Abstract Deepwater rice is grown widely in areas flooding every year to depths of 0.5 to 3.5 m in the valleys and deltas of South and Southeast Asia. In Bangladesh alone, 1.7 million ha are planted to this especially adapted type of rice. The agronomic practices and yield performance of deepwater rice were investigated in Bangladesh from 1977 to 1980 by farmer survey and crop-cut assessment. Land preparation consists usually of seven ploughings and eight harrowings in January to April with simple cattle-drawn implements; usually dry seed, but occasionally pre-germinated seed, is broadcast at 87–95 kg/ha in March to May and a small amount of transplanting takes place; 58% of the fields were sown to pure stands of deepwater rice, 32% were mixed sown with a short duration rice (‘aus’), and 10% were mixed with other non-rice crops. Small amounts of dung and compost are applied; fertilizer was used on one-quarter of the fields, mainly in the eastern, high-rainfall areas; virtually no pesticides are applied. Weeds were controlled in the preflood period (April to mid June) by two cultivations and two handweedings. Harvesting (October to December) is by hand and by cattle are used for threshing; the straw serves mainly as animal fodder and the more bulky stubble is used for fuel in the village or as a mulch for dry season crops. From December to March 80–87% of the fields were planted to dry season crops; more than 90% of the fields were double- or triple-cropped during the year. Favourable preflood rainfall and mild flooding regimes in 1977 and 1978 produced average yields of 2.5 and 2.3 t/ha unmilled paddy, respectively. In 1979 an early drought, and in 1980 severe late flooding, were mainly responsible for reducing the average yield to 2.1 t/ha. The overall 4-year yield average of 2.3 t/ha represents a higher yield level than that quoted in many previous reports. Deepwater rice yields appear to be higher and more stable than the short duration aus crop which is planted at the same time. Higher yields were associated with: (a) the Meghna floodplain having higher early rainfall and a milder flooding pattern; (b) vareties such as Pankaish, Khama, and Kartik Sail yielding > 3 t/ha in one-third of the fields; (c) maximum water depths of 1.5–1.8 (although some varieties yielded well in 3.4 m of water); and (d) pure stands of deepwater rice. The mean yield of fertilized fields was similar to that of unfertilized ones. Farmers use simple wooden implements and few inputs except for the relatively high labour requirements of land preparation, weeding and harvesting. The stable yields are probably due to: (a) natural sources of fertility (crop residues, pulse crops, alluvium and algae); (b) the development of intensive cropping patterns using different crops and special varieties accompanied by a set of versatile agronomic practices; and (c) the absence of a water deficit during the reproductive phase. Yields are clearly governed by the distribution of preflood rainfall, flooding regime, the deepwater rice variety selected, the cropping pattern followed, and the incidence of several pests.