OBJECTIVE To examine how socioeconomic position influences physical and mental health dynamics. METHODS The Whitehall II study of civil servants collected questionnaires on six occasions from 1991/93 to 2006/07. Civil service grade measured socioeconomic position and Short Form 36 General Health Survey component scores rated physical and mental health. Bivariate growth curve models of physical and mental health over 15 years were estimated for high, medium, and low grades (n = 8309). RESULTS At baseline, levels of physical and mental health were correlated for participants in low grades only. Among study participants in medium grades, mental health was maintained over time, even as physical health decreased. Restoring mental health after a negative response to poor physical health was more difficult for some in low grades. Recovery from downturns in physical health associated with poorer mental health also depended on better socioeconomic circumstances There was greater variability in baseline levels and rates of change in the mental and physical health of those in lower grades compared with higher grades. CONCLUSIONS Homeostatic mechanisms may vary by socioeconomic position. The greater variability of change in health function for those in lower grades implies considerable scope for improvement if sources of variation in health within disadvantaged groups that are amenable to intervention can be identified.