Howells and FORDISC Vindicated

April 21, 2005

Afrocentrists who don't like the work of W.W. Howells on the non-African racial character of pre-historic East Africans often cite research attempting to discredit his methods of classifying crania. But a new paper demonstrates that inaccurate results yielded by computer programs used for this purpose are the result of human error, vindicating Howells and silencing his critics.

The truth is out there: how NOT to use FORDISC


Freid et al. (2005)

FORDISC is an interactive computer program designed to classify an unknown adult cranium based on the reference samples in its database. FORDISC uses discriminant functions to construct a classification matrix and assign group membership of the unknown cranium into one of the selected reference groups. The researcher guides the analysis by choosing the populations against which to classify the unknown, choosing from eleven population samples from the Forensic Anthropology Data Bank or twenty-eight population samples from Howells' (1989) worldwide database. The utility and efficacy of FORDISC has been criticized for providing 'incorrect' classifications, however these disputed results are often due to inappropriate reference samples and failure to properly evaluate the typicality and posterior probabilities provided by the program. In this paper, unknown crania from populations known not to belong to any of the reference samples will be analyzed, demonstrating the interpretation of posterior and typicality probabilities provided in the FORDISC output and the importance of the use of an appropriate reference sample.

Link