This study provides empirical information on intrapopulation gene dispersal via pollen, the size of genetic ‘neighbourhoods’, and interpopulation gene flow in a long-lived conifer, limber pine (Pinus flexilis). We used allozyme loci for a paternity analysis of 518 seeds produced in an isolated population of limber pine located in north-eastern Colorado, U.S.A., separated by 2 km from the nearest conspecific trees and nearly 100 km from populations in the Rocky Mountains. We also used indirect techniques (FST analyses) to estimate gene flow rates among subdivisions of the study population and among five widely separated populations. Within the main study population limber pine exhibited a polymorphism level of 50%, observed heterozygosity of 0.159, and 2.36 alleles per polymorphic locus. Mountain populations were slightly more variable. The main study population showed significant differentiation in allozymes among neighbouring subpopulations. The mean FST was 0.031 and the gene flow rate among subpopulations was estimated as 7.8 migrants per generation. Among widely separated populations the mean FST was 0.035 and the gene flow rate was estimated as 6.9 migrants per generation. The paternity analysis indicated a best estimate of 6.5% pollen immigration (minimum 1.1%) from populations 2 km to 100+ km away. For 4% of the seeds examined, paternity could be ascribed to a single tree in the study population. Fractional paternity and likelihood methods were used to estimate pollen dispersal distances for the remainder of the seeds. Mean pollen dispersal distance was estimated at 140 m using the fractional method, similar to results from the other techniques. This compares with a mean distance of 172 m between potential mates. These results suggest near-panmictic pollen dispersal over this population, which covers about 15 ha. The observed allozyme differences and surprisingly low estimates of among-subpopulation gene flow are ascribed to a probable restriction of gene dispersal by seeds.