RNA editing, a novel and unexpected type of information processing, was first demonstrated in the kinetoplasts of certain protozoans. It is a remarkable phenomenon: certain species of messenger RNA have nucleotide sequences that differ greatly from the sequences of the genes from which they are presumably transcribed. The differences are usually due to addition of uridylate residues, but occasionally also to their deletion. The most spectacular case of editing known occurs in the mRNA of the mitochondrial gene for subunit III of cytochrome oxidase (COIII) in Trypanosoma brucei. This mRNA is twice the length of its gene owing to the addition of several hundred uridylate residues and a few deletions, spread over the entire length of the RNA molecule. Whereas unedited RNA molecules, the nucleotide sequences of which correspond to the genomic sequence, have been isolated, no DNA template corresponding to the edited RNA sequence has been detected in either the mitochondrial or nuclear genome. It was suggested, therefore, that unedited mRNAs are transcribed from mitochondrial DNA and then edited post-transcriptionally by endonucleolytic cleavage of the primary transcript at specific sites, followed by insertion or deletion of uridylate residues and religation. We have now examined the general nature of the RNA editing process to determine whether it involves the insertion of nucleotide residues into pre-existing molecules or the continuous de novo synthesis of edited mRNA. The results of our experiments rule out an insertion mechanism and strongly indicate that edited mRNA is synthesized as a unit.