In this study we investigate to what extent the effects of motor imagery on postural sway are constrained by movement features and the subject's imagery ability. Twenty-three subjects were asked to imagine three movements using the kinesthetic modality: rising on tiptoes, whole-body forward reaching, and whole-body lateral reaching. After each task, subjects reported the level of imagery vividness and were subsequently grouped into a HIGH group (scores ≥3, "moderately intense" imagery) or a LOW group (scores ≤2, "mildly intense" imagery). An eyes closed trial was used as a control task. Center of gravity (COG) coordinates were collected, along with surface EMG of the deltoid (medial and anterior portion) and lateral gastrocnemius muscles. COG variability was quantified as the amount of fluctuations in position and velocity in the forward-backward and lateral directions. Changes in COG variability during motor imagery were observed only for the HIGH group. COG variability in the forward-backward direction was increased during the rising on tiptoes imagery, compared with the control task (p = 0.01) and the lateral reaching imagery (p = 0.02). Conversely, COG variability in the lateral direction was higher in rising on tiptoes and lateral reaching imagery than during the control task (p < 0.01); in addition, COG variability was higher during the lateral reaching imagery than in the forward reaching imagery (p = 0.02). EMG analysis revealed no effects of group (p > 0.08) or task (p > 0.46) for any of the tested muscles. In summary, motor imagery influences body sway dynamics in a task-dependent manner, and relies on the subject' imagery ability.