Even single sessions of motor practice can lead to long-term storage of movement representations in the brain . It is now clear that after practice has ended the functional properties and representation of skilled movement continues to evolve in the brain [4–7]. These changes are evident in the gradual development of resistance to interference from other behaviours as time passes after task practice [4, 5]. In some cases motor skills are not merely stabilized but can be improved through this consolidation process [6, 7]. Indeed, this is what we discovered when we paired 5 Hz excitatory rTMS to left PMd with motor task practice; motor skill continued to improve off-line after practice. Conversely, participants who practiced the motor task and received either inhibitory or sham stimulation showed only memory stabilization, there was no further between session improvement, but rather a relative preservation in motor skill level acquired via practice . Though these two forms of memory consolidation are not mutually exclusive, our data suggest that PMd has a role in off-line motor skill enhancement. Importantly, we separated the short-term effects of practice from more permanent changes in behaviour demonstrated at retention by performing these tests on different days . This experimental feature allowed us to view off-line learning in the Excite group without any interference effects from practice .
Critically, the off-line motor consolidation demonstrated by the Excite group was related to sequence-specific motor learning rather than to generalized improvements in motor control associated with task practice. Illustration of this point is evident in the difference in tracking error across groups for the repeated sequence at retention (Figure 4); the three groups demonstrated equivalent performance on random sequences at the same time. Random sequence tracking reflects generalized motor execution whereas repeated sequence performance shows motor learning [27, 28, 43]. Thus, the role of PMd in motor consolidation in the present work related to implicit sequence-specific learning rather than an overall improvement in the generalized ability to track continuous sequences.
There has been debate as to whether PMd activity relates to motor learning  or to the recall of already learned movements [5, 20]. Our data support the hypothesis that PMd activity facilitates motor learning, specifically by aiding memory consolidation. Two features of our data support our conclusion. Critically, excitatory stimulation to PMd promoted off-line learning. We expected that if PMd played a role in recall of learned movements rather than in motor learning, we would have noted memory stabilization rather than off-line improvements. Second, the improvements associated with excitatory stimulation to PMd were sequence-specific and not simply related to generalized motor control improvements.
The influence of PMd activity on motor learning and memory consolidation likely operated though a network of brain regions. PMd is ideally situated to impact a broad range of cortico-cortical and cortic-subcortical networks. On-line rTMS-fMRI imaging has shown that excitatory stimulation of PMd increases the BOLD signal both locally (in PMd, PMv, supplementary motor area, somatosensory cortex, and cingulate motor area) and distantly (in contralateral PMd, cerebellum, putamen and caudate; ). Further, these rTMS driven modifications in hemodynamics occur even in the absence of overt motor responses. This pattern of brain activation associated with rTMS to PMd reflects the known anatomical and functional connectivity amongst these regions [14, 15, 45]. Though we cannot ascribe the offline learning we documented to any single region within this broad network, it is evident that 5 Hz rTMS stimulated motor memory consolidation most likely via up-regulating at least some elements of both local and distantly connected brain regions.
We expected that the Inhibit group might have demonstrated worse behaviour than those participants in the Sham stimulation condition. However, it may be that the positive effects of practice on accuracy of motor tracking performance countered the impact of 1 Hz stimulation. Indeed, we and others [26–28, 46] have shown that motor task practice of continuous tracking tasks may be well learned over as few as three practice sessions. It is possible that in the present study the effects of 1 Hz TMS was either overcome by motor practice or that the network of brain regions associated with motor learning [5, 21, 45] was able to compensate for less PMd function following inhibitory stimulation.
Though sleep may have played a role in the consolidation we noted across our experimental groups it cannot explain the lower tracking error for repeated sequences shown only by the Excite group at retention. Each of our groups slept in between the last practice day and the retention test; conferring the benefit of sleep on motor skill consolidation regardless of group assignment. Further, past work has demonstrated that off-line improvements in implicit motor learning in young, healthy controls are not sleep dependent . Instead, sleep related improvements in motor skill may develop equally well over the day as they do over the night . Thus, we do not believe that the sleep-induced benefits that are associated with consolidation can account for our findings.
It is also unlikely that differences in explicit knowledge explain any of our group differences across practice or at retention; none of the groups gained explicit awareness of the repeating sequence. In addition, past work [27, 48] has not shown a benefit of explicit knowledge for motor learning of tracking tasks. Based on the results of our explicit tests we are confident that our data reflect changes associated with the implicit motor learning system.
We were surprised at the large number (n = 16) of individuals who required TMS intensity to be reduced owing to inadvertent motor twitching during PMd stimulation. These individuals were from the Excite and Inhibit groups alike. One possible explanation is that the threshold for stimulating primary motor cortex (from which we derived our resting motor threshold) is not the same as the threshold in other brain regions [45, 49]. Specifically, it is possible that PMd has a lower threshold for stimulation than M1; thus, stimulation of PMd may have either activated M1 via PMd-M1 connections or recruited descending tracts from PMd that normally would not fire at lower intensities. Future work will have to endeavor to develop methods for thresholding stimulation intensity more appropriately for regions outside motor cortex.