Our data show that single TMS pulses applied to the left primary motor cortex evoke stronger MEP amplitudes in the contralateral APB muscle during mental rotation as compared with a resting baseline. This holds for all categories of rotated objects except hands. Furthermore, the extent of the MEP increase is similar for the rotation of 3-D Shepard & Metzler figures, houses, tools and 2-D figures. This finding is in line with previous results from our lab that additionally showed the exclusiveness of increased MEP amplitudes for mental rotation compared with several control conditions .
The strategy hypothesis [3, 6] predicted that pictures of hands as well as tools are more likely to trigger the use of an internal strategy. This is based on the proposition that it is easier to imagine rotating ones own hands or tools. The use of such a body-related rotation strategy is supposed to cause primary motor cortex involvement and should therefore lead to higher corticospinal excitability as reflected by higher MEP amplitudes. On the other hand, we hypothesized that the abstract 3-D Shepard & Metzler figures, 2-D figures as well as pictures of houses would not cause M1 activation, because these stimuli would be more likely to trigger the external strategy: Subjects would imagine the objects being rotated by external forces. However, our results do not support this hypothesis. We did not find a significant difference between the MEPs obtained during mental rotation of the different figures.
One might argue that it is not the particular mental rotation strategy that modifies M1 activation but the difficulty of the mental rotation task. If task difficulty was the main reason, the increase in effort to mentally rotate objects would have been expected to cause the entire neural circuit to operate at a higher activation level compared with the less demanding task. However, we did not find significant correlations between task difficulty (indicated by RTs and error rates) and MEP amplitudes. Given that task difficulty is also a subjective experience, we also examined whether there is a relationship between subjective task difficulty and MEP amplitudes. Again, there was no relationship indicating the independence of MEP amplitudes and, thus, M1 activation (including activation of the corticospinal tract) from subjectively experienced task difficulty.
It could also be argued that subjects did not follow the instructions to use mental rotation for the task, since we did not examine the relationship between angle disparity between the objects and reaction times. This, however, is very unlikely because all subjects reported having used the general strategy of mental rotation during the experiment. They all completed the post-experimental questionnaire and agreed with at least one of the given descriptions of mental rotation as corresponding with their performed operation. Examining the post-experimental ratings of strategies used for mental rotation in more detail, we found that the subjects all but exclusively reported the use of an external strategy for Shepard & Metzler figures, houses, tools and 2-D figures, with the exception of the hand stimuli for which only a marginal difference was found between the frequency of used strategies. Even though the questionnaire additionally asked for other strategies not assed by the two given descriptions, none of the subjects indicated the use of a different strategy. Obviously, the so-called external strategy is the most common used strategy in our study more or less independent from the rotated objects.
Given that task difficulty as well as differences between strategies cannot explain why M1 and the adjacent corticospinal tract are activated during mental rotation, what else could be a satisfying explanation? In some way, this study confirms the conclusions from Eisenegger et al.  that mental rotation itself causes the excitability of M1. One possibility is that M1 is directly involved in mental rotation. Results of Georgopoulos et al.  demonstrated the existence of direction-sensitive neurons in M1. These neurons could also play a role in planning and imagining of the mental rotation process. Single pulse TMS applied over the hand as well as the foot area of M1 showed that simply listening to sentences involving hand and foot actions modulated MEP amplitudes measured at the corresponding muscles, respectively . MEP amplitudes evoked by TMS applied to M1 were also modulated by only visualizing motor actions without acting them out , and performance in mental rotation of body parts could be disturbed by TMS and intra-cortical stimulation [14–16]. These findings point towards a direct involvement of M1 in imagining the rotation process but could also be restricted to body parts and depend on the exact task.
The other possibility is that the strong excitability of M1 during mental rotation simply reflects a spill-over effect from adjacent and strongly activated brain regions [6, 17]. Based on findings demonstrating that posterior parietal cortex and premotor cortex are activated in mental rotation, it was concluded that these areas are involved in spatial transformations and operations [2, 4, 27–29]. From studies in monkeys and humans it is known that the lateral premotor cortex and M1 are densely interconnected both anatomically and functionally [30–36]. In addition, it has been shown that TMS-induced neural activity does show spill-over effects, even at a sub-threshold level [37–39]. Bestmann et al.  used simultaneous TMS and fMRI to show that rTMS applied to the primary motor cortex led to changes of the haemodynamic response not only in M1, but also in anatomically and functionally connected regions such as primary sensory cortex (S1), SMA, dPMC, Cingulum, Putamen and Thalamus. In a later paper of the same group, similar functional remote effects were reported for rTMS applied over the dPMC . Altogether, these findings provide strong support for the hypothesis that task-inducted neural activity could also lead to spill-over effects. That there may be no causal involvement of M1 in mental rotation is therefore possible. This explanation would be in line with a brain imaging study that reported only premotor cortex but not M1 activation during mental rotation using motor imagery . M1 may be excitable only because of the activation of premotor cortex. Under real-life conditions, many cognitive processes prepare for subsequent actions. Premotor cortex activation during the mental rotation of objects could therefore prepare subjects for the possibility of acting on them . Such preparation must not be bound to the strategy used for mental rotation; it could be executed for all objects generally. Following this argument, pictures of hands might represent an exception since they do not require the preparation of object-related, manipulative actions. Subjects may have simply perceived them as body parts, thus rendering a preparatory response to act on them unnecessary. One can speculate that premotor cortex activation for hands did not have to be as strong as for the mental rotation of the other objects, especially since subjects indicated that they did not make predominant use of a strategy that involved direct motor imagery, and that the over-spilling activation to M1 as well as the excitability of M1 was therefore comparably smaller. Another explanation is that visualizing a rotating body part (e.g. a hand) led to inhibition of subjects' action schemas and therefore to decreased MEP amplitudes, such as observed when subjects listen to sentences involving action , and that this may have occurred because the instructions required subjects not to act during the TMS session. However, finding a compelling explanation for the differences between objects and body parts is beyond the focus of this study and should be the target of more specific and systematic studies in the future.