Assessing the relevance of a stratigraphy by network analysis of conjoin relations between archaeological objects
Session 5H
Résumé
Considered at a fundamental level of analysis, the time dimension in archaeology is stated from an examination of spatial relations between material objects which are distributed into multiple spatial units (e.g. stratigraphic units). We propose a graph-based method to address these relations. We argue that the structural properties of a conjoin/refitting network can be a proxy to measure the admixture of materials between stratigraphic units and to assess the sediment dynamics and the validity of a stratigraphy.
In the last two decades, one can observe a renewed interest in analysis of archaeological object refitting [Cziesla & al. 1990; Hofman, Enloe 1992; Schurmans & De Bie 2007]. Multiple ways to define relations between fragmented pieces have been proposed, as Bollong's [1994] six items typology. However, most of these methods rely on a count of the number of relations and do not take into consideration the structure of a set of refitting links [see Vila 1982]. Network analysis is relevant to do so and has not yet been applied to this purpose [Brughman 2012].
Firstly, we consider the formalization of relations between fragments of a same object as a graph and the related issues. It is argued that archaeology can take advantage of the insights taken from formal ontology [Casati, Varzi 1999]. Two types of relations are defined and implemented in a graph formalism. Secondly, we discuss the relevant network measurements for both intra-level and inter-level analysis. Details concerning two cases studies are then given. Initially, our method was developed to answer the problems we encountered in Liang Abu Cave, Indonesia. We then perform our method on a bigger dataset taken from the Neolithic Taï Cave, South-east of France. All the functions we run have been written in R and built upon the igraph [Csardi & Nepusz 2006] and tnet packages [Opsahl 2009] in order to allow an easier application on other datasets.
Finally, we aim to emphasize some aspects of our approach with regard to the more general debates in archaeology. From actor-network theory to graph analysis, network-based approach in archaeology pushed archaeologists to encounter “ontological” issues. But ontology is then understood in two ways: either following the view developed by philosophical anthropology, as in works by T. Ingold or E. Viveiros de Castro [see Alberti, Marshall 2009] or in the perspective closer to what ontology means in computer science and formal ontology. We firmly consider our approach of fragmentation as a contribution to a human-decentralized [Lucas 2012] perspective in archaeology and a tool for an archaeology considered not only as the discipline of things [Olsen & al. 2012] but of fragmented things.