Journal of Human Evolution 59 (2010) 321e339 Contents lists available at ScienceDirect Journal of Human Evolution journal homepage: www.elsevier.com/locate/jhevol Taphonomic analysis of the Middle Stone Age faunal assemblage from Pinnacle Point Cave 13B, Western Cape, South Africaq Jessica C. Thompson School of Social Science, University of Queensland, Brisbane, QLD 4072, Australia a r t i c l e i n f o a b s t r a c t Article history: Received 26 May 2008 Accepted 2 July 2009 A detailed taphonomic analysis is provided for the mammalian and tortoise faunal assemblages from Pinnacle Point Cave 13B (PP13B). It is the first of several reports on the fauna from this site, and must necessarily precede analyses focused on higher level interpretations of Middle Stone Age (MSA) butchery, transport, and hunting behavior. The taphonomic work shows that the faunal assemblage is well preserved and there are discernable differences in the taphonomic pathways to which the fauna was subjected at PP13B between the Middle and Late Pleistocene, between the front and back of the cave, and between body size classes. The largest mammals (size classes 2e5, body weight >24 kg) were mainly accumulated by MSA hominins. Size class 1 ungulates also exhibit a degree of hominin modification consistent with some hominin accumulation of fresh carcasses, but this is more variable through time and includes an observable degree of independent carnivore contribution. Basic taxonomic comparisons reveal a low representation of small mammals, tortoises, and marine mammals at PP13B relative to larger (>4.5 kg) terrestrial mammals. This is a different pattern from other MSA sites along the southwestern coast of South Africa, where small mammals and tortoises are abundant. A microscopic study of the bone surfaces confirms that MSA hominins exploited these small faunal components opportunistically, while focusing most heavily on large terrestrial ungulates. All faunal components show evidence of carnivore scavenging of hominin food debris and a high degree of density mediated destruction. Raptors are at no point implicated as major accumulators of any fauna. The study demonstrates that the full spectrum of MSA faunal exploitation can only be understood when the large mammal, small mammal, and tortoise components of fossil assemblages have all been subjected to comprehensive taphonomic analyses. Ó 2010 Published by Elsevier Ltd. Keywords: Zooarchaeology Taphonomy Introduction Faunal exploitation strategies during the Middle Stone Age (MSA) and Middle Paleolithic have figured prominently in discussions of the timing and nature of the origins of modern human behavior (Stiner, 1993; Marean and Assefa, 1999; McBrearty and Brooks, 2000). These exchanges have focused on ability to hunt large or dangerous ungulates (Klein, 1976, 1989; Binford, 1984; Milo, 1998; Marean et al., 2000; Clark and Plug, 2008; Faith, 2008), ability to make intensive use of seasonal resources (Klein et al., 1987), and capacity to exploit flying and swimming animals (Klein, 1975). In addition to epistemological complications with this list making approach (Henshilwood and Marean, 2003), there are also persistent methodological problems in the collection and analysis of zooarchaeological data that have interfered with the ability of researchers to make confident interpretations about MSA faunal exploitation behavior. q This article is part of ‘The Middle Stone Age at Pinnacle Point Site 13B, a Coastal Cave near Mossel Bay (Western Cape Province, South Africa)’ Special Issue. E-mail address: jessica.thompson@uq.edu.au. 0047-2484/$ e see front matter Ó 2010 Published by Elsevier Ltd. doi:10.1016/j.jhevol.2010.07.004 Most conclusions have been based on assemblages that have had no or limited published taphonomic analyses (Klein, 1975, 1976, 1977, 1978a,b; Binford, 1984; Milo, 1998; Klein et al., 1999; Henshilwood et al., 2001; Clark and Plug, 2008). Without such analyses unjustified attribution of patterning in zooarchaeological data may be made to hominin behavior. In order to avoid such an issue, this paper provides the necessary first component of an overall zooarchaeological analysis of the fossil assemblage recovered from Pinnacle Point Cave 13B (PP13B) during the 2000e2006 excavations. Through analyses of bone preservation, fragmentation, differential destruction, and surface modification it reconstructs the taphonomic pathways undergone by different species and body size classes of animals during different time periods. The analyses provide a basis for understanding the timing and agencies of bone accumulation and modification. They reveal the roles played by different faunal components in MSA diet and recommend adjustments that are necessary before proceeding to higher level inferences about hominin behavior. This is the first taphonomic study of an MSA faunal assemblage in southern Africa to be presented using methods identical to those from the more recent MSA site of Die Kelders Cave 1 (DK1; Marean 322 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 et al., 2000). Explicit comparisons are now possible between these two assemblages, but until similar studies are done for other sites only general comparisons can be made to other important collections such as Blombos Cave (BBC), Boomplaas (BPS), Klasies River (KR), and Ysterfontein 1 (YS1; Fig. 1). This study is also the first to undertake a comprehensive microscopic analysis of the small mammal and tortoise components of an MSA assemblage, thus offering a more comprehensive picture of MSA faunal exploitation. Fish and birds were present in the sample, but were in such small quantities that they are not considered here (n ¼ 19 and n ¼ 57, respectively). Because of space limitations, separate papers will present more detailed interpretations of hominin butchery, transport, and hunting at PP13B through analyses of cut marks, percussion marks, and skeletal element representation. Background to Pinnacle Point Cave 13B PP13B is a large coastal cave located near the modern town of Mossel Bay, Western Cape Province, South Africa. The site is within a few minutes walk of several other caves containing MSA deposits and is to date the most intensively studied of these (Nilssen and Marean, 2002; Marean et al., 2004b, 2007; Brown et al., 2009). Although PP13B is a cavity cut by wave action into cliffs of quartzitic Table Mountain Sandstone, a capping calcrete at the top has contributed to the good preservation of fossil fauna at the site. The artifactual assemblage is dominated by unretouched flakes and flake blades on quartzite, although very early examples of bladelet technology are also present on both quartzite and silcrete in older consolidated deposits at the site (Marean et al., 2007). Ochre is not abundant but is present and was clearly being transported to PP13B for pigment extraction by rubbing or grinding (Marean et al., 2007). The archaeological deposits likely began between 174 ka and 153 ka, although a small amount of anthropogenic deposit below the location of this age estimate can only be maximally dated to the same age as the non-anthropogenic layers at the base of the sequence at 349 ka (Jacobs, 2010; Marean et al., 2010). External dune formation invaded and blocked the mouth of the cave, precluding human occupation between ca. 91 and 39 ka, after which the cave reopened (Marean et al., 2010). Optically stimulated luminescence (OSL) and uranium-series dates from PP13B indicate that there was not much reoccupation at the site after the cave reopened. Therefore, the majority of fossils reported here, began their depositional history with the cold period of early Marine Isotope Stage (MIS) 6, continued through the very warm MIS 5e, and eventually ended near the close of a cooler MIS 5b. Most sediment at PP13B is unconsolidated, except for sections along the sides of the cave where Lightly Cemented MSA (LC-MSA) deposits capped by a flowstone have been attached to the walls. Excavations have sampled relatively shallow (less than 1 m) deposits at the front (Eastern area) of the cave and the deeper deposits at the rear (Western area) of the cave. The LC-MSA units are also located at the front of the cave. Both the front and back contain deposits that date to MIS 6 and to MIS 5, enabling both spatial and chronological evaluations of the faunal assemblagedalthough the sample of fragments from MIS 6 at the back of the cave is so small (n ¼ 120) it limits some analyses. Fig. 1. Location of PP13B relative to sites discussed in the text. 323 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 Methods The faunal assemblage reported here consists of both piece plotted specimens and specimens that have been recovered from the 10 mm and 3 mm screens. It does not include pieces of trabecular bone, bits of cortical bone with no evidence of having a medullary cavity or having otherwise come from a long bone, or fragments of cortical bone with no facets, diagnostic shape, or muscle markings that could indicate even an approximate location in the skeleton. The analyzed assemblage does include less identifiable fragments such as pieces of enamel, long bone, crania, and tortoise carapace/plastron that are not assignable to a specific element. No size cut off was used. All specimens were cleaned with fresh water and given individual numbers for a total Number of Identified Specimens (NISP) of 18,590 larger mammals (>4.5 kg live weight), 407 small mammals, and 4418 tortoises. Some of this material came from disturbed or filled areas, recent surface sediments, or non-anthropogenic basal layers. Unless otherwise indicated all data analyzed here are from only well-dated and well-provenienced sediments (NISP ¼ 12,889 larger mammals, 281 small mammals, and 2822 tortoises). The stratigraphic aggregates as defined by Marean et al. (2010) have been divided into eight well-understood analytical units. These underlie all analyses of the fauna from PP13B (Table 1). Because many analyses focus on subcomponents of the assemblage for which sample sizes are small, some results are presented in larger terms of MIS 5 (a warmer period <128 ka) versus MIS 6 (a colder period >128 ka). These cases are identified in the text. All data collection was conducted at Iziko: South African Museums of Cape Town. Each numbered specimen was entered as an individual record into a customized Microsoft Access database. Attributes such as element, side, taxon, and body size were entered along with macroscopically visible taphonomic attributes such as fragmentation (Villa and Mahieu, 1991) and burning (used here in terms of present/absent and based on basic color changes such as carbonization or calcination). Each specimen was then examined under a 10e40 binocular zoom light microscope for more subtle taphonomic indicators such as weathering (Behrensmeyer, 1978), rodent gnawing, and post-depositional surface destruction (Thompson, 2005). Tooth, percussion, and cut marks were identified using the criteria of Blumenschine et al. (1996), for which the author has taken blind tests on specimens modified by known agents. These data were collected during the publication of arguments regarding the revision of criteria for distinguishing between carnivore tooth marks and geochemical etching and bioerosion (Domínguez-Rodrigo and Barba, 2006, 2007; Blumenschine et al., 2007). As such, only high confidence marks of any kind are presented here. Each fragment that could be identified to element and confidently placed on a template of that element was drawn into a GIS image analysis program developed by Marean et al. (2001) and linked by specimen number to the external database created in Access. The minimum number of elements (MNE) was estimated from these GIS images on the principle of overlaps: where two fragments overlap on a given element they cannot be from the same individual (Marean et al., 2001). This technique provides a way to rapidly and effectively estimate MNE values from any given subset of data without manually revisiting the collection, but as with any method of estimating MNE values it may miss an overlap or indicate a very slight overlap (in terms of pixels) where one is lacking (Lyman, 2008). All images produced by the program were, therefore, inspected for areas where the MNE value may have been inflated in this way, and the smallest number of definite overlaps was the one recorded as the MNE. The digital fragment records underlying each image were used to quantitatively evaluate the relationship between bone portion representation and bone density using the long bone zones described by Abe et al. (2002). The area represented within each zone was calculated and multiplied by its MNE value. For example, if the distal shaft zone was represented by some areas with an MNE of one, some with an MNE of two, and some with an MNE of three, these areas were tabulated and multiplied by one, two, and three, respectively. These were then added to obtain the total area count (in pixels) within that zone. The proportion of the total area that falls within the five zones was then expressed as a percentage. Once quantified, these proportions were input as the y-axis into a regression analysis with density as the x-axis (density values are from computed tomography [CT] scans of a sheep skeleton in Lam et al. [1998]). Spearman’s Rho was then used to assess the degree of correlation between bone portion density and representation (Thompson and Marean, 2009), as this non-parametric test is less susceptible to small sample sizes and influence by outlying points. Three other statistical tests were employed in analysis. A Chi-square test was used to examine the likelihood of independence between two nominal or ordinal variables (e.g., body size and analytical unit). Fisher’s Exact Test was used on two-way tables of data where small sample sizes made a Chi-square test inappropriate. This provided a way to assess if a given proportion of data in one analytical unit was statistically different from the proportion of data in another analytical unit. The 95% confidence limits were determined using Wald’s adjusted method (Agresti and Coull, 1998) for proportional data for which large numbers of analytical units required comparison (e.g., proportion of fracture angles that are right or proportion of shaft fragments bearing a percussion mark). Where these confidence intervals overlapped, two proportions Table 1 Minimum and maximum ages of sediments and NISP from analytical units used in the PP13B faunal analysisa MIS Overall climate Min age (ka) Max age (ka) PP13B Analytical unit Lg. mamm. Sm. mamm. Tortoise 3 5b 5b 5c 5b & 5d 5c 5d 5e 6 6 Cold Cool Cool Warm Cool Warm Cool Very Warm Cold Cold 38.9 91.6 91 91 94 91 106 120 164 (94) 153 38.9 91.6 94 102 134 98 114 130 159 (349) 174 Re-opening of cave Initial sealing of cave Light Brown Sand 1 (Western) Upper Dark Brown Sand Units (Western) Light Brown Gray Sand (Western)b Shelly Brown Sand/Upper Roof Spall (Eastern) Lower Roof Spall (Eastern) LC-MSA Middle/Lower Dune Lower Dark Brown Sand Units (Western)c LC-MSA Lower N/A N/A 639 2449 1161 4423 1283 277 120 2537 N/A N/A 9 54 22 101 49 3 2 41 N/A N/A 245 536 114 701 292 65 24 845 a b c Ages and stratigraphic assignments are summarized from U-series and OSL data in Marean et al. (2010) and Jacobs (2010). Includes at least two occupations (at w98 and w122) with a hiatus or truncation between. Most likely ages based on context with radiometric minima and maxima given in brackets. 324 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 were determined to be statistically indistinguishable and differences between the two may be attributed to sampling error. Basic assemblage description Overall taxonomic abundances Taxonomic representation at PP13B is heavily weighted in favor of larger mammals using Brain’s (1981) size classes 1e5 (>4.5 kg live weight). Relative to other MSA faunal accumulations in the Western Cape, the collection from PP13B is impoverished in small fauna (Klein and Cruz-Uribe, 2000; Henshilwood et al., 2001; Halkett et al., 2003; Klein et al., 2004). By NISP, a full 80% of the total identifiable fragments are large mammals, while only 2% are small mammals such as hares (Lepus spp.), hyraxes (Procaviidae), and small carnivores. The remaining 18% are tortoises such as the angulate tortoise (Chersina angulata) and the pancake tortoise (Homopus areolatus). Table 2 shows the relative abundances of small mammals, large mammals, and tortoises for the entire sample of fauna from PP13B in comparison to other MSA sites in the Cape Provinces of South Africa. Small fauna are further broken down by genus and species where available. Unfortunately, only counts for the Minimum Number of Individuals (MNI) are available for KR and BPS (Klein, 1976, 1978a). This is problematic because MNI values cannot be accurately added for all the layers of a site for a grand total, and MNI values based on separate layers are often so small that they are relatively meaningless (Klein, 1976; Grayson, 1984; Lyman, 2008). The totals given here for KR and BPS are therefore to be used as a rough measure of comparison only. Two values are reported for PP13B in Table 2. The first is a true value of the NISP of each category, with ID specimens referring to those taxonomically identifiable to the order level or below. The second is an adjusted value to ensure the data from PP13B are comparable to other MSA sites. At these sites mammal counts are based on epiphyseal portions of elements, horn core tips and bases, and teeth identifiable to the family level or below (Klein and CruzUribe, 1984). The adjusted values for tortoises in Table 2 represent only complete distal humeri, following the reporting methods used at other MSA sites (Klein and Cruz-Uribe, 2000; Henshilwood et al., 2001; Klein et al., 2004). PP13B is the first site for which complete tortoise NISP data are available for all elements, including less identifiable carapace/ plastron fragments. Because very tiny fragments of tortoise bone can easily be identified as such, a straight NISP comparison to identifiable large mammalian fauna should show inflation of tortoise representation. Small mammal fragments may also be better represented than large mammals because small absolute fragment sizes can still preserve a large portion of identifiable bone morphology. This makes the relative paucity of small mammal and tortoise remains at PP13B even more striking. Taxonomic abundances of large mammals Because of the relatively small numbers of fragments that could be reliably identified to species or even genus, all large mammal data are given here in terms of body size and general taxonomic category at the family level or above. This is a standard way to present important basic taxonomic data in African zooarchaeology, where the diversity of ungulates, particularly bovids, makes it difficult to identify the species based on fragmented bone (Brain, 1981; Klein and Cruz-Uribe, 1984). Some examples of species from each body size class that were identified from the same sample (Rector and Reed, 2010) include, but are not limited to: grysbok/steenbok (Raphicerus spp., size 1), springbok (Antidorcas spp., size 2), wildebeest (Connochaetes spp., size 3), eland (Taurotragus oryx, size 4), and hippopotamus (Hippopotamus amphibius, size 5). Table 3 shows a summary by NISP of the sizes 1e5 mammals from the eight well-understood analytical units at PP13B. It is immediately apparent that the site has low representation of marine versus terrestrial mammal bone. Marine mammals lack long bones with a medullary cavity that is surrounded by a dense cortical shaft. Relative to terrestrial mammals, this may reduce the degree of differential density throughout the skeleton, resulting in fewer outstandingly dense element portions in the long bonesdalthough this has not been quantified using the same density measures as for terrestrial mammal bones (Lam et al., 1998, 2003). Relative proportions of dental fragments, which are among some of the densest elements in the skeleton, show that of the 136 tooth fragments identifiable beyond the generic ‘Mammal’ category only one is from a marine mammal (a seal). If there was no reason that terrestrial mammal crania were transported more often than marine mammal crania, then the very low proportion of marine mammals is not an artifact of a high degree of differential destruction. Reported proportions of marine mammals relative to large terrestrial mammals by NISP are also low at other coastal MSA sites, Table 2 NISP of large mammals, small mammals, and tortoises at PP13B compared to data from nearby MSA sitesa,b Taxon Carnivora Lepus spp. Procaviidae Bathyergus suillus Erinaceus frontalis Hystrix africaeaustralis Chersina angulata Homopus areolatus N/A N/A N/A N/A N/A a b c d Common name Carnivores Hares Hyraxes Cape dune mole rat Hedgehog Porcupine Angulate tortoise Pancake tortoise Small Mammal (indet.) Tortoise (indet.) Total ID Small Mammal Total ID Tortoise Total ID Large Mammal Middle Stone Age site PP13B Raw PP13B Adjusted YF1 DK1 BBC KR1c KR1Ac BPSc 5 14 33 0 0 0 349 158 355 3911 52 507 2576 5 12 31 0 0 0 5 4 N/A 1 48 10 1596 0 1 0 172 0 4 N/A N/A N/A N/A 177 34 101 352 16,128 2753 150,167 69 25 N/A N/A N/A N/A 169,494 4213 30,199 35 72 767 890 20 2 N/A N/A N/A N/A 1786 620 2323 21 1 51 7d 0 22 N/A N/A N/A N/A 102 N/A 530 6 0 33 0 0 4 N/A N/A N/A N/A 43 N/A 216 1 5 0 0 0 0 N/A N/A N/A N/A 6 N/A 174 Klein, 1976, 1978a; Klein and Cruz-Uribe, 2000; Henshilwood et al., 2001; Halkett et al., 2003; Klein et al., 2004. YF1 ¼ Ysterfontein 1, DK1 ¼ Die Kelders Cave 1, BBC ¼ Blombos Cave, KR ¼ Klasies River, BPS ¼ Boomplaas. These sites are only reported by the minimum number of individuals (MNI). Only Cape mole rats (Georychus capensis) are reported from this site. 325 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 Table 3 Basic taxonomic and body size representation by NISP for each analytical unit Body size Analytical unit MIS 5 Class Mammalia Terrestrial mammal Marine mammal Order Carnivora Family Hyenidae Family Otariidae/Phocidae Order Primates Superorder ungulata Order Artiodactyla Family Bovidae Family Suidae Family Hippopotamidae Order Perissodactyla Family Equidae Family Rhinocerontidae 1 2 3 4 5 Indet. 1 2 3 4 5 Indet. 1 2 3 5 Indet. 1 2 3 2 1 2 3 2 1 2 3 4 5 Indet. 1 2 3 4 5 1 2 3 4 5 Indet. 2 5 4 4 5 MIS 6 LB Sand 1 Upper DBS LBG Sand SB Sand/Upper RS Lower RS LC-MSA Middle/Lower Dune Lower DBS LC-MSA Lower 32 20 15 6 3 198 49 44 56 10 4 72 1 1 0 0 1 0 2 0 0 0 0 0 1 5 4 7 9 0 1 0 3 0 0 0 21 22 19 11 4 17 0 0 0 1 0 144 77 76 31 14 779 168 162 206 74 13 192 4 0 3 0 5 0 2 0 0 1 7 1 0 14 23 39 12 4 10 4 4 2 0 0 75 101 87 40 1 70 1 1 0 1 1 27 26 23 8 3 483 39 56 106 23 3 118 0 0 1 0 0 4 3 0 2 1 1 1 1 4 6 15 9 5 3 0 3 2 1 0 30 40 34 11 4 60 1 0 1 3 0 294 167 147 38 3 1145 410 405 495 102 13 221 5 4 3 0 7 8 6 2 0 0 7 4 0 34 47 46 17 2 7 9 16 11 1 0 188 173 175 16 2 192 0 0 0 1 0 88 38 20 5 1 469 134 128 74 9 3 113 0 0 0 1 6 1 1 0 0 0 1 0 0 10 8 9 1 0 2 3 1 3 0 1 61 50 17 3 0 20 1 0 0 1 0 7 4 2 0 0 46 46 39 26 7 0 63 0 1 0 0 0 0 0 0 0 0 0 0 0 3 2 3 1 0 0 0 0 1 0 0 12 6 6 0 0 2 0 0 0 0 0 4 5 1 0 0 46 1 8 6 0 0 12 0 0 1 0 0 3 0 0 1 0 0 0 0 0 1 1 0 0 2 0 1 0 0 0 5 4 3 2 0 13 0 0 0 0 0 71 46 63 17 1 689 276 342 304 61 11 357 5 1 1 1 8 2 0 0 0 0 0 0 0 13 17 30 13 1 13 2 10 5 0 0 30 57 47 8 1 34 0 0 0 0 0 but not as low as at PP13B: 5% at DK1, (Klein and Cruz-Uribe, 2000), 10% at BBC, (Henshilwood et al., 2001), and 8% at YS1, (Halkett et al., 2003)dthough the last sample is very small. Proportions appear to be much higher at Klasies River, but these data are based on the summed MNI and are likely overestimates: 15% at KR Cave 1, 37% at KR Cave 1A, and 38% at KR Cave 1B (Klein, 1976). Despite a fluctuating shoreline, the ocean was within marine resource transport distance to PP13B more often during MIS 5 than MIS 6 (Marean et al., 2007: their Supplementary Video). Marine resources of all types might therefore be expected to be most abundant in the deposits dating to MIS 5. However, at PP13B there is very little variation in marine mammal representation between analytical units (proportions of fragments that could be identified as either marine or terrestrial mammal that are marine range between 1.6% and 0.5%) and there are shellfish throughout (Marean et al., 2007). There is slightly higher marine mammal representation in MIS 5 (1.2%) than MIS 6 (1.0%). Although this is the expected direction of difference, Fisher’s Exact Test does not show this to be a significant difference (p ¼ 0.6947). The low overall abundance of marine mammals during both time periods shows that although MSA populations may have had access to marine resources at various intervals throughout the course of the site’s occupation, acquiring and transporting marine mammals to PP13B did not comprise a substantial portion of the overall subsistence strategy. Ungulate taxa are 97.4% by NISP of terrestrial mammal specimens identifiable to the superorder or below. Ungulates are distributed nearly evenly between size classes 1 and 3 (89.9% of all ungulates identifiable to body size), with a small contribution of larger sizes 4 and 5 specimens (10.1% of all ungulates). A full 99% of specimens that could be identified to the family level or below were bovids (with the remainder belonging to the families Equidae, Suidae, Rhinocerontidae, and Hippopotamidae). The uniformity of the assemblage suggests that the majority of less identifiable specimens are also from sizes 1e3 bovids. The remainder includes primates (0.1%) and large carnivores (1.6%)dmost of which are represented by small tooth fragments identifiable only to the order 326 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 level. The three specimens attributed to the Hyenidae are two halves of a complete fibula and a single scapholunar. The low carnivore representation is an excellent indication that the site was not extensively used by denning carnivores such as hyenas, as den sites often have high proportions of carnivore remainsdmany of which are juveniles (Stiner, 1991). However, it does not rule out the possibility that carnivores used the site transiently either as a temporary lair or as scavengers after human occupation. Patterning in body size abundances is apparent when the data are divided into spatial (i.e., front versus back), chronological (i.e., MIS 5 versus MIS 6), and body size groupings (Fig. 2). Small fauna (sizes 1 and 2) are more abundant at the entrance of the cave and large fauna (sizes 3, 4, and 5) are relatively more abundant at the rear. A Chi-square test shows a highly significant association between location in the cave and proportions of different body size classes (Chi-square value ¼ 1675.3; D.F. ¼ 5; p-value <0.0001), and does not appear to be influenced by wildly high or low proportions in any single square. This spatial difference could be because the relative contributions of potential human, carnivore, and raptor accumulators differed between the front and the back of the cave, because there was one primary accumulator that differentially distributed larger mammal bones toward the rear, or because larger mammal bones were differentially fragmented according to body size in the two areas. Body sizes are also significantly different in their distributions between MIS 5 and MIS 6 (Chi-square value ¼ 13763; D.F. ¼ 5, p-value <0.0001). Specifically, size 1 ungulates are more common during MIS 5, and very large ungulates (size 5) are more abundant in MIS 6dalthough samples for this largest size class are quite small overall. An obvious question is whether the spatial and chronological patterns are related, with the smallest fauna being more common at the front of the cave where a larger sample from MIS 5 is also available. When the NISP of size 1 versus all other body sizes is examined within the more detailed categories of front/MIS 5, back/MIS 5, front/MIS 6, and back/MIS 6, only front/MIS 5 has a higher proportion of size 1 fauna (26.2% versus 21.7%e21.8% in the other three groupings). Fisher’s Exact Test shows that this difference is highly significant when front/MIS 5 is compared to all other groupings (p < 0.0001). This indicates that size 1 fauna is truly more abundant in MIS 5, and during this time period there was additional input of size 1 fauna at the front of the cave. The slightly higher representation of size 1 fauna from MIS 5 may have been climatically driven: local environments during warmer periods could have resulted in habitats favoring small, pair bonded antelope such as grysbok and steenbok, which use elusive hiding as their primary predator defense system (Skinner and Chimimba, 2005). During colder MIS 6 times, local vegetation may have favored herding ungulates that preferred aggregation and speed for predator defense. The predator in question could have been either MSA hominins or carnivores. In the case of the size 1 fauna, the primary predator may even have been raptors, as at DK1 (Marean et al., 2000). The pattern could also be because the agent of accumulation differed during these periods (or contributed differently to the faunal assemblage). A Chi-square test was used to examine how the distributions of body sizes between two analytical units compared to one another individually, without being placed into larger aggregates (Table 4). For these analyses, size 5 fragments are not included because some analytical units would have a value of zero, which cannot be used in a Chi-square test. Most p-values are highly significant below the a ¼ 0.05 level, indicating a high likelihood of association between analytical unit and body size distribution. This shows that although the relative body size abundances from various analytical units appear similar, they are subtly and significantly different both within and between chronological and spatial aggregates. These results suggest that the large mammals at PP13B have had a complex taphonomic history that was in part guided by prey body size and is not easily explained by broad-scale environmental changes. The data also raise several important questions about the agent(s) of accumulation that must be answered for each analytical unit and body size class at PP13B. Taxonomic abundances of small fauna Fig. 2. Large mammal body size representation broken down by front versus back of the cave (a) and MIS 5 versus MIS 6 (b). Because of small sample sizes, it is difficult to interpret taxonomic abundances for the small mammal assemblage from PP13B; however, the presence of hares and hyraxes indicate relative proximity to both rocky outcrops and more level, vegetated environments (Skinner and Chimimba, 2005). One anomaly in the taxonomic patterning is the absence of Cape dune mole rats (Bathyergus suillus). These large rodents average about 1 kg in weight and are a common component of other coastal MSA faunal assemblages in the Cape (Klein and Cruz-Uribe, 2000; Henshilwood et al., 2001; Klein et al., 2004). These rats are sensitive ecological indicators of the surrounding landscape, and today they are restricted to the littoral zone in the Western Cape Province in areas with sand dunes or loose, sandy soil (Jarvis and Bennett, 1991). Where available, they have been a documented part of the diets of raptors, mammalian carnivores, and humans (Henshilwood, 1997; Skinner and Chimimba, 2005). The tops of the cliffs in which PP13B is set would have provided suitable habitat at several points in the cave’s history, so the absence of Bathyergus suillus begs a taphonomic explanation that will be explored in the next section. 327 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 Table 4 Results of Chi-square tests comparing the body size distributions between analytical units D.F. Analytical unit 1 Analytical unit 2 Test statistic MIS 5 vs. MIS 5 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 Light Brown Sand 1 Light Brown Sand 1 Light Brown Sand 1 Light Brown Sand 1 Light Brown Sand 1 Upper Dark Brown Sand Upper Dark Brown Sand Upper Dark Brown Sand Upper Dark Brown Sand Light Brown Gray Sand Light Brown Gray Sand Light Brown Gray Sand Shelly Brown Sand/Upper RS Shelly Brown Sand/Upper RS Lower Roof Spall Upper Dark Brown Sand Light Brown Gray Sand Shelly Brown Sand/Upper RS Lower Roof Spall LC-MSA Middle/Lower Dune Light Brown Gray Sand Shelly Brown Sand/Upper RS Lower Roof Spall LC-MSA Middle/Lower Dune Shelly Brown Sand/Upper RS Lower Roof Spall LC-MSA Middle/Lower Dune Lower Roof Spall LC-MSA Middle/Lower Dune LC-MSA Middle/Lower Dune 0.7560 12.7750 11.0610 48.0240 8.9393 14.9460 38.1510 96.5070 15.3330 40.7470 116.1800 31.0910 64.3710 6.8325 3.6631 0.9443 0.0124 0.0259 <0.0001 0.0626 0.0048 <0.0001 <0.0001 0.0041 <0.0001 <0.0001 <0.0001 <0.0001 0.1450 0.4535 MIS 5 vs. MIS 6 4 4 4 4 4 4 4 4 4 4 4 4 Light Brown Sand 1 Upper Dark Brown Sand Light Brown Gray Sand Shelly Brown Sand/Upper RS Lower Roof Spall LC-MSA Middle/Lower Dune Light Brown Sand 1 Upper Dark Brown Sand Light Brown Gray Sand Shelly Brown Sand/Upper RS Lower Roof Spall LC-MSA Middle/Lower Dune Lower Dark Brown Lower Dark Brown Lower Dark Brown Lower Dark Brown Lower Dark Brown Lower Dark Brown LC-MSA Lower LC-MSA Lower LC-MSA Lower LC-MSA Lower LC-MSA Lower LC-MSA Lower 4.8287 6.1586 7.2061 4.0573 5.2261 3.2046 9.5438 24.6660 21.1800 15.0320 79.5170 13.2990 0.3053 0.1876 0.1254 0.3983 0.2649 0.5242 0.0489 < 0.0001 0.0003 0.0046 <0.0001 0.0099 MIS 6 vs. MIS 6 4 Lower Dark Brown Sand LC-MSA Lower 2.1967 0.6996 The tortoise sample was large enough for a detailed analysis of fragments from the eight well understood analytical units at PP13B. Of these, only 13% (n ¼ 356 of 2822) could be identified to species. By NISP, the pancake tortoise makes up 27% of this portion of the assemblage. The angulate tortoise comprises 73%, and five fragments in the assemblage are unidentified large chelonians. These species abundances at PP13B are quite unusual in comparison to other MSA sites such as DK1, BBC, and YF1 where only one species, the angulate tortoise, is reported (Klein and Cruz-Uribe, 2000; Henshilwood et al., 2001; Halkett et al., 2003; Klein et al., 2004). Although tortoise representation at PP13B is unusual compared to other MSA sites, it is the others that have the unexpected pattern. Today, the Western Cape lies within the geographic ranges of the angulate tortoise, the pancake tortoise, and the leopard tortoise (Geochelone pardalis). The helmeted turtle (Pelomedusa subrufa) is also found here (Boycott and Bourquin, 1988). In most instances, if a faunal collection samples the immediate faunal community, species richness is expected to decrease with decreasing sample size (Grayson, 1984, 1991). However, the comparatively small tortoise sample from PP13B has higher species richness than do other MSA sites and is a closer approximation of what is today the local ecological situation. During MIS 6, the angulate tortoise makes up a relatively higher proportion of the total NISP identifiable to species than in MIS 5 (72% versus 82%, respectively), but Fisher’s Exact Test does not find this difference to be significant (p ¼ 0.1682). MNE and MNI estimates were conducted at PP13B for all identifiable elements, including carapace and plastron fragments that could be assigned to their correct number within the skeleton (Fig. 3). The total MNI for the angulate tortoise is eight (on xiphiplastra and femora), the total MNI for the pancake tortoise is four, (on xiphiplastra and marginals), and the total MNI for all tortoises at PP13B is 12 individuals based on femora. Some tortoise elements Sand Sand Sand Sand Sand Sand p-Value (such as costal and neural fragments) are less easily assigned a specific location in the skeleton, and thus appear underrepresented in the MNE estimates. For all species, the highest MNE estimates occur on both shell and limb elements. This indicates that an understanding of tortoise representation and the potential role of small fauna in MSA diet at other sites requires a thorough taxonomic study that includes representative elements from all portions of the skeleton. Agent of accumulation Large mammals Unraveling the complex taphonomic histories of assemblages with multiple possible accumulators is one of the most challenging tasks in zooarchaeology. It is also one of the most important, because interpretations of hominin subsistence behavior rely on analyses that use only components of a given assemblage that were actually collected and/or modified by the hominins in question. Cave assemblages are particularly suspect in having had multiple accumulators because caves act as focal points on the landscape to which carnivores, raptors, rodents, hominins, and abiotic processes (such as running water) all bring bones that will eventually become fossils (Brain, 1981; Lyman, 1994). Large mammal zooarchaeology benefits from a variety of actualistic and naturalistic studies that bridge the interpretive gap between the behavior of various bone modifiers and the physical effects they have on a bone accumulation. These analyses include the sizes and types of elements that are present (Marean and Spencer, 1991; Marean et al., 1992; Pickering et al., 2003) and the microscopic traces that are left on bone surfaces (Blumenschine, 1986, 1995; Blumenschine and Selvaggio, 1988; Capaldo and Blumenschine, 1994; Selvaggio, 1994; Blumenschine et al., 1996; Dominguez-Rodrigo, 1997, 2008; 328 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 Fig. 3. MNE estimates on carapace and plastron elements for the pancake tortoise (Homopus areolatus), angulate tortoise (Chersina angulata), and all tortoises combined. Capaldo, 1998; Selvaggio and Wilder, 2001; Domínguez-Rodrigo and Piqueras, 2003; Pickering et al., 2004; Domínguez-Rodrigo and Yravedra, 2009). Initial assessment of taphonomic processes A taphonomic analysis must begin with the identification of any systematic biases in the assemblage that may have been caused by differential preservation, destruction, or analytical bias. One problem in MSA zooarchaeological studies has been the interpretation of data based on analyses that are limited to the most readily identified elements and element portions, such as the epiphyseal ends of long bones (Klein and Cruz-Uribe, 1984; Bartram and Marean, 1999; Klein et al., 1999). This approach has been critiqued (Bartram and Marean, 1999) because although epiphyses and elements which are made mainly of trabecular bone, are highly identifiable, they are also the most susceptible to density mediated destructive processes such as carnivore ravaging and sediment leaching (Klein and Cruz-Uribe, 1984; Marean and Spencer, 1991; Lam et al., 2003; Cleghorn et al., 2004). Skeletal element abundances in collections recovered from numerous contexts and time periods that have undergone such processes are characterized by what Marean et al. (2004a) call a “Type II” pattern, where head and foot elements are overrepresented relative to their expected occurrence in a complete skeleton (Marean and Frey, 1997; Bartram and Marean, 1999). Marean et al. (2004a) explain the pattern in terms of differential survivorship of highly identifiable portions of head and foot elements rather than purposeful selection or transport by hominins. Under some conditions dense long bone shafts can also be secondarily modified or fragmented past the point of identification by repeated episodes of wetting/drying and heating/cooling (Conard et al., 2008), selection by rodents for gnawing (Thompson et al., 2008), or shaping into bone tools by hominins. However, in most cases these less easily identified portions survive well in a manner that is predicted by their relatively high density, and they are one of the most common components of a zooarchaeological assemblage (Lyman, 1993; Lam et al., 2003). PP13B is a sheltered environment with limited rodent gnawing from which only a single bone tool has been recovered, so long bone shafts should be well represented there (Thompson, 2008). Because of the impact any form of differential destruction has on interpretations of hominin behavior and involvement in a fossil assemblage, the possibility of density-mediated destruction must first be examined at PP13B before higher level inferences about the agent(s) of accumulation can proceed. PP13B strongly resembles one of the more typical shaft dominated assemblages. Seventy-six percent of long bone fragments in the total sample from PP13B are shaft fragments, with 13% near epiphysis shaft fragments (defined as a length of shaft with a medullary cavity and some attached trabecular bone) and a mere 11% represented by epiphyseal portions. When ungulate bone portion representation is plotted against bone portion density for the entire assemblage, Spearman’s Rho confirms a highly significant positive correlation between the two variables (Rs ¼ 0.59867, p ¼ 0.0005). This quantitatively documents an overall high degree of differential destruction at the site in accordance with what one would expect if this destruction was mediated by bone density. Sample sizes were not sufficient to examine each analytical unit separately, but when data were broken down chronologically and by body size, a simple linear regression and Spearman’s Rho both show a positive and highly significant correlation between bone density and bone portion representation for all but one data subset (Table 5). Figure 4 shows this pattern for all analytical units and body sizes combined and for small and large ungulates in MIS 5 and MIS 6. Sample sizes from MIS 6 are extremely small, as evidenced by Table 5 Spearman’s Rho correlations and p-values for bone portion representation versus bone density for small (sizes 1 and 2) and large (sizes 3 and 4) ungulates during MIS 5 and MIS 6 Ungulate size Dataset Spearman’s Rho p-Value Small Large Small Large MIS MIS MIS MIS 0.4315 0.6772 0.4587 0.5442 0.0160 <0.0001 0.0130 0.0040 5 5 6 6 the wide spread of data points, which indicates that the correlations should be interpreted cautiously. Within the large sample from MIS 5, the residuals for midshafts almost always fall on or below the linear regression line. Thus, very dense portions are often less well represented in the assemblage than predicted by their density for all body sizes. In addition, for large ungulates a clear pattern emerges for the epiphyses and near epiphysis shafts. Epiphyses are even more poorly represented than expected, while the residuals for near epiphysis shafts are nearly always positive. This illustrates the complex interactions between bone density and bone portion ‘identifiability’ in analysesdparticularly for animals of a larger body size. Near epiphysis shafts often possess both a dense midshaft portion that is likely to survive in addition to a less dense but generally more easily identified near epiphysis portion. Overall, the correlations in both MIS 5 and MIS 6 suggest that smaller ungulates at PP13B have undergone less densitymediated attrition than have larger ungulates. This again may be related to issues of relative fragmentation and ‘identifiability.’ The details of these relationships and the implications this has for interpretations of hominin behavior will be discussed in detail in a separate paper. Experimental and ethnoarchaeological studies have shown that carnivore ravaging of a bone assemblage results in such a pattern of density-mediated destruction and that ravaging is an extremely common process even at sites where humans have nearly depleted all nutrients from bones through marrow and/or grease extraction (Marean and Spencer, 1991; Marean et al., 1992; Bartram, 1993; Lyman, 1993, 1994; Lupo, 1995; Lam et al., 1998, 2003; Bartram and Marean, 1999; Pickering et al., 2003; Cleghorn et al., 2004; Thompson and Lee-Gorishti, 2007). The high degree of densitymediated destruction at PP13B suggests that carnivores were at work and show that only the densest archaeological fragmentsdlong bone midshaftsdcan be used to confirm this through analyses of surface modification (Blumenschine, 1986, 1995; Blumenschine and Selvaggio, 1988; Capaldo and Blumenschine, 1994; Selvaggio, 1994; Blumenschine et al., 1996; DominguezRodrigo, 1997, 2008; Capaldo, 1998; Selvaggio and Wilder, 2001; Domínguez-Rodrigo and Piqueras, 2003; Marean et al., 2004a; Pickering et al., 2004; Domínguez-Rodrigo and Yravedra, 2009). Preservation and fragmentation Because proportions of surface modification on archaeological bones will be lowered by a high degree of weathering, surface destruction, or post-depositional fragmentation (i.e., when the bone is in a non-nutritive state), the second stage of analysis must first assess and correct for these factors (Abe et al., 2002; Thompson, 2005). One way to assess the relative degrees of post-depositional fragmentation between analytical units is by analysis of the completeness of compact bones (Marean, 1991). These are rarely fragmented by humans or carnivores while the bone is in a nutritive state, and if so they bear a percussion mark, tooth mark, or gastric etching as evidence. Bones bearing such modifications can then be excluded from analysis, ensuring that fragmentation can be most reliably attributed to processes such as sediment compaction or leaching (Marean, 1991). At PP13B samples of ungulate compact bones identifiable to element were relatively small, and extremely small if each of the eight analytical units is considered separately (n ¼ 74 for all units combined). By following the procedures outlined by Marean (1991), completeness indices were derived for unmodified carpals and tarsals from MIS 5 versus MIS 6. The calcaneum was excluded because the size of its marrow cavity invites differential fragmentation relative to the other bones of the wrist and ankle. The average completeness index for MIS 5 is 80% (n ¼ 67), relative to MIS 6 at 75% (n ¼ 7). Compact bones from small and large ungulates may be differentially susceptible to post-depositional fragmentation (Marean, 1991) and there are more small ungulates in the MIS 5 deposits. Unfortunately, the sample from MIS 6 was too small to assess this separately for both time periods, but this pattern is apparent within MIS 5 (small ungulate average completeness ¼ 81%; n ¼ 49 and large ungulate average completeness ¼ 72%; n ¼ 18). The larger proportion of size 1 fauna in MIS 5 deposits may, therefore, have resulted in less overall fragmentation of compact bones. Another measure of post-depositional fragmentation is long bone fracture patterns: shafts that are broken while fresh tend to retain oblique fracture angles and curved or V-shaped fracture outlines, while shafts broken while in a ‘dry’ state tend to have right fracture angles and transverse fracture outlines (Villa and Mahieu, 1991). The well-provenienced portion of the PP13B assemblage had 5795 long bone fragments, resulting in a potential 11,590 long bone fractures that could be used for analysis. Of these, 6820 remained after elimination of unbroken ends, indeterminate ends, ends from fragments that could not be assigned to a body size, or fractures that suffered excavation damage and thus are uninformative about ancient breakage patterns (Table 6). At PP13B there is no directional trend in the proportions of ‘green’ and ‘dry’ breaks by body size and only small samples of size 5 animals are available for some analytical units (Table 6). The angle and outline data also track one another closely within all data subsets. These factors justify dividing the PP13B fragmentation data into groupings of size 1/2 and size 3/4 for analytical purposes and using proportions of right-angled breaks as an overall proxy for the degree of post-depositional long bone fragmentation (Fig. 5). In nearly all cases at PP13B there is an overlap between the 95% confidence intervals for the proportions of right-angled long bone breaks found in the small (size 1/2) and large (size 3/4) ungulate assemblages from the same analytical unit. In cases where there is no overlap, large ungulates may have greater or smaller proportions, suggesting variability between depositional units in differential fragmentation between small and large ungulates. When taken overall it appears that post-depositional fragmentation was approximately equal between the body sizes during MIS 5, but during MIS 6 large ungulates underwent less post-depositional breakage than smaller ungulates. This was also found to be the case within the large sample from DK1 Layer 10 (Fig. 5). The compact bone fragmentation data suggest the opposite patterndthat large ungulates from MIS 6 suffered the greatest degree of post-depositional fragmentation. These results illustrate the importance of using multiple lines of evidence to understand taphonomic processes. They also show how various assemblage subsets are differentially sensitive to these processes. Together, the data converge on the single conclusion that MIS 6 experienced more post-depositional fragmentation than did MIS 5. This is expected given that the MIS 6 deposits have had more time for cumulative processes of destruction to be in operation. Differences in long bone fragmentation may be attributable to variability between analytical units in other factors such as subaerial weathering and burning. These processes can weaken the bone and lead to greater susceptibility to breakage by other means 330 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 Fig. 4. Long bone portion representation at PP13B (y-axis) versus bone density (x-axis) as measured by CT scans (Lam et al., 1998). Open squares ¼ proximal end, open circles ¼ distal end, gray squares ¼ proximal shaft, gray circles ¼ distal shaft, black diamonds ¼ midshaft. Figure includes overall PP13B (a), MIS 5 small ungulates (b), MIS 5 large ungulates (c), MIS 6 small ungulates (d), and MIS 6 large ungulates (e). (Marean et al., 2000). At PP13B only 0.1% of all bones were weathered beyond Behrensmeyer’s (1978) stage 1 (99.1% were in stage 0 and 0.8% were in stage 1). The proportion of bones showing any weathering at all is slightly higher at the back of the cave (1.4% rather than 0.6%). More subaerial weathering might be expected at the mouth of the cave if it took place within the confines of the site. Alternatively, very occasionally bones may have been transported from outside the cave after being exposed to subaerial weathering, but this weathering was almost never sufficient to move the bone beyond stage 1. However, weathering was so rare in the assemblage that it cannot account for differences in post-depositional fragmentation, nor were bones regularly scavenged by hominins or other accumulators from carcasses that had been exposed for longer than a few days (Behrensmeyer, 1978). The slightly greater degree of post-depositional fragmentation observed at the front of PP13B may instead be explained by more fragments from the front of the cave being burned (7.5% in the front and 4.7% in the back). Marean et al. (2000) subtracted burned fragments from the DK1 long bone fragmentation analysis and found that this increased the proportion of fragments showing ‘green bone’ breaks. At PP13B only 12% of all fragment types showed any evidence of burning. When burned fragments are eliminated from the long bone fragmentation data at PP13B, proportions of ‘dry’ breaks in the back decrease by only 1% for small Table 6 Fracture angle and outline data for long bones in each analytical unit Fracture angles Size 1 Oblique Oblique/right MIS 5 LB Sand 1 Upper DBS LBG Sand SB Sand/Upper RS Lower RS LC-MSA Middle/Lower Dune 53 129 57 451 153 57 1 5 3 7 5 1 MIS 6 Lower DBS LC-MSA Lower 0 363 0 8 Fracture outlines MIS 5 LB Sand 1 Upper DBS LBG Sand SB Sand/ Upper RS Lower RS LC-MSA Middle/ Lower Dune MIS 6 Lower DBS LC-MSA Lower Size 2 Right Intermed. 59 131 0 0 59 463 Size 5 Right Oblique Oblique/right Right Oblique Oblique/right Right Oblique Oblique/right Right 23 39 13 97 30 30 64 251 100 522 181 45 0 9 1 17 6 2 24 53 32 159 37 24 66 297 125 652 99 36 2 9 4 28 2 2 15 84 48 197 36 16 25 123 36 128 11 6 1 4 1 7 1 0 0 28 6 29 3 3 5 17 6 14 6 0 0 0 1 3 0 0 5 8 0 3 0 0 3 139 13 503 0 13 2 170 8 459 0 12 2 119 0 84 1 3 0 16 0 16 0 0 0 7 Size 2 Transverse Size 4 Oblique/right Size 1 Curved/ V-shaped Size 3 Oblique Size 3 Trans./ curved Curved/ V-shaped Intermed. 17 38 1 4 70 251 0 2 17 53 0 14 0 102 0 1 83 8 529 3 Transverse Trans./ curved Size 4 Curved/ V-shaped Intermed. 1 7 62 301 1 0 18 82 28 3 128 2 151 15 655 4 Transverse Trans./ curved Size 5 Curved/ V-shaped Intermed. Transverse Trans./ curved Curved/ V-shaped Intermed. Transverse Trans./ curved 2 7 23 118 1 0 2 33 0 4 6 18 0 0 4 7 0 0 44 3 35 0 7 1 4 0 2 1 196 22 121 1 39 3 15 0 5 0 153 1 31 3 188 0 32 4 107 0 28 2 9 0 6 0 5 0 1 0 56 0 31 1 48 1 22 0 41 0 11 2 5 0 4 0 0 0 0 0 0 0 3 0 14 0 1 0 7 0 3 0 0 1 0 0 0 0 0 0 376 3 124 7 506 2 160 18 456 1 122 11 86 0 16 1 19 0 4 0 332 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 Fig. 5. Proportions of long bone breaks that exhibit a right fracture angle and their 95% confidence intervals for individual analytical units at PP13B, spatial aggregates (front vs. back), and chronological aggregates (MIS 5 vs. MIS 6). Right-angled breaks are a proxy for post-depositional destruction, illustrated by comparison with a freshly broken experimental assemblage and the DK1 assemblage (data from Marean et al. [2000: 209]). Crosses ¼ small (size 1/2) mammals, Black squares ¼ large (size 3/4) mammals. ungulates in both MIS 5 and MIS 6. Proportions at the front lower by only 1% for small ungulates in MIS 5 and by 2% for both small and large ungulates in MIS 6. All proportions remain far outside the 95% confidence interval for a freshly broken assemblage. Thus, although a small amount of additional post-depositional breakage occurred in the front, likely assisted by burning, it was not substantially greater than that at the rear. Importantly, all archaeological data exhibit a degree of postdepositional fragmentation that exceeds the 95% confidence limits reported for three modern known-agent experimental assemblages (Fig. 5; Marean et al., 2000: 210e211). This indicates that proportions of surface modification such as cut, percussion, and tooth marks have been depressed from their original values at the time of discard. Fragments exhibiting evidence of post-depositional breakage should therefore be eliminated from long bone surface modification analyses (Abe et al., 2002). It is also necessary to eliminate fragments with poorly preserved surfaces (Thompson, 2005). At PP13B bone surfaces are preserved quite well, with less than 0.1% of the assemblage showing dendritic etching, pocking, or sheen that can erase diagnostic marks (Thompson, 2005). Smoothed surfaces that can indicate microabrasion by water-or wind-borne particles were only present in 1.4% of the assemblage, and most of these were recovered from the Laminated Facies below the archaeological horizons and are not considered here. Thus, abiotic factors such as water were not likely responsible for the deposition or substantial transport of faunal material at PP13B. The only relatively common type of surface destruction was exfoliation, which was severe in 5.5% of the assemblage. Crystal formation was apparent on much of the assemblage, although the minerals involved were not specifically identified. These crystals often formed a matrix over the fossils, with 21% of all fragments exhibiting matrix sufficient to cover half or more of the surface. Fragments with severely exfoliated surfaces or more than 70% of their surface covered in matrix are not included in the surface modification analysis. Surface modification Proportions of gastric etching and rodent gnawing were negligible throughout the PP13B assemblage (1.3% and <0.1% for all body sizes, respectively). This etching occurs more commonly on smaller fauna (2.5% for size 1, 1.9% for size 2, 1.1% for size 3, and 0.5% for size 4). These differences are small but statistically significant below the a ¼ 0.05 level using Fisher’s Exact Test between sizes 1 and 3 (p ¼ 0.0005), sizes 1 and 4 (p ¼ 0.0026), sizes 2 and 3 (p ¼ 0.0448), and sizes 2 and 4 (p ¼ 0.0227). The differences are not significant between sizes 1 and 2 (p ¼ 0.1564) or sizes 3 and 4 (p ¼ 0.3355). This indicates that a small proportion of the fauna at PP13B was contributed as fecal matter or regurgitated by raptors or carnivores and that this contribution was significantly higher for the two smallest body size classes. This result is reminiscent of the faunal assemblage from DK1, where size 1 bovids from Layers 10 and 11 displayed a disproportionately high degree of gastric etching relative to other size classes (Marean et al., 2000). The authors interpreted this as an indication of a substantial raptor contribution to the size 1 faunal assemblage. They supported the inference with a spatial analysis showing gastrically etched size 1 bovid fragments concentrated under solution cavities that would have made good roosting sites. The modern day physical configuration of PP13B also provides roosting sites along the north wall (on the left facing out of the cave). However, when the percentages of gastrically etched size 1 bone and gastrically etched bone from all body sizes are mapped by square they do not concentrate in the area that today contains the most potential raptor roosts (Fig. 6). Instead, more gastric etching occurs at the rear of the cave. If raptors were responsible for this etching, then smaller animals that are their common prey (e.g., hares, hyraxes, and tortoises) Fig. 6. Schematic map of the site and excavated area showing the incidence of gastric etching by excavated square meter. 333 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 should also be concentrated at the rear. However, small fauna makes up a larger proportion of the total from the front of the cave (19.8%) than from the back (18.7%). In addition, the fragment size data showed that most gastric etching on fauna too large to be prey for raptors is also concentrated at the rear of the cave. The sum evidence points to a small contribution of gastrically etched bone by carnivores, with raptors having an inconsequential contribution. Cut, percussion, and tooth marks are three classes of surface modification that have also been used in conjunction with actualistic studies to understand human and carnivore interactions with a fossil assemblage (Blumenschine, 1988, 1995; Blumenschine et al., 1996; Dominguez-Rodrigo, 1997; Capaldo, 1998; Marean and Kim, 1998; Marean et al., 2000; Selvaggio and Wilder, 2001; Pickering, 2002; Domínguez-Rodrigo and Barba, 2006). Examples of these marks from PP13B are given in Supplementary Online Material (SOM) Figure 1. Cut marks are excellent indicators of hominin activity, but there has been much debate over the utility of using proportions of cut marked bone to directly address the accumulator of an assemblage. The frequencies with which cut marks are produced can vary dramatically depending on factors such as butcher skill, amount of flesh remaining on the bone, taxon, and the material that comprises the cutting tool (Domínguez-Rodrigo, 2003a,b, 2008; O’Connell et al., 2003; O’Connell and Lupo, 2003; Domínguez-Rodrigo and Yravedra, 2009). These complexities demand that cut marks from PP13B be dealt with in a separate study that also specifically details MSA butchery strategies. The relative proportions of tooth marked and percussion marked midshafts have been shown to be informative about the sequence of carcass access for carnivores and hominins in scenarios where spotted hyenas (Crocuta crocuta) are the predominate carnivore (Blumenschine, 1988; Blumenschine and Marean, 1993; Capaldo, 1995, 1998; Marean and Kim, 1998; Selvaggio, 1998). However, tooth marks can be problematic because they are less activity specific and can be left both during extraction of nutrients within the bone and of those outside the bone. Also, based on morphology alone a single mark cannot be diagnosed as being left by a felid, canid, hyenid, or other carnivore (Domínguez-Rodrigo and Piqueras, 2003). Even suids, hominins, and crocodiles leave tooth marks that cannot always be differentiated from other bone modifiers (Njau and Blumenschine, 2006; Landt, 2007; Domínguez-Solera and Domínguez-Rodrigo, 2009). Proportions of tooth marked midshafts can deviate from models of carcass access if the numbers and species of carnivores accessing a carcass vary. For example, felids leave substantially fewer tooth marks when they have primary access and some carnivores leave extensive tooth marking even when scavenging the remains of hominin kills (Domínguez-Rodrigo et al., 2007a,b). The tooth mark data are, therefore, best viewed as a general guide to assist in interpretation of the percussion mark data, which provide an excellent indication of a specific activity performed by tool-using hominins (marrow and grease extraction). There were 4529 long bone midshaft fragments from the PP13B sample, but only 1667 fit the rigorous criteria demanded by the results of the preceding taphonomic analysis (no dry breaks, recent breaks, poorly preserved surfaces, or heavily obscured surfaces). Because most actualistic data have been reported only for specimens 2 cm or larger in their maximum dimension, this sample had to be further reduced to 888 in order to be comparable to experimental studies in which the agents of bone modification are known (summarized in Marean et al. [2000: 215]). This reduced sample hinders comparisons between body sizes within all eight analytical units at PP13B, necessitating that these data be aggregated into MIS 5 and MIS 6 for interpretation (Table 7). Taken overall, the PP13B assemblage most closely fits a human first scenario for all body sizes (Fig. 7). Proportions of tooth marked Table 7 Numbers of percussion marked and tooth marked midshaft fragments and numbers of all midshaft fragmentsa Marked shafts in sample Size indet. Size 1 PM TM PM TM PM TM PM TM PM TM PM TM 0 5 3 1 1 0 2 5 0 25 5 1 1 3 1 3 1 0 4 30 7 49 10 1 1 10 4 11 6 0 5 37 12 77 5 3 1 18 4 7 1 0 2 12 5 15 1 0 2 3 2 4 1 0 0 2 0 1 1 0 0 0 0 0 0 0 0 1 0 9 0 6 1 40 0 8 0 37 0 7 0 9 0 1 0 4 0 0 MIS 5 LB Sand 1 3 Upper DBS 17 LBG Sand 5 SB Sand/Upper RS 15 Lower RS 2 LC-MSA Middle/ 2 Lower Dune MIS 6 Lower DBS LC-MSA Lower 0 9 Size 2 Size 3 Size 4 Size 5 Total shafts in sample MIS 5 LB Sand 1 Upper DBS LBG Sand SB Sand/Upper RS Lower RS LC-MSA Middle/ Lower Dune MIS 6 Lower DBS LC-MSA Lower Size indet. Size 1 Size 2 13 40 21 27 13 7 5 19 6 37 15 4 7 47 19 75 22 5 2 28 0 25 1 70 Size 3 Size 4 Size 5 9 58 25 112 9 5 4 24 10 25 1 1 0 3 1 1 1 0 0 72 0 15 0 4 a Fragments have no post-depositional breaks, well preserved and visible surfaces, and are >2 cm in maximum length. midshafts are statistically indistinguishable from proportions obtained in all scenarios that include both hominins and carnivores, but all fall below the 95% confidence intervals for experimental assemblages where only carnivores are involved. In contrast, proportions of percussion marked midshafts all either fall within or (more commonly) above the 95% confidence intervals for both Blumenschine and Marean’s (1993) ‘hominin only’ and ‘homininthen-carnivore’ scenarios (summarized in Marean et al. [2000: 215]). The extensive degree of percussion marking on midshafts combined with the low incidence of tooth marking demonstrates that MSA hominins had primary access to marrow and likely meat. It further shows that hominins accumulated the majority of the large mammal bones at PP13B and extracted the marrow on site, because long bone shafts from which marrow has been removed are not commonly transported by scavenging carnivores or human hunters (Marean and Spencer, 1991; Marean and Bertino, 1994). One exception in the PP13B data is the size 1 animals. The surface modification data show lower proportions of percussion marked size 1 midshafts during both time periods, and in MIS 6 this is combined with an elevated proportion of tooth-marked midshafts. The body size abundance data showed that there were significantly fewer size 1 ungulates during MIS 6 as well. This supports the suggestion by the gastric etching data that size 1 ungulates in the earlier occupation of the site underwent a markedly separate taphonomic pathway from other faunal components. Although the proportion of percussion marked midshafts falls within the 95% confidence intervals for ‘hominin first’ scenarios, they still fall below the ranges of all other body sizes combined at PP13B. An important point is that most experimental data derive from ungulates with a body size of 2, 3, or 4 (Blumenschine, 1988; Blumenschine and Marean, 1993; Capaldo, 1995, 1998; Marean and Kim, 1998; Selvaggio, 1998). Size 1 bones naturally create smaller fragments during marrow extraction. Many that preserve marks are 334 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 flakes ¼ 400 and 33.8% at the back; n total flakes ¼ 148), this difference is not statistically significant (p ¼ 0.4589). Because flakes detached through hammerstone percussion have been shown to be less susceptible to post-depositional movement by scavenging carnivores, this indicates that the original locus of most marrow extraction by hominins was likely in various parts of the cave (Marean and Bertino, 1994). A larger sample from MIS 6 may enable comparisons within both time periods and spatial designations. The small degree of carnivore input into the PP13B sample could be attributable to scavenging carnivores, an independent carnivore-accumulated component, or some combination of both. Experimental studies that replicate a ‘hominin first’ scenario have been shown to result in a proportion of midshafts with both a percussion and tooth mark on the same fragment that is close to 5% (Capaldo, 1997, 1998; Marean et al., 2000; Egeland et al., 2004). The corrected data set from PP13B for which both marks occur on the same fragment is very small (n ¼ 31, with only 2 fragments from MIS 6). Within MIS 5 at PP13B the size 1 ungulates have a very low proportion of fragments with both mark types (1.2%), while sizes 2, 3, and 4 ungulates have proportions of 6.3%, 6.4%, and 4.6%, respectively. Proportions in MIS 6 overall are 1.1% with sample sizes too small to determine differences between body sizes. This implies that some of the fauna were contributed by carnivores independently of human activity in MIS 6 (likely the size 1 ungulates), that the smallest ungulates also had the highest degree of independent carnivore input in MIS 5, and that on larger ungulates in MIS 5 the majority of tooth marking can be accounted for by hominin accumulation and subsequent carnivore modification. Small fauna Fig. 7. Proportions of midshafts bearing a percussion mark (black circles) or tooth mark (open triangles) and their 95% confidence intervals for chronological aggregates at PP13B, compared to actualistic data from known agent carcass access scenarios (Blumenschine, 1988; Blumenschine and Marean, 1993; Capaldo, 1995, 1998; Marean and Kim, 1998; Selvaggio, 1998). less than 2 cm in maximum length. Using this arbitrary cut off allows the sizes 2, 3, and 4 data to become comparable to the experimental data sets, but may introduce a systematic bias against size 1 ungulates analyzed using the same method. However, when taken together, all lines of evidence implicate size 1 ungulatesdparticularly during MIS 6das being a subset of fauna at PP13B that had an elevated degree of carnivore input. Only the sample of midshafts from MIS 5 was adequate for further division by position within the cave (i.e., front vs. back). Within these deposits a higher proportion of percussion marked midshafts was found at the cave front (61.7% at the front; total shaft n ¼ 311 versus 51.9% at the back; total shaft n ¼ 233). However, Fisher’s Exact Test did not find this to be a significant difference (p ¼ 0.2493). A corresponding test between the proportions of tooth marked midshafts in these areas (10.9% in the front and 21.5% at the back) was found to be significant (p ¼ 0. 0047). This suggests that hominin activity may have been evenly distributed throughout the cave, while carnivore activity was concentrated more at the back. However, post-depositional disturbance may have resulted in the statistically insignificant distribution throughout the site of shafts bearing evidence of marrow processing. The fragment size data showed that for both time periods combined larger fragments were concentrated at the back. When all time periods and body sizes are considered, the front also has a significantly higher proportion of small (under 3 cm of maximum length) percussion marked midshafts (67.5%) than the back (50.3%). Fisher’s Exact Test found this to be a significant difference (p ¼ 0.0378). However, although bone flakes from any body size that bear a percussion mark are more common at the front (39.9% at the front; n total Documenting the relative roles of small versus large fauna in the MSA diet provides an important approach for understanding larger questions of MSA sociality, technology, and foraging decisions. In modern hunteregatherer groups the contributions of big game hunters are considered to have more value than the more reliable contributions of gatherers. Big game carries a higher status and it is almost universally provided by adult males (Hawkes et al., 1991; Kelly, 1995; Hawkes and Bliege, 2002; Burger et al., 2005). A more equable social structure is expected if meat resources are not the exclusive result of men’s efforts (Hawkes, 1996; Hawkes and Bliege, 2002). In the southern African MSA, tortoises could be collected and small mammals obtained with minimal personal risk. All group members may have participated in the acquisition of critical fats and proteins. This is particularly true for tortoises, which can be collected even by children and stored for later use because a tortoise has a slow metabolism and lacks an escape mechanism. The exploitation of small, less aggressive animals by women and children has been documented among modern hunteregatherers (Anell, 1960, 1969; Romanoff, 1983; Hurtado et al., 1985), but it is likely that these efforts are opportunistic and conditional on other foraging opportunities (Lupo and Schmitt, 2002). Although small fauna have been a documented part of the hominin diet for over 1.7 million years (Fernandez-Jalvo et al., 1999), and despite the abundance of small fauna at other MSA sites in the Western Cape such as DK1, Blombos Cave, and YF1 (Klein and Cruz-Uribe, 2000; Henshilwood et al., 2001; Halkett et al., 2003), detailed taphonomic analyses to identify the collector of these fauna at MSA sites are rare. Such high abundances and this lack of study raise the possibility that the roles small mammals and tortoises played in the MSA diet have been underestimated at these sites. In contrast, taxonomic abundances at PP13B show very low representation of small fauna. This may suggest that an emphasis on large mammal hunting was J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 established in South Africa by as early as 150 ka, with younger sites revealing a broader dietary spectrum that included both fast and more sedentary prey items. This is a pattern that would be generally analogous to subsistence shifts observed in the Middle and Upper Paleolithic of Europe and the Levant (Stiner et al., 2000; Stiner, 2001). However, this can only be ascertained after a full taphonomic analysis aimed at determining the agent(s) responsible for the accumulation of large mammals, small mammals, and tortoises at PP13B and other MSA sites. Small mammals At DK1, where small mammals make up 85% of the fauna by NISP, Klein and Cruz-Uribe (2000) report macroscopically visible gastric etching and polish on Cape dune mole rat postcrania, implying that the Cape eagle owl (Bubo capensis) may have been the agent of accumulation. Klein and Cruz-Uribe (2000) also show statistically that there are high frequencies of Cape dune mole rats in layers with little evidence of human occupation and low frequencies in layers with abundant archaeological evidence (Avery et al., 1997; Cruz-Uribe and Klein, 1998; Klein and Cruz-Uribe, 2000). Hyrax and hare remains tend to vary inversely with the presence of Cape dune mole rats and positively with layers containing abundant evidence of human occupation. Cruz-Uribe and Klein (1998) also used skeletal element abundances to rule out several species of eagles in the accumulation of hyraxes and hares at DK1. This could point to a human accumulator, despite these small mammals being approximately the same size as Cape dune mole rats. In contrast, at Blombos Cave Cape dune mole rat remains are spread relatively consistently throughout the stratigraphy, which is taken as evidence for a potential human accumulator (Henshilwood et al., 2001). At YF1, Cape dune mole rats are the most abundant mammalian species. Halkett et al. (2003) argue that the geological context of the site would not have provided suitable raptor roosting areas, and, therefore, these large rodents were accumulated by MSA hominins. None of these studies have included a microscopic examination of small mammal fossil surfaces, nor have they ruled out the possibility that small carnivores may have accumulated the other small mammals. The severity and distribution of gastric etching throughout the skeleton can identify specific raptor species, mammalian carnivores, or even humans as potential accumulators of a small mammal assemblage (Andrews, 1990; Crandall and Stahl, 1995). The sample from PP13B is too small to do this in detail, but the overall incidence of gastric etching is quite low at only 6.0% (n ¼ 17 of 281 small mammal fossils from the eight analytical units at PP13B). This suggests that small mammal fossils that do make a rare appearance at PP13B are mainly the result of raptor accumulation or discarded (not fecal) human refuse rather than carnivore activity (Andrews, 1990). Very low proportions (6.4% or 18 fragments) of small mammal elements retain tooth marks. These may be caused by either mammalian carnivores or humans, as there is no morphological distinction between the two (Landt, 2007). This is much smaller than the 20.3% of marked elements Landt (2007) documented in an ethnoarchaeological study of human mastication of small mammal bones. More obvious human modification is also present at a low degree at PP13B, with 10 (3.6%) cut marked and four (1.4%) percussion-marked fragments. Sharp lithic edges were used for defleshing and blunt abrasive stones were occasionally employed to assist in processing (Blumenschine and Selvaggio, 1988; Pickering and Egeland, 2006). Levels of burning are also quite low, appearing on 26 (9.3%) small mammal fragments. This discoloration is distributed randomly throughout the skeleton and all burning stages are represented evenly, unlike in the LSA deposits at Blombos Cave where burning patterns indicated a mode of processing that included 335 cooking the animal whole (Henshilwood, 1997). Thus, most of the burning at PP13B can be attributed to post-depositional incorporation of small mammal bones into sediments that were subsequently burned. Although these samples from PP13B are too small to explore specific butchery patterns, several certain facts about hominin behavior emerge. The sheer under-representation of small mammals shows that at PP13B they were not normally transported, whereas taphonomic studies at sites such as Blombos Cave and DK1 may or may not show transportation of small fauna. However, hominin modification on a few specimens does indicate that the MSA inhabitants of PP13B were at least occasionally making use of small game. Such a pattern points either to a flexible and opportunistic foraging strategy in which small mammals figured quite insignificantly, or to a strategy in which small mammals were processed and consumed at the spot of encounter and only rarely transported back to PP13B. Tortoises The sample of tortoises is larger than that of small mammals but there is very little actualistic or ethnoarchaeological work to which the archaeological results from PP13B can be compared. Sampson (2000) used skeletal element abundances from modern raptor kill and roost sites, inferred raptor accumulations, and tortoises from the Later Stone Age site of Volstruisfontein to identify unique patterns in element representation for each of these processes. This is not an ideal situation because the accumulator is inferred in two of these cases, but it does provide a comparative framework (Sampson, 2000). In this study, raptor roosts had a preponderance of cranial and axial elements, as well as relatively low frequencies of shoulder and pelvic girdle elements. Human accumulations had much higher representation of carapace and plastron fragments and very low representation of cranial and axial elements. When presented next to NISP counts of each element category it is clear that PP13B does not resemble a raptor assemblage (Fig. 8). The low proportion of gastrically etched specimens (<1%) is further evidence against nesting birds as the main accumulator. Sampson (2000) suggests that tortoise assemblages made by humans are heavily fragmented and exhibit 30e40% charring. He does not quantify a ‘heavily fragmented’ assemblage, but PP13B tortoise elements certainly fit this criterion. A full 77% of the assemblage was fragmented to such an extent that it could not be identified to element, and 13% of the assemblage was identifiable to element but still incomplete. Only 7% of the assemblage was complete elements: 5% carapace/plastron elements and 2% limb/ axial/girdle elements. However, PP13B falls somewhat short of Fig. 8. Proportions of tortoise skeletal elements at PP13B compared to the inferred raptor accumulation capping the LSA site of Volstruistfontein (raptor data from Sampson [2000]). 336 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 Sampson’s (2000) predictions for proportions of burned bone, with only about 19% (n ¼ 533) charred fragments. A simple method for processing tortoises is to place the live animal upside down in the fire and allow it to cook in the shell (Sampson, 1998). If the majority of burning in a zooarchaeological collection was not related to tortoise processing and instead occurred post-depositionally, then burning should occur with equal frequency on both the inside and the outside of the shell. If tortoises were normally processed in the manner suggested by Sampson (1998), then burning should occur preferentially on the exterior aspect. At PP13B both shell and limb fragments had similar proportions of burned and unburned fragments (24% and 17% burned, respectively). On the shell only, 51% (n ¼ 263) of burned carapace and plastron fragments were burned on both the exterior and interior aspects. This indicates that much of the burning on tortoise fragments was the result of post-depositional fires. However, four times as many fragments that were burned only on either one aspect or the other had the burning concentrated on the exterior (n ¼ 141 on the exterior versus n ¼ 35 on the interior). This suggests that cooking in the shell was occasionally employed as a part of tortoise processing. Sampson (2000) found that small carnivore accumulations have similar skeletal element representation to human accumulations, but possibly with higher representation of forelimb and shoulder girdle elements. Unfortunately, there are no actualistic data on modern tortoise butchery and carnivore consumption that provide the proportions of cut, percussion, or tooth marks that would be expected if humans or carnivores were the predominant collector. There is a further problem that humans chewing on tortoise limbs may leave tooth marks that are morphologically indistinct from those of carnivores, much as they do with small mammal elements (Landt, 2007). Percussion marks (n ¼ 19 fragments) were rare but present, and found predominately on the external aspect of carapace and plastron fragments. This indicates that hammerstone percussion of tortoises was a cost-saving method employed to open the shell. Cut marks were also present but rare (n ¼ 27 fragments). Of these, all occurred either on the interior aspect of shell (n ¼ 20) or on limb or girdle elements (n ¼ 7), indicating that stone tools were employed both for defleshing/disarticulating limbs and for removing adhering tissue from the interior of shells. Tooth marks ranged from large (ca. 2 mm) to very tiny (<0.25 mm), and were apparent on 2% (n ¼ 57) of specimens. About half (n ¼ 29) of tooth marked elements were limbs and the other half was carapace and plastron. This suggests that both humans and carnivores may have participated in the gnawing, as humans do not need to gnaw the hard carapace and plastron to access the tortoise inside. Unfortunately, tooth mark size alone has not been found to reliably indicate the carnivore at work, and human tooth mark sizes on small fauna overlap with the confidence intervals established for smaller carnivores such as jackals (Domínguez-Rodrigo and Piqueras, 2003; Landt, 2007). Additional modifications also occasionally made an appearance. Some elements retained smooth striae that may be the result of gnawing by rodents or a tiny carnivore. Three more fragments displayed multiple sharp indentations that may be the result of multiple stabbings by raptor beaks. SOM Figure 2 shows examples of all these types. At present, the majority of the evidence indicates that raptors may have made a very small contribution, but that the majority of the tortoises were accumulated and modified by both humans and mammalian predators. Where humans were involved they broke open the shell with hammerstones or cooked the tortoise intact, but this was likely an opportunistic supplement to the main fat and protein sources provided by large mammals. The overall poor representation of tortoises at PP13B and the relatively high representation of very small species such as the pancake tortoise indicate that over the course of the site’s occupation tortoise exploitation was not a major component of the MSA hominin diet. The abundance and taxonomic homogeneity of tortoises at other MSA sites contrasts sharply with this conclusion from PP13B and begs that a similar taphonomic study elsewhere be a priority in future work. Summary and conclusions The analysis from PP13B shows several distinctive patterns. The faunal assemblage is impoverished in marine mammals, small mammals, and tortoises relative to other MSA sites in the Western Cape. Large bodied ungulates, divided nearly evenly between size classes 1 and 3, dominate the faunal assemblage. As is typical at most archaeological sites, there has been a substantial degree of density-mediated destruction. This has resulted in a high representation of long bone midshafts relative to epiphyseal ends and trabecular elements. Smaller ungulates have been less subject to this destruction than larger ungulates, but the pattern is still present. This, combined with a moderate degree of post-depositional fragmentation, indicates that behavioral inferences of skeletal element transport and butchery strategies that will be presented elsewhere cannot reliably proceed without adjustments for surface area and bone portion representation. The taphonomic pathways of the fauna from MIS 6 and MIS 5 were distinctive. During MIS 6 at PP13B small quantities of ungulates were brought to the cave by carnivores and deposited at both the front and rear. Most of these were the smallest antelope in the size class of grysbok and steenbok. Low levels of tooth marking, combined with an understanding of the modern ecology of predators that regularly take the smallest antelope, suggests that felids such as leopards (Panthera pardus) were likely involved (Skinner and Chimimba, 2005; Domínguez-Rodrigo et al., 2007a,b). MSA hominins independently contributed large ungulates in the size classes of springbok, wildebeest, and eland. These hominins also exploited size 1 ungulates, but it was on an opportunistic basis. Details of skeletal element butchery and transport will be dealt with in separate papers, but it is clear that in MIS 6 hominins engaged in at least the final stages of marrow extraction on site for all body sizesdpossibly more often at the front of the cave. During MIS 5 a low level of independent carnivore contribution of size 1 ungulates continued, but hominin exploitation of this size class intensified. This could have been stimulated by a change in hunting technology, foraging strategy, or diet breadth that was in turn related to transformation of the local environment and its available resources. Despite these changes, MSA hominins in MIS 5 continued to hunt larger prey and bring portions back to PP13B for processing. Marrow was an important resource that was exploited on site. Bones that were abandoned by hominins attracted scavenging carnivores, but the low levels of tooth marking indicate that spotted hyenas were not the predominate modifiers. The very small frequencies of carnivore remains in general show that PP13B was never regularly used as a carnivore den (Stiner, 1991). More large ungulate bones eventually were deposited at the rear of the cave than at the front. Larger ungulates may have been subjected to greater peri-depositional fragmentation during nutrient extraction, but once the bones were deposited more subsequent fragmentation occurred on smaller ungulatesdparticularly within the older MIS 6 deposits. As at the nearby site of DK1 it was also found that the size 1 fauna had the highest degree of contribution by non-human agents (Marean et al., 2000). The data from both sites provide a word of caution to researchers who have interpreted patterning in larger mammal fauna at other MSA sites without first providing a welldescribed taphonomic analysis that is broken down by body size class (Klein, 1975, 1976, 1977; Henshilwood et al., 2001; Klein et al., J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 2004; Clark and Plug, 2008). The analysis at PP13B also revealed that MSA hominins opportunistically exploited small mammals and tortoises. This indicates that an analytical bias toward large mammalian fauna at other published MSA sites hinders the ability of researchers to understand the full complement of MSA faunal exploitation strategies and further illustrates the necessity of providing a full taphonomic analysis prior to embarking on behavioral interpretations of any zooarchaeological assemblage. The work presented here provides the basis for more detailed studies of MSA butchery, transport, and prey acquisition strategies at PP13B. These will be presented separately through analyses of skeletal element abundances and distributions of butchery marks throughout the skeleton. In addition to facilitating these more complex analyses, the present study has revealed important variability in taxonomic and body size abundances of homininaccumulated fauna. This shows that even within a single site faunal exploitation strategies were not as uniform during the MSA as has previously been implied, a point that has also recently been made (Lombard and Clark, 2008; Thompson, 2010). Furthermore, differences between MSA localities indicate that this observed range of variability will increase even more once comparable, taphonomically informed data sets are produced from other sites. Acknowledgements This work would not have been possible without Curtis Marean’s generous permission for access to the collection, support, and advice. The data were collected with the aid of National Science Foundation (NSF) Dissertation Improvement Grant number 0620317, an NSF Graduate Research Fellowship, and funding from the School of Human Evolution and Social Change at Arizona State University. Further support was provided by the National Science Foundation (USA; grants #BCS-9912465, BCS-0130713, and BCS0524087 to Marean), funding from the Huxleys, the Hyde Family Foundation, the Institute for Human Origins, and Arizona State University. All work was carried out at Iziko: South African Museums of Cape Town, where Sarah Wurz facilitated access to space and comparative osteological specimens and provided unwavering encouragement throughout. Finally, three anonymous reviewers greatly assisted in the organization and presentation of the analyses provided here. Appendix. Supplementary data Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.jhevol.2010.07.004. References Abe, Y., Marean, C.W., Nilssen, P.J., Stone, E.C., Assefa, Z., 2002. The analysis of cutmarks on archaeofauna: a review and critique of quantification procedures and a new image-analysis GIS approach. Am. Antiq. 67, 643e663. Agresti, A., Coull, B.A., 1998. Approximate is better than “Exact” for interval estimation of binomial proportions. Am. Stat. 52, 119e126. Andrews, P., 1990. Owls, Caves and Fossils: Predation, Preservation and Accumulation of Small Mammal Bones in Caves, with an Analysis of the Pleistocene Cave Faunas from Westbury-Sub-Mendip, Somerset, U.K. University of Chicago Press, Chicago. Anell, B., 1960. Hunting and Trapping Methods in Australia and Oceania. Hakan Ohlssons Boktryckeri, Lund, Sweden. Anell, B., 1969. Running Down and Driving Game in North America. Studia Ethnographica Upsaliensia, Lund, Sweden. Avery, G., Cruz-Uribe, K., Goldberg, P., Grine, F.E., Klein, R.G., Lenardi, M.J., Marean, C.W., Rink, W.J., Schwarcz, H.P., Thackeray, A.I., Wilson, M.L., 1997. The 1992e1993 excavations at the Die Kelders Middle and Later Stone Age Cave Site, South Africa. J. Field Archaeol. 24, 1e29. Bartram, L.E., 1993. An Ethnoarchaeological Analysis of Kua San (Botswana) Bone Food Refuse. Ph.D. Dissertation, University of Wisconsin at Madison. 337 Bartram, L.E., Marean, C.W., 1999. Explaining the “Klasies pattern”: Kua ethnoarchaeology, the Die Kelders Middle Stone Age archaeofauna, long bone fragmentation and carnivore ravaging. J. Archaeol. Sci. 26, 9e29. Behrensmeyer, A.K., 1978. Taphonomic and ecologic information from bone weathering. Paleobiology 4, 150e162. Binford, L.R., 1984. The Faunal Remains from Klasies River Mouth. Academic Press, New York. Blumenschine, R.J., 1986. Carcass consumption sequences and the archaeological distinction of scavenging and hunting. J. Hum. Evol. 15, 639e659. Blumenschine, R.J., 1988. An experimental model of the timing of hominid and carnivore influence on archaeological bone assemblages. J. Archaeol. Sci. 15, 483e502. Blumenschine, R.J., 1995. Percussion marks, tooth marks, and experimental determinations of the timing of hominid and carnivore access to long bones at FLK Zinjanthropus, Olduvai Gorge, Tanzania. J. Hum. Evol. 29, 21e51. Blumenschine, R.J., Marean, C.W., 1993. A carnivore’s view of archaeological bone assemblages. In: Hudson, J. (Ed.), From Bones to Behavior. Center for Archaeological Investigations, Carbondale, pp. 273e301. Blumenschine, R.J., Marean, C.W., Capaldo, S.D., 1996. Blind tests of inter-analyst correspondence and accuracy in the identification of cut marks, percussion marks, and carnivore tooth marks on bone surfaces. J. Archaeol. Sci. 23, 493e507. Blumenschine, R.J., Prassack, K.A., Kreger, C.D., Pante, M.C., 2007. Carnivore toothmarks, microbial bioerosion, and the invalidation of Dominguez-Rodrigo and Barba’s (2006) test of Oldowan hominin scavenging behavior. J. Hum. Evol. 53, 420e426. Blumenschine, R.J., Selvaggio, M.M., 1988. Percussion marks on bone surfaces as a new diagnostic of hominid behavior. Nature 333, 763e765. Boycott, R.C., Bourquin, O., 1988. The South African Tortoise Book. Southern Book Publishers, Johannesburg. Brain, C.K., 1981. The Hunters or the Hunted? An Introduction to African Cave Taphonomy. University of Chicago Press, Chicago. Brown, K.S., Marean, C.W., Herries, A.I.R., Jacobs, Z., Tribolo, C., Braun, D., Roberts, D.L., Meyer, M.C., Bernatchez, J., 2009. Fire as an engineering tool of Early Modern Humans. Science 325, 859e862. Burger, O., Hamilton, M.J., Walker, R., 2005. The prey as patch model: optimal handling of resources with diminishing returns. J. Archaeol. Sci. 32, 1147e1158. Capaldo, S.D., 1995. Inferring Hominid and Carnivore Behavior from Dual-patterned Archaeofaunal Assemblages. Ph.D. Dissertation, Rutgers University. Capaldo, S.D., 1997. Experimental determinations of carcass processing by PlioPleistocene hominids and carnivores at FLK 22 (Zinjanthropus), Olduvai Gorge, Tanzania. J. Hum. Evol. 33, 555e597. Capaldo, S.D., 1998. News and views: methods, marks, and models for inferring hominid and carnivore behavior. J. Hum. Evol. 35, 323e326. Capaldo, S.D., Blumenschine, R.J., 1994. A quantitative diagnosis of notches made by hammerstone percussion and carnivore gnawing on bovid long bones. Am. Antiq. 59, 724e748. Clark, J.L., Plug, I., 2008. Animal exploitation stategies during the South African Middle Stone Age: Howiesons Poort and post-Howiesons Poort fauna from Sibudu Cave. J. Hum. Evol. 54, 886e898. Cleghorn, N., Marean, C.W., 2004. The destruction of skeletal elements by carnivores: the growth of a general model for skeletal element destruction and survival in zooarchaeological assemblages. In: Toth, N., Schick, K. (Eds.), African Taphonomy: A Tribute to the Career of C.K. “Bob” Brain. Stone Age Institute Press, Bloomington, pp. 13e42. Conard, N., Walker, S.J., Kandel, A.W., 2008. How heating and cooling and wetting and drying can destroy dense faunal elements and lead to differential preservation. Palaeogeogr. Palaeoclimatol. Palaeoecol. 266, 236e245. Crandall, B.D., Stahl, P.W., 1995. Human digestive effects on a micromammalian skeleton. J. Archaeol. Sci. 22, 789e797. Cruz-Uribe, K., Klein, R.G., 1998. Hyrax and hare bones from Modern South African eagle roosts and the detection of eagle involvement in fossil bone assemblages. J. Archaeol. Sci. 25, 135e147. Dominguez-Rodrigo, M., 1997. Meat-eating by early hominids at the FLK 22 Zinjanthropus site, Olduvai Gorge (Tanzania): an experimental approach using cutmark data. J. Hum. Evol., 669e690. Domínguez-Rodrigo, M., 2003a. Bone surface modifications, power scavenging and the “display” model at early archaeological sites: a critical review. J. Hum. Evol. 45, 411e415. Domínguez-Rodrigo, M., 2003b. On cut marks and statistical inferences: methodological comments on Lupo and O’Connell (2002). J. Archaeol. Sci. 30, 381e386. Domínguez-Rodrigo, M., 2008. Conceptual premises in experimental design and their bearing on the use of analogy: an example from experiments on cut marks. World Archaeol. 40, 67e82. Domínguez-Rodrigo, M., Barba, R., 2006. New estimates of tooth mark and percussion mark frequencies at the FLK Zinj site: the carnivore-hominidcarnivore hypothesis falsified. J. Hum. Evol. 50, 170e194. Domínguez-Rodrigo, M., Barba, R., 2007. Five more arguments to invalidate the passive scavenging version of the carnivore-hominid-carnivore model: a reply to Blumenschine et al. (2007a). J. Hum. Evol. 53, 427e433. Domínguez-Rodrigo, M., Barba, R., Egeland, C.P., 2007a. Deconstructing Olduvai: a Taphonomic Study of the Bed I Sites. Springer, Dordrecht, Netherlands. Domínguez-Rodrigo, M., Pickering, T.R., Egeland, C.P., 2007b. Models of passive scavenging by early hominids: problems arising from equifinality in carnivore tooth mark frequencies and the extended concept of archaeological palimpsests. In: Pickering, T.R., Toth, N., Schick, K. (Eds.), African Taphonomy: 338 J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 a Tribute to the Career of C.K. Brain. CRAFT Press, Bloomington, Indiana, pp. 255e268. Domínguez-Rodrigo, M., Piqueras, A., 2003. The use of tooth pits to identify carnivore taxa in tooth-marked archaeofaunas and their relevance to reconstruct hominid carcass processing behaviours. J. Archaeol. Sci. 30, 1385e1391. Domínguez-Rodrigo, M., Yravedra, J., 2009. Why are cut mark frequencies in archaeofaunal assemblages so variable? A multivariate analysis. J. Archaeol. Sci. 36, 884e894. Domínguez-Solera, S.D., Domínguez-Rodrigo, M., 2009. A taphonomic study of bone modification and of tooth-mark patterns on long limb bone portions by suids. Intl. J. Osteoarchaeol. 19 (3), 345e363. Egeland, C.P., Pickering, T.R., Domínguez-Rodrigo, M., Brain, C.K., 2004. Disentangling Early Stone Age palimpsests: determining the functional independence of hominid- and carnivore-derived portions of archaeofaunas. J. Hum. Evol. 47, 343e357. Faith, J.T., 2008. Eland, buffalo, and wild pigs: were Middle Stone Age humans ineffective hunters? J. Hum. Evol. 55, 24e36. Fernandez-Jalvo, Y., Andrews, P., Denys, C., 1999. Cut marks on small mammals from Olduvai Gorge Bed I. J. Hum. Evol. 36, 587e589. Grayson, D.K., 1984. Quantitative Zooarchaeology. Academic Press, New York. Grayson, D.K., 1991. Alpine faunas from the White Mountains, California: adaptive change in the Late Prehistoric Great Basin? J. Archaeol. Sci. 18, 483e506. Halkett, D., Hart, T., Yates, R., Volman, T.P., Parkington, J., Orton, J., Klein, R.G., CruzUribe, K., Avery, G., 2003. First excavation of intact Middle Stone Age layers at Ysterfontein, Western Cape Province, South Africa: implications for Middle Stone Age ecology. J. Archaeol. Sci. 30, 955e973. Hawkes, K., 1996. The evolutionary basis of sex variations in the use of natural resources: human examples. Popul. Environ. 18, 161e173. Hawkes, K., Bliege, Bird R, 2002. Showing off, handicap signaling, and the evolution of men’s work. Evol. Anthropol. 11, 58e67. Hawkes, K., O’Connell, J.F., Blurton-Jones, N.G., 1991. Hunting income patterns among Hadza: big game, common goods, foraging goals and the evolution of the human diet. Phil. Trans. R. Soc. Lond. 334, 243e251. Henshilwood, C.S., 1997. Identifying the collector: evidence for human processing of the cape dune mole-rat, Bathyergus suillus, from Blombos Cave, Southern Cape, South Africa. J. Archaeol. Sci. 24, 659e662. Henshilwood, C.S., Marean, C.W., 2003. The origin of modern human behavior: critique of the models and their test implications. Curr. Anthropol. 44, 627e651. Henshilwood, C.S., Sealy, J.C., Yates, R.J., Cruz-Uribe, K., Goldberg, P., Grine, F.E., Klein, R.G., Poggenpoel, C., van Niekerk, K., Watts, I., 2001. Blombos Cave, southern Cape, South Africa: preliminary report on the 1992e1999 excavations of the Middle Stone Age levels. J. Archaeol. Sci. 28, 421e448. Hurtado, A.M., Hawkes, K., Hill, K., Kaplan, H., 1985. Female subsistence strategies among Ache hunter-gatherers of eastern Paraguay. J. Hum. Ecol. 13, 1e47. Jacobs, Z., 2010. An OSL chronology for the sedimentary deposits from Pinnacle Point Cave 13B - a punctuated presence. J. Hum. Evol 59 (3-4), 289e305. Jarvis, J.U.M., Bennett, N.C., 1991. Ecology and behavior of the family Bathyergidae. In: Sherman, P.W., Jarvis, J.U.M., Alexander, R.D. (Eds.), The Biology of the Naked Mole Rat. Princeton University Press, Oxford, pp. 66e96. Kelly, R.L., 1995. The Foraging Spectrum: Diversity in Hunter-Gatherer Lifeways. Smithsonian Institution Press, Washington, D.C. Klein, R.G., 1975. Middle stone age man-animal relationships in southern africa: evidence from Die Kelders and Klasies River Mouth. Science 190, 265e267. Klein, R.G., 1976. The mammalian fauna of the Klasies River Mouth sites, southern Cape Province, South Africa. S. Afr. Archaeol. Bull. 31, 75e98. Klein, R.G., 1977. The mammalian fauna from the Middle and Later Stone Age (Later Pleistocene) levels of Border Cave, Natal Province, South Africa. S. Afr. Archaeol. Bull. 32, 14e27. Klein, R.G., 1978a. A preliminary report on the larger mammals from the Boomplaas stone age cave site, Cango Valley, Oudtshoorn District, South Africa. S. Afr. Archaeol. Bull. 33, 66e75. Klein, R.G., 1978b. Stone age predation on large African bovids. J. Archaeol. Sci. 5, 195e217. Klein, R.G., 1989. Biological and behavioural perspectives on modern human origins in Southern Africa. In: Mellars, P., Stringer, C. (Eds.), The Human Revolution. Edinburgh University Press, Edinburgh, pp. 529e546. Klein, R.G., Avery, G., Cruz-Uribe, K., Halkett, D., Parkington, J.E., Steele, T., Volman, T.P., Yates, R., 2004. The Ysterfontein 1 Middle Stone Age site, South Africa, and early human exploitation of coastal resources. Proc. Natl. Acad. Sci. 101, 5708e5715. Klein, R.G., Cruz-Uribe, K., 1984. The Analysis of Animal Bones from Archaeological Sites. University of Chicago Press, Chicago. Klein, R.G., Cruz-Uribe, K., 2000. Middle and Later Stone Age large mammal and tortoise remains from Die Kelders Cave 1, Western Cape Province, South Africa. J. Hum. Evol. 38, 169e196. Klein, R.G., Cruz-Uribe, K., Milo, R.G., 1999. Skeletal part representation in archaeofaunas: comments on “Explaining the ‘Klasies pattern’: Kua ethnoarchaeology, the Die Kelders Middle Stone Age archaeofauna, long bone fragmentation and carnivore ravaging” by Bartram and Marean. J. Archaeol. Sci. 26, 1225e1234. Klein, R.G., Nitecki, M.H., Nitecki, D.V., 1987. Reconstructing how early people exploited animals: problems and prospects. In: Nitecki, M.H., Nietcki, D.V. (Eds.), The Evolution of Human Hunting. Plenum Press, New York, pp. 11e45. Lam, Y.M., Chen, X., Marean, C.W., Frey, C.J., 1998. Bone density and long bone representation in archaeological faunas: comparing results from CT and photon densitometry. J. Archaeol. Sci. 25, 559e570. Lam, Y.M., Pearson, O.M., Marean, C.W., Chen, X., 2003. Bone density studies in zooarchaeology. J. Archaeol. Sci. 30, 1701e1708. Landt, M.J., 2007. Tooth marks and human consumption: ethnoarchaeological mastication research among foragers of the Central African Republic. J. Archaeol. Sci. 34, 1629e1640. Lombard, M., Clark, J.L., 2008. Variability and change in Middle Stone Age hunting behaviour: aspects from the lithic and faunal records. In: Badenhorst, S., Mitchell, P., Driver, J.C. (Eds.), Animals and People: Archaeozoological Papers in Honour of Ina Plug. Archaeopress, Oxford, pp. 46e56. Lupo, K.D., 1995. Hadza bone assemblages and hyena attrition: an ethnographic example of the influence of cooking and mode of discard on the intensity of scavenger ravaging. J. Anthropol. Archaeol. 14, 288e314. Lupo, K.D., Schmitt, D.N., 2002. Upper Paleolithic net-hunting, small prey exploitation, and women’s work effort: a view from the ethnographic and ethnoarchaeological record of the Congo basin. J. Archaeol. Method Theory 9, 147e179. Lyman, R.L., 1993. Density-mediated attrition of bone assemblages: new insights. In: Hudson, J. (Ed.), From Bones to Behavior. Center for Archaeological Investigations, Carbondale, pp. 324e341. Lyman, R.L., 1994. Vertebrate Taphonomy. Cambridge University Press, New York. Lyman, R.L., 2008. Quantitative Paleozoology. Cambridge University Press, Cambridge/New York. Marean, C.W., 1991. Measuring the postdepositional destruction of bone in archaeological assemblages. J. Archaeol. Sci. 18, 677e694. Marean, C.W., Abe, Y., Frey, C.J., Randall, R.C., 2000. Zooarchaeological and taphonomic analysis of the Die Kelders Cave 1 layers 10 and 11 Middle Stone Age larger mammal fauna. J. Hum. Evol. 38, 197e233. Marean, C.W., Abe, Y., Nilssen, P.J., Stone, E.C., 2001. Estimating the minimum number of skeletal elements (MNE) in zooarchaeology: a review and a new image-analysis GIS approach. Am. Antiq. 66, 333e348. Marean, C.W., Assefa, Z., 1999. Zooarchaeological evidence for the faunal exploitation behavior of Neandertals and Early Modern Humans. Evol. Anthropol. 8, 22e37. Marean, C.W., Bar-Matthews, M., Bernatchez, J., Fisher, E., Goldberg, P., Herries, A.I.R., Jacobs, Z., Jerardino, A., Karkanas, P., Minichillo, T., Nilssen, P.J., Thompson, E., Watts, I., Williams, H.M., 2007. Early human use of marine resources and pigment in South Africa during the Middle Pleistocene. Nature 449, 905e908. Marean, C.W., Bertino, L., 1994. Intrasite spatial analysis of bone: subtracting the effect of secondary carnivore consumers. Am. Antiq. 59, 748e768. Marean, C.W., Domínguez-Rodrigo, M., Pickering, T.R., 2004a. Skeletal element equifinality begins with method. J. Taph. 2, 69e98. Marean, C.W., Frey, C.J., 1997. The animal bones from caves to cities: reverse utility curves as methodological artifacts. Am. Antiq. 62, 698e711. Marean, C.W., Bar-Matthews, M., Fisher, E., Goldberg, P., Herries, A.I.R., Karkanas, P., Nilssen, P., Thompson, E., 2010. The stratigraphy of the Middle Stone Age sediments at Pinnacle Point Cave 13B (Mossel Bay, Western Cape Province, South Africa). J. Hum. Evol 58. Marean, C.W., Kim, S.Y., 1998. Mousterian large-mammal remains from Kobeh Cave: behavioral implications for Neanderthals and early modern humans. Curr. Anthropol. 39, S79eS113. Marean, C.W., Nilssen, P.J., Brown, K., Jerardino, A., Stynder, D., 2004b. Paleoanthropological investigations of Middle Stone Age sites at Pinnacle Point, Mossel Bay (South Africa): archaeology and hominid remains from the 2000 field season. J. Paleoanthropology 1, 14e83. Marean, C.W., Spencer, L.M., 1991. Impact of carnivore ravaging on zooarchaeological measures of element abundance. Am. Antiq. 56, 645e658. Marean, C.W., Spencer, L.M., Blumenschine, R.J., Capaldo, S.D., 1992. Captive hyena bone choice and destruction, the Schlepp effect, and Olduvai archaeofaunas. J. Archaeol. Sci. 19, 101e121. McBrearty, S., Brooks, A.S., 2000. The revolution that wasn’t: a new interpretation of the origin of modern human behavior. J. Hum. Evol. 39, 453e563. Milo, R.G., 1998. Evidence for hominid predation at Klasies River Mouth, South Africa, and its implications for the behavior of early modern humans. J. Archaeol. Sci. 25, 99e133. Nilssen, P., Marean, C.W., 2002. Background and results from test excavations of Middle Stone Age sites at Pinnacle Point, Mossel Bay. Quatern. Res. Centre Newsl. 10, 11e38. Njau, J.K., Blumenschine, R.J., 2006. A diagnosis of crocodile feeding traces on larger mammal bone, with fossil examples from the Plio-Pleistocene Olduvai Basin, Tanzania. J. Hum. Evol. 50, 142e162. O’Connell, J.F., Hawkes, K., Lupo, K.D., Blurton Jones, N.G., 2003. Another reply to Dominguez-Rodrigo. J. Hum. Evol. 45, 417e419. O’Connell, J.F., Lupo, K.D., 2003. Reply to Dominguez-Rodrigo. J. Archaeol. Sci. 30, 387e390. Pickering, T.R., 2002. Reconsideration of critieria for differentiating faunal assemblages accumulated by hyenas and hominids. Intl. J. Osteoarchaeol. 12, 127e141. Pickering, T.R., Domínguez-Rodrigo, M., Egeland, C.P., Brain, C.K., 2004. Beyond leopards: tooth marks and the contribution of multiple carnivore taxa to the accumulation of the Swartkrans Member 3 fossil assemblage. J. Hum. Evol. 46, 595e604. J.C. Thompson / Journal of Human Evolution 59 (2010) 321e339 Pickering, T.R., Egeland, C.P., 2006. Experimental patterns of hammerstone percussion damage on bones: implications for inferences of carcass processing by humans. J. Archaeol. Sci. 33, 459e469. Pickering, T.R., Marean, C.W., Domínguez-Rodrigo, M., 2003. Importance of limb bone shaft fragments in zooarchaeology: a response to “On in situ attrition and vertebrate body part profiles” (2002), by M.C. Stiner. J. Archaeol. Sci. 30, 1469e1482. Rector, A., Reed, K., 2010. Middle and late Pleistocene faunas of Pinnacle Point and their paleoecological implications. J. Hum. Evol 59 (3-4), 231e233. Romanoff, S., 1983. Women as hunters among the Matses of the Peruvian Amazon. Hum. Ecol. 11, 339e343. Sampson, C.G., 1998. Tortoise remains from a Later Stone Age rock shelter in the upper Karoo, South Africa. J. Archaeol. Sci. 25, 985e1000. Sampson, C.G., 2000. Taphonomy of tortoises deposited by birds and Bushmen. J. Archaeol. Sci. 27, 779e788. Selvaggio, M.M., 1994. Carnivore tooth marks and stone tool butchery marks on scavenged bones: archaeological implications. J. Hum. Evol. 27, 215e228. Selvaggio, M.M., 1998. Concerning the three stage model of carcass processing at FLK Zinjanthropus: a reply to Capaldo. J. Hum. Evol. 35, 319e321. Selvaggio, M.M., Wilder, J., 2001. Identifying the involvement of multiple carnivore taxa with archaeological bone assemblages. J. Archaeol. Sci. 28, 465e470. Skinner, J., Chimimba, C., 2005. The Mammals of the Southern African Subregion. Cambridge University Press, Cambridge. Stiner, M.C., 1991. Food procurement and transport by human and non-human predators. J. Archaeol. Sci. 18, 455e482. Stiner, M.C., 1993. Modern human origins e faunal perspectives. A. Rev. Anthropol. 22, 55e82. 339 Stiner, M.C., 2001. Thirty years on the “Broad Spectrum Revolution” and Paleolithic demography. Proc. Natl. Acad. Sci. 98, 6993e6996. Stiner, M.C., Munro, N.D., Surovell, T.A., 2000. The tortoise and the hare: small-game use, the broad-spectrum revolution, and Paleolithic demography. Curr. Anthropol. 41, 39e73. Thompson, J.C., 2005. The impact of post-depositional processes on bone surface modification frequencies: a corrective strategy and its application to the Loiyangalani site, Serengeti Plain, Tanzania. J. Taph. 3, 57e80. Thompson, J.C., 2008. Zooarchaeological Tests for Modern Human Behavior at Blombos Cave and Pinnacle Point Cave 13B, Southwestern Cape, South Africa. Ph.D. Dissertation, Arizona State University. Thompson, J.C., 2010. Variability in Middle Stone Age faunal exploitation and use of the physical and social landscapes in the southwestern Cape, South Africa. In: Delagnes, A., Conard, N. (Eds.), Settlement Dynamics of the Middle Paleolithic and Middle Stone Age III. Kerns Verlag, Tubingen, pp. 11e38. Thompson, J.C., Lee-Gorishti, Y., 2007. Carnivore bone portion choice and surface modifrication on modern experimental boiled bone assemblages. J. Taph. 5, 121e135. Thompson, J.C., Marean, C.W., 2009. Using image analysis to quantify relative degrees of density-mediated attrition in Middle Stone Age archaeofaunas. Soc. Archaeol. Sci. Bull. 32, 18e23. Thompson, J.C., Sugiyama, N., Morgan, G.S., 2008. Taphonomic analysis of the mammalian fauna from Sandia Cave, New Mexico, and the “Sandia Man” controversy. Am. Antiq. 73, 337e360. Villa, P., Mahieu, E., 1991. Breakage patterns of human long bones. J. Hum. Evol. 21, 27e48.