Across Atlantic Ice

by Dennis J. Stanford

  If Clovis did not appear fully developed, where and in what technologies should we be looking for an antecedent? Based on what we know about flaked stone traditions at the end of the Last Glacial Maximum, we find three possibilities: Beringia, inset blade / thick biface; Eurasia, blade; and southwestern Europe, thinned biface/blade. In the present chapter we begin to evaluate these possibilities with quantitative comparisons of pre-Clovis, Clovis, Beringian, and European Upper Palaeolithic flaked stone technologies. Then, in chapter 7 we move to more qualitative comparisons of other aspects of these cultures: artifact form and function, cultural innovation, subsistence patterns, environmental settings, art, and specific human behaviors.

  LIMITS AND METHODS OF COMPARISON

  As we have pursued this study, the limitations of the available data have become abundantly clear. For the most part they follow from the small number of investigated sites and their poor state of preservation. With few exceptions only stone and the most durable of organic artifacts, such as bone tools, have survived, and then in only a few sites. This is true for not only whole time periods but whole regions. There is also great variation in the manner in which sites have been investigated and reported and their artifact assemblages analyzed. In many excavations before World War II, only the artifacts that were considered important or interesting were kept. Many “lesser” and broken pieces were discarded before being studied. It is difficult to determine from reports whether or not all artifacts were retained or what the selection criteria were. More than 85 percent of all French Solutrean sites were excavated before 1940, and few have detailed reports. The spoil heaps from these excavations are littered with flakes, bone scraps, and broken tools. Fortunately, by the late 1950s the work of François Bordes, Hallam Movius, and André Leroi-Gourhan had transformed the methodology of Paleolithic archaeology in France, and the more recent Solutrean excavations, such as those by Lawrence Straus and his colleagues in northern Spain and Thierry Aubry and his colleagues in France, have been painstakingly documented, with all artifacts and many environmental samples retained.4

  The sizes of samples are highly variable because of differential preservation, recovery method, and the size of the area excavated. Unfortunately, there seems to be an inverse relationship between the care of excavation and the size of the sample. For example, Straus and crew worked meticulously at La Riera, but by the time they got down to the lower Solutrean deposits they were able to excavate only five square meters. However, today’s archaeology has so many analytical methods available that large samples can be an impediment to study. We want to know everything we can about what we recover, so we tend to recover less and analyze it more thoroughly. This approach has led to advances in understanding technology, ancient environments, and intra-site organization, but it has also meant that fewer sites and smaller areas have been investigated. This raises the thorny issue of how to realistically compare site assemblages consisting of tens of objects—including those collected in previous decades, when only portions of the artifacts were taken—with those having tens of thousands.

  Another difficulty is our having to rely on secondary publications in situations where either the primary data was not published or we didn’t have access to it. Some sites have been summarized in several books, with major discrepancies among the reports. These usually take the form of different numbers and types of artifacts being reported. Which do we use as our reference? Generally speaking we have relied on the most recent reports. Although these discrepancies do not alter the overall interpretations of the major archaeological culture reconstructions, they inhibit detailed inter-site comparisons. But we contend that even with all of these limitations, archaeologists must keep going back to the available materials and information. We may be able to apply new analysis techniques and tease more information out of the existing collections. This exercise also helps to identify gaps in our data.

  Considering the types of archaeological data we have or can acquire through additional analyses, how do we make our comparisons? Does it make sense to compare an assemblage of a score of artifacts from a special-function site with a sample of thousands from a long-term habitation? The obvious answer is that it does not. Why not group the sites into classes and compare within those categories? The answer is once again sample size. When it is possible to identify sites with a similar function, there are often only a few in each category, and these frequently come from huge geographic areas and are of different ages.

  With all of these problems, should we just give up and admit the task is impossible or fall back on our subjective feel for similarities? Obviously we should do neither. It is possible to find methods that allow comparisons to be made from which we can draw conclusions, albeit with many qualifiers. This is what we have chosen to do.

  To achieve a more robust analytical comparison, we have combined assessments of typological and technological traits, because these are most likely to represent cultural norms across sites that may not include the whole range of activities of an archaeological culture. For example, it is easy enough to say that all the archaeological assemblages we are considering as possibly related to Clovis have flaked stone tools. Everybody agrees, though, that this is not enough similarity upon which to base an interpretation of historical relationship. However, archaeologists commonly compare artifact inventories in conjunction with time and space relationships to identify historical connections, such as Denali deriving from Dyuktai.

  The next requirement for establishing the historical connections among sites or assemblages is identifying their relationships in time. For example, it may be possible to make a case for a historical connection between archaeological cultures in Siberia and Alaska, but if the Alaskan material is older than the Siberian, the obvious interpretation is that the culture originated in Alaska and spread into Siberia rather than the other way around. Many sites and assemblages have not been dated and are chronologically placed by their geological context or typological similarity to dated assemblages, some of which are themselves indirectly dated. It is not our purpose here to reevaluate the dating of all these sites and assemblages. The dates we use are the most current we could find in the published literature. We do not, however, include some that may be chronologically mixed or whose dating depends solely on possible similarities to other poorly dated assemblages. (See the appendix for table A.1, which shows the assemblages we use.) We also group dates into large categories, primarily indicating whether they are earlier than, contemporaneous with, or later than 13,000–14,000 years ago. Although our conclusions about historical relationships among archaeological cultures are based on what we know about their chronological nearness or distance, this is not our overriding criterion.

  Physical closeness among sites or assemblages is commonly used by many researchers to make interpretations about possible historical connections, although this is also based on our individual assumptions about what physical connections were possible in the past. It is a function not just of distance but also of what the perceived difficulties of travel might have been. The land connection—not the distance—between Alaska and Siberia is the main argument for the current dominant paradigm of Beringia as the origin of people in the New World. Of course physical distance is an important consideration, but we contend that the best criterion for comparison is typological and technological similarity combined with dating, which informs us about the probability of physical connections. We think that New World origin issues have frequently been approached backward: an evaluation of likely routes of movement led to a theory, which archaeological information was then gathered to support.

  Tool typology has been used for decades to define and compare archaeological cultures. A standard approach of many archaeologists is to determine tool type proportions within site assemblages and then construct cumulative graphs. They then use correspondences and differences to evaluate relationships among the assemblages. Wide divergences are explained as the product of different archaeological cultures or indica
tions of different site functions or both.

  In this study we expand these comparisons by adding technological attributes. Our basic assumption is that there are different ways to make tools and different tools to serve the same purpose and that choices in different technological traditions distinguish archaeological cultures. In addition, we accept that early humans incorporated abstract concepts into the forms and designs of the tools they made. Some designs and even the process of manufacture itself may have had symbolic meaning or even ritual function. The less a particular tool attribute can be shown to be necessary to its physical use, the more likely this attribute was culturally rather than functionally determined. An example of this is corner versus side notching of projectile points: although both served the same function and were fit to purpose, they were produced in these different forms.

  FLAKED STONE TECHNOLOGIES: DYNAMIC SYSTEMS ANALYSIS

  Archaeological cultures with a high degree of similarity in the way their artifacts were made, as well as in the abstract and symbolic aspects of their artifacts’ forms and other characteristics, are likely to have a common ancestor. James Adovasio has shown us how this fact can be graphically applied to textile analysis and interpretation, and the same is true for flaked stone.5 It is possible to devise a hierarchical chart with the simplest attributes at the top and the more complex ones toward the bottom. Symbolic aspects of the artifact (e.g., design styles) add even more specificity. Correspondences at the top may be fortuitous, but as they continue down the chart the more likely they are to be historically related.

  Since raw material type and initial fracturing are universal features of stone flaking, their similarities in different assemblages do not inform us about possible historical relationships except at the dawn of stone tool making. How materials were selected and manipulated, however, does have the potential to help us identify more recent historical relatedness. This approach is most informative when applied to complex manufacturing technologies, such as bifacial projectile point or blade and microblade production.

  We have designed a simple dynamical systems analysis (DSA) diagram that illustrates where two technologies are the same and where they diverge. Flaked stone tool production requires a series of decisions among a number of possibilities. Each action modifies the object, and the following decisions have to take the former results into account. Flaked stone technology is a sequence of causes and effects. Since there are almost always viable options, the choices of action represent the technological framework in which the knapper operates.

  FIGURE 6.1.

  Dynamic systems analysis chart of Beringian Sluiceway biface, Clovis point, and French Solutrean laurel leaf manufacturing sequences.

  Since each manufacturing process begins with the simplest choices, correspondences between different technologies at these steps do not inform us about possible historical relationships. But as flaking continues, correspondence between two technologies is increasingly likely to be the result of a common ancestral technology. Along with sequencing, it is equally important to see where and how the processes diverge. The greater the number of divergences, the earlier they begin to occur in the sequence, and the more divergences there are in a sequence, the more likely the technologies are unrelated.

  To illustrate this method we compared the manufacturing sequences of French Solutrean laurel leafs, Clovis points, and Beringian Sluiceway bifaces (one of the earliest kinds of Alaskan bifacial points). We examined identifiable flaking actions in the generalized production sequence from the beginning of the process to the finished piece.6 The resulting DSA diagrams show distinct differences and similarities (figure 6.1).7 The Solutrean and Clovis sequences are identical nearly to the end and diverge in only one attribute (basal treatment), whereas the Beringian sequence deviates early (with proportional shaping versus overshot thinning) from both the Solutrean and Clovis, maintains the divergence through several steps, corresponds for one trait (bifacial pressure flaking), diverges again, and finally corresponds for the last trait (lower margin grinding). These diagrams indicate that there is a much greater likelihood of a relationship between Clovis point and Solutrean laurel leaf technologies than between Sluiceway and laurel leaf or Sluiceway and Clovis technologies.

  The similarities between Solutrean laurel leaf and Clovis point manufacture are remarkable, from the initial selection of raw material, which displays a preference for exotic stones, through the final edge treatment. Even the details of flaking are virtually identical (figure 6.2). Both technologies incorporated overshot flaking as the main method of biface thinning, especially during the early and middle stages (table 6.1). They also used the overshot technique to remove square edges. The spacing of the thinning flakes was wide, so only two to four needed to be removed from each face to produce a flat biface. In both systems, thinning flakes were frequently removed from the same face by alternating between edges. Even the preparation of flake platforms was the same. Both used isolated, projected, released ground platforms that were designed to be straight rather than convex.8 Both even had platform grinding that extended from the area of contact on the flake platform to the adjacent flake removal surface. In North America, flakes with these platform attributes are diagnostic of Clovis, and this is true for Solutrean in Europe as well. Limited use of intentional heat treatment of the stone during the flaking process has also been proposed for both technologies.9

  FIGURE 6.2.

  Technological reduction comparison of middle phase manufacturing of bifaces showing overshot flake scars: (a) Clovis; (b) Solutrean.

  Finishing techniques were also virtually identical and included pressure thinning and shaping. The only real difference between the two technologies is that Solutrean laurel leafs exhibit a higher proportion of diving flake thinning than Clovis in the final phase of manufacture, whereas Clovis points were thinned from the base throughout the production process, ending with the distinctive basal flaking known as fluting. Both laurel leafs and Clovis points were finished with edge grinding from the greatest width to the base. Considering the time and space differences, this level of correspondence between technologies is amazing.

  TABLE 6.1 Occurrence of Overshot Flaking on Bifaces

  CLUSTER ANALYSIS

  It is useful to determine possible historical relationships between individual artifact types, but how do we do the same with whole assemblages? Cluster analysis is a simple yet informative technique for finding natural groupings of objects on the basis of similarities and dissimilarities in their observed variables. No knowledge of group membership or possible number of natural clusters need be predetermined. Cluster analysis of flaked stone assemblages is less than ideal when applied to small collections or those with few attributes or characteristics. We fully recognize these limitations. Nevertheless, it’s a more objective estimation than “they look alike to us.”

  As with any statistical method, the selection of which attributes to use influences cluster analysis. Because we think manufacturing technology is highly variable and most likely to reflect cultural choice, we have emphasized this aspect of the collections.

  We tested out the cluster analysis method on North American fluted point assemblages, and the results gave us confidence that this is a viable approach to discovering relationships between these traditions and the nearly contemporaneous traditions in Beringia and Western Europe (see the appendix for the first, practice test). However, the large range of time periods, great geographic distance, and small number of sites in some areas required different criteria than were used for the fluted point sites.

  Rather than include tool counts converted into proportions, we worked exclusively with the presence and absence of tool types. We also grouped sites and assemblages into major archaeological complexes. For the most part, we considered a type present if it had at least one unambiguous occurrence. For example, if there was a single good graver within the Late Dyuktai assemblages, we marked this type as present. It could certainly be argued that this was overly
generous, but the sample sizes are so small that we feel one occurrence is significant.

  Although the shaft wrench found at the Murray Springs Clovis site is unique and there is no direct evidence that this type of tool was widely used, almost everybody who has written about Clovis assemblages considers it a Clovis tool type. An ivory flaking tool called a billet, found at Blackwater Draw, is another single occurrence, but in this case we can be sure that it was a common tool type, because thousands of identifiably biface thinning flakes in Clovis sites were made with just such a tool.

  As with the fluted point assemblages, for the most part we used the identifications of tool types as they have been published. The only area where we consistently deviated from this rule was in what we labeled bifaces, which were originally called preforms, knives, or just bifaces; however, we did not include bifacial microblade precores in this category. Where the illustrations or descriptions allowed, we made some reclassifications.

  Cluster analysis allows certain attributes to be assigned more weight than others, so it would be possible to give more emphasis to common tool types or primary technologies. We have not pursued this, because we feel that the results we achieved without weighting different categories were sufficient and that emphasizing specific categories would probably have produced tighter clusters rather than different ones.

  We did two separate analyses, one with stone tool type assemblages and another with technological traits. Appendix table A.2 lists the assemblages and tool type presence and absence assessments we used in the first analysis. Some of these assemblages contain bone, ivory, and antler tools, but their absence from many, most commonly because of poor preservation, influenced us not to include them.