This narrative and meta-analytic review of the effectiveness of traditional birth attendant (TBA) training to improve access to skilled birth attendance for obstetric emergencies produced mixed results. Among 16 studies that fit the inclusion criteria, there is a medium, positive, non-significant association between training and TBA knowledge of risk factors and conditions requiring referral; and small, positive, significant associations between TBA referral behavior and maternal service use. These results cannot be causally attributed to TBA training because of the overall quality of studies; moreover, in several studies TBA training was a component of integrated intervention packages. The effort and expense of more rigorous research focusing on TBA training to improve access to emergency obstetric care are difficult to justify. The referral process is complex; the real effects of TBA training on TBA and maternal behavior are likely to be small; and while the proportion of TBA-attended births worldwide varies, it is, on average, quite low. The behavioral determinants and logistical barriers to care seeking for emergency obstetric care are generally well known. We suggest a more promising research agenda would reposition the questions surrounding referral into a broader ecological perspective.