Scientific artifact classification developed initially as an adjunct to culture classification. The early prehistorians, like Thomsen and de Mortillet, were not really interested in tools and pots as evidence of activities or of technologies or of thought-patterns, but only as identifiers of chronological horizons. The objective in artifact classification was to identify those types that could be associated specifically with the Stone Age, the Bronze Age, or the Iron Age, and could help in the allocation of sites to one or another of those periods. Individual types were of course defined on the basis of formal characteristics, but they were then grouped together on the basis of chronological contexts rather than on the internal evidence of form or function.
So long as they were undertaken for purposes of culture definition and of site attribution, artifact classifications were dominated by the concept of the ‘index fossil’. That is, primary attention was given to those artifact types that were found to be diagnostic of specific periods. On the other hand, artifact types that were considered non-diagnostic were often ignored. Thus, for example, Mediterranean archaeologists gave names and definition to a few pottery types, like Minyan ware, that had a high degree of historical significance, while a great many other and more ubiquitous pottery wares went unnamed. In the same way, North American prehistorians developed comprehensive and highly detailed classifications of projectile points, while nothing comparable was done for more generalized stone tools like scrapers and choppers.
A long step forward was taken when Montelius introduced the ‘typological method’, in which cultures and chronological horizons were defined on the basis of total assemblages rather than of a few diagnostic types. As a result, artifact classifications became more comprehensive. There was not, however, any attempt to introduce formal systematics in the classifications.
However, not all nineteenth-century artifact classifications were instrumentalist. A great many museum collections had been donated by amateurs, with little or no accompanying provenience information. As a result, museums often developed purely formalistic artifact classifications, based on criteria of size, shape, and color alone. Under this procedure, groups of similar-appearing artifacts were displayed together, without any consideration for their times or places of origin. This tradition persisted in the display of many kinds of museum materials until well into the twentieth century.
General A. H. Pitt Rivers, often regarded as the father of scientific archaeology in Britain, went beyond other museum curators in combining classifica-tory formalism with evolutionism. So committed was he to a unilinear evolutionary perspective that he arranged all of the objects in his vast collection - both ethnographic and archaeological - into what he believed to be developmental sequences, without any reference to their place of origin. Pitt Rivers may thus have been the first prehistorian to employ purely logical seriation, though the technique had certainly been used earlier by art historians. Combined with more formal typology, it would later be used by scholars such as Petrie and Kroeber to develop culture sequences in areas where direct chronological evidence was lacking. That is, types were initially defined on formal grounds, and then were arranged into what appeared logically to be developmental sequences.
The single most important revolution in artifact classification came about as a result of the introduction of frequency seriation, in the early decades of the twentieth century. This procedure grew out of a recognition that cultures generally change gradually rather than cladistically. At any given time, some types of artifacts are always coming into use while others are going out. After the passage of a generation or two, all of the same artifact types may still be in use as at the beginning, but some will have increased in frequency while others will have decreased. Accordingly, cultures and their various developmental phases need not always be recognized by diagnostic artifact types or assemblages; they can also be recognized by diagnostic frequencies of artifacts, even when there are no individually diagnostic types.
Frequency seriation introduced for the first time the procedure of quantification in artifact classification, and this involved two important methodological correlates. First of all, if types were to be treated quantitatively, classifications had to be fully comprehensive; there had to be a type category for every sherd or point. Moreover, the categories had to be mutually exclusive. In other words, the classification had to be a true typology. Second, the frequency seriation approach had the effect of partially decoupling artifact classification from culture classification. That is, types might not necessarily be diagnostic of any one period or even of any one culture, yet their relative frequency could help in the definition of periods and cultures.
The frequency seriation method was originally developed in the study of pottery distributions, and it is still much more often applied to pottery than to other materials. In the 1940s and 1950s, however, Francois Bordes introduced a similar approach in the study of lithic materials from French Palaeolithic sites. The method requires a far more precise and comprehensive typology of stone tool types than had previously been attempted. The Bordes method has proved to be generally effective on its home ground, and attempts have been made to employ it also in many other parts of Europe, the Near East, and Africa. Field workers tried originally to employ the same actual tool types that had been designated by Bordes in France, assuming them to be universal in distribution, but this has not proved to be successful. The general conclusion now is that the frequency seriation method itself works well with Palaeolithic stone tools, but the actual types must be separately classified for each area.
The partial decoupling of artifact classification from culture classification had a kind of liberating effect on the study of artifacts. No longer treated simply as culture-markers, they could be studied more nearly as autonomous data, giving evidence of technologies, of activity patterns, and even of thought patterns that were not necessarily culture-specific.
Moreover, many North American prehistorians by 1940 had come to feel that the older, instrumental artifact classifications had now done their job, insofar as a series of cultures and culture sequences had now been devised for nearly the whole continent. It was time therefore to develop new classifications for new purposes. The result was a series of self-proclaimed ‘revolutions’ in American archaeology, each of which had an impact, at least theoretically, on the processes of artifact classification.
Beginning around 1940, it was argued that archaeologists had been splitting hairs over minute differences of pottery temper or of length-width ratio in arrowheads, because these differences proved to be useful for the chronological ordering of sites, when they might have been accidental or meaningless to the actual makers of the artifacts. There was, as a result, a certain turning away from the strict formalism that had characterized the previous generation of classifiers. The basic idea now was that artifact types should be essentialist rather than instrumentalist; they should represent ‘mental templates’ in the minds of the makers, and should ignore variations and variables that seemed to be unintentional. However, there was never complete agreement as to how this was to be determined. One school, championed by Alex Krieger and Irving Rouse, argued that the most sharply demarcated types could be safely assumed to represent the intent of the makers. Another school, best represented by Walter Taylor, apparently believed that prehistoric peoples thought more in functional than in formal terms, and therefore recommended an approach to classification based more on the function than on the form of objects.
The supposed functionalist revolution had not proceeded very far when it was overtaken, in the 1960s, by the self-proclaimed ‘scientific revolution’ ushered in by new archaeology. The new archaeologists shared with their predecessors the belief that artifact types should in some sense be ‘real’, rather than mere heuristic constructs of the archaeologist, and there was, at least for a time, a continued assumption that ‘reality’ must reflect the intent of the makers. However, ‘reality’ was now to be determined by strictly scientific and empirical procedures, without recourse to interpretation or context. The touchstone of reality was replicability. Any type proposed by one archaeologist should be capable of confirmation by others, using properly scientific measures. Types, like other scientific propositions, should be regarded as propositions to be tested. As a result, there was, necessarily, a renewed emphasis on formal systematics, exemplified especially in the work of David Clarke, Robert Dunnell (1971), and Dwight Read.
The renewed emphasis on formality and rigor led to an important methodological innovation: the use of statistics, and the definition of types by attribute clustering rather than by object clustering. The archaeologist’s traditional procedure of partitioning a body of collected material into types through visual inspection and the observation of similarities was now thought to be too intuitive. Instead, a list was to be made of all the attributes of size, shape, color, and so on that were exhibited by all the objects in a collection individually. Types were then to be defined by clusters of attributes that occurred together with a frequency greater than chance, as revealed through the use of accepted statistical measures.
In the beginning, the major difficulty in statistically based classification lay in the enormous number of separate calculations that had to be made in order to determine the randomness or non-randomness of a nearly infinite number of attribute combinations. That difficulty was overcome in the 1970s, however, with the introduction of computers. There followed a decade of experimentation with computerized systems of classification, which, it was hoped, would finally achieve the elusive goal of automatic classification. The search for automatic and absolutely ‘natural’ classification before long became an end in itself, rather than a means to an end, and as such it considerably outlived the new archaeology paradigm that had given it birth.
It was eventually realized, however, that the goal of automatic classification was not practically attainable. The coding of more and more variables simply resulted in the generation of more and more types, far more than were useful for any practical purpose. Indeed, the ultimate logical outcome of computerized classification was a series of classifications in which every object constituted a separate type. In the end there was a general, though not universal, acknowledgement that types that were not produced for any specific purpose were also not useful for any specific purpose.
The successive changes of direction and of interest that have taken place since 1940 have given rise far more to theoretical and programmatic literature than to actual, in-use classifications. The concern of most authors has been to propose new methods of classification, rather than to develop actual classifications based on those methods. Insofar as the new methods have been put to practical use, it has been almost entirely at the level of individual assemblages.
There has been, as a result, a proliferation of ad hoc classifications, each archaeologist developing his or her own system as a way of dealing with, and publishing, his or her own finds. For the most part, these have not proved to be capable of generative use, and have not passed into what might be called the public domain. The archaeological type concepts and typological systems that remain in general, region-wide use are still very largely those that were developed, chiefly for the instrumentalist purpose of culture classification, more than half a century ago.
World History
