Quasi-experimental, nonrandomized, nonequivalent, parallel group-controlled study involving before and after telephone surveys of the general population. To measure the magnitude of any sustained change in population beliefs about back pain 3 years after cessation of a population-based intervention designed to alter beliefs about back pain. A media campaign, which ran between September 1997 and December 1999 in Victoria, Australia, provided simple evidence-based advice about back pain. Following the campaign, there were significant improvements in both community and physicians' beliefs about back pain, as well as a decline in number of workers' compensation back claims. The campaign's impact on population beliefs about back pain and fear-avoidance beliefs was measured using telephone surveys. Demographically identical population groups in Victoria and the adjacent control state, New South Wales, were surveyed at 4 times: before, during, immediately after, and 3 years after the media campaign in Victoria (surveys 1-4, respectively). Back beliefs were measured using the Back Beliefs Questionnaire (BBQ) (possible score 9-45, for which a higher score indicates a more positive belief about low back trouble). Fear-avoidance beliefs were measured using the Fear-Avoidance Beliefs Questionnaire consisting of 2 subscales for physical activity and work activity (possible scores 0-24 and 0-36, respectively), for which a lower score indicates fewer fear-avoidance beliefs. A total of 6230 individuals in the general population completed the surveys. Large statistically significant improvements in population beliefs about back pain were still observed in Victoria 3 years after cessation of the campaign (mean scores on the BBQ were 26.5, 28.4, 29.7, and 28.8 for surveys 1-4, respectively). No changes were observed in New South Wales (mean scores on the BBQ were 26.3, 26.2, 26.3, and 26.1 for surveys 1-4, respectively). Of those individuals who reported back pain during the previous year, fear-avoidance beliefs about physical activity in Victoria were also significantly better than at baseline (mean scores on the Fear-Avoidance Beliefs Questionnaire for physical activity were 14.0, 12.5, 11.6, and 12.3 for surveys 1-4, respectively). Significant sustained improvements in population beliefs about back pain were observed 3 years after cessation of a media campaign of provision of positive messages about back pain. This result provides further evidence that a primary preventive strategy of altering population beliefs about back pain may be a highly effective way for reducing back-related disability.