The aim of this study was to investigate the responses of biomarkers in feral and caged fish and the capacity of these biomarkers to discriminate contamination levels along a stream located in an agricultural area in Southern Brazil. Specimens of the Neotropical fish, Astyanax altiparanae, were confined for 168h in three lakes along the stream. Additionally, during the weeks of in situ exposure, wild specimens of this species were collected from the same sites. Biochemical biomarkers were analyzed, such as phase I biotransformation enzyme 7-ethoxyresorufin-O-deethylase (EROD) and phase II biotransformation enzyme glutathione S-transferase, and we also determined hepatic and branchial levels of non-protein thiols (NPSH), oxidative damage such as lipid peroxidation (LPO), and acetylcholinesterase (AChE) activity in muscle and brain. Genetic biomarkers such as DNA breaks (comet assay), frequency of micronuclei (MN) and erythrocytic nuclear abnormalities (ENA) were also examined. The results indicate that the most sensitive biomarkers for discriminating contamination levels are DNA breaks, LPO and AChE activity. Similar results were obtained for both caged and feral fish. The biomarkers that reflect the results of cumulative events, such as ENA, were more discriminative for chronically exposed specimens (feral fishes). Analyzing biomarkers using an integrated response index showed that both approaches (using feral and caged A. altiparanae) were effective for discriminating contamination levels along the stream, corroborating the results of chemical analyses for selected pesticides. Taken together, these results highlight the importance of biomarker selection and show that both approaches (caged and feral fish) are satisfactory for evaluating water quality in streams impacted by agricultural activities.