Network analysis of archaeological data: a systematic approach

Network theory can be employed in two ways in archaeology: it can be used to analyse archaeological data, or it can be used to model a historical process for the purpose of simulating the data. This paper focuses on the first approach. In such analyses, similar archaeological contexts are often connected to form a similarity network. Similarity is treated as a proxy for social or causal relationships. Most often, similarity is defined by the presence of the same kind of find in two contexts. However, to detect relationships effectively, we have to allow any kind of similarity relation to be a criterion for connection, in which different kinds of attributes that characterise the contexts may be mixed. We discuss how such general similarity networks can be used to disclose relational patterns hidden in archaeological data. Statistical tests are necessary to distinguish significant patterns from random patterns. We argue that random permutation tests are well suited for this task, and we introduce appropriate tests of this kind. The methods outlined are compared to other kinds of quantitative data analysis, such as correspondence analysis. We discuss which approach is more suitable for which kind of data. The choice of approach also depends on the questions addressed to the archaeological material.