11Molecular data are now commonly used in taxonomy for delimiting cryptic species. In 12 the case of giraffes, which were treated as a single species (Giraffa camelopardalis) during 13 half of a century, several molecular studies have suggested a splitting into four to seven 14 species, but the criteria applied for taxonomic delimitation were not fully described.
15In this study, we have analysed all multi-locus DNA sequences available for giraffes 16 using multispecies coalescent (MSC: *BEAST, BPP and STACEY), population genetic 17 (STRUCTURE, allelic networks, haplotype network and bootstrapping) and phylogenetic 18 (MrBayes, PhyML, SuperTRI) methods to identify the number of species. Our results show 19 that depending on the method chosen, different taxonomic hypotheses, recognizing from two 20 to six species, can be considered for the genus Giraffa. Our results confirm that MSC methods 21 can lead to taxonomic over-splitting, as they delimit geographic structure rather than species.22 The 3-species hypothesis, which recognizes G. camelopardalis sensu strico, G. giraffa, and G. 23 tippelskirchi, is highly supported by phylogenetic analyses and also corroborated by most 24 population genetic and MSC analyses. The three species show high levels of nucleotide 25 divergence in both nuclear (0.35-0.51 %) and mitochondrial sequences (3-4 %), and they are 26 characterised by 7 to 12 exclusive synapomorphies (ES) detected in nine of the 21 nuclear 27 introns analysed for this study. By contrast, other putative species, such as G. peralta, G. 2 28 reticulata, G. thornicrofti or G. tippelskirchi sensu stricto, do not exhibit any ES in nuclear 29 genes. 30 A robust mito-nuclear conflict was found for the position and monophyly of G. giraffa 31 and G. tippelskirchi, which is explained firstly by a mitochondrial introgression from Masai 32 giraffe to southeastern giraffe during the Pleistocene, and secondly, by gene flow mediated by 33 male dispersal between southern populations (subspecies G.g. giraffa and G.g. angolensis). 34 35 37 38 46 (cytochrome c oxidase subunit 1), have been intensively used for species delimitation [3-4]. 47 However, numerous molecular studies have revealed that the mitochondrial tree may deviate 48 from the species tree. Indeed, the maternal inheritance of the mtDNA genome can be 49 misleading for species delimitation because females and males have generally different 50 dispersal behaviours (female philopatry versus male dispersal) [5,6], and because interspecific 51 hybrid females are generally fertile, whereas hybrid males are often sterile (Haldane's rule), 52 facilitating mitochondrial introgression between closely related species [7-9]. To overcome 3 53 these limitations, most recent taxonomic studies dealing with the delimitation between cryptic 54 mammal species have focused on multi-locus datasets [10-12], as the use of multiple 55 independent DNA markers has been shown to provide a strong and reliable signal for 56 deciphering relationships among closely related taxa [13-14]. However, i...