The chemotaxonomic relationships between Coffea (subgenus Coffea) species have been poorly studied to date and the compounds tested so far - chlorogenic acids, diterpenoids and purine alkaloids - did not enable the establishment of phylogenetic relationships analogous to those revealed by chloroplast and nuclear DNA studies. In the present study, the relationships between African Coffea species were assessed on the basis of their seed lipid composition. Fatty acids and sterols were determined in 59 genotypes belonging to 17 distinct Coffea species/origins. Principal Component Analysis of fatty acid and sterol data enabled easy identification of the few species for which one or several compounds could serve as a quantitative signature. Hierarchical Clustering classified the Coffea species in seven groups with both fatty acids and sterols. However, while groupings based on seed fatty acid composition showed remarkable ecological and geographical coherence, no phylogeographic explanation was found for the clusters retrieved from sterol data. When compared with previous phylogenetic studies, the groups deduced from seed fatty acid composition were remarkably congruent with the clades inferred from nuclear and plastid DNA sequences.