hairs from the same person is small, about 1 in 4,500.73 This assignment of probabilities has since been shown to be unreliable.74 Moreover, the study does not confirm the chance of asserting a match between two dissimilar hairs, and the authors acknowledge that, “due to the fact that so many of the characteristics coded are subjective—for example, color, texture—it was not possible to get complete reproducibility between two or more examiners coding the same hair.”75

Barnett and Ogle raised four concerns with the Gaudette and Keeping study: (1) it relied on idealized (not from real life) test scenarios; (2) there was no objective basis for selecting the features; (3) the statistical analysis of data from the study was questionable; and (4) there was a possible examiner bias.76 Gaudette attempted to address these concerns through a further study. However, this additional study involved only three hair examiners, in addition to the author. The author concluded that:

… whereas hair is not generally a basis for positive personal identification, the presence of abnormalities or unusual features or the presence of a large number of different unknown hairs all similar to the standard can lead to a more positive conclusion. The problem, at present, lies in finding suitable additional characteristics [of hair, for effecting individualization]. Although there is basic agreement as to the value of the macroscopic and microscopic characteristics used, other characteristics are either unreliable or controversial. Physical characteristics such as refractive index, density, scale counts, tensile strength, and electrical properties have been proposed by some workers but have been attacked by others, and the general consensus is that they are of little use in hair comparison.77

In 1990, Wickenheiser and Hepworth attempted a study to address examiner bias in a small study with only two examiners. They reported that “no incorrect associations were made by either examiner.”78 But a study with only two examiners cannot offer accurate and precise estimates of bias in the population of examiners.

An attempt at an objective system for identifying “matches” among hair samples is presented in Verma et al., based on a neural network.79


A later study on human pubic hairs (Caucasian only) estimated this probability as “about 1 in 800.” B.D. Gaudette. 1976. Probabilities and human pubic hair comparisons. Journal of Forensic Sciences 21(3):514-517.


P.D. Barnett and R.R. Ogle. 1982. Probabilities and human hair comparison. Journal of Forensic Sciences 27(2):272-278.


Gaudette and Keeping, op. cit.


Barnett and Ogle, op. cit.


B.D. Gaudette. 1978. Some further thoughts on probabilities and human hair comparisons. Journal of Forensic Sciences 23(4):758-763, pp. 761-762.


Wickenheiser and Hepworth, op. cit., p. 1327.


M.S. Verma, L. Pratt, C. Ganesh, and C. Medina. 2002. Hair-MAP: A prototype automated system for forensic hair comparison and analysis. Forensic Science International 129:168-186.

The National Academies of Sciences, Engineering, and Medicine
500 Fifth St. N.W. | Washington, D.C. 20001

Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement