Sex and age estimation in human osteoarchaeology: improving standards with advanced methodologies

The dissertation “Sex and age estimation in human osteoarchaeology: improving standards with advanced methodologies” aimed to enhance methods for accurately estimating sex and age in human skeletal remains from archaeological samples. The study used a large sample of early medieval adult skeletons from ‘Hermsheimer Bösfeld’ as a test case, focusing on three specific topics.
The first topic was the systematic selection of traits for osteological sex estimation. While many traits are available for this purpose, reliable sex estimation does not require evaluating all of them. Given limited resources, it is essential to identify subsets of traits that are most informative, yet limited in size. The aim of the first research paper was to find out how to compile these subsets of traits, allowing a combination of traits from different anatomical regions as well as metric and morphological traits. An exemplary set of characteristics was used. In order to compare different subsets of traits with respect to their utility for sex estimation a simple measure for the utility was introduced: the fraction of subjects for which a posterior probability of 95% of being male or female could be reached, based on application of a discriminant analysis. To apply the latter, a reference standard was created using LCA (latent class analysis). However, also the few existing approaches to combine the information from multiple traits were included in the comparison, as well as an evaluation of each single trait. It was found that a combination of several traits from different anatomical regions yields the best results, i.e. the highest utility for estimating sex in adult individuals. The maximally possible utility of 0.76 was reached when combining traits from three anatomical regions.
The second and third topics related to age-mimicry in age estimation. Age-mimicry describes the fact that age estimates based on traditional age estimation methods to some degree reflect the age distribution of the reference population used in developing the specific methods, in particular its mean age. The investigations of the second and third topic were based on the application of 15 well-established and noninvasive methods in the field of paleodemographic studies (Wittwer-Backofen et al., 2008; Falys & Lewis, 2011; Buckberry, 2015), that can be applied in large archaeological series by macroscopic inspection and without application of imaging techniques or further technical means of investigation. All selected methods also allow translating a specific age indicator into an age estimate. The 15 different age-estimation methods based on nine different osteological/dental traits were once again applied to the Bösfeld sample.
The well-known phenomenon of age-mimicry is often considered a major obstacle against the application of formal age estimation methods. This bias in population mean estimates implied by age-mimicry has never been quantified before, hence the aim of the second research paper was to do so. The results showed that a difference of 3.57 years between the mean ages in two reference samples predicts a difference of 1 year in the corresponding mean ages of the target population. Thus, the bias implied by age-mimicry could be successfully quantified.
Methods to estimate the population age mean in an archaeological sample without suffering from age-mimicry have existed for over 20 years (Konigsberg & Frankenberg, 1992; Müller et al., 2002), but their use remains rather limited. The third investigation aimed to apply the available methodology to a large sample to demonstrate its potential and identify challenges. Results indicated successful application of the age estimation methodology in the Bösfeld sample avoiding age-mimicry, though challenges remain in terms of data base, statistical methodology and software, as well as suitable reference samples.
In summary, much has already been accomplished in the development of reliable sex and age estimation methods and their application in human osteoarchaeology. Now, there is the need to agree on a best practice.
For sex estimation we are already close to this objective - if the needed software is provided. In age estimation there remains a longer way to go, as we are still in need for consistent and reliable age indicators. For a routine use of the available methodology for age estimation and avoiding age-mimicry, we need more data from relevant reference samples. These data must be made publicly available in order to enable the efficient use of the general methodology on a large scale. With increasing insights into the potential biases in sex and age estimation methods (in adults), a more consequent use of adequate statistical methods, and better input there is distinct hope for the future.