This paper reports the outcome of a 17-month follow-up of structured, community-based, offence-focused, intervention programmes designed to reduce rates of re-conviction amongst adjudicated offenders under probation supervision. Three separate programmes were examined, all derived from a cognitive social learning model of risk factors for repeated involvement in crime. Using a quasi-experimental design, the study compared male offenders who had completed programmes (n∈=∈215) with a non-completion group (n∈=∈181), a group allocated to programmes but who had not commenced them (n∈=∈339), and a control sample (n∈=∈194) not allocated to the programmes. Outcome data analysis employed (a) an "intent to treat" between-group comparison, (b) "treatment received" methodology. In order to take account of selection bias, data were further analysed using instrumental variables and propensity scores; results suggested a possible treatment effect for moderate and higher-risk cases. Factors influencing different interpretations of these findings were considered.