Background: Self-Rated Health (SRH) correlates with risk of illness and death. But how are different questions of SRH to be interpreted? Does it matter whether one asks: "How would you assess your general state of health?"(General SRH) or "How would you assess your general state of health compared to persons of your own age?"(Comparative SRH)? Does the context in a questionnaire affect the answers? The aim of this paper is to examine the meaning of two questions on self-rated health, the statistical distribution of the answers, and whether the context of the question in a questionnaire affects the answers. Methods: Statistical and semantic methodologies were used to analyse the answers of two different SRH questions in a cross-sectional survey, the MONICA-project of northern Sweden. Results: The answers from 3504 persons were analysed. The statistical distributions of answers differed. The most common answer to the General SRH was "good", while the most common answer to the Comparative SRH was "similar". The semantic analysis showed that what is assessed in SRH is not health in a medical and lexical sense but fields of association connected to health, for example health behaviour, functional ability, youth, looks, way of life. The meaning and function of the two questions differ - mainly due to the comparing reference in Comparative SRH. The context in the questionnaire may have affected the statistics. Conclusions: Health is primarily assessed in terms of its sense-relations (associations) and Comparative SRH and General SRH contain different information on SRH. Comparative SRH is semantically more distinct. The context of the questions in a questionnaire may affect the way self-rated health questions are answered. Comparative SRH should not be eliminated from use in questionnaires. Its usefulness in clinical encounters should be investigated.