Bois Laiterie Cave and the Magdalenian of Belgium.

Straus, Lawrence Guy ; Otte, Marcel

The interest in conducting an excavation of the small site of Bois Laiterie Cave in the context of the Belgian Magdalenian lies precisely both in its size and in its chrono-cultural attribution. The study of the Belgian Magdalenian is currently in ferment. More generally, the question of the Tardiglacial recolonization of northern Europe has stimulated much new research. Just within the last few years a large number of radiocarbon dates have been produced on materials associated with Magdalenian artefacts in several cave sites on the edges of the Belgian Ardennes. Twenty dates from nine sites place the Magdalenian occupation of Belgium within the traditional temporal range of Bolling between 13 and 12.2 kya (uncalibrated). If only AMS dates on artefacts or cut-marked bones are considered, the range is even narrower: 12.9-12.3 kya (Charles 1994; 1996; Germonpre 1997; Housley et al. 1997). We are thus looking at a relatively short 'moment' in time, perhaps no more than one climatic phase, some 600 years, during which humans, re-extending their range after the southward contraction imposed by the Last Glacial Maximum, re-found and re-learned how to use the environments and resources of the Meuse Basin. The nature and degree of permanency of their re-settlement in Magdalenian times have been the subjects of speculation and research since the phenomenon was first discovered and defined some 139 years ago by E. Dupont. Competing models of permanent occupation versus seasonal migration, major versus minor degrees of contact with and dependency on the Magdalenian territory of the Paris Basin, have been put forth ever since. We now possess a database that is adequate to begin to test these ideas.

Yet many of the cave sites that have been radiometrically dated recently were excavated long ago. Re-excavation of a few of those sites has provided much valuable information of all sorts, but often from very limited remnant deposits. The recent excavations at open-air sites in adjacent areas of Middle Belgium, Dutch Limburg and French Ardennes have provided a great wealth of information of lithic technology. Yet they are totally lacking in faunal remains and have poor chronological resolution. Cave sites are still required, even if the range of activities conducted therein clearly differed from that of the open-air flint quarry-workshop sites of Brabant and Limburg. Among the recently (re-)excavated cave sites, Chaleux and most others are all fairly large. Little was known about smaller cave sites. In addition to the larger, possibly residential cave sites and the open-air flint workshop sites, we needed smaller, possibly limited-activity cave sites. Bois Laiterie (BL), a small cave above the Burnot gorge near its confluence with the Meuse on the edge of the Ardennes upland in Namur Province (Otte & Straus 1997), filled this bill [ILLUSTRATION FOR FIGURE 1 OMITTED]. The cave is located at c. 120 m a.s.l. (c. 35 m above the valley floor). Its coordinates are 50 [degrees] 21 [minutes] 45 [seconds] N latitude and 4 [degrees] 52 [minutes] E longitude [ILLUSTRATION FOR FIGURE 2 OMITTED].

In addition to the above stated reasons, small sites are advantageous in that they can be totally excavated, thus obviating the skewing effects of sampling that inevitably come with the partial excavation of large cave sites. With BL we were able to recover essentially the entire lithic and faunal contents of a 'place' that had formed a part of the Magdalenian settlement system in Belgium. And because that place has some very definite physical characteristics of both positive and negative character, the nature of its human use could be hypothesized to have been limited. The location, dominating a gorge that is a strategic avenue of communication between the Meuse and the Meuse-Sambre interfluve plateau, suggests the importance of game spotting and ambush hunting at BL. But with its north-facing exposure, small area, steeply sloping bedrock floor and draught-producing upper mouth, BL is a dark, cold, uncomfortable cave, suggesting the hypothesis of short-term, limited-function human visits [ILLUSTRATION FOR FIGURE 3 OMITTED]. Theoretically, it is in such sites that archeologists are most likely to obtain a relatively clear 'reading' of site function, as contents should be fairly simple and unblurred by the accumulation of residues from many different kinds of occupations. The archeological decipherment of such small, possibly limited-use sites should be more straightforward than that of large, complex sites with their massive palimpsest deposits.

Site integrity

The study of site formation/disturbance processes in BL Cave reveals a series of rather perplexing apparent contradictions. Yet these must somehow be dealt with, not only to try to determine whether BL had one or many types and seasons of use, but also to try to understand its 'strange' faunal assemblages that combine both arctic steppe/tundra taxa and more temperate or humid, woodland-dwelling animals. The question is whether this 'combination' is the result of mechanical mixture or of azonal, mosaic cohabitation during the Late Glacial (e.g. FAUNMAP 1996).

On the one hand, most of the Magdalenian lithic artefacts have sharp, fresh edges without crushing or rolling damage. Antler and bone artefacts are very well preserved. Large mammal bones are in good condition and include some fragile elements. Bird and fish remains are also well preserved, indicating the absence of significant sediment movement. Spatial analyses show the existence of distinct activity areas involving at least fire and flint knapping (Straus & Martinez 1997). The fact that these had maintained their spatial integrity despite the cave floor's slopes suggests the absence of catastrophic solifluction or erosion by running water. The micromorphological analyses by Courty (1997) also argue against massive slope failure or runoff erosion, although localized movements caused by such processes as human trampling are not ruled out. Importantly, three accelerator radiocarbon determinations (one on an antler sagaie and two on individual bones) from Stratum YSS in different parts of the cave and at different depths yielded statistically identical dates of 12.6 kya (Charles 1994; Krueger 1997). This would suggest that the Magdalenian occupation was a short episode or series of closely-spaced episodes.

On the other hand, there are inescapable indicators of some disturbance in what at times may have been a rather wet, plastic sedimentary matrix [ILLUSTRATION FOR FIGURE 4 OMITTED]. Many of the stone slab manuports were found lying at steep angles or even vertically, and many of the fragments refit. There are lithic refits across the thickness of YSS. These facts suggest both vertical migration within the deposit and limited 'sliding' from the rear toward the exterior terrace, which makes sense given the slopes of the bedrock cave floor. The existence of 'preferred' orientations among blades and long bones lining up against the eastern cave wall, together with the steep inclinations of some of these elongated objects, are also indicators of movement (Straus & Martinez 1997).

Common sense and these facts require us to admit that the site is not 'pristine'. Nor, however, does it lie toward the totally redeposited, mixed end of the intactness spectrum. Astonishingly, the scatter of artefacts and faunal remains continues within YSS all the way to the rim of the bedrock ledge in front of the cave. Although completely exposed to the elements outside the cave proper, even this sector of the cultural layer survived remarkably intact, with distinct evidence of burning and knapping areas on the terrace. There must have been very rapid but gentle silt burial. So the situation at BL cannot be characterized as one of major disturbance.

Palaeoenvironments and faunas

BL provided a half-dozen sources of information on the environments during, before and/or after Magdalenian occupation. The micromorphological analysis (Courty 1997) tells of cold conditions with intense cryclastic activity during the formation of strata LGS (lower grey sand) and UGS (upper grey sand), which respectively underlie and overlie YSS (yellowish/reddish sandy silt) in the cave interior. These cold periods are tentatively attributed to Dryas I and II respectively. Primary loess was deposited in YSS times under somewhat less cold conditions, with especially warmer summers. There was greater seasonality than in the immediately preceding or succeeding periods. YSS sedimentary characteristics are fully consonant with the Bolling age indicated by the AMS dates. Given the northerly site location, it is not surprising that conditions during Bolling were still cool, albeit less so than in Dryas I.

It was pollen and wood evidence from the recent re-excavation of the Magdalenian horizon at Chaleux that first suggested the relatively moderate nature of Bolling environments, with the presence of localized woods or thickets in at least favoured microhabitats of the deep Ardennes valleys, alongside the continued existence of open steppe-tundra landscapes on the plateaux (Noirel-Schutz 1994; Schoch 1994; see also Leotard 1993, in regard to wood charcoal from the Magdalenian at the nearby site of Trou Abri). Limited analyses (Emery-Barbier 1997; Pernaud 1997) show the presence of trees: alder, hazel, juniper, pine and walnut among the BL pollen; charcoal of birch and alder, plus another tree in the birch family that could be alder or hornbeam. Pollen from the rose family and grasses, as well as fern and horsetail spores, were also found in Stratum YSS. All these taxa are represented among the pene-contemporaneous pollen and charcoal spectra at Chaleux. The combination of sedimentological and botanical data paints a very mixed picture for the Bolling, that is reasonable given the northerly (and still North Sea-less, continental) location of Belgium. It was still cool and loess was being deposited by the winds, but summers were relatively warm and there was enough local humidity in the valleys of Wallonia to permit the growth of a variety of both coniferous and deciduous trees, including some relatively warm-loving taxa probably confined to sheltered, south-facing slopes. Yet the woodlands had not spread to the plateaux of the Ardennes or Middle Belgium, which were still apparently covered with steppe-tundra vegetation. Because of Wallonia's relief, wooded microhabitats would have been adjacent to open grasslands; both types of environments and their respective resources were accessible from the Magdalenian sites.

It is in this context that one should not necessarily disbelieve the apparently anomolous co-occurrence of cold, dry, open steppe-tundra ungulates and others today associated with relatively temperate, often wooded environments, although it is also possible that the mix of species at BL (and at other Belgian Magdalenian sites) is the result of stratigraphic telescoping of faunas from both relatively temperate/humid and cold/dry phases, such as Bolling combined with Dryas I and/or II. Gautier (1997), who favours the palimpsest explanation and who disagrees with Straus' ecological mosaic interpretation, documents the presence in Strata YSS+BSC of reindeer, musk-ox (also present in radiometrically contemporaneous Magdalenian faunas found by Dupont at Chaleux and Goyet), bison, European wild ass and horse, along with red deer and moose. The first group (notably musk-ox) signifies the presence of open vegetation and cold, dry conditions, whereas the second group (especially moose) indicates locally much more humid conditions with woods. The presence of chamois (also sometimes a woodland-dweller) and especially ibex mainly indicates the existence of steep, rocky slopes (Van den Brink 1971). Both arctic and common fox are present in YSS+BSC. Such a faunal mosaic does not seem impossible given the local relief surrounding Bois Laiterie and the other cave sites of Wallonia. Although the glacial faunas were still present in Belgium during Bolling, it appears that the most favourable habitats of the region were being populated by taxa more at home in relatively temperate, humid woodlands. Humans, as part of this northward biotic displacement, preyed on all the ungulates available within a short radius of each site such as Bols Laiterie. Because it overlooks an obligatory passage between the 85-m a.s.l. Meuse canyon floor and the 250 m a.s.l. plateau, this cave was ideally situated for hunters to 'sample' species that probably lived contiguously, but ecologically separated by vegetation and terrain type in the complex mosaic situation of the Belgian Bolling.

The remaining sources of information on palaeoenvironments are microfaunal: terrestrial molluscs, lagomorphs, rodents, insectivores, bats and batrachians. Like pollen, but unlike wood charcoal or large mammal bones, these remains are not in the cave primarily due to human action. The samples of microvertebrates from Stratum YSS yielded the following information to Cordy & Lacroix (1997): an abundance of Dicrostonyx gulielmi and Microtus gregalis (a lemming and a vole, both cold, open-vegetation forms) and presence of Ochotona pusilla (pika, another cold, open-country dweller), together with Clethrionomys glareolus or C. rutilus (bank or ruddy vole, both today woodland dwellers) and Apodemus sylvaticus-flavicollis (yellow-necked field mouse: also a woodland taxon). Microtus oeconomus (root vole), M. arvalis-agrestis (common or short-tailed vole) and Arvicola terrestris (ground vole) are also represented, together with Sorex araneus-coronatus (common shrew). All these presently live in humid habitats, sometimes with substantial arboreal cover, the former being the 'coldest' in its modern distribution in western Europe (Van den Brink 1971). In short, the YSS micromammals are a mix of open, arctic species and more temperate, humid woodland taxa - like the macromammal fauna from this stratum.

The malacofauna from Stratum YSS (Lopez Bayon et al. 1997) includes only one strictly woodland-dwelling species (Discus ruderatus) and two semi-woodland species (D. rotundatus and Retinella hamonis). There are also a very few individuals of taxa that prefer humid or even marshy environments. On the other hand, there are also a few molluscs preferring open, sunny habitats and some xerophiles. Taxa tolerant of a wider range of environments are abundant and diverse. These facts again may suggest the existence of mosaic environments during the time of human occupation, with areas of woodland or thickets and local humidity, but close to drier, more open patches, perhaps on this north-facing slope. Indeed, one would expect the low-mobility landsnails to give the 'coldest', most open vegetation picture of all the environmental indicators. Humans may also have even further denuded the area in front of the cave mouth by their activity.

The birds represented in YSS (Deville & Gautier 1997) are consonant with a mosaic environment. There are open-country grouse and partridge (living on the plateau above the cave), water birds (geese and whimbrel, favouring the Meuse and Burnot stream below the cave), rocky cliff-dwellers around the cave (magpie, jay, passerines), and either a woodland-preferring owl (Asio otus) or (and?) an open vegetation one (A. flammeus).

In short, all the biological data can be interpreted to suggest either very different, but closely spaced, climatic phases that have been telescoped within the BL Magdalenian horizon, or genuinely very mixed ecological conditions in southern, upland Belgium c. 12,600 radiocarbon years ago. According to the latter hypothesis, while loess was still being deposited and the plateaux were still vegetated by steppetundra plants and the north-facing slope around the cave may have been partly bare, partly covered with low grasses and shrubs, the valley floor and south- and west-facing slopes of the Meuse and Burnot probably harboured not only conifers, but also some deciduous trees and bushes - made possible by higher summer temperatures and local humidity. This set of conditions nonetheless represented a dramatic change vis a vis the much more uniformly open or even barren environments of full Dryas I at 50 [degrees] N. These conditions seem to have made possible the human recolonization of Belgium and northwest Europe in general during Bolling in its traditional definition. Recent debates call into question the existence of a Dryas II phase (e.g. Kolstrup 1991; Barton et al. 1991; Sanchez Goni 1996; contra Cordy 1991), while referring to a Late Glacial Interstadial (i.e. Bolling+Allerod) interrupted by several short cold, dry pulses. Semantics aside, it was in the context of general amelioration of climatic conditions, with complex biotopes, that humans were able to reoccupy northwest Europe c. 13 kya.

Nature of the human occupations

BL seems to have had only one period of actual human habitation: during the Upper Magdalenian. While it became a Mesolithic burial place at 9.2 kya, there is no evidence for use of the cave either before or immediately after the episode(s) centred fairly tightly around 12-6 kya (uncalibrated).

Our interpretation of the Magdalenian uses of BL Cave is based on the lithic raw material composition, technological analysis and typology of its lithic assemblage, characteristics of its osseous artefacts, make-up and seasonality indicators of its macromammalian, bird and fish faunas, site structure and location information - and common sense. As detailed by Straus & Orphal (1997), almost all the chipped stone artefacts are of non-local flint. This high-quality material was probably brought to BL from one of the Upper Cretaceous chalk sources in Middle Belgium - either the Mons-Spiennes area of Hainaut to the West or the closer Orp area in Brabant to the north. (On-going research by R. Miller (pers. comm.) suggests that the former case is more likely.) The flint was mainly transported in the form of blade blanks. Only three small, exhausted cores were abandoned at Bois Laiterie and probably very few had been brought there, as there is very little cortical debris either - almost none of it primary. There are no hammerstones. The debris and implements are almost all diminutive, also indicating transport from a distance. The large number of tiny trimming flakes suggests in situ transformation of some blanks into tools and weapons, as well as resharpening or recycling. The high tool/debris ratio is added evidence that only the last stages of the lithic operational chain took place at BL. People arrived here with their light flint blanks and tools, and abandoned a minimum when they moved on to other places.

The tools are a specialized lot: many backed bladelets/small blades, plus several burins and only a few endscrapers and perforators [ILLUSTRATION FOR FIGURE 5 OMITTED]. Together with a few stone points, there are also four magnificent large antler sagaie fragments [ILLUSTRATION FOR FIGURE 6 OMITTED] and other bits of possible sagaies (Lopez Bayon et al. 1996). Since the backed blade(let)s are generally shown by microwear analysis to have been used as weapon tips or barbs (and there is some support for this here too; see Jardon 1997), all these data together support the hypothesis that Bois Laiterie was fundamentally a hunting camp. There is scant evidence of maintenance activities (e.g. hidescraping, wood-working), although the burins (with some microwear evidence for hard substratum scraping and grooving) may have been used in osseous weapon manufacture or reworking, just as the lithic assemblage composition suggests blank transformation and implement/armature resharpening. That animal carcasses were processed is suggested by the organic residue analyses of a few lithic artefacts (Newman 1997).

Spatial analyses (Straus & Martinez 1997) show the existence of a rudimentary site structure, with burning and flint-knapping areas in part corresponding to small slab-paved zones at and in front of the cave mouth, and a disposal area at the cave rear. However there is no evidence for elaborate organization or substantial investment in infrastructure [ILLUSTRATION FOR FIGURE 7 & 8 OMITTED]. The stone slabs are not very numerous and paved only a small surface in a potentially muddy area. They were available strictly locally on the Bois Laiterie hilltop and slopes. None are artistically engraved (Lejeune 1997), unlike in the cases of Chaleux, Frontal, Dasomme and Roc-la-Tour in the Ardennes or at Gonnersdorf and Andernach in the not-too-distant German Rhineland. There is no evidence of constructed hearths or other pits. Fires had simply been built on the ground surface. Bones, lithics and even broken sagaies were simply tossed toward the dark, otherwise little-used rear of the cave and presumably also down the steep talus in front of the cave. In short, there is no indication of major, long-term, multi-purpose residence at this site.

The macromammalian faunal assemblage (Gautier 1997) is small and, unlike that of Chaleux (dominated by horse), is quite diverse: reindeer (NISP = 70; MNI = 4), horse (NISP = 61; MNI = 3), ibex (NISP = 37; MNI = 2), musk-ox (NISP = 15; MNI = 2), ass, red deer, moose, chamois and bison (1 MNI each, with NISP ranging from 1 to 11) from combined Strata YSS+BSC. Such an assemblage probably speaks of opportunistic hunting of small numbers of animals during a few repeated visits to the cave, taking advantage of its strategic location on a cliff above the Burnet talweg connecting the Meuse and plateau. Humans not only took equids, bovines and cervids moving up and down that gorge, but also exploited the rocky slopes around the cave itself for caprines. Although none of the large mammals are represented by very large numbers of bones and teeth, the reindeer, horses and ibex (in contrast to the musk-oxen and other animals) do have enough remains (including thoracic bones) to suggest that more or less whole carcasses were brought into the cave, perhaps after initial dismemberment at kill spots. It is possible that other parts of the carcasses, especially meat and fat, were removed from BL to more commodious residential sites for consumption by larger human groups. In any event, the reindeer dental evidence (Stutz 1997; Gautier 1997) is unanimous in indicating warm-season (summer-fall) kills.

Some of the larger water birds may have been prey to humans, also possibly during the warm season according to Deville & Gautier (1997). Many or most of the smaller birds may have been prey to owls and foxes, both also present at Bols Laiterie. There is no evidence that humans had anything to do with the foxes, which probably used the cave as a den when humans were not there and vice versa. Arguments about fish availability during spawning suggest that humans were responsible for catching brown trout, burbot and grayling in late winter or early spring (Van Neer 1997). This might extend the time of human visits to the very end of the cold season or beginning of the warm season. Yet common sense still militates against the notion of humans spending very much, if any, time in BL Cave during winter.

In any event, Bols Laiterie was clearly a limited-use site, which saw one or a few short-term visits by humans during Bolling. Although such occupations were minor both in duration and in numbers of participants, some minimal investment was made in providing for a dry, solid living surface by paving a small area with locally available stone slabs. This paved area in the downslope strip at the front of the cave may have supplemented a bare bedrock floor upslope. But in no case, either in the small cave or on the narrow ledge in front of it, would there have been space for extensive occupations or activities,

The place of Bois Laiterie Cave in the Belgian Magdalenian system

The Magdalenian of Belgium (and contiguous enclaves of Netherlands and France) is represented by two kinds of sites:

1 Radiocarbon-dated cave sites along the northern and western fringes of the Ardennes and

2 Open-air sites on the low plateaux of Middle Belgium and on the high plateaux of the Ardennes. Only two of the open-air sites have been dated (imprecisely, by TL): Orp East (11.8-12.9 kya) and West (13.113.7 kya - all with standard deviations between 1-2 and 1.7 ky) (Vermeersch 1991). (TL dates are likely to be older by nature than 14C dates from the same deposits, as shown at Pincevent and Etiolles, where both methods were used; Valladas 1994.) The open-air sites fall into two general categories: loci at high-quality flint sources that seem mainly to have been quarry-workshops (Orp, Kanne, Mesch, Eyserheide, Schweikhuizen and possibly Obourg-Saint Macaire) and a site located at a strategic look-out location at the rim of a cliff dominating the Meuse-Semois confluence (Roc-la-Tour). Arguments (summarized by Rensink 1993) have been made that most (all?) of these sites were occupied during the traditional Bolling. (Cultural evidence could support either a ponecontemporaneous Magdalenian or slightly more recent 'Creswellian' age for St Macaire; Letocart 1970; Charles 1994: 77-8.)

The chronological data for the cave sites are most recently summarized by Charles (1994; 1996). The Ourthe Valley cave cluster south of Liege consists of Coleoptere, Sy Verlaine, Fontde-Foret and Walou. The first site has two conventional and one accelerator radiocarbon date that place the Magdalenian occupation in Bolling (i.e. 13-12.2 kya). The most reliable date (AMS on a cut-marked horse bone) from Verlaine is of early Bolling age. Walou has two conventional dates that, with their standard deviations, straddle the traditional 13-kya Dryas I/Bolling boundary. The Magdalenian cave sites of the Lesse-Meuse confluence area (Chaleux, Frontal, Nutons, Magrite, Abri, Dasomme and Vaucelles/Blaireaux) are probably all of traditional Bolling age, since all but Magrite and Abri have now produced radiocarbon dates that lie between about 12.9 and 12.3 kya. Germonpre (1997) has recently obtained AMS dates of 12.6 and 12.7 kya on cut-marked musk-ox and horse bones in the Dupont collection from Goyet on the Samson, a tributary of the Middle Meuse. The three AMS dates from Bols Laiterie (12-6 kya) fall squarely within the traditional Bolling time period. The dates from the cave sites suggest a main (or sole) Magdalenian colonization of the Belgian Meuse basin during a period of no more than about 600 years. The problem we face is how to correlate the cave sites with the open-air sites that are argued to also be of traditional Bolling age. Assuming that the two groups of sites were at least roughly contemporaneous and pertained to the same set of environmental conditions, we may possess samples from different aspects of a regional settlement system. Even if the open-air sites are not strictly contemporaneous with the cave sites (and this could only be demonstrated by intersite lithic refitting of the sort pioneered by A. Scheer (1993) among three Gravettian sites in the Ach Valley of southwest Germany), it is easy to imagine that there are/were other such sites either waiting to be discovered or long ago destroyed. What follows is the sketch of a model for an Upper Magdalenian settlement - subsistence system based on the territory of what is today Belgium, but with contacts both to the south (Paris Basin) and east (Neuwied Basin).

Seasonality evidence from dental cementum analyses of specimens from modern, controlled excavations in Chaleux (ibex), Trou Dasomme (ibex or chamois) and Bois Laiterie (reindeer, confirmed by dental eruption/wear sequence analyses) and reindeer from Dupont's excavation in Trou des Nutons, indicates:

1 winter-early spring and summer-fall kills (hence, presumably, human presence) in the upland Lesse-Meuse confluence area along the Ardennes and

2 summer/summer-fall kills around Bols Laiterie at the very northern edge of the uplands near the Sambre-Meuse confluence and the plains of Middle Belgium (Stutz 1997).

Following the logic that people would naturally prefer to take advantage of available shelter (i.e. caves) especially in winter, that game would seek shelter during winter in the protected, well-watered valleys of the Ardennes fringes rather than on the open, windswept plains of Middle Belgium, and that flint nodules would be difficult or impossible to obtain under the snow or in frozen earth, it can be hypothesized that the open-air Magdalenian sites of Limburg, Brabant and Hainaut were mainly occupied by people in the warm season. These places would be repeatedly visited specifically for their abundant, high-quality flint. The aim of flint-knapping activities at such sites as Orp and Kanne was the specialized production of laminar blanks (e.g. Vermeersch & Symens 1988). Our analyses (Straus & Orphal 1997) show that precisely those kinds of small blades/bladelets that predominate in the upland cave assemblages of Chaleux and Bois Laiterie are the ones that are under-represented at Orp and Kanne. This might suggest the selective transport of lightweight blanks from the quarry - workshop sites to the caves. The absolute scarcity of cores, cortical debris, crested blades, hammerstones or other evidence for primary reduction in the cave sites and, to the contrary, their abundance at the open-air sites, all clearly support an hypothesis of functional (and perhaps seasonal) complementarity between the two classes of sites. That flint transport was involved is indicated by the large average size of artefacts at the open-air sites and the very small average size thereof at Chaleux and BL. The general scarcity of backed blade(let)s and points (presumed weapon elements) at the open-air sites is in complementary opposition to their abundance at the cave sites - suggesting that, while some hunting must have been conducted around the open-air sites, it was not the principal activity at the quarry-workshops. In contrast, hunting was very important around the cave sites.

Given its geographic location and physical characteristics, BL might have been a transit camp en route between the open-air flint sources of Middle Belgium and the Ardennes. Though its lithic inventory is smaller and somewhat less diverse than that of Chaleux, there are strong similarities between their debris and tool assemblages - perhaps because BL was on the logistical supply-line between the flint sources and Chaleux (or sites like it in the uplands). Given the moderate distances involved (maximally 63 km from Chaleux to Orp and 80 km from Chaleux to Obourg/Spiennes via the easiest routes), humans could use the Ardennes caves at times even during the warm season and still go (or send logistical parties) to the flint sources also in summer. Logically, stricter seasonal constraints would be imposed on human use of the open-air sites, for reasons mentioned above. Better than no shelter at all, BL could have served both as an unspecialized hunting camp during the warm season and, if the fish were really caught by people, as a minor site at the very end of the cold season or beginning of the warm season.

Obviously we need seasonality information from Goyet on the border between the hill country and the Hesbaye plateau along the Middle Meuse, from other upland sites in the Ourthe drainage (such as Trou Walou), and from open-air loci - if any are eventually found that contain faunal remains. Likewise we need petrographic 'fingerprinting' of the flints in order to be able to distinguish with certainty among the very similar Upper Cretaceous flints of the Mens, Hesbaye and Maastricht areas (see Casper 1984), a task that so far has defied several serious efforts (P. Vermeersch pers. comm.). If this could be done, we would be able to determine where the people who used different cave sites procured their flints - presumably via either logistical trips and/or as a result of their annual residential moves (see discussion in Rensink 1993; 1995). The presumption of direct group visits to source areas is based on the relatively short distances involved. In the case of BL, the presence of pyrite, whose closest known source is in the Mens area (Lozouet & Gautier 1997), is another indicator of contacts with that flint-rich area, but does not prove the nature of the contact. The presence of pieces of pyrite at Chaleux (Otte 1994) could suggest that BL lay along the route between it and Mons. This brings us to the question of the fossil shells at BL and other penecontemporaneous Magdalenian cave sites.

Although Dupont (1873) had thought that the Magdalenian inhabitants of the caves in the Upper Meuse and Lesse valleys procured their flints in north-central France, this has been shown to be highly unlikely in the face of much more feasible and parsimonious 'Belgian' origins (Teheux 1994). Yet the fossil shells (64 in Chaleux, over a dozen at Frontal, one at Dasomme and unspecified numbers at Goyet - as well as at Verlaine and Coleoptere in the Ourthe drainage; Rensink 1993), also important to Dupont's argument, are now joined by 8 at BL (5-6 of which are artificially perforated) [ILLUSTRATION FOR FIGURE 9 & 10 OMITTED]. Most of these are probably from Eocene deposits in the Paris Basin, although one taxon's closest source is the Loire Basin, even further south (Lozouet & Gautier 1907). These shells are clear evidence of direct (visits/intermarriage) or indirect (exchange) contacts between the 'Belgian' regional band (perhaps a 'daughter' offshoot) and the Parisian area, location of the penecontemporaneous sites of Pincevent, Verberie and Etiolles. The fact that they are only found in the caves and not in the open-air sites is not an argument against possible contemporaneity of the two classes of sites or hence their potential participation in the same settlement - subsistence system. According to P. Vermeersch (pers. comm.), such fossils would likely not have survived in the loess milieux of sites such as Orp and Kanne. On the other hand, the fact that they are found both in major 'residential' cave sites such as Chaleux and Goyet and at minor cave sites such as Dasomme and Bois Laiterie helps to tie these sites together in one territorial system, into which the fossils were introduced as a consequence of social relationships and movements that linked the Belgian and Parisian areas and groups. The closest possible geological sources of the Paris Basin fossils are believed to be no less than about 150 km from Chaleux. It is one thing to transport small numbers of mostly small fossils (either in long, point-to-point trips or in down-the-line, group-to-group exchanges); it is a different matter to transport significant amounts of lithic raw materials over distances of this magnitude.

On the other hand, there are now known to be Gravettian-age fossils at nearby Goyet (Polymesoda convexa and Granulolabium plicatum) that are probably from Oligocene beds near Tongres in Belgian Limburg (Lozouet & Gautier 1997), suggesting connections, at least at that time, between the upland cave area and the flintrich plains (also the location of the Gravettian site of Huccorgne between Goyet and Tongres). Such connections between caves of the northern Ardennes flanks and quarry-workshop sites in Belgian and Dutch Limburg (the settings of Kanne, Mesch, Eyserheide and Schweikhuizen) no doubt also existed 15-12 ky later in the Magdalenian.

It has been convincingly argued (e.g. Bosinski 1988; Floss 1991) that substantial proportions of the flints used in the open-air sites of Gonnersdorf and Andernach at the northern end of the Neuwied Basin of the Middle Rhine came from Upper Cretaceous sources in the region of Maastricht-Liege (eastern end of the Hesbaye, a distance of at least 80 km). The Neuwied sites are radiometrically penecontemporaneous with the Belgian cave sites and most of the Paris Basin sites, although some researchers in Germany prefer to calibrate radiocarbon dates to ice cores and to claim that the northward migration had already begun in late Dryas I (Street et al. 1994). While there is no evidence of contacts between inhabitants of the Belgian cave sites and those of the Neuwied Basin sites, one can formulate the hypothesis that there existed in late Dryas I-Bolling at least three regional bands in this part of western Europe: Paris Seine-Oise-Basin, Belgian Meuse Basin and Neuwied Rhine Basin. The Belgian group was clearly not an isolate, yet it was separated and distinct from the Paris Basin group, from which it may have split off as ameliorating environmental conditions and food resources made settlement of the Meuse Basin possible.

Unlike earlier visions thereof, it is clear that the Magdalenian of Belgium (and northwest Europe in general) was not just dependent on specialized reindeer hunting. Although there is some tendency for horse to dominate the fauna of Chaleux, there was clearly a very wide diversity of game in the Bolling-age landscapes. The relief of southern Belgium provided rich food, lithic and shelter resources. Physical separation of the Belgian territory was apparently 'dictated' by the exposed, shelterless, flintless and then food-resource-poor plains of Champagne (Rozoy 1988), which people must have crossed quickly en route to/from the Paris Basin to maintain social contacts, seek mates, obtain possibly socially and/or symbolically important fossils, etc.

Thus two types of 'traffic' seem to have made up the human landscape of northwest Europe:

I procurement of non-local flints for tool manufacture either by means of

A embedding in group movements governed by subsistence activities

1 seasonal or

2 non-seasonal or

B essentially specialized logistical supply expeditions; and

II long-distance contacts whereby fossil shells (and presumably other things such as mates) circulated either by

A direct visits or

B down-the-line exchange.

Bois Laiterie helps to prove that there was at least one essentially whole Magdalenian settlement-subsistence system on the territory of Belgium in Bolling times and that its group members maintained contacts with the Paris Basin. The Belgian system was apparently self-sufficient in terms of game, fuel, some vegetal food, lithic resources and natural shelters. But demographically this (or these) regional band(s) may have depended on the ability to obtain mates among 'adjacent' bands, even if these were at some distance (see Wobst 1976). They may also have needed 'insurance policies': i.e., kin or fictive kin to whom they might hope to turn in times of crisis such as a game crash, severe winter, etc., as was done in the ethnohistoric Arctic.

BL, a small, uncomfortable but usefully situated cave, was part of this system. It was a good place to stop en route between the flint sources of the plains and the larger residential caves deeper in the hill country. It was a good place from which to hunt local caprines or a variety of other large game often likely to be moving between the Meuse and the inter-Meuse-Sambre plateau. It may have been a good place near which to fish in the Meuse or Burnot stream. Since its qualities were known and since it was at the intersection of two excellent routes between Upper and Middle Belgium, BL may have been visited several times - enough to warrant at least minimal paving slightly to ameliorate its uncomfortable living conditions. Lightweight flint blanks were brought to the site; some were transformed into tools and weapons. Antler weapons may also have been fabricated or reworked at the site, possibly using burins. Spent weapons were abandoned. But much of the flint may have been taken out by humans as they moved on to their next destination, although some that was left behind may have been reused/resharpened during possible later visits. Animal carcasses or parts thereof were also brought into the cave presumably from nearby killspots. Some may have been consumed in situ, but others may have been removed to other (residential) sites. The few endscrapers, perforators and needles testify to the fact that at least some other activities may have been conducted around the simple campfires in Bols Laiterie, such as clothes-making or repair. Meat or hides may have been cut against some of the stone slabs. Perhaps skins were scraped with ochre, traces of which were found on other slabs (Lejeune 1997). And then there are the bored circular stone and the eight fossil shells, most of which are perforated. Why they were abandoned or forgotten at Bois Laiterie, we will never know. Yet they testify to the fact that there were other aspects to the activities conducted at this small cave and that it was in turn part of a wider world. Belgium was on the fringe of that Magdalenian world and community. But it was a vital, viable fringe, thanks to its peculiarly advantageous combination of resources.

Acknowledgements. Research at Bols Laiterie was authorized by the Regional Government of Wallonia and the Township of Profondeville (Namur Province, Belgium). It was generously funded by the National Geographic Society (USA), the Regional Government of Wallonia and the Universities of Liege and New Mexico (USA). We wish to thank Philippe Lacroix, the discoverer of Bols Laiterie and indispensible member of the excavation and research team. We are grateful to all members of the 1994 and 1995 field crews, to Captain P. Francois of Namur who lodged them in 1995, and to all our scientific collaborators. Figures were drafted by A. Martinez and R. Stauber. The authors wish to congratulate Gerhard Bosinski on his 60th birthday and dedicate this article to him.

References

BARTON, N., A. ROBERTS & D. ROE. 1991a. Preface, in Barton et al. (ed.): xi-xii.

(Ed.). 1991b. The Late Glacial in northwest Europe. London: Council for British Archaeology. Research Report 77.

BOSINSI, G. 1988. Upper and Final Paleolithic settlement patterns in the Rhineland, in H. Dibble & A. Montet-White (ed.), Upper Pleistocene prehistory of Western Eurasia: 375-86. Philadelphia (PA): University of Pennsylvania Museum.

CASPAR, J.-P. 1984. Materiaux lithiques de la prehistoire, in D. Cohen & P. Haesaerts (ed.), Peuples chasseurs de la Belgique prehistoriques dans leur cadre nature]: 107-14. Brussels: Institut Royal des Sciences Naturelles de Belgique.

CHARLES, R. 1994. Food for thought. Unpublished doctoral thesis, Oxford University.

1996. Back to the north, Proceedings of the Prehistoric Society 62: 1-17.

CORDY, J.-M. 1991. Palaeoecology of the Late Glacial and early Postglacial of Belgium and neighbouring areas, in Barton et al. (ed.): 40-47.

CORDY, J-M. & P. LACROIX. 1997. Bio- et chronostratigraphie a partir des microvertebres, in Otte & Straus (ed.): 161-76.

COURTY, M.-A. 1997. Etude micro-stratigraphique, in Otte & Straus (ed.): 113-40.

DEVILLE, J. & A. GAUTIER. 1997. The avifauna, in Otte & Straus (ed.): 215-18.

DUPONT, E. 1873. L'homme pendant los Ages de la Pierre dans les environs de Dinant-sur-Meuse. 2nd edition. Brussels: Muquardt.

EMERY-BARBIER, A. 1997. Analyses palynologiques, in Otto & Straus (ed.): 141-2.

FAUNMAP. 1996. Spatial response of mammals to Late Quaternary environmental fluctuations, Science 272: 1601-6.

FLOSS, H. 1991. Sur l'approvisionnement des matieres premieres au Magdalenien et Paleolithique final en Rhenanie, in M. Otte (ed.), Los Bassins du Rhin et du Danube au Paleolithique Superieur: 104-13. Liege: Universite de Liege. ERAUL 43.

GAUTIER, A. 1997. The macromammal remains, in Otto & Straus (ed.): 177-96.

GERMONPRE, M. 1997. AMS dating of the Upper Magdalenian horizon I at Goyet Cave 3, Bulletin de l'Institut Royal des Sciences Naturelles de Belgique (Sciences de la Terre) 67: 167-82.

HOUSLEY, R., C. GAMBLE, M. STREET & P. PETTITT. 1997. Radiocarbon for the late glacial human recolonisation of northern Europe, Proceedings of the Prehistoric Society 63: 25-54.

JARDON, P. 1997. Analyse traceologique du materiel lithique, in Otte & Straus (ed.): 325-8.

KOLSTRUP, E. 1991. Palaeoenvironmental developments during the Late Glacial of the Weichselian, in Barton et al. (ed.): 1-6.

KRUEGER, H. 1997. Radiocarbon dating and isotopic analyses of the human remains, in Otte & Straus (ed.): 365-8.

LEJEUNE, M. 1997. L'art mobilier magdalenien final de la Grotte du Bois Laiterie, in Otte & Straus (ed.): 293-318.

LETOCART, L. 1970. Un gisement du Paleolithique final a Obourg 'St Macaire', Fruhe Menschheit und Umwelt 1: 352-61.

LOPEZ BAYON, I., E. TEHEUX, L. STRAUS & J.-M. LEOTARD. 1996. Pointes de sagaies au Magdalenien du Bois Laiterie, Prehistoire Europeenne 8: 125-41.

LOPEZ BAYON, I., P. LACROIX & J.-M. LEOTARD. 1997. Etudes des restes malacologiques, in Otte & Straus (ed.): 145-60.

LOZOUET, P. & A. GAUTIER. 1997. Coquillages fossiles et restes de 'briquet', in Otte & Straus (ed.): 319-324.

NEWMAN, M. 1997. Immunological analysis of lithic artifacts, in Otte & Straus (ed.): 329-36.

OTTE, M. (ed.). 1988. De la Loire d l'Oder. Oxford: British Archaeological Reports. International series S444(i).

(Ed.) 1994. Le Magdalenien du Trou de Chaleux. Liege: Universite de Liege. ERAUL 60.

OTTE, M. & L.G. STRAUS (ed.). 1997. La Grotte du Bois Laiterie. Liege: Universite de Liege. ERAUL 80.

PERNAUD, J.-M. 1997. Analyse anthracologique, in Otte & Straus (ed.): 143-4.

RENSINK, E. 1993. Moving into the north. Unpublished doctoral thesis, Universiteit Leiden.

1995. On Magdalenian mobility and land use in north-west Europe, Archaeological Dialogues 2: 85-119.

ROZOY, J.-G. 1988. Le Magdalenien superiour de Roc-la-Tour I dans le contexte franco-belgo-rhenan, in Otte (ed.): 137-56.

SANCHEZ GONI, M. 1996. The Older Dryas of northern France in a West European context, Revue de Paleobiologie 15: 519-31.

SCHEER, A. 1993. The organization of lithic resource use during the Gravettian in Germany, in H. Knecht, A. Pike-Tay & R. White (ed.), Before Lascaux: 193-210. Boca Raton (FL): CRC Press.

STRAUS, L. & A. MARTINEZ. 1997. Site formation/disturbance processes, spatial distributions, site structure and activity areas, in Otte & Straus (ed.): 65-112.

STRAUS, L. & J. ORPHAL. 1997. Bols Laiterie Cave and the Magdalenian of Belgium, in Otte & Straus (ed.): 337-60.

STREET, M., M. BAALES & B. WENINGER. 1994. Absolute chronologie des spaten Palaolithikums und des Fruhmesolithikums im Nordlichen Rheinland. Archaologisches Korrespondenzblatt 24: 1-28.

STUTZ, A. 1997. Seasonality of Magdalenian cave occupations in the Mosan Basin, in Otte & Straus (ed.): 197-204.

TEHEUX, E. 1994. Le Magdalenien de la Vallee de la Lesse. Unpublished these de licence, Universite de Liege.

VALLADAS, H. 1994. Chronologie des sites du Magdalenien final du Bassin parisien, in Y. Taborin (ed.), Environnements et habitats Magdaleniens dans le Centre du Bassin Parisien: 65-8. Paris: CNRS. Documents d'Archeologie Francaise 43.

VAN DEN BRINK, F. 1971. Guia de Campo de los Momiferos Salvajes de Europa Occidental. Barcelona: Omega.

VAN NEER, W. 1997. Fish remains, in Otte & Straus (ed.): 205-14.

VERMEERSCH, P. & N. SYMENS. 1988. Le Magdalenien de piein air en Belgique, in Otte (ed.): 243-58.

WOBST, H. 1976. Locational relationships in Paleolithic society, Journal of Human Evolution 5: 49-58.