SIX

Torralba and Ambrona

A REVIEW OF DISCOVERIES

INTRODUCTION

In the 1960s, F. Clark Howell began a program of multidisciplinary investigations at the Spanish Mesetan sites of Torralba and Ambrona that quickly became classic. Torralba and Ambrona retain among the best-preserved, most carefully excavated, and informative mid-Pleistocene localities known from Western Europe to the present day. It is my belief that in the future these excavations will be increasingly recognized as among Howell’s foremost contributions.

This chapter reviews the work of the team that excavated and analyzed Acheulean residues and bones at Torralba and Ambrona under Howell’s supervision and outlines the implications of the analysis of those residues.

The conclusions reached by Howell’s team in the 1960s seemed interesting but unexceptionable at the time. Careful attention to microstratigraphy revealed several stratified levels of paleontological and archeological materials. Intimate spatial associations of tools and faunal materials and some otherwise seemingly inexplicable marks on the bones suggested that humans had visited the site to hunt or at least to butcher large game animals, although it was always recognized that some of the animals could have died natural deaths without human intervention and might have had nothing to do with hominid scavenging or butchering at all. We stated that elephants were neither the exclusive nor the principal object of human attention: that other animals, especially horses, were as abundant or more so in some levels. We detected and recognized geologically caused rearrangements of residues, some due to faulting and more impressive ones due to freeze-thaw cycles in a harsh climate, and we were attentive to the possibility of winnowing and realignment due to flow in sheets and channels, though we could not detect edges of channels in any part of the Torralba Acheulean deposits. We carried out extensive analyses for paleoenvironmental reconstruction, including the contour-mapping of old temporary surfaces. We knew that carnivores had been present at least occasionally at both sites and had occasionally gnawed at a bone. Worked wood—cut and hacked, and sometimes charred—was recovered, as was charcoal in abundance, but nothing we could definitely identify as a hearth.

Statistical analyses indicated that the visible spatial associations we could see were part of larger patterns of consistent and repeated frequency relationships. In the 1970s, T. P. Volman showed that the frequency relationships detected when whole levels were compared had a spatial component in individual levels, that different sets of stone tools and body parts were consistently found in different parts of the ancient landscape: marshy waterlogged low-lying areas were the loci of death and discovery of carcasses, and the loci of preliminary disjointing of body parts. Higher areas were the setting of intermediate stages of butchering and bone breaking. Still higher and drier were the few situations where final processing of carcasses took place, with some amount of stone flaking or tool repair (Freeman 1978).

Throughout the 1960s and 1970s there was little in the way of challenge or contradiction of those interpretations, although new studies of site formation processes and taphonomy suggested by the late 1970s that some revision of interpretations was necessary. Beginning in 1981, however, those conclusions were disputed, often with little or no justification and less regard for the facts. Certainly conclusions reached 30 years ago can stand a deal of revision in light of new information and criticism. But the conclusions Howell’s team reached about Torralba—conclusions for which I take a major share of responsibility—were not simply evaluated and evenhandedly criticized; the interpretations were distorted by the critics to become unrecognizable caricatures, and the caricature then savaged. Now some say that results from the Torralba/Ambrona excavations, where “faunal assemblages are in disturbed context” (Villa 1991: 206), are unreliable or at least suspect. To advance science, those critics advocate dismissing our results, to rely instead on other sites, whose deposits are in fact no more intact, whose stratigraphy is no less complex, whose age is no less uncertain, whose samples are smaller, whose excavations were if anything less carefully controlled, and whose excavators have proposed interpretations no less “simplistic” or “anecdotal” and “unsystematic” (Villa 1991: 202, 204) than those we proffered.

The best answer to criticism comes from the sites themselves; were the data they provide better known, much of the debate about them would evaporate. A chapter of this length unhappily cannot do justice to excavations whose results require substantial monographic publication. The final monograph on Torralba, finished in the early 1980s, has been ready for press for some time, and at one time was even accepted for publication. Its appearance has paradoxically been delayed several years due to just such misconceptions about the site and its residues as a prompter publication might have dispelled.

Despite the deplorable impression that very little about the sites has seen print, part of the information to be reviewed here has long been available. For Torralba alone, there are more than a dozen largely nonrepetitive articles in English, based specifically on the analysis of recovered materials and distribution patterns from the site; together they total more than 300 pages. Though there are fewer sources about Ambrona, some quite extensive preliminary treatments of our work there have appeared (for example, Howell, Butzer, and Aguirre 1963; Howell and Freeman 1982; Howell, Freeman, Butzer, and Klein 1992). This chapter reviews aspects of the research conducted at Torralba and Ambrona during the 1960s and 1980s, in light of the most salient questions that have been raised.

The chapter is intended as a clarification of the record, not a debate with critics: truth is usually not well served by rhetoric. In the few passages where irritation mars the presentation, I beg the reader’s indulgence with my loss of patience. I intend simply to state the facts about Howell’s (and later, our) excavations as I understand them and to make it clear that for any understanding of mid-Pleistocene adaptations in mid-latitude Europe the data they offer must be taken into consideration.

Dismissing the results of research at Torralba and Ambrona is unwise—it would mean casting aside a great deal of important information about the nature of environmental change, site-formation processes, and hominid adaptations. Even the most vocal critic of our work cannot help admitting, in the midst of a slighting comment about my procedural inadequacies, that “the interaction between hominids and faunal remains seems clear. In fact, the results are not in conflict with the results that Freeman obtained” (Binford 1987: 95). Encouraged by so forceful an advocate, even an analyst as short-sighted as I cannot fail to be hopeful that a new overview will sharpen our vision of the significance of these Mesetan sites.

THE EXCAVATIONS

History of Research

On June 17, 1909, the Marqués de Cerralbo visited the hamlet of Torralba del Moral near Medinaceli in Soria, Spain. There, in 1888, trenches cut for the Madrid-Zaragoza railway had revealed bones of extinct Pleistocene mammals, including huge elephants. To his surprise, Cerralbo found Acheulean handaxes and other stone tools in association with these remains (Cerralbo 1909, 1973a, 1973b). This high-altitude site (1,113 meters above m.s.l.) was soon famous as one of the earliest human hunting stations known from Europe, though Cerralbo did not live to see it published in extenso. After Cerrabo’s death in 1922, despite the site’s recognized importance, no one returned to explore it until the 1960s.

It was of course Howell who initiated new fieldwork. In 1961 he also rediscovered the Ambrona site, an analogue to Torralba, situated at a slightly higher elevation (1,140 meters above m.s.l.) about 3 kilometers away. Though Cerralbo located and tested Ambrona some time prior to 1916, it was only known from the briefest published references (Obermaier 1976: 190, 1925: 180), before Howell’s work.

Beginning in 1961, Howell directed three seasons’ excavations at Torralba, removing most of the site sediments left intact after Cerralbo’s extensive excavations. The seven-week 1961 season proved that a portion of the site was still undisturbed, yielding hundreds of animal fossils and scores of stone artifacts. However, the site stratigraphy proved much more complex than suspected in 1961, when most finds seemed to come from a single archeological level. As a third-year graduate student at Chicago, I joined Howell’s team as an assistant in 1962, and after an initial three-week period excavating at Ambrona with Howell and Dr. Pierre Biberson, I spent six weeks working at the Torralba site. Again in 1963, I spent a month digging at Ambrona—Thomas Lynch supervised work there until his departure in June—before undertaking ten weeks’ work at Torralba as site supervisor. Emiliano Aguirre undertook limited excavations at Ambrona in 1973 to improve the on-site museum. As co-director with Howell and the late Dr. Martín Almagro, I returned for full-scale excavations at Ambrona in 1980–1981. Howell alone directed one last season there (1983), in which other excavations at el Juyo in Cantabrian Spain kept me from participating. In total, the 1980s excavations lasted 203 days.

Size of Sites and Exposures

The Torralba site was much smaller than its sister, Ambrona. When Cerralbo began work, it may have extended over as much as 3,000 square meters or perhaps slightly more. Although he gave a much lower estimate of his exposures at Torralba, we learned that he had in fact opened at least 1,000 square meters. For a careful, modern excavation, Howell’s fieldwork was also undertaken on a very large scale, as its duration would suggest. In 1961, he exposed approximately 450 square meters over the site surface, and during 1962 and 1963, we dug another 576 square meters all told, evidently in a richer and stratigraphically more complex part of the site. (While we left some intact sediments at Torralba as a witness, they are neither contiguous nor easily accessible.) At Ambrona, the largest European Acheulean site known at present (more than 6,000 square meters were “intact,” in 1962), Howell’s exposures through 1983 attained the truly impressive extent of some 2,800 square meters.

Excavation Techniques

Methods of excavation employed at Torralba and Ambrona from the 1960s on were as close to state-of-the-art as Howell or I could make them under the circumstances. Obviously, appropriate tools and techniques must vary with the nature of deposits, the availability of water, and other factors. At both Torralba and Ambrona, some levels are fine clays that when dry come away in chunks, separating cleanly from the finds they encase, while others are fluviatile/colluvial deposits that are sometimes sandy and friable, sometimes indurated to a near rock-hard consistency. At both sites, excavators used small pick-hammers on the indurated and clayey sediments, as well as knives and trowels, “crochets,” and brushes of several kinds as digging tools.

The excavation was mostly done by workers—farmers from the surrounding hamlets—but they were as well trained and capable as most students I have since had on field crews of my own; some became as technically virtuous excavators as any I have ever known. They were adequately overseen. One trained student assistant supervised the four workers excavating in two contiguous squares, advising them as needed, measuring, drawing, excavating in particularly delicate situations, and so on.

In any excavation (if the excavator is honest), some materials are inevitably recovered “out of context,” and Torralba/Ambrona are no exception: some finds were made whose level was known but whose exact horizontal position, orientation, and so on were indeterminate. Others were recovered in screening. These pieces were not plotted, but bagged by square, sector, and level.

At all times, our excavations were conducted with careful attention to provenience and microstratigraphy. The procedures used were not perfect. They never are. But I do not see how anyone could have done a much better job of excavating Torralba and Ambrona than Howell and his crews. Speaking as one who has at least as much experience directing meticulous excavations as any other Old World prehistorian, I find no contradiction of that evaluation in the work of others since then. The excavations were visited and inspected by a large number of first-class excavators and sedimentologists. The methods used were praised at the time—in the 1980s no less than in the 1960s. Only one person has seen fit to challenge our procedures; I can only characterize her comments as both uninformed and unreasonable (Villa 1990).

Sediments and Stratigraphy

At both sites natural levels of deposition are primarily differentiated by texture, and we were as scrupulous as possible in detecting minor changes in sediments as we proceeded. At Torralba, there are 10 “major” archeological horizons that seem to have accumulated in colluvial screes, or in and on dry channel deposits, or in fans along a pond margin, or in the marshy shallows of a pond. Many recovered bones show localized—rarely complete—polish or abrasion, perhaps by waterborne sand flowing over partly buried pieces; some elongated pieces have polish or abrasion restricted to one or both ends, like the wear on expedient butchering tools of bone from some U.S. buffalo jumps described by Frison (1991: 302–308). Most archeological residues were found atop former temporarily stable surfaces, at the contact between layers of sediment that differed texturally. Very occasionally, the presence of a continuous sheetlike horizon of bones and artifacts within an otherwise uniform level was the only indication of a former temporary surface.

Field designations of levels differ from the final occupation designations (these are “final” only at Torralba). As excavation progressed, and some levels were subdivided (or in some cases where relationships between different spreads of material were temporarily unclear), field designations became cumbersome (“B4aa,” “Occupation X,” etc.). Microfaulting required that final correlation of Torralba levels be done in the laboratory, using all maps and sections as well as three-dimensional stereoscopic plots and models of the site. Publications that appeared at different times reflect these changes, which has produced some confusion on the part of readers.

The earlier excavations of Cerralbo, in the form of wide trenches, cut through the upper site deposits, producing an interruption in distributions in all the levels affected. Though his trench did not always remove the basal levels, where it was present it appears as a blank in the distributions, and that gap reappears in the same area in maps of all the levels affected. It appears, and is clearly labeled, on the partial map of Occupation 7 published by Freeman and Butzer (1966). Binford, in the course of an ad hominem attack, suggested that the distribution gap, and the resulting apparent alignment of materials along its edges in several levels, may be “the structured result of differential erosion,” adding that “Freeman never considered this possibility since he already assumed the hominid behavioral cause of his structured results” (Binford 1987: 58). On the contrary, I have never suggested that the gap means anything other than Cerralbo’s trench, or that its structure is due to prehistoric cultural behavior. It is characteristically careless of Binford to suggest both that I have done so and to offer the innovative “reinterpretation” that the disturbance is instead really some sort of stream channel.

There were several cases of detached islets or spreads of archeological material that occurred atop a single temporary surface but were separated from each other horizontally by interruptions or large gaps in item distributions, sometimes caused by removal of the intervening surface by Cerralbo’s excavations, or happening to coincide with a zone of microfaulting. Since the Torralba stratigraphy is so complex, with fluviatile/colluvial levels pinching out laterally or merging to produce a single surface where there were two before, I still believe that the only safe practice, in the absence of some obvious proof that the islets are contemporaneous (such as finding, in different islets, conjoinable bone or stone fragments, or bones from the same identifiable individual), was to keep their contents separate for analytical purposes, even when it seemed likely that they had accumulated at “approximately the same time.” Where solid evidence of contemporaneity was lacking, separation was consistently our practice.

There were nine such “sublevels” all told: four (designated 1a–1d) on the Level 1 surface, apart from the major contiguous expanse of Level 1 itself; two in each of Levels 2 and 4; and one in Level 3. Four of the individual spreads involved were quite small, but five were large enough to provide considerable material. Even without counting these segregated islands of material that occurred in the same deposits, the natural archeological strata distinguished in the field were finer and more numerous than the geological units of deposition recognized by Karl Butzer, who analyzed the sedimentology at both sites.

Vertical and Horizontal Control

We excavated by natural archeological strata, also recording absolute depths to the nearest centimeter below an arbitrary horizontal datum, whose position was marked on stakes in each square. Leveling (within a square) was sometimes done with line levels; at other times “parallax triangles” were used, following the practice of the late François Bordes. In the 1980s excavations at Ambrona, an optical level was used for vertical control.

Horizontal control was provided by a grid of 3-meter squares, and all visible finds in each square were located with tapes and plumb-bob and piece-plotted at a scale of 1:20. When two pieces were found in direct and intimate contact, they were often given the same feature number. This was explained in notes on the plans, and in the site log or inventory such finds were differentiated as necessary by adding letters to the feature number. Unless the pieces were themselves very similar, the letters were not always placed on the piece labels themselves (there was little reason to do so, since the inventory was expected to resolve any possible confusion).

Orientations and inclinations of pieces were generally visible from or noted on the plans, but where recovered pieces were markedly disconformable to the lay of the stratum that contained them, special measurements, photographs, and notations were made. Naturally, we drew continuous sections showing both geological and archeological levels following all square walls, and abundant photography documents our procedures and finds and the stratigraphic distinctions we made.

Screening

At Torralba there was no available water for wet-sieving or washing finds (and in the 1960s, the needs of the Ambrona farmers for garden irrigation kept us from using the trickle of water seasonally available in the Río Ambrona below that site). Contrary to some of the critics (surprisingly, these include Klein: see 1987: 22–23), we did dry-screen samples of sediments at Torralba. In the 1962 excavations at Torralba, small amounts of sediment were sporadically passed through round screens with a mesh of about 5 millimeters. In 1963, we more systematically screened 15–20 percent samples of sediment (by square and level) from the archeological horizons, and 100 percent of the sediment from three selected squares designated as controls (screening did not include any of the later culturally sterile deposits overlying the archeological levels, though that procedure might also have been informative). In that year, the screens used were specially constructed large rectangular ones, still with a 5-millimeter mesh. The requirements of backdirt disposal dictated the technique employed: we unbolted the screens from their stands, lay them over wheelbarrows, and shoveled excavated sediment through them directly into the wheelbarrows. (Figure 6.1, taken to document the appearance of one edge of Cerralbo’s trench through the area we excavated, shows a screen and its stand.) Screening at Torralba yielded a disappointingly small amount of material.

At Ambrona, in the 1980s, in addition to dry-screening, it became possible to wash sediments in bulk through fine-mesh screens in the stream below the site. This, of course, permitted more complete retrieval of small finds, including microfaunal remains. The richest source of finds was the clayey pond/marsh sediment, much better represented at Ambrona than at Torralba. It is, however, noteworthy that washing did not yield appreciable quantities of small flaking debris.

RESULTS AND INTERPRETATION

Paleoenvironments and Site Formation Processes

The archeological deposits at Torralba and Ambrona were studied by K. W. Butzer in 1962 and 1963, and he returned to Ambrona during the 1980–1981 field seasons. What follows is a brief summary of his results, focused particularly on the Torralba site, digested from his most recent treatment to appear in the forthcoming monograph. His interpretation of the nature of the sedimentary column at that site is based both on his field examination of morphology, sediment sizes, and particle or item orientations, as well as on macroscopic and microscopic analysis of 77 sediment samples taken during the course of excavation and later processed by Dr. Réné Tavernier in Ghent. I have intercalated the results of the pollen analysis, based on identifications by the late F. Florschütz and J. Menéndez-Amor, to add relevant vegetational detail to the paleoclimatic picture. They analyzed 161 samples (of which many were sterile) taken in two partially overlapping series at 10-centimeter intervals through the Torralba column.

The archeological horizons are found in cold-indicative Pleistocene sediments lying above Triassic Keuper clays. Later lubrication and deformation of the plastic Keuper resulted in a series of microfaults, with thrusts of a few centimeters to as much as a meter, that affected the site sediments.

Following a series of sterile units formed under cold conditions, Member II^b of the Torralba Formation, up to 30 centimeters of coarse, subangular to subrounded gravel, incorporating fine lenses of clay, was formed (“A-Gravel”). The deposit suggests a frost-weathered detritus transported over some distance. Cobbles and larger rocks have been rearranged into stone rings of 25- to 40-centimeter diameter on slopes of 2–5 degrees, elongated into ellipsoidal “garlands” of rock on slopes of 5–10 degrees, and on even steeper slopes torn apart into stone stripes, perpendicular to the contours, or scatters in which individual pebbles, either point downhill or lie parallel to the contours. These are typical “patterned ground” phenomena of periglacial upland environments, attributed to seasonal or diurnal freeze-thaw cycles. The stone rings and garlands are contemporary with the accumulation of the A-Gravel or with the human occupation directly on top of them. Rare artifacts are found reworked in the gravels and clay lenses of this unit. However, the earliest archeological level coincides with the immediate surface of this gravel and appears to be coeval with local lenticles of light gray clay that indicate a shift from high-energy slope mobilization to low-energy subaqueous sedimentation. A single pollen sample from this clayey layer shows high AP values (76 percent), predominantly Pinus silvestris. Sphagnum spores suggest poorly drained or boggy ground near the site, while sedges are absent. The fact that the NAP is essentially all grasses, with a trace of Artemisia, indicates open vegetation on drier plateau surfaces nearby.

Size distribution histograms for rocks from all the archeological levels at Torralba reveal an abnormal frequency of large stones in this level, either indicating far more effective frost-shattering than can be found in recent analogues, or that the larger stones were concentrated in the site through human activity.

FIGURE 6.1. Screens at Torralba (1963)

The presence of stone rings, garlands, and stripes more or less contemporaneous with the earliest archeological level raises the question whether solifluidal transport or sheetwash disturbed the cultural associations of this particular horizon. From the orientation and dispersal of bones, it is obvious that some sliding has taken place, particularly on slopes exceeding 10 degrees. However, the limited rolling or wear of articular bone surfaces, the lack of size sorting of bones or artifacts, and the nearly articulated position of bones of single animals, all argue that, in general, such sliding has not destroyed the validity of cultural associations—independent of orientation.

Other archeological levels lack soil-frost structures and have not been so extensively disturbed. The best occurrences are in semiprimary context.

Most of the Acheulean occupations are concentrated in levels in Member IIc (Lower Gray Colluvium), disconformably deposited atop the “A-Gravel.” The A-Gravel is absent in the western sector of the site, where Unit IIc rests directly on earlier deposits, the contact distorted by congeliturbation structures. There are several well-stratified subunits and facies in Member IIc that range from gravel layers to unconsolidated, white to light gray or pale brown gritty sands with lenses of fine gravel and sandy silts. Periodic halts in deposition or episodes of erosion interrupt these deposits. This unit is a quasi-horizontal graded valley fill, with abundant fragments of thin-shelled aquatic gastropods (see later discussion) in most finer facies. Current-bedding is visible in some of the fine-sediment subunits. The gravel facies of unit IIc is characterized by angular to subangular shapes, containing some 19 percent pebbles fractured during transport. This suggests very short transport distances but only an intermediate intensity of frost-weathering. There is no evidence of soil-frost structures. Limonitic staining and mottling band the sediments, showing water-table fluctuations. Since these stains do not conform to the lay of the deposits, the water-table changes happened after the site deposits accumulated and even after some microfaulting took place.

A number of cobbles and boulders, varying in major diameter from 20 to 55 centimeters, were probably carried into the site area by Acheulean people, and numerous archeological horizons of variable area are found throughout this unit.

The range of horizontal facies from sandy clays to gritty sands, with some current bedding and discontinuous rubble bands, combined with the aquatic gastropods, suggests a predominantly fluvial depositional environment. A low- to moderate-energy stream crossed parts of the site, and incorporated some slope rubble during periods of intense overland flow, while ponding was not uncommon farther down-valley, at least during the early phases of accumulation. Climate was quite cold, but not as severe as during accumulation of the A-Gravel, and surface denudation was less vigorous.

Pollen spectra attest a cycle of shrinkage and later recovery of a swamp or lake near the site, and continued very cold conditions. Arboreal pollen drops to 36 percent before rising again to its former level. At that point pine forests must have been reduced to scrubby stands in a largely grassland environment. Other (rare) tree species are those that would fringe watercourses or ponds/lakes nearby. Preservation is unusually good for plant material, and bits of wood as well as other material are preserved. Identifications of macrobotanical remains, by Dr. B. F. Kukachka of the Wood Products Laboratory in Madison, Wisconsin, are mostly of conifers, among which Pinus silvestris is predominant—presumably brought to the site by humans—but they additionally include one bit of birch and another of Salix or Populus, that could have been obtained locally. Chenopod pollen increases with grasses in the middle of the series, when attractive and nutritious pasturage was most abundant. Sedges are represented in the earliest and latest pollen samples in the sequence, while Artemisia is always present. Increasing desiccation in the midpart of this member was likely due to physiological drought during the cold season, not to a total drop in precipitation: the presence of water lily in one of the grass-rich samples betrays the (perennial) presence of standing water from 1 to 3 meters deep.

The final bed of Member II rests on an eroded surface, attaining a thickness of 90 centimeters in a former topographic hollow in the northeastern part of the site. This “Brown Marl” bed is a compact, light gray to brownish gray marl, intermixed with lime-sand or grit. Diffuse limonitic staining as well as reddish-yellow mottling indicate oxidation in a zone of fluctuating water table. Some cryoturbation festooning is present. A ponded stream channel, spring seep, or the margin of a swampy floodplain is implied. Slope denudation was minimal and the environment was more temperate as well as wetter than during accumulation of Unit IIc, but never as benign as it would be during a full interglacial. Archeological materials in the Brown Marl are very localized in their occurrence. Pollen samples show an initial peak of AP (80 percent +), declining thereafter. The last Acheulean occupation at Torralba occurs in this unit.

At Ambrona, however, occupations continue into Units IV (Upper Gray Colluvium and Gray Marls) and V (Rubefied Colluvium) of the Torralba Formation. Unit IV begins with moderate-energy fluvial deposition, becoming increasingly low-energy, and attests cold conditions with intensive seasonally concentrated runoff at first. At Ambrona this unit is terminated by gravels indicating a return to higher-energy conditions. There follow marly mixed slope and fluvial accumulations, indicating intensive seasonally concentrated runoff under cold conditions (Gritty Gray Marls) and then the Upper Gray Marls, low-energy ponded or lacustrine deposits in more temperate conditions (though temperate, climate was still some 5°C colder than today). Last come the stratified, in part lenticular, deposits of Unit V, resulting from moderate-energy footslope and valley-margin accumulation by surface runoff and frost-assisted gravity transfer, including alluvial fans at Ambrona. Intensive frost-weathering and vigorous denudation took place on higher slopes, with incomplete vegetative mat (very cold). Dr. Thure Cerling noted that small red quartz crystals in the gravels of this unit were so fresh, and their surfaces so free from abrasion, that they could not possibly have traveled far by hydraulic action (personal communication in Toth, in litt.). The final Acheulean occupation at Ambrona took place during the first of the moister episodes in this unit.

There are several distinct Acheulean occupations in the Ambrona deposits just as at Torralba (though they may be fewer in number); since the distributions are still not completely analyzed, they have been grouped into two larger sets in earlier descriptions: those from the Lower Unit and those from the Upper Unit.

Butzer notes that most of the major archeological horizons at both sites are found in seasonally active, valley-margin deposits, in close proximity to permanently wet ground. However, a minority of archeological levels—more at Ambrona than at Torralba—occur within more clayey swamp- or pond-edge sediments themselves, as though shallow water or waterlogged marshy areas were sometimes used for the accumulation of or disposal of archeological residues.

Though Butzer estimates that the accumulation of the Torralba Formation sediments may have taken some 125,000 years, and the Acheulean deposits may date between very roughly 420,000 and 450,000 BP, it must be noted that the deposits and the archeological materials they contain were not accumulated continuously, as Binford (1987) seems to suggest, but rather episodically; long periods of nondeposition and some erosion, and even longer periods when neither artifacts nor animal remains were accumulating in the site deposits, were followed by relatively brief moments of active site use by animals and/or humans, and then by other periods of disuse.

None of the occupations at Torralba is a pristine intact association in true “primary” archeological context, and if earlier papers have not made that sufficiently clear, it has not been our intention to deny it, as some secondary sources seem to suggest (see later discussion).

Size of Samples

If the density of finds at Torralba and Ambrona is not particularly high for well-excavated sites of their age and type, neither is it especially low. The very large size of exposures, coupled with good preservation of organic materials, should suggest that sample sizes of recovered artifacts, bones, and other materials of all kinds are likely to be larger, not smaller, than “average.” At Torralba, 2,141 bones and 689 stone artifacts were excavated during the 1962–1963 field seasons alone.

I find Villa’s (1990: 307) observation that this sample size is too small and sparse for reliable statistical analysis puzzling to say the least. It betrays a surprising ignorance of statistics; worse, it is fundamentally illogical, since she finds no such fault with the much smaller samples from the Aridos quarry localities, which together are less than half that size. In fact, from the published evidence, I see no more reason to believe Aridos a convincing intact butchery site than to consider Torralba the same. At Ambrona, Howell’s investigations produced vastly larger quantities of varied occupation residues: over 2,085 fragmentary remains of the single taxon Elephas, and more than 1,400 stone artifacts, have been found in the Lower depositional unit alone to date.

Lithic Artifacts

Stone artifacts are, of course, one principal evidence of human activity at Torralba and Ambrona. Various aspects of the lithic assemblages at these sites are interesting: the raw materials used, the composition of assemblages, the presence of wear traces, spatial associations with other evidence, including conjoinability, and relationships in abundance of specific sets of tools and particular animal species or body parts, are all informative in their respective ways.

Freeman (1991) provides a more detailed discussion of raw material use at Torralba. None of the raw materials used for stone tool manufacture at either site is local. Three basic kinds of stone are represented: cherts/chalcedonous flints, quartzites of variable grain size, and limestones. Although there are outcrops of porous limestone a few hundred meters from either site, they are not really suitable for tool manufacture and were not used. The Triassic clays underlying the site contain no stone raw material. The closest stone sources are suitable limestones a few kilometers from the site; the quartzites used are found no closer than 10 kilometers away, and the flints and cherts would have had to be transported scores of kilometers to the sites. One distinctive and rare kind of flint seems to have been imported from the Jalón drainage, more than 50 kilometers from the site. The Río Ambrona flowing past that site has none of this material in its bed—it could not, for the source is across the divide separating the site from the Ebro drainage, several kilometers downstream on that side. The most probable sources of commoner raw materials are downstream from the Torralba site in the Tajo/Duero drainages. Raw material from any of these sources would have had to be transported upstream to reach the sites, so it must have been imported by humans. At Torralba, aside from the fact that flints are not frequently used to make bifaces, the finer cryptocrystalline materials—the flints and cherts—were not especially chosen for the manufacture of smoothly retouched working edges such as sidescrapers.

From the 1962–1963 excavations at Torralba, there are 689 stone artifacts, of which 63, or about 9 percent of the total, are geologically crushed (rather cryoturbated than rolled) pieces on flakes. Though they are or once were artifacts, their original typology is indeterminate, so they have always been excluded from detailed analysis of the stone artifact collections, leaving 626 identifiable artifacts. The total includes 1 battered polyhedron and 5 hammerstones (1 percent). Thirty cores and discs make up 4.8 percent of the collection. There are 36 or 5.8 percent bifaces, and 212 or 34 percent shaped flake tools. Minimally retouched/utilized flakes, 160, are 25.6 percent, and unretouched so-called waste, another 159 pieces or 25.4 percent: together they compose 51 percent of the artifacts in the combined collection. When just the shaped tool collection—the 212 flake tools plus 36 bifaces—is considered, bifaces are 14.5 percent of the total for all levels. Scraping tools (60) are 24.2 percent, notches (21) 8.5 percent, and denticulates (48) 19.4 percent of the shaped tool series. There are small proportions of burins (5.2 percent) and backed knives (0.8 percent), while perforators and becs are more frequent (10.9 percent). Two points were recovered. About 4 percent of the pieces are raclette-like artifacts with continuous abrupt retouch on much or all of the circumference. Unclassifiable variants (usually multiple-edged, prismatic-sectioned pieces) are quite numerous—10.1 percent of shaped tools.

From my counts, the lithic collection from the 1962–1963 excavations at Ambrona (all units) is more than twice as large: 1,520 total pieces. These were apparently not all included in Howell et al.’s earlier (1992) summary. The counts that follow are complete for the years in question: I studied the Ambrona artifacts piece by piece when they were on loan to the University of California in the 1970s.

My records show geological crushing to be much less evident than it was in the Torralba series: most of the 199 pieces with coarse abrupt retouch may well be heavily utilized, rather than cryoturbated. But, since the threshold of differentiation between deliberate, irregular, coarse retouch, and geological crushing is hard to draw consistently, they are excluded from the remaining calculations, leaving 1,321 undoubted artifacts. The 50 cores make up about 3.8 percent of that total. Minimally utilized flakes are 212 (16.1 percent) and waste flakes another 636 (48.2 percent) of these: together they constitute just over 64 percent of the collection. The “waste” series included 14 biface trimming flakes and a pseudo-Levallois point. Shaped tools are 391, or 26 percent of the total. The proportion of shaped tools is smaller than at Torralba, and other differences between the two sites also appear. The 47 bifaces (including 3 roughouts) make up 12 percent of the shaped tool collection, scrapers are 36.6 percent (more than at Torralba), notches 13.3 percent, and denticulates 14.8 percent. While notches are more numerous and denticulates less so than at Torralba, their summed percentage representation is about the same at the two sites. The proportion of unclassifiable tools is smaller (only 1.5 percent—multiple-edged pieces are rarer), while burins, perforators, and alternate burinating becs (1.2 percent) are about equally well represented in this shaped tool collection.

Despite the opinion of some authors, such figures—particularly the proportion of bifaces and ratios of unretouched or minimally utilized pieces to shaped tools—are not in any way anomalous for well-excavated Acheulean assemblages from stratified contexts. The proportion of bifacial tools is not particularly low, nor is it uniform from occupation to occupation, While in some units at both sites, there are few bifaces or none at all, there are major occupations with more than 15 percent bifaces (Torralba Level 3), and in Torralba Level 2a the total is nearly twice that (the Level 2a collection is very small). The proportion of waste and minimally retouched pieces would probably be considered low for sites located near contemporary sources of good raw material, but the stone at Torralba (as at Ambrona) was all imported from some distance—some of it from scores of kilometers away, as noted. There is very little evidence for primary flaking or workshop activities at either site, as one might expect from that fact alone. Nor would one expect a great many (but see later) conjoinable pieces at these sites, as compared to the situation at the Aridos or Pinedo quarries, where sources of good stone in reasonably large sizes were readily available locally as cobbles from river terraces—a point that I have tried previously to make, apparently without much effect (Freeman 1991).

While we have called the rather idiosyncratic Torralba artifact assemblage “Late Early Acheulean or Early Middle Acheulean” (in litt.), Santonja and Villa consider them typologically later in the Middle Achuelean, comparing our better formed bifaces to the cruder pieces from Pinedo. Pinedo’s age is itself in question, though it is respectably old, but even if it were Early Acheulean, the comparison would still not be conclusive. At Pinedo, a quarry-workshop site near Toledo, the biface series consists mostly of abandoned roughouts, not finished pieces, many of them on obviously flawed raw material. Naturally they look crude. An earlier (1987) study by Carbonell et al. also suggests that the Torralba series, though it may overlap in age with Aridos, is later than Pinedo, and possibly later than Aridos as well. They provide no new evidence for their assessment.

Wear Traces on Stone

Dr. Nicholas Toth of Indiana University examined the Ambrona artifact collections for traces of wear-polish (in litt.). He found that none of the tools from atop and in the “pebble-pavement” in the earlier part of the Lower Unit was suited to study: all had a “frosted” surface lustre that obliterated any use-polish.

Artifacts in clayey and sandy deposits of the Upper Unit (Va and Vb), including the fan sediments, were relatively fresh and 37 pieces were chosen as suitable for analysis. Of the larger flakes and retouched pieces, most had use-wear polishes, and where striations were present they were normally parallel to working edges, suggesting slicing. All wear patterns found are consistent with hide, meat, and (rarely) bone being the material operated on. In only one case was there wear indicative of “heavy” hide working, and no plant polish was observed. Toth concludes that the presence of little “unused” waste suggests minimal on-site flaking, and since micro-wear patterns are consistent, indicating animal butchery, while other patterns are lacking, the site seems to have been specialized, rather than a base camp or some other general station.

Conjoinable Lithics

The study of conjoinable stone artifacts is an informative addition to the analytical battery of the prehistorian; despite a widespread misapprehension, it was not ignored in our work at Torralba and Ambrona. Dr. Nicholas Toth has had the Ambrona study under way for some time, but my knowledge of his results is too sketchy to include. I do know that there were conjoinable pieces in the 1962–1963 collections from that site: my notes indicate that feature 50D, IV, 7a, and 7b (two fragments of a quartzite chunk) can be rejoined and refit to 50F, IV, 13, and that 50F, IV, 1 is also attributable to this chunk (but will not join); another pair of refittable pieces is 48E, IV, 6 and 48F, IV, 6. I presume that Toth may have identified other cases.

For Torralba, my information is relatively complete, since we had the lithics in Chicago for study (and replication) for an extended period. The series includes a relatively small number of conjoinable stone artifacts. Of the 626 classifiable artifacts (excluding congelifracts), 29 are conjoinable fragments. We were quite aware of the potential information to be gained from such pieces, and most of them were detected during the course of excavations. The field identifications were all verified in Chicago. There were only two cases (totaling 6 flakes), where the conjoinability of pieces was first recognized in Chicago. It is possible of course that the collections still contain one or more conjoinable pieces that I missed, but I would not expect their number to be large. Nor would I expect there to be many such pieces in the smaller 1961 collection that I have not examined as closely. The following list does not include the several cases discovered of artifacts that are probably attributable to the same core or chunk of raw material, but could not actually be physically conjoined.

The 29 conjoinable artifacts found in 1962–1963 are from 12 occurrences in 7 levels at Torralba. Their provenience and separation are shown in Table 6.1.

The data in Table 6.1 are remarkable. Virtually all the conjoinable materials identified in the Torralba collections are pieces that were found with very little lateral separation between them or none at all (the four pieces level-bagged from Occupation 2 were found very close together and placed in a matchbox, but the markers indicating find positions were accidentally disturbed before they could be mapped). The unusually small lateral distance between the pieces would seem to imply that neither during deposition nor afterwards were they affected by any appreciable lateral transport. The separations noted are in fact small in comparison with average distances separating conjoinable finds in other situations where there is no possible question of fluviatile transport, where distributions are universally agreed to be “human-made,” and have always been interpreted as such. That would seem to be a datum to bear in mind in evaluating the possibility that long-distance water transport and rolling have altered bone surfaces or materially affected the original distribution of recovered materials at Torralba.

TABLE 6.1. Conjoinable pieces from Torralba

In two cases only at Torralba, fragments of the same original piece were found separated by 4.2–4.5 meters. But those cases are unique. Each involves a pair of complete sidescrapers made on two refittable pieces of a single large flake. In both cases, the flake was broken before the final shaping of the individual sidescraper edges took place. With such data, human agency seems the most likely explanation for the separation of the find spots.

Faunal Samples and MNIs

The Torralba fauna—its makeup, condition, significance, and abundance—has been the subject of some debate, partly because of differences of opinion about identifications and individual estimates provided by the two principal faunal analysts, Emiliano Aguirre and Richard Klein. I believe that a significant part of the disagreement between them can be resolved at this time. A certain amount of disagreement will remain unexplained, particularly where a single feature seems to have been attributed to two different taxa. Even in that case, part of the difference is due to the assignment of a single feature number to two (rarely three or four) pieces found together in a level, in intimate contact.

Sometimes, curation procedures that are beyond the control of Howell or the excavators were the cause of later analytical problems. Materials once excavated were removed (after plaster jacketing, where necessary) for shipment to the Museo Nacional de Ciencias Naturales by workers under Aguirre’s direction. Some faunal materials—and this is particularly true for the shafts of ribs—were discarded by that team as “requiring excessive museum space for their limited scientific interest.” All such items were identified and thoroughly examined by Aguirre beforehand. While I have no reason to question his identifications, such pieces will of course not have been available to Klein for his later study. There is some reason to believe that among the bones so treated were some that bore possible marks of human modification.

After arrival at the museum, several of the bigger and more impressive bones—particularly elephant bones—were selected for display. Those pieces were repaired—sometimes separately found fragments of the same bone were rejoined—and their surfaces smoothed where necessary and coated with preservative. Pieces so treated often or usually lost their identifying labels in the process. And, the surface treatment they received obliterated what I had identified as cutmarks in some cases, or made it impossible for Klein to differentiate modern damage from ancient modification. While the number of pieces so affected is not large, most of the information that they might have provided is forever lost. A larger number of plaster-jacketed pieces—some tusks, skulls, mandibles, pelvis, and scapula fragments, as well as the bigger and more complete limb bones—were stored in their jackets and remain in them. Consequently, Klein was unable to examine and identify them, and any information they provide about human agency or carnivore action is for the time being inaccessible. A still more important problem has been that most of the pieces have been relocated and relabeled on several occasions during periodic museum reorganization, and an even larger number of (usually smaller) items has become detached from its labels, misplaced, confused with other materials, or outright lost in the process. Last, the 1960s collections are reported to have been partly dispersed due to overlap in function between museums on an intra- (should these remains be regarded as primarily paleontological with tools, or primarily archeological, with bones?) or interregional (do they belong in Madrid or in Soria?) scale.

In cases where finds can no longer be identified, we have no recourse except to accept Aguirre’s faunal identifications and his, Howell’s, and my observations recorded in our field and laboratory notes.

The discrepancy between Aguirre’s counts of taxa and Klein’s can be partly explained on this basis. Klein, after all, saw only 1,521 (71 percent) of the 2,141 bone fragments recovered, and among the bones he could not examine were a substantial part of the largest, most readily identifiable skeletal elements.

That by itself will probably not account for most of the discrepancy. For the 1962–1963 Torralba excavations, Aguirre calculates an estimated minimum of 116 individual animals (112 mammals) for all levels, of which 37 are elephants, 23 equids, 21 red deer, 15 aurochs, 7 Dama, 5 rhinoceros, 2 lions, 2 small carnivores, and 4 Aves. (Azzaroli in litt. identifies one of the cervid mandibles as Megaceros sp.) Klein, in contrast, estimates only 64 individual mammals: 15 horses, 14 elephants, 10 red deer, 10 aurochs, 8 Dama, 4 rhinos, 2 lions, and 1 lagomorph. Klein then has 48 fewer individuals (excluding the birds) than Aguirre. Another factor helps resolve most of this difference.

Klein’s MNI calculations were derived on the basis of counting repetitions of the best represented body part for each taxon in each “level”—surely accepted practice, and the most conservative, justifiable way to proceed. However, when Klein calculated MNIs, he combined remains from the sublevels or spreads discussed earlier with the major horizon with which they were associated: all sublevels of Level 1 were united in his Level 1, and so on. When levels are combined, the MNI count invariably drops, as Klein himself illustrates in his chapter in the forthcoming monograph: uniting all Torralba levels drops his total MNI by almost 50 percent—from 64 to 34! Combining sublevels as he did by itself eliminates from Klein’s level-by-level counts 42 animals that would be called different individuals were the subhorizons differentiated, reducing the overall discrepancy between Klein and Aguirre to 7 animals. Since Klein only saw 70 percent of the bones, a difference of this order of magnitude is scarcely cause for alarm. Some unexplainable differences still remain: Klein’s list, though shorter, has one more Dama than Aguirre’s, and a lagomorph (which may be one of the otherwise missing “small carnivores” in Aguirre’s list).

Klein originally characterized the mortality profiles for Torralba elephants as catastrophic (in litt.) but has later stated that the sample size was probably too small for reliable estimation, suggesting that “if the Torralba and Ambrona ‘Lower’ samples are combined, the case for attritional mortality is especially strong” (Klein 1987: 29). However, if combining remains from different sublevels is likely to be misleading, combining remains from different sites is much more perilous. In fact, when the remains of all bones (not just teeth) from the larger sample of ageable “individuals” obtained from the separated sublevels are examined, the Torralba mortality profiles once more become catastrophic rather than attritional. If that is a correct diagnosis, the observation made by Santonja and Villa (1990: 61), that “the mortality profiles . . . cannot be reconciled with Freeman’s and Howell’s view of the sites,” is wrong. (I believe that it is best to reserve judgment about the shape of the age distribution at Ambrona until the final level distinctions have been established, and the occupation contents correlated across the site.)

At various times Klein has suggested that even catastrophic profiles might be explained by nonhuman agency, suggesting the drying of water holes or flash-flooding as likely alternatives. However, there is not the least geological or paleoenvironmental evidence for either phenomenon at either site. In the prehistoric environmental settings as they are now understood, truly catastrophic age profiles would almost certainly imply human agency.

Birds

Bird remains from Torralba and Ambrona have been identified by Antonio Sanchez and E. Aguirre (Sanchez and Aguirre in litt.). At Torralba, the four specimens recovered are all water birds: a “wishbone” of Tadorna ferruginea, the ruddy shelduck; a scapula of Mergus serrator, the red-breasted merganser; a humerus of Porphyrio porphyrio, the purple swamphen; and a coracoid from an unidentified anatid. There is no reason to believe that these creatures were captured by humans—such small, light remains may have been dropped nearby by kites or other predators and washed into the site deposits, and none is cut or otherwise altered. At the right season, all could have been found nesting in the reedy edges of lakes or slow-moving streams at Torralba—the merganser would normally be found near more northerly seacoasts, far from Torralba, at other seasons, and the swamphen, a partial migrant, though occasionally reported as far from its southerly range as Norway, would not ordinarily be found in as cold conditions as those at either Torralba or Ambrona during the winter season (Vaurie 1965: 138–39, 357–58).

Twelve bird bones were recovered from Ambrona; the provenience label is missing from one of them. In addition to the swamphen and the merganser represented at Torralba, the provenienced items are bones of Anser anser (graylag), Anas acuta (pintail), Fulica cf. atra (coot), and Vanellus vanellus (lapwing). All but the lapwing are waterfowl, and it too inhabits the banks of ponds and shores as frequently as moist meadows. The coot prefers large, open bodies of water. Like the merganser, the pintail is tolerant of brackish water (Vaurie 1965: 116–17, 359–360, 389). Again, there is no evidence that these bones are related to any human activity at the site.

The avifauna tells us something about local environments, but the species list is chronologically uninformative. It is interesting that most bird remains were detected in the course of excavation, even at Ambrona; few specimens were recovered by washing.

Small Fauna

The Torralba deposits did not yield much in the way of small fauna, aside from the often intact remains of tiny freshwater snails, some specifically pond dwellers, dominated by Hydrobia sp., denizens of streams, ponds, marshes, and backwaters. It is notable that this genus is a recent invader of fresh water and is salt tolerant. They and the birds confirm the presence of bodies of water near the site but are otherwise climatically uninformative. My notes also indicate a 1960s identification of a pelobatid (spadefoot) toad from the site, but it is unclear and the material is not mentioned in later references.

At Ambrona, where the 1980s sediments were washed, samples of small animals were recovered in some abundance. They were identified by Drs. C. Sese, B. Sanchiz, and I. Doadrio in Madrid. Sese recognized the insectivore Crocidura sp., the rodents Arvicola aff. sapidus, Microtus brecciensis, and Apodemus aff. sylvaticus, as well as the leporid Oryctolagus (Sese in litt.). (Lepus was said by Aguirre to be represented in the 1960s material.) Arvicola, the water vole, is a strong swimmer that prefers to live in cool, humid ground near bodies of water—I would be surprised if A. sapidus can be differentiated from the more northern form A. amphibius from the material recovered. This surprisingly impoverished fauna suggests a post-Biharian age for the site but is not otherwise very informative.

Sanchiz identified anurids including Discoglossus pictus, Pelobates cultripes, Pelodytes punctatus, Bufo bufo, Bufo calamita, a Hyla (H. arborea or H. meridionalis), and Rana perezi, as well as the water snake cf. Natrix (Sanchiz in litt.). Discoglossus is usually found in bodies of water or their damp grassy banks. Pelobates, the spadefoot “toad,” lives in dry, sandy ground close to bodies of water (Salvador 1974), excavating galleries in which it can survive long dry or cold periods.

Fish remains were found in considerable numbers, but all may probably represent a single species: Rutilus arcasii—its first documented fossil occurrence; less precisely identifiable remains were all attributable to Rutilus/Chondrostoma sp. or to indeterminate cyprinids (Doadrio in litt.), all of which may very well be from the same species. That in itself is interesting since R. arcasii has been found as the exclusive fish colonizing some interior drainage lakes in Spain (Doadrio in litt.). The species prefers to live in and near the reedy shallows of sluggish or tranquil waters and is absent from turbulent streams or very cold water. The waters of lakes deep enough not to freeze solid may be warm enough for them to survive year-round even in cold climates.

For the number of remains that were recovered by washing, the poverty of small mammal, reptile, and fish taxa is noteworthy. These creatures were all most probably resident at the site during its formation. The species found coincide in showing that the site environment was characterized by lakes, ponds, and marshy ground. As far as refinements of dating are concerned, they are unfortunately banal. There is no indication that any of them were used by people at Ambrona.

Carnivores as Agents of Bone Accumulation

Binford and others have suggested that the accumulations of animal remains at Torralba and Ambrona may be due to natural causes having nothing at all to do with the human presence seemingly attested by the stone tools. The excavators (and later, Shipman) detected traces of animal gnawing on a few bones. Discussions by Klein have reinforced the impression that carnivore remains or coprolites are quite abundant at the sites. Klein characterizes coprolites as “numerous—although artifacts are more numerous than coprolites” (1987: 18), thus giving the unfortunate impression that there must be many hundreds of large carnivore coprolites at Torralba and Ambrona, when in fact that has never been demonstrated. These observations have suggested to some that animals may be the major agents involved in the bone accumulations.

To the contrary, carnivore remains—bones as well as coprolites—while present, are rare at both sites. Even where present, specimens that are apparently coprolites must be further analyzed before their meaning is clear. Most of the fragments considered to be coprolites are not well-formed scats, but fragments of clayey sediment containing small bits of bone. At some mid-Pleistocene sites near Madrid, I have seen small clumps of clay filled with crushed or whole remains of the bones of small mammals and reptiles that are probably fossil pellets of raptorial birds. In the case of true coprolites, only detailed analysis of their contents can determine which carnivore is responsible: even some amount of decayed “bone-meal” (which may be present in scat of foxes and smaller carnivores) is no guarantee that hyenas are responsible. Furthermore, the feces of several small carnivores contains bone fragments. Klein has certainly identified coprolites at Ambrona. I have seen some of them myself but I don’t think that analysis of the specimens has been thorough enough to show that all the bits of bone-rich clayey sediment from the site were produced by large carnivores, or in particular, hyenas.

Bones of carnivores large enough to have killed the animals represented at either site or to have gotten their jaws around the bones of the larger ungulates to gnaw them are very few indeed, and marks of gnawing at Torralba have been said to be as rare as cutmarks apparently due to human modification. There are just two lion bones at Torralba: one in Occupation 1c and one in Occupation 4 (Klein lists the latter in Level 3). No wolves, no bears, no hyenas—in fact, no other large carnivores at all—are represented at that site. There are, of course, possibly two small (mustelid-sized) carnivores in Aguirre’s list, one from Level 4b and one from Level 10 but even if both are carnivores, they are certainly not the bone accumulators at Torralba. Even the Torralba lion, a respectably large cat, could not have dealt with a healthy adult elephant the size of those at Torralba—with shoulder heights verging on 11 to 12 feet—though lions could certainly have killed some of the other animals, and they might very well have—probably did—scavenge from carcasses of animals dead from other causes. How any of the carnivores represented could have managed to remove the appendicular bones of the large elephants, as Klein (1987: 25) suggests to explain their rareness compared to the abundance of innominates, is quite unclear; the imbalance must be due to some other agency, and among the alternative possibilities human activity seems the strongest.

At Ambrona, in the Lower Unit, both hyena and lynx are represented by but a single individual each, while indications of carnivore activity are not abundant at Ambrona, and Klein and Cruz-Uribe identified just three bones as bearing marks of carnivore chewing (Klein 1987). Such figures as these do attest a carnivore presence but are scarcely convincing evidence of a major carnivore role in the accumulation or alteration of the mammal remains from either site.

One might object that marks of carnivore activity could have been obliterated by natural alterations of the bone surfaces during or after their deposition. But if that is the case, as many marks of human alteration could have been obliterated at the same time. Arguments that postulate that a mechanism that is inherently non-selective is responsible for selective destruction of particular kinds of data are inherently fallacious.

Implications of the Surface Condition of Bones

Emiliano Aguirre, in his original study of the faunal remains from Torralba (in litt.), said:

The preservation of the vertebrate remains at Torralba varies from good, even sometimes excellent, to specimens having been altered in various ways, some prior to the process of fossilization and others, clearly subsequent to that process. In respect to the latter situations it is worth noting that there is relatively little breakage attributable to processes—such as mechanical deformation due to tectonic events or other such causes—within the sedimentary body itself. . . . On the other hand, in not a few instances, there are clear evidences of modification to faunal elements as a consequence of post-depositional chemical or biological processes, which hamper the identification of features of interest on a number of pieces.

Superficial decay or degradation of the bone and dendritic patterns produced by invertebrates and roots occur with some frequency, indicating interruptions in the process of sedimentation, deflation, and even periods of atmospheric exposure. He noted that exceptionally, bones were seen to exhibit a uniform polish all over, or all over one flat surface, but observed that “relatively few bones exhibit erosive traces over the entire surface, such as might result from water washing over a fossiliferous horizon, and leading to smoothing of protruding body parts through transport and rolling, or more rarely, aeolian processes” (Aguirre in litt.). He goes on to say “the great majority of modifications of bony elements fall into regular patterns,” particularly patterns of breakage, incision, and percussion, “that can be attributed to cultural activities.” Aguirre thus suggested that the bone was in good enough condition, despite surface alteration, so that traces of deliberate cultural modification could still be recognized on some—perhaps many—bones, and in this I concur. From the outset, Aguirre and all other analysts have recognized that surface abrasion exists on a number of specimens from Torralba (and Ambrona). However, Aguirre’s assessment of the general state of the bones is much more positive than the later diagnosis by Richard Klein.

Klein (1987: 19–21) states that at both Torralba and Ambrona

intense post-depositional leaching . . . has corroded bone surfaces. . . . The alterations introduced by leaching and corrosion are compounded by the massive fragmentation that occurred during and after burial at both Torralba and Ambrona. . . . It is notable that one-third of the 1779 bones at Torralba and one-sixth of the 4326 bones from Ambrona “Lower” exhibit edge-rounding that Butzer (pers. comm.) suggests occurred during limited fluvial transport on seasonally activated valley-margins or during net transport of sandy alluvium that partially buried the bones. Many bones that are not conspicuously rounded show a distinctive polish or luster and probably would exhibit abrasion or edge rounding under magnification. . . . Using a hand-held glass on a sample of lustrous Torralba and Ambrona bones, Butzer (pers. comm.) found parallel microstriations from abrasion by sand-sized particles on every one. (1987: 20)

He notes that Shipman and Rose also found “rounding” (under greater magnification) on nearly every specimen they examined from the two sites and goes on to say that “excepting abrasion and corrosion, Cruz-Uribe and I found little other damage on the Torralba and Ambrona bones” (1987: 21). The total number of carnivore-chewed pieces they detected was 14 from Torralba and 3 from Ambrona Lower, while the number of possible stone tool cutmarks was 22 at Torralba and none from Ambrona Lower. (Klein recognizes, of course, that surface corrosion may have obliterated other traces of both kinds.)

Butzer’s observations on this subject are recorded in an appendix to his final faunal chapter in the Torralba monograph. It is worth quoting in extenso. He reports:

The conspicuous concentration of archeological materials in such coarser-grained, intermediate energy horizons cautions strongly against diagnosing these as intact, primary associations. Instead, it is probable not only that there has been a measure of pre-depositional dispersal, but that at least some of the archeological micro-horizons are telescoped lag levels. This is strongly supported by my 1981 examination of the Torralba bone in the Madrid museum. Every specimen selected at random under low-power magnification showed systematic, very fine, longitudinal and parallel striations and had a “sandpapered” feel. This systematic striation was noted on all sides of each bone and was strongest on the most-exposed ends. It can only be explained by sand transport below, above and around the bone, resulting from energy conditions adequate to transport sand but mainly inefficient to move large bone; repeated burial and exposure is therefore probable. This is not incompatible with my conclusion that the archeological occurrences may retain their basic associations, i.e. between bone and bone, or bone and artifact, despite some horizontal displacement and changes in orientation. But the problem of telescoping bone and artifacts into “pseudo-floor” lags is more serious than I had anticipated. Trampling and sinking of heavy objects in wet clayey sediments is less problematical than at Ambrona, although it bedevils interpretation of those archeological materials at Torralba that are found in or at the base of clayey deposits. In effect, like all other Paleolithic open-air sites that I have examined since 1961–1963, the best associations at Torralba are in sediment taphonomic terms, semi-primary. (See Butzer 1982: 120–22)

Some more or less significant differences in these three observations call for comment. Sometimes they are quite subtle, but the differences have such important consequences for interpretation that it is essential to be quite careful about language. Aguirre’s description makes the Torralba fauna sound relatively intact, and relatively informative about cultural behavior. Klein in contrast talks of the “intense pre- and post-depositional destruction that affected the Ambrona bones” (1987: 27) (a description that can only be fairly applied to the Ambrona Upper series, where intense leaching has removed most bone).

Butzer’s description does not make it clear whether his sample was chosen from all bones or all visibly polished bones, as Klein suggests, but that is of less consequence than the conclusions he derives. His term “semi-primary” implies limited dispersal of cultural materials prior to burial, after which the buried deposits are subject to some disturbance (Butzer 1982: 121). In the depositional unit at Torralba bearing most of the archeological materials, though its sediments deposited under cold conditions in valley-bottom deposits, there is little cryoturbation and transport distances must have been quite short, Surface abrasion of bone could be ascribed to sediments passing around the bones, rather than to lateral movement of the bones in the sediments. The lack of preferential orientations or size-sorting would seem to support this possibility. Archeological associations, as Butzer points out, could survive this degree of disturbance and still be recognizable. Only his conclusion that the depositional environment is one in which different archeological levels might have been telescoped into “pseudo-floor lags” poses any substantial theoretical problem to cultural interpretation.

Butzer’s conclusions are borne out by the archeological field observations. The local merging of elsewhere discrete levels shows that even the thinnest, apparently most pristine level might contain materials originally deposited in several separate episodes. But lag deposits have a geo-archeological signature. Ordinarily lag deposits built up over any length of time may be expected to be heterogeneous in content, and different lag deposits should differ in random ways. That is because, as a rule, the depositional conditions were different for each of the discrete “moments” that later telescope to form a single apparent “floor.” Ordinarily, the materials in one lag deposit don’t differ from those in another in patterned ways, unless the landscape and the conditions of deposition have remained so constant that the local depositional environment has repeatedly caused accumulations of materials of the same size and shape to be dropped in essentially “the same spot.” Only then should telescoping of formerly disparate levels produce a horizon (or horizons) whose contents are both internally homogeneous in their characteristics and different from others in regularly repeated and predictable ways. Such cases are by no means geologically exceptional; nevertheless, careful examination should reveal the essentially “geological” nature of the accumulation (due to similar behavior of items whose sizes or shapes are analogous when waterborne or moved by gravity, etc.). What is more, in the archeological case, the original cultural behavior that produced the residues forming the lag would of course have had to be essentially similar during each episode of accumulation, implying the repeated performance of the same set of activities in the same part of the changing prehistoric landscape (whether this is the actual area excavated or other areas which served as sources to the lag). The evidence of the accumulations called “occupations” at Torralba and Ambrona runs contrary to such an interpretation.

Another problematic situation he mentions is that of the “sinking of heavy objects in wet, clayey sediments.” At Ambrona, there are some situations in clayey sediments in which skeletal remains of several animals were found lying one above another in layer-cake fashion, and in the absence of other evidence, it would be a mistake to interpret these as single cultural accumulations. At Torralba, this is less a potential problem than at Ambrona, since the major accumulation at the base of clays (the clay facies in the north half of Occupation 7) consists principally of the bones from one side of one individual animal, in a somewhat rearranged “near-anatomical” position. Since that individual died but once, the question of whether or not it sank, and at what rate, is immaterial. The large stones in the same horizon that are interpreted as part of this accumulation were pretty evidently positioned in relation to the bones: again, sinking provides no objection to previous interpretation. I see no reason to believe that all else is a culturally meaningful association, while the stone tools in intimate juxtaposition to the bones are extraneous.

The concentration of accumulations within or at the base of clayey deposits certainly does impose peculiar restraints on interpretation—in some cases it may even rule out explanations in purely geological terms. It is hard to account for differences in the distributions of materials deposited in still-water or marshy sediments, particularly the sorting of large, dense items such as elephant bone, in terms of geological agency. If the accumulations are found at some distance from the edge of a prehistoric lake or bog, and there are no nearby channels in which flow would have sorted them as they were swept along, the discovery of sorting by body part or bone size or shape may well have cultural rather than simply geological significance.

Figure 6.2 shows an example of this sort from Ambrona. In 1980, we found a group of five elephant tusks of different sizes, lying in close proximity in clays (not an isolated example—other tusks were found grouped together not far away in the same deposits). One of the tusks was near vertical in the clayey sediment. There are no faults or other disturbances in the deposits that could account for its attitude. It must have been buried that way, fast enough so that it was not weathered to pieces. Its position may very likely be due to a heavier tusk having sunk more rapidly, trapping the point of the smaller one and pulling it down into the angle it maintained at discovery. While the attitude of this single find may be purely a result of depositional processes, I do not see how any natural agent other than human activity can explain the spatial segregation of the tusks from other bones in these deposits.

The sediments are still-water beds, not stream deposits, and assortment by channeled flow is out of the question. No geological force as far from a contemporary channel would have separated these five tusks from other relatively same-shaped body parts and dumped them all together.

Non-human biotic agencies are also improbable agents. The tusks are uninteresting to carnivores, who in any case would scarcely have dragged them all into a separate pile in muck or standing water. As Villa notes, elephants today often pick up and carry about bones of their dead congeners, and anyone who has seen filmed behavior of this sort must admit that it is remarkable. However, they do not sort the bones and dispose of them in piles segregated by body part. Rather, they seem to carry or drag the bones about for a bit, then toss them away apparently at random. Peter Beard (1977) has published scores of photographs of dead elephants, including some astonishing natural accumulations of bone, but in the few cases he shows where bones are segregated by body part (or arranged into tidy localized piles) the hands of humans were responsible.

Butzer’s concern about the problematic effects of sinking in clayey deposits is doubtless well placed. On the other hand, such sediments may, in special cases such as the ones just described, constrain geological interpretation in directions that pave the way toward an understanding of cultural phenomena.

Marks of Human Alteration on Bone

On many of the bones from Torralba and Ambrona, there are marks that I do not believe could have been made by any non-hominid agency. The marks are gross enough in most cases so that surface alteration has not obliterated them or rendered them unrecognizable. Of course, those marks will never pass muster as evidence for hominid alteration if one insists that the bone surface topography must be essentially fresh for the markings to be studied at all. That requirement has been both one of the strengths and, at the same time, one of the weaknesses of a recent study of some Torralba and Ambrona bones undertaken by Shipman and Rose (1983).

FIGURE 6.2. Group of elephant tusks—one vertical—in Ambrona Lower Unit (1980)

They subjected replicas of surfaces of some of the smaller Torralba bone fragments to reexamination for microscopic evidence of cutting and gnawing, using the scanning electron microscope. Shipman’s criteria for identifying cutmarks are very exacting, and her procedure rigorous. Therefore, there is very little doubt that a bone she identifies as cut actually bears marks that most would find convincing evidence of such alteration. She found such marks on some of the Torralba bone, and I suppose that I should be pleased that there is some support for my belief that hominids altered some of the Torralba bones. While it may seem contrary of me, I have several reservations about the Shipman and Rose study.

Most important, I believe that their description of procedures as published is unreliable, and the estimate of the proportion of sliced bones in the collection they offer is therefore unusable. I don’t mean that the marks identified do not exist or were wrongly counted, but that other statements about the study and the size of samples examined incorporate serious errors.

For one thing, Shipman claims to have examined all the Torralba bones. However, there is no way she could have examined the whole collection, since the plaster-jacketed bones Klein was unable to see are still in the same jackets. (Many of the marks I find most convincing are found on those larger bones—major parts of elephant pelvis, whole tusks, mandibles, or elephant and bovid crania—under the plaster jackets.) Second, she claims to have found convincing marks of hominid alteration on a total of 12—some 1.2 percent—of the replicas examined from Torralba. The figure cannot possibly be correct.

After Klein’s reclassification of the Torralba bone, the museum collections were reorganized, replacing the finds in shelved lots by square, level, and feature number, rather than by species and body part. (The only exception is a lot of 22 bones Klein suspected might be cut and had shelved separately for future study.) Any thorough examination of all bone finds would first require opening every box, locating the label (square and feature number) on each piece, and then identifying it from Klein’s inventory of taxa and body parts, a process that by itself would necessitate several days’ work. Then the surface of each bone would have had to be completely examined under proper lighting, even on occasion under low magnification. Next, suspect bone surfaces would need replication, in itself a time-consuming process. To examine this bulk of material carefully and replicate the specimens that seemed altered would require a minimum of several weeks’ time. This estimate may be approximately doubled because of the shortness of the museum’s hours—ordinarily only 4 to 5 hours of access to its warehoused collections are permitted each day.

Shipman spent in all several hours with the collections, not several weeks. In such limited time it is not possible that she could have had time to examine, let alone replicate, more than a few bones from the Torralba collection. For 12 to be 1.2 percent of the replicas made, Shipman would have had to make a thousand of them. In the short time available, this is an unrealizably high number, even if several replicas were made of any single bone.

It seems to me most probable that, given the time restrictions of her study, Shipman must in fact have spent almost all her time on the two dozen bones Klein had set aside, looking at others only as (or if) time permitted. Perhaps this is actually what Shipman and Rose intended to say. Whatever the explanation, the account they give of sample size and procedures is inconsistent with the nature and size of the collections.

If what Shipman and Rose really examined was just the collection Klein thought might be worked, their sample was doubly constrained by any preconceptions he may have had at the time about the nature of bone working, and by his ability under the less-than-ideal conditions in the museum to distinguish marks on bone. Results of the Shipman and Rose study would then be unintentionally biased, no matter what their remaining procedures.

The Shipman-Rose study provides some information of qualified interest. They did find 12 convincing marks of human alteration on 4 bones of Paleoloxodon, 3 of Equus, and 1 of Cervus, as well as on 1 bone of indeterminate species. In a clear misunderstanding of the evidence, Santonja and Villa state that “the rarity of cutmarks on the bones . . . cannot be reconciled with Freeman’s and Howell’s view of the sites as places where herds of elephants were killed and butchered” (1990: 61). But Shipman and Rose actually said that their study offers “only limited” support, not “no support” for the fact that Torralba (and Ambrona) were butchery sites. There is a real difference. And, there is no reason to suppose that butchering marks need to be abundant even in a culturally modified faunal collection. Visible cutmarks may be very rare—even virtually absent—in more recent butchery sites, such as some “buffalo jumps” in the United States, where humans are known to be the principal or only agent of bone accumulation and/or alteration. As mentioned previously, Shipman and Rose also detected carnivore tooth scratches in comparable frequency. They were characterized as less abundant than might be expected of assemblages where carnivores were the primary agents of bone alteration. Shipman and Rose’s conclusions do not correspond to Santonja and Villa’s summary.

In sum, despite its problematic nature, the work of Shipman and Rose is nonetheless interesting insofar as it provides some direct evidence of apparent human intervention in the alteration of the Torralba bones. However, theirs is far from the “last word” on the subject. There are other kinds of apparently cultural marks on bones from these two sites that were not considered in their study. By far the most abundant marks that were earlier interpreted as signs of hominid alteration are grosser traces than the fine slicing studied by Shipman. They consist of large-scale scars of gross damage—marks of battering; chopping with a large, sharp, wedge-shaped edge; scraping, or abrasion with a smooth, blunter stone edge; and deep slicing, gouging, or grooving. Though they occur on bones whose surfaces have also been altered by natural postdepositional phenomena, they have resisted obliteration. Quite comparable coarse marks of hacking are identified as butchering traces at the Casper site (Frison 1974: 36–37) and elsewhere and have been interpreted as important evidence about butchering techniques at those sites. Shipman’s methods simply ignore all such evidence, which to me seems as obvious and as convincingly indicative of human handiwork as the pristine fine slicemarks she studies. The coarser topography of such marks was the basis for my own field counts of worked bones, in the majority of cases.

In collections from sites where bone surfaces have undergone more than minimal postdepositional alteration, as at Torralba, macroscopic butchery marks may be the only ones that can survive. Most bone under such conditions cannot preserve the diagnostic microtopographic features, the fresh traces of fine slicing, that Shipman’s microscopic study relies on. Marks of gross damage certainly merit further investigation, instead of summary dismissal.

I examined the bones from Torralba while they were being excavated, while the surfaces were “fresh,” unvarnished, and still unjacketed. It was then still easy to tell fresh excavation damage from ancient alteration. Workers alerted me as they recognized apparent human modification, so I watched many of the surfaces as they were cleaned and excavated not a few myself. In the field, I identified four types of modifications that seemed to be cultural: slicing, hacking with a wedge-shaped edge, scraping or abrasion, and battering or repeated percussion. My notes show 56 sliced surfaces, 6 cases of hackmarks, 1 abraded bone, and 4 battered specimens. There were in addition a number of charred bones, 2 so heavily burnt that I thought it unlikely that grassfire could be responsible. There were also some large bones that had apparently been deliberately flaked while “green” in such ways that carnivore gnawing as responsible agency was out of the question. Those were not counted; we relied on Aguirre to study them (which he did, in a chapter in the forthcoming monograph). Klein saw the collections only after they were jacketed/warehoused/preserved, when it was much harder or impossible to differentiate fresh damage from ancient modification, and so he quite properly excluded several by then “dubious” cases from his accounts. Nonetheless, he recorded 22 bones as potentially cut, 4 charred (possibly naturally), and 10 from which flakes had apparently been struck in the “green” state. In fact, the disagreement between Klein’s figures and mine is really not serious, considering what had happened to the collections between the excavation and the time he saw them.

In the 1980 excavation in the Ambrona Lower Unit, I found that about 50 percent of the larger bones bore marks suggestive of cultural alteration. A selection of pieces from both sites is illustrated. Figure 6.3 shows an immense elephant left innominate with subparallel grooves attesting extensive scraping. Figure 6.4 shows hacking and slicing on the premaxilla of an elephant skull. In Figure 6.5, an elephant mandible whose ascending ramus was removed, by repeated chopping with a wedge-shaped edge, is shown. Details of the hacking are illustrated in Figure 6.6. The remainder of the ramus was found just behind the mandible (it can be seen in the first photograph), and bore matching scars (Fig. 6.7). Despite the evident surface corrosion on these pieces, the marks are still easily identifiable, and in no case do they seem explicable by carnivore activity. None of these pieces would have been replicated by Shipman: their surfaces are too corroded and the marks they bear are not the sort she studies. Three apparently sliced specimens from Torralba are shown in Figures 6.8 to 6.10 and a hacked bone from the same site in Figure 6.11. Only space limits keep me from illustrating a score of other altered bones, including skulls, scapulae, ribs, innominates, and longbones.

FIGURE 6.3. Large elephant innominate, showing marks of scraping (Amb 80)

Spatial Associations

The discovery of items in close juxtaposition in an archeological level has traditionally been seen as evidence that there is a real relationship between them. While some apparent spatial associations that are detected by eye are misleading, at least in the absence of statistical demonstration that the associations are unlikely to have arisen by chance, other visual associations are quite valid. No one would believe that an association of the bones of the skeleton of a single individual (such as the focal association in the part of Occupation 7 at Torralba) needs statistical validation. Nor is that an isolated instance.

The separate accumulations of elephant tusks in the Ambrona Lower Unit are statistically significant associations. So are the repeated concentrations of bovine horncores and bifaces in squares G15, G12, H12, and I3 in Torralba’s Occupation 3 (Fig. 6.12).

Still other associations seem so unlikely that even though their probability cannot be directly determined because each is almost unique, their nature and number still persuasively suggest a direct relationship between the animal bones and the implements found at these two sites. In Torralba Occupation 1 we found one particularly striking case: square J12 held an elephant pelvis with a convergent denticulate tucked inside the acetabulum; a limestone battered polyhedron lay just outside the socket (Fig. 6.13). In I9 in the same level, a flint utilized flake lay atop an elephant right pyramidal. In Level 7 a small flint biface lay next to an elephant radio-ulna in square M12. At Ambrona, in the Lower Unit we repeatedly found bifaces right beside tusks (Square G99: Fig. 6.14) or elephant vertebrae (STE 4: 2 vertebrae with two handaxes; Fig. 6.15). Other cases are too numerous to mention. The sheer number of such finds and their coherence with results of the statistical study of frequency relationships (see later discussion) cannot fail to impress a reasonable analyst.

FIGURE 6.4. Elephant skull, hacked on premaxilla (Amb 80)

FIGURE 6.5. Elephant mandible with marks of hacking (to remove ramus?) (Amb 80)

FIGURE 6.6. Details of hacking, visible despite corrosion, on mandible (Amb 80)

Statistical Analyses

If this were not enough, there is more abundant—and, to my mind, more convincing—evidence that there is a meaningful, culturally mediated relationship between the remains of large animals found at these sites and the artifacts left there by humans. That is the evidence provided by multivariate statistical analyses of relationships between these different kinds of data, analyses that have been carried furthest (and criticized most) at Torralba.

There is only one problem with the results of statistical testing. Most people still really don’t understand the tests or their results, and so they will either reject the whole process as less meaningful than the solid, tangible “real” data an excavator digs up, or—even worse—will uncritically accept any and all statistical manipulations as valid, only to reject each in turn in favor of the latest test claimed to have produced contradictory results. It is an unfortunate problem, but one that eventually must vanish with education. In fact, used properly and evaluated critically, statistical tests are just so many more among the tools—knives, brushes, and so on—that excavators use to gather data, and their results are just as real and meaningful as any finds made with those tools. Just as one can pick the wrong tool for excavation, using shovels where trowels are called for, so one can use an inappropriate statistical procedure. Not all statistical procedures are equally justifiable. Just as one can excavate badly, producing erroneous information, one can also use statistics inappropriately to produce wrong or misleading information. Not all statistical results are equally reliable.

The use of any statistical test requires that the data to be analyzed be error free, that if the data must be transformed it be done in a justifiable and appropriate way that will neither invalidate the calculations nor hinder interpretation, and that the measures chosen be suited to the kind of data being studied. The tests chosen here produce measures of bivariate relationship—a matrix of correlation coefficients, in this case—and then use those measures as a basis for further computation. Any measure of the strength of a relationship between two variables should remain the same whether or not a third variable is present; that is not the case for some measures, but it is for the coefficients used here. Some variables are unrepresented in some samples. The problem of zeros in the data was handled by treating them as missing values and deleting any pairwise comparison where a zero occurred. Sometimes statistical software packages perform a multivariate test in different ways, producing different solutions from the same data. Obviously, that is undesirable: it ought to be the case that any analyst, using the same data and the same tests, should get the same results. The tests we used are fully replicable.

Table 6.2 lists the more abundant artifact types and MNIs for the major species represented in the Torralba occupations. Since use of edge counts in a previous work (Freeman 1978) drew criticism and, more important, caused confusion, the artifact counts used here are of whole tools tabulated by level.

FIGURE 6.7. Ascending ramus of same mandible with matching hack marks

Whether edge counts or whole tool counts are used, significant patterned relationships appear in the data. The solutions are not identical. Multiple tools often combine different kinds of edges, but where the particular combination is not abundant enough to be considered a significant “new” type, they are placed into the type of the best-made edge. There are inevitably differences between solutions based on edge counts (which I still consider more meaningful) and those based on whole tool counts, but the differences are less important than the fact that significant patterning is detected no matter which data are used.

FIGURE 6.8. Torralba bone showing marks of slicing

The counts were ranked and used to calculate the rank-order correlation coefficient, Spearman’s rho. Rho works with ranks of frequencies, rather than the raw frequencies themselves, and ranking is far and away the most mathematically defensible transformation for these data, where it is inappropriate to make assumptions about the underlying shape of the data distributions. In the past, I have transformed frequencies to square roots and used the more common bivariate correlation coefficient, Pearson’s r (with larger data sets the transformation had essentially no effect on results); the results were slightly different from those presented here, but both tests coincide in showing relationship between similar sets of variables. There are no significant patterns of replacement or inverse relationship in the data.

FIGURE 6.9. Torralba bone showing apparent slicing

The rank-order correlation coefficients in Table 6.3 were used as measures of nearness in a cluster analysis (Fig. 6.16). The structure of variability was simple enough so as to be discernable in most of its details on visual inspection of the coefficient matrix, but the dendrogram, based on a single-linkage procedure, shows its characteristics more clearly. Two data categories are really unlike the rest: congelifracts and bovid MNIs. Since the number of cattle is nearly invariant, this is as one would expect. The fact that congelifracts are unlike other data categories is reassuring. The animal species remaining are in fact related to each other, and also to particular stone tool types. Notches and denticulates stand apart from the rest of the tools, but choppers are related in frequency to equids and cervids, while bifaces, end- and sidescrapers, and cores relate more closely to elephant counts. That is not to say that elephants, cervids, and equids are unrelated to each other—all, and the other tools, form an interrelated group at a more distant level.

This simple test indicates beyond any doubt that there are meaningful relationships between the abundance of particular stone tools and the abundance of particular animal species. That simply would not be the case if it were true as some allege that the human presence at Torralba was essentially unrelated to the presence of animals at the site. But the tests are not in every respect satisfactory explorations of the data. Bovid MNIs, as noted, are small and nearly invariant. There are fewer cervids and equids than one would ideally prefer, and there are a lot of tied ranks. In these respects, the correlation matrix and cluster procedure, though certainly conclusive, leave something to be desired.

FIGURE 6.10. Torralba bone showing apparent slicing and battering

FIGURE 6.11. Fragment of “chopped” elephant bone from Torralba

When body part counts are used rather than counts of MNIs both counts and variability in the faunal categories increase. The picture of relationships is both strengthened and clarified as details are added. At the same time, new dimensions of variability appear that are not adequately depicted in the essentially two-dimensional cluster analysis; fortunately, the related but more elegant principal components analysis can show these relationships quite well.

FIGURE 6.12. Association of bifaces and horncores, Occ. 3, Torralba

In the following test I have used the stone artifact frequencies from the previous table, dropped the MNIs, and added the body part counts shown in Table 6.4. Differences in frequency are not so great that raw frequencies and Pearson’s r could not have been used, and it might very well have been appropriate to do so, but since nothing is known of the nature of the underlying distribution of these data, to be safe the same nonparametric measure of correlation, Spearman’s rho, was chosen.

The matrix of bivariate rank-order correlations is given in Table 6.5. Binford (1987) claims his analyses show that the Torralba deposits show a palimpsest of two major patterns: one in which bovid, equid, and cervid remains were deposited with tools while elephant remains were deposited in unrelated fashion; the other in which elephant bones were deposited in association with stone tools, but in which bones were broken into unidentifiable bits by forces other than human agency (1987: 66). More detailed examination of frequency relationships including body parts leads him to identify one pattern as potentially due to hominids, only to reject that possibility in the following terms: “No matter how we interpret the patterning, the case for ‘activity areas’ is very hard to sustain. The elephant carcass material is inversely related to remains of other species, making it difficult to argue that the differences in tools represent tools appropriate to sequential processing steps in the butchering of a single animal” (Binford 1987: 90). In his detailed “analysis,” in fact, he claims to find in the Torralba data an inverse relationship between frequencies of a kind of pseudo “Mousterian of Acheulean Tradition” tool set, including especially bifaces, notches, and denticulates, on the one hand, and those of waste on the other (1987: 55); an association of scrapers and choppers—which he also wrongly calls “corescrapers or core axes” (1987: 77); elephants varying inversely with other animals (1987: 83, 85, 91); and an association between bifaces, sidescrapers, and elephant remains that he explains away as partly related to the paucity of those tool types (sidescrapers are, on the contrary, the next most abundant flake tool category—whether whole tools or edges are counted—in the collection and only five fewer in number than denticulates), and partly due to the fact that the sidescraper counts are elevated because they are “compound edged tools” (1987: 89). While it is true that counts of working edges were used instead of whole tool counts for most flake tool types in the study (Freeman 1978) that was the source of data reanalyzed by Binford, patterned relationships between sidescrapers, other tools, and bones appear just as clearly when whole tool counts are used, as the present study shows. One could go on to contest other “results” of Binford’s “analysis” in detail, but it is pointless. No matter what one feels about the logical coherence of his explanations of patterning (there, too, I find much that is questionable), the statistical results on which the arguments are based are worthless, since he used erroneous data, unjustified and unnecessarily convoluted data transformations, and inappropriate analytical procedures—no one whose hand was not guided by Binford could repeat his test and obtain the same results.

FIGURE 6.13. Elephant pelvis, with denticulate inside acetabulum; polyhedron nearby, Torralba

In fact, a simple inspection of the correlation matrices in this chapter is enough to show that Binford’s claims are wrong. Aside from the association of notches and denticulates, which is only part of a more heterogeneous group he defines, not a single one of his claimed relationships has any validity—a very unfortunate state of affairs, since his results have been uncritically accepted at face value by Villa (1990, 1991—though she interprets them differently) among others.

True, pairs of items that cluster in Binford’s solution sometimes also cluster in mine, but not in the ways or for the reasons he specifies, and there is no evidence of any substantial “inverse relationship” between variables. The numerous inverse relationships that so preoccupy Binford are in fact mathematical fictions. They occur neither in a correctly calculated matrix of correlation coefficients—product-moment or rank-order—nor are they at all numerous in an appropriate matrix of component loadings. Any real inverse relationship between variables has to be reflected in one increasing as the other decreases—producing at least a partial inversion of their numbers or rank orders—and that must result in a significant negative correlation. This simply is very unusual in the Torralba data: only one of the small number of negative coefficients (notches vs. elephant feet) reaches significance at the .05 level. It doesn’t even happen when the erroneously copied data Binford presents are analyzed correctly.

I can only explain the large number of negative loadings in Binford’s tables by assuming that either his “chi-square” transforms were inappropriately calculated from percentages (I suspect this may be the case, since Binford has been so fond of percentages in the past), or that he has presented an incomplete solution, which, had he allowed the test to continue to iterate until it reached a unique solution, would have eliminated the negative loadings. (There may be other mathematical explanations for his results, but no one could isolate them from Binford’s almost deliberately obtuse procedural description.) Whatever the case, the statistical procedure is—has to be—invalid, as one can determine just by inspection of his data tables.

Our table of rotated factor loadings (Table 6.6) shows that seven factors or components are adequate to account for over 92 percent of the variance in the matrix of correlation coefficients. As is my usual practice, I rotated one more component (as a possible “error component”) than the number with eigenvalues of 1.0 or greater. The last component does not principally determine variation in any variable; that is as one would hope. The seventh component only loads highly on geologically crushed pieces. That is also an encouraging sign.

Such tests as these are most justifiable when applied to data about whose structure there are some prior expectations. We had some idea beforehand what the statistical results at Torralba might show. Field observations of spatial associations suggested that cores and scrapers should each be related to elephant tusk, ribs, limbs, vertebrae, scapula, and pelvis (these two were combined in the statistical test); that bifaces and perforators should be related to elephant skull; that denticulates and notches were related (a pattern also incidentally found by Binford); and that both bifaces and denticulates were related to bovid skull (but bovid skull was too infrequent to be used alone in the test). Cervid metapodials and bovid and elephant foot bones were suspected to be related to waste and minimally utilized pieces, but waste flakes often occurred near skull fragments. (Note that expectations would be different for Ambrona, where other spatial associations were observed with greater frequency.) In my previously published analysis based on Pearson’s r, several of these associations were confirmed statistically.

FIGURE 6.14. Associated biface and elephant tusk, Ambrona

FIGURE 6.15. Association of vertebrae and handaxes, Ambrona

In this test, using a less powerful but more justifiable measure of association, and a slightly different set of data, with fewer collapsed categories, fewer correspondences occur, but the general picture remains the same.

The first and largest tendency for variation is associated very strongly with sidescrapers, elephant teeth, tusks, limbs, ribs, feet and vertebrae, scapula/pelvis, and equid teeth and feet, and less strongly with cores, elephant skull, equid skull, and bovid limbs. The second is still less strongly associated with variance in bifaces and endscrapers but strongly determines variation in cervid antler. The third is highly associated with cervid limbs, less strongly with cervid skull, and less still with equid limbs, while equid skull and endscrapers show moderate negative loadings on this factor—that suggests simply that we may be sampling different aspects of the “cultural landscape” in each level, and that the places of discovery of these latter items are different from those of the former (a fact that has no evident geological explanation). Perforators, “waste,” and elephant skull fragments are strongly associated with Factor 4. The fifth tendency for variation strongly determines variation in choppers and horse scapula/pelvis, “explaining” to a smaller degree variation in bovid limbs. Notches and denticulates are found alone to be determined by Factor 6.

TABLE 6.2. Torralba major data categories by level (lithics are whole pieces; taxa are Aguirre’s MNIs)

Using Pearson’s r (results not shown), the number of meaningful factors isolated was 7. The associations detected remained essentially similar, but one factor determined most variation in both waste and equid limbs, and elephant foot bones and cervid limbs were found related to another.

TABLE 6.3. Torralba artifacts and species matrix of Spearman Rank Correlation Coefficients (RHO) (zeros treated as missing data; pairs with any zero member eliminated)

The results of the principal components analysis indicate that there are in fact patterned relationships between stone artifact types and particular animal body parts, for all the major species represented at Torralba. They are nontrivial: a trivial association would be, for example, a single tendency that determined variation in all the variables, which would indicate that sample size was the only operative variable. They do not bear out Villa’s (1990: 304) claim that at Torralba people butchered “elephant carcass leftovers” rather than whole carcasses—all major elephant body parts are involved in these patterned relationships. The statistical tests demonstrate that the human presence and the animal remains really cannot be independent of each other. In some cases, at least, they correspond to spatial associations viewed during the course of excavation, and their contents could not have been acted on similarly by natural depositional agencies other than humans, so that no simple explanation of site formation processes that excludes human agency can adequately account for them.

FIGURE 6.16. Dendrogram of relationships between artifact types and animal species, Torralba; single lineage method; distance measure = Spearman’s rho

When the results of the level-by-level statistical analysis are evaluated in light of all other evidence from Torralba, the most economical way to account for the presence of the different components, the distinctive clusters of variables associated with each, and the fact that different clusters were frequently found in different areas is to ascribe them largely to the organization of human activities. That is not to say that all materials from Torralba reflect human behavior, for there are many kinds of data that were not included in the tests, and some that were not adequately explained in terms of the factors isolated. Nor is it to claim that there has been no natural disturbance of the original patterns in the residues. Despite these processes, however, a picture of human activity emerges among the other pictures reflected in the Torralba finds.

Binford’s dubious statistical procedures and errors and his mistranscriptions—perhaps better, “manipulations”—of artifact and bone counts from the site have misled readers about the nature and composition of the Torralba assemblages, and about the relationships between data categories. As Howell noted in a review in the Journal of Human Evolution (1989), 14.3 percent (10) of the 70 cells in the matrix Binford supposedly compiled from my earlier published figures are wrong: even had he used identical tests, he would therefore have obtained different results from mine. The situation is aggravated by questionable transformations of the data—over-interpretation of mathematical results that are not statistically significant to begin with, and the use of analytical procedures that I defy anyone (other than Binford) to understand or replicate. There are perfectly appropriate ways of transforming the data for his purposes—simply ranking the raw counts and using a rank-order correlation procedure as has been done here is the simplest and probably the best, while square root or log transformations of all the data and the use of Pearson’s r are probably also defensible in this case—and when error-free data, transformed appropriately, are used as input to ordinary principal components analysis and rotation (or related multivariate tests whose results are free of operator bias and equally insensitive to the order of data entry), the results obtained are the ones I have published here and elsewhere, not those Binford presents.

TABLE 6.4. Torralba body parts count by level

CONCLUSIONS

I hope that I have presented enough information regarding Howell’s work at Torralba and Ambrona in the 1960s and later to indicate the significance of those investigations once and for all, and to lay the less well-founded criticisms of our work to rest. I do not imply that our interpretations—specifically, my own—have always been impeccable and infallible. They have certainly not. In my earlier work, I seriously misjudged the extent of cultural elaboration expectable in a mid-Pleistocene site and underestimated the difficulties in unraveling what cultural information there is from the overlay of other processes—geological, mechanical, chemical, and biological—that may embed and hide it. Nor have I always expressed myself as well as I could have done. Excavations in 1963, conducted under my guidance, while good enough, could nonetheless have been better; I paid too little attention in the 1960s to marks of gnawing or to marks of butchery; it is probably my own fault that no one knows that screens were used at Torralba; I should undoubtedly have indicated more clearly that we knew that carnivores were present at the sites, or that we had taken geological processes such as slopewash and channel flow into consideration; I never stressed enough that our statistical tests only included some of the Torralba data, or that some of the tested variables were not adequately explained. Of course there are animal remains at Torralba and Ambrona that were not manipulated by humans, or even evident to them. That should also have been made clearer.

TABLE 6.5. Matrix of Spearman correlation coefficients—ranked data

TABLE 6.6. PC analysis with rotation Torralba artifacts and body parts

Questions about the causes of patterning in these residues are not simple black-and-white issues. It is irresponsible and nonscientific to decide that either all the patterning detected must result from human cultural agency or none of it can. There is patterning due to nonhuman agency at the two sites. At the same time, a substantial basis for cultural interpretation can still be recognized at both. The archeological record of hominid activity at Torralba and Ambrona is not pristine and free from pre- and postdepositional distortions. However, even if these obscure the message, they do not obliterate it entirely, and enough remains to tell at least part of a story of hominid-animal interactions at Torralba and Ambrona.

No single kind of evidence tells the whole story. The sediments and fauna pose questions that must be answered with conjoinable stone tools, wear-polishes, skeletal dispersal, and patterned regularities discerned statistically. No single line of evidence—sinking in clayey deposits, MNIs, age distributions, tooth-marks, stone tools—tells its own story unambiguously. To decipher what Binford has called the “palimpsest” of Torralba requires assembling and comparing all these multitudinous kinds of information and trying to reconcile each with the rest. But when that is done, the outline of a message about human adaptation appears, behind other messages, it is true, but nevertheless still legible.

I hope that these observations will help in some small way to clarify the importance and potential of Torralba and Ambrona and to secure for them the recognition they deserve. It is unfortunately true that we still know all too little about hominid adaptations of the mid-Pleistocene. That is so in spite of the number of new mid-Pleistocene sites that have been discovered and carefully excavated since the 1960s. Despite the high quality of excavations at sites like Aridos and Isernia, much more information will be needed before any satisfactory idea of the nature of mid-Pleistocene adaptations in any region can be derived. Each site we now know is like an irreplaceable piece of a huge and variable picture puzzle most of whose pieces are missing. Each site we know so far has proven to be unique in scale, in scope, and in quality of information; it would be absolutely senseless to discard or ignore any of the pieces we have so laboriously assembled, assuming that it is replicated by any other. The pattern on each and every piece is damaged or obscure—every other European site from this time range presents at least as many problems of interpretation as do Torralba and Ambrona. To progress, we must try to understand every piece in its own terms, and to see how each relates to all the rest.

In this interpretive process, Torralba and Ambrona will continue to play a large part. Few European mid-Pleistocene sites are nearly as informative as they. In the last analysis, that we owe to the vision, care, and scholarship of F. Clark Howell.

REFERENCES

Beard, P. 1977. The End of the Game. Garden City, Doubleday.

Binford, L. R. 1981. Bones: Ancient Men and Modern Myths. New York, Academic.

———. 1987. Were There Elephant Hunters at Torralba? In The Evolution of Human Hunting, ed. M. Nitecki and D. Nitecki, 47–105. New York, Plenum.

Butzer, K. W. 1982. Archaeology as Human Ecology. Cambridge, Cambridge University.

Carbonell, E., C. Diaz Fernández-Lomana, J. Enamorado, and A. Ortega. 1984. Análisis morfotécnico de la industria lítica de Torralba (Soria). Antropolgía/Etnografía 4: 203–216.

Cerralbo, E. Aguilera y Gamboa, Marqués de. 1909. El Alto Jalqon: Descubrímientos Arqueológicos. Madrid, Fortanet.

———. 1913a. Torralba, la plus ancienne station humaine de l’Europe? Congrès Internacional d’Anthropologie, Archéologie, et Prehistoire, Genève, C. R. XIV Session, 227–90.

———. 1913b. Torralba, la estación mas antigua de Europa entre las hoy conocidas. Asociación Española para el Progreso de Ciencias 1(4): 197–210.

Freeman, L. G. 1975. Acheulean sites and stratigraphy in Iberia and the Maghreb. In After the Australopithecines, ed. K. B. Butzer and G. L. Isaac, 661–743. The Hague, Moutory.

———. 1978. The Analysis of Some Occupation Floor Distributions from Earlier and Middle Paleolithic Sites in Spain. In Views of the Past: Essays in Old World Prehistory and Paleoanthropology, ed. L. G. Freeman, 57–115. The Hague, Mouton.

———. 1981. The Fat of the Land: Notes on Paleolithic Diet in Iberia. In Omnivorous Primates: Gathering and Hunting in Human Evolution, ed. R. S. Harding and G. Teleki, 104–65. New York, Columbia University Press.

———. 1991. What Mean These Stones? Remarks on Raw Material Use in the Spanish Paleolithic. In Raw Material Economy among Prehistoric Hunter-Gatherers, ed. A. Montet-White and S. Holen, 73–125. Lawrence, University Press of Kansas.

Freeman L. G., and K. W. Butzer. 1966. The Acheulean Station of Torralba, Spain: A Progress Report. Quaternaria 8: 9–21.

Frison, G., ed. 1974. The Casper Site. New York, Academic.

———. 1991. Prehistoric Hunters of the High Plains. New York, Academic.

Howell, F. C. 1989. Review of M. Nitecki and D. Nitecki, The Evolution of Human Hunting. Journal of Human Evolution 18: 583–94.

Howell, F. G., K. W. Butzer, and E. Aguirre. 1963. Noticia preliminar sobre el emplazamiento Achelense de Torralba. Excavaciones Arqueológicas en España 10: 1–38.

Howell, F. C., and L. G. Freeman. 1982. Ambrona, an Early Stone Age Site on the Spanish Meseta. L.S.B. Leakey Foundation News 22: 11–13.

Howell, F. C., L. G. Freeman, K. W. Butzer, and R. Klein. 1992. Observations on the Acheulean Occupation Site of Ambrona (Soria Province, Spain), with Particular Reference to Recent (1980–1983) Investigations and the Lower Occupation. In Die altest Besiedlung Europas, ed. G. Bosinski. Jahrbuch des Romisch-Germamischen Zentralmuseums, Mainz: (1992):33–82 + 2 Tafeln.

Klein, R. 1987. Reconstructing How Early People Exploited Animals: Problems and Prospects. In The Evolution of Human Hunting, ed. M. Nitecki and D. Nitecki, 11–43. New York, Plenum.

Obermaier, H. 1916. El hombre fósil. Comisión de Investigaciones Paleontológicas y Prehistóricas. Memoria 9.

———. 1925. El hombre fósíl. Comisión de Investigaciones Paleontológicas y Prehistóricas. Memoria 9 (Second Spanish Edition).

Salvador, A. 1974. Guía de los anfibios y reptiles españoles. Madrid, Instituto Nacional para la Conservación de la Naturaleza.

Santonja, M., and P. Villa. 1990. The Lower Paleolithic of Spain and Portugal. Journal of World Prehistory 4: 45–94.

Shipman, P., and J. Rose. 1983. Evidence of Butchery and Hominid Activities at Torralba and Ambrona: An Evaluation Using Microscopic Techniques. Journal of Archaeological Science 10: 465–74.

Vaurie, C. 1965. Birds of the Palearctic Fauna: Non-passeriformes. London, H.F. & G. Witherby.

Villa, P. 1990. Torralba and Aridos: Elephant Exploitation in Middle Pleistocene Spain. Journal of Human Evolution 19: 299–309.

———. 1991. Middle Pleistocene Prehistory in Southwestern Europe. Journal of Anthropological Research 47: 193–217.