Participants were cued by an auditory tone to grasp a target object from within a size-contrast display. The peak grip aperture was unaffected by the perceptual size illusion when the target array was visible between the response cue and movement onset (vision trials). The grasp was sensitive to the illusion, however, when the target array was occluded from view when the response was cued (occlusion trials). This was true when the occlusion occurred 2.5 s before the response cue (delay), but also when the occlusion coincided with the response cue (no-delay). Unlike previous experiments, vision and occlusion trials were presented in random sequence. The results suggest that dedicated, real-time visuomotor mechanisms are engaged for the control of action only after the response is cued, and only if the target is visible. These visuomotor mechanisms compute the absolute metrics of the target object and therefore resist size-contrast illusions. In other situations (e.g. prior to the response cue, or if the target is no longer visible), a perceptual representation of the target object can be used for action planning. Unlike the real-time visuomotor mechanisms, perception-based movement planning makes use of relational metrics, and is therefore sensitive to size-contrast illusions.