In this thesis, I use experimental and quasi-experimental data from 25 villages and a total of 1,105 farmers from eastern DRC to investigate the relationship among agricultural training, the adoption of agricultural technologies, crop productivity, and household food insecurity and dietary diversity. I present evidence that contributes to narrow the gap in the literature on the role of input subsidies fostering small-scale farmers' uptake of productivity-enhancing technologies, how farmer field school and farmer-to-farmer trainings affect the adoption of agricultural technologies, how F2F training may reduce the costs of FFS implementation, how adoption materializes on yields of food crops, and how training through the adoption of improved agricultural technologies impacts household food insecurity and the diet diversification of target households. As a complement to econometric evidence and in order to understand the main findings, I also discuss behavioral features and farmer driven initiatives which somehow condition these impacts. Throughout the four main chapters, I identify practical implications that are highly important for the design and implementation of new programs and policies aimed to address agricultural productivity issues and reduce household food insecurity. In Chapter 1 I develop a general introduction to the research which discusses the evolution of agricultural extension in the last few decades, and describe FFS and F2F training methodologies. Chapter 2 provides a detailed description of the project intervention, technologies promoted, research settings and the data collection process. In Chapter 3, I report the results of an experimental study that analyses the impact of one-shot input starter packs on the adoption of productivity-enhancing complementary practices, which have the potential to maximize the impact of starter pack inputs. Additionally, I assess the levels of persistence on farmers’ use of improved crop seeds which are included in the starter packs. Overall, I find no evidence of starter packs’ impact on small-scale farmers’ adoption of productivity-enhancing technologies. Similarly, the levels of persistence regarding the use of seeds following the delivery of starter packs were not significant. These results are consistent with studies that have found minimal or no persistence on the use of inputs following the provision of subsidies, including Duflo, Kremer et al. (2011). The limited impact that starter packs had on yields in the first year may logically explain that farmers refrained from using improved seeds subsequently because the inputs are not economically attractive. Chapter 4 studies the effectiveness of knowledge transmission from farmers trained in FFS through farmer-to-farmer training (F2F), which could potentially result in lower extension costs and higher impacts. I find that FFS training has a higher impact than F2F training in the first period, but the magnitude of the treatment effect in the second period is not statistically different between the two training methods. I argue that the dissemination of technologies promoted in FFS groups can well be formalized through farmer-to-farmer deliberate training attached to the FFS approach. Given the low costs of F2F training compared to FFS, the introduction of F2F training may substantially alleviate a major constraint to the large-scale introduction of FFS as a training method, its high costs. In Chapter 5, I study the impact of farmer’s participation in FFS and F2F training on small-scale agricultural productivity. A multi-crop yield-index and the yields of cassava were used as impact indicators. The results indicate that both FFS and F2F trainings contribute to a significant increase in farmers’ yields, especially in the second period when the magnitude of the effect substantially increased. We also learned that the effect size does not differ between the two training approaches in neither period, suggesting that F2F communications are a suitable alternative or complement to FFS training. While the chapter was unable to confirm if training materializes in higher yields through technology adoption, I argue that in the context of the sample the adoption of productivity-enhancing practices and inputs are likely the most important impact mechanism. I also study the relationship between agricultural training, the adoption of improved technologies and household food insecurity. I find that farmers’ participation in agricultural trainings has a positive effect, through the adoption of improved technologies, on improvements in household dietary diversity (HDDS). Nonetheless, the impact on household access to food (HFIAS) is less evident. These results suggest that FFS/F2F training can well reduce household food insecurity, which is mostly achieved through the adoption of improved agricultural technologies. Yet, there are farm and household specific factors which constrain how training impacts technology adoption and how adoption affect household food insecurity and diet diversification. In Chapter 7, I synthesize the results of the four main chapters and articulate the sequence of results from training to adoption to productivity to food security.