In this paper we analyze the problem of optimizing the switching levels of adaptive modulation for a two-user system under constant power condition, in order to maximize the overall system throughput for a target average bit-error-rate (BER) for each user. Since the optimal switching levels depend on the probability density of the signal-to-interference-noise ratio (SINR), expression for this probability density is derived and the optimal switch levels are obtained through a Lagrange optimization technique. This approach can be readily extended to multiple-user case. It is shown by simulation that the proposed jointly optimizing method achieves the average BER target for each user while an individually optimizing method fails. The efiects of these two methods and of the interference coe-cients on the average system throughput are also analyzed. DOI: 10.2529/PIERS061113041157 Wireless communication channels typically exhibit channel quality ∞uctuations, and to mitigate their detrimental efiects a number of approaches have been proposed in the literature. Methods based on adaptive modulation according to the near-instantaneous channel quality information have gained popularity. For Rayleigh fading channel, Chung and Goldsmith (1) showed that variable power variable-rate schemes are optimal, but they also found that the extra throughput achieved by the additional variable power assisted adaptation over the constant power variable rate schemes is marginal for most types of fading channels. Constant power variable rate adaptive scheme is also analyzed by Choi and Hanzo (2), where the target average BER forms the constraint condition. They established the relationship among the optimal mode switching levels, and found that although such relationship holds regardless of the underlying channel scenarios, the switching levels do depend on the statistics of the channel quality. The above studies were focused on single user scenario. Yet in wireless communication, a user typically sufiers from the interference of other users, which in the GSM system may be due to multipath channel conditions, and in the CDMA system may stem from the loss of orthogonality among the spread sequences. Several approaches have been proposed based on power control and base station selection (3,4), with little consideration for adaptive modulation. In (5) Qiu and Chawla approached the multiuser problem by using power control and adaptive modulation. They showed that adaptive modulation without power control scheme outperforms SINR-balance power control scheme in terms of overall throughput, a flnding that indicates the signiflcance of constant power adaptive modulation for multiuser system. In their approach, the switching levels were obtained by using the method due to Webb and Steele, and a parametric approximation for throughput for all modulation modes was also adopted. In this paper we propose a constant-power adaptive modulation scheme for downlink trans- mission in a multiuser system. It is based on a Lagrangian optimization technique. We consider two-user case here to highlight the concept. Extension to multiuser case can be readily carried out. The downlink transmission model is shown in Fig. 1, which is similar to the one used in (6). The received signals are for user 1