A field trial was conducted to evaluate the reliability of four widely used health risk appraisals (HRAs). A total of 338 randomly selected adults aged 25 to 65 years residing in the Boston metropolitan area completed an HRA on two occasions seven to 12 weeks apart. Test-retest reliability was assessed by comparing self-reported baseline risk scores to follow-up scores for heart attack risk and selected coronary heart disease risk factors. Respondents generally gave consistent reports for family history, cigarette smoking, and relative weight (test-retest r greater than .75), but self-reported scores for physiologic status (blood pressure and cholesterol) and lifestyle (diet, physical activity, and stress) were much less stable. Coefficients for heart attack risk and appraised age ranged from .43 to .87 for the four HRAs. The reliabilities of two self-scored instruments were greatly reduced by mathematical errors made by respondents when computing their heart attack risk scores. These results were not affected by the length of the follow-up period (seven to 12 weeks).