Two separate meta-analyses were conducted to examine (1) the effects of training programs on the behaviour of direct care staff working with clients with ID who present challenging behaviour problems (predominantly aggressive and violent behaviour), and (2) the effects of staff training on the challenging behaviour of their clients with ID. A 3-level random effects model was used for both meta-analyses to account for both within and between study variance. Results showed that staff training was moderately effective in changing staff behaviour, but no convincing evidence was found for an effect on the reduction of challenging behaviour of persons with ID. The type, content and goal of training did not moderate the effects of staff training, whereas sample and study characteristics (e.g., sex participant or year of publication) did. The way a training program is delivered to staff may be much more important than characteristics of a training. •Staff training for challenging behaviour of clients with ID changes staff behaviour.•No evidence that staff training changes challenging behaviour of clients with ID.•Sample and study characteristics moderated the effects of staff training.