Correlates of self-rated health among a randomly selected sample of 1863 Australian-born women 45-55 years of age were examined in two logistic regression analyses: one comparing a self-rated health of worse than one's peers with a self-rated health the same as one's peers; and, one comparing a self-rated health of better than one's peers with a self-rated health the same as one's peers. The final model for worse health was largely a reflection of the physical experience of ill health while that for better health was a more complex construct including not only the absence of illness but also markers of sociodemographic advantage and self-image. The two models had only three variables in common. Notably, the relationship between the outcome measures and one common variable, body mass index, differed markedly. It is suggested that previous analyses of self-rated health have had their power to adequately describe correlates and determinants of health status constrained. By assuming that the various self-rated health states are part of a continuum and employing statistical methods consistent with that assumption, previous studies have been unable to demonstrate the discontinuity among such states. In particular, it is suggested that self-rated health is at least in part a reflection of social role and as there is no basis for assuming that such roles form a continuum as the use of correlation-based analyses imply, then such analyses are inappropriate.