Figure 1. (a) The dance steps consisted of six-step movements in which the left and right legs always alternated. The pattern in the left panel was used for the Metric and Non-Metric dance tasks; the other pattern was used for the self-paced dance task (Motor). The dotted lines in the left panel show the path of each limb as it passes near a position stepped to by the other limb. (b) A subject illustrating the dance task.
Thesis, Antithesis, Synthesis
Dancers and Neurologists get together
We send thanks and gratitude to the dancers who volunteered to do this. For no benefit for themselves, volunteers danced inside the $4 million machine, as it detected the movements of information inside their brains. Showing that the brain, when dancing, is using circuits that are not the same as the ones that are damaged in Parkinson's patients. But this study was not about Parkinson's; it was not funded by the $200 million spent every year on Parkinson's research. This is dancers and neurologists trying to speak the same language, because they know that something is going on in the brain when people dance to the music. Enjoy the following, it is a different form of poetry, but it is definitely poetry, the poetry of science. The dreamer and the man of action: we need you all and we need you now:
The Neural Basis of Human Dance
Human dance was investigated with positron emission tomography to identify its systems-level organization. Three core aspects of dance were examined: entrainment, meter and patterned movement. Dancers performed small-scale, cyclically repeated tango steps on an inclined surface to the beat of tango music, without visual guidance. Entrainment of dance steps to music, compared to self-pacing of movement, was supported by anterior cerebellar vermis. Movement to a regular, metric rhythm, compared to movement to an irregular rhythm, implicated the right putamen in the voluntary control of metric motion. Spatial navigation of leg movement during dance, when controlling for muscle contraction, activated the medial superior parietal lobule, reflecting proprioceptive and somatosensory contributions to spatial cognition in dance.
Finally, additional cortical, subcortical and cerebellar regions were
active at the systems level. Consistent with recent work on simpler,
rhythmic, motor-sensory behaviors,
these data reveal the interacting network
of brain areas active during spatially patterned,
bipedal, rhythmic movements that are integrated in dance.
Many natural, complex sensorimotor activities involve the integration of rhythm, spatial pattern, synchronization to
external stimuli and coordination of the whole body. Such
activities include old evolutionary adaptations such as hunting,
fighting and play, as well as more recent ones such as group
physical labor, marching, musical performance and sport.
Neuroimaging studies have examined some components of these complex actions, such as the entrainment of movement to external timekeepers or spatial patterning of limb movement. However, this research has typically studied elementary processes such as ankle rotation or finger tapping (e.g. Penhune et al., 1998; Lutz et al., 2000; Debaere et al., 2001; Ehrsson et al., 2003; Sahyoun et al., 2004). A central issue is whether the
neural systems implicated in these elementary processes
‘scale up’ and ‘scale out’ to complex ecological activities. Are the
mechanisms controlling complex sensorimotor processes the same ones as those that underlie elementary processes like ankle rotation and finger tapping or are new mechanisms recruited? For example, dance is a complex sensorimotor action: do known elementary processes underlying simple movements ‘scale up’ to rhythmically timed, spatially patterned whole-body movements seen in human dance? The aim of the present study was to explore these and related issues in the context of examining for the first time the neural basis of dance.
Dance is a universal human behavior, one associated with
group rituals (Sachs, 1937; Farnell, 1999). Although it is
depicted in cave art from more than 20 000 years ago
(Appenzeller, 1998), dance may be much more ancient than
that. Dance may in fact be as old as the human capacities for
bipedal walking and running, which date back 2--5 million years
(Ward, 2002; Bramble and Lieberman, 2004). One of the principal properties of dance is that body movements are organized
into spatial patterns. This patterning of movement encompasses
a trajectory map of the body in exocentric space (Longstaff,
2000) as well as a kinesthetic and visual map of the body schema
in egocentric space (Haggard and Wolpert, 2005). The displacement patterns of dance can involve any body part; every dance
can be characterized by the identity and number of its
participating movement-units. In addition, dances tend to be
modular in organization, being composed of discrete sections
that are concatenated or interleaved with one another cyclically. Because of this combinatoric organization, dances are
amenable to grammatical analysis and description (Hutchinson-
A second property of dance is the synchronization of movements with timekeepers such as musical beats, a capacity that is
apparently specific to humans. Indeed, it is striking how our
bodies can spontaneously move in response to a musical beat.
Virtually all dancing is done to musical rhythms, thereby
permitting a temporal synchronization among dancers. Dance
gestures generally mirror the hierarchical arrangement of
strong and weak beats found in musical rhythm patterns. In
waltz music, for example, the first beat is stressed while the
second and third beats are weaker; likewise in waltz movements, the first step is the broadest and most forceful, while the
second and third steps are shorter and weaker. Thus, the
entrainment of dance to music not only involves synchronization in time but a spatial element related to equating hierarchies
in the motor pattern with those in the musical rhythm.
We conducted a positron emission tomography (PET) study
with amateur dancers performing small-scale, bipedal dance
steps on an inclined surface, as compared to auditory, motor and
rhythmic control tasks. In addition to working towards
a systems-level view of the neural basis of the complex sensorimotor processes underlying dance, we attempted to isolate
the foregoing individual processes, using planned comparisons
in a subtractive design. First, we sought to localize brain areas
involved in the synchronization of leg movement to the rhythm
of an auditory stimulus. For this, we made a planned comparison
of patterned leg movement performed to a musical beat (Metric
condition) with a matched motor pattern performed in a self-
paced though metric manner without an external timekeeper
(Motor condition). A second goal was to identify the brain areas
involved in the voluntary control of metric movements, that is,
dance steps occurring in an equal-time-interval rhythm. For this,
we made a planned comparison of patterned leg movement
paced to music possessing a regular, metric rhythm (Metric
condition) with that paced to music possessing a highly irregular, unpredictable rhythm (Non-Metric condition). Finally, we
sought to isolate the neural basis of spatial patterning of lower-
limb movement by making a planned comparison of the
conditions in which the legs moved in space (Metric condition)
with a condition in which the leg muscles contracted isometrically but without leg movement, also to a metric rhythm
(Contractions condition). By controlling for parameters related
to muscle contraction, we expected this contrast to reveal
brain areas supporting spatial cognition in dance, especially as
related to the lower extremities. Two control conditions
involved passive listening to music (Listening condition) and
The neurologists talk funny, don't you think ? But finally they are getting together. Dancers and Parkinson's patients have the anecdotes, but we need the science.
Thank you to the researchers. We thanked the dancers first, because it is surely no fun to be dancing attached to that machine. But we also send our thanks and gratitude to those who wrote this scientific report, which is like poetry to us, beautiful and elegant and true.
Thank you: Lawrence M. Parsons, Ph.D. Professor of Cognitive Neuroscience, Department of Psychology, The University of Sheffield, Sheffield, United Kingdom.
And Steven Brown and Michael J. Martinez of the Research Imaging Centre at San Antonio, Texas (Steven Brown is now at the Department of Psychology, Simon Fraser University, Burnaby B.C.)We thank Julie Grezes, Petr Janata, Katz Sakai and two anonymous reviewers for very helpful comments on this manuscript.
Let's get it on:
Materials and Methods
Five male and five female amateur dancers, with a mean age of 33.8 years
(range 19--46 years), participated in the study after giving their informed
consent (in accord with the Declaration of Helsinki and the Institutional
Review Board of the University of Texas Health Science Center). Each
individual was without neurological or psychiatric illness. All participants were right-handed, as confirmed by the Edinburgh Handedness
Inventory. Nine of the 10 subjects indicated that they would use their
right leg to kick a ball (Elias and Bryden, 1998). The subjects were
currently active amateur tango dancers, with a mean of 8.5 years of
recent recreational dance experience, of which 2.5 years were of
Argentine tango (range 1--4 years) and the remainder was of a wide
variety of other dance forms (e.g. Latin, ballroom, ballet). In spite of
these individual differences in age and years of overall dance experiences, the dancers were of comparable proficiency in Argentine tango
and in the relatively simple tango tasks used in the study. The subjects
had minimal musical experience.
Commercial recordings of instrumental Argentine tango songs, with
a typical tempo of 60 beats per minute, were presented for the tasks
containing music. All songs were matched for instrumentation, tempo
and tonality, and were presented to subjects using CoolEdit (Syntrillium) from a laptop computer. For the Non-Metric condition, songs
were edited to produce an irregular and unpredictable beat but without
altering the average tempo (i.e. each song had an equal number of
accelerations and decelerations of the tempo at random junctures). The
stimuli for the listening condition were Greek ‘rembetika’ songs that
were matched for the tempo, instrumentation and tonality of the tango
songs. By using music other than tango songs, we sought to minimize
the tendency for stimuli to elicit motor imagery of the dance steps
performed in the movement tasks (see Ehrsson et al., 2003). The
activations observed for this condition, as contrasted to rest, are
described elsewhere (Brown et al., 2004).
To make it clear: these dancers and scientists were being dancers and scientists for the sake of climbing the mountain because it is there. They were not fighting Parkinson's, or any other disease. It was for the pure thing itself, the art of dance and the science of dance. We are borrowing this information for our own purposes, to fight Parkinson's. Other people can borrow it for other reasons. That's the beauty of pure art and of pure science. The results of the research are not political or pre-ordained. No one in that group was thinking about Parkinson's disease at all. They went to find the truth, the facts of the case, to know about what is going on when we dance. Not for something or against something. Just the thing itself, in all its beauty and its complexity and its simplicity.
I challenge you to read this science paper. It is good for you. I was absent the day they taught science in school, so I don't understand it all and can hardly fathom how people can think like that - and that is why it is important to read about things that happen when scientists and artists try to figure out the same thing. John Lee Hooker was right.
Let's get it on:
There will be a test, so pay attention
The tasks involved the performance of simple bipedal dance movements on a laminated grid (see Procedure). The subjects were trained to be proficient at these dance steps in advance of the scanning session, so very little motor learning likely occurred during the experiment. Six conditions were tested, all of them with the eyes closed: (i) a patterned leg movement synchronized to the beat of metric, regularly timed tango music (Metric); (ii) the same patterned leg movement executed to the beat of non-metric, irregularly timed tango music (Non-Metric); (iii) a matched, patterned leg movement performed with no music (Motor);
4) a condition of isometric leg-muscle contractions synchronized to the
beat of metric tango music but with no leg displacement, in which the
left and right leg-muscles contract in alternation (Contractions); 5)
passive music listening with no movement (Listening); and (vi) silent
motionless ‘Rest’. Subjects experienced little visual stimulation during
trials, as they were lying in a dimly light room with their eyes closed. The dance step for the metric and non-metric conditions consisted of a simple 6-step box pattern (Fig. 1a, left panel) — derived from the basic ‘salida’ step of the Argentine tango — that involved alternation of the left and right feet. The pattern for the Motor condition (Fig. 1a, right panel) was different from that of the Metric and Non-Metric conditions in order to minimize the use of mental imagery of the music from the latter two tasks. For the Metric, Non-Metric and Contractions conditions, subjects were instructed to take one step or make one leg-muscle contraction per strong beat of the music, be that to a metric or non-metric rhythm. For the Motor condition, there was no music for the movement to be entrained to, but subjects were instructed to practice the step, and were given feedback to either quicken or slow the movement if their tempo was too dissimilar from that of the Metric condition. As per instruction, and as verified by observation and feedback, all foot movement occurred in a smooth gliding manner on the surface rather than in a stepping manner; the feet never completely lifted off the surface.
While lying supine in the PET scanner, subjects flexed their knees ~90. and placed their feet on a stable, flat, smooth surface, which was inclined ~30. above the horizontal (Fig. 1b). The surface was laminated to reduce friction during foot movement and to minimize compensatory body or head movement. Subjects wore stockings to further reduce friction. Each big toe was visibly marked to facilitate coding of foot position on a grid with centimeter markings, as recorded using a camcorder. The music stimuli were presented via headphones to the subject.
Behavioral analysis of the rate and extent of videotaped leg movement was performed by one of the authors (M.J.M.) for the three movement tasks from the 40 s PET scan. First, the dance steps were performed quite accurately, often with 1 cm of error. Second, the mean number of steps during the 40 s task interval in the Metric, Non-Metric and Motor conditions was 34.5 ± 4.0, 40.9 ± 8.5 and 36.0 ± 1.3 (mean ± SD). There was no difference in the rates between the Metric and Motor conditions (P > 0.05). However, the average movement rate for Non-Metric dance was higher than that for the other two conditions, indicating that the dancers misinterpreted some secondary beats as primary ones; subjects were most likely susceptible to such errors because tango music is highly syncopated. Third, the perimeter of the triangular movement-path outlined by each foot during the dance tasks was measured. The means across the two feet for the Metric, Non-Metric and Motor conditions were 38.1 ± 8.1, 39.5 ± 6.9 and 30.5 ± 6.4 cm2. The perimeter values for the self-paced dance task were significantly shorter than those for the other two tasks. In absolute terms, this amounted to a difference of ~0.7 cm in each direction, which is small compared to the overall extent of movement (on the order of 6 cm in each direction). The shorter paths likely relate to the use of a different pattern for the Motor condition compared to the other two. This matched pattern was designed so as to preserve the same combination of foot movements (forward, backward, left and right) as the pattern for Metric and Non-Metric. Movement extent for these patterns was not predictable a priori. There was no relationship between years of tango experience and age of subject with respect to performance in the scanner as measured by rate of stepping and by extent and accuracy of steps during the task (data not shown). A correlation analysis of the relationship between movement perimeter and regional cerebral blood flow failed to find significant covariance with any of the observed activations.
Comments: Science is an extreme mental discipline. It is one of the hardest things that a person can do. Every variable must be controlled; every word requires proof, every result must be repeatable by other researchers; every sentence is rigorously challenged by your scientific peers before you can publish it; every concept will be challenged by other schools of thought and other avenues of research. The Scientific Method saves us all from superstition and quackery. My neurologist, Dr. Calvin Melmed, a great spirit, a real DOCTOR, a man of science and a humanitarian, has reminded me three times: "Where is the science? Dancing is good for you, and good for everybody. But to fight a disease? You need the science. You cannot change the medical treatment of millions of people just because you feel good. You must have scientific proof." Dig it. More to come:
During the PET session, the subject’s head was immobilized using a closely fitted thermal-plastic facial mask with openings for the eyes, ears, nose and mouth. Auditory stimuli were presented through earpieces taped over the subjects’ ears. During scanning, subjects were instructed to close their eyes and to minimize head and body movement as much as possible. During a session in advance of the scan date and during a practice session on the day of the scan, subjects practiced performing the dance steps in a highly controlled manner while minimizing head movement. Subjects had two PET scans for each of the six tasks. Each scanning session began with the Motor task, and the tasks were thereafter counterbalanced across subjects such that one replication of each task was performed before the next set began. Subjects began each task 30 s prior to injection of the H215O bolus. Bolus uptake required ~20 s to reach the brain, at which time a 40 s scan was triggered by a sufficient rate of coincidence-counts, as measured by the PET camera. At the end of the 40 s scan, the auditory stimulus was terminated and the subject was asked to discontinue the task and lie still during an immediately following 50 s scan.
“Subjects began each task 30 s prior to injection of the H215O bolus. Bolus uptake required ~20 s to reach the brain …” Whoa! Are these people serious about it, or what? That’s why we thanked the dancers first. As dancers do, they put their bodies on the line, for truth.
Positron emission tomography was performed with a CTI HR+ camera, which had a pixel spacing of 2.0 mm, an inter-plane, center-to-center distance of 2.4 mm and 63 transaxial scan planes. Images were reconstructed using a Hanning filter with a cut off frequency of 0.5, resulting in images with a spatial resolution of ~4.3 mm (full-width at half-maximum). The data were smoothed with an isotropic 10 mm Gaussian kernel. Anatomical MRI data was acquired with an Elscint 1.9 T Prestige system with an in-plane resolution of 1 mm2 and 1.5 mm slice thickness. Convex-hull spatial normalization was performed prior to group subtraction (n = 10) using change distribution analysis’ methods (Raichle et al., 1983; Fox et al., 1988; Mintun et al., 1989). Thus, significant changes in cerebral blood flow indicating neural activity were detected with a region-of-interest-free image subtraction strategy. Intra-subject image averaging was performed within conditions (Fox et al., 1988). Images were spatially normalized (Fox et al., 1985) into proportional, bicommissural coordinate space relative to the Talairach atlas (Talairach and Tournoux, 1988) using spatial normalizationperformed by using SN. Inter-scan, intra-subject movement was assessed and corrected using the Woods’ algorithm (Woods et al.,
1993). A search algorithm (Mintun et al., 1989) was used to identify local extrema within a 5 3 5 3 5 voxel search cube. A beta-2 statistic measuring kurtosis and a beta-1 statistic measuring skewness of the extrema histogram (Fox et al., 1988) were used as omnibus tests to assess overall significance (D’Agostino et al., 1990). Critical values for beta statistics were chosen at P <> 4.27, P <>
The analyses of functional images for factors in our experimental design were conducted employing the following planned comparisons. A systems-level view of dancing to music was analyzed by comparing Metric and Rest conditions. The functional neuroanatomy of movement entrainment was analyzed by comparing Metric and Motor conditions, Metric and Music conditions, and Music and Rest conditions. Metric movement was analyzed by comparing Metric and Non-Metric conditions. Spatial pattern of leg movement was analyzed by contrasting Metric and Non-Metric, respectively, with the Contractions conditions. Spatial patterning of movement was also examined in the comparison of Motor and rest conditions. Finally, to eliminate spurious activations resulting from the subtractions of deactivations, we verified every activation in each higher-level contrast by examination of the relevant condition from rest.
A systems-level view of the brain areas contributing to comparativelynatural, although supine, dance performance (Metric
dancing minus Rest, Table 1) revealed activations in bilateral motor, somatosensory and premotor areas, right supplementary motor area, right frontal operculum, left medial superior parietal cortex, superior temporal regions, right cingulate motor area, basal ganglia, and bilateral anterior vermal and posterior-lateral cerebellum. The following planned and post hoc comparisons examine the specific subsystems activated during dance. There were no significant differences detected between male and female dancers in the profiles of activations across conditions(data not shown).
Entrainment of Movement to Music
When dance in a self-paced manner without music (Motor) was subtracted from dance entrained to a musical beat (Metric), the principal signal for this subtraction outside of auditory areas was seen in the vermis of anterior cerebellar lobule III (Fig. 2, Table 2). This activation was also present in the analysis of each task minus rest. These data implicate the anterior vermis in entrainment of movement to music. No such increases for entrainment were observed in other regions activated in common between entrained and self-paced dancing, such as sensorimotor, premotor, superior parietal, cingulate or frontal opercular areas. The increase in vermal activity for entrained dancing was not due to the addition of the accompanying music per se, as the subtraction of music listening from metric dance had no effect on the intensity of the anterior vermal activity (not shown). In addition, this latter subtraction, while eliminating all activity in cortical auditory areas found in both the music-listening condition alone and in entrained dance minus self-paced dance, revealed a significant signal in the right medial geniculate nucleus (Fig. 3a,b). The medial geniculate nucleus was not active during self-paced dance without music. Moreover, metric dance minus self-paced dance and, especially, metric dance minus music listening revealed activity in lateral and vermal aspects of posterior left cerebellar lobules V and VI (Figs 2 and 3b). As suggested later, audiomotor entrainment may be
mediated through the transmission of coarsely processed beat information from subcortical auditory areas to the cerebellum.
For all tables, Talairach atlas coordinates are in millimeters along the left--right (x), anterior--posterior (y) and superior--inferior (z) axes. In parentheses after each brain region is the approximate Brodmann area, except in the case of the cerebellum, in which the anatomical labels of Schmahmann et al. (2000) are used. The intensity threshold for this
table is Z [ 4.27, P \ 0.00001 (one-tailed).
Figure 2 The activation (–2,–44,–12) in anterior cerebellar vermis (lobuleIII) in the analysis of Metric dance minus Motor. Also shown is the activation in lobule V. The group mean activations are shown registered onto an averaged brain in all figures. The right side of the figure is the right side of the brain in all figures. At the left end of the figure is a colorscale for the intensityof the activations. The intensity threshold in Figures2—5 is Z>2.58, P<0.005m/s1600-h/table+2.jpg">
Figure 3. (a) Activation in the medial geniculate nucleus in the analyses of Metric dance minus Motor and (b) Metric dance minus music Listening. The peak coordinate for the medial geniculate nucleus in the former subtraction is at (14, –28, –4) whereas that for the latter is at (14, –30, –6). Activity in cerebellar lobule V and VI in the analysis of Metric dance minus music Listening. In coronal and axial views throughout, the left side of the image represents the left side of the brain.
(Figs 2 and 3b). As suggested later, audiomotor entrainment may be mediated through the transmission of coarsely processed beat information from subcortical auditory areas to the cerebellum.
In comparing the dance condition entrained to a metric rhythm with a dance condition entrained to an irregular rhythm (Table 3), we observed a strong signal bilaterally in the putamen for the metric condition, with an emphasis in the right putamen (see Fig. 4, which shows this same pattern when each condition is contrasted with rest). This activity appeared to occur in the somatotopic representation of the leg (Maillard et al., 2000). No such basal ganglia activation was detected for non-metric dance. By contrast, in non-metric dance minus rest, a strong activation was seen in right ventral thalamus, specifically, in the ventral posterior nucleus bordering on the pulvinar. This area was not activated above threshold for metric dance. To explore this pattern of findings, we noted that in a post hoc comparison there were intermediate levels of activity in both putamen and thalamus observed in the performance of both self-paced dancing without music (Motor) and isometric leg-muscle contractions to music (Contractions condition) (Fig. 4, legend). This overall pattern points to a reciprocal relationship in activity between the putamen and ventral thalamus, with metric movement exhibiting strong putamen and weak thalamus responses, and non-metric movement showing the reverse profile. This reciprocity may be causal, as mediated, for example, by the indirect pathway of the basal ganglia circuit, where the ventral thalamus is the major output structure of the basal ganglia (e.g. Rao et al., 1997). Alternatively, it may be incidental, mediated through the intervention of other structures. In either case, the results suggest that the putamen is preferentially activated by movement patterns that are regular and predictable, and that irregular and unpredictable patterns activate other pathways, including those containing the ventral thalamus.
Spatial Patterning of Leg Movement
The analysis of metric dance minus the contractions condition (where the leg muscles were contracted isometrically in an alternating fashion to the beat of metric tango music but without there being leg movement along the surface) revealed activity in the medial portion of the superior parietal lobule [Brodmann’s area (BA) 5/7; precuneus] (Fig. 5, Table 4). Superior parietal activity was equally strong in the subtraction of the contractions condition from either metric dancing or non-metric dancing and for self-paced dancing minus rest, thereby suggesting that this area is specifically involved in spatial guidance of leg movement independent of temporal parameters related to movement timing or entrainment. The dance tasks were performed quite accurately without visual guidance (see Behavioral Analysis), and thus the foregoing activations suggest that the medial superior parietal lobule encodes proprioceptive or somatosensory information about spatial coordinates for leg movement.
Activations seen in each of the movement tasks (minus rest) occurred in the primary motor and sensory cortices (paracentral lobule), premotor cortex,supplementary motor area (SMA), and somatotopic leg areas of the cerebellum (lobules IV and VIII) (Fig. 5; see Table 1 for the coordinates of these activations in metric dance minus rest). In addition, the right frontal operculum (BA 44/6) — near the right-hemisphere homologue of Broca’s area — was activated in all of the tasks involving motor production (minus rest) but not in passive music listening (Fig. 6). Because the intensity of this activation did not vary in each of the movement tasks (minus rest), we attribute its function more to motor sequencing in general (Ehrsson et al., 2000; Janata and Grafton, 2003) than to spatial patterning of limb movement. Finally, activation in the cingulate motor area was also detected for all of the motor conditions (minus rest), although much more so for the movement conditions than for the condition of isometric leg-muscle contractions (see the middle panel of Fig. 5). This activity likely reflects a somatotopic map in this region, as the dorsal bank of the cingulate sulcus in monkeys contains a leg representation, and electrical stimulation of this region leads to hindlimb movement (Luppino et al., 1991).
These findings illustrate the coordination of distributed neural systems that underlie bipedal, cyclically repeated dance steps entrained to a musical rhythm. The functional subsystems can be summarized as follows. The superior temporal gyrus and superior temporal pole represent the melodic and harmonic
aspects of the heard music. In parallel, the medial geniculate nucleus appears to send inputs, via brainstem relay nuclei, to the anterior cerebellar vermis and lobules V and VI regarding beat information, to support the entrainment of movement to a musical beat. The basal ganglia, and particularly the putamen,subserve the selection and organization of segments of action,
right cingulate motor area processes aspects of movement intention and the allocation of motor resources. Finally, medial aspects of the superior parietal lobule subserve kinesthetically mediated spatial guidance of leg movement during navigation in dance. This network of brain areas controlling dance will require confirmation and refinement in future studies. There was a trend for right hemisphere dominance for many of our unilateral activations, including the frontal operculum, cingulate motor area, putamen, ventral thalamus and medial
geniculate nucleus, along with a corresponding focus in the left cerebellar vermis. Although no such trend has been observed for natural gait (Fukuyama et al., 1997; Miyai et al., 2001), a study that compared flexion/extension of the ankle, flexion/ extension of the wrist and finger movement, all on the right side of the body (Luft et al., 2002), found that knee movement was accompanied by strong ipsilateral activation in the primary motor cortex and primary sensory cortex. Likewise, a study of imagined or executed flexion/extension of the toes (Ehrsson et al., 2003) reported right frontal operculum activation for right-sided movements.
In addition, recent studies of postural deficits in stroke patients suggest the presence of a distributed system, primarily in the right hemisphere, for representing trunk posture relative to the environment (e.g. Spinazzola et al., 2003). Further research will be needed to clarify the extent to which horizontal placement of the body during the dance tasks might have contributed to our observed right-hemisphere lateralization effects. In future, studies may be able to circumvent the neuroimaging constraints that currently prevent an investigation of important issues such as upright whole-body movements and interpersonal coordination processes involved in pair dancing
(e.g. as in celebrated instances of Rogers doing Astaire’s steps backward and in high heels).
In the following comments, we discuss these observations in more detail, with a view toward assessing how the mechanisms observed in paradigms focusing on controlled elementary processes ‘scale up’ to those for the more complex natural activity of dancing. At a general level, we note that elements of both discrete and rhythmic movements (Schaal et al., 2004) are present in dance, itself a gestural system. As such, the patterns of activations we observed are broadly consistent with the possibility that the subcortical systems activated are involved in the timing and coordination of discontinuous movements, whereas the specific cortical systems activated here may be supporting the control of the continuous movements (Miall and Ivry, 2004).
Figure 4. Reciprocal activation in the putamen and ventral thalamus in Metric and Non-Metric dance. The thalamic activation occurs at the junction of the ventral posterior and pulvinar nuclei. Direct subtraction of Non-Metric from Metric retains the activation in the putamen, and direct subtraction of Metric from Non-Metric retains the activation in the thalamus (not shown). The Z-score values for the putamen and ventral thalamus, respectively, for the four motor conditions were: Metric (5.00, 2.49), Non-Metric (undetectable, 5.78), Motor (3.44, 3.10) and Contractions (3.27, 3.19).
Figure 5. Metric dance minus leg-muscle Contractions eliminates the activations in the premotor and motor-sensory cortices, and leaves signal in the superior parietal lobule, spanning the medial part of BA 5 and 7 (precuneus). Activation in the cingulate motor area is seen to be just above threshold. The activations in the motor-sensory cortex and SMA/ premotor cortex are bilateral even though the figure shows only one hemisphere. The mesial motor strip comprised of SMA/premotor cortex and motor-sensory cortex is active in (a) and (b). (c) The peak coordinate of the anterior focus in the precuneus is (8, –56, 60).
Figure 6. Cerebral blood flow increases in the right frontal operculum (BA 44, see arrows) during the four motor tasks but not music listening. The peak voxel for Motor minus Rest is at (54, 6, 8), for Contractions minus Rest is at (54, 6, 6), and for Metric minus Rest is at (54, 8, 6). The activation in the Non-Metric minus Rest subtraction was slightly more posterior to those for the preceding three conditions and mapped more onto BA 6 than BA 44, with a peak voxel at (58, 0, 6).
Three key facets of dance were selectively analyzed in our study design: audiomotor entrainment, meter and patterning of movement. For the first facet, a comparison between two matched dance patterns performed at the same rate — one requiring entrainment to a musical beat and the other one self-paced highlighted the importance of the anterior cerebellar vermis (central lobule, III), but not other parts of the motor or sensory system, to entrainment processing. The vermal activation was equally strong in the Metric and Non-Metric dance conditions (data not shown), both of which were based on temporal entrainment. This suggests that the vermis functions in entrainment per se, independent of the nature of the temporal pattern being entrained to. Very similar activations have been observed in recent studies of repetitive lower-limb movement entrained to metric cues in either auditory or visual form. In a functional magnetic resonance imaging (fMRI) study of right-foot flexion/extension timed to a metric auditory cue (Debaere et al., 2001), activity was detected in the anterior cerebellar vermis (3, –45, –18, as compared to 0, –48, –16 here), as it was in a similar fMRI study of right-foot flexion/extension timed to a metric visual cue (Sahyoun et al., 2004; i.e. 2, 48, –20, as compared to 0, –48, –16 here). This region was also activated by rhythmic self-paced walking (Fukuyama et al., 1997: 0, –50, –20, as compared to 0, –48, –16 here) and finger tapping without an ongoing external stimulus (Penhune et al., 1998: 8, –48, –21 and 1, –50, –15, as compared to 0, –48, –16 here). These similarities between the controlled elementary paradigms and dancing imply that common mechanisms are involved in entrainment. This pattern of data highlights the importance of the anterior cerebellar vermis (III) for the entrainment of movement to external timing cues.Interestingly, in our analysis of functional activation data for metric dance minus passive music listening, activity in cortical auditory areas was eliminated, leaving behind a significant signal in the right medial geniculate nucleus as well as posterior cerebellar lobules V and VI. These regions were not found to be activated during self-paced dance steps performed without music (when contrasted to rest). There are substantial reciprocal projections between the thalamic nuclei and cerebellum via brainstem relays (see reviews Schmahmann, 1997). In addition, cerebellar lobules V and VI have been specifically implicated in neuroimaging studies of pitch and melody discrimination, as dissociated from motor coordination or cortical motor activity (e.g. Holcomb et al., 1998; Griffiths et al., 1999; Gaab et al., 2003; Parsons, 2003a; Petacchi et al., 2005). It is thus possible that the sensory input to the anterior cerebellar vermis for entrainment processing involves coarsely processed auditory information from subcortical sites. If so, this would imply that entrainment in dance does not require higher-level musical content (e.g. tonality, harmony, timbre) but may simply depend on low-level beat information, as mediated by subcortical pathways. This hypothesis may in part account for the sharing of entrainment mechanisms between dance and simple sensorimotorbehaviors like finger tapping and ankle rotation.
There is strenuous ongoing debate about the function of the cerebellum, which was classically viewed as a motor-control and coordination structure only but which has recently been implicated in non-motor processes by a wide range of findings (e.g. see reviews in Schmahmann, 1997; Ivry and Fiez, 2000; Rapoport et al., 2000; Vokaer et al., 2002; Bower and Parsons, 2003). The following three hypotheses of cerebellar function are likely to be most pertinent to the findings of this study. One account assumes that the cerebellum embodies internal forward-inverse model pairs (Wolpert et al., 1998); in the present case, such model pairs would need to include the sensory aspects of movement, the movements per se and perception of the auditory beat. Another account (Ivry, 1997) emphasizes the role of cerebellum in supporting timing processes in both the preparation and coordination of motor responses (in vermal and anterior cerebellum) and the sensory perception of duration on the order of hundreds of milliseconds (in lateral cerebellum). In a third view, the role of the cerebellum is to optimize the control of the acquisition of sensory data (Bower, 1997; Bower and Parsons, 2003). In the current case, the cerebellum would be hypothesized to assist cortical, subcortical and peripheral neural structures in collecting optimal auditory and somatosensory information in order to influence the cortical motor system to better synchronize the execution of movement with the auditory rhythm. Further research is needed to clarify the functions of the foregoing cerebellar regions. Metric and Non-metric Movement Another principal feature of our results was seen in the contrast between dance steps entrained to a metric rhythm and the same steps entrained to a non-metric rhythm. We found that metric dance movement induced strong activity bilaterally in the putamen, and especially the right putamen. Non-metric dance movement, in contrast, showed no activity in the putamen but instead displayed a strong signal in the right ventral thalamus. A variety of prior research affirms a role for the basal ganglia in the control of metric movement in rhythmic tapping tasks (e.g. Rao et al., 1997; Penhune et al., 1998) and in piano performance of memorized musical pieces (Parsons et al., 2005). The involvement of the putamen in metric movement is supported by an fMRI study of visually cued, metric right-foot flexion/extension (Sahyoun et al., 2004); the thalamus was much less active. Likewise, in a PET study of the same task (Ehrsson et al., 2000), activity in the putamen, but not the thalamus, was reported. The involvement of the ventral thalamus in non-metric rhythms agrees with similar findings from an fMRI study (Lutz et al., 2000) of tapping the right index finger to a non-metric, randomly timed visual cue (–20, –16, 12, as compared to 18, –24, 8 here).
Overall, the reciprocal activity between putamen and ventral thalamus just described suggests that for both dance and elementary movements, the basal ganglia are preferentially activated in the execution of motor activities having a regular, predictable rhythm and that unpredictable unfamiliar temporal patterns recruit other pathways. This is also congruent with an fMRI study of self-paced finger tapping showing that the basal ganglia were principally active for simple rhythms and that their activity decreased with greater rhythmic complexity, whereas the thalamus (and anterior cerebellar vermis) increased in activation with increasing complexity (Dhamala et al., 2003). In our study, intermediate levels of activity in both the putamen and ventral thalamus were seen for both self-paced dancing without music and for the performance of isometric leg-muscle contractions to metric tango music. Thus, activity in the basal ganglia circuit is modulated by limb displacement and entrainment as well as by the presence or absence of metric regularity. This complex functionality suggests that the basal ganglia might be one part of the brain sensitive to the interactions amongst entrainment, meter and spatial patterning specifically seen in dance.
Somatotopy and Control of Lower Limbs
The third principal aspect of our data highlights the topographic representation of the lower extremity in the motor-sensory cortex as well as in the superior parietal lobule, cingulate motor area, cerebellum, and putamen. Activation of a mesial strip encompassing the leg representation in the primary motor cortex, somatosensory cortex, SMA and premotor cortex was present for all four tasks involving motor activity. Activations in very similar somatotopic regions for the lower extremity have been found in a number of reports using a variety of techniques and paradigms, including the following: a SPECT study of upright walking (Fukuyama et al., 1997); a near-infrared spectroscopy study of bipedal walking on a treadmill (Miyai et al., 2001); an fMRI study of right-foot flexion/extension timed to a metric auditory cue (Debaere et al., 2001); a PET study of this same task (Ehrsson et al., 2000); an fMRI study of metric, visually cued right-foot flexion/extension (Sahyoun et al., 2004); an fMRI study of unipedal flexion/extension of either the left or right knee joint (Luft et al., 2002); an fMRI study of the placement of either foot into visually presented foot postures (Chaminade et al., 2005); and an fMRI study of imagined and executed flexion/extension of the toes timed to a metric auditory cue (Ehrsson et al., 2003). The foregoing motor, premotor, and SMA areas most likely encode parameters related to muscle group, contractile force, initial and final position, and movement direction (Graziano et al., 2002). The SMA, the cingulate motor area and possibly the cerebellum (Ivry, 1997; Wolpert et al., 1998) are likely involved in interhemispheric coupling supporting cyclically repeated coordination of the two homologous limbs, as suggested by studies of bimanual coordination (e.g. Ja¨ ncke et al., 2000). Very similar activations were observed for coordinated unilateral movements of the hand and foot (Ehrsson et al., 2000; Debaere et al., 2001).
The right frontal operculum (BA 44/6) was activated by the four tasks involving motor production but not by music listening, suggesting a general role for this area in motor sequencing rather than a specific role in either spatial patterning or metric entrainment. Responses in the frontal operculum were observed in an fMRI study (Ehrsson et al., 2003) of flexion/ extension of the toe timed to a metric auditory cue (56, 8, 4 and 56, 8, 0, as compared to 54, 8, 6 here), both during mental imagery of movement and actual movement. Comparable activity was also reported in an fMRI study of visually cued metric right-foot flexion/extension (Sahyoun et al., 2004). In a PET study of finger tapping timed to imitate the rhythm of brief sequences of visual stimuli with long or short elements (Penhune et al., 1998), the right frontal operculum was also activated (46, 18, 3, as compared to 54, 8, 6 here). In addition, this region shows activations for motor mental imagery, perception, and imitation tasks involving the hands (Parsons et al., 1995; Grafton et al., 1996; Heiser et al., 2003). Equally pertinent, a region anterior to this one was activated in ballet dancers while observing ballet movements and in capoeira dancers while observing capoeira movements (peaking at Talairach coordinates 54, 35, 1: Calvo-Merino et al., 2005), thereby demonstrating expertise-dependent activity in the frontal operculum. These results support a role for this region in both elementary motor sequencing and in dance, during both perception and production. This activation pattern may also bear on new functional hypotheses that propose supralinguistic sequencing and syntax operations for the region broadly defined as Broca’s region and its right homologue.
The right cingulate motor area (cingulate sulcus) was activated in all four motor tasks, with trends toward larger extents in the three tasks requiring movement of the legs. Similar activity was observed in an fMRI study (Sahyoun et al., 2004) of visually cued metric right-foot flexion/extension (6, 14, 44 and –10, –6, 48, as compared to 8, –6, 44 here). The location of this activation corresponds to cytoachitectonic area 24dd (Vogt and Vogt, 2003), which in monkeys contains a topographic representation of the hindlimb and lower trunk (Luppino et al., 1991; Rizzolatti et al., 1996). This area may encode aspects of movement intention and the allocation of motor resources (Ball et al., 1999), processes required in both elementary motor activities and in dance.
Posterior parietal cortex is regarded as subserving a variety of spatial cognitive functions (Colby and Goldberg, 1999; Parsons, 2003b) including those related to body schema (Berlucchi and Aglioti, 1997; Halligan et al., 2003). The inferior and superior parietal lobules receive both somatosensory and visual inputs. The posterior parts of both lobules process visual information, the anterior superior parietal lobule processes somatosensory information, and the anterior inferior parietal lobule integrates somatosensory and visual information (Rizzolatti et al., 1997; Colby and Olsen, 2003). The performance of dance steps with the eyes closed was reported by some of our subjects to be accompanied by mental imagery of their body. Thus, the foregoing parietal activation was likely involved in spatial cognitive functions based on proprioceptive processing of leg position and joint angle and on somatosensory contact of the feet with the surface (Parsons, 1987). Little is known about leg representations in posterior parietal cortex of either humans or monkeys. The monkey homologue of BA 5, area PE, contains leg representations (Rizzolatti et al., 1996). PE is a high-level somatosensory area that does not receive visual inputs and that projects to primary motor cortex. Studies of hand movement find that PE neurons encode limb location in space using a body-centered coordinate system (Lacquaniti et al., 1995). Thus, the medial superior parietal lobule may possess a map of peripersonal (egocentric) space based on proprioceptive cues related to lower limb position. Somesthetic guidance of navigation is key to dance, where vision provides a support role indicating whether space is sufficient to carry out particular movements. The fact that superior parietal lobule is activated in some of the studies of elementary ankle and wrist rotation discussed earlier might suggest that, unlike isometric muscle contraction, these simple movements still have a basic element of spatial patterning to them, simple though it might be. This suggests that activity in this region increases as the spatial and navigational demands of the movement increase.
Dance, like numerous natural, complex sensorimotor activities (e.g. sport, group physical labor, and musical performance), requires the integration of spatial pattern, rhythm, synchronization to external stimuli and whole-body coordination. Our findings suggest that many of the brain areas activated for dance are also recruited in elementary sensorimotor activities. However, the present methods can only show proximity or overlap in the location of neural activity. These results set the stage for more precise techniques that compare the detailed neural computations (e.g. Gold and Shadlen, 2001) performed in regions localized in common for simpler and more complex sensorimotor activities. Moreover, the present findings are based on observations of relatively skilled dancers who are well practiced at performing the dance steps in the study. Thus, our data do not reveal the role of learning in organizing the various elementary and complexity-related neural mechanisms. It is likely that learning or refinement of natural complex tasks would entail changes in functional and effective connectivity, and in the reorganization and redistribution of processes (Garraux et al., 2005; Kelly and Garavan, 2005). Indeed, we observed greater variety and number of anterior cerebellar activations during the most unfamiliar condition (Non-Metric, Table 4), suggesting a role in adjusting fine, complex sensorimotor coordination to relatively novel entrainment signals (Ivry, 1997; Wolpert et al., 1998; Bower and Parsons, 2003). Future studies will surely harvest significant information from studies of these and other aspects of complex natural activities. Our findings specifically elucidate for the first time the neural ystems and subsystems that underlie dance. These observations imply that dance, as a universal human activity, involves a complex combination of processes related to the patterning of bipedal motion and to metric entrainment to musical rhythms. More broadly, this study brings us closer to a richer understanding of the neural and psychological bases of complex, species-specific creative and artistic behaviors. This study is part of a contemporary wave of research exploring new neuroscientific hypotheses in the context of activities such as musical performance, drawing, visual aesthetics, dance observation and the viewing of cinematic narratives (Ino et al., 2003; Kawabata and Zeki, 2003; Makuuchi et al., 2003; Cela-Conde et al., 2004; Hassan et al., 2004; Calvo-Merino et al., 2005; Parsons et al., 2005).
This research was supported by grants from the ChevronTexaco Foundation and the International Foundation for Music Research. None of the two hundred million dollars a year spent on Parkinson's research was spent on this.
Address correspondence to Lawrence Parsons, Department of Psychology, University of Sheffield, Western Bank, Sheffield S10 2TP, UK. Email: L.firstname.lastname@example.org.
Cerebral Cortex V 16 N 8 1165Blessed are the dancers
Blessed are the scientists
They so love the world
and we thank you
to know that we are not alone