<jats:title>Abstract</jats:title><jats:p>Bacteria require two class-I release factors, RF1 and RF2, that recognize stop codons and promote peptide release from the ribosome. RF1 and RF2 were most likely established through gene duplication followed by altering their stop codon specificities in the common ancestor of extant bacteria. This scenario expects that the two RF gene families have taken independent evolutionary trajectories after the ancestral gene duplication event. However, we here report two independent cases of conversion between <jats:italic>RF1</jats:italic> and <jats:italic>RF2</jats:italic> genes (<jats:italic>RF1</jats:italic>-<jats:italic>RF2</jats:italic> gene conversion), which were severely examined by procedures incorporating the maximum-likelihood phylogenetic method. In both cases, <jats:italic>RF1</jats:italic>-<jats:italic>RF2</jats:italic> gene conversion was predicted to occur in the region encoding nearly entire domain 3, of which functions are common between RF paralogues. Nevertheless, the ‘direction’ of gene conversion appeared to be opposite from one another—from <jats:italic>RF2</jats:italic> gene to <jats:italic>RF1</jats:italic> gene in one case, while from <jats:italic>RF1</jats:italic> gene to <jats:italic>RF2</jats:italic> gene in the other. The two cases of <jats:italic>RF1-RF2</jats:italic> gene conversion prompt us to propose two novel aspects in the evolution of bacterial class-I release factors: (i) domain 3 is interchangeable between RF paralogues and (ii) <jats:italic>RF1-RF2</jats:italic> gene conversion have occurred frequently in bacterial genome evolution.</jats:p>