The joint development of polymorphic molecular markers and paternity analysis methods provides new approaches to investigate ongoing patterns of pollen flow in natural plant populations. However, paternity studies are hindered by false paternity assignment and the nondetection of true fathers. To gauge the risk of these two types of errors, we performed a simulation study to investigate the impact on paternity analysis of: (i) the assumed values for the size of the breeding male population (NBMP), and (ii) the rate of scoring error in genotype assessment. Our simulations were based on microsatellite data obtained from a natural population of the entomophilous wild service tree, Sorbus torminalis (L.) Crantz. We show that an accurate estimate of NBMP is required to minimize both types of errors, and we assess the reliability of a technique used to estimate NBMP based on parent–offspring genetic data. We then show that scoring errors in genotype assessment only slightly affect the assessment of paternity relationships, and conclude that it is generally better to neglect the scoring error rate in paternity analyses within a nonisolated population.