Introduction

One of the most overlooked, but basic, sources of error in skeletal age estimation comes from problems with the human osteological reference collection on which the method is based. The purpose of this chapter is to review the need for known-age and sex human skeletal collections to be used both for testing current age estimation methods and for developing new ones, and to present a database of such reference collections world-wide.

Bocquet-Appel and Masset (1982) were the first researchers to stress the importance of the reference sample. They showed that, because of the simple regression methods being used for age estimation methods, the age structure of the reference population was reflected in the estimated age distribution in the target (unknown or archaeological) sample. The authors saw this as one of the fatal flaws of paleodemography, and although many did not agree that it was insurmountable, the paper did spur a reanalysis of age estimation methods (see e.g., Konigsberg and Frankenberg 1992, 1994; Milner et al. 1997). However, none of these concentrated on the reliability and representativeness of reference collections, which are usually taken for granted.

There are two ways to deal with the problems with the structure of the reference population. The first is to use a reference collection of skeletons with a uniform distribution of ages. This has been the most logical method in new age estimation methods (see e.g., Boldsen et al., Chapter 5, this volume).