Since the early 2000s conservation agriculture (CA) has been promoted in the Lake Alaotra region of Madagascar for a more sustainable and profitable agriculture. There is, however, little known about its performance in low-input rainfed rice-based cropping systems. We conducted a study during two growing seasons (2013/14 and 2014/15) on an experiment that was established in 2009 at the agricultural research station of the National Center for Applied Research and Rural Development (FOFIFA) in the Lake Alaotra region of Madagascar. The experimental setup was a randomized block design with four replications. Two soil/residue management treatments were studied, conventional tillage without residue retention (CT) and no-tillage with residue retention (NT). These two treatments were tested for a 2-year rotation of maize + Dolichos lablab followed by rice (MD//R) and a 3-year rotation of maize + Stylosanthes guianensis, followed by S. guianensis in the second year, and rice in the third year (MS//S//R). During the 2013/14 and 2014/15 seasons, two levels of weed pressure: 'high' and 'low' were introduced as a split-plot design on the rice plots. The main determining factors of rice yield in the study region were studied: radiation interception, weed infestation, soil moisture and soil mineral nitrogen (N). Our results showed that five to six years of continuous practice of NT with retention of high amounts of crop residues (more than 5¿Mg¿DM¿ha-1) on the soil surface had a significant (p¿=¿0.02) positive effect on rice yield, irrespective of the level of weed pressure and type of crop rotation. CA systems significantly (p¿<¿0.05) reduced weed density and biomass as compared to CT particularly during the vegetative stage of the rice crop in the two growing seasons, which to a certain extent explained the yield gains under CA. In contrast, treatment effects on soil moisture and mineral N contents were marginal. The positive effects of CA on reduced weed pressure may constitut