Developing more effective behavioural interventions requires an understanding of the mechanisms of behaviour change, and methods to rigorously test their theoretical basis. The delivery and theoretical basis of an intervention protocol were assessed in ProActive, a UK trial of an intervention to increase the physical activity of those at risk of Type 2 diabetes (N = 365). In 108 intervention sessions, behaviours of facilitators were mapped to four theories that informed intervention development and behaviours of participants were mapped to 17 theoretical components of these four theories. The theory base of the intervention specified by the protocol was different than that delivered by facilitators, and that received by participants. Of the intervention techniques delivered, 25% were associated with theory of planned behaviour (TPB), 42% with self-regulation theory (SRT), 24% with operant learning theory (OLT) and 9% with relapse prevention theory (RPT). The theoretical classification of participant talk showed a different pattern, with twice the proportion associated with OLT (48%), 21% associated with TPB, 31% with SRT and no talk associated with RPT. This study demonstrates one approach to assessing the extent to which the theories used to guide intervention development account for any changes observed.