river by-passed the weir at its southern end.
On the basis of their reading of the geomorphological evidence, it appeared to Butzer et al. (1979) that bedrock, having been buried for most of the Holocene, began to be exposed in places as a result of rapid channel-cutting, from a moist flood-plain at +7 m, after 2500 BP. Renewed aggradation of flood silts and sands took place between about 2200 and 1300 BP, stabilizing as a mainly dry flood-plain 2-3 m above the present river level, with little or no bedrock exposure. Channel cutting recommenced after 1300-1200 BP, resulting in considerable bedrock exposure - with relatively minor silt build-up and erosion since then. "As a general framework," they predicted, "the Driekops Eiland engravings could only have been executed c. 2500-2200 BP or after 1300 BP" (Butzer et al. 1979; Fock & Fock 1989:141).
Reading together the environmental history and the evidence of the engravings themselves, the highly abraded representational (animal and human) images and 'older' geometric designs (75% of 1543 motifs) restricted mainly to higher, convex surfaces in the western (downstream) part of the site probably date from an earlier period, most likely (in terms of the above scenario) c. 2500-2200 BP, when, seemingly, only portions of bedrock were exposed. The less abraded predominantly geometric engravings (99% of 2004 motifs) in the flatter eastern part of the site, 120 m upstream, by contrast, appear younger, exhibit some formal differences from the 'older' geometrics, and were probably made (in terms of this model) from perhaps the end of the first millennium AD onwards, after much more of the glacial pavement was exposed (Butzer et al. 1979; Fock & Fock 1989).
Since the late 1970s progress has been made in defining the environmental history of South Africa's interior, so that the interpretation of Butzer et al. (ibid.) can be, if not yet fine-tuned, then certainly qualified relative to new data.
The climate proxy record based on shifting oxygen and carbon isotope values and the grey index of a well dated stalagmite from Cold Air Cave in the Makapansgat Valley (Lee-Thorp et al. 2001) provides the most remarkable high-resolution indication so far of Holocene climate change in the Southern African interior. The evidence derived from it of lower d13C, high humate (grey index) and positive d18O values before 5200 years ago suggest warmer more humid conditions than today, with denser vegetation-cover, during the Holocene Climatic Optimum (Lee-Thorp et al. 2001:4508). Independent pollen data from Wonderkrater, with temperature proxy data in the Uitenhague aquifer, and Mozambique Channel sea surface temperatures similarly suggest warmer (which in the summer rainfall region also means wetter) conditions in the period approximately 8000 - 5000 years ago (ibid.). In the Northern Cape, the coprolite-derived pollen spectrum from Stratum 1a (lower) at Equus Cave shows that after about 8000 BP a tree-rich savanna, indicative of warmer and wetter conditions than now, replaced a 12 000-9000 BP karroid open Acacia grassland (Beaumont et al. 1992; cf. Scott 1987). This pattern is mirrored in the terminal Pleistocene-early Holocene levels at Wonderwerk Cave where both pollen evidence and that from micromammals shows that Strata 4c-4b, dated about 8000 - 5000 BP accumulated under warmer and wetter conditions that climaxed at circa 6000 - 5000 BP with a woodland savanna setting (Beaumont et al. 1992; cf. Avery 1981). There is further corroboration at Kathu Pan 2, also based on pollen evidence, while the 'Younger Silts' at Pniel 6 appear to reflect the impacts of the same phenomenon in a riverine setting (Beaumont 1990a; Beaumont et al. 1992). The contention that bedrock at Driekopseiland was buried in the first part of the Holocene seems plausible in light of this evidence, the more so if Unit 5A can be confirmed as relating to this time.
The Cold Air Cave stalagmite indicates that a more variable moist/warm climate prevailed from 4300 to 3200 years ago; but this was followed by a rapid high amplitude shift, marked by a sharp vegetation change that is evident in d13C values and the grey index series. The trend thereafter was towards increasing aridity and cooling, with two clusters of Oxygen isotope minima at 3100 and 2600 years ago respectively. In this period C4 grasses became widespread, their abundance peaking about 2100 years ago. However, C4 grass abundance dipped after 1400 years ago, at which time there is evidence for a marked drying and cooling trend which culminated in the strongest Oxygen isotope minimum 247 years ago [AD 1750]. "The cooling phenomenon generally known as the Little Ice Age," comment Lee-Thorp et al. (2001:4509) "was manifest in southern Africa as a long-lasting low- rainfall anomaly, accompanied by cooling." This trend was, however interrupted by the highly variable Medieval Warm Epoch between AD 1000 and AD 1300 (Tyson et al. 2000).
Partly because temporal controls at sites relevant to this period are not generally adequate for resolving palaeoenvironmental variables, the matching of Cold Air Cave data with trends noticed in other sites is more difficult. However, at Wonderwerk Cave from 5000 to 3000 BP there is evidence - both from climate- sensitive small mammals and pollen spectra - of dry savanna grassland conditions; and some indication of oscillating wetter and drier periods thereafter (Avery 1981; van Zinderen Bakker 1982; Beaumont et al. 1992). A similar scenario is apparent at Kathu Pan 2. The colluvial Unit 5B 'Drab Sand' at Driekopseiland could well reflect such an environment of correspondingly increased aridity, following the wetter Holocene Climatic Optimum.
Beaumont et al. (ibid.) suggest that there were wetter spells at Wonderwerk Cave at about 1900-1400 BP and 900-800 BP, evidenced by travertines in the sequence (Beaumont 1990b). These dates show remarkable concordance with evidence reviewed by Tyson and Lindesay (1992), Huffman (1996) and Vogel and Vuls (1999), who posit warmer, wetter events in the first half of the first millennium AD, and again, in the Medieval Warm Epoch (Tyson et al. 2000) - which is now quite accurately dated (suggest Vogel & Vuls 1999:100) by way of Iron Age dates in the otherwise eschewed Limpopo-Shashi region - from AD 890 to AD 1290. Similar warmer/wetter conditions could have led to a degree of renewed aggradation at Driekopseiland as posited in Unit 5C (the Medieval warming has been found to be reflected in the sedimentary record of Namibian rivers - Tyson et al. 2000:122).
Finally, the "marked drying and cooling trend" of the last millennium evident at Cold Air Cave (Lee-Thorp et al. 2001) is matched by signs of increasing aridity at the same period in the micromammal evidence at Wonderwerk Cave (Avery 1981:269). The Unit 5D 'Pale Cover Sand' regional marker horizon of aeolian origin noted at Driekopseiland would seem to correspond well with such a "long-lasting low-rainfall anomaly" - one of "some consequence throughout the subcontinent of southern Africa" (Tyson et al. 2000:125) - when channel cutting is reckoned to have exposed bedrock and left open the expanse of rock on which the second major episode of engraving activity took place. The culmination of this trend at about AD 1750 was followed by a rapid shift from dry/cool conditions to wet and warm, in a period of less than 50 years (Lee-Thorp et al. ibid.).
In summary, this review of the model of Butzer et al. (1979; Fock & Fock 1989) relative to subsequent palaeoenvironmental research (Fig. 2), shows that, in all probability:
* All the engravings at Driekopseiland were made during the Holocene, given that environmental conditions here were largely unfavourable for human occupation for several tens of millennia before that. Indeed the last episode of high archaeological visibility locally, prior to the Holocene, was seemingly during Oxygen Isotope Stage 5e, some 125 000 years ago, when up to 15 m of sediment filled the river bed here.
* Aggradation of silts, up to +10 m, took place during warmer and wetter conditions which peaked at the Holocene Climatic Optimum, probably obscuring bedrock until after the mid Holocene.
* Downcutting may have commenced after 5000 years ago, possibly becoming more accentuated after about 3000 years ago, resulting, at an uncertain point in the later Holocene (possibly including the second half of the first millennium AD), in the partial exposure of the higher convex surfaces of bedrock, where engravings were then made that included geometric and animal imagery.
* Warmer/wetter interludes, possibly those known from the early first millennium AD and/or the Medieval Warm Epoch, may have resulted in renewed aggradation of silts to about +5 m in the Riet River channel.
* The marked low-rainfall anomaly associated with a somewhat drawn-out Southern Hemisphere Little Ice Age evidently led to the exposure of a far greater extent of bedrock, upon which the 'younger' geometric engravings were then made. Some of the engravings are below the modern low-water level, suggesting that "the river was prone to drying up almost completely at the time the bulk of the geometric designs were produced" (Fock & Fock 1989:143).
Inskeep (1971:102) had written that "...on the shelter walls, and in their floors, we have two worlds" which could not yet be brought together. Several projects since the end of the 1960s have enabled the beginnings of an engagement of these 'two worlds' - with work at Driekopseiland showing that a geo-archaeological approach in some relevant situations has the potential to yield useful results - even if the dating itself remains for now somewhat coarse-grained.
A changing social landscape
Still further data require to be integrated, however; for, at the same time that the physical landscape at Driekopseiland was changing - the glaciated rock surfaces becoming exposed in the river channel - indications are that the social landscape was also in a state of flux (e.g. Humphreys 1988). In this history the environment plays no doubt a significant role, but is just one of many variables (cf. Mazel 1989:25-26).
A first detailed approximation of the Holocene archaeology of the Northern Cape has been derived from research projects of the 1970s-early 1980s, the findings of which were synthesised by Humphreys and Thackeray (1983) in their monograph Ghaap and Gariep. It was based on studies of sequences from shelters along the Ghaap Escarpment (Humphreys 1979) and Wonderwerk Cave in the Kuruman Hills (Thackeray 1981), and assessed with reference to material from the Orange River Scheme, and in relation to other Holocene syntheses from the subcontinent. The major focus was on identifying trends and patterns in lithic artefact and related technologies, as a necessary initial step towards elucidating a history of the Holocene in the region. Humphreys and Thackeray (1983) show that in the early Holocene, from before 10 000 to about 8500 years BP, the cultural sequence is represented by a Kuruman Industry, a regional expression of the wider Oakhurst Complex, typified by large scrapers. (The 'Smithfield A' of earlier literature would be broadly its equivalent). The earliest of the Wonderwerk Cave engraved stones is associated with this part of the sequence. Material that is characteristic of the Wilton Complex, but with local industries as yet awaiting definitive differentiation, replaced the Kuruman Industry and persisted through the middle and late Holocene to protohistoric times. Distinguished by the use of a wider variety of raw materials than in the Kuruman Industry, and by the presence of a range of retouched artefacts such as scrapers and backed tools, its features change gradually through time and exhibit continuity in stone tool manufacture after the introduction of pottery in the west and south west some 2000 years ago. Humphreys (1979) had found no evidence to warrant the separate definition of a Smithfield Complex for Orange River Scheme material dating from the last 500 years (Sampson 1974), and this material is subsumed by Humphreys and Thackeray (1983:281) within the Wilton Complex pending future possible temporal and spatial differentiation.
All the material described as 'Smithfield', 'Smithfield B' or 'Smithfield C', or indeed as 'Wilton', in the earlier literature on Driekopseiland, is effectively part of this Wilton Complex of middle to late Holocene age. In chronological terms it represents the period in which the engravings at Driekopseiland were made. As mentioned, the industries in question are by no means static (indeed, a "very variable feast", is how Parkington has characterised the Wilton - 1980:83); their variability through time and space signalling processes that are as yet not fully understood. Two recent assessments of the diversity in post-8000 BP lithic assemblages in South Africa (Mitchell 1996; Wadley 2000) identify essentially two kinds of variability: inter- regional stylistic variability; and intra-regional, inter-site activity or functional differences, with one scenario being a site undergoing a change of 'place' (sensu Parkington 1980) in relation to changing social or environmental contexts. Latterly the practice of a herding life-style was emerging in the wider region, and ceramic technology appeared during the first millennium AD (Beaumont & Vogel 1984; 1989). A distinctive pattern of Type R stone-walled sites evolved later along the Riet River, dating between about AD 1380 and 1780, and distributed mainly further upstream, but with isolated settlements along the lower Riet (Maggs 1971; Humphreys 1972; 1988; 1997; Brink et al. 1992). During this period, apparently after 350 BP [AD 1640] (Vogel & Fuls 1999), there was a significant expansion of Late Iron Age settlement onto the Southern Highveld (Maggs 1976; Vogel & Fuls ibid.). Frontier complexity was considerably augmented by the early nineteenth century and following the upheavals of the Difaqane period: the historical records indicate that 'Bushmen', Griqua, 'Bastards', 'Bechuana', Xhosa ('Caffers' in contemporary sources), Boers, and travellers from the Cape and between missionary centres, lived in, or passed through, the valley of the Riet (Humphreys 1972; 1997).
Pertinent evidence on the ground at Driekopseiland consists of late Holocene lithic scatters on the banks overlooking the engraving site (Battiss 1948; Van Riet Lowe 1952; Willcox 1963; Butzer et al. 1979); at least two burials (Battiss 1948; Mason 1954; Butzer et al. 1979; Humphreys 1987); and a Type R stone-walled settlement unit at the foot of the hills some way off (Du Toit 1964; Humphreys 1972; 1987; Fock & Fock 1989). Van Riet Lowe and others characterized the lithic scatters as 'Smithfield B' and Wilton - in other words, Wilton Complex (Humphreys & Thackeray 1983). Associated with these scatters, in places, are relatively rare sherds of fine grit-tempered pottery. Substantial assemblages of probably comparable ceramics occur (along with evidence of livestock) at Blinkklipkop and Doornfontein near Postmasburg from at least c. AD 790 to around AD 1630, after which time the pottery becomes coarser (Beaumont & Vogel 1984:82, 92; Beaumont et al. 1995). By contrast, much thicker sherds from essentially undecorated pots and bowls are associated with the Type R stone walling (Humphreys 1972; 1988).
Whereas ceramics on Later Stone Age sites may signal a herder presence, it is not clear to what extent (and when) pottery and livestock remains in a site reflect true pastoralism (Sadr 1998). Debate revolves equally now on the question of whether migration of pastoralists (e.g. Smith 1990a; Boonzaier et al. 1996), or diffusion of ceramic technology and herder practices (Sadr 1998; Sadr & Smith 2001; Sadr & Plug 2002), account for the emergence of these elements in Later Stone Age sites in the last 2000 years. In Sadr's (1998) view, the seemingly unsynchronised first appearances of sheep remains and of pottery, together with the lack of a clear 'stylistic chain' linking the Cape ceramics with those in a putative area of migratory point of origin, count against the migration theory. So, too, might the formal continuity in Later Stone Age lithic artefacts after the appearance of sheep and pottery (Sadr & Smith 2001). But there is evidence, which Sadr (1998) reviews, for a possible later migration by Khoekhoe speakers at the end of the first millennium AD. A major obstacle to the spread-by-diffusion scenario has been the theoretical contention that the sharing ethic amongst hunter-gatherers would tend to prevent the adoption of herding. But Sadr shows that one of the few traits common to all Kalahari San groups is that of flexibility - in social and political spheres as well as in lifestyle (cf. Guenther 1999) - and that such fluidity and opportunism would have enabled hunter-gatherers to adopt and maintain livestock if circumstances made this desirable. Evidence from south-eastern Botswana (Sadr & Plug 2002:5) indicates a situation where foragers may well have considered livestock as "equivalent to socially unimportant smaller game, rather than to the meat of 'significant' large and medium game", suggesting that the sharing ethic would not need to be compromised by herding on this scale. Indeed, Sadr (1998) questions the extent to which true pastoralism generally arose in South Africa, in the light of evidence of usually less than substantial quantities of livestock on 'herder' sites. To what extent, he enquires, were Later Stone Age groups with livestock and pottery, in fact, real herders? Schrire (1980) had previously pointed to repeated paradoxical references in the literature of the Cape interior to so-called 'Bushmen' hunter-gatherers keeping stock.
The Type R settlements along the Riet River have been accounted for by Humphreys (1972; 1988) in broadly these kinds of terms; of erstwhile hunter-gatherers having adopted pastoralism, and being engaged in trade and cultural exchange with other groups, probably Sotho-Tswana to the north (see Brink et al. 1992; Humphreys 1997). An analogue for the process resulting in the form of their stone-walled settlements is given with reference to twentieth-century stock-keeping Dobe !Kung (Humphreys 1988; cf. Yellen 1984) whose settlement organization underwent change, with similar eventual spatial layout, after the incorporation of herding into their lifestyle.
But there may be indicators for immigration of a distinct identity conscious group at some sites in South Africa dating from about 900-1200 BP, where there is evidence of new elements of ceramic style (lugged pots), changes in lithic technology (more macrolithic, a drop in formal tool percentages, coarser raw materials; also grooved grinding stones), and in the sizes of ostrich eggshell beads (Sadr 1998; cf Kinahan 1986; Jacobson 1987a; 1987b; Smith 1990b:11) . Even here, though, in situ cultural change cannot yet be ruled out (Sadr 1998). In the Northern Cape some assemblages with ceramics are distinguished by these features (Morris & Beaumont 1991; Beaumont et al. 1995).
The evidence of burials along the Riet River could have a bearing on the question as to whether changes in local Wilton Complex industries, and the introduction of ceramics, reflects some degree of population replacement, or the kind of process that Humphreys has posited. Numbers of such burials, having various features in common, are linked in several instances with the Type R sites, by virtue of their placement, their dating and the nature of grave goods (e.g. copper objects). But these conform to a burial pattern noted in a grave from Weltevreden, near Driekopseiland, dated to 3360 60 (Humphreys 1974), prompting Humphreys (ibid.:271) to suggest that the Riet River burials represent "a relatively stable aspect of a cultural system extending from at least 3000 years ago." The human remains themselves cannot be distinguished morphologically beyond the level of 'Khoisan', but Morris's (1992:102) analysis of the Riet River sample - including the Weltevreden burial - indicates that they do constitute a morphologically homogeneous population, with a certain amount of anatomical distinctiveness: "the Riet River population as a whole experienced a degree of genetic isolation from other South African groups," he concludes. It is also to be noted that mortuary data for the Riet River sample are consistent with what Morris distinguishes as San burial practices - and none with those linked specifically with Khoekhoe (ibid.: 70).
Against this background, it would appear not only simplistic, but indeed unwarranted, to suggest - as some earlier writers have done - that the animal engravings and 'older' geometrics at Driekopseiland were made by 'Bushmen', linked with the 'Smithfield B' sites, while the 'younger' geometric motifs were the work of Korana [Khoekhoe] or Korana/Bush hybrids (see discussion in Chapter 3). Rather, it seems likely that a complex 'mosaic' (Beaumont & Vogel 1984; Humphreys 1988) of economic, technological and ideological responses became manifest in the region through the last 2000 years, characterized by a variety of social and cultural interactions between differing subsistence modes (Humphreys 1988; cf. Denbow 1990; Maggs & Whitelaw 1991; Lewis-Williams & Blundell 1997: note 17). Pottery or stone tool assemblages, rock art, or other aspects of material culture are, in this view, unlikely to be a straightforward reflection of social entities such as ethnic or even techno-economic groupings (Jacobson et al. 1994a; 1994b; cf. Deacon & Deacon 1999:184). In fact, it has been shown that the ethnic groups recorded in the colonial era might well be of uncertain time depth and perhaps even of doubtful relevance (Humphreys 1998). In the nineteenth century, 'Korana', for example, became something of a catch-all frontier category - at least in colonial literature; different from 'Korana' herders along the middle Orange River in the previous century (Barnard 1992; cf. Legassick 1979:251). The term 'Bushman' as a clearly identifiable group with its own culture, languages and physical type likewise became fixed in that era; and even terms such as 'Sonqua' and 'abaThwa', previously having more diffuse significances, were glossed in the colonial literature as the equivalents of 'Bushman' (Wright 1996).
Quite how the changes outlined may have impacted on rock-art traditions and beliefs locally is not clear. It has been suggested that regional differences in rock-art repertoires of the last two millennia may reflect different trajectories of change and forms of interaction between hunter-gatherers and other social groups in their respective areas (e.g. Prins 1991; Parkington 1996; Jolly 1998). Referring to such interaction, Wilmsen has pointed to the way the content of mythology "transcends time and tribe and ethnicity", indicating a complex history of social relations, through which elements of cosmologies "were constructed and transmitted in a less segmented social environment than presently exists" (Wilmsen 1986:358; cf. Schmidt 1979; Dowson 1998; Jolly 1998). In their relations with others, foragers may have been particularly receptive to outside ideas, suggests Guenther (1999:87), who also believes it is nearly impossible, in the Khoisan context, to sift imported beliefs and rituals from those that are not: "the 'Common Bushman' core of beliefs that constitutes the benchmark for differentiating indigenous from derived items is too varied and fluid," he adds, "to serve as a standard for evaluation." Schapera (1930) and Barnard (1992) have both noted "striking resemblances" in religion across the Khoisan spectrum, while Biesele (1993:34-37) finds San and Khoekhoe 'traditions' to be "practically indistinguishable." In his study of Tricksters and trancers: Bushman religion and society, Guenther (1999:87-88, 128-129) cites both San and Khoekhoe sources, noting how histories of contact will have resulted in myths permeating and blending across boundaries. One particular example is the story of the moon and the hare, which Bleek (1875:10, cited in Guenther 1999:128) characterised as a "veritable Hottentot myth" because of its apparent preponderance in Namaqualand (in fact, an artefact of missionary linguistic work in that region) - notwithstanding his having collected nine versions of the same myth from the /Xam San. Moreover, Bleek was unable to differentiate Khoekhoe from San elements in these latter versions. Lewis-Williams has similarly drawn attention to the close correspondences between Korana rituals and beliefs and those of the /Xam (Lewis-Williams 1981:105-106; Prins 1991). Guenther (1999:86-93) demonstrates that Khoisan were receptive also to the myths and even rituals of other neighbours including both Bantu-speakers and Europeans, with whom their interactions have been as substantial as they have been varied (e.g. Wright 1971; Vinnicombe 1976; Lewis-Williams 1981; Campbell 1987; Prins 1990; 1991; Prins & Lewis 1992; Mazel 1992; Jolly 1992; 1996a; 1996b).
It is in relation to this contact period reality of "profound commonalities and extensive social relations" that Lewis-Williams and Dowson (1994:207, 220) have moved away from considering a 'pan-San cognitive system' as an appropriate unit of analysis for rock art, recognising rather a "subcontinental cognitive set" (ibid. 220); one where "some beliefs are pan-southern African in that they are held by people other than the San" (Lewis-Williams 1988:86-87).
Remarkably, aspects of the beliefs in question have been sustained to this day in the lives of some descendants of Khoisan in the Northern Cape, with a degree of regional variability (Van Vreeden 1955; 1957; Waldman 1989; 2001; Prins & Rousseau 1992; Hoff 1993; 1995; 1997; 1998; Lange 1998).
Environment, history and rock art
What emerges from this review of the environmental and archaeological settings of Driekopseiland and the wider region is that the historical context(s) of the engravings here probably have a measure of complexity commensurate with the array of processes that were unfolding in the region in the last 2000 years and more. Precise answers to the question of authorship remain elusive, with the options being, at most, between Khoisan groupings, where the influence of emergent pastoralism and contact with Iron Age groups may have played a role. But insights with respect to significant cross-cultural continuity in the sphere of beliefs and ritual, mean that the pertinent questions on diversity in the engravings here may in fact relate more to changing emphases in the expression of widely shared beliefs (cf. Lewis-Williams 1988), and including dynamic landscape temporalities (Ingold 1993) and variable contemporary uses of places, than to the relative merits of different ethnic authorships.
Establishing the temporal linkages between rock art and its archaeological and environmental contexts has been a key objective in this Chapter. If the dating itself leaves much to be desired, at least some understanding has been gained of the parameters involved. "Ultimately what is of primary relevance about dating in archaeology," suggest Rosenfeld and Smith (1997:409), "is not the age as such, but the temporal positioning of the thing dated in relation to other archaeological material." A broad contemporaneity between certain kinds of environmental processes, historical trends and rock engraving events has been suggested for this site. This temporal positioning hints at a dynamic interplay between history, rock art and environment, where Driekopseiland, more than a mere location in space, has represented for people at different periods rather different kinds of opportunities and intersecting contingencies (Morris 1988:117; cf. Parkington 1980). It is suggested that one may look afresh at this engraving site by drawing upon what Ingold (1993) refers to as a 'dwelling perspective', where relevant ethnography may provide important insights.
1. A letter by J.A. van den Heever, Zoology Department, University of Stellenbosch to E.A. Voigt, McGregor Museum, dated 8 Aug 1991, and brought to my attention in June 2002, referred to the finding of a fossil elephant occiput at Driekopseiland by a Mrs Edith Schwenk. The accompanying sketch indicates a "sand bank" which is most likely in fact the alluvial conglomerate at the base of the Pleistocene sequence. A search in the vicinity in June 2002 revealed no further similar material.
8004visitors since 4 April 2003