Recent behavioral studies in monkeys and humans have shown that holding an item in spatial working memory may lead to sustained and spatially selective prolongation of reaction times (RTs) to visual stimuli presented during the memory delay. In order to resolve the seeming contradiction between these findings and current theories on the interaction of working memory and attentional orienting, it has been hypothesized that memory-dependent modulation of orienting may be the net effect of superposed facilitatory and inhibitory mechanisms. Their relative strength during the memory delay may determine whether RTs to visual stimuli presented during the memory delay are shortened or prolonged. Here, we expand on this hypothesis by investigating the spatial distribution of memory-dependent inhibition with behavioral data from normal human subjects. The experiment consisted of a combination of an oculomotor spatial working memory task (memory-guided saccade task, 6-s delay) and a visual discrimination task (performed 1500, 2500, or 3500 ms after presentation of the memory cue). RTs to discrimination stimuli were analyzed as a function of memory-guided saccade amplitude. By fitting polynomial approximations to our data we show that the spatial distribution of memory-dependent inhibition of orienting significantly differs from a monotonic gradient across the visual field. Instead, we demonstrate the existence of a central inhibitory peak surrounded by a facilitatory annulus, forming a transient "inverted Mexican hat" profile, which mirror-images findings from recent studies on the spatial distribution of attention. These findings are consistent with the hypothesis of a highly flexible modulation of orienting in which both the signs and spatial distribution of memory-dependent bias signals are adapted to behavioral demands.