The uncertainty of parameters of the mathematical model for the control system of balloon-borne gondola attitude is analysed in this paper. The effects of such uncertainty on the system performance are also verified through digital simulations and experiments. To search the ideal control characteristics, a model reference adaptive control for gondola attitude is developed based on Popov's hyperstability theory. This paper presents the discrete adaptive algorithm which gives time varying adaptive gain. Large amount of digital simulations are conducted accordingly. The results show the advantages of this new method over the traditional one in coping with the problems such as parameter variations over a wide arrange, unknown parameters and nonlinearities. The new approach also proves to have kigher adaptive speed and to be implemented more easily as compared with other adaptive control methods.