Retroviruses encapsidate their genome as a dimer of homologous RNA molecules noncovalently linked close to their 5′ ends. The dimerization initiation site (DIS) of human immunodeficiency virus type 1 (HIV-1) RNA is a hairpin structure that contains in the loop a 6-nt self-complementary sequence flanked by two 5′ and one 3′ purines. The self-complementary sequence, as well as the flanking purines, are crucial for dimerization of HIV-1 RNA, which is mediated by formation of a “kissing-loop” complex between the DIS of each monomer. Here, we used chemical modification interference, lead-induced cleavage, and three-dimensional modeling to compare dimerization of subtype A and B HIV-1 RNAs. The DIS loop sequences of these RNAs are AGGUGCACA and AAGCGCGCA, respectively. In both RNAs, ethylation of most but not all phosphate groups in the loop and methylation of the N7 position of the G residues in the self-complementary sequence inhibited dimerization. These results demonstrate that small perturbations of the loop structure are detrimental to dimerization. Conversely, methylation of the N1 position of the first and last As in the loop were neutral or enhanced dimerization, a result consistent with these residues forming a noncanonical sheared base pair. Phosphorothioate interference, lead-induced cleavage, and Brownian-dynamics simulation revealed an unexpected difference in the dimerization mechanism of these RNAs. Unlike subtype B, subtype A requires binding of a divalent cation in the loop to promote RNA dimerization. This difference should be taken into consideration in the design of antidimerization molecules aimed at inhibiting HIV-1 replication.