We previously used in vitro selection to identify the 7S11 deoxyribozyme, which catalyzes formation of 2',5'-branched RNA using a branch-site adenosine nucleophile and a 5'-triphosphate electrophile. An unanswered question is whether the use of branch-site adenosine is inherently preferred or a chance event during the particular selection experiment. Here we have found that deoxyribozymes newly selected to use uridine as the branch-site RNA nucleotide in a structural context that resembles natural RNA splicing instead prefer a branch-site adenosine, although adenosine was never available during the selection itself. Our results support a chemical basis for nature's choice of the branch-site nucleotide, which is almost always adenosine in group II introns and the spliceosome.