Time and Space in the Middle Paleolithic: Spatial Structure and Occupation Dynamics of Seven Open-Air Sites

Evolutionary Anthropology 25:153–163 (2016)

ARTICLE

AMY E. CLARK

The spatial structure of archeological sites can help reconstruct the settlement ces were exploited. For example, how dynamics of hunter-gatherers by providing information on the number and length are we to make inferences about what of occupations. This study seeks to access this information through a compari- behaviors occurred at a site and how son of seven sites. These sites are open-air and were all excavated over large strategies of resource procurement spatial areas, up to 2,000 m2, and are therefore ideal for spatial analysis, which were organized if we do not know was done using two complementary methods, lithic refitting and density zones. whether the analyzed assemblage Both methods were assessed statistically using confidence intervals. The statis- derives from a single long-term occu- tically significant results from each site were then compiled to evaluate trends pation or many short occupations? that occur across the seven sites. These results were used to assess the Reconstructions of human mobility “spatial consistency” of each assemblage and, through that, the number and based on raw material proveniences duration of occupations. This study demonstrates that spatial analysis can be a also run into the same problem when powerful tool in research on occupation dynamics and can help disentangle the we attempt to define territory or many occupations that often make up an archeological assemblage. home-range size for an archeological assemblage. This paper demonstrates that the Hunter-gatherer mobility is a popu- gatherers is such an intrinsic part of spatial patterning of artifacts can be lar subject among archeologists.1–6 Its their livelihood that it is difficult to used to assess the temporal character- popularity can perhaps be partially study any other part of their economy istics of an assemblage. The organiza- attributed to a romantic impression of and culture without understanding tion of artifacts and features relative a mobile lifestyle in contrast to the the underlying physical movement.7,8 to one another can indicate whether concretely (and, some may argue, Archeologists traditionally address an assemblage was the result of a sin- increasingly) sedentary life with which mobility by sourcing raw materials,3,4,9,10 gle or multiple occupations, as well as we are familiar. More importantly, assessing the degree of site provisioning the relative duration of occupations. however, the mobility of hunter- with raw materials,1,11–13 and deter- This is particularly the case when the mining the season(s) of occupation and sites are open-air and were excavated occupation intensity through faunal at very large scales, as was the case analysis.5,14,15 Each of these methods for the sites in this study. Because the gives a slightly different perspective on duration of occupation and the num- Amy E. Clark is a recent Ph.D. graduate of the mobility of groups occupying a site ber of occupations occur along a con- the University of Arizona. She is currently a postdoctoral researcher at INRAP, l’institut and can provide some insight into the tinuum, such a study should take national de recherches archeologiques number and duration of occupations. place in a comparative format. preventives , in Campagne, Dordogne, However, differentiating between many The sites included in this study France. Her main research interests include hunter-gatherer mobility and settle- short-term occupations and one long range in occupation intensity, meas- ment dynamics. Email: aeclark@email. occupation, or any combination of the ured in artifact density and assem- arizona.edu two variables, remains a major obstacle blage size, from sites located directly in the reconstruction of mobility and on raw-material sources to a low- land use patterns. density site on a floodplain. The two Key words: mobility; spatial analysis; Middle The number of occupations, the interrelated analyses used to deter- Paleolithic; France; lithic refitting size of the group, and the duration of mine the spatial structure of each site the time spent at a site is not only partition a site into density groups

VC 2016 Wiley Periodicals, Inc. salient to any discussion of hunter- and refitting groups based on where DOI: 10.1002/evan.21486 gatherer mobility,16 but also pertinent they are located relative to the general Published online in Wiley Online Library (wileyonlinelibrary.com). to our understanding of how resour- distribution of materials, as well as to 154 Clark ARTICLE

Figure 1. A map of France showing the location of all seven sites. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] pieces that belong to the same refit- give basic site details. Two of the sources near the city of Bergerac.19 ting set. These two analyses provide sites, La Doline de Cantalouette II Like Bossuet, the exploitation of flint information about how the sites were and Champs de Bossuet, are located at this site was concentrated within organized, as well as how long they directly on raw material sources and, a depression in the landscape, this were occupied, and the relative num- accordingly, boast assemblage sizes time within a doline. ber of occupations. In this paper, of more than 15,000 lithics. The other Fresnoy, Bettencourt, and Villiers “occupation” refers to any period of sites are located further from raw Adam are located in the loess belt of time spent at a site, whether that material sources, but all of them have northern France.20–23 All three sites time be minutes, hours, days, or high-quality flint sources within rela- are situated within paleosols on a months or, potentially, years. There- tively close proximity. None of these northeastern facing slope. Fresnoy fore, “occupation” should not carry sites contain preserved bone. La Folie, and Bettencourt contain dense any connotations of “home” in the a small site located on a floodplain, is patches of artifacts, but also areas of sense that a site was used as a living the only site to contain a preserved low and medium density. The depos- place; instead, “occupation” implies fire feature. its at Bettencourt are divided into only that a space was occupied by the The general spatial structure of three sectors by channels of erosion. physical presence of at least one these sites presents a range of varia- Villiers Adam is a very large site that human at one point in time. tion (Fig. 2). Cantalouette and Bos- contains many small clusters, but suet, the quarry sites, have uniformly has a low density over all. THE DATA SET dense accumulations. At Bossuet, Le Prisse de Bayonne and La Folie lithic nodules were exploited within have the smallest assemblage sizes in The sites used in this study were a partially filled paleo-channel on a the data set; both contain two princi- excavated as rescue or contract proj- terrace of the Isle River, northeast of ple clusters of high-density accumu- ects by l’institut de recherches arch- Bordeaux.17,18 Because the lithic lations. La Folie, outside of Poitiers, eologiques preventives (INRAP). All nodules were exploited where they is arguably the best preserved site in sites were excavated over large spatial were found, the site follows the con- this sample, with evidence of a extents and therefore are well suited tours of the paleo-channel. La Doline hearth, a nonpedogenic organic hori- for a study of spatial organization. de Cantalouette II is one of a series zon (likely bedding), and postholes Figure 1 provides a map of site loca- of several sites set atop the high- that may have been used for a wind- tions within France; Tables 1 and 2 quality and prolific Bergeracois flint break (they enclose an area thought ARTICLE Time and Space in the Middle Paleolithic 155

TABLE 1. Basic Information on the Seven Sites Landscape Position General Spatial Structure Le Prisse Plateau near the confluence of two major rivers Two closely spaced clusters, large area of low density scatter Bettencourt Northeast facing gentle slope, forest soil Three sectors separated by erosion, sev- within loess deposits eral very high density clusters Bossuet Paleo channel on river terrace Extremely high density distribution, follows contours of the drainage Cantalouette Doline (sinkhole) Extremely high density distribution with some more concentrated clusters La Folie Floodplain Small site with two clusters connected by many refits. Evidence for windbreak and bedding (excellent preservation) Fresnoy Northeast facing moderate slope, forest soil Large scatter over large area, refits con- within loess deposits nect the entire area Villiers Adam Northeast facing gentle slope, forest soil Many small, distinct clusters over a large within loess deposits area

to be too large for a covered struc- medium-, and low-density areas. The are placed into refitting group 1, ture).24–26 Le Prisse is located on a artifacts are then organized into the lithics within 2 meters of that group plateau outside of the city of three density groups based on where into group 2, and lithics more than Bayonne.27 While this site has two they are located. This analysis 2 meters from group 1 into group 3 distinct concentrations of material, excluded the two quarry sites, Canta- (Fig. 4). Group 1 is determined by like La Folie, the majority of the louette and Bossuet, because they do locating the smallest area where site’s area is made up of very low- not have technological data associ- the highest concentration of lithics density scatter. ated with each coordinated artifact. occurs. This area most likely corre- The data sets were simply too large; sponds to the location where they METHODS thus, the artifacts were segregated into were knapped. technological categories and counted, Thus, these analyses produce two The spatial analysis of these seven rather than recording each individual sets of groups: the high-, medium-, sites used two related approaches. artifact into the database. and low-density groups determined Both analyses were done using Arc The second analysis, the refitting by the density contour analysis and VR GIS 10.0 and combine the spatial location analysis, makes use of the refitting groups 1, 2, and 3 deter- location of lithics and their techno- extensive refitting data that was col- mined by the refitting location analy- typological attributes. The first analy- lected for these sites. The percent of sis. Once the artifacts are divided into sis, the density contour analysis, the assemblage that was refitted is density and refitting groups, the pro- begins by constructing a density map outlined in Table 2. It ranges from portion of technological categories of all coordinated artifacts (lithics 38% at La Folie to just above 5% at was calculated for each group. This greater than 2 cm).28 A density map is Bossuet. The small assemblage size value was then subtracted from the a raster image with color shading val- and low refit percentage (4%) at Vil- breakdown of technological groups ues depending on the relative density liers Adam make it unsuitable for this for the assemblage as a whole to of artifacts. The next step is to create analysis. Like the density contour assess whether categories were over- contours for the density map in the analysis, the refitted lithics are divided or under-represented in each group. same way that might be done for an into three groups.28 Each set of refit- For example, there were 60 total cores elevation map (Fig. 3). The generated ted pieces is analyzed separately: all at Bayonne, 22 of them located in the contours are used to isolate high-, lithics within one meter of each other low-density zone, which gives a

TABLE 2. Quantitative data for each site in the studya

Le Prisse Bettencourt Bossuet Cantalouette La Folie Fresnoy Villiers Adam Site density (lithics/m2) 0.81 25.13 69.29 54.62 6.1 3.74 0.84 Site area (m2) 1075 915 228 282 207 1143 1928 Total lithics 870 5729 15,797 15,404 1,262 4,270 1,619 Densest square (# lithics) 66 225 527 443 189 183 108 Refit percentage 20% 14% 5% 10% 38% 9% 4% aThe site size quoted here was calculated in ArcGIS and reflects the minimum area occupied by the cloud of coordinated artifacts. 156 Clark ARTICLE

Figure 2. All seven sites included in this analysis displayed at the same scale. This image also provides a sense of the general site structure, including the refitting connections (colored lines). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

proportion of 0.37. In order to evalu- CENTRIFUGAL MOVEMENT disentangle the temporal factors that ate whether this number was statisti- AND THE FORMATION controlled site structure formation. cally significant, 95% confidence OF SITE STRUCTURE The density contour analysis maps intervals were established around the the location of artifacts within con- total lithics that fell into the high-, The two methods presented are centric rings of artifact density. The medium-, and low-density zones. The essentially mapping the formation of areas highest in density are most likely proportion of cores that were located site structure through centrifugal where the majority of core reduction within the low-density zones fell out- movement of lithic artifacts. The occurred; the medium- and low- side of this interval and was therefore underlying concept here is that the density areas roughly coincide with deemed significant. This method was spatial structure of lithic artifacts areas where lithics were moved at used for both the density and refitting within archeological sites is primarily some point after they were knapped. groups. formed by the introduction of lithic Knapping events, of course, may have The density contour analysis and raw materials, the knapping of these taken place in the medium- and low- refitting location analysis are simply materials at one or several discrete density areas and some lithics may ways of dividing coordinated lithics spatial locations, and the subsequent have been moved from the low- and into groupings based on where they centrifugal movement of these arti- medium-density areas into the high- were mapped during excavation. The facts to other locations of the site density areas. However, these are density contour analysis divides all through human or geologic agents. exceptions to this idealized model. lithics into zones based on whether By tracking the relative number of This model will be tested by looking at the artifact was located within a low- times this process occurred (that is, whether the high-density areas largely , medium-, or high-density part of the number of times raw materials consist of core reduction debris. the site. The refitting location analy- were introduced to a site and then There will also be a certain num- sis, on the other hand, divides refit- reduced), the extent to which the arti- ber of lithics that were not knapped ted lithics based on where they are facts were then redistributed, and on site but were transported to the located relative to other lithics with- how these events are configured rela- site from some other location. These in a given refitting set. tive to one another, we can begin to lithics, however, are not numerous ARTICLE Time and Space in the Middle Paleolithic 157

Figure 3. An example of how spatial groups are divided in the density contour analysis. This figure displays the density map from Betten- court overlain by contours chosen to delimit the low-, medium-, and high-density zones. Artifacts found within each of these zones are then assigned to the corresponding group. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] enough to contribute significantly to there they will not be overwhelmed by on the extent of their spatial spread the overall site structure. They can be reduction debris. from where they were knapped. This expected to occur in higher propor- The refitting location analysis also analysis, however, differs from the tions in low-density areas because tracks the location of artifacts based density contour analysis in that here there is a concrete, rather than im- plied, indication of artifact movement. This is because at one point in time each artifact in a refitting group was located at the same point in space (that is, within the core, before knapping). The location of refitting group 1 may not always correspond with the location of knapping for every refitting set, but we can expect the two to co-occur in the majority of cases. Both analyses track the centrifugal movement of artifacts within the site and, therefore, the formation of the site structure as a whole. The density contour analysis documents movement by locating artifacts within high-, medium-, and low-density zones of the site. The refitting location analysis, however, documents the movement of individual reduction events. Once these Figure 4. An example of how refitted lithics are divided into groups. This figure shows the analyses are performed at each site, we spatial location of artifacts from refitting set 12 from Bettencourt. Three lithics tightly clus- tered together fall into group 1, two more slightly farther away are in group 2, and a final can begin to deconstruct the site struc- lithic some meters from the others falls into group 3. The grid is in square meters. [Color ture formation and, through that, the figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] temporal dynamics of occupations. 158 Clark ARTICLE

TABLE 3. Significant Values for Both Analyses Indicating Whether an Artifact their conspicuous nature, which ren- Category Was Over- or Underrepresenteda ders them more mobile than other technological types. They are large Refitting High Refitting Medium Refitting Low objects that receive a great deal of att- Group 1 Density Group 2 Density Group 3 Density ention from the knapper and, likely, other group members as well (such as Cortical Flake "" " "" flfl fl fl children). They could have been moved Partially Cortical Flake "" flflfl" for future knapping or for other uses. Naturally Backed Flake fl Maintenance Flake "" "flflflfl" In addition, their rounded geometry Debordant Flake fl" flfl and generally larger sizes makes them Debris "" """ fl fl fl flflflfl more susceptible to movement by site Non-Cortical Flake fl flflfl """ " " reuse or other surface disturbances. Levallois Flake flflfl" fl " Flakes, blades, and Levallois flakes Blade flfl " fl" were also overrepresented in refitting Nodule " fl " group3andthelow-densitygroup, Core """" flfl flfl fl but underrepresented in refitting group Tool #"" flfl# fl " """ 1 and the high-density group. Bifaces Biface/Cleaver flfl" and cleavers were present only at the site of Le Prisse, but they also followed aSignificant values for both analyses indicating whether an artifact category was over or underrepresented. A upwards facing arrow indicates overrepresentation this pattern in the density contour while a bolded downwards facing arrow indicates underrepresentation. The num- analysis. In addition, retouched tools ber of arrows corresponds to the number of significant results per site (for example, were strongly overrepresented in the partially cortical flakes are overrepresented in refitting group 1 at two of the sites in low-density group and underrepre- the analysis). sented in the high density group, although the results of the refitting RESULTS: REFITTING LOCATION group 3 and low-density areas were analysis (specifically, group 1) were ANALYSIS AND DENSITY moved a longer distance from where mixed, showing more variability. Corti- cal flakes, partially cortical flakes, and CONTOUR ANALYSIS they were knapped. Much of this “movement” is, of course, assumed maintenance flakes displayed a pattern The number of statistically signifi- rather than proved, but this is the gen- that was generally the inverse of these cant results for the two analyses is eral model that most artifacts should “planned” pieces, but their results were shown in Table 3. This table breaks the follow. Whether these assumptions mixed. significant results down by technolo- are correct will be tested, to an extent, gical category and whether they were by which artifacts are over and under- SPATIAL CONSISTENCY over or underrepresented. Both analy- represented in each group. AND ITS USE IN DETERMINING ses are placed together in a single table The categories of debris and cores THE NUMBER AND DURATION to facilitate comparison. In theory, yield the strongest patterns. Debris are results from the high-density group overrepresented in the first group of OF OCCUPATIONS should be similar to results from refit- both analyses and underrepresented The results produced by the two ting group 1, as should results from the in the third. The debris category is analyses can be used to evaluate the medium-density group and refitting made up of small flakes (<3 cm), shat- spatial consistency of each site in the group 2 and results from the low- ter, and split pieces; therefore, an over- study. Spatial consistency simply refers density group and refitting group 3. representation of these pieces in to the clarity of spatial patterning at The alignment in results between these refitting group 1 and the high-density thesiteinquestion:Canpatternsbe two analyses may vary by site because group support the assumption that identified and do these patterns agree differing site use may yield different these areas can be linked with knap- with each other? The spatial consis- patterns, but the results from equiva- ping events. Although there are some tency of a site can first be evaluated by lent groups should be more or less sim- examples of knapping events occur- the number of statistically significant ilar within any particular site. ring in medium-density zones, deter- results from the spatial analyses. A The high-density area and refitting mined by the location of refitting sets, high number of significant values indi- group 1 are both assumed to be loca- most artifacts located in the medium- cates that the spatial patterns resulting tions where the majority of core reduc- and low- density areas were moved to from site use were clear and not jum- tion occurred, so the artifacts there their location after being knapped in bled through site reuse or prolonged should reflect this behavior. Artifacts the high-density area or off site. occupation. The clarity of spatial pat- found in refitting group 2 and the In contrast to debris, cores are over- terning can also be evaluated by medium-density group would have represented in the low-density group whether the two analyses agreed with moved, by a variety of processes, a but underrepresented in the high- each other for each of the three levels short distance from where they were density group and refitting group 1. of spatial groupings. For example, knapped. Artifacts found in refitting The location of cores likely reflects were cortical flakes overrepresented in ARTICLE Time and Space in the Middle Paleolithic 159

TABLE 4. The Total Number of Significant Results (Counting Both Analyses) for the This scatter could have been partially Seven Sitesa the result of a “background scatter” of material, suggesting a continued human Total results Total Significant presence on the landscape.16,29,30 The subjected to significant results/ density contour analysis captures the significance test results total possible Disagreements signal from the background scatter Le Prisse 75 14 .19 0 from the low-density group, which can Bettencourt 72 23 .32 4 be expected to disproportionally repre- Bossuet 36 2 .06 0 sent lithics left during short-term visits Cantalouette 36 1 .03 0 to the site. Indeed, Le Prissehasapar- La Folie 72 7 .10 0 ticularly high number of statistically sig- Fresnoy 72 20 .28 2 nificant results showing that retouched Villiers Adam 36 18 .50 0 tools, as well as bifaces and cleavers, are overrepresented in the low-density area. aThe total possible results are different for each site because both analyses were These lithics likely represent discarded not performed on all sites and not all sites had the same artifact categories (that is, elements of a mobile toolkit.31 Le Prisse is the only site to include the category “biface/cleaver”). For this reason, In contrast to Le Prisse, the site the third column contains a ratio of the significant results and the total possible area of La Folie was very small; it was results. The final column displays the number of times the results of the two analyses located on a floodplain and therefore contradict each other. was covered too quickly to accumulate a background scatter. La Folie the high-density group but underrepre- time and not reoccupied. The inverse, also had a much more concentrated sented in refitting group 1 (considering however, could be related to the jum- density than did Le Prisse. Both sites only statistically significant patterns)? bling and disorder of artifacts that have low overall artifact counts, but If so, this would count as a disagree- might occur when sites were occupied La Folie contained 189 lithics in its ment. A disagreement between the two on numerous occasions and for longer densest square, a number nearly three analyses might indicate that there is times. These results can be compli- times as high as that at Le Prisse. The discordance in spatial patterning at the mented by other site characteristics concentrations of lithics at La Folie site. found in Table 2. were covered more quickly that than Table 4 presents the number of stat- Bossuet and Cantalouette, the two those at Le Prisse, and thus not sub- istically significant results, as well as sites located directly on raw material jected to as many processes that may the number of disagreements between sources, have very few significant have caused their diffusion. However, analyses for each site in this study. results. These sites likely experienced La Folie had a higher overall artifact This table can be used to evaluate spa- many short occupations that left the count and thus was also likely occu- tial consistency. The first column sites with fragmented and inconsis- pied for a longer time Le Prisse. Dis- presents the number of statistically tent spatial structures. These two sites counting the background scatter significant results found at each site; were located on top of high-quality found at Le Prisse, the two sites likely the second presents the number of raw material sources, so not only had only one occupation. Both sites categories that were submitted to the were they nodes on the landscape that had a medium number of statistically statistical test; and the third column attracted multiple occupations, but significant results, a number that displays a ratio of these two values. A lithic debris accumulated at a much was difficult to achieve, given their ratio is important because not all sites faster rate than at other sites. This is smaller assemblage sizes. In addition, were subjected to both analyses. Fur- shown by their extremely high artifact both sites have high refitting percen- thermore, Le Prisse is the only site to counts, which were spread over a tages, which indicate well-ordered contain the category “biface/cleaver.” fairly small area, making their den- assemblages that had not been reor- The final column of Table 4 displays sities very high and evenly distributed ganized through long-term or multi- the number of times these two analy- (Table 2). ple occupations. ses disagree with each other at any La Folie and Le Prisse, the sites Villiers Adam has the highest ratio particular site. This occurred, for with the smallest assemblage sizes, of statistically significant results (50% example, at Bettencourt, where corti- have a medium number of significant of the total results subjected to the cal flakes were found to be signifi- results. Le Prisse had a very low site- significance test). It is also distinct cantly overrepresented in refitting wide artifact density. Artifacts were from the other sites because of its group 1 but significantly underrepre- mainly concentrated in two clusters very large area and low artifact den- sented in the high-density group. in the northeast portion of the site. sity. Although it is larger than all of Spatial consistency can give insights These clusters were not particularly the other sites in the study by nearly into the number and duration of occu- dense, however; the densest square 800 m2, it contained only 1,619 lithics. pations. Agreement between the two contained only 66 lithics. The re- The high number of significant re- analyses and a high number of statisti- maining portion of the site, which sults indicates clear spatial pattern- cally significant results may indicate amounted to quite a large area, was ing, but this was likely the result of that a site was occupied for a short made up of a very low-density scatter. multiple short-term occupations that 160 Clark ARTICLE

TABLE 5. The disagreements in statistically significant results between spatial analyses

Bettencourt Cortical Flake Density Contour Analysis High Density Under Refitting Location Analysis Refitting Group 1 Over Bettencourt Maintenance Flake Density Contour Analysis High Density Under Refitting Location Analysis Refitting Group 1 Over Fresnoy Partially Cortical Flake Density Contour Analysis High Density Under Refitting Location Analysis Refitting Group 1 Over Bettencourt Cortical Flake Density Contour Analysis Low Density Over Refitting Location Analysis Refitting Group 3 Under Bettencourt Maintenance Flake Density Contour Analysis Low Density Over Refitting Location Analysis Refitting Group 3 Under Fresnoy Partially Cortical Flake Density Contour Analysis Low Density Over Refitting Location Analysis Refitting Group 3 Under

were spatially distinct. Subsequent the entire site. The refitting location beginning of the occupation were occupations did not occur on top of analysis, however, tracks only the more likely than others to be moved abandoned debris but, instead, were location of lithics that have been refit- away from their knapping location to spatially segregated so that the site is ted. The refitted lithics are likely other parts of the site. This movement made up of a number of distinct clus- biased toward reduction events that was likely the result of both inten- ters, all of which display spatial con- occurred later in time during the site’s tional action through selection of sistency. This means that sample sizes occupation; earlier reduction events lithics for use and, perhaps more im- were able to become large enough to are more likely to have holes in their portantly, unintentional movement obtain significance, but patterns were sequence from lithics discarded off- through site reuse. This differential not jumbled during reoccupation. It is site. Therefore, the results of the den- representation of time explains the likely that each of these occupations sity contour analysis display patterns discordance in the results. It also indi- was relatively short; the densest sq- that document the entire occupation cates that Fresnoy and Bettencourt uare at Villiers Adam contains 108 span, while the refitting location anal- had much longer occupations than lithic pieces, about halfway between ysis is skewed toward representation did the other sites in the study. This the numbers at Le Prisse and La Folie. of events occurring later. conclusion is supported by other lines One might question why these clus- At Bettencourt and Fresnoy, lithic of evidence, as shown in Table 2. ters were produced through reoccu- classes were systematically moved Compared to the other sites, Betten- pation when they could have been around the site, but the patterns found court has a very high artifact density, produced by the occupation of one for the overall distribution of artifacts the highest without being situated large group. Large group size could (tracked by the density contour analy- atop a raw material source. In addi- be one explanation for the observed sis) did not always match the pattern tion, its density is quite concentrated, spatial patterning at the site. How- found for the events occurring later in with its densest square containing ever, with only 4% of the assemblage the site’s occupation (tracked by the 225 lithics. At Fresnoy, the artifact part of a refit, the site does not reflect refitting location analysis). Examining density was lower over a larger area. interconnectivity. Furthermore, most the data presented in Table 5, the The densest square contained only of these refits occur at small scales results of the refitting location analysis 183 artifacts, a value just slightly within rather than between clusters. always follow the dominant results lower than the densest square found This does not exclude the possibility from other sites in the study; that is, at La Folie. Fresnoy is situated on a of a large group, but it merely makes although cortical flakes are overrepre- slightly steeper slope than other sites, spatially distinct reoccupations a sented in the first group and underre- however, which might have contrib- more parsimonious explanation. presented in the third, the density uted to a less concentrated distribu- Bettencourt and Fresnoy also have contour analysis found the opposite tion of material. However, its total high numbers of statistically signifi- results. artifact count is still high, with 4,270 cant results, but they are also the only The refitting location analysis pieces. Bettencourt was therefore two sites that exhibit disagreements reveals patterns consistent with other occupied for a longer period than was among these results. This indicates sites in the study because they have Fresnoy. It has a higher artifact that these sites have some consistency been less obscured by long occupa- count, a higher ratio of statistically in their spatial organization, but tion. The density contour analysis in- significant results, and also more dis- discordance in the spatial patterning. dicates that Fresnoy and Bettencourt agreements between the two analyses. In order to understand what this were occupied long enough for a sig- It can be concluded from these discordance means, we must con- nificant amount of centrifugal disper- analyses that Cantalouette and Bos- sider what each analysis is tracking. sion to occur. In other words, as the suet had many reoccupations, all The density contour analysis tracks site was occupied for a longer period located on top of one another, in the distribution of all lithics across of time, the lithics knapped at the areas where abundant raw materials ARTICLE Time and Space in the Middle Paleolithic 161 could be found. This resulted in in- trifugal dispersion of reduction events CONCLUSIONS consistent spatial patterning and few can be a powerful tool for conceptual- The spatial analysis presented here statistically significant results from izing the effect of time on the spatial shows that a repeated pattern of spatial the two spatial analyses. Villiers Adam structure of archeological sites. organization is found among the Nean- also had numerous reoccupations but, One confounding factor that makes derthal sites in this dataset. The factors because raw materials were not being this exercise so difficult is the differ- that control this spatial organization exploited in place (though likely ent degrees of access to raw materials, are repeated events of core reduction imported from close by), the succes- which can distort the relationship of followed by the centrifugal dispersion sive occupations were separated in artifact quantity to time. For example, of lithics from this initial location. The space. Thus, spatial patterning was a site located close to a source of raw centrifugal dispersion of lithics is not obscured through reuse and the materials could accumulate a large caused by several agents. First, humans site exhibited a strong spatial consis- quantity of lithic materials in a rela- select lithics to be used elsewhere on tency. Le Prisse has the smallest lithic tively short time. At sites located far- the site, usually “preferential” pieces assemblage in the dataset and a ratio ther away from raw materials, the such as noncortical flakes, blades, or of statistically significant results that cost of transporting raw materials retouched pieces. Second, centrifugal falls in the middle. This site therefore would result in a much slower rate of dispersion can result from nonhuman likely had a relatively short principal accumulation of lithic materials. In addition, much of the initial stages of agents, usually when a site is uncov- occupation, but with a background ered on the landscape. Such agents scatter suggesting repeated human core reduction may have taken place closer to the raw material source, include fluvial or colluvial processes, presence on the land surface, which bioturbation, or use of the land surface likely was exposed for a longer time reducing even further the amount of lithic debris discarded at the site by nonhumans. Finally, prolonged than other sites in the dataset. La Folie site use causes centrifugal dispersion has a lower ratio of statistically signifi- under study. However, the sites dis- cussed here are all relatively close to through accidental movement by cant results than Le Prisseanda humans or through cleaning and other higher assemblage size, with a much high-quality raw materials. The only major difference is whether or not behaviors. However, structuring of denser concentration of lithics. There- space through systematic cleaning or fore, La Folie was likely occupied for a they were physically located on top of the source. partitioning of activities was not found longer time. Finally, Fresnoy and Bet- to be a major component of the spatial tencourt had higher ratios of statisti- A major disadvantage of this dataset is the lack of evidence from faunal organization of lithics for sites in this cally significant results but, as noted, remains. Faunal analysis can help dataset. It should be pointed out, how- there were disagreements between the determine whether an occupation ever, that La Folie does have evidence two analyses. This indicates that the was repeated seasonally or across of a structure, as well as bedding. sites were occupied long enough that many seasons.14,15,32 However, many Therefore, a certain cultural organi- the centrifugal dispersion of knapping sites do not have preservation of fau- zation of space could be evidenced events created a pattern that was at nal remains. Moreover, many that do through other materials, but the spatial odds with more recent reduction are caves or rockshelters, which yield organization of lithics at La Folie fol- events, revealed through the refitting assemblages that are arguably even lowed the same rules that applied at location analysis. These more recent more difficult to disentangle (but see other sites in the dataset. events followed patterning exhibited work at Abric Romanı33–36). The rules of centrifugal dispersion at other sites in the dataset. Given its The duration and number of occu- and site structure formation are clo- high density and largest number of pations directly influence the size and sely tied to the impact of time on an disagreements between analyses, Bet- density of an archeological assem- archeological assemblage. Therefore, tencourt was likely occupied for the blage. However, an equally important the spatial structure of lithics can be longest time. variable in this puzzle is group size. used to disentangle the elements of This factor was mentioned earlier in time (that is, the number and duration THE DIFFICULTIES OF reference to Villiers Adam. The spa- of occupations) on a lithic assemblage. DIFFERENTIATING BETWEEN tially segregated concentrations of This is vital if we are to understand OCCUPATION NUMBER debris at Villiers Adam could poten- the mobility and landscape use of hunter-gatherers in the past This is AND DURATION tially be the result of a larger group rather than the product of reoccupa- especially true for the study of popu- The methods presented here do not tions. However, small group size is lations of hunter-gatherers who are completely solve the problem of dif- the more parsimonious explanation, removedfromusbytensofthousands ferentiating between many short-term particularly because it is supported of years of geologic processes and mo- occupations and one or several long- by the majority of other studies.37–39 dification of the landscape by humans term occupations. They do, however, It is a question that deserves further and other organisms. provide some additional lines of attention, however, and will most This study used the spatial organiza- inquiry that can be used along with likely be accessed by comparing sites tion of lithics to address two of the fac- other indicators. In particular, break- from a population that we know dis- tors that control the size and density ing down site formation through cen- play a range of group sizes. of an archeological assemblage. A 162 Clark ARTICLE third factor, access to raw materials, Jean-Luc Locht and David Colonge. 15 Rendu W. 2007. Planification des activites Steven Kuhn, my dissertation advisor, de subsistance au sien du territoire des derniers occurred primarily on two scales: raw Mousteriens. Cementochronologie et approache materials were either located within was instrumental throughout this pro- archeozoologique de gisements du Paleolithique sites or within easy access. The final ject. Funding was provided by a Cha- moyen (Pech-de-l’Aze I, La Quina, Mauren) et Paleolithique superieur ancien (Isturitz). The`se major factor, group size, was found to teaubriand Fellowship. I would also de doctorat, l’Universite Bordeaux 1, Bordeaux, be more or less consistent among the like to thank the anonymous reviewers France. sites in this dataset, though this factor who provided helpful comments on an 16 Conard NJ. 2001. Advances and problems in the study of Paleolithic settlement systems. In: must be addressed more rigorously in earlier version of this manuscript. Conard NJ, editor. Settlement dynamics of the the future. However, if we are to Middle Paleolithic and Middle Stone Age. accept the results of this study at face REFERENCES Tubingen:€ Kerns Verlag. p VII-XX. value, the major access of variation in 17 Bourguignon L, Ortega I, Lenoble A, et al. 1 Blades BS. 1999. Aurignacian lithic economy 2000. Le gisement mousterien de Champs de terms of the size and density of Middle and early modern human mobility: new per- Bossuet. Bordeaux: AFAN, ASF et SRA Paleolithic archeological assemblages spectives from classic sites in the Veze `re valley Aquitaine. is the number and duration of occu- of France. J Hum Evol 37:91–120. 18 Lenoble A, Ortega I, Bourguignon L. 2000. 2 Delagnes A, Rendu W. 2011. Shifts in Nean- Processus de formation du site mousterien de pations. This indicates that certain dertal mobility, technology and subsistence Champs-de-Bossuet (Gironde). Paleo 12:413– locations were used repeatedly for strategies in western France. J Archaeol Sci 38: 425. different lengths of time. Numerous 1771–1783. 19 Bourguignon L, Blaser F, Rios J, et al. 2008. 3 Feblot-Augustins J. 1993. Mobility strategies L’occupation mousterienne de la Doline de Can- studies have found that Neanderthal talouette II (Creysse, Dordogne): specificit es tech- 40–42 in the Late Middle Palaeolithic of Central population densities were low ; Europe and Western Europe: elements of sta- nologiques et economiques, premiers resultats recent genetic work supports these bility and variability. J Anthropol Archaeol 12: d’une analyse integr ee. In: Jaubert J, Bordes J-G, 211–265. Ortega I, editors. Les societ es du Paleolithique findings. Genetic sequencing of a dans un Grand Sud-Ouest de la France: nouveaux 4 Fernandez P, Raynal J-P, Moncel M-H. 2008. gisements, nouveaux resultats, nouvelles meth- female Neanderthal from Siberia Middle Paleolithic raw material gathering terri- ods. Bordeaux: Memoire XLVII de la Societ e tories and human mobility in the southern shows that her parents were half- prehistorique franc¸aise. p 133–150. Massif Central, France: first results from a siblings and that many of her recent petro-archaeological study on flint. J Archaeol 20 Goval E, Locht J-L. 2009. Remontages, sys- ancestors were the product of inbreed- Sci 35:2357–2370. te`mes techniques et reparitions spatiales dans 43 l’analyse du site weichselien ancien de Fresnoy- ing. This suggests a long tradition of 5 Niven L, Steele TE, Rendu W, et al. 2012. au-Val (Somme, France). Bull Soc prehist Neandertal mobility and large-game hunting: small group sizes. If population sizes franc¸aise 106:653–678. the exploitation of reindeer during the Quina were low, we cannot assume that the Mousterian at Chez-Pinaud Jonzac (Charente- 21 Locht J-L. 2001. Modalites d’implantation et repeated use of specific locations was Maritime, France). J Hum Evol 63:624–635. fonctionnement interne des sites. L’apport de trois gisements de plein air de la phase recente simplyarandomresultofhavingmany 6 Rios-Garaizar J, Garcia Moreno A. 2015. Mid- du paleolithique moyen dans le nord de la hominid groups on the landscape. dle Paleolithic mobility patterns and settlement France (Bettencourt-Saint-Ouen, Villiers-Adam system variability in the eastern Cantabrian et Beauvais). In: Conard NJ, editor. Settlement Rather,itisreasonabletoassumethat region (Iberian Peninsula): a GIS-based resource dynamics of the Middle Paleolithic and Middle many of these locations were used by patching model. In: Conard NJ, Delagnes A, edi- Stone Age. Tubingen:€ Kerns Verlag. p 361–394. tors. Settl. Dyn. Middle Paleolithic Middle Stone thesamegroupasapartoftheirsea- Age. Tubingen:€ Kerns Verlag. p 329–360. 22 Locht J-L, Antoine P, Bahain J, et al. 2003. sonal rounds. This suggests that Nean- Le gisement paleolithique moyen et les 7 Binford LR. 1980. Willow smoke and dogs’ sequencess pleistoce `nes de Villiers-Adam (Val-d derthals had a complex system of land tails: hunter-gatherer settlement systems and ’Oise): chronostratigraphie, environnement et archaeological site formation. Am Antiq 45(1): implantations humaines. Gall prehist 45:1–111. use in which certain prime locations 4–20. were used repeatedly and others were 23 Locht J-L, Goval E, Antoine P. 2010. Recon- 8 Kelly RL. 1995. The foraging spectrum: diver- structing Middle Palaeolithic hominid behav- used for more prolonged occupation. sity in hunter-gatherer lifeways. Washington, iour during OIS 5 in northern France. In: This conclusion is not a major reve- DC: Smithsonian Institution Press. Conard NJ, Delanges A, editors. Settlement lation. Archeologists have known for 9 Geneste J-M. 1988. Syste`mes d’approvisionne- Dynamics of the Middle Paleolithic and Middle ment en matie`res premie`res au Paleolithique Stone Age. Vol. III. Tubingen:€ Kerns Verlag. some time that Neanderthals likely moyen et au Paleolithique superieur en Aqui- p 329–356. had a complex system of mobility and taine. In: Kozolowski JK, editor. L’Homme de 24 Bourguignon L. 2010. Le campement Neanderthal. 8, La Mutation. Lie`ge: ERAUL 35. resource exploitation. However, using mousterien de la Folie (Vienne). In: Buisson- p 61–70. Catil J, Primault J, editors. Prehistoire entre the methods presented here, we can 10 Tomasso A, Porraz G. 2016. Hunter-gather- Vienne et Charente: Hommes et societ es du identify features of the landscape that ers’ mobility and embedded raw material pro- Paleolithique. Memoire XXXVIII. Bordeaux: curement in the Mediterranean Upper Ministe`re de la Culture et de la Communica- led to prolonged occupation versus Paleolithic. Evol Anthropol this issue. tion. p 169–174. repeated ones, then combine this 11 Clark AE. 2015. Changes in land use and 25 Bourguignon L, Sellami F, Deloze V, et al. with information about territory and occupation intensity at the onset of the Middle 2002. L’habitat mousterien de “La Folie” (Poit- home range sizes to achieve a more Paleolithic? A view from Tabun Cave, Israel. In: iers, Vienne): synthe`se des premiers resultats. Conard N, Delagnes A, editors. Settlement Paleo 14:29–48. nuanced view of the specifics of these dynamics of the Middle Paleolithic and Middle 26 Bourguignon L, Vieillevigne E, Guibert P, systems of land use. Stone Age. Tubingen:€ Kerns Verlag. p 127–144. et al. 2006. Complements d’informations chro- 12 Kuhn SL. 2004. Upper Paleolithic raw mate- nologiques sur le campment mousterien de tra- rial economies at Uc€ ¸agızlı cave, Turkey. dition acheuleenne du gisement de la Folie ACKNOWLEDGMENTS J Anthropol Archaeol 23:431–448. (Poitiers, Vienne). Paleo 18:37–44. 13 Kuhn SL, Clark AE. 2015. Artifact densities 27 Colonge D, Claud E, Deschamps M, et al. I am grateful to Laurence Bour- and assemblage formation: Evidence from 2015. Preliminary results from new Palaeolithic guignon, who not only gave me much Tabun Cave. J Anthropol Archaeol 38:8–16. open-air sites near Bayonne (south-western of the data used in this study, but also 14 Costamagno S, Liliane M, Cedric B, et al. France). Quaternary Int 364:109–125. 2006. Les Pradelles (Marillac-le-Franc, France): 28 Clark AE. 2015. Spatial structure and the provided me with valuable mentorship. a mousterian reindeer hunting camp? temporality of assemblage formation: a com- Data was also generously provided by J Anthropol Archaeol 25:466–484. parative study of seven open air Middle ARTICLE Time and Space in the Middle Paleolithic 163

Paleolithic sites in France. Ph.D. thesis, Univer- 34 @@Vaquero M. 2008. The history of stones: 40 Mellars P, French JC. 2011. Tenfold popula- sity of Arizona, Tucson, Arizona. behavioural inferences and temporal resolution tion increase in Western Europe at the 29 Roebroeks W, De Loecker D, Hennekens P, of an archaeological assemblage from the Mid- Neandertal-to-modern human transition. Sci- et al. 1996. A veil of stones: on the interpreta- dle Palaeolithic. J Archaeol Sci 35:3178–3185. ence 333:623–627. tion of an early Middle Paleolithic low density 35 Vaquero M, Allue E, Vallverdu-Poch, J. 41 Stiner MC, Kuhn SL. 2006. Changes in the scatter at Maastricht-Belvede `re (The Nether- 2012. Conclusions: Landscapes, Campsites, lands). Analecta Praehist Leidensia 25:1–16. Time, and Neanderthal Behavior. In: Carbonell ‘connectedness’ and resilience of Paleolithic i Roura E, editor. High Resolution Archaeology societies in Mediterranean ecosystems. Hum 30 De Loecker D. 2004. Beyond the site: the Ecol 34:693–712. Saalian archaeological record at Maastricht- and Neanderthal Behavior: Time and Space in Belvedere, The Neatherlands. Analecta Praehist Level J of Abric Romanı (Capellades, Spain). 42 Mirazon Lahr M, Foley R. 2003. Demogra- Leidensia 35/36: 300. New York: Springer. p. 389–405. phy, dispersal and human evolution in the Last 31 Kuhn SL. 1994. A formal approach to the 36 Vaquero M, Chacon MG, Garcıa-Anton MD, Glacial period. In: van Andel TH, Davies W, design and assembly of mobile toolkits. Am et al. 2012. Time and space in the formation of editors. Neanderthals and modern humans in Antiq 59:426–442. lithic assemblages: The example of Abric the European landscape during the Last Glacia- Romanı Level J. Quat Int 247:162–181. 32 Adler DS, Prindiville TJ, Conard NJ. 2003. tion. Cambridge: McDonald Institute for Patterns of spatial organization and land use 37 Mellars PA. 1996. The Neanderthal legacy: an Archaeological Research, University of Cam- during the Eemian interglacial in the Rhine- archaeological perspective from Western bridge. p 241–256. land: new data from Wallertheim, Germany. Europe. Princeton: Princeton University Press. 43 Prufer€ K, Racimo F, Patterson N, et al. Eurasian Prehist 1:25–78. 38 Gamble C. 1999. The Palaeolithic societies of Europe. Cambridge: Cambridge University Press. 2014. The complete genome sequence of a 33 Vallerdu J, Allue E, Bischoff JL, et al. 2005. Neanderthal from the Altai Mountains. Nature Short human occupations in the Middle Palaeo- 39 Kuhn SL, Stiner MC. 2006. What’s a mother 505:43–49. lithic level I of the Abric Romanı rock-shelter to do? The division of labor among Neander- (Capellades, Barcelona, Spain). J Hum Evol 48: thals and modern humans in Eurasia. Curr 157–174. Anthropol 47:953–980. VC 2016 Wiley Periodicals, Inc.