Research has shown that video modelling can improve social behaviours in children with ASD. In addition, research in behaviour modelling training from the field of organisational psychology has shown that using a mix of positive and negative exemplars can assist in acquisition and generalisation of a skill. The current study compared the use of one negative and one positive exemplar, with the use of two positive exemplars to determine which combination would result in faster acquisition and/or superior generalisation of a skill. No other studies have examined this with children diagnosed with ASD. Seven children, aged between 5 and 15 years, and diagnosed with ASD participated in a multiple baseline design across children; within child across two modelling conditions; and within each modelling condition across two tasks. In one condition, a participant watched a video containing one exemplar of a model (same sex and of similar age but with normal development) perform a task the wrong way, and one exemplar of the same model perform the same task the right way (wrong/right). In another condition, the participant watched a video containing two different exemplars of the model perform a matched task the right way (right/right). During the intervention, 1 participant refused to watch the videos. For 13 of the 16 tasks, where training was completed, participants either reached criterion or made some gains in acquisition of the social skills. However, for seven of the tasks criterion was not reached. Generally, neither modelling condition was superior in acquisition or generalisation of the targeted social skills. Confounds occurring during the course of the study may have contributed to the equivocal results. For some children with ASD, video modelling in combination with the delivery of preferred reinforcers may be required for successful skill acquisition. Further implications, particularly the potential negative effects of vicarious reinforcement when an observer does not gain reinforcement for imitation are discussed, as are recommendations for future research.