To explain now-numerous cases of codon reassignment (departure from th
e ''universal'' code), we suggest a pathway in which the transformed c
odon is temporarily ambiguous. All the unusual tRNA activities require
d have been demonstrated. In addition, the repetitive use of certain r
eassignments, the phylogenetic distribution of reassignments, and the
properties of present-day reassigned tRNAs are each consistent with ev
olution of the code via an ambiguous translational intermediate.