Resumo. The objective of this study was to correlate the Biuret, Lowry, Bradford, and Mlikoscan methods used to determine the level of protein in goat's milk and compare them with the Kjeldahl method. Milk samples collected from 18 goats of breeds Saanen (n=5), Anglo-Nubian (n=4), Pardo-alpina (n=5), and Creole (n=4) were analyzed. Correlation analyses between the different methods were performed using the Pearson test and the protein content between different breeds was compared using ANOVA. Total protein (TP) and casein (CN) concentrations obtained by all methods studied were positively correlated with the Kjeldahl method (p<0.01). The Bradford method presented the lowest correlation coefficient. All methods apart from Bradford showed significant differences in TP and CN between the Anglo-Nubian and Pardo-alpina breeds (p<0.05). The Bradford method showed significant differences in TP between the Saanen and Pardo-alpina breeds (p<0.05), and no differences were found between breeds reegarding CN. In conclusion, the Bradford method had the lowest correlation value (TP and CN) with the Kjeldahl, and the Bradford method did not show the same pattern of differences between breeds in TP and CN, as was evidenced in all other methods. Therefore, the use of the Bradford method is not recommended to determine proteins in goat’s milk.