Ecological studies of species, such as the stomach content analysis, allow us to recognize different trophic groups, the importance of trophic levels and the interrelationships among species and other members of the community. In this investigation, we studied food habits, feeding variation and trophic relationships of the fishes present in streams of the Corral de San Luis drainage, Tubará, Atlántico Department, a part of the lower Magdalena River Basin in Colombian Caribbean. Fish samples of Awaous banana, Agonostomus monticola, Andinoacara latifrons, Hyphessobrycon proteus, Poecilia gillii, Gobiomorus dormitor and Synbranchus marmoratus were obtained using a seine (2x5 m, mesh 0.5 cm), from November 2012 to October 2013. To analyze their stomach contents, we used numeric (% N), volumetric (% V) and frequency of occurrence (% FO) methods, an emptiness coefficient (C.V), index of food item importance (I.A). Besides, physical and chemical habitat parameters were recorded on site. Information obtained was processed using multivariate statistical analysis, ecological indices, and null models: canonical correspondence analysis (ACC), principal component analysis (ACP), trophic niche amplitude (Shannon-Weaver H´) and trophic overlap (Morisita-Horn). We observed significant differences on food resources consumption (K-W= 20.86; p<0.05) among the studied species. They were classified according to their food habits as omnivores with a tendency towards insectivory (A. monticola H´0.60; A. latifrons H´0.43), herbivores with a tendency towards the consumption of algae (A. banana H´0.50; P. gillii H´0.54) and carnivores with a tendency towards insectivory (H. proteus H´0.23); benthic invertebrates and microalgae were found the most important food sources. A total of 65 food items were identified in this study: 21 for A. banana (2 unique, 19 shared), 40 for A. monticola (21 unique, 19 shared), 19 for A. latifrons (5 unique, 14 shared), 6 for H. proteus (1 unique, 5 shared) and P. gillii with 28 (4 unique, 24 shared). The canonical correspondence analysis showed that water conductivity, salinity and pH were the variables that directly influenced fish community structure at the sampled sites. The null model analyses showed that the group of fishes was significantly segregated (p= 0.001) along the trophic axis, with respect to shared food items, and that the segregation was not influenced or generated by competition. The Morisita-Horn index showed false trophic overlap (similarity of about 80 %) between A. banana and P. gillii. The first component of the PCA analysis was explained mainly by phytoplankton, and component two was correlated with items of animal origin. The fishes associated with PC1 were P. gillii and A. banana, with high ingestion values of microalgae. PC2 was explained by A. monticola with high numbers of food items of animal origin. The group of fishes studied behaved as an assemblage; given that the trophic interrelationships showed false trophic overlap, and that they did not exclude one another from the ecosystems, but instead, used different food resources and different physical spaces within their habitat.

Keywords: Trophic ecology, diet, fish assemblage, Neotropical fishes, diversity, Colombia.