The identification of the mechanisms underlying co-occurrence patterns of species is a way to identify which processes (niche, neutral or both) structure metacommunities. In this paper, our goals are to identify patterns of co-occurrence in neotropical stream fish and determine which processes structure the metacommunity and the gradients that underlie this structure.  Our results pointed out that the metacommunity formed by the total pool of species is structured by a nested pattern (Hyperdispersed Species Loss) of co-occurrence and the mass effect mechanism. On the other hand, a set of core species displays a Clementisian pattern and is structured by the species sorting mechanism. Both, hyperdispersed species loss and the Clementisian patterns point to a discrete set of communities in the metacommunity. These communities could be isolated by physicochemical conditions, or physical barriers, like dams or waterfalls.

We used a data set formed by 66 streams , composed by a pool of 143 species and richness varying from 1 to 27 species. Stream data entered the database when meet the following criteria : (i) the collection was carried out in a 1^st – 3^rd order stream; (ii) the stream stretch was georeferenced; (iii) fish sampling was carried out by electric fishing, trawl nets, and/or hand nets; (iv) a minimum of 50-meter stretch was sampled by stream; (v) a single sample site per stream; (vi) there was a stream species list; (vii) environmental variables (turbidity, conductivity, pH, dissolved oxygen, water velocity, stream channel depth and width) were measured in situ at three or more points in each stream; (viii) fish sampling, transport and preservation of sampled specimens were authorized by the Sistema de Autorização e Informação em Biodiversidade (SISBIO),  Instituto Chico Mendes de Conservação da Biodiversidade, Ministerio do Meio Ambiente and (ix) all streams was located in the Cerrado Biome. Constraining the data for streams under the same vegetation condition ensure that the features linked to vegetation type, like substrate and temperature, are not so different decreasing the amount of variability.