In recent years scholarship on the Viking Age rune-stones has tended to focus on single aspects of the stones and their features, without a ‘bigger-picture’ view. This paper sets out to begin filling in this gap, through a focus on a larger-scale interpretation of the rune-stones and what they disclose (implicitly) about the people(s) of Sweden in the late Viking Age, rather than what they were intended to reflect socially. Temporal and regional differentiation was identified as a vehicle to explore this approach. Data analysis, e.g. PCA and Bayesian statistics were established as the best methods to address these goals, and to that end a database (of sorts) was established, based on Rundata and incorporating other information from a variety of sources. The most important components of this analysis were Linn Lager’s cross-classification system, Birgit Sawyer’s catalogue from The Viking-Age Rune-Stones, and the inscription text files provided along with Rundata. While well established as separate social groups in the historical record, there was found to be very little work attempting to identify large-scale differentiation between the rune-stones of Svealand and Götaland. This was attempted using a variety of techniques, including PCA and cross-type analysis, and some differences were identified.
In order to explore temporal differentiation in the rune-stones, a way was sought to investigate just how to rectify the large number of undated stones in the corpus. Using Gräslund’s stylistic dating, a distribution pattern of four separate ‘phases’ (Early, M1, M2, Late) was established, and a Naīve Bayes Classifier was built. Bayes Classifiers are seeing more and more use in archaeological analysis, often in situations where a date-range for an assemblage has been established, but dating an entire corpus of artefacts is either impossible, or prohibitively expensive. Data assembled from the database was input into a test-learner, and the classifier attempted to assign values to the undated rune-stones based on the information obtained from those already assigned to a phase. Ultimately very low levels of differentiation were detected by the classifier, making it difficult to assign dates, but this homogeneity is in itself significant, as it can be interpreted as being symptomatic of high levels of cultural similarity across the various Scandinavian communities of the Viking Age. The implications of the use of the Bayes Classifier, among other forms of analysis, will be discussed, and possibilities for future use explored.