Correspondence analysis and principal components analysis as methods for integrating archaeological plant and animal remains

Abstract

Multivariate analyses, such as correspondence analysis (CA) and principal-components analysis (PCA), have gained prominence in archaeology over the past several decades, with applications to a variety of archaeological datasets, including subsistence data (Hollenbach 2005; VanDerwarker et al. 2007; Whitridge 2001, 2002; see also Hollenbach and Walker, this volume and Peres et al., this volume). These analytical techniques are especially useful in that they allow for the consideration of multiple cases (e.g., contexts, sites, periods) along with multiple variables, producing solutions that can map associations between the two. The implication of these techniques for integrating archaeological plant and animal datasets is the ability to consider how different plant and animal resources cluster together in relation to different contexts or time periods. Thus, we can imagine a single solution that differentiates between the diets of different periods (see below) or diets related to different social or spatial contexts (see Peres et al., this volume), just to name a few possibilities. This chapter, offers a method for the direct quantitative integration of-zooarchaeological and paleoethnobotanical remains. I demonstrate the utility of correspondence analysis and principal components analysis for different types of datasets (e.g., CA for nominal data, PCA for ratio/interval data). Using the site of La Joya as a case study from Formative Gulf Coastal Mexico, I argue that multivariate statistics are well-suited for exploring the covariance of large zooarchaeological and paleoethnobotanical datasets. Indeed, these techniques have the potential to crosscut subsistence specialties by enabling a broader consideration of foodways at different levels of-analysis; we can examine the degree of correlation between different categories of plant and animal remains that originate from multiple sites, multiple features at a single site, or multiple temporal periods. Thus, in situations in which we are interested in exploring multiple contexts alongside multiple food taxa, multivariate analysis is an ideal method for integrating-zooarchaeological and paleoethnobotanical data. © 2010 Springer-Verlag New York.