The ventriloquism effect is a critical phenomenon for understanding the underlying mechanisms of multisensory integration. Cross-modal spatial attention causes a distortion of sound localization, although the neural basis of the effect remains an unanswered question. We hypothesized that top-down and bottom-up visual-spatial attention causes the ventriloquism effect with different modulations of ongoing neural oscillation. To test this hypothesis, human scalp electroencephalography (EEG) was measured during a sound localization task. Top-down attention suppressed the EEG amplitude in the alpha frequency (10 Hz) over the contralateral temporal electrode sites to visual cue hemifields. Bottom-up attention shifted the EEG phase to the theta frequency (7 Hz), rather than suppressing the amplitude. Two different neural mechanisms of ongoing neural oscillation contributed toward the ventriloquism effect, with different spatial attention.