Sexing skulls using discriminant function analysis of visually assessed traits.

Area of Science:

• Forensic anthropology
• Biological anthropology
• Bioarchaeology

Background:

• Visual assessment of cranial traits is a common method for sex determination in skeletal remains.
• Previous methods have varying accuracy and may exhibit sex biases.
• Understanding population-specific variations in cranial morphology is crucial for accurate sex estimation.

Purpose of the Study:

• To evaluate the accuracy of sex determination using visual assessments of specific cranial traits.
• To compare the performance of different statistical models in sex estimation.
• To investigate population-specific differences in cranial sexual dimorphism.

Main Methods:

• Analysis of 304 modern skulls (European American, African American, English) and 156 ancient Native American skulls.
• Scoring of five sexually dimorphic cranial traits: mental eminence, orbital margin, glabellar area, nuchal area, and mastoid process.
• Application and evaluation of linear, kth-nearest-neighbor, logistic, and quadratic discriminant analysis models.

Main Results:

• Logistic regression discriminant analysis yielded the highest accuracy, correctly classifying 88% of modern skulls with minimal sex bias (0.1%).
• The best logistic model correctly assigned sex to 78% of ancient Native Americans, with a sex bias of 0.2%.
• Cranial trait distributions showed similar patterns within modern populations but contrasted significantly with ancient Native Americans, indicating population-specific dimorphism.

Conclusions:

• Logistic regression using five cranial traits provides a robust method for sex determination.
• The accuracy is high in modern populations but requires caution when applied to different ancestral groups due to variations in sexual dimorphism.
• Discriminant functions developed for one population should be used cautiously on others.