What we see is what we touch? Sex estimation on the pelvis in virtual anthropology

Computed tomography (CT) scans are a convenient means to study 3D reconstructions of bones. However, errors associated with the different nature of the observation, e.g. visual and tactile (on dry bone) versus visual only (on a screen) have not been thoroughly investigated.

We quantified the errors between modalities for sex estimation protocols of nonmetric (categorical and ordinal) and metric data, using 200 dry pelves of archaeological origin and the CT reconstructions of the same bones. In addition, we 3D surface scanned a subsample of 39 pelves to compare observations with dry bone and CT data. We did not focus on the sex estimation accuracy but solely on the consistency of the scoring, hence, the interchangeability of the modalities.

Metric data yielded the most consistent results. Among the nonmetric protocols, ordinal data performed better than categorical data. We applied a slightly modified description for the trait with the highest errors and grouped the traits according to consistency and availability in good, intermediate, and poor.

The investigated modalities were interchangeable as long as the trait definition was not arbitrary. Dry bone (gold standard) performed well, and CT and 3D surface scans performed better. We recommend researchers test their affinity for using virtual modalities. Future studies could use our consistency analysis and combine the best traits, validating their accuracy on various modalities.