In late summer, in a semi-arid mountain range in Nepal, we compared 3 field methods for determining the botanical composition of herbivore diets. Data were collected from the same animals belonging to 1 herd of domestic yak (Bos grunniens) and 2 herds of mixed smallstock, consisting of domestic goats (Capra hircus) and sheep (Ovis aries). Bite count, feeding site examination, and microhistological analysis of feces gave different estimates of forage categories and plant species in both animal groups. Because yaks grazed in other vegetation communities when not observed for bite-counts and feeding signs, the results from the latter methods could not be compared directly with that from fecal analysis. In smallstock, feeding site examination gave higher estimates of graminoids and lower estimates of shrubs than the other 2 methods, probably because all feeding signs on shrubs were not detected. Bite-counts and fecal analysis gave comparable results, except that forbs were underestimated by fecal analysis, presumably due to their more complete digestion. Owing to the difficulty in collecting samples that are representative of the entire grazing period and the problem of recording feeding signs correctly, both feeding site examination and bite-counts are unsuitable methods for studying the food habits of free ranging domestic and wild herbivores. Microhistological analysis of feces appears to be the most appropriate method, but correction factors are needed to adjust for differential digestion. The systematic use of photomicrographs improves the speed and accuracy of the fecal analysis.