Awareness of self-generated movements arises from comparing motor plans, and the accompanying (hypothetical) efference copy, with the visual and proprioceptive consequences of movement. Here we used repetitive transcranial magnetic stimulation (rTMS) to investigate the role of a posterior region in the superior parietal lobule (SPL) in this process. Nine healthy volunteers performed a finger extension actively and passively while wearing a CyberGlove; the glove recorded these (actual) finger movements and used this information in real time to move a virtual hand displayed on a computer screen. To assess the participant’s awareness of movement onset, we introduced a delay between the onset of the actual and virtual movement (60–270 ms, 30 ms increments); the task was to judge whether the virtual hand movements were delayed relative to the actual hand movements. Low-frequency rTMS (15 min, 0.6 Hz) was applied either over the left SPL or the left temporal cortex (control site) to decrease excitability of these regions and, in turn, test their role in the awareness of self-generated movement. Following the SPL stimulation, participants’ assessments of asynchrony were impaired for active but not passive movements. No significant changes were observed after rTMS applied over the control site. We suggest that these findings are consistent with the role of the SPL in evaluating the temporal congruency of peripheral (visual) and central (efference copy) signals associated with self-generated movements. As such, this region may contribute to the sense of ‘agency’ and its disturbances in disorders such as apraxia and schizophrenia.
To determine whether or not an action is self-generated requires comparison of the motor plan, and the hypothetical efference copy (Sperry, 1950; von Holst and Mittelstadt, 1950), to the sensory consequences resulting from the production of the action (Teuber, 1964; Soechting and Flanders, 2001). Both visual (i.e. seeing one’s hand move) and proprioceptive (i.e. feeling one’s hand move) inputs are relevant for this decision. This arrangement allows for a comparison between the actual versus the predicted sensory consequences of an action. Several lines of evidence suggest that posterior parietal cortex is involved in this comparison and, in turn, might give rise to the awareness of self-determined actions. The present study tested this hypothesis by applying transcranial magnetic stimulation (TMS) over one of the parietal subregions and, subsequently, assessing consequences of such stimulation on the awareness of self-generated movements. This experiment was inspired by previous observations made by Sirigu et al. (Sirigu et al., 1999) that patients with lesions to the left parietal lobe are impaired in this respect. Before describing in detail the Sirigu study, we will briefly review current evidence implicating the posterior parietal cortex, and the related fronto-parietal circuits, in the integration of visual and somatosensory inputs with motor outputs.
Neuroimaging studies carried out in healthy human participants consistently revealed significant increases in cerebral blood flow in the posterior parietal cortex during the performance of many types of movements (Grafton et al., 1992, 1996; Kawashima et al., 1996; Rizzolatti et al., 1996; Faillenot et al., 1997; Binkofski et al., 1998; Gordon et al., 1998; Ramnani et al., 2001). Fink et al. (Fink et al., 1999) investigated specifically the role of the parietal cortex in monitoring self-generated movements; they examined the effect of providing false visual feedback (i.e. presenting the participant’s mirror-imaged fingers to represent the opposite hand) during in-phase and out-of-phase bimanual finger opening and closing. Significant activations in posterior parietal cortex occurred for both in-phase and out-of-phase movements, with increased activation during the false visual feedback. This pattern of activation suggests that posterior parietal cortex is involved in the comparison of sensory and motor information during self-generated movements.
Single cell recordings in non-human primates provide further evidence that posterior parietal cortex participates in the production and evaluation of self-generated movements. Studies have shown that many neurons in the medial intraparietal sulcus (MIP) respond preferentially to movement (Johnson et al., 1996) but also respond to visual and somatosensory stimuli, especially when these stimuli are within reaching distance of the monkey (Colby and Duhamel, 1991, 1996; Colby and Goldberg, 1999). Furthermore, experiments with macaques have revealed that visual receptive fields of the bimodal neurons enlarge with tool use, extending to represent the new distance in extrapersonal space attainable by reach with the tool (Iriki et al., 1996). Finally, some neurons in MIP change their activity simply in relation to the intention to perform reaching movements (i.e. motor plan), prior to the presentation of the target of the movement and prior to the enactment of the movement itself. The latter observation suggests that, in addition to sensory and tactile/proprioceptive information, MIP receives an efference copy generated in relation to the preparation and/or execution of movement (Anderson et al., 1997; Colby and Goldberg, 1999).
Studies of neural connectivity, using injections of retrograde tracers in macaque monkeys, have further revealed that MIP, along with adjacent posterior parietal areas (e.g. V6A, PEc, PEip) project predominantly to dorsal premotor regions [i.e. PMd or dorso-rostral area F2 (Cavada and Goldman-Rakic, 1989; Rizzolatti et al., 1997, 1998; Wise et al., 1997; Matelli et al., 1998)] and, to a lesser extent, to the primary motor cortex (Wise et al., 1997). In human participants, Chouinard et al. (Chouinard et al., 2002) also demonstrated connectivity between premotor and parietal regions. Combining positron emission tomography (PET) with TMS, they showed that TMS of dorsal premotor areas (i.e. PMd/F2) resulted in a distal blood-flow response in the posterior superior parietal cortex.
A final line of research investigating the role that posterior parietal cortex might play in the determination of self-initiated actions arises from the study of patients with lesions of the posterior parietal lobe. Lesions in the parietal lobe, particularly in the left hemisphere, result in various impairments in the performance and sequencing of actions, namely apraxia (Critchley, 1953; Heilman and Rothi, 1982; Poizner et al., 1995; Sirigu et al., 1995; Haaland et al., 2000). As mentioned previously and particularly relevant to the aim of the current study are findings by Sirigu et al. (Sirigu et al., 1999) suggesting that patients with apraxia are impaired relative to control participants in differentiating self-generated movements from experimenter-generated actions. They tested three such patients with lesions to the left parietal lobe; they were asked to perform movements with their right and left hands while they were prevented from directly observing these actions. Visual feedback was provided via a video monitor and consisted of (a) the participant’s own gloved hand performing the instructed action, (b) the gloved hand of the experimenter performing the instructed action, or (c) the gloved hand of the experimenter performing an action different from that performed by the participant. The participant’s task was to indicate whether the movement presented on the video monitor was his/her own. When visual, proprioceptive, and motor information were identical (i.e. when a participant actually saw his or her own hand performing the instructed movement) patients with apraxia and control participants were highly accurate. Conversely, when visual information was clearly discrepant from proprioceptive and motor information (i.e. when participants saw the hand of the experimenter performing a different action) both groups performed comparably and correctly. However, when greater, although imperfect, overlap existed between visual, proprioceptive, and motor information and finer discriminations were required (i.e. when the experimenter’s hand was presented performing the instructed action) all participants performed more poorly, with the performance of the patients being significantly impaired relative to that of controls. This pattern occurred for both right and left hand gestures, and is consistent with a view that (the left) posterior parietal cortex is necessary for distinguishing actions generated by self from those generated by others, particularly when only subtle differences exist between movements.
The aim of the current study was to test experimentally in healthy participants whether the left posterior parietal cortex is crucially involved in the determination of agency of movements. This study focused on the left parietal cortex in light of the higher prevalence of apraxia following lesions to the left vs right hemisphere, respectively (Freund, 1992). Further, the study design did not permit inclusion of the right hemisphere in the same groups of participants (see below). The approach was similar to that employed by Sirigu et al. (Sirigu et al., 1999). Healthy participants performed finger movements, either extension of the index or middle finger, while wearing a CyberGlove that recorded, in real time, their actual hand movements and used this information to move a virtual hand displayed on a computer screen. Although the movements of the virtual hand matched those of the actual hand exactly, the former were delayed by various intervals relative to the latter. The participants’ task was to identify the trials on which they detected a delay between the actual onset of the action and the onset of the virtual hand movement.
Participants performed one block of these asynchrony-detection trials before and another block after low-frequency stimulation of the left posterior parietal cortex with repetitive transcranial magnetic stimulation (rTMS). Low-frequency (1 Hz) rTMS has been shown to decrease excitability of the motor system (Pascual-Leone et al., 1995; Chen et al., 1997; Gerschlager et al., 2001; van der Werf and Paus, 2002; Chouinard et al., 2003), and has also been used to investigate brain– behaviour relationships [e.g. (Kosslyn et al., 1999; Hilgetag et al., 2001)]. Given the proposed role of posterior parietal cortex in the integration of proprioceptive, visual, and motor information for the evaluation of movements, we expected that asynchrony detection would be impaired following rTMS. This finding would support the necessary function of parietal cortex in the determination of self-generated movements.
Twelve McGill University students (nine females and three males) volunteered to participate in this experiment. Three female participants did not return for the second session of the study and were therefore excluded from analyses. All participants were right-handed and had normal or corrected to normal vision. They ranged from 22 to 27 years of age, with a mean age of 24 years. The study was approved by the Research Ethics Board of the Montreal Neurological Institute and Hospital, and written informed consent was obtained from all participants.
We obtained T1-weighted magnetic-resonance (MR) images from all participants prior to their participation in this experiment. The experiment comprised two sessions on separate days: parietal cortex was stimulated in one session while the temporal cortex (a control site) was stimulated in the other session (Fig. 1). During the first session, we determined the motor threshold for each participant as well as the localization of parietal and temporal rTMS stimulation sites. Otherwise, both sessions proceeded as follows: (a) asynchrony-detection trials, (b) rTMS, and (c) asynchrony-detection trials. The rTMS was applied to the experimental region on 1 day and to the control region on the other in counterbalanced order across participants.
Transcranial Magnetic Stimulation
Determination of Motor Threshold
The threshold was defined as the minimal stimulation intensity required to elicit reliably a twitch in the first dorsal interosseus muscle of the participant’s right hand. Single TMS pulses were applied using a Magstim 200 magnetic stimulator (Magstim, Whitland, Dyfed, UK) connected through the BiStim module, and a figure-of-eight coil to the scalp in the region above the primary motor cortex. The stimulation intensity was varied until a level was reached at which (a) observable muscle twitches were reliably elicited for at least 50% of stimulations, and (b) lesser levels of stimulation failed to consistently elicit these observable muscle contractions. This intensity, the TMS motor threshold, was used to set the intensity of stimulation during the rTMS phases of the experiment.
TMS was applied using the same set up described above. The centre of the figure-of-eight coil was positioned over the experimental (i.e. left parietal cortex) region during one session, and over the control (i.e. left temporal cortex) region on another, occurring in counterbalanced order across participants. Four participants received stimulation of the parietal lobe in the first session and five participants received stimulation of the temporal lobe during the first session. Figure 1 presents an example counter-balancing order and serves as an illustration of the study design. The short axis of the figure-eight coil was perpendicular to the interhemispherics fissure, with the coil current flowing in the latero-medial direction; the coil was held in place by a mechanical holder. Participants were asked to remain immobile and were placed in a chin rest to reduce the probability of movement during stimulation.
The intensity of TMS stimulation was 90% of motor threshold (50–70% of the maximum stimulator output). A total of 550 TMS pulses were administered at a rate of 0.6 Hz. The rTMS phase lasted ~15 min.
Localization of Parietal and Temporal rTMS Stimulation Sites
Frameless stereotaxy was used to target the stimulation sites (see Fig. 2). First, the MRI of the participant’s brain was transformed into standardized stereotaxic space with an automatic feature-detection algorithm (Collins et al., 1994). Next, the exact locations of the experimental region in the left posterior parietal cortex and the control region of secondary auditory cortex in the temporal lobe were identified in standardized stereotaxic space (parietal cortex: x = −36, y = −64, z = 54; temporal cortex: x = −66, y = −16, z = 8); the parietal coordinates are based on the results of our recent TMS/PET study of fronto-parietal connectivity (Chouinard et al., 2003). These MRI locations were then transformed to the participant’s native brain coordinate space with an inverse version of the native-to-stereotaxic transformation matrix (Paus, 1999). Following the determination on the MRI of the experimental and control regions using these participant-specific coordinates, localization of the scalp positions overlying the site was next achieved using a three-dimensional infrared optical-tracking system (Polaris System, Northern Digital Inc., and Brainsight software, Rogue Research Inc.). Using an optically tracked pointer, positioned perpendicularly to the head, we determined and marked with a grease pen the scalp positions overlying the cortical regions of interest. In both sessions, the centre of the coil was placed against these locations on the scalp. The participant was asked not to wash off these marks so that there was no need to perform the registration on the second day of the experiment.
To perform the asynchrony-detection task, participants wore a Cyber-Glove (Immersion 3D Interaction Inc.; www.immersion.com/products/3d/interaction/cyberglove.html) that recorded, in real time, the movements of their wrists and fingers and used this information to move a virtual hand on a computer screen. In this way, movements of the virtual hand corresponded exactly to the movements of the participant’s actual hand. The participants saw only the virtual hand displayed on the computer screen, while their actual hand, fitted with the data glove, was obscured from view inside a black box positioned directly in front of them. The visual feedback presented on the computer screen was asynchronous with the actual onset of movement. The onset of the virtual hand was delayed at various intervals ranging from 60 to 270 ms, in 30 ms increments, relative to that of the actual hand. The participant’s task was to indicate on each trial whether they detected an asynchrony between the onset of movements of the virtual hand and those of their actual hand.
During each session (i.e. Parietal vs Temporal), all participants performed 128 asynchrony-detection trials both pre-and post-rTMS. Participants performed the task actively for one block of trials (i.e. 64 trials) and passively for another. The order of tasks (i.e. Active vs Passive) pre- and post-rTMS was counterbalanced across participants (See Fig. 1). In a given subject, the same order of tasks for the pre- and post-rTMS blocks was used in the Parietal and Temporal sessions. The inclusion of the passive task allowed us to evaluate possible impairment in the processing of movement-related proprioceptive inputs. For each task, there were eight trials at each of the delay durations, from 60 ms to 270 ms in 30 ms increments. The trials at these various delays were presented in randomized order. For trials that were performed actively, the participant extended either the index or the middle finger of his or her right hand, depending on the oral instruction given by the experimenter at the beginning of the trial. For the passive trials, the participant’s index or middle finger of his or her right hand was extended by the experimenter. This was accomplished with padded finger rings placed on the ends of the participant’s index and middle fingers, connected with clear nylon wire via a pulley system to analogous finger rings that were placed on the experimenter’s fingers. As the experimenter flexed her index or middle finger, the participant’s corresponding finger was passively extended.
Each trial proceeded as follows. The virtual hand appeared on the screen. In the active block of trials, the experimenter gave the oral instruction to extend the index or the middle finger. The participant actively performed the movement. In the passive block of trials, the participant’s index or middle finger was extended passively without prior notice as to the finger that would be extended; no prior notice was given to minimize the participant’s active involvement in preparing to move the finger. The virtual hand displayed the corresponding action on the computer monitor at various intervals of delay relative to the participant’s actual finger movement. The participant responded ‘yes’ or ‘no’ to indicate whether or not a delay was detected. A blank screen then appeared for 1 s prior to the onset of the next trial. Each trial lasted on average 5 s. In each block of trials, there were an equal number of index and middle finger extensions, occurring in random order.
A t-test was conducted on the proportion of delay-detected trials per task prior to rTMS, irrespective of cortical region stimulated. The proportion of delay-detected trials was calculated as the number of trials per task on which the participant detected a delay between the actual finger movement and the virtual finger movement, divided by the total number of trials for that task. The analysis revealed no difference in terms of accuracy of delay-detection between the active and passive tasks, t = 1.261, P > 0.200.
Effects of rTMS
Our hypothesis was that rTMS applied over the parietal site would impair the detection of movement delay for active trials only. Therefore, we used an a priori single degree-of-freedom contrast of the parietal-active condition relative to all other Region × Task conditions (i.e. parietal-passive, temporal-active, and temporal-passive). This contrast was highly significant, F(1,8) = 19.823, P < 0.002, reflecting a decrease in the proportion of delay-detected trials in the active condition after rTMS stimulation of the parietal region (Table 1). Student t-tests confirm that a significant difference in performance of the detection of asynchrony arose only in the parietal-active conditions when comparing pre-rTMS and post-rTMS scores, t = 4.966, P < 0.001, whereas for all other conditions (i.e. parietal-passive, temporal-active, and temporal-passive), t < 1. Post-rTMS of posterior parietal cortex, participants appreciated the delay between their actual hand movements and those of the virtual hand significantly less frequently than pre-rTMS. The difference scores (post-rTMS minus pre-rTMS) of the proportion of delay-detected trials per task, for each region of TMS stimulation are presented in Figure 3. Figure 4 presents the difference scores (post-rTMS minus pre-rTMS) of the proportion of delay detected trials per participant for each task and each region of TMS stimulation.
A second a priori single degree-of-freedom contrast examining the difference scores of the proportion of delay-detected trials in the parietal-active condition with the parietal-passive condition approached significance, F(1, 8) = 4.638, P = 0.063; the post minus pre-rTMS difference was larger in the parietal-active, as compared with the parietal-passive, condition. Again, the latter finding reflects the decrease in accuracy of asynchrony detection in the active condition following rTMS stimulation of the parietal cortex compared to equal performance in the passive condition pre- and post-parietal rTMS.
Finally, the analogous contrast of the active-temporal with the passive-temporal condition was not significant, F < 1. As predicted, performance of the asynchrony-detection task was unaffected in either the active or the passive condition by stimulation of the temporal cortex.
There are two principal findings in the current experiment. First, participants performed comparably at baseline in the active and the passive tasks. This equivalence confirms the appropriateness of the passive task as a control measure. Second, rTMS of a region in left posterior parietal cortex impaired detection of asynchrony between actual and virtual hand movements when these movements were self-generated. Given that rTMS of this region did not alter performance of asynchrony detection when the hand movements were passively performed, a task with equivalent cognitive demands, we can conclude that this change in performance was not due simply to a general impairment in the processing of visual information associated with the virtual-hand movement on the computer screen. Further, the unaltered performance in the passive task suggests that stimulation of the superior parietal lobule (SPL) did not produce a gross impairment in the processing of movement-related proprioceptive input. By including a control session, in which rTMS was applied to a region in temporal cortex without effect on performance, we can also refute the possibility that the main finding in the current study was due to a general effect of rTMS on active movements. Finally, this control session enabled us to appreciate how performance in the active task changed simply with practice and fatigue due to repetition of the task after rTMS. We can therefore dismiss the possibility that our finding was simply a function of task repetition.
The result of main interest in the current experiment is in line with the finding of Sirigu et al. (Sirigu et al., 1999). They found that patients with lesions to the left posterior parietal lobe were more impaired than controls in distinguishing whether movements presented via a video monitor were self-generated versus experimenter-generated. The current study differed from that of Sirigu et al. in a number of important ways, however. Because the disruption of parietal cortex was focal, controlled and temporary, this study was not subject to the confounding effects of relatively large cortical lesions, and to the possible functional reorganization of the affected neural circuits. Second, baseline measures of performance for each participant could be sought before the experimental manipulation (i.e. rTMS). Finally, in the current experiment, we isolated temporal information from kinematic information. The CyberGlove recorded hand position, amplitude, direction, and rotation of movements from distal and proximal inter-phalangeal joints, from metacarpal phalangeal joints, as well as from the wrist, in x, y, z, pitch, roll and yaw coordinates. In this way, the movements of the virtual hand replicated the kinematics of the actual hand movements with great fidelity, and any minor imprecisions in the representation were consistent across trials. Consequently, participants could only rely on the temporal relation between the actual and the virtual hand in evaluating movements. This is in contrast to the Sirigu et al. study in which the experimenter-generated movements varied in terms of kinematic, as well as temporal information relative to the self-generated movements.
In the current experiment, temporal information about movement provided the basis for consistently performing the behavioural tasks. Theoretically, this temporal information about hand movements could be derived in different ways. Participants could compare the position of their actual hand at any point in time with the position of the virtual hand. In this way, asynchrony would be determined by noting discrepancies between proprioceptive (hand) and visual (screen) feedback. Alternatively, participants could use the initiation of a motor programme as a reference point for comparing visual feedback of the virtual hand. That is, asynchrony judgements could be achieved by comparing efference information with the visual feedback in our experiment.
A number of studies have addressed the question of awareness of movement initiation. First, Libet et al. (Libet et al., 1983) asked participants to estimate the onset of their actions. Participants were instructed to note the time on a clock when they initiated a movement. This subjective assessment of movement initiation was then correlated with electromyographic (EMG) information about the actual onset of movement. They found that participants’ estimates preceded the beginning of the actual movement, on average, by 86 ms. Supporting this finding, McCloskey et al. (McCloskey et al., 1983), asked participants to judge the order of occurrence of an auditory tone and a voluntary hand movement. Again using EMG as an objective measure of movement onset, they found that participants only deemed the tone and the movement to be synchronous when the tone in fact preceded the movement. Finally, Haggard and Magno (Haggard and Magno, 1999) replicated the finding of Libet et al. that judgements of movement onset preceded actual movement. In addition, they demonstrated that applying rTMS to premotor cortex increased the delay between a ‘go’ signal and the participants’ awareness that a movement had been initiated. Despite delaying the appreciation of movement onset, the initiation of the actual movement was not delayed. This interesting dissociation between performance and awareness of movement was further maintained by the finding that rTMS of primary motor cortex delayed the onset of movement but not the subjective appreciation of the beginning of movement, which again preceded actual movement, relative to a control condition.
Together, these findings are inconsistent with a view that ascribes decisions about self-initiation of movement to sensory (proprioceptive) feedback. This view would require awareness of action onset to follow actual enactment of a movement. Rather, the studies reviewed here suggest that determinations of onset of self-generated actions are anticipatory, favouring an account whereby decisions about the timing of actions result from reference to efference information. That is, the efference copy acts in some manner as a time-stamp for the initiation of movement. The current findings are also interpreted in this light.
Following rTMS of the left SPL in human participants, virtual hand movements that were delayed relative to actual, self-generated hand movements were judged as synchronous significantly more often than before rTMS of this region. Stimulation of SPL with rTMS caused a delay in the awareness of self-generated movements. Participants became more likely to judge the movements of the virtual hand, which were in fact delayed by various intervals ranging from 60 to 270 ms, as occurring simultaneously with the initiation of the actual movement. This finding is convergent with the results obtained by Haggard and Magno (Haggard and Magno, 1999) following rTMS of premotor cortex. In fact, we speculate that the stimulated parietal site is part of the same fronto-parietal network investigated by Haggard and Magno, the function of which is to plan, predict, and evaluate movements of self. This view is plausible given the sensori-motor integration functions that have been ascribed to MIP (Rushworth et al., 1997aRushworth et al., 1999; Wise et al., 1997; Colby and Goldberg, 1999) in addition to the reciprocal connectivity that exists between MIP and premotor regions (Rizzolatti et al., 1997; Wise et al., 1997; Matelli et al., 1998; Chouinard et al., 2003). Recall that TMS of a premotor region (PMd/F2) resulted in significant activation of a distal region in the posterior superior parietal cortex (Chouinard et al., 2003); the coordinates for the parietal stimulation site in the current study in fact were derived from that study. Ours is a study among a number of others coupling premotor and parietal regions in the production, assessment, and correction of movement (Critchley, 1953; Hyvarinen, 1982; Kalaska, 1991; Rizzolatti et al., 1997; Rushworth et al., 1997a; Wolpert et al., 1998; Desmurget et al., 1999; Iacoboni et al., 1999).
In summary, our findings suggest that a region of posterior parietal cortex, which may be analogous with macaque MIP, is crucially involved in the assessment of self-generated movements. That is, decreased cortical excitability of this region impairs awareness of self-initiated movements. We contend that this impairment arises due to interference with efference information about a self-generated movement along a fronto-parietal pathway involving premotor regions and MIP. Future studies will further explore this pathway as well as investigate any hemispheric asymmetry that might exist.
This study was supported by the Canadian Institutes of Health Research and the Canadian Foundation for Innovation. We thank Drs Ysbrand van der Werf, Marie-Helene Grosbras and Kate Watkins for their help in determining motor thresholds and positioning the stimulating coil and Jason Lerch for programming the CyberGlove software. We also thank Drs Kate Watkins and Marie-Helene Grosbras for useful comments on the manuscript.
Address correspondence to Tomáš Paus, Montreal Neurological Institute, 3801 University, Montreal, Quebec, H3A 2B4, Canada. Email: firstname.lastname@example.org.
|Region of rTMS||Active||Passive|
|Parietal||0.47 (0.04)||0.37 (0.04)||0.42 (0.04)||0.41 (0.04)|
|Temporal||0.39 (0.03)||0.38 (0.05)||0.37 (0.04)||0.39 (0.05)|
|Region of rTMS||Active||Passive|
|Parietal||0.47 (0.04)||0.37 (0.04)||0.42 (0.04)||0.41 (0.04)|
|Temporal||0.39 (0.03)||0.38 (0.05)||0.37 (0.04)||0.39 (0.05)|