I study the evolution of human culture, within the framework of Darwinian evolutionary theory, and using mathematical models as tools. I am trained as an anthropologist, and primarily as an archaeologist, so I am mostly interested in major cultural differences that persist and differentiate over long spans of time. In other words, my specialty is understanding how micro-scale, individual level processes of social learning, strategic interaction, and the social structure of populations translate into the types of macroscopic, persistent differences that we talk about as “cultures” or “societies.”
My current research aims to link formal models of cultural transmission (social learning) with standard archaeological methods and data types, in order to better infer evolutionary processes from long-term archaeological data. Specifically, I aim to understand whether and how we can differentiate conformist, prestige bias, and other models of social learning in the evolutionary record of human history, using seriation, spatial analytic, and phylogenetic methods.
Much of the existing literature on cultural transmission models focus on well-mixed or mean-field approximations. In part, this reflects the need for simplicity if we wish models which are fully solvable, which is always a desirable goal. But it also reflects the fact that many models have been developed by anthropologists who study cultural transmission in living populations and through experimental studies within psychology.
I argue that further progress in formal modeling of cultural transmission will require dropping the well-mixed approximation, and understanding CT as an inherently spatio-temporal process. This is important for two reasons. First, we know that evolutionary processes can display different, and richer, dynamics when space or interaction networks are explicitly modeled, given the immense literature on spatial evolutionary game theory. Thus, we will not fully understand the behavior of cultural transmission theories until we model culture as a spatiotemporal process. Second, when we look at historical and archaeological data, we face data of reduced dimensionality (compared to examining living populations), and we need models which take advantage of the fact that we only have spatiotemporal distributions to work with.
To this end, I am working on transmission models which represent both geometric space, and social contact networks, as well as different models of social learning bias. Such models are most easily analyzed using the framework provided by statistical physics, with observables defined by spatial statistics. Specifically, I am attempting to connect transmission models to their quasistationary consequences for seriation and phylogenetic methods, by analyzing spatial/network observables within ensembles of networks representing communicating individuals, who follow specific rules for social interaction. This involves constructing formal models of standard archaeological methods such as artifact classification, data collection and sampling, and understanding their effects on the distribution of observables.
I am using several case studies in my research. The first is a large and detailed data set of ceramic artifact variation from the Mississippi River Valley. The main data set was originally collected by Phillips, Ford, and Griffin in 1951, and augmented by Carl Lipo for his dissertation research. I have previously used this data set for precursor studies, including my 1997 joint paper in the Journal of Anthropological Archaeology.
The second case study is a fairly complete inventory of formal variation present on the famous statues (moai) of Easter Island (Rapa Nui). These data were collected by the University of Hawaii-Manoa/California State University Long Beach expedition to Rapa Nui. Each statue has detailed photographs, geographic data, and a recent classification of its formal variation.
These cases were chosen because they differ in the methods required to quantify cultural change and difference. In the case of Rapa Nui moai, statues are a communal project, representing periodic snapshots of cultural information present within a local community. In connecting transmission models to moai variation, the key linkage will be to object-scale seriation, while in the case of Miss. Valley ceramic variation, the key linkage is to frequency seriation. The spatial statistics used will differ as well due to these differences, ranging from joint-counts to statistics applicable to continuous variables.