Problem based learning (PBL) arguably represents the most significant development in education over the past five decades. It has been promoted as the curriculum of choice, and since its introduction in the 1960's, has been widely adopted by many medical and dental schools. PBL has been the subject of much published literature but ironically, very little high quality evidence exists to advocate its efficacy and subsequently justify the widespread curriculum change. The purpose of this review is to classify and interpret the available evidence and extract relevant conclusions. In addition, it is the intent to propose recommendations regarding the relative benefits of PBL compared with conventional teaching. The literature was searched using PubMed, ERIC and PsycLIT. Further articles were retrieved from the reference lists of selected papers. Articles were chosen and included according to specific selection criteria. Studies were further classified as randomised controlled trials (RCTs) or comparative studies. These studies were then analysed according to intervention type: whole curricula comparisons and single educational interventions of shorter duration. At the level of RCTs and comparative studies (whole curricula), no clear difference was observed between PBL and conventional teaching. Paradoxically, it was only comparative studies of single PBL intervention in a traditional curriculum that yielded results that were consistently in favour of PBL. Further research is needed to investigate the possibility that multiple PBL interventions in a traditional curriculum could be more effective than an exclusively PBL programme. In addition, it is important to address the potential benefits of PBL in relation to life-long learning of health care professionals.