Denitrification is an important soil process for assessing nitrogen cycling and controlling nitrogen pollution in the environment. Numerous studies of denitrification rates in soils have been reported over the last decade, many with sampling protocols that are more reliable than in the past. In this paper, we review denitrification rates for agricultural and forest soils that have been reported in the literature, discuss factors that appear to be important in controlling the amount of denitrification that occurs in these soils, and summarise modelling approaches that have been used to predict annual denitrification rates. Most studies of in situ denitrification in upland soils have been conducted in agricultural grassland and forest ecosystems, with a paucity of studies reported from other ecosystems. A large range of annual, in situ, denitrification rates have been reported (0–239 kg N/ha.year), with the highest rates typically occurring in irrigated, nitrogen-fertilised soils. However, most annual denitrification rates reported in the literature appear to be fairly low, with over half of the rates in forest soils being 1 kg N/ha.year, and a mean rate of 13 kg N/ha.year. Numerous soil, site, and management factors have been reported to affect the denitrification process in situ. The literature indicates that the highest rates of denitrification can be expected in nitrogen-fertilised soils, or where site management increases soil nitrate availability. Where nitrate is non-limiting, denitrification rates appear to be highest in irrigated loam soils. The review suggests that it is difficult to predict denitrification rates based on our current understanding, and that pilot studies should still be conducted if soil nitrogen balances are required.