Client satisfaction surveys in developing countries are increasingly being promoted as a means of understanding healthcare service quality and the demand for these services. However, concerns have been raised about the reliability of responses in such surveys: for example, 'courtesy bias' may lead clients, especially if interviewed upon exiting clinics, to provide misleadingly favorable responses. This study investigates this and other issues using unusual data from Madagascar in which identical questions about satisfaction with local health care centers were asked in user exit surveys and in a population-based household survey, the latter being less contaminated by courtesy bias as well as changes in provider behavior in response to being observed. The findings suggest that reported satisfaction in exit surveys is biased strongly upward for subjective questions regarding (for example) treatment by staff and consultation quality, but not for relatively objective questions about facility condition and supplies. The surveys do provide useful information on the determinants of consumer satisfaction with various dimensions of provider quality. Still, to obtain reliable estimates of consumer perceptions of health service quality, household-based sampling is strongly preferred.