The E site (exit site for deacyl-tRNA) has been shown to be allosterically linked to the A site (aminoacyl-tRNA binding site), in that occupation of the E site reduces the affinity of the A site, and vice versa, whereas the intervening peptidyl-tRNA binding sire (P site) keeps its high affinity. Here the question is analysed of whether or not the low affinity state of the A site caused by an occupied E site is of importance for the ribosomal accuracy of the amimoacyl-tRNA selection. In a poly (U) dependent system with high accuracy in poly (Phe) synthesis, the acceptance of the cognate ternary complex Phe-tRNA - EF-Tu - GTP (which has the correct anticodon with respect to the codon at the A site) was compared with the competing acceptance of ternary complexes with near-cognate Leu-tRNA(Leu) (which has a similar anticodon) or non-cognate Asp-tRNA(Asp) (which has a dissimilar anticodon), by monitoring the formation of AcPhePhe, AcPheLeu or AcPheAsp, respectively. Cognate (but not near-cognate) occupation of the E site reduced synthesis of the 'wrong' dipeptide AcPheLeu only marginally relative to that of the cognate AcPhe2, whereas the formation of AcPheAsp was decreased as much as 14-fold, thereby reducing it to the background level. It follows that the allosteric interplay between E and A sites, i.e. the low affinity of the A site induced by the occupation of the E site, excludes the interference of non-cognate complexes in the decoding process and thus reduces the number of aminoacyl-tRNA species competing for A site binding by an order of magnitude. Furthermore, near- or noncognate AcPhe-aminoacyl-tRNA molecules readily fall off the A site and can be efficiently chased from the A site by ternary complexes. This process might be related to the 'editing' mechanism proposed previously.