In 2006, an experiment examining families belonging to the first selection stage of the Sugar Cane Breeding Program of Universidade Federal Rural do Rio de Janeiro/Rede Interuniversitária para o Desenvolvimento do Setor Sucroalcooleiro was conducted. Families and plants within families were evaluated to select superior plants for subsequent stages of the breeding program. The experiment was arranged in a randomized block design, in which progenies were grouped into 4 sets, each with 4 replicates and 100 seedlings per plot. The following traits were evaluated: average stem diameter, total plot weight, number of stems, Brix of the lower stem, and Brix of the upper stem. The study of families used the restricted maximum likelihood/best linear unbiased procedure mixed models. After selection, families were genotyped via inter-simple sequence repeat to assess the genetic distance of genotypes. This approach was found to be efficient for selecting new genotypes.