Health policy-makers in developing countries are often disturbed and to a degree surprised by the phenomenon of the ill travelling past a free or subsidized local public clinic (or other public facility) to get to an alternative source of care at which they often pay a considerable amount for health care. That a person bypasses a facility is almost certainly indicative either of significant problems with the quality of care at the bypassed facility or of significantly better care at the alternative source of care chosen. When it is a poor person choosing to bypass a free public facility and pay for care further away, such action is especially bothersome to public policy-makers. This paper uses a unique data set, with a health facility survey in which all health facilities are identified, surveyed, and located geographically; and a household survey in which a sample of households from the same health district is also both surveyed and located geographically. The data are analyzed to examine patterns of health care choice related to the characteristics and locations of both the facilities and actual and potential clients. Rather than using the distance travelled or some other general choice of type of care variable as the dependent variable, we are able actually to analyze which specific facilities are bypassed and which chosen. The findings are instructive. That bypassing behaviour is not very different across income groups is certainly noteworthy, as is the fact that the more severely ill tend to bypass and to travel further for care than do the less severely ill. In multivariate analysis almost all characteristics of both providers and facilities are found to have the a priori expected relationships to facility choice. Prices tend to deter use, and improved quality of services to increase the likelihood of a facility being chosen. The answer to the bypassing dilemma seems to be for providers to provide as good quality care relative to the money charged (if any), as other, often further away, providers.