Macrolide antibiotics bind to the large subunit of procaryotic ribosomes and perturb protein synthesis. There are two competing models to explain this perturbation: (1) shortly after initiation of the polypeptide chain, peptide bond formation and/or translocation is inhibited by the presence of macrolides that are bound in the ribosome 'tunnel' through which the nascent peptide travels; (2) bound macrolides loosen the interaction between the ribosome and peptidyl-tRNA, which therefore, dissociates with a higher probability. The former view cannot easily explain the observed enhancement by macrolides of the dissociation of peptidyl-tRNAs from ribosomes, while the latter view is consistent with the available data. Peptidyl-tRNAs are bound to ribosomes through non-specific and decoding-specific interactions. If macrolides preferentially weaken the non-specific interactions, a greater fraction of the binding energy will be due to decoding-specific interactions and better discrimination between erroneous and correct peptidyl-tRNAs should result. This idea has been tested with low doses of erythromycin, which was observed to counteract the error-inducing effects of streptomycin and of ethanol on the synthesis of beta-galactosidase by Escherichia coli. A specific error near the C-terminus of the enzyme was also responsive to this effect of erythromycin, which therefore must have exerted its influence long after the initiation of the polypeptide synthesis. These results are more easily explained by the idea that the primary mechanism of inhibition of protein synthesis by macrolides is to stimulate the dissociation of peptidyl-tRNA from the ribosome.