We chose the Apistogramma agassizii model to test for rapid, potentially sympatric, speciation in Amazonian cichlids. A. agassizii shares many similarities with Haplochromines cichlids which are remarkable examples of adaptive radiations in vertebrates. Like the Haplochromines, A. agassizii is philopatric, and show: an important colour polymorphism with marked sexual dimorphism, a differential partner selection, and parental care. We sampled 1170 individuals of A. agassizii from 36 popsites (smallest collection unit) located in the Loreto region of the Peruvian Amazon. The genetic variability of A. agassizii throughout the study area was studied using 2 mitochondrial markers (COI and Cyt b) and 10 microsatellite loci.The 1170 individuals was genotyped, of which 104 were sequenced for both mitochondrial markers. A total of 44 haplotypes were obtained from the concatenated sequences. The results revealed 3 vicarious "species flocks" (named, Sp1, Sp2 and Sp3) that would have started to diverge from their most recent common ancestor (MRCA) 1.83 Ma ago (calibration from fossils).The degrees of genetic differentiation of the 3 "species flocks" would be fractally organised according to the hierarchy of the hydrographic network. The values of genetic differentiation (FST estimator θ) that we observed within the "species flocks" had never before been observed on such small geographical scales for a fish in the Amazon (Sp1: 0.04 - 0.37, avg. = 0.16 ± 0.06 (σ); Sp2: 0.08 - 0.40, avg. = 0.17 ± 0.09 (σ)). Thus, large rivers within sub-watersheds, such as the Marañon, Ucayali and Amazon, would constitute the hierarchical level of rank 1. They would limit the dispersion of the 3 evidenced "species flocks" Sp1, Sp2, Sp3, by constituting barriers to genes flow and possible allopatric speciation. The micro-watersheds would constitute the lower hierarchical level of rank 2, with genetic differentiations of variable amplitude structuring the species flocks in sympatry. The term "sympatry" being used here in the broad sense, considering that these micro-watersheds are very close to each other and are connected by a main river on the same bank. Finally, within the micro-watersheds, the streams would constitute the hierarchical level of rank 3 in which proto-species (genetic unit whose speciation process may be reversible) would diverge in strict sympatry, between connected streams, or even within the same stream (within the same popsite or collecting site). We have shown that the evolutionary history of the "species flock" Sp1 had been conditioned by multiple hydro-geomorphological events which would have alternately favoured fragmentation or dispersion events inducing mixtures of populations in secondary contact. Mate choice experiments were carried out taking into account the 3 identified hierarchical levels. At hierarchy level 1 (sub-watersheds), females Sp1 and Sp2 preferentially chose males of their own "species flock" (Sp1: p-value = 0.0005; Sp2: p-value = 0.0029). Mate choice experiment for level 2 (micro-watersheds) and 3 (stream) were carried out within the species flock Sp1. We tested the choice of females by offering them a male from their stream (same micro-watersheds) and a male from a stream in another micro-watersheds. For level 3 we tested the choice of females by offering them a male from their stream and a male from another stream belonging to the same micro-watersheds. In both configurations, females generally showed sexual preference according to male origin, but this could not be tested in sufficient depth to corroborate that sexual selection would play a driving role in the process of sympatric speciation.The evolutionary mechanisms highlighted in the Apistogramma agassizii model could also be involved in the diversification of other Cichlid species, or even other groups of organisms in the Amazon, and partly explain the exceptional species diversity in the world's largest watershed.