ContextRice yield is low at 2.1 t ha−1 in sub-Saharan Africa. Increased yield is a critical challenge to food security and environmental conservation in this region. However, smallholder farmers have limited access to irrigation, mineral fertilizers, and improved crop varieties. One approach that even resource-limited farmers can easily manipulate is to optimize planting densities. However, there is limited empirical evidence to provide technical recommendations under such low-yielding conditions. ObjectiveThis study aimed to identify the effect of dense transplanting on lowland rice yields under low-yielding conditions, with a target range below 5 t ha−1. MethodsMulti-field trials were implemented with transplanting densities of a regular rate at 25–26.7 hills m−2, a doubled rate at 50–53.3 hills m−2, and a tripled rate at 88.9 hills m−2 in the central highlands of Madagascar, where rice yields are limited by nutrient deficiency and low temperature. Canopy coverage and cumulative intercepted radiation (CIR) were monitored from transplantation to maturity using digital imagery analysis. Field observations (n=306) and four-year household surveys (n=356) were combined to calculate the costs and benefits of changing transplanting densities. ResultsDoubling densities from 25.0–26.7 hills m−2 to 50.0–53.3 hills m−2 had a consistent yield advantage by approximately 0.4 t ha−1 across a yield range of 1.8 t ha−1–4.4 t ha−1. The yield was further increased by tripling the transplanting densities to 88.9 hills m−2 when the yield range was 1.9–2.3 t ha−1. The yield advantage of higher transplanting densities was attributed to a greater CIR at the initial growth stages and a significantly greater panicle number. Household surveys and field observations indicated that the benefit of yield gain was more than three times greater than the additional cost of doubling the seed amounts. No significant yield differences were observed by changing the transplanting densities when the yield level was higher than 5 t ha−1 or lower than 1.3 t ha−1 where substantial reductions in grain fertility occurred owing to low-temperature stress. ConclusionsA relatively high transplanting density of 50–53.3 hills m−2 or even higher is recommended to ensure initial canopy development and panicle number in low-yielding conditions where individual plant growth is stagnant, except in fields with high risks of grain set failure. ImplicationsThis study provides an easy-to-use opportunity for smallholder farmers to increase their rice yield. Further studies are required to determine whether these findings apply to warmer climatic conditions.