Chapter 7 ©Copyright 2015 All other materials ©Copyright 2020 Mark Ernest Madsen Measuring Cultural Transmission at Archaeological Scales: How Can We Improve Empirical Sufficiency? Mark Ernest Madsen A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy University of Washington 2020 Reading Committee: James K. Feathers, co-Chair Benjamin Marwick, co-Chair Carl P. Lipo, Binghamton University Program Authorized to Offer Degree: Anthropology University of Washington Abstract Measuring Cultural Transmission at Archaeological Scales: How Can We Improve Empirical Sufficiency? Mark Ernest Madsen Co-Chairs of the Supervisory Committee: Research Associate Professor James K. Feathers Associate Professor Benjamin Marwick Anthropology Cultural transmission has long been a key organizing principle within anthropology, but the effort to for- malize cultural transmission models and fit them to archaeological data is more recent, stimulated by work by Robert Dunnell in the 1970’s. Since then, the use of cultural transmission modeling in archaeology has branched into several research programs: one macroevolutionary, employing phylogenetic methods; and one microevolutionary, employing models derived from population genetics. A third research program, focused on intermediate or ”mesoscopic” scales and seriation as a finer-grained counterpart to phylogenetic and cladis- tics, is being developed by Carl Lipo and the present author. This dissertation collects research papers by the author since 2012 which examine two questions. First, are equifinality issues encountered in the microevolutionary research program solvable or do they prevent us from employing individual-scale models? Second, to the extent that equifinality cannot be circumvented, can we construct better approaches at the mesosopic scale appropriate to coarse grained, time averaged data? Two papers examine the first question, using simulation modeling and statistical methods to test whether theoretical models can be distinguished even in principle. The first paper examines the effects of temporal aggregation, which is ubiquitous in the archaeological record, on our ability to distinguish between cultural transmission models, and finds significant issues in doing so with time averaged data. The second paper examines the effects of population heterogeneity in social learning modes, which is well documented from living human and animal populations. I find that heterogeneous mixtures of social learning rules can be identified statistically, but only with synchronic censusing of the population, and that time averaging and small samples render mixtures indistinguishable from pure unbiased copying. Turning to the second question, three papers continue my long-term research into reshaping the classical seriation method into a tool for tracing the structure of cultural transmision at regional scales. One short paper examines the combinatorial structure of the seriation problem when we admit multiple subsolutions. A second paper seeks to increase the size of possible seriations, which is necessary to incorporate significant spatial variation and yield a tool usable for investigating the history of cultural transmission in a region. We increase the potential size of seriation solutions by switching from unimodality to distance minimization as the ordering criterion, yielding “continuity” seriation as a distinct method. A third paper in this group then applies continuity seriation graphs as the observable variable, in a methodological study of how to construct models of how cultural transmission was structured at the regional scale. This paper introduces “interval temporal networks” as a way to formalize our hypotheses about regional interaction and transmission, and explores a statistical method for summarizing the topology of seriation graphs, to assess their fit to our regional interaction models. A final paper examines a different kind of mesoscale question: how do we begin to model not just the spatiotemporal structure of past cultural transmission, but its content as well. The chapter models the depen- dency structure of the knowledge required to construct complex artifact types, through the “prerequisites” needed for each step, and introduces a model where transmission of subsequent traits requires learning their prerequisites first. This simplified model of “structured” cultural traits is then used to explore the “learn- ing hypothesis” for behavioral modernity, by looking at the richness and depth of knowledge gained when transmission is mostly accomplished by simple imitation compared to learning via a teacher. The results are suggestive that the learning hypothesis can account for the increased richness of “behaviorally modern” hominids, and more importantly, points the way to more substantive and technologically informed cultural transmission models. Contents List of Figures v List of Tables xi Acknowledgements xiii 1 Introduction and Research Problem 1 1.1 Introduction .......................................... 1 1.2 Attempts to Assess Equifinality in the Microevolutionary Program .......... 8 1.3 Are There Structural Equifinalities We Cannot “Correct”? . 14 1.4 Seriation and the Mesoscopic Approach to Cultural Transmission Modeling . 19 1.5 Dependency Graphs and Incorporating Structured Information In Cultural Trans- mission Studies ......................................... 24 2 Neutral Cultural Transmission in Time Averaged Archaeological Assemblages 29 2.1 Introduction .......................................... 30 2.2 Conceptual Structure of Neutral Cultural Transmission . 32 2.3 Unbiased Transmission: The Wright-Fisher Infinite-Alleles Model . 37 2.3.1 Statistical Tests for Neutrality ........................... 39 2.3.2 Estimation of Innovation Rates .......................... 41 2.3.3 Diversity Measures .................................. 42 2.4 Methods ............................................. 44 2.4.1 Model Verification .................................. 44 2.4.2 Time-Averaging and Simulation Parameter Space . 46 2.5 Results .............................................. 48 2.5.1 Time Scales and Time averaging .......................... 50 2.5.2 Neutrality Testing .................................. 53 2.5.3 Theta Estimation and Innovation Rates ...................... 56 2.5.4 Diversity Measures .................................. 61 2.6 Discussion and Conclusions ................................. 63 2.7 Acknowledgements ...................................... 67 i Contents 3 Can We Identify Biased Cultural Transmission in the Archaeological Record? 69 3.1 Introduction .......................................... 69 3.2 Within Population Variation in Social Learning: A Cause of Structural Equifinality? 73 3.3 Methods ............................................. 76 3.3.1 Study Design ..................................... 76 3.3.2 Measuring Equifinality Through Classification Error . 76 3.3.3 Simulation Modeling of Cultural Transmission Mixtures . 81 3.3.4 Summary Statistic Selection ............................ 83 3.3.5 Data Collection Treatments ............................. 85 3.3.6 Classifier Selection and Training .......................... 88 3.3.7 Classification Error and Equifinality Assessment . 91 3.4 Results .............................................. 92 3.4.1 Classification Error and Equifinality Results ................... 92 3.4.2 Which Predictor Variables Help Discriminate Models? . 96 3.4.3 Time Averaging Makes Identification of Bias More Likely . 97 3.5 Discussion ........................................... 99 4 Combinatorial Structure of the Deterministic Seriation Method with Multiple Subset So- lutions 103 4.1 Single Seriation Combinatorics ............................... 104 4.2 Deterministic Seriation with Multiple Solution Groups . 105 4.3 Discussion ........................................... 109 5 Measuring Cultural Relatedness Using Multiple Seriation Ordering Algorithms 111 5.1 Introduction .......................................... 112 5.2 Seriation and the Frequency Principle . 115 5.2.1 Unimodality and Cultural Transmission Processes . 117 5.2.2 Continuity: An Alternative to Unimodality . 120 5.2.3 Statistical Seriation Methods ............................ 122 5.2.4 Exact Distance Minimization Ordering: “Continuity” Seriation . 124 5.3 Comparing Frequency and Continuity Seriation . 125 5.3.1 Examining a Solution Which Differs . 128 5.4 Discussion ........................................... 131 6 A Computational Method for Identifying Regional Interaction Patterns From Seriation Solutions 137 6.1 Introduction .......................................... 137 6.2 Documenting the Regional History of Cultural Transmission With Seriation Graphs 140 6.3 Representing Hypotheses About Regional Transmission History With Temporal Net- works .............................................. 144 6.3.1 Transmission Scenarios Studied . 150 6.4 Methods ............................................. 153 ii Contents 6.4.1 Study Design ..................................... 153 6.4.2 Quantifying The Structure of Seriation Solution Graphs . 156 6.4.3 Simulation of Cultural Transmission on Interval Temporal Networks . 157 6.4.4 Classifier Training and Accuracy Evaluation . 160 6.5 Results .............................................. 161 6.5.1 Equifinality Analysis of Transmission Scenarios with Simulated Data . 161 6.5.2 Analysis of Lower Mississippi River Valley Ceramic Data . 162 6.6 Discussion ..........................................