In the above paper, Tsuji et al. (see ibid. vol.28 (1998)) proposed an interesting scheme for adaptive control that uses only one neural network (NN). Moreover, the stability of the parametric and identification errors was analyzed and a sufficient condition for it was presented. However, since the selected signals for the NN input were not the natural choice, their scheme presents some problems. This study intends to circumvent these problems by modifying the scheme proposed in the work of Tsuji et al., and to highlight some advantages of the modified scheme.