The objective of this study was to examine the spatial genetic relationships of the Lake Qinghai scaleless carp Gymnocypris przewalskii within the Lake Qinghai system, determining whether genetic evidence supports the current taxonomy of Gymnocypris przewalskii przewalskii and Gymnocypris przewalskii ganzihonensis and whether Gymnocypris przewalskii przewalskii are returning to their natal rivers to spawn. Comparison of mitochondrial (control region) variation (42 haplotypes in 203 fish) of G. przewalskii with the postulated ancestral species found in the Yellow River, Gymnocypris eckloni (10 haplotypes in 23 fish), indicated no haplotype sharing, but incomplete lineage sorting. Consistent with the sub-species status, an AMOVA indicated that the Ganzi River population was significantly different from all other river populations (FST = 0·1671, P < 0·001). No genetic structure was found among the other rivers in the Lake Qinghai catchment. An AMOVA of amplified fragment length polymorphism (AFLP) loci, however, revealed significant genetic differences between most spawning populations (FST = 0·0721, P < 0·001). Both mitochondrial and AFLP data found significant differences among G. p. przewalskii, G. p. ganzihonensis and G. eckloni (FST values of 0·1959 and 0·1431, respectively, P < 0·001). Consistent with the incomplete lineage sorting, Structure analysis of AFLP loci showed evidence of five clusters. One cluster is shared among all sample locations, one is unique to G. p. ganzihonensis and G. eckloni, and the others are mostly found in G. p. przewalskii. Genetic evidence therefore supports the current taxonomy, including the sub-species status of G. p. ganzihonensis, and is consistent with natal homing of most Lake Qinghai populations. These findings have significant implications for the conservation and management of this unique and threatened species. The evidence suggests that G. p. przewalskii should be treated as a single population for conservation purposes. Exchangeability of the populations, however, should not be used to promote homogenization of fish spawning in the different rivers. As some degree of genetic divergence was detected in this study, it is recommended that the spawning groups be treated as separate management units.