The endocellular bacterium Wolbachia manipulates its hosts’ reproduction in an impressive variety of ways: it can induce male killing, feminization, thelytokous parthenogenesis and cytoplasmic incompatibility (CI; embryonic mortality resulting from crosses between Wolbachia-infected males and uninfected females). Based on the distribution of these different phenotypes within Wolbachia phylogeny, it is classically argued that CI, the most widespread of the four, is likely to be ancestral, whereas the three others evolved several times. Such a view seemed acceptable, until Jiggins et al. 1xRecombination confounds interpretations of Wolbachia evolution. Jiggins, F.M. et al. Proc. R. Soc. London B Biol. Sci. 2001; 268: 1423–1427Crossref | Scopus (69)See all References, following Werren and Bartos 2xRecombination in Wolbachia. Werren, J.H. and Bartos, J.D. Curr. Biol. 2001; 11: 431–435Abstract | Full Text | Full Text PDF | PubMed | Scopus (109)See all References, showed that Wolbachia can recombine. As noted by these authors, Wolbachia recombination makes it very tricky to infer the evolutionary history of the four different phenotypes from phylogeny, for two main reasons. First, gene phylogenies can be misleading, because they reflect the history of single genes, not necessarily that of the whole genome: Wolbachia strains appearing very distant in a phylogenetic tree might be closely related in reality. Second, the genetic machinery allowing a given Wolbachia to induce a particular phenotype can move between distant strains, so that distant Wolbachia could induce the same phenotype, although this phenotype might have evolved only once.Similar problems emerge when considering CI. The mechanism underlying this phenomenon is unknown, but the current model is that it occurs when a sort of poison (termed mod, for modification), presumably deposited by Wolbachia in the germline of the males, is not neutralized by its antidote (termed resc, for rescue), presumably produced by Wolbachia in infected eggs, resulting in embryo death. The mod–resc (or poison–antidote) interaction is specific: several different mod–resc pairs (or compatibility types) exist. The evolutionary process involved in the emergence of new compatibility types from ancestral ones is unknown, but a basic prediction is that compatibility types will be more similar between closely related strains than between phylogenetically distant ones. This prediction can be tested through interspecific injection experiments, allowing the confrontation of mod–resc pairs of more or less closely related symbionts in a single host species. Recombination greatly complicates the interpretation of such experiments, for the two reasons presented above: first, strains might be more (or less) distant than inferred from phylogenetic markers and second, the mod–resc genes themselves can move between strains.Finally, recombination might affect our understanding of Wolbachia horizontal transfers (here we refer to the interspecific transfers of Wolbachia bacteria, not to Wolbachia genes), which is mainly based on phylogeny. It is generally observed that Wolbachia and host phylogenies are not congruent: closely related Wolbachia strains can be observed in very distant host species. This pattern is usually interpreted as a result of Wolbachia bacteria being transferred between distant hosts. Recombination suggests an alternative interpretation: if genes can move between different host species through vectors such as viruses, horizontal transfers involving whole and viable bacteria might be much less frequent than generally believed.Now, does recombination throw Wolbachia research into chaos? Not necessarily. Phylogenetic inferences based on unique genes must be considered very cautiously, but the use of several independent markers, sufficiently distant within the genome, should circumvent these difficulties.