Three direct methods (refractive index and weighing method, Tergitol method, and filtering method) for the determination of percentage fat in butter were investigated. For the refractive index and weighing method, the accuracy of the determination of percentage fat was predicted by calculating the theoretical changes in refractive index of fat-solvent mixture and in the weight of 5 ㎖ of mixture for a unit change in the fat percentage in butter. Because of the low accuracies predicted by the calculation and the toxicity of solvents used, this method was not used for further investigation. The fat percentages of seven samples were determined by the Tergitol method and compared with the Mojonnier method for accuracy and time requirement. The precision was ±0.2091 for the Tergitol and ±0.0766 for the Mojonnier method, but the former required only 30-35 minutes while the latter required 40-45 minutes for duplicate analyses. The Tergitol method is as simple as the Kohman method and permits more analyses to he conducted simultaneously with a given amount of space and equipment. If any device or technique which would reduce the final reading error could be achieved, this method could be practical for measuring the fat content of butter during the manufacturing process. For the filtering method an attempt was made in a preliminary investigation to determine its feasibility, and in further investigations an attempt was made to modify the initial procedure in order to reduce the time requirement with a minimum sacrifice of accuracy. The effects of drying temperature, cooling time, drying method and amount of sample on accuracy and time requirement were investigated. The initial conditions of loser temperature (80℃) drying in a convection oven and adequate time (20 minutes) for cooling in an aluminum desiccator gave results with a standard deviation of difference and a mean difference from the Mojonnier method of ±0.1782 and -0.002 respectively and a precision ±0.1375. However, this procedure took approximately 50 minutes and this was considered to be too time-consuming to be practical for a test during the manufacture of butter. Drying at higher temperature resulted in a saving of time, and a loss of accuracy. The time allowed for cooling the dried fat and its container had an effect on the accuracy of the results but the accuracy achieved after 10 minutes cooling was as good as that achieved after 20 minutes cooling. To reduce the time requirement for the drying an infra-red heat lamp was used instead of a convection oven, and this would found to be a satisfactory method for fast drying. Two types of balance, an analytical balance and a moisture balance, were used for the final weighing. Using the analytical balance gave much better accuracy. The determination could be done in 20 minutes with the moisture balance and in 20-25 minutes with the analytical balance. Two different amounts of sample (1g and 5g) were tried when using the analytical balance. Better results were obtained with 5g samples although 25 rather than 20 minutes was required for a single analysis. When using the analytical balance for weighing, the infra-red lamp for drying, and 5g samples, the standard deviation of difference. form the Mojonnier method was ±0.2846, the mean difference was -0.012, and the precision was ±0.1028. This procedure was considered to be the most suitable filtering procedure for the determination of percentage fat in a creamery due to the relatively accurate results and the shorter time requirement. When compared with the presently available Kohman method, it gave slightly less accurate results, but the times required for determinations with these two methods were the same. For the Kohman method, the precision was ±0.0946, and the standard deviation of difference, and mean difference from Mojonnier results were ±0.1398 and ±0.031 respectively.