Four subsets of apple (Malus Mill.) germplasm representing modern and old cultivars from the repository and apple genetics population of the Horticulture and Food Research Institute of New Zealand Limited were used in this study. A total of 155 genotypes randomly chosen from the four subsets were analyzed for random amplified polymorphic DNA (RAPD) variation. Nine decamer primers generated a total of 43 fragments, 42 of which were polymorphic across the 155 genotypes. Pairwise distances were calculated between germplasm subsets using the distance metric algorithm in S-PLUS, and used to examine intra-and inter-subset variance components by analysis of molecular variation (AMOVAR). A phenogram based on unweighted pair group method with arithmetic average (UPGMA) cluster analysis was constructed from the pairwise distances and a scatter plot was generated from principal coordinate analysis. The AMOVAR showed that most of the variation in the germplasm (94.6%) was found within subsets, suggesting that there is significant variation among the germplasm. The grouping of genotypes based on the phenogram and scatter plot generally did not reflect the pedigree or provenance of the genotypes. It is possible that more RAPD markers are needed for determining genetic relationships in apple germplasm. Nevertheless, the variation observed in the study suggests that the current practice of sublining populations in the first generation to control inbreeding may not be necessary in subsequent generations. If these results are confirmed by fully informative molecular markers, germplasm managers should reassess the structure of their genetics populations. There may be a need to combine sublines in order to capture the maximum genetic diversity available and to streamline breeding efforts.