Abstract: The multiposition test Is a common method for obtaining all gyro drift coefficiente from measured data.In this paper a design criterion of multiposition tests is discussed from the view-point of optimization and the error theory.In accordance with the criterion a series of different tests in common use have been analyzed and compared.In result it is concluded that the 30-position test is an optimum one for the error equation and the 16-position test is an optimum one for the simplified error equation.Then it is also pointed out that although 8-position test is a common method for determining the drift coefficients of the dynamically tuned gyro,it is impossible to obtain the mean square deviations of different estimates and the coefficients related to a2 from one test.Moreover,in relation to the dynamically tuned gyro D（X）XZ and D（Y）YZ are bigger than D（X）Z and D（Y）Z,so D（X）XZ and D（Y）YZ are not to be negleted.In consequence,only the multiposition test can reflect the coefficients of the dynamically tuned gyro actually.The experimental results given show that the design criterion of the multiposition tests is valid.