Genetic diversity amongst 76 plantain landraces has been studied using RAPD analysis at two levels of intensity and compared with groupings based on phenotypic indices and morphotype. There was a good correlation (R²=0.78) between estimates of genetic diversity based on 76 RAPD bands and 164 RAPD bands. However, there was a poor correlation between RAPD-based estimates of genetic diversity and a phenotypic index based on agronomic characters. There was also a poor correlation between RAPD analyses and morphotype group (based on bunch type and stature). These results suggest that the traditional designations of plantain landraces based on morphotype do not provide a true reflection of overall genetic divergence. Similarly, classification systems using phenotypic indices based on agronomic characters may not provide accurate taxonomic differentiation. The level of genetic divergence within morphogroups based on bunch type suggests that True Horn plantains are derived from False Horn plantains which in turn are derived from French plantains. Genetic divergence was found to be generally quite low within the plantain landrace genepool, which is consistent with the proposed evolution of this germplasm through somatic mutation of a relatively small number of introductions. However, putative synonyms/duplicates have been shown to be genetically distinct. In contrast, a group of 12 landraces have been identified that are highly distinct from one another (showing 20–35% dissimilarity). Fertile members of this group may be useful for generating genetically diverse 2x and 4x breeding populations that can be used in breeding secondary triploid hybrid plantain varieties.