The impact of different approaches to methadone maintenance on the level of crime committed by heroin addicts was examined in a cohort of addicts entering methadone treatment. The cohort comprises three groups: 72 subjects (group 1) who were approved for treatment and referred to a long-term programme which tolerated continued illicit drug use in treatment; 159 subjects (group 2) who were referred to an abstinence orientated programme from which clients who continued to use heroin were expelled; and 84 subjects who were rejected as unsuitable or failed to complete the assessment process. Official records of convictions were used to calculate conviction rates in the pre- and post-assessment periods. Differences between groups in conviction rates for drug and property crimes were analysed using Poisson regression. Three variables--age, sex and age of first criminal conviction--were significant predictors of conviction rates and all analyses controlled for these variables. Most of the rejected subjects entered treatment during the study period and it was, therefore, not possible to interpret differences between treated and untreated subjects. Among those who entered treatment, property offence rates actually rose, due to a significant increase in conviction rates in group 2. Subjects in group 1 were retained significantly longer in treatment than those in group 2. Among those who remained in treatment less than 12 months, most offences occurred after leaving treatment. When conviction rates were adjusted for time spent in treatment, the differences between the clinics disappeared.(ABSTRACT TRUNCATED AT 250 WORDS)