Introduction Since many studies in epidemiology rely on self-rated health (SRH) as a generic measure of health status, studying this indicator is highly relevant for epidemiological research. Problematically, due to SRH's vagueness and the great room for interpretation for respondents, researchers usually do not know what exactly they measure when using this indicator. Firstly, if SRH, as studies suggest, is based on diverging aspects in different social groups (e.g., age-groups or cohorts), interpretations of respective group-differences are called into question. If, for example, younger respondents lay more weight on mental health than older respondents in rating their health, a comparison of SRH for these groups is disputable. Secondly, it is still unclear whether a (lack of) change in subjective health is due to real changes in health or changes in health aspirations. Accordingly, respondents might provide stable health-ratings despite deteriorating health because of lowered standards or changing frames of reference due to older age. For both reasons, more research on group-differences in response behavior and the sensitivity of SRH to changes in health is needed. Method In order to shed light on these issues, this paper analyzes data from the Canadian National Population Health Survey (NPHS). In the NPHS, around 17,000 people have been extensively surveyed biannually from 1994/1995 till 2012 regarding their health. This allows for a detailed analysis of which health-indicators and health-domains respondents use in rating their health and how reported changes in health over time affect changes in SRH. Utilizing dominance analyses, the contribution of five health-domains to R2 in explaining SRH are quantified: functioning, diseases, pain, depression, and risk-behavior. Employing both cross-sectional (OLS-regression models) and longitudinal methods (FE-regression models), this paper analyzes which health-domains are relevant for both SRH at a point in time as well as for their changes in SRH over time. Results Both cross-sectionally (R2 = 0.32 for women and 0.27 for men) and longitudinally (R2 = 0.12 for women and 0.11 for men), functioning was the most important health domain in explaining SRH, followed by diseases and pain. Depression and risk-behavior only played a minor role in evaluating one's current health status as well as changes. There were no meaningful differences in this ranking by gender. However, separate cross-sectional analyses by birth cohort show clear disparities in that functioning and diseases steadily increase in relevance for explaining SRH for older cohorts. The same is true, albeit to a lesser extent, for assessing changes in health. Conclusions This paper shows that there is a clear pattern in how respondents rate their health and that this pattern applies both to rating one's health at a point in time as well as to evaluating changes in health. However, the analyses also show that there are marked differences in how birth cohorts rate their health, highlighting the need to take these differences in rating behavior into account. The results are of great relevance to meaningfully measure generic health status via SRH and thus analyze health trajectories and health inequality.