Danesfort 5 in County Kilkenny, where some of the radiocarbon dates used in the study came from. (PHOTO: IAC Archaeology Ltd)

Many people, myself included, have been drawn to archaeology because of its emphasis on understanding past human behaviour – and perceptions of its distinct lack of mathematics. But this is quickly changing, as the use of statistics increases. One area where this is most evident is in population estimation. Previously, large-scale changes in population were quite difficult, if not impossible, to discern from the archaeological record. But while there are still many biases and pitfalls, new statistical techniques are starting to provide innovative ways to determine movement and migration patterns. In this month’s ‘Science Notes’, we explore some of these new techniques, and examine recent research that has utilised them to assess population fluctuations in Ireland.

Determining an estimate of a historical population largely depends on the period in question. More recent eras obviously benefit from the survival of documents, which may preserve information on housing density and census records, or at least provide substantial clues from which such estimates can be made. As we go further back in time, though, these estimates become increasingly difficult to make as the archaeological record becomes increasingly sparse. Despite the challenges, there are several tools available and they are becoming more and more accurate as further research is dedicated to this area.

In their paper, recently published in the Journal of Archaeological Science (https://doi.org/10.1016/j.jas.2019.04.001), Emma Hannah and Rowan McLaughlin from Queen’s University Belfast use a modelling technique called kernel density estimation (KDE), supported by the examination of other evidence, to explore the changing population density of Ireland throughout history, but particularly focusing on the early medieval period.

KDE is a well-established statistical tool used for calculating the relative magnitude of a particular variable of interest – in this case, human activity in Ireland. As a proxy for human activity, Emma and Rowan used a robust database of over 8,800 radiocarbon dates taken from archaeological contexts across Ireland, including Danesfort 5 in Co. Kilkenny. While using radiocarbon dates will have inherent biases, such as only being able to include in the model sites that have been excavated and successfully dated using this method, Ireland is a good test-case because many of the archaeological sites in the region were excavated as part of intense rescue projects during the 1990s and 2000s that, as Emma and Rowan describe, were ‘quasi-random… [and] unbiased by the pre-existing research interests of archaeologists’.

Radiocarbon-based reconstruction of ancient population levels in Ireland. (IMAGE: Emma Hannah and Rowan McLaughlin)

The KDE test uses the input data to produce a gradient from which rates of growth and decline can be determined and, as Emma and Rowan explain in the paper, ‘sudden events like migrations of large numbers of people would manifest in the models as abrupt change to the kernel density’. This is exactly what their models showed, highlighting a period of widespread decline starting c.AD 700 and continuing until c.AD 1100 (their radiocarbon-based reconstruction is shown). What was even more significant was that the results were surprisingly consistent, showing peaks and troughs at the same point in time across the entirety of Ireland and even when isolated to specific dating material or archaeological contexts (e.g. settlements, enclosures, and cemeteries). This suggests that they were not just finding a decline in a particular type of cultural practice, but a genuine decline in the whole population.

These findings are quite different from the previous consensus, which assumed that the early medieval period was a time of sustained growth, both economically and socially. Although the impetus for this decline – whether driven by disease, famine, migration, or a combination of all three – is not apparent, the story is enhanced by the inclusion of other research that can model migration, particularly work using genetics.

In Ireland, recent studies have modelled migration by looking at the rates of genetic admixture in the contemporary population. They have confirmed that the inclusion of the Scandinavian haplotypes in Ireland can be traced to the Viking period, c.AD 900-1200. And while this is, of course, no surprise, it does suggest that when combined with a declining local population, it would not have taken a very large Norse migration to reach the levels of admixture seen in today’s modern population.

As Emma and Rowan explained: ‘These models of decline are speculative and not mutually exclusive. From the empirical archaeological data, we can only state that the process was gradual, drawn over some 500 years, and quite possibly imperceptible even to those living under its shadow. Under these circumstances a significant but relatively small group of Viking Age Scandinavian migrants introduced to Ireland a genomic signal still detectable today.’

While it is doubtful that all archaeologists will suddenly become statisticians, this growing area of research shows the benefits that including mathematical models in archaeological studies can have. It is just another example of how the addition of other disciplines into archaeology can enhance our overall understanding of the past and how it led us to where we are today.

This article appeared in CA 354.

Leave a Reply