Résumé : The general objective of this work is to better understand the mechanisms of diversification of the African flora on the Gulf of Guinea islands. We focused on orchids, one of the three predominant plant families on São Tomé and Príncipe. We selected the genus Tridactyle, wich presents a high level of diversity and of endemism in the archipelago.We first redefined the taxonomical framework, as the genus belongs to a clade in which the taxonomical classification did not correspond to the phylogenetic tree obtained by previous studies. To address these classification problems (paraphyly and polyphyly of nominal genera), we firstly used Sanger sequencing to obtain more molecular markers to better estimate the phylogenetic tree of the Tridactyle-Cyrtorchis clade. We used one nuclear marker, the internal transcribed spacer (ITS), and five chloroplastic markers (matK, rps16, trnC-petN intergenic spacer, trnL-trnF intergenic spacer, ycf1). Then, we combined the phylogenetic information to a morphological survey, including as many specimens as possible for the genera concerned. We recircumscribed the previously paraphyletic genus Tridactyle, as well as three other genera within the Tridactyle-Cyrtorchis clade (Rangaeris, Ypsilopus and Podangis), and we described two genera with our collaborators, Aziza and Planetangis. We also described six species new to science, two from East Africa and four from São Tomé and Príncipe. Indeed, the morphological diversity of the Tridactyle in the Gulf of Guinea islands has been misevaluated, such that wrong names have been attributed to species new to science.Once this taxonomic work was achieved, we have focused on two other studies: a biogeographic analysis of Tridactyle, to understand the origin and mechanisms generating its diversity in São Tomé and Príncipe, and a phylogeographic study to analyse the genetic variation and geographical distribution of Tridactyle tridactylites, distributed on the archipelago as well as on the continent. These studies were based on DNA sequence variation of the chloroplast genome and ribosomal DNA genes and the data were generated through Illumina Next Generation Sequencing (NGS), which allowed us to include herbarium specimens for which the classic Sanger method did not give satisfactory results.Our study showed that all Tridactyle species currently found on São Tomé and Príncipe colonised the archipelago independently, and that the current species diversity on the islands is the result of allopatric divergence between the islands and the continent, following island colonisation.- 20 -The intraspecific study revealed a high genetic diversity for Tridactyle tridactylites individuals present on Príncipe, which is not common on oceanic islands, but could be a signal that the island was a refuge for the species during the climatic changes related to ice ages. An approximate Bayesian computation analysis (ABC) of the geographic distribution of genetic variation in Atlantic Central Africa and West Africa favoured a hypothesis of recolonisation of the continent from the island rather than a colonisation of the island from the continent. It is possible that the dust-like seeds used wind currents moving from the islands to Central and West Africa as a mean of travel.The work presented here stresses the importance of conducting such studies on more orchid genera, but also on the two other main families of São Tomé and Príncipe, Rubiaceae and Euphorbiaceae. We made a first step toward a better understanding of the mechanisms of diversifications on the Gulf of Guinea islands, but only with a larger number of studies on diverse families and genera, we could draw more general conclusions about these mechanisms for the flora of the archipelago.With regard to the limitations of our study, we were not able to include all species of the genus, especially species from the Democratic Republic of Congo or East Africa. Sampling in the African rainforests is currently not uniformly carried out, and could be improved. Another way to increase sampling is to use new NGS sequencing methods to extract DNA from herbaria preserved in European Herbarium institutions, and obtain genetic information from the chloroplast and ribosome as we have done, but potentially from low-copy nuclear genes as well.