There has been considerable debate over the importance of the health selection hypothesis for explaining social gradients in health. Although studies have argued that it may not be an important explanation of social gradients in health, previous analyses have not estimated, simultaneously, the relative effect of health on changes in social position and of social position on changes in health (social causation). Cross-lagged longitudinal analyses using structural equation models enable the estimation of the relative size of these pathways which would be useful in determining the relative importance of the health selection hypothesis over the social causation hypothesis. Data from four phases of the Whitehall II study (initially consisting of 10,308 men and women aged 35-55 in the British civil service) were collected over a 10 year period. There was no evidence for an effect of mental (GHQ-30 and SF36) or physical health (SF-36) on changes in employment grade. When financial deprivation was used as a measure of social position, there was a significant effect of mental health on changes in social position among men although this health selection effect was over two and a half times smaller than the effect of social position on changes in health. The results suggest that the development of social gradients in health in the Whitehall II study may not be primarily explained in terms of a health selection effect.