Nine laboratories participated in a collaborative study on determination of crude protein in animal feeds to compare a generically described combustion method with the AOAC mercury catalyst Kjeldahl method (7.015). The combustion method was written in general terms of method principle, apparatus specifications, and performance requirements. The sample set comprised closely matched pairs of feed ingredients and mixed products ranging from 10 to 90% protein. Ten pairs ground to 0.5 mm were the focus of the study; 4 pairs were ground to 1.0 mm for comparison. Nicotinic acid and lysine monohydrochloride were included as standards. Collaborators were instructed to report their results for performance checks using materials supplied. Only one laboratory failed to meet the proposed limits. Seven laboratories used the LECO Model FP-228 analyzer and 2 used the LECO CHN 600 analyzer. For the 0.5 mm pairs, repeatability standard deviations (Sr) ranged from 0.09 to 0.58 for the Kjeldahl method and from 0.14 to 0.33 for the combustion method, with a pooled Sr value of 0.28 and relative standard deviation (RSDr) of 0.59%. Reproducibility standard deviations (Sg) ranged from 0.23 to 0.86 (Kjeldahl) and from 0.30 to 0.61 (combustion), with a pooled Sg value of 0.52 and RSDg of 1.10%. Grand means for the samples ground to 0.5 mm were 47.65% protein by the combustion method and 47.41% protein by the Kjeldahl method. For samples ground to 1.0 mm, corresponding values were 31.82 and 31.50% protein. The generic combustion method has been approved interim official first action.