The System of Rice Intensification (SRI) is being promoted worldwide, but relatively little is yet known about its impacts at farm level. This article reviews available evidence on the impact of SRI practices in terms of yield and productivity. Adoption of SRI practices necessarily changes the mix and allocation of inputs, in particular of water, seeds, fertiliser and labour. However, SRI impact studies have generally failed to distinguish between technological change – a more productive use of inputs, evidenced by a change in total factor productivity – increases in input use, or selection effects and their respective effects on yields. The studies reviewed point not only to modest increases in rice yields associated with SRI adoption, but also to concurrent increases in labour and fertiliser use. Often SRI is selectively practised on more fertile plots. As a result, no firm evidence on changes in total factor productivity can be discerned, while partial productivities of land and labour show mixed results. Though yields tend to be higher under SRI management, risk also seems to increase, which initially favours adoption by better-endowed farmers and on better soils. Evidence on SRI impact is further complicated by the large diversity of SRI practices associated with different biophysical, socio-economic and institutional circumstances. We conclude by identifying knowledge gaps surrounding the SRI phenomenon, encompassing agro-technical aspects, socio-economic issues and (dis)adoption behaviour.