--- - |2+ Abstract Background Religious beliefs and practices have been shown to have some association with desistance from offending, as have cognitive processes, including emotion regulation, self‐control, reasoning, learning, and empathy. Religious behaviours may, however, be moderated or mediated by cognitive abilities, but few studies take this into account. Aim The aim of this study was to assess the role of cognitive ability in any relationship between religious beliefs and behaviours and desistance from substance use among adolescents. Method Data were extracted from the Pathways to Desistance Study dataset (n = 1,354). Religious measures were self‐rated importance of religion, helpfulness of religion in dealing with problems, and attendance at religious services. Cognitive measures were the Wechsler Abbreviated Scale of Intelligence and Trail Making Tests. A count of numbers of mind‐altering substances used constituted the main substance measure. Control variables included sex, age, and age of onset of offending. Results Mixed‐effects longitudinal analyses revealed that religious behaviours did constitute a significant predictor of lower substance use for young people after a criminal conviction, after controlling for changes in employment, social support, and delinquent peer association, but only among those with average or above cognitive function. Conclusions These findings may explain why there have been some discrepancies in previous studies of religion and substance misuse. They may also have important implications for the training of anyone delivering programmes to young people in prison that have faith‐based elements, including community faith group leaders and volunteers. - Criminal Behaviour and Mental Health, Volume 28, Issue 4, Page 350-360, August 2018.