Although the theoretical effects of a severe reduction in effective population size (i.e. a bottleneck) are well known, relatively few empirical studies of bottlenecks have been based on extensive temporally spaced samples of a population both before and after a bottleneck. Here we describe the results of one such study, utilising the Jenolan Caves (JC) population of the brush-tailed rock-wallaby (Petrogale penicillata). When first sampled in 1985 (n = 20) the JC population comprised ~90 individuals. Subsequently the population crashed, and by 1992 only seven individuals remained. In 1996 the entire population (n = 10) was again sampled. Genetic diversity in the pre- and post-crash JC population was compared using 11 polymorphic microsatellite loci and PCR–SSCP analysis of the mitochondrial DNA control region. Only a single unique control region haplotype was detected in the pre- and post-crash JC population, although variant haplotypes were present in other P. penicillata populations. Of the 35 microsatellite alleles present in the pre-crash population, nine (26%) were lost during the bottleneck. The average number of rare alleles declined by 72%, allelic diversity was reduced by 30% and average heterozygosity declined by 10%. These observations are consistent with theoretical predictions. Additional analyses revealed that a P.�penicillata female at Wombeyan Caves was the only survivor of a 1990/91 reintroduction attempt using animals from JC. Of the microsatellite alleles detected in this female, 21% (4/19) were no longer present in the post-crash JC population. Furthermore, the genetic profiles of animals from the recently discovered Taralga population indicate that they are not derived from JC stock, but represent a threatened remnant of a hitherto undetected natural P. penicillata population.