Measuring the temperature dependence of conductivity of a series doped samples of different thickness, we find three activation energies in the range of 300-500 K. After prolonged annealing in 180-250℃, the positions of two kinks formed by these activation energies shift to high temperature side. In thinner samples the shifting rate is larger than that in thicker samples. In a very thin sample the positions of the kinks even shift spontaneously after being kept at room temperature for a long-period.Changing the voltage, drop across the sample during measuring significantly influences the values of high and medium temperature activation energies. In thicker samples this kind of influence is stronger than that in thinner ones.Most of these results are interpreted by a tentative model which has been checked by computer graphics rather successfully.