Perception can prime action (visuomotor priming), and action can prime perception (motorvisual priming). According to ideomotor theory both effects rely on the overlap of mental representations between perception and action. This implies that both effects get more pronounced the more features they share. We tested this hypothesis by employing in a motorvisual (Exp. 1) and in a visuomotor (Exp. 2) setting, three different pairs of left/right target stimuli (hand pictures, arrows, and words) varying in how strongly they overlap with the pair of left/right responses. For two stimulus pairs (hands and words) the hypothesis was confirmed: hand pictures share more features with the responses than words, consequently hand pictures produced a stronger visuomotor and a stronger motorvisual priming effect than words. However, arrow stimuli showed a different pattern: the temporal dynamics of both priming effects, as well as the direction of the effect seen in motorvisual priming, were significant but opposite to that of the hand and word stimuli. This suggests that the arrows’ representations were not involved in ideomotor processes, and we propose instead that they were represented in a spatial or scalar fashion, outside the representations assumed in ideomotor theory. The results are discussed in the context of ideomotor theory, and the planning and control model of motorvisual priming.