New Guinea is considered the most important secondary centre of diversity for sweet potato (Ipomoea batatas). We analysed nuclear and chloroplast genetic diversity of 417 New Guinea sweet potato landraces, representing agro-morphological diversity collected throughout the island, and compared this diversity with that in tropical America. The molecular data reveal moderate diversity across all accessions analysed, lower than that found in tropical America. Nuclear data confirm previous results, suggesting that New Guinea landraces are principally derived from the Northern neotropical genepool (Camote and Batata lines, from the Caribbean and Central America). However, chloroplast data suggest that South American clones (early Kumara line clones or, more probably, later reintroductions) were also introduced into New Guinea and then recombined with existing genotypes. The frequency distribution of pairwise distances between New Guinea landraces suggests that sexual reproduction, rather than somaclonal variation, has played a predominant role in the diversification of sweet potato. The frequent incorporation of plants issued from true seed by farmers, and the geographical and cultural barriers constraining crop diffusion in this topographically and linguistically heterogeneous island, has led to the accumulation of an impressive number of variants. As the diversification of sweet potato in New Guinea is primarily the result of farmers' management of the reproductive biology of their crop, we argue that on-farm conservation programmes that implement distribution of core samples (clones representing the useful diversity of the species) and promote on-farm selection of locally adapted variants may allow local communities to fashion relatively autonomous strategies for coping with ongoing global change. (Résumé d'auteur)