We aimed to determine the relative effectiveness of an education intervention and a threat-of-enforcement intervention in reducing sales of cigarettes to under-age youth by randomly allocating 300 retailers in a nonmetropolitan region of New South Wales to: a control group with no intervention; a minimal-intervention group, which received an educational letter; and a maximal-intervention group, which received a threat of enforcement followed by a visit from a public health officer. Retailers were checked for compliance at pretest and post-test, six months apart, by twelve 18-year-olds who were judged by independent raters to look younger. The retailers were surveyed by telephone at both times for knowledge, attitudes and self-reported sales practices. Neither intervention achieved significant improvements for the two key behavioural outcomes: requiring proof of age and display of a warning sign. Neither was there an intervention effect on knowledge about the law. The greatest improvement in the proportion of retailers who believed that the legal age should be 18 or over was in the minimal-intervention group, and both intervention groups were less likely than the control group at post-test to think that it was acceptable to sell to a person who was nearly 18. There was poor overall compliance with the revised legislation at pre-test. The finding of a pretest-to-post-test improvement but no differential intervention effect highlights the methodological difficulties of such research. The interventions may, however, have been partly successful in modifying the attitudes of retailers.