The genetic divergence of 27 common bean (Phaseolus vulgaris L.) cultivars from different races was evaluated using RAPD markers. The association of this divergence with some indirect variability estimates obtained in field experiments was also assessed. Genetic distances were calculated using Sorensen-Dice's similarity coefficient, displayed in a dendrogram (UPGMA method), and by projection in two-dimensional space. Analysis of molecular variance (AMOVA) was used to evaluate the distribution of variance between and within domestication centers. RAPD marker genetic diversity within these centers was also calculated. The correlations between estimates of genetic distances obtained with RAPD markers and estimates of Mahalanobis' generalized distances, specific combination ability, and midparent heterosis obtained from morpho-agronomical traits were also determined. RAPD markers were efficient in separating cultivars according to domestication centers. However, only for Middle American domestication center was there good agreement between the grouping obtained and the proposed classification of races. The genetic divergence obtained with these markers was similar to that provided by morpho-agronomical traits. RAPD markers, however, were not efficient in predicting the specific combination ability and midparent heterosis expressed in crosses.
