Recommendation of: Bridging Material Culture Networks and Social Networks in Archaeology. Round#2

This paper proceeds from a very important question, one that remains under-theorized in network approaches in archaeology: to what extent do patterns in material culture (especially material culture similarity) reflect underlying social interactions?The authors situate this question within a growing body of network science scholarship in which similarity based material culture networks are used as proxies for past social connectivity.Although relational thinking has increasingly been mobilized to move beyond rigid culture-historical categories, the connection between artifact similarity and social interaction is often treated as self-evident rather than explicitly examined.In this paper, the authors aim to evaluate how similarity networks might approximate interaction networks, and in so doing, clarify where and why this similarity is less evident.I was particularly interested in the central question of whether similarity in material culture can reflect underlying social networks, and if so, under what conditions.This is a foundational issue for archaeological network research.If similarity in material culture does not directly reflect past social interaction but instead results from differences in how material culture was made and used, then we need clearer theoretical and methodological criteria for assessing when similarity can be interpreted as evidence of social connectivity and when it cannot (see Gheorghiade et al. forthcoming).In this paper, Bischoff and Padilla-Iglesias rely on an agent-based model (ABM) (ArchMatNet) originally designed to replicate simple social interaction among hunter-gatherers, to explore how material culture networks and social interaction networks correlate.Their analysis draws on more than 30,000 simulations to produce two different types of networks: social interaction networks based on visiting, trading, and learning; and material culture networks where similarity is constructed from ceramic and point assemblages.Similarity 1 in the latter is calculated using a weighted Jaccard index.The resulting networks are compared using three additional methods: graph correlation, quadratic assignment procedure (QAP), and Exponential Random Graph Models (ERGMs).This multi-method strategy moves the discussion beyond abstract debate toward experimentally grounded inference.By including ERGMs the authors can also assess whether the resulting networks share similar generative structures, rather than only measuring correlation.The results show that material culture networks more closely resemble interaction networks when variation in assemblages is sufficiently high, but that this correspondence weakens with lower variability, extended time depth, or with uneven transmission patterns.Taken together, the results in this paper point to three main conclusions.First, material culture networks can provide good representations of interaction networks when sufficient variation exists in the assemblages, particularly when camps begin with distinct traits (the uniqueTraits parameter).Second, material culture networks were inherently "fuzzier" than interaction networks.For example, the relationship between them was not as strong as those between different types of social networks.Third, different types of material culture (e.g., pots and points) could produce different results and network structures.This led the authors to suggest that combining multiple categories of material culture might provide a more accurate representation of interaction patterns.I particularly appreciated the authors' experimental approach to isolating the conditions under which material culture can function as a proxy for social interaction.By combining the ABM with multiple other network comparison methods, this paper tests, in a systematic way, when material similarity networks correspond to interaction patterns.The peer-review process further improved the clarity of the manuscript.Some of the suggestion submitted by reviewers included: clarifications on the QAP procedure, further discussion of the application of the ERGM and its implications for analysis, and further reflection on how certain limitations might stem from the generalized structure of the model, or from structural assumptions.The reviewer suggestions and resulting changes provided a clearer discussion of the methods and their limitations.The resulting paper provides a careful and well documented contribution for scholars working with similarity-based network approaches in archaeology.