A new layered symbol detection algorithm for multiple-input multiple-output systems is presented. In this scheme, the layers are divided into two groups and detected differently. For the layer with the smallest post-detection signal-to-noise ratio, an exhaustive search is performed over the signal constellation; for the remaining layers the conventional iterative vertical Bell laboratories layered space-time (V-BLAST) technique is utilised. The proposed algorithm also uses a different symbol detection ordering from that recommended in the original V-BLAST algorithm. Simulation results show that the error rate performance of the proposed detection algorithm approaches closely that of an optimal maximum likelihood detector with no reduction in the symbol detection throughput, while the computational complexity is less than |Q| times that of V-BLAST, where |Q| is the cardinality of signal constellation Q.