Catching falling objects: The role of the cerebellum in processing sensory-motor errors that may influence updating of feedforward commands. An fMRI study

Abstract

The human motor system continuously adapts to changes in the environment by comparing differences between the brain's predicted outcome of a certain behavior and the observed outcome. This discrepancy signal triggers a sensory-motor error and it is assumed that the cerebellum is a key structure in updating this error and associated feedforward commands. Using fMRI, the aim of the present study was to determine the main cerebellar structures that are involved in the processing of sensory-motor errors and in updating feedforward commands when simply catching a falling ball without displacement of the hand. Subjects only grasped the ball with their fingers when receiving it in their hand. By contrasting functional imaging signal obtained in conditions in which it was possible and impossible to predict the weight of the ball, we aimed to highlight sensory-motor error processing which we expected to be more marked in the conditions without prediction (less accurate feedforward process or more important feedback corrections) with respect to conditions with prediction (more accurate feedforward process or less important feedback corrections). When catching a falling ball and the possibility of prediction about the ball weight was manipulated, our results showed that both the right and left cerebellum is engaged in processing sensory-motor errors. It may also be involved in updating feedforward motor commands, perhaps on a trial by trial basis. In addition, when subjects were blindfolded, we observed a similar network but centered in a more anterior portion of the right cerebellum and we noted the presence of a cerebellar-thalamo-prefrontral network that may be involved in cognitive prediction (rather than sensory prediction) about ball weight.

Full text

CATCHING FALLING OBJECTS: THE ROLE OF THE CEREBELLUM IN
PROCESSING SENSORY–MOTOR ERRORS THAT MAY INFLUENCE
UPDATING OF FEEDFORWARD COMMANDS. AN fMRI STUDY
L. FAUTRELLE,
a,b
C. PICHAT,
c
F. RICOLFI,
b,d
C. PEYRIN
c
AND F. BONNETBLANC
a,b
*
a
Université de Bourgogne, Dijon, Campus Universitaire, UFR STAPS,
BP 27877, F-21078 Dijon, France
b
INSERM, U887, Motricité-Plasticité, UFR STAPS, Dijon, F-21078,
France
c
Laboratoire de Psychologie et NeuroCognition, CNRS UMR5105,
Université Pierre Mendès France, 38040 Grenoble Cedex 09, France
d
Centre Hospitalier Universitaire de Dijon, Service de Neuroradiologie,
Hôpital Général, 21033 Dijon, France
Abstract—The human motor system continuously adapts to
changes in the environment by comparing differences be-
tween the brain’s predicted outcome of a certain behavior
and the observed outcome. This discrepancy signal triggers
a sensory–motor error and it is assumed that the cerebellum
is a key structure in updating this error and associated feed-
forward commands. Using fMRI, the aim of the present study
was to determine the main cerebellar structures that are
involved in the processing of sensory–motor errors and in
updating feedforward commands when simply catching a
falling ball without displacement of the hand. Subjects only
grasped the ball with their fingers when receiving it in their
hand. By contrasting functional imaging signal obtained in
conditions in which it was possible and impossible to
predict the weight of the ball, we aimed to highlight sen-
sory–motor error processing which we expected to be
more marked in the conditions without prediction (less
accurate feedforward process or more important feedback
corrections) with respect to conditions with prediction
(more accurate feedforward process or less important
feedback corrections). When catching a falling ball and the
possibility of prediction about the ball weight was manipu-
lated, our results showed that both the right and left cerebel-
lum is engaged in processing sensory–motor errors. It may
also be involved in updating feedforward motor commands,
perhaps on a trial by trial basis. In addition, when subjects
were blindfolded, we observed a similar network but centered
in a more anterior portion of the right cerebellum and we
noted the presence of a cerebellar–thalamo–prefrontral net-
work that may be involved in cognitive prediction (rather than
sensory prediction) about ball weight. © 2011 IBRO. Pub-
lished by Elsevier Ltd. All rights reserved.
Key words: fMRI, cerebellum, feedforward control, sensory–
motor error.
When we are catching falling or thrown objects with our
hand and the object can be seen, electromyographic acti-
vations occur in the arm in advance of the estimated time
to contact and before feedback processes are able to
compensate a posteriori for the perturbation (Lacquaniti
and Maioli, 1989a,b). This classical observation proves
that the brain is able to exert some level of feedforward
control over any movement we perform. More specifically,
these so-called anticipatory postural adjustments (APA)
are fine-tuned according to the weight of the object and
illustrate the brain’s ability to anticipate and predict the
upcoming motor perturbation and sensory outcomes be-
fore they occur. It has been shown that this prediction
capacity is severely impaired in cerebellar patients, as
APAs are cancelled before perturbation onset (Babin-
Ratté et al., 1999; Lang and Bastian, 1999, 2001; Nowak et
al., 2002, 2007).
Predicting the sensory consequences of a gesture is
necessary for on-line control. The difference between the
brain’s predicted outcome of the behavior (or efferent
copy) and the observed outcome is called the sensory–
motor error. This error allows motor corrections to be trig-
gered more rapidly by compensating for biophysical trans-
duction, transmission and processing delays (Desmurget
and Grafton, 2000).
Another important function of this sensory–motor error
is to drive motor adaptations that occur on a trial by trial
basis during the completion of repetitive actions (Tseng et
al., 2007). This process is used to update feedforward
commands and calibrate internal models or representa-
tions of the dynamics of our own body or of our physical
interaction with the environment (Wolpert and Miall, 1996).
It is assumed that the cerebellum stores a motor memory
in the form of internal models and is a key structure in
updating these internal models (Marr, 1969; Blomfield and
Marr, 1970; Kawato et al., 1987; Johansson and Cole,
1992; Wolpert and Miall, 1996; Kawato and Wolpert, 1998;
Wolpert and Kawato, 1998; Wolpert et al., 1998; Johans-
son, 1998; Ito, 2000; Blakemore et al., 2001).
In most experiments, the cerebellum-dependent adap-
tation is investigated by inducing systematic sensory–mo-
tor conflicts (Bock, 1992; Shadmehr and Mussa-Ivaldi,
1994;Martin et al., 1996; Krakauer et al., 2000; Morton and
Bastian, 2004, 2006), for instance discrepancies between
vision and proprioception (Tseng et al., 2007) or between
vestibular and non vestibular sensory graviceptor (Dha-
rani, 2005). In the case of a simple change in the dynamics
of the environment, and when the system is not fed by a
constant error (unpredictable event), it has also been sug-
*Correspondence to: F. Bonnetblanc, Université de Bourgogne, Dijon,
Campus Universitaire, UFR STAPS, BP 27877, F-21078 Dijon,
France. Fax: ⫹33-(0)3-80-39-67-02.
E-mail address: francois.bonnetblanc@u-bourgogne.fr (F. Bonnet-
blanc).
Abbreviations: BA, Brodman area; ROI, regions of interest; SMA,
supplementary motor area.
Neuroscience 190 (2011) 135–144
0306-4522/11 $ - see front matter © 2011 IBRO. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.neuroscience.2011.06.034
135

gested that the predicted state and/or motor command is
constantly updated using past experiences (Bhushan and
Shadmehr, 1999; Donchin et al., 2003; Pasalar et al.,
2006). In a process similar to pure motor adaptation, the
cerebellum generates a discrepancy signal between the
predicted sensory consequences as compared to periph-
eral feedback, and the sensory prediction error is itera-
tively modified (Bastian, 2006). This entire process may be
the origin of permanent updating of feedforward com-
mands and more particularly of anticipatory postural ad-
justments. Both functions are summed-up in Fig. 1.
In other words, the cerebellum is presumably the loca-
tion where the efference copy is transformed to a sensory
prediction that is compared to the sensory input. The dis-
crepancy signal (sensory prediction error) is believed to be
the output of the cerebellum. The sensory prediction error
influence adaptation of the forward model (in the cerebel-
lum), online corrections (Tseng et al., 2007, through cor-
tex), and probably changes of the inverse model (maybe
also at the cortical level).
To further investigate the hypothesis which holds that
unexpected and variable errors are processed in the cer-
ebellum to update feedforward commands when repeti-
tively catching a ball, we used a block design fMRI exper-
iment in which balls looked exactly the same but the level
of incertitude (i.e. prediction) about their weight varied. In
certain conditions, the weight of the ball was predictable, in
others, it was not. By contrasting functional imaging signal
obtained in conditions in which it was possible and impos-
sible to predict the weight of the ball, we aimed to highlight
sensory–motor error processing which we expected to be
more marked in the conditions without prediction (less
accurate feedforward process or more important feedback
corrections, i.e. more important sensory–motor error) with
respect to conditions with prediction (more accurate feed-
forward process or less important feedback corrections,
i.e. less important sensory–motor error). In this vein and
according to the Bayesian framework it has been demon-
strated that the nervous system adapts more when its state
estimate is more uncertain (Wei and Körding, 2010). We
hypothesized that we would solicit a cerebellar network
that drives the processing of sensory–motor errors that
may update feedforward commands. In addition, before
catching an object with the hand, vision allows for the
critical estimation of time to contact (Lacquaniti and Maioli,
1989a). In order to determine whether cerebellar networks
involved in the updating of sensory–motor errors are de-
pendent on visual cues, we suppressed sight in a final
condition without prediction of the ball weight. In addition,
in this latter “blind” condition, we also hypothesized that
prediction of sensory consequences would be less accu-
rate in comparison to the condition in which vision was not
removed, and that in this case, sensory–motor errors
would be compensated by feedback processes in a greater
extent. In other words, when sight is suppressed, the pre-
diction of sensory consequences may also be impaired
due to poorest sensory estimations that may be cumulated
with impaired updating of forward modeling.
EXPERIMENTAL PROCEDURES
Participants
Sixteen healthy participants [all males, 27.9⫾4.7 years old,
180.75⫾5.6 cm, and 76.5⫾7.6 kg] volunteered for the experiment.
All participants had normal or corrected to normal vision (lenses)
and none of them had a previous history of neuromuscular or
neurological disorders. All subjects were right handed as as-
sessed by the Edinburgh Handedness Inventory (Oldfield, 1971).
The entire experiment conformed to the Declaration of Helsinki
and informed consent was obtained from all participants according
to the guidelines of the clinical ethic committee of the University of
Burgundy.
Experimental design and fMRI paradigm
A block design paradigm was used that alternated periods of rest
and a motor task. It resulted in functional imaging signal of the
right-hand repetitive ball catching task as contrasted to the rest
periods in the four different catching conditions.
Participants were lying on their back in the scanner and the
upper right limb was elevated by 5 cm so that the right hand was
not in contact with anything during the ball catching sessions. A
double-mirror mounted on the MRI head coil was adjusted to allow
the participants to clearly see (with no inversion) the 25 cm vertical
trajectory of the falling balls as well as their right hand.
Fig. 1. Adapted from Johansson (1998), Desmurget and Grafton (2000) and Bastian (2008). The motor control needs to continuously adapt to
changes by comparing differences between the brain’s predicted outcome of a certain behavior (“What I want to do”) and the real produced outcome
(“What is done”). This discrepancy signal triggers a sensory–motor error that allows updating feedforward motor command or producing on line-motor
corrections. In the literature, it is assumed that the cerebellum is a key structure to ensure such processes.
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144136

During the rest period, participants were instructed to remain
quiet and motionless and to keep their eyes open without thinking
of anything in particular. Participants were observed during this
period to check that no movement was performed. During the ball
catching period, participants were simply instructed to catch falling
balls with their right hand. For each recording session, 10 scans at
rest were followed by 10 scans during which the catching task was
performed and this alternation of rest and catching periods was
repeated four times in one session, for a total of 80 scans for each
experimental condition (4⫻10 scans at rest and 4⫻10 scans of
ball catching). Each subject participated in four different sessions
in which the condition of the repetitive falling balls differed. In each
experimental condition, 10 balls fell during the 10 scans. In the first
experimental condition (light ball condition), participants were re-
quired to catch light balls (6.5 cm in diameter, weighing 30 g, and
black). In the second experimental condition (heavy ball condi-
tion), participants were required to catch heavy balls (6.5 cm in
diameter, weighing 300 g, and black in color). In the third condition
(random condition), five light balls and five heavy balls were
randomly dropped. Note that the light and heavy balls looked the
same and differed only in their mass, so that participants could not
know if the falling ball was heavy or light before catching it. In
addition, to determine whether cerebellar networks involved in the
updating of sensory–motor errors are dependent on visual cues,
we blindfolded our subjects in a final condition without prediction
of the ball weight. In this condition (blind condition), participants
had their eyes blindfolded and either light or heavy balls were
randomly dropped. Consequently, the subjects were not able to
anticipate the time to contact of the falling ball with their hand. The
orders of the session were totally randomized from one subject to
another. Before each block a few practice trials were performed to
recall the balls weight in advance. In a similar paradigm, Lac-
quaniti and Maioli (1989b) observed no adaptation on electromyo-
graphic (EMG) traces in a condition with vision and a single trial
adaptation was observed in no-vision condition. As such 10 trials
seem sufficient to make the condition with constant ball weight
predictable.
The task was very simple. The balls were released by the
experimenter from a 25 cm height at the vertical of the hand.
Subjects grasp the ball with their fingers when receiving it with no
displacement of the arm. The movement was authorized one
degree of freedom at the level of the wrist (horizontal rotation) but
many more at the level of the hand. As such subjects performed
the task very easily in all the conditions and always succeeded to
catch it with no particular noticeable difficulty.
Among the potential limitations of this paradigm, one may
suggest that some adaptation occurred on a trial by trial basis
within the constant weight conditions. This effect illustrates a
decrease of the sensory–motor error and allows highlighting even
more the cerebellar networks when contrasting the random and
no-vision conditions with the light and heavy conditions.
One may also suggest that feedback (e.g. a different manip-
ulation of the ball due to an inaccurate prediction of the motor
perturbation) and/or proactive strategies (e.g. subjects adopt a
stiffness strategy to cope with the variability of the ball weight) may
be different between conditions. However, even if true, the sen-
sory–motor errors would increase in these cases. And contrasting
the random and no-vision conditions with the light and heavy
conditions would highlight even more the cerebellar networks we
sought to observe.
In the same vein, the randomness may be insufficient (only
two balls were used) and there may be also a possibility that
subjects adapted a plan to deal with the uncertainty at the begin-
ning of the first random block and stick to that plan (for instance a
more passive reaction, with less anticipation movements). How-
ever, in both cases, it would play against our hypothesis and
would limit the size and detection of the sensory–motor errors and
the associated contrasts between the random and fixed condi-
tions. Additionally, these a priori changes in the motor strategies
would exhibit different levels of activation in M1. Indeed, this is
sustained by Dai et al. (2001) who demonstrated both in an
isometric and a dynamic task that the level of activation in M1 was
proportional to the force produced. Because the weights of the
balls were very different in our task (30 vs. 300 g), changes in
motor strategies would be observable. For instance, this is typi-
cally illustrated as a classical effect in some fMRI studies in which
parasite activations can be observed in the ipsilateral motor area
when the urgency button is slightly hold in the non working hand.
In other words, the weights of the balls were very different (30
g vs. 300 g) in order to determine clearly using functional imaging
signals whether subjects adopted a conscious motor strategy
before catching the balls in the conditions without predictions. We
hypothesized that if subjects had averaged their motor strategy
before catching the balls we would have clearly observed different
level of activation in motor areas between the conditions with and
without prediction. In addition, they were also asked to perform the
catching task naturally with a relaxed hand.
After debriefing, subjects were asked whether they were able
to predict the ball weight in the random and blind conditions. All
subjects confirmed that they were unable which suggested that
sensory–motor errors are greater in these two conditions. They
also confirmed that they did not use different strategies between
all the blocks and performed the task without conscious motor
strategies between all blocks but with a relaxed hand.
In the present task, error signals were very difficult to measure
efficiently for several reasons. Indeed, the movement was very
limited, very usual and natural. Due to its nature, error signals are
probably minored in this type of task. As such, it limits the capa-
bilities to detect and record potential biases during the sensory–
motor transformations. In addition, the hand offers many possibil-
ities (as a system with many degrees of freedom) to compensate
for the errors of catching and there are probably too few trials to
observe an error when averaging the data. This latter aspect limits
the possibilities to systematically measure an error. However, a
reasonable assumption would be that in the conditions with con-
stant weight (light or heavy) the probability to make an error of ball
weight is inferior to that in the random or blind conditions. As such
behavioral differences may be very subtle. Any supposed similar-
ities between these two sets of behavioral conditions would in fact
play against our hypothesis and diminish the contrast observed
between these two sets of conditions.
Our method is a constrained choice to try to exhibit networks
associated with processing of sensory–motor errors in complex
(several degrees of freedom) and very usual movements in which
errors may be difficult to observe behaviorally and with no exper-
imental modification of the sensory to motor mapping. Our method
thus contrasts with most common paradigms used in this case
(e.g. saccadic adaptation, prismatic adaptation, force field adap-
tation, etc).
MRI acquisition
Whole-brain fMRI was performed using EPI ona3TMagnetom
Trio system (Siemens AG, Munich, Germany), equipped with a
standard head coil configuration. The imaging volume was ori-
ented parallel to the bicommissural (AC–PC) plane. First a T1-
weighted high-resolution three-dimensional volume (repetition
time⫽1700 ms, echo time⫽2.93 ms, flip angle⫽90°; 144 adjacent
axial slices, 1.09 mm thickness; in-plane voxel size⫽1⫻1⫻1
mm
3
) was acquired. Second, functional volumes composed of
43-mm adjacent, interlaced horizontal slices were acquired using
a gradient Echoplanar T2*-weighted EPI sequence (repetition
time⫽3050 ms, echo time⫽45 ms, flip angle⫽90°, matrix
sizes⫽64⫻64⫻40, voxel sizes⫽3⫻3⫻3mm
3
). Each participant
performed four consecutive block fMRI sessions devoted to each
experimental condition. In each functional session, 80 scans were
acquired (i.e. a total of 320 scans per participant). The averaged
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144 137

inter-trial interval was 3 s. The total duration of each functional
scan was 4’36’’.
MRI data processing
Data were analyzed using the general linear model for block
design as implemented in SPM5 (Wellcome Department of Imag-
ing Neuroscience, London, UK). Individual scans were time cor-
rected; T1-weighted anatomical volume was co-registered to
mean images created by the realignment procedure and was
normalized to the MNI space using an affine registration which
was followed by estimating nonlinear deformations, whereby the
deformations was defined by a linear combination of three dimen-
sional discrete cosine transform (DCT) basis functions, as men-
tioned by Ashburner and Friston (1999). The anatomical normal-
ization parameters were subsequently used for the normalization
of functional volumes. Finally, each functional volume was
smoothed by an 8-mm FWHM (Full Width at Half Maximum)
Gaussian kernel. Time series for each voxel were high-pass fil-
tered (1/128 Hz cutoff) to remove low frequency noise and signal
drift.
After pre-processing, statistical analysis was first performed
on functional images for each participant and each session indi-
vidually. For each participant, the ball catching and rest periods in
the four conditions (light ball, heavy ball, random and blind) were
modeled as eight regressors convolved with a canonical hemody-
namic response function (HRF). Movement parameters derived
from realignment corrections for each session were also entered
in the design matrix as additional regressors of no interest. The
general linear model was used to generate parameter estimates of
activity at each voxel, for each condition, and each participant.
Statistical parametric maps were generated from linear contrasts
between the HRF parameter estimates for the different experi-
mental conditions.
At the individual level, we first assessed the whole network of
cerebral areas involved in the processing of each ball catching
condition by contrasting the ball catching blocks with the rest
blocks in each session. We then highlighted brain correlates as-
sociated with the processing of sensory–motor errors by contrast-
ing [two random conditions⬎(light ball⫹heavy ball)] conditions,
and [two random blind⬎(light ball⫹heavy ball)] conditions on the
other hand. We next performed a group analysis and applied
sample t-tests to all contrasts. Clusters of activated voxels were
then identified, based on the intensity of the individual response
(P⬍0.05, FWE corrected for multiple comparisons, T⬎7.49 for
contrasts calculated relative to the rest period and P⬍0.001 un-
corrected, T⬎3.73 for contrasts between conditions of interest).
An extended threshold of 20 voxels was determined empirically
and then used for all contrasts. Moreover, the regions highlighted
in both contrasts of interest were defined as regions of interest
(ROI) in order to confirm that mean ROI parameter estimates were
significantly different in light ball compared to random and blind
conditions separately, and in heavy ball compared to random and
blind conditions separately. In this way, mean ROI parameter
estimates were compared between the studied conditions using
t-tests. Brain regions were reported according to the stereotaxic
atlas of Talairach and Tournoux (1988).
RESULTS
In order to determine whether the cerebellum was involved
in the processing of sensory–motor errors, we first identi-
fied all the cerebral networks activated in the catching task
in each experimental condition (light ball, heavy ball, ran-
dom and blind conditions). Secondly, to verify the involve-
ment of the cerebellum in sensory–motor error, we con-
trasted images obtained in the conditions in which predic-
tion was impossible (random and blind conditions) with
those in which prediction of the ball weight was possible
(light ball and heavy ball conditions). Such contrasts al-
lowed us to isolate feedback processes from feedforward
processes linked to the prediction of the upcoming me-
chanical perturbation.
Motor tasks contrasted with rest periods revealed
similar networks in the right posterior cerebellum
and in the left primary motor cortex in all conditions
The first step was to identify the activations corresponding
to the motor task in the four experimental conditions
(namely light ball, heavy ball, random, and blind conditions
vs. rest, P⬍0.05 corrected). Similar neural networks were
recruited in all conditions. The two largest clusters of acti-
vation were found in the right posterior cerebellum (lobule
V, Schmahmann et al., 1999) and in the left primary motor
cortex extending to the left primary somatosensory cortex
(Brodman area (BA) 3, 4) (see Fig. 2). However, in the
random and the blind conditions, additional neural net-
works including the supplementary motor area (SMA), the
premotor cortex, and the left posterior cerebellum were
also activated. Finally, significant additional activations of
the insula, the superior temporal and supramarginal gyrus,
and the thalamus were also noted in the blind condition
only (see Table 1).
Activation in the right and left cerebellum increased
with task uncertainty
The second step was to determine the neural network
involved in the processing of sensory–motor errors by
contrasting [two random⬎(light ball⫹heavy ball)] condi-
tions, and [two blind⬎(light ball⫹heavy ball)] conditions
(P⬍0.001 uncorrected). In these contrasts, the same num-
ber of identical stimuli on both sides of the subtraction was
maintained. For the sake of clarity, they are named “ran-
dom” and “blind” contrasts respectively in the following.
The significant clusters which were found were thus not
due to a different level of tactile and proprioceptive feed-
back but to the task uncertainty. The results identified the
right and left cerebellum and the right thalamus in both
contrasts. More precisely, left cerebellum (lobule VI) and
right thalamus networks shared some overlap for the ran-
dom and the blind contrasts.
In the right cerebellum, activation networks were dif-
ferently centered. The right posterior cerebellum (24x,
⫺65y, ⫺24z, lobule VI) was indeed significantly activated
in the random condition whereas the anterior portion of the
right cerebellum (24x, ⫺39y, ⫺23z, lobule IV) was acti-
vated when vision was removed (blind condition; see Figs.
3and 4). However, no significant activation was observed
when the blind and random conditions were contrasted
together. Note that in the random condition, 5/33 of the
significant voxels were observed in the anterior portion of
the right cerebellum.
Additional and specific networks were noted for each
contrast (see Table 2). In the random contrast, the asso-
ciative visual cortex and the left supramarginal gyrus were
highlighted. The blind contrast revealed the functional con-
tribution of the right frontal gyrus, the anterior prefrontal
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144138

cortex in both hemispheres and the left primary somato-
sensory cortex (see Table 2). Note that the inverse con-
trasts did not show any significant activation.
Finally, all the regions highlighted by these two con-
trasts were defined as ROI. Mean ROI parameter esti-
mates were extracted from the ROI clusters and the values
were submitted to t-tests in order to compare the light ball
and the heavy ball conditions separately, with the random
and the blind conditions. At this point, it is important to
underline that ROI analyses serve to verify whether ob-
served networks were activated independently of the ball
weight. Results revealed that all ROI activities defined in
Table 2 increased significantly in the random in compari-
son to the light (All t⬍⫺2.79, P⬍0.014) and heavy ball
conditions (All t⬍⫺3.41, P⬍0.004). Similarly, ROI activi-
ties increased significantly in the blind condition in com-
parison to the light (All t⬍⫺2.32, P⬍0.03) and heavy ball
conditions (All t⬍⫺2.24, P⬍0.04). In addition, no signifi-
cant difference was found when light ball and heavy ball
conditions were compared for both contrasts.
DISCUSSION
Our aim here was to identify the main cerebellar structures
involved in the processing of sensory–motor errors when
catching a falling object. By contrasting functional imaging
signals obtained in conditions in which it was possible and
impossible to predict the weight of the ball, we aimed to
highlight sensory–motor error processing which we ex-
pected to be more marked in the conditions without pre-
diction (less accurate feedforward process or more impor-
tant feedback corrections, i.e. more important sensory–
motor error) with respect to conditions with prediction
(more accurate feedforward process or less important
feedback corrections, i.e. less important sensory–motor
error).
In order to determine whether networks activated in the
catching task were similar to those generally reported in
the literature, we first assessed the whole network of ce-
rebral areas involved in the processing of each ball catch-
ing condition by contrasting the ball catching blocks with
the rest blocks obtained at each session. We found com-
mon activations in left motor and sensory areas and in the
right cerebellum for all conditions. Additional activations
were observed in the random and blind conditions. Impor-
tantly, when subjects were blindfolded, the time to contact
could not be estimated and we may have expected some
differences between the activations observed in the ran-
dom vs. blind condition. Interestingly, however, we also
observed similar activations within the cerebellum between
the two conditions suggesting that the sensory–motor pro-
cessing shared some common networks independently of
visual cues. These common activations were similar to
those generally obtained for similar tasks in the literature
(Field and Wann, 2005; Senot et al., 2008; Bédard and
Sanes, 2009).
When we contrasted conditions of greater uncertainty
about the ball weight with conditions of no uncertainty, our
main results clearly demonstrated that when the ball is
caught with the right hand without possible prediction
about its weight, two networks in both the right and left
cerebellum are highlighted. Interestingly, when we per-
formed these contrasts, we did not notice any activation at
the level of the primary motor cortices. Nor did we observe
different levels of activation in the primary sensory cortex
(post central gyrus) when we contrasted the random con-
dition (with sight) with the light and heavy ball conditions. In
contrast, this level was different when we contrasted the
blind condition with the light and heavy ball conditions,
suggesting that the manipulation of the ball induced slightly
different reafferences (Table 2).
The subtraction of the baseline activity from the activity
in the blind condition extracted activity related to the esti-
mation of the weight and the time to contact, while the
Fig. 2. Significant activations in the four experimental conditions (light
ball, heavy ball, random and blind conditions vs. rest, P⬍0.05 cor-
rected for multiple comparisons) when the motor task period was
contrasted with the rest period. The two largest clusters of activation
were found in the right posterior cerebellum and in the left primary
motor cortex in each condition.
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144 139

subtraction of the baseline activity from the activity in the
random condition extracted activity related to the estima-
tion of the weight only. This may explain greater activations
in the sensory cortex for the blind condition which suggests
greater manipulation of the ball when caught with the hand.
Interestingly, however, despite these disparities there
seem to be some overlap within the cerebellum between
activations in the random vs. blind conditions.
No significant activation was observed when the blind
and random conditions were contrasted together. Clusters
of the right cerebellum networks were aligned on the same
antero–posterior axis, so in the same horizontal plane.
Activations were respectively centered in a more posterior
(24x, ⫺65y, ⫺24z) and in a more anterior portion (24x,
⫺39y, ⫺23z) of the right cerebellum (see Fig. 5).
Using PET, Desmurget et al. (1998) demonstrated that
in humans, the medioposterior cerebellum was involved in
the control of saccadic adaptation. In their study, they
reported a network whose coordinates (1.5x, ⫺62y, ⫺18z)
might involve some overlap with the more posterior one we
obtained (note that their y- and z-coordinates were quite
similar to ours). In their study, they demonstrated in-
creased activity in the medioposterior cerebellum after
adaptation compared to random target presentation. One
may suggest that this result is opposite to our; however, in
saccadic adaptation a systematic target jump is made
during the saccadic blind period. In consequence, a sys-
tematic bias and sensory–motor error is introduced uncon-
sciously on a trial by trial basis. In our experiment, sen-
sory–motor errors are conscious and variable between
trials but more important in the condition without prediction.
As a whole, this comparison suggests that at this level, the
cerebellum may be involved in the processing of sensory–
motor errors for different effectors and in updating the
associated feedforward commands whatever the error is
systematic and unconscious or variable and conscious.
Alternatively, in the study of Desmurget et al. (1999) a new
internal model was probably built, following a constant
perturbation. In this case, cerebellar activations may indi-
cate a new internal model that could not be consolidated
during the random perturbation. In our experiment, the
difference in cerebellar activations may rather be propor-
tional to the error and not necessarily to the modification of
the internal model. As the error signal is computed as the
difference between what is predicted and real sensory
feedbacks (see Fig. 1), this error may both trigger on-line
motor corrections and/or update feedforward commands.
As such we could not disambiguate whether the presented
errors should lead to changes of internal model or to
initiation of on-line corrections. However, these responses
seem to originate from the same source and identical
sensory–motor error processing. In this vein, Diedrichsen
et al. (2005) demonstrated that execution errors (assigned
to the movement) were systematically associated with
feedforward correction on the next trial in a pointing task.
This is the kind of error that may occur in our experimental
design. In accordance with our result, they also observed,
that errors were associated with cerebellar activation in
lobule IV and V whatever the perturbation (i.e. mechanical
perturbation, visual feedback perturbation, or target dis-
placement) and even if there was no behavioral adaptation
during the task, but only feedback on-line corrections. In
addition, Tseng et al. (2007) demonstrated that adaptation
to visuo–motor perturbation depends on the cerebellum
Table 1. Cerebral regions specifically activated during ball catching in the four experimental conditions (P⬍0.05 corrected for multiple comparisons). For
each cluster, the region showing the maximum T-value is listed first, followed by the others belonging to the cluster [between brackets]. The cerebellum
lobules are reported according to Schmahmann et al. (1999). The Talairach coordinates (x, y, z), the corresponding Brodman area (BA), the laterality of the
hemisphere (H; L, left hemisphere; R, right hemisphere) and the number of voxels in the cluster (k) are reported. Voxel size: 3⫻3⫻3mm
3
Contrast Region H BA kxyzT
Light ball⬎Rest Right posterior cerebellum (lobule V) R 111 21 ⫺53 ⫺18 14.15
Primary sensory or motor cortex L 2/3 196 ⫺42 ⫺29 51 11.26
[Primary motor cortex] L 4 ⫺45 ⫺32 62 11.14
Heavy ball⬎Rest Right posterior cerebellum (lobule V) R 49 21 ⫺51 ⫺27 9.46
Primary sensory–motor cortex L 2/3 55 ⫺36 ⫺30 48 9.27
[Primary motor cortex] L 4 ⫺36 ⫺38 40 9.11
Random⬎Rest Primary sensory cortex L 3 683 ⫺39 ⫺27 48 14.87
[Primary motor cortex] L 4 ⫺39 ⫺23 62 14.43
Right posterior cerebellum (lobule V) R 427 21 ⫺50 ⫺18 14.70
Primary sensory–motor cortex R 1/2 159 65 ⫺30 40 11.94
Occipito–temporal cortex L 19/39 46 ⫺50 ⫺64 9 10.60
Premotor cortex and SMA L 6 23 ⫺42 ⫺2 11 10.36
Left posterior cerebellum (lobule VI) L 42 ⫺30 ⫺56 ⫺17 10.12
Blind⬎Rest Primary sensory cortex L 3/4/6 855 ⫺56 ⫺19 20 18.47
[Primary motor cortex] L ⫺39 ⫺38 60 17.72
[SMA] L ⫺39 ⫺20 62 14.85
Right posterior cerebellum (lobule V) R 261 21 ⫺50 ⫺18 13.56
Insula L 85 ⫺42 ⫺5 14 13.12
Superior temporal gyrus R 21/22 116 56 ⫺52 3 12.63
Supramarginal gyrus R 40 110 59 ⫺39 32 11.31
Left posterior cerebellum (lobule VI) L 72 ⫺30 ⫺53 ⫺20 10.27
Thalamus L 47 ⫺18 ⫺20 15 9.30
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144140

and is driven by the mismatch between expected and
actual sensory feedbacks independently of the occurrence
of on-line corrections. More specifically, for ball catching
tasks and to adapt to ball weight changes, anticipatory
muscles activities must be modified and timed to occur
before the impact (Bennett et al., 1994; Lang and Bastian,
1999). It was shown that cerebellar subjects adapted
slowly or not at all to modifications of the ball weight for
light to heavy or heavy to light balls. Finally, Lang and
Bastian (2001) showed that cerebellar patients remained
slow or unable to adapt to the change of the ball weight
even with on-line information. Altogether it confirmed the
idea that the processing of the sensory–motor error is
computed in the same cerebellar circuits despite different
behavioral effects and types of errors, and independently
of the occurrence of on-line corrections.
Some studies have shown that cerebellar activity spe-
cifically reflects the operations of internal models in the
prediction of dynamic constraints during movement control
(Imamizu et al., 2000; Kawato et al., 2003; Diedrichsen et
al., 2005). At the cellular level, the role of Purkinje cells has
been evoked to explain cerebellar plasticity that occurs
during learning and motor adaptation. More particularly, it
has been suggested that Purkinje cells could modify the
gain of motor commands in many adaptive mechanisms by
comparing signals conveyed by parallel fibers and a teach-
ing signal conveyed by climbing fibers (Ito, 1993, 2001,
2002; Boyden et al., 2004; De Zeeuw and Yeo, 2005). This
process could involve the cerebellum in updating sensory
Fig. 3. Significant activations in the [two random⬎(light ball⫹heavy
ball)] contrast (P⬍0.001, uncorrected). L, left; R, right; post Cer, left
posterior portion of the cerebellum; Ass Vis cx, associative visual
cortex; Supmarg Gyrus, supramarginal gyrus.
Fig. 4. Significant activations in the [two blind⬎(light ball⫹heavy
ball)] contrast (P⬍0.001, uncorrected). L, left; R, right; post Cer, left
posterior portion of the cerebellum; Ant Cer, anterior portion of the
cerebellum; a PFrontal Cx, anterior prefrontal cortex; dlPFrontal cx,
dorsolatéral prefrontal cortex; SM cx, primary sensory–motor
cortex.
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144 141

prediction based on error detection between true reaffer-
ences and the prediction itself (Ramnani, 2006).
Interestingly, in a quite similar experimental design as
our, Schmitz et al. (2005) investigated brain activations in
a task in which subjects had to lift an object whose mass
could be unexpectedly varied. Recognizing the role of
cerebellum in anticipatory control and prediction of sensory
consequences, they hypothesized that if cerebellar activa-
tions were reflecting rapid grip force corrections during
conditions without predictions, strongest activations should
have been observed in this case. However, they did not
observe activations in the cerebellum when the conditions
without and with prediction of the object weight were con-
trasted. This result is in contradiction with our and with
neuroimaging studies that have revealed the presence of
activity in the human cerebellum related to error signals
(Ramnani et al., 2000; Imamizu et al., 2000; Diedrichsen et
al., 2005). Schmitz et al. suggested that depending on
whether the weight was lighter or heavier than expected,
the corrections were not the same and should have prob-
ably not activated the same areas. Because both events
were mixed they could not make the distinction. Another
explanation could occur from the fact that in their experi-
ment the ratio between the heavy and the light object was
not important enough (230 and 830 g, i.e. ⫻3.6 vs. 30 and
300 g, i.e. ⫻10).
We also observed one network in the left cerebellum
that was centered roughly on the same location indepen-
dently of the manipulation of sight (⫺30x, ⫺62y, ⫺22z and
⫺21x, ⫺77y, ⫺26z for the random and blind conditions,
respectively). These activations within the right posterior
cerebellum may involve the flocculo nodular lobe. This
structure, under the influence of vestibular signals, is
known to send motor commands to axial antigravity mus-
cles and extensors (with bilateral projections). These acti-
vations suggest that the effects of the perturbation asso-
ciated with the ball catching task were bilateral and may
also trigger some postural responses in the left side. In this
case, different and more axial feedforward motor com-
mands may be involved, as catching the ball may imply
greater perturbation of the axial and proximal muscular
control. This will be readily accepted if we consider the
idea that the human hand manipulates many objects of
different weights every day from a standing or sitting po-
sition, but that catching an object while lying on one’s back
is much less frequent and may unusually challenge axial
movements. In any case, these left cerebellar activations
also suggest that the left cerebellum could be involved in
the processing of sensory–motor errors that may influence
the updating of feedforward commands.
A possible limitation of our study could be that the
predicted changes in error magnitude would not be asso-
ciated with changes in prediction error, as the internal
model used will be different in the random conditions com-
Table 2. Cerebral regions specifically activated to process the sensory motor errors (P⬍0.001 uncorrected). For each cluster, the region showing the
maximum T-value is listed first, followed by the others belonging to the cluster [between brackets]. The cerebellum lobules are reported according to
Schmahmann et al. (1999). The Talairach coordinates (x, y, z), the corresponding Brodman area (BA), the laterality of the hemisphere (H; L, left
hemisphere; R, right hemisphere) and the number of voxels in the cluster (k) are reported. Voxel size: 3⫻3⫻3mm
3
Contrast Region H BA kxyzT
[2*random⬎(light
ball⫹heavy ball)]
Right posterior cerebellum (lobule VI) R 33 24 ⫺65 ⫺24 5.46
Occipito–temporal cortex L 19/39 40 ⫺45 ⫺61 9 5.13
Left posterior cerebellum (lobule VI) L 52 ⫺30 ⫺62 ⫺22 4.90
Supramarginal gyrus L 40 25 ⫺48 ⫺32 51 4.63
Thalamus R 23 30 ⫺26 4 4.52
[2*blind⬎(light
ball⫹heavy ball)]
Insula L 817 ⫺33 ⫺20 9 6.93
[Premotor cortex and SMA] L ⫺56 4 22 6.09
Dorsolateral prefrontal cortex R 9 188 50 16 32 6.62
Frontal gyrus R 8 143 9 18 60 6.54
Right anterior cerebellum (lobule IV) R 29 24 ⫺39 ⫺23 6.13
Anterior prefrontal cortex R 10 83 24 53 14 5.98
Anterior prefrontal cortex L 10 47 ⫺42 42 15 5.75
Left posterior cerebellum (lobule VI) L 79 ⫺21 ⫺77 ⫺26 5.49
Thalamus R 120 3 ⫺18 12 5.45
Primary sensory motor cortex L ½ 110 ⫺48 ⫺29 57 5.33
Fig. 5. Cerebellar networks significantly activated in [two
random⬎(light ball⫹heavy ball)] (in yellow) and in [two blind⬎(light
ball⫹heavy ball)] (in red). Voxels common to both contrasts appear in
orange.
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144142

pared to the blocked conditions. However, this interpreta-
tion is limited for at least two reasons that are linked
together. First, it suggests that different controllers are
involved each time the dynamics is modified despite the
movement kinematics remains roughly the same. This
raised an important problem of storage, as many control-
lers would be differentiated and stored in the brain and
especially in the cerebellum. Second and related to the
previous remark, this interpretation is not compatible with
the role that plays the cerebellum in the adaptation to the
environment dynamics. Indeed, the cerebellum rather
adapts the same controller to various dynamical con-
straints (Wolpert et al., 1998 for a review).
Interestingly, when subjects were blindfolded ([two
blind⬎(light ball⫹heavy ball)] contrast), we also observed
right activations in the thalamus and the prefrontal cortex
when we contrasted the conditions without and with pre-
diction. It is known that the prefrontal cortex sends projec-
tions to the cerebellum via the cortico–olivo–cerebellar
pathways and the inferior olive and reciprocally that the
cerebellum sends projections back to the prefrontal cortex
via the cerebellar nuclei and the thalamus. This is rather in
contradiction with the traditional view that considers the
cerebellum and its canonical circuitry as a strictly motor
structure providing error correction during motor adapta-
tion. However, Ramnani (2006) has suggested that the
cerebellum may compute some predictive cognitive con-
sequences of cognitive operations, and that the error be-
tween predicted and actual cognitive outcomes is used to
refine future predictions. In our task, the prefrontal cortex
may be involved in the cognitive operation that assigns
probabilities for each of the ball weights. This structure
may thus influence motor error signals according to the
conscious prediction of the ball weight. This network was
mainly exhibited in the blind contrast maybe because sub-
jects could not estimate visually the time to contact and
might rely more importantly to cognitive predictions (rather
than vision) in relation to sensory predictions. Note that
small activations in the thalamus were also noticeable for
the random contrast that might not reach the prefrontal
cortex. Nevertheless, this motor and cognitive network that
seems to link the cerebellum, the thalamus and the pre-
frontal cortex was observed only in the right hemisphere,
suggesting that adaptations to perturbations were more
challenging for the left or axial part of the body (via the
flocculo–nodular lobe) although the ball was caught with
the right hand. In the blind condition, we also observed left
activations within the insula, the premotor and supplemen-
tary motor areas. Premotor and supplementary areas are
known to be involved when the motor sequence becomes
more difficult (Roland et al., 1980). These structures may
be more engaged in the task when vision is suppressed.
Similarly, the insula has been shown to store an internal
model of gravity which is involved in the representation of
visual gravitational motion (Indovina et al., 2005; Maffei et
al., 2010). The fact that we observed greater activations
when sight was suppressed may suggest that the repre-
sentation of gravity is challenged in this situation.
Finally, when we contrasted the random condition (with
sight) with the light and heavy ball conditions, we observed
activations in the left occipito–temporal cortex. This area is
known to be involved in visual attention (Barnes et al.,
2000). It may be that when ball weight varied unexpect-
edly, subjects engaged more attentional resources to try to
optimize their catching.
In conclusion, in a situation of repetitive catching of a
falling ball when the possibility of prediction about ball
weight is manipulated, our results showed that both the
right and left cerebellum is engaged to process sensory–
motor errors and to update feedforward motor commands,
perhaps on a trial by trial basis. In addition, when subjects
were blindfolded, we observed a more anterior network in
the right cerebellum and we identified a cerebellar–
thalamo–prefrontral network that may be involved in cog-
nitive prediction (rather than sensory prediction) about the
ball weight.
Acknowledgments—This work was supported by the Conseil Re-
gional de Bourgogne. We thank the CENTRE Hospitalier Univer-
sitaire de Dijon to have allowed us to realize the data acquisitions.
We thank the two anonymous reviewers for their stimulating com-
ments about this work.
REFERENCES
Ashburner J, Friston KJ (1999) Nonlinear spatial normalization using
basis functions. Hum Brain Mapp 7(4):254–266.
Babin-Ratté S, Sirigu A, Gilles M, Wing A (1999) Impaired anticipatory
finger grip-force adjustments in a case of cerebellar degeneration.
Exp Brain Res 128:81–85.
Barnes J, Howard RJ, Senior C, Brammer M, Bullmore ET, Simmons
A, Woodruff P, David AS (2000) Cortical activity during rotational
and linear transformations. Neuropsychologia 38(8):1148–1156.
Bastian AJ (2006) Learning to predict the future: the cerebellum
adapts feedforward movement control. Curr Opin Neurobiol
16(6):645–649.
Bastian AJ (2008) Understanding sensorimotor adaptation and learn-
ing for rehabilitation. Curr Opin Neurol 21(6):628–633.
Bédard P, Sanes JN (2009) Gaze and hand position effects on finger-
movement-related human brain activation. J Neurophysiol 101(2):
834–842.
Bennett DJ, Gorassini M, Prochazka A (1994) Catching a ball; contri-
butions of intrinsic muscle stiffness, reflexes, and higher order
responses. Can J Physiol Pharmacol 72:525–534.
Bhushan N, Shadmehr R (1999) Computational nature of human
adaptive control during learning of reaching movements in force
fields. Biol Cybern 81:39–60.
Blakemore SJ, Frith CD, Wolpert DM (2001) The cerebellum is in-
volved in predicting the sensory consequences of action. Neurore-
port 12:1879–1884.
Blomfield S, Marr D (1970) How the cerebellum may be used. Nature
227(5264):1224–1228.
Bock O (1992) Adaptation of aimed arm movements to sensorimotor
discordance: evidence for direction-independent gain control. Be-
hav Brain Res 51:41–50.
Boyden ES, Katoh A, Raymond JL (2004) Cerebellum-dependent
learning: the role of multiple plasticity mechanisms. Annu Rev
Neurosci 27:581–609.
Dai TH, Liu JZ, Sahgal V, Brown RW, Yue GH (2001) Relationship
between muscle output and functional MRI-measured brain acti-
vation. Exp Brain Res 140(3):290–300.
De Zeeuw CI, Yeo CH (2005) Time and tide in cerebellar memory
formation. Curr Opin Neurobiol 15:667–674.
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144 143

Desmurget M, Epstein CM, Turner RS, Prablanc C, Alexander GE,
Grafton ST (1999) Role of the posterior parietal cortex in upda-
ting reaching movements to a visual target. Nat Neurosci
2(6):563–567.
Desmurget M, Grafton S (2000) Forward modeling allows feedback
control for fast reaching movements. Trends Cogn Sci 4(11):423–
431.
Desmurget M, Pélisson D, Urquizar C, Prablanc C, Alexander GE,
Grafton ST (1998) Functional anatomy of saccadic adaptations in
humans. Nat Neurosci 1(8):743.
Dharani NE (2005) The role of vestibular system and the cerebellum in
adapting to gravitoinertial, spatial orientation and postural chal-
lenges of REM sleep. Med Hypotheses 65(1):83–89.
Diedrichsen J, Hashambhoy Y, Rane T, Shadmehr R (2005) Neural
correlates of reach errors. J Neurosci 25:9919–9931.
Donchin O, Francis JT, Shadmehr R (2003) Quantifying generalization
from trial-by-trial behavior of adaptive systems that learn with basis
functions: theory and experiments in human motor control. J Neu-
rosci 23:9032–9045.
Field DT, Wann JP (2005) Perceiving time to collision activates the
sensorimotor cortex. Curr Biol 15(5):453–458.
Imamizu H, Miyauchi S, Tamada T, Sasaki Y, Takino R, Pütz B,
Yoshioka T, Kawato M (2000) Human cerebellar activity reflecting
an acquired internal model of a new tool. Nature 403(6766):
192–195.
Indovina I, Maffei V, Bosco G, Zago M, Macaluso E, Lacquaniti F
(2005) Representation of visual gravitational motion in the human
vestibular cortex. Science 308(5720):416–419.
Ito M (1993) Synaptic plasticity in the cerebellar cortex and its role in
motor learning. Can J Neurol Sci 20:S70–S74.
Ito M (2000) Mechanisms of motor learning in the cerebellum. Brain
Res 886(1–2):237–245.
Ito M (2001) Cerebellar long-term depression: characterization, signal
transduction, and functional roles. Physiol Rev 81:1143–1195.
Ito M (2002) Historical review of the significance of the cerebellum and
the role of Purkinje cells in motor learning. Ann N Y Acad Sci
978:273–288.
Johansson RS (1998) Sensory input and control of grip. Novartis
Found Symp 218:45–59.
Johansson RS, Cole KJ (1992) Sensory-motor coordination during
grasping and manipulative actions. Curr Opin Neurobiol 2:815–
823.
Kawato M, Furukawa K, Suzuki R (1987) A hierarchical neural-network
model for control and learning of voluntary movement. Biol Cybern
57(3):169–185.
Kawato M, Wolpert D (1998) Internal models for motor control. No-
vartis Found Symp 218:291–307.
Kawato M, Kuroda T, Imamizu H, Nakano E, Miyauchi S, Yoshioka T
(2003) Internal forward models in the cerebellum: fMRI study on
grip force and load force coupling.. Prog Brain Res 142:171–188.
Krakauer JW, Pine ZM, Ghilardi MF, Ghez C (2000) Learning of
visuomotor transformations for vectorial planning of reaching tra-
jectories. J Neurosci 20(23):8916–8924.
Lacquaniti F, Maioli C (1989a) The role of preparation in tuning antic-
ipatory and reflex responses during catching. J Neurosci 9:134–
148.
Lacquaniti F, Maioli C (1989b) Adaptation to suppression of visual
information during catching. J Neurosci 9(1):149–159.
Lang CE, Bastian AJ (1999) Cerebellar subjects show impaired adap-
tation of anticipatory EMG during catching. J Neurophysiol 82(5):
2108–2119.
Lang CE, Bastian AJ (2001) Additional somatosensory information
does not improve cerebellar adaptation during catching. Clin Neu-
rophysiol 112(5):895–907.
Maffei V, Macaluso E, Indovina I, Orban G, Lacquaniti F (2010) Pro-
cessing of targets in smooth or apparent motion along the vertical
in the human brain: an fMRI study. J Neurophysiol 103(1):360–
370.
Marr D (1969) A theory of cerebellar cortex. J Physiol 202(2):437–470.
Martin TA, Keating JG, Goodkin HP, Bastian AJ, Thach WT (1996)
Throwing while looking through prisms. I. Focal olivocerebellar
lesions impair adaptation. Brain 119:1183–1198.
Morton SM, Bastian AJ (2004) Prism adaptation during walking gen-
eralizes to reaching and requires the cerebellum. J Neurophysiol
92:2497–2509.
Morton SM, Bastian AJ (2006) Cerebellar contributions to locomotor
adaptations during splitbelt treadmill walking. J Neurosci 26:
9107–9116.
Nowak DA, Hermsdörfer J, Marquardt C, Fuchs HH (2002) Grip and
load force coupling during discrete vertical arm movements with a
grasped object in cerebellar atrophy. Exp Brain Res 145:28–39.
Nowak DA, Topka H, Timmann D, Boecker H, Hermsdörfer J (2007)
The role of the cerebellum for predictive control of grasping. Cer-
ebellum 6(1):7–17.
Oldfield RC (1971) The assessment and analysis of handedness: the
Edinburgh inventory. Neuropsychologia 9:97–113.
Pasalar S, Roitman AV, Durfee WK, Ebner TJ (2006) Force field
effects on cerebellar Purkinje cell discharge with implications for
internal models. Nat Neurosci 9:1404–1411.
Ramnani N (2006) The primate cortico-cerebellar system: anatomy
and function. Nat Rev Neurosci 7(7):511–522.
Ramnani N, Toni I, Josephs O, Ashburner J, Passingham RE (2000)
Learning- and expectation-related changes in the human brain
during motor learning. J Neurophysiol 84:3026–3035.
Roland PE, Larsen B, Lassen NA, Skinhøj E (1980) Supplementary
motor area and other cortical areas in organization of voluntary
movements in man. J Neurophysiol 43(1):118–136.
Schmahmann JD, Doyon J, McDonald D, Holmes C, Lavoie K, Hurwitz
AS, Kabani N, Toga A, Evans A, Petrides M (1999) Three-dimen-
sional MRI atlas of the human cerebellum in proportional stereo-
taxic space. Neuroimage 10(3 Pt 1):233–260.
Schmitz C, Jenmalm P, Ehrsson HH, Forssberg H (2005) Brain activity
during predictable and unpredictable weight changes when lifting
objects. J Neurophysiol 93(3):1498–1509.
Senot P, Baillet S, Renault B, Berthoz A (2008) Cortical dynamics of
anticipatory mechanisms in interception: a neuromagnetic study. J
Cogn Neurosci 20(10):1827–1838.
Shadmehr R, Mussa-Ivaldi FA (1994) Adaptive representation of dy-
namics during learning of a motor task. J Neurosci 14:3208–3224.
Talairach J, Tournoux P (1988) Co-planar stereotaxic atlas of the
human brain. 3-dimensional proportional system: an approach to
cerebral imaging. New York: Thieme.
Tseng YW, Diedrichsen J, Krakauer JW, Shadmehr R, Bastian AJ
(2007) Sensory prediction errors drive cerebellum-dependent ad-
aptation of reaching. J Neurophysiol 98(1):54–62.
Wei K, Körding K (2010) Uncertainty of feedback and state estimation
determines the speed of motor adaptation. Front Comput Neurosci
4:11.
Wolpert DM, Miall RC (1996) Forward models for physiological motor
control. Neural Netw 9(8):1265–1279.
Wolpert DM, Kawato M (1998) Multiple paired forward and inverse
models for motor control. Neural Netw 11(7–8):1317–1329.
Wolpert DM, Miall RC, Kawato M (1998) Internal models in the cere-
bellum. Trends Cogn Sci 2:338–347.
(Accepted 9 June 2011)
(Available online 28 June 2011)
L. Fautrelle et al. / Neuroscience 190 (2011) 135–144144