Osteological age-at-death estimation in an archaeological sample avoiding age-mimicry: a six-step approach

Archaeological and Anthropological Sciences https://doi.org/10.1007/s12520-024-02034-0 (2024) 16:126 RESEARCH Osteological age-at-death estimation in an archaeological sample avoiding age-mimicry: a six-step approach Denise U. Navitainuck1 · Werner Vach1,2 · Kurt W. Alt1,3 · Sandra L. Pichler1 Received: 24 March 2023 / Accepted: 3 July 2024 © The Author(s) 2024 Abstract In human osteoarchaeology, individual age-at-death is traditionally assigned by scoring characteristic expressions of specific traits and applying formulas or algorithms. However, it is well known that the resulting age estimates in target samples suffer from bias due to the effect of age-mimicry, both at the individual and at the population level. A general statistical methodology to obtain unbiased estimates of the age-at-death distribution in skeletal samples has been available for more than two decades. Even so, the procedure is rarely used. This may be related to the fact that this methodology requires external input which has selection of distributional characteristics to be reported. In this paper, we translate the general methodology into a clearly stated six-step procedure to be followed. We illustrate the six steps using an archaeological sample of 675 adult individuals and 15 scoring methods from traditional age-estimation procedures. By clearly identifying the actions that are necessary for its application we intend to make the approach more accessible for osteoarchaeologists while at the same time highlighting some challenges that need to be addressed in the future. Our study demonstrates that the approach is feasible and illustrates the absence of age-mimicry. A combined analysis of five informative traits allowed to obtain estimates of several characteristics of the target sample age distribution. However, its routine use will benefit from improved input from relevant reference samples and improved statistical software. Keywords Human osteoarchaeology · Aging methods · Age-mimicry · Population age mean · Population standard deviation · Individual age estimation · Rostock manifesto Introduction To describe either an individual or a population, age-at-death is one of the most basic and important pieces of information in both forensics and bioarchaeology. Various features describing morphological changes and signs of wear/degeneration in the human skeleton have been identified for age diagnosis and form the basis for various methods for osteological age estimation. Denise U. Navitainuck 1 Integrative Prehistory and Archeological Science (IPAS), Department of Environmental Sciences, University of Basel, Spalenring 145, Basel CH-4055, Switzerland 2 Basel Academy for Quality and Research in Medicine, Steinenring 6, Basel CH-4051, Switzerland 3 Center of Natural and Cultural Human History, Danube Private University, Krems, Austria The starting point of formal osteological age-estimation methods is typically a scoring procedure, translating discrete characteristics of a given trait into numerical values. The score values may also arise from direct, metric measurements. Traditionally, formal age-estimation methods predict the age of an individual based on the score value obtained. It is, however, well known that this approach suffers age-mimicry: The resulting age estimates tend to reflect the age structure of the reference sample (Bocquet-Appel and Masset 1982; Mensforth 1990; Konigsberg and Frankenberg 1992; Aykroyd et al. 1997). Only recently has it been shown that age-mimicry can explain differences in population mean age estimates of 15 years or more if different age-estimation methods are applied to the same archaeological sample (Navitainuck et al. 2022). In principle, it is also known how age-estimation is to be performed in order to obtain unbiased estimates of individual age and of population age distributions, the basic statistical framework of which has been propagated by the Rostock Manifesto (Hoppa and Vaupel 2002). The key step 13 126 Page 2 of 17 is to use a reference sample to describe the distribution of the score as a function of age, and not the distribution of age as a function of the score – as done by a prediction rule. Such a description can be combined with a suitable distributional assumption of the age distribution in the target sample in order to obtain unbiased estimates of the age-atdeath distribution. In a final step, individual age estimates can be obtained. However, this framework has rarely been used to derive population age distribution estimates in archaeological samples (Herrmann and Konigsberg 2002; Caussinus and Courgeau 2010; Seguy et al. 2013). Only transition analysis introduced by Boldsen et al. (2002) has gained some popularity, as illustrated by its application in diverse samples worldwide. However, transition analysis only considers those aspects of the general framework allowing to compute individual age estimates based on additional assumptions. In order to make the whole framework more accessible to osteoarchaeologists, we will translate the general methodology into a clear six-step procedure. We apply these six steps in an archaeological sample of 675 adult individuals, using 15 scoring methods from traditional age-estimation methods based on nine different traits. In this way we illustrate their applicability as well as addressing the challenges inherent to the approach. We further illustrate the absence of age-mimicry when the six-step procedure is used. A discussion on the future perspectives of applying the general framework concludes this paper. Materials and methods The target population: an archaeological sample The early medieval cemetery of Mannheim-Seckenheim (‘Hermsheimer Bösfeld’), Germany, comprised 908 individuals dating to the 6th to 8th centuries. Although the majority of these were buried as simple inhumations with few or no grave goods, individuals in stone cists and wooden chambers regularly exhibited rich furnishings. In this study, we included all 675 adult individuals, i.e., individuals who had lived to at least 20 years of age. Classification as an adult individual was based on complete epiphyseal closure (especially the spheno-basilar junction). Basic statistical methodology The general statistical framework to obtain valid estimates of individual age has been clearly described by the Rostock Manifesto (Hoppa and Vaupel 2002). The first essential point is to use the reference sample to estimate the aging process of the characteristic of interest at the population 13 Archaeological and Anthropological Sciences (2024) 16:126 level, i.e., how the distribution of the characteristic is changing with age. Formally, this mean that we estimate the conditional distribution of the characteristic given age, i.e., the probabilities P r (c| a) to observe the value c of the characteristic given the age a of the individual. This is contrary to traditional age-estimation methods, which try to predict an individual’s (...truncated)