The genetic diversity of cocoa has been studied using morphological, enzymatic and molecular descriptors. It has often proved difficult to obtain a clear classification due to factors including the significant genetic mixing that has occurred over the past three centuries, bias in the samples analysed (e.g. the limited number of Forastero samples collected in Peru for Witches' broom resistance), the uncertain origin of some accessions (e.g. some early studies were based on material believed to be Criollo, but which was in fact of hybrid origin), different populations have been used in different studies. Breeding has been hampered by a lack of knowledge of the genetic diversity and level of heterozygosity of the accessions. The main results of the diversity studies are: - In Forastero populations there is significant diversity between and within populations, with continuous variation between them. The greatest diversity was observed among Ecuadorian LCTEEN populations, and the least among a few populations such as Peruvian NA or GU from French Guiana. However, the Ecuadorian populations studied by Allen were collected from a larger number of trees and from a wider area than those collected by Pound, and very few Colombian and Brazilian samples were used in these studies. - Almost completely homozygous "ancestral" Criollo and Nacional genotypes that were probably at the origin of "modern" Criollo and Nacional varieties were identified. Modern Criollo and Nacional varieties are hybrid types resulting from introgression of a few Lower Amazon Forastero genotypes into ancestral Criollo, and of Trinitario into ancestral Nacional varieties, respectively. - The specificity of some populations or varieties has been recognised, for example wild French Guiana, ancient Criollo and Nacional varieties. The founder effect or refuge areas may be responsible for the differences between these populations. - The narrow genetic base of cocoa genotypes used in breeding programs is well k