FRUITS of the sea? A cautionary tale regarding Bayesian modelling of palaeodiets using stable isotope data

Abstract

A recent paper by Bownes et al. (2017, Radiocarbon 59(5): 1275–1294) used the Bayesian modelling software package FRUITS (Fernandes et al. 2014, PLoS ONE 9(2): e87436) to argue that Neolithic individuals from Carding Mill Bay on the west coast of Scotland obtained up to ca. 21% dietary protein from marine sources. This is in contrast to previous interpretations of the same isotopic (δ13C and δ15N) data, which concluded that these individuals showed little if any use of marine resources (Schulting and Richards 2002, European Journal of Archaeology 5(2): 147–189). Resolving this discrepancy is important for our understanding of the nature of the Mesolithic–Neolithic transition not only on the west coast of Scotland, but along all of Atlantic Europe, since similar isotopic results to those obtained at Carding Mill Bay are widespread throughout the Neolithic and indeed later periods. We suggest that greater caution needs to be excercised in the interpretation of the output of Bayesian palaeodietary modelling, which can be very useful heuristically, but should not always be taken at face value. Given the large number of parameters employed in such models (diet isotopic values and nutrient concentrations, trophic level and tissue fractionations, etc.) and the uncertainty involved in almost all of them, a wide range of outcomes are possible, as we demonstrate in this paper. We reaffirm the overwhelmingly terrestrial nature of diet at Carding Mill Bay both through FRUITS modelling and through the new application of compound-specific δ13C analysis of single amino acids of humans and fauna from the site.