Conventional multistage multiuser detectors (MUDs) such as parallel interference cancellation (PIC) suffer from error propagation. In this paper, we propose a generalized multistage MUD with receiver diversity, which is referred to as weighted interference cancellation (WIC), in which reliability weighting factors and user-decision ordering are applied to reduce error propagation. Two flavors (the parallelizable and nonparallelizabe WIC) are presented in this paper. An algorithm is next derived for the joint determination of the reliability weighting factors and weightings for linear diversity combining. Two selection diversity schemes are also presented as lower complexity alternatives to linear diversity combining. Finally, sorting algorithms for determining the user decision order are also proposed. Simulation results show that the proposed MUD achieves very good performance. For Eb/N0<20 dB, using just a few stages, certain variations of the proposed MUD almost achieve the single user bound with one antenna when the number of active users is smaller than 15 and with two antennas when the number of users is smaller than 30.