Particle size distribution of lithic assemblages and taphonomy of Palaeolithic sites

Journal of Archaeological Science 39 (2012) 3148e3166

Contents lists available at SciVerse ScienceDirect

Journal of Archaeological Science journal homepage: http://www.elsevier.com/locate/jas

Particle size distribution of lithic assemblages and taphonomy of Palaeolithic sites Pascal Bertran a, *, Arnaud Lenoble b, Dominique Todisco c, Pierre M. Desrosiers d, Mikkel Sørensen e a

INRAP, 156 avenue Jean Jaurès, 33600 Pessac, France PACEA, UMR 5199 CNRS, Université de Bordeaux 1, bâtiment de géologie, avenue des facultés, 33405 Talence, France c Département de Géographie, IDEES, UMR 6266 CNRS, Université de Rouen, rue Lavoisier, 76821 Mont Saint Aignan, France d Avataq Cultural Institute, 215 Redfern, Suite 400 Westmount, Québec H3Z 3L5, Canada e Department of Prehistoric Archaeology, University of Copenhagen, Njalsgade 80, DK-2300 Copenhagen S, Denmark b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 20 April 2011 Received in revised form 27 April 2012 Accepted 28 April 2012

Lithic debris derived from knapping and used tools can be assimilated to simple sedimentary particles that may undergo size sorting when exposed to geomorphic processes such as streamflow or overland flow. Sorting can be identified by comparing the size distribution of archaeological assemblages to that of experimental core reduction sequences. A new database including different types of raw material (mainly flint and quartzite) and Palaeolithic debitage (blade, Levallois, discoid, on anvil, and shaping) has been built for this purpose. Palaeoeskimo data have also been added to illustrate microlithic industries. For all the debitages and raw materials, the particle size of knapping products >2 mm in width fits with a power-law distribution and shows only minor fluctuations, the range of which is always <15% between experiments (all steps of the chaîne opératoire included up to the final tool). A lithic assemblage derived from block/core knapping or blank/preform production will display a particle size distribution close to the experimental distributions if not subsequently modified. Modifications may originate either from sedimentary processes or from anthropogenic factors. To help distinguishing amongst these, data on the impact of both water flows on sedimentary particles or experimental assemblages, and anthropogenic processes such as importation-exportation (of core, preforms or finished tools) or uneven spatial distribution of the different steps in core reduction and tool production within a site, are reviewed. By contrast to anthropogenic modifications, sedimentary processes are generally typified by strong impoverishment in or selective accumulation of fine-grained (<10 mm) artefacts together with a low intra-site variability (spatial homogenization) or a downslope size trend. Archaeological case studies taken from French Palaeolithic site are then detailed. Evidence for lithic redistribution implies that care should be taken in archaeological site analysis since sorting may impact significantly the initial technotypological balance of the assemblage. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Lithic taphonomy Water flow Experimental knapping Size sorting Palaeolithic Site formation processes

1. Introduction Once abandoned on the ground by prehistoric people, lithic debris derived from knapping and used tools can be assimilated to simple sedimentary particles that may undergo spatial redistribution by superficial geomorphic processes before being completely buried. These, and particularly water flows, are often able to transport selectively the particles according to their size, weight and shape. Particle size sorting is a feature that can be easily recognised in the sediments by grain size analysis, and is used by geomorphologists to evaluate flow velocity or distance to sediment * Corresponding author. E-mail addresses: [email protected] (P. Bertran), [email protected] (A. Lenoble), [email protected] (D. Todisco), [email protected] (P.M. Desrosiers), [email protected] (M. Sørensen). 0305-4403/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.jas.2012.04.055

sources. As already shown by pioneer works of Isaac (1967), Schick (1986, 1987), and later by Lenoble (2005), hydraulic modification can also be identified from archaeological lithic remains provided that some assumptions are made on the initial composition of the assemblages. Sorting evidence of artefacts has major implications for interpreting their spatial distribution, site patterning as well as palaeo-economy. For instance, lack (or under-representation) of coarse-grained items like blocks of raw material, large flakes derived from the first steps of block reduction, or large poorly exhausted cores in a site may be the result of either importation of already prepared cores or performs, or selective accumulation of fine to medium-sized lithics by flows. Similarly, lack of fine-grained artefacts due to hydraulic sorting may introduce a significant taphonomic bias in assemblage interpretation, particularly with respect to strongly microblade Upper Palaeolithic industries. On poorly vegetated slopes (typically during the cold periods of the

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

A hand-measured width (cm)

Pleistocene in western Europe) or in fluvial environments, where flows may have played a significant role in site formation, detailed taphonomic analysis of lithic assemblages is thus required to assess whether hydraulic sorting occurred or not, and hence, to avoid misleading behavioural interpretations based on poorly reliable archaeological material. We present here a review of the present knowledge on particle size analysis of lithic assemblages based on both experimental data and some archaeological case studies. Emphasis is given on methodology, experimental database, and potential interpretative difficulties of grain size data when applied to archaeological assemblages. Case studies taken from French Palaeolithic sites are then discussed. The general aim of the paper is to propose a tool that could serve for taphonomic analysis of Palaeolithic sites and complement those already available like reffitings (Villa, 1982; Bordes, 2003), fabrics (Bertran and Texier, 1995; Lenoble and Bertran, 2003; McPherron, 2005), intra-site spatial analysis (Le Grand, 1994; Lenoble and Bertran, 2008), and artefact abrasion (Petraglia and Potts, 1994).

3149

sieving

10 9 8 7 6 5 4 3 2 1 0

N (d > 5.0 cm) = 3

5 2 N (3.15 < d < 5.0 cm) = 14

3.15 2 N (2.0 < d < 3.15 cm) = 23

2 2

20

0

40

60

80

100

Artefact number

B 35

d = 20 mm d = 10 mm d = 5 mm

30

20 2

2. Particle size analysis applied to lithic assemblages: methodology Size analysis of coarse-grained (>2 mm) sediments consists in separating the particles into dimensional classes using a column of standard sieve meshes. The meshes follow usually a geometric progression (2 mm, 2.5 mm, 3.15 mm, 4 mm, 5 mm, 6.3 mm, 8 mm, 10 mm, 12.5 mm, 16 mm, 20 mm etc.). For round elements, the mesh d corresponds precisely to the diameter of the coarsest particle able to pass through, whereas for angular elements, d corresponds approximately to the particle width (Fig. 1). When the thickness is low or negligible as for flint flakes, the maximal width wmax that can pass through a mesh d is equal to dO2, i.e. wmax ¼ 1.414 cm if d ¼ 1 cm. Comparison between sieving and manually measured width w of large sets of archaeological material shows that the best estimate of wmax is close to dO2 even if the thickness of the pieces varies significantly (Fig. 2). Since sieving may damage the fragile edges of some artefacts, the following procedure has been adopted as a compromise between time-consuming width measurement of the pieces with a rule or a calliper rule and quicker but possibly piece-damaging sieving. Artefacts with a width lower than 1 cm, which are often very abundant in Palaeolithic sites and usually collected by sieving of the sediment by archaeologists, are separated using a standard sieve column. The simplified mesh series retained in this work is 2, 4, 5, 10 and 20 mm. For coarser material, piece width is manually measured with a rule or lithics are individually passed through sieve meshes that serve as gauges. To insure correct fit between sieving and measurement data, the size class boundaries of hand-

d

w~ ~d w

d ~~ 2 Fig. 1. Influence of shape on maximum particle width w able to pass through a sieve mesh d.

width (mm)

25 20 10 2

15 10

5 2

5 0 0

10

20

30 40 50 length (mm)

60

70

80

Fig. 2. A e Width w of 80 artefacts from a levallois debitage manually measured with a ruler. The size classes are indicated by symbols. Similar results are obtained by measurement and sieving if the class limits d0 ¼ dO2. B e Relationship between length and width of Aurignacian artefacts and sieving classes (from Lenoble, 2005).

measured w has to be equivalent to dO2, i.e. artefacts having w ranging from 14.1 to 28.3 mm correspond to the sieve class 10e20 mm. Table 1 gives rounded equivalents between d and w for usual class boundaries. In sedimentology, size distribution of particles is given by weighing the content of each dimensional class. For archaeological material, counting rather than weighting the pieces has to be preferred for the following reasons: (1) The shape of a weight distribution is mostly influenced by the coarsest and heaviest elements, which are the less numerous. In case of small archaeological assemblages (100e200 artefacts), small variation in the number of large pieces may lead to significant modification of the weight distribution, making comparison between assemblages unreliable. Numeric distribution makes it possible to eliminate such a disadvantage. (2) The smaller the lithics, the more prone to flow entrainment they are. A size distribution expressed in weight will be little sensitive to a moderate impoverishment in small artefacts. By contrast, a distribution expressed in piece number will make it possible to detect small modifications due to hydraulic sorting. (3) Most of the tools or blanks are medium or large-sized (except for microlithic industries). Their abundance within an assemblage depends on human behaviour and particularly site function, goals of the lithic production, and raw material availability and economy. By contrast, small lithics are mainly waste material that is abandoned where it is produced. As part of a taphonomic study focusing on assessing the role played by natural processes in site formation, numeric particle size

3150

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

Table 1 Size class limits to be taken for comparison between manually measured artefact width w and sieve mesh d. d (mm) dO2 (mm)

2 2.83

2.5 3.54

3.15 4.45

4 5.66

5 7.07

6.3 8.91

distributions minimizing the contribution of the coarse fraction seems of greatest interest. In sedimentology, representation of the particle size distribution is made using histograms, cumulative curves or triangular diagrams. Curves make it possible to calculate graphically parameters such as the median Md (d50), the size for which 50% of the material is finer, and sorting indices that reflect the tightening of the distribution around the median. Similar diagrams can be used for archaeological assemblages, but most of the classical sedimentological statistical parameters are thought to be of little interest, since the number of dimensional classes used for archaeological material analysis remains limited. As a consequence, comparison between archaeological and experimental lithic assemblages can be easily achieved using simple histograms and triangular diagrams. 3. Particle size distribution of experimental lithic assemblages Available data on particle size distribution of lithic assemblages are limited and lack standardization, mainly with respect to the grain size class limits, use of either width or length measurements, and counting or weighing the artefacts (Shott, 1994). Therefore, new experimental data were collected to build a homogeneous Palaeolithic database including different types of raw material (mainly flint and quartzite), and blank production techniques (blade, Levallois, discoid, on anvil, and bifacial shaping) the closest as possible to archaeological examples. Seven experienced knappers were involved (B. Aubry, L. Bourguignon, M. Brenet, A. Delagnes, P. Fouéré, J. Pelegrin and V. Mourre). For all the experiments, the hammers used were either hard or semi-hard (limestone). The experimenters knapped over plastic tarps and collected all knapping products in each replication. Block reduction was stopped usually before full exhaustion of the core when a large enough number of products were obtained. The products were then sieved, the smallest mesh used being 2 mm. Additional data from Palaeoeskimo knapping experiments including various raw materials (chert, flint, crystal quartz and slate) are also presented for comparison to document microlithic industries. Palaeoeskimo from the eastern Canadian Arctic (Desrosiers et al., 2008) have some similarities with French Mesolithic or the chaînes opératoires dedicated to microblade production typical of some Upper Palaeolithic cultures. Experimenters were J. Pelegrin, M. Sørensen and P.M. Desrosiers. Except in the shaping and some Palaeoeskimo experiments, no tools were produced from the blanks. The results together with those obtained by Stahle and Dunn (1982), Hansen and Madsen (1983), Schick (1986), Patterson (1990), Brown (2001) and Lenoble (2005) show a similar particle size distribution of the knapping products (Table 2, Figs. 3 and 4). The amount of fragments decreases rapidly when size increases and tends asymptotically to zero, i.e. lithic reduction via knapping yields a huge quantity of small-sized debris and few large artefacts. This asymmetric distribution fits a power law (Weibull distribution: Stahle and Dunn, 1982 or fractal distribution: Brown, 2001) that obeys the relation Nd ¼ dx, where Nd is the number of fragments larger than d. Few differences were found according to the raw material and production type. This may reflect an intrinsic property of hard rock fragmentation since strong similarity in size

8 11.31

10 14.14

12.5 17.68

16 22.63

20 28.28

31.5 44.55

50 70.71

distribution can be observed between the data obtained from knapping of flint blocks and those from rock fragmentation by natural phenomena (Turcotte, 1986; Bertran et al., 2006). Although weak, some differences between experiments are observed. Higher amounts of fine-grained (2e4 mm) particles are provided by flint as compared to quartzite. Means are 64.8% (N ¼ 10) and 60.5% (N ¼ 9) respectively for the discoid production (Fig. 5). Schick (1986) also reports small differences according to the tested rock type (quartz, obsidian, basalt, chert, ignimbrite). Coarsetextured rocks such as quartz and basalt that tend to give irregular fractures yield more fine-grained (0e10 mm) particles than finertextured homogeneous rocks. Clear trends also emerge due to the influence of the blank production mode. Among all the experimented Palaeolithic knapping techniques, the discoid production yields lesser amounts of fine-grained particles (mean 2e4 mm ¼ 64.8% for flint, N ¼ 10) than the other modes, and particularly the blade mode (mean 2e4 mm ¼ 67.5%, N ¼ 19). As expected, Palaeoeskimo knapping experiments give size distributions skewed toward fine particles (Table 3). In addition, they also illustrate the homogeneity of size distributions regardless the raw material tested and the techniques used (Fig. 6). For any production mode, different factors (yet poorly experimentally explored) may influence the size distribution of lithic assemblages including: (1) The size of the desired end products. Shaping of bifacial tools produces few large flakes (Patterson, 1990). Stahle and Dunn (1982) data indicate a deficiency in pieces >10 mm when compared with a discoid production, which increases along preform reduction. In our database, particle size data of shaping of a Middle-Palaeolithic-type biface of flint and a Neolithic-type handaxe of volcanic chert are given. The corresponding distributions remain as a whole within the variation range of the other experiments. Microblade production from small cores or flakes that are typical of Upper Palaeolithic industries will probably lead to under-representation of coarse-grained fractions. The Palaeoeskimo experiments are thought to yield a good approximation for this kind of chaîne opératoire. (2) Careful preparation of the striking platform may produce higher quantities of small debris. However, our data indicate that this has only a minor (although actual) impact on the size distribution. Two Levallois productions made with gesture economy and low preparation of the striking platform according to the modalities observed at the Middle Palaeolithic site of Hermies yield among the lowest amounts of the 2e4 mm fraction in our dataset. Although the highest values are provided by blade productions for which the preparation of the striking platform is advanced, the distribution range of the different production modes largely overlap. (3) A third also influencing factor on distribution variability is the size of the raw material blocks. This will have obviously an impact on the maximal size of the artefacts produced. However, since the coarsest fractions do not account for much in the distribution shape, it can be expected that this effect will remain rather marginal. (4) For a given production, the size of the artefacts produced during the successive stages of core reduction or blank preparation may vary significantly. Stahle and Dunn (1982) show

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

3151

Table 2 Particle size composition of experimental Palaeolithic debitages and shaping, fraction >2 mm. Sieve mesh d (mm) Piece width w (mm)

2e4 2.8e5.7

4e5 5.7e7

5e10 7e14

10e20 14e28

On anvil debitage On anvil debitage On anvil debitage Mean Unifacial discoid Unifacial discoid Unifacial discoid Failed discoid (fissured block) Discoid Discoid Discoid Discoid Discoid Mean Standard deviation Max. Min. Discoid Discoid Discoid Discoid Discoid Discoid Discoid Discoid Discoid Discoid Mean Standard deviation Max. Min. Levallois Levallois Levallois Levallois Centripetal reccurrent levallois Centripetal reccurrent levallois Centripetal reccurrent levallois Centripetal reccurrent levallois Convergent unipolar levallois Convergent unipolar levallois Convergent unipolar levallois Convergent unipolar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Mean Standard deviation Max. Min. Laminar Laminar Laminar Laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Bipolar laminar Bipolar laminar Bipolar laminar

67.82 65.93 61.57 65.10 55.07 55.26 65.05 51.72 62.41 68.50 67.22 55.53 63.35 60.46 6.13 68.50 51.72 69.67 63.90 64.43 63.99 63.67 70.17 64.26 64.67 58.58 64.80 64.81 3.24 70.17 58.58 66.11 68.30 58.95 78.80 65.19 64.79 65.57 66.68 66.81 69.95 70.19 69.28 72.24 68.27 68.09 68.11 63.64 67.86 67.17 68.89 67.81 4.01 78.80 58.95 68.79 75.37 58.87 69.94 71.41 68.87 56.07 69.69 67.26 68.03 66.78 69.37 69.20 69.61 66.38 71.30 68.40

10.73 12.22 12.23 11.73 10.14 13.82 8.65 17.24 8.81 10.44 8.17 8.36 11.55 10.80 3.02 17.24 8.17 8.74 9.09 7.22 11.66 8.30 6.08 10.47 8.08 12.62 7.89 9.02 2.02 12.62 6.08 9.41 9.71 13.57 7.41 10.52 9.50 10.60 9.51 10.99 9.28 9.99 9.43 9.44 9.92 10.15 10.51 10.95 9.59 10.54 9.52 10.08 1.20 13.57 7.41 11.52 13.83 16.54 10.74 8.69 10.34 11.48 8.98 9.37 8.41 9.86 8.69 8.91 6.98 10.11 10.19 9.66

10.34 12.96 16.59 13.30 16.22 15.13 10.03 17.24 14.03 13.10 13.39 19.14 13.94 14.69 2.64 19.14 10.03 14.75 14.55 13.40 14.77 16.26 15.47 16.97 12.87 15.21 14.47 14.87 1.21 16.97 12.87

6.90 4.81 4.37 5.36 11.15 9.21 6.92 0.00 6.29 4.76 6.79 8.63 4.38 6.46 3.23 11.15 0.00 2.73 4.16 8.76 5.44 6.57 6.08 3.61 7.49 9.06 6.58 6.05 2.11 9.06 2.73 21.68 15.52 6.72 3.33 4.29 5.11 4.75 5.09 4.50 3.77 4.23 4.67 3.46 5.81 5.35 4.55 5.67 3.93 5.39 5.56 4.79 0.89 6.72 3.33 4.47 1.50 3.75 4.50 4.54 4.98 12.46 4.47 5.93 4.08 6.12 5.26 6.13 5.75 4.47 4.70 4.99

18.16 9.21 14.94 16.00 15.13 15.76 13.51 13.91 12.77 13.60 12.06 12.68 13.51 14.00 15.35 15.38 14.52 13.78 14.13 1.89 18.16 9.21 14.40 8.35 17.17 12.10 12.12 13.22 15.08 14.53 14.71 16.31 14.17 13.83 13.58 15.40 17.02 11.89 14.75

20e31.5 28e44 1.53 2.22 2.62 2.12 6.08 5.26 7.27 3.45 4.32 1.47 2.07 4.04 4.38 4.26 1.83 7.27 1.47 1.91 3.38 5.15 2.85 3.46 1.10 2.53 2.99 3.24 5.26 3.19 1.28 5.26 1.10 1.50 2.18 1.49 0.79 2.86 1.68 2.05 1.26 2.41 1.45 1.53 1.69 1.21 2.41 2.18 1.50 2.64 2.07 1.34 1.23 1.77 0.56 2.86 0.79

1.83 1.63 3.61 1.43 2.02 2.29 2.83 1.83 0.88 1.03 1.28 1.57 1.34

31.5e50 44e71

>50 >71

2.68 0.74 1.75 1.72 1.01 0.66 1.38 3.45 3.78 1.47 1.87 4.04 2.39 2.23 1.25 4.04 0.66 0.82 4.68 1.03 1.30 1.73 1.10 1.81 3.59 1.29 0.99 1.83 1.28 4.68 0.82 1.06 0.99 0.90 0.37 2.08 2.26 1.61 1.21 1.47 1.64 1.04 1.04 1.31 1.06 0.63 0.93 1.56 1.03 0.82 0.68 1.20 0.50 2.26 0.37 0.82 0.96 3.67 2.71 1.01 0.86 0.66 0.82 0.59 0.89 0.11 1.03 1.17 1.03 0.53 0.21 0.65

0.00 1.11 0.87 0.66 0.34 0.66 0.69 6.90 0.36 0.27 0.49 0.27 0.00 1.11 2.18 6.90 0.00 1.37 0.26 0.00 0.00 0.00 0.00 0.36 0.30 0.00 0.00 0.23 0.43 1.37 0.00 0.25 3.30 0.19 0.09 0.13 0.66 0.29 0.48 0.31 0.00 0.25 0.30 0.28 0.07 0.09 0.40 0.20 0.14 0.23 0.34 0.25 0.16 0.66 0.00

0.39 0.10 0.66 0.09 0.12 0.00 0.11 0.00 0.15 0.21 0.21 0.14 0.21

Total

Raw material

Experimenter

Quartzite Quartzite Quartzite

V. Mourre V. Mourre V. Mourre

Quartzite Quartzite Quartz Quartzite Quartzite Quartz Quartzite Quartzite Quartzite

V. Mourre V. Mourre V. Mourre V. Mourre M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet

366 385 194 386 289 181 277 334 309 304 3025

Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint

L. Bourguignon J. Pelegrin M. Brenet M. Brenet M. Brenet L. Bourguignon L. Bourguignon L. Bourguignon M. Brenet M. Brenet

1605 1514 1547 3278 770 1369 1368 2062 955 1035 2412 2015 1070 1415 1103 2264 1023 1450 3064 2932 31132

Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint

B. Aubry B. Aubry L. Bourguignon B. Aubry M. Brenet M. Brenet L. Bourguignon L. Bourguignon M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet

2570 2408 1415 1843 2071 1044 305 2306 843 785 882 875 685 974 940 1404 4326

Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint

In Lenoble, In Lenoble, In Lenoble, In Lenoble, M. Brenet J. Pelegrin J. Pelegrin M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet

261 270 229 760 296 152 289 29 556 1092 1016 371 251 4052

2005 2005 2005 2005

(continued on next page)

3152

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

Table 2 (continued ) Sieve mesh d (mm) Piece width w (mm)

2e4 2.8e5.7

4e5 5.7e7

5e10 7e14

10e20 14e28

20e31.5 28e44

31.5e50 44e71

>50 >71

Bipolar laminar Bipolar laminar Mean Standard deviation Max. Min. Neolithic handaxe Biface Biface Biface Biface Biface Biface Biface Biface Biface Biface Biface Biface Biface Mean Standard deviation Max. Min.

63.91 65.62 67.46 3.75 71.41 56.07 58.73 64.15 62.39 68.43 64.59 66.25 62.94 68.47 68.98 68.39 68.85 70.79 69.98 56.80 65.70 4.28 58.73 70.79

8.67 10.79 9.41 1.11 11.48 6.98 13.89 11.01 11.10 9.86 11.21 7.75 4.98 10.10 10.23 10.81 9.78 11.24 11.35 10.28 10.26 2.01 13.89 4.98

16.53 14.83 14.53 1.48 17.02 11.89 19.05 15.60 16.84 14.60 15.30 18.40 20.40 14.45 15.18 13.71 16.09 14.60 13.22 12.56 15.71 2.26 20.40 12.56

7.66 5.62 5.81 2.05 12.46 4.08 6.22 6.18 6.99 5.12 5.87 5.39 6.72 5.42 4.29 3.87 5.15 2.51 4.04 4.15 5.14 1.25 6.99 2.51

2.22 2.02 1.85 0.71 3.61 0.88 1.72 1.69 2.11 1.43 2.76 1.80 2.99 1.23 0.99 2.74 0.00 0.79 1.09 15.78 2.65 3.87 15.78 0.00

0.81 1.01 0.76 0.30 1.17 0.11 0.13 1.05 0.48 0.37 0.27 0.28 1.24 0.25 0.11 0.32 0.00 0.00 0.16 0.31 0.35 0.36 1.24 0.00

0.20 0.11 0.18 0.16 0.66 0.00 0.26 0.30 0.10 0.19 0.00 0.14 0.75 0.08 0.22 0.16 0.13 0.07 0.16 0.10 0.19 0.18 0.75 0.00

that the amount of fine-grained (3.2e6.4 mm) particles increases as shaping of a handaxe progresses, by about 8% between the early step of flaking and the late shaping stage. This is also illustrated by the Palaeoeskimo experiments. The last steps of tool production including the short blank transformation process leading to the final tool (e.g. end and side scrapers) and the secondary steps of the blank transformation involving the final trimming of a bifacial preform into an end blade point tend to produce larger percentages of small-sized lithics. These amount to 86.7% (2e4 mm, N ¼ 10) in comparison to 68.7% (2e4 mm, N ¼ 12) when all the steps of the chaîne opératoire are included.

Total

Raw material

Experimenter

496 890 27062

Flint Flint

M. Brenet M. Brenet

756 2007 1045 1603 1124 723 402 1218 909 620 777 1397 643 963 14187

Chert Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint

P. Fouéré J. Pelegrin M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet

d > 10 mm (w > 14 mm)

A

C 1

0

laminar discoid anvil levallois façonnage

0.2

0.8

0.4

0.6

0.6

0.4

0.8

0.2

90

1

F

max.

80

0.6

0.8

0

0.2

0.4

min.

60

d > 20 mm (w > 28 mm)

C

B

50

M

10 mm > d > 4 mm (14 mm > w > 5.7 mm)

anvil discoid levallois laminar façonnage

4 mm > d > 2 mm (5.7 mm > w > 2.8 mm)

70

percentage

1

0

1

0

laminar discoid anvil levallois façonnage

0.2

0.8

40 0.4

0.6

30 0.6

0.4

20

0.8

0.2

10

1

0

0 2-5

5 - 10

10 - 20

20 - 31.5 31.5 - 50

> 50

dimensional class (mm) Fig. 3. Compared mean particle size distributions of different Palaeolithic experimental debitages.

F

1

10 mm > d > 5 mm (14 mm > w > 7 mm)

0.8

0.6

0.4

0.2

0

M

20 mm > d > 10 mm (28 mm > w > 14 mm)

Fig. 4. Particle size composition of the experimental Palaeolithic debitages, triangular diagrams. A e fraction d > 2 mm, B e fraction d > 5 mm.

C

anvil (N=3)

80

discoid (N=20)

100

1

2 - 4 mm (%)

2 - 4 mm (%)

40 20 0

0

blank/preform production blank/preform transformation into tool

0.2

0.8

0.4

0.6

Palaeolithic debitages

0.4

60

3153

d > 10 mm (w > 14 mm)

levallois (N=20)

80

B laminar (N=19)

100

flint (discoid, N=10)

A

quartzite (discoid, N=9)

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

0.6

60

0.8

0.2

40

1

0

F

20

1

0.8

0.6

0.2

0.4

4 mm > d > 2 mm (5.7 mm > w > 2.8 mm)

0

M

10 mm > d > 4 mm (14 mm > w > 5.7 mm)

Fig. 6. Particle size composition of the Palaeoeskimo debitages, triangular diagram. The grey area corresponds to the range of Palaeolithic debitages.

0

Fig. 5. Proportion of the 2e4 mm fraction as a function of raw material (A) and debitage types (B). The bars indicate the standard deviation.

Table 3 Particle size composition of Palaeoeskimo debitages and shaping, fraction >2 mm. The experiments are based on the analysis of Palaeoeskimo lithic technology from various sites in Nunavik (Canada). Sieve mesh d (mm)w (mm) Piece width w (mm) b

Lamellar debitage Flake production (opportunistic scheme) Flake production (opportunistic scheme) Lamellar debitage Mean Standard deviation Max. Min. Bifacial shaping Bifacial shaping Bifacial shaping Bifacial shaping Mean Standard deviation Max. Min. Flake production and bifacial shaping Flaking, bifacial shaping and lamellar debitage Flaking, bifacial shaping and lamellar debitage Flake production and bifacial shaping Mean Standard deviation Max. Min. Tool retouching Tool retouching Bifacial tool retouching Tool retouching Tool retouching Tool retouching Tool retouching Tool retouching Tool retouching Tool retouching Mean Standard deviation Max. Min. Bifacial tool retouching a

2e4 2.8e5.7

4e5 5.7e7

5e10 7e14

10e16 14e22.6

>16 >22.6

74.42 60.27 62.50 70.00 66.80 6.57 74.42 60.27 76.57 75.68 71.25 73.29 74.20 2.40 76.57 71.25 71.22 67.55 60.87 60.70 65.09 5.19 71.22 60.70 86.36 94.74 96.61 80.00 93.42 80.00 77.46 83.50 82.07 92.86 86.70 7.09 96.61 77.46 83.15

11.63 10.04 8.95 11.25 10.47 1.22 11.63 8.95 9.34 7.03 10.70 7.92 8.75 1.61 10.70 7.03 8.18 8.38 7.07 8.11 7.93 0.59 8.38 7.07 0.00 1.64 1.06 4.55 3.29 7.50 12.32 7.89 9.05 0.00 4.73 4.27 12.32 0.00 6.32

11.63 20.31 21.05 16.75 17.44 4.30 21.05 11.63 13.69 13.34 15.60 14.00 14.15 1.00 15.60 13.34 15.52 18.25 20.33 22.41 19.13 2.95 22.41 15.52 13.64 0.99 2.33 13.64 1.97 8.50 8.80 7.65 8.19 7.14 7.28 4.45 13.64 0.99 9.97

2.33 4.02 4.32 2.00 3.17 1.17 4.32 2.00 0.40 2.74 2.14 4.45 2.43 1.67 4.45 0.40 3.63 4.76 10.87 6.67 6.48 3.18 10.87 3.63 0.00 0.00 0.00 1.82 1.32 4.00 0.00 0.97 0.69 0.00 0.88 1.28 4.00 0.00 0.00

0.00 5.36 3.18 0.00 2.13 2.62 5.36 0.00 0.00 1.22 0.31 0.34 0.47 0.52 1.22 0.00 1.45 1.06 0.87 2.11 1.37 0.55 2.11 0.87 0.00 2.63 0.00 0.00 0.00 0.00 1.41 0.00 0.00 0.00 0.40 0.90 2.63 0.00 0.56

Raw material

Experimenter

Stepsa

Quartz crystal Chert Chert Quartz crystal

J. Pelegrin M. Sørensen M. Sørensen M. Sørensen

BP BP BP BP

495 329 327 292 1443

Chert Slate Slate Slate

M. Sørensen P.M. Desrosiers P.M. Desrosiers P.M. Desrosiers

BT1 BT1 BT1 BT1

688 567 230 285 1770

Flint Flint Flint Flint

J. J. J. J.

BP BP BP BP

22 38 59 55 76 25 71 103 145 14 608

Flint Chert Chert Chert Chert Chert Chert Chert Chert Quartz crystal

J. Pelegrin M. Sørensen M. Sørensen M. Sørensen P.M. Desrosiers P.M. Desrosiers M. Sørensen M. Sørensen M. Sørensen M. Sørensen

SBT SBT SBT SBT SBT SBT SBT SBT SBT SBT

178

Flint

J. Pelegrin

BT2

Total 43 224 440 50 757

Pelegrin Pelegrin Pelegrin Pelegrin

þ þ þ þ

BT1 BT1 BT1 BT1

BP ¼ blank production including core preparation; BT1 ¼ blank transformation into preform; SBT ¼ short blank transformation into tool; BT2 ¼ last steps of preform transformation into tool. b Lamellar debitage ¼ microblades production.

3154

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

(5) The knapper’s skills may also exert a noticeable influence. This is suggested by the blade production made by J. Pelegrin who was able to manage the core with high efficiency and with relatively little waste while obtaining homogenous products. The amount of fine-grained (2e4 mm) fraction is lower by about 10% than that of the other blade productions. (6) Finally, a secondary fragmentation of the material may occur as a consequence of trampling (Vallin et al., 2005). Trampling, known to be spatially widespread on the surface of archaeological sites (Rasic, 2004), could have a significant impact in artefact concentrations, especially where the pieces are in contact to each other. Even if size reduction of flakes (or other artefacts) by this process is often thought to be limited (Schiffer, 1987), it remains to be accurately quantified. In the present state of knowledge, the particle size distribution of knapping products >2 mm appears to be rather constant, with only minor fluctuations, the range of which is always <15% between experiments (all steps of the chaîne opératoire included up to the final tool). If a distinction is made between experimental assemblages according to the steps of the chaîne opératoire, the variability (based on Palaeoeskimo knapping data) can reach 20% and up to w35% between experiments. When considering all steps of tool production together, the mean experimental size distribution can be considered as a good approximation for defining nonsorted assemblages whatever the production mode or raw material type. As a consequence, each lithic assemblage derived from block/core knapping as well as blank/preform production will display a particle size distribution close to the mean experimental distribution if not subsequently modified. One of the main difficulties for studying archaeological assemblages is the need to extensively collect all lithic artefacts. This implies careful water sieving of the archaeological sediment. Since post-depositional processes such as bioturbation may lead to vertical scatter of the artefacts (particularly the smallest: Van Nest, 2002), the whole sedimentary volume that includes archaeological pieces has to be sieved. Even if the best conditions are set up, recovering of the whole original assemblage is probably never achieved. Estimation of the recovery rate was made by Lenoble (2005) for experiments in natural environments (a sand quarry near Fumel, southwest France). A known number of artefacts was disseminated on ca. 10 m2 surfaces and recollected after exposition to rain and overland flow over some months. The recovery rate, i.e. the number of pieces found back to the initial number, reached 96.5% all fractions mixed. For the 2e4 mm fraction, the rate was 95.2%, and 100% for the 5e10 mm fraction. A conservative value of 90% can therefore be taken as the recovery rate for artefacts larger than 5 mm in case of careful sieving. This means that deviation of the archaeological distribution lower or equal to 10% from the experimental data have to be considered as non-significant. Results obtained here on Palaeolithic site that gave distributions similar as the experimental ones (see infra) seem to validate this assumption and indicate thus minor bias due to incomplete recovery of smallsized lithics in usual conditions of excavation. Sieving with a 2 mm mesh is time-consuming even if made on selected squares considered as representative of the whole site and, as a consequence, not routinely made on excavations. In accordance, experimental distributions were also calculated on the fraction d  5 mm to fit with the most commonly used mesh diameter by archaeologists (Table 4). The obtained distributions are strongly similar to the previous ones except for a slightly larger scatter of data according to the production mode (Fig. 4). The mean 5e10 mm fraction amounts for 63.2% for blade production that tends to produce long but narrow artefacts, while it is ca. 57.3% and 56.9% respectively for discoid and on anvil production.

4. Size sorting by water flow: the geomorphological dataset and archaeological experiences Since Hjülström (1939), abundant experimental data on the mobility of particles in flowing fluid-sediment mixtures have been collected. Two kinds of flows are usually distinguished by geomorphologists: (1) dense sedimentary flows with a high sediment concentration (30e80%) giving a flow a high viscosity, i.e. debris flows (Coussot and Meunier, 1996; Fig. 7), and (2) diluted (common) to hyperconcentrated flows where the suspended particle load remains weak, respectively 0e5% and 5e30%. Debris flows occur typically on steep slopes (7e35 ) and their ability to sort the particles is low, leading typically to the emplacement of diamictic layers (Fig. 8). Electro-chemical interactions between clay particles and friction between coarsest elements lead to muddy behaviour that prevents particle settling. However, it has been observed that coarse particles tend to concentrate at the surface and to be pushed aside during flow. Few data are available on the resulting sorting. Fig. 9 shows the relative downslope impoverishment in coarse-grained particles in the levees of an alpine debris flow over a 50 m distance. Fine-grained (clay to silt) particles remain suspended by turbulence in diluted and hyperconcentrated flows whereas coarser material is transported by rolling, sliding and saltation on the bed. The particles are respectively in permanent and intermittent contact with the bed, both phenomena being referred to as bedload transport. The flow ability to transport particles depends on the tractive force it exerts on the bed. This is a function of flow speed, depth, bed rugosity and channel configuration (Middleton and Southard, 1984). When the grain size range of the bed is large, particle motion is controlled by the largest elements, finer grains being protected from entrainment by the former. The tractive force applied to a particle the size of which largely overcomes that of the mean bed grain size is greater than if protected by others with a similar size. Therefore, at equivalent size, flow ability for lithic artefact entrainment must be larger when these lie on a sandy substrate than on gravel. Usually, particle transport occurs intermittently as a function of flow variation and local bed topography. In sandy rivers, bedforms correspond mainly to ripples, dunes or antidunes for increasing flow velocity (Reineck and Singh, 1980; Middleton and Southard, 1984). When gravel abounds, they tend to form localized accumulations with a current-parallel or transverse elongation, which grow or migrate progressively upon successive floods. Overland flow on slopes is typified by shallow water thickness usually in the order of a few millimetres. Within such conditions, the threshold for particle detachment and transport is modified in comparison to what occurs in streamflow. The main processes involved are rainsplash that favours particle motion and low submersion of the coarse elements (bed rugosity is high with respect to water depth). A large panel of factors influence overland flow efficiency, such as soil infiltration capacity, soil moisture content prior to rainfall or snowmelt, microtopography (including vegetation), and slope length and steepness (Campebll, 1989). Diffuse (inter-rill or unconcentrated) overland flow does not create special macroscopic bedforms, but coarse poorly mobile particles tend to remain near their outcropping area and form lag deposits, i.e. residual concentration at the ground surface (Fig. 10). A longitudinal size gradient can sometimes be observed at the whole slope scale where overland flow is the dominant geomorphic process, with upslope pavements and a sheet of fine-grained material at the slope foot. For concentrated overland flow (in rills or gullies), bedforms are close to those observed in steep rivers. Coarsegrained material forms localized accumulations either because of flow inability to transport it (pavement of outsized blocks), or

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

3155

Table 4 Particle size composition of experimental debitages and shaping, fraction >5 mm. Sieve mesh d (mm) Piece width w (mm)

5e10 7e14

10e20 14e28

20e31.5 28e44

31.5e50 44e71

>50 >71

On anvil On anvil On anvil Mean Unifacial discoid Unifacial discoid Unifacial discoid Discoid Discoid Discoid Discoid Discoid Discoid Mean Standard deviation Max. Min. Discoid Discoid Discoid Discoid Discoid Discoid Discoid Discoid Discoid Discoid Mean Standard deviation Max. Min. Levallois Levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Laminar levallois Convergent unipolar levallois Convergent unipolar levallois Convergent unipolar levallois Convergent unipolar levallois Centripetal reccurrent levallois Centripetal reccurrent levallois Centripetal reccurrent levallois Centripetal reccurrent levallois Mean Standard deviation Max. Min. Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Unipolar laminar Bipolar laminar Bipolar laminar Bipolar laminar Bipolar laminar Bipolar laminar Mean Standard deviation Max. Min.

48.21 59.32 63.33 56.96 46.60 48.94 38.16 48.75 62.17 54.40 52.99 53.85 55.56 51.27 6.73 62.17 38.16 68.35 53.85 47.27 60.64 58.02 65.12 67.14 47.25 52.81 53.01 57.35 7.77 68.35 47.25 66.12 66.81 65.82 57.56 62.08 65.50 60.38 68.20 65.15 63.82 60.85 66.98 64.44 63.87 61.50 62.22 63.50 66.19 63.94 2.75 68.20 57.56 60.92 63.13 45.00 68.09 62.94 69.19 60.68 63.02 62.00 65.79 72.40 64.23 67.23 60.29 62.86 63.19 6.10 72.40 45.00

32.14 22.03 16.67 23.61 32.04 29.79 26.32 21.88 22.61 27.60 23.88 25.64 17.46 25.25 4.39 32.04 17.46 12.66 15.38 30.91 22.34 23.46 25.58 14.29 27.47 31.46 24.10 22.76 6.69 31.46 12.66 24.47 24.12 18.88 26.37 24.58 21.28 22.31 17.43 24.16 25.75 20.28 18.14 21.34 21.91 17.65 19.89 19.94 21.38 21.66 2.77 26.37 17.43 22.82 23.96 38.00 20.93 25.38 17.30 26.21 23.96 28.00 24.56 19.00 25.38 22.76 27.94 23.81 24.67 4.72 38.00 17.30

7.14 10.17 10.00 9.10 17.48 17.02 27.63 15.00 6.96 8.40 11.19 12.82 17.46 14.88 6.16 27.63 6.96 8.86 12.50 18.18 11.70 12.35 4.65 10.00 10.99 11.24 19.28 11.97 4.23 19.28 4.65 5.41 5.75 6.63 10.93 10.00 7.02 10.38 9.17 6.00 5.69 10.85 6.98 7.74 7.93 11.76 6.53 8.59 5.30 7.93 2.12 11.76 5.30 9.22 7.83 11.00 6.71 8.63 9.73 12.14 8.33 4.00 4.39 5.43 8.46 6.11 8.09 8.57 7.91 2.27 12.14 4.00

12.50 3.39 6.67 7.52 2.91 2.13 5.26 13.13 6.96 7.60 11.19 6.84 9.52 7.28 3.62 13.13 2.13 3.80 17.31 3.64 5.32 6.17 4.65 7.14 13.19 4.49 3.61 6.93 4.63 17.31 3.61 3.29 2.65 7.14 4.82 2.92 4.34 6.15 4.59 3.66 3.16 6.60 7.91 5.23 4.90 8.56 8.81 6.75 5.09 5.36 1.93 8.81 2.65 5.10 4.15 2.00 3.86 2.54 3.78 0.49 4.69 5.33 4.39 2.26 1.15 2.95 2.94 4.29 3.33 1.43 5.33 0.49

0.00 5.08 3.33 2.81 0.97 2.13 2.63 1.25 1.30 2.00 0.75 0.85 0.00 1.32 0.81 2.63 0.00 6.33 0.96 0.00 0.00 0.00 0.00 1.43 1.10 0.00 0.00 0.98 1.96 6.33 0.00 0.71 0.66 1.53 0.32 0.42 1.86 0.77 0.61 1.02 1.58 1.42 0.00 1.26 1.40 0.53 2.56 1.23 2.04 1.11 0.67 2.56 0.00 1.94 0.46 2.00 0.41 0.51 0.00 0.49 0.00 0.67 0.88 0.90 0.77 0.95 0.74 0.48 0.75 0.57 2.00 0.00

Total

Raw material

Experimenter

56 59 60

Quartzite Quartzite Quartzite

V. Mourre V. Mourre V. Mourre

103 47 76 160 230 250 134 117 63

Quartzite Quartzite Quartz Quartzite Quartzite Quartzite Quartzite Quartzite Quartzite

V. Mourre V. Mourre V. Mourre M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet

79 104 55 94 81 43 70 91 89 83

Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint

L. Bourguignon J. Pelegrin M. Brenet M. Brenet M. Brenet L. Bourguignon L. Bourguignon L. Bourguignon M. Brenet M. Brenet

425 452 196 311 240 484 260 327 683 633 212 215 478 429 187 352 326 491

Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint

L. Bourguignon B. Aubry M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet L. Bourguignon L. Bourguignon M. Brenet

412 217 99 1607 567 534 589 607 474 678 624 1001 2959 317 584

Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint

M. Brenet J. Pelegrin J. Pelegrin M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet

(continued on next page)

3156

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

Table 4 (continued ) Sieve mesh d (mm) Piece width w (mm)

5e10 7e14

10e20 14e28

20e31.5 28e44

Neolithic handaxe Biface Biface Biface Biface Biface Biface Biface Biface Biface Biface Biface Biface Biface Mean Standard deviation Max. Min.

62.85 69.57 63.54 67.24 63.24 70.74 63.57 67.43 73.02 65.89 75.30 81.27 70.83 38.17 66.21 9.73 81.27 38.17

24.90 22.71 26.35 23.56 24.26 20.74 20.93 25.29 20.63 18.60 24.10 13.94 21.67 12.62 23.80 4.07 26.35 13.94

6.83 6.28 7.94 6.61 11.40 6.91 9.30 5.75 4.76 13.18 0.00 4.38 5.83 47.95 6.55 11.42 47.95 4.38

because of reconcentration of gravel material (tightly packed, usually imbricated accumulations of homometric gravel giving birth to step-like bedforms) (Fig. 11). Small-scale (a dm to a m in length) coarse-grained bedforms alternate with sandy areas (Fig. 12). Sorting is usually lower than for equivalent bedforms in rivers. The main processes involved are (1) mixing of bedload material and local inputs derived from bank erosion and sliding, (2) infiltration of fine-grained particles in between the gravel during low water stages (the “sieve deposition”, Moss and Walker, 1978), and (3) interaction with other geomorphic or pedogenic processes such as wind action or bioturbation during phases of overland flow inactivity.

Fig. 7. Debris flow, Vallon Laugier, French Alps.

31.5e50 44e71 4.22 0.48 1.81 1.72 1.10 1.06 3.88 1.15 0.53 1.55 0.00 0.00 0.83 0.95 2.35 1.26 4.22 0.48

>50 >71 1.20 0.97 0.36 0.86 0.00 0.53 2.33 0.38 1.06 0.78 0.60 0.40 0.83 0.32 1.09 0.56 2.33 0.00

Total 498 207 277 348 272 188 129 261 189 129 166 251 120 317

Raw material

Experimenter

Chert Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint Flint

P. Fouéré J. Pelegrin M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet M. Brenet

The use of geomorphological data for archaeological purposes raises some issues. These are mainly: (1) Natural deposits may have undergone long lasting sorting upon repeated phases of transport and redeposition before final burial. By contrast, rapidly buried archaeological levels after occupation imply minimal exposition to flow before complete burial. Therefore, documentation of the early stages of redistribution by flow is of prime importance for archaeology and makes it necessary to get appropriate experimental data.

Fig. 8. Section view of poorly sorted (diamictic) layer with scattered archaeological remains interpreted as debris flow deposits, Artenac cave, southwest France (excavation A. Delagne).

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

3157

0 - 5 cm 5 - 10 cm 10 - 20 cm > 20 cm 40° 37° 35° 34° 32° 100 %

30°

upslope channel

28° 25° 23° 18° 19° 0

9°

8°

13°

14°

16° coarse-grained levees axial erosion

30 m

1m levees, no erosion

frontal lobe Fig. 9. Evolution of the size of pebbles at the surface of debris flow levees, the French Alps (from Bertran and Texier, 1994).

Fig. 10. Lag deposits and rill, La Mortice, French Alps.

Fig. 11. Gravel accumulation in a steep alluvial bed, Gavarnie, French Pyrenees.

3158

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

Fig. 12. Overland flow sand and gravel sheets, cultivated field, Grenoble, French Alps.

To date, experimental data on archaeological material taphonomy are rather poor. Most derive from the experiments performed by Schick (1986, 1987) in artificial canals and ephemeral rivers of Africa, and Frostick and Reid (1983), Reid and Frostick (1985), Petraglia and Nash (1987), and Lenoble (2005) for overland flow. The influence of artefact morphology on transport has been documented by Schick (1986). Thick artefacts, and particularly cores, may be as mobile as much smaller flakes on a sandy substrate because of the larger surface exposed to flow traction. Angular bladed artefacts tend to adopt a stable position, flat side downward, whereas approximately spherical artefacts (cores, choppers) are more prone to rolling. The experiments also show that interactions between artefacts (e.g. shadow and cluster effects) play an important role on movement, either because of tight packing and mutual blockage or because of local flow disturbance and acceleration around large artefacts that favour entrainment of the surrounding small particles. All these factors suggest that size sorting of archaeological material will be globally weaker than that of rounded alluvial gravel. Schick’s (1986) experimental data mainly show that: (1) Modifications of lithic assemblages due to floods are variable according to site location (e.g. floodplain, river bank, channel bed) and flow characteristics. They range from simple burial by mud drapes to long-distance displacements (some tens of metres) of the whole assemblage. Artefact deposition does not occur randomly but local secondary concentrations of sorted artefacts are usually observed. (2) Distance of movement for individual particles is primarily a function of artefact dimension, the shape factor playing only a secondary role. Grain size analysis is thus the best way to test hydraulic sorting of the material. (3) After remobilization, distinct areas typified by a specific particle size can be observed. In the upstream part of the so-

called “Flooded workshop site” (i.e. near the original location of the experimental assemblage), the assemblage showed strong impoverishment in artefacts <2 cm, which amount for ca. 20% versus 70% for the reference material. By contrast, cores and choppers amount to 50% of the modified assemblage. The fraction <2 cm decreases downstream in a jigsaw way to become predominant some 20 m away from the original location (Fig. 13). Strong variability is also observed beside the general trend to size reduction with increasing distance. The experiments indicate that the whole assemblage can be modified very quickly on a river bed, even in relatively low-energy environments like a floodplain. Both small and large artefacts are

“ Flooded Workshop site " 100 % small artefacts (< 2 cm)

(2) Most of the lithic artefacts (e.g. flakes, blades) are flat-shaped and that should have a significant influence on particle behaviour within a flow.

80 60 40 20 0 0

10

20

30

40

distance to site (m) Fig. 13. Proportion of artefacts <2 cm as a function of the distance to the original site in a river bed (from Schick, 1986).

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

put in motion, although at different rates. Simple ratios such as the proportion of artefacts <2 cm or that of flakes to cores plus the large tools (bifaces, choppers) provide robust indication of hydraulic sorting. The grain size composition of the remobilized assemblages shows the following pattern (Fig. 14): (1) At the original site location, the particle size curve of the residual assemblage remains approximately similar to the initial one for the early stages of redistribution by flow (less than 50% of the artefacts have been removed away from the site). However, the proportion of fine-grained particles is significantly reduced. For further stages of residualisation, the particle-size histogram tends to become flat, the proportion of coarse artefacts increasing progressively. (2) Farther downstream, the size composition of the redeposited artefacts fits a normal distribution because of sorting. Near the original site, the mode is close to 20e40 mm or 40e80 mm. A shift toward smaller sizes occurs as the distance to the source increases. Size distribution evolves only slightly as a consequence of longer exposition to flow, but the amount of recovered artefacts by unit surface decreases.

percentage

residual site 30 20 10

percentage

0

coarse-grained concentration

30

3159

The most detailed experiments for overland flow have been made by Lenoble (2005). The main conclusions that can be drawn from these are the following: (1) Overland flow, both diffuse on bare slopes and concentrated in rills, is able to transport the full size range of the archaeological material. Like in Schick’s experiments, size is the best explaining factor for the length of transport while shape plays only a minor role. (2) Size sorting develops quickly and become significant from the first rain event. Sorting increases as a function of the length of exposition to flow. (3) Spatial and temporal fluctuations of flow conditions and threshold for particle entrainment exceed those in rivers, because of rapid variation of both depth due to microtopography and water flow due to fluctuation of rain intensity and soil saturation. Therefore, the ability of overland flow for sorting is weaker. The distance of displacement for particles of a given size is strongly variable during a single event. Small (2e4 mm) artefacts are always present in significant amounts in the downslope accumulation area but also in the upslope residual site. (4) The initial artefact arrangement plays a determinant role. Heaps of flakes tend to deviate the water trickles and the isolated peripheral artefacts are preferentially removed. As a consequence, the size composition of the heap is only slightly modified during the first stages, but the area of the heap decreases progressively. The particle size composition of the experimental material is illustrated on Fig. 15 according to Lenoble (2005).

20

5. Particle size analysis of archaeological lithic material

10

Interpretation of the particle-size composition of archaeological material is primarily based on a comparison with the experimental

0

d > 10 mm (w > 14 mm)

percentage

40

flake streak

30

C 1

10

0.2

0.8

0

0.4 F M C

30

downslope redeposited 0.2 material

20

F M C

0.8

F M C

F

1 4 mm > d > 2 mm (5.7 mm > w > 2.8 mm)

1 0.8

0.6

0.4

lag deposits, residual concentrations (first stages of residualisation)

0.2

0 M 10 mm > d > 4 mm 14 mm > w > 5.7 mm)

+

c ha ore m s m er s

cm 6

cm -1 8

-8 4

-4

cm

cm 2

-2 1

-1

cm

0 5

0.6

0

10

0.

increasing sorting

depositional area

40

rillwash material ultimate stages of residualisation

0.4

0.6

50 percentage

0

20

Fig. 14. Particle size distribution of a lithic assemblage modified by floods (from Schick, 1986). The histograms have typically a bell shape, the mode of which shifts towards small sizes as the distance to the site increases. ‘Coarse-grained concentrations’ are secondary accumulations of coarse-grained lithics close to the original site location, ‘Flake streaks’ are ribbon-like concentrations of lithics, while ‘accumulation area’ corresponds to the most distal area of lithic accumulation.

F M C

Fig. 15. Particle size evolution of lithic assemblages affected by overland flow, modified from Lenoble (2005). The hatched area corresponds to experimental Palaeolithic debitages.

3160

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

database. Such a comparison is valid only if the whole size spectrum of the assemblage has been retrieved, usually from test areas large enough to provide a representative number of artefacts. Deviation of the archaeological composition from the experimental one implies sorting. Sorting may originate from anthropogenic factors due to specific activities, or may be induced by sedimentary processes. Anthropogenic factors are related to site function and techno-economic management of raw material and tools by humans. Also important is the intra-site spatial distribution of activity areas of core preparation, blank production, bifacial reduction, shaping and use of tools. Sites close to raw material outcrops and dominated by core preparation and reduction to provide blanks and preforms are likely well described by experimental data. However these may be less representative for other sites typified by the consumption of tools for butchering and others activities, with subordinate flaking but abundant products yielded by tool reshaping or breakage. Importation of already prepared cores, blanks or preforms to the site rather than blocks of raw material and intense trimming may produce under-representation of large flakes provided by the early steps of core exploitation. In contrast, importation of finished tools may increase the number of coarse elements. The impact of such factors can be evaluated by coupling particle size and technological analysis of the artefacts. At the Palaeoeskimo Tayara site, for instance, the size distribution of four test units in level II is largely asymmetric and positively skewed, i.e. the assemblage is largely dominated by small-sized lithics (Todisco et al., 2009). Technological analysis shows that natural blocks were managed to principally manufacture small microblade blanks. Flake by-products induced by core preparation and rejuvenation were almost systematically collected in order to serve themselves as blanks (Desrosiers, 2009), the transformation of which yielded small-sized elements (<1 cm long). Such technological specificities lead to an underrepresentation of large lithics in the assemblage and the hypothesis of natural sorting has thus been rejected. As mentioned above, human-induced sorting is assumed to modify only slightly the numeric particle-size composition of the assemblage. For instance, doubling the number of artefacts > 3 cm to simulate importation of half of the coarse-grained material does not alter significantly the general composition if documented on the fraction >2 mm because of the much greater number of finegrained artefacts. If working only on coarse material (>1 cm or more), discrepancies with the experimental data are likely to appear due to exportation-importation of specific products by humans. As a consequence, interpretation of the deviation from the reference composition needs to take into account the technological analysis, as described by Schick (1986). Another potential factor concerning human-induced sorting is the presence in a single site of activity areas dedicated to either production, reshaping or specific use of lithic material producing significant intra-site particle size variability. Evidence for periodic cleaning of the knapping areas, which were often located close to the fireplaces, and subsequent discharge of the coarsest debris at the periphery of the dwelling have been found in French Palaeolithic sites (Pigeot, 1984). This led to a concentration of finegrained artefacts in the production area, whereas larger pieces were lacking. It is assumed that the particle size composition of the discarded material lacks small artefacts (particularly if the detritus have been manually evacuated one by one to the dumping area) (cf. Fladmark, 1982; Behm, 1983; Healan, 1995). Port-de-Penne, SW France (Fig. 16), is an Epipalaeolithic site in a floodplain which was buried by silty overbank deposits (Detrain et al., 1996). This site has been used to test the potential impact of anthropogenic intra-site sorting. The sedimentary context is favourable to good site preservation because of low flow energy

0° Hermies

Paris

Le Casseux

45° Port-de-Penne

100 km

Auriac-Duclos Romentères

Fig. 16. Location of the study sites.

and high sedimentation rate, so that particle size composition and spatial distribution of the artefacts are assumed to reflect mainly anthropogenic factors. Level 2 exhibits a fireplace, heaps of burned cobble gravels, 650 flint artefacts and abundant faunal remains (red deer, horse, Bos or Bison, and wild boar) that are concentrated within two horn-like ribbons around the fireplace (Fig. 17). Technological study of the artefacts indicates that production of blade blanks was only a subordinate activity at the site, and did not induce a knapping heap (at least in the excavated area). By contrast, knapping debris are scattered over the whole area in between the fireplace and the bone accumulations. Raw material was mainly introduced into the site as already prepared cores and finished tools, as testified by the incomplete core reduction sequence. The high amount of scrapers among the retouched tools (79%), and that of points (among these backed pieces: 62%, and Malaurie points: 20%) suggest that this dwelling was primarily used for game treatment and repairing of hunting weapons. It was thus a specialized site that testifies to a short occupation with only in situ occasional knapping. Despite massive importation to the site of prepared blanks, the particle size composition of the whole assemblage remains close to the experimental data (Fig. 18A). However, when considering each square metre in isolation, large fluctuations of the fraction >5 mm become obvious. A high concentration of small-sized pieces (174 artefacts/m2) occurs in the vicinity of the fireplace (Fig. 17). This concentration, which comprises few large artefacts, is interpreted as an area where the backed pieces have been processed or resharpened, giving rise to abundant small fragments. Comparison between the squares that contain more than 20 artefacts shows that deviation from the mean can be large when the artefact number is low (Fig. 18B). This suggests that: (1) Intra-site variability implies multiplication of the number of samples for obtaining a representative size composition of the assemblage. (2) The experimental particle size curves can be used as a basis for a large panel of sites, even if blank production and/or preform shaping account for only part of the activities carried out at the site. Similarly, the experimental data remain valid when parts of the core reduction sequence are lacking (e.g. first steps of lithic reduction).

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

A

concentration 2

XIII

lithics burnt pieces fauna pebbles

XIV

3161

C 1

experimental debitages Port-de-Penne, full assemblage Port-de-Penne, square meters (N > 20)

0

XV 0,2

0,8

XVI

concentration 1

0,4

0,6

b

XVII fireplace

XVIII

0,6

0,4

g

0,2

1m

XIX

f

c 41

40

39

38

37

36

35

34

33

32

31

0 1

F cores points blades, bladelets

XIII

h

d e

0,8

a 1

4

0,8

0,6

0,4

0

0,2

M

concentration 2

XIV

B

XV

80

concentration 1 XVI

90 70

2-4 mm (%)

fireplace

XVII XVIII

60 50 40 30 20 10

XIX

0 41

40

39

38

37

36

35

34

33

32

31

0

50

100

150

200

250

artefact number per square meter XIII artefacts < 5 mm

XV

Fig. 18. Particle size distribution of the lithic artefacts, Port-de-Penne, level 2. The letters a to h refer to the squares labelled on Fig. 17.

concentration 2

>1 > 10 > 25 > 50 > 100

XIV

f g

c

a

d

b

fireplace

h

concentration 1

compared, and convergence allows a decision whether the hypothesis of artefact redistribution by sedimentary processes can be accepted or rejected.

e fau

na

XVI

fauna

XVII XVIII XIX

Port-de-Penne, level 2 41

40

39

38

37

36

35

34

33

32

31

Fig. 17. Artefact distribution maps, Port-de-Penne, level 2 (from Detrain et al., 1996).

Minimizing the uncertainties relative to the original particle size composition of the assemblage in a site needs to be done with care, first in the general taphonomic approach of the site, second in the choice of robust criteria for particle size analysis. The taphonomic analysis of a site rests mainly on a “confrontation to the geoarchaeological model” as proposed by Colcutt et al. (1990) and Lenoble (2005). The main steps are: (1) Identification of the sedimentary mechanisms that were at work in the site by using a classical sedimentological approach. (2) Analyzing the archaeological patterns (i.e. spatial distribution and associations of artefacts, particle size composition, fabrics, .) and comparison to those we can expect if site formation was uniquely due to the recognized sedimentary processes. The results yielded by all the taphonomic analyses are then

Each type of data, and particularly particle size data, gives only partial and sometimes ambiguous results. Robust evaluation of the degree of site modification by sedimentary processes and drawing up of the best plausible scenario for site formation can only be achieved from a confrontation between the different approaches (Petraglia and Potts, 1994). Since the available database on the size composition of lithic assemblages does not cover all the variability that can be potentially found in archaeological sites, the following conditions for a particle size analysis must be fulfilled: (1) The smallest sieve mesh used must be 2 or 5 mm to consider a large size spectrum, which is less sensitive to anthropogenic factors. Sieving must be made carefully in order to avoid any bias in the particle size composition. (2) The number of artefacts in each sample have to be >N ¼ 100 to limit statistical bias. (3) The sample number must be large enough to collect an assemblage representative of the particle size composition of the whole site. This number can be small when no activity area can be identified from the excavation map. In that case, homogenization of the spatial distribution of the lithics can be assumed, because of either anthropogenic factors (trampling, relocation of the activities through time within the site and mixing of different core reduction/tool production events) or

3162

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

Hermies - le Tio Marché level R/S 79.0

U

sieved square meter core levallois flake 78.3 contour line

T

1m

S

concentrations of levallois flakes

R

78.3

78.4

79.0

.8

.7

78.5

78.6

78

78

Q

78.9

knapping areas

P knapping area 78

.8

O

20

21

22

23

24

25

1

2

3

4

5

6

Fig. 19. Artefact distribution map and location of the sieved squares, Hermies-Le Tio Marché, level R/S (from Vallin and Masson, 2005). The contour interval is 5 cm (elevation in m a.s.l.). d > 10 mm (w > 14 mm)

A

C

post-abandonment sedimentary processes. Such a hypothesis can be considered as valid only if all samples yield a similar particle size composition.

archaeological sites (dominant debitage type) laminar discoid levallois experimental debitages

0

1

0.2

0.8

16e 18

0.4

0.6 16d 16b

6. Archaeological case studies

13 16c

0.4

6.1. Hermies e Le Tio Marché

17

0.2

Hermies e Le Tio Marché is located on the flanks of a small palaeovalley incised in a loessic plateau at Hermies, northern France (Vallin and Masson, 2005) (Fig. 16). The site yielded a Mousterian industry typified by the quasi-exclusive production of large levallois flakes. The main archaeological level lies within the Saint-Acheul soil (MIS 3) and shows well-delineated lithic concentrations, which correspond to knapping heaps and clusters of levallois flakes that served for butchery activities, separated by low-density or empty areas (Fig. 19). Particle size analysis indicates that the amount of small artefacts (d ¼ 2e4 mm) is similar for all the test squares located at the valley bottom, in knapping heaps, clusters of levallois flakes as well as low-density areas (Fig. 20). The mean reaches 84.5%, i.e. significantly more than for experimental data, and suggests enrichment of the archaeological level in small pieces. By contrast, particle size composition of squares located at the valley border does not show similar enrichment but plot within the experimental range. It has been previously suggested that secondary fragmentation of the artefacts occurred due to trampling (Vallin et al., 2005), since lithic comminution have been observed where pieces are at contact. However, generalized excess of smallsized lithics whatever the artefact clustering does not support the hypothesis that trampling was the main factor involved. Microscopic examination of small artefacts shows that they do not display homogenous characteristics. Weathering ranges from null (i.e. fresh pieces) to strong patination with dissolution pits. This suggests partial redistribution of ‘old’ material located upslope towards the valley bottom and mixing with in situ (younger)

0.6

16a

0

F

1

14

11 10 5 2 1 38 6 47

15

9

0.8

0.8 14

14

12

1 0.6

0.4

0

0.2

M

10 mm > d > 4 mm (14 mm > w > 5.7 mm)

4 mm > d > 2 mm (5.7 mm > w > 2.8 mm) d > 20 mm (w > 28 mm)

C

B

0

Bichou B Romentères (whole assemblage) Bel Soleil 0.2 0.8 archaeological sites Auriac Hermies experimental debitages Lanne-Darré 0.4 0.6 1

0.6

0.4

Montaut

Hermies (mean) 0.2 Romentères square F

F

0.8

Les Bosses Port-de-Penne 0 1

10 mm > d > 5 mm (14 mm > w > 7 mm)

0.8

0.6

Moulin Neuf 2

0.4

0.2

1 0

M

20 mm > d > 10 mm (28 mm > w > 14 mm)

Fig. 20. Compared particle size distribution of experimental Palaeolithic debitages and assemblages. The main debitage type for each site is indicated by a symbol. A e Particles >2 mm, B e Particles >5 mm. Sites: 1 to 12 e Hermies,13 e Les Bosses,14 e Combe Menue, 15 e Moulin Neuf, 16 e Le Casseux (a: upper level, e: lower level), 17 e Port-de-Penne.

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

3163

90 Auriac-Duclos Romentères, square F Romentères, whole assemblage Exp. Discoïde (N=9) Exp. Levallois (N=18)

80 70

percentages

60

max standard deviation + mean standard deviation min

50 40 30 20 10 0 5-10 mm

10-20 mm

20-31.5 mm

31.5-50 mm

> 50 mm

sieve mesh Fig. 21. Particle size distribution of archaeological assemblages (Auriac-Duclos, Romentères) and experimental debitages (discoid, levallois).

artefacts. Redistribution by water was minimal and affected mainly fine-grained pieces (the composition of the fraction >5 mm is similar to the experimental one for most of the test squares), possibly because of rapid burial under loess deposits, leading to good preservation of the anthropogenic patterning of the site. 6.2. Le Casseux Le Casseux site is located at Mareuil-sur-Cher, centre of France (Fig. 16) in the floodplain of a small river, and delivered Recent Mesolithic to Upper Palaeolithic industries (Fourloubey, 2004). The lithic material does not form well-defined archaeological levels but is scattered in sandy loams overlying alluvial gravels. Typotechnological study of the artefacts points to mixing in variable amounts of Mesolithic and Magdalenian material in the whole sequence. Mechanical weathering (polish, edge fragmentation) together with the bimodal fabric (i.e. two preferred orientations at 90 ) argues for redistribution of the artefacts by water flows. As expected, particle size analysis made on test squares clearly shows low percentages of small to medium-sized pieces that do not match with the experimental data. Significant fluctuations in the composition occur vertically. The deepest sample (16e, Fig. 20A) exhibits strong impoverishment and is mainly composed of

Fig. 22. Distribution map of artefact size classes, Romentères.

artefacts >10 mm, showing that artefacts provided by site erosion due to bank slumping were first redistributed by powerful currents during floods, whereas the top sample (16a) is closer to the experimental distribution. Such variations are interpreted as the result of decreasing transport capacity of the flows as

Fig. 23. Vertical projections of artefact and corresponding size classes, Romentères.

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

heavy tools (bifaces + cleavers + choppers) (%)

3164

20 18

Lower Palaeolithic Orgnac 3 Middle Palaeolithic

Septsos

16 14 12

Auriac-Duclos 10

Lanne-Daré 8

Cantalouette 1-IVb Cazalège Bichou B Romentères Petit-Bost 0 Les Bosses Bel Soleil St-Anne

6 4 2 0 0

5 Soucy 1 Grotte Noisetier

10

15

20

25

retouched flakes (%)

Bichou B (cores=26%) 18

Auriac-Duclos 16

Cazalège

Romentères 14

Septsos

cores (%)

12

maximal value of core to total experimental assemblage (discoid debitage) calculated for fraction:

Les Bosses

Cantalouette 1-IVb

10 8

Bel Soleil 6

Petit-Bost 0

4 2

d > 20 mm Lanne-Daré

St-Anne

d > 10 mm

Grotte Noisetier

d > 5 mm

Soucy 1

0 0

5

10

15

20

25

retouched flakes (%) Fig. 24. Compared ratios of heavy tools to retouched flakes (A) and cores to retouched flakes (B) for French Lower and Middle Palaeolithic sites. Percentages were calculated on the whole assemblages, excluding the unmodified or only tested blocks and the hammers from the sum. The size of the assemblages ranges between 358 and 27765 pieces. Bel Soleil (Bruxelles et al., 2008), Bichou série B (Jarry, unpublished data), Les Bosses (Jarry et al., 2007), Cantalouette 1 (Brenet et al., 2006), Cantalouette 4 (level 2, Blazer et al., 2006), Cazalège (Millet et al., 1999), Combebrune 3 (Folgado et al., 2006), Etoutteville (Delagnes and Kuntzmann, in Delagnes and Ropars, 1996), Lanne-Daré (Collonge and Texier, 2005), Lailly-Le Domaine de Beauregard (level B, Locht and Ferdouel, in Deloze et al., 1994), Lailly-Le Fond de la Tournerie (level A, Depaepe et al., in Deloze et al., 1994), Molinons-Le Grand Chanteloup (level A, Locht et al., in Deloze et al., 1994), Orgnac 3 (Moncel, 1989), Petit-Bost (levels 0 and 1, Bourguignon et al., 2006), Le Pucheuil (assemblages B and C, Ropars et al., in Delagnes and Ropars, 1996), Saint-Anne (Raynal et al., 2007), Septsos (Fourloubey and Colonge, unpublished data), Soucy 1 (Lhomme et al., 2001), Villeneuve-l’Archevêque (level B, Locht et al., in Deloze et al., 1994), Vineuf-Les Hauts Massars (level 1, J.M. Guedo, in Deloze et al., 1994).

a consequence of floodplain accretion and correlative lowered flood influence.

6.3. Auriac and Romentères Auriac-Duclos and Romentères are two Palaeolithic sites located in colluviated loess on a plateau near Aire-sur-l’Adour, southwest France (Fig. 16). They have yielded industries dated by TL and OSL to the penultimate interglacial (MIS 7) (M. Hernandez, unpublished data) and typified by dominantly discoid production, lack (AuriacDuclos) or subdued (Romentères) levallois production, biface and cleaver production, and overwhelming use of quartzite pebble gravel as raw material (Turq et al., 2010; Colonge et al., unpublished report). Auriac-Duclos industry is ascribed to Upper Acheulean whereas Romentères is considered as Old Mousterian.

Because of the colluvial setting, a taphonomic study has been made to test whether the assemblages were significantly affected by overland flow. Three squares (ca. 2 m2 each) were water sieved at d ¼ 5 mm in both sites. Results of the particle size analysis are shown on Fig. 21. Both assemblages exhibit a size composition that fits a normal distribution, with a mode at d ¼ 31.5e50 mm. Comparison with experimental data (here the mean of 9 discoid productions) indicates strong sorting. Impoverishment in fine to medium-grained artefacts took place and the ratio of the fraction <20 mm to that between 20 and 50 mm is about 15e20 times lower than that of experimental productions. Such an impoverishment particularly affected flint material because of the small size of raw blocks (and therefore, that of artefacts). At Romentères, the size distribution is not homogeneous over the whole surface of the excavation. Artefact maps and projections show an overall downslope fining trend together with localized selective accumulations

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

of either coarse-grained or fine-grained pieces (Figs. 22 and 23). Sieving of a fine-grained accumulation indicates overrepresentation of the fraction 5e10 mm and an almost total lack of coarse pieces (Fig. 21, square F). About 80% of the >20 mm expected pieces are lacking with respect to the experimental distribution. Since pieces belonging to the all steps of the chaîne opératoire have been recovered and testifies to in situ knapping, and since waste material were obviously not subject to importation or exportation by humans, the best hypothesis that can explain sorting is the impact of overland flow. Plotting the data in Lenoble’s (2005) experimental diagram (Fig. 20B) suggests that AuriacDuclos and Romentères correspond to residual sites where a significant part of the artefacts
3165

categories), may remain uncertain, particularly in case of underrepresentation of coarse-grained pieces in the assemblage. Significant progress can be expected from the study of well-preserved archaeological cases where the spatial variability of particle size due to areas where specific activities took place can be precisely documented. At the moment, robust assumption for assemblage modification by flows can be derived from particle size analysis coupled with other geoarchaeological/taphonomical investigations (e.g. piece abrasion, sedimentary features, fabrics, refitting, spatial distribution and associations of artefacts, see for example Petraglia and Potts, 1994). Evidence for lithic redistribution may imply that care should be taken in archaeological analysis of the site since sorting may impact significantly the initial techno-typological balance of any lithic assemblage.

Acknowledgements We acknowledge the knappers that contributed to the experimental database, B. Aubry, L. Bourguignon, M. Brenet, A. Delagnes, P. Fouéré, J. Pelegrin and V. Mourre. D. Colonge, L. Detrain, C. Fourloubey, M. Jarry, L.A. Lelouvier, B. Masson and L. Vallin are also thanked for providing unpublished data from their excavations. M.J. Shott and an anonymous reviewer are warmly acknowledged for their helpful comments.

References Behm, J.A., 1983. Flake concentrations: distinguishing between flintworking activity areas and secondary deposits. Lithic Technology 12, 9e16. Bertran, P., Texier, J.P., 1994. Structure sédimentaire d’un cône de flot de débris (Vars, Alpes françaises méridionales). Permafrost and Periglacial Processes 5, 155e170. Bertran, P., Texier, J.P., 1995. Fabric analysis: application to Palaeolithic sites. Journal of Archaeological Science 22, 521e535. Bertran, P., Claud, E., Detrain, L., Lenoble, A., Masson, B., Vallin, L., 2006. Composition granulométrique des assemblages lithiques. Application à l’étude taphonomique des sites paléolithiques. Paleo 18, 7e36. Bordes, J.G., 2003. Lithic taphonomy of the Châtelperronian/Aurignacian interstratifications in Roc de Combe and Le Piage (Lot, France). In: The chronology of the Aurignacian and transitional technocomplexes. Dating, stratigraphies, cultural implications. Trabalhos de Arqueologia, vol. 33. Instituto Português de Arqueologia, Lisboa, 223e244 pp. Brown, C.T., 2001. The fractal dimensions of lithic reduction. Journal of Archaeological Science 28, 619e631. Campebll, I.A., 1989. Badlands and badland gullies. In: Thomas, D.S.E. (Ed.), Arid Zone Geomorphology. Belhaven Press, London, pp. 159e183. Colcutt, S.N., Barton, N.R.E., Bergman, C.A., 1990. Reffiting in context: a taphonomic case study from a late Upper Palaeolithic site in sands on Hengistburry Head, Dorset, Great Britain. In: Ciesla, E., Eickhoff, S., Arts, N., Winter, D. (Eds.), The Big Puzzle, pp. 219e235 (Bonn). Coussot, P., Meunier, M., 1996. Recognition, classification and mechanical description of debris flows. Earth Science Reviews 40, 209e227. Desrosiers, P.M., 2009. A l’origine du Dorsétien. Apport de la technologie lithique des sites GhGk-63 et Tayara (KbFk-7) au Nunavik. Unpublished PhD thesis, University Paris 1, Panthéon-Sorbonne, Paris. Desrosiers, P.M., Gendron, D., Todisco, D., Monchot, H., Rahmani, N., Bhiry, N., Houmard, C., 2008. Tayara (KbFk-7) et le Dorsétien: Recherche pluridisciplinaire sur un site-clé du Paléoesquimau du détroit d’Hudson (Nunavik, Canada). L’Anthropologie 112, 757e779. Detrain, L., Chauvière, F.X., Diot, M.F., Kervazo, B., Martin, H., O’Yl, W., 1996. Camping du Saut, Port-de-Penne, Penne d’Agenais, Lot-et-Garonne. Document Final de Synthèse, AFAN Grand-Sud-Ouest, Pessac. Fladmark, K.R., 1982. Microdebitage analysis: initial considerations. Journal of Archaeological Science 9, 205e220. Fourloubey, C., 2004. Les occupations magdalénienne et néolithique du “Casseux” à Mareuil-sur-Cher (Loir-et-Cher). Site A85 n 42. Notice intermédiaire de diagnostic archéologique. Unpublished report, Service Régional de l’Archéologie, Orléans. Frostick, L.E., Reid, I., 1983. Taphonomic significance of sub-aerial transport of vertebrate fossils on steep semi-arid slopes. Lethaia 16, 157e164. Hansen, P.V., Madsen, B.O., 1983. Flint axe manufacture in the Neolithic. An experimental investigation of a flint axe manufacture at Hastrup Vaenget, East Zealand. Journal of Danish Archaeology 2, 43e59. Healan, D.M., 1995. Identifying lithic reduction loci with size-graded macrodebitage: a multivariate approach. American Antiquity 60 (4), 689e699.

3166

P. Bertran et al. / Journal of Archaeological Science 39 (2012) 3148e3166

Hjülström, F., 1939. In: Trash, P.D. (Ed.), Transportation of Detritus by Moving Water. Recent Marine Sediments. American Association of Petrology and Geology, pp. 5e31. Isaac, G.L., 1967. Towards the interpretation of occupation debris: some experiments and observations. Kroeber Anthropological Society Papers 37, 31e57. Le Grand, Y., 1994. Processus de formation des dépôts archéologiques du Pléistocène moyen de Lunel-Viel 1 (Hérault). L’utilisation des remontages. Préhistoire et Anthropologie Méditerranéenne 3, 57e63. Lenoble, A., 2005. Ruissellement et formation des sites préhistoriques: référentiel actualiste et exemples d’application au fossile. B.A.R. International Series. Oxford, n 1363. Lenoble, A., Bertran, P., 2003. Fabric of Palaeolithic levels: methods and implications for site formation processes. Journal of Archaeological Science 31, 457e469. Lenoble, A., Bertran, P., Lacrampe, F., 2008. Solifluction-induced modifications of archaeological levels: simulation based on experimental data from a modern periglacial slope and application to French palaeolithic sites. Journal of Archaeological Science 35, 99e110. McPherron, S.J.P., 2005. Artifact orientations and site formation processes from total station proveniences. Journal of Archaeological Science 32 (7), 1003e1014. Middleton, G.V., Southard, J.B., 1984. Mechanics of Sediments Movement. Lecture Notes for Short Course. n 3. Society of Economic Palaeontologists and Mineralogists, Rhode Island. Moss, A.J., Walker, P.H., 1978. Particle transport by continental water flows in relation to erosion, deposition, soils and human activities. Sedimentary Geology 20, 81e139. Patterson, L.W., 1990. Characteristics of bifacial-reduction flake size distribution. American Antiquity 55 (3), 550e558. Petraglia, M.D., Nash, D.T., 1987. The impact of fluvial processes on experimental sites. In: Nash, D., Petraglia, M. (Eds.), 1987. Natural Formation Processes and the Archaeological Record, vol. 352. B.A.R. International series, Oxford, pp. 108e130. Petraglia, M.D., Potts, R., 1994. Water flow and the formation of early Pleistocene artifact sites in Olduvai Gorge, Tanzania. Journal of Anthropological Archaeology 13, 228e254. Pigeot, N., 1984. Les derniers Magdaléniens d’Etiolles. Perspectives culturelles et paléohistoriques. Gallia Préhistoire XXXVIIe (supplément) (CNRS, Paris). Rasic, J.T., 2004. Debitage taphonomy. In: Hall, C.T., Larson, M.L. (Eds.), Aggregate Analysis in Chipped Stone. University of Utah Press, Salt Lake City, pp. 112e135. Reid, I., Frostick, L.E., 1985. Arid zone slopes and their archaeological materials. In: Pitty, A.E. (Ed.), Themes in Geomorphology. Croom Helm, Beckenham, pp.141e157.

Reineck, H.E., Singh, I.B., 1980. Depositional Sedimentary Environments, second ed. Springer-Verlag, Berlin. Schick, K.D., 1986. Stone Age Sites in the Making. In: Experiments in the Formation and Transformation of Archaeological Occurrences, vol. 319. B.A.R. International Series. Schick, K.D., 1987. Experimentally-derived criteria for assessing hydrological disturbance of archaeological sites. In: Nash, D.T., Petraglia, M.D. (Eds.), 1987. Natural Formation Processes and the Archaeological Record, vol. 352. B.A.R. International Series, pp. 86e107. Schiffer, M.B., 1987. Formation Processes of the Archaeological Record. University of New Mexico Press, Albuquerque, NM. Shott, M.J., 1994. Size and form in the analysis of flake debris: review and recent approaches. Journal of Archaeological Method and Theory 1 (1), 69e110. Stahle, D.W., Dunn, J.E., 1982. An analysis and application of the size distribution of waste flakes from the manufacture of bifacial stone tools. World Archaeology 14 (1), 84e97. Todisco, D., Bhiry, N., Desrosiers, P.M., 2009. Paleoeskimo site taphonomy: an assessment of the integrity of the Tayara site, Qikirtaq Island (Nunavik, Canada). Geoarchaeology: An International Journal 24 (6), 743e791. Turcotte, D.L., 1986. Fractals and fragmentation. Journal of Geophysical Research 91, 1921e1926. Turq, A., Brenet, M., Colonge, D., Jarry, M., Lelouvier, L.A., O’Farrell, M., Jaubert, J., 2010. The first human occupations in southwestern France: a revised summary twenty years after the Abbeville/Saint Riquier colloquium. Quaternary International 223e224, 383e398. Vallin, L., Masson, B., 2005. Le gisement moustérien d’Hermies e le Tio Marché. Rapport de synthèse sur les fouilles programmées 1997e2003. Service Régional de l’Archéologie du Nord-Pas-de-Calais, Villeneuve d’Ascq. Vallin, L., Antoine, P., Aubry, B., Caspar, J.P., Cliquet, D., Delagnes, A., Depaepe, P., Guillemet, G., Laignel, B., Locht, J.L., Masson, B., Ozouf, J.C., Swinnen, C., 2005. Taphonomie des assemblages lithiques du Paléolithique moyen en contexte périglaciaire: approches expérimentale et archéologique à partir de sites du nord-ouest européen. Rapport non publié, Service Régional de l’Archéologie du Nord-Pas-de-Calais, Villeneuve d’Ascq. Van Nest, J., 2002. The good earthworm: how natural processes preserve upland Archaic archaeological sites of Western Illinois, U.S.A. Geoarchaeology 17 (1), 53e90. Villa, P., 1982. Conjoinable pieces and site formation processes. American Antiquity 47 (2), 276e290.