Growing awareness of widespread hunger and poverty in many countries in the SSA is spurring a focus on productivity increase in smallholder farming systems. The rationale is that with current production systems many SSA countries are not keeping pace with population growth and changing of peoples’ lifestyles. To respond to this challenge the Government of Mozambique developed its Strategic Plan for Agricultural Development (PEDSA) aiming to improve agricultural productivity of the majority of smallholder farmers who depend on agriculture for their livelihoods. Smallholder farmers are diverse in terms of resources and aspirations. The main objectives of this study are first to understand the diversity among maize-based smallholder farms and their current constraints in improving agricultural productivity in the Manica Plateau, Central Mozambique, and second, building on that understanding to explore options for biomass production either for food, cash or biofuel at farm level and contributions to maize availability in the region. The study was conducted in the Dombe and Zembe Administrative Posts. Farmers in the two posts cultivate both food and cash crops using the same resources, however, distances to the urban market differ, with Zembe close and Dombe far away from the markets. In addition, the agroecological conditions for crop production are more favourable in Dombe compared with Zembe. Using farm surveys, direct observations and on-farm measurements, followed by Principal Components Analysis (PCA) I identified land and labour as the variables that can best explain the variability found among smallholder farms (Chapter 2). Based on these variables I categorised farms into four Farm Types (FT): FT1. Large farms (4.4 ha in Dombe and 2.2 ha in Zembe), hiring in labour; FT2. Intermediate sized farms (1.9-1.2 ha), hiring in and out labour; FT3a. Small farms (1.1-0.9 ha), sharing labour; and FT3b. Small farms (1.0-0.7 ha), hiring out labour. The maize yield and maize labour productivities were higher on large farms (2.3 t ha-1 in Dombe and 2.0 t ha-1 in Zembe; 2.5×10-3 t h-1 in Dombe and 2.6 ×10-3 t h-1 in Zembe) compared with small farms (1.5 t ha-1 in Dombe and 1.1 t ha-1 in Zembe; 1.4×10-3 t h-1 in Dombe and 0.9×10-3 t h-1 in Zembe). The hiring in labour from small farms allowed large farms to timely weed their fields. Small farms were resource constrained and hired out labour (mutrakita) for cash or food to the detriment of weeding their own fields, resulting in poor crop yields. Excessive alcohol consumption by small farms also raised concerns on labour quality. Chapter 3 explored options aiming at addressing farmers’ objectives of being maize self-sufficient and increased gross margin and the contribution to national objective of producing food. A bio-economic farm model was used to investigate two pathways to increase agricultural production: (i) extensification, expanding the current cultivated area; and (ii) intensification, increasing input use and output per unit of land. In the extensification pathway I considered the use of animal traction, herbicides and cultivators to save labour, whereas in the intensification pathway I explored the use improved varieties of maize, sesame, sunflower, pigeonpea and fertilizers. I focused on the large farms and the small farms hiring out labour as they represent both sides of the spectrum. The simulated results showed that combining labour and labour saving technologies substantially increased both gross margin and maize yields of large and small farms in both posts. Minor trade-offs is observed on large farms between the two goals whereas for small farms we see synergies between the goals. We concluded that prospects for increasing gross margin and food production are much better for large farms in Dombe compared with other farms. In Dombe, the maximum gross margin of large farms was 7530 $ y-1 per farm and maximum maize sales of 30.4 t y-1 per farm. In Zembe, the maximum gross margin of large farms (2410 $ y-1 per farm) and maximum maize sales (9.5 t y-1 per farm) were comparable to small farms in Dombe. I further assessed the impact of two biofuel investments (jatropha plantation and sunflower outgrower schemes) on farm level food security (food availability, access to food, stability of food, utilization of food). The results showed positive impact on small farms from employment on a jatropha plantation by increasing access to food and no impacts on intermediate and large farms. Impacts on food security from the sunﬂower outgrower scheme were minor which may be explained by the poor yields. The need to link smallholder farmers to markets has been increasingly recognized as important strategy to promote rural development and poverty reduction. I developed an analytical framework, the Windmill Approach that looked at decision making at farm level to grow certain crops and at transaction strategies (Chapter 5). Through this framework I showed that a farmer decision to participate in a particular (new) value chain is determined by (a) the suitability of the new crop in the farm system (including the adaptability of the current farm system), and (b) the farmer’s experience with selling in various value chains. This has major policy implications as it highlights that to support smallholder farmers access to markets a holistic approach is needed that combines farming systems analysis and transaction cost theory. In order to explore the opportunities for smallholder development there is need to understand the diversity of farms and farmers’ social and economic context. For large farms, in Central Mozambique farms with on average 2-4 ha of land, opportunities to improve their livelihoods through crop production can follow two pathways: intensification and extensification. Smallholders continue to produce staple food crops even when working on a plantation or participating in outgrower schemes. For small farms, off-farm opportunities such as those in a biofuel plantation are the best options to improve their livelihoods.