Prior studies have confirmed a bias against disconfirmatory evidence (BADE) in schizophrenia which has been associated with delusions. However, its role in the pathogenesis of psychosis is yet unclear. The objective was to investigate BADE for the first time in subjects with an at-risk-mental-state for psychosis (ARMS), patients with a first episode of psychosis without antipsychotic treatment (FEP) and healthy controls (HC). A standard BADE test presenting written scenarios was employed. In addition, psychometric rating scales and a neuropsychological test battery were applied. A three-staged image was revealed. FEP-patients showed a significant BADE compared to the other groups. The performance of ARMS-patients lay in between HC and FEP-patients. A trend towards significance became evident for a bias against confirmatory evidence (BACE) in FEP-patients. Results were not attributable to antipsychotic or other medication or depressive symptoms. Correlations with delusions reached medium effect sizes but failed significance after Bonferroni-corrections. These results provide evidence for aberrations in evidence integration in the pathogenesis of psychosis and contribute to our knowledge of metacognitive functioning which can be used for (meta-)cognitive intervention in psychosis.