Please cite this article in press as: Dresler et al.

, Dreamed Movement Elicits Activation in the Sensorimotor Cortex, Current Biology (2011), doi:10.1016/j.cub.2011.09.029
Current Biology 21, 15, November 8, 2011 2011 Elsevier Ltd All rights reserved DOI 10.1016/j.cub.2011.09.029

Report Dreamed Movement Elicits Activation in the Sensorimotor Cortex

Martin Dresler,1,5 Stefan P. Koch,2,5 Renate Wehrle,1,5 Victor I. Spoormaker,1 Florian Holsboer,1 Axel Steiger,1 Philipp G. Samann,1 Hellmuth Obrig,2,3,4 and Michael Czisch1,* 1Max Planck Institute of Psychiatry, Kraepelinstrasse 2-10, 80804 Munich, Germany 2Berlin Neuroimaging Center, Charite University Hospital, Humboldt University of Berlin, 10099 Berlin, Germany 3Max Planck Institute for Human Cognitive and Brain Sciences, Stephanstrasse 1, 04103 Leipzig, Germany 4Clinic for Cognitive Neurology, University Hospital Leipzig, 04103 Leipzig, Germany case, dreamed hand clenching. Predecided eye movements served as temporal markers for the onset of hand clenching and for hand switching. Previous studies have shown that muscle atonia prevents the overt execution of dreamed hand movements, which are visible as minor muscle twitches at most [3, 9]. Subjects were instructed to make series of left and right hand movements separated by sets of left-right-left-right (LRLR) eye movements upon becoming lucid (see Experimental Procedures below; see also Supplemental Information available online). Out of six highly experienced lucid dreamers who participated in the study, two subjects succeeded in performing the task during lucid REM sleep under combined electroencephalography (EEG)-fMRI or combined EEG-NIRS conditions, respectively. As a control condition, the subjects additionally performed both an actually executed and an imagined hand-clenching task during wakefulness. Increases in blood oxygen level-dependent (BOLD) signal in the sensorimotor cortex contralateral to the indicated movement side were found in the successful fMRI experiments during the two lucid REM states in one subject (Figure 2). Given the short duration of the lucid dreams, the activation of the sensorimotor cortical regions was very specic compared to wakefulness: executed hand clenching during wakefulness (WE) elicited widespread activation clusters in the contralateral pre- and postcentral gyrus (pFWE < 0.005). Collection threshold for both imagined hand movements during wakefulness (WI) and dreamed hand movements (LD) were set at puncorr < 0.005, with voxel extent 50. Cluster signicance was assessed in a volume of interest dened by the activation during executed hand movement in wakefulness. The strength of the activation was evaluated by analyzing the difference between the maximum and minimum BOLD signal amplitude in the peak voxels time course of the respective motor regions. Strongest BOLD signal uctuations of the peak voxels time course can be seen during WE (left hand: 5.62%; right hand: 4.49%), whereas WI and LD showed mean uctuations of only 1.3%1.4% and 1.8%2.5%, respectively (Figure 2). FMRI results were conrmed by an independent imaging method in a second subject: NIRS data showed a typical hemodynamic response pattern of increased contralateral oxygenation over the sensorimotor region during successful task performance in lucid REM sleep (Figure 3; Figure 4). Notably, during dreaming, the hemodynamic responses were smaller in the sensorimotor cortex but of similar amplitude in the supplementary motor area (SMA) when compared to overt motor performance during wakefulness. Discussion The discovery of the close relationship between REM sleep and dream reports [10, 11] boosted decades of neuroscientic research into the brain state potentially evoking vivid and intense dreams. Neurophysiological studies suggest that during REM sleep, the brain functions as a closed loop system, in which activation is triggered in pontine regions while sensory input is gated by enhanced thalamic inhibition and motor output is suppressed by atonia generated at the brain stem

Summary Since the discovery of the close association between rapid eye movement (REM) sleep and dreaming, much effort has been devoted to link physiological signatures of REM sleep to the contents of associated dreams [14]. Due to the impossibility of experimentally controlling spontaneous dream activity, however, a direct demonstration of dream contents by neuroimaging methods is lacking. By combining brain imaging with polysomnography and exploiting the state of lucid dreaming, we show here that a predened motor task performed during dreaming elicits neuronal activation in the sensorimotor cortex. In lucid dreams, the subject is aware of the dreaming state and capable of performing predened actions while all standard polysomnographic criteria of REM sleep are fullled [5, 6]. Using eye signals as temporal markers, neural activity measured by functional magnetic resonance imaging (fMRI) and near-infrared spectroscopy (NIRS) was related to dreamed hand movements during lucid REM sleep. Though preliminary, we provide rst evidence that specic contents of REM-associated dreaming can be visualized by neuroimaging. Results Lucid dreaming is a rare but robust state of sleep that can be trained [5]. Phenomenologically, it comprises features of both waking and dreaming [7]: in lucid dreams, the sleeping subject becomes aware of his or her dreaming state, has full access to memory, and is able to volitionally control dreamed actions [6]. Although all standard polysomnographic criteria of rapid eye movement (REM) sleep [8] are maintained and REM sleep muscle atonia prevents overt motor behavior, lucid dreamers are able to communicate their state by predened volitional eye movements [6], clearly discernable in the electrooculogram (EOG) (Figure 1). Combining the techniques of lucid dreaming, polysomnography, and brain imaging via functional magnetic resonance imaging (fMRI) or near-infrared spectroscopy (NIRS), we demonstrate the possibility to investigate the neural underpinnings of specic dream contentsin this

5These

authors contributed equally to this work *Correspondence: czisch@mpipsykl.mpg.de

Please cite this article in press as: Dresler et al., Dreamed Movement Elicits Activation in the Sensorimotor Cortex, Current Biology (2011), doi:10.1016/j.cub.2011.09.029
Current Biology Vol 21 No 21 2

Figure 1. Exemplary Lucid REM Sleep as Captured by Polysomnography during Simultaneous fMRI Note high-frequency electroencephalogram (EEG) and minimal electromyogram (EMG) amplitude due to muscle atonia characteristic of rapid eye movement (REM) sleep (left), with wakefulness for comparison (right). Subjects were instructed to communicate the state of lucidity by quick left-right-left-right (LRLR) eye movements. Filter settings are as follows: EEG, bandpass lter 0.5270 Hz, with additional notch lter at 50 Hz; electrooculogram (EOG), bandpass lter 0.130 Hz; EMG, bandpass lter 16250 Hz.

level [4, 12]. Cortical synthesis of internally generated sensorimotor perceptions may explain partially coherent narrative dream mentation [13]. REM sleep is associated with increased cerebral blood ow in visual association areas and limbic regions as well as attenuated metabolism in dorsolateral prefrontal cortex, primary visual cortex, and precuneus [14, 15]. Such a pattern of activation has been proposed to underlie general dream features like vivid imagery, emotional experiences, and lack of insight and volition [4, 16]. The neural correlates of specic dream mentation, however, still had to be disclosed. Efforts have been made to correlate REMs to gaze direction during dreamsthe scanning hypothesis [1, 2]and indeed similar cortical areas are involved in eye movement generation in wake and REM sleep [17]. In a similar vein, small muscle twitches during REM sleep were presumed to signal a change in the dream content [3]. In REM behavior disorder, patients even seem to enact dream fragments [18], suggesting that dreamed motor actions involve activation of similar brain regions as during waking actions. Also in non-REM sleep, dreaming of task-related content after an intense skill-learning session hints to the possibility that the same neural systems control executed and merely dreamed motor activity [19]. Nonetheless, the main obstacle in the direct neuroscientic study of specic dream content is that spontaneous dream activity cannot be experimentally controlled because subjects typically cannot perform predecided mental actions during sleep. Hence, dream research methodology mostly relies on the evaluation of subjective reports of very diverse dream contents. However, such memory traces collected after awakening are often imprecise. Therefore it seems difcult, if not impossible, to pinpoint specic dream content in a precise temporal frame, a prerequisite for the analysis of imaging data acquired during dreaming. In the present study, we employed the skill of lucid dreaming to overcome some of these obstacles. In line with the conception that sleep atonia is generated at the brainstem level [4] and conrming earlier EEG studies [20], fMRI BOLD responses were observed bilaterally in the same sensorimotor cortical regions for lucid dreaming and wakefulness. However, during dreaming, activation was much more localized in small clusters representing either generally weaker activation or focal activation of hand areas only, with signal uctuations only in the order of 50% as compared to the actually executed task during wakefulness. Although in the fMRI analysis activation

in the SMA was not observed, the NIRS-measured hemodynamic responses during dreaming were smaller in the sensorimotor cortex but of similar amplitude in the SMA when compared to overt motor performance during wakefulness. The SMA is involved in timing, preparation, and monitoring of movements [21], and linked to the retrieval of a learned motor sequence especially in the absence of external cues [22]. The SMA has been attributed as a programming area active only during complex movements [23]; however, our NIRS data speak for an activation of SMA even during simple movements in this subject. This is in line with several PET and fMRI studies reporting SMA activations for simple tasks such as hand clenching, single nger-tapping, and alternated nger-tapping (see Supplemental References and Supplemental Discussion of the NIRS and fMRI mismatch). In general, our data support the assumption that the pattern of activation in motor imagery largely overlaps with activity corresponding to motor execution [24]. Moreover, the nding of a similar action representation for hand clenching during lucid dreaming extends the view of a common neural substrate for actions also for covert movements during this particular dreaming state. In this regard, dreamed motor activity might help to elucidate the controversy about the locus of control for the inhibition of motor output in motor imagery: because it is known that sleep atonia is generated at the brainstem level and spinal cord [4], the similarity between imagined and dreamed motor activity might be interpreted as conrmative for explanations in terms of a blocking mechanism downstream of the motor cortex in motor imagery rather than a prefrontal blockade [24]. Controversial reports exist on the involvement of M1 activation in motor imagery [23, 25, 26]: although it is typically smaller compared with that in execution [26], transient responses might be interpreted to reect the onset of a specic motor imagination not leading to sustained M1 activation [23]. The close interaction between somatosensory and motor cortices during the performance of a movement differs between waking and dreaming. Lack of sensory feedback due to REM sleep atonia may further reduce the activation in M1 and primary somatosensory cortex. However, small muscle twitches might well occur during dreamed hand clenching [3, 9], potentially leading to residual M1 activation. Because several limitations of our study complicate the interpretation of the activation pattern, the data have to be considered as very preliminary. Due to the scarcity of the phenomenon and the complexity of the methods, we basically present two case studies of two dreams each, assessed with different methods, which naturally leaves the possibility of individual differences strongly inuencing the results. Based on this, no statistical comparisons were made between conditions (WE, WI, LD) or regions. The lack of hand electromyogram (EMG) data further complicates the interpretation, and the mismatch between the two subjects concerning SMA activation can hardly be explained satisfactorily. Hence, our data could not and should not be interpreted as a thorough clarication of the neural correlates of motor activity during dreams but rather as a methodological proof of concept on how to measure neural correlates of dream content with neuroimaging methods. Although lucid dreaming comprises all dening markers of REM sleep proper [8] and all basal dream features such as hallucinations, it differs from nonlucid dreaming in its metacognitive insight into the hallucinatory nature of the dream state and full access to cognitive capabilities [7, 27]. Recent sophisticated EEG analyses have shown increased gamma band

Please cite this article in press as: Dresler et al., Dreamed Movement Elicits Activation in the Sensorimotor Cortex, Current Biology (2011), doi:10.1016/j.cub.2011.09.029
Dreamed Movement Activates the Sensorimotor Cortex 3

Figure 2. Comparison of Sensorimotor Activation during Wakefulness and Sleep Functional magnetic resonance imaging (fMRI) blood oxygen level-dependent (BOLD)-response increases were contrasted between left and right hand movements (columns) in the three conditions (rows): executed hand movement during wakefulness (WE) (A), imagined hand movement during wakefulness (WI) (B), and dreamed hand movement during lucid REM sleep (LD) (C). Effects of left (right) hand movements were calculated in a xed-effects analysis as a contrast left > right and right > left, respectively. Subpanels depict results in an SPM glass-brain view (sagital and coronal orientation) to demonstrate the regional specicity of the associated cortical activation, along with sensorimotor activation overlaid on an axial slice of the subjects T1-weighted anatomical scan (position indicated on the glass brain for condition A). Clusters of activation in the glass-brain views are marked using the numbering given in Table S1. Red outlines in the glass-brain views mark the extent of activation found in the WE condition. This region of interest (ROI) was derived from the

Please cite this article in press as: Dresler et al., Dreamed Movement Elicits Activation in the Sensorimotor Cortex, Current Biology (2011), doi:10.1016/j.cub.2011.09.029
Current Biology Vol 21 No 21 4

Figure 3. Near-Infrared Spectroscopy Topography Concentration changes of oxygenated (D[HbO], upper panel) and deoxygenated hemoglobin (D[HbR], lower panel) during executed (WE) and imagined (WI) hand clenching in the awake state and dreamed hand clenching (LD). The optical probe array covered an area of w7.5 3 12.5 cm2 over the right sensorimotor area. The solid box indicates the ROI over the right sensorimotor cortex with near-infrared spectroscopy (NIRS)-channels surrounding the C4-EEG electrode position. NIRS channels located centrally over midline and more anterior compared to sensorimotor ROI were chosen as ROI for the supplementary motor area (SMA, dotted box).

activity over frontal and frontolateral areas when comparing lucid against nonlucid REM sleep, suggesting a neural correlate of this subjectively experienced cognitive control [28]. Hence, it cannot be excluded that neural activity associated with nonlucid dreams differs from that of dream content deliberately induced by a lucid dreamer. To assess regular dream mentation in more detail, lucid dreamers might reenact certain simple dream reports and the accordant activation patterns of lucid and nonlucid dreams could be compared. Although the neuroimaging of lucid dreaming is currently hampered by the scarcity of the phenomenon even in trained subjects, promising attempts exist to enhance lucid dreaming frequency, e.g., by transcranial direct current stimulation [27]. In summary, we provide the rst demonstration of imaging of specic dream contents by using the technique of lucid dreaming. This technique could be used to inversely infer specic dream content from its underlying neural activity, allowing for true dream reading. Similar to rst brain reading experiments in the wake state [29], we chose a simple hand-clenching task because motor activity can be reliably located by both imaging modalities applied. The combination of lucid dreaming with neuroimaging and polysomnography is a promising technique to also transfer more sophisticated brain reading tasks [30] to the dreaming state. Future applications might therefore lead to the imaging of prototypical dream content like visual dream imagery or dreamed emotions.
Experimental Procedures Six healthy male subjects (aged 2138 years, mean age 29.6 years) who had trained and practiced lucid dreaming for several years participated in the study. Whole-brain BOLD-fMRI (1.5 Tesla, 25 slices, 4 mm thickness, in-

plane resolution 3.125 3 3.125 mm2, TR 2 s, TE 40 ms) and NIRS were performed in the early morning hours during which REM sleep incidence is highest. The NIRS device was a custom-built monitor with 22 measuring channels unilaterally covering right central sensorimotor areas. Five of the six subjects underwent NIRS measurements with concurrent polysomnographic recordings during sleep for 13 nights, and four of the six subjects were assessed with fMRI with concurrent polysomnographic recordings for 26 nights. Subjects were instructed to signal the onset of a lucid dream by left-right-left-right (LRLR) eye movements and to immediately start clenching their left hand in their dream. They were further instructed to repeat the LRLR signal after ten clenches and start to clench the right hand. The alternating hand-clenching task was repeated for as long as possible, with the LRLR signal indicating each changeover. Whereas a at electromyogram (EMG) indicated REM sleep atonia, eye signals were clearly visible and distinguishable from spontaneous REMs in the EOG, providing unequivocal temporal markers for imaging analysis (Figure 1). Polysomnographic data were monitored online by a trained experimenter throughout the recording session. Immediately after participants awakened, a short dream report was prompted to verify lucidity and task compliance. Task performance was considered successful when the following three conditions were concurrently fullled: (1) the subject was in REM sleep, (2) EOG signals showed at least four sets of LRLR eye signals, and (3) both lucidity and dreamed hand clenching were conrmed by the subject in the subsequent verbal report. When the EEG showed clear signs of awakening after a LRLR signal, one of the investigators entered the scanner or NIRS room and inquired about task performance. Alternatively, when LRLR signals were followed by a period of about 40 s without any further signals, one of the investigators entered the scanner or NIRS room and inquired about task performance, thereby denitely awakening a potentially sleeping participant. During fMRI, two subjects signaled and reported lucid dreams of sufcient length, i.e., four sets of LRLR eye signals and hand clenchings. Polysomnographic evaluation, however, revealed that one of the two subjects showed increased EMG amplitudes, which could indicate a transition to wakefulness. This subject was excluded from the analysis. The remaining subject showed two lucid dreams during which the task was correctly performed, veried by polysomnographic data matching the dream report. Two further subjects reported lucidity during fMRI; however, one was not able to perform the signaling or task as a result of sudden dream termination, whereas the EEG of the other was compromised by strong sweating artefacts, thereby making any objective evaluation of sleep stage, lucidity, or task performance impossible. During NIRS, two subjects signaled and reported lucidity, but only one subject was able to successfully conduct the hand-clenching task (in two different dreams). We compared the activation pattern of lucid hand movements to executed and imagined hand movements, collected during wakefulness with both neuroimaging modalities (see Supplemental Information for details). Supplemental Information Supplemental Information includes two tables, Supplemental Results, Supplemental Experimental Procedures, and Supplemental Discussion and can be found with this article online at doi:10.1016/j.cub.2011.09.029. Acknowledgments We thank J. Allan Hobson for his advice and discussions on this topic. Received: February 9, 2011 Revised: August 31, 2011 Accepted: September 15, 2011 Published online: October 27, 2011 References 1. Dement, W., and Wolpert, E.A. (1958). The relation of eye movements, body motility, and external stimuli to dream content. J. Exp. Psychol. 55, 543553.

respective activation map during executed hand movement (A), thresholded at whole-brain corrected pFWE < 0.005, cluster extent >50 voxels, and served as a ROI for analysis of the WI and LD conditions in (B) and (C), respectively. T values are color-coded as indicated. The time course of the peak voxel inside the ROI is depicted (black) along with the predicted hemodynamic response based on the external pacing (A and B) or the predened LRLR-eye signals during (C). The maximal difference in activation of the peak voxel between conditions is indicated as percentage of BOLD signal uctuations of the predicted time course (gray).

Please cite this article in press as: Dresler et al., Dreamed Movement Elicits Activation in the Sensorimotor Cortex, Current Biology (2011), doi:10.1016/j.cub.2011.09.029
Dreamed Movement Activates the Sensorimotor Cortex 5

Figure 4. Condition-Related NIRS Time Courses Time courses of HbO (red traces) and HbR (blue traces) from the right sensorimotor ROI (left panel) and the supplementary motor ROI SMA (right panel) for executed (WE) and imagined (WI) hand clenching in the awake state and dreamed hand clenching (LD). The time courses represent averaged time courses from NIRS channels within the respective ROI (Figure 3). For each condition, 0 s denotes the onset of hand clenching indicated by LRLR-signals. Note that the temporal dynamics, i.e., an increase in HbO and a decrease in HbR, are in line with the typical hemodynamic response. Overt movement during wakefulness (dark red/blue traces) showed the strongest hemodynamic response, whereas the motor-task during dreaming leads to smaller changes (light red/blue traces). In the SMA, the hemodynamic response was stronger during the dreamed task when compared to imagery movement during wakefulness.

2. Roffwarg, H.P., Dement, W.C., Muzio, J.N., and Fisher, C. (1962). Dream imagery: relationship to rapid eye movements of sleep. Arch. Gen. Psychiatry 7, 235258. 3. Gardner, R., Jr., Grossman, W.I., Roffwarg, H.P., and Weiner, H. (1975). The relationship of small limb movements during REM sleep to dreamed limb action. Psychosom. Med. 37, 147159. 4. Hobson, J.A., and Pace-Schott, E.F. (2002). The cognitive neuroscience of sleep: neuronal systems, consciousness and learning. Nat. Rev. Neurosci. 3, 679693. 5. LaBerge, S.P. (1980). Lucid dreaming as a learnable skill: A case study. Percept. Mot. Skills 51, 10391042. 6. LaBerge, S., Nagel, L., Dement, W.C., and Zarcone, V. (1981). Lucid dreaming veried by volitional communication during REM sleep. Percept. Mot. Skills 52, 727732. 7. Hobson, J.A. (2009). REM sleep and dreaming: towards a theory of protoconsciousness. Nat. Rev. Neurosci. 10, 803813. 8. Rechtschaffen, A., and Kales, A. (1968). A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects (Washington, DC: Public Health Service, US Government Printing Ofce). 9. Erlacher, D., and Schredl, M. (2008). Do REM (lucid) dreamed and executed actions share the same neural substrate? Int. J. Dream. Res. 1, 714. 10. Aserinsky, E., and Kleitman, N. (1953). Regularly occurring periods of eye motility, and concomitant phenomena, during sleep. Science 118, 273274. 11. Dement, W., and Kleitman, N. (1957). The relation of eye movements during sleep to dream activity: an objective method for the study of dreaming. J. Exp. Psychol. 53, 339346. 12. Wehrle, R., Kaufmann, C., Wetter, T.C., Holsboer, F., Auer, D.P., Pollmacher, T., and Czisch, M. (2007). Functional microstates within human REM sleep: rst evidence from fMRI of a thalamocortical network specic for phasic REM periods. Eur. J. Neurosci. 25, 863871. 13. Hobson, J.A., and McCarley, R.W. (1977). The brain as a dream state generator: an activation-synthesis hypothesis of the dream process. Am. J. Psychiatry 134, 13351348. 14. Maquet, P., Peters, J., Aerts, J., Delore, G., Degueldre, C., Luxen, A., and Franck, G. (1996). Functional neuroanatomy of human rapid-eyemovement sleep and dreaming. Nature 383, 163166. 15. Braun, A.R., Balkin, T.J., Wesensten, N.J., Gwadry, F., Carson, R.E., Varga, M., Baldwin, P., Belenky, G., and Herscovitch, P. (1998). Dissociated pattern of activity in visual cortices and their projections during human rapid eye movement sleep. Science 279, 9195. 16. Schwartz, S., and Maquet, P. (2002). Sleep imaging and the neuropsychological assessment of dreams. Trends Cogn. Sci. 6, 2330. 17. Hong, C.C., Harris, J.C., Pearlson, G.D., Kim, J.S., Calhoun, V.D., Fallon, J.H., Golay, X., Gillen, J.S., Simmonds, D.J., van Zijl, P.C., et al. (2009). fMRI evidence for multisensory recruitment associated with rapid eye movements during sleep. Hum. Brain Mapp. 30, 17051722. 18. Boeve, B.F. (2010). REM sleep behaviour disorder. Ann. N Y Acad. Sci. 1184, 1554. 19. Wamsley, E.J., Tucker, M., Payne, J.D., Benavides, J.A., and Stickgold, R. (2010). Dreaming of a learning task is associated with enhanced sleep-dependent memory consolidation. Curr. Biol. 20, 850855. 20. Erlacher, D., Schredl, M., and LaBerge, S. (2003). Motor area activation during dreamed hand clenching: A pilot study on EEG alpha band. Sleep Hypn. 5, 182187.

21. Shima, K., and Tanji, J. (1998). Both supplementary and presupplementary motor areas are crucial for the temporal organization of multiple movements. J. Neurophysiol. 80, 32473260. 22. Tanji, J., and Mushiake, H. (1996). Comparison of neuronal activity in the supplementary motor area and primary motor cortex. Brain Res. Cogn. Brain Res. 3, 143150. 23. Roland, P.E., Larsen, B., Lassen, N.A., and Skinhj, E. (1980). Supplementary motor area and other cortical areas in organization of voluntary movements in man. J. Neurophysiol. 43, 118136. 24. Jeannerod, M. (2006). Motor Cognition: What Actions Tell the Self (Oxford: Oxford University Press). 25. Dechent, P., Merboldt, K.-D., and Frahm, J. (2004). Is the human primary motor cortex involved in motor imagery? Cogn. Brain Res. 19, 138144. 26. Munzert, J., Lorey, B., and Zentgraf, K. (2009). Cognitive motor processes: the role of motor imagery in the study of motor representations. Brain Res. Rev. 60, 306326. 27. Noreika, V., Windt, J.M., and Lenggenhager, B. (2010). New perspectives fort he study of lucid dreaming: From brain stimulation to philosophical theories of self-consciousness. Int. J. Dream Res. 3, 3645. 28. Voss, U., Holzmann, R., Tuin, I., and Hobson, J.A. (2009). Lucid dreaming: a state of consciousness with features of both waking and non-lucid dreaming. Sleep 32, 11911200. 29. Dehaene, S., Le ClecH, G., Cohen, L., Poline, J.B., van de Moortele, P.F., and Le Bihan, D. (1998). Inferring behavior from functional brain images. Nat. Neurosci. 1, 549550. 30. Haynes, J.D., and Rees, G. (2006). Decoding mental states from brain activity in humans. Nat. Rev. Neurosci. 7, 523534.