The results of 86 replicated experiments on soybean (Glycine max) cultivation practices in eight provinces – planned, conducted and analysed by farmers – were examined to identify local differences in the effects of cultural practices on yield, and to explore the potential for farmers to improve their practices. Plant spacing consistently improved yield relative to broadcast seeding. The comparison between a moderate and a high dose of N, and between local spraying practices and integrated pest management, also had a relatively consistent effect on yield. Conversely, the effects of straw mulch, of moderate doses of N compared to zero N, and of weeding, were highly location-specific. A change in cultural practices influenced yield and economic benefits more strongly in low- than in high-yield situations. Modifications of current cultivation practices often resulted in a significant yield increase. This suggests that farmers need training in how to improve their cultural practices and that it should utilize locally conducted field experiments. Training in experimental skills made farmers less dependent on external measures and advisers, and enabled them to become ‘experts’ who utilize science.