Motor imagery may use the same mental representations as overt actions, or it may be performed using a nonmotoric cognitive estimation strategy. These competing hypotheses were tested by having participants perform either simple overt pointing tasks or analogous motor imagery tasks while manipulating the visual context and the availability of visual information. In three pairs of experiments, visual illusions, word labels, and numeric labels were all found to have comparable effects on overt pointing and motor imagery. In each case, effects of the contextual variables on overt performance and imagery were larger when vision was removed and a delay imposed before movement initiation. These findings support the hypothesis that a common mental representation is used in both motor imagery and overt actions. In contrast, the results were inconsistent with the view that motor imagery is performed using a cognitive estimation strategy. Limitations in the ability of motor imagery to faithfully simulate overt actions are discussed.