In a quasi-experimental study, hospital-treated traffic accident injuries were recorded prospectively for 7 1/2 years in the two Norwegian cities, Harstad and Trondheim. In Harstad the recorded data were used actively in analysis, planning, and implementation of a community-based injury prevention program. Trondheim was the nonequivalent control city. The intervention was divided into three periods, each of 30 months duration. Preventive efforts were implemented to some extent in period 1, increasingly in period 2 and period 3. Traffic safety was promoted in an extensive community program based on the Ottawa charter for health promotion. A 26.6% overall reduction of traffic injury rates was found in Harstad from period 1 to period 3 (p < 0.01), whereas a corresponding significant increase was found in the comparison city. Analysis of data from other sources were not conclusive in supporting the Trondheim data as showing the national trend. Alternative explanations for the injury rate reduction in Harstad were assessed by means of other available relevant data. The exact mechanisms that brought about the reduction of injury rates were hard to elucidate because so many intervention elements were implemented at the same time. It is concluded that at least some of the reduction was due to behavioural and structural changes brought about by health promotion. Important factors for the effect of and participation in the prevention program were local relevance and continuous feedback of accident injury data.