Several cellular processes involve alignment of three nucleic acids strands, in which the third strand (DNA or RNA) is identical and in a parallel orientation to one of the DNA duplex strands. Earlier, using 2-aminopurine as a fluorescent reporter base, we demonstrated that a self-folding oligonucleotide forms a recombination-like structure consistent with the R-triplex. Here, we extended this approach, placing the reporter 2-aminopurine either in the 5'- or 3'-strand. We obtained direct evidence that the 3'-strand forms a stable duplex with the complementary central strand, while the 5'-strand participates in non-Watson-Crick interactions. Substituting 2,6-diaminopurine or 7-deazaadenine for adenine, we tested and confirmed the proposed hydrogen bonding scheme of the A*(T.A) R-type triplet. The adenine substitutions expected to provide additional H-bonds led to triplex structures with increased stability, whereas the substitutions consistent with a decrease in the number of H-bonds destabilized the triplex. The triplex formation enthalpies and free energies exhibited linear dependences on the number of H-bonds predicted from the A*(T.A) triplet scheme. The enthalpy of the 10 nt long intramolecular triplex of -100 kJ x mol(-1) demonstrates that the R-triplex is relatively unstable and thus an ideal candidate for a transient intermediate in homologous recombination, t-loop formation at the mammalian telomere ends, and short RNA invasion into a duplex. On the other hand, the impact of a single H-bond, 18 kJ x mol(-1), is high compared with the overall triplex formation enthalpy. The observed energy advantage of a 'correct' base in the third strand opposite the Watson-Crick base pair may be a powerful mechanism for securing selectivity of recognition between the single strand and the duplex.