This study investigated cross-modal interactions in spatial attention by means of recording event-related brain potentials (ERPs). Noise bursts and light flashes were presented in random order to both left and right field locations separated by 60 degrees in free-field. One group of subjects was instructed to attend selectively to the noise bursts (attend-auditory group), and a second group attended only to the flashes (attend-visual group). On different runs attention was directed to either the right or left field stimuli of the designated modality. In the attend-auditory group, noise bursts at the attended location elicited a broad, biphasic negativity (Nd) beginning at 70 ms. The cross-modal spatial attention effect on the auditory ERPs in the attend-visual group was very similar in morphology, but the Nd was reduced in amplitude relative to the intra-modal effect. In the attend-visual group, flashes at the attended location elicited enhanced early (100-200 ms) and late (200-350 ms) ERP components relative to unattended-location flashes. The cross-modal effect in the attend-auditory group included small but significant enhancements of early components of the visual ERPs. It was concluded that spatial attention has a multi-modal organization such that the processing of stimuli at attended locations is facilitated at an early, sensory level, even for stimuli of an unattended modality.