Objective: to summarise the available published and unpublished studies on traditional birth attendant (TBA) training effectiveness. Design: a meta-analysis. Study sample: sixty studies (n=60) spanning 1971-1999 from 24 countries and three regions. Measurements: the effect size index, Cohen's h for each outcome; the variance-weighted mean effect size and 95% confidence interval for sub-group of outcomes; homogeneity tests on the distribution of the weighted mean effect sizes; and sensitivity analysis to detect the presence of publication bias. Findings: TBA training was associated with significant increases in attributes such as TBA 'knowledge' (90%), 'attitude' (74%), 'behaviour' (63%) and 'advice' (90%) over the untrained TBA baseline. Results for 'behaviour' and 'advice' in specific content areas related to peri-neonatal health outcome, however, reveal sources of variability and underscore the conflicting evidence on TBA training. TBA training was also associated with small but significant decreases in peri-neonatal mortality (8%) and birth asphyxia mortality (11%). Incomplete reporting limited the assessment of neonatal mortality due to tetanus and acute respiratory infection, maternal mortality, as well as assessment of the relationship between intervention characteristics and outcomes. The quality of studies included in the meta-analysis lack sufficient rigour to address the question of causality. Thus, while the data suggest that TBA training is effective in terms of the outcomes measured, we are unable to demonstrate that it is a cost-effective intervention. Implications: skilled attendance at birth is a distant reality in many developing countries and effective community-based strategies are needed to help reduce high levels of mortality. Given the magnitude of peri-neonatal mortality, the associations observed between TBA training peri-neonatal and birth asphyxia mortality, and TBA attributes in content relevant to peri-neonatal survival, we suggest that these strategies may usefully include TBA training in appropriate settings. If TBAs are to be trained, however, it is imperative that their training be adequately evaluated in order to develop the strong evidence base that is lacking to-date and that is necessary for sound policy and programming.