An experiment was conducted over seven sites in eastern England sown to winter cereals, whereby the soil mineral nitrogen (N) and N recovery in the crop were measured frequently during the active growth period. Plots received 100 kg/ha fertilizer N on one of three dates in spring (28 March, 11 April and 25 April) on each site in each of the three years (1994 to 1996), and these were compared with controls with no applied N. At one site, uncropped plots were included, and measurements of gaseous N emission, soil respiration, soil microbial biomass and root tissue N concentration were also made. The fertilizer applications boosted yields by at least 3 t/ha at all sites, but apparent recovery of fertilizer N varied throughout the previously known range of 45–85%. Soil type and timing of application had no effect on N recovery, or on final yields. N was mineralized from soil sources throughout the growth period, but mineralization was outweighed by ‘‘immobilization’’ of large amounts of N (around 30 kg/ha) in the soil, chiefly during May when crop growth was most rapid. This occurred on all sites with a crop, but not where the crop was absent. Measured losses of N2 and N2O were very small (<70 g/ha/dayN) on the site where they were measured, and no other evidence of loss or storage of N was found. The apparent recovery of fertilizer N at each site, was almost exactly explained by the amount immobilized during May. We conclude that the poor recovery of spring-applied N was due, not to losses as previously assumed but to temporary immobilization during the period of most active crop uptake. Immobilization was caused primarily by the presence of the crop.