The Genetic Structure

The infra-structure of three relatively undisturbed tribes of American Indians (Yanomama, Makiritare, Xavante) has been investigated by means of the F-statistics of Wright, using 8, 9 and 6 codominant systems respectively. The data for the first two mentioned tribes are much more extensive (37 and 7 villages) than for the third (3 villages), and much of the argument is based on the first two. An additive model partitioning F,, into an average effect ( F A ) and deviations due to deme size, systems effects, village effects, and random error has been employed. The Cannings-Edwards formulation suggests that the small size of the demes alone would result in an 8,, of -0.008 for the Yanomama and -0.007 for the Makiritare. There is no evidence for significant village or systems effects. Despite considerable scatter, 8, values are not significantly heterogeneous and tend to be negative (-0.012 to -0.023). On the basis of a computer simulation model, it appears that there is an excess of consanguineous marriage over random expectation, i.e. the negative F A values are probably not due to avoidance of close inbreeding in a subdivided population in which demes are small. Aspects of population structure which could contribute to negative k, values are identified. These include unequal gene frequencies in the sexes and occasional marked differential fertility. I t is a t this point unnecessary to introduce overdominance as a cause of the negative F A values, since a computer simulation program which does not incorporate selection satisfactorily reproduces the observed F,, values. If population breeding structure alone can result in negative F,, values, then this may constitute a mechanism for retarding random fixation.-Mean FST values are 0.063 for the Yanomama and 0 036 for the Makiritare. While truly comparable data are lacking, it seems likely these will be found to be relatively high values for human populations. krT values have been calculated by both direct and indirect approaches. The direct approach yields a value of 0.045 for the Yanomama and -0.009 for the Makiritare; the respective indirect values are 0.085 and 0.017. The primary identifiable reason for this difference between tribes is the greater genetic heterogeneity among Yanomama villages. The assumptions ' The support of the U.S. Atomic Energy Commission and National Science Foundation is gratefully acknowledged. a It is a pleasure to acknowledge the extensive cooperation of the Venezulean Institute of Scientific Investigation. This program of investigation represents a contribution of both this Institute and the University of Michigan to the International Biological F'rogram. Genetics 72: 639-666 December 1972. 640 J. V. NEEL A N D R. H. WARD underlying the indirect approach to the calculation of FIT do not appear to be met in these populations. AN understanding of the organization and breeding structure of natural populations is prerequisite to any formulation which purports to account for the large amount of genetic variability encountered in all adequately studied plant and animal species, including man. Human populations present unusual advantages for studies of population structure, since not only can a single blood specimen yield data on some 15 systems at which codominant alleles may occur in polymorphic proportions, but members of the populations can supply a variety of data on history and mating patterns not readily obtained from members of other free-living populations. Unfortunately, as with many other animal species, relatively undisturbed, “natural” (i.e. primitive) human populations whose mode of life reflects the circumstances under which this variability arose, are rapidly decreasing in number. During the past 10 years, we have engaged in intensive, multidisciplinary studies of three South American Indian tribes (Xavante, Yanomama, Makiritare) which, while certainly not still in the pre-Columbian state (NEEL 1972), appear to have retained their primeval, tribal population structure essentially intact at the time of study. The available formulations for characterizing that structure are essentially five in number: the F-statistics of WRIGHT (1943 et seq.), the distance-kinship assay of MALBCOT (1948, see 1969), the matrix characterization of migration of KIMURA and WEISS (1964) and BODMER and CAVALLI-SFORZA (1968) , the genetic distance matrices and dendrograms of EDWARDS and CAVALLI-SFORZA (1964 et seq.) , and the analyses of variance of gene frequencies of COCKERHAM (1967,1969). The overlap between these formulations is substantial. Recognizing the formidable difficulties in reasoning with any rigor from a set of field data to an underlying mathematical model, we feel the prospects for so doing are better for man than any other species, and propose in time to analyze the data from these three tribes by all these approaches, followed by a systematic comparison of the results, since this is one way in which the strengths and limitations of each approach become apparent. The present paper will be entirely devoted to the problems and results which issue from an attempt to derive F-statistics as measures of the breeding structure of these tribal populations.