Background and context: the 1997 Safe Motherhood Initiative effectively eliminated support for training traditional birth attendants (TBAs) in safe childbirth. Despite this, TBAs are still active in many countries such as Bangladesh, where 88% of deliveries occur at home. Renewed interest in community-based approaches and the urgent need to improve birth care has necessitated a re-examination of how provider training should be conducted and evaluated. Objective: to demonstrate how a simple evaluation tool can provide a quantitative measure of knowledge acquisition and intended behaviour following a TBA training program. Design: background data were collected from 45 TBAs attending two separate training sessions conducted by Bangladeshi non-governmental organization (NGO) Gonoshasthaya Kendra (GK). A semi- structured survey was conducted before and after each training session to assess the TBAs’ knowledge and reported practices related to home-based management of childbirth. Setting: two training sessions conducted in Vatshala and Sreepur in rural Bangladesh. Participants: 45 active TBAs were recruited for this training evaluation. Findings: there were significant improvements following the training sessions regarding how TBAs reported they would: (a) measure blood loss, (b) handle an apneic newborn, (c) refer women with convulsions and (d) refer women who are bleeding heavily. A greater degree of improvement, and higher scores overall, were observed among TBAs with no prior training and with less birth experience. Key conclusions and recommendations for practice: as the Safe Motherhood community strives to improve safe childbirth care, the quality of care in pregnancy and childbirth for women who rely on less-skilled providers should not be ignored. These communities need assistance from governments and NGOs to help improve the knowledge and skill levels of the providers upon which they depend. Gonoshasthaya Kendra’s extensive efforts to train and involve TBAs, with the aim of improving the quality of care provided to Bangladeshi women, is a good example of how to effectively integrate TBAs into safe motherhood efforts in resource-poor settings. The evaluation methodology described in this paper demonstrates how trainees’ prior experiences and beliefs may affect knowledge acquisition, and highlights the need for more attention to course content and pedagogic style.