AbstractHerbage growth was measured in two contrasting years on five farms representing a range of soil types. On fields cut for conservation in 1982 DM production varied between farms from 12.3 tha‐1to 14.1 tha‐1in line with fertilizer N levels, which ranged from 220 to 333 kg ha‐1. In 1983 production was from 10.3 tha‐1to 12.3 tha‐1; this again corresponded with fertilizer N, except on farm B which had the shallowest soils. On fields used for grazing all farmers used rotational management. Herbage accumulation was measured by trimming to 25 mm and harvesting after 4 weeks. As on conservation fields, herbage accumulation was generally greater on farms using more N, at least up to 300 kg ha‐1.In 1982 growth consistently exceeded predictions based on plot experiments. In 1983 growth was on average less than predicted on grazing fields but slightly more than predicted on cutting fields‐We conclude that such predictions of annual production are a valuable aid to farm planning, and need not be reduced to allow for on‐farm conditions, although they are least good at low levels of N. Predictions of individual cuts were not satisfactory, with a strong tendency to overestimate first cut and underestimate second cut.There was no indication that annual herbage production was less on badly drained than on well drained soils. Indeed, in the dry summer of 1983 badly drained land appeared to have an advantage, despite considerable poaching damage in the spring. Furthermore, there was no evidence that production was deficient on swards dominated by Holcus a