Many authors report on the effect of rice–fish culture on rice yields. Some reports show increased rice yields, others show no effect or even decreased yields. To verify the impact of rice–fish culture, data gathered over eight experiments (1995–1999) at the Co Do experimental rice–fish station, Vietnam, were analysed through multiple regressions with rice yield and different yield components as the dependent variables. We used eight independent variables: season, water depth, rice variety, rice seeding rate, and the effective stocking density (ESD) of silver barb, Nile tilapia, common carp and snakeskin gourami. Season had the biggest impact on nearly all dependent variables. Rice yields in the wet season were on average 2·42 t/ha lower as compared to the dry season. Higher water levels decreased the number of panicles/m2 and the rice yield. The rice seeding rate affected the yield component variables, but had no impact on the rice yield. Silver barb feeding decreased the number of panicles/m2 but this was compensated by an increased number of grains/panicle. Snakeskin gourami had a positive effect on panicles/m2 and, together with common carp, on sink capacity, probably through improved nutrient cycling in presence of fish and improved nutrient uptake by the rice plant. Snakeskin gourami had a negative effect on the ripening ability, whereas Nile tilapia had a positive effect. Fish did not have any impact on the rice yield. Increased water level and reduced rice arable area, the main two requirements for rice–fish culture, result in lower rice yields from rice–fish systems as compared to monoculture systems.