We investigated whether the conventional Spanish translation of the self-rated health survey question helps explain why Latinos' self-rated health is worse than Whites' despite more objective health measures showing them to be as healthy as or healthier than are Whites. We analyzed the relationship between language of interview and self-rated health in the Chicago Community Adult Health Study (2001-2003) and the 2003 Behavioral Risk Factor Surveillance System. Being interviewed in Spanish was associated with significantly higher odds of rating health as fair or poor in both data sets. Moreover, adjusting for language of interview substantially reduced the gap between Whites and Latinos. Spanish-language interviewees were more likely to rate their health as fair (regular in Spanish) than as any other choice, and this preference was strongest when compared with categories representing better health (good, very good, and excellent). Our findings suggest that translation of the English word ''fair'' to regular induces Spanish-language respondents to report poorer health than they would in English. Self-rated health should be interpreted with caution, especially in racial/ethnic comparisons, and research should explore alternative translations.