The relationship between protein nutrition and milk urea N was investigated in three experiments with a total of 125 cows. After 4 wk of pretreatment, cows received 1 of 13 diets with different ratios of protein to energy for 16 wk. Milk was sampled individually for urea analyses during pretreatment and during wk 1, 5, 10, and 15 of treatment. Results were compared with N losses estimated from rumen fermentation and with N losses of metabolic origin. The mean milk urea N concentration was 12.6 mg/100 ml of milk (range, 9.0 to 18.3 mg). For bulk samples especially, the rumen efflux of crude protein intake was the main determinant of the variation in milk urea N (r2 = 0.81; residual SD = 1.1). However, N losses from the rumen explained only about 50% of the variation in the milk urea N content of samples from individual cows. The N losses of metabolic origin, which, in these experiments, were responsible for 47 to 100% of urinary N losses, were not related to milk urea N. Results showed that regular measurement of milk urea N in bulk samples can be used to monitor N losses from rumen fermentation. However, the value does not give an indication of the efficiency with which the absorbed protein is utilized.