Mimosa scabrella Bentham (Bracatinga), is a native tree species of Atlantic Forest, of restricted occurrence to the Southwest and South of Brazil (RS, SC, PR and SP). The species comes being cultivated in the Brazil south and in the other countries as Costa Rica and Ruanda, for the most varied ends. But, although the species has a varied ends, lack studies in the genetic breeding. Thus, in May of 2003, a M. scabrella provenance progeny test was implanted in the Experimental Station of Itatinga (ESALQ/USP). The design used was the Compact Family Block with eight provenances, 10 to 20 families per provenance, six plants per subplot and five replications (block). Total height and survival at six months old were measured in the trials. Concomitantly, a sampled of about 50 plants per provenance was used to analyses of isozymes electrophoresis, reveling seven polymorphic loci. The results showed that the species has a mixed mating system with outcrossing predominance (average t m = 0.971). In the provenance average was detected 7.6% of biparental inbreeding and 51.3% of biparental outcrossing, indicating strong deviations of random mating and that the families are mixtures of different relatedness kinds. In agreement, the average relatedness coefficient among plant within families was high (0.392). The analyses of genetic diversity detected ˆ ˆ high levels of observed and expected heterozygosities ( H e = 0.580; H o = 0.530 , respectively), although in two provenances (Caçador-SC and Ituporanga comercial-PR) were detected significant excess of homozygotes in relationship to expected in Hardy- Weinberg equilibrium. The studied of association among Nei (1978) genetic distance and geographic distances among provenances, using Mantel test, no detected significant ˆ ˆ correlation ( r = 0.198 ; P = 0.159 ), indicating that the observed genetic distance among provenance patters is not possible to attribute the isolation distance model. The comparison of genetic divergence among provenance measured by quantitative traits ˆ ˆ ( Q p = 0.039) with the measure by isozymes loci ( θ p = 0.018) suggest that the quantitative genetic divergence among provenance at six months old to height trait was ˆ ˆ caused by divergent selection ( Q p > θ p ). Genetic parameters were estimated to height trait, using the random (assume open-pollinated arrays with half-sibs) and mixed (assume open-pollinated arrays with mixed of different relatedness kinds) mating model. The additive genetic variance, heritabilities coefficients and expected gain selection were 36%, 48% and 41%, respectively, super-estimates, where a random model was used instead of mixed mating model.
