Speech is an attractive interface for mobile equipment if it is clearly received. A microphone array aids in reducing noise and enhancing speech. A small-scale microphone array generally adopts the subtractive beamforming technique, which constructs sharp notches in a beampattern in either an active or a passive way, so that target sound sources are assumed as point sound sources. However, the shapes of actual sound sources are various and not always points. In particular, the mouth needs to be modeled as line sound source or plane sound source. Therefore, beamformers should target line sound sources. A passive subtractive beamformer is proposed to expand width of a sharp notch by combining several 2ch subtractive beamformers. In multiple sound source conditions, it is necessary to avoid the effect of spatial aliasing against nontarget sound sources. Then, a hybrid technique is applied to realize an optimal connection of the quadruple, double, and single subtractive beamformer for the low, middle, and high frequency region. The feasibility of the proposed subtractive beamformer is confirmed by performance evaluation in suppressing the signals from line sound sources. [Work supported by the Telecommunications Advancement Organization of Japan.]