A quasi-experimental research design is used to evaluate Carolina Healthstyle, a health promotion project for South Carolina state employees. A 10% stratified random sample of employees was surveyed in the Spring of 1983 and again in 1984. Eighteen agencies were intervention agencies and the rest comparison that year. This article reports changes with simple before-after comparisons in the intervention agencies and matched pair analysis and randomized block designs to compare intervention and comparison agencies. Results are reported for smoking, exercise safety, nutrition, stress, and alcohol. Significant increases in exercise were found in both intervention and comparison agencies. The number of smokers decreased in intervention from 30%-26% with no change in comparison agencies. Safety practices changed at similar rates in both intervention and comparison groups. Consumption of chicken increased significantly only in the intervention agencies. Few other diet or stress changes were found. There were changes in alcohol consumption in intervention agencies only. The presence of the comparison group helps to separate the program effects from secular trends.