If the prevalence of sick building syndrome (SBS) is estimated before intervention begins, then a reduction in the estimate may later be used to measure success of the intervention, and in particular, those efforts toward improving air quality. However, the measure of success will be distorted if factors other than air quality affect the SBS prevalence estimate. In this study the background prevalence of SBS was estimated and study factors identified that alone affected the estimate. Two symptom questionnaires were randomly administered to workers from 39 offices before routine physical examinations; one questionnaire described the SBS study, the other did not. SBS was defined as a symptom in the prior 24-hour or 7-day recall period that was more severe at work and not related to suspected confounders--allergy, cold, flu. Prevalence and prevalence ratios were estimated along with 95% confidence intervals (CI). Symptoms were reported by 45% of 1088 workers surveyed, but most reported them as more severe outside work or related them to confounders. SBS prevalence was 5%. It was 3.2 times higher (95% CI: 1.8, 5.7) among workers cognizant of the study relative to those blinded, 2.2 times higher (95% CI: 1.2, 4.1) for the 7-day relative to the 24-hour recall period, and 2.5 times higher (95% CI: 1.4, 5.0) for females. SBS prevalence did not differ by workday or age. Since study factors alone affected prevalence estimates, a standardized assessment method seems necessary for SBS.