In array processing, the spatial smoothing technique and its variations are known to be effective in combatting coherent interferences. However, they are disadvantageous in that they either reduce the effective array aperture or require the formation of covariance matrices which causes numerical difficulties when finite-precision computations are involved and given array data are ill-conditioned. In this paper, we present a data-domain spatial preprocessing algorithm, by which the effective array aperture is expanded without forming covariance matrices. Also, we propose a parallel spatial smoothing technique in which spatial subarray data are rearranged before processing. The incorporation of the data-domain spatial preprocessing algorithm and the parallel spatial smoothing technique is simple, and constitutes a parallel modified spatial smoothing technique which is a parallel implementation method of the modified spatial smoothing technique. The proposed parallel modified spatial smoothing method is highly fast, numerically stable, and capable of nulling out coherent interferences. Since the proposed method can readily be combined with least-squares solving systems using orthogonal transformations, one can take full advantages of systolic/wavefront arrays to get high throughput.