While knowledge of the degree of inter-population genetic differentiation underlies the understanding of microevolutionary processes in any organism, its calculation through G ST , F ST , or h (which, unlike the previous two, was designed to correct for unequal and small sample sizes) is often based in severely restricted intra-population samples, which are nonetheless tacitly assumed adequate to their accurate estimation. Empirical assessment of the influence of the number and intra-population distribution of samples on the values of G ST and h for several Canarian endemic plants compellingly suggests that (1) contrary to expectations based on simulated datasets, h does not account for empirical sampling bias better than G ST ; (2) sample sizes being equal, collections scattered across each population's occupancy area entail significantly lower over-estimates of G ST and h than if they only consider one of the population extremes, especially in narrow allogamous taxa with small populations; (3) in small samples, a scattered sampling strategy is significantly less sensitive to G ST inflation than sampling in one of the population extremes; and (4) a software-related component of bias should be considered when pooling values of G ST from different studies to calculate averages. Thus, unlike the sampling methods used for many plant endemics from the Canaries and other regions, collections for a reliable estimation of inter-population differentiation using molecular markers should encompass the whole occupancy area of each population, and include a higher proportion of individuals respect to the total size in narrow endemics than in widespread congeners. Critically, the high average allozyme inter-population differentiation reported for the Canarian endemic Flora is possibly an over-estimate, and could be explained predominantly by the generally biased intrapopulation sampling associated with G ST estimates, rather than by specific factors of insularity that restrict gene flow radically, as it has been hitherto assumed.