The retreating edge hypothesis for species responding to climate change predicts severe bottlenecks and eventual extinction. The disjunct Lost Pines population at the westernmost edge of the widespread Pinus taeda range is well suited for testing this prediction. The occurrence of one or more genetic bottlenecks in the Lost Pines population was tested using 34 nuclear microsatellite markers and a control sample from the larger, more continuous east Texas P. taeda forests. The Lost Pines population has undergone drastic contractions in effective population size between 3000 and 30 000 years bp. These results were supported by: (i) detection of transient heterozygosity excess, (ii) a mode-shift indicator of allele frequencies, and (iii) a ratio of allele number to allele size range. No bottleneck was detected for the east Texas control using any of the three methods. The distribution of allele frequencies was skewed for the Lost Pines population compared to the control, indicating a loss of rare alleles. However, allelic diversity was similar between the Lost Pines population and its east Texas control; the mean allele number per locus was 5.29 and 5.38, respectively. It is proposed that the Lost Pines population was the western refugium for P. taeda during Pleistocene glaciation and that East Texas P. taeda forests descended from the bottlenecked Lost Pines population.