A 1H-NMR investigation was carried out on the tetranucleotides U-m6(2)A-U-m6(2)A and m6(2)A-m6(2)A-U-m6(2)A (m6(2) = N6-dimethyladenosine) as well as on the hybrid trinucleotide dA-r(U-A). An extensive comparison with m6(2)A-U-m6(2)A and other relevant compounds is made. Previous proton NMR studies on trinucleotides have shown that purine-pyrimidine-purine sequences prefer to adopt a mixture of states which have as a common feature that the interior pyrimidine residue bulges out, whereas the flanking purine residues stack upon each other. A stacking interaction on the 3' side of the bulge is known to have no measurable effect on the bulge population. Chemical-shift data, ribose ring conformational analysis and information from NOE experiments now show unambiguously that the moderate U(1)-m6(2)A(2) stack in U-m6(2)A-U-m6(2)A diminishes the population of bulged-out structures in favour of a regular stack. This tendency towards conformational transmission in the downstream 5'----3' direction is fully confirmed by the fact that the strong m6(2)A(1)-m6(2)A(2) stack in the tetranucleotide m6(2)A-m6(2)A-U-m6(2)A virtually precludes the formation of bulged-out structures. The conformational characteristics of dA-r(U-A) appear comparable with those of m6(2)A-U-m6(2)A, which indicates that the presence of a 2'-hydroxyl group in the first purine residue is not a necessary prerequisite for the formation of a bulge.