Supporting Online Material for

www.sciencemag.org/cgi/content/full/330/6004/659/dc1 Supporting Online Material for Early Use of Pressure Flaking on Lithic Artifacts at Blombos Cave, South Africa Vincent Mourre, Paola Villa, * Christopher S. Henshilwood *To whom correspondence should be addressed. E mail: villap@colorado.edu This PDF file includes: Materials and Methods Figs. S1 to S7 Tables S1 and S2 References Published 29 October 2010, Science 330, 659 (2010) DOI: 10.1126/science.1195550

1 Supporting Online Material Early Use of Pressure Flaking on Lithic Artifacts at Blombos Cave, South Africa 1. Materials and Methods Materials. Our data base of Still Bay bifacial points and point fragments recovered at Blombos Cave consists in total of 127 silcrete specimens from layers CA to CF and equivalent layers and sub-layers, excavated between 1993 and 2010 by CSH and coworkers. Of these 51 relate to production phase 2b and 76 to phase 3. For each production phase the number of attributes observed is less than the total for the phase because specific attributes could not be observed on small fragments. The sample of flakes comes from quadrants G5d and E5b of layer CC. The experimental materials include six bifacial points made on unheated silcrete and eight made on heated silcrete, four of which are made with soft hammer only and four finished with soft hammer followed by pressure. CSH collected the silcrete samples and heat-treated some pieces in co-operation with K. Brown. VM performed the knapping experiments. Microscopic analysis Scars on archaeological and experimentally replicated points and flakes were examined and photographed with a motorized Leica M 125 equipped with a DFC 295 digital camera and a Leica Application Suite with Multifocus module. Scar dimensions were measured using the Interactive Measurement function of the Leica Application Suite software for scars less than 10 mm long. A digital caliper was used for the others. The dimensions were considered only when the scars were complete and not reduced by later removals. Identification of heat treatment by non-destructive methods Brown et al. (S1) used a Novo Curve Glossmeter that measure the reflectivity of an artifact surface. The instrument is passed over the surface in a continuous reading mode until a maximum gloss value, measured in gloss units, is achieved. The histograms of unheated and heated archaeological samples, compared to the experimental samples, show weak bimodalities. Postdepositional weathering and patination can interfere with the measurements. Gloss can occur on artifacts due to wind action at open air sites or by sand abrasion in sediments but it will normally occur on the whole surface of the artifacts or on the whole exposed face. We preferred a method commonly used to identify heat treatment on Paleoindian (S2-S4) and Solutrean points (S5-S6). The method is based on the visual identification of a contrast between matt, dull surfaces which were flaked prior to heating and glossy surfaces flaked after heating, on the same artifact. With this method natural fracture surfaces must be distinguished from flaked surfaces and the sequence of removals must be established since only post-heating removals provide clear evidence of a contrast with pre-heating surfaces.

2 Table S1 provides results for a large sample of points (159) retouched pieces (175) and flakes (703). Note that pieces completely flaked after heating would not show different surfaces, thus the number of heated artifacts is only a minimum figure. The advantage of the method, if applied to a rather large sample, is that it tells us at what phase of the knapping process heating occurred. At Blombos our results indicate that the heating was on blanks extracted from unheated cores but prior to the advanced flaking phases of the artifact. Heat treatment was thus a consistent phase in the knapping process. Tip angle and thickness The tip penetrating angle is the tip seen in plan view and measured in degrees (also called the front angle). This angle can be measured on complete and incomplete points as long as the distal part is preserved. We measured the tip angle using the caliper method based on measuring width at a fixed distance (1 cm) from the apex. The angle was then calculated using a trigonometric formula (S7). Tip thickness was also measured at 1 cm from the apex. Chi square and t-tests for Tables 1-2 in the text. Table 1. The chi square results (S8) of experimental flakes by pressure vs. experimental flakes by soft hammer (df = 1 in all cases) are: for prominent bulb: 166.15; p = 0.000 for bulb without a lip: 40.63; p = 0.000 for hackles on the bulb: 171.03; p = 0.000 These values indicate a very significant difference between the two samples. For regular ridges: 9.187; p = 0.002 For small but not punctiform platform: 5.208; p = 0.022 These values indicate a difference at the 0.01 level only for regular ridges. Table 2. t-tests from parameters (S8) Tip angle V-shaped tips vs. arched tips: df = 39, p < 0.001. V-shaped tips with straight edges vs. unifacial points: df = 102, p < 0.001. V-shaped tips vs. production phase 2b: df = 40, p < 0.001 Production phase 3 arched tips vs. unifacial points: df = 101, p < 0.001 Tip thickness Production phase 3 points vs. production phase 2b: df = 53, p = 0.002 Production phase 3 points vs. unifacial points df = 126, p < 0.001

3 Flintknapping and replicating experiments Replicating experiments on heated and unheated blocks of silcrete have been performed by VM who has 20 years of experience in flintknapping different raw materials such as quartz, quartzite, volcanic rocks, and silcrete. More than a flintknapper, VM is actually a skilled replicator. The two concepts should not be confused. A flintknapper is a person who can make stone artifacts using knapping tools available to prehistoric craftsmen and documented in the archaeological records (S9-S12). A replicator uses flintknapping to recreate consistently, with the same lithic materials, the same reduction technology and end products as the prehistoric craftsmen (S13). A modern flintknapper may have limited technological skills and a limited repertoire of techniques. He may be able to reproduce artifacts using certain techniques and if results match it may be concluded that those particular techniques were used in the past as well. However there are potentially many way of getting similar results from flintknapping. This is why a replicator must produce not only a good copy of the artifacts but must consistently replicate the manufacturing sequence and the technical features of the associated products. It is the entire manufacturing process, rather than the end product, that is important for archaeological research. Results of the replicative experiments must be compared technologically with the prehistoric materials to determine the validity of the experimental techniques for the identification of prehistoric reduction techniques (S13). Our experimental replication was preceded by a systematic analysis of the Blombos artifacts. We had previously established the point manufacturing sequence and the frequency distributions of technical attributes by phase (S14). The debitage, cores, and retouched pieces of the Still Bay assemblages have also been studied for this paper. The success in replicating the Blombos Still Bay points with the same techniques is documented by evidence of similarity between VM s experimental lithics and the Still Bay points and shaping flakes. This is demonstrated by similar morphologies (Fig. 2, Figs, S5-S7) and supported by statistics on attributes and dimensions of flakes in the experimental and archaeological samples (Table 1 with associated chi-square tests and Table S2 with associated t-tests).

4 2. Figures Fig. S1. Stratigraphy of Blombos Cave showing the Still Bay levels. The principal levels for the Still Bay are CA, CB, CC, CD and CF. Subdivisions of these levels are denoted by the addition of a letter in alphabetical order e.g. CAA, CAB for level CA and CBA, CBB for level CB etc.

5 Phase 1 Initial shaping (N = 51) Phase 2 (N = 57 + 57 + 56 phase 2 indet. = 170) phase 2a Advanced shaping (N = 57) phase 2b Advanced shaping (N = 57) Phase 3 Phase 4 Finished product (N = 107) Recycled, modified (N = 17) Product Blombos points Percussion technique Internal percussion (hard hammer) Marginal percussion (soft hammer) Marginal percussion (soft hammer) Marginal percussion (soft hammer) and pressure Internal percussion (hard hammer) Fig. S2. Phases of manufacture and knapping techniques of the Blombos Still Bay bifacial points (modified after S14). About 80% of the points are broken. Excluding 9 broken points with distal impacts or haft fractures, most fractures are production failures. Direct percussion (by hard then soft hammer) was employed in the early production phases of reduction while pressure flaking was used in the finishing stage. In our previous work we had identified only the use of direct marginal percussion by soft hammer in the last phase (see Text). Scale = 1 cm.

Fig. S3. Examples of heat-treated artifacts from the Still Bay levels at Blombos. The red arrows indicate a surface flaked prior to heating; the white arrow indicates cortex. Note the smooth and compact surface of removals done after heating. (A) Denticulate on flake, layer CC no. 66, the arrow is on the ventral surface. (B, D) Retouched flake, layer CAB no. 53. The red arrow rests on the ventral surface, hence heating was done on this small flake, not on the core. (C, E) Two views of a bifacial fragment at the initial shaping phase, PVN 183, layer CC. (F) Bifacial fragment in the initial shaping phase (no. 54, layer CAA). Scale of A-C, E-F = 1 cm; scale of D = 1 mm. 6

Fig. S4. Pressure-flaked Paleoindian points and Solutrean shouldered points. (A, C) from the Jurgens site in Colorado (9070 ± 90 B.P. S15). (B) from the Olsen-Chubbuck site in Colorado (10,150 ± 500 B.P. S16). (D) Upper Solutrean points from the site Fourneau du Diable (Dordogne). (E, F) Upper Solutrean point from the site of le Placard (Charente, c. 18 ka). Scale = 1 cm. A-C photos P. Villa; D courtesy of A. Morala, photo Ph. Jugie, Musée National de Préhistoire, Les Eyzies. E, F modified after S17. 7

8 Fig. S5. Replicates of Still Bay bifacial points made by soft hammer (B-C) compared to Blombos bifacial points of production phase 2b (A, D) and diagnostic attributes of soft hammer scars on the experimental replicates (E-H). (A, D) Points PVN 64 (coarse silcrete, tip is broken) and PVN 47 (matricial silcrete) Blombos, layers CA, CC. The white arrow indicates a very large scar clearly done by soft hammer.(e-h) show large scars with irregular ridges, often expanding distally. Scale of A-D = 1 cm; scale of E-H = 1 mm.

Fig. S6. Experimental pressure flakes and pressure flakes from the Blombos Still Bay levels showing the attributes which are the counterparts of those present on the points. A,C,E and G experimental flakes; B,D,F, H flakes from Blombos, layer CC. Note the prominent bulb on A-C and F, the small but not punctiform platform on all flakes, the absence of a lip (typical of soft hammer percussion) and the hackles on the bulb indicated by the white arrows. Scale = 1 mm. 9

Fig. S7. Experimental soft hammer flakes (A, C, E, G) and flakes from Blombos (B, D, F, H, layer CC) showing the attributes diagnostic of soft hammer percussion. Note the lipped bulb (A, C-H), the diffuse (A-G) or absent (H) bulb of percussion and the irregular ridges, some expanding distally (A, C, D). Scale = 1mm. 10

11 3. Tables Table S1. Frequencies of artifacts with different surfaces indicative of heat treatment. The number of heated artifacts should be considered a minimum figure, since pieces completely flaked after heating would not show different surfaces. The frequencies are higher for retouched pieces because they are not intensively flaked. Among the Still Bay bifacial points with different surfaces, three correspond to initial shaping (production phase 1) and four are finished products (production phase 3). The analyzed sample of Still Bay points does not include fragments < 3 cm because they are too small for a visual diagnosis. The retouched pieces include a few unifacial points and bifacial points of non- Still Bay morphology. The flake sample comes from 10 quadrants of layer CC. There are only 16 silcrete cores and core fragments in layers CA to CF and none shows different surfaces. This suggests that flakes and blanks extracted from larger cores were heated but not the cores. The heated pieces vary in size between 6 and 3 cm; even small flakes (Fig. S3: B) were heated prior to further retouch. Silcrete artifacts only Analyzed sample With different surfaces Still Bay points and fragments 159 7 4.4 Retouched pieces Layer CA 42 5 11.9 Layer CB 22 2 9.1 Layer CC 53 5 9.4 Layer CD-CE 47 7 14.9 Layer CF 11 2 18.2 Total 175 21 12.0 Flakes Layer CC, 10 quadrants 703 6 0.9 %

12 Table S2. Scar width and length Width of scars in mm on experimental points made by pressure, by soft hammer and on Blombos points of production phase 2b and 3. Mean SD Min Max N Experimental soft hammer 7.59 5.28 1.26 23.95 26 Blombos production phase 2b points 7.84 3.86 2.9 21.31 29 Experimental pressure 3.76 1.23 1.79 5.86 27 Blombos production phase 3 points 3.83 1.96 1.07 8.66 31 Length of scars in mm on experimental points made by pressure, by soft hammer and on Blombos points of production phase 2b and 3. Mean SD Min Max N Experimental soft hammer 10.61 6.91 1.21 26 26 Blombos production phase 2b points 8.38 4.73 2.45 23.45 29 Experimental pressure 6.26 2.36 2.84 12.99 27 Blombos production phase 3 points 4.91 2.46 1.42 10.45 31 t-tests for width of scars (from parameters; S8) experimental soft hammer vs. Blombos points production phase 2b: df = 53, p = 0.841 (not significant at the 0.01 level) experimental pressure vs. Blombos points production phase 3: df = 56, p = 0.874 (not significant at the 0.01 level) experimental pressure vs. experimental soft hammer: df = 51, p < 0.001 (very significant) Blombos points production phase 3 vs. phase 2b: df = 58, p < 0.001 (very significant) t-tests for length of scars (from parameters; S8) experimental soft hammer vs. Blombos points production phase 2b: df = 53, p = 0.165 (not significant at the 0.01 level) experimental pressure vs. Blombos points production phase 3: df = 56, p = 0.038 (not significant at the 0.01 level) experimental pressure vs. experimental soft hammer: df = 51, p = 0.003 (very significant). Blombos points production phase 3 vs. phase 2b: df = 58, p < 0.001 (very significant).

4. References S1. K. S. Brown et al., Science 325, 859 (2009). S2. D.E. Crabtree, B.R. Butler, Tebiwa 7, 1 (1964). S3. H.G. Nami, in Folsom Lithic Technology. Explorations in Structure and Variation, D. S. Amick, Ed. (International Monograph in Prehistory, Ann Arbor, 1999) pp. 82-97. S4. M.J. Root, in J.E. Clark, M.B. Collins, Eds.(Special Publication no. 4 Lithic Technology, University of Tulsa, Oklahoma, 2002), pp. 315-331. S5. M.-L. Inizan, J. Tixier, Paléorient 26/2, 23 (2001). S6. M.-L. Inizan, M. Reduron-Ballinger, H. Roche, J. Tixier, Technology and terminology of flaked stone (CREP, Nanterre, 1999). S7. H.L. Dibble, M.C. Bernard, American Antiquity 45, 857 (1980) S8. Ø. Hammer, D. A. T. Harper, P.D. Ryan, PAST: Paleontological Statistics Software Package, version 2.01, 2001) available online at folk.uio.no/ohammer/past/ S9. F. Bordes, in Premier colloque international sur l industrie de l os dans la préhistoire, H. Camps-Fabrer, Ed. (Univ. of Provence, 1974) pp. 97-100. S10. F.F. Wenban-Smith, in Boxgrove. A Middle Pleistocene hominid site at Eartham Quarry, Boxgrove, West Sussex, M.B. Roberts, S.A. Parfitt, Eds.(English Heritage, London, 1999) pp. 384-394. S11. M. Soressi et al., in Les sociétés du Paléolithique dans un Grand Sud-Ouest de la France: nouveaux gisements, nouveaux résultats, nouvelles méthodes, J Jaubert,, J.G., Bordes, I. Ortega, Eds. (Mém. XLVII Soc. Préhist. Fr, Paris, 2008) pp. 95-132. S12. A. I. Yevtushenko, in The Middle Paleolithic of Western Crimea-vol. 1, A.E. Marks, V. P. Chabai, Eds. (ERAUL 84, Liege, 1999) p. 287-322. S13. Flenniken, J. J., Ann. Rev. Anthropol. 13, 187-203 (1984). S14. P. Villa, M. Soressi, C. S. Henshilwood, V. Mourre, J. Archaeol. Sci. 36, 441 (2009). S15. J. B. Wheat, The Jurgens site (Plains Anthropologist Memoir 15, Plains Anthropological Society, Lincoln, Nebraska, 1979). S16. J. B. Wheat, The Olsen-Chubbuck site. A Paleoindian bison kill (Society for American Archaeology Memoir 26, Washington, D.C. 1972). S17. H. Plisson, J. M.Geneste, Paléo 1, 65 (1989). 13