Contents lists available at ScienceDirect

Journal of Archaeological Science: Reports

journal homepage: www.elsevier.com/locate/jasrep

Later Stone Age toolstone acquisition in the Central Rift Valley of Kenya:Portable XRF of Eburran obsidian artifacts from Leakey's excavations atGamble's Cave II

Ellery Frahma,b,⁎, Christian A. Tryonb,⁎⁎

a Yale Initiative for the Study of Ancient Pyrotechnology, Council on Archaeological Studies, Department of Anthropology, Yale University, New Haven, CT, United StatesbDepartment of Anthropology, Harvard University, Peabody Museum of Archaeology and Ethnology, Cambridge, MA, United States

A R T I C L E I N F O

Keywords:Obsidian sourcingRaw material transferNaivasha-Nakuru BasinAfrican Humid PeriodHunter-gatherer mobilityHuman-environment interactions

A B S T R A C T

The complexities of Later Stone Age environmental and behavioral variability in East Africa remain poorlydefined, and toolstone sourcing is essential to understand the scale of the social and natural landscapes en-countered by earlier human populations. The Naivasha-Nakuru Basin in Kenya's Rift Valley is a region that is notonly highly sensitive to climatic changes but also one of the world's most obsidian-rich landscapes. We usedportable X-ray fluorescence (pXRF) analyses of obsidian artifacts and geological specimens to understand pat-terns of toolstone acquisition and consumption reflected in the early/middle Holocene strata (Phases 3–4 of theEburran industry) at Gamble's Cave II. Our analyses represent the first geochemical source identifications ofobsidian artifacts from the Eburran industry and indicate the persistent selection over time for high-qualityobsidian from Mt. Eburru, ~20 km distant, despite changes in site occupation intensity that apparently correlatewith changes in the local environment. This result may indicate resilience of Eburran foraging strategies duringenvironmental shifts and, potentially, a cultural preference for a specific lithic material that overcame its ac-cessibility changes. Testing such hypotheses requires a more extensive program of obsidian artifact sourcing. Ourfindings demonstrate the great potential for sourcing studies in the Rift Valley as well as underscore the amountof work that remains to be done.

1. Introduction

The complex dynamics of Holocene interglacial environmental andbehavioral variability in East Africa remain poorly defined. Significantand often abrupt environmental changes include the return to near-glacial conditions during the Younger Dryas event (∼12.9–11.7 ka),followed by the African Humid Period (AHP, ~11–6 ka), and a return tomore arid conditions (∼6–4 ka; e.g., Gasse, 2000; Tierney anddeMenocal, 2013). Changes during the AHP are particularly striking,including the “Green Sahara” phase of northern Africa, with a numberof East African lakes expanding in size and depth, coincident with theexpansion of more forested ecotones (Butzer, 1972; Hamilton, 1982;Bergner et al., 2009; Trauth et al., 2010). Archaeologically, a number ofregionally distinct artifact traditions are recognized, including theKansyore fisher-foragers of the Lake Victoria basin (Dale and Ashley,2010), fisher-forager communities who produced the “dotted-wavyline” pottery and barbed harpoons found throughout parts of the Sahara

and the Sahel as far south as Lake Turkana (Wright et al., 2015; cf.Sutton, 1974; Holl, 2005), and the Eburran industry produced by for-agers in the Central Kenyan Rift Valley (e.g., Ambrose, 1984; Wilshaw,2016). Cattle, sheep, and goat spread southward from northern toeastern Africa by ~7 ka, resulting in a complex cultural and economicmosaic that was in place by 3 ka (di Lernia, 2013; Kusimba, 2003; Lane,2013; Skoglund et al., 2017).

A number of unanswered questions remain about the archaeology ofthis timeframe. For example, do distinctive regional behavioral entities(e.g., archaeological industries like the Eburran) form because ofadaptations to specific habitats, as a result of isolation from neigh-boring groups, or as a deliberate expression of what Wiessner (1983)referred to as emblemic style? Does the spread of pastoralism reflect thedemic diffusion of groups along an expanding frontier or a series oflocal adaptations as stock is transferred along existing social networksamong foraging groups (cf. Lane, 2004; Prendergast, 2010; Skoglundet al., 2017)? Answering these questions ultimately requires

https://doi.org/10.1016/j.jasrep.2018.01.042Received 24 July 2017; Received in revised form 23 January 2018; Accepted 28 January 2018

⁎ Correspondence to: E. Frahm, Yale Initiative for the Study of Ancient Pyrotechnology, Council on Archaeological Studies, Department of Anthropology, Yale University, New Haven,CT, United States.

⁎⁎ Corresponding author.E-mail addresses: ellery.frahm@yale.edu (E. Frahm), christiantryon@fas.harvard.edu (C.A. Tryon).

Journal of Archaeological Science: Reports 18 (2018) 475–486

2352-409X/ © 2018 Elsevier Ltd. All rights reserved.

T

establishing the nature and strength of inter-group connections, whichis not straightforward, particularly using archaeological data. For-tuitously, artifacts made of obsidian are common at many AHP sites inEast Africa, particularly in Kenya, and identifying the geologicalsources of these artifacts can enhance our understanding of the extentto which different regions – and, in turn, the populations within them –were connected in the past.

There are> 80 geochemically distinct obsidian sources along an800-km stretch of Kenya between Lake Turkana on the Ethiopianborder and Lake Natron on the Tanzanian border (Brown et al., 2013).Of these, roughly two-thirds lie within Kenya's Naivasha-Nakuru Basinand its surroundings, and there is now a fairly robust geochemicaldatabase for source outcrops as well as Pleistocene-Holocene artifactsmade by foragers and pastoralists from sites in Kenya and Tanzania(e.g., Merrick and Brown, 1984a, 1984b; Merrick et al., 1988, 1994;Mehlman, 1989; Coleman et al., 2008, 2009; Nash et al., 2011; Ndiemaet al., 2011; Ambrose et al., 2012a, 2012b; Ferguson, 2012; Prendergastet al., 2013; Faith et al., 2015; Blegen, 2017; Blegen et al., 2017; Frahmet al., 2017). From the extant data, a few generalizations can be made.First, foragers and pastoralists in the Lake Turkana basin appear to haveused locally available obsidian sources, which means there is no strongevidence in these data to support cultural connections between theTurkana region and the Naivasha-Nakuru Basin. Second, pastoralistgroups relied extensively on sources within the Naivasha-Nakuru Basin,transporting large quantities of obsidian across long distances (often≥100 km), particularly to the south. Third, there has been persistentmovement of obsidian from the Naivasha-Nakuru Basin into the LakeVictoria basin. Connections between these regions date to at least theLate Pleistocene (Faith et al., 2015; Blegen et al., 2017) and persistthroughout the Holocene, including portions of the AHP (Merrick andBrown, 1984a, 1984b; Frahm et al., 2017) that overlap in time withKansyore sites in the Lake Victoria basin and assemblages attributed tothe Eburran industry in the Naivasha-Nakuru Basin.

Apparent connections between foragers in the Lake Victoria basin(e.g., Kansyore groups) and groups in the Naivasha-Nakuru Basin (e.g.,Eburran producers) are based entirely on obsidian sourcing data fromsites in the Lake Victoria basin, that is, from sites ~200 km from thesources. As yet, there are no published geochemical data on obsidianartifacts found at Eburran industry sites. Such data can serve as a meansto better understand the extent of connections between these two areas.To initiate this, we focus on collections from Gamble's Cave II (GC2;Kenya), the type-site for Phases 3, 4, and 5 of the Eburran industry. Inparticular, we focus on a collection of GC2 obsidian artifacts from L.S.B.Leakey's 1920s excavations housed at Harvard University's PeabodyMuseum of Archaeology and Ethnology. This subsample of the GC2assemblage reflects what Leakey (1931) termed the “fourth occupationlevel” of GC2. He initially subdivided it into the Kenyan AurignacianPhases a and b, which were subsequently designated as the type de-posits for Phases 3 and 4 of the Eburran industry (see Ambrose et al.,1980; Ambrose, 1984; Wilshaw, 2016).

Using state-of-the-art portable X-ray fluorescence analysis (pXRF),we chemically identified the geological sources of 239 GC2 obsidianartifacts. pXRF instruments offer archaeologists several advantages overtechniques traditionally used for obsidian sourcing. First, pXRF isnondestructive, meaning that artifacts do not need to be polished,powdered, dissolved, or discarded. Second, it can be conducted at amuseum, in a field house, or even at an archaeological site. Third, it canbe rapid, often needing just a minute or two to measure dozens ofelements. The first two advantages were paramount in this study.Recent studies demonstrate that newer pXRF instruments have tech-nical advances (e.g., large-area Si drift detectors with high spectralresolution, adaptive signal-processing electronics) that enable excellentaccuracy, reproducibility, and sensitivity (e.g., Frahm, 2014; Milić,2014; Frahm and Feinberg, 2015; Newlander et al., 2015; LeBourdonnec et al., 2015; Campbell and Healey, 2016; Bonsall et al.,2017; Orange et al., 2017; cf. older instruments in Potts and West,

2008; Drake et al., 2009; and Liritzis and Zacharias, 2011). In thisstudy, our pXRF measurements were directly calibrated and comparedto the published data of Brown et al. (2013) in order to determine theartifacts' origins. We have previously documented that the XRF data-base of Brown et al. (2013) and our pXRF measurements are directlycompatible (Frahm et al., 2017).

Our results for the Eburran Phases 3 and 4 at GC2 suggest that thesite's occupation history or local environment had no discernable effect– at least in this sample – on the variety or range of obsidian sourcesused for tool manufacture. Previously unpublished data for the site's useindicate a higher occupation intensity during the more humid EburranPhase 3, when the lake's shoreline was likely closer at GC2 than atpresent. As the lakeshore retreated and, with it, the site's position at anecotone important to Eburran groups, the occupation intensity at GC2decreased, as indicated by the lower frequency of retouched tools andartifact discard rates in the later deposits. In both phases, the closestsource – Mt. Eburru at ~20 km distant – was consistently exploited tomeet the demands for lithic material. This implies that demand for thistoolstone outweighed ecologically linked changes in its accessibility toGC2 occupants. This result might also indicate the resilience of Eburranforaging strategies during the transition from an earlier, more humidinterval and a later, drier one. Speculatively, these GC2 data might alsomark the beginning of the widespread use of Mt. Eburru obsidian byforagers across Kenya and Tanzania. Testing such hypotheses requires amore extensive program of obsidian sourcing in the region, and ourfindings not only reveal the potential for such analyses but also un-derscore the amount of work that remains to be done.

2. Background

The GC2 rockshelter (0.55525° S, 36.08936° E, 1934m) is anoverhang cut into an interface between Quaternary lacustrine sedi-ments and Pleistocene pyroclastics, where it formed during an earlyHolocene highstand of Lake Nakuru (Figs. 1–2; Washbourn-Kamau,1971). Leakey excavated ~45m2 at GC2 to a depth of ~8.5 m in1926–1929, and he recognized four “occupation levels” that include anumber of early-to-late Holocene LSA obsidian blade-based lithic in-dustries that span the transition to food production (i.e. pastoralism)and the local adoption of ceramics. These different occupation levelswere only summarily described in his 1931 book The Stone Age Culturesof Kenya Colony. As Leakey (1971:iv) notes in a new introduction to thesecond printing of that book, “a very detailed report… on Gamble'sCave II, was sent to press in Paris, just before the outbreak of World WarII, but the whole report and all the illustrations and diagrams weremislaid or destroyed during the German Occupation.”

In 1964, Glynn Isaac and Ronald Clarke excavated a 1×2m stra-tigraphic section at GC2 to collect material for radiocarbon dating.Ambrose (1984) provides a summary of lithic and faunal assemblagesfrom the 1964 excavation. Materials from the Leakey and Isaac/Clarkeexcavations formed the basis for a major revision of the Holocene LaterStone Age (LSA) sequence in the Central Kenya Rift. The divisions arebased on stratigraphy as well as changes in retouched tool type and, inparticular, a size decline in blade and backed piece length over time.The early so-called “giant blade” Eburran Phases 1 and 2, which are notpresent at GC2, are dated to ~12 ka and 11.0–10.3 ka elsewhere in theCentral Kenya Rift (Wilshaw, 2016). These are followed by EburranPhase 3 and Phase 4, which are not very well dated at GC2. Reliabledates are restricted to a number of similarly aged radiocarbon de-terminations near the base of the Phase 3 deposits. Estimates from GC2(based partially on inferred sedimentation rates) are generally con-sistent with those from other Eburran Phase 3 and Phase 4 assemblages,which suggest an age of ~8.5–8 ka for Phase 3 and ~7.8–6 ka for Phase4 (Ambrose, 1984; Ambrose et al., 1980; Bower et al., 1977; Protsch,1978; Wilshaw, 2016). These are, in turn, overlain by the ‘small blade’Eburran Phase 5 (~4.5–1.8 ka) and by the Pastoral Neolithic Elmen-teitan industries from ~3.2–1.4 ka (Goldstein and Munyiri, 2017).

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

476

GC2 is considered the type site for Eburran Phases 3 and 4 (as wellas the Eburran Phase 5 and the Elmenteitan Neolithic industry;Ambrose, 1984), and Phases 3 and 4 are the focus of our research. TheEburran Phase 3 and 4 lithic technological system is geared towards theproduction of blades and bladelets from single or opposed platform

cores made on angular spalls. These blades and bladelets were subse-quently modified into several tool types, in particular a variety ofbacked pieces, with the larger blades segmented and transformed intosmaller tools such as endscrapers and truncated/facetted pieces for theproduction of small flakes, bladelets, and burin spalls (Nelson, 1973;

Fig. 1. Map of the Central Rift Valley of Kenya, including the location of Gamble's Cave II (GC2) and the obsidian sources identified at GC2 in the Eburran Phases 3 and 4 deposits. Thelight blue shaded areas correspond to Wilshaw's (2012) lake level reconstructions ∼8 ka. Before that, ∼9.5 ka, the Nakuru-Elmenteita palaeo-lake reached all the way to – and formed –the GC2 rockshelter. See also the lake level reconstructions in Bergner et al. (2009) for their maximum extents. (For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

Fig. 2. Stratigraphic section from Leakey's original excavations at GC2 (redrawn from Leakey, 1931: Figure 15; colors are arbitrary, selected for clarity).

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

477

Newcomer and Hivernel-Guerre, 1974; Ambrose, 1984; Wilshaw, 2012,2016).

Considerable research has focused on potential links among theEburran-producing groups' subsistence strategies, settlement locations,and shifting altitudes of savannah-bush-forest ecotones within thisbasin (e.g., Isaac, 1972; Ambrose et al., 1980; Ambrose, 1984, 1986,2001; Ambrose and DeNiro, 1989; Ambrose and Sikes, 1991; Marean,1992; Marean et al., 1994). Three main ecological zones exist at dif-ferent altitude ranges in the Nakuru-Naivasha Basin: savannah (low-land), bush (mid-range) and montane forest (highland). Human settle-ment focus appears to track the position of the boundaries betweenthem over time as they shifted in response to climatic changes (e.g., thebush and montane forest expands into lowland areas during wet times,the savannah expanded into higher altitudes in drier times).

The GC2 faunal assemblage appears to reflect at least some of thesechanges. Provenience data are poor for original faunal materials re-ported in Leakey (1931). Data reported from the Isaac and Clarke ex-cavations consist of identifiable teeth only: 87 specimens from Phase 3and 27 from Phase 4. Phases 3 and 4 are both dominated by smallbovids (Ambrose, 1984; Marean, 1992). Ambrose (1984) interprets theEburran Phase 3 deposits as indicative of forest or bush conditions,based on the presence of bushbuck (Tragelaphus scriptus) and variousduikers (Cephalophus sp.), consistent with evidence from Lake Nakuruand Lake Naivasha for more humid conditions when the lakes had ex-panded at this time and the lake margin was closer than at present(Richardson and Dussinger, 1986; Bergner et al., 2003, 2009). In con-trast, the Phase 4 fauna are interpreted by Ambrose (1984) as indicativeof a slightly more open bush and/or grassland landscape, suggested, inpart, by higher proportions of reedbuck (Redunca cf. fulvorofula) in thePhase 4 deposits. This interpretation would be consistent with a generaldecline in lake size due to increased aridity that began ~6 ka or earlier(Richardson and Dussinger, 1986; Maitima, 1991), which, therefore,would have positioned GC2 farther from the lake margin in Phase 4than during Phase 3.

3. Hypotheses

Because we are interested in the nature of diachronic changeswithin the Eburran, we use archaeological and obsidian source datasetsto test three hypotheses about the occupation history of GC2.

Hypothesis 1. Occupation intensity was greater at GC2 during EburranPhase 3. This hypothesis is based on paleoenvironmental data from GC2and neighboring Eburran sites as well as theoretical models drawn frombehavioral ecology. In general, we expect more wooded nearshoreenvironments of Eburran Phase 3 at GC2 to provide a more dense andpredictable resource base than the inferred drier and more openhabitats during Eburran Phase 4, thereby supporting more intensiveoccupation at the site. A number of ethnographic and theoreticalstudies (e.g., Dyson-Hudson and Smith, 1978; Wiessner, 1982, 1983;Ambrose and Lorenz, 1990; Kelly, 2013; Marean, 2016) suggest that(1), with predictable and dense resources on the landscape, hunter-gatherers tend to have small ranges and make less frequent moves but(2), when resource productivity and predictability decrease, their homerange sizes and mobility tend to rise.

Hypothesis 2. Occupation intensity was greater at GC2 during EburranPhase 4. Based on data from the Isaac and Clarke excavations, Ambrose(1984) suggested, on the basis of patterning within the GC2 lithic andfaunal data, more intensive occupation at GC2 occurred during EburranPhase 4 relative to Phase 3. He noted there were considerably moreretouched tools in Phase 4 relative to Phase 3, which he interpreted asincreased occupation intensity. The Phase 4 retouched artifactassemblage showed reduced typological diversity relative to Phase 3,interpreted as the outcome of greater spatial segregation of activitiestied to more intensive occupation (Ambrose, 1984). Finally, the greaternumber of complete and identifiable faunal elements (measured by

teeth) in Eburran 3 levels were used to support this scenario, with theexpectation of greater fragmentation (and thus fewer completespecimens) with increased occupation intensity during Eburran Phase4.

Hypothesis 3. Greater occupation intensity correlates to increased relianceon local obsidian sources and greater diversity in the number of sources used.This hypothesis predicts an increased reliance on local obsidian sources,as tools, which are subject to more intensive and/or prolonged uses, areincreasingly made or replaced at the site as needed. Similarly, thediversity of toolstone materials brought to the site should increase withoccupation intensity as wider portions of the landscape are encounteredduring extended forays to procure toolstone or other resources (cf.Kuhn, 1991; Surovell, 2009), as foraging range tends to increase withoccupation duration (Hovers, 2009:158).

4. Materials and methods

Here we report how we tested these hypotheses, in particular thedetails of the sourced GC2 artifact assemblage, the pXRF instrumentand analytical protocols, our specialized calibration for direct com-patibility with the published datasets of Brown et al. (2013), and ourmeans of matching GC2 artifacts to their sources. We followed theprocedures documented in Frahm et al. (2017), so readers interested inadditional details regarding our protocols and their quality assessmentsare directed to that publication.

4.1. Archaeological assessments of mobility and occupation intensity

Ambrose's (1984) estimates of occupation intensity and frequencyare based on the number of retouched tools and on an inverse re-lationship between artifact and faunal densities at the site, with faunaldensity estimated entirely on the number of recovered teeth. The in-ferences were that more such artifacts reflect more use of a site and thatgreater, more intensive occupation results in increased faunal frag-mentation and, in turn, fewer identifiable teeth for the Eburran 4 de-posits. In terms of the lithic assemblage, research since Ambrose's(1984) hypotheses were formulated have emphasized that the numberof retouched tools alone is insufficient to infer either the frequency orintensity of site use. Specifically, Riel-Salvatore and Barton (2004;Barton and Riel-Salvatore, 2014) have shown that a superior measure isthe frequency of retouched tools relative to the “artifact volumetricdensity,” that is, the total number of artifacts per excavated cubic meterfor a particular assemblage or deposit. Because retouched tools areoften components of the transported toolkit, assemblages made byhighly mobile groups that stay at a given locality for only a short timetend to have relatively more retouched tools than sites occupied bygroups that stay longer at a particular place on the landscape. In thelatter case, with increased site occupation duration, the number of re-touched tools becomes swamped by the more numerous flakes, flakefragments, and other debris associated with on-site tool production andupkeep.

Here we estimate occupation duration and intensity at GC2 bycalculating average artifact discard rates, measured by dividing artifacttotals by estimated time span, inferred from available radiometric andstratigraphic data. We also calculate the ratios of retouched tools tolithic artifact volumetric densities, as described by Riel-Salvatore andBarton (2004) and by Barton and Riel-Salvatore (2014). Age spansfollow those provided by Ambrose (1984). Artifact abundances werecalculated using retouched tool counts reported by Ambrose (1984) andpreviously unpublished data on the total number of artifacts recoveredfrom the 1964 Isaac and Clarke excavation at GC2 (collected in 1978 byCharles Nelson and provided by him to us in 2017). Our obsidiansourcing data focus on an available sample excavated by Leakey. Weuse abundance data from the Isaac and Clarke excavations because theirimproved excavation and recording techniques reduce the sample

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

478

biases due to the loss of smaller debris. Unfortunately, lithics from theIsaac and Clarke excavation appear to have been lost (Wilshaw, 2012)and, thus, are no longer available for study.

4.2. Analyzed GC2 assemblage

In 1935–1936, the Peabody Museum acquired 241 GC2 obsidianartifacts from L.S.B. Leakey, at the request of then-Director HughHencken, who sought a “strictly typical series representing the variousancient cultures of Kenya” (PMAE Accession File 36-6). All but two ofthe artifacts still had Leakey's attribution to Upper Kenyan AurignacianPhase a or b, which today correspond to Eburran Phases 3 and 4, re-spectively (Fig. 3). We sourced 239 obsidian artifacts: 121 (51%) fromlower deposits (Eburran Phase 3) and 118 (49%) from the upper de-posits (Eburran Phase 4). Thus, the assemblage is nearly evenly splitbetween the phases, with only 2.5% relative difference. The totalmasses of the artifacts are also similar: 434.3 g in Phase 3 and 481.4 g inPhase 4 – a 10% difference (Table 1). In addition, the proportions ofartifact types identified in the phases, while not identical, are suffi-ciently similar to allow their comparison for our purposes (Table 2).

The Peabody's sample consists largely of retouched artifacts(n=203), with fewer cores (n=31, including burins) and 16 un-retouched pieces. It is clearly biased, attested by the near-absence of

unretouched flaking debris. However, retouched tools are often con-sidered to be among the most portable elements of the toolkit trans-ported by foragers. For example, Eerkens et al. (2007) report that,among hunter-gatherer populations of the North American West, ob-sidian retouched pieces typically exhibit greater source diversity and,

Fig. 3. Examples of sourced obsidian tools from the Eburran Phase 3 (left) and Phase 4 (right) deposits, showing that these tools, principally backed blades and lunates, decreased in sizeover time (backed blades shrank from 34 ± 11mm to 27 ± 8mm in length; Ambrose, 2002).

Table 1Summary of the obsidian sources for GC2 artifacts from the Eburran Phases 3 and 4 deposits. Additional details are available in the supplementary online materials.

Distance Elevation Eburran Phase 3 Eburran Phase 4

Obsidian source (km) (m asl) n % mass (g) % n % mass (g) %

Mt. Eburru outcrops 20 ∼2590 105 87% 368.0 85% 93 79% 433.2 90%Mundui/Sonanchi 30 ∼1950 11 9.1% 48.2 11% 15 13% 28.7 6.0%Baixia Estate/Ilkek 30 ∼2030 3 2.5% 14.0 3.2% 9 7.6% 19.0 3.9%Masai Gorge Rockshelter 30 ∼2050 1 0.8% 0.5 0.1%Menengai 35 ∼2020 1 0.8% 0.8 0.2%Hell's Gate 1/Ololbutot 1 40 ∼2100 1 0.8% 3.3 0.8%Total 121 434.3 118 481.4

Table 2Summary of the typological classification of GC2 artifacts from the Eburran Phases 3 and4 deposits. Additional details are available in the supplementary online materials.

Eburran Phase 3 Eburran Phase 4

Simplified types n % n %

Backed piece, complete 62 51% 54 46%Backed piece, fragment 31 26% 21 18%Core 11 9% 20 17%Scraper 8 7% 11 9%Core edge maintenance flake 6 5% 7 6%Blade, outrepasse 2 2%Flake fragment 1 1%Retouched blade, complete 1 1%Retouched blade, fragment 2 2%Misc. retouched piece 1 1% 1 1%Total 121 118

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

479

on average, are found farther from their sources, whereas unmodifiedflakes tend to reflect fewer, closer sources. Similar trends have alsobeen noted by Conard and Adler (1997), Turq et al. (2013), and othersamong Middle Paleolithic assemblages across Western Europe. Re-touched or “finished” tools, particularly backed microliths (like thosethat dominate Eburran assemblages) may also be more suited for thepurposes of exchange between groups (Ambrose, 2002). Despite itsshortcomings, our sample is therefore appropriate to explore obsidiansource and landscape use in the early Holocene of the Central KenyanRift Valley.

Hivernel (1974) studied all of the remaining material known fromLeakey's excavations of Eburran Phases 3 and 4 at GC2, stored in var-ious museums across the U.K. and Kenya. She reports a total of 1551artifacts from Phase 3 and 1982 artifacts from Phase 4. Therefore, ouranalysis of 239 artifacts represents a ~5–9% sample of the knownEburran Phases 3 and 4 material excavated and retained by Leakey. Interms of absolute (rather than relative) size, our sample contains thelargest number of obsidian artifacts with geochemical data and sourceattributions currently published for any Holocene LSA site in Kenya (cf.Merrick and Brown, 1984a, 1984b; Merrick et al., 1990; Nash et al.,2011; Ndiema et al., 2011).

4.3. Instrument and settings

We used a Thermo Scientific Niton XL3t GOLDD+ analyzer in thisstudy. It is outfitted with a 2-W, Ag-anode tube to produce an incidentX-ray beam. The tube's voltage and current change in combination withdifferent X-ray filters to optimally fluoresce elements in different por-tions of the periodic table. The elements of primary interest – the so-called “mid-Z” elements – were measured for 30–45 s using the “main”X-ray filter with a tube voltage of 40 kV and current of ≤50 μA. Thismodel is also equipped with a 25-mm2 silicon drift detector that has aresolution ≤155 eV. The X-ray beam has a diameter of ∼8-mm. Theinternal camera facilitated positioning the artifacts over the beam, anda test stand was used for the most stable measurement conditions.

4.4. Correction and calibration

Measured X-rays must be “corrected” for various physical phe-nomena in a specimen (X-ray absorption and attenuation, secondaryand tertiary X-ray fluorescence, photoelectric emission, and so on). Weused the fundamental parameters (FP) approach, which has been de-monstrated to yield greater accuracy than empirical methods(Heginbotham et al., 2010). The instrument has a factory-set calibrationtested with certified reference materials (CRMs) principally from theUnited States' National Institute of Standards and Technology. We “fine-tuned” our data with a custom calibration tailored to the datasets ofBrown et al. (2013). Specifically, to maximize compatibility with Brownet al. (2013), we used 27 specimens originally analyzed in their study.As documented in Frahm et al. (2017), twelve of the specimens wereused as calibration standards, and the other fifteen served as secondarystandards to evaluate our custom calibration and show how well thecalibrated pXRF data replicates values in Brown et al. (2013). As shownin Fig. 4, pXRF data for the mid-Z elements are highly correlated totheir wavelength-dispersive XRF (WDXRF) data. After calibration,slopes of the best-fit lines nearly equal 1. As established in Frahm et al.(2017), these elements exhibit only small differences (2–4% relative)compared to the data in Brown et al. (2013), and our calibration allowsdirect compatibility between our pXRF data and those in Brown et al.(2013).

4.5. Procedures for small artifacts

Many of the GC2 artifacts are small, a known challenge for XRFtechniques (e.g., Davis et al., 1998; Eerkens et al., 2007; Shackley,2011, 2012; Freund, 2014; Escola et al., 2016). Consequently,

archaeologists typically exclude small artifacts from pXRF-based sour-cing studies (e.g., Sheppard et al., 2011; Golitko, 2011; Goodale et al.,2012; Kellett et al., 2013; Galipaud et al., 2014; Coffman and Rasic,2015; Millhauser et al., 2015). Frahm (2016), though, developed andtested two methods for pXRF-based sourcing of small obsidian artifactsusing well-measured mid-Z elements with similar X-ray energies. Weused one of these techniques to minimize errors associated with XRF ofsmall lithic size classes. In short, elements such as Rb, Sr, Y, Zr, and Nbare well measured using XRF and, due to their similar characteristic X-ray energies, are affected by small specimen sizes in the same way,meaning that ratios between them can largely cancel out the size ef-fects. Sr occurs at very low concentrations in most Kenyan obsidians, soits utility for discerning sources is small. Therefore, we base our sourceidentifications on ratios of Zr, Y, and Nb to Rb. As a result, theseidentifications are based on four independent variables while size ef-fects have been minimized.

Fig. 5 is a scatterplot of Zr/Rb, Y/Rb, and Nb/Rb. Note that, for fourof the six identified obsidian sources, we plot not only the WDXRFvalues of Brown et al. (2013) but also our pXRF data for geologicalspecimens from the same sources. We also plot their WDXRF data andour pXRF data for seven additional sources that were not identifiedamong the GC2 artifacts. Analyzing geological specimens offers an extracheck on our ability to reproduce the values in Brown et al. (2013). Thepublished and our newly measured values directly overlap in thisscatterplot, demonstrating how well our measurements duplicate thosefrom Brown et al. (2013). Thus, we have high confidence in our iden-tifications, and all of these data are included in the supplementarymaterial.

5. Results

Here we document our estimates of occupation intensity as well asthe identified obsidian sources and differences in their exploitationbetween the two site phases. It is worth noting that Central Rift Valleyobsidian sources principally occur as discrete primary outcrops – moreoften as localized flows, less often as blocks in welded tuffs and ig-nimbrites – rather than as secondary deposits of transported nodules(Ambrose, 2012). There is, of course, variability in the surface expres-sions of these sources: some are obsidian flows that extend ∼2000m,while others are exposures< 200m2. On occasion, where obsidian isexposed on a hillside, there is a colluvial scatter across the slope. Theregional topography, however, limits the alluvial transport of obsidian(i.e., blocks cannot escape closed basins without outlets via gravity orwater transport). Particularly during the AHP, when the lakes greatlyexpanded, secondary deposits more than a few kilometers from theprimary sources would have been underwater. Secondary obsidian de-posits at the base of Mt. Eburru attest to this. Given the lack of op-portunities for obsidians to be spread over large areas by natural pro-cesses, we have confidence in the collection locations.

5.1. Archaeological measures of occupation duration and intensity

Table 3 synthesizes artifact abundance and accumulation data forGC2, based on material from the 1964 Isaac and Clarke excavations andon age estimates from inferred sedimentation rates calculated byAmbrose (1984). In this sample, the proportions of retouched tools andtotal lithic debris differ significantly between the Eburran Phase 3 andPhase 4 deposits (χ2= 101.24, p < 0.01). These data show that theEburran 3 strata (spits #1–13) have lower numbers of retouched tools(n=322) but a higher artifact volumetric density (2953/m3), whereasEburran 4 strata (spits #14–37) show more retouched tools (n=407)but a lower artifact volumetric density (1034 artifacts/m3). Therefore,the retouched tool-to-artifact volumetric density ratio is 0.1 for theEburran Phase 3 strata and 0.4 for the Eburran 4 strata. Although thesenumbers should be regarded with caution because of the many as-sumptions underlying them, estimated artifact discard rates are nearly

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

480

twice as high in the Eburran Phase 3 levels (∼6.2 artifacts/year) as inEburran Phase 4 (∼3.2 artifacts/year), implying that more intensiveoccupation occurred earlier rather than later in the site's history.

These findings suggest lower intensity occupations (i.e., fewer ar-tifacts discarded per year) and increased mobility (i.e., relatively fewerretouched tools) during Eburran Phase 4 relative to Eburran Phase 3 atGC2. Such results do not support Ambrose's (1984) original inter-pretation of GC2's occupation history (our Hypothesis #2) but dosupport interpretations that drier, more open habitats associated withEburran Phase 4 structured settlement strategies (Hypothesis #1). Wewould expect greater occupation intensity during Eburran Phase 3 tocorrelate with increased lithic reduction intensity. Testing this hy-pothesis, however, is difficult given the available sample, which isdominated by backed pieces that are replaced, not resharpened, duringtheir use (Hiscock, 2006). Cores are rare and mostly (n=30; 97%)made on relatively thin blade segments that have little potential forextensive reduction prior to discard. While Goldstein (2014) provides auseful approach to estimating the extent of endscraper reduction, manyof the GC2 artifacts were made on blade blanks that were snapped,perhaps deliberately, so comparisons of endscraper size at discard toreconstructed original blade length provides a poor estimate of reduc-tion intensity.

Obsidian source identification provides a means to assess the shapeand size of the territory used by Eburran Phase 3 and 4 foragers. For themore intensive Eburran Phase 3 occupations, we predict increased

reliance on local sources and use of a greater diversity of sources than inPhase 4. Toolstone from these diverse obsidian sources may have beenacquired either by direct access or through contact with neighboringgroups.

5.2. Identified obsidian sources

Six obsidian sources are reflected among the artifacts (Figs. 6 to 8).The most abundant source is also the closest: Mt. Eburru, ~20 km fromGC2, is the source for 79–87% of the artifacts in the two phases. Thereare multiple primary outcrops and secondary deposits on the north-eastern slope of Eburru (Brown et al., 2013), including a Late Holoceneobsidian quarry (i.e., GsJj50) discussed by Goldstein and Munyiri(2017). At this quarry, which lies at ∼2570m asl, excellent qualityobsidian crops out as very large blocks, and there is a sizable debrisscatter from reducing them (Brown et al., 2013; Goldstein and Munyiri,2017). Above this quarry, up to elevations of ∼2600m asl, there areadditional outcrops, some with quarrying debris (Brown et al., 2013).Below it, as far as∼4–5 km to the northeast, there are a few outcrops oflow-quality obsidian as well as secondary deposits, specifically water-rolled, -rounded, and -abraded cobbles within lacustrine sediments(Brown et al., 2013). The obsidian textures that occur in the GC2 toolsappear more consistent with the high-elevation quarrying sites (i.e.,“quality is excellent,” Brown et al., 2013) than low-elevation deposits(e.g., “granular,” “flow banding and small phenocrysts are common,”

R = 0.986

500

1000

1500

2000

2500

3000

3500

500 1500 2500 3500

WD

XR

F (

Brow

n e

t al. 2

013)

pXRF (ppm, calibrated)

Zr

R = 0.988

100

200

300

400

100 200 300 400

WD

XR

F (

Brow

n e

t al. 2

013)

pXRF (ppm, calibrated)

Rb

R = 0.969

200

300

400

500

600

200 300 400 500 600

WD

XR

F (

Brow

n e

t al. 2

013)

pXRF (ppm, calibrated)

Nb

R = 0.994

0

100

200

300

0 100 200 300

WD

XR

F (

Brow

n e

t al. 2

013)

pXRF (ppm, calibrated)

Y

Fig. 4. Our pXRF data for the elements of interest are highly correlated to the WDXRF data of Brown et al. (2013), and after calibration, slopes of the best-fit lines nearly equal 1 (dottedlines). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

481

“opaque green with a granular texture,” Brown et al., 2013). Given thisand that there is evidence of extensive quarrying at the outcrops, ourworking hypothesis is that the GC2 artifacts are most likely to reflectacquisition from these quarrying sites or similar ones.

Five obsidian sources are supplementary to Mt. Eburru, all ≥30 kmfrom GC2. The “Mundui/Sonanchi” source has two chemically indis-tinguishable exposures ~3 km apart (Brown et al., 2013). The “MunduiEstate” exposure is a ~40-m-long ridge, partly covered by lacustrinesediments, that contains fair-quality obsidian as loose, “football-sizedblocks… with numerous pumice inclusions” (Brown et al., 2013). The“Lake Sonanchi Crater” exposure, however, has excellent-quality ob-sidian blocks, occasionally half a meter in size, accessible within thecrater wall (Brown et al., 2013). The “Baixia Estate/Ilkek” sourceconsists of two obsidian exposures, approximately ∼9 km apart, thatcannot be elementally distinguished. The “Baixia Estate” facies has“very high quality obsidian” exposed in a road cut into a quarry, inwhich volcanic tuff containing obsidian blocks was extracted, and thisgeological unit extends at least several hundred meters (Brown et al.,

2013). In contrast, the “Ilkek” exposure is an isolated outcrop of pre-dominately low-quality obsidian in small nodules (Brown et al., 2013).Thus, we hypothesize that “Baixia Estate” represents the origin of thisobsidian variety. The “Masai Gorge Rockshelter” source is a chill zone(i.e., where the lava rapidly cooled via contact with the existing rock)exposed in a cliff face (Brown et al., 2013). The “Menengai” obsidiansource is exposed in a caldera wall, where fair-quality obsidian seamsoccur throughout pumiceous layers (Brown et al., 2013). Lastly, the“Hell's Gate 1/Ololbutot 1” obsidian source has two elementally in-distinguishable exposures ~6 km apart. The “Hell's Gate 1” exposure isa chill zone in a cliff wall, and the observed material was “very poorobsidian for artifact manufacture, having many inclusions” (Brownet al., 2013). The “Ololbutot 1” facies, however, consists of large blocks,occasionally a meter in diameter, some of which are excellent quality(Brown et al., 2013). Consequently, we presume that “Ololbutot 1” wasthe exposure exploited in antiquity. Note that, for both the Baixia Es-tate/Ilkek and the Hell's Gate 1/Ololbutot 1 obsidian varieties, thehigher-quality obsidian exposures are marginally closer to GC2,meaning that we are also using the more conservative geographic ori-gins.

5.3. Diachronic view

Within our sample, both phases are dominated (~79–87%) by theclosest obsidian source (Mt. Eburru), which lies at or near the margin(~20 km) of what is often considered the daily range for foragers (e.g.,Surovell, 2009). In both phases, almost all of the artifacts (∼98–99%)came from the same three obsidian sources: Mt. Eburru, Mundui/So-nanchi, and Baixia Estate/Ilkek, all ~30 km or less from GC2. Each ofthe other three sources that are ≥30 km away – Masai Gorge Rock-shelter, Menengai, and Hell's Gate 1/Ololbutot 1 – is reflected by asingle artifact. Of these three sources, the farther two only occur inPhase 3, and the nearest one only occurs in Phase 4. However, given theabundance of Mt. Eburru obsidian (20 km away), the mean source-to-site distances in both phases are essentially the same: 21.4 ± 3.8 km inEburran Phase 3 and 22.1 ± 4.1 km in Eburran Phase 4. Note thatthese values use linear distances between GC2 and the sources. An

Fig. 5. Three-dimensional scatterplot of Zr/Rb, Y/Rb, and Nb/Rb for the GC2 artifacts and a variety of Kenyan obsidian sources, including published data and our new pXRF analyses.Table 1 summarizes the results. All data plotted here are available in the supplementary online materials.

Table 3Summary of artifact composition, density, and accumulation rate for GC2 based on the1964 Isaac and Clarke excavations. Artifact totals from Ambrose (1984) and Nelson(personal communication, 2017), and age spans from Ambrose (1984).

Eburran 3 Eburran 4

Spit numbers 1–13 14–37Excavated area (m2) 2 2Excavated thickness (m) 1.3 2.3Excavated volume (m3) 2.6 4.6Minimum age (yr BP) 7266 5784Maximum age (yr BP) 8510 7266Age span (yr) 1244 1482Artifacts/year 6.2 3.2Retouched tools (n) 322 407Debris (n) 7357 4351Artifact total (n) 7679 4758Retouched tools (%) 4.2 8.6Artifacts volumetric density (artifacts/m3) 2953 1034Retouched tools: artifact volumetric density 0.1 0.4

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

482

actual path on the ground between this site and each source would belonger. Even without Mt. Eburru, the difference between the phasesoverlaps within a standard deviation: 30.9 ± 2.7 km during EburranPhase 3 and 30 km (i.e., these sources are roughly equidistant) duringEburran Phase 4. Still, the two most distant obsidian sources – Me-nengai (35 km away) and Hell's Gate 1/Ololbutot 1 (40 km away) – areonly found among artifacts from the earlier Eburran Phase 3.

Based on our estimates of artifact discard rates and retouched toolfrequencies, we expect the more intensive Eburran 3 occupations tosample a more diverse array of obsidian sources, as the diversity oftoolstone tends to increase as greater portions of the landscape are usedduring more prolonged occupations. We also expect the Eburran 3 oc-cupations to show increased reliance on local (> 20 km distant)sources. This is, in part, due to the inferred greater occupation intensity

as well as the drier condition during Eburran Phase 4, as suggested bythe fauna, which may have compelled larger home ranges and, thus,more interactions with distant areas. Neither expectation is stronglysupported by our available data. Artifacts from Phase 3 originate frommore sources (n=5) relative to Phase 4 (n=4); however, this does notsignificantly differ from an even or random distribution between thetwo phases (χ2= 0.45, p=0.83). Similarly, more of the Eburran 4sample originated from non-local obsidian sources (21.2%) relative tothe Eburran 3 sample (13.2%), but this difference is not supported atthe p < 0.05 level (χ2= 2.67, p=0.10).

We note also that that there is no significant difference in the ele-vation of the sources used between Phases 3 and 4. Overall, the meanelevations for these two phases are similar, with means overlapping atone standard deviation: 2509 ± 209m asl in Eburran Phase 3 and

2%9%

87%

Mt. Eburru outcropsMundui/SonanchiBaixia Estate/IlkekMasai Gorge RockshelterMenengaiHell's Gate 1/Ololbutot 1

8%

13%

79%

Eburran Phase 3 Eburran Phase 4

By number

3%11%

85%

4%6%

90%

By mass

Fig. 6. Proportions of the obsidian sources identified in the Eburran Phases 3 and 4 deposits, both by number of artifacts and by mass.

Mt. Eburru outcrops 93 49 27 13 10 2 1 1 1 1

Mundui/Sonanchi 15 2 4 4 1

Baixia Estate/Ilkek 7 2 2 1

Masai Gorge Rockshelter 1

Menengai 1

Hell's Gate 1/Ololbutot 1 1

Bac

ked

piec

e, c

ompl

ete

Bac

ked

piec

e, fr

agm

ent

Cor

e

Scr

aper

Cor

e ed

ge m

aint

enan

ce fl

ake

Bla

de, o

utre

pass

e

Fla

ke fr

agm

ent

Ret

ouch

ed b

lade

, com

plet

e

Ret

ouch

ed b

lade

, fra

gmen

t

Mis

c. r

etou

ched

pie

ce

Ob

sid

ian

so

urc

e

Simplified types

Fig. 7. Number of GC2 obsidian artifacts classified both by lithic type and by source.

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

483

2461 ± 250 asl in Eburran Phase 4. These mean values are skewed bythe abundance of Mt. Eburru obsidian at 2590m asl. Excluding Mt.Eburru from the calculations results in lower mean altitudes:1979 ± 47m in Eburran Phase 3 and 1983 ± 41 km in Eburran Phase4. These elevation ranges, however, clearly overlap within a singlestandard deviation. It is also worth noting that both the highest ob-sidian source (Mt. Eburru at ∼2590m) and lowest source (Mundui/Sonanchi at ∼1950m) were the two most commonly exploited, at leastbased on the Peabody Museum's sample.

6. Discussion

Paleoenvironmental proxies from GC2 and neighboring lakes in-dicate a habitat shift from a more wooded, closed nearshore settingduring Eburran Phase 3 to a more open, grassy, and drier setting duringEburran Phase 4. The results here suggest that this change in habitatwas paralleled by a decline in occupation intensity over time. Given theapparent preference for occupations near the forest-savanna ecotone byEburran foragers, the GC2 pattern can reasonably be interpreted as achange in the site's position or role within a broader settlementstrategy, shifting from an intensely used site near the ecotone boundaryto one farther from the boundary and, hence, used more rarely (cf.Ambrose, 1986; Ambrose and Sikes, 1991; Ambrose, 2001). We wouldexpect this change to be reflected in the types of lithic resources usedgiven that a number of chemically distinct obsidian outcrops occurwithin the vicinity of the site and would have been available to its pastoccupants. However, despite being able to attribute all 239 GC2 arti-facts in our sample to their sources, our results indicate no significantdifferences between Eburran Phases 3 and 4 in terms of the diversity ofsources used, use of non-local sources, or altitude of the outcrops used.In short, changes in the settlement pattern appear to have had nomeasurable impact on the types of toolstone being used, at least withinthe biased sample that Leakey sent to the Peabody Museum.

The lack of temporal differences in obsidian sources may indicatethe resilience of Eburran foraging strategies and the ability of a culturalpreference for particular types of stone to overcome changes within thelocal environment that would have impacted its availability. Whilespeculative, the persistent use Mt. Eburru obsidian observed in EburranPhases 3 and 4 at GC2 may indicate the early stages of a wider phe-nomenon of the preferred use of this lithic material at many Kenyan andTanzanian sites, a pattern that reached its apogee with Elmenteitanpastoralist groups. Mt. Eburru obsidian is the dominant lithic materialat Elmenteitan sites regardless of source distance (at times ~200 kmaway), suggesting regular mechanisms for the supply and movement ofthis toolstone, to the near-exclusion of contemporary “Savanna PastoralNeolithic” groups, who relied instead on the obsidian sources south of

Lake Naivasha ~10–20 km from Mt. Eburru (see Goldstein and Munyiri,2017 for review). This pattern has been interpreted as an example ofcontrolled access to specific resources or landscape facets via socialand/or physical means (cf. Robertshaw, 1990; Ambrose, 2012;Goldstein and Munyiri, 2017). The antiquity of this pattern is difficultto discern, particularly as the exploited obsidian sources for EburranPhase 5 assemblages (which precede the Elmenteitan strata) have notyet been determined. Limited support for this speculative scenario maybe found in the Lake Victoria basin ~200 km away, where Mt. Eburruobsidian accounts for about 25% of the obsidian during the early Ho-locene (when only foragers are present) but 53% of the late Holoceneassemblages made by foragers and 67% of the studied pastoralist El-menteitan sites (Merrick and Brown, 1984a, 1984b; Frahm et al., 2017).These data may suggest either (a) increased use over time of Mt. Eburruobsidians by foragers in the Lake Victoria basin or (b) obsidian acqui-sition by late Holocene foragers through some form of interaction withElmenteitan groups.

7. Conclusions

The Naivasha-Nakuru Basin of Kenya's Central Rift Valley is a regionthat is not only highly sensitive to climatic changes but also one of theworld's most obsidian-rich landscapes, allowing us to investigate therelationship between environmental change and behavioral adaptationsover the course of the Holocene. The effect of these changes on EburranLSA hunter-gatherer populations has been one such focus of research inpast studies, in particular those of Ambrose and colleagues (e.g.,Ambrose, 1984, 1986; Ambrose and Sikes, 1991). GC2 serves as thetype site for the early to middle Holocene Eburran Phases 3 and 4,which likely sample an earlier, more humid interval and a later, drierone, respectively. Our analysis of occupation data for GC2 suggests agreater intensity during the more humid Eburran Phase 3, when thelake's shoreline was likely closer to the site than it is at present. As thelake retreated (and, with it, the site's position at an important ecotone),occupation intensity declined, as measured by both the lower frequencyof retouched tools and by the reduced artifact discard rates in theEburran Phase 4 deposits.

Using pXRF, we analyzed 239 obsidian artifacts from GC2 housed atHarvard's Peabody Museum of Archaeology and Ethnology and attrib-uted them to six obsidian sources using a WDXRF database published byBrown et al. (2013). Comparison between the two phases suggest thatthe site's occupation history or local environment had no discernableeffect – at least within this sample – on the variety or range of obsidiansources used for tool manufacture, with the closest source (Mt. Eburruat ~20 km away) consistently used to meet the supply demands forlithic raw material. This implies that the demand for this material

5%7%

9%

26%

51%

6%

9%

17%

18%

46%

Backed piece, completeBacked piece, fragmentCoreScraper

Others

Eburran Phase 3 Eburran Phase 4

Fig. 8. Proportions of the lithic types identified in the Eburran Phases 3 and 4 deposits.

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

484

outweighed any changes in the difficulty of its acquisition by GC2 oc-cupants. Speculatively, the GC2 data might also mark the beginning ofthe widespread use of Mt. Eburru obsidian by groups of foragersthroughout central and southern Kenya and northern Tanzania. Testingsuch a hypothesis requires, in part, a new, more extensive program todetermine the obsidian sources exploited by Eburran foragers. Our re-sults here demonstrate the potential for such analyses and underscorethe amount of work that remains to be done.

Acknowledgements

Frank Brown passed away while this manuscript was under review,and given the importance of his decades of work to the success of thisproject, we wish to dedicate this paper to his memory. Harry Merrickalso generously contributed the collection of geological obsidian spe-cimens used for our calibration to Brown et al. (2013). Viva Fisher(Senior Registrar), Kara Schneiderman (Director of Collections), DianaLoren (Director of Academic Partnerships) and Emily Pierce Rose(Curatorial Assistant for Academic Partnerships) facilitated access tothe artifacts in the Peabody's collections. The Peabody's artifacts were agift of the East African Archaeological Expedition via Dr. L.S.B. Leakey,F.S.A., 1936. Charles Nelson generously provided his unpublished dataon the total number of artifacts recovered from the 1964 Isaac andClarke excavations. Ravid Ekshtain provided valuable research supportfor the pXRF and lithic analyses, and she was key in getting this projectstarted. Support was provided by the American School of PrehistoricResearch, Harvard University and the University of Minnesota's De-partment of Anthropology, Department of Earth Sciences, and Institutefor Rock Magnetism. The pXRF instrument utilized for this study is partof the research infrastructure of the University of Minnesota's Institutefor Rock Magnetism (IRM). Funding for the instrument came, in part,from the University of Minnesota's Grant-in-Aid of Research, Artistry,and Scholarship (GIA) Program, awarded to Joshua Feinberg, GilbertTostevin, and Kyungsoo Yoo.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jasrep.2018.01.042.

References

Ambrose, S.H., 1984. Holocene Environments and Human Adaptations in the Central RiftValley, Kenya (PhD). University of California.

Ambrose, S.H., 1986. Hunter-gatherer adaptations to non-marginal environments: anecological and archaeological assessment of the dorobo model. Sprache undGeschichte in Afrika 7, 11–42.

Ambrose, S.H., 2001. Middle and Later Stone Age settlement patterns in the central RiftValley, Kenya: Comparisons and contrasts. In: Conard, N. (Ed.), Settlement Dynamicsof the Middle Paleolithic and Middle Stone Age. Kerns Verlag, Tübingen, Germany,pp. 21–43.

Ambrose, S.H., 2002. Small things remembered: Origins of early microlithic industries inSubsaharan Africa. In: Elston, R., Kuhn, S. (Eds.), Thinking Small: Global Perspectiveson Microlithic Technologies. Archaeological Papers of the American AnthropologicalAssociation, No. 12, pp. 9–29.

Ambrose, S.H., 2012. Obsidian dating and source exploitation studies in Africa: im-plications for the evolution of human behavior. In: Liritzis, I., Stevenson, C. (Eds.),Obsidian and Ancient Manufactured Glasses. University of New Mexico Press,Albuquerque, pp. 56–72.

Ambrose, S.H., DeNiro, M.J., 1989. Climate and habitat reconstruction using stablecarbon and nitrogen isotope ratios of collagen in prehistoric herbivore teeth fromKenya. Quat. Res. 407–422.

Ambrose, S.H., Lorenz, K.G., 1990. Social and ecological models for the Middle Stone Agein southern Africa. In: Mellars, P. (Ed.), The Human Revolution: Behavioral andBiological Perspectives in the Origins of Modern Humans. Edinburgh UniversityPress, pp. 3–33.

Ambrose, S.H., Sikes, N.E., 1991. Soil carbon isotope evidence for Holocene habitatchange in the Kenya Rift valley. Science 253, 1402–1405.

Ambrose, S.H., Hivernel, F., Nelson, C.M., 1980. The taxonomic status of the KenyaCapsian. In: Leakey, R.E.F., Ogot, B.A. (Eds.), Proceedings of the 8th PanafricanCongress of Prehistory and Quaternary Studies, 1977. The International Louis LeakeyMemorial Institute for African Prehistory, Nairobi, pp. 248–252.

Ambrose, S.H., Ferguson, J., Glascock, M., Slater, P., 2012a. Chemical fingerprinting of

Kenyan obsidian sources and Late Quaternary artifacts with NAA and XRF. In:Presented at the Annual Meeting of the Society for American Archaeology, Memphis,Tennessee.

Ambrose, S.H., Slater, P., Ferguson, J., Glascock, M., Steele, I., 2012b. Obsidian sourcesurvey and Late Quaternary artifact sourcing in Kenya: implications for early modernhuman behaviour. In: Presented at the Annual Meeting of the PalaeoanthropologySociety, Memphis, Tennessee.

Barton, C.M., Riel-Salvatore, J., 2014. The formation of lithic assemblages. J. Archaeol.Sci. 46, 334–352.

Bergner, A.G.N., Trauth, M.H., Bookhagen, B., 2003. Paleoprecipitation estimates for theLake Naivasha basin (Kenya) during the last 175 k.y. using a lake-balance model.Glob. Planet. Chang. 36, 117–136.

Bergner, A.G.N., Strecker, M.R., Trauth, M.H., Deino, A., Gasse, F., Blisniuk, P.,Dühnforth, M., 2009. Tectonic and climatic control on evolution of rift lakes in theCentral Kenya Rift, East Africa. Quat. Sci. Rev. 28, 2804–2816.

Blegen, N., 2017. The earliest long-distance obsidian transport: evidence from the∼200 ka Middle Stone Age Sibilo School Road Site, Baringo, Kenya. J. Hum. Evol.103, 1–19.

Blegen, N., Faith, J.T., Mant-Melville, A., Peppe, D.J., Tryon, C.A., 2017. The middle stoneage after 50,000 years ago: new evidence from the Late Pleistocene sediments of theEastern Lake Victoria Basin, Western Kenya. PaleoAnthropology 2017, 139–169.

Bonsall, C., Elenski, N., Ganecovski, G., Gurova, M., Ivanov, G., Slavchev, V., Zlateva-Uzanova, R., 2017. Investigating the provenance of obsidian from Neolithic andChalcolithic sites in Bulgaria. Antiquity 91 (356), e3.

Bower, J.R.F., Nelson, C.M., Waibel, A.F., Wandibba, S., 1977. The University ofMassachusetts' Later Stone Age/Pastoral ‘Neolithic’ comparative study in CentralKenya: an overview. Azania 12, 119–146.

Brown, F.H., Nash, B.P., Fernandez, D.P., Merrick, H.V., Thomas, R.J., 2013. Geochemicalcomposition of source obsidians from Kenya. J. Archaeol. Sci. 40, 3233–3251.

Butzer, K.W., 1972. Environment and Archaeology. Methuen & Co., London.Campbell, S., Healey, E., 2016. Multiple sources: the pXRF analysis of obsidian from

Kenan Tepe, S.E. Turkey. J. Archaeol. Sci. Rep. 10, 377–389.Coffman, S., Rasic, J.T., 2015. Rhyolite characterization and distribution in central

Alaska. J. Archaeol. Sci. 57, 142–157.Coleman, M.E., Ferguson, J.R., Glascock, M.D., Robertson, J.D., Ambrose, S.H., 2008. A

New Look at the Geochemistry of Obsidian From East Africa. 39. InternationalAssociation for Obsidian Studies Bulletin, pp. 11–14.

Coleman, M., Robertson, J.D., Ferguson, J.R., Glascock, M.D., Ambrose, S.H., 2009.Further studies into the geochemistry of obsidian from Kenya. In: Presented at theAnnual Meeting of the Society for American Archaeology, Atlanta, Georgia.

Conard, N.J., Adler, D.S, 1997. Lithic reduction and hominid behavior in the MiddlePaleolithic of the Rhineland. J. Anthropol. Res. 53 (2), 147–175.

Dale, D., Ashley, C.Z., 2010. Holocene hunter-fisher-gatherer communities: new per-spectives on Kansyore using communities of Western Kenya. Azania: Archaeol. Res.Afr. 45, 24–48.

Davis, M.K., Jackson, T.L., Shackley, M.S., Teague, T., Hampel, J.H., 1998. Factors af-fecting the energy-dispersive X-ray fluorescence (EDXRF) analysis of archaeologicalobsidian. In: Shackley, M.S. (Ed.), Archaeological Obsidian Studies: Method andTheory. Plenum, New York, pp. 159–180.

di Lernia, S., 2013. The emergence and spread of herding in northern Africa: A criticalreappraisal. In: Mitchell, P., Lane, P. (Eds.), The Oxford Handbook of AfricanArchaeology. Oxford University Press, Oxford, pp. 527–540.

Drake, B.L., Nazaroff, A.J., Prufer, K.M., 2009. Error assessment of portable X-ray flor-escence spectrometry in geochemical sourcing. SAS Bull. 32 (3), 14–17.

Dyson-Hudson, R., Smith, E.A., 1978. Human territoriality: an ecological reassessment.Am. Anthropol. 80, 21–41.

Eerkens, J., Ferguson, J., Glascock, M.D., Skinner, C., Waechter, S., 2007. Reductionstrategies and geochemical characterization of lithic assemblages: a comparison ofthree case studies from western North America. Am. Antiq. 72, 585–597.

Escola, P.S., Hocsman, S., Babot, M.P., 2016. Moving obsidian: the case of Antofagasta dela Sierra basin (Southern Argentinean Puna) during the late Middle and LateHolocene. Quat. Int. 422, 109–122.

Faith, J.T., Tryon, C.A., Peppe, D.J., Beverly, E.J., Blegen, N., Blumenthal, S., Chritz, K.L.,Driese, S.G., Patterson, D., 2015. Paleoenvironmental context of the Middle StoneAge record from Karungu, Lake Victoria Basin, Kenya, and its implications for humanand faunal dispersals in East Africa. J. Hum. Evol. 83, 28–45.

Ferguson, J.R., 2012. Chapter 12: X-ray fluorescence of obsidian: approaches to calibra-tion and the analysis of small samples. In: Shugar, A.N., Mass, J.L. (Eds.), HandheldXRF for Art and Archaeology. Leuven University Press, Leuven, pp. 401–422.

Frahm, E., 2014. Characterizing obsidian sources with portable XRF: accuracy, re-producibility, and field relationships in a case study from Armenia. J. Archaeol. Sci.49, 105–125.

Frahm, E., 2016. Can I get chips with that? Obsidian sourcing down to the microdebitagewith portable XRF. J. Archaeol. Sci. Rep. 9, 448–467.

Frahm, E., Feinberg, J.M., 2015. Reassessing obsidian field relationships at Glass Buttes,Oregon. J. Archaeol. Sci. Rep. 2, 654–665.

Frahm, E., Goldstein, S.T., Tryon, C.A., 2017. Late Holocene forager-fisher and pastoralistinteractions along the Lake Victoria shores, Kenya: perspectives from portable XRF ofobsidian artifacts. J. Archaeol. Sci. Rep. 11, 717–742.

Freund, K.P., 2014. Obsidian consumption in Chalcolithic Sardinia: a view from Bingia 'eMonti. J. Archaeol. Sci. 41, 242–250.

Galipaud, J.-C., Reepmeyer, C., Torrence, R., Kelloway, S., White, P., 2014. Long-distanceconnections in Vanuatu: new obsidian characterisations for the Makué site, AoreIsland. Archaeol. Ocean. 49, 110–116.

Gasse, F., 2000. Hydrological changes in the African tropics since the Last GlacialMaximum. Quat. Sci. Rev. 19, 189–211.

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

485

Goldstein, S.T., 2014. Quantifying endscraper reduction in the context of obsidian ex-change among early pastoralists in southwestern Kenya. Lithic Technol. 39, 3–19.

Goldstein, S.T., Munyiri, J.M., 2017. The Elmenteitan obsidian quarry (GsJj50): newperspectives on obsidian access and exchange during the pastoral Neolithic inSouthern Kenya. Afr. Archaeol. Rev. 34, 43–73.

Golitko, M., 2011. Provenience investigations of ceramic and obsidian samples using laserablation inductively coupled plasma mass spectrometry and portable X-ray fluores-cence. Fieldiana Anthropol. 42, 251–287.

Goodale, N., Bailey, D., Jones, G., Prescott, C., Scholz, E., Stagliano, N., Lewis, C., 2012.pXRF: a study of inter-instrument performance. J. Archaeol. Sci. 39, 875–883.

Hamilton, A.C., 1982. Environmental History of East Africa: A Study of the Quaternary.Academic Press, London.

Heginbotham, A., Bezur, A., Bouchard, M., Davis, J.M., Eremin, K., Frantz, J.H.,Glinsman, L., Hayek, L.-A., Hook, D., Kantarelou, V., Germanos Karydas, A., Lee, L.,Mass, J., Matsen, C., McCarthy, B., McGath, M., Shugar, A., Sirois, J., Smith, D.,Speakman, R.J., 2010. An evaluation of inter-laboratory reproducibility for quanti-tative XRF of historic copper alloys. In: Mardikian, P. (Ed.), Proceedings of theInternational Conference on Metal Conservation, Charleston, South Carolina, USA,October 11–15, 2010. Clemson University Press, pp. 244–255.

Hiscock, P., 2006. Blunt and to the point: changing technological strategies in HoloceneAustralia. In: Lilley, I. (Ed.), Archaeology of Oceania: Australia and the PacificIslands. Blackwell Publishing, Malden, MA, pp. 69–95.

Hivernel, F., 1974. A study of the Kenya Capsian of Gamble's Cave (Kenya) and its re-lationship with other industries. MA Thesis. University College, London.

Holl, A.F.C., 2005. Holocene “aquatic” adaptations in North tropical Africa. In: Stahl, A.B.(Ed.), African Archaeology. Blackwell Publishing, Malden, MA, pp. 174–186.

Hovers, E., 2009. The Lithic Assemblages of Qafzeh Cave. Oxford University Press.Isaac, G.Ll., 1972. Comparative studies of site locations in East Africa. In: Ucko, P.,

Tringham, R., Dimbelby, G.W. (Eds.), Man, Settlement, and Urbanism. Duckworth,London, pp. 165–176.

Kellett, L.C., Golitko, M., Bauer, B.S., 2013. A provenance study of archaeological ob-sidian from the Andahuaylas region of southern Peru. J. Archaeol. Sci. 40,1890–1902.

Kelly, R.L., 2013. The Lifeways of Hunter-Gatherers: the Foraging Spectrum, 2nd edition.Cambridge University Press.

Kuhn, S.L., 1991. "Unpacking" reduction: lithic raw material economy in the mousterianof West-Central Italy. J. Anthropol. Archaeol. 10, 76–106.

Kusimba, S.B., 2003. African Foragers: Environment, Technology, Interactions. AltamiraPress, Walnut Creek, CA.

Lane, P., 2004. The ‘moving frontier’ and the transition to food production in Kenya.Azania 39, 243–264.

Lane, P., 2013. The archaeology of pastoralism and stock-keeping in East Africa. In:Mitchell, P., Lane, P. (Eds.), The Oxford Handbook of African Archaeology. OxfordUniversity Press, Oxford, pp. 585–602.

Le Bourdonnec, F.-X., Bellot-Gurlet, L., Lugliè, C., Bressy-Leandri, C., 2015. Archéométriede l'obsidienne: déchiffrer la circulation d'une matière première. Nouv. Archeol.(138), 23–27.

Leakey, L.S.B., 1931. The Stone Age Cultures of Kenya Colony. Cambridge UniversityPress, Cambridge.

Leakey, L.S.B., 1971 [1931]. The Stone Age Cultures of Kenya Colony, Reprinted with aNew Introductory Note. Frank Cass and Company Limited, London (1971 [1931]).

Liritzis, I., Zacharias, N., 2011. Chapter 6: Portable XRF of Archaeological Artifacts:Current Research, Potentials and Limitations. In: Shackley, M.S. (Ed.), X-RayFluorescence Spectrometry (XRF) in Geoarchaeology. Springer, pp. 109–142.

Maitima, J.M., 1991. Vegetation response to climatic change in Central Rift Valley,Kenya. Quat. Res. 35, 234–245.

Marean, C.W., 1992. Hunter to herder: large mammal remains from the hunter-gathereroccupation at Enkapune ya Muto rock-shelter, Central Rift Valley, Kenya. Afr.Archaeol. Rev. 10, 65–127.

Marean, C.W., 2016. The transition to foraging for dense and predictable resources and itsimpact on the evolution of modern humans. Philos. Trans. R. Soc. B 20150239.

Marean, C.W., Mudida, N., Reed, K.E., 1994. Holocene paleoenvironmental change in theKenyan Central Rift as indicated by micromammals from Enkapune Ya MutoRockshelter. Quat. Res. 41, 376–389.

Mehlman, M.J., 1989. Late Quaternary Archaeological Sequences in Northern Tanzania(Ph.D. thesis). University of Illinois, Urbana.

Merrick, H.V., Brown, F.H., 1984a. Rapid chemical characterization of obsidian artifactsby electron microprobe analysis. Archaeometry 26 (2), 230–236.

Merrick, H.V., Brown, F.H., 1984b. Obsidian sources and patterns of source utilization inKenya and Northern Tanzania: some initial findings. Afr. Archaeol. Rev. 2, 129–152.

Merrick, H.V., Brown, F.H., Connelly, M., 1988. Sources of the obsidian at Ngamuriak andother Southwestern Kenyan sites. In: Robertshaw, P. (Ed.), Early Pastoralists of South-western Kenya. British Institute of Eastern Africa, Nairobi.

Merrick, H.V., Brown, F.H., Nash, W.P., 1994. Use and movement of obsidian in the Earlyand Middle Stone Ages of Kenya and Northern Tanzania. In: Childs, S.T. (Ed.),Society, Culture, and Technology in Africa. Masca Research Papers in Science andArchaeologypp. 29–44.

Milić, M., 2014. PXRF characterisation of obsidian from central Anatolia, the Aegean andcentral Europe. J. Archaeol. Sci. 41, 285–296.

Millhauser, J.K., Fargher, L.F., Heredia Espinoza, V.Y., Blanton, R.E., 2015. The geopo-litics of obsidian supply in Postclassic Tlaxcallan: a portable X-ray fluorescence study.J. Archaeol. Sci. 58, 133–146.

Nash, B.P., Merrick, H.V., Brown, F.H., 2011. Obsidian types from Holocene sites aroundLake Turkana, and other localities in northern Kenya. J. Archaeol. Sci. 38 (6),1371–1376.

Ndiema, K.E., Dillian, C.D., Braun, D.R., Harris, J.W.K., Kiura, P.W., 2011. Transport andsubsistence patterns at the transition to pastoralism, Koobi Fora, Kenya.Archaeometry 53, 1085–1098.

Nelson, C.M, 1973. A Comparative Analysis of Twenty-Nine Later Stone Age Occurrencesin East Africa. Ph.D. Dissertation. Department of Anthropology, University ofCalifornia at Berkeley.

Newcomer, M.H., Hivernel-Guerre, F., 1974. Nucleus Sur Eclat: technologie et utilizationpar differentes cultures prehistoriques. Bull. Soc. Prehist. Fr. 71, 119–128.

Newlander, K., Goodale, N., Jones, G.T., Bailey, D.G., 2015. Empirical study of the effectof count time on the precision and accuracy of pXRF data. J. Archaeol. Sci. Rep. 3,534–548.

Orange, M., Le Bourdonnec, F.-X., Bellot-Gurlet, L., Luglie, C., Dubernet, S., Bressy-Leandri, C., Scheffers, A., Joannes-Boyau, R., 2017. On sourcing obsidian assem-blages from the Mediterranean area: analytical strategies for their exhaustive geo-chemical characterization. J. Archaeol. Sci. Rep. 12, 834–844.

Potts, P.J., West, M., 2008. Portable X-ray Fluorescence Spectrometry: Capabilities for InSitu Analysis. Royal Society of Chemistry.

Prendergast, M.E., 2010. Kansyore fisher-foragers and transitions to food production inEast Africa: the view fromWadh Lang'o, Nyanza Province, Western Kenya. Azania 45,83–111.

Prendergast, M.E., Mabulla, A.Z.P., Grillo, K.M., Broderick, L.G., Seitsonen, O., Gidna,A.O., Gifford-Gonzalez, D., 2013. Pastoral Neolithic sites on the southern MbuluPlateau, Tanzania. Azania: Archaeological Research in Africa 48, 498–520.

Protsch, R., 1978. The chronological position of Gamble's Cave II and Bromhead's Site(Elmenteita) of the Rift Valley, Kenya. J. Hum. Evol. 7, 101–109.

Richardson, J.L., Dussinger, R.A., 1986. Paleolimnology of mid-elevation lakes in theKenya Rift Valley. Hydrobiologia 143, 167–174.

Riel-Salvatore, J., Barton, C.M., 2004. Late pleistocene technology, economic behaviorand land-use dynamics in southern Italy. Am. Antiq. 69, 257–274.

Robertshaw, P., 1990. Early Pastoralists of South-western Kenya. British Institute inEastern Africa, Memoir 11, Nairobi.

Shackley, M.S., 2011. Chapter 2: an introduction to X-ray fluorescence (XRF) analysis inarchaeology. In: Shackley, M.S. (Ed.), X-Ray Fluorescence Spectrometry (XRF) inGeoarchaeology. Springer, pp. 7–44.

Shackley, M.S., 2012. Portable X-ray fluorescence spectrometry (pXRF): the good, thebad, and the ugly (online essay). In: Archaeology Southwest Magazine. Spring issue,pp. 26.

Sheppard, P., Irwin, G., Lin, S., McCaffrey, C., 2011. Characterization of New Zealandobsidian using PXRF. J. Archaeol. Sci. 38, 45–56.

Skoglund, P., Thompson, J.C., Prendergast, M.E., ... Pinhasi, R., Krause, J., Reich, D.,2017. Reconstructing prehistoric African population structure. Cell 171, 59–71.

Surovell, T.A., 2009. Toward a Behavioral Ecology of Lithic Technology: Cases fromPaleoindian Archaeology. University of Arizona Press.

Sutton, J.E.G., 1974. The aquatic civilization of Middle Africa. J. Afr. Hist. 15 (4),527–546.

Tierney, J.E., deMenocal, P.B., 2013. Abrupt shifts in Horn of Africa hydroclimate sincethe Last Glacial Maximum. Science 342, 843–846.

Trauth, M.H., Maslin, M.A., Deino, A.L., Junginger, A., Lesoloyia, M., Odada, E.O., Olago,D.O., Olaka, L.A., Stretcker, M.R., Tiedemann, R., 2010. Human evolution in avariable environment: the amplifier lakes of Eastern Africa. Quat. Sci. Rev. 29,2981–2988.

Turq, A., Roebroeks, W., Bourguignon, L., Faivre, J.-P., 2013. The fragmented characterof Middle Palaeolithic stone tool technology. J. Hum. Evol. 65, 641–655.

Washbourn-Kamau, C.K., 1971. Late Quaternary lakes in the Nakuru-Elmenteita basin,Kenya. Geogr. J. 137, 522–535.

Wiessner, P., 1982. Risk, reciprocity and social influences on !Kung San economics. In:Leacock, E., Lee, R. (Eds.), Politics and History in Band Societies. Cam-bridgeUniversity Press and Editions de la Maison des Sciences de l'Homme, Cambridge andParis, pp. 61–84.

Wiessner, P., 1983. Style and social information in Kalahari san projectile points. Am.Antiq. 48, 253–276.

Wilshaw, A., 2012. An Investigation into the LSA of the Nakuru-Naivasha Basin andsurround. In: Central Rift Valley, Kenya: Technological Classifications and PopulationConsiderations. Cambridge University (Ph.D. dissertation).

Wilshaw, A., 2016. The current status of the Kenya Capsian. Afr. Archaeol. Rev. 33,13–27.

Wright, D.K., Forman, S.L., Kiura, P., Bloszies, C., Beyin, A., 2015. Lakeside view: so-ciocultural responses to changing water levels of Lake Turkana, Kenya. Afr. Archaeol.Rev. 32, 335–367.

E. Frahm, C.A. Tryon Journal of Archaeological Science: Reports 18 (2018) 475–486

486