Improvements in agricultural land and labour productivity are needed to meet the growing food demand and reduce farmer poverty in sub-Saharan Africa. The objectives of this study were to (i) quantify variation in labour inputs, yield and labour productivity among rice fields; (ii) elicit factors associated with this variation; and (iii) identify opportunities for improving yield and labour productivity. The study was carried out in two contrasting Beninese villages: Zonmon in the south and Pelebina in the north-west. In Zonmon 82 irrigated rice fields were surveyed during the 2013 and 2014 dry seasons. In Pelebina 50 rainfed lowland rice fields were surveyed over three rainy seasons (2012–2014). Data on farmer field management practices and field conditions were recorded through interviews with farmers, on-farm observations and measurements. Stepwise regression analyses were used to identify variables associated with variation in yield, labour inputs and labour productivity. Average yields were 4.8 ± 2.0 t ha-1 in Zonmon and 2.3 ± 1.2 t ha-1 in Pelebina. Average labour productivity, however, was larger in Pelebina (17 kg of paddy rice person-day-1) than in Zonmon (8 kg of paddy rice person-day-1). Relative yield gaps (43–48%) and labour productivity gaps (59–63%) were similar in the villages. There was no trade-off between yield and labour or labour productivity within the villages, suggesting that in many cases rice yields can be increased without additional labour inputs. The major labour-demanding farming operations were bird scaring in Zonmon and harvesting and threshing in Pelebina. We identified opportunities to improve rice yield and labour productivity, given current farmer knowledge and resource endowment. Based on the statistical models fitted per village, increasing the average hill density would result in up to 1.2 t ha-1 more yield, and up to 4 kg person-day-1 greater labour productivity for Zonmon. Increasing the average field size and avoiding rice sh