BACKGROUND AND OBJECTIVES: Cognitive models of psychosis implicate interpretation biases as one of the mechanisms involved in the formation and maintenance of symptoms. First we measured the strength of association between interpretation biases and psychosis-relevant traits. Next we manipulated these biases and quantified the effects of doing so on psychosis-relevant outcomes. METHODS: Experiment 1 used two measures of interpretation bias in a healthy sample (n = 70). Experiment 2 used a novel cognitive bias modification procedure (CMB-pa) in individuals with moderate trait paranoia (n = 60). RESULTS: Experiment 1 revealed that over a third of the variance in interpretation bias could be explained by the combined effect of trait measures of paranoia/psychosis. In Experiment 2, CBM-pa produced training-congruent changes in the interpretation of new ambiguous information and influenced the interpretation, attribution and distress associated with a real-life social event. LIMITATIONS: The potentially confounding effects of elevated anxiety and depression on interpretation bias and the restricted range of outcome measures to assess the wider effects of CBM-pa. CONCLUSIONS: These studies are consistent with interpretation biases contributing to the maintenance of paranoia. CBM-pa could next be adapted and evaluated to test its efficacy as a therapeutic intervention.