Direct electrical stimulation of the premotor cortex shuts down awareness of voluntary actions

The study addresses how conscious motor awareness of voluntary actions emerges from brain processes that generate willed actions. Neuropsychological evidence from anosognosia for hemiplegia (AHP) suggests that lesions to the premotor cortex (PMC) impair awareness by preventing detection of mismatches between intended and executed movements. Based on comparator models positing that motor awareness arises from matching predicted and actual motor outcomes, the authors hypothesize that PMC is a key node supporting both motor execution and motor awareness. They test whether transient disruption of PMC during voluntary movement interferes not only with motor execution but also with awareness of movement, contrasting this with stimulation of primary somatosensory cortex (S1), which is expected to affect execution but spare awareness. The work is conducted intraoperatively in awake brain tumor patients, offering a direct causal test of PMC’s role in human motor awareness.

Background work indicates AHP patients may deny paralysis despite motor failure, implicating fronto-parietal networks including PMC, insula, and temporo-parietal junction in motor awareness deficits. Comparator-system theories propose that efference copies and forward models generate predictions that are compared with sensory feedback to yield motor awareness. Prior DES studies showed negative motor responses (movement arrest) in premotor and parietal cortices, and PPC’s role in online control and movement intention. However, evidence also suggests that motor awareness can be independent of somatosensory feedback from S1 in both healthy individuals and patients, and that ventral PMC may support motor monitoring, with increased activity during impossible movements and effects of noninvasive stimulation on uncertainty about performance. The authors position PMC, particularly its ventral sector, as a candidate locus where predicted and sensed outcomes are compared to generate conscious motor awareness.

Design: Two intraoperative experiments during awake brain surgery in 12 patients with low-grade gliomas affecting frontal, parietal, or temporal regions in the left hemisphere. Patients had no preoperative language, apraxia, motor, or somatosensory deficits. All provided informed consent; protocol approved by ethics committee. Tasks: Hand-manipulation task (HMt): rhythmic grasp–hold–turn of a cylindrical handle with the right hand. Motor-monitoring task (MMt): two versions. Online MMt (n=8 patients: 4 with PMC stimulation, 4 with S1 stimulation): patients verbally reported performance in real time, saying OK when execution felt normal and STOP when difficulty was experienced. Delayed MMt (n=4 patients; PMC stimulation): immediately after stimulation, patients answered YES if they believed they had executed the task correctly and NO otherwise. Preoperative blindfolded training included occasional mechanical blocks of the manipulandum to train STOP/NO responses. Intraoperatively, an initial mechanical block trial verified correct MMt performance. Stimulation sites: Target area: ventrolateral PMC (BA6), selected based on expected effects on both motor execution and awareness. Control area: S1 hand/finger sector, immediately posterior to the hand-knob, selected to elicit comparable motor arrest without altering awareness. Due to clinical constraints, PMC and S1 were not exposed/tested in the same patient. Direct electrical stimulation (DES): Low-frequency DES (LF-DES) used during tasks. Parameters: bipolar probe (2-mm ball tips, 5-mm separation), 60 Hz trains of biphasic pulses (0.5 ms each phase) lasting 2–5 s; intensity set to working current that disrupted counting (2–4.5 mA in effective HMt trials reported). High-frequency DES was used only for routine mapping outside the experimental task. Movement-related EMG recorded from hand muscles (FDI, APB), and patients’ vocal signal was digitized for alignment with MMt. Neurophysiological monitoring: EEG/ECoG to monitor cortical activity and detect after-discharges; motor-evoked potentials monitored outside mapping. Free-running EMG captured voluntary and stimulation-evoked activity. Outcome measures: HMt execution (video and EMG-defined movement vs arrest); MMt awareness reports (OK/STOP online; YES/NO delayed). Trials with phonoarticulation disruption during PMC stimulation were excluded from awareness statistics for online MMt. Sample/Trials: Across 12 patients, 47 cortical sites were stimulated; 27 sites (17 PMC, 10 S1) produced complete motor arrest during HMt. Statistics: EMG root mean square (RMS) during DES vs baseline compared with Kruskal–Wallis test and multiple comparisons; EMG normalized within muscle and patient. Awareness outcomes compared between PMC and S1 using Fisher’s exact test, conducted for all PMC trials (pooling online and delayed) and for online-only trials. Disconnection analyses: Indirect structural approach to identify white matter tracts associated with stimulation effects. Generated probability density maps of effective stimulation sites to create virtual lesion volumes for PMC (extended to precentral sulcus fundus) and S1 (extended to postcentral sulcus fundus). Lesion-based disconnection analysis performed in five AHP patients (right hemisphere lesions flipped to left for comparison) using CT/MRI-FLAIR segmentations normalized to MNI space. Compared presence/probability and voxel proportion of tract disconnections across arcuate fasciculus, SLF II, SLF III, anterior thalamic radiation, fronto-insular tracts, fronto-striatal tracts, inferior fronto-occipital fasciculus, corticofugal/pontine tracts, and U-shaped postcentral–precentral fibers.

• Motor execution: DES on both PMC and S1 reliably and reversibly arrested right-hand movements during HMt at 27/47 stimulated sites (17 PMC, 10 S1). EMG confirmed significant muscle activity suppression vs baseline: PMC H=43.29, p=0.000; S1 H=28.38, p=0.000. No movements or EMG activation were evoked at rest by DES on PMC or S1.
• Motor awareness during online MMt: During PMC stimulation, in 88.9% (8/9) of effective trials patients reported OK (normal execution) despite complete motor arrest. During S1 stimulation, in 100% of effective trials patients reported STOP, correctly recognizing motor arrest. Trials with transient phonoarticulation disruption during PMC stimulation (4 trials) were excluded; in those, patients reported normal execution once speech returned, consistent with other PMC trials.
• Motor awareness during delayed MMt: In 4/4 PMC trials (4 additional patients), patients answered YES (believed they executed correctly) despite movement arrest.
• Statistical comparison: Significant association between altered MMt trials and stimulated area (PMC vs S1) for both analyses (all PMC trials pooled and online-only), Fisher’s exact test p<0.001, indicating impaired awareness specific to PMC stimulation.
• Disconnection analyses: Lesion-based disconnection across five AHP patients commonly involved nine tracts, prominently the arcuate fasciculus and SLF II/III. PMC virtual disconnection overlapped with five tracts; high-probability disconnections included arcuate fasciculus and SLF II/III; medium probability included fronto-insular and fronto-striatal tracts. S1 virtual disconnection predominantly involved U-shaped postcentral–precentral fibers (>80%) with minimal involvement of SLF II/III (<20%). This indicates that temporo-parieto-premotor long-range networks (SLF II/III, arcuate) are common to AHP lesions and PMC stimulation effects, whereas S1 effects map to short-range primary sensorimotor connections.

The data provide causal evidence that ventral PMC is necessary for conscious awareness of voluntary action execution. While both PMC and S1 stimulation produced comparable negative motor responses (transient arrest of ongoing voluntary movement), only PMC disruption led to erroneous awareness judgments: patients believed they were moving normally despite complete motor arrest. This intraoperative phenomenon mirrors explicit unawareness in anosognosia for hemiplegia and supports models positing a PMC-based comparator that integrates predicted consequences of motor commands with sensory feedback to yield veridical motor awareness. The preserved awareness during S1 stimulation indicates that conscious motor monitoring does not solely depend on somatosensory feedback from S1 and can be maintained when somatosensory input is perturbed, consistent with prediction-based accounts. Comparisons with PPC stimulation literature suggest functional specialization: PPC appears key for online action control and can inhibit volitional movement while preserving awareness, whereas PMC disruption here simultaneously affected execution and awareness during voluntary, intentional action. Anatomical considerations point to effective stimulation sites in dorsal ventral PMC (putative human homolog of monkey area F5), involved in grasp-related sensorimotor integration. Disconnection analyses further link PMC’s role in awareness to long-range fronto-parietal pathways (SLF II/III, arcuate), overlapping with networks implicated in AHP, whereas S1 effects relate mainly to short-range U-shaped fibers to M1. Clinically, identifying and sparing motor-awareness-critical cortex during tumor resections may improve rehabilitation outcomes, analogous to the negative impact of AHP on stroke recovery.

During voluntary hand movements, direct electrical stimulation of PMC and S1 both interrupted movement execution, but only PMC stimulation profoundly disrupted patients’ awareness of this motor arrest. These results identify the premotor cortex as a shared neural substrate for motor execution and conscious motor awareness, acting as a key hub of a temporo-parieto-premotor comparator network. Future work should test hemispheric and effector specificity (right vs left PMC; contralateral vs ipsilateral hand), include larger samples with more stimulation sites and control conditions (e.g., catch trials), and further delineate interactions with PPC and cerebellar contributions to motor monitoring. Incorporating motor-monitoring assessments into intraoperative mapping may help preserve awareness-related functions and support postoperative recovery.

• Clinical constraints limited sample size per site and number of stimulations (17 effective PMC trials across 8 patients; 10 effective S1 trials across 4 patients); simplified design without catch trials.
• Only left-hemisphere tumor patients were studied; right-hemisphere effects and ipsilateral/contralateral interactions were not tested.
• PMC and S1 were not exposed/stimulated within the same patient, preventing within-subject comparisons.
• Intraoperative setting precluded visual feedback of the moving hand; reliance on verbal reports may be affected by transient phonoarticulation disruptions during some PMC stimulations (such trials were excluded from awareness analyses).
• Anatomical heterogeneity of stimulation sites within PMC may limit precise subregional attribution.