Supplementary motor area provides an efferent signal for sensory suppression

Cognitive Brain Research 19 (2004) 52 58 Research report Supplementary motor area provides an efferent signal for sensory suppression Patrick Haggard*, Ben Whitford Institute of Cognitive Neuroscience and Department of Psychology, University College London, Alexandra House, 17 Queen Square, WC1N3AR, London, UK Accepted 30 October 2003 www.elsevier.com/locate/cogbrainres Abstract Voluntary actions produce suppression of neural activity in sensory areas, and reduced levels of conscious sensation. Recent computational models of motor control have linked sensory suppression to motor prediction: an efferent signal from motor areas may cancel the sensory reafferences predicted as a consequence of movement. Direct evidence for the efferent mechanism in sensory suppression has been lacking. We investigated the perceived size of finger-muscle twitches (MEPs) evoked by TMS in eight normal subjects. Subjects freely chose on each trial whether to make or withhold a voluntary flexion of the right index finger, in synchrony with an instructional stimulus. A test MEP occurred at the instructed time of action. The subject then relaxed and a second reference MEP occurred a few seconds later. Subjects judged which of the two MEPs was larger. Subjects perceived the first test MEP to be smaller in trials where they made voluntary actions than on trials where they did not, demonstrating sensory suppression. On randomly selected trials, a conditioning prepulse was delivered over the supplementary motor area (SMA) 10 ms before the pulse producing the test MEP. The SMA prepulse reduced and almost abolished the sensory suppression effect in voluntary action trials. We suggest the SMA may provide an efferent signal which is used by other brain areas to modulate somatosensory activity during self-generated movement. D 2003 Elsevier B.V. All rights reserved. Theme: Motor systems and sensorimotor integration Topic: Cortex Keywords: Action; Sensation; Motor control; Frontal lobes; Human; Sensorimotor integration 1. Introduction Recent theories of motor control emphasise neural prediction of movement outcomes from an internal efference copy [19]. Recently, motor prediction has also been suggested to explain the widespread finding of suppressed somatosensation during voluntary movement [3]. For example, the sensations produced when trying to tickle oneself are perceptually less ticklish than when the same movements are made by an external agent. An efference copy of the motor command sent to the muscles may be used by an internal predictive model to predict the sensory consequences of the command. The predicted consequences are compared to delayed somatosensory feedback: if these match perfectly, cortical perceptual systems may not fully * Corresponding author. Tel.: +44-20-7679-1164/1177; fax: +44-20- 7813-2835. E-mail address: p.haggard@ucl.ac.uk (P. Haggard). process the afferent signal, as it adds no new information to the prediction [3,17]. Such models depend critically on an efference copy signal. However, the precise neural source of this signal remains unclear. The motor areas of the frontal cortex have appropriate anatomical connections with two brain structures that may provide internal predictive models, namely the cerebellum and parietal cortex. For example, motor cortical areas are known to send axon collaterals to the cerebellum, and patients with cerebellar damage behave as if they had no predictive representation of their own limb movements [9]. In addition, the parietal cortex plays an important role in monitoring actions: patients with parietal lesions are impaired in action-recognition tasks [16]. The parietal cortex also receives frontal signals which update sensory processing [13,16]. Nevertheless, several questions about the relation between sensory suppression and motor prediction remain unanswered: is all sensory processing suppressed during 0926-6410/$ - see front matter D 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.cogbrainres.2003.10.018

P. Haggard, B. Whitford / Cognitive Brain Research 19 (2004) 52 58 53 movement, or only those sensory afferences predicted by the model? Where in the brain is the comparator circuit that performs the cancellation? What is the source of the efferent signal used for cancellation? This paper focuses on the latter question. The efferent signal used to modify sensory processing might be the descending volley from the motor cortex to the spinal cord, or it might be an earlier signal involved in the preparation of action. Several studies point to the latter conclusion. Chronicle and Glover [5] studied the ticklishness of voluntary actions of muscle twitches induced by transcranial magnetic stimulation over the primary motor cortex (MI) and of externally applied stimuli. They found that twitches were perceived to be as ticklish as external stimuli, and more ticklish than voluntary actions. This finding suggests that efference from MI is insufficient for sensory suppression, and that the critical efferent signal must arise prior to MI. However, their method implicitly assumes that TMS-induced twitches mimic the normal physiological efference from MI. Haggard and Magno [8] investigated the perceived time of manual actions, and reached a similar conclusion regarding the relation between efferent signals and conscious experience. Subjects reported the time at which they made a simple manual response to a tone. On some trials, the response time was artificially increased by applying TMS over MI just before the expected reaction [6]. Subjects were largely unaware of delays caused by intervention at this level in the system. In contrast, similar stimuli applied over the SMA produced shorter delays in actual response time, but the subjects were aware of a greater proportion of the delay that did occur. This result suggests that awareness of the manual response was formed partly upstream of the MI, possibly in the SMA. The present study attempted a more direct investigation of the role of SMA as the source of the efferent signal for sensory suppression. We studied the perceived magnitude of a somatosensory stimulus controlled by the experimenter, when the stimulus was embedded in a voluntary action and when it was not. We chose TMS-induced twitches evoked by MI stimulation as suitable somatosensory stimuli for subjects to judge, as their magnitude can be directly quantified, their timing precisely controlled, and they involve precisely the sensory receptors that are also affected by voluntary action. We predicted sensory suppression in voluntary action trials. If the SMA generates an efferent signal during voluntary action which is used for sensory suppression, we reasoned that transient disruption of the SMA just prior to the to-be-judged stimulus should disrupt the putative efferent signal, prevent cancellation of sensory reafference, and thus reduce the sensory suppression. 2. Methods Eight healthy subjects (aged 21.7 28.1 years, five female) participated with ethical approval. Subjects heard a sequence of three tones at 1-s intervals and were trained to produce a voluntary right index finger flexion at the onset of the third tone. In the experimental trials, subjects freely decided whether to make or withhold the voluntary action on that trial, aiming for roughly equal proportions of action and no-action trials. TMS over the optimal area of the left motor cortex for exciting the right index finger flexor (first dorsal interosseus muscle: 1DI) was delivered at the onset of the 3rd tone, causing an involuntary Motor Evoked Potential (MEP1). A Magstim 200 stimulator and figure of eight coil were used, at intensities between 100% and 120% of the relaxed motor threshold of the right 1DI. The field strength was varied randomly from trial to trial between these limits under computer control. The resulting MEP superimposed on any voluntary action the subject might make. On 50% of prepulse trials selected at random, a conditioning TMS pulse at 100% of the relaxed motor threshold of the right 1DI was applied over the supplementary motor area (SMA), using a circular coil and a second stimulator. To localise the SMA, we first identified the anterior boundary of the leg area, by using a cone coil and finding the most anterior site from which we could evoke MEPs from the preactivated right gastrocnemius at intensities of up to 65% of maximum output. A circular coil was then placed with its handleoriented anterior-posteriorly, and its most posterior tip just touching the anterior boundary of the leg area. The effective field maximum was thus 20 mm anterior from this boundary. In 50% of randomly selected experimental trials, the SMA prepulse was applied 10 ms before motor cortical TMS. This latency has been found to induce intracortical facilitation locally within the motor cortex in paired-pulse experiments [11]. The timing of TMS pulses was controlled by dedicated timing hardware which ran independently from the computer s operating system. Approximately 1 s after the motor cortical TMS, the subject received a verbal instruction to relax. This was followed a few seconds later by a further series of three tones and a second comparison twitch in the right 1DI (MEP2) caused by motor cortical TMS through the same figure of eight coil. The stimulation intensity for MEP2 was fixed at 110% of relaxed motor threshold for the 1DI, in the centre of the range used for MEP1. Thus, MEP2 served as a fixed reference against which a variable MEP1 could be compared. Subjects never made voluntary actions during MEP2, nor was there any conditioning pulse for MEP2. Subjects compared the perceived intensity of MEP1 and MEP2, and made unspeeded keypresses with their left hand to indicate which was larger. Up to 10 training trials were given at the start of the experimental session, to familiarise subjects with the sequence of trial events and the decision whether to make voluntary actions or not. Subjects then performed a total of 150 trials, giving approximately 38 in each cell of a factorial design made by crossing the two factors occurring at the time of MEP1. These factors were voluntary action (present, absent) and SMA prepulse (present, absent). Our interest

54 P. Haggard, B. Whitford / Cognitive Brain Research 19 (2004) 52 58 focused on how these two factors would modulate the perceived size of the test MEP1 in comparison to the reference MEP2. In two additional conditions, the conditioning pulse occurred after the MI pulse. However, all subjects had very poor judgements of the MEP amplitude in those conditions. They are not therefore reported here, but form the focus of ongoing investigation. The EMG of the right first dorsal interosseus muscle was recorded with bipolar recording using surface electrodes, amplified and digitised at 5 khz. The time of all TMS pulses, the EMG signals around the times of MEP1 and MEP2, and the subject s judgement for each trial were stored on a computer. An offline analysis algorithm then searched for the minimum and maximum EMG levels in the window 15 80 ms after the TMS pulse. The difference between these levels was defined as the MEP amplitude. In addition, each trial was checked visually by the experimenters, to classify each trial as either including a voluntary movement or not, depending on whether voluntary background EMG was seen in a time window extending from 60 ms before to 60 ms after the third synchronisation tone. The experimenter verified the form of each MEP at the same time. 3. Results Fig. 1 shows typical data for MEP1 from one subject, for a trial containing neither prepulse nor voluntary action (A) and for a further trial with both a prepulse and a voluntary action (B). The MEPs show the classical monophasic pattern, with a large initial direct wave, followed by further peaks and valleys corresponding to indirect waves. When subjects made voluntary actions, MEP1 was larger than when they did not [peak to valley 2.81 vs. 1.82 mv, F(1,7) = 21.4, p < 0.01]. This result was predictable from the known increase in excitability of motor cortex around the time of voluntary contraction [4]. There was also a significant facilitatory effect of the SMA prepulse on MEP1 amplitude [2.55 mv vs. 2.08 mv, F(1,7) = 21.43, p < 0.01], consistent with a facilitatory influence of the SMA on the primary motor cortex. The interaction between the factors of voluntary action and prepulse was close to significance [ F(1,7) = 4.647, p = 0.067], with the prepulse causing less facilitation in voluntary action trials (0.33 mv) than in nonaction trials (0.62 mv). Analysis of MEP2 amplitudes showed no effects of whether there had been a voluntary action or a prepulse at the time of the preceding MEP1 and no interaction (range of means 2.35 2.43 mv, all F < 1). Subject s judgements regarding whether MEP1 was greater or than MEP2 or not were correct on 57.2% of trials over the whole experiment. Although subjects clearly found the judgement task difficult, overall performance was nevertheless significantly better than chance. To further investigate effects of voluntary action and of SMA prepulses on somatosensory processing and conscious perception, we calculated the difference in twitch sizes (MEP1 MEP2) in each trial. We then grouped these trials into 10 frequency bins, and plotted histograms according to the judgement of which MEP was larger. The results (Fig. 2) showed a general tendency to judge MEP2 larger than MEP1 (red area>green area), perhaps due to a recency effect in memory leading to MEP2 seeming more intense. In addition, MEP1 was less likely to be judged larger than MEP2 when subjects made a voluntary action than when they did not. This result is consistent with a sensory suppression of MEP1 during voluntary actions. More importantly, the degree of sensory suppression was reduced by the SMA prepulse (note more green trials in bottom right panel than in top right panel of Fig. 2). To analyse this effect statistically, we obtained a psychophysical function by fitting logistic regression to each Fig. 1. (A) Typical test MEPs in the right first dorsal interosseus muscle produced by motor-cortical TMS in conditions without voluntary action or SMA prepulse. The dashed line indicates the time of motor-cortical TMS stimulation. Stimulator output is 120% of relaxed motor threshold. (B) Typical MEP with voluntary action and SMA prepulse. Note the larger MEP during voluntary action. The dashed line indicates the time of motor-cortical TMS stimulation. Stimulator output is 100% of relaxed motor threshold. The dot-dashed line indicates the time of the conditioning pulse over SMA, also at 100% of relaxed motor threshold. Note that the MEP latency is comparable in both traces, showing that the SMA prepulse did not itself cause MEPs.

P. Haggard, B. Whitford / Cognitive Brain Research 19 (2004) 52 58 55 Fig. 2. Psychophysical judgements of relative MEP size. Each histogram shows binned data pooled across subjects for the difference between the experimental and comparison MEPs (MEP1 MEP2). The green column shows the number of trials in which subjects judged MEP1 larger than MEP2, and the stacked red column shows the number of trials in which subjects judged MEP2 larger than MEP1. Note the larger values of MEP1 MEP2 on trials in which subjects made voluntary actions (right hand column), due to increased size of MEP1. Note also the greater probability of judging MEP1 larger than MEP2 during voluntary action trials when a prepulse over SMA is delivered, compared to when it is not. subject s binned judgement data in each condition of the 2 2 factorial design. The function estimated the probability of judging MEP1 larger than MEP2, as a function of the amplitude difference MEP1 MEP2. Fig. 3 shows mean curves obtained by averaging coefficients from individual fits. Numerically similar coefficients were obtained by fitting a single regression to all subjects pooled data, suggesting the estimations are robust. Psychophysical functions for the trials without voluntary actions (Fig. 3 solid lines) follow the expected pattern, with a clear positive gradient. The point of subjective equality, at which MEP1 and MEP2 are judged equally intense, occurs when MEP1 is slightly larger than MEP2, again suggesting a recency benefit in sensory memory for MEP2. The SMA prepulse has only a small effect on the curve in the absence of voluntary action, but produces a large shift in the intercept in trials where a voluntary action was made. The intercepts and slope coefficients of the individual logistic regressions were subjected to 2 2 factorial ANOVA with factors of voluntary action (present or absent) and SMA prepulse (present or absent). First, a planned t-test on the overall mean of the slopes confirmed that they were positive, implying that subjects were able to perform the perceptual judgement [t(7) = 3.868, p < 0.01]. Voluntary action significantly reduced the slope of the psychophysical function compared to trials without voluntary action, consistent with sensory suppression [ F(1,7) = 9.007, p < 0.05]. The slope reduction could reflect both specific neural processes such as reduction of afferent information for perception due to cancellation by efference copy, but also general factors, such as division of attention between the voluntary action and the twitch. The SMA prepulse had no effect on the psychophysical function slope, and there was no interaction between the prepulse and the voluntary action factor [ F(1,7) < 1 in both cases]. Analysis of the psychophysical curve intercepts showed no main effect of voluntary action [ F(1,7) < 1], but a significant effect of the prepulse [ F(1,7) = 7.117, p < 0.05]. Most interestingly, there was an interaction between these two factors, which can be clearly seen in Fig. 3. In trials without conditioning pulses, voluntary action introduced a bias in sensory judgement: during voluntary action, MEP1 was generally perceived as smaller than MEP2. This is again consistent with sensory suppression and is particularly striking given that voluntary action generally increased the actual size of MEP1 (Fig. 2). Crucially, this bias was reduced, and almost abolished by the prepulse over SMA producing the interaction. Since the direction of the effect of SMA stimulation was predicted, a one-tailed test of the interaction is appropriate [ F(1,7) = 3.617, p < 0.05]. Followup simple effects t-tests confirmed a significant effect of prepulse on the intercept in the voluntary action trials

56 P. Haggard, B. Whitford / Cognitive Brain Research 19 (2004) 52 58 Fig. 3. Logistic regressions fitted to psychophysical data. The curves are made by averaging coefficients from individual fits to each subject s data in each condition. Note that voluntary actions generally impair judgement of MEP size (reduced slope). Voluntary actions also introduce a bias (rightward shift) towards judging MEP2 to be larger than MEP1, due to sensory suppression of MEP1. However, this bias is reduced and almost abolished by a TMS prepulse over SMA. [t(7) = 3.00, p < 0.05] but not in the absence of voluntary actions [t(7) = 1.69, p>0.1]. 4. Discussion Our results showed that voluntary actions caused sensory suppression of the perceived magnitude of a somatic stimulus (a muscle-twitch caused by MI stimulation). Sensory suppression was seen as changes in both intercepts and slopes of the psychometric function. Voluntary action can produce both an underestimation bias and an attenuation, or reduction in the amount of available perceptual information. While sensory suppression has been reported before (see Section 1), few previous studies have distinguished between its bias and attenuation components. Tsakiris and Haggard [17] obtained Likert scale judgements of the perceived size of a TMS-induced muscle-twitch occurring 270 ms after either a voluntary action, or an equivalent passive movement. They found an intercept change but no slope change, suggesting a biasing but not an attenuating effect of sensory suppression. Their design deliberately kept the voluntary action separate from the somatic stimulus in space and time, reducing the need to divide attention between these events. In the present study, action and somatosensory stimulus were superimposed, and significant attenuation was observed, as measured by the slope of the psychophysical curves. We therefore suggest that attenuation effects in sensory suppression may partly reflect division of attention between the action and the to-be-judged stimulus. The remainder of the discussion focuses on the effect of an SMA prepulse on sensory suppression. We found that transient disruption of the SMA with a TMS prepulse reduced and almost abolished the sensory suppression bias in the perceived magnitude of an MEP during voluntary action. Fig. 3 shows that the SMA prepulse restored the point of subjective equality to a similar level to that in trials where the subject withheld voluntary action, effectively abolishing sensory suppression. Thus, processing in the frontal cortex preceding the motor command itself contributes to predictive suppression of the sensory events during voluntary action. The SMA stimulation did not, however, produce a general impairment of somatosensation or of judgement: its effects were confined to the voluntary action condition. In addition, SMA prepulses modulated the intercept but not the slope of the psychometric function. This suggests SMA stimulation influenced a perceptual system specifically

P. Haggard, B. Whitford / Cognitive Brain Research 19 (2004) 52 58 57 associated with ongoing actions, rather than operating indirectly via a general change in arousal or divided attention. This specificity also raises an important methodological point. In many psychological studies, TMS is used to produce virtual lesions, influencing behaviour. Lack of effects following stimulation at a control site is used to rule out non-specific effects such as arousal [18]. In conditioning pulse paradigms, the specificity of the conditioning effect offers an alternative for control for arousal. In our study, the effect of the conditioning pulse was specific to the trials where the subject chose to make a voluntary action. Computational models of motor control suggest that internal models may precisely predict and thus cancel reafference caused by voluntary action [3]. The cancellation process explains reduced sensation of those somatic events linked to the motor command. But cancellation also implies that other somatic events occurring during movement but not predicted by the motor command should not show sensory suppression. The present experiment can clarify how precise these efferent effects on sensation may be. Our results show that the perceived size of an involuntary movement is reduced during voluntary action of the same digit at the same time. Since the MEP was not a part of the subject s intention, this suggests that somatosensory suppression during action is partly a general suppression of afferent input from the relevant body part at the relevant time, including both the intended and unintended events. Suppression may be spatially and temporally specific, but may not be specific to the values of sensory parameters. Two brain regions, the cerebellum and the parietal cortex, are thought to be involved in predicting the consequences of action. The cerebellar cortex contains an adaptive circuit which may predict the sensory consequences of motor commands [19]. Since actual sensory experience presumably depends on somatosensory cortical processing rather than the cerebellum, an interaction between cerebellar prediction and somatosensory cortical activity may underlie the subjective effect of sensory suppression. However, many studies show, as we do here, that sensory events unrelated to the movement command are also suppressed during actions [2,7]. We believe this general suppression is not consistent with the specificity of cerebellar prediction in current computational motor control models. Although our results do not exclude some cerebellar contribution to sensory suppression, we suggest that other circuits, and specifically the SMA, are additionally involved. Could the present results reflect spinal, as opposed to cortical, effects? Efferent signals from the SMA during voluntary action could, in principle, depress transmission at spinal synapses conveying afferent information, thus causing sensory suppression. Our data cannot rule out this interpretation conclusively. However, two pieces of evidence suggest it is unlikely. First, the spinal interpretation would imply that patients with lesions to the SMA should have altered (specifically, heightened) somatic sensation, yet this has not been reported to our knowledge. Second, Abbruzzese et al. [1] reported that motor imagery tasks known to involve activation of the SMA produced changes in motor cortical excitability without modulating spinal excitability on H-reflex tests. The suppression effects observed here presumably arise by a different route. We suggest somatosensory processing in parietal cortex is modulated by an efference copy signal arising in the frontal cortex, specifically the SMA. The primary and secondary somatosensory areas of the parietal cortex are involved in conscious perception of somatosensory events [14]. Parietal and frontal brain regions are heavily and reciprocally interconnected by corticocortical fibres [12]. Many neurophysiological studies have indicated that the parietal areas compute spatial descriptions of objects, which are then used in frontal cortex to control grasping actions (e.g., see Ref. [10]). More recent studies have shown a role for the reciprocal, fronto-parietal connections in attention for action, ensuring that perceptual processing is directed to the requirements of impending voluntary actions. For example, signals from the frontal eye field may be responsible for attentional enhancement of visual responses in parietal neurons [13]. Somatosensory suppression during voluntary action is a second kind of attention for action, since it involves deselection of predicted sensory inputs. Our results suggest that the SMA plays a special role in controlling perceptual processing during voluntary action. An efference copy signal within the internal model for sensorimotor control may thus arise in the SMA. This signal is presumably used to regulate somatosensory processing in the parietal lobes at the level of conscious perception. Finally, we speculate on why sensory suppression occurs. Neurophysiological accounts typically assume that sensory suppression, like saccadic suppression in the oculomotor system, serves to reduce the cognitive load of the barrage of sensory information triggered by voluntary movements. However, sensory information is both available and important during skilled action, since loss of sensory information causes severe movement control deficits in deafferented individuals [15]. Sensory suppression therefore seems quite paradoxical. More recent computational accounts [3] suggest that sensory suppression serves to reduce the cognitive load of those sensory events that can be predicted from the motor command. However, this account cannot simply explain why external stimuli unrelated to the motor command are also suppressed. We suggest a possible resolution of this paradox: the CNS may suppress sensory information related to the movement, but enhance sensory information related to achieving the action goal. This approach would ensure that limited cognitive resources for sensory processing are used where they contribute most to our goals, and would reflect an optimal information-processing strategy. This approach implies a sophisticated intentional selection process in the CNS, which would adjust the strength of specific sensory pathways in advance of each action. We plan to test this account in future research.

58 P. Haggard, B. Whitford / Cognitive Brain Research 19 (2004) 52 58 Acknowledgements This research was supported by a Leverhulme Research Fellowship to PH. Additional support was provided by a Wellcome Trust Equipment Grant and an MRC Cooperative Group Grant. We are grateful to two anonymous reviewers for helpful comments and to Clare Press for assistance. References [1] G. Abbruzzese, C. Trompetto, M. Schieppati, The excitability of the human motor cortex increases during execution and mental imagination of sequential but not repetitive finger movements, Exp. Brain Res. 111 (1996) 465 472. [2] R.W. Angel, R.C. Malenka, Velocity-dependent suppression of cutaneous sensitivity during movement, Exp. Neurol. 77 (1982) 266 274. [3] S.J. Blakemore, D.M. Wolpert, C.D. Frith, Central cancellation of self-produced tickle sensation, Nat. Neurosci. 1 (1998) 635 640. [4] R. Chen, M. Hallett, The time course of changes in motor cortex excitability associated with voluntary movement, Can. J. Neurol. Sci., (1999) 163 169. [5] E.P. Chronicle, J. Glover, A ticklish question: does magnetic stimulation of the primary motor cortex give rise to an efference copy? Cortex, (2003) 105 110. [6] B.L. Day, J.C. Rothwell, P.D. Thompson, A.M. Denoordhout, K. Nakashima, K. Shannon, C.D. Marsden, Delay in the execution of voluntary movement by electrical or magnetic brain-stimulation in intact man-evidence for the storage of motor programs in the brain, Brain, (1989) 649 663. [7] N. Forss, T. Silen, Temporal organization of cerebral events: neuromagnetic studies of the sensorimotor system, Rev. Neurol. 157 (2001) 816 821. [8] P. Haggard, E. Magno, Localising awareness of action with transcranial magnetic stimulation, Exp. Brain Res. 127 (1999) 102 107. [9] P. Haggard, R.C. Miall, D. Wade, S. Fowler, A. Richardson, P. Anslow, J. Stein, Damage to cerebellocortical pathways after closed head injury: a behavioural and magnetic resonance imaging study, J. Neurol. Neurosurg. Psychiatry 58 (1995) 433 438. [10] M. Jeannerod, M.A. Arbib, G. Rizzolatti, H. Sakata, Grasping objects: the cortical mechanisms of visuomotor transformation, Trends Neurosci. 18 (1995) 314 320. [11] J. Liepert, J. Classen, L.G. Cohen, M. Hallett, Task-dependent changes of intracortical inhibition, Exp. Brain Res. 118 (1998) 421 426. [12] G. Luppino, M. Matelli, R. Camarda, G. Rizzolatti, Corticocortical connections of area F3 (SMA-proper) and area F6 (pre-sma) in the macaque monkey, J. Comp. Neurol. 338 (1993) 114 140. [13] T. Moore, K.M. Armstrong, Selective gating of visual signals by microstimulation of frontal cortex, Nature 421 (2003) 370 373. [14] W. Penfield, T. Rasmussen, The Cerebral Cortex of Man: A Clinical Study of Localisation of Function, Macmillan, New York, 1950. [15] J.C. Rothwell, M.M. Traub, B.L. Day, J.A. Obeso, P.K. Thomas, C.D. Marsden, Manual motor performance in a deafferented man, Brain 105 (1982) 515 542. [16] A. Sirigu, E. Daprati, P. Pradat-Diehl, N. Franck, M. Jeannerod, Perception of self-generated movement following left parietal lesion, Brain 122 (1999) 1867 1874. [17] M. Tsakiris, P. Haggard, Awareness of somatic events associated with a voluntary action, Exp. Brain Res. 149 (2003) 439 446. [18] V. Walsh, A. Pascual-Leone, Transcranial Magnetic Stimulation: A Neurochronometrics of Mind, in press. [19] D.M. Wolpert, Computational approaches to motor control, Trends Cogn. Sci. (1997) 209 216.