A soft decision decoding method in a multiple-input multiple-output system which respectively receives two simultaneously transmitted data streams, a first transmitting symbol and a second transmitting symbol, in a first receiving signal and a second receiving signal is provided. The soft decision decoding method includes calculating common metrics corresponding to distances between first candidate symbols, which correspond to combinations of all bits that the first transmitting symbol can have, and the first receiving signal using a channel matrix corresponding to a communication channel through which the first and second transmitting symbols are transmitted and determining intermediate nodes between the second receiving signal and the first candidate symbols and nodes between the second receiving signal and second candidate symbols corresponding to combinations of all bits that the second transmitting symbol can have; selecting first nearest candidate symbols that are at a shortest distance from the intermediate nodes on a signal constellation from among the second candidate symbols and determining a soft decision metric for each bit value of the first transmitting symbol based on the first nearest candidate symbols and the common metrics; and determining second nearest candidate symbols based on a logic value of a predetermined bit of each of the second candidate symbols and the positions of the intermediate nodes on the signal constellation and determining a soft decision metric for each bit value of the second transmitting symbol using the second nearest candidate symbols and the common metrics.