A statistical method for dealing with isolated teeth: ageing pig teeth from Hagoshrim, Israel

Abstract

Reconstructing kill-off patterns from faunal remains is an essential step in deciphering the association between humans and animals in ancient societies. Specifically, demographic patterns and trends are considered key parameters in studying the process of domestication as revealed in the archaeozoological record (Davis 1987; Clutton-Brock 1999; Reitz & Wing 1999). This is especially true in the case of pig domestication, where both archaeozoological and ethnoarchaeological evidence point to a prolonged process that has included, and still includes today, a wide diversity of associations between pigs and humans (Grigson 1982; Griffin 1998; Hongo & Meadow 1998; Lobban 1998; Redding & Rosenberg 1998; Rosman & Rubel 1989). Although several methods exist for the recording of tooth eruption and wear for the purpose of ageing teeth, these methods rarely include statistical means by which to compare the resulting kill-off patterns. A further complication, seldom addressed in the literature, is the use of isolated teeth. In this chapter I suggest a simple procedure to overcome these problems. This procedure is illustrated using data recorded for pig remains from Hagoshrim, a Neolithic site from Israel, and discuss its implications for the study of pig domestication in the southern Levant. However, this method is applicable to any mammalian assemblages where similar ageing methods are used. Analysing kill-off patterns based on tooth eruption and wear usually involves the following stages (e.g. Hesse 1986, 2002; Rolett & Chiu 1994; Hongo & Meadow 1998; Ervynck et al. 2001; Horwitz 2001; Davis in press): (1) collecting the raw data by recording eruption and wear stages, most commonly (for pigs) based on the scheme provided by Grant (1982); (2) correlating the archaeozoological data with absolute ages from recent populations, based (for pigs) primarily on the works of Matschke (1967), Silver (1969), and Bull & Payne (1982); and (3) generating histograms or survivorship curves that summarize the raw data. However, Grant’s method and other common methods for generating survivorship curves produce information that does not lend itself easily to statistical testing. Furthermore, the method outlined by Grant (1982) is designed specifically for the analysis of half-mandibles, which are scarce in most archaeozoological assemblages.