Random amplified polymorphic DNA (RAPD) markers are a useful tool for plant population genetic studies. However, as these markers are dominant, heterozygotes and homozygotes for a particular locus are indistinguishable as they will produce a band of similar amplification intensity. This makes it impossible to calculate allelic frequencies directly from RAPD profile data and limited their usefulness in the survey of genetic differentiation among plant populations. To obviate this problem, a Bayesian hierarchical model was presented in terms of the classical F-statistics for partitioning genetic diversity as assessed with dominant markers; and, through the use of Markov chain Monte Carlo (MCMC) simulation, the posterior distribution of θB, an analogue to FST, can be numerically approximated. Pseudotaxus chienii (Taxaceae) is an endangered conifer endemic to China. In this research, 11 of its relictual populations located in Zhejiang, Jiangxi, Hunan and Guangxi provinces were examined using RAPD markers. The degree of genetic differentiation among these populations was estimated by Bayesian approaches and several conventional methods as well. With a burn-in of 50000 iterations and sampling run of 250000 iterations from which every 50th sample is retained for posterior calculations, the estimate for θB was obtained as 0.5018 and its 95% credible interval is [0.4672, 0.5322]. The GST and ΦST values are 0.5865 and 0.5436, respectively. Interestingly, when using Shannon’s index of phenetic diversity to quantify and further partition the amount of genetic diversity, it was found that 48.39% of the total genetic diversity occurred among populations, very close to the θB estimate. Comparison with the estimates based on RAPD analyses of other conifer species indicated that a significantly high degree of genetic differentiation occurred among populations of P. chienii, possibly because: (1) its natural distribution has long been restricted in disjunct isolates characteristic of a fragmented population; (2) the species is dioecious, tending to grow under dense canopy, which hinders the female cone from being pollinated; and (3) the sample collection covers almost the entire range of the species, a large geographical distance. Mantel test showed that the degree of population genetic differentiation is significantly related with geographic distance (r=0.719, P=0.003). In addition, a low level of RAPD variation was revealed within P. chienii. We propose that the low amount of genetic variation is the result of bottleneck effects as well as genetic drift from small population size. In terms of the above results, the development of conservation strategies for P. chienii was discussed.