Another parallel to the visual system exists with respect to studies on the influence of imagery on cortical activity. We have so far detailed how properties of motor output can be traced in corresponding activity changes in M1, and we have reviewed evidence showing that factors over and above motor output contribute to and modulate Ml activity. But a question that has tickled many researchers' minds for a long time has been to what extent does the mental simulation or rehearsal of movements share neural activity patterns with actual movement execution. This is clearly the case for many of the upstream areas involved in motor planning and preparation, but the contribution of M1 has remained controversial.
Several of the earlier and in part anecdotal reports on motor imagery studied by fMRI were negative with respect to Ml activation. However, Roth et al.107 reported M1 activation during mental execution of a finger-to-thumb opposition task in four of six subjects studied, in addition to effects in the premotor and supplementary motor areas. This was also found in a more detailed study the same year by Porro et al.108 using the same motor task but with a control task of visual imagery. The same group later studied the issue of ipsilateral motor cortex activation during motor imagery.109 The ipsilateral effect was just significant in a region-of-interest group analysis and, due to the analytical methods applied, it remains unknown in how many of the individuals it was significant. Yet, the analysis provided convincing data showing that the motor imagery effect occurred in those ipisilateral voxels that were also active during motor performance. Furthermore, the observation of greater effects in the left hemisphere (i.e., of ipsilateral activation in the case of left-hand movements) parallels the pattern observed for ipsilateral activation during overt motor performance. This notion receives further support from studies that have found a similar congruence for the activations during executed and imagined movements in relation to the somatotopic representation of hand, foot, and tongue movements.110111 Together, these studies suggest that motor imagery yields quantitatively smaller but qualitatively similar patterns of activation in the M1 as motor execution.
However, not all laboratories have obtained this type of result. Across different but probably comparably sensitive approaches there have also been recent studies that reported negative findings regarding M1 activation during motor imagery.112-114 How can these apparent discrepancies be accounted for? A simple view would conclude that some experiments or laboratories simply did not have enough sensitivity or power, or that they applied too conservative statistical thresholds to detect effects in M1. This appears rather unlikely because the degree of signal change in those studies that did observe activation in M1, and its relation to the signal change during movement execution, should have been readily detectable by other laboratories as well.
The opposite approach would be to assume that the cases with positive findings are accounted for by movement execution during (and despite the instruction of) motor imagery. In contrast with the visual system, which can easily be deprived of input, the situation is far more complicated for motor output. There is continuous output, and imagery easily elicits electromyographic (EMG) activity above this resting level. Voluntary relaxation, which in itself may activate M1 as discussed above, and suppression of such imagery-induced activity is difficult to achieve and usually requires training. The problem of controlling for involuntary movement during imagery has been noted by several groups. One way of ensuring "pure" imagery would be to perform extensive EMG monitoring. EMG has indeed been used in the context of imagery studies. In the study by Porro et al.,109 EMG recordings showed some activity increases during the imagery condition in roughly half of the subjects. However, these recordings only covered two sites and were obtained offline. In that sense, it is doubtful whether the lack of correlation observed with the fMRI findings is conclusive. Similarly, Lotze et al.115 used EMG to train subjects via biofeedback to minimize muscular activity during imagery, but obtained no EMG recordings during the fMRI sessions. It is therefore not clear to what extent the learned pattern may have progressively decayed during those sessions. In that sense, it is probably fair to say that to date no study that has reported M1 activation during motor imagery has provided sufficient support for the claimed absence of muscular activity during that condition.
So far, one of the few studies using on-line EMG recordings during fMRI of motor imagery was that of Hanakawa et al.,116 and they did not find significant M1 activation. However, they used an interesting analytical approach. Instead of qualitatively mapping activation under different conditions with a somewhat arbitrary threshold, the authors addressed the quantitative relation of activation effects under imagery and execution of movement. They determined areas with movement-predominant activity, imagery-predominant activity, and activity common to both movement and imagery modes of performance (movement-and-imagery activity). The movement-predominant activity included the primary sensory and motor areas, the parietal operculum, and the anterior cerebellum, which had little imagery-related activity (-0.1~0.1%), and the caudal premotor areas and Brodmann area 5, which had mild to moderate imagery-related activity (0.2~0.7%). Many frontoparietal areas and the posterior cerebellum demonstrated movement-and-imagery activity. Imagery-predominant areas included the precentral sulcus at the level of the middle frontal gyrus and the posterior superior parietal cortex/precuneus.
One of us used a different approach for dissociating the effects of motor imagery and actual movements during fMRI measurements.114 In this study, subjects were presented with drawings of hands and asked to quickly report whether they were seeing a left hand or a right hand, regardless of the angle of rotation of each stimulus from its upright position (rotation). Several psychophysical studies117119 have demonstrated that subjects solve this task by imagining their own hand moving from its current position into the stimulus orientation for comparison. This motor imagery task was paired with a task known to evoke visual imagery, in which subjects were presented with typographical characters and asked to quickly report whether they were seeing a canonical letter or its mirror image, regardless of its rotation.120 Behavioral (reaction times) and neural correlates (BOLD) of motor and visual imagery were quantified on a trial-by-trial basis, while EMG recordings controlled for muscular activity during task performance. Using a fast event-related fMRI protocol, imagery load was parametrically manipulated from trial to trial, while the type of imagery (motor, visual) was blocked across several trials. This experimental design permitted to isolate modulations of neural activity driven by motor imagery, over and above generic imagery- and performance-related effects. In other words, the distribution of neural variance was assessed along multiple dimensions, namely the overall effects of task performance, the specific effects of motor imagery, and the residual trial-by-trial variability in reaction times unaccounted for by the previous factors. With this approach, it was found that portions of posterior parietal and precentral cortex increased their activity as a function of mental rotation only during the motor imagery task. Within these regions, parietal cortex was visually responsive, whereas dorsal precentral cortex was not. Crucially, fMRI responses around the knob of the central sulcus, i.e., around the hand representation of primary sensorimotor cortex,56,121 correlated significantly with the actual motor responses, but neither showed any relationship with stimulus rotation, nor did they distinguish between motor and visual imagery. This result indicates that, at the mesoscopic level of analysis by fMRI, putative primary motor cortex deals with movement execution, rather than motor planning. However, it remains to be seen whether this finding is limited to a precise experimental context, namely implicit motor imagery, or whether it represents a general modus operandi of the human M1.
So should one conclude that those studies that did find activations in M1 during motor imagery were confounded, e.g., by associated motor output? Let us again turn to the visual system for an analogy. In visual cortices, sensory effects are readily detected in early areas and become progressively difficult to follow the deeper one ascends into the cortical hierarchy. Conversely, the participation of primary visual cortex in mental imagery has been far more difficult to demonstrate and does not reach the strength of effects that visual imagery evokes in higher-order areas.122 So far, this nicely parallels the pattern described in the studies by Hanakawa et al.116 and Gerardin et al.112 that were discussed above. Nonetheless, there is now a consensus that the primary visual cortex can participate in imagery, and this may depend on specific aspects of the paradigm employed, such as the requirement of processing capacities that are best represented at this cortical level. If we attempt to transfer this analogy to the sensorimotor system we must analyze in greater depth the paradigms employed across the various motor imagery studies.
Motor imagery can be carried out in a predominantly visual mode (imagining seeing one's own hand moving) or in a kinesthetic mode (imagining the proprio-ceptive sensations one would experience if one moved the hand in the mentally simulated way). Indeed, it seems to be the case that only studies employing the latter strategy have reported robust effects in M1. This means that the fMRI responses would then be accounted for, not necessarily by the executive neural elements in M1, but by those dealing with proprioceptive input in the context of movement. Psychophysically, it was found that motor imagery affects the illusory perception of movement created by a purely proprioceptive stimulus.102 However, in a related functional neuroimaging experiment, the authors found no M1 response during imagery and, accordingly, overlap of activations from these two conditions was confined to nonprimary motor areas. It should be noted that this experiment was carried out using PET, and it may therefore have suffered from sensitivity or spatial resolution limitations. At the same time, the authors reproduced their finding of M1 activation from the illusion, and it hence seems unlikely that this should be accounted for by the movement illusion rather than by proprioceptive processing. The issue therefore awaits further investigation.
Was this article helpful?