The genetic code is defined by the specific aminoacylations of tRNAs by aminoacyl-tRNA synthetases. Although the synthetases are widely conserved through evolution, aminoacylation of a given tRNA is often system specific-a synthetase from one source will not acylate its cognate tRNA from another. This system specificity is due commonly to variations in the sequence of a critical tRNA identity element. In bacteria and the cytoplasm of eukaryotes, an acceptor stem G3:U70 base pair marks a tRNA for aminoacylation with alanine. In contrast, Drosophila melanogaster (Dm) mitochondrial (mt) tRNA(Ala) has a G2:U71 but not a G3:U70 pair. Here we show that this translocated G:U and the adjacent G3:C70 are major determinants for recognition by Dm mt alanyl-tRNA synthetase (AlaRS). Additionally, G:U at the 3:70 position serves as an anti-determinant for Dm mt AlaRS. Consequently, the mitochondrial enzyme cannot charge cytoplasmic tRNA(Ala). All insect mitochondrial AlaRSs appear to have split apart recognition of mitochondrial from cytoplasmic tRNA(Ala) by translocation of G:U. This split may be essential for preventing introduction of ambiguous states into the genetic code.