A Study of Fractured Proboscidean Bones in Recent and Fossil Assemblages

Abstract

Reliable methods are needed to distinguish anthropogenic from non-anthropogenic causes of proboscidean limb bone breakage in fossil assemblages because of theoretical uncertainty about human-proboscidean relationships in the Pleistocene. This paper compares experimentally broken bones of African elephants (Loxodonta africana) and mammoths (Mammuthus spp.) after establishing that limb bone fracture dynamics are the same for those proboscidean taxa. We show that features thought exclusively diagnostic of percussive fracturing of green proboscidean long bones such as notched fracture edges, smooth fracture surfaces, and curvilinear fracture outlines also can be created on non-green bones and on bones affected by non-anthropogenic processes. The information reported here can be applied in analyses or re-analyses of fossil proboscidean bone assemblages and may either support or potentially alter current interpretations of hominin behavior.

Introduction

Archeological sites with proboscidean bones and lithic artifacts “occur regularly from the Plio/Pleistocene transition throughout the entire Pleistocene of Africa, Europe, and Asia” (Gaudzinski et al. 2005:181) and cluster in time in Late Glacial North America. Many sites in the world contain evidence that proboscideans such as Palaeoxodon (a.k.a. Elephas) spp. and Mammuthus spp. were sources of food and bones used as raw materials. Fragments of proboscidean bones were made into tools as early as the Lower Paleolithic in African sites such as Olduvai Site BK (Tanzania) (Leakey 1971) and in European sites such as La Polledrara (Italy) (Villa et al. 1999). Bones also might have been broken to extract marrow fats (e.g., at EDAR Culebro 1 and PRERESA in Spain; Yravedra et al. 2012, 2014), which would have been a critical tactic for Pleistocene hunter-gatherers whose diets were seasonally low in carbohydrates (see Jochim 1976; Speth 1987) and which possibly also contributed to the evolutionary expansion of hominin brain capacity (Ben-Dor et al. 2011; Thompson et al. 2019).

Anthropogenically modified bones have figured prominently in theoretical debates about the importance of proboscideans to Paleolithic hominins. Ben-Dor et al. (2011:1) considered the evidence for hominin utilization of elephant (Palaeoloxodon/Elephas antiquus) to be abundant in Acheulian sites in Africa, Asia, and Europe, and contended that elephant was the “main calorie contributor to the diet of Homo erectus in the Levant.” Furthermore, Ben-Dor et al. (2011:9) proposed that the disappearance of elephants from the Levant’s faunal record at the end of the Acheulean just before 400 ka “triggered the evolution of a [hominin] species that was more adept, both physically and mentally, to obtain dense energy (such as fat) from a higher number of smaller, more evasive animals.” Agam and Barkai (2018:20) proposed that early humans regularly hunted proboscideans at will and elephants “played a main role in the Paleolithic diet.” Barkai (2019; Lev and Barkai 2015) proposed that proboscideans were not only essential suppliers of food resources for Paleolithic people but were also symbolically humanized as equals in the world. Zutovski and Barkai (2016) suggested that Lower Paleolithic handaxes made from proboscidean bones were symbolic links between the hominins who made the tools and the animals whose bones provided the raw material.

Other archeologists have made a different case that animals as large as proboscideans would not have been regularly or preferentially killed in spite of the promise of high caloric returns, because of low encounter rates and very high handling costs in terms of time and energy (e.g., Byers and Ugan 2005; Lupo and Schmidt 2016). Wolfe and Broughton (2020) analyzed North American records of Late Glacial extinct megafauna and found mammoths occur in archeological sites in proportion to their occurrence in non-archeological/paleontological sites, as do other extinct species, and therefore were not disproportionately preferred human prey. No comparable in-depth study has been done of the worldwide numbers of single- and multiple-proboscidean archeological and paleontological sites compared to the numbers of sites with smaller prey, but it would be a worthwhile endeavor. A list posted online (Haynes 2020 doi: https://doi.org/10.13140/RG.2.2.19976.37121) contains > 150 worldwide fossil sites with possible evidence for human interactions with proboscideans, although it is unclear if all the animals were indeed killed and butchered by humans. More sites could be added to that list, but the evidence is often equivocal.

Cut marking is considered indisputable evidence that hominins butchered carcasses, and broken proboscidean bones sometimes are considered obvious evidence that hominins fractured the bones by percussion. Bone items modified into tools are unambiguously anthropogenic, but unmodified or minimally modified fragments can be contentious to interpret if they lack clear traces of human actions such as cut marks. A particularly divisive debate about the earliest human presence in the Americas centers on the possibility that fractured but otherwise unmodified mammoth and mastodon bones were broken by people before the Last Glacial Maximum (e.g., Bonnichsen 1979; Holen et al. 2017, b; Irving 1987; Johnson 2006, 2007; Morlan 1980; Waters and Stafford 2013).

The interpretive problems with the unmodified bone fragments have arisen because proboscidean bones in some sites might have been broken by deliberate human actions such as percussive fracturing, but conceivably some fragments also might have been produced during natural processes such as animal trampling. The ability to distinguish between cultural and non-cultural breakage is an essential starting point for archeological interpretation of proboscidean assemblages, and especially important for interpreting how humans interacted with the largest terrestrial mammals in their ranges.

While it may seem apparent that hominins recognized proboscideans as a resource for raw material and food since even before the emergence of the genus Homo, neotaphonomic research to support interpretations of proboscidean bone fracturing has been limited. Several experimental studies of elephant-bone fracturing have been carried out (Aguirre 1984; Backwell and d’Errico 2004; Biberson and Aguirre 1965; Bonnichsen 1979; Holen and Holen 2007, 2010, Holen K and Holen S 2017, Holen S and Holen K 2017; Holen et al. 2017; Stanford et al. 1981), but most were limited to few elements of varying freshness. Elephant bones are not readily available for experimental and analytical research due to severe recent declines in elephant populations in Africa and Asia and restrictions on international movement of elephant products such as bones. In spite of the difficulties, we think more studies are needed of both anthropogenic and non-anthropogenic fracturing of proboscidean bones, to provide archeologists with clearer standards for interpreting fragmentation patterns in fossil assemblages.

In this paper, we describe experimental studies of recent Loxodonta africana bones and comparative studies of Upper Paleolithic mammoth bones. Our aim is to provide a comprehensive view of proboscidean bone fracturing by comparing various taphonomic contexts in which broken bones are found. The data may help archeologists either strengthen or rethink claims that human actions broke fossil proboscidean bones, thus contributing to the development of theory about the evolution of human behavior. We also briefly discuss documented or potential non-anthropogenic bone-breaking processes such as traumatic injury in life, geological events such as debris flow, and disturbance from heavy earthmoving equipment. A separate paper has been completed on elephant bones fractured by trampling elephants (Haynes et al. 2020).

Materials and Methods

We report qualitative and quantitative records of (1) cortical bone thicknesses of elephant and mammoth long bone fragments; (2) the features created by experimental hammerstone impact on green, dry, and sub-fossil proboscidean bones, such as fracture outlines and fracture angles; and (3) the characteristic features seen on traumatically fractured long bones. We discuss and compare the fractured recent and fossil bones.

The materials we examined consisted of (1) 15 long bones of L. africana which we experimentally broke with hammerstone impact; (2) a sub-sample of eight of those long bones for which entire medullary interiors could be examined after breakage, along with four other L. africana long bones that had been traumatically fractured in life, exposing medullary interiors: (3) an estimated 100 long limb bones broken by trampling animals or gnawing carnivores with exposed interiors at localities in southern Africa where 403 elephants had died during drought (cited as 402 in Haynes 2018:121, table 6); (4) two Mammuthus cf. Mammuthus primigenius tibiae (one with unfused epiphyses and the other with both epiphyses fused) that were experimentally broken with hammerstone impacts; and (5) 21 specimens of archeological long bone diaphyseal fragments with exposed medullary interiors examined in curated collections at institutions in the USA, Poland, the Czech Republic, and Russia (listed in “Acknowledgments”).

We were interested in testing what features differentiate fractures made on green elephant bones from those on dry bones, such fracture angles, fracture surfaces, and other features. The features expected from percussive fracturing of green bones have been characterized mostly from experiments with non-proboscidean bones. Fewer experiments with proboscidean long bones have been described in the literature, as cited above. Taken together, these studies have established that certain identifiable features may appear on green-bone fragments of elements deliberately broken by hard hammers, such as spiral fracture fronts, acute fracture angles, smooth fracture surfaces, and impact marks on cortical surfaces (e.g., Blumenschine 1995; Blumenschine and Selvaggio 1988; Holen and Holen 2007, 2010, 2017a, b; Villa and Mahieu 1991).

The experimental fracturing was done three ways: (1) forcibly tossing a ~ 15-kg boulder from overhead at bones on the ground, (2) striking with hand-held 4-kg hammerstones against bones supported by stones at both ends or in the center, and (3) striking with hafted 2-kg and 4-kg hammerstones against bones with stone supports under each end or an anvil stone under the mid-diaphysis. Archeological finds of broken proboscidean bones sometimes have associated large stones, spheroids, or polyhedrons that might have served as hand-held, free-thrown, or hafted impactors; evidence for organic handles and binding has not been found in Lower and Middle Paleolithic assemblages. We wanted our experiments to create bone breaks with features similar to what has been found in fossil assemblages, namely diaphyseal fragments which are not extensively scarred by unsuccessful impacts. We also wanted to maintain control of impact placement, directions, and force. A stone-tossing experiment described below required >20 impacts before the bone fragmented, and the impacts left large deep marks on the bone surface. Such scarring has not been noted in archeological descriptions of fragmented proboscidean bones. The hand-held hammerstone experiments also required multiple blows to fragment sub-fossil bones, leaving numerous impact points. Another reported experiment (Backwell and d’Errico 2004:111) could not successfully fracture semi-fresh elephant limb bones with hand-held hammerstones. Because of the advantage to be gained in control of aim and force and the need to avoid badly scarring bone surfaces, we predominantly used hafted hammerstones to break elephant limb bones.

We evaluated the condition of each bone in the samples as green (defleshed but less than 3 weeks out of a carcass), unweathered but dry, or weathered, according to a scale of differences expanded from Behrensmeyer’s (1978) model of sequential weathering stages, as detailed in the next section.

Characteristics of Proboscidean Long Limb Bones

We begin by discussing (1) weathering stages specific to elephant limb bones, (2) the relatively unusual architecture of proboscidean long bone medullary interiors, and (3) variable cortical thickness in late Quaternary proboscidean long bone. These have important effects on the fracture dynamics of proboscidean long bones.

Weathering: a Process Which Affects How Proboscidean Bones Break

Weathering is the cumulative process of bone decomposition, usually occurring on subaerial surfaces. Bones in different weathering stages break in different ways. Behrensmeyer (1978) presented an often used descriptive categorization of the stages of bone deterioration, based on observations made in equatorial east Africa. The rate and degree of weathering destruction are affected by the length of exposure after death, the bone size, the type of skeletal element, the depositional context, daily and seasonal variations in temperature and humidity, chemistry of contacting sediments, and other variables (Andrews and Whybrow 2005; Behrensmeyer 1978; Conard et al. 2008; Gifford 1981; Lyman and Fox 1989; Tappen 1994). Studies in temperate regions and in cool settings (e.g., Andrews and Armour-Chelu 1998; Andrews and Cook 1985; Fosse et al. 2004; Haynes 1981, 2018) have documented that large-mammal bones undergo weathering processes at rates slower than those of bones in the tropics. In non-tropical latitudes, ultraviolet radiation is less intense than in the tropics, reducing the rate of bone deterioration. In higher latitudes, natural three-season refrigeration probably retards microbial decay.

Proboscidean bones in southern Africa deteriorate in a progressive sequence much like what has been described for other mammalian bones. Many elements of juvenile elephants weather faster than those of adults, and individual elements in a single skeleton will show different effects of weathering, as seen in other vertebrate taxa (Lyman and Fox 1989). Equally important is the depositional microenvironment of bones. The upward sides of proboscidean bones and tusks in southern African savannas at the southernmost limit of the tropics weather faster than the downward sides, sometimes being two or more weathering stages ahead of the downward side. Major differences in preservation and weathering between upward and downward sides of surficial bones also have been recorded in other biomes; for example, a Thule snow knife thought to be undisturbed for 800 years was found in the Canadian High Arctic with lichens and decay on its upside while the lower surface still looked fresh (Sutcliffe 1990:178, 179, fig. 5).

Brain (1981: 141) pointed out that drying cracks from weathering on some animal bones may be curvilinear. Many such cracks are noticeably curved on proboscidean humerus, ulna, mandible, innominate, and scapula, because the composite structuring of collagen fibrils and hydroxyapatite mineral crystals has curving morphology on different parts of the elements (e.g., see Brain 1981:141, fig. 150). The cracks reflect cortical bone ultrastructure, which is the arrangement of mineralized collagen fibril bundles, or the bone’s “split-line orientation,” meaning the structural orientation of microscopic bone matrix in cortical bone lamellae (see, e.g., N. Tappen 1969). If the weathered bones are trampled by large animals, the curving cracks may create curvilinear fracture outlines. Fracture surfaces may be mostly rough, although smooth parts of the surfaces are also possible in places.

The split-line orientation of the cortical tissue on relatively straight limb elements such as the femur and tibia are oriented mostly parallel to the long axis of the bone; therefore, those elements may develop the least curved drying cracks. Longitudinal microcracking of periosteal surfaces of long bones may lead over time to severe exfoliation of the surface.

Deterioration of bone surfaces due to weathering is a continuum. Behrensmeyer (1978) and others have divided the continuum into six stages, 0 and 1–5. Table 1 briefly describes the weathering stages through which L. africana long bones pass through in southern Africa. We use the Behrensmeyer (1978) stages as guides, but we also define intermediate steps in Table 1 to indicate transitional categories.

Table 1 Weathering characteristics of modern African elephant long bones in Zimbabwe. The stage designations are assigned when at least one aspect of the element (such as caudal, lateral, etc.) possesses these characteristics, even when other aspects of the element are classifiable into lower stages

Other elements such as scapulae, innominates, and ribs have different weathering characteristics. Ribs develop abundant microcracks in early weathering and may split lengthwise for half or more of their length as soon as weathering stage 1.

The effects of weathering on some subchondral bone can differ from the effects on limb elements and ribs. For example, the glenoid cavity of scapulae may develop a rougher “bubbly” texture in weathering stages 1 and 2 instead of microcracking.

Internal Anatomy of Proboscidean Long Bones

The medullary interiors of modern Elephas and Loxodonta long bones contain dense three-dimensional networks of hard tissue called trabeculae or substantia spongiosa anchored to endosteal surfaces (Fowler and Mikota 2008; Nganvongpanit et al. 2017; Shoshani 1996) (Fig. 1). The yellow marrow is contained within the thick trabecular network. This characteristic sets apart proboscidean long bones from those of most other ungulates except equidae and rhinocerotidae. The trabecular tissue varies in density and morphology throughout the lengths of proboscidean long bones, and on a microscale the structure may be unique to every individual. In some broken long bones relatively large trabecular bone struts extend from cranial to caudal or lateral to medial sides, such as shown in Fig. 1(a) at the proximal part of a femoral diaphysis. These thicker trabecular structures are additional reinforcements against stresses, such as at the proximal femur where the shaft is narrow from front to back. These enlarged trabeculae probably account for the numerous unsuccessful attempts made by Holen and Holen (2012) to break a Loxodonta africana femur near the proximal end using a hafted heavy stone. Other forms of trabeculae include mesh-like networks of fine bony tissue and relatively long subparallel spicules (Fig. 1(b) and (c)); both types attach to relatively smooth endosteal layers. Figure 1(d) shows a section of mesh-like trabecular tissue in a broken femur and Fig. 1(d¹) shows the same section after the mesh-like bone was removed by hand, revealing the smooth innermost (endosteal) part of the cortical tissue.

Fig. 1

Different trabecular structures in a broken left femur of an African elephant (see text). Frame (a) is a medial view; frames (b) and (c) are views of trabecular bone structure inside the element; frames (d) and (d¹) are views before and after trabeculae in one spot on a diaphyseal fragment not near a hammerstone impact point have been gently removed by hand. Note that in frame (d¹), the endosteal surface appears smooth after removal of the trabeculae seen in frame (d). Photographed by G. Haynes

Proboscidean long bone diaphyses are not filled with only marrow as are the long bones of most hoofed mammals. Edible marrow is contained within small spaces of the trabecular tissue. The marrow in diaphyses is called yellow marrow. Another kind of marrow, red marrow, is found near the ends of proboscidean long bones where blood cells are created, which is typical of mammals grown beyond the early postnatal phase of life (Mikota 2006:325). As discussed in Vettese et al. (2020, and references therein), edible yellow marrow has been and still is an important motivation for hunter-gatherers to fragment non-proboscidean prey bones as it is nutritious, easily digested, and pleasing to eat. From earliest times in prehistory, the yellow marrow in non-proboscidean bones have been exploited by hominins, perhaps by scavenging from carnivore-kills, an activity termed “percussive scavenging” (Thompson et al. 2019). The medullary interiors of long bones from most ungulates are packed with yellow marrow that is almost completely free of trabeculae and that can be accessed by either pushing or scraping it out from fragmented elements. But because of the dense trabecular material in proboscidean limb bones the edible yellow marrow cannot be extracted the same way as is done with bones of other herbivores; it is exceptionally difficult to extract in substantial and edible form.

It has been suggested that some proboscidean limb bones do have marrow cavities in them, analogous to the marrow-filled and trabeculae-free interiors of non-proboscidean limb bones. Boschian et al. (2019) found limited spaces in two CT-scanned femora and one tibia of Palaeoloxodon antiquus. The two largest cavities were < 1% of total bone volumes (Boschian et al. 2019:92). These spaces in the scanned Palaeoloxodon limb bones were much smaller than the spaces called marrow cavities in non-proboscidean limb bones. A photographic view into a broken Palaeoloxodon femur in Boschian et al. (2019:95, fig. 11) is captioned as having a “wide medullary cavity,” but we caution that dense needle-like trabeculae are dislodged when limb bones are broken, exposing smooth endosteal surfaces and making apparent open spaces appear much larger than they originally were (see Fig. 1 (d¹)).

Boschian et al. (2019:94) refer to a scene in an archived documentary film online showing the interior of an elephant long limb bone being chopped by an axe while human hands reach into the exposed medullary interior to grab dislodged chunks of pale material, described by a narrator as “soft, raw, and ready to eat” marrow, “eaten like candy” (https://archive.org/details/pygmiesoftherainforest, ~ minute 24:20). The little picked-out chunks may have contained trabecular bone, as suggested by the actions of a child who places chunks in the mouth, chews or sucks on them, and then throws away solid and presumably inedible residue that may have been trabeculae. The edible softer material would have been yellow marrow.

So far our studies have not found elephant long bones with spaces large enough to be called “marrow cavities,” which is the term customarily applied to trabeculae-free interiors of the long bones of most non-proboscidean ungulates. One L. africana femur in our experimentally broken sample from an adult female >30 yrs of age does have a midshaft space with no trabecular bone (Fig. 2), but this is not a true marrow cavity. The bone was in transitional weathering stages 1–2 (dry and without marrow) when broken by a single toss of an angular cobble with estimated mass > 5 kg. The trabeculae-free pocket is ~ 14–15 cm long, located directly below where the angular cobble had fragmented the bone with one blow. The impact dislodged the trabecular bone which fell out when the bone fragments were collected. We have seen in several other experiments that hard-hammer impact on whole dry bones dislodges trabecular tissue adjacent to the impact. All other Loxodonta long bones in our sample which were broken when green or dry contained no open spaces that were or would have been full of yellow marrow, and only very small voids in trabeculae were seen, varying in size from ~ 5 mm × ~ 3 mm × ~ 2 mm to almost unmeasurably small.

Fig. 2

Refitted fragments of the adult female African elephant’s left femur also seen in Fig. 1, broken by hammerstone when dry. The view (a) is medial, showing the mid-shaft region impacted by hammerstone at the point indicated by the arrow in (b), which shows the interior of the element; trabeculae in this area had been dislodged by hammerstone impact. Photographed by G. Haynes

We did find one M. primigenius long bone fragment with a small pocket free of trabeculae and which possibly contained only yellow marrow. Part of the space without trabeculae can be seen in Fig. 3; this was recorded in a fragment of a post-depositionally broken M. primigenius right femur from the Ruppersthal site (Austria) (Bachmayer et al. 1971; Fladerer 1997; Kubiak 1990) which was associated with undiagnostic lithics and had conflicting dates (21.5 rcy BP on loess, 11.6 rcy BP on bone). We do not know if this space was originally present in the bone or if it had been cleared of trabeculae after the bone was fragmented. The greatest length of the cavity is ~ 5.7 cm and its greatest width is ~ 2.0 cm, but the full breadth is not measurable. A rough estimate of volume is < 70 cm³, which is much less than 1% of what would have been the whole bone’s total volume.

Fig. 3

A possible marrow cavity in a diaphyseal fragment from a distal right femur of M. primigenius from the Ruppersthal site (Austria). Specimen 312-C4-1H, photographed by G. Haynes at the Vienna Natural History Museum in 2019, courtesy of Ursula Göhlich

The yellow marrow in proboscidean trabeculae would be a nutritious but barbed snack. A substantial dietary reliance on fresh yellow marrow contained within proboscidean long bones might have been possible if containers were available to catch marrow dripping out of split bones exposed to the sun or indirect heat, a practice observed in nineteenth-century southern Africa (cited in Yravedra et al. 2012:1064). Speth (2015) speculated that even as early as Lower Paleolithic times foraging hominins might have been able to boil bones in containers of perishable materials using heated rocks, but the evidence for wet-cooking is “limited and largely circumstantial,” in Speth’s words. McNabb (2019:805) has written that “sophisticated techniques…(would have been) necessary to extract the marrow,” such as solid cooking vessels placed over fires to hold melting fats from heated bone fragments. This technology does not appear in the archeological record until the Upper Paleolithic.

We do not doubt that “significant quantities of marrow” which could provide fat to hominin diets must be present in proboscidean limb bones (Boschian et al. 2019:93-94) but extracting it in edible form is not a simple process. Ben-Dor et al. (2011:5) assumed that cancellous fat in proboscidean bone marrow “probably would not have been extractable (by Homo erectus) in the absence of cooking.” Yravedra et al. (2012:1064) recognized that “extracting marrow from elephant bones could be laborious.” We agree with Boschian et al. (2019) who thought studies are needed of the costs and benefits of persistent breakage of proboscidean bones. Although we have found no published sources that record anything other than small spaces which might have contained marrow in proboscidean bones, much more research is needed to test bones from different proboscidean taxa and perhaps discover if any had relatively larger pockets of marrow without dense trabeculae. Such studies can settle uncertainty about proboscidean bones being a major source of inside-bone nutrients before the Middle Upper Paleolithic.

Cortical Thickness of Proboscidean Long Bones

Figure 4 a is a lateral cross section of a femoral distal diaphysis from an old adult female L. africana and Fig. 4b is a cross section from the femur of a mid-life adult male M. primigenius, illustrating cortical thicknesses at different places on the circumferences. The mammoth bone trabeculae were probably dislodged and removed when the bone was sawn apart. The measurements of cortical thickness were taken with calipers held at a right angle to the contact of imaginary tangents to the periosteal and endosteal surfaces. The transition between endosteal and periosteal bone is irregular and diffuse, and the measurements are the closest possible values. If made carefully, the measurements can be considered accurate if not precise.

Fig. 4

Frame (a) is the interior of a right femoral diaphysis from a small adult female African elephant in Zimbabwe, showing very different cortical thicknesses at three places; the inset shows where the cross section was made. Frame (b) is the interior of a sawn-off femoral diaphysis from an adult male M. primigenius, also showing very different cortical thicknesses at three places; unnumbered specimen from Dobrkow in Galicia, Poland/Ukraine, stored at 312/C4/3f in the Vienna Museum of Natural History, photographed by G. Haynes, courtesy of Ursula Göhlich. The innermost trabecular bone in this fossil specimen had been partially removed during post-excavation preparation

Table 2 reports measured thickness from a sample of long bone fragments of three proboscidean taxa, as directly measured or extrapolated from published photographs having scale bars. The mammoth bone fragments in Table 2 have been interpreted by some archeologists as broken in prehistory by humans using hard impactors. Direct measurements of cortical thicknesses were made with digital calipers; the caliper jaws were held open at right angles to the cortical surface. Extrapolated measurements were made using the same approach but taken from high-resolution photographs with scale bars.

Table 2 Cortical bone thicknesses of long bone fragments from L. africana, M. columbi, and M. primigenius. The L. africana fragments are from either drilled/cut complete elements that were in weathering stages 0–1 or 1, or green elements (weathering stage 0) that were experimentally broken by hammerstone. The mammoth fragments were measured at or near curvilinear fracture outlines where fracture surfaces were relatively smooth, which some archeologists might interpret as results of hammerstone-breakage when the bones were green. Thickness measurements for mammoth bones are followed by data in this order: (1) element name, sex (known or estimated, or unknown), age of the animal (as known or estimated), (2) part of the element measured, (3) site name and specimen number, (4) where and when measured. Measurements are rounded to one decimal place. ‘ISEZ PAN’ in the table stands for Instytut Sytematyki i Ewolucji Zwierząt, Polskiej Akademii Nauk (Institute of Systematics and Evolution of Animals, Polish Academy of Sciences)

Table 2 provides (1) 16 measurements of cortical thicknesses of long bone fragments from modern L. africana, either from complete elements drilled or cut through to expose the periosteal-to-endosteal distance, or experimentally broken in weathering stage 0 under controlled conditions in Africa (Haynes and Krasinski 2010; Krasinski 2010); (2) nine measurements of Mammuthus columbi long bone fragments with apparent green-bone breaks; and (3) 35 measurements from fragments of M. primigenius long bones with apparent green-bone breaks.

Morlan (1980, tables B1, B2, B4) provided other measurements of “wall thickness” on 127 “green-fractured” fragments of M. primigenius bones; these numbers are not included in Table 2 because it is not clear how the measurements were taken, such as along fracture surfaces at oblique or acute angles to cortical surfaces rather than at right angles, and how the contact between cortical and trabecular bone was determined. The largest “wall thickness” in Morlan’s tables are 46 and 42.8 mm, but ~ 2/3 of the maximal measurements are < 23 mm. The largest means of wall thicknesses of different kinds of fragments are 20.1, 22.7, and 23.3 mm, which are smaller than the mean of L. africana cortical thickness (= 25.1 mm) in Table 2.

Cortical thickness on some limb elements from large adult proboscideans such as Palaeoloxodon antiquus and M. columbi clearly are greater than thicknesses on analogous parts of some African elephant long bones, measuring up to 60 mm in the case of Palaeoloxodon (Boschian et al. 2019:94, referring to an unidentified element). But the point we want to make here is that mammoth bones which might be interpreted as human-fractured when green often do have the same ranges of cortical wall thicknesses as the experimentally broken bones of African elephants. The fracture dynamics of many fossil proboscidean specimens claimed to have been broken by human actions are substantially the same as the fracture dynamics of modern African elephant bones. Cortical bone tissue in proboscideans is a polyphasic solid which is somewhat plastic rather than elastic (that is, it can deform irreversibly up to a point, rather than reversibly, before breaking), but it largely fractures like a brittle material (Alms 1961; also see Rooney 1977:106). The bottom line is that cortical bone in extant and extinct proboscideans is made of the same materials which would have the same fracture responses to stresses and impacts.

Experimental Bone Breakage

A common prehistoric and historic method of extracting the nutrient-rich fats in non-proboscidean long bones starts with cracking open an element by striking it with a hard percussor, then removing the marrow (which is semisolid in healthy individuals) and sometimes breaking the fragments into smaller pieces which can be boiled to release fats and oil. Proboscidean bones are so much larger and denser than the bones of other animals and so much more packed with hard trabecular tissue that their fracturing must be directly studied; the breakage patterns and fracture characteristics of non-proboscidean bones are not applicable.

A limited number of experiments have been carried out with hammerstones to produce breaks in modern elephant long bones, with the aim of recording the physical characteristics of the fragments, such as shapes of fracture fronts, fracture surface textures and angles, and marks produced on cortical surfaces (e.g., Backwell and d’Errico 2004; Haynes and Krasinski 2010; Holen et al. 2017; Holen and Holen 2007, 2017a, b; Krasinski 2010; Krasinski and Haynes 2009). In one experiment Holen and Holen. (2007; also see Holen and Holen 2012) broke an African elephant femur with a hafted 4.5-kg hammerstone. The bone was described as “amazingly durable” (Holen and Holen 2012:97) and required 10 blows against a wooden anvil to break with a 4.5-kg hammerstone attached to a 1.2-m-long wooden handle. Most of the unsuccessful blows were directed onto the proximal end of the diaphysis just below the femoral head. Later blows to the middle of the diaphysis after the initial breakage further fragmented the element. These actions to break a limb element may not be identical to the actions used in prehistory. The locations chosen for hard hammer impacts may have varied on limb elements of different sizes and shapes, and also may have varied when the intention was to maximize medullary exposure versus producing suitable pieces of raw material to shape into implements. Studies of non-proboscidean bones have shown that percussive breakage patterns often do vary; for example, Neanderthal actions to access marrow from Rangifer tarandus (reindeer/caribou) bones were not standardized at two Middle Paleolithic sites in France (Vettese et al. 2018), whereas modern Nunamiut Eskimo people broke elements from caribou in very standardized ways (Binford 1981). The Nunamiut people impacted limb elements adjacent to distal and proximal epiphyses so that the epiphyses detached, leaving unbroken diaphyseal cylinders without bone chips or splinters driven into the marrow, which then could be extracted with a “marrow pusher” (Binford 1981: 158–163). This would not be a useful tactic with proboscidean bone because of the dense trabecular tissue in long bone interiors.

Experiments described in Backwell and d’Errico (2004) involved breaking eight limb bones taken from a young adult elephant five months after death and one weathered bone from a teenaged elephant. Breakage in those experiments was done in three ways: batting each bone against a rock, throwing each bone against a rock, or throwing a rock against each bone. Striking the bones with a hand-held hammerstone did not succeed in breaking any of the sample bones. These and other experiments with proboscidean bones have been too limited to reveal whether the different methods used to break the bones result in different frequencies and shapes of impact marks and notches on fracture edges, variables which have been recorded in experiments with non-proboscidean bones. For example, Blasco et al. (2014) compared non-proboscidean bones broken by hammerstones to specimens broken by batting against an anvil stone. The batted non-proboscidean specimens had fewer impact flakes than hammerstone-broken specimens, and irregular and overlapping notches located more often on transverse planes.

In our experiments, we deliberately placed impact points in the mid-diaphysis of green and dried African elephant long limb bones, which in most cases broke elements into two segments of roughly equal lengths and several smaller fragments. Our experiments with green elements produced fracture fronts which carried through most of the length of the diaphysis in both directions, stopping short of the epiphyses/metaphyses. This kind of fracturing is optimal for accessing most of the bone interior and typically produces long cortical bone fragments.

The Breaking of Proboscidean Long Bones by Hard-Hammer Impact

Figure 5 illustrates terms used here to classify fracture outlines on limb bones. Figure 6 shows how fracture angles on broken limb bones were measured with a goniometer/protractor. The trabeculae which had filled the medullary space in the particular elements in Fig. 6 were dislodged postmortem after the fragments dried, indicated by the abundant spicules of loose trabecular bone found with these diaphyseal fragments in the field; more of the dried and brittle trabeculae continued to fall out when the bones were transported and shipped for study in the USA in the 1980s. Figure 7 shows different kinds of fracture surfaces on bones broken when green or dry.

Fig. 5

Descriptive terms applied in this paper for common fracture outlines on limb bones. Illustration by G. Haynes

Fig. 6

Measurement of fracture angles on elephant limb bone fragments: obtuse (a), right (b), and acute (c). Frames (a¹), (b¹), and (c¹) show how those fractures angles are measured with a goniometer. Photographed by G. Haynes

Fig. 7

Frame (a) shows a smooth fracture surface on an African elephant’s long bone fragment broken when green by hammerstone impact. Frame (b) shows a mostly smooth with irregularities surface. Frame (c) show a rough surface. Frame (d) shows a very rough surface. Frame (e) shows a rippled surface (arrow) on a green bone fragment. Photographed by K. Krasinski and G. Haynes

A specialized lexicon has grown out of studies of percussion breakage of animal bones. Literature reviews are in Morin and Soulier (2017) and Vettese et al. (2020). Binford (1981), Blumenschine and Selvaggio (1988), Martin (1910), Pickering and Egeland (2006), and others have described features created by experimental hammerstone percussion on non-proboscidean long bones, such as percussion marks a.k.a. percussion pits, which are roughly circular superficial pits and/or striae (Blumenschine and Selvaggio 1988; Yravedra et al. 2018), and often impact flakes (a.k.a. cone flakes) made at striking points and sometimes driven into bone interiors and sometimes not detached. If detached, they leave behind “inner conchoidal percussion scars” (Pickering and Egeland 2006:463). Impact flakes were described by Binford (1981:163) as “short but rapidly expanding flakes inside the bone cylinder.” Binford (1981:164, 160, fig. 4.53) also described “depressed margins” where fractures are not carried through the bone. The word “splinters” has been used to refer to percussion-produced shaft fragments which retain < 100% of their original diaphyseal circumference (e.g., Buckland 1823; White 1992; Pickering and Egeland 2006:462), and are sometimes further defined as being much longer than they are wide (e.g., Turner II et al. 2001:25). Shipman (2018:121) thought the term is not “generally used to describe spirally fractured material,” although it is accepted by some researchers (e.g., Vettese et al. 2020). Hard-hammer impact also often produces notched fracture edges (also see Capaldo and Blumenschine 1994; de Juana and Domínguez-Rodrigo 2011; Galán et al. 2009), curvilinear fracture fronts, smooth fracture surfaces, acute fracture angles, and shaft fragments which preserve < 50% of the element’s original cross-sectional circumferences. Vettese et al. (2020) discuss the terms applied to percussion-broken long bones of non-proboscideans. We illustrate and define similar features recorded in our experiments with elephant bones below.

Krasinski (2010; Krasinski and Haynes 2009; Haynes and Krasinski 2010) described bone-breaking experiments with limb elements from an African elephant found dead near the umKwizizi river in northwestern Zimbabwe. Three long bones were collected a few days after the elephant died, and five others were collected two weeks after that, when most muscle and soft tissue except skin had disappeared due to autolysis and consumption by small carnivores, vultures, and arthropods. Three other elephant limb bones that were dry and slightly weathered also were collected from a different locality.

Most bones that we collected from the umKwizizi elephant carcass (a tibia and an ulna) were in weathering stage 0, but two elements had developed extremely fine microcracks after drying for about three weeks and were beginning the transition to weathering stage 1. All bones from the carcass had trabeculae-dense marrow inside that was semi-solid and probably edible. The elephant carcass had been subjected to temperatures that were mild in the daytime and cool at night, and forelimb bones had not been exposed to sunlight or drying for the first week after the animal died. Both femora had been cleaned of muscle and soft tissue by lions feeding over the course of ~ 3–7 days immediately after the elephant died of unknown cause.

Forensic studies (e.g., Green and Schultz 2017; Wheatley 2008; Wieberg and Wescott 2008) have shown that non-proboscidean animal bones in temperate and subtropical settings retain the specific properties of freshness when experimentally broken for up to four or more weeks after death; that is, the fracture angles, fracture surfaces, and fracture outlines are typical of green-bone fracturing characteristics defined by multiple authors (see Green and Schultz 2017 for references). Based on our observations of African elephant bones in Zimbabwe, we classify proboscidean long bones in weathering stage 0 and transitional stage 0–1 as green, in other words capable of fracturing as if green.

A defleshed right ulna in weathering stage 0 was fragmented by impacts from a 15-kg angular sandstone boulder which was tossed freehand from overhead against the lateral diaphysis while the bone was supported at each end by cobbles. More than two dozen hard throws were needed before the ulna broke apart, partly because each impact did not strike exactly the same spot on the diaphysis each time (Fig. 8), and partly because a stone support was not placed directly under the midshaft of the bone, where the impacts occurred, which allowed the substrate and epiphyseal ends of the element to absorb some impact force (Krasinski and Haynes 2009). The element broke from the combined stresses of repeated impact and bending of the shaft. Two shaft fragments with cortical surface were generated, both having large hammer or anvil marks.

Fig. 8

In frame (a), Bernard Sibanda tosses a 15-kg boulder against the lateral aspect of an African elephant’s right ulna in Zimbabwe. Frame (b) shows the lateral aspect of the ulna after it fully fragmented. Frame (c) shows the medial aspect after successful breakage. Photographed by K. Krasinski

In another experiment, an unmodified rounded 4-kg quartzite cobble was attached to a 1 m-long forked tree branch handle, and three blows from it were needed to fragment a green left femur from the umKwizizi elephant carcass. The first blow was partly deflected off the caudal aspect’s cortical surface, the second blow made a much larger percussion mark but a very short visible fracture front, and the third blow broke the diaphysis apart (Fig. 9). This femur had no support under the midshaft opposite the hammerstone blows. The element broke into five fragments > 1 cm in maximum dimension, and all had at least one hammer mark or one anvil mark.

Fig. 9

In frame (a), Zimbabwe National Park Ranger Anderson Munkuli uses a hafted 4-kg hammerstone to break an African elephant’s green left femur in weathering stage 0; (b) shows the impact marks on the caudal surface and fractures created by three blows. Photographed by K. Krasinski

A defleshed green left ulna in weathering stage 0 from the elephant carcass was broken by two blows from a hafted 4-kg hammerstone. The defleshed green right femur in weathering stage 0 from the carcass was first cracked by two blows from a 2-kg hammerstone hafted to the meter-long wood handle and broken apart by a third blow (Fig. 10a). Also broken were a defleshed green humerus in weathering stage 0, requiring four blows with the 2-kg hafted hammerstone (Fig. 10b) and a defleshed green tibia in weathering stage 0 requiring six blows from a person of smaller frame using the 4-kg hafted hammerstone (Fig. 10c). Other bones that were experimentally broken include two defleshed green scapulae in weathering stage 0, each of which required three blows of the 4-kg hafted hammerstone (Fig. 11).

Fig. 10

Frame (a) is an African elephant’s green right femur (in weathering stage 0) broken after three blows from a hafted 2-kg hammerstone. The arrow indicates the successful last impact. Frame (b) is an African elephant green humerus broken after four blows from a hafted 2-kg hammerstone. The arrow indicates the successful last impact point. Frame (c) is an African elephant’s green tibia with a large sidestruck diaphyseal flake removed after six blows from a hafted 4-kg hammerstone. The arrow indicates the direction and location of the decisive blow. Note the trabecular bone fragments displaced by the hammerstone impacts in frames (b) and (c). Photographed by K. Krasinski

Fig. 11

Frame (a) is a costal view of an African elephant’s green left scapula (in weathering stage 0) broken by two blows of a hafted 4-kg hammerstone; frame (b) is the same element after a third blow. Arrows indicate a notch on the fracture margin at an impact point. Photographed by K. Krasinski

Two other elephant femora and an articulated radius-ulna transitioning to weathering stage 2 were collected from separate localities when they were dry and without signs of grease, showing incipient microcracking. These were broken by single hammerstone blows. These elements had cobble supports placed under each end and another cobble support (= an anvil) under the diaphysis directly below where the hammerstone was aimed.

We hypothesized that accidental recent fractures of well preserved mammoth bones such as made by earthmoving equipment might resemble the features seen on broken green elephant bones. To test this possibility, two sub-fossilized mammoth tibiae (donated by the University of Alaska, Fairbanks, Museum of the North, Earth Science Department) were experimentally broken by hammerstone impacts. The unprovenienced adult right tibia and juvenile right tibia were presumed to be from Alaska, and likely date between > 40,000 cal BP and 10,000 cal (Guthrie 2006; Krasinski and Haynes 2010).

The juvenile tibia was broken using an unhafted hammerstone weighing 4.082 kg. The tibia midshaft, anterior aspect, was placed on an anvil and the tibia was handheld in place. Nine impacts were required to completely expose the medullary cavity of the specimen with the handheld unmodified hammerstone. The first four blows impacted the midshaft, posterior aspect, but failed to expand weathering cracks or create new cracks. Because the hammerstone was difficult to grip during impact, it occasionally slipped from the hand, reducing full striking force to the bone. However, the actor was able to repeatedly strike the same location on the bone with the hammerstone. Breakage initially occurred along the distal shaft and epiphysis with the seventh blow, exposing a small fraction of the medullary cavity. The eighth blow did not respond with breakage or cracking, and upon the ninth blow, the posterior aspect of the shaft shattered. In this instance, the anvil did not facilitate breakage because the bone was not directly placed on the anvil during the last blow.

The adult mammoth tibia was broken using the 4.082-kg unmodified hammerstone hafted to a handle constructed from modern materials. The bone’s midshaft anterior aspect was placed against an anvil and each epiphysis was placed on supports to stabilize the tibia and concentrate force. The specimen shattered on the first forceful blow. Substantial quantities of sediment within the medullary cavities poured from the cancellous bone as fragments were rotated and manipulated during analysis, thus demonstrating one of the ways in which depositional environments affect bone preservation and coloring. While it was buried, sediment seeped through bone pores and the nearly completely fused proximal epiphysis.

Results of the Experiments

Evidence of Percussion from Hard Hammers or Anvils

A typical percussion mark (a.k.a. impact mark) is a pit produced by hammerstone or anvil impact, defined by Backwell and d’Errico (2004:108) as an irregular depression with lifted or detached microflakes adjacent to the depression, and a field of composite striae within or near the depression.

In our experiments, every blow from a hammerstone produced a percussion mark on a green bone surface (Fig. 12a). Marks on the dry bones were less conspicuous, but the dark brown discoloration of the experimentally fractured mammoth bones made percussion marks easy to identify. Notably, the number of hammerstone blows per element and percussion marks on the fragments of each element are not linearly related, although there appears to be a weak bivariate relationship: some of the separate fragments of the broken elements did not have a percussion mark for each hammerstone blow to the element, depending on whether the number of blows required to fragment an element was greater than six (Krasinski 2010; Krasinski and Haynes 2009). We did not perform chi-square tests to predict if the number of impacts was related to the specific types of bone elements because cell values were too small for the different types of bones and also because different individuals wielding percussors had different strengths and aiming abilities in the experiments.

Fig. 12

Two percussion marks (= impact marks) on an African elephant’s green femur (in weathering stage 0), both unsuccessful in fragmenting the element, although very fine microcracking was created to the left of the larger impact mark. Hammerstone percussion marks are shown in (a) and an impact mark with striae is shown in (b). Photographed by K. Krasinski

In the experiments by Backwell and d’Errico (2004:112) mentioned above, in which less green (“semi fresh”) elephant bones were broken by free-thrown rocks or by smashing against rocks, half of the 107 flakes analyzed had “multiple large percussion marks.” In our experiments with green and dry bones, 319 fragments > 1-cm length were produced of which 220 had cortical surface on them and 99 (31%) of all fragments had percussion marks created by the hammerstone or the stone supports under the bones (Krasinski 2010). The 2-kg hafted hammerstone created 2–3-mm-deep percussion marks ~ 20 mm × 30 mm in diameter. The irregular ends of some hammerstones and the uneven surfaces of anvil stones created variably shaped pits and striae fields (Fig. 12b). Hammerstone impact produced striae fields much more often than anvil impact, but both hammerstones and anvils produced equal frequencies of pits without striae (Krasinski 2010; Krasinski and Haynes 2009).

The presence of hard-impactor marks was also noted on several fragments of the experimentally broken and flaked bones of a butchered captive African elephant named Ginsberg (Stanford et al. 1981; Stanford 1987; also summarized in Krasinski 2010:222-223). The results of elephant-bone experiments are similar in many ways to results of experiments with non-proboscidean bones (e.g., Blumenschine 1995; Blumenschine and Selvaggio 1988; Domínguez-Rodrigo and Barba 2006; Pickering and Egeland 2006). Pickering and Egeland (2006) suggested that a hominin-affected faunal assemblage’s total limb bone MNE from non-proboscideans will have identifiable percussion marks on bone surfaces. The same expectation holds for proboscidean bones broken by humans. Features such as percussion marks are further described below.

Ring cracks are semi-concentric arcuate cracks on cortical surfaces created adjacent to fracture margins. They may outline a percussive impact point or outline a point where static pressure was applied (Villa and Mahieu 1991:27, footnote) (Fig. 13).

Fig. 13

Semi-concentric arcuate cracks (= ring cracks) on a fragment of an African elephant’s green long bone broken by hammerstone impact. Photographed by K. Krasinski

Fracture Angles and Fracture Surface Textures Resulting from Percussion

Our experimental fracturing of green and dry/weathered proboscidean long bones allowed us to record three kinds of fracture angles on fragments: right, acute, and obtuse (see Fig. 6). The term fracture angle refers to the angle between the cortical face (periosteal surface) and the breakage face (fracture surface) of a fragment, as measured with a goniometer or recorded after scanning with 3D hardware and software. Fracture angles may be difficult to measure precisely because fracture surfaces are large and sometimes concave and wavy on proboscidean limb bone fragments, thus affecting the seating and placement of a goniometer arm. They also vary along the length of a fracture, as they do on non-proboscidean bones (Coil et al. 2017). We emphasize that we report only our experimental results, and that more experimental work may have varying results. More research needs to be done on a variety of different elephant bones to record fracture angle patterning when the bones are broken by hammerstones or other agencies (Moclán et al. 2019).

We also recorded four kinds of fracture surfaces on the experimentally produced fragments: (1) flat/smooth (which is really only a relative term since bone surfaces are never perfectly flat or smooth), (2) mostly smooth with irregularities such as minor undulations, (3) rough and very rough, and (4) rippled (see Fig. 7e). Any of the fracture surface types may be concave (Fig. 14), in which case measurements of fracture angles should be taken at multiple points along the fracture edge to account for ranges of degree variation.

Fig. 14

Frame (a) shows an African elephant’s green-fractured femur fragment with concave smooth fracture surface (indicated by arrow) and an acute fracture angle at the point of measurement. Photographed by K. Krasinski. Frame (b) shows a M. primigenius long bone fragment also with concave smooth fracture surfaces and acute fracture angles; photograph of specimen number 2412 from Trench B at the Gravettian site Kraków Spazista (Poland). Photographed by G. Haynes

In the experiments, specific combinations of fracture angle types and fracture surface textures were reflective of bone condition when fractured, either green/well preserved or dry/weathered (Table 3). For example, concave + flat/smooth and concave + mostly flat/smooth surfaces were often created when green bones were fragmented; however, concave + flat/mostly smooth surfaces also were recorded on parts of fragments broken when bones were dry. Fragments with fracture angles of ~90^o and flat/smooth fracture surfaces were recorded on bones broken when green, while some fragments of bones broken when dry had fracture angles of ~ 90^o but only with mostly smooth or rough fracture surfaces. Acute fracture angles were recorded on green bone fragments with flat/smooth and mostly smooth fracture surfaces. Acute fracture angles on dry or weathered bone fragments were never recorded with flat/smooth or mostly smooth fracture surfaces. Obtuse fracture angles were recorded on fragments broken when bones were green, with adjacent flat/smooth and mostly smooth fracture surfaces. Fragments of bones broken when dry had obtuse fracture angles only with rough or rippled fracture surfaces. Concave fracture surfaces were recorded with flat/smooth or mostly smooth surfaces on green bone fragments, and with mostly smooth or rough surfaces on dry bone fragments.

Table 3 Co-varying fracture angles and fracture surfaces on fragments of green and dry/weathered bones. The lower-case letter d in the table indicates the feature was seen on dry bones; the upper-case letter G indicates a characteristic recorded on green bones. The dash mark indicates no instances were seen

Notches Created by Hammerstone and Anvil Impact

With green proboscidean bones, both hard-hammer impact and hard-anvil impact may create semi-circular/arcuate offsets on fracture margins, called notches, and Hertzian cone-like flakes that are either released into the medullary cavity or remain partly attached (Fig. 15).

Fig. 15

Frame (a) is a medullary view of an African elephant’s green-fractured femur diaphyseal fragment with notches indicated by arrows at the right end; frame (b) is a cortical view of this fragment with arrows indicating anvil marks. Frame (c) is a closer view of the right end of the fragment in (a), showing a notch made by hammerstone (upper arrow) and a notch made by contact with the supporting anvil stone (lower arrow). Frame (d) is a close-up of the right end of the cortical view in (b), showing anvil marks near the notches. Photographed by K. Krasinski

The term notch is not a synonym of percussion mark or impact mark because notches are produced not only by impact but also by other processes, such as trampling or carnivore gnawing. Here we use the terms percussion mark or impact mark to refer only to marks on cortical bone surfaces, while the term notch refers only to a feature which could have been made by a variety of causes on the fracture margin. Domínguez-Rodrigo et al. (2014:140) speculated that “the extreme thickness of the dominant megafaunal remains [at the 1.34 Ma site BK4b, Upper Bed II, Olduvai Gorge, Tanzania] lend themselves to the creation of spiral fractures rather than notches when broken when fresh” (meaning green or recently defleshed). While our experiments with green elephant limb bones have produced ample numbers of notches, notches may not be created with every impact. Hammerstone-produced notches on proboscidean bones therefore may not be as numerous as on non-proboscidean bones.

Notches may have variable shapes and sizes, depending on the morphology of an impactor’s working end and the anvil surface, the force of impact, and bone freshness. Long bone cortical tissue is laminar and ply-like layers may be definable in the walls of impact notches, even on fragmented green bones, and this surface feature is especially visible on impact-fractured bones that are lightly weathered (Fig. 16). Turner II et al. (2001:25) termed this sort of fracture surface “stepped-and-ledge” caused by “breakage that occurs months or years after death.”

Fig. 16

Frame (a) shows an African elephant’s green-fractured tibia with notches made by hammerstone impact against the caudal aspect. Frame (b) is a closer view of impact notches (indicated by arrows). Frame (c) is a medullary view of a fragment broken off the tibia diaphysis with notches (indicated by arrows) made by contact with an anvil stone. Photographed by K. Krasinski. Frame (d) is a medullary view of a notch on a fragment of an elephant long bone broken by hammerstone impact when in weathering stage 1. Note that fracture surfaces are tiered where cone flakes detached on both the green bone and the dry bone; also note that trabeculae have been displaced by the percussive impact. Photographed by G. Haynes

Notch shapes on non-proboscidean and elephant bones can be similarly described. They are usually “broad and arcuate, with an acute release angle and negative flake scars” (Blumenschine and Selvaggio 1991:30; also see Capaldo and Blumenschine 1994 and Domínguez-Rodrigo and Barba 2006).

Pseudo-notches also may be present. They are arcuate, lack negative flake scars and adjacent impact marks, and have small changes in outline setting them apart from the rest of the fracture outline (Capaldo and Blumenschine 1994:744 – 745; Pickering and Egeland 2006:464). Most are only undulations in fracture margins; they are not produced by impact. Figure 17 shows a fracture margin on an elephant long bone with a pseudo-notch.

Fig. 17

An African elephant’s femur with a traumatic fracture that occurred in life, showing in (a) a pseudo-notch (indicated by arrow) on one of the refitted fragments, which was not created by impact. Three other views of the same pseudo-notch are shown in (b), (c), and (d). This elephant died shortly after fracturing its leg and was not butchered. Photographed by G. Haynes

Capaldo and Blumenschine (1994) analyzed the shapes and proportions of notches on non-proboscidean bone fragments broken by percussion and by gnawing carnivores, finding diagnostically different results. For example, percussion-created notches were broader than notches made by carnivores, and the carnivores in the sample tended to make notches on opposing sides of fragmented non-proboscidean bones. Villa and Bartram (1996) provided photographs of large-mammal bone fragments that were spirally fractured, notched, and flaked by cave hyenas in a French cave site. The hyena-made notches were present on opposing fracture edges but also along a single fracture edge. Buckland (1823:277, plate 23) illustrated a single tooth-notch on a piece of an ox tibia broken by a captive spotted hyena and a single notch on a piece of fossil bone probably chewed by a Pleistocene cave hyena (Fig. 18), again showing that carnivore notches are not always present on opposing fracture edges. Figure 19 shows examples of notches created by modern large carnivores which gnawed fresh non-proboscidean limb bones. Fernández-Jalvo and Andrews (2016, chapter 4) provide high-quality photographs and photomicrographs of notches created by carnivore teeth and other effectors such as hammerstones.

Fig. 18

Two fragments of large mammal bones. Specimen (a) is a piece of a modern ox tibia gnawed by a captive spotted hyena in Oxford, U.K., and specimen (b) is a piece of a Pleistocene bone from Kirkdale Cave in the U.K., chewed probably by a cave hyena. This digital image was edited from Plate 23 in Buckland (1823: 277), which appeared in Reliquiae Diluvianae published by John Murray of London. The source of the image is no longer in copyright

Fig. 19

The modern non-proboscidean long bone diaphysis fragments in frames (a) and (b) have fracture margins that were flaked between upper and lower teeth of spotted hyenas in Africa, as indicated by the arrows; similar examples of chipped non-proboscidean bones are illustrated in Villa and Bartram (1996). The modern non-proboscidean diaphysis fragment in (c) has a notch with a diameter ~ 1 cm, and also has flaking on the margin made by spotted hyena cheek teeth. Frames (d) and (e) show notches on a fragment of modern bison (B. bison) humerus gnawed by a wild wolf (Canis lupus). Note the superficial tooth-drag marks on the cortical surface near the notches in frame (e). All bones are from animals that died from predation and were not butchered by people. Photographed by G. Haynes

Domínguez-Rodrigo et al. (2007) found that notches on opposite fracture margins and adjacent incomplete notches with overlapping flake scars were much more abundant on non-proboscidean bones when carnivores created the notches than when hammerstones did. Domínguez-Rodrigo et al. (2015) suggested that the creators of notches may not be so clearly differentiated when equid bones are involved, because they are denser than most other non-proboscidean bones. Proboscidean bone is even denser than equid bone.

Holen et al. (2017:143, fig. 1) measured eight notches on fragments of M. cf. M. primigenius limb bones directly dated > 40 ka rcy BP from Yukon (Canada), and compared these to notches on experimentally broken L. africana bone fragments and on fragments of M. columbi bones from the Clovis-era Lange-Ferguson archeological site (USA). The notch breadths on mammoth bones varied from 13.19 to 49.54 mm and the depths varied from 2.62 to 9.35 mm; the range of breadth:depth ratios varied from 3.89 to 10.95, averaging 6.58. Krasinski (2010) measured notch breadths and depths on L. africana and M. primigenius bones which were experimentally fragmented by hafted hammerstones and free-thrown hammerstones and found the notch breadth:depth ratio always under 5. No quantitative studies have been done to distinguish how notches differ on proboscidean bones when made by different sizes of percussors or by other processes.

Thirty-seven fragmented pieces from our hammerstone breakage experiments on L. africana long bones had 41 individual notches where the hammerstones or anvils had contacted cortical bone to create the fractures. Figure 20 a shows one of our hafted hammerstones and a broken femur diaphysis with impact marks and a notch made by this hammerstone (Fig. 20b). Some notches were wholly semicircular in shape; some were incomplete. Both hammerstone-broken mammoth long bones possessed notches as well. Over 83% of notches were associated with percussion marks on green, dry, and subfossil bones. In Krasinski’s (2010) experiments, the mean maximum breadth of complete hammerstone-created notches on proboscidean bones (22.92 mm; 95 % confidence interval = 16.10–29.74 mm) was much larger than the mean maximum breadth of carnivore-created complete notches (11.87 mm; 95 % confidence interval = 10.08–13.66) on a sample of non-proboscidean bones (Krasinski 2010:326, table 5.15).

Fig. 20

Frame (a) shows the hafted hammerstone used to fragment the African elephant green femur shown in frame (b). Both (a) and (b) are shown at about the same scale. Note in (b) the pseudo-notches on the undulating fracture front to the right of the large impact-made notch. Photographed by K. Krasinski

Fracture Fronts/Breakage Planes

Villa and Mahieu (1991) summarized and annotated the criteria which are often put forward for distinguishing breakage made when human bones are green versus when they are in a dry or subfossil state. Dry-broken human long bones often have fracture fronts that are parallel and/or transverse to the long axis of the elements, as well as sometimes jagged fracture edges and right-angle offset fractures; fracture surfaces may be uneven and pebbly with micro-step fractures, while the fracture angles between the cortical bone surfaces and the fracture surfaces are mostly ~ 90^o. Green human bones when broken, according to this set of criteria, usually have curved fracture fronts and smoother fracture surfaces, with obtuse or acute fracture angles between cortical and fracture surfaces, and no right-angle offsets in spiral/helical fracture fronts. Some features usually considered diagnostic of dry-bone breakage on human bones also have been produced on green proboscidean bones broken by hard-hammer impact. For example, Fig. 21 shows a distal elephant femur broken by an unhafted 4.5-kg cobble in the Ginsberg experiment while the bone was green (Stanford et al. 1981).

Fig. 21

Distal part of an African elephant femur broken when green by an unhafted 4.5-kg cobble in the Ginsberg experiment (Stanford et al. 1981), showing a right-angle offset in a spiral fracture front. Photographed by G. Haynes in 2000 In the Paleo Bone Lab of the Smithsonian Institution’s Natural History Museum, courtesy of Dennis Stanford

The most common features used to diagnose green-bone breakage by humans such as fracture angle or breadth:length ratios of shaft fragments “are useful at the statistical, assemblage level but are not diagnostic for individual pieces” unless another support for hammerstone breakage is found, such as the presence of impact notches and percussion pits and grooves (Villa and Mahieu 1991:45). Domínguez-Rodrigo et al. (2015:80) recommended that measurements of angles on fracture surfaces/breakage planes on equid bones “must be interpreted cautiously” because of the relatively greater density of equid limb elements, and that “further experiments and a broader set of referential data” are needed. Likewise, some of the generally accepted criteria for distinguishing breakage of green and dry bones do not strictly apply to the even denser bones of proboscideans.

Different Results When Dry Proboscidean Bones Are Broken

Boldurian (2007) broke off a piece of an Elephas maximus femur in an experiment to replicate a mammoth-bone rod found at the Late Glacial archeological site Blackwater Locality No. 1 (New Mexico, USA). One proposition arising from the experiment was that dried bone would have been unusable for making reliable tools because of brittleness. This is not completely accurate. Even weathered elephant cortical bone is dense and can be worked into usable tools, although bones with the most advanced weathering would be too brittle for making into thrown or thrust projectiles.

Breaking Dry Bones

The breakage patterns from impact and static loading differ between dry and green proboscidean bones, as has been documented in our experiments carried out in Africa and the USA. When broken by dynamic impact or static loading, dry bones in weathering stages 0 and 1 still may have spiral/helical fracture fronts with relatively smooth fracture surfaces on some fragments, but there also may be linear (longitudinal) and irregular (transverse, sawtooth) fracture fronts. Right angle offsets in the outlines of spiral fractures have also been seen in the experiments on bones with light weathering. Microcracks are usually oriented linearly or curvilinearly lengthwise along diaphyses on dry long bones in weathering stages 1 and 2 and will re-direct fractures fronts that were generated by impact or static-loading, often creating more longitudinal fractures. The fracture surfaces on dry bones in weathering stages 1 and 2 which have been fragmented by dynamic or static loading are predominantly uneven and rough (Fig. 22a), even on curvilinear fracture fronts, and fracture lines may continue into or through epiphyses (Fig. 22b), which we have not seen on green proboscidean bones broken by hammerstone impact. Percussion marks on cortical surfaces of bones in weathering stages 1 and 2 and higher may be almost superficial and not as well defined as on green bone surfaces. Striae also may be created by hammerstones and anvils or trampling of dry bones. Impact/cone flakes and notches on fracture edges are possible on bones in weathering stages above 0, but the notching may be inconspicuous, irregularly shaped, and only suggestively arcuate (Fig. 23). Flakes from impact or static loading are not always identifiable as Hertzian cones (Figs. 23b and 24) in bones weathered above stage 0. Ring cracking is possible but not expectable on hammerstone-broken dry bones and is usually inconspicuous.

Fig. 22

Frame (a) shows part of the rough fracture surface on an African elephant dry femur (weathering stage 0–1) that was collected and broken by hammerstone 1–2 years after death. Frame (b) shows an adult African elephant ulna (distal end) that was in weathering stage 1–2 when broken by hammerstone 1–2 years after death; the fracture fronts continued through the fused distal epiphysis. Photographed by K. Krasinski

Fig. 23

Frame (a) is a cortical view of an African elephant ulnar fragment and (b) is a view of the mostly smooth but irregular fracture surface of a femoral fragment; both bones have inconspicuous notches produced by hammerstone impact 1–2 years after death. Photographed by K. Krasinski

Fig. 24

Partially refitted fragments of an adult female African elephant femur, collected in weathering stage 1–2 three years after death, broken 13 years later after storage in a climate-controlled environment. On the right is a close-up with an arrow indicating part of the impact area from a tossed ~ 5-kg cobble. Part of an impact flake is still attached below the impact point; this impact-made spall is not cone-shaped. The elephant had died of starvation and was not butchered. Note the lack of trabeculae directly below the percussor impact point. Photographed by G. Haynes

Experimental breakage of the two subfossil mammoth tibiae produced 97 fragments. The tibia from an adult mammoth broke into 27 pieces produced by a single blow from a hand-held hammerstone; all were > 1 cm in maximum dimension. The juvenile mammoth’s tibia required nine blows to fully fragment, and 56% of the fragments were > 1 cm in maximum dimension. Cortical bone was present on 36% of the juvenile mammoth tibia fragments and percussion marks were present on 31% of the fragments. Cortical bone was present on 74% of fragments from the adult mammoth tibia, but only 15% contained percussion marks from the single blow. The majority of the percussion marks consisted of discrete areas of battering or crushing, but occasionally percussion pitting was present, and striation fields were produced to a lesser extent. Battered percussion areas were consistently lighter in color than the darkly brown-stained cortical surfaces and were primarily associated with impact from the anvil rather than the hammerstone. All subfossil mammoth fracture surfaces were rough or rippled, with most fracture planes roughly at right angles (often slightly obtuse as compared to slightly acute). As with fresh bones, 83% of the percussion marks were associated with notches.

We introduce the term post-fracture notch (PFN) for an arcuate notch-like feature on a fracture edge which was not created by the dynamic impact that fragmented a bone, but afterwards. These are more clearly defined in outline than pseudo-notches. Figure 25 shows an example of a PFN on a femur from a free-ranging elephant in the Chirisa Safari Area of Zimbabwe. The bone had been traumatically fractured into multiple fragments which were still in place in the limb before death. Figure 25a shows three refitted fragments from the element. The elephant had been killed before any bone healing began. The bones were collected ~ 1 year after death, while some skin and soft tissue remaining on epiphyses were still greasy. Note there are no notches on the original fracture edges. The edges of fragments had been pressed against each other while the elephant was still alive, producing pressure flakes and rounded fracture edges, as seen in Fig. 25b and c. After death and natural skeletonization at the death site, a small piece of cortical bone had broken off along a split-line/fracture front, leaving a semicircular notch-like feature (=PFN) (Fig. 25d) which was not made by impact or carnivore teeth.

Fig. 25

A traumatically fractured femur of an adult female African elephant that was shot and killed in Zimbabwe to end its suffering; it was not butchered. Frame (a) is a view of three refitted fragments, collected ~ 1 year after death. The view in frame (b) shows pressure-flake scars on refitted opposing edges of two fragments which pressed against each other probably as the elephant attempted to walk after the break. In frame (c), one refitted fragment has been slightly pulled away from contact with the other one to show some of the rounding of the fracture edge on the other fragment. The view in frame (d) shows a circular notch-like feature (a PFN) created when a piece of cortical bone broke along a drying crack or an extension of a fracture front. The small piece of bone shown out of place in (d) was still in place when the bone was collected, as seen in (a), (b), and (c), and it has no indication of impact. It is a result of a brittle break after the bone was naturally cleaned of soft tissue. The PFN has inflection points but no negative flake scar. Note the lack of trabeculae which had been displaced postmortem. Photographed by G. Haynes

Static pressure is known to collapse and produce PFNs on fracture edges of brittle solids such as lithic flakes experimentally trampled by humans on ground surfaces (Driscoll et al. 2016), especially if the flake edge angles are acute (McPherron et al. 2014). We suggest the much heavier trampling of proboscideans would have been able to create notch-like features on existing fracture edges of proboscidean long bone fragments.

Summary of Implications of Bone-Breaking Experiments

The simplest one-step bone breakage requires direct impact with a hard hammer, probably at least 2–4 kg or more in mass for proboscidean long limb elements. Hafted hammerstones are most efficient, and thus produce the fewest percussion marks relative to handheld hammerstones. Unhafted impactors can be held in the hands, but the force of impact is hard on the hands and wrists. Prehistoric humans might have been stronger and more tolerant of the stresses from striking proboscidean bones with hand-held hammerstones. Heavy impactors also may be thrown at bones, if the aim is carefully made to strike a targeted spot on the diaphysis. An optimal breakage process involves placing the bone against a hard anvil opposite the intended point of impact.

Impact notches are not produced from every blow, which is true for all sizes of animal bones, including those of proboscideans (Krasinski 2010:252). Notably, they can be produced in similar frequencies on green or dry bone. Bending failure also can produce notches (Krasinski 2010:252). Impact-produced notches may be multiple or contiguous if more than one blow was required or if hammerstones had multiple striking points. An impact notch may or may not have a percussion mark immediately adjacent to it (Krasinski 2010:255). A successful impact that fractures a bone may drive a flake into the medullary cavity and not leave an impact mark on the surviving cortical surface, only a notch. Unsuccessful impacts on green bones do leave clear percussion marks. Impact notches may have smooth fracture surfaces or “tiered” surfaces (as in Fig. 16; also see Krasinski 2010:253, fig. 4.23b and d). The relationship between notch breadth and notch depth is a function of cortical bone thickness and the contact area of an impactor’s surface. A small notch breadth:depth ratio (< 2) on proboscidean long limb elements might result from a sharp impactor or from carnivore teeth.

Fracture fronts parallel to long axes do occur on hammerstone-impacted green bone, along with spiral/helical/curvilinear fracture fronts (Fig. 26); lateral/transverse fracture fronts also may occur, especially if fracture fronts reach metaphyses, as in Fig. 26a. Extreme bending can produce complex breakage morphology, with intersecting spiral/helical/curvilinear, lateral/transverse, and longitudinal fracture fronts (e.g., Fig. 26b); the midshaft diaphyses fragments often retain 100% of pre-breakage circumference. Bending failure is further discussed in a separate paper (Haynes et al. 2020). Pebbly fracture surfaces were prevalent on the sub-fossilized and dry remains, although we suggest that fracture patterns on the mammoth bones may have been different had the bones been broken immediately after their removal from permafrost and before drying.

Fig. 26

Frame (a) shows longitudinal and transverse fracture fronts from hammerstone breakage of an African elephant’s green femur (in weathering stage 0). The lateral/transverse fracture front occurred ~ 7 cm from the suture line between diaphysis and distal epiphysis, in the part of this element where the internal trabeculae were densest. Photographed by K. Krasinski. Frame (b) shows fractures from experimental bending of an African elephant’s green radius (in weathering stage 0). The distal half of the element was broken off in the experiment, described in Haynes et al. (2020). Photographed by K. Krasinski

Cone flakes (= Hertzian cone flakes, impact flakes, notch flakes) are produced at dynamic or static loading points on bones and tend to be more conical in shape on green bones. A less conical shape may be present on dry bones broken by dynamic hard-hammer impact. On dry bones, a definable bulb of percussion is usually reduced or absent and notch flakes may be wider and less arcuate. The shape of the impactor/indenter affects the flake produced; very sharp indenters may create fracture fronts which radiate outward from the center of loading on the more brittle bones (see Lawn 1993: chapter 8). Hammerstone impact-fracturing may remove patches of both endosteal and periosteal surfaces near or at points of impact (Fig. 27).

Fig. 27

Distal fragment of the African elephant green femur in Fig. 26a which was broken with a hammerstone. In frame (a) a small periosteal fragment that had been loosened by the hammerstone impact has been taken away to reveal partial removal of cortical surface, shown in closeup in (b). Photographed by K. Krasinski

Table 4 compares features observed on hammerstone-broken green bones and dry bones in our experiments that were not in advanced weathering stages. No single one of the features in Table 4 is exclusively diagnostic of green proboscidean bones broken by hard-hammer percussion.

Table 4 Proboscidean bones in weathering stage 0 and transitional weathering stage 0–1 had these marks and morphological characteristics after experimental breakage with hard percussors, using support (anvil) stones under the bones

Discussion

Empirical Evidence Put Forward in the Literature that Proboscidean Bones Were Deliberately Broken by Human Actions

Using criteria from experimental studies of percussive bone fracturing, Shipman (2018) examined 42 mammoth bone fragments alleged to show evidence for human modification in the Lange/Ferguson Clovis-era assemblage (South Dakota, USA). More than half of the fragments had spiral/helical/curvilinear fractures, assumed to have been created when the bones were green. The features (“macroscopic criteria” in Shipman 2018:131–133) which are claimed to support human-caused breakage of Lange/Ferguson mammoth bones include “predominantly fractured obliquely,” predominant “curvilinear or V-shaped fracture surfaces,” fragments preserving less than 50% of the shaft circumference, and fractures predominantly not perpendicular or transverse to the long axis of long bones. Although perpendicular and transverse fractures are thus considered as the expected results of breakage by heavy earthmoving equipment, under certain conditions fossil bones affected by earthmoving equipment do spirally fracture in the same ways as green or almost green bones do. The green-like breakage may occur if bones in sedimentary matrices which are waterlogged and anaerobic are affected by crushing, deformation, or impact from excavating tools, as seen or suspected at a number of non-anthropogenic sites containing broken mammoth or mastodon bones, further discussed below.

Does the Presence of Notches on Fragments Always Prove Hominin Involvement?

We recognize that some features on broken mammoth bones which are not associated with unquestioned artifacts are similar to features on dynamically broken recent elephant bone fragments— specifically notches, curvilinear fracture fronts, ring cracks, cone flake scars, and smooth fracture surfaces—and we acknowledge it is plausible that those mammoth bones were broken by human actions. But, in many cases, we are not completely convinced. We would like to see the fracture properties evaluated on all mammoth bone fragments from assemblages with broken limb elements, not just the few which have been emphasized in publications. One method to evaluate the fractures would be to calculate each fragment’s Freshness Fracture Index (FFI) number (Outram 2001), although this may not be the most ideal test since the Index was derived from study of non-proboscidean bones. The FFI number can be between zero and six and suggests the possible “freshness” ( = green-ness) of a fracture, based on assignment of values for each fracture surface texture (only two kinds are used in this method, 0 = smooth, 2 = rough), outline (0 = helical, 2 = non-helical), and angle formed with the cortical surface (0 = acute or obtuse, 2 = right angle).

A mammoth site with fractured and notched proboscidean bones but no unambiguous evidence for human involvement is La Sena (Nebraska, USA), dated ~ 22 ka; refittable fragments of a long bone were interpreted as broken when green by people using a hammerstone (Holen 2006). Haynes examined bone specimens at the Denver Museum of Nature and Science in 2008 and saw no clear percussion pits on cortical surfaces; however, some bone surfaces were weathered and root-etched, which would have obscured any percussion pits. No cut marks or scraping marks were seen. None of the bone fragments with flake-like morphology (one thick [= “proximal”]) end opposite a feather termination and possible bulb of percussion) have clearly visible crushing or deformation on the platform end from hard-hammer impact. No carnivore tooth marks are identifiable, leading to doubt that large carnivores were responsible for notching or breakage, although these marks also could have been obscured by the weathering and root-etching.

We applied the FFI system to La Sena femoral fragments such as the specimen in Fig. 28. Figure 28a shows that several fracture surfaces on this specimen are at right angles to cortical surfaces, and some of the fracture surfaces are very rough (Fig. 28b). There are notches on a fracture edge of another fragment from the site (Fig. 29a), caused either by dynamic impact or static pressure, but the adjacent periosteal cortical surface is too rough for possible percussion marks to be reliably identified (Fig. 29b). An FFI calculation for the fragment would be closer to 6 than to 0, which—if the FFI system is applicable to proboscidean bones—suggests the bone was not green when broken.

Fig. 28

A fragment of a La Sena M. columbi femur. The area outlined in frame (a) is seen in closeup in (b). This does not appear to be an impact notch. See text. Photographed by G. Haynes in 2008 at the Denver Museum of Science and Nature, courtesy of Steven Holen

Fig. 29

Frame (a) is a view of refitted fragments of a La Sena M. columbi long bone, looking into the medullary interior; one apparent notch has a tiered fracture surface, while the fracture surfaces on either side of it are smooth and may be wavy parts of the fracture front which we would call pseudo-notches not produced by impact. Frame (b) is a view of the periosteal cortical surface near the two notches seen in (a), showing rough patches (indicated by arrows) affected by weathering or diagenesis; they cannot be reliably identified as tooth marks or impact points. Photographed by G. Haynes in 2008 at the Denver Museum of Science and Nature, courtesy of Steven Holen

A more in-depth analysis using new methods such as applying the data we have reported and perhaps an elephant-bone-specific version of the FFI system may uphold the claims that the La Sena bones were indeed broken by human actions. New analysis also may weaken that claim. We encourage an active taphonomic interest in the assemblage and others like it in North America.

Examples of other ambiguously non-anthropogenic and clearly anthropogenic proboscidean sites have notched long bone fragments. The notches may be impact points from hard hammers or chopping tools, which would be evidence that people in prehistory broke the bones either for extracting marrow or to acquire bone fragments as raw material. One example is in Fladerer (2001:436, fig. 5), showing a multiply notched mammoth cortical bone fragment from the Gravettian archeological site Krems-Wachtberg (Austria), dated ~ 30.8 ka. Another example is from the Gravettian archeological site Milovice I (Czech Republic), dated 30.5–28.7 ka, where two femora were found with multiply notched fracture outlines (Oliva 2009:150, fig. 38; plate VI, figs. XXXIV and XXXV). A younger Gravettian site with a multiply notched mammoth bone fragment (Fig. 30) is Langmannersdorf (Austria), dated ~ 21 ka BP (Angeli 1952-53; Salcher-Jedrasiak 2012).

Fig. 30

Views of a fragment of a left femur from an M. primigenius excavated in 1919 at the Late Pleistocene Paleolithic site Langmannersdorf (Austria). In frame (b) the fragment has been rotated 90 degrees from (a), and in (c) another 45 degrees. The fracture surfaces are rough or mostly smooth with irregularities. The notches across the top of the fragment in these views were made on a longitudinal fracture edge and not on a spiral fracture edge. The two places on the cortical surface indicated by arrows in frame (a) are shown in closeup in frame (d). The right-most arrow in (d) indicates parallel grooves left by a hard edged object, possibly a lithic chopping tool or a steel shovel or trowel, which also may have made the other notches. Specimen 2009z0029-0269 at the Vienna Museum of Natural Science photographed by G. Haynes in 2019, courtesy of Ursula Göhlich

A fourth example is from the single mammoth found near Tomsk (Russia) (Late Glacial age) (Kashchenko 1901), where multiply notched bones (Fig. 31) were found near or with Upper Paleolithic lithic artifacts (summarized in Goebel 2004:314). Photographs of the bones in situ (Kashchenko 1901) (Fig. 32) show that shovels and pickaxes were used to uncover the skeleton along a steep bank-cutting. It is possible that breaks, notches, and surface marks resulted from the actions of the excavators in this and possibly other sites such as those mentioned above.

Fig. 31

Notches on fracture edges of a left femur and right humerus from the Tomsk (Russia) M. primigenius site. Note the lack of trabeculae. Images scanned and edited from Kashchenko (1901), published by the (Russian) Imperial Academy of Sciences. The source of the images is no longer in copyright

Fig. 32

Tomsk mammoth bones in situ during excavation. Note the shovels, picks, and digging bars. Scanned from Kashchenko (1901), published by the (Russian) Imperial Academy of Sciences. The source of the images is no longer in copyright

Notched bones of a Columbian mammoth were found at Duewall-Newberry (Texas, USA) (Steele and Carlson 1989), a Late Glacial site, but were not associated with lithic artifacts or features. The bones were recovered from a steep bank over a reservoir. Krasinski (2010:318-323) examined the bones and found that some notches on fragment edges compared well with carnivore-inflicted damage, but others were more similar to the kinds produced by humans using hammerstones on green or near-green bones. One possible bulb of percussion was noted. Steele and Carlson (1989:424, 425, figs. 15 and 16) illustrated fragments of a femur diaphysis with ring cracks. Krasinski (2010:371) identified five marks on three vertebrae from the assemblage as likely cut marks (probability greater than 50%) (we note that several “scores” on a rib and a humerus were inadvertently labeled as cut marks in two figures in Krasinski 2010:373, figs. 5.45 and 5.46). The lack of percussion pits, the lack of clear platform preparation or deformation on the mammoth bone flakes which were interpreted as knapped, and the ambiguity about notches do not disprove humans broke the bones, but it does allow room for reasonable doubt about that possibility.

Other possibilities do exist to account for notching on fossil proboscidean elements: For example, if percussion marks are missing on broken proboscidean bone surfaces, especially those with notched edges, and carnivore toothmarking is also absent, an often overlooked possibility is postmortem/post-burial pressure creating the fractures through dynamic or static point-loading on pre-existing fracture margins, such as would be done by heavy earth-moving equipment or excavation tools, as demonstrated for the Inglewood mammoth bone site (Haynes 2015a, b, 2017). A related process that might notch fracture edges is trampling—especially by proboscideans—which Mosquera et al. (2015:661) thought responsible for fragmenting and notching a scapula from M. meridionalis at the Lower Paleolithic site Barranc de la Boella (Spain).

Other Processes that Break Proboscidean Bones, and Which Can Be Mistaken for the Effects of Percussive Fracturing by Humans

Other processes can break elephant bones besides percussive fracturing by people using hammerstones. Some results of those nonhuman processes may be mistaken for deliberate human actions. Agenbroad (1989) suggested that eleven mammoth bones with green bone breaks at the pre-LGM Hot Springs Mammoth site (SD, USA) (now dated to > 100 ka) were fragmented either by animals falling, by torsional stress as trapped animals struggled to escape, or by trampling by living mammoths. Whatever the cause(s) of the breakage, “a small, but conspicuous percentage of bones” (Agenbroad 1989:141) had spiral fractures, which might be an expectable pattern in multi-mammoth assemblages affected by non-anthropic processes.

Other processes that might fracture bones and create features similar to those on hammerstone-impacted bones are traumatic fracturing in life, ice breakup on large rivers, debris flow, and disturbances from earth moving equipment. An additional possibility, post-depositional breakage by trampling, is discussed in another paper (Haynes et al. 2020) which also describes experiments to replicate trampling pressure.

Antemortem/Perimortem Fracturing of Proboscidean Limb Bones

Traumatic fracturing of elephant bones is not frequent but does occur in free-roaming L. africana populations. We use the term antemortem here to refer to fractures which occurred during life and which may or may not show complete healing; included in the category are cases of fractured elements showing some reactive bone growth that indicates survival for days to weeks after the injury. In forensics and the study of fossil bones the term perimortem is appropriate to use when referring to fractures which occurred at a time close to death and which show no signs of healing.

During 30 field seasons of neo-taphonomic studies in Zimbabwe’s protected areas that have large populations of free-roaming elephants, >1000 elephant carcasses or bone sites were examined; detailed notes could not be taken at all sites, so precise frequencies of antemortem/perimortem fractures of different elements were not systematically made. Examples of various skeletal elements with traumatic fracturing in life include an antemortem break on a juvenile mandible which was incompletely healed at the time of death (Haynes and Klimowicz 2015:139, fig. 3) and one or more ribs with antemortem breaks with partial or complete healing in an estimated 5–10% of adult elephant skeletal sites (Haynes and Klimowicz 2015:139, fig. 4). We estimate that less than 0.08% of adult elephants whose skeletal remains were most closely recorded in the field studies had suffered a traumatic break of a long limb element; it is possible that other skeletal sites which were much less closely examined also had traumatic breaks. Haynes and Klimowicz (2015:140, fig. 5) illustrated three examples of adult elephant limb bones with breaks that occurred in life.

No examples have been recorded in the wild of the complete healing of a fractured elephant long limb bone. Traumatic fracturing of a large limb bone in adult L. africana is thought to lead inevitably to death (Sikes 1971) within very few weeks at most; such a fracture in a fossil assemblage would be considered perimortem if no signs of healing are present. The breakage may result from a misstep, twisting of a leg while running, or falling (Douglas-Hamilton and Douglas-Hamilton 1975) (Fig. 33). Fractures in Zimbabwe elephant bones suggest that three stresses cause such breaks—twisting, axial compression (from the elephant’s weight), and bending—termed “torsion fractures” when seen in Equus caballus (Rooney 1977:110, 111, and figs. 177–181).

Fig. 33

A traumatically fractured African elephant’s femur in Zimbabwe showing intersecting fracture fronts. Photographed by G. Haynes

Spirally fractured limb elements with fragments having smooth fracture edges are especially interesting results of natural injuries documented among Zimbabwe elephants. The edge-smoothing might be mistaken for wear from tool use by humans. Figure 34a and b show an example of refittable fragments from a traumatically fractured L. africana limb element with smooth fracture edges created in life. Figure 34c shows three views of one fragment to illustrate differing fracture surfaces, outlines, and angles.

Fig. 34

Frame (a) shows refittable fragments of an adult female African elephant’s femur which had fractured in life; the elephant died before healing began and was not butchered. Frame (b) shows three refitted fragments from frame (a). Frame (c) shows a fragment with intersecting fracture fronts in three views (cranial orientation at the top, lateral in the center, medullary at the bottom). The V-shaped projection indicated by arrows is the only part of any fragment abraded by animals trampling the bone fragments against a fine-grained substrate. This femur was collected in the Chirisa Safari Area of Zimbabwe, about six months after the elephant died. This specimen is also pictured in Fig. 25. Photographed by G. Haynes

In-life fracture patterns on the broken limb elements are similar to patterns produced by hard-hammer breakage. The fracture morphology of breaks may be complex, with spiral, sawtooth, and linear fracture outlines, because of such factors as the existence of periosteum and other soft tissue impeding fracture fronts as well as post-fracture stresses exerted when animals attempt to use the broken limb. An example in Haynes (2002, p.142, fig. 3.14) shows three pieces of a traumatically broken proximal femur of an adult L. africana which was killed when discovered with the broken limb. With no supporting information at hand except this photograph, Karr (2015:101) described the illustrated specimens as bones that probably had experienced considerable drying and degradation. In fact, two of the three fragments in the illustration were broken apart while the bone was inside a living elephant; the third piece did separate after weathering. Figure 35 shows parts of this proximal femur with visible fracture fronts resulting from a major torsion break of the element while the elephant was alive.

Fig. 35

Fragments of an African elephant’s femur that was traumatically broken in life; the two fragments on the left split apart along drying cracks after weathering. The largest fragment has fracture lines (indicated by arrows) created by a torsion break in life. Photographed by G. Haynes

An adult male woolly mammoth dating 14.5–14 ka from Condover (U.K.) is known to have survived a major fracture of one scapula, possibly a tusking injury (Lister 2009). Some associated vertebrae also showed lesions that may have resulted from the spread of infection after the injury. The fracture had healed, leaving the scapula deformed. It is possible that proboscideans in other populations survived antemortem traumatic fracturing of major elements and managed to survive, although we think only juveniles and young adults are likely to live long enough for complete healing of largest bones, which is in keeping with what Bulstrode (1990; Bulstrode et al. 1986) suggested for wild primates.

Ice Breakup and Debris Flow as Possible Causes of Bone Breakage

Re-deposited large-mammal bones with fractures suggestive of green-bone breakage are numerous in the Old Crow River basin of Yukon (Canada), and have been controversially interpreted and popularized as broken by humans > 20 ka (e.g., Estabrook 1982; Holen et al. 2017; Irving 1987; Irving and Harington 1973; Irving et al. 1986, 1989; Morlan 1980). Large northern rivers in winter may be covered by ice that is 1–2-m thick, and during the usually powerful and violent process of spring ice breakup, ice floes with entrained sediments are capable of faceting, striating, and flaking boulders and cobbles in channels and on shores (Thorson and Guthrie 1984). According to Thorson and Guthrie (1984:188), “significant abrasion and violent impacts below debris-laden ice occur” during breakup. Although it would be extremely difficult to study proboscidean bones experimentally placed in river ice or in sediments that are eroded and redeposited during spring breakup on large rivers, we hope such necessary work will be done by dedicated taphonomists in the future, as this process seems capable of causing fragmentation of even the largest proboscidean bones.

Another possible non-anthropogenic cause of bone fracturing is debris flow, a sometimes violent geomorphic process in nature. Debris flow is a general term for the fast downslope movement of watery slurries of sediments, rocks, and entrained material such as bones, usually following heavy rains or rapid snowmelt. Debris flows may entrain undisturbed/articulated bones or reworked materials incorporated into older or younger strata (e.g., Hill 2018). A variant is called a lahar, specifically referring to movement of mud, water, rocks, and other entrained material initiated either by pyroclastic flow during volcanic eruptions or after eruptions when intense rainstorms activate volcanic slope sediments. An intense rainstorm, especially on sediments already water saturated, can result in semifluid landslides of churned debris moving 20 m/s or faster. A high-energy debris flow which involves alluvium and adjoining bedrock terrain may be a “mélange” (Radbruch-Hall 1978:631; Stanistreet et al. 2018:41) of randomly dispersed large clasts, cobble-sized and larger, which had been entrained in the flow along with the finer-grained material.

Whatever is entrained in debris flows will roll and tumble, and at higher velocities large clasts would be forcibly smashed together against other clasts. We suggest that sites with fragmented proboscidean bones should be evaluated in terms of the potential for breakage during debris flow, if evidence indicates such a geological event happened in the past (see Voight 1978). There are several possible examples, such as Byzovaya (Russia) with > 20 woolly mammoths dated ~ 33 ka (Heggen et al. 2012; Slimak et al. 2011), Hartley (USA) with one Columbian mammoth dated ~33 ka (Huckell et al. 2016), and Tocuila (Mexico) with seven M. columbi, dated ~ 11,000 rcy BP (Arroyo-Cabrales et al. 2003) (Fig. 36).

Fig. 36

Tocuila M. columbi bones in situ. The arrow points to an apparent articular contact of a mammoth femur and tibia. Photographed by G. Haynes in 2000, courtesy of Joaquín Arroyo Cabrales

The Tocuila assemblage was deposited by a volcanic mudflow (lahar) within lake margin sediments (Gonzalez et al. 2014; Morett Alatorre and Arroyo Cabrales 2001; Siebe et al. 1999;). Three fragments of mammoth long bone appear to have flake-and-core morphology (Arroyo-Cabrales et al. 2001, 2003) (Fig. 37), which has been interpreted as produced by humans quarrying the bones to make tools. The process of high-energy bone-on-bone impact and bending of some elements in the debris flow is here suggested as another possible cause of the bone breakage. A mammoth femur and tibia appear to have been still articulated when deposited, but other elements are jumbled.

Fig. 37

Frame (a) is a cortical view of a Tocuila M. columbi bone spall with a refittable smaller spall; frame (b) is a medullary view of the larger piece. The larger piece is specimen I-281; the smaller spall had no number. Photographed by G. Haynes in the Instituto Nacional de Antropología e Historia, Mexico City, courtesy of Joaquín Arroyo Cabrales

A possibly similar bone assemblage affected by high-energy sediment movement is Untermassfeld (Germany). The deposition of some anatomically connected limb bones in a mudflow has been observed for the heaviest animals at the 1 Ma Untermassfeld fossil site (Germany) (Kahlke 1999), as also seen at Tocuila. Limestone and chert rocks were redeposited at Untermassfeld into fossil-bone-bearing sediments through high-energy processes that caused “mechanical splintering and breakage of the…rocks” (Roebroeks et al. 2017, and references therein). The careful analytical study of this site’s fossiliferous deposits can serve as a model for examining other assemblages with associated broken bones and lithic clasts in fine-grained sediments. High-energy processes at Untermassfeld and other sites might have mixed together fine and coarse sediments to create the fossil deposits; some sites also could contain animal bones which were well enough preserved to be spirally fractured by the high-energy tumbling even while other limb segments remained articulated.

Post-Depositional Fracturing by Earthmoving Activities

Several discoveries of mastodon or mammoth bones are on record for which bones were claimed to be surprisingly well preserved. Warren (1855:185) discussed the remarkable preservation of some mastodon skeletons. The Oak Creek mammoth bones reported by Barbour (1925:108) were described as “tough as modern bones,” with “elastic” splinters. Another example of apparently extraordinary preservation is a nineteenth-century find in Indiana (USA) of large mastodon bones that were split open and their preserved marrow “utilized by the bog cutters to ‘grease’ their boots” (Collett 1881:18); if this is true, it implies the cortical bone might have been in a similarly excellent state of preservation.

Shipman (2018:121) has pointed out that no reports exist that those well-preserved bones actually did fracture as if green upon discovery, perhaps unwittingly implying that the only way to prove bones with preserved organic material could break as if green is to break them. Shipman is correct to be skeptical. The bones from Warren’s mastodon, the Oak Creek mammoth, and the Indiana mastodon (and others like them) have long since dried out, if they still exist, so they cannot be subjected to valid experimental hammerstone breakage. The only way to demonstrate that ancient proboscidean bones which were just pulled out of a preserving burial matrix would break as if green is to immediately break them before they dry out, which is pretty much unthinkable.

However, several fossil proboscidean finds do come closer to demonstrating that ancient bones in certain circumstances are well enough preserved to fracture as if green. Haynes was made aware of this possibility in 1982 when excavating the Inglewood (Maryland, USA) Columbian mammoth bones (Haynes 1991:235), and the possibility became a surety in 1994 while Haynes watched Pleistocene bones being exposed as a dragline bucket scraped up waterlogged interglacial sediments in a quarry pit near Shropham, Norfolk, UK. Figure 38 shows associated bones that were freshly exposed by a drag of the bucket and shovel cleaning. Figure 39 shows one of the exposed bones, a proximal left femur from an extinct rhinoceros (cf. Coelodonta). The bones came from sediments of Ipswichian age (MIS 5e, ~ 130–115 ka). The distal half of this femur was broken off by the earthmoving scraper. Two associated ribs were unbroken and two others were broken. Also associated were vertebral fragments and pieces of wood.

Fig. 38

Some associated bones of an extinct rhinoceros (cf. Coelodonta) dating probably to MIS 5, uncovered by mechanical excavator and shovel in a quarry pit near Shropham, U.K., including ribs, the proximal half of a left femur, and fragments of vertebrae. A piece of wood is also visible among the rib fragments. Photographed by G. Haynes, courtesy of J. Lightwing

Fig. 39

Frame (a) shows the proximal left femur from an adult extinct rhinoceros (cf. Coelodonta sp.) seen in Fig. 38, after breakage by a mechanical excavator. Frame (b) is a closeup showing ~ 58% of the mostly smooth helical fracture surface. The bone also has a series of experimental shovel-strike marks along the cranio-medial margin, inflicted after it was recovered. Photographed by G. Haynes, courtesy of J. Lightwing

The lower part of the proximal femur in Fig. 39b shows the recent contiguous damage from the earthmoving activity, including abrasive striae, ring cracking, and helical fracture morphology, as seen in the closeup. The Freshness Fracture Index (FFI; Outram 2001) for this break is between 0 and 1: the fracture outline is helical (0 value), the fracture surfaces are at acute and obtuse angles to the cortical surface (0 value), and most of the length of the helical fracture surface is not rough (total FFI value = 0.54). The helical fracture outline, the acute and obtuse fracture angles, and the relatively smooth fracture surfaces of this fragment appear comparable to the characteristics of the horse bone fragment illustrated in Karr and Outram (2012:557, fig. 3) which was labeled as exhibiting a “fresh [ = green] fracture.” The unbroken antero-medial edge of the fossil bone in Fig. 39a also has experimental marks made by striking the working edge of a short-handled steel garden spade (square-point shovel, unsharpened) against the diaphysis after the specimen had been cleaned of sediment.

Another case of green-bone breakage thousands of years after the bones were deposited is the Columbian mammoth site called Inglewood (Maryland, USA), where bones were fractured during earthmoving activity. The site was briefly mentioned in Haynes (1991:235-236) as a noncultural site where the breakage occurred long after burial, but more than two decades after that publication appeared some breaks were interpreted as ancient by Karr (2015), based on incomplete data from the site and experimental studies of non-proboscidean bones subjected to conditions very different from those that affected the Inglewood bones.

Another site where buried proboscidean bones were possibly broken by heavy equipment is Orleton Farms (Ohio, USA). A mastodon’s bones were found in “limy clay” below wet “muck or peaty material,” in a farm field in Ohio (Goldthwaite 1952; Thomas 1952:1, 3;). The heavy equipment may have been farm tractors or excavating machines. Thomas (1952:3) described the skeleton as having a “wide dislocation of the parts,” but a figure in the publication (Thomas 1952, fig. 2) shows a compact clustering of bones not significantly larger than the body size of a recumbent mastodon. The bones also appear to be in rough anatomical order. Thomas (1952:3) described the skeleton as “badly disturbed and the bones crushed and broken.” At least one long limb bone (a femur) was broken in midshaft. Although that break was described as being “squarely across” (Thomas 1952:3, fig. 3 caption), the visible break in a photograph appears curvilinear (Thomas 1952:3, fig. 3) (Fig. 40). As Thomas (1952:4) proposed, “some considerable force had been exerted upon (the broken bones),” but we think his suggestion of trampling by other mastodons is not a likely cause, because trampling has been observed to displace fragments (Haynes unpubl. field notes 1983-1997; Haynes et al. 2020). The broken halves of the femur were still tightly fitted together when exposed, which agrees with the speculation of Holen et al. (2017) who predicted that bone fragments should be adjacent to each other if they had been broken by extreme downward pressure while buried.

Fig. 40

Figure 3 modified from Thomas (1952), showing exposed Mammut americanum bones at the Orleton Farms site (Ohio, USA), used by permission. The digital closeup (frame b) shows the curvilinear fracture in the femur (indicated by arrow). The image in frame (a) was taken by a photographer for the Columbus Dispatch newspaper in 1949, later published in 1952 in the Ohio Journal of Science 11(7):3. Permission to reproduce the image has been granted by the Columbus Dispatch newspaper

Koons (2014) examined the Orleton Farms bones for cut marks, but the illustrated marks thought to be worthy of study are natural foramina and two small gouges. One gouge is ~ 1-cm long and the other is of unknown size (shown without a scale bar). None of the illustrated marks are cuts or gnawing marks made by rodents or carnivores.

Conclusion

We have pointed out the comparable fracture dynamics of bones from extant and extinct proboscidean. Experimental and neo-taphonomic studies of recent Loxodonta africana bones are valid data sources for understanding broken fossil proboscidean bones. Much more research of that type needs to be done, but we think our results are a useful start as cautionary guides for taphonomic interpretations of fossil proboscidean bone assemblages.

Our experimental breakage of elephant bones has shown that (1) percussion marks are always present on at least some of the fragments in an assemblage; (2) there are fewer notches than percussion marks on hammerstone-broken fragments of green bones; (3) fracture outlines can vary from spiral to linear/longitudinal/lateral on hammerstone-fragmented green bones, although spiral is likely; (4) fracture surfaces and fracture angles also can vary on hammerstone-broken green bones, but smooth/mostly smooth fracture surfaces and acute or oblique fracture angles are always present on some fragments; (5) dry bones broken by impact or pressure can have some of the same characteristic features of broken green bones.

Proboscidean long bones can be broken by a number of different agencies and processes, such as deliberate percussion by hominins when bones were green, trauma in life, violent debris flow events, and post-depositional pressure on well-preserved materials. Taphonomic analysts look for significantly different features on fractured proboscidean bones that will set apart the pieces broken by human actions from those broken by other processes. However, some of those features may be ambiguously defined or wrongly attributed, such as the possible causes of notches on fracture edges, the typical surface textures of fractures, and potentially diagnostic shapes of fracture fronts. Possible alternative causes of green-bone breakage should be further investigated.

The information reported in this paper can be a methodological tool applied to analyses of assemblages with fractured proboscidean bones, or in re-analysis of classic deposits such as at Torralba and Ambrona (Spain) and FLK N6 (Tanzania). Further research and analyses may strengthen the existing claims that humans were responsible for fractured proboscidean bones in those and other sites; another outcome may be a rethinking of some claims.