In this paper we follow the fate of single applications of 15 N‐labeled fertilizer to old grassland, over a period of nearly 20 yr. In 1980 and 1981, 15 N‐labeled N was applied to two of the treatments on the Park Grass Continuous Hay Experiment at Rothamsted, started in 1856. The labeled N was applied at the same rate (nominally 96 kg ha −1 yr −1 ) and in the same chemical form (NH 4 or NO 3 ) as the unlabeled N normally applied as fertilizer to the selected treatments. After 19 yr, 69.6% of the N applied in 1980 as 15 NH 4 had been harvested in successive cuts of herbage, with a further 16.5% remaining in the soil. For 15 NO 3 , 64.3% had been harvested and 13.8% remained in the soil. The 15 N data were then used to calculate annual inputs of nonfertilizer N, annual losses of N and N turnover times in old grassland, assuming that the selected treatments were under steady‐state conditions. The annual input of N from nonfertilizer sources (rain, dry deposition, N fixation by leguminous components of the herbage, etc.) was large: 39 kg N ha −1 yr −1 for the NH 4 treatment and 31 kg for the NO 3 treatment. Leguminous plants made up <2% of the herbage in both the NH 4 and NO 3 treatments. The annual loss from the NH 4 treatment was 19 kg N ha −1 yr −1 and 24 from the NO 3 treatment. The gross turnover time of N in the root compartment (which included plant crowns) was 1.41 yr for the NH 4 treatment and 0.42 yr for NO 3 The gross turnover time of soil microbial N was 2.13 yr (NH 4 ) and 1.83 yr (NO 3 ): for humus N (i.e., soil N not in roots or microbial biomass) it was 181 yr (NH 4 ) and 116 yr (NO 3 ).