The Kjeldahl and Dumas (combustion) methods were compared in 11 laboratories analyzing samples of milk, skim milk powder, whole milk powder, whey protein concentrate, infant formula, casein, caseinate, 2 reference compounds (glycine and EDTA), and a secondary reference skim milk powder. The comparison was conducted by using international standards where applicable. Overall means were 8.818 g N/100 g by the Kjeldahl method and 8.810 g N/100 g by the Dumas method. No evidence was found for a consistent bias between methods that may be of concern in the trading of dairy produce. A review of more than 10 related trials revealed a lack of consensus in the bias between the 2 methods, suggesting that differences in methodology and sources of systematic error may be contributors. For samples containing > 2 g N/100 g, the Dumas relative repeatability and reproducibility standard deviations were consistently about 0.35 and 0.75%, respectively, whereas the corresponding Kjeldahl values declined generally with N content and were significantly larger. The Dumas precision characteristics may be due to the dominance of Leco analyzers in this trials, and in most other recent trials, rather than an inherent method attribute. Protein determination methods for dairy products need to be reviewed and updated. The Dumas method needs Codex Alimentarius status as a recognized test method.