Background: Childhood trauma has been associated with a more severe clinical expression of bipolar disorder (BD). However, the results that specifically associated traumatic events and psychotic features in BD have been inconsistent, possibly due to the low resolution of the phenotypes being used.Methods: 270 normothymic patients with BD completed the Childhood Trauma Questionnaire (CTQ) and the Peters Delusion Inventory (PDI) that assessed 21 delusional beliefs. Patients were characterized for the lifetime presence of psychotic features during episodes and cannabis misuse in accordance with DSM-IV. We performed a series of path analyses to investigate the links from three types of childhood abuse (physical, sexual and emotional) directly to delusional beliefs and psychotic features, and indirectly through cannabis misuse.Results: A first path analysis showed no link between any of the childhood abuse types and psychotic features when only a categorical definition of psychosis was used. When incorporating the quasi-dimensional measure of delusional beliefs in a second path analysis, we found that emotional and physical abuse and cannabis misuse were each directly associated with PDI score. PDI score and psychotic features were strongly correlated. Childhood abuse did not operate through cannabis misuse to increase delusional beliefs. Including type of BD in the model did not alter the results.Conclusion: Emotional and physical abuse, but also cannabis misuse, increased delusional beliefs in patients with BD. Using a quasi-dimensional measure of psychotic symptoms in BD provided higher resolution of the psychosis phenotype and helped reconcile ambiguous findings from previous studies.