Abstract The aim of the present experiment was to determine the frame of reference used to encode a visually specified target location, in a simple, visuo-spatial short-term-memory task, using delayed manual pointing. Subjects were required to remember the location of a single, visually presented stimulus, over delays of 3 and 30 s, and to recall its location by pointing to it with the tip of a stylus with eyes closed. There were two starting positions for the hand: one near the body, proximal to the target area, the other far from the body, i.e., beyond the target area. Two groups of twelve subjects participated in the experiment. One group observed the target locations in the light, the other group observed the target LEDs in the dark. The results showed a large (between 10–30%) and consistent undershoot of the remembered target location with respect to the hands starting position, in the sagittal axis, both when the hand started from the near and from the far position. That is, from the far starting position, pointing errors lay consistently beyond the actual target location as seen from the subjects viewpoint. Further, this undershoot error in the remembered target location increased over time delay. These results are in favour of the hypothesis that the visual target locations were encoded in a hand-centred frame of reference in spatial short-term memory. They also implicate the use of motor- or action-related representations in visuo-spatial working memory.