Since the 1980s there has been a decline or stagnation in public expenditure on agriculture in most developing countries (Akroyd & Smith, 2007). Likewise, the proportion of official development assistance (ODA) going to agriculture is estimated to have declined from around 20 per cent in 1979 to a low of 3.7 per cent in 2006, and has remained around 5 per cent since (Cabral & Howell, 2012). As noted in the World Development Report on Agriculture, "extension services, after a period of neglect, are now back on the development agenda... [but] more evaluation, learning, and knowledge sharing are required to capitalize on this renewed momentum" (World Bank, 2007, p. 175). Poverty reduction strategies in 24 African countries also listed extension as a top agricultural priority (InterAcademy Council, 2004; cited in Davis, 2006). Nevertheless, age-old questions in agriculture remain, including how to raise yields and farmer incomes, how to ensure environmentally sustainable development, and how to empower the poorest farmers and particular groups such as women farmers in developing skills in adoption and resilience to shocks. There is increasing criticism as to whether extension services are capable of achieving these broad objectives, or whether a more intensive approach is required such as that provided by the farmer field school initiative. Farmer field schools (FFS) are a common approach used to transfer specialist knowledge, promote skills and empower farmers around the world. At least 10 million farmers in 90 countries have attended such schools. FFS are implemented by facilitators using participatory "discovery-based" learning based on adult education principles. Many different implementing bodies have been involved. Field schools have a range of objectives, including tackling overuse of pesticides and other harmful practices, improving agricultural and environmental outcomes, and empowering disadvantaged farmers such as women. The authors conducted a systematic review of evidence on FFS implementation to investigate whether FFS make a difference, to which farmers, and why or why not. The authors synthesised quantitative evidence on intervention effects using statistical meta-analysis, and qualitative evidence on the barriers and enablers of effectiveness using a theory of change framework. The results of statistical meta-analysis provide evidence that FFS are beneficial in improving intermediate outcomes relating to knowledge learned and adoption of beneficial practices, as well as final outcomes relating to agricultural production and farmers' incomes. The findings suggest this to be the case for FFS promoting integrated pest management (IPM) technology, as well as other techniques. However, the rigorous impact evaluation evidence base is small and there are no studies that the authors were able to identify as having a low risk of bias. There is no evidence that neighbouring non-participant farmers benefit from diffusion of IPM knowledge from FFS participants. Therefore, they do not experience improvements in IPM adoption and agriculture outcomes. The evidence of positive effects on agricultural outcomes is largely limited to short-term evaluations of pilot programmes. In the few examples where FFS have been scaled up, the evidence does not suggest they have been effective in improving agricultural outcomes among participating farmers or neighbouring non-participants. Although empowerment is a major objective of many FFS, very few studies have collected information on this outcome in a rigorous manner. A few studies suggest farmers feel greater self-confidence. What explains the lack of scalable effects among FFS participants, or diffusion of IPM practices among the community? FFS differ from standard agricultural extension interventions, which tend to focus on disseminating knowledge of more simple practices such as application of fertiliser and pesticides, or adoption of improved seeds. The experiential nature of the training, and the need for the benefits of the FFS technology to be observed, are barriers to spontaneous diffusion. Furthermore, the effectiveness of scaled-up interventions has been hampered by problems in recruiting and training appropriate facilitators at scale. The review provides implications for policy, practice and research. The references are organised into the following groups: (1) Included Effectiveness Studies; (2) Included Qualitative Studies; (3) Studies Excluded from Synthesis; and (4) Additional References. Reasons for exclusion of marginal studies are given in Appendix B. Appendices Contain: (1) Global project portfolio review; (2) Example search strategy; (3) Record of database searches; (4) Record of internet searches; (5) Journals handsearched; (6) Reasons for exclusion of marginal studies; (7) Data collection codes; (8) Critical appraisal methods; (9) Effect size calculations; (10) Synthetic effect sizes; (11) Meta-analysis of bivariate and partial effect sizes; (12) Included effectiveness study descriptives; (13) Detailed outcomes reported; (14) Results of critical appraisal: impact evaluations; (15) Results of critical appraisal: qualitative evaluations; (16) Meta-analyses by programme name; (17) Meta-analyses including all standard errors corrected for possible unit of analysis errors; (18) Meta-analysis findings: additional analysis; and (19) Descriptive synthesis of findings from qualitative studies.