We used event-related functional magnetic resonance imaging to study the neural correlates of endogenous spatial attention for vision and touch. We examined activity associated with attention-directing cues (central auditory pure tones), symbolically instructing subjects to attend to one hemifield or the other prior to upcoming stimuli, for a visual or tactile task. In different sessions, subjects discriminated either visual or tactile stimuli at the covertly attended side, during bilateral visuotactile stimulation. To distinguish cue-related preparatory activity from any modulation of stimulus processing, unpredictably on some trials only the auditory cue was presented. The use of attend-vision and attend-touch blocks revealed whether preparatory attentional effects were modality-specific or multimodal. Unimodal effects of spatial attention were found in somatosensory cortex for attention to touch, and in occipital areas for attention to vision, both contralateral to the attended side. Multimodal spatial effects (i.e. effects of attended side irrespective of task-relevant modality) were detected in contralateral intraparietal sulcus, traditionally considered a multimodal brain region; and also in the middle occipital gyrus, an area traditionally considered purely visual. Critically, all these activations were observed even on cue-only trials, when no visual or tactile stimuli were subsequently presented. Endogenous shifts of spatial attention result in changes of brain activity prior to the presentation of target stimulation (baseline shifts). Here, we show for the first time the separable multimodal and unimodal components of such preparatory activations. Additionally, irrespective of the attended side and modality, attention-directing auditory cues activated a network of superior frontal and parietal association areas that may play a role in voluntary control of spatial attention for both vision and touch.