Briefly, social network analysis maps the relations of different groups or people or objects to each other through specific attributes. Social networks as visualized in archaeology consist of nodes (or dots) and edges (or lines), in which the nodes represent a person, group, or other entity and the edges represent a relationship between two nodes. The type of relationship can vary (e.g. trade, political relations, or geographic distance) as well as the nodes (e.g. individual graves in a cemetery, sites with bronze drums in southern Thailand, or individual polities in Metal Age Southeast Asia). The example below shows a hypothetical network of glass trade in South and Southeast Asia during the first millennium based on similarities in glass chemistry.
The importance of SNA for archaeology and anthropology as a whole lies in its ability to map any type of relationship between anything. It can be used to understand the degree of friendship between various individuals, the similarity in chemical composition between various obsidian blades, or whether specific characters in an ancient tale are associated with specific places. So long as the nodes and relationships are defined, SNA can provide a visualization of the relational structure.
This leads to two other important features of SNA. The first is an emphasis on relationships. In using SNA, we are concerned with who or what the nodes represent, but we are equally and perhaps more concerned with the relationship between them. Since we are looking to recreate the structure of relationships between people or things, SNA therefore becomes highly useful when speaking of things like landscapes, heterarchy, or any other relational model.
The other important feature of SNA is its ability to test hypotheses concerning relationships. Should we wish to look at a specific type of heterarchy or landscape or other relation, we can create a model by combining SNA with graphical models and test the archaeological data against it. While this does not necessarily work on a predictive level, it does provide a reliable and testable means through which we can begin to identify and speak of past relationships.
Let’s use glass beads as an example, since that’s what this blog is all about. Glass and glass beads require specific materials and skills in acquiring raw materials, manufacturing each product, selling the goods, use of the objects, and discard. This process requires natural sand and other mineral deposits large enough to sustain such a craft. It also requires large amounts of timber for firing the furnaces as well as clay for crucibles, furnaces, and other tools. Metal also becomes highly important, because it is the only material that will not burn under the high temperatures required for glass manufacture. Trade would require these sites to be located near major trade routes, probably maritime trade routes.
We may try to look at the differences between glass and bronze use, or the differences in importance of different types of glass. We can even speak of the use of different types of beads or colors of beads, as well as different uses for each. Based on the differences in each of these categories, we can begin to form an understanding of the different social systems connected to the glass trade, such as that between manufacturers, between manufacturers and merchant, or between merchants and consumers, among others.
Unfortunately, though, we do not have information about the specific natural deposits of ore or the location of early trade routes for these sites. Nor do we know of the specific uses for glass beads or the meaning embedded in the various types, colors, or ingredients. Yet, surely there must be some patterning of glass beads in the archaeological record, given the number of types colors, and chemical compositions. Herein lies the importance of SNA.
We can input data about a single characteristic of beads, like the number of specific colors, shapes, or chemical types present at a site. We don’t necessarily need to know anything else about the site except that characteristic. The chart above showed a social network analysis of first millennium AD glass bead sites in Southeast Asia on the basis of the number of each chemical type present at a site. The following chart shows the exact same thing, but uses the basis of colour instead.
When viewing this data in SNA, we need not have a set model in mind as to how the data will connect. The user sets a threshold at whatever level, which tells the program to draw a line between nodes (or sites) with a certain level of similarity. Thus we can connect those sites with 46% similarity or higher or – since 46% similarity means a potential 54% dissimilarity, which isn’t all that great – we can draw connections only between sites with 90% similarity or higher (below). Each threshold makes certain assumptions about the data and its connectedness.
Since we do not know anything about how connected the glass bead trade was, let us begin by finding the threshold at which every site becomes connected to the network. This clearly makes the assumption that every site we have is somehow connected. This may not be true, but then neither might anything else we might say about glass trade so far. Thus, let us view the network that appears when every site is connected to at least one other site when looking at similarity in color composition. This is where the above chart with a threshold of 46% comes from, as does the chart analyzing chemistry with a threshold of 31%. The fact that certain connections remain when we increase the color threshold to 90% indicates that the patterns seen at 46% similarity may be significant.
Yet, both charts use the same sites, but measures similarity in two difference categories of bead data. The chart analyzing chemical similarity demonstrates a different relationship than that seen for color prevalence, suggesting the potential for either characteristic to generate relationships and play a role in social organization. It might be useful, then, to look at the relations created by similarities in chemical type and color composition within the same chart. The following chart shows each site plotted in relation to the others in which the threshold has been set at 50% similarity. Red lines connect sites which are 50% or more similar in chemical composition, blue lines connect sites which are 50% or more similar in color composition, and pink lines connect sites that are 50% or more similar in both categories. If the threshold were 0% or greater, then all ties would be pink.
Notice that at 50% similarity there are still many sites that are similar in only one category. Let us also look at the threshold for these relations: what threshold, when viewing both chemical and color composition, serves as the point at which all nodes are connected? Interestingly, though probably not surprisingly, this threshold falls at 55% similarity (chart below), rather than the lower similarities in the individual networks (color: 46%; chemistry: 31%).
Interestingly, most of the sites with connections to others based on 55% similarity in both color and chemistry are also potential if not definite manufacture sites for glass and/or glass beads. More research is needed before any conclusions can be made, but this would indicate that manufacture sites are well-connected in both categories for glass bead data, therefore remaining potentially well-connected at the manufacture as consumer level. While this is not terribly surprising for the logic of a manufacture site, SNA has allowed us to visualize this data and come to this conclusion using simply chemical and color composition and each site.
All of this further demonstrates the point made earlier concerning the differences between the social relations created when looking at chemical composition versus those formed when looking at color assemblages. Chemical composition may be highly important to those working the glass into beads or recycling them into something else, while color would probably remain highly important in consumption. Thus, these two factors alone create different – though not necessarily separate – social landscapes.
Social network analysis and the theories surrounding it are far from perfect, but so far they have allowed for an expanded notion of the concept of “relationship” as well as a more systematic means for hypothesizing and testing these relational models. By archaeological analysis with SNA, archaeologists can use data that may not fit current models and begin to create a more complex and probably more accurate picture of the reality of social and geographical relationships at work within ancient societies.
Hanneman, Robert A. and Mark Riddle. 2005. Introduction to social network methods. Riverside, CA: University of California, Riverside (published in digital form at http://faculty.ucr.edu/~hanneman/ ).