A decision feedback equalizer (DFE) using complex backpropagation (CBP) algorithm is presented in this paper. In passband multilevel signaling systems, learning algorithms with complex-valued parameters are generally required in adaptive equalizers. Conventional approaches using complex least mean square (CLMS) algorithm cannot be satisfactory due to their lack of capability in partitioning the signal space. A new approach using CBP is used to equip a DFE with better performance. Computer simulations show significant performance improvement compared to the conventional one.