The measurement of radiated sound power from a complex underwater sound source in a non-anechoic pool is investigated, taking account of the facts that a non-anechoic pool does not satisfy the conditions for existence of a diffuse field and that the theory applicable to a reverberation room in the air cannot be used for a non-anechoic pool. Based on normal wave theory, a relationship is derived between the spatial average sound pressure level in the reverberation control area far from the source and the sound source radiation power level. The influence on the measurement results of the number of points (i.e., hydrophones) evenly distributed in the non-anechoic pool over which the spatial average is taken is also investigated. The radiated sound power from a marine propulsion system including a propeller is measured using this technique. Both the theoretical and experimental results show that above the cutoff frequency of the non-anechoic pool, the radiated sound power from a complex underwater sound source can be measured accurately using the spatial average method. The radiated sound power from a marine propulsion system including a propeller measured by this method differs from that obtained using the enveloping surface method by no more than 2 dB. The accuracy of the proposed method is related to the number of averaging points evenly distributed in the non-anechoic pool: the greater the number of evenly distributed points, the higher is the accuracy, and the radiated sound power measured using 256 points evenly distributed in the non-anechoic pool differs from the free field result by no more than 2 dB.