This chapter focuses on the genetic background leading to the “RF-tRNA mimicry hypothesis” and the progress made after its proposal by using genetic probes of bacterial release factors (RFs). It should be a surprising fact that it took more than 30 years after the discovery of the mechanism of sense-codon decoding by tRNA before the authors began to understand the mechanism of stop codon decoding by protein RFs. A major breakthrough in this long-standing coding problem was achieved by the discovery of universally conserved structures in prokaryotic and eukaryotic RFs that led to a novel hypothesis of “molecular mimicry” between RFs and tRNA. An intriguing RF2 variant that acquired omnipotent decoding activity has been isolated as a plasmid-borne prfB mutant that restored the growth of a temperature-sensitive prfA1 strain of Escherichia coli. Of the proposed seven domains of prokaryotic RF1 and RF2, the C-terminal tRNA mimicry portion, including domains D and E, is highly conservative, whereas the N-terminal portion, including domains A and B, is much less conservative. Intriguingly, E. coli RF2 is an exception with Thr at this position. Due to this abnormal residue, E. coli RF2 terminates translation very weakly at UAA or UGA, and an excess of E. coli RF2 was toxic to cells. The RF-tRNA mimicry hypothesis has proven useful to solve the problem of how RF recognizes the stop codon.