In this study, we examined the molecular evolution of the fusion protein (F) gene in human respiratory syncytial virus subgroup B (HRSV-B). First, we performed time-scale evolution analyses using the Bayesian Markov chain Monte Carlo (MCMC) method. Next, we performed genetic distance, linear B-cell epitope prediction, N-glycosylation, positive/negative selection site, and Bayesian skyline plot analyses. We also constructed a structural model of the F protein and mapped the amino acid substitutions and the predicted B-cell epitopes. The MCMC-constructed phylogenetic tree indicated that the HRSV F gene diverged from the bovine respiratory syncytial virus gene approximately 580years ago and had a relatively low evolutionary rate (7.14×10−4substitutions/site/year). Furthermore, a common ancestor of HRSV-A and -B diverged approximately 290years ago, while HRSV-B diverged into three clusters for approximately 60years. The genetic similarity of the present strains was very high. Although a maximum of 11 amino acid substitutions were observed in the structural model of the F protein, only one strain possessed an amino acid substitution located within the palivizumab epitope. Four epitopes were predicted, although these did not correspond to the neutralization sites of the F protein including the palivizumab epitope. In addition, five N-glycosylation sites of the present HRSV-B strains were inferred. No positive selection sites were identified; however, many sites were found to be under negative selection. The effective population size of the gene has remained almost constant. On the basis of these results, it can be concluded that the HRSV-B F gene is highly conserved, as is the F protein of HRSV-A. Moreover, our prediction of B-cell epitopes does not show that the palivizumab reaction site may be recognized as an epitope during naturally occurring infections.