In the present paper, we consider the nonparametric regression and study the integrated square error for nonparametric estimate of the unknown function. For the case that the errors of the regression model are martingale differences, the asymptotic normality and consistency of the integrated square error are established. These results improve the works in Ioannides [5].