Investigation of the application of different methods for riboflavin determination proved the lumiflavin method recommended by Hausheer et al. to be the most suitable one for determining riboflavin in materials of complex composition, such as soybean. However, due to inadequate preparation of the blank sample, the method is burdened by a positive error. By modifying the blank sample and certain modifications in the procedure, more precise results, greater speed and simplification of the method were achieved. The soybean sample analysed by this modified procedure gave a mean recovery and standard error of 102.9 +/- 2.1 for added riboflavin, which is very satisfactory.