Abstract Anthropometric, serological and previous skeletal studies bearing on the question of Bushman‐Hottentot similarities and the relationships of these populations to South African Negroes are summarized briefly. Measurements of well documented crania representative of these three groups are then examined using the multivariate statistical techniques of generalized distance and discriminant function analysis, and the following points emerge: The Bushman‐Hottentot generalized distance as computed on Mahalanobis' D2for six characters, and again on D2and Penrose's C2Husing 12 measurements, appears small (and non‐significant) relative to the Bushman‐Negro and Hottentot‐Negro separations, which achieve statistical significance at the 5% level; in general the analysis suggests Bushmen and Hottentots to be fully as similar in cranial form as in blood group and serum protein distributions. That these peoples should be “lumped” in a Negro category appears doubtful on the evidence available. Simple two‐group discriminant functions, computed on subsets of variables selected from the 35 original measurements, correctly assign “race” in 89% of cases when Bushman and Hottentot males are compared, and in 98–100% of cases when Bushman skulls are paired with those of Zulus. The latter discriminants also perform well when confronted with “new” (Sotho) skulls of both sexes. By showing which measurements or aspects of morphology are important to group separation, discriminant analysis provides an interpretation of differences between these series, and the functions themselves should prove efficient aids for classifying material of unknown of questionable origin, assuming that such skulls in fact belong in one of the parent groups included in the analysis. A multiple discriminant approach, whereby seven groups of Bushman, Hottentot and Negro crania are located in a multidimensional space or statistical framework defined by six 29‐variable functions as axes, permits simultaneous discrimination by both sex and race and allows correct assignments to be made for nearly 80% of all (179) individuals tested. Examination of the scaled vector versions of the functions suggests that measurements of over‐all vault size are generally important in distinguishing between the sexes, while various aspects of skull form are the more efficient race discriminators. It seems likely that other shape‐directed measurements, more sensitive to racial variation in cranial morphology than many currently employed, could and should be devised in order to exploit fully the capabilities of discriminants as tools not only for assigning individual specimens to a population but also for classifying the populations themselves in objective terms.
Stored only when both papers are in the corpus — 60 citations exist upstream.