TEN

Research on the Middle Paleolithic in the Cantabrian Region

WHERE HAVE WE COME FROM? WHERE ARE WE NOW?*

INTRODUCTION

The story of investigations of Mousterian sites in Cantabria has respectable antiquity, and Cantabrian research since its commencement has made contributions of great consequence to our understanding of the Mousterian complex of industries. Eduardo de la Pedraja first excavated Mousterian levels at the site of Covalejos in 1872 and Fuente del Francés in 1880; between 1878 and 1880 Sanz de Sautuola himself discovered the archeological deposits in the Cueva del Pendo or San Pantaleón (though without excavating its Mousterian materials). The recovery of Mousterian materials continued sporadically throughout the earlier part of the twentieth century. During the 1950s, the leading authority on the Spanish Mousterian was Prof. Francisco Jordá Cerdá, director of the Museo Arqueológico Provincial in Oviedo, who himself had developed an overview of the peninsular Mousterian, including the Cantabrian collections. Under the direction of J. M. de Barandiarán, J. Altuna, and others, some research had been done at the Basque sites of Lezetxiki, Axlor, and Amalda. Other specialists had shown an interest in the Mousterian assemblages from some Cantabrian sites, particularly the well-known caves at Castillo, el Pendo, Cueva Morín, and Hornos de la Peña, and the lesser-known sites of la Mora, the Fuente del Francés, the Abrigo de San Vitores, and the railroad cut at Unquera. Most of the investigators of Cantabrian sites were local, but others, including Breuil, Obermaier, Shallcross, and Wernert, were foreign: Cantabria has always received serious international collaborators cordially, placing very few restrictions on the activities of those willing to share their research responsibilities with their Spanish colleagues. The result has been a fruitful cross-fertilization of theoretical approaches, which, when coupled with the richness of the region in Paleolithic remains, has placed Cantabrian research at the forefront of Mousterian studies.

Needless to say, some important Mousterian sites discovered in Cantabria were first investigated many years before my own studies began (some during the nineteenth century), often with techniques that today are considered unacceptably primitive, and in a few cases, they have since been destroyed and their collections lost. Other excellent investigations remained poorly published. Investigations of the Cantabrian Mousterian were scarcely the central focus of attention for the majority of prehistorians. In fact, as a student, I was told that “most prehistorians seem to be afraid of the Mousterian. Whenever a set of tools that are large, or even smaller but crudely made, but are not obviously Acheulean handaxes are found, instead of trying to understand it they call the collection ‘Mousterian’: that effectively sweeps it under the rug and out of sight.”

Excavations at the Cueva del Castillo in the early part of the last century were very respectable for their time, and yielded spectacular and rich Mousterian collections. A timely and full publication of those excavations would almost certainly have stimulated Middle Paleolithic investigations in Cantabria. But monographic publication of research results at Castillo had to await the painstaking synthesis of the late Dr. Victoria Cabrera in 1984 (Cabrera Valdés 1984). Mousterian studies would surely have received renewed impetus had the truly superb investigations undertaken during 1953–1957 by an international team under the direction of Dr. Julio Martínez Santa-Olalla been published expeditiously, but that was not to be, and no monographic description of that work appeared until 1980 (González Echegaray et al. 1980). It was not, in fact, until the last quarter of the twentieth century that new generations of scholars began to focus attention once more on this fascinating period of prehistory. Then, under the direction of the author and Dr. Joaquín González Echegaray (at Cueva Morín and elsewhere), Drs. Victoria Cabrera and Federico Bernaldo de Quirós (esp. Castillo and el Pendo [Montes Barquín, J. Sanguino et al. 2001]), and Ramón Montes (esp. at el Pendo), new teams of investigators once more began excavating Mousterian levels and analyzing the assemblages they yielded. We have every reason to expect that the new excavations at Castillo will add appreciably to our knowledge of Cantabrian Mousterian adaptations when they are fully published.

The situation outside Cantabria at the time I began my own studies of the Mousterian was summarized by the Neanderthal Centenary Conference in Düsseldorf in 1956, published in 1958. Contributions to that conference volume (Von Koenigswald 1958) raised questions of two sorts. First of all, what is the relationship of the Neandertals to modern humans in Europe and Asia? Second, what is the nature, and what are the relationships, of the Mousterian complex of industries? And what, if anything, can answers to the second question tell us about the first?

WHAT HAVE WE LEARNED?

Studies of Neandertal skeletons elsewhere in Europe reveal a great many differences between the physical characteristics of those skeletons and our own. Apparently, Neandertals used their front teeth to help hold and pull things in ways that would be quite unusual for modern humans. Biomechanical studies suggest that Neandertals were capable of throwing things very forcefully, and both of hugging items close to their bodies more powerfully, and with a stronger, more vice-like grasp between thumb and forefinger than is the case for the average modern. But, with the exception of the promising remains from el Sidrón in Asturias, there are no Neandertal skeletal remains from Cantabria.¹ Certainly, material that most agree is Neandertal has been found in some abundance elsewhere in Spain, but the question of the relationship between Neandertals and modern humans can still not be answered using bones from the Cantabrian sites. Local evidence all bears on the nature and relationships of the Mousterian complex of industries.

Despite all the contrary claimants, the analysis of this evidence must build on the pioneering work of François Bordes and a few of his dedicated French colleagues, such as his wife Mme. Denise de Sonneville-Bordes, and his teachers and collaborators Denis and Elie Peyrony, Maurice Bourgon, and Paul Fitte. Instead of studying single guide-artifacts, Bordes had shown that one could greatly improve our understanding of Mousterian collections by introducing systematic principles of artifact classification to replace the chaotic, unsystematic, and overlapping type definitions in previous use and by studying artifacts as parts of whole assemblages. Bordes had examined thousands of Mousterian artifacts from scores of archeological sites, many of them well excavated. He had discovered that the majority of these French assemblages fell into a few modal types he called the facies. The facies, he thought, were the stylistically distinctive products of different identity-conscious socio-cultural groups, rather than groups of tools with differing technological functions. He had started to extend his classification to other regions, including Italy, Central and Eastern Europe, and China. His work on climatic succession documented by loess deposits in the Seine basin showed that some of the facies were not chronological stages of assemblage evolution but rather long endured side-by-side, and he was able to dismiss some supposedly independent “cultures” such as the Levalloisian as really only technical variants of ordinary Acheulean or Mousterian industries, showing that others such as the Tayacian were mostly cryoturbated or geologically crushed pieces. His study of Paleolithic industries led him to question how the earliest Western European Upper Paleolithic artifact industries, then all assumed to have been produced exclusively by modern men, might relate to earlier Mousterian. Others, too, were fascinated by this question, and were divided into several camps, some of which spoke of transitional Mousterian–Upper Paleolithic industries. To Bordes, the only obvious transitional industries were the Chatelperronian and the Aurignacian “0.”

In France, by the 1950s most prehistorians were satisfied that a broad picture of the evolution of Paleolithic artifact industries had been formulated, as had a framework of temporal periods and regional paleoenvironmental successions against which industrial development could be viewed. Crude though these schemes were known to be, they were judged reliable enough and adequate in general outlines. What remained to be done was to fill in the gaps in this picture: most Paleolithic prehistorians regarded the principal task facing them to be an increasing refinement of chronology and the ever more refined study of the microevolution of regional climates and artifact industries. No one then seemed to realize the extent to which Bordes’s concept of the facies, his recognition that they could not be explained as chronologically successive stages of industrial development, and his alternative proposal that they were instead the stylistically distinctive products of different social groups that developed independently of each other had begun to undermine this plan. In my own research, I was at first no exception.

In Cantabria, no one trained in the use of his methods but Bordes himself and his colleague Jacques Tixier (they had together studied the partial Castillo collection in Paris) had attempted to classify the Mousterian collections, or to examine the supposedly transitional industries. (On account of the striking abundance of cleaver flakes in the Castillo Mousterian Level Alpha, Bordes had assigned that collection to a postulated new facies, which he misleadingly called the “Vasconian.”) Consequently, when I began my doctoral research on the Cantabrian Mousterian in 1962, my first task seemed logically to be the classification of all the tools in museum collections from the local Mousterian levels according to the Bordes typology, and the assignment of the collections to the facies that Bordes had defined, wherever possible. Only after having assigned the Cantabrian assemblages to Mousterian facies, establishing “Quantas maneras son dellas,” as Alfonso “El Sabio” would have said, could one proceed to determine why they differed and how they were related. Secondarily, the evidence for changing Middle Paleolithic paleoenvironmental conditions would have to be reevaluated, so that the relationship, if any, between them and the contemporary Mousterian industries could be determined. In the process, it would also be necessary to examine all the so-called transitional Mousterian–Upper Paleolithic industries, to determine their makeup and relationships.

In 1962/3 I classified all the museum collections of Mousterian artifacts from Cantabria, and beginning with our excavations at Cueva Morín in 1969, González Echegaray and I eventually managed to study two dozen trustworthy Mousterian artifact collections, including nineteen from sites we had excavated or tested ourselves. We found that they included representative assemblages of Typical Mousterian, Quina Charentian, and Denticulate Mousterian, including a super-denticulate type with a much higher proportion of denticulate tools than was usual in French collections. Cleaver flake–bearing collections or assemblages from Castillo, el Pendo, and Morín proved to be otherwise heterogeneous, and so assignable to different facies, which effectively eliminated the so-called Vasconian from Bordes’s facies list. Despite the contrary opinion of some early “authorities,” the cleaver flake–bearing Mousterian with tools so reminiscent of Acheulean bifaces was not our oldest Mousterian, but as earlier excavators had correctly noted, was stratified atop levels of such “Mousterian of small types” as the Quina Charentian at Castillo and the Denticulate at Morín. True Chatelperronian was present in one level at Cueva Morín, while none of the supposedly transitional Aurignaco-Mousterian cases was trustworthy. Most such “transitional” levels were poorly excavated, mixing Mousterian with Upper Paleolithic materials, and others, such as the old collections made by the Conde de la Vega del Sella at the Cueva del Conde in Asturias, included purely Mousterian or Upper Paleolithic levels containing tools that had simply been misunderstood. What is more, the Chatelperronian from Cueva Morín, and that later identified in the collection of tools from Level VIII, overlying the earlier Aurignacian Level VIIIa at el Pendo, did not look like it sprang from the local Mousterian.

It has been both a strength and a weakness of my own research on the Cantabrian Mousterian that it has consisted entirely of the reinvestigation of sites that were previously known and excavated. Where I was restricted to information from older Museum collections, the available paleoenvironmental data were most often deficient. Such information, where present at all, was fragmentarily reported and only selectively saved, and so, working with materials that González Echegaray and I had not gathered ourselves, the collections of stone tools necessarily became our major focus.

In our investigations, we introduced statistical procedures such as the Kolmogorov-Smirnov test, rank-order correlation, and principal components analysis for the evaluation of the contribution of random sampling error to difference between assemblages. We were first able to confirm the lack of evidence for chronological development in the collections studied: no tool type ever replaced any other through time. That of course was just what we had expected. What we learned next was a surprise. We had worked to assign the collections we studied to Bordes’s facies. But the statistical tests that showed that some types of tools were related to each other, and not related to other types, soon also indicated that the Bordes facies intergraded along axes of scraper richness as opposed to denticulate richness. The facies, we showed, were completely arbitrary constructs of the classifier. That becomes obvious to the eye when the Bordes diagrams of cumulative percentages of tool types in each are compared (Freeman 1994: 37–54, esp. fig. 4.2). What is more, the differences between these differently covarying tool categories seemed to be functional rather than stylistic. To clinch this interpretation, we found tools of two facies—Typical and Denticulate—in substantial numbers in separate restricted areas within a single archeological level (Freeman 1992b), where the areal distributions of tools were so small that no one could rationally suggest that two tribes had occupied that one level at the same time.

Despite my continuing insistence, which I shall return to below, that there is no necessary parallelism between the body forms of the makers of prehistoric industries and the kinds of tools they fabricated, there are nonetheless some very intriguing characteristics of Mousterian assemblages that seem to indicate that their makers’ mental capacity, degree of cultural complexity, and socialization procedures, and the rapidity of cultural process, must have been quite different from those of fully modern peoples. Several Mousterian stone tool types such as cleaver flakes seem to endure immensely longer in the archeological record than do comparable Upper Paleolithic types, and the measurements of such tools from different levels or sites—even sites separated by great distances—do not seem to differentiate them well; comparable Upper Paleolithic types from neighboring sites or adjacent levels are more often quite distinctive in their metric characteristics. In my opinion, some Mousterian levels (e.g., Morín, el Pendo) contain tools such as sidescrapers made on irregular pieces of longbone. The working edges of such pieces replicate in shape and in representation the several types of stone tools in the assemblage, but no attempt has been made to give a predetermined overall shape to the bone support. Since this technique is so different from that used to make Upper Paleolithic bone tools, where the overall shape of the piece is more often conformable to a regular pattern, some prehistorians have doubted that the Mousterian bone pieces are tools at all. Nevertheless, that is what they seem to be. Manifestly, the degree of regularity of Mousterian “models” and the application of Mousterian “standards” of form and measurement must have been much less rigid, and tolerated much more variability, than was the case during the Upper Paleolithic. That suggests that the “socialization process”—Neandertal modes of teaching and learning—must have been markedly different from our own.

There are comparable differences between Mousterian and Upper Paleolithic cave occupants in the treatment of the space they lived in and the rigidity of patterning of scatters of tools they left over the living floors (Freeman 1992a). Upper Paleolithic peoples subjected their living space to violent alterations, periodically performing a drastic “housecleaning,” artificially lowering and leveling the cave floors, and shoveling previous accumulations of debris out of the living area. They dug house foundations, graves, postholes, and storage pits into earlier levels, sometimes moving tons of earth and stone. While in a few sites Mousterian peoples did some of those same things, by and large they took the cave surface as a given, building atop it despite its irregularities, rather than digging into it to rearrange it and discard whatever got in their way. And last, while all Neandertals seem to have at least occasionally made some use of coloring materials, and some scratched regular marks on nummulites or bones (Freeman and González Echegaray 1983; Henshilwood et al. 2004), while a few Middle Paleolithic people may have perforated shells for suspension, or buried their dead, such behavior is remarkably rare in comparison with the abundance of remains of Paleolithic art, adornment, and even human burials that are regularly found once fully modern Homo sapiens sapiens appears on the scene. These bits of evidence all suggest that Neandertals were behaviorally quite different from their successors. But though it is suggestive, that, I insist, is only the very most tenuous and indirect evidence against their relationship to moderns.

WHERE DID WE GO WRONG?

Unfortunately, and this is particularly the case now that so much information is gleaned from the Internet, the great amount of true wisdom that has been accumulated is passed on with a lot that is not truly wisdom, but misinformation masquerading as such. In the last years there has been an explosion of both kinds of “data,” misinformation as well as truth. It has always been the case that we should reexamine or challenge past ideas, since much of the accepted wisdom from the past is mistaken, though from the stature of its authors it has often been accepted as unimpeachable law. It is the duty of every generation to test the “truths” passed on by our predecessors by comparing them to what our own experience has taught us. This is not just the task of the archeologist or the anthropologist, by the way; it is the task of everyone who pretends to make a real contribution to knowledge, whether that be in the arts or the sciences. We should learn from the past what it offers that is good, but never fear to test its conclusions, and we must reject those that in our experience do not ring true. In the course of development of Mousterian research, all of us have unfortunately been led astray by a number of errors that we should now do our best to eliminate.

Let me begin by enumerating those errors that are easiest to understand. Any decent archeologist probably knows at least one case of modern work that suffers from these mistakes. I must note that these potential pitfalls are not all unrelated. In fact, errors of one kind are often partly due to misconceptions of another kind; as in all aspects of human endeavor, the potential errors are complex and intimately interrelated. With a few exceptions, I do not intend to mention names of living archeologists who have fallen into blatant error, or to embarrass any of our colleagues, none of whom, I am sure, would ever be guilty of such gross blunders.

(1) First of all, it is deplorably still true that some excavations are not conducted with adequate care. Poor excavation technique and inadequate care for stratigraphic distinctions still plague the study of Middle Paleolithic industries. I do not mean to imply that there is only one way to dig a site, for sometimes it may be justifiable to ignore a good deal of minor detail in the interest of illuminating the grander aspects of prehistoric life.² It would make little sense to search for the outlines of an open-air settlement and the walls of the huts within it with the same painstaking dental-pick and paintbrush techniques and concern for microstratigraphy used in the excavation of a small-scale Paleolithic occupation in a cave or shelter. Nor would the agencies that fund our research be likely to support such slow and painstaking work at such an obviously complex site. But until more intact open-air Mousterian sites are found, the observations below will apply.

Unfortunately, even otherwise well-trained excavators may continue to follow the mediocre methods of traditional practice even while acknowledging that they are deficient. It is still common practice to dig a small window deep into the site sediments in order to “trace the evolution of climate and industries,” or to find the earliest occupation of a site, despite the fact that it is impossible to detect stratigraphic inversions, or delineate the distribution patterns that might indicate discrete activity areas or the presence of structural features, from such areally limited exposures. It would be much better to forget all about finding “the oldest level in the site,” and to concentrate instead on exposing as extensive a surface as possible within each natural level excavated. In fact, my experience suggests that an exposure less than 20 meters square in a single level is too small to let us appreciate any single aspect of how people actually used a site, and even that area, larger than many current excavations, is too small to show us all the activities that they may have performed there.

To keep track of the positions of small recovered finds, it is essential not only to excavate and process the earth from each natural level separately, but also to subdivide a level’s contents into sublevels of about 5 centimeters’ thickness, and to subdivide the surface of each meter square excavated into 9 sectors (sizes that have been shown to be productive in practice). Earth recovered from each such unit should be processed separately for best results.

Too many excavators still seem to believe that in archeology the word “level” means “horizontal.” It does not! It simply means “layer,” and the natural layers of deposition may be highly irregular in thickness and orientation: they lie conformably at base with the surfaces on which they are deposited, and usually have upper surfaces that are also (though differently) irregular in their contours, as the thickness of the layer varies. There is no possible excuse for continuing to excavate horizontal “spits” in a site where such spits crosscut the orientation of the sediments.

In most sites, I have found that stratigraphic changes are more often identifiable from differences in the texture of the sediments in the levels than they are from differences in sediment color. Furthermore, the edges of dugouts or pits and the lots of debris in a shell-midden may be difficult to detect unless the excavator pays careful attention to the orientation of the bones and stone artifacts in the levels. They often follow slopes or lie oriented along ancient cuts into the sediments. Testing sediments regularly for differences in pH value (acidity) and other chemical characteristics can also help delineate such features.

Sad to say, although we know that very small finds may be crucial to the understanding of activities undertaken at a site (Freeman et al. 1998), most archeologists still make little or no effort to recover them, even where they are present, where water is readily available, and where it would not be difficult to wash sediments through ultrafine mesh screens. We know almost nothing about the plants used by Mousterian communities despite the fact that seeds are well preserved in some sites: they have been recovered sporadically from some Mousterian occupations, such as the horizons at Abric Agut in Cataluña, by simple water flotation.³ We now know that our approach to the recovery of small finds from Mousterian occupations at Agut and Morín was flawed: we should have processed all the sediment excavated from every level, rather than just that small sample whose characteristics suggested that it would be the most productive. In fact, the extra effort and cost of processing all the sediments would almost certainly have been amply compensated by the important new data recovered, as we later learned at el Juyo.

Last, the utmost care in applying the best modern principles of excavation is useless without extremely careful and detailed recording, using field notebooks, plans, sections, and good photographic documentation as well. Too often, this is where our practice fails.

(2) A related problem is often manifest in the improper “dating” of assemblages or of unusual specimens. An especially insidious example is the dating of surface finds by the assumed ages of the surfaces on which they are found. Many archeologists do not understand the complications involved in the formation of river terraces or raised beaches, and interpret them oversimply.⁴ Materials found in the various levels of a river terrace or beach do not all reflect uniform environmental conditions, and the sediments found in upstream deposits may have formed under climatic conditions that were the reverse of those responsible for contemporary deposition downstream. What is more, archeologists often confuse a date that establishes a terminus ex quo with a true, precise, and accurate age for finds.⁵ An item actually dropped by a human atop a river terrace is no older than the terrace but may be many thousands of years younger. (But it is also true that materials derived from older deposits may be eroded from them and deposited atop much younger terraces.) In fact, we should all be aware that artifacts of many different ages may be found intermingled in contexts that cannot be differentiated in or on terraces. Attempts to date assemblages by the nature of biological materials found in deposits with them can be problematic for other reasons. Postulated stages of microfaunal evolution or the development of vegetation are often based on unverified (or obviously flawed) assumptions.

One of our major problems is the lack of suitable radiometric dating techniques for deposits of Mousterian age. Decisions about the age of Mousterian assemblages are made instead on “relative,” indirect criteria, such as the supposed geological age of the sediments in or on which they are found, or on a few uncertain direct “absolute dating” techniques such as amino acid racemization, thermoluminescence, or electron spin resonance, which require us to make prior assumptions that may be totally invalid. Partly because of these uncertainties, we are inclined to discard what may prove to be accurate dates that do not correspond to our present preconceptions, either adjusting the parameters of the tests until the results “seem reasonable” or accepting only those ages that can at best be considered “consensus dates,” in better agreement with what we think we know. I remember that, when potassium-argon dating was first applied to them, there was a great debate about the age of Villafranchian deposits and the Plio-Pleistocene boundary in which many insisted that both had to be much younger than the new (and approximately accurate) dates showed. Then, too, we are still hampered by the idea that one can date specimens by the apparent crudeness or primitive characteristics of their form, which leads us to postulate a great age for hastily made expedient artifacts or pieces found as roughouts that were discarded at quarry sites because of their imperfections.

Inaccurate dating is only one of the problems we must avoid. It appears so large only because we have been taught that to study an assemblage, one must know its age.⁶ That is only true for some purposes. Many important questions about the past can be answered by studying well-excavated assemblages whose absolute or relative ages are unknown.

(3) There are several other potential traps for the unwary. One of the most insidious of these is our human desire for the simplest possible explanation of complex phenomena. We behave as though we believe that as “scientists” we should emulate physicists or astronomers, for whom, in our distorted caricature of them, simple explanations have historically proven more powerful than more complex ones.⁷ As archeologists, we try to explain such occurrences as the radiation of hominids out of Africa by the invention of fire, the introduction of a single miraculously powerful tool type, or some convulsive climatic change. It is currently fashionable to ascribe the origin of spoken language to the shape of a single bone (Clegg 2004), or to a single gene. While a defect in a single gene can undoubtedly prevent one from being able to speak, having that gene in normal form will not automatically produce articulate language. The ability of a Cervantes was certainly not the result of a single gene. Humans are popularly supposed to be killer apes whose genetically determined instinctual aggressiveness accounts for wars. In fact, the decision to wage war is not individual but collective; it is a political decision, not an instinctive reaction. Simple-minded environmental determinism is one of the most pervasive of these errors, mostly because so often the results attributed to environmental conditions are themselves complex and hard to analyze. Human behavior always involves too many interacting factors to be explained in simple terms. The explanations for complex cultural behaviors are never monofactorial. Sometimes it is easier to understand how silly such monofactorial explanations can be by reducing them to imagined modern examples. When Javier wins a public office, the outcome may seem to be triggered by a sudden event, but it always involves a multitude of contributory factors; it is not simply due to climatic change.

Now, let me discuss some problems that are subtler but no less dangerous.

(4) While the “hypothetico-deductive” model of archeological investigations, so important among archeologists trained in American and British institutions, rightly claims that the particular theory (preconceived working hypothesis) that one sets out to test determines the kinds of data one collects, it is especially important to keep in mind that the data one excavates may have nothing at all to do with one’s pet theory, in which case it would be the gravest error to discard or disregard the data as meaningless. They may well prove to be of the utmost relevance to some other, equally important but unanticipated, theory. As a corollary, one must never let one’s “theory” assume such importance that it leads one to deny contrary data where they are found. In the Middle Ages, no one knew how shellfish could become fossilized. But fossil shellfish existed despite that ignorance. I have been told that unless I can explain the formation of the soil pseudomorph recovered in 1969 at Cueva Morín, it simply can’t exist. I am unable to explain completely just how the Morín pseudomorph formed.⁸ Nevertheless, it is undeniably true that it does exist.

(5) An inadequate appreciation of the potential and limits of artifact typology brings as a consequence a number of particularly complicated problems.

• Many archeologists still fail to understand that the observations that can be made of artifacts are of only three kinds: (1) one can observe the physico/chemical attributes of artifacts and the traces they bear; (2) one can observe the positions in which they were found; (3) one can observe the relations (both in space and in number) between the artifacts themselves, and between them and all contextual information. The level of detail afforded by the study of their physico/chemical attributes and that of their positions of discovery provide a very limited amount of information in comparison with all that can be learned from relational data. But studies of relational data depend on careful excavation, detailed recording, precise and intelligible labeling of the finds, and their meticulous conservation.

• A major misunderstanding of the potential of artifact classification stems from a belief that it is easy to distinguish “stylistic” attributes—those that differentiate tools of the same sort made at different times, or by different societies, social subgroups, or individuals at the same time—from those “functional” attributes that reflect the artifact’s technological uses, a priori. Ordinarily, however, the distinction of “stylistic” from “functional” attributes is a late stage of interpretation that can usually not be done until after the classification and much of the study of assemblages has been completed. We do know a great deal about the influence of raw material on style. Much stylistic information is superficial decoration that does not affect artifact function, such as weaving patterns, carving in wood or bone, or painting, glazing, and texturing of ceramics. Textiles, wood, bone, and ceramics are media for the manufacture of tools whose functions are relatively limited, but whose stylistic attributes can be much more extensively and freely varied. That situation contrasts with stone tools that lend themselves to a range of stylistic modifications that is extremely limited, in comparison with their extremely varied functions.

• It would be possible to design a classificatory scheme for stone artifacts that minimized the role of stylistic factors, even where those theoretically exist but are hard to identify, and that is precisely what François Bordes did for Mousterian stone artifacts. In fact, he even went further. He insisted that one should always call a knife a knife, regardless of the “accidental attributes” that made a knife from one region such as China look different from a knife made in another such as France. In so limiting his type definitions, Bordes defined his types “functionally,” defining types in terms of their “knifeness” in this case, or their suitability for other tasks in others, and in so doing he virtually eliminated from consideration any attributes of the stone tools that might possibly prove to be stylistic. His classificatory practice, if applied to modern products, would have led him to ignore the very stylistic differences of which he was so proud, that set the folding knives made in Carsac that he used in excavation off from those made in the Charente, or Paris. Having done this, despite the implicit logical contradiction, he defined the Mousterian facies as stylistically distinctive: the products of different “tribes” of Mousterians. If type definitions are not “stylistic,” then the tool types they define can’t be, and neither can the artifact assemblages or “facies” types based on them. These implications of his classificatory practice were obviously unanticipated and unintended, and are still misleading classifiers today.

• Two important alternatives to the Bordes artifact classification, using “traditional” approaches, have been proposed: that of Leroi-Gourhan and that of Laplace (a third approach will be dealt with in a later section). The first of the two is logically flawed, and ridiculously impractical. Leroi-Gourhan proposed listing all the possible variations in form that could possibly be taken by a tool or its edges, and then, after observing what variants actually occurred, to use those as the basis of one’s typology. The number of possible variants being infinite, Leroi-Gourhan will perhaps have enough free time between harp lessons to list them all, now that he is in heaven. The Laplace classification really yields virtually the same types as does the Bordes classification. But it has the disadvantage of disguising types whose names should be comprehensible to any citizen in the street under the cloak of pseudo-mathematical formulae. The meaning of his “equations” cannot always be remembered even by their inventor, who has had occasionally to refer to previous publications of their definitions in order to recall what particular combinations of symbols signified. If our field is to survive, and to find continued support from governmental or private donors, the results of prehistoric investigations must be made intelligible to the broadest audience possible, not just to an elite “priesthood” using an esoteric pseudo-scientific jargon for their description that can only be understood by other initiates.

• But until artifact classification appropriately considers the operational chain of artifact manufacturing—both the nature and staging of materials processing and sequences of core reduction—as well as the progressive changes of artifact edges due to continued use-wear and breakage, further progress in artifact classification will be limited. There have already been attempts, though not always realistic ones, to take these factors into consideration, and there will surely be more and better ones in future.

• Another source of error lies, paradoxically, in our current focus of attention on “whole” artifact assemblages. For a long time that was a strength of the Bordes approach to the classification of Mousterian industries. Artifacts found together in a single discrete archeological level constitute true assemblages. Assemblages, as we now understand them, are composed of the multitude of toolkits used by the occupants of a discrete level. Artifacts found together in museum drawers are always collections, but they can never be considered “toolkits,” and seldom do they constitute assemblages as the excavator found them. Collections are sometimes “split” between two or more museums, often must be “consolidated” due to lack of storage space, and can become detached from relevant information about their provenience, or may become mixed through movement during cleaning, or building reforms (rebuilding, adding a room, etc.), or during careless study. Consequently studies based on museum collections, particularly older ones from less than adequately controlled excavations, but even excavations conducted according to the best modern criteria only to suffer from careless warehousing and slipshod curation, are often overambitious and overoptimistic as to the significance of their results. My own doctoral dissertation was to some extent flawed because of such effects, but fortunately they did not alter its major conclusions.

• Another problem faces the student of assemblages from the best, most careful, and up-to-date excavations. We owe to François Bordes a pioneering insistence that the attributes of all tools in a whole assemblage should be considered by the classifier. While that sounds laudable, it is less so in practice. No assemblage is ever truly “whole,” in the sense of representing all the activities undertaken by a human group, or the tools used in them. In fact, because of the disappearance of perishable items, and the areal limits of excavation, they cannot be more than a sample of the materials made and used in a single archeological level. To regard them as “complete assemblages” is simply self-delusion.

• Bordes’s insistence led historically to a great improvement in our understanding of Mousterian assemblages and their differences, including the definition of the supposedly distinctive “facies.” But, we have now reached the limits of progress in that direction. To progress further, it seems clear that we must shift the focus of our efforts to attempts to understand the different “toolkits” that go to make up an assemblage, rather than on the “study of assemblages as wholes.” (Of course, our study must still take into account all the artifacts recovered from each level excavated.)

(6) One of the most important potential causes of error and one that is still very much with us is due to an inadequate appreciation of the random factors that always cause difference between assemblages. The use of well-tested statistical procedures is essential for their evaluation but those procedures are not well understood by most prehistorians. It is commonly held by a surprising number of prehistorians that a simple subjective comparison of lists of percentages is “The Statistical Method.” That is a naïve misunderstanding. Most such scholars believe themselves incompetent to understand any more complicated statistical tests, and so whenever they read a publication that includes mathematical calculations, they may reject its results as completely unintelligible and thus irrelevant (when in fact it may be quite correct and crucial to their own research). On the other hand, they may accept its results as totally accurate without trying to evaluate them (and in the latter case they seldom dare to contradict the results of such calculations even when they disagree with all the reader’s past experience or with common sense).

But those who publish such calculations often do not understand them any better than their naïve readers. Their work may excusably incorporate honest mistakes in the use or interpretation of tests, or simple errors in calculation. One glaring example from the past was a well-meaning but misguided attempt to alter random sampling procedures, to distribute the samples taken in a program of surface collection more regularly over the ground, so that they could be used both as random samples and as systematic samples for mapping purposes. The proposed modification, called “Systematic Stratified Non-Aligned Sampling,” actually violated the principles of the random sampling method to such a degree as to make the samples useless for statistical comparisons.⁹

On the contrary, it is inexcusable when results that make little sense are presented dishonestly, and when, if they contradict previous interpretations, they are falsely claimed to be “complementary” rather than contradictory. Unforgivably, a very few of our colleagues have tried to make their reputations on quantitative obfuscation, assuming that most of the rest of us are too mathematically unsophisticated to catch them.

If you feel mathematically incompetent, but will at least try to learn a few standard descriptive and analytical statistical tests, you will soon be surprised at how easy they are. In my experience it is easier for a prehistorian to train him- or herself as a statistician than it is to get a professional statistician to understand the kinds of problems a prehistorian is faced with, and to select appropriate statistical procedures for use in our research.¹⁰ Unfortunately for those who are ill-at-ease with mathematics, the proper use of statistical testing is becoming an essential part of prehistoric research. That is because difference between archeological assemblages may be due to any of several factors, but among them one that is always involved is random sampling error.

Random error is partly responsible for differences between assemblages, but it is also one cause of difference within an assemblage: differences in content between groups of items found in different areas within a single archeological occupation. Prehistorians seem to assume that artifacts are usually distributed uniformly over an occupied surface, or if not uniformly, “at random,” and that collecting a “large enough” sample of artifacts from each level will somehow “compensate” for these differences in distribution. That is not the case. No matter how large they may be, no two samples drawn from a single population of artifacts are ever identical, any more than two handsful of black or white marbles from a jar whose contents are mixed are ever expected to be identical. Only by the use of standard statistical tests is it possible to estimate the likely amount of difference between them that is due to “random sampling error” alone. That error is meaningless in cultural or depositional terms, and so of no further interest to the prehistorian. Many prehistorians (and other specialists, such as palynologists) attempt to compensate for the effects of random sampling error by restricting comparison to samples larger than some arbitrary size, such as 100 or 150 items. But this practice does not have the desired effect. Even differences between samples of 150 pieces that appear to be meaningful may not be significant, while it may be shown that some differences between samples that are much smaller (30–40 pieces, say) are highly unlikely to be due just to chance when they are evaluated by standard statistical tests.

This prompts me to mention yet another approach to artifact classification: what the late Hallam Movius and his students called “attribute cluster analysis.” Its use depends on having enough pieces bearing some particular combination of attributes to demonstrate that the cluster of attributes is not “accidental,” but statistically significant. But most clusters of attributes are not represented with sufficient frequency in archeological assemblages of “average” to large size to permit statistical tests such as chi-square to show that they are significant. Sometimes all that is possible is to prove that retouched pieces are significantly different from unretouched products, by-products of the flaking process, and waste. Though a valid approach, attribute cluster analysis would in the best case imaginable lead to descriptions of assemblages each of which would be quite different and none of which could be codified in the same conventional set of terms, making results hard to communicate to other professionals. Artifact types that really do exist but are rarely represented in any assemblage would be impossible to define with this procedure. That leaves us on the “middle ground” of compromise classifications that combine attribute clustering analysis with both analyses of the operational chain of artifact production and use and traditional procedures for tool typology.

For the evaluation of the significance of differences between long lists of items arranged in a traditionally accepted order, such as the Bordes list of Mousterian artifact types, I have long favored the use of the Kolmogorov-Smirnov two-sample test, which the statistician Leo Goodman (1954) has shown to give results that are at least as strong and reliable as chi-square. Of course, the Kolmogorov-Smirnov test takes difference of sample size into account, and for long lists of different classes of items, it is very much easier to calculate than is chi-square or any suitable alternative.

Once the contribution of random error has been evaluated, the remaining difference is due to factors that are of interest for interpretation. Here, one caveat is called for. The larger the samples compared, the smaller the difference between them need be to reach “significance.” Consequently, statistically significant differences between some very large assemblages may be so tiny that they are essentially useless for interpretation. (It may be fortunate that archeological samples are seldom that large.) Archeologically interpretable “significant” differences can be due to phenomena at the time the artifacts were deposited, or to forces that only acted after the deposits were laid down. Then, one must eliminate from consideration those differences that are likely due to rolling or realignment by flowing water; gravity movement downslope; the action of scavengers, bacteria, and fossorial animals; or redeposition by human activity such as the construction of dugouts or “housecleaning” activities, etc. When that is done, one is left with differences due to phenomena contemporary with deposition, including geological and biotic agency, as well as past human behavior, whose reconstruction is of the greatest possible interest to us. Even in their study, standard statistical tests are essential. Instead of interpreting every item found in a grave as a deliberate offering to the dead, one must first show statistically that the items in question are not accidental inclusions. At Cueva Morín, a Mousterian cleaver flake was recovered from the sediments in a prehistoric grave, and consequently one famous prehistorian insisted that it was a grave offering and the burial must be Mousterian. But its stratigraphic situation proved that the burial was Aurignacian, not Mousterian. The cleaver flake was accidentally included in earth from levels into which the grave had been dug.¹¹

(7) After the contribution of random factors has been evaluated, differences within assemblages still remain, as do differences between assemblages. From the very remote past, our ancestors performed different tasks in different places. Mousterian peoples arranged the ground surfaces on which they worked to some extent to fit these different purposes, sometimes digging graves and pits, or building walls to divide the surface into different zones. Upper Paleolithic people altered the earth they lived on more drastically, shoveling sediments out of their living areas during periodic housecleaning, or to smooth out irregularities, lower the level of their living floors, or build dugout structures. Nor do the forces of nature act uniformly. As one example, less buoyant pollen grains fall from the air currents that carry them closer to the entryways of caves than do more buoyant grains. Sometimes the differences in pollen fallout that occurred simultaneously in two parts of a cave vestibule can be as great as the differences that most pollen analysts would attribute to major climatic change. The implications of these factors contradict a widely held tenet of prehistory. One cannot assume that a single long section cut into the deepest part of its deposits ever contains all the important strata that are represented in the site, nor can one ever assume that the activities undertaken during each of the occupations exposed in such a section were identical. There is no such thing as a “typical stratigraphy” for an archeological site.¹² I recognize that this means that our current practice of leaving an intact “witness section” to be excavated by future generations with improved techniques, so that they can check the accuracy of our stratigraphic conclusions, is of dubious value.

Another, related error stems from the fact that the fauna, flora, or sediments recovered from a particular level are sometimes thought to be adequately “representative” of environmental conditions when the level formed. In other cases, there have been attempts to reconstruct prehistoric “diet” and hunting practices based solely on the faunal remains found in a level. Not only does this ignore other components of diet than animals, it also assumes that the bones found represent all the animals hunted. Obviously, this is wrong. Some sites are specialized hunting or butchering camps, where only the remains of those animals scavenged or hunted on a limited number of occasions are to be found, and then the bones are unlikely to represent all elements of their skeletons. Others are base camps or dwelling sites, or sites of other types. Some items such as berries or small animals may have been consumed on the spot, and not brought back to a dwelling site at all. A mammoth, whose bones are huge, is likely to have been butchered where it was found or killed, the bones left on the spot, and only the meat carried to the site where the hunters lived at the time. Until other indications than bones of the amount of meat represented are considered, this means that smaller, more readily transported animals will be wrongly thought the most abundant elements of diet when they are actually overrepresented in the faunal sample.¹³

Yet another related error is involved when one attempts to study potential land use by early prehistoric groups, based on the characteristics of the modern landscape (the procedure of “site catchment analysis”¹⁴). True, advocates of this procedure insist that one should first evaluate changes that the landscape has undergone, but how can one really appreciate the potential uses of different parts of the landscape in Mousterian times, let alone “walk over” any reconstruction of it as it was then? Several other assumptions involved in the procedure, such as the distances prehistoric people might have been able to travel in a given time, and the total time they might have been willing to travel for a particular purpose, are also unrealistic. This technique is best restricted to the study of post-Pleistocene prehistory, for which it was designed, and then “taken with a grain of salt.”

Since one cannot assume that the assemblage from a single excavated level represents all the activities undertaken by a prehistoric group, different assemblages cannot ever be assumed to be related to the same range of tasks. What is more, because the artifacts the excavator recovers are found in different “activity areas,” because these areas are not uniformly distributed over the surface of any site, and because sites are almost never completely excavated in any case, the sites themselves cannot simply be assumed to be equivalent; the prehistorian must prove that they are before considering them to be such. The foregoing observations, incidentally, invalidate all unsophisticated attempts to reconstruct stages of the “microevolution” of environments and industries based simply on relative stratigraphic position.

As a corollary, if one cannot assume that excavated areas or assemblages are equivalent, it is always inappropriate to “increase samples” by combining materials from different discrete levels, since the materials may represent completely unrelated sets of site uses or activities. It is often assumed that one can simply add the faunal remains from different levels of Mousterian to determine the climatic conditions under which the deposits formed, as long as the levels “belong to the same facies.” But since the different levels in all probability witnessed different sets of activities, this practice is fallacious. Equally often, it is assumed that if one finds Denticulate Mousterian assemblages in levels indicative of a particular set of environmental conditions, and Typical Mousterian in other levels indicating similar environmental conditions, there can be no “functional” difference between assemblages of the two types. Since different activities can be and often are undertaken in a single environmental setting, that too is just wrong, and indicates an incorrect understanding of the word “function.”

(8) A most pernicious error is due to the lack among many of our colleagues trained in Departments of “Integrative Biology” in the United States or Faculties of Geology or Paleontology, or even Departments of Archaeology, in other countries, of any real understanding of socio-cultural anthropology, its terminological usage, or its findings. Most frequently, this is manifest in the equation of an industrial complex with “culture” or “a culture.”¹⁵

Bordes thought that all unmixed Mousterian assemblages belonged to one of a few “facies,” which he wrongly thought of as mutually exclusive, non-overlapping phyla of industrial development, the products of different groups of “tribes.” He wrongly thought that “people exchange their genes more willingly than their customary behavior.” His facies were unlike the facies of geomorphology that were his model, since different geological facies can and do intergrade. But the products of his different Mousterian tribes did not. “If a woman from the Quina-type Mousterian was carried off by a man from the Mousterian of Acheulean Tradition, she might perhaps have continued to make the thick scrapers of her own tribe (they are found sporadically in the Mousterian of Acheulean Tradition), but it is more than probable that after her death no one would continue making them.”¹⁶ In order to define the facies the way he did, however, Bordes had to ignore a number of collections from earlier excavations, dismissing them as mixed or otherwise contaminated, since they did show considerable intergradation. Later, excavations by de Lumley in France and Ripoll and Freeman in Spain showed that even well-excavated assemblages whose stratigraphic position was adequately controlled often intergrade as though they represented two or more facies at the same time.

Bordes (like his predecessors Breuil and Denis Peyrony) also had a mistaken notion of the nature of “tribes,” or of how different socio-cultural groups that are aware of their distinctiveness differentiate themselves from their neighbors. When two individuals meet, it seems ridiculous to assume that the only way they will be able to determine whether they belong to the same “tribe” and should befriend each other, or to different hostile “tribes,” is by the proportions of different types of stone tools in the assemblages of pieces left on the floors of the caves they live in. Yet that is not much of an exaggeration of Bordes’s position regarding the distinguishing characteristics of the Mousterian facies. Even when “tribes” are defined as polities rather than linguistic units, anthropologists know that they use many other marks of their identity (some intangible, but others material and potentially visible archeologically), and different tribes may use tools that are largely indistinguishable. When their social identity is important to them it is signaled in more obvious ways than the makeup of their tool assemblages.¹⁷ Outward signs of identity are given by such aspects of cultural behavior as gestures and body carriage, body mutilation and decoration, clothing, hairstyles, house form, food, the use of flags and banners, and many others. Several of these signs can be distinguished at a distance.

A tribe is a multiperson social category whose activities require substantial space. At Cueva Morín, we have shown that large but partial artifact assemblages that would have to be classified as belonging to different facies are found together, occupying different areas, within one single occupation level, and the coexistence of two separate “tribes” occupying such small spaces would be a physical impossibility (Freeman 1992b). In this case, a functional interpretation of the facies differences is the only possible alternative. But what is more, once the arbitrary nature of the distinction between facies was recognized, it became clear that it had been a mistake to spend so much effort in a search for the reasons for the difference between them. If facies intergrade, the distinctions between them are arbitrary constructs; the facies as Bordes defined them really do not exist, and so their differences are trivial, and a search for the causes of those differences is fruitless. The facies concept should be abandoned.

(9) Last, one of the commonest and most dangerous misconceptions in prehistoric research is that artifact assemblages are equivalent to, and behave like, human types. Human bodies are human bodies, and tools are, after all, only tools. Tools have no genes, and they cannot reproduce. It was once believed that every human type used a different set of tools,¹⁸ so that European Homo erectus was coterminal with large Acheulean bifaces (choppers and chopping tools in Asia), now-debunked “Piltdown Man” made crude flake tools, while Neandertals made and used either small bifaces made on flakes (early) or such small tools as Mousterian points and sidescrapers (late), and “Cro-Magnon man” was the author of the Aurignacian. We now know that this was simply not so. Except at the very dawn of toolmaking some two and a half million years ago, when the simple flakes and cores called “Oldowan” tools do seem to have been the only types made by the Australopithecines (or earlier hominids), it seems always to have been true that any human type can be found in association with more than one kind of artifact industry, and some industrial complexes such as the Chatelperronian, for a long time considered the earliest Western European “Upper Paleolithic” industrial complex, seem to have been made by more than one kind of hominid. Both Neandertals and fully modern people are now thought to have made the distinctive tools of that complex. In any case, it has been shown that there is no necessary one-to-one correspondence between artifact types and human physical types.

Consequently, when we ask what became of the Neandertals, the answer must be based on the nature of their physical remains and their DNA. When we want to know if they are related to fully modern men, and if so, how, any proof or disproof of relationship must rest on skeletal morphology and other direct evidence of their biology. Studies of the relationship between the Mousterian and Chatelperronian industrial complexes are irrelevant to that question. The fact that both Neandertals and moderns used backed knives (or other similar tools) must be regarded as absolutely meaningless in that respect. Continuing to rely on artifact comparisons to elucidate genetic relationships is to confuse the mechanisms of biological and cultural evolution.

Any kind of evolution is basically just adaptation, but when human bodies are involved its mechanisms or forces are very different from those at work when we consider the evolution of systems of belief and behavior. People who look different may each learn how to behave like the others. While the nature of their bodies can only change relatively gradually, by passing on changes or recombinations of their genetic material from one generation to the next, their behavior can change much more rapidly by the “inheritance of acquired characteristics.” We pass on what we have learned to others, and that process is almost instantaneous when compared to genetic inheritance, since we can and do learn new behavior not just from our ancestors, but also from our contemporaries (including those younger than ourselves) during the course of our lifetimes. François Bordes wrote that the evolution of Paleolithic industrial complexes was “ramifying evolution” (évolution buisonnante), in which the various branches crossed and split in complex fashion. His precocious conclusion was regarded as strange and little used by prehistorians, who preferred to think of cultural evolution as operating by mechanisms in general analogous to the “forces” of neo-Darwinian biological evolution: selection, mutation, gene flow, and genetic drift. Nevertheless, it should have surprised no one. Because of the possibilities for the rapid exchange of information between peoples whose backgrounds and experiences are very different, all cultural evolution is, in fact, “ramifying evolution.”

CONCLUSIONS

This list of pitfalls for the unwary omits many, but highlights some of the most important. Despite the fact that the list of potential causes of error we must still overcome is daunting, the progress that had been made in understanding the Mousterian complex of industries nonetheless surpasses it.

While museum collections from earlier excavations throw considerable light on Mousterian adaptations, much more can be gotten from careful excavation with up-to-date techniques. Our experience shows that there is still much unexpected information that can be learned from the excavation of intact sediments in many sites that were discovered and first (partially) excavated long ago, and it is obviously economical to investigate sites that are already known. But that experience also suggests that more effort should be devoted to searching for and investigating new, previously unknown sites,¹⁹ and particularly those in the open air rather than in caves or rock-shelters. No doubt this search will be difficult, in light of the lush vegetation cover that characterizes the Cantabrian landscape, but road and railway cuts, foundations for new construction, wells, streambanks, beaches, and, in fact, any natural or artificial feature that exposes sediments of appropriate age should be thoroughly examined. This effort should be accompanied by an exploration of sunken undersea grottoes and caves that were partially drowned by freshwater lakes and siphons; French divers have shown that underwater Upper Paleolithic sites exist, and there is no reason to believe that changes in drainage and sea level will not have preserved some Cantabrian Mousterian sites as well. But careful survey and mapping are not all that is needed.

If Mousterian studies are to continue to progress, painstakingly careful excavation will have to become the rule. This will have to combine the best techniques of horizontal as well as vertical control: we must attend assiduously not only to vertical stratification, but also to those spatial variations in the strata that betray the presence of natural differences in or artificial interference with the “orderly sequence of deposition.” Thorough recording, coupling detailed note-taking, precise drawing of maps and sections, and thorough photographic documentation, must unfailingly accompany the excavation. New kinds of artifact classification will have to be developed. Subjecting their results to appropriate statistical procedures, we can first classify the toolkits they compose, discovering how they relate to all aspects of the contextual data found with them. Then we can compare them and the activity areas in which they were found. Working outward, we should be able to demonstrate that comparable areas exist in multiple sites, and perhaps that the functions of some sites are in fact equivalent. By means of comparisons of comparables, we should be able to reconstruct subsistence and settlement systems in relatively broad and accurate terms, and then to discover how functionally similar activity areas and sites differ stylistically. Attention to minor differences in artifact manufacture and wear may someday permit us to identify the different “signatures” of different members of a team and overlap in team membership.

The criticisms of past investigations made or implied in this chapter are not meant to denigrate the contributions of the pioneers of Cantabrian prehistory. As I said earlier, we have all been guilty of at least some of the errors I have outlined above. Despite this fact, each generation of investigators in the past has truly added something to our wisdom about the Mousterian. Cantabria has contributed more than its share of this wisdom. Can we continue to make as much progress in Mousterian studies in the future as we have in the past? I for one am most optimistic. But in the last analysis, the answer to that question depends not on my own opinion, but on the behavior of the members of this and future generations of investigators.

NOTES

1. Of course, there are bones from Axlor and Lezetxiki in Basque country—a fragment of maxilla, several teeth, and a long bone—that have been diagnosed as Neandertal. They have, however, not helped in the determination of relationships between Neandertals and moderns. A fragmentary left parietal found at the cave of la Flecha and possibly from its Denticulate Mousterian levels is thin and fully modern in appearance (Freeman 1964: 269). More information on the Spanish Neandertals is given in González Echegaray and Freeman (1998). I have not yet seen the recent publication of the discovery of a large number of Neandertal remains in the site of el Sidrón in Asturias by Javier Fortea’s team (Fortea et al. 2003). Because I only know them from the presentation to this workshop, I omit any discussion of these finds, although they promise to be of great importance. They were found with some 30 stone tools, a collection that is still too small for accurate diagnosis.

2. This point was made more elegantly more than a half century ago by Sir Mortimer Wheeler (1950).

3. A fuller description of recovery at Abric Agut is published in González Echegaray and Freeman (1998). A previous attempt was made to find plant materials at Cueva Morín, using cruder methods that produced no useful result.

4. “Each terrace corresponds to a period of warmer climate, to an interglacial period” (Leroi-Gourhan 1957: 17).

5. A clearly fallacious example was an attempt to establish the age of a Lower Paleolithic stone industry by dating the basalt of which its stone tools were made. The resulting age might be correct for the basalt, but obviously could be millions of years earlier than the date of toolmaking.

6. “Comme la Géologie et l’Histoire qui l’encadrent, la Préhistoire est avant tout une science de la chronologie. Les comparaisons et les explications ne sont valables que pour des faits insérés dans le déroulement général du temps et datés les uns par rapport aux autres” (de Sonneville–Bordes 1981: 5).

7. Such caricatures are wrong. The current state of quantum dynamics, gravitational theory, or current debates about cosmology, should make that evident.

8. That is, aside from analogies with the mechanisms of formation of other, later soil shadows found in Northern Europe (see, e.g., Clark 1957: plate 17b).

9. For an explanation of statistical requirements of “simple random sampling” programs and their uses, see any good text on basic statistics, or Deming 1966: a probability sampling program “is carried out according to a statistical plan embodying automatic selection of the elements . . . concerning which information is to be obtained. In a probability sample neither the interviewer [the one drawing the sample—LGF] nor the elements of the sample have any choice about who is in the sample. . . . [I]n a probability sample the procedure for forming the estimates is automatic, being laid down beforehand as part of the sample design. Unless these conditions are met, probability theory can not be used to appraise the precision of the results, and a survey can not be characterized as a probability sample” (p. 10). The regular spacing of sampling units that is needed for the best surface gradient mapping can never be random sampling, since some set of elements of the sample determines what other elements are to be included.

10. One of the best introductory works that deals with tests of use to prehistorians is still Siegel (1956) (later editions are available, but not necessarily preferable).

11. Some of the causes of interpretive error related to statistical testing are not so obvious. There are limits to the use of the frequently used chi-square test. The data used should be raw counts, not percentages. When the data are cast in a table larger than 2 × 2, no cell can have an expected frequency less than 1, and fewer than 20 percent may have expected frequencies less than 5. For the 2 × 2 case, all expected frequencies must be 5 or more. When these criteria are not met, the Fisher Exact Probability test must be used instead. Correlation analysis and principal components analysis are two of our most powerful approaches to the recognition of types that vary together in related fashion because they were used for related tasks, or for other reasons. Correlation coefficients calculated for the correlation test itself, or as data for principal components analysis, that are based on seven or fewer cases, are mathematically invalid. H. Harman’s (1967) Modern Factor Analysis, though now old, is still to my way of thinking the best introduction to the theory of and approaches to principal components analysis. Harman additionally presents the rationale for the rotation of axes, mathematical transformations that do not alter the solution but do facilitate the interpretation of the resulting structure of the solution. It is often assumed that because assemblages often vary widely in size, large differences in the numbers of tools of any particular type in different samples will lead to false correlations. That is certainly a possibility. Several means have been proposed to “standardize” sample sizes in order to compensate for this problem. Many of them, such as the transformation of the raw counts to their square roots or logarithms, are essentially useless—they really do not eliminate the size difference between the samples. Some, such as the “chi-square” transformation, yield results that are hard to evaluate. Others, such as the transformation of the counts of tools to percentages of collection totals, “constrain variance” in ways that always produce spurious negative correlations, wrongly making it appear that some types take the place of others, when in fact no such replacement actually occurs. The easiest and most justifiable, rational, and effective way to transform data is the use of rank-order correlation (I use Spearman’s r) instead of the Pearson product-moment correlation coefficient. If data are ranked, so that the most abundant type in each sample is given the highest rank and the least, the lowest, the formula for the ordinary Pearson’s correlation coefficient automatically produces Spearman’s r. However, the dangers of using untransformed raw counts and Pearson’s correlation coefficient with archeological data are often more theoretical than real, and often very nearly identical results will be obtained from the use of either approach to correlation.

12. This affirmation is made as strongly and justified in greater detail in the chapter by J. González Echegaray (1984: esp. 265–67).

13. I have dealt with this topic in greater detail in my article “The Fat of the Land” (1981).

14. This procedure was pioneered by a group under the direction of Eric Higgs (1975: 223–24).

15. “Artifacts made by the same people are an industry. A number of industries belonging closely together in time and space are a ‘culture.’ Ideally, of course, a culture will consist of far more than just so many groups of artifacts, burial places, and settlement types” (Roe 1970: 27). Despite the fact that this definition includes more than stone and bone tools, such “enumerative” definitions of culture are now regarded as anthropologically outmoded. A more modern definition is that culture consists of a society’s system of beliefs and behavior.

16. “l’Homme échange plus volontiers ses gènes que ses coutumes. . . .” “Si une femme du Moustérien type Quina était enlevée par un homme du Moustérienne de tradition acheuléene, peut-être continuait-elle a fabriquer les racloirs épais de sa tribu (on en trouve sporadiquemment dans le Moustérien de tradition acheuléene), mais il est plus que probable qu’après sa mort personne n’en fabriquait plus” (Bordes 1968: 144–45). This quotation illuminates several misunderstandings, from one regarding the relative rapidity of exchange of information and genetic material (even the French wear Levi’s™), to a strange idea of tribes, and the now discredited concept of marriage by capture.

17. The study of such signing is the subject of the field of semiotics. Specialists have identified many kinds of signs, nonverbal as well as linguistic, that are used to differentiate social groups. See, for example, Barthes 1970 (esp. 25–30). Bogatyrev (1971) is an exemplary study of clothing that attends to its signing functions.

18. This idea is still not completely dead. Derek Roe (1970: 28) observes that “on occasion there may even be evidence for correlation between a culture and some specific human physical type.”

19. Exemplary work along these lines has in fact already begun. See such publications as Arquenas, Sautuola, the Boletín Cántabro de Espeleología, the Revista Arqueológica, and other publications produced by local speleological and/or archeological groups, such as the Colectivo para la Ampliación de Estudios de Arqueología Prehistórica and the Grupo de Espeleología e Investigaciones Subterráneas Carballo/Raba, for illustrations.

REFERENCES

Barthes, R. 1970. Elements of Semiology. Boston, Beacon Press.

Bogatyrev, P. 1971. The Functions of Folk Costume in Moravian Slovakia. The Hague, Mouton.

Bordes, F. 1968. Le Paléolithique dans le monde. Paris, Hachette.

Cabrera Valdes, V. 1984. El yacimiento de la Cueva del Castillo (Puente Viesgo, Santander). Bibliotheca Praehistorica Hispana 22. Madrid.

Clark, Grahame. 1957. Archaeology and Society. Cambridge, Harvard University Press.

Clegg, M. 2004. Modern Approaches to the Evolution of Speech and Language. General Anthropology 10(2): 1, 8–11.

Deming, W. E. 1966. Some Theory of Sampling. New York, Dover Publications.

de Sonneville–Bordes, D. 1981. L’Age de la Pierre. Series Que sais-je? Paris, Presses Universitaires de France.

Fortea, J., et al. 2003. La Cueva de el Sidrón (Borines, Pilona, Asturias). Estudios Geológicos 59: 159–79.

Freeman, L. G. 1964. Mousterian Developments in Cantabrian Spain. Ph.D. dissertation, Department of Anthropology, University of Chicago.

———. 1981. The Fat of the Land. In Omnivorous Primates, ed. R. Harding and G. Teleki, 104–65. New York, Columbia University Press.

———. 1992a. La “Transición” en Cantabria: La importancia de Cueva Morín y sus vecinos en el debate actual. In El orígen del hombre moderno en el suroeste de Europa, ed. M. D. Garralda, 171–93. Madrid, U.N.E.D.

———. 1992b. Mousterian Facies in Space: New Data from Morín Level 16. In The Middle Paleolithic: Adaptation, Behavior and Variability, ed. H. Dibble and P. Mellars, 113–25. University Museum Monograph 78. Philadelphia, University of Pennsylvania.

———. 1994. Kaleidoscope or Tarnished Mirror? Thirty Years of Mousterian Investigations in Cantabria. In Homenaje al Dr. Joaquín González Echegaray, ed. J. A. Lasheras, 37–54. Museo y Centro de Investigacion de Altamira, Monografias 17.

Freeman, L. G., and J. González Echegaray. 1983. Tally-Marked Bone from Mousterian Levels at Cueva Morín (Santander, Spain). In Homenaje al Prof. Martín Almagro Basch 1: 143–47. Madrid, Ministerio de Cultura.

Freeman, L. G., J. González Echegaray, J. Pokines, H. Stettler, and M. Krupa. 1998. Tamisage ultra fin et récupération de l’outillage, observations réalisées à el Juyo (Espagne cantabrique). L’Anthropologie 102: 35–44.

González Echegaray, J. 1984. Reflexiones sobre el momento actual en la investigación del Paleolítico superior cantábrico. In Scripta Praehistorica Francisco Jorda Oblata, ed. J. Fortea, 259–69. Salamanticae.

González Echegaray, J., and L. G. Freeman. 1998. Le Paléolithique inférieur et moyen en Espagne. Grenoble, Jérôme Millon.

González Echegaray, J., et al. 1980. El yacimiento de la cueva de “el Pendo” (excavaciones 1953–57). Bibliotheca Praehistorica Hispana 17. Madrid, CSIC.

Goodman, L. 1954. Kolmogorov-Smirnov Tests for Psychological Research. Psychological Bulletin 51(2): 160–68.

Harman, H. 1967. Modern Factor Analysis. Chicago, University of Chicago Press.

Henshilwood, C., F. d’Errico, M. Vanhaeren, K. van Niekirk, and Z. Jacobs. 2004. Middle Stone Age Shell Beads from South Africa. Science 304: 404.

Higgs, E. S. 1975. Paleoeconomy. Cambridge, Cambridge University Press.

Leroi-Gourhan, A. 1957. Prehistoric Man. New York, Philosophical Library.

Montes Barquín, R. E., J. Sanguino, et al. 2001. La Cueva del Pendo: Actuaciones arqueológicas 1994–2000. Santander, Ayuntamiento de Camargo, Gobierno de Cantabria, and Parlamento de Cantabria.

Roe, D. 1970. Prehistory. Berkeley, University of California Press.

Siegel, Sidney. 1956. Non-Parametric Statistics for the Behavioral Sciences. New York, McGraw-Hill.

Von Koenigswald, G.H.R., ed. 1958. Hundert Jahre Neanderthaler. Utrecht, Kemink en Zoon.

Wheeler, Sir Mortimer. 1950. What Matters in Archaeology? Antiquity 24: 122–30.

* Published in Spanish in 2005 as La investigación del Paleolítico Medio en la región cantábrica, in Actas de la Reunión Científica Neandertales cantábricos, estado de la cuestión, ed. R. Montes Barquín and J. A. Lasheras Corruchaga (Santillana del Mar, Museo de Altamira), 21–38.