Cytochrome b DNA sequence data (ca. 1,140 bp) of 66 Alouatta belzebul from the Amazonian and the Atlantic Forests of Brazil were used for phylogenetic reconstructions and population studies. Our sample consisted of 60 specimens from the Amazonian Forest (captured in 1984 and 1998 in Pará-PA state) and six specimens from the Atlantic Forest (Paraíba-PB state). We found 32 haplotypes, 23 in PA-1984 (with 12 present in more than one individual), 11 in PA-1998 (with two present in more than one individual), and a single haplotype in the PB sample. Animals from PA-1984 and PA-1998 shared three haplotypes while animals from Pará and Paraíba did not share any haplotype. We found 57 variable sites, consisting of 53 transitions and four transversions, with most replacements occurring at third codon position (77.19%) and less frequently at first and second positions (10.53 and 12.28%, respectively). Genetic distance between all haplotypes varied between 0 and 1.2%. Nucleotide diversity estimates between PA-1984 haplotypes and PA-1998 haplotypes were the same (π=0.01), and haplotype diversity estimates were very similar (h=0.96 and 0.93 for PA-1984 and PA-1998, respectively). Maximum parsimony, median-joining, split decomposition, and TCS showed that PA and PB haplotypes had not drastically diverged and that subsequent radiation within these regions was not apparent. No temporal structure was found between PA-1984 and PA-1998. The sum of square deviation estimate for PA-1984 equaled 0.01 (P=0.23), in agreement with a hypothetical model of sudden expansion contrary to PA-1998 whose sum of square deviation estimate (0.40; P=0.04) was not compatible with this model, although the small sample size of PA-1998 as well as the smaller area of capture could have also accounted for this result. Fu's F(s) and R(2) statistical neutrality tests corroborated these propositions. Lack of drastic differentiation was attributable to the once existing connection between the Atlantic and the Amazonian forests at a non-distant past.