Milk urea content (MUC) is used to manage protein nutrition and predict nitrogen excretion of dairy cows. However, MUC might depend on the roughage type offered and hence, for comparable MUC values, different N-excretions might be found. To evaluate this, three diets were compared in a feeding trial with 18 lactating Holstein cows in a Latin square design with as roughages 100% maize silage (treatment 100 MS), 5050% maize silage/prewilted grass silage (treatment 50 MS) and 100% prewilted grass silage (treatment 100 PGS). For all treatments, cows were fed to supply 105% of their net energy and digestible protein requirements and to have a daily rumen degraded protein balance (RDPB) intake of 100g. This was only possible by feeding soybean meal as a protein corrector to 100 MS and 50 MS and by feeding citruspulp as an energy corrector in 100 PGS. The same balanced concentrate was fed to all groups. In a separate trial, N-balance was determined for both 100% rations. In the feeding trial, the MUC of 100MS (230mg/1) and 50MS treatment (214mg/1) were significantly (P <0.001) different from that of 100PGS (171 mg/l). Cows on treatments 50 MS and 100 PGS ingested the same amount of RDPB (71 and 73g/day), but when fed lOOMS cows ingested -16g/day. After correction for differences in energy and protein supply, MUC of the 100 MS was 71 mg/l higher than that of 100 PGS. N-balances indicated that total N-excretion (faecal, urinary and milk) was almost identical for both treatments: 392 for 100MS versus 389g/day for 100PGS, as was environmental N-excretion (faecal and urinary): 259 for 100 MS versus 272g/day for 100 PGS. However, the MUC content for 100MS was significantly higher: 248mg/l versus 180mg/l for 100PGS. From a correction for differences in energy and protein supply, this difference increased up to 84mg/l between 100MS and 100PGS. These results suggest that MUC is roughage dependent and that a system to predict N-excretion should account for these differences. Therefore the exact mechanism behind the determined roughage influence should be investigated further. (c) 2006 Elsevier B.V. All rights reserved.