We investigated how efficient spatial attention was oriented to pain in two experiments. Participants detected whether painful (pain group) or non-painful somatosensory stimuli (control group) were delivered to the left or right hand. Each stimulus was preceded by a visual cue presented near to the stimulated hand (valid trial), the opposite hand (invalid trial), or centrally between hands. In order to examine both exogenous and endogenous orienting of attention, the spatial predictability of somatosensory targets was manipulated. In the first experiment, visual cues were non-predictive for the location of the pain stimulus as a result of which orienting was purely exogenous, i.e., resulting from the occurrence of the visual cue at the location of somatosensory input. In the second experiment, visual cues were spatially predictive as a result of which endogenous control was added, i.e., attention driven by expectations where the somatosensory target will occur. The results showed that only in experiment 1 spatial attention was oriented more efficiently to painful compared to non-painful somatosensory stimulation. This effect was due to faster responses on valid relative to baseline trials (engagement), rather than slower responses on invalid relative to baseline trials (disengagement), and was significantly correlated with self-reported bodily threat. In experiment 2, prioritization of the pain location was probably overridden by task strategies, as it was advantageous for participants" task performance to attend to the cued location irrespective of whether stimulation was painful or not. Implications of these findings for theories of hypervigilance and attentional management of pain are discussed.