Academic journal article Southeastern Archaeology

Seriation: Classic Problems and Multivariate Applications

Academic journal article Southeastern Archaeology

Seriation: Classic Problems and Multivariate Applications

Article excerpt

This article examines current advocacy and use of relative frequency seriation, employing the classic "battleship" curve (unimodal distribution or popularity assumption) made popular by James A. Ford (Phillips, Ford, and Griffin 1951). This approach particularly is pronounced in recent writings of scholars advocating a Darwinian evolutionistic approach to archaeological data, but it is hardly limited to them. There are major difficulties inherent in relative frequency seriation as well as in the Brainerd-Robinson index. For the future, exploring the potential of modern multivariate techniques is strongly advocated. Fearing that Ford's (Phillips, Ford, and Griffin 1951) methods will not be banished completely by this article, extreme caution is urged on practitioners of this approach.

Many years ago I published a critique of then-popular seriation techniques that employed relative frequency percentages to construct "battleship" curves (e.g., James A. Ford in Phillips, Ford, and Griffin 1951; henceforth cited as PFG) and other uses of relative frequency percentages to order sites (e.g., Brainerd-Robinson technique) that purported to create chronological series (McNutt 1973). That critique did not apply to the various multivariate techniques then coming into fashion. All uses of the term "seriation" in this article refer only to classic uses of "battleship" curves and the unimodal distribution assumption, unless specified otherwise.

There are three major reasons for this article: (1) the use of battleship curves and Brainerd-Robinson techniques have persisted; (2) the use of battleship curves and Brainerd-Robinson techniques has been espoused in the voluminous literature of "evolutionary archaeology" (Fox 1998; O'Brien and Fox 1994; O'Brien and Lyman 1999, 2000; Teltser 1995, etc.); and (3) easily available multivariate techniques provide superior means to achieve the goals of seriation.

The primary goal of evolutionary archaeology is to build and explain sets of related cultural lineages (e.g., Lyman and O'Brien 1998:615). The latter have described seriation as their "preferred method" of ensuring heritable continuity in artifact lineages (O'Brien and Lyman 2000:352). In short, seriation is proposed as a basic method for evolutionary archaeology.

Modern use of seriation is hardly confined to the evolutionary archaeologists. A very casual and incomplete survey of recent theses and dissertations confirmed continuing use of relative percentages, battleship curves, and/or Brainerd-Robinson for basic analyses at a variety of institutions (e.g., Castano 2000; Crowell 1994; Kojo 1991), sometimes in conjunction with multivariate techniques (e.g., Maher 1996; Neff 1984) and sometimes simply for descriptive purposes (e.g., Yakubik 1990). Also, it must be recalled that many current texts on archaeology still present frequency seriation as a relative chronology technique.

Some definitions and observations are in order. Seriation simply means placing things in an ordered series. The order is derived on the basis of some shared attribute of the things involved-weight, cost, length, count (frequency), and so on. "Relative frequency seriation" refers to forming a series on the basis of relative frequencies, typically percentages. As used today and in the past several decades, relative frequency seriation refers to an analytical technique that attempts to order sets of artifact types, in hopes of producing a series of interest. The significance of the series depends on assumptions about the set of types utilized; typically they involve assumptions that would justify a chronological interpretation of the series (cf. Rowe 1961:326).

Lyman, Wolverton, and O'Brien (1998:243), in an admirable and overdue attempt to resolve the terminological confusion attendant to seriation and allied methods, make a useful distinction between "percentage seriation" as defined above and "percentage stratigraphy," which involves plotting relative frequencies of a class of artifact types (e. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.