Microstratigraphic evidence of in situ fire in theAcheulean strata of Wonderwerk Cave, NorthernCape province, South AfricaFrancesco Bernaa,1, Paul Goldberga,b, Liora Kolska Horwitzc, James Brinkd,e, Sharon Holtd, Marion Bamfordf,and Michael Chazang

aDepartment of Archaeology, Boston University, Boston, MA 02215; bRole of Culture in Early Expansions of Humans, Heidelberg Academy of Scienceand Humanities, 72070 Tübingen, Germany; cNatural History Collections, Faculty of Life Sciences, Hebrew University, Jerusalem 91904, Israel; dFlorisbadQuaternary Research Department, National Museum, Bloemfontein 9300, South Africa; eCentre for Environmental Management, Bloemfontein 9300, SouthAfrica; fBernard Price Institute for Palaeontological Research, University of the Witwatersrand, Johannesburg 2050, South Africa; and gDepartment ofAnthropology, University of Toronto, Toronto, ON, Canada M5S 2S2

Edited by Donald K. Grayson, University of Washington, Seattle, WA, and approved February 24, 2012 (received for review October 25, 2011)

The ability to control fire was a crucial turning point in humanevolution, but the question when hominins first developed thisability still remains. Here we show that micromorphological andFourier transform infrared microspectroscopy (mFTIR) analyses ofintact sediments at the site of Wonderwerk Cave, Northern Capeprovince, South Africa, provide unambiguous evidence—in theform of burned bone and ashed plant remains—that burning tookplace in the cave during the early Acheulean occupation, approx-imately 1.0 Ma. To the best of our knowledge, this is the earliestsecure evidence for burning in an archaeological context.

micromorphology | cooking hypothesis | Homo erectus

The ability to control fire was a crucial turning point in humanevolution, but there is no consensus as to when hominins first

developed this ability. According to Richard Wrangham’s “cook-ing hypothesis,” Homo erectus was adapted to a diet of cookedfood and therefore was capable of controlling fire (1). Recentphylogenetic studies on nonhuman and human primates basedon associated trends in body mass, feeding time, and molar sizesupport the hypothesis of the adoption of a cooked diet at leastas early as the first appearance of H. erectus approximately 1.9Ma (2). However, to date, the evidence for controlled use of firein association with H. erectus is scant and inconclusive, aspointed out in a recent review of the archaeological record byRoebroeks and Villa (3). Unequivocal evidence for the habitualuse of fire in early hominin sites, such as that reported for QesemCave (4), is so far found in sites dated after 0.4 Ma, thus associ-ating the earliest control of fire primarily with earlyHomo sapiensand Neanderthals (3).Through the application of micromorphological analysis and

Fourier transform infrared microspectroscopy (mFTIR) of intactsediments and examination of associated archaeological finds—fauna, lithics, and macrobotanical remains—we provide un-ambiguous evidence in the form of burned bone and ashed plantremains that burning events took place in Wonderwerk Caveduring the early Acheulean occupation, approximately 1.0 Ma.To date, to the best of our knowledge, this is the earliest secureevidence for burning in an archaeological context.

Previous Research on Early Evidence of FireClaims for early traces of fire have been made for sites in Africa,Asia, and Europe (5). In East Africa, sites that have producedevidence of fire include Gadeb 8E, Koobi Fora FxJj 20 East, andChesowanja GnJi 1/6E. At Chesowanja, dated to more than 1.42±0.07 Ma based on K-Ar dating of an overlying basalt, 40 pieces ofdiscolored clay aggregates were found intermingled with De-veloped Oldowan lithics and fauna (6). Magnetic susceptibilityanalysis of the rubefied clay aggregates indicates that they wereburned. At Gadeb 8e, magnetic properties of cobbles of weldedtuff indicate that they were also burned (7), and at Koobi Fora FxJj20—dated to 1.5 Ma—similarly discolored sediment patches were

identified as having been burned based on thermoluminescenceproperties (8, 9). Comparable analyses have been made on sites inthe Middle Awash (10). At Swartkrans (South Africa), burnedbones were identified from member 3, dated to ca. 1.0 to 1.5 Ma,based on histological characteristics and chemical identification ofchar (11–13). However, at Swartkrans, the burned bones appear tobe in secondary context in the fill of a gully (11). Some of the mostintensive research on early use of fire has focused on the site ofGesher Benot Ya’akov in the Jordan Valley (Israel), dated tobetween 0.7 and 0.8 Ma, where pot-lid fractures, characteristicrounded concave scars produced by heat-induced removal ofplanoconvex flakes, have been used to identify burned micro-debitage (14). Thermoluminescence analysis supports the identi-fication of burned microdebitage, and its spatial distribution,together with the presence of charred wood, seeds, and grains ledto the identification of “phantom hearths” (5, 15, 16). Neverthe-less, the evidence and acceptance for controlled use of fire at anyof the Acheulean sites noted earlier remains controversial. Thecontroversies stem from the fact that these are open-air sites and itis not possible to completely exclude the action of wildfires (3).Moreover, in none of the Acheulean contexts reviewed earlier hasresearch included microstratigraphic analysis of the deposits thatencase the burned objects. There is no evidence of, nor wereattempts made to look for, calcareous wood ash (i.e., ashed planttissues and oxalate pseudomorphs) as reported inQesemCave (4).Interestingly, at Zhoukoudian in China, microstratigraphic

analysis demonstrated that features as old as 0.6 Ma originallyconsidered evidence of in situ combustion (e.g., layer 10) or woodash residues (e.g., layer 4) are actually the result of water-de-posited organic-rich sediment and colluvial reworking of loess,respectively (17). Although Fourier transform infrared spectros-copy (FTIR) analysis supports the presence of burned bones as-sociated with burned flint at Zhoukoudian (18), these remains arenot directly associated with in situ anthropogenic combustionfeatures. Thus, any reasonable statement about their unambiguousassociation to hominin behavior remains inconclusive (17, 18).The use of high-resolution microscopic analysis of intact sedi-

ments has been used extensively in the Middle Stone Age of Africaand theMiddle Paleolithic of theMiddle East and Europe (cited in

Author contributions: F.B., P.G., L.K.H., and M.C. designed research; F.B., P.G., L.K.H., J.B.,S.H., M.B., and M.C. performed research; F.B. contributed new analytic tools; F.B., P.G.,L.K.H., J.B., S.H., M.B., and M.C. analyzed data; and F.B., P.G., L.K.H., and M.C. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: fberna@bu.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1117620109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1117620109 PNAS Early Edition | 1 of 6

ANTH

ROPO

LOGY

PNASPL

US

refs. 19, 20). Numerous microscopic studies of combustion featuresclearly reveal information not readily obtainable with other ana-lytical techniques, including fuel composition, internal organizationof the feature, and combustion conditions (19, 20). Such studiesillustrate the range of ways that Neanderthals and early modernhumans were associated with fire in their occupation sites (21).Historically, with the exception ofZhoukoudian, research that usedmicrostratigraphic analyses of contextualized intact sediments havebeen absent in the study of early hominin use of fire.

Wonderwerk CaveWonderwerk Cave (Fig. 1) is an approximately 140-m-longphreatic tube that formed in Precambrian dolostones of theKuruman Hills (Northern Cape province, South Africa). Be-ginning in the 1970s and ending in the 1990s, extensive archae-ological excavations were carried out by P. B. Beaumont in sevendifferent areas within the cave (Fig. 1D). The longest EarlierStone Age (ESA) sequence, approximately 2 m deep, is found inexcavation 1, currently located approximately 30 m in from the

cave mouth, immediately behind a large, active stalagmite whichdeveloped during the past 35,000 y (22, 23).Since 2004, our team has renewed fieldwork, performed chro-

nometric dating. and reanalyzed the archaeological record ofthe site (24, 25). Details on the archaeological, lithological, andchronological stratigraphy of excavation 1 are given in SI Textand illustrated in Figs. S1–S4. The archaeological sequencebegins with a small tool industry attributed to the Oldowan inbasal stratum 12, which is overlain by an Acheulean sequence.This sequence shows developments from rare protobifaces(stratum 11) through bifaces with noninvasive retouch in stratum10 (Fig. 1 B and C), to highly refined biface production beginningin stratum 9 (detailed in SI Text and Fig. S2). The evidence forfire presented in this study comes from stratum 10. Beaumont’sexcavation of this stratum covered 48 square yards (Fig. S1), andyielded a low density of lithic artifacts comprising only sevenbifaces (Fig. 1 B and C and Fig. S3 A–D), 36 flakes, 15 cores, and23 slabs of banded ironstone showing flake removals that wereclassified as “modified slabs.”

Fig. 1. (A) Map showing the lo-cation of Wonderwerk Cave. (B–C)Handaxes characteristic of theAcheulean of stratum 10, excava-tion 1, Wonderwerk Cave. (D) Planof Wonderwerk Cave generated bylaser scanning shows the locationof excavation areas discussed inthis study (courtesy of H. Rüther,Zamani project).

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1117620109 Berna et al.

ResultsSite Formation and in Situ Evidence of Fire in ESA at Wonderwerk Cave.Microstratigraphic investigation combining the use of sedimentmicromorphology and mFTIR was conducted to investigate siteformation processes and human activities in excavation 1 (Fig. 2).Micromorphological analysis indicates that the lithological se-quence in excavation 1 began with an archaeologically sterilephreatic sediment. This is overlain by the earliest archaeologicaloccupation (stratum 12) characterized by low-energy water de-position of sand and fine gravel, probably caused by sheet flow(Fig. 2B). From the top of stratum 12, the depositional processesinvolved the accumulation of aeolian material composed ofrounded aggregates of silty clay (Fig. 2D) that formed in dryingponds outside the cave, as well as fine sand (Fig. 2E); in some ofthe strata, the aeolian aggregates appear reworked by gravity ortrampling. The aeolian deposition is interspersed with episodes ofcave roof/wall collapse and successive diagenesis of dolostone orflowstone that produced grayish-white phosphatized layers, richin dark oxide nodules and degraded rock fragments (Fig. 2C).These white layers were at first erroneously interpreted asremains of combustion features (23).Field and microscopic observations of thin sections revealed

that stratum 10 is composed of a complex sequence of lithologicalcentimeter-scale microstratigraphic units (i.e., microfacies) (26).Three of these are illustrated in the thin-section scans shown in

Fig. 3. Micromorphological analysis shows that the upper part ofthe basal microfacies 1 is a clearly defined surface that exhibitsabundant remains of ashed plants and minute bone fragments(Fig. 3). As a number of intact blocks of sediment encompassedthis part of the stratigraphic sequence, it was possible to docu-ment a few surfaces and the evidence for burning over 1 meteralong the section (Fig. S5). FTIR microspectroscopy applied di-rectly to thin sections made from these blocks shows that some ofthe bones lying on these surfaces had been heated to ca. 500 °C(Fig. 4 and Fig. S5). Significantly, the angularity of bone frag-ments and the exceptional state of preservation of the ashed plantmaterial (Fig. 3 C and D and Fig. S5) indicate that both compo-nents were not transported from a distance into the cave by wateror wind, but were combusted and accumulated locally. Moreover,micromorphology and mFTIR did not show evidence for remainsof guano and/or high-temperature phosphate mineral phases (i.e.,berlinite and hydroxylellestadite); these minerals characteristi-cally form during spontaneous combustion of bat guano—a rareevent but one documented inside caves (27).The ashed plant remains are situated in the middle of ar-

chaeological stratum 10, which shows a normal magnetic orien-tation and is bracketed between two cosmogenic burial ages of1.27 ± 0.19 Ma and 0.98 ± 0.19 Ma (Fig. S4 and Table S1) (25).The Normal event can therefore be assigned to the Jaramillosubchron (1.07–0.99 Ma), a time range that fits with current

Fig. 2. (A) Photograph of the east section in excavation 1 with boundaries between archaeological strata 12 and 9. Numbered boxes indicate location ofintact block sampled for microstratigraphic analysis. (Scale bar: 10 cm) (B) Representative micrograph of low-energy, water-bedded silt, sand, and 0.5-cm-thickgravel (lag) from stratum 12. (Scale bar: 1 mm.) (C) Micrograph of microfacies from white layer close to top of stratum 12 composed of diagenetically altereddolostone and flowstone, with nodules of montgomeryite (arrows). (D) Micrograph of reddish-brown, wind-blown silty clay aggregates that comprise severallithological units starting from the top of stratum 12. (Scale bar: 1 mm.) (E) Representative micrograph of wind-blown, fine sand mixed with millimeter-sizedbone fragments from tan lithological units in strata 11, 10, and 9. (Scale bar: 1 mm.)

Berna et al. PNAS Early Edition | 3 of 6

ANTH

ROPO

LOGY

PNASPL

US

understanding of the chronological position of the early Acheu-lean within the ESA in Southern Africa (28).

Macroscopic Evidence of Fire in Stratum 10 at Wonderwerk Cave. Ina recent publication, based on his field observations during ex-cavation, Beaumont (23) reported macroscopic evidence forburning in the excavation 1 ESA assemblage. Our investigationis based on the study of faunal, lithic, and macrobotanicalassemblages from Beaumont’s excavation in stratum 10, in con-cert with the analysis of their sedimentary contexts.A sample of stratum 10 fauna (total number of identified

specimens, 675; includes 80 teeth and/or tooth fragments and595 complete bones and/or bone splinters) was studied in detailfor macroscopic traces of burning, namely surface darkening andcalcination. Much of the fauna from this stratum shows discol-oration typical of burning as a result of charring and calcination(Fig. 5). Color changes, interpreted as the result of exposure tofire, were identified on 43.7% (total number of identifiedspecimens, 295) of the bones and teeth in this sample (Table S2).Traces of burning were found on faunal samples from excavationspits (arbitrary 5- or 10-cm-deep levels within a stratum) acrossthe entire excavation area and in all depths within stratum 10.In square R28, which directly abuts the location with ashed

plant material and burned bones on a surface, the frequency ofburned bone reached 80% of the sample. FTIR analysis per-

formed on several bone fragments from this square shows thatsome of the discolored bone fragments (namely black, gray, andwhite fragments) display FTIR absorption characteristics of bonemineral heated to more than 400 °C (Fig. 5), thus supporting themicrostratigraphic observation (Fig. 4 and Fig. S5).None of the bones analyzed shows IR patterns characteristic of

complete calcination, namely the complete removal of the carbo-nates within the bone carbonate-hydoxylapatite. Thus, none of thespecimensanalyzed reached temperaturesof, or greater than, 700 °C.Sediment adhering to some of the gray bone fragments exhibit FTIRspectral characteristics of clay minerals heated between 400 °Cand 700 °C (Fig. S6), again supporting the hypothesis of in situburning of sediment during the Acheulean in this area of the cave.Banded ironstone is the main raw material for artifacts found in

the Acheulean of excavation 1. Banded ironstone is also found inthe assemblage as decimeter-size unworked slabs. These couldhave entered the cave only as manuports, as the cave is situated indolostone and there were no site formation processes that couldhave transported these slabs into the cave (Fig. 2 and Fig. S5). Instratum 10, banded ironstone artifacts and manuports show char-acteristic pot-lid fractures (Fig. S7A); of 633 spits (arbitrary units of1 × 1 square yards and 5 cm depth) excavated in stratum 10, 61produced unworked ironstone with pot-lid fractures. The distri-bution of the spits that produced ironstone with pot-lid fracturescovers the entire excavation area and all depths within stratum 10.

Fig. 3. (A) Photograph of east profile corresponding to squares R/Q28 showing a detailed view of stratum 10. Box indicates approximate location of thinsection (B), exhibiting three microstratigraphic units (microfacies): (1) bottom sandy silt and clay mixed with ashed plant material, dispersed wood ash, andbone fragments; (2) clay aggregates and fragments; and (3) rounded aggregates of sandy silt. Large red rectangular box indicates the location of boundarybetween microfacies 1 and 2, which is enlarged immediately above in C. Small blue box indicates subsurface location with wood ash pieces dispersed in siltysediment (enlarged in H). (C) Magnification of contact area between microfacies 1 and 2 in thin section shown in B shows characteristic of erosion andsuccessive stabilized surface, on top of which are ashed plant material and bone fragments shown in micrographs (D–J). Boxes mark the location ofmicrophotographs shown in D–G, I, and J (PPL). (Scale bar: 1 cm.). (D) Micrograph of fragments of ashed plant material (PPL). (Scale bars: 1 mm.) (E)Micrographs of lump of calcitic wood ash with typical ash rhombs (oxalate pseudomorphs) and prisms at the contact between microfacies 1 and 2 (PPL). (Scalebar: 500 μm.). (F and G) Micrographs of fragments of ashed plant material (PPL) (Scale bars: 1 mm.) (H) Sediment from microfacies 1, with fragmented ashedplant material and dispersed wood ash rhombs and prisms (oxalate pseudomorphs; PPL). (Scale bar: 1 mm.) (I) Micrograph of contact area betweenmicrofacies showing clay aggregates in microfacies 2 and organometallic area (degraded charred material?) and bone fragments resting on the surface ofmicrofacies 1. (Scale bar: 1 mm.). (J) Micrograph of bone fragment. (Scale bar: 100 μm.)

4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1117620109 Berna et al.

Several pot-lid flakes (some refittable to the original slabs; Fig. S8)were found, indicating that fracturing of the ironstone occurredinside the cave. The sizes of pot-lid fractures vary from approxi-mately 1 mm to larger than 4 mm, and are consistent with featuresproduced by us in experimental heating of ironstone at temper-

atures in excess of 500 °C (Fig. S7B). Although pot-lid fracture canform as a result of agents other than exposure to fire, we have noevidence to support an alternative interpretation for their abun-dance in the stratum 10 assemblage; pot-lids were rarely observedin surface samples outside the cave.

Discussion and ConclusionsMicrostratigraphic investigations of the Acheulean stratum 10 atWonderwerk Cave show the presence of well preserved ashedplant material and burned bone fragments deposited in situ ondiscrete surfaces and mixed within sediment in between thesesurfaces. The good preservation and angularity of the particlessuggest that these materials were the products of local combus-tion episodes that occurred in the proximity of the find spot,which is 30 m in from the present-day entrance.The significant amounts and extensive distribution of macro-

scopic burned bone fragments and ironstone manuports withpot-lid fractures, along with evidence of heated sediment, sug-gest widespread burning events inside the cave. The prevalenceof burning throughout the entire thickness of stratum 10 mini-mizes the likelihood that repeated wildfires were the source ofthe burning in the cave. Similarly, the absence of guano remainsand the lack of characteristic high-temperature phosphates sug-gest that the burning in stratum 10 was not a result of guano self-ignition episodes. Finally, the possibility of the material beingcombusted as a result of heat transfer from combustion eventsoccurring in an overlying Holocene layer is highly unlikely be-cause of the interposition of overlying Acheulean deposits.Thus, our data, although they do not show evidence of con-

structed combustion features, as listed by Roebroek and Villa asa criterion of controlled burning (3), demonstrate a very close as-

Fig. 4. Representative mFTIR reflectance spectra (red line) of bone frag-ments shown in micrographs of Fig. 4 and Fig. S1, and of an unheated andexperimentally heated bone processed in the thin section (black lines). Ap-pearance of infrared bands at 1,096 cm−1 and 630 cm−1 are used as heatingtemperature indicators, showing that the archaeological fragment washeated to more than 400 °C but less than 550 °C.

Fig. 5. Selection of bone fragments recovered close to wood ash identified in thin section (excavation 1, stratum 10, square R28, elevation from top of stratum10 of 15–20 cm) and their representative FTIR spectra. Gray and black bones (samplesA, C, andD) show the presence of IR absorptions at 630 cm−1 and 1,090 cm−1

characteristic of bone mineral heated to more than 400 °C (32). Yellow (B) and white bone (E) fragments show IR spectral pattern characteristic of unheatedbone or heated below 400 °C. The circular and irregular opaque nodules are composed of Fe and Mn oxides and a result of diagenetic impregnation.

Berna et al. PNAS Early Edition | 5 of 6

ANTH

ROPO

LOGY

PNASPL

US

sociation between hominin occupation and the presence of firedeep inside Wonderwerk Cave during the Early Acheulean. Thisassociation strongly suggests that hominins at this site had knowl-edge of fire 1.0 Ma. This is the most compelling evidence to dateoffering some support for the cooking hypothesis ofWrangham (1).Preliminary data suggest that the fuel used in Wonderwerk

was composed mainly of “light” plant material such as grasses,brushes, and leaves. Only a small number of identifiable calcifiedmacrobotanical specimens was recovered from stratum 10, in-cluding two small fragments of grass culms, two fragments ofsedge culm (possibly Eleocharis spp.), and very small fragmentsof dicot stem or root with diameters that are too small to permitidentification (Fig. S9). Interestingly, no large wood charcoalfragments were found in stratum 10. The absence of charredwood could be also a result of diagenetic processes leading to theselective preservation of charred organic materials. Nevertheless,the heating temperatures estimated by FTIR analysis of bonesand sediments do not exceed 700 °C (32) and therefore arecompatible with fires fueled with leaves and grasses.Our approach demonstrates that the composition and fabric of

combustion deposits are best documented at a microscopic scale,and it offers a partial explanation why traces of early fire havebeen so difficult to document. It is, in fact, striking that, as op-posed to the easily visible combustion features found in MiddleStone Age and Middle Paleolithic cave occupations in Africa,Europe, and the Middle East, the most conclusive evidence forfire was visible only through the use of soil micromorphology.Moreover, the unambiguous identification of burning was possi-ble only with the use of mFTIR directly on thin sections. Addi-tionally, the macroscopic evidence from burned bones and lithicmaterial supports the micromorphological evidence and high-lights the need for further research in the investigation of earlyfire in the archaeological record. We believe microstratigraphicinvestigations at Wonderwerk cave and other early hominin sitesin Asia and South and East Africa will have a significant impact inproviding fundamental evidence for the appearance of use of fireand its role in hominin adaptation and evolution.

Materials and MethodsMicromorphology. Samples were taken as intact blocks and loose samples,oven-dried for several days at 60 °C, and then impregnated with unpromotedpolyester resin, diluted with styrene and catalyzed with methyl-ethyl-ketoneperoxide. Hardened blocks were trimmed (50 × 75 mm by 10 mm thick)by using a rock saw and then sent to Spectrum Petrographics (Vancouver,WA) for processing into 30-μm-thick petrographic thin sections. The thinsections were examined with binocular and petrographic microscopes inplane-polarized (PPL) and cross-polarized light at magnifications rangingfrom 20× to 200×. Descriptive nomenclature follows that of Courty et al. (29)and Stoops (30).

FTIR Spectroscopy and Microspectroscopy. FTIR spectroscopy is a molecularanalytical technique well suited to identify heat-related transformation inmaterials of different nature such as clay minerals (ref. 31 and refs. therein)and bone (32). In particular, because of high temperature, the bone mineral—namely carbonate-hydroxylapatite—undergoes characteristic recrystal-lization. The recrystallization occurs at approximately 500 °C and above, andis appreciable by FTIR spectroscopy via the sharpening of the ν4PO4 (565–630cm−1) and ν3PO4 (1,020–1,100 cm−1) bands. The sharpening caused by hightemperature results in the appearance of two characteristic peaks at 630cm−1 and 1,096 cm−1 that are absent in fresh, archaeological, and fossil bone(32). Samples processed in thin section were analyzed by FTIR micro-spectroscopy by using a Thermo Spectra-Tech Continuum IR microscope at-tached to a Thermo-Nicolet Nexus 470 IR spectrometer. Spectra of particleswith diameter of approximately 150 μm were collected in transmission andtotal reflectance mode with a Reflectocromat 15× objective between 2,000cm−1 and 450 cm−1 at 8 cm−1 resolution.

ACKNOWLEDGMENTS. Fieldwork at Wonderwerk Cave is carried out underpermit to M. Chazan from SAHRA and museum analysis under the termsof an agreement with the McGregor Museum. We are grateful to ColinFortune, Director, and David Morris, Head of Archaeology, and othermembers of the staff of the McGregor Museum for their assistance. Weacknowledge the earlier work carried out at Wonderwerk Cave by PeterBeaumont, which served as the touchstone for this research. We thank AnnaPhilips for her work on the ironstone heating experiments. This research wasfunded by grants from the Canadian Social Sciences and HumanitiesResearch Council, the Wenner Gren Foundation, and National ScienceFoundation Grants 0917739 and 0551927.

1. Wrangham R (2009) Catching Fire: How Cooking Made Us Human (Basic Books, NewYork).

2. Organ C, Nunn CL, Machanda Z, Wrangham RW (2011) Phylogenetic rate shifts infeeding time during the evolution of Homo. Proc Natl Acad Sci USA 108:14555–14559.

3. Roebroeks W, Villa P (2011) On the earliest evidence for habitual use of fire in Europe.Proc Natl Acad Sci USA 108:5209–5214.

4. Karkanas P, et al. (2007) Evidence for habitual use of fire at the end of the LowerPaleolithic: site-formation processes at Qesem Cave, Israel. J Hum Evol 53:197–212.

5. Alperson-Afil N, Goren-Inbar N (2010) The Acheulian Site of Gesher Benot Ya’aqov:Ancient Flames and Controlled Use of Fire (Springer, New York), Vol 2.

6. Gowlett JAJ, Harris JWK, Walton D, Wood BA (1981) Early archaeological sites,hominid remains, and traces of fire from Chesowanja, Kenya. Nature 294:125–129.

7. Barbetti M (1986) Traces of fire in the archaeological record before one million yearsago. J Hum Evol 15:771–781.

8. Bellomo R (1994) Methods of determining early hominid behavioral activitiesassociated with the controlled use of fire at FxJj20 Main, Koobi Fora, Kenya. J HumEvol 27:173–195.

9. Rowlett RM (2000) Fire control by Homo erectus in East Africa and Asia. ActaAnthropologica Sinica Supplement to 19:198–208.

10. Clark JD, Harris JWK (1985) Fire and its roles in early hominid lifeways. Afr ArchaeolRev 3:3–27.

11. Brain CK (1993) The occurrence of burnt bones at Swartkrans and their implicationsfor the control of fire by Early Hominids. Swartkrans: A Cave’s Chronicle of Early Man,ed Brain CK (Transvaal Museum, Pretoria, South Africa), pp 229–242.

12. Sillen A, Hoering T (1993) Chemical characterization of burnt bones from Swartkrans.Swartkrans: A Cave’s Chronicle of Early Man, ed Brain CK (Transvaal Museum,Pretoria, South Africa), pp 243–249.

13. Brain CK, Sillen A (1988) Evidence from the Swartkrans Cave for the earliest use offire. Nature 336:464–466.

14. Goren-Inbar N, et al. (2004) Evidence of hominin control of fire at Gesher BenotYa’aqov, Israel. Science 304:725–727.

15. Richter D, Alperson-Afil N, Goren-Inbar N (2011) Employing TL methods for theverification of macroscopically determined heat alteration of flint artefacts fromPalaeolithic contexts. Archaeometry 53:842–857.

16. Alperson-Afil N, et al. (2009) Spatial organization of hominin activities at GesherBenot Ya’aqov, Israel. Science 326:1677–1680.

17. Goldberg P, Weiner S, Bar-Yosef O, Xu Q, Liu J (2001) Site formation processes atZhoukoudian, China. J Hum Evol 41:483–530.

18. Weiner S, Xu Q, Goldberg P, Liu J, Bar-Yosef O (1998) Evidence for the use of fire atZhoukoudian, China. Science 281:251–253.

19. Berna F, Goldberg P (2007) Assessing Paleolithic pyrotechnology and associatedhominin behavior in Israel. Isr J Earth Sci 56:107–121.

20. Goldberg P, Berna F (2010) Micromorphology and context. Quat Int 214:56–62.21. Sandgathe DM, et al. (2011) On the Role of fire in Neandertal adaptations in Western

Europe: Evidence from Pech de l’Azé and Roc de Marsal, France. PaleoAnthropology2011:216–242.

22. Beaumont PB, Vogel JC (2006) On a timescale for the past million years of humanhistory in central South Africa. S Afr J Sci 102:217–228.

23. Beaumont PB (2011) The edge: More on fire-making by about 1.7 million years ago atWonderwerk Cave in South Africa. Curr. Ant. 52:585–595.

24. Chazan M, et al. (2008) Radiometric dating of the Earlier Stone Age sequence inexcavation I at Wonderwerk Cave, South Africa: Preliminary results. J Hum Evol 55:1–11.

25. Matmon A, Ron H, Chazan M, Porat N, Horwitz LK (2012) Reconstructing the historyof sediment deposition in caves: A case study from Wonderwerk Cave. GSA Bull124(3-4):611–625.

26. Flügel E (2004) Microfacies of Carbonate Rocks: Analysis, Interpretation andApplication (Springer, New York).

27. Onac BP, Effnberger HS, Breban RC (2007) High-temperature and “exotic” mineralsfrom the Cioclovina Cave, Romania: A review. Studia Universitatis Babeş-Bolyai,Geologia 52:3–10.

28. Kuman K, Field AS, McNabb AJ (2005) La Préhistoire ancienne de l’Afriqueméridionale: contributions des sites à hominidés d’Afrique du Sud. La Paléolithiqueen Afrique: L’Histoire la Plus Longue, eds Sahnouni M, Pigeaud R (Errance, Paris), pp53–82.

29. Courty MA, Goldberg P, Macphail RI (1989) Soils and Micromorphology inArchaeology (Cambridge Univ Press, Cambridge).

30. Stoops G (2003) Guidelines for Analysis and Description of Soil and Regolith ThinSections (Soil Science Society of America, Madison, WI).

31. Berna F, et al. (2007) Sediments exposed to high temperatures: Reconstructingpyrotechnological practices in Late Bronze and Iron Age strata at Tel Dor (Israel). JArchaeol Sci 34:358–373.

32. Berna F (2010) Scientific Methods and Cultural Heritage. An Introduction to theApplication of Materials Science to Archaeometry and Conservation Science, edArtioli G (Oxford Univ Press, Oxford), pp 364–367.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1117620109 Berna et al.

Supporting InformationBerna et al. 10.1073/pnas.1117620109SI TextArcheological Stratigraphic Sequence in Excavation 1. A plan ofexcavation 1 is given in Fig. S1. Archaeological strata were de-fined by Beaumont during excavation (1). In publications,Beaumont presents subdivisions of Strata (St.)—St.12 a-c; 10a-b;9a-f; 8a-e;7a-b; 5a-b—but the basis for these subdivisions has notbeen described by him, nor is it possible to pinpoint the tran-sitions between subdivisions in the stratigraphic profiles. Asa consequence, we have grouped artifacts by strata. It is howeverpossible to track depth within the stratum for each artifact.Presentations of Wonderwerk Cave by Beaumont (i.e., refs. 1,

2) also use site-wide Major Units, which are based on strata-specific correlations between excavation areas. The correlations,however, are not based on detailed lithic, faunal, or botanicalanalyses, continuous stratigraphic profiles, or rigorous absolutedates, so we have chosen not to adopt the Major Units describedby Beaumont, and we treat each excavation area separatelybased on its excavation-specific strata.The lithic assemblage fromBeaumont’s excavation is curated in

the McGregor Museum, Kimberley (South Africa). This sampleincludes all lithic material recovered during excavation and isdominated by fragments of dolostone from roof collapse in ad-dition to artifacts and manuports. Throughout the sequence ofexcavation 1, the lithic assemblage size is small and the number offlakes is very limited. All flaked material and pieces with pot-lidfracture were cataloged for documentation and analysis.Stratum 12.The lithic assemblage of stratum 12 is characterized bysmall flakes and small cores (3). This assemblage is assigned tothe Oldowan and appears comparable to the assemblage fromSterkfontein, although there is a significant difference in rawmaterial between Sterkfontein where the dominant material isquartzite and Wonderwerk stratum 12, where the dominant rawmaterial is chert.Stratum 11. Stratum 11 has yielded a very small assemblage of onlytwo bifaces and 19 tools in total (3). One biface (SPL 37) that wasfound near the base of stratum 11 is very crudely worked and isbest classified as a protobiface (Fig. S2A). The definition ofa protobiface, following Leakey (4), includes bifacial flaking alongboth edges and at the tip. On SPL 37, bifacial flaking is limited toone edge and the tip; however, the nature of the fine removals onthe tip fits better with the classification of a protobiface than withclassification as a chopper. The second biface (SPL 36) was re-covered near the top of the stratum and is characterized by a se-ries of short removals around the circumference. The tip is brokenand the piece is damaged by conchoidal fractures associated withthe development of calcium carbonate crusts.Stratum 10. Seven bifaces were recovered from stratum 10 (Fig. 1 Band C and Fig. S3 A–D). With one exception, none of the bifacesshow invasive flake scars. Biface tips tend to be cleaver-like, asthey form a working edge rather than a pointed tip, although theoverall dimensions of the bifaces do not fit with the definition forcleavers. All are on slabs of ironstone, with the exception of twobifaces (SPL 39 and 702), which are on large cortical flakes (Fig.S3 B and D). One piece (SPL 45) included as a biface could beclassified as a chopper, as it consists of three large removals withdeep bulbs of percussion on an ironstone cobble (Fig. S3C).Several of the bifaces have more than one working edge. Themost striking of these is SPL 702 (Fig. S3D), which is on a largesemicortical flake of homogenous gray ironstone. This biface istrapezoidal in section, with one edge formed by a preexistingdorsal flake scar and the other by abrupt removals off the op-posing edge. There are just a few invasive flake scars, all with

prominent bulbs. The tip is offset from the main axis and hasbeen formed with a series of flat removals off one face. There isalso limited working on the base which has a cleaver-like edgethat is quite sharp. From a functional perspective, this can beseen as a double cleaver/handaxe tool. There are a total of 36flakes and 15 cores from stratum 10; however, none of these canbe considered diagnostic of a specific production method. Asshown in Fig. S1, the bifaces are mostly found in the central andsouthern part of the excavated area in excavation 1, whereasflakes are more widely distributed.Strata 9 and 8.Beginning with stratum 9, there is decrease in bifacethickness and increase in length-to-thickness ratio (i.e., refine-ment) associated with the development of invasive flat removalsfor thinning bifaces which first appears in stratum 9 and thenbecomes a regular aspect of the assemblage in stratum 8. In theupper part of stratum 9 and in stratum 8, there are several cleaversthat appear consistent with the Victoria West technique, as theyare made on large flakes struck from apparently prepared cores(5). One of these cleavers is made on igneous rock (Fig. S2B).One handaxe from stratum 8 is made on chert, the first large toolon this raw material in the sequence (Fig. S2C). Throughout thesequence assemblage size is small and the number of flakes isvery limited. There is no evidence of either reduction in bifacesize in the excavation 1 sequence or systematic blade productionthat would support the attribution of stratum 8 to stratum 5 tothe Fauresmith (1).

Chronostratigraphic Sequence in Excavation 1. Paleomagnetic ori-entation and cosmogenic 26Al/10Be ratio were analyzed along theexcavation 1 sequence in the main north and main east profiles(Fig. S1). An integrated overview of the paleomagnetic data andthe cosmogenic dates from the stratigraphic sequence is given inFig. S4. The burial ages calculated from the measured 26Al/10Beratio are given in Table S1. These ages are calculated using26Al/10Be ratios of 3.98 ± 0.24 and 4.08 ± 0.22 of samples col-lected from the surface outside the cave, which imply an initialburial signal corresponding to 0.78 ± 0.15 Ma (see ref. 6).

SI Materials and MethodsFaunal Sample. A sample of mammalian bones and teeth, repre-senting just more than one quarter of the total stratum 10 faunalassemblage, has been studied to date. For each item, surface colorwas scored by using five categories that have been shown to followa progression during burning (7, 8) (Table S2): brown-red, black,gray-white, and white (i.e., calcined).

Heat Experiment with Lithic Raw Material. Four slabs of localironstone were collected from outside Wonderwerk Cave and cutwith a petrographic saw into four subsamples, each approximately2 × 2 × 1 cm in size. Subsamples from each slab were heated in aprogrammable laboratory kiln to temperatures of 250 °C, 500 °C,750 °C, and 1,000 °C, respectively, for a duration of 1 hour. Thesamples were then examined for changes to color and texture aswell as the presence and characteristics of pot-lid fractures (Fig. S7).

Botany. Three samples of macrobotanical remains were analyzedfrom stratum 10. Two derive from square R25 and one is fromsquare T33 (Fig. S9).Grass culm. The epidermis has been partially removed from theculm, revealing several longitudinal peripheral vascular bundles.This fragment is 3 mm long and just less than 1 mm in diameter.Another fragment is 0.5 mm in diameter and has a solid outer

Berna et al. www.pnas.org/cgi/content/short/1117620109 1 of 8

cortexwith large cells toward the center that formacentral canal ormore likely a central hollow because the cells have broken down.Sedge culm.Another two examples with a diameter of 0.7 mm havea very thin epidermis and aerenchymatous cells in the center(like a sponge). The vascular bundles are not visible. The sedgeEleocharis spp. has this type of culm.

Dicot stem or root. These are also 0.5 and 0.7 mm in diameter, buthave a different type of tissues. They typically crack radially (i.e.,like slices of a pie) and comprise fibers and parenchyma as theground tissue (i.e., smallest cells), with solitary vessels (i.e., largerempty cells). It is not possible to identify such small-diameter stemsor roots.

1. Beaumont PB, Vogel JC (2006) On a timescale for the past million years of humanhistory in central South Africa. S Afr J Sci 102:217–228.

2. Beaumont PB (2011) The edge: more on fire-making by about 1.7 million years ago atWonderwerk Cave in South Africa. Curr. Ant. 52:585–595.

3. Chazan M, et al. (2008) Radiometric dating of the Earlier Stone Age sequence inexcavation I at Wonderwerk Cave, South Africa: preliminary results. J Hum Evol 55:1–11.

4. Leakey M (1971) Olduvai Gorge: Excavations in Beds I and II, 1960-1963 (CambridgeUniv Press, Cambridge).

5. Lycett SJ, von Cramon-Taubadel N, Gowlett JAJ (2010) A comparative 3D geometricmorphometric analysis of Victoria West cores: Implications for the origins of Levalloistechnology. J. Arch. Sci. 37:1110–1117.

6. Matmon A, et al. (2012) Reconstructing the history of sediment deposition in caves: Acase study from Wonderwerk Cave, South Africa. Geol Soc Am Bull 124:611–625.

7. Stiner MC (1991) The faunal remains from Grotta Guattari: A taphonomic perspective.Curr. Anth. 32:103–117.

8. Stiner MC, Weiner S, Bar-Yosef O, Kuhn S (1995) Differential burning, fragmentation,and preservation of archaeological bone. J. Arch. Sci 22:223–237.

Fig. S1. Distribution of artifacts in stratum 10, excavation 1, in Wonderwerk Cave, relative to the location where micromorphology revealed the presence ofashed plant material (red arrow).

Berna et al. www.pnas.org/cgi/content/short/1117620109 2 of 8

Fig. S2. Bifaces from excavation 1, Wonderwerk Cave: (A) stratum 11 and (B and C) stratum 9.

Berna et al. www.pnas.org/cgi/content/short/1117620109 3 of 8

Fig. S3. Bifaces from stratum 10, excavation 1, Wonderwerk Cave. (A) SPL 43 on slab of ironstone. (B) SPL 39 on a large semicortical flake of homogenous grayironstone. (C) SPL 45 on ironstone cobble. (D) SPL 702 on large ironstone cortical flake.

Fig. S4. Photograph of main north and main east profiles of excavation 1 show subdivision of archaeological strata 12 to 9, the corresponding sedimentaryunits (i.e., numbers 1–9), and the locations of cosmogenic dates (in yellow boxes) and paleomagnetic normal and reverse data (black circle, undetermined; redcircle, reverse; blue circle, normal). Further details can be found in refs. 3, 6.

Berna et al. www.pnas.org/cgi/content/short/1117620109 4 of 8

Fig. S5. (A) Photograph of stratum 10 in the trench east profile of excavation 1. Boxes mark approximate locations of thin sections shown in images B–D. (B)Scan of 5 × 7.5 cm petrographic thin section showing two major microstratigraphic units. (i) At bottom, there is white phosphatized flowstone or degradeddolomite with iron manganese dendritic nodules. (ii) At top, separated by sharp boundary, is orange-red sand with few partially disaggregated rounded siltaggregates and millimeter-sized bone fragments and opaque millimeter-size oxide nodules. Note a large angular fragment of ironstone. Small dashed boxindicates location of bone fragments shown in micrograph [E; plane-polarized (PPL)] (Scale bar: 1 cm). (C) Scan of 5 × 7.5 cm petrographic thin section showingtwo major microstratigraphic units separated by a sharp contact onto which large (i.e., centimeter-size) fragments of bone were deposited. The unit at thebottom (i) is composed of a dusty silty clay groundmass, with quartz grains coated with clay and local impregnations of Fe-Mn. The top portion of the unit islocally bedded with some phytoliths. (In cross-polarized light; not shown here—the sand beneath the contact appears compressed, as if it were trampled.) Inthe top unit (ii), several bone fragments and phosphatized wood ash fragments are present. Small dashed box indicates location of bone fragments shown inmicrograph. (D) Scan of petrographic thin section exhibiting three microstratigraphic units: (i) bottom sand silt and clay mixed with ashed plant material,dispersed wood ash, and bone fragments; (ii) reddish clay aggregates and fragments; and (iii) rounded aggregates of sandy silt. Small dashed box indicateslocation of bone fragments shown in micrograph (G). (E) Micrograph of charred bone (PPL). (Scale bar: 1 mm.) (F) Micrograph of white spongy bone fragmentlaying on the surface between two microstratigraphic units (PPL). (Scale bar: 1 mm.) (G) Angular fragments of partially charred bone (PPL). (Scale bar: 1 mm.)(H) Representative reflectance Fourier transform IR microspectroscopy spectra collected from bone fragments in micrographs E–G showing the presence of IRbands at 630 cm−1 and 1,090 cm−1 characteristic of bone mineral heated above 400 °C.

Berna et al. www.pnas.org/cgi/content/short/1117620109 5 of 8

Fig. S6. (A) Fragment of discolored (gray) bone with adhering rubefied sediment from excavation 1, stratum 10, square Q31. (B) Representative Fouriertransform IR spectrum of rubefied sediment showing IR pattern characteristic of quartz, calcite, and heat-altered clay as described previously (1). (C) Repre-sentative Fourier transform IR spectrum of the gray bone material showing presence of IR absorptions at 630, 1,090, and 3,570 cm−1 characteristic of bonemineral heated above 400 °C.

Fig. S7. (A) Photograph of banded ironstone fragment from location adjacent to wood ash (Square R28). Note characteristic pot-lid fracture. (B–D) Samples ofbanded ironstone slabs collected on the hillside outside Wonderwerk Cave and heated under experimental conditions at temperatures ranging from 500 °C to1000 °C for one hour, showing the formation of pot-lid fractures comparable to the ones found in the archaeological assemblage.

1. Berna F, et al. (2007) Sediments exposed to high temperatures: Reconstructing pyrotechnological practices in Late Bronze and Iron Age Strata at Tel Dor (Israel). J. Arch. Sci. 34:358–373.

Berna et al. www.pnas.org/cgi/content/short/1117620109 6 of 8

Fig. S8. Ironstone slabs with refitted pot-lid flake from excavation 1, stratum 10, square Q33, spit 10 to 15 cm.

Fig. S9. Macrobotanical remains from stratum 10. (A) Culm with a diameter of 0.7 mm, very thin epidermis, and aerenchymatous cells in the center. Thevascular bundles are not visible. The sedge Eleocharis spp. has this type of culm. (B) Dicot stem or root, 0.5 and 0.7 mm in diameter, with different type oftissues. Dicots typically crack radially and comprise fibers and parenchyma as the ground tissue (i.e., smallest cells), with solitary vessels (i.e., larger empty cells).It is not possible to identify such small-diameter stems or roots. (C) Dicot stem or root external view with a knot where a “branch”was attached. (D) Grass culm:the epidermis has been partially removed from the culm, revealing several longitudinal peripheral vascular bundles (light lines in a dark background). Thisfragment is 3 mm long and just less than 1 mm in diameter.

Berna et al. www.pnas.org/cgi/content/short/1117620109 7 of 8

Table S1. Cosmogenic burial ages for excavation 1, based on26Al/10Be ratio measured in sediment samples collected alongthe stratigraphic column

Sample Archaeological stratum Sedimentary unit Age, Myr

COS 6 9 2 0.99 ± 0.19COS 5 10 4 top 0.98 ± 0.19COS 4 10 4 1.27 ± 0.19COS 3 11 6 1.17 ± 0.19COS 2 12 9 top 1.39 ± 0.19COS 1 12 9 bottom 1.66 ± 0.20WWD 1 12 9 bottom 1.58 ± 0.19

For complete discussion of these ages see Matmon et al (6).

Table S2. Burned and unburned bone and teeth from a sample of fauna recovered from stratum10 in excavation 1 at Wonderwerk Cave and a breakdown of burned bone by color

NISP bone % bone NISP teeth % teeth Teeth and bone NISP

Unburned 349 91.8 31 8.2 380Burned 246 83.3 49 16.6 295Total 595 88.1 80 11.8 675

Burning color NISP bone % of total bone NISP teeth % of total teeth Teeth and bone NISP

Brown-red 6 1.0 2 2.5 8Black 58 9.7 33 41.2 91Grey 103 17.3 14 17.5 117White 79 13.2 0 0 79

NISP, number of identified specimens.

Berna et al. www.pnas.org/cgi/content/short/1117620109 8 of 8