Human respiratory syncytial virus (RSV) is the most important viral cause of serious pediatric respiratory illness worldwide. Currently, the most promising live-attenuated vaccine candidate is a temperature-sensitive (ts) cDNA-derived virus named rA2cp248/404/1030ΔSH, in reference to its set of attenuating mutations. In a previous clinical study, more than one-third of postvaccination nasal wash isolates exhibited partial loss of the ts phenotype. Most of this instability appeared to be due to reversion at a missense point mutation called 1030. This 1030 mutation is a single-nucleotide tyrosine-to-asparagine substitution at position 1321 (Y1321N) of the polymerase L protein that contributes to the ts and attenuation phenotypes of the vaccine candidate. The goals of the present study were to identify a reversion-resistant codon at position 1321 conferring a comparable level of attenuation and to use this to develop a genetically stable version of the vaccine virus. We modified wild-type (wt) RSV to insert each of the 20 possible amino acids at position 1321; 19 viruses were recoverable. We also investigated small deletions at or near this position, but these viruses were not recoverable. Phenotypic analysis identified alternative attenuating amino acids for position 1321. Several of these amino acids were predicted, based on the genetic code, to be refractory to deattenuation. Classical genetics, using temperature stress tests in vitro combined with nucleotide sequencing, confirmed this stability but identified a second site with a compensatory mutation at position 1313. It was possible to stabilize the 1313 site as well, providing a stable 1030 mutation. Further stress tests identified additional incidental mutations, but these did not reverse the ts/attenuation phenotype. An improved version of the vaccine candidate virus was constructed and validated in vitro by temperature stress tests and in vivo by evaluation of attenuation in seronegative chimpanzees. In addition to developing an improved version of this promising live-attenuated RSV vaccine candidate, this study demonstrated the propensity of an RNA virus to escape from attenuation but also showed that, through systematic analysis, genetics can be used to cut off the routes of escape.