Interpreting Past Human Diets Using Stable Isotope Mixing Models

Abstract

Palaeodietary reconstruction using stable isotope analysis is becoming increasingly common, as is the practice of using mixing models to quantify ancient dietary compositions. However, many archaeologists may be unaware of the complexities and pitfalls of stable isotope mixing models (SIMMs). This study serves to provide an overview of the basic principles of SIMMs, evaluates the performances of several of the most commonly used SIMM software packages, and offers some field-specific guidelines for the application of SIMMs in archaeological contexts. We present a series of simulated and published archaeological data to demonstrate and evaluate the different types of SIMMs. We compared the outputs of linear mixing models, simple probabilistic models (IsoSource), and conditional probabilistic models (FRUITS and MixSIAR). Our results show that each mixing model has its pros and cons, and archaeologists should select the best model based on a number of factors, including familiarity with coding languages, sample characteristics (i.e. sample size and normality) of the consumer groups, and research questions.