The intercropping of rapeseed with frost-sensitive companion plants (CP) has recently been proposed as a way to mitigate the negative environmental impact of rapeseed crops. Using mixed-effect linear models, we compared the yield and weed amounts of rapeseed intercropped with different CP species with that of rapeseed as a sole crop in an unique dataset of 79 field experiments covering a wide range of climate, soil and practices conditions in the northwestern part of France, from 2009 to 2015. Bayesian model averaging procedure was used to determine the relative contributions of sites characteristics to the effects of intercropping.& para;& para;Before winter, field pea and faba bean had accumulated the largest amounts of dry mass, with more than 100 g m(-2). Rapeseed biomass was reduced by 56% by non-legume CPs and by only 18% by legume CPs, the largest decrease being caused by pea. Non-legumes decreased the nitrogen nutrition index of rapeseed by 7%, whereas pea and faba bean increased this index by 6% and 3%, respectively. Intercropping with non-legume and legume CPs reduced weed amounts by 52% and 38% respectively, with no difference between CP species. Non legume CPs decreased rapeseed yield at harvest by 0.581 ha(-1), whereas faba bean and faba bean + lentil in creased yield by 0.16 and 0.12 t ha(-1) respectively, when fertilized at the recommended rate. Intercropping with faba bean, lentil or a mixture of both made it possible to reduce nitrogen applications by 30-40 kg ha(-1) with no significant decrease in rapeseed yield. Faba bean and faba bean + lentil mixtures had the best overall performance. This work suggests that intercropping rapeseed is promising, particularly in soils with low nitrogen content with an early sowing date in the late summer.