In the initial stage of SAD phasing, the essential point is to break the intrinsic phase ambiguity. The presence of two kinds of phase information enables the discrimination of phase doublets from SAD data prior to density modification. One is from the heavy atoms (anomalous scatterers), while the other is from the direct-methods phase relationships. The former can be expressed by the Sim distribution, while the latter can be expressed by the Cochran distribution. Typically, only the Sim distribution has been used to yield initial phases for subsequent density modification. However, it has been demonstrated that using direct-methods phases based on the product of the Sim and Cochran distributions can lead to improved initial phases. In this paper, the direct-methods phasing procedure OASIS has been improved and combined with the SOLVE/RESOLVE procedure. Experimental SAD data from three known proteins with expected Bijvoet ratios </DeltaF/>/<F> in the range 1.4-7.0% were used as test cases. In all cases, the phases obtained using the program RESOLVE beginning with initial phases based on experimental phases plus Sim and direct-methods information were more accurate than those based on experimental plus Sim phase information alone.