Methods Matter: Assessing the Seriation Multiverse in the U.S. Southwest

Abstract As a bedrock of archaeological research, seriation and pottery dating methods have proliferated and become increasingly statistically sophisticated. Yet, rarely do archaeologists critically evaluate the effectiveness of our methods in relation to each other to ensure that results are meaningful. This study compares four techniques—Pecos Classification, Correspondence Analysis, Uniform Probability Density Analysis, and a Village Ecodynamics Project (VEP) Bayesian Statistical Model—using the same site data from Park Mesa on the Mesa Verde Cuesta in the U.S. Southwest to assess the comparability, strengths, sensitivities, and weaknesses of these methods. Importantly, the occupation histories produced by each of these methods differed in ways that had significant implications for their interpretive narratives underscoring the critical need for us to understand how the methods we use bias outcomes. In this case, the methods’ outputs varied due to a combination of factors including changes in what data were used as inputs, how phases were defined, how pottery types were weighted, and how multicomponent sites were handled. Methods varied most significantly during well-documented periods of social change, generating a multiverse of possible histories of what these Ancestral Pueblo settlements and communities were like in the past. This study serves as a caution for archaeologists to consider how the methods they select bias results and to select their methods with care.