We propose zero-forcing (ZF) nulling followed by fuzzy parallel multistage detection for multi-input multi-output system. The parallel detector consists of two or more stages, the first stage give estimate of the substreams and the second stage refines this estimate. The output can also be further iterated to improve the performance. We use ZF for the first stage and then the adaptive partial weighting for parallel detector is evaluated using the fuzzy logic. The simulation results show the uncoded error probability of our proposed method is better than that of ZF followed by improved PIC with hyperbolic tangent function.