Babies followed manual pointing by an adult to locate the more peripheral of a pair of identical targets, whereas a change of head and gaze direction allowed accurate target identification only for the first target to enter the infant's field of view. Babies accurately followed manual pointing into the periphery of their right visual field some 3 months before they did so to their left. To investigate whether following pointing involves precise linear vector extrapolation, different combinations of head, eye and pointing cues were studied with 4.5-year-old children and adults. Neither head movements, nor head and eye movements, nor eye movements alone were accurate cues for children, whereas pointing allowed accurate spatial localization to peripheral positions. Adults accurately followed pointing to the outer periphery, but not to inner positions. These studies suggest that comprehension of pointing is unlikely to depend on extrapolating precise linear vectors along the pointing arm because a precise vector system should yield accurate localization irrespective of the position of the referent. Instead, eyes, head, nose and manual pointing cues may differentially refer to broad zones of visual space. They serve as differential cues because each varies in the extent of lateral movement. Thus, babies and children may follow pointing more accurately than head and eye movements simply because this is the most effective cue for redirecting attention to the outer periphery of each visual hemifield.