5.2 Modelling the terrestrial δ34S

There is a statistically significant positive correlation between δ13C and δ34S for humans (Spearman's r= 0.55, p<0.001), which is also obvious from Figure 3. The implication is that the marine component of the diet elevates the sulphur isotope value, thus obscuring mobility patterns. In order to discern mobility from such data, it is therefore necessary to isolate the δ34S of the terrestrial dietary protein. This is required because only the terrestrial δ34S values reflect the local environment, whereas marine δ34S values are consistently elevated. Employing a model first suggested by Fornander et al. (in press) for 87Sr/86Sr, and employed on δ34S data by Eriksson et al. (2013), an estimate of the original terrestrial δ34S value can be calculated, which can subsequently be compared to the local terrestrial range.

Previous δ13C and δ15N analyses (Eriksson et al. 2008) have demonstrated that the Neolithic and Bronze Age humans on Öland relied essentially on only two main protein sources - terrestrial herbivores and marine mammals. This was based on extensive faunal data showing a clear isotopic separation between fish and marine mammals, where the strong correlation between human δ13C and δ15N ruled out any substantial contribution of fish protein to the human diet (whereas dogs were suggested as the main consumers of fish, providing a better fit with the isotopic data cf. Eriksson 2004).

Because there are only two main protein sources, and because of the demonstrated association between δ34S and δ13C, a linear mixing model can be employed (making the use of more complex models superfluous). Accordingly, the estimated terrestrial δ34S value, Rt, is calculated as:

Rt = Robs + (Robs - Rmar) × M

where Robs is the observed δ34S value, and Rmar is the marine δ34S end-point value, +15‰. M is the marine factor, which is calculated as


where Cmar is the percentage of marine dietary protein, which is calculated from δ13C, using -22‰ and -13‰ as terrestrial and marine end-points, respectively.

The calculation of the marine δ34S mean (Rmar, +15‰) is based on marine mammals only, that is, harbour porpoise (n=3), ringed seal (n=2) and harp seal (n=1). The δ13C and δ15N evidence shows that the marine protein consumption derived mainly from marine mammals, and apparently not from fish or birds to any large extent. Moreover, the marine bird (+8.4‰) was not determined to species, and therefore whether it was migratory could not be established. This does not exclude the possibility that its flesh or eggs constituted a minor part of the diet, and its value is accordingly informative. Much the same could be said about the garfish (+17.3‰), with the addition that its C/S ratio indicates that it should be treated with caution. Including the bird and the fish would not substantially alter the mean marine δ34S value (from +14.8 to +15.2‰), and it was deemed reasonable to exclude them from the calculation.

The local terrestrial range, reflecting the bioavailable δ34S composition of the island, is calculated as the δ34S mean ± 2 standard deviations of wild local terrestrial fauna. The resulting range, +6.3‰ to +13.7‰, is thus based on δ34S data from moose (n=3), roe deer (n=1), pine marten (n=1), mountain hare (n=5) and wild/feral pigs (n=3) (cf. Table 2). All these species are likely to have fed both at the coast and the interior, and their home ranges to have been naturally delimited by, and in most cases covered, the whole island, which is more or less geologically homogeneous. The mountain hare displays the widest range of δ34S values, from 7.8 to 13.6‰, probably reflecting more limited home ranges of individual hares. In view of the limited size of the island, it is likely that the terrestrial plants, and by implication also the local fauna, are to some extent influenced by the sea spray effect - the hare displaying the highest terrestrial isotopic value, +13.6‰, is conceivably an example of this. However, the sea spray effect cannot be expected to be as substantial on Öland as on islands of comparable size in the ocean or the Mediterranean, first because of its proximity to the mainland, but above all because the size, morphology, salinity and history of the Baltic Sea differ radically from the oceans. The established local terrestrial range is consistent with these facts, also supported by δ34S analysis of modern cod, demonstrating that Baltic cods have lower δ34S than Atlantic ones (Nehlich et al. 2013). The impact of the sea spray effect is also considerably lower because of the lower marine δ34S range.

Domestic species were not included in the calculation of the local terrestrial range, mainly because of the risk that they were imported, or subject to specific cultural practices affecting what they fed on, hence resulting in δ34S values not representative of the bioavailable sulphur of the local environment. The Middle Neolithic cattle specimen from Köpingsvik - clearly contemporaneous with the Pitted Ware population, but culturally an 'exotic' - is an example of the former, dogs of the latter. Nevertheless, the majority of the domestic animals - including one pig from Resmo - display rather homogeneous δ34S values, within the local terrestrial range, consonant with a local origin. This is not surprising, as it is highly unlikely that people in Neolithic and Bronze Age Scandinavia relied to any large extent on imported livestock. Two specimens from Resmo fall below the local terrestrial range; an ovicaprid tibia of historical date (thus probably imported) and a pig tooth (not dated).

The application of the model allows the identification of genuine mobility patterns at the individual level. As for any model, there are of course intrinsic uncertainties and errors, and these increase with the percentage of marine protein input. At a certain point there is accordingly so much imprecision that the estimate is no longer informative. Our assessment is that this point is reached at around 60% marine dietary input. Consequently it is not possible to make any estimates for the Köpingsvik individuals. Their δ34S values correspond with previously published data from the Pitted Ware site Korsnäs in eastern central Sweden (Fornander et al. 2008), indicating that high marine protein consumers have uniform values throughout the Baltic Proper. Residential changes will accordingly not be discernible.


Internet Archaeology is an open access journal. Except where otherwise noted, content from this work may be used under the terms of the Creative Commons Attribution 3.0 (CC BY) Unported licence, which permits unrestricted use, distribution, and reproduction in any medium, provided that attribution to the author(s), the title of the work, the Internet Archaeology journal and the relevant URL/DOI are given.

University of York legal statements

File last updated: Thu Nov 13 2014