NCM 22nd Annual Meeting NCM Posters Authors, Titles, Affiliations & Abstracts

NCM 22nd Annual Meeting
HILTON MOLINO STUCKY | VENICE, ITALY
NCM Posters
Authors, Titles, Affiliations
& Abstracts
Sessions 1 & 2
Satellite Meeting
April 22 – 23, 2012
22nd Annual Meeting
April 23 – 28, 2012
Society for the Neural Control of Movement
| www.ncm-society.org
8:00
8:15
8:30
8:45
9:00
9:15
9:30
9:45
10:00
10:15
10:30
10:45
11:00
11:15
11:30
11:45
12:00
12:15
12:30
12:45
13:00
13:15
13:30
13:45
14:00
14:15
14:30
14:45
15:00
15:15
15:30
15:45
16:00
16:15
16:30
16:45
17:00
17:15
17:30
17:45
18:00
18:15
18:30
18:45
19:00
19:15
19:30
19:45
20:00
20:15
20:30
20:45
21:00
21:15
21:30
21:45
22:00
22:15
22:30
(19:00 - 21:00)
Professor Jose Maria
Delgado Garcia
in honor of
Satellite Dinner
(17:00 - 19:00)
Satellite
Registration
Arrivals, Free Time, Excursions
22-Apr
Registration /Information Desk Open
(19:00 - 21:00)
Opening Dinner
(17:00 - 19:00)
Conference Registration
(8:30 - 17:30)
Satellite Meeting
Satellite Registration
23-Apr
Monday
Free Time and/or
Excursions
(15:00 - 17:15)
Session 3
Panel
AZIM
(13:00 - 15:00)
Poster Session
1a and Lunch
(10:45 - 13:00)
Session 2
Panel
SABES
(10:15 - 10:45)
Break
(8:00 - 10:15)
Session 1
Panel
SCHIEBER
24-Apr
(16:00 - 17:20)
Session 7
Individual
Presentations
(14:30 - 16:00)
Session 6
Panel
GRIBBLE
(12:45 - 14:30)
Poster Session
1b and Lunch
(10:45 - 12:45)
Session 5
Individual
Presentations
(10:15 - 10:45)
Break
(8:00 - 10:15)
Session 4
Panel
FLASH
25-Apr
Free Time and/or
Excursions
Exhibits on Display
Wednesday
Exhibits on Display
Registration /Information Desk Open
(19:45 - 21:15)
*NCM Special Excursion*
Private Evening of
St. Mark’s Church
FreeTime and/or
Excursions
(12:05 - 13:00)
Business Meeting
(10:45 - 12:05)
Session 9
Individual
Presentations
(10:15 - 10:45)
Break
(8:00 - 10:15)
Session 8
Panel
POZZO
26-Apr
Thursday
Exhibits on Display
Free Time and/or
Excursions
(15:00 - 17:00)
Session 12
Individual
Presentations
(12:45 - 15:00)
Poster Session
2a and Lunch
(10:45 - 12:45)
Session 11
Individual
Presentations
(10:15 - 10:45)
Break
(8:00 - 10:15)
Session 10
Panel
SUMINSKI
27-Apr
Friday
29-Apr
28-Apr
Free Time and/or
Excursions
(14:00 - 15:00)
Session 15
BIZZI
(12:15 - 14:00)
Poster Session
2b and Lunch
(10:45 - 12:15)
Session 14
Panel
BURDET
(10:15 - 10:45)
Break
(8:00 - 10:15)
Session 13
Panel
KRAKAUER
Sunday
Saturday
Exhibits on Display
Tuesday
Exhibits on Display
Sunday
Departures, Free Time, Excursions
Time
2012 At-A-Glance Satellite
and Annual Conference Schedule
Hilton Molino Stucky, Venice, Italy
Posters on Display (Session 2)
Registration /Information Desk Open
Posters on Display (Session 2)
Registration /Information Desk Open
Posters on Display (Session 1)
Registration /Information Desk Open
Posters on Display (Session 1)
Registration /Information Desk Open
Poster Sessions
Full Abstracts
SESSION 1
A - Adaptation & Plasticity in Motor Control
1-A-1
Paradoxical arm stiffness: Humans actively
reduce endpoint arm stiffness in force tasks
Bram Onneweer1, Erwin de Vlugt1, Alfred C Schouten1, Winfred
Mugge1, Carel G.M. Meskers2, Frans C. T. van der Helm1
1
TU Delft, 2Leiden University Medical Center
For humans to efficiently interact with their surroundings, the
central nervous system (CNS) manipulates the stiffness of the
neuromuscular system. Joint stiffness consists of an intrinsic and
reflexive component. Intrinsic stiffness is the stiffness of a
contracted muscle that can be increased with level of muscle
activation. Reflexive stiffness originates from reflexive muscle
activations as a result of the imposed perturbations that are
sensed and fed back to the CNS by the muscle afferents like
muscle spindles and Golgi tendon organs. A recent study on a
single joint found that humans, when asked to give way to
external forces, actively reduce joint stiffness to values below
those found in the relaxed state. Here we assess if humans are
capable of tuning the stiffness over multiple joints and at endpoint
space during 2D (planar) hand tasks. Subjects (n=9, all male,
mean age 34,7) held the handle of a planar manipulandum, in the
horizontal plane in front of their shoulder with the elbow in 90
degrees, which applied unpredictable continuous isotropic forces
to their hand. The subjects performed three tasks with their
dominant arm, 1) force task (FT) where 'hand reaction forces
need to be minimized'; 2) position task (PT), where 'hand
displacement need to be minimized' and 3) relax task (RT) where
subjects were instructed to 'do nothing' and which was used as a
reference condition. Endpoint stiffness was calculated from the
recorded movements and interaction forces at the handle. Joint
stiffness of the shoulder and elbow, as well as the stiffness
contribution of the bi-articular muscles, were derived from the
endpoint stiffness using the Jacobian of the kinematic relations
between joint angles and endpoint coordinates. As expected, both
joint and endpoint stiffness was magnitudes larger in PT than in
RT and FT. Joint and endpoint stiffness was smaller in FT than
RT, although not as pronounced as in the single joint case. With
the change in stiffness, the shape and orientation of the stiffness
ellipse changed, indicative of a change in the contribution of the
muscle groups. Due to the different stiffness directions of the
muscle groups, the shape of the ellipse widens if more, and
narrows if less muscle groups contribute to the endpoint stiffness.
The orientation of ellipse is directed to the shoulder for PT and
rotates clockwise for RT and more for FT. In addition to
orientation, the shape of the ellipse narrows for RT and more for
FT. The reduction of stiffness was more pronounced for the single
joint shoulder and elbow muscles compared with the bi-articular
shoulder-elbow muscles. Muscle activation levels (monitored with
EMG) were higher for FT than RT, indicating that humans actively
reduce joint stiffness. In conclusion, during a force task humans
actively reduce endpoint arm stiffness to levels below the relaxed
state, suggesting the usage of continuous inhibitory feedback
mechanisms.
1-A-2
Internal model reference frames and
generalizability
Max Berniker1, Konrad Kording1
1
Northwestern University
How the central nervous system represents and stores the
information necessary to control our bodies is one of the
central questions in movement science. This information is
usually assumed to reside in an internal model, a
representation of how movements are generated. However,
fundamental questions such as which variables they
encode, and how they generalize to new circumstances, for
example, are still unclear. Depending on the kind of motor
behavior subjects adapt to (e.g. force field, inertial
perturbation, visuomotor rotation) and the type of
generalization examined (e.g. interlimb or intralimb) different
groups find different results. In one of the earliest studies of
this kind (Shadmehr, '94) it was found that subjects were
able to successfully adapt to a force field, and appeared to
use intrinsic (i.e. joint-based) variables to generalize that
knowledge to a new area of their workspace. However,
subsequent studies have found evidence for both intrinsic
and extrinsic (i.e. Cartesian-based) variables in internal
models. Therefore, we reproduced this early study in an
effort to carefully examine how subjects adapt and
generalize new motor behaviors. As in the original study, we
had subjects adapt to either a force field based on the
velocity of their hand (extrinsic) or the velocity of their limb's
joints (intrinsic). During adaptation, subjects made 1,000
reaches in the force field to randomly generated targets,
forming a pseudo-random walk. Subjects were then
examined for their performance when generalizing in a new
area of the workspace. In this generalization phase,
subjects were randomly exposed to either the intrinsic,
extrinsic or null fields. Thus if the subjects made straight
reaches in the intrinsic/extrinsic field, it would imply the
internal model acquired during adaptation used
intrinsic/extrinsic coordinates to adapt to the force field.
Surprisingly our results were distinct from what was
originally reported. We found subjects did not adapt to the
level observed in the previous study and continued to
display "hooked" reaches even after 1,000 trials (indeed, a
second group of subjects that adapted for 1,500 trials
continued to display hooked reaches). Perhaps more
importantly, when the subjects were examined for their
ability to generalize in a new area of the workspace, their
performance was equally poor in both the intrinsic and
extrinsic fields. Our findings suggest that the ability to adapt
is more pronounced when the range of movements is
restricted. Similarly, we find that subject's ability to
generalize is very modest, suggesting an internal model
may be formed locally and not capable of global
generalization. On the whole, the many contrasting findings
on adaptation and generalization suggest our basic
understanding of the variables internal models represent,
and how they extrapolate to new circumstances, needs
further clarification and examination before a clear
understanding of internal models can be found.
1-A-3
Properties of a grasping tool affect grasping
behavior
Raoul Bongers1, Leonoor J Mouton1, Frank Zaal1
1
University of Groningen
The use of objects as tools is a hallmark in evolution of
which we only start to understand the basic underlying
processes. The current study examines changes in the
grasping pattern of healthy participants when picking up
objects with pairs of pliers that differed in length and that
differed in the relation between digit movement and
movement of the beak of the pair of pliers. The goal of this
study was to reveal how the kinematics of the grasp profile
change as a function of properties of the manipulated
1
Poster Sessions
Full Abstracts
relation between beak and digit movement and to link these
changes to neural mechanisms of prehension and tool use.
In the first experiment participants picked up a cylindrical object
with a pincer-like pair of pliers, with the beak at 3 cm, 8 cm or 16
cm from the fingers. Our findings showed that contrary to natural
finger grasping, the grasping pattern with the pair of pliers
showed a clear plateau phase. Moreover, the duration of this
plateau phase was independent of the length of the pair of pliers.
Finally, the maximum opening of the beak of the pliers was larger
for longer tools.
A second experiment was designed to investigate if the plateau
phase would disappear when participants use the pair of pliers
repetitively for huge number of trials. The analysis of pilot data
suggested that this is not the case.
In the third experiment, different participants used a pincer-like
pair of pliers, a pair of pliers in which the digits were crossed with
the beak, and a reverse pair of pliers, all of a length of 20 cm or
40 cm. Results showed that also with this set of pair of pliers the
grasp profile showed a plateau phase. Importantly, the plateau
phase was shortest for pincer pliers and longest for reverse pliers.
Also, the duration of the plateau phase was longer for longer
tools. Furthermore, hand opening was larger with pincer pair of
pliers, which follows from the functioning of that particular pair of
pliers.
Importantly, participants showed that they were able to use each
pair of pliers immediately in an appropriate way, independent of
the complexity of the required transformations. Moreover,
participants adapted the hand opening to the transformations
between digit movement and movement of the beak of the pair of
pliers. However, pliers grasping was characterized by a plateau
phase in the grasping profile, something that is not often reported
for fingers grasping. Interestingly, the plateau phase increased as
a function of complexity of transformations between digits and
beak.
The discussion will focus on the possible origins of this change in
the plateau phase as function of properties of tools. It will be
discussed how interactions between neural sites active during
tool use and during prehension predict the emergence of a
plateau phase in using a pair of pliers. In particular, the focus will
be on parietal networks coding for grasping of tools vs grasping of
non-tools, and those coding for objects within reach by hand and
tool, and those networks in dorsal premotor cortex that are
hypothesized to be involved in coordinating the reach and the
grasp.
1-A-4
Oral and pharyngeal sensory processing have
differential control over hyo-laryngeal kinematics
Ianessa Humbert1, Akshay Lokhande1, Heather Christopherson1,
Rebecca German1, Alice Stone1
1
Johns Hopkins University
Before a bolus is pushed into the pharynx, oral sensory
processing is critical for planning movements of the subsequent
pharyngeal swallow, including hyoid bone and laryngeal (hyolaryngeal) kinematics. However, oral and pharyngeal sensory
processing for hyo-laryngeal kinematics is not fully understood. In
11 healthy adults, we examined changes in kinematics with
sensory adaptation, sensitivity shifting, with oropharyngeal
swallows versus pharyngeal swallows (no oral processing), and
with various bolus volumes and tastes. Only pharyngeal swallows
showed sensory adaptation (gradual changes in kinematics with
repeated exposure to the same bolus) and sensitivity shifting
(changing sensory threshold for a small bolus when it immediately
follows several very large boluses). Conversely, only
oropharyngeal swallows distinguished volume differences, while
2
pharyngeal swallows did not. No taste effects were
observed for either swallow type. The hyo-laryngeal
kinematics were very similar between oropharyngeal
swallows and pharyngeal swallows with a comparable
bolus. These findings indicate that once oral sensory
processing has set a motor program for a specific kind of
bolus (i.e. 5ml water), hyo-laryngeal movements are already
highly standardized and optimized, showing no shifting or
adaptation regardless of repeated exposure (sensory
adaptation) or previous sensory experiences (sensitivity
shifting). Also, the oral cavity is highly specialized for
differentiating certain properties of a bolus (volume) that
might require a specific motor plan to ensure swallowing
safety, while the pharyngeal cavity does not make the same
distinctions. Pharyngeal sensory processing might not be
able to adjust motor plans created by the oral cavity once
the swallow has already been triggered.
1-A-5
Interference between geometric cues and
sensorimotor memories for anticipatory control of
manipulation
Qiushi Fu1, Marco Santello1
1
Arizona State University
Numerous studies using force fields and visuomotor
rotations have shown that sensorimotor memory of one
learned action can exert an anterograde interference on the
subsequent learning of an opposite action. However, few
studies have investigated whether an anterograde effect
occurs also when visual cues about object properties are
congruent with the action to be performed. For object
manipulation, visual geometric cues are normally used to
anticipate the properties of the object and the dynamics of
the upcoming manipulation. Our recent study showed that
sensorimotor memory built through manipulation with Ushaped object, requiring a torque in a given direction,
interferes with the ability of using congruent geometric cues
when learning to manipulate the same object requiring a
torque in the opposite direction. This result led us to ask
whether the interference is object-dependent or actiondependent. In the former case, an anterograde interference
might be found only within manipulations on one object,
whereas the latter case would occur every time that
sensorimotor memories conflict with geometric cues and
would therefore point to fundamental limitations in motor
planning. To address this question, we asked subjects (n =
16) to reach and lift one of the two L-shaped objects while
minimizing the roll by grasping the vertical handle with three
digits. The two objects have the same visual and
mechanical properties except the vertical handle being
either on the right (R) or left (L) side. The shape of the
object and the task require that, when grasping the L or R
object, subjects generate a counterclockwise or clockwise
compensatory torque (320 Nmm), respectively. Each
subject performed four blocks of eight trials. Within each
block, subjects grasped the same object and were
instructed to switch to the other object after the last trial of
each block. We evaluated subjects' anticipatory control of
manipulation by measuring compensatory torque produced
at object lift onset. The geometric cues were effective in
eliciting an anticipatory compensatory torque in the very first
trial (174.8±14.7 Nmm) and quickly improved to a level
close to 320 Nmm in subsequent trials. However, in the first
trial of the second block after switching to the other object,
the compensatory torque was poorly scaled (83.9±16.7
Nmm) indicating anterograde interference. Subsequently,
subjects were able to re-adapt to the second object to the
same performance level attained at the end of the first
Poster Sessions
Full Abstracts
block. The third and fourth block also showed similar interference
and re-adaptation, but the magnitude of interference decreased
as the total number of trials increased (165.5±15.1 Nmm and
205.9±12.1 Nmm of compensatory torques for the beginning of
third and fourth blocks, respectively). These results confirm earlier
observations that subjects are able to use object geometric cues
for anticipatory control of manipulation. Most importantly,
however, the present findings significantly extend previous work
by revealing that sensorimotor memory of previous manipulations
strongly interfere with visually-based estimation of object
properties for planning manipulations even when manipulating
different objects.
1-A-6
Patterns of hand muscle activation in pianists:
From amateur to expert
Sara Winges1, Shinichi Furuya1, Martha Flanders1
1
University of Minnesota
The effect of training can be observed in the quality of the
performance of a task or movement but it may also be discerned
at the cortical and neuromuscular levels. For piano playing, the
effect of training on the performance of skilled finger movements
can be demonstrated in the quality of play and the economy of
movements and improved accuracy. Highly trained pianists are
also more capable of restricting movement and force production
at non-striking digits when compared to amateur pianists. At the
cortical level, the effect of training is associated with enhanced
motor cortical excitability and plasticity and can be observed
specifically as changes in patterns of cortical activation for
specific finger movements. Presumably with training these cortical
changes would allow the pianist to perform highly practiced finger
movements more precisely with less effort. However, at the
neuromuscular level, the muscle activation patterns that
accompany changes in performance with respect to training are
not known.
(FPB), as well as four intrinsic finger muscles: first dorsal
interosseus (FDI), middle and ring lumbricals (LUMm,
LUMr), and abductor digiti minimi (ADM). The extent of
phasic coactivation across these hand muscles was
assessed for each musical piece and compared across
subjects and training levels.
1-A-7
Neural correlates of sensory plasticity
following motor learning
Sazzad Nasir1, Mohammad Darainy2, David J Ostry2
1
Northwestern University, 2McGill University
Though it is reasonable to assume that the neuroplasticity
associated with motor learning might affect sensory as well
as motor systems, there is limited neurophysiological
evidence supporting this idea. We show in the present study
that motor learning indeed alters sensory areas of the brain.
We adopted a force-field adaptation paradigm in which, as
subjects reached out to a visual target, a robot pushed the
hand laterally in proportion to its velocity in the direction of
the target. We used somatosensory evoked potentials
(SEPs), in response to brief force-pulses elicited by a
robotic device, as a probe to assess sensory plasticity. The
brain's response to these perturbations was measured
before and after learning. Our experiments show that motor
learning produces short latency changes to somatosensory
areas of the brain and in particular, to second
somatosensory cortex, an area of the brain presumably
involved in sensory aspects of tactile learning. The changes
we observe are substantially linked to motor learning. They
were not obtained when subjects passively traverse the
same movement trajectories and experience the same
spatial distribution of movements, but do not experience
learning. Moreover, the magnitude of the response is
correlated with the amount of learning; the more subjects
learn the greater the change in the sensory evoked
potential. These results are consistent with the idea that the
effects of motor learning extend into somatosensory areas
of the brain.
The purpose of this study was to characterize the temporal multimuscle activation patterns of pianists with different levels of
training during performance. The goal of the experiment was to
determine how training changes patterns of muscle activation
during skilled finger movement sequences and thus demonstrate
the plasticity within the patterns of neuromuscular control.
Therefore, we tested the specific hypothesis that asynchronous
phasic EMG bursts to limit movement at the non-striking digits
would characterize higher levels of training, while tonic coactivation of hand muscles would characterize lower levels of
training.
1-A-8
Sensory preference in speech motor
learning revealed by simultaneous alteration of
auditory and somatosensory feedback
Ten subjects (4 Male; 35±10 yrs) were included in the study. The
training level for each subject was determined by the number of
years playing (range: 7-48 yrs). Other training related information
was also recorded such as age at start of training, days per week
and hours per day spent practicing. Each subject performed 14
different musical scores that required the use of the right hand.
Excerpts were chosen to be challenging but also easy enough
that subjects with all levels of training included in the study could
play them. Subjects were allowed to familiarize themselves with
the piano prior to the start of data collection. Prior to data
recording for each new score, subjects were allowed to practice
the musical score in order to play consistently and accurately.
Each of the 14 scores was played for 10 successful trials at a
normal tempo provided by a metronome. Trials with miss-strikes
were repeated. Rest periods were given between each score to
prevent fatigue. Surface EMG was recorded from three portions
of extrinsic finger muscles: a central portion of extensor digitorum
(ED) and two portions of flexor digitorum superficialis, one closer
to the middle finger (FDS) and the other closer to the ring finger
(FDS2). EMG was also recorded from two intrinsic thumb
muscles: abductor pollicis brevis (APB) and flexor pollicis brevis
The idea that the nervous system learns and maintains
accurate speech by carefully monitoring auditory feedback
is widely held. But this view neglects the fact that auditory
feedback is highly correlated with somatosensory feedback
during speech production. Somatosensory feedback from
speech movements could be a primary means by which
cortical speech areas monitor the accuracy of produced
speech. We tested this idea by placing the somatosensory
and auditory systems in competition during speech
production. To do this, we combined two speech motor
learning paradigms to simultaneously alter somatosensory
and auditory feedback in real-time as subjects spoke. A
robotic device was used to perturb the motion path of the
jaw, altering somatosensory feedback during speech; an
acoustical effects processor was used to change the
frequency of vowel sounds, altering the sound of the voice
so that subjects heard something different from what they
produced. The amount of compensation for each
perturbation was used as a measure of sensory reliance. In
a large sample of subjects, we found that everyone
corrected for at least one of the two perturbations. By
applying the perturbation alone and then in combinations we
Sazzad Nasir1, Daniel R Lametti2, David J Ostry2
1
Northwestern University, 2McGill University
3
Poster Sessions
Full Abstracts
found that subjects have a trait-like preference for either
somatosensory or auditory feedback during speech. These
results have two surprising and important implications for our
understanding of how the brain produces accurate speech. The
first is that auditory feedback does not dominate speech
production. The second is that, in contrast to studies of
sensorimotor adaptation in limb movements, where all subjects
are observed to integrate sensory feedback in the same way, in
speech production the weighting of sensory feedback appears to
differ on an individual basis.
1-A-9
Task-specific effect of Transcranial Direct
Current Stimulation on motor learning
Cinthia Saucedo1, Xue Zhang1
1
K.U. Leuven
Anodal transcranial Direct Current Stimulation (tDCS) applied to
the human primary motor cortex (M1) proves to have beneficial
effects on motor skill learning in both: healthy controls [1-2] and
patients [3-4]. However, it remains unclear whether tDCS
improves motor learning in a general manner or whether there is
a task-specific effect. In this study, we tested the effect of tDCS in
two different motor tasks: (1) explicit sequence learning and (2)
visually guided force control task. Our hypothesis was that anodal
tDCS would lead to greater motor learning in both tasks, with
these improvements being task-dependent. Thirty-two healthy
subjects participated in this double-blind, sham-controlled crossover designed study. All subjects were randomly assigned to an
anodal-tDCS group or sham-group. We applied tDCS over the
primary motor cortex (M1) while subjects performed the motor
task. Two different sessions (session interval>1 month) were
performed, with the task-order randomized across participants.
Motor training of each task consisted of 20 min training for 3
continuous days. Retention tests were performed on the final
training day and one week after the training. Learning scores
were calculated and compared using a mixed model ANOVA
analysis. The results showed that for both tDCS groups there was
an overall improvement of scores across training (**p<0.001).
Anodal tDCS showed more improvement compared to sham, in
both motor tasks, but not to a significant level. In the explicit
sequence task, a significant interaction between the TIME of
stimulation (pre, training, post, RT) and TYPE of stimulation
(anodal/sham) was found (*p=0.01), with the greatest
improvement by anodal tDCS being at the 20 min retention test
(*p=0.01). On the other hand, the visually guided control motor
task showed the greatest improvement in the 1 week retention
test (*p=0.03). This findings suggest that anodal tDCS does lead
to an increase in motor learning most likely by improving
consolidation. Further, the exact expression of this effect seems
to be dependent on the motor task itself.
1-A-10 Enhanced locomotor adaptation after-effect in
the 'broken escalator' phenomenon using anodal tDCS
Diego Kaski1, Shamim Quadir1, Nada Yousif1, Adolfo M Bonstein1
1
Imperial College London
The everyday experience of stepping onto a stationary escalator
causes a stumble, despite our full awareness that the escalator is
broken. In the laboratory, this 'broken escalator' phenomenon is
reproduced when subjects step onto an obviously stationary
platform (AFTER trials) that was previously experienced as
moving (MOVING trials), and attests to a process of motor
adaptation. Given the critical role of M1 in upper limb motor
adaptation, and the potential for transcranial direct current
stimulation (tDCS) to increase cortical excitability, we
hypothesised that anodal tDCS over leg M1 and premotor
cortices would increase the size and duration of the locomotor
after-effect. Thirty healthy volunteers received either sham or real
4
tDCS (anodal bihemispheric tDCS; 2mA for 15 minutes at
rest) to induce excitatory effects over the primary motor and
premotor cortex, prior to walking onto the moving platform.
The real tDCS group - compared to sham - displayed larger
trunk sway, and increased gait velocity in the 1st AFTER
trial and a persistence of the trunk sway after-effect into the
2nd AFTER trial. We also used transcranial magnetic
stimulation to probe changes in cortical leg excitability using
different electrode montages and eye blink conditioning,
before and after tDCS, as well as simulating the current flow
of tDCS on the human brain using a computational model of
these different tDCS montages. Our data show that anodal
tDCS induces excitability changes in lower limb motor
cortex, with resultant enhancement of locomotor adaptation
after-effects. These findings might encourage the use of
tDCS over leg motor and premotor regions to improve
locomotor control in patients with neurological gait
disorders.
1-A-11 Enhancing voluntary control of neural
oscillatory activity driving a brain-machine interface
(BMI)
Surjo Soekadar1, Matthias Witkowski1, Niels P Birbaumer1,
Leonardo G Cohen2
1
University of Tübingen, 2NINDS / National Institutes of
Health
Introduction: Movement of a hand prosthesis after hand
paralysis following stroke can be established using brainmachine interfaces (BMI) that translate electric or metabolic
brain activity into movements of robotic or prosthetic
devices. Buch et al. (2008 and 2012) have shown that
severely affected chronic stroke patients are able to use murhythms, a form of oscillatory brain activity recordable over
motor areas with a frequency of 7-13Hz, to open and close
a mechanical hand orthosis. However, learning to control a
BMI based on modulation of oscillatory brain activity
requires training and is characterized by high variance of
trial-to-trial performance. Previous work demonstrated that
transcranial application of direct currents (tDCS) can
improve motor performance and motor learning (Hummel et
al. 2006; Reis et al. 2009). Thus, we tested the hypothesis
that application of weak electric currents preceding BMI
training will facilitate learning to control neural oscillatory
activity driving a brain-machine interface, and long-term
retention of this skill.
Methods: Thirty healthy participants (n=30) were randomly
assigned to one of three groups. They all trained over five
consecutive daily sessions imagining hand-opening motions
under EMG monitoring to modulate their mu-rhythm.
Immediately before training, depending on group
assignment, either anodal, cathodal or sham tDCS was
applied. During training, successful mu-rhythm modulation
was reflected in proportional online feedback delivered
through an orthotic device that passively opened the
subject's hand, mimicking the imagined training task.
Changes in successful mu-rhythm modulation were
analyzed across the training days and 30 days later to
assess long-term retention.
Results: Training in the anodal tDCS group improved murhythm control compared to the sham and the cathodal
tDCS groups. Cathodal tDCS cancelled the beneficial
effects of training alone evidenced in the sham group. One
month later, this skill remained significantly superior in the
anodal relative to both the sham and cathodal tDCS groups
(p<.05 and p<.01, respectively).
Poster Sessions
Full Abstracts
Conclusion: Our results show that application of painless
noninvasive weak anodal tDCS over a relevant primary cortical
region can improve learning voluntary control of neural oscillatory
activity.
1-A-12 How action shapes space and body
representations
Michela Bassolino1, Alessandra Finisguerra1, Andrea Serino2,
Thierry Pozzo1
1
Istituto Italiano di Tecnologia, 2CsrNC, Centro studi e ricerche in
Neuroscienze Cognitive
In order to interact with the external world, our brain relies on
multisensory representations of the body in space, related both to
the position and dimension of the body parts (the body schema,
BS) and to the location of external objects in the space
surrounding the body (the peripersonal space, PPS). Both the
PPS and the BS are dynamically shaped by actions (e.g. Làdavas
& Serino, 2008): interacting with far objects by means of a tool
extends the boundaries of the PPS and affects the BS. In order to
better characterize the relation between action, PPS and BS, here
we investigate if the absence of motion may alter these
representations. We compare PPS and BS before and after 10
hours of right arm immobilization in healthy subjects.
Modifications of PPS are tested by means of an audio-tactile
integration task: we consider the boundaries of the PPS as the
critical distance at which an auditory stimulus, approaching or
receding to the body affects the processing of a tactile stimulus
on the arm (Serino et al., 2007; Bassolino et al., 2010). If action is
necessary to shape space representation, the boundaries of the
PPS should shift closer to the body after immobilization,
suggesting a contraction of the PPS around the arm. Moreover,
modifications of the BS are tested by means of a tactile distance
perception task: we consider the perceived distance between two
tactile stimuli on the forearm as an implicit measure of the
perceived arm length (Canzoneri et al, under revision). Since after
tool use subjects perceived their forearm as elongated, after
immobilization we expect a shrinkage of the arm representation in
the BS. Further, the same measurements are performed on the
left, unrestricted, arm in order to evaluate if the compensatory
overuse of the free limb during non-use (Avanzino et al. 2011),
can modify space and body representation, probably in the
opposite direction as compared to the effect of immobilization.
Our results are discussed in terms of plasticity and mechanisms
underlying PPS and BS representations.
1-A-13 Adaptation of surround inhibition in the human
motor system
Panagiotis Kassavetis1, Tabish A Saifee1, Anna Sadnicka1, Isabel
Pareés1, Maja Kojovic1, John C Rothwell1, Mark J Edwards1
1
University College London, Institute of Neurology
Motor surround inhibition (SI) is proposed as a mechanism to
enhance motor performance by actively inhibiting the excitability
of the corticospinal pathways which control surround muscles that
potentially interfere with desired movements. Motor performance
can be modified through a feed-forward motor adaptation process
when a mismatch of the motor command and the sensory
feedback occurs. Motor adaptation has been extensively studied
with behavioural paradigms but electrophysiological evidence is
lacking especially from the scope of surround inhibition. We
probed adaptation of SI by introducing a false sensory feedback
from a surround muscle during a training session when
participants were requested to repetitively perform a brief index
finger flexion movement while stimulation of the muscle spindles
with vibration was applied to a surround muscle either at the
onset of movement or with a delay of 100ms. We assessed motor
SI before and after the training session with transcranial magnetic
stimulation (TMS). Statistical analysis of the results showed
that SI measured after the training session is increased
compared to baseline when vibration was timed at the onset
of the movement but not when vibration was delayed. The
effect persisted for a while after withdrawal of vibration and
slowly returned to baseline. The present study demonstrates
that SI can be modified by experience. The timing of the
sensory stimulation was found to be critical for the
modification of SI, suggesting that only sensory signals
closely related to the movement onset can induce adaptive
changes, presumably through a feed-forward process.
1-A-14 Transitions between finger keypresses
during acquisition of novel piano sequences are
related to patterns of low frequency local field
potential activity in the human globus pallidus
internus
Maria Herrojo Ruiz1, Christof Brücke1, Gerd-Helge
Schneider1, Andrea A Kühn1
1
Charité - University of Medicine
The regulation of temporal and spatial order in sequence
production engages cortical and subcortical structures to a
different degree depending on the stage of skill acquisition.
The current study explores further the role of the basal
ganglia in organizing and enacting novel action sequences
during learning. To this aim, we investigated the temporal
patterns of local field potential (LFP) activity in the human
internal globus pallidus (GPi) in nine patients undergoing
bilateral DBS for idiopathic cervical or segmental dystonia
(hands not affected) during the performance of novel piano
sequences. The GPi is the main output station of the BG
and sends inhibitory projections to the thalamus.
Specifically, we tested whether patterns of pallidal
oscillatory LFP activity prior to correct or erroneous (in pitch)
keypresses in a finger sequence differ. Such a phenomenon
would be an indication of early gating mechanisms of
cortical representations of the sequence, leading to
facilitation of the execution of correct sequence items and
avoidance of upcoming errors.
Patients had to learn (~20min) five finger sequences for the
right hand, consisting of 4-10 notes at a rate of 3 Hz, on a
digital piano. Following the initial learning session, patients
performed 10 trials of 25 seconds of each sequence type,
while the performance was recorded as MIDI (music
instruments digital interface) files. Simultaneously, LFPs
were recorded bipolarly in the motor area of the GPi at
sampling rate of 1kHz, with a bandpass filter of 0.5-400Hz.
Analysis of the performance data revealed an average interonset-interval (IOI) of 400ms (standard deviation [SD] 60m)
in correct pitch notes. Interestingly, prior to and following
wrong pitch keypresses the IOI was significantly larger
reflecting pre- and post-error slowing mechanisms (pre: 560
[70] ms, post: 700 [200] ms). Moreover, the MIDI velocity in
pitch errors was significantly reduced as compared with the
values of correct pitch keypresses. In sum, performance
analysis pointed to an early detection of errors during motor
skill acquisition which manifested as performance
adjustments in pitch errors. Correspondingly, analysis of the
LFP 4-100Hz oscillatory activity time-locked to correct and
wrong pitch keypresses demonstrated two salient effects:
first, a significant decrease in theta and alpha (4-13Hz)
oscillatory activity 300-100ms prior to correct keypresses in
the contralateral GPi, which was not present prior to errors.
Second, in pitch errors, the reduced MIDI velocity
converged with significantly less enhanced gamma (40-
5
Poster Sessions
Full Abstracts
90Hz) oscillations at key onset, as compared to correct
keypresses.
In sum, decrease in pallidal low frequency LFP activity before
correct keypresses might facilitate the selective execution of
upcoming correct elements in a sequence during skill acquisition.
This putative gating mechanism possibly depends on the degree
of coherence in a prior cortical and striatal representation of the
sequence serial order.
1-A-15 Limited visuomotor adaptation to variable
amplitude gains within a movement trajectory
trajectory. Instead, the results suggest that a single gain is
employed over the entire movement in conjunction with a
reactive control strategy. Decomposing the overall
movement into smaller, fixed-gain movements might allow
the motor system to efficiently compensate for tracking
errors due to gain-mismatches using feedback control.
1-A-16 Physiological evidence of reduced
neuroplasticity in human adolescents who were born
preterm
Julia Pitcher1, Alysha M Riley1, Michael C Ridding1
Deborah Barany1, Shivakumar Viswanathan1, Scott T Grafton1
1
1
Background: Neuroplasticity is the ability of the brain to
alter neuronal synaptic strength in response to activity and
experiences. It is widely accepted to be the mechanism
underlying learning and memory formation, but also plays a
key role in brain growth and development. Preterm children
have alterations in cortical development, functional
connectivity and neural activation patterns that suggests
their capacity for neuroplastic reorganization may also be
reduced, critically contributing to their common difficulties
with learning and memory. Hypothesis: Preterm birth is
associated with a reduced response to a non-invasive
neuroplasticity induction intervention designed to induce a
short-term LTD-like (i.e. inhibitory) change in motor cortex
(M1) excitability.
University of California, Santa Barbara
Gain adaption is the motor system's ability to appropriately scale
movement amplitude to match the physical demands of the
environment with practice. Does this ability extend to learning
different visuomotor gains within the same environment? The
motor system can learn to use different amplitude gains along
different movement directions (Pearson, Krakauer, & Mazzoni,
2010). However, it remains unclear whether the motor system can
adapt to amplitude gains that vary along different segments within
the movement trajectory itself. We evaluated this possibility in two
experiments using a novel virtual-reality paradigm where
participants tracked the visual movement of a metronome.
The metronome was a virtual upright bar pivoted at the lower end,
while the upper end oscillated along the perimeter of a notional
semicircle with constant angular velocity (75 degrees/s).
Participants tracked the metronome with a virtual right hand
grasping a bar displayed on a computer screen. The virtual
hand's orientation was updated continuously based on the current
orientation of the participant's out-of-sight right hand. The gain
(the ratio of virtual to actual movement amplitude) was equal to
1.5 within a central sector of width 60°; and 6 everywhere else.
On each trial, participants tracked the metronome over 20
oscillations, followed by feedback indicating the Root-MeanSquare-Error (RMSE) on that trial and the percent improvement in
RMSE over the previous trial. Participants were exposed to 40
adaptation trials interleaved with probe trials that occurred every
10 trials. In Experiment 1, the probes had a constant gain (equal
to 6) over the entire movement trajectory. In Experiment 2, the
probes only showed the metronome without any visual feedback
about the participant's current position.
In both experiments, participants' RMSE decreased over the 40
trials, showing that the task was indeed learnable. Due to the
variable gains over the movement trajectory, an additional
measure of adaptation was the extent to which the virtual hand
"overshot" the metronome's position at the terminal points of the
oscillation. As predicted, the overshoot error at these points was
substantial on early trials but decreased with practice.
Were two visuomotor maps used to minimize tracking error? The
behavior on the probe trials suggested otherwise. In Experiment
1, the RMSE on the constant-gain probe trials remained high and
largely unaffected by the decreasing error on the variable gain
trials. Importantly, we found no evidence of anticipatory velocity
changes at orientations corresponding to the previously
experienced transition boundaries between gain fields. In
Experiment 2, movement amplitudes on the no-feedback probe
trials were substantially and symmetrically larger than on the
learning trials, consistent with movements scaled using a single,
low-valued gain over the entire movement trajectory. As in
Experiment 1, we found no evidence of anticipatory effects at the
gain-field transition boundaries.
Together the results are inconsistent with the use of different
spatially defined visuomotor mappings along the movement
6
University of Adelaide
Methods: 25 children (15 females) aged 12-15 years (13.67
± 0.48 years) participated; Term born (37-41 wks GA) N=6,
Late preterm (33-36 wks GA) N=9 and Early preterm (24-32
wks GA) N=9. Continuous theta burst stimulation (cTBS)
was applied to the M1 to induce LTD-like neuroplasticity. To
assess changes in M1 excitability (an indicator of
neuroplasticity), transcranial magnetic brain stimulation was
used to evoke motor evoked potentials (MEPs) from a hand
muscle before and up to 60 min following cTBS.
Results: Term-born children showed robust MEP amplitude
suppression immediately following cTBS that was greater
and more persistent than that previously consistently
recorded in adults. In comparison, MEP suppression in both
preterm groups was significantly less than term born
children and returned to baseline within 40 min of cTBS
ceasing. GA correlated negatively with the mean MEP
suppression following cTBS, i.e. the least suppression was
evoked in the most preterm children.
Conclusions: These data provide the first physiological
evidence of reduced neuroplasticity in preterm children.
While different types of neuroplasticity induction (i.e. LTPlike, behavioural) are yet to be assessed, these results
demonstrate that even modest levels of prematurity are
associated with significant impairments that persist at least
into early adolescence. The underlying mechanisms are not
yet clear, but may include synapse specific dysfunction
and/or altered cortisol secretion patterns which are known to
influence neuroplasticity.
1-A-17 The presence of multiple potential visual
targets affects the retrieval of motor memory for a
reaching movement
Masaya Hirashima1, Daichi Nozaki1, Gaku D Yamawaki1
1
The University of Tokyo
Humans can adapt movements to a novel dynamic
environment by acquiring an internal model of the dynamics.
It is suggested that the internal model is constructed by a
population of the primitives encoding the desired joint
Poster Sessions
Full Abstracts
kinematics. In this scheme, the desired joint kinematics
determines the combination of the primitives to be retrieved for
executing an upcoming movement, suggesting that as long as a
physically identical movement is intended, identical primitives are
always retrieved. However, a recent study demonstrated that
distinct motor memories can be flexibly retrieved for executing a
physically identical movement by planning distinct motions in a
visual space (Hirashima and Nozaki, Curr Biol 2012). This finding
suggests that the primitives to be retrieved are determined not
only by the desired joint kinematics but also by the motor planning
process. Although there is ample neurophysiological evidence
showing that physically identical movement could be generated
through distinct preparatory neural activities, it remains largely
unknown how such preparatory activity affects the control of
upcoming movement.
Here, we investigate how motor planning process affects the
retrieval of motor memory for an identical movement, by using
force field adaptation paradigm with delayed reaching task. At the
beginning of a delay period, one or two potential target locations
were provided as prior target information available for motor
planning (i.e., single-target and double-target condition), and then
at the end of the delay period, the final target was randomly
selected from the potential targets and illuminated as the GO
signal. In experiment 1, after participants adapted to a curl force
field with a particular reaching movement (toward 90º target) in
single-target condition, they were occasionally required to perform
double-target task (with 90º and 120º targets). By using errorcramp trials, we assessed the transfer of learning from single- to
double-target condition in the identical 90º movement. We found
that the degree of adaptation was significantly lower in doubletarget condition than single-target condition (p < 0.05).
Considering the neurophysiological finding that relatively lower
population activity was observed during the delay period when
multiple potential targets were provided (Bastian et al. Eur J
Neurosci 2003), the present results might be explained by
assuming that preparatory activity level determines the degree of
the retrieval of motor memory. However, it should be noted that
this assumption predicts that a hyper generalization occurs from
double- to single-target condition. To test this prediction, we
conducted experiment 2 in which participants first adapted to the
force field with two reaching movements (90º and 120º) in doubletarget condition, and then they were required to perform them in
single-target condition. Results exhibited no evidence of hypergeneralization; the degree of adaptation decreased in singletarget condition for 90º target (p < 0.05) and there was no
significant difference for 120º target (p > 0.05). Taken together,
the results suggest that motor planning process affects the
retrieval of motor memory in a way that activation "pattern" of the
network rather than the level of population activity determines the
primitives to be retrieved.
1-A-18 Plasticity in amputees: Reorganization in
somatosensory and motor cortices varies with adaptive
strategies for limb use
Tamar Makin1, Jan Scholz2, Nicola Filippini1, David Henderson
Slater3, Irene Tracey1, Heidi Johansen-Berg1
1
Oxford University, 2Hospital for Sick Children, 3Nuffield
Orthopaedic Centre
Our brain's ability to reorganize itself by forming new neural
connections throughout life is a key mechanism that enables
adjustments to novel situations, as well as compensation for
nervous system injury. Following hand-amputation, two types of
reorganization could occur. First, reorganization associated with
sensory deprivation resulting from the loss of the hand, which is
commonly viewed as maladaptive. Second, reorganization due to
altered patterns of use of the remaining limbs, which could play
an adaptive role; to overcome their impairment, individuals
will adopt new strategies for using their intact hand or their
residual arm (on the amputated side). The majority of
research on amputees to date has focused on maladaptive
plasticity, whereas we aimed to test whether plasticity plays
an adaptive role following amputation. We tested for
functional variations associated with different patterns of
limb use in individuals with a unilateral congenital or
traumatic upper-limb deficit. Using an adapted version of the
motor activity log (MAL) inventory, we identified different
limb use strategies in these two groups: whereas
congenitals were likely to use their residual arm during daily
tasks, the late amputees predominantly relied on their intact
hand. Accordingly, a fMRI based somatomotor-mapping test
showed a dissociation in remapping patterns in these two
populations: congenitals showed contralateral overrepresentation of the residual arm, compared with both late
amputees and controls. In the primary somatomotor cortex,
this over-representation overlapped with activations relating
to movements of the phantom (absent) hand in the late
amputees, as well as movements of the non-dominant hand
in controls. By contrast, late amputees showed ipsilateral
over-representation of the intact hand, compared with the
other groups, in the primary somatomotor cortex associated
with the phantom hand. Moreover, increases in ipsilateral
representation of the intact hand correlated negatively with
MAL scores across both amputees populations: individuals
that were less able to utilize their residual arm in daily tasks
were more likely to show increased over-representation of
the intact hand in the phantom cortex. These results
suggest that functional plasticity is not restricted to a critical
period, but is rather contingent upon the limb use strategy
adopted by individuals.
1-A-19 Cortical networks involved in speech
recovery after glossectomy: Preliminary results of
an fMRI study
Audrey Acher1, Marc Sato1, Laurent Lamalle2, Coriandre
Vilain1, Alexandre Krainik2, Pascal Perrier1
1
Gipsa-lab - UMR CNRS 5216, 2CHU de Grenoble
Tongue surgery, also called glossectomy, is often necessary
in the clinical treatment of tongue cancer. It is often
associated with radiotherapy. In some cases these
treatments induce a noticeable reduction of tongue mobility,
and capacities of the patient to shape his vocal tract.
Intelligibility of speech is strongly altered. Hence, speech
production after tongue surgery often requires the patient to
go through a long speech recovery process. We investigate
cortical networks involved in the recovery process. It is
assumed that a major step is the learning of new relations
between motor commands, oro-sensory feedback and
spectral characteristics of speech. It has been proposed that
these relations could be stored in the form of internal
representations of the motor system, which could be
implemented in the cerebellum. We expect the speech
recovery process to be associated with significant changes
in cerebellar activations: they should be stronger during the
first part of the process, as the results of the learning, before
returning to a normal level in the second part. Once these
new internal representations acquired, patients should
elaborate new compensation strategies, in order to reach
their usual acoustic speech goals. Intact articulators, i.e. jaw
and lips, should be used more systematically and possibly
with a greater accuracy than before surgery, in order to
overcome the limits of tongue mobility. This could induce
activity changes in ventral premotor and primary motor
cortices, associated with orofacial motor control (Grabski et
7
Poster Sessions
Full Abstracts
al., in press). For some speech units (phonemes or syllables), the
usual goals of speech production are likely to be not reachable
any more. For these units, speech recovery would then require
the definition of new goals, both in oro-sensory and auditory
domains. From this view, auditory speech goals have been
suggested to be located in the left temporo-parietal junction
(Hickok & Poppel, 2004) while brain regions sensitive to
somatosensory goals are hypothesized to be located in the
somatosensory cortex and the antero-dorsal part of the inferior
lobule and adjacent intraparietal sulcus (Golfinopoulos et al.,
2011). We expect all these regions to be strongly activated during
the first part of the learning process. Finally, learning new
coordination strategies between speech articulators should be
reflected in increased activity in the left anterior insula. To assess
these hypotheses, patients were recorded longitudinally in 1 presurgery and 3 post-surgery sessions, within 9 months. Using
sparse sampling acquisition in order to minimize movementrelated artifacts and scanner noise, patients were asked to
produce speech and non speech orofacial motor tasks in each
fMRI session. Recordings are currently in progress and
preliminary results will be presented at the conference. Hickok, G.
& Poeppel, D. (2004). Dorsal and ventral streams: A framework
for understanding aspects of the functional anatomy of language.
Cognition, 92, 67-99. Grabski, K., Lamalle, L., Vilain, C.,
Schwartz, J.-L, Vallée, N. Troprès, I., Baciu, M. Le Bas, J.-F &
Sato, M. (In Press). Functional MRI assessment of orofacial
articulators: neural correlates of lip, jaw, larynx and tongue
movements. Human Brain Mapping. Golfinopoulos, E., Tourville,
J.A., Bohland, J.W., Ghosh, S.S., Nieto-Castanon, A. & Guenther,
F.H. (2011). fMRI investigation of unexpected somatosensory
feedback perturbation during speech. NeuroImage, 55(3):132438.
1-A-20 Comparing neural activity for repetitive finger
and foot movements in cerebellar patients and healthy
controls reveals lesion induced changes in specific
regions of cortex
Paul Pope1, Roxana Burciu2, Maria Dagioglou1, Nina Theysohn2,
R Chris Miall1, Dagmar Timmann2
1
University of Birmingham, 2University of Duisburg-Essen
Objective: The extent to which cerebellar-cortical connectivity is
modified during post-stroke cortical reorganization is little
understood, yet potentially useful for understanding lesioninduced network plasticity. Here we used functional Magnetic
Resonance Imaging (fMRI) to investigate differential patterns of
neural activity in cerebellar patients and healthy controls using
tasks that have previous been shown to activate the cerebellum
and motor regions of cortex.
Method: Blood Oxygenation Level Dependent (BOLD) signals
were recorded using a fast fMRI sequence in cerebellar patients
at least one year following their stroke (N = 14), and
age/education/gender matched controls (N = 14), whilst they
reproduced either repetitive sequence of finger-thumb opposition,
or foot-tapping movements with the left or right limbs in four
separate runs of 372 s, cued by a video inside an MR scanner
(1.5 T). Covariates of motor behaviour (time and force) were also
recorded simultaneously. Patterns of neural activity for each
condition were compared between groups with General Linear
Model (GLM) analyses after flipping data sets so that lesion sites
were on the same side, and the locus of neural activity was
complementary between groups.
Results: Inspection of motor behaviour revealed accurate task
compliance in the scanner. Functional data associated with this
performance revealed a network of task-specific brain regions,
including: the cerebellum, motor and pre-motor cortices,
supplementary motor area, together with frontal and parietal
8
regions of cortex in all four conditions. However, activity in
post central gyrus was different in patients than controls for
sequential hand and foot movements, which may be
explained in terms of post-stroke reorganization processes
that accompany the partial recovery of cerebellar function in
patients relative to controls. Further analyses will correlate
task-specific brain activity in local regions of cortex with
covariates of motor behaviour such as movement time and
response force, together with scores for the assessment of
motor and cognitive functions.
Conclusion: Repetitive and sequential movements of the
fingers and feet are accompanied by task-specific MR
activations, which are affected differently in cerebellar
patients and healthy controls, presumably due to cortical
reorganization after stoke. Exploring the relationship
between brain activation and covariates of motor behaviour
in local regions of interest will further expand our
understanding of specific changes in the brain after
cerebellar stoke. Supported by the Wellcome Trust and a
Marie Curie Initial Training Network C7 (Cerebellar-Cortical
Control: Cells, Circuits, Computation and Clinic).
1-A-21 Learning finger coordination patterns by
altering dimensionality
Robert Scheidt1, Rajiv Ranganathan2, Jon A Wieser1,
Kristine M Mosier3, Ferdinando A Mussa-Ivaldi2
1
Marquette University, 2Rehabilitation Institute of Chicago,
Indiana University School of Medicine
3
Motor skill acquisition can sometimes be facilitated by
breaking a task down into simpler components (part-whole
practice). However, in cases where multiple degrees of
freedom are involved, learning task components may result
in the suboptimal use redundant degrees of freedom when
performing the whole task. Here, we investigate whether
learning of a redundant 2-D task can be facilitated by
manipulating (i.e. either reducing or augmenting) the
dimensionality of training. 51 subjects wore a data glove to
perform a virtual reaching task using their fingers. 19-signals
from the glove were linearly mapped onto the 2-D position of
a cursor on a display screen. The screen was a 5x5 grid of
25 target squares. Subjects were instructed to move the
cursor into the specified target as quickly and accurately as
possible. They practiced this task for three days. The
mapping of finger motions onto cursor motion was formed
from the first two principal components obtained during a
finger-spelling task (hereafter referred to as the "original"
map). Subjects were divided into five groups that differed in
the map practiced on Day 1: Two groups practiced a 2-D
task on Day 1 - (a) the 2D group used the original map, (b)
the 2D -Null group practiced with two dimensions that were
in the null space of the original map (PC3 and PC4). Two
groups practiced with only one-dimension of the task - (c)
the 1D group used a map with the first dimension of the
original map (i.e., PC1) whereas (d) the 1D-Null group
practiced the task with PC3 (in the null space of the original
map). Finally, (e) the 3D group practiced the task using
three dimensions - two from the original map that controlled
the location of the cursor and a third (PC3) that controlled
the size of the cursor. After practicing the different maps on
Day 1, all groups performed the 2D task using the original
map on Days 2 and 3. The extent of learning and transfer
was assessed using the performance on Day 2 when all
groups performed the same task. On Day 2, there was a
systematic trend in task performance depending on the
number of Day 1 task components that were shared with the
"original" Day 2 map. The best performance was achieved
by the 2D and the 3D groups, where two components were
Poster Sessions
Full Abstracts
shared. This was followed by the 1D group which had one
component in common - but the performance of this group on Day
2 was not different from the naive performance of the 2D group
on Day 1. Finally, the 1D-Null group and the 2D-Null groups that
had no shared components had the worst levels of performance
on Day 2. In particular, the 2D-Null group showed sustained
interference even after 2 days of practice. Data analysis revealed
that subjects in the 2D-Null group were exploring the null space
even after the map had changed, indicating that the map learned
by the participants on Day 1 was interfering with the acquisition of
the map on Days 2 and 3. These results demonstrate that when
subjects perform a redundant kinematic task, they develop a
representation of the degrees of freedom that are relevant to task
performance. This representation has strong interference with
maps of that are based on orthogonal subspaces. Accordingly,
altering dimensionality during training must be conducted with
care in tasks with multiple degrees of freedom. In particular,
uncontrolled recruitment of null-space motions during practice
may lead to the use of suboptimal coordination patterns in the fulldimensional task.
1-A-22 Ontogeny of vestibulo-ocular reflex following
genetic or environmental alteration of gravity-perception
Mathieu Beraneck1, Mickael Bojados2, Anne Le Séach1, Marc
Jamon2, Pierre-Paul Vidal1
1
CNRS - University Paris Descartes, 2Aix-Marseille Univ, UMR
1106
The vestibular organs consist of complementary sensors: the
semicircular canals detect rotations while the otoliths detect linear
accelerations, including the constant pull of gravity. Several
fundamental questions remain on how the vestibular system
would develop and/or adapt to prolonged changes in gravity such
as during long-term space journey. The aim of present work was
to evaluate the role of otolith information during ontogeny of the
vestibular system. Inner ear defect (ied) mutant mice suffer from
otoconial agenesis and thus develop ontogenetically in the
absence of the vestibular perception of gravity. In ied mice
maculo-ocular reflexes were absent. While canals-related reflexes
were present, the ied deficit also led to the abnormal gain, timing,
and spatial tuning of the horizontal angular vestibulo-ocular reflex.
The perturbation of the encoding of gravity by the otolith organs
could therefore have impaired the development of canal-related
pathways. To test this hypothesis and identify putative otolithrelated critical periods, C57Bl/6 mice were subjected to 2G
hypergravity by chronic centrifugation during different periods of
development and compared to adult centrifuged mice. One month
after the end of the centrifugation, horizontal angular vestibuloocular reflex of all centrifuged animals was normal, while maculoocular reflexes was transitorily impaired in mice which completely
developed in hypergravity. The maculo-ocular reflexes of adult
centrifuged mice were also impaired; however long-term effects
were found to persist in a subset of adult mice, suggesting
chronic alteration in the processing of vestibular signals. In
summary, genetic suppression of gravity-related signals indicated
that otolith-related signals might be necessary to ensure proper
functioning of canal-related vestibular reflexes. On the other
hand, exposure to hypergravity was not sufficient to alter durably
motor behaviour, probably because the otolithic perception of
gravity was altered but not suppressed.
1-A-23 The impact of limb agenesis on brain structure
Erika Rodrigues1, Fernanda Tovar-Moll2, Ivanei Bramati2, Claudia
Vargas3, Jorge Moll2, Angela Sirigu4
1
Augusto Motta University Center (UNISUAM), 2D Or Institute for
Research and Education (IDOR), 3Federal University of Rio de
Janeiro, 4Center for Cognitive Neuroscience
Motor areas in the brain are continuously under the effect of
sensory-motor learning. Here we asked whether
undeveloped body parts following injuries occurring in utero
impact on the way the corresponding movement
representation develops during ontogenesis. We addressed
this question by examining gray and white matter changes
in individuals (N=5) with left or right upper limb agenesis
and compared them with traumatic upper limb amputees
(N=9) and healthy controls (N=10). Conventional MRI
(FLAIR, T1- and T2-weighted images) showed no
anatomical differences for the three groups. Cortical
thickness and diffusion tensor imaging (DTI) analyses,
however, revealed clear cortical and white matter changes
in agenesic amputees compared to controls and the
traumatic group. Whole brain voxelwise tract-based spatial
statistics analysis (TBSS; FSL 4.0, FMRIB software)
revealed extensive fractional anisotropy (FA) reduction in
the cortico-spinal tract, the superior longitudinal fasciculus,
the corpus callosum and the cingulate bundle in the
hemisphere contralateral to the missing limb in agenesics
subjects, but not in traumatic amputees who did not differed
from the healthy group. Whole brain analysis further showed
reduced grey matter thickness in the agenesics' hand and
arm area of the primary motor cortex (M1) contralateral to
the absent limb (Freesurfer image analysis suite, version
4.4). These structural differences were not found in the
hemisphere ipsilateral to the healthy limb. Thus, having
never experienced movements of the limb has an impact on
the structural organization of the sensorimotor cortical areas
and related white-matter pathways. Such massive
reorganization may explain why these patients contrary to
traumatic amputees never develop phantom movements,
which are supposed to be caused by the activation of a
persistent representation of the lost limb in M1. Remarkably,
all changes were exclusively contralateral to the absent
limb. Our findings suggest that motor experience is crucial
for the development of the hand motor representation in the
primary motor cortex. We can speculate that patients with
limb agenesis have never activated motor commands
towards muscles. This may have a consequence on the way
cortico-spinal pathways are shaped during ontogenesis.
1-A-24 Recent action determines the encoding of
motor memory
Ian Howard1, David W Franklin1, James N Ingram1, Daniel M
Wolpert1
1
University of Cambridge
Real world tasks often require movements that follow on
from a previous action. Here we investigate the contextual
effect of immediate prior motion using an interference task.
Subjects performed trials in which they made movements in
a randomly selected clockwise or counter-clockwise
velocity-dependent curl force-field. Movements during this
adaptation phase were preceded by a contextual phase that
determined which of the two fields would be experienced.
As expected, when a static visual cue was used in the
contextual phase to indicate the direction of the field, strong
interference was observed. However, when the contextual
phase involved subjects making a movement that was
continuous with the adaptation phase movement, a
substantial reduction in interference was seen. As the time
between these two movements increased, so did the
interference, reaching a level similar to the interference
seen for static visual cues for delays greater than 600 ms.
The contextual effect for such recent motion generalized to
purely visual motion, active movement without vision and
passive movement. In addition, isometric force generation
9
Poster Sessions
Full Abstracts
without movement also provided the same contextual effect. Our
results show that active, passive or visual prior motion, as well as
static force production, can each engage different representations
in motor memory. This suggests that recent action in a variety of
modalities has a special role in determining the context for motor
learning, as compared to static cues. However, as the delay
between recent action and the movement is increased, the
contextual effect progressively decays, suggesting that the
encoding of motor memory depends strongly on the recency of
the action.
1-A-25 Adaptation to force perturbations alters
feedforward and feedback signals in Purkinje cell simple
spike firing
Angela Hewitt1, Laurentiu S Popa1, Timothy J Ebner1
1
University of Minnesota
Motor control theorists have postulated that the nervous system
employs a forward internal model to predict the consequences of
a motor command. A forward internal model has many useful
properties, including bypassing sensory feedback delays,
coordinating movements between effectors, and providing training
signals. Although the neurophysiological substrate for a forward
internal model remains unknown, psychophysical, imaging, and
patient case studies suggest it is implemented in the cerebellum.
If true, Purkinje cell (PC) simple spike firing should behave as a
forward internal model and "learn" new input-output properties of
the motor apparatus during adaptation to predictable force
perturbations.
To test this, extracellular firing from single PCs (lobules IV-VI of
the intermediate/lateral zones) was recorded while two rhesus
monkeys used a robotic manipulandum to move a cursor from a
start target to a cue target, resulting in very fast (< 750 ms)
reaching movements along a 10 cm path. Animals completed four
blocks of movements. First, baseline trials obtained the cell's
preferred direction and the firing discharge in an error clamp with
no force perturbations. The error clamp consisted of robotgenerated virtual walls that required the animal to move along a
highly defined path towards the target, minimizing small tangential
kinematic errors that might mask learning effects. The second
block adapted the animal to the force perturbation. Perturbation
parameters, including magnitude, start position, and duration,
were randomized between recording days. The third block
continued using the perturbation, but also included catch trials
(randomized at 10-15%), where the perturbation was
unexpectedly absent. A fourth block repeated baseline conditions.
Kinematic analyses illustrate that both animals make large
changes in arm velocity and force to rapidly correct for the
perturbation within several trials, and then gradually continue
adjusting parameters to optimize their movements. Similarly, 6070% of task-modulated PCs show steady, progressive firing
changes with adaptation. Catch trials also demonstrate strong
adaptation, as reaching movements often over- or undershoot the
target as expected. Firing activity during a catch trial closely
matches that of adaptation trials in the period before the
perturbation should occur, but large differences arise thereafter.
Preliminary modeling fit the firing data to multiple linear
regression models of hand kinematics (position and velocity) and
kinetics (force parallel or tangential to movement). An additional
model parameter τ estimated the time lead or lag between PC
firing and hand kinematics or kinetics. During baseline conditions,
maximum model R2adj values peak at both feedforward and
feedback τ timepoints for many PCs. Initial exposure to the
perturbation disrupts these depictions, and the best model fit
often reflects a feedback representation of kinematics or force.
With adaptation, best model fits frequently return to including both
feedforward and feedback signals that may be shifted in time
10
compared to baseline data. Overall, adaptation appears to
alter the relationships between feedforward and feedback
signals present in PC firing. This is consistent with a forward
internal model learning new input-output properties of the
motor apparatus.
Supported in part by NIH grants RO1 NS18338, F30
NS071686, T32 GM008244
1-A-26 Motor learning increases the distinctiveness
of cortical sequence representations: Evidence from
multi-voxel pattern analysis
Tobias Wiestler1, Jörn Diedrichsen1
1
University College London
The study of motor learning using functional magnetic
resonance imaging (fMRI) has revealed a complex picture of
activity changes in the human brain. Some regions increase
activity, often interpreted as evidence for increased neural
recruitment through the learning process. Activity decreases
with learning are also found in many regions. These may
either indicate that an area has become less important for
the skilled behavior, or, alternatively, that the same regions
now encodes the learned behavior more efficiently. The
complex overlap of these possible different mechanisms
makes the interpretation of overall activity changes during
learning problematic. Here we test the hypothesis that motor
skills are associated with the development of neuronal units
within motor cortical network that are specialized for the
trained behavior. We therefore predict that local activity
patterns for different behaviors should become more distinct
from each other when a person acquires higher levels of
performance. We show here that these changes are
detectable using functional magnetic resonance imaging
and multi-voxel pattern analysis, and that they can occur in
the absence of overall activity changes. We trained
participants over 4 days to produce 4 fast sequences that
are comprised of a series of 5 finger presses in different
orders. The sequences were matched for difficulty and
force, and each sequence included each finger once.
Training decreased movement times from 2s to 1.1s over
the four days. The learning generalized partially to untrained
sequences, as well as to the other hand. Above and beyond
this general learning we found a sequence-specific
improvement of 280ms for the trained sequences only. To
study neuronal representational changes participants
produced trained and untrained finger sequences while
undergoing fMRI. Due to the matching of sequences, the
overall activity level elicited by each sequence was nearly
identical. Using local multi-voxel pattern analysis, however,
we could detect regions in which the fine-grained activation
patterns differed significantly between different sequences.
We could show that the right motor cortex, and bihemispheric premotor, supplementary motor and parietal
areas encode sequential aspects of left hand actions. We
also show that a single behavioral confound, such as force
or movement speed alone cannot explain the classification
results. We then predicted that the natural skill level of
participants to produce fast finger sequence movements
should correlate with our ability to decode these sequences
from the activity patterns in cortical motor areas. Indeed,
faster participants showed clearer differences between the
activity patterns for different sequences, even if they did not
perform faster during the scan. Interestingly no such
correlations could be found between performance and the
overall BOLD signal. Comparing the classification
accuracies for trained sequences with those for untrained
sequences, we also show that motor training leads to
specific increases in decoding accuracy in the
Poster Sessions
Full Abstracts
supplementary motor area (SMA). This result suggests that SMA
contains highly specialized neuronal units, which represent
specific chunks of the trained sequences and therefore underlies
the sequence specific-learning effects. No changes in overall
activity were found in SMA. Our results therefore provide a novel
method for assessing the development of skilled representations
in the human brain.
1-A-27 The effect of acquisition of an internal forward
model on an exploration task
Maria Dagioglou1, Joaquin Bugella1, Tom Walton2, Tom Stafford2,
Peter Redgrave2, Chris Miall1
1
University of Birmingham, 2University of Sheffield
Fast and accurate execution of movements is achieved to a large
extent because the brain is able to predict their outcome. The
cerebellum is believed to contain internal forward models that
simulate our movements and calculate the predicted outcome. In
parallel, the acquisition and selection of actions is believed to
depend on the basal ganglia, which also subserve reinforcement
learning. We hypothesize that participants can perform better in a
novel exploration task (e.g. a task where detection of a target is
via a "tool" whose dynamics are uncertain), if they can first learn
the relationship between movements and outcomes. If supported,
this hypothesis suggests cerebellar learning of forward models
can be exploited by reinforcement-based learning within the basal
ganglia. Participants used a joystick under two different kinds of
trials: exploration trials and tracking trials. During exploration,
participants were asked to find two hidden targets and hit them
alternately, as many times as possible. The workspace flashed
150 ms after a target was hit, imposing a delayed relationship
between current action state and reward. During these trials
participants had no visual feedback of their position in the
workspace. In the tracking trials the subjects were asked to follow
a continuous, pseudorandomly moving target. Their position in
the workspace was displayed as a cursor representing either the
current position of the joystick (Group A, n=10) or its position
delayed by 150 ms (Group B, n=10). Block 1 consisted of sixteen
30-s exploration trials with new random target positions in each
trial. In each of Blocks 2-8, participants completed ten 30-s
tracking trials, followed by two exploration trials. During the final
Block 9 participants again performed 16 exploration trials, with
150 ms delayed reward. An Index of Performance (IoP) was
calculated for each of the exploration trials, equal to the number
of hits per trial over the total distance travelled in each trial
(normalized by the separation of the targets). The results showed
a significant interaction of Block x Delay (F(1,18)=5, p=0.038).We
observed an increase of the mean IoP in Group B (delay in both
exploration and tracking trials) and a decrease in Group A (delay
only in the exploration trials). The preliminary results support our
hypothesis that experience in delayed tracking (Group B) would
improve performance in the delayed exploration task. Hence
adapting the state estimate of movement outcome, putatively in
the cerebellum, facilitates reinforcement learning, putatively in the
basal ganglia. Funded by the Wellcome Trust and EU-fP7-ITN
"C7"
1-A-28 Transfer of ballistic motor skills between bilateral
and unilateral contexts in young and older adults: Neural
adaptations and behavioural implications
Mark Hinder1, Timothy J Carroll2, Jeffery J Summers1
1
University of Tasmania, 2University of Queensland
Bilateral movement rehabilitation therapies are gaining popularity
and aim not only to improve the recovery of bimanual actions, but
also aim to improve unilateral motor functions. Despite this, the
neural mechanisms mediating the transfer of bilateral training
gains into unimanual task contexts are not fully understood.
Converging evidence from behavioural, neurophysiological
and imaging studies suggests that bimanual movements are
not simply the superposition of unimanual tasks undertaken
with both (upper) limbs. However, bilateral practice may
release inhibition within both hemispheres, conceivably
facilitating subsequent unilateral actions. The current study
was designed to investigate the neural mechanisms
associated with a bilateral ballistic motor task, and to assess
the extent to which improvements in performance
transferred to a unimanual task. We have previously used
this motor task to investigate the transfer of practice related
gains between the hands. The task is appealing as training
related improvements are mediated substantially within
primary motor cortex, and thus can be assessed by way of
transcranial magnetic stimulation (TMS). Moreover, recent
work has shown that the mechanisms mediating
performance gains in ballistic motor task are akin to those
mechanisms mediating strength gains during resistance
training. Information from such experiments is therefore
applicable to rehabilitation contexts where, following stroke
or limb immobilisation, muscle atrophy and loss of strength
are ubiquitous. Young (n=9; mean age 19.4 years) and
older (n=9; 66.3 years) adults were trained in a bilateral
motor task in which they were required to simultaneously
abduct their left and right index fingers as quickly as
possible with visual feedback of task performance. Before,
during and after bilateral training we assessed performance
in a bimanual and unilateral task context (in the absence of
feedback) and measured corticospinal excitability and
intracortical inhibition using TMS. We observed a high
degree of transfer between the bimanual and unimanual
contexts for both groups, i.e., bimanual training resulted in
significant performance gains in both bimanual and
unimanual tasks. However, regression analyses revealed
that only for the older adults did the bimanual training gain
predict subsequent unimanual performance for either hand.
Task-related increases in corticospinal excitability and
releases of inhibition (which we have previously consistently
shown in both hemispheres following unilateral ballistic
motor tasks) were minimal and only observable for older
adults. Overall, the results indicate a greater overlap of the
neural mechanisms mediating bilateral and unilateral
ballistic motor learning in older adults.
1-A-29 Risk-sensitivity in motor learning
Michael Trent1, Alaa A Ahmed1
1
University of Colorado
Not all movement errors are created equal. For example,
compare a 5 cm error in foot placement when approaching
the edge of a curb, to the same error approaching the edge
of a cliff. One would likely avoid the cliff edge more than the
edge of the curb. Thus, movement planning would be risksensitive and depend on the subjective value of an error,
rather than its actual value. Interestingly, models of
movement adaptation have traditionally assumed that
adaptation is proportional to movement error1. In recent
years the notion of proportionality been challenged2,3.
However, the role of risk-sensitivity, which emerges from a
distortion between the subjective and actual value of an
error, has not been investigated in movement adaptation.
Here we quantified adaptation in a unique cliff-like virtual
environment that was stable, but only within certain limits, to
influence the subjective value associated with a given
movement error. Seated subjects (N=8) made reaching
movements to a target directly ahead of them while grasping
the handle of a robotic arm. Feedback was presented on a
monitor at eye-level. The protocol consisted of: 50 no-force
11
Poster Sessions
Full Abstracts
trials, 200 in a viscous curl-field (stable condition), 450 in an
identical curl-field with a cliff-like region of instability (unstable
condition), and 50 no- force. Curl-field gains varied from trial to
trial and were biased to the left. During the unstable condition,
curl-field dynamics were maintained but rightward errors greater
than 2.5cm were penalized, with a line indicating edge of the
boundary. In the first 50 trials of the unstable condition a strong
rightward force was applied if the cursor crossed the cliff. After
these 50 trials, this force was de-activated leaving only an
audiovisual warning should the subject cross the boundary.
Importantly, because most movement errors were less than 2.5
cm, subjects were merely alerted to the presence of the cliff. To
adapt subjects must produce a force towards the instability. We
hypothesized that errors closer to the edge would be penalized
more heavily than errors of the same magnitude that occurred
when the instability was not present. This would lead to reduced
adaptation to the stronger gains. Adaptation was calculated in
both conditions using previously developed methods2. Movement
error was similar across conditions (P > 0.1027). Adaptation to
the strongest perturbation was greater in the stable compared to
the unstable condition (P = 0.0050), but similar for the remaining
gains. Model results corroborate these findings. Sensitivity, fit
using a state-space model2, was larger in the stable compared to
the unstable condition only for the largest gains (P < 0.01), and
similar for all others (P > 0.123). This asymmetry suggests that
subjects were responding to the fact that over-compensation was
penalized only for rightward errors. These results imply that
movement adaptation is risk-sensitive. It is not solely dependent
on movement error, but can be modulated by the subjective value
associated with the error. 1. Scheidt et al. (2001) Learning to
move amid uncertainty. Journal of Neurophysiology. 2. Fine and
Thoroughman (2007) Journal of Neurophysiology. 3. Wei and
Kording (2009) Journal of Neurophysiology.
1-A-30 Our brain sees and learns during fast eye
movements
Muriel Panouilleres1, Valérie Gaveau1, Christian Urquizar1, Denis
Pelisson1
1
CRNL, INSERM U1028
Motor adaptation relies on our ability to detect errors when we
perform actions and to correct them with subsequent movements.
To study adaptation of saccadic eye movements in human, a trick
is used to create an artificial saccade inaccuracy, thanks to the
double-step target paradigm (McLaughlin, 1967). This paradigm
consists in stepping the target a first time to trigger the saccade
and a second time to a new location during the eye movement.
Because of saccadic suppression, subjects are usually unaware
of the second intra-saccadic target step. It is thus thought that the
post-saccadic error between eyes and target positions is the
signal necessary to recalibrate the saccadic system (postsaccadic hypothesis, e.g.: Bahcall et Kowler, 2000). In a previous
study, we showed that very short post-saccadic visual information
(stepped target visible only for saccade duration 15 ms) was
sufficient for inducing optimal saccadic adaptation (Panouillères
et al, 2011). Moreover, Gaveau et al. (2003) showed that some
retinal information is processed during large eye movement.
Based on these two previous studies, we then made the
assumption that our brain could process and learn from an error
available only during the saccade, but not after its completion
(intra-saccadic hypothesis). To test this hypothesis, we used a
modified double-step target paradigm in which the second
stepped target was switched off at the end of saccades. In two
separate experiments, the intra-saccadic target step was directed
toward the fixation point to induce a decrease of saccade
amplitude or away from the fixation point to produce an increase
of saccade amplitude. Twenty subjects took part to this study, and
were randomly assigned to one of the two experiments (10
12
subjects each). Each experiment comprised 3 recording
sessions, performed with at least 5-7 days in-between. In
the intra-saccadic adaptation session, we tested our intrasaccadic hypothesis by stepping the target at the beginning
of the saccade and then switching it off at the end of it. In
the post-saccadic adaptation session, the stepped target
was presented for the same duration as in the intra-saccadic
adaptation session, but immediately after the end of the
saccade. According to the classical post-saccadic
hypothesis of error processing, this session should provide
the largest amount of saccadic adaptation. In the control
session, the target did not jump, but remained at its location
for the duration of the saccade and was then turned off. For
the first experiment, we found that the same amount of
saccadic adaptation was achieved for the intra-saccadic and
the post-saccadic adaptation sessions. Even more
interestingly, this adaptation was highly similar to the one
we found in our previous study, where the error signal was
presented for the duration of the saccade plus 15ms in the
post-saccade period. This suggests that our brain can
process error signals during an eye movement and that it
can learn from this error. Preliminary data in the second
experiment suggest that this conclusion can be extended to
large errors requiring an increase of saccade amplitude.
B - Integrative Control of Movement
1-B-89 A study of the influence of biomechanics on
decisions between reaching movements
Ignasi Cos1, Paul Cisek1
1
Université de Montreal
There is still considerable debate about the influence of the
arm's biomechanical properties on the planning of voluntary
movements. While several models describe reach planning
as primarily kinematic, some studies have suggested that
implicit knowledge about biomechanics may also exert
influence on the planning and preparation of reaching
movements. Here we present the results of three related
experiments showing first that biomechanics and factors
related to movement stability are predicted during motor
planning, because they influence choices between voluntary
movements, and second, that these influence the choice at
different latencies. In the first experiment, human subjects
made free choices between two potential reaching
movements, each defined by the origin, a via-point and a
target, varying in path distance and biomechanical factors
related to movement energy and stability. Our results show
that subjects preferred movements whose final trajectory
was better aligned with the direction of lowest
biomechanical cost, even when the launching properties
were very similar. This reveals that the nervous system can
predict biomechanical properties of potential actions prior to
movement onset and that these predictions, in addition to
purely abstract criteria, exert an influence on the decisionmaking process. Next, we performed a second set of
experiments to test whether cost factors related to arm
stability were also predicted and whether they exerted a
bias on reach selection. To this end, we quantified the
modulation of the baseline pattern of choices of each
subject, as a function of two additional factors related to the
ease of control of the end-point in the vicinity of the target:
the ease of aiming and the requirement of stopping. The
results show that the strongest preference for the low
biomechanical cost choice appears in the absence of these
Poster Sessions
Full Abstracts
constraints. As the control constraints are gradually imposed, this
preference is gradually reduced. This shows that factors related
to movement stability modulate the influence of biomechanical
cost on the subjects' choices already during movement
preparation, prior to movement onset. Finally, our last set of
experiments examined the latencies at which different factors
influence motor decision-making prior to movement onset.
Hypothetically, if the encoding of factors such as distance or
direction is attributed to the fronto-parietal loop and the
computation of biomechanics to the cerebellum, their latencies to
reach the motor plan should be significantly different. To test this,
we constrained the time subjects could observe the targets and
via-points prior to movement onset to a range between 0.2 and
1s. Unexpectedly, the subjects' choices showed that the influence
of biomechanical factors precedes the influence of path distance,
therefore suggesting that the calculation of biomechanical factors
may take advantage of priors including knowledge of
biomechanics along different arrival directions. In summary, these
three experiments provide insights into decision-making between
reaching movements, demonstrating that: 1) biomechanics and
stability are two influential factors, even prior to movement onset;
and 2) motor decision-making is a gradual process, first
incorporating intrinsic and stability properties and secondly
additional kinematic factors such as path distance. Presumably,
this process yields choices that are most comfortable and easiest
to control. Support: NSERC, CRCNS.
1-B-90 Measurable improvements in dexterous
manipulation throughout adolescence reveal previously
undetected functional effects of neuromaturation
Sudarshan Dayanidhi1, Åsa Hedberg2, Francisco J ValeroCuevas1, Hans Forssberg2
1
University of Southern California, 2Karolinska Institutet
Neural control of dexterous manipulation is attributed to specific
neuroanatomical structures whose connectivity and function are
known to have a prolonged period of development into late
adolescence[1-3]. In contrast, functional improvements in
dexterous manipulation--as measureable by current
developmental and clinical milestones--show few changes past
the age of eight[4] because most measures of hand function
saturate. We now show that an extension of our prior work[5, 6]
bridges this apparent discrepancy and establishes a novel and
clear link between known neuroanatomical development and
dexterous manipulation well in to late adolescence. Importantly,
musculoskeletal growth and strength are poorly correlated with
these functional improvements in dexterity. These results begin to
clarify the behavioral benefits of such neural maturation, enable
the systematic study of specific neuroanatomical structures, their
connectivity, and plasticity. For example, neuroimaging studies to
disambiguate the differential roles and contributions of maturation
of the corticospinal tract vs. the emergence of fronto-parietal and
cortico-striatal-cerebellar networks. Clinically, this extends the
ages for which therapeutic interventions can be considered
fruitful, and provides a clinically-practical means to chart
functional development of dexterous manipulation in typically
developing children, and children with neurological conditions.
References 1. Paus, T., et al., Structural maturation of neural
pathways in children and adolescents: in vivo study. Science,
1999. 283(5409): p. 1908-11. 2. Lebel, C., et al., Microstructural
maturation of the human brain from childhood to adulthood.
Neuroimage, 2008. 40(3): p. 1044-55. 3. Muller, K., V. Homberg,
and H.G. Lenard, Magnetic stimulation of motor cortex and nerve
roots in children. Maturation of cortico-motoneuronal projections.
Electroencephalogr Clin Neurophysiol, 1991. 81(1): p. 63-70. 4.
Forssberg, H., et al., Development of human precision grip. I:
Basic coordination of force. Exp Brain Res, 1991. 85(2): p. 451-7.
5. Vollmer, B., et al., Evidence of validity in a new method for
measurement of dexterity in children and adolescents.
Developmental Medicine & Child Neurology, 2010. 52(10):
p. 948-954. 6. Valero-Cuevas, F.J., et al., The strengthdexterity test as a measure of dynamic pinch performance. J
Biomech, 2003. 36(2): p. 265-70.
1-B-91 Identification of human limb impedance in 5
DoF
Patrick van der Smagt1, Dominic Lakatos1, Daniel Rüschen1,
Jörn Vogel1
1
DLR / RM
Dynamic interaction with the environment means handling
impacts and unknown contact forces. Therefore compliant
systems are active topics of research in the field of robotics.
Surpassing traditional rigid robots, the control loops of
modern robotic systems are extended with additional
impedance parameter, viz. stiffness and damping.
Even though the implementation of impedance control in
robotics is resolved to a large part, one important issue still
needs to be resolved: how are the impedance parameters
set to optimally perform a predefined task? Traditionally,
robotic tasks are only defined in target end-effector positions
or, in some cases, end-effector trajectories; but the
impedance around these positions or trajectories remains a
matter of common sense, at best. For instance, when
performing a peg-in-hole task, high stiffness in the
perpendicular and low stiffness in the lateral directions, so
as to allow for imprecise positioning while solving the task,
appears to be useful. But how do we find general rules-ofthumb for setting these extra parameters?
Beside heuristic methods tuning the impedance parameters,
mimicking the behaviour of the human arm is an auspicious
field of research, and leads to what we call biologicallyinspired robotics. By measuring and subsequently analysing
human arm impedance parameters, we can attempt to
extract general rules and project these to the robotic
domain.
The only direct method to measure stiffness in a functioning
feedback system is to apply external force perturbations to
the limb and to measure the resulting displacements; such
measurements have only been satisfactorily realised in
planar (2D) movements. To date, no fully satisfactory
method exists to investigate the time-varying impedance
during movements. Early efforts were subject to error
because they assume that subjects perform the same
movement on repeated trials and they ignore the non-linear
properties of the musculo-skeletal system.
The human arm's capability to alter its impedance has
motivated multiple developments of robotic manipulators
and control methods. It provides advantages during
manipulation such as robustness against external
disturbances and task adaptability. However, how the
impedance of the arm is set depends on the manipulation
situation; a general procedure is lacking.
We provide a method to identify human arm impedance in
more than 2 degrees of freedom. We do this by initially
identifying the kinematic and inertial parameters of the arm
through movement. Subsequently we identify stiffness
parameters of the human arm in 5 degrees of freedom
(shoulder, elbow, and lower arm rotation), while taking the
numerical stability of the data into account. Confidence
criteria to determine the accuracy of the estimated
parameters are given. The data are related to a
representation of the stiffness by EMG which, in
13
Poster Sessions
Full Abstracts
combination with the kinematics, gives us a 3D Cartesian
identification of the impedance parameters of the human arm.
1-B-92 Simulated hemianopia drives eyes to distraction
Liana Brown1, Carina La Mantia1
1
Trent University
We used gaze contingent displays to explore how people adapt
their visual exploration and reaching behaviour to simulated
homonymous hemianopia (SH). Participants were asked to
search for and point to a target displayed with distractors, and
gaze location at reach onset was measured. Previously, we found
that participants tended to look short of blind-field targets and
reach blind. This study revealed that participants no longer
pointed blind when good-field distractors were removed or when a
hand was placed in the blind-field target to anchor gaze there.
Partial visual loss leaves people vulnerable to distractors, but this
can be reduced somewhat by placing one's hand in the blind field.
1-B-93 Different brain pathways for strategic control
versus sensorimotor recalibration: Evidence from a dual
task reach paradigm
Joshua Granek1, Lauren E Sergio1
1
York University
We have previously demonstrated that patients with optic ataxia
(OA), having damage to their superior parietal lobules, rely on
explicit strategic control when reaching under a 90° cursor
feedback rotation condition (i.e. non-standard, "decoupled"
reaching). We observed that OA patients displayed improved
performance when moving towards ordinal (on-axis) as opposed
to oblique (off-axis) visual targets. We assume that for these
rotated cursor feedback situations a) oblique targets require an
implicit realignment between proprioception and vision, referred to
as sensorimotor recalibration, while b) ordinal targets rely on the
use of explicit rule integration, referred to as strategic control. Our
OA patient data suggests that the network used for sensorimotor
recalibration is to some extent independent of that used for
strategic control. This observation motivates the general question:
Do neurologically intact adults have difficulty with non-standard
reaching when strategic control, but not sensorimotor
recalibration, has been interfered with? If so, this would support
the notion of independent brain pathways for the processing of
these different types of visuomotor mappings. Here, we trained
healthy adults on visually-guided reaching with both veridical and
rotated (90° clockwise rotation) cursor feedback, while
simultaneously performing an attentionally- demanding task
(backwards counting by different amounts). We tested the
hypothesis that different pathways are used for these two classes
of movement control by increasing the neural load associated
with explicit rule integration, but not implicit visual-proprioceptive
alignment. We predicted that performance in the dual task would
decline when making a non-standard reach towards ordinal
targets (where explicit rules are more useful) relative to oblique
targets (relying on sensorimotor recalibration).
Overall, simultaneous performance of the two tasks had
detrimental effects on hand movement timing (increased
movement timing), accuracy (increased constant error) and
precision (increased variable error). Importantly, the time to
prepare the reaches having rotated visual feedback (reaction
time) increased when performing the dual task towards only the
ordinal targets, relative to controls. Similarly, subjects' trajectories
for reaches having rotated feedback were longer (increased path
length) and were initiated towards the wrong direction (increased
angle at peak velocity) during the dual task towards the ordinal
targets, while they remained unaffected relative to the controls
towards the oblique targets. These observations support our
14
assumption that certain types of non-standard reaches rely
on explicit strategic control, while others rely on implicit
sensorimotor recalibration. Our results suggest that
independent neural pathways may underlie the control of
these different types of reach, since one class of movement
was interfered with while the other was not.
1-B-94 Separating standard and non-standard
reaches: Topographical differences within PMd
Patricia Sayegh1, Kara M Hawkins1, Lauren E Sergio1
1
York University
It is well established that reaching movements rely on a
network of brain regions including the dorsal premotor
(PMd) and superior parietal lobule (SPL), which are regions
located within the dorsomedial parieto-frontal network.
However, what is less understood is how the regions within
this network are modulated during a reaching movement
when there is a dissociation between the action of the eye
and the hand, termed a non-standard movement. There is a
large body of evidence to suggest that the action of the eye
and hand are tightly linked (1-3), that decoupled eye-hand
coordination is not innate(4), and that the accuracy and
movement profile of a non-standard reaches are altered
when compared to standard reaches (5). Additionally,
patients with neurological disorders demonstrate
deteriorated reaching performance on non-standard tasks
while leaving reaching performance on standard tasks
unaffected (6). This suggests that during non-standard
reaching movements, a specific set of processes must occur
in order to break the tight linkage between the eye and the
hand. As a result, non-standard reaching movements likely
depend on neural circuitry that is different albeit
interconnected with the circuitry important for controlling
natural (standard) reaching movements (7). We recorded
the local field potential (LFP) within PMdr and PMdc during
the planning phase of two types of visually-guided reach
movements. During the standard condition, a visually guided
reach was performed in which the visual stimulus guiding
the movement was the target of the reach itself. During the
non-standard condition, the visual stimulus provided
information about the direction of a required movement, but
was not the target of the motor output. We observed distinct
task related differences as well as topographical differences
between PMdr and PMdc. Our results support functional
differences between PMdr and PMdc during visually-guided
reaching. PMdr activity appears more involved in integrating
the rule-based aspects of a visually-guided reach, while
PMdc is more involved in movement related activity. More
broadly, our results highlight the necessity of accounting for
the non-standard nature of a motor task when interpreting
movement control research data.1. Henriques, et al. J.
Neurosci. (1998). 2.Gorbet, L. E. Sergio. Brain Res. (2009).
3. Neggers et al. J. Neurophysiol. (2000). 4. Bo, et al. Hum.
Mov. Sci. (2006). 5. Messier, et al. Exp. Brain Res. (1997).
6. Tippett, L. E. Sergio. Brain Res. (2006). 7. Clavagnier, et
al. Neuroscientist (2007).
1-B-95 Mapping the spatio-temporal structure of
motor cortical LFP and spiking activity during reach
and grasp movements
Alexa Riehle1, Thomas Brochier1
1
CNRS
Grasping an object involves shaping the hand and fingers in
relation to the object's physical properties. Following object
contact, it also requires a fine adjustment of grasp forces for
secure manipulation. Earlier studies suggest that the control
Poster Sessions
Full Abstracts
of hand shaping and grasp force involve partially segregated
motor cortical networks both during preparation and execution.
However, it is still unclear how information originating from these
networks is integrated over motor cortical areas. We addressed
this issue by analyzing massively parallel activity of LFPs and
single neurons recorded in primary motor (MI) and dorsal
premotor (PMd) cortices of macaque monkeys performing a
delayed reach to grasp task. They are trained to grasp and pull an
object using either a Side Grip or a Precision Grip. The object is
either heavy or light. The trial starts by the monkey pressing a
switch. A first cue provides the instruction about the grip (object
load). After a delay of 1s, a second cue provides additional
information about the object load (grip) and serves as GO signal.
The monkey has then to release the switch, grip the object, pull it
and hold it in a narrow position window for 500ms to receive a
food reward. Neuronal activity was recorded in two monkeys by
using a 100 electrode "Utah" array, chronically implanted at the
MI/PMd border. In motor cortical areas, LFPs exhibit a large multicomponent movement-related potential (MRP) around movement
onset. However, little is known about its spatio-temporal
distribution across motor cortex in relation to task requirements.
Our data show that MRP amplitude and shape depend both on
grip type and electrode position. The strength and the generation
of each of the 5 individual MRP components depend on the
location on the cortical surface. The first two components, P1
occurring just before and N1 around movement onset, were
related to reaching and generated in the proximal representation,
towards the precentral dimple, whereas the later components, P2,
N2 and P3, were closely related to grasping and generated in the
distal representation, the most posterior part of the array towards
the central sulcus. Single neuron activity related to the four trial
types was analyzed during movement preparation and execution.
During both periods, a much larger proportion of neurons was
selective to the grip than to the object load. During movement
preparation, the highest amount of selective grip- and forcerelated activity was found in PMd, and the smallest amount in the
proximal representation of MI. Grip-related activity during
movement execution was widely distributed over the array, but
highest at the location of distal representation, thus overlapping
with that of the late MRP components of the LFP. A close
matching was observed between the distribution of distal
somatosensory receptive fields on hand and fingers and the
distribution of grip-related activity of both LFP and single neurons.
These data show that in motor cortex, a precise spatio-temporal
pattern of activation is involved for the control and hand position
and force during the preparation and execution of reach to grasp
movements. These data provide some new insight about the
functional organization of motor cortex for the control of hand grip
and contact forces during object grasp and manipulation.
Funding: Collaborative Research Agreement Riken-CNRS, CNRS
(PEPS, Neuro_IC2010)
1-B-96 Reach to grasp movements: A combined EEG
and kinematic study
Teresa De Sanctis1
1
Università degli studi di Padova
Introduction: Reach-to-grasp movements have been widely
investigated in both humans and monkeys with a variety of tasks
and techniques (Filimon, 2010; Grafton, 2010). These studies aim
to integrate information from various domains to reveal the neural
circuits that underlie reach-to-grasp movement. The fast
development of techniques might allow researchers to investigate
how the various grasping related areas in the human brain act in
concert (Castiello & Begliomini, 2008). To address this issue here
we co-register electrical and kinematic signals with the intent to
ascertain how and whether they show some form of correlation
The combination of different techniques offers the unique
opportunity to understand more fully the processes
underlying reach-to-grasp movements.
Methods: Participants. Fourteen right-handed participants
(9 female; age 19-27 years old) participated in the study.
Apparatus. The electroencephalographic activity was
recorded by means of a 30-channel set up placed on the
scalp according to the international 10-20 system.
Movement kinematics was recorded by means of a 3D
motion analysis system (SMART-D, BTS Bio Engineering).
Stimuli. Two spherical wooden objects (large object, 7 cm
diameter; small object, 3 cm diameter), placed at 30 cm
along the participants' mid sagittal plane. Procedures.
Participants were instructed to reach toward and grasp the
object. The action started at the time the stimulus became
visually available. Vision was controlled by means of liquid
crystal goggles worn by participants. There were two
experimental conditions: reach-to-grasp towards the small
object, and reach-to-grasp toward the large object. Forty
trials per condition were administered in a semi-random
order. Data analysis. An analysis of variance with type of
stimulus (small, large) as within-subject factor was
conducted for several dependent measures. For the EEG
measures, we considered amplitude and latency of eventrelated potentials (ERP) for each channel. For the
kinematics we considered key reach-to-grasp kinematic
landmarks. They were time to (i) peak velocity; (ii) peak
acceleration; and (iii) peak deceleration for the reaching
component. And, the time of maximum grip aperture (the
time at which the maximum distance for the markers placed
upon the thumb and index finger occurred) for the grasping
component. Furthermore, ERP components were correlated
with kinematic landmarks.
Results: Object size affected both EEG and kinematic
signals. The ERP component related to movement planning
was evident as a negative deflection frontally distributed,
and was differentially modulated in amplitude and latency.
At parietal sites, a sustained positive ERP variation was
found, likely related to visuo-spatial processing. Both these
components were evoked by the object appearance. As
expected, there were kinematic differences depending on
stimulus size. Concerning the approach phase, time to peak
velocity occurred earlier and the time to peak deceleration
occurred later for the small than for the large stimulus. For
the grasping component, the time of maximum grip aperture
occurred earlier for the small than for the large stimulus.
ERP components and temporal kinematic events correlated
depending on object size.
Conclusions: These findings provide the first evidence of
how kinematic landmarks characterizing reach-to-grasp
movements are synchronously related to electrical neural
activity recorded within the fronto-parietal network.
1-B-97 On the relationship between spiking activity
and low-frequency local field potentials in primate
motor cortex
Thomas Hall1, Kianoush Nazarpour1, Andrew Jackson1
1
University of Newcastle-upon-Tyne, UK
There is increasing interest in the low-frequency (0.5 to 5 Hz
or 'delta band') components of the local field potential (LFP)
in motor cortex. Recent studies have demonstrated that
movement-relevant information can be decoded from these
frequencies, with application in brain-machine interfaces for
restoring motor function. However it is not yet clear how
these LFP components relate to the spiking activity of
neurons in the motor networks.
15
Poster Sessions
Full Abstracts
We have recorded simultaneous spikes and LFPs using chronic
intracortical electrode arrays in the primary motor (M1) and
ventral premotor (PMv) areas of awake macaque monkeys
performing wrist torque-tracking and brain-machine interface
tasks. Spike-triggered averages (STAs) of LFPs exhibited lowfrequency (1 to 2 Hz) features that have not previously been well
described. These features have a remarkably consistent
morphology in two subjects across a variety of task
configurations, consisting usually of a biphasic potential with initial
negativity at latencies between -200 and 200ms, followed a
positivity between 100 and 500ms. Unlike higher frequencies in
the beta band, these low-frequency features do not correspond to
precise phase-locking of individual spikes to a global rhythm. By
cross-correlating event-related potentials with peri-event time
histograms, we further showed that the spike-field relationship is
not explained by covariation with motor- or sensory-evoked
potentials but instead reflects slow, stochastic fluctuations in
spike rate and LFP.
distance as fast as possible without missing and sustain the
cursor in the target for 0.5 s. The relationship between
movement time and target difficulty was calculated. Fitts
Law predicts that movement time will be linear in the log of
the target difficulty. Preliminary results show that subjects
are able to adapt to the task quickly, and their movement
times after practice are well described by Fitts' Law. In
ongoing work we will compare the calculated index of
performance for myocontrol with the index of performance
for a similar task based on position or force control of the
index finger.
1-B-99 When to move and how to move:
Information about the movement type and the timing
of movement are concurrently processed in the
human motor system
Nobuhiro Hagura1, Yosuke Goto2, Michikazu Matsumura2
1
We studied the localisation of spike-related low-frequency LFP
components by comparing STAs for spikes and LFPs recorded on
the same electrode, pairs of electrodes within the same cortical
area and across different areas. In general, the amplitude of lowfrequency features tended to reduce with increasing electrode
separation, suggesting they reflect local sources. Moreover the
shape and latency of features in STAs triggered by the same cell
differed significantly for LFPs across and, to some extent, within
areas suggesting multiple, distributed sources may correlate with
the recorded neuron. Using a double-spike triggered averaging
technique with pairs of simultaneously recorded neurons, we
found that each cell contributes an independent component to the
LFP at a third site. Extending this to multiple units, we attempted
to predict the low-frequency LFP from a convolution of each spike
train with a unique kernel. We found that the LFP was welldescribed by a linear sum of individual contributions, with the
goodness of fit improving with increasing numbers of units.
The LFP is often conceived as a global phenomenon that drives
synchronous firing of broad neuronal populations. By contrast, we
hypothesise that the low-frequency LFP is the summation of
contributions from a large number of relatively independent
sources, each reflecting the correlated firing of a local or
distributed neuronal ensemble. Because multiple electrodes will
sample a slightly different combination of sources, this may
explain why simultaneous LFP recordings can be decoded to
yield a wealth of movement-related information at these
frequencies.
1-B-98 Nonlinear EMG estimation as a control signal in a
one-muscle myocontrol task
Adam Feinman1, Terence D Sanger1
1
University of Southern California
Nonlinear filtering of surface EMG (sEMG), using a recursive
Bayesian method sidesteps some of the issues in using sEMG as
a control signal, such as the trade-off between smoothing and
signal delays. The use of nonlinear filters raises the possibility of
using myocontrol for precision tasks for which sEMG has
traditionally been considered too noisy. In order to quantify the
control abilities of sEMG in terms of both precision and speed of
responses, we test using a Fitts Law paradigm and calculate the
index of performance for myocontrol. We have implemented this
methodology in a simple one-muscle myocontrol task to examine
a person's ability to isolate control of one muscle and to
determine the bandwidth of the control signal. Subjects were
asked to control the FDI muscle of their dominant hand with visual
feedback of filtered sEMG. Vertical position of a bar cursor was
proportional to the level of contraction of the FDI. Subjects were
instructed to move the cursor into targets of varying size and
16
University College London, 2Kyoto University
On the tennis court, when waiting for the opponent to serve,
you have to prepare for a movement by taking 1) the kind of
movement type (forehand or backhand), and 2) the timing of
movement initiation (when to swing), into account.
Intuitively, the timing of movement initiation cannot be
prepared before deciding what kind of movement to make.
Here, we show that two information are concurrently
processed during motor preparation, both giving benefit for
efficient action preparation by each independently
modulating the motor system. 14 subjects participated in the
study. In a trial, an arrow indicating either left of right was
presented on a monitor (GO signal). Subjects made flexion
movement of their right wrist for the left arrow and extended
for the right in a ballistic manner. In each block, the
presentation ratio of left and right arrows was changed (7:1,
3:1, 1:1, 1:3, 1:7), enabling the subjects to estimate which
movement (flexion or extension) is more likely to be
performed during that block. In each block, for half of the
trials, a warning signal was presented 500ms before the GO
signal. This allowed the subjects to predict the initiation
timing of the movement. To evaluate the cortico-spinal
excitability during motor preparation, single pulse
transcranial magnetic stimulation (TMS) was applied to the
left primary motor cortex 50~500ms before the GO signal,
and motor evoked potentials (MEP) were measured from
flexor carpi radialis (FCR) and extensor carpi radialis (ECR).
Reaction time (RT) was reduced both by information about
the movement type and the movement timing, without
showing any effect of interaction. This indicated that two
information independently acted on the motor system for
efficient motor preparation. This behavioural pattern of
independent information processing was also observed as
modulation of MEP. Overall amplitude increased depending
on the probability of the movement type (effect of movement
type), but on top of that, reduction of MEP towards the GO
signal was observed when the warning signal was
presented (effect of movement timing). Critically, both
component of MEP modulation directly explained the
variance of RT; reduction of individual RTs were able to be
explained by the increase in MEPs by the probability of
movement, and also by the amount of reduction of MEP due
to the warning signal. Our result of increased MEPs
depending on the probability of movement resembles with
the notion that selection of movement is achieved through
competition of neuronal firing that can be biased by the
probability of future movements. Also, as in the present
study, MEP, including the excitability of spinal motoneurons,
is reported to reduce when the warning signal is presented,
suggested to function for increasing the sensitivity to the
Poster Sessions
Full Abstracts
descending motor command at the critical timing. We propose
that the information about movement types and movement
timings are concurrently processed, possibly in different brain
regions, acting on the cortico-spinal pathway from the motor
cortex at different stages. Even when you are not sure which
swing to make, we can still prepare for the timing of the swing.
reward, TMS and compatibility on the independent variables
reaction time and accuracy were tested in a 2 × 2 × 2
repeated measures analysis of variance (ANOVA).
Accuracy rates were analyzed separately for fast and slow
responses after a median split to assess the effects of rTMS
on premature incorrect responses.
1-B-100 Hand-related sensory-motor activity in secondary
somatosensory cortex of the macaque monkey
Results: In reward trials, participants significantly increased
reaction speed compared to neutral trials (F(1,13) = 16.65, p
= 0.001). For trials following an incompatible trial, fast
responses were more error-prone than for trials following a
compatible trial (F(1,13) = 8.321, p = 0.013). Real rTMS
over pre-SMA reduced subjects' tendency to make such
premature incorrect responses, but only in the context of
reward (F(1,13) = 4.711, p = 0.049).
Laura Grandi1, Hiroaki Ishida2, Luca Fornia1, Vittorio Gallese2
1
University of Parma, 2Italian Institute of Technology (IIT), Unit
Parma
Dexterous use of the hand requires continuous update of
somatosensory input to the motor system. Both in humans and in
non-human primates, the secondary somatosensory cortex (SII)
has been implicated in sensory-motor integration during active
touch. However, physiological properties of reaching-grasping
movements in this area remain largely unclear. Here we
investigated whether, and to what extent, sensory and visual
information of target can integrate at the single neuron level in the
posterior SII of non-human primate. To explore this issue, two
monkeys were trained to perform a reaching-to-grasping task,
which required three different hand prehension types (grasping
objects from a grove, a cup, a plate) tested both in light and dark
conditions. Our results showed that within SII hand representation
region, although most neurons responded to passive
somatosensory stimulation and during the motor task, a
consistent percentage of them was only activated during
monkeys' active hand grasping/manipulation. Approximately half
of them was active before the onset of object contact.
Furthermore, few neurons showed stronger activation during
manipulation in the dark in comparison with that performed under
visual control. The present findings show the existence of
grasping- and manipulation-related neurons in SII. This suggests
that neuronal activity of SII hand region likely contributes to hand
motor control both in the initial and finale phases of grasping and
object manipulation motor acts. Such motor-related activity could
perhaps also provide the sensory-motor binding principal enabling
the translation of diachronic somatosensory inputs into a coherent
image of the explored object.
1-B-101 Improved control over impulsive but
inappropriate response tendencies following 1 Hz
repetitive TMS of pre-supplementary motor area
Damian Herz1, Hartwig R Siebner1, Mark S Christensen1, Norbert
Bruggemann1, Brian N Haagensen1, Kristoffer H Madsen1
1
Copenhagen Unversity Hospital Hvidovre
Background: In choice-reaction tasks, participants respond
slower and less accurately to an arbitrary cue when the position
of the cue and the response are spatially incongruent. This
response interference caused by spatial incongruency between
cue and response is flexibly modulated by the compatibility
between consecutive trials: Responses are slower and less
accurate if an incongruent trial follows a congruent trial or a
congruent trial follows an incongruent trial. Aim. To study the role
of the pre-supplementary motor area (pre-SMA) in the flexible
control of response inhibition.
Methods: On two sessions, 14 young healthy participants
received either 1 Hz repetitive transcranial magnetic stimulation
(rTMS) to induce a lasting suppression of cortical excitability in
the pre-SMA or sham stimulation. Following rTMS, participants
performed a Simon task in which two arbitrary cues prompted
spatially incompatible or compatible responses. To increase the
motivation to respond, participants were rewarded in 50% of the
trials in case of accurate and fast responses. Main effects of
Conclusion: Suppressing cortical excitability in pre-SMA
leads to a context-dependent improvement in interference
control by enhancing inhibition of premature response
tendencies. This finding suggests a critical role of pre-SMA
in flexible response inhibition.
1-B-102 Intra-hemispheric parietal-motor paired
associative stimulation induces bidirectional
plasticity
Giacomo Koch1, Domenica Veniero1, Francesco Di
Lorenzo1, Carlo Caltagirone1
1
Santa Lucia Foundation
According to the Hebbian rule, synapses increase their
efficacy if the synapse consistently assists the postsynaptic
target neuron to generate action potentials. This type of
plasticity is referred to as spike-timing-dependent plasticity
(STDP). STDP-like plasticity can be induced in the intact
human cortex using paired associative stimulation (PAS).
Conventional PAS protocols consistently pairs electrical
stimulation of the median nerve with TMS of the
contralateral primary motor or sensory cortex (Stefan et al.
2000). Here we sought to investigate whether PAS of intrahemispheric cortico-cortical connections between the
posterior parietal cortex (PPC) and the primary motor cortex
(M1) may induce plastic changes in the excitability of M1.
Ten healthy subjects were enrolled. To activate the
ipsilateral PPC-M1 connection within the left hemisphere, a
conditioning stimulus (CS) was applied over the PPC at an
intensity of 90% of the ipsilateral resting motor threshold
(RMT) (Koch et al., 2007, 2008). The CS preceded or
followed the M1 test stimulus (TS) by 5-50 ms. In each
protocol (PAS-5 and PAS+5) one hundred pairs of stimuli
were continuously delivered at a rate of 0.2 Hz for 8 min.
Each pair of stimuli consisted of a monophasic stimulus
given to the left PPC followed by another monophasic
stimulus given to the left M1. In a second set of experiments
we also evaluated the changes occurring in intracortical
circuits over the target M1 testing both short intracortical
inhibition (SICI) at 2 and 3 ms ISI and intracortical facilitation
(ICF) at 15 ms ISI using the standard paired pulse protocol
(Kujirai et al., 1993). We found that intra-hemispheric PPCM1 PAS was able to induce bidirectional long lasting plastic
changes in the excitability of the target M1. When the PPC
TMS preceded the M1 TS by 5 ms we found evidence for a
long lasting inhibition of the MEP that lasted for up to 20
minutes after the end of the protocol (ANOVA with TIME as
main factor (F(5,45)=3.81; p=0.009), implicating a long-term
depression (LTD)-like phenomenon. On the other hand,
When the PPC TMS followed the M1 TS by 5 ms we found
evidence for a long lasting increase of the MEP that lasted
for up to 25 minutes after the end of the protocol (ANOVA
with TIME as main factor (F(5,45)=2.70; p=0.021),
17
Poster Sessions
Full Abstracts
implicating a long-term potentiation (LTP)-like phenomenon.
These data demonstrate the possibility to explore bidirectional
hebbian-like plasticity between cortical areas of the same
hemisphere. The current protocol could be important to study
plastic mechanisms related to motor learning and to develop
specific rehabilitative approaches
1-B-103 UCM analysis of human multi-joint coordination
and mechanical arm impedance
Shunta Togo1, Takahiro Kagawa1, Yoji Uno1
1
Nagoya University
Humans can naturally perform many types of dexterous tasks. It
is considered that the humans coordinate their redundant multijoints to achieve the dexterous task. We investigated a human
dexterous task "carrying a cup with water without spilling it". This
task requires the individual to dampen hand vibration while
walking, and in generally, the human can naturally perform it.
Moreover, this simply task has an interesting problem of human
control e.g. redundant whole body movement, control of
mechanical impedance of an arm. We hypothesize that the
human reduces the hand jerk (the rate of change of acceleration)
and maintains the cup angle constant by multi-joint coordination
to achieve the task. We conducted an experiment in which human
subjects walked on a treadmill holding a cup with water (WW
task) and with stones (WS task). Their kinematics data were
measured using a three-dimensional position measurement
system. We empirically confirmed that the value of the hand jerk
and the variance of the cup angle in the WW task were smaller
than those in the WS task. The jerks of some body parts were
gradually dampened from a hip to a hand, indicating that the jerk
might be dampened by use of redundant multi-joints. To quantify
the multi-joint coordinated movements, we used UCM
(UnControlled Manifold) analysis. This method divided joint
variance into two orthogonal components: UCM component which
does not affect the hand jerk and the cup angle, and ORT
component which directly affects them. Using these components,
we defined the index of hand jerk coordination and that of cup
angle coordination. According to these indices in the WW task,
we found that the human reduces hand jerk and maintains the
cup angle constant by multi-joint coordination. To consider the
mechanism of coordinated movements especially the mechanism
of reducing hand jerk, we focused on the mechanical impedance
of the human arm. We considered the two-link arm with joint
stiffness and viscosity, and simulated human walking by moving
the origin of the two link arm. In the simulation, we changed the
value of joint stiffness and viscosity, and calculated the hand jerk.
We found that the smaller value of joint stiffness or the larger
value of joint viscosity was assigned, the smaller hand jerk was
appeared. To qualitatively confirm these results, we conducted an
experiment in which subject's hand was perturbed while carrying
a cup. In this experiment, subjects performed the WW task and
the WS task, and a held cup was quickly perturbed by the tension
in the almost same range. We measured subject's kinematic data
and strength of the perturbation. Our results showed that the
displacement of the hand in the WW task was larger than that in
the WS task. These results suggest that the human recovers the
effect of the perturbation by changing the characteristics of
mechanical impedance of the arm. This suggestion also indicates
that the human reduces hand jerk by controlling their mechanical
impedance of the arm.
1-B-104 Two hands, one perception: Bimanual integration
of haptic information
1
1
1
Valentina Squeri , Alessandra Sciutti , Monica Gori , Lorenzo
Masia1, Giulio Sandini1, Juergen Konczak2
18
1
Universitá degli Studi di Genova - DIST-IIT, 2University of
Minnesota
Introduction: Humans routinely use their hands to
haptically explore objects in the environment. In many
cases, both hands are used to gain information about the
properties of the object. However we know very little about
how the brain combines the haptic information of our two
hands to achieve a single percept of an object and about the
underlying mechanism of how the nervous system
integrates or fuses information from two haptic systems.
This study systematically measured the haptic precision of
humans exploring a virtual curved object contour with one or
both hands to understand if the brain integrates haptic
information from the two hemispheres. Evidence on
unimanual visuo-haptic integration showed that vision and
the haptic information from one hand are integrated
optimally. The extension of the Bayesian perception theory
would predict that redundant information from both hands
should improve haptic estimates.
Method: Twenty right-handed adult participated to this
study. A bimanual robotic manipulandum passively moved
the hands of the blindfolded subjects along virtual curved
contours. Passive movements constrained by the robot
assured that both hands performed simultaneous and
synchronous displacements with the same speed profile.
Thus, any possible differences in perceptual sensitivity
between unimanual and bimanual conditions cannot be
explained by differences in the motor performance of the
two hands. In a single trial two different contours were
presented: a standard (constant) and a comparison
(variable) stimulus. Within each trial, the order of
presentation of the two stimuli was random. After probing
both contours, subjects indicated which contour was more
"curved" (forced choice). Participants explored contours in 4
unimanual and 6 bimanual conditions (left, right, both hands
coupled; both hands moving independently). Respective
psychophysical discrimination thresholds were computed.
Results: Analysis of the unimanual left and right hand
exploration showed a tendency for one hand being more
sensitive than the other with most of the subjects exhibiting
a non-dominant (left) hand bias. Bimanual thresholds were
mostly within the range of the corresponding unimanual
thresholds and were not predicted by a maximum-likelihood
estimation model. Moreover, during bimanual exploration
perception tended to be biased towards the motorically
dominant, right hand, not towards the haptically more
sensitive left hand.
Conclusion: Two-handed exploration does not necessarily
improve haptic sensitivity. We found no evidence that haptic
information from both hands is integrated using a maximumlikelihood estimation mechanism. Rather, results are
indicative of a process of "sensory selection", where
information from the dominant, right hand is used, although
the left non-dominant hand may yield more precise haptic
estimates.
1-B-105 Predicting sensory consequences of
intended movements in monkey posterior parietal
cortex
Alexander Gail1, Shenbing Kuang1
1
German Primate Center
The ability to predict the sensory outcomes of our own
actions is considered essential, since it allows to
compensate for system delays during adaptive motor
control. While clinical and neuropsychophysical studies
Poster Sessions
Full Abstracts
suggest that the posterior parietal cortex (PPC) of primates could
be involved in such forward predictions, convincing direct
neurophysiological evidence of predictive signals in PPC is still
lacking. During movement execution such evidence is also very
difficult to provide, since feed-forward motor commands, sensory
feedback about the movement, and putative efferent copy or
forward prediction signals change dynamically and co-vary
substantially. This almost unavoidably leads to various potential
confounds when trying to interpret neural data in highly integrative
sensorimotor areas during movement execution. Beyond
immediate motor control, the encoding of anticipated sensory
effects of an intended movement, prior to execution, has been
postulated as a possible mechanism of action selection and
planning. Here we directly test this hypothesis for neurons in
monkey PPC.
We designed a novel delayed and memory-guided reach tasks in
which monkeys had to perform either pro or anti reaches in either
normal viewing conditions or under reversed vision through
inverting prisms. With this task we can double-dissociate the
spatial information contained in the sensory instruction cue, the
intended physical movement, and the predicted visual feedback
about the movement. The pro vs. anti comparisons dissociated
the visual memory of the cue from the movement intention ('motor
goal'), in both the normal viewing context and the prism viewing
context. The prism vs. non-prism comparisons dissociated the
physical intention from the visual prediction in the pro and anti
contexts. Especially, the combined prism and anti-reach task
created situations where the sensory cue and physical reach
directions were identical, but the visual predictions were opposite.
About one fourth of the neurons in PPC which were movement
related and directionally selective in the prism and no-prism
condition encoded the direction of the anticipated visual feedback,
rather than the intended physical arm movement direction. The
remaining neurons encoded the physical direction of planned
movement.
Our results provide direct evidence for the notion that motor
planning not only evokes neuronal representations linked to the
planned action per se, but also to its anticipated perceivable
sensory consequences. These observed predictive
representations establish a role of forward model prediction in the
parietal cortex, even before a motor command is issued, as an
integral part of the planning and probably also selection process
for movements, one of the distinguishing features of the
ideomotor principle. This idea dates back to the early 19th century
and is prominent until today in research of goal-directed behavior,
not least since it tightly relates to the ecological approach of
perception, motor awareness, the mirror-neuron system, and
modern concepts of neural adaptive motor control.
C - Control of Eye & Head Movement
1-C-31 Where cats look during walking
Irina Beloozerova1, Trevor J Rivers1, Mikhail G Sirota1, Neet A
Shah1, Andrew I Guttentag1, Dmitri A Ogorodnikov2
1
Barrow Neurological Institute, 2Mount Sinai School of Medicine
Vision is important for locomotion, particularly when an individual
is moving through complex natural environments. How and what
visual information is collected during locomotion in complex
environments is far from being understood, however.
In this study we used freely walking cats. Cats walked in five
different conditions: (1) on a flat surface in complete darkness, (2)
on flat surface in the light, (3) on a horizontal ladder with
crosspiece of different widths (5, 12, 18 cm), (4) on a
pathway cluttered with many small stones, and (5)
overstepping different series of thin 7 cm high obstacles. We
recorded cat's eye movement using a scleral coil and headmounted magnetic field generator, and head and limbs
movement using an active-marker 3-dimensional tracking
system. We calculated cat's gaze, determined where it
intercepts walking surface, and compared gazing behaviors
and coordination of gaze with stepping movements across
different walking conditions.
We found that as the density of obstructions along the
pathway increases, cats look closer, fixate more, and also
shift gaze more often. In agreement with previous studies in
cats, we found that cats make an overwhelming majority of
gaze shifts two strides prior to stepping on the site looked
at. There is a strong correlation between vertical gaze shifts
and stride phase. A typical stride has two gaze shifts. Gaze
shifts ending further away from the cat occur near the end of
the stance phase of either forelimb (peaking 200 ms before
the onset of swing phase), while gaze shifts ending closer to
the cat, occur near the beginning of the either forelimb
swing (peaking 100-200 ms after the onset of swing). In
addition to correlating with stepping movements, vertical
gaze shifts also correlate with head pitch rotation. While
overstepping barriers, cats tend to look at the top edge, but
not at the center or the bottom edge of the barriers. On the
ladder, cats' gaze is typically directed at the front or back
edge of the crosspiece, regardless of the crosspiece width,
and they rarely look either at the center of a crosspiece or at
the space between rungs. Thus, we found that for
successful stepping in complex environments cats as
people are more concerned with obtaining visual information
about the boundaries within which they can step, rather than
with examining the exact spots of future foot placement with
their central vision.
Thanks to the use of scleral coil and computerized 3
dimensional head tracking techniques, this study provides
the first accurate numeric characterization of cat gazing
behavior during locomotion and sets a stage for analysis of
neuronal mechanisms for processing of visual information
during locomotion.
1-C-32 Robotic models for neural control of eye
movements
Dinesh Pai1, Martin Lesmana1, Per-Erik Forssén2
1
University of British Columbia, 2Linköping University
Theoretical models of neural control are essential for
developing insights into physiological experiments but it is
difficult to test whether these models are actually effective
for control, and not just for describing the data. Here we
describe an approach which uses robotic models of the
oculomotor control system to investigate theories of neural
control of movement. We focus on the oculomotor system
as it is a relatively simple yet complete sensorimotor
system, with a wealth of neurophysiological data. In
particular, we look at two types of eye movements: saccadic
gaze shifts and gaze stabilization (with VOR and OKR).
Saccades are too fast to control using the usual engineering
approach of relying on online sensory feedback. This stems
from the delays inherent in the nervous system; a saccade
of less than 10 degrees can be completed in less than 50
milliseconds, while low level visual processing alone takes
40 milliseconds. To achieve fast movements in the face of
large delays, we developed a biologically inspired controller
that learns a non-linear dynamic model of the eye and
predicts the required pulse-step activation signal. We tested
19
Poster Sessions
Full Abstracts
the controller on a robotic eye with complicated non-linearities
(similar to those due to orbital tissues) and demonstrate saccade
performance comparable to that observed in the human eye.
Interestingly, the controller also reproduces the characteristic the
main sequence relationship. Gaze stabilization enables clear
vision during locomotion by moving the eyes to compensate for
head movements. This is achieved through combining head
movement information measured by the fast vestibular system
and the slower visual information from the eye. We develop a
gaze stabilization controller that uses vestibular input as the
primary input while visual error is used as a teaching signal and to
prevent drift. We implemented the controller on a robotic head
mounted on a high speed robotic testing platform. The ''head''
contains the equivalent of an eye (actuated camera) and the
vestibular system (inertial measurement unit). We demonstrate
the controller's effectiveness in stabilizing the camera to high
speed perturbations.
1-C-57 Smooth pursuit eye movement (SPEM) in
patients with idiopathic Parkinson's disease (PD):
Movement preparation and execution is impaired but not
visual motion working memory
exhibited initial pursuit (latency 128±19SD ms) followed by
corrective saccades (212±56SD ms), and latencies of these
components were similar to those of controls (p>0.6).
Postsaccadic SPEM enhancement was also seen in most
PD (13/16). (4) In both tasks, peak SPEM velocity during
pursuit maintenance was lower in PD than controls
(p<0.01).
Discussion: Since initial SPEM of most PD during task B
was basically similar, though reduced, to that of controls, the
initial SPEM difficulty during task A was unlikely to reflect
impaired SPEM per se; possibly it reflects difficulty in
inducing priming effects using cue information (Fukushima
et al. 2011) and/or generating extra retinal components for
SPEM (Barnes and Collins 2008). The difficulty in the latter
may also be involved in impairing pursuit maintenance.
Difficulty in terminating fixation may also contribute to the
prolonged latency of the first saccades during memory
pursuit (Warabi et al. 2011).
Conclusion: The working memory of motion-direction and
go/no-go selection was normal in PD but they had difficulty
in initiating SPEM using the cue information.
Norie Ito1, Nozomi Tamaki1, Ayumi Masuno1, Kunihiro Ikeno1,
Sachiyo Onishi1, Nobuyoshi Kobayashi1, Hidetoshi Takei1, Peter
M Olley1, Susumu Chiba1, Kiyoharu Inoue1, Grahamu Barnes2,
Kikuro Fukushima1, Tateo Warabi1
1-C-58 Oculomotor adaptation of sound localization
depends upon the temporal relationship between
targets and eye movements
1
Gary Paige1, Emily C Clark1, William O'Neill1
Sapporo Yamanoue Hospital, 2University of Manchester
1
Purpose: To examine whether the impaired SPEM in PD includes
impaired visual motion working memory and/or movement
preparation.
Background: Clear vision of slowly moving objects depends on
SPEM during which target images are maintained on the fovea by
predictive compensation of the inherent delays in responding to
target motion. Various factors influence prediction such as cues
and working memory of stimulus trajectory (Barnes 2008 for
review). Prediction could occur not only in motor commands to
prepare for and maintain ongoing movements but also in the
sensory and/or perception pathways. A memory-based smooth
pursuit task separates neuronal activity related to visual motiondirection memory from movement preparation (e.g. Shichinohe et
al. 2009). SPEM is impaired in most patients with PD (Leigh and
Zee 2006 for review), though the nature of the impairment is
poorly understood. Impaired working memory during cognitive
tasks has also been reported in PD (e.g. Lee et al. 2010).
Methods: We applied this task (task A) to 16 PD patients (Hoehn
and Yahr stage 2.6±0.6SD) and 7 age-matched controls. In task
A, cue 1 indicated visual motion-direction (rightward or leftward)
while cue 2 instructed the subject to prepare to pursue (go) or not
to pursue (no-go). Based on their motion direction memory and
go/no-go instruction, subjects had to select the correct spot from
2 oppositely moving spots at 10°/s and to pursue it or not pursue
by selecting a 3rd stationary spot. For comparison, simple ramp
pursuit using a single spot with the identical spot velocity (task B)
was examined. Infrared oculography was used to record eye
movements.
Results: (1) In task A, error rates of all PD patients tested during
go and no-go trials were low (<5%) and were similar to those of
controls (p>0.3). (2) During go trials, controls exhibited SPEM
with a mean latency of 158±40SD ms followed by corrective
saccades which were further followed by enhanced SPEM. Most
PD (13/16) tracked the correct spot predominantly with saccades;
SPEM was absent before the saccades, and 9/16 PD lacked
postsaccadic SPEM enhancement. The mean latency of the first
saccades in PD (326±57SD ms) was longer than that of controls
(289±32SD ms, p= 0.02). (3) During task B, most PD (11/16)
20
University of Rochester
We have described a unique oculomotor adaptation in which
prolonged changes in eye position shift the perception of
auditory space in the same direction as ocular deflection
(mean magnitude ~40%, time-constant ~1 min.). One
consequence is that localization of ongoing sounds
overshoots target azimuth when the head is fixed but the
eyes are free to guide laser-pointing responses, entailing
gaze eccentricity for a few seconds. This overshoot is
largely eliminated when the eyes are fixed. However, this
distinction disappears for transient targets, whether the eyes
are fixed or free to guide localization from memory. Perhaps
adaptation is limited to the state of eye position at the time
targets are presented. Ongoing targets provide a continuous
sampling regardless of eye movements, but transient
targets register at a single moment with respect to eye
movements. To test this concept, normal human subjects
(age <30) localized transient targets (5 noise bursts over 1
s) while fixating a visual reference. Trials began with the
eyes centered, followed by sustained eccentric fixation (20°
left or right). Subjects waited 10 seconds after the ocular
deflection before localizing targets, each presented during
one of three different times in the trial: A) eyes fixed
centrally, immediately before eccentric fixation, B)
immediately after re-fixation eccentrically, or C) after 10
seconds of eccentric fixation. These conditions were
compared to a baseline control condition with only center
fixation and no delay in localizing targets, and another
control in which eye position was held eccentrically for all
targets. The key variable is the time between the attainment
of new eye position and target presentation. Results for
condition A were indistinguishable from baseline. In
contrast, a small response overshoot arose in condition B,
when targets were presented just after eccentric fixation,
which increased for condition C, when targets were
presented 10 seconds after eccentric fixation, and proved
greatest for the persistent eccentric gaze task. We conclude
that oculomotor adaptation of auditory spatial perception
depends upon current eye position and duration at the time
of transient target presentation, but not subsequent changes
Poster Sessions
Full Abstracts
in eye position. The role of oculomotor adaptation of auditory
space presumably reflects a need to establish and maintain a
sense of 'straight ahead,' which includes eye, head and body
orientation as well as sensory counterparts. A cohesive alignment
system, together with co-calibration of spatial gain, secures our
concept of space constancy for both spatial offset and gain
across modalities.
1-C-59 Comparing models for eye-head trajectory
interactions during head-free gaze shifts
1
2
Iman Haji Abolhassani , Daniel Guitton , Henrietta L Galiana
Radboud Univ Nijmegen, 2University of Hamburg
The gaze (eye-in-space) orientation system uses multiple
platforms, i.e., eyes, head, and body, to move gaze in space.
Body-fixed head-free gaze shifts, which are the main focus of this
study, comprise two phases: the fast phase (saccade) and the
slow phase (fixation). A complete head-free gaze shift model
needs to include both phases and a mechanism for switching
between the two. Two major types of models have been proposed
for head free-gaze shifts: the independent eye-head control
models and the shared gaze feedback control models. In the
former, it is presumed that plant controllers operate in parallel and
are independent, whereas, in the shared gaze feedback control
models, the updated gaze motor error affects both plants. The
model tested here is an updated version of the Prsa-Galiana
using shared gaze error feedback. In this study the efficiency of
this model is compared to other available popular models for
head-free gaze shifts, in comparison to experimental data. It will
be shown that our model has prominent advantages to replicate
numerous experiments, including:
‐
‐
‐
Pieter Medendorp1, Frank T Leone1, Ivan I Toni1, Tobias
Heed2
1
McGill University, 2Montreal Neurological Institute
‐
1-C-60 Testing effector-specificity of human
posterior parietal cortex for eye, hand and foot
movements: A multivariate analysis
1
1
‐
neurophysiology regarding their pathways and
elements.
The switch: The transition between the fast and slow
phases of gaze shifts must be consistent with behaviour
and with data on OPN/Burst cells, e.g., the fact that the
eyes can reach their peak deviation before the end of
the gaze saccade (end of OPN pause) and produce a
'VOR-like' turn around during a saccade. In contrast,
other models consider the moment of maximum
deviation of conjugate eyes as the switching moment for
OPNs.
Head perturbations & brakes: The gaze-shift system is
known to show high robustness to head perturbations
both during the fast and slow phases. Passive head
rotations during the fast phase especially, have drastic
impact on eye and head trajectories. However, their
impact on the overall gaze trajectory is far less
noticeable. The ability of various models to predict this
robustness in gaze shift trajectory will be compared.
OPN lesions: In some patients, due to the failure of the
mechanism responsible for switching between phases
(Burst neuron lesions), the gaze shift system gets stuck
in the slow phase. While these patients lose their ability
to perform fast saccadic movements, they are still
capable of moving their gaze, although in a slower
fashion. The ability of the models to replicate this
phenomenon is also explored.
Vestibular lesions: The predictions of the models in
acute vestibular lesions will be compared with
experimental results. A scheme will also be proposed
for our model to replicate compensated gaze shifts for
these cases.
Network topology/neurophysiology: Finally, the models
will be compared on the basis of consistency with
Monkey neurophysiology and human neuroimaging have
reported segregated processing in the posterior parietal
cortex (PPC) for the planning of goal-directed eye and hand
movements. These findings have been interpreted to
indicate effector-specific processing for movement planning
in the PPC. However, when comparing eye, hand, and foot
motor planning in humans using fMRI, our group recently
found that processing in the PPC did not differ between
hand and foot motor planning, whereas limb processing
differed from saccade planning (Heed et al. JNS, 2011).
While these results may reflect a shared representation for
hand and foot movements, it is also possible that there are
separate hand and foot representations, which are hidden to
standard methods of analysis. To distinguish between these
possibilities, we applied multivariate analysis. Participants
were instructed to make delayed eye, right hand, or right
foot movements to targets, presented at three different
locations on either side of the body midline. Delay times
were randomized between 2 and 6 s. In 12 separate runs
we acquired multi-echo BOLD-sensitive echo-planar MR
images (TR=2.0 s, 26 slices, 3.0x3.0x3.0 mm voxel size)
using a Siemens 3-Tesla MRI system. Multiple regression
analysis was applied to the unsmoothed data per run,
focusing on the delay activation, with separate regressors
for each effector-side-repetition combination. The t-values
characterizing the respective regressors were used as input
for a searchlight classification procedure, which tested local
effector-specific content using a leave-one-run-out SVM to
each possible sphere with a radius of 2 voxels. The analysis
revealed the frontoparietal motor network, with effectorspecific tuning in PPC, precuneus, (pre-) motor cortex,
supplementary motor cortex, and the frontal eye fields.
There was a spatial gradient from caudal PPC regions,
mainly concerned with planning eye movements, to more
rostral PPC regions, mainly concerned with planning hand
and foot movements. Crucially, activity in caudal PPC
regions predicted whether subjects were planning to move
their eyes or their limbs, but did not discriminate between
hand or foot movements. We conclude that PPC contains a
shared representation for hand and foot movements, distinct
from that involved in planning eye movements. Hence,
processing in the PPC is not primarily organized according
to the effectors, but rather according to functional criteria,
which differ markedly between eyes and limbs.
1-C-61 A nonlinear model of the horizontal Angular
Vestibulo-Ocular Reflex (AVOR) - A mechanism for
context-dependent responses
Mina Ranjbaran Hesarmaskan1, Henrietta L Galiana1
1
McGill University
The vestibulo-ocular reflex (VOR) produces involuntary eye
movements that are evoked in response to vestibular
stimuli. The VOR is responsible for stabilizing retinal images
during head perturbations to maintain clear vision. This
reflex is composed of slow phases where the eyes move in
the opposite direction to the head and fast phases when
21
Poster Sessions
Full Abstracts
they usually move in the direction of the head. The gain of the
VOR is defined as the ratio of peak eye velocity to peak head
velocity during harmonic testing or high-frequency pulse
perturbations.
Testing the VOR after fixation at different depths shows that the
gain of the VOR depends on the location of the fixation target
(real or imaginary) relative to the observer, since the eyes are not
concentric with the head center of rotation [1]. It can be
demonstrated geometrically that the magnitude of the ocular
deviations required for compensating a translation of the eyes
relative to a near target increases as a function of decreasing
fixation distance.
The common approach in modeling studies that focus on the
context-dependence of the VOR is to use an estimate of target
position to scale the gain of the reflex in the neural circuit.
However, the underlying neural site and mechanism are still
unknown. Recent works experimentally [2] and theoretically [3]
suggest that nonlinear computations in VOR premotor circuits
could generate target-distance related changes in VOR gain. The
work presented here expands on this concept, with a simple
bilateral model for slow phase horizontal angular VOR in the dark,
based on recent knowledge of its premotor anatomy and
physiology. This model is able to replicate binocular responses for
targets at different depths and eccentricity during short duration
head velocity 'bumps'.
In this model it is postulated that nonlinear neural computations
appear at the level of premotor cells in the vestibular nuclei to
generate target-distance-dependent VOR gains. This nonlinear
function is placed in eye-head-velocity (EHV) neurons that
receive direct sensory information from the ipsilateral semicircular
canals as well as efferent copies of the ipsilateral eye position
and vergence signals. EHV cells are reported to have variable
sensitivity to head perturbations depending on current monocular
eye position and vergence angle and they project directly to
motoneurons. The model also includes nonlinear vestibular
sensors as described in the literature, for a more realistic sensory
system in the VOR. As a result, it is possible to explore different
mechanisms for the fusion of vestibular and reafference signals
on premotor cells. Using simulations, the new model is shown to
be consistent with observations in unilaterally-deficient or canalplugged animals [4].
References: 1. Viirre, E., et al., A reexamination of the gain of the
vestibuloocular reflex. Journal of Neurophysiology, 1986. 56(2): p.
439-50. 2. Zhou, W., et al., Multiplicative computation in the
vestibulo-ocular reflex (VOR). Journal of Neurophysiology, 2007.
97(4): p. 2780-2789. 3. Khojasteh, E. and H.L. Galiana,
Implications of gain modulation in brainstem circuits: VOR control
system. J Comput Neurosci, 2009. 27(3): p. 437-51. 4. Migliaccio,
A.A., L.B. Minor, and J.P. Carey, Vergence-mediated modulation
of the human angular vestibulo-ocular reflex is unaffected by
canal plugging. Experimental brain research. Experimentelle
Hirnforschung. Experimentation cerebrale, 2008. 186(4): p. 581-7
1-C-62 Analysis of visual behavior in a continuous
visual-motor coordination task
Cristian Arellano1, Pablo Burgos1, Pedro Maldonado1
1
University of Chile
Objectives: Currently there is a broad understanding of the role
of perceptual vision, but there is less clarity about the role of
vision in the guidance of motor actions. The aim of our study was
to characterize visual behavior during the performance of visually
guided motor tasks and observe how it adapts to conditions of
varying difficulty (number of obstacles in a maze).
22
Methods: Using a computer 10 normal vision subject, went
through a series of spiral mazes contained variable number
of obstacles in its path. Five maze levels of difficulty were
analyzed (level 0 to 4) and two experimental conditions
were compared visuo-motor (VM) v/s just gaze (JG). In VM
condition subjects went through the mazes using a digital
pen (G-Pen 4500). In JG condition subjects performed same
task only using the gaze to went through the mazes. In both
conditions the instructions and goal task were the same.
Time duration task for each difficulty level and eye
movements (fixations and saccades) of the subjects were
recorded using an eyetracker (EyelinkII).
Results: Our results showed as more difficult becomes the
task more time was used. In VM condition the number of
fixations and saccades increased progressively with task
difficult, while saccades duration and amplitude decreased.
All these variations were significantly different between the
lowest and higest difficulty levels (p< 0.05). By contrast time
fixations duration was similar for all difficulty levels without
significant differences. When comparing visuomotor (VM) v /
s just gaze (JG) conditions, we found that in the JG
condition the task duration, fixation number and saccades
number showed a consistent decrease for the 5 levels of
difficulty assessed, although this difference was significant
only at level 4 for task duration, and level 2 of difficulty for
the number of fixations and saccades (p <0.05). By contrast
when we compared the duration and amplitude of saccades
we observed a consistent increase in these variables for JG
condition for 3 lower levels of difficulty (p <0.05). Again the
duration of fixations in the JG condition is similar for all
levels of difficulty, and there were no significant differences
compared with the VM condition.
Conclusions: The visual strategy used in performing the
task is similar for different subjects and consistently adapts
to different levels of difficulty. The increased time task
duration is given by the increase of the number of fixations
and saccades. The duration of fixations is similar for both
analyzed conditions (VM-JG). This results suggest that the
time of capture and processing of the visual information
(fixations) is relatively fixed and could be a physiological
constraint to increase speed execution of analyzed task.
1-C-63 Dynamic probabilistic control of audiovisual
saccades
John van Opstal1, Marc Van Wanrooij2, Brian Corneil3, Doug
Munoz4
1
Donders Institute, 2Radboud University Nijmegen,
University of Western Ontario, 4Queens University
3
Due to its high resolution, vision is commonly thought to
dominate spatial perception over other sensory systems,
such as audition. Recent studies have suggested that visual
capture could result from statistically-optimal (Bayesian)
integration of unreliable auditory and precise visual cues
that are typically assumed constant. We here report on how
cue reliabilities for auditory, visual, and audiovisual targets,
hidden within a cluttered audiovisual background, evolve on
a millisecond timescale in a saccadic-search task.
Interestingly, not only response variability (conversely,
precision) depended systematically on sensory modality,
saccadic-reaction time, and signal-to-noise ratio, also the
response gain (accuracy) was systematically affected. We
reveal a universal near-optimal precision-accuracy
relationship, based on posterior probability matching, which
shows constrained responses when cue reliability is poor
(e.g. for fast visual responses, and for low-level sounds).
This study shows for the first time that saccadic eye
Poster Sessions
Full Abstracts
movements, and multisensory integration in particular, obey
Bayesian inference on a millisecond timescale.
1-C-64 Properties of FEF projection neurons in smooth
pursuit
Michael Mustari1
1
University of Washington
The frontal pursuit area of primates, which is located in the fundus
of the arcuate sulcus plays a role in smooth pursuit (SP). Neurons
in FEF discharge during smooth pursuit with appropriate timing to
play a role in SP initiation. Our studies were conducted in 2
monkeys (Macaca mulatta) trained to perform SP and other
visual-oculomotor tasks. Eye movements were recorded with
scleral search coil technology and single units recorded with
standard extracellular recording methods. For the FEF to
generate SP signals must be delivered to appropriate brainstem
centers. The FEF is known to project to the rostral nucleus
reticularis tegmenti pontis (rNRTP), dorsolateral pontine nucleus
(DLPN), superior colliculus (SC) and basal ganglia. However, we
still have relatively little specific information about the signals sent
from FEF to each of these regions. To address this we first
mapped the SP regions of FEF and rNRTP before conducting
antidromic activation studies to identify projection neurons. We
tested all SP neurons in FEF for activation and characterized their
response dynamics during smooth pursuit eye movements. SP
testing was with step-ramp target motion so that we could
consider neuronal response during open and closed-loop
components of SP. We delivered micro-electrical stimulation
(single biphasic pulse 200 µS, 50-200 µA) to the rNRTP and
recorded FEF neurons that were antidromically activated
(collision testing verified) or not activated. We used multiple linear
regression modeling to account for the sensitivity of individual
FEF neurons to components of SP motion. We also examined the
time to peak firing during SP for antidromically activated or nonactivated neurons. We found differences in projection neurons
(antidromically activated) and other neurons. The main findings
were that FEF projection neurons showed strong eye acceleration
sensitivity. When we attempted to reconstruct unit firing rate
without eye acceleration parameters our multiple-regression
models produced significantly poorer fits then when acceleration
was included. Furthermore, individual FEF projection neurons had
a similar early times-to-peak firing, which were appropriate for SP
initiation. We found that this timing was mirrored in the properties
of rNRTP neurons, which receive a major input from FEF cortex.
In contrast neurons that were not antidromically activated often
had more spread out times-to-peak firing during SP. Our results
suggest that FEF plays a critical role in the initiation of SP
presumably through its strong connections to the oculomotor
vermis. Further studies are required to determine the FEF signals
sent to other brainstem targets that might play a role in SP>
Supported by: EY013308; RR000166
1-C-65 Adaptation of intrinsic and synaptic properties of
medial vestibular neurons during visual-vestibular
mismatch
Erwin Idoux1, Daniel Eugène1, Antoine Mialot1, Mathieu
Beraneck1
1
Université Paris Descartes
The vestibulo-ocular reflex (VOR) stabilizes the gaze during
unwanted movements of the head. It relies on the conversion of a
head velocity signal to an eye velocity signal at the level of the
medial vestibular nuclei. However, these brainstem nuclei receive
information from not only the vestibular afferents, but also from
visual and proprioceptive origins. A mismatch between the
information from the different sources will result in the adaptation
of the gain of the VOR, so that the conflict is minimized.
Here, we generate a visual-vestibular mismatch (VVM) to
trigger the adaptive plasticity, with an innovative protocol
where the animal is wearing a translucent stripped helmet
for 2 weeks. While a fully opaque helmet prevented the
formation of any image altogether and a homogenously
translucent helmet produced an image without salient
features, the image formed when wearing the stripped
helmet is fixed on the retina yet highly contrasted. Thus
instead of being stimulated, head fixed, at a given frequency
for a short period of time, the animal is free to experience its
natural repertoire of behaviors, while constantly undergoing
the VVM. The 3 protocols decreased the gain of the VOR
drastically (60-70%) at all tested frequencies (0.2 to 2 Hz).
After 2 days without the fully opaque or fully translucent
VVM creating helmet, the gain of the VOR is restored at all
frequencies. Interestingly, the same recovery pattern was
found for the stripped helmet VVM at high frequencies (12Hz) but not at low frequencies (0.2-0.5Hz), and stays like
that for at least a week.
We thus looked for the neural correlates of these gain
changes and recorded medial vestibular neurons in acute
brainstem slices of control and VVMed animals. We found
typical type A and B neurons in both sets, as defined
quantitatively in Beraneck et al 2003. Both in control and
VVMed animals, type A and B were in comparable
proportions with almost identical intrinsic properties (spike
height, width, concavity, convexity, AHPR and DAHP),
except for changes in opposite directions for the frequency
of the spontaneous discharge of both type (increase for type
A and decrease for type B).
We also investigated the plastic potential of the synapse
between the vestibular afferent and the medial vestibular
neurons, both for long-term potentiation and depression
(LTP and LTD respectively). While this synapse could be
potentiated or depressed in control animal, the VVM
protocol abolished most of its modulation. This could be part
of the mechanism by which the gain is decreased for low
frequencies.
1-C-66 When the hand drives the eye: Hand-eye
coordination when reaching to a manifold
Thierry Pozzo1, Ambra Bisio1, Marco Jacono1, Bastien
Berret1
1
Istituto Italiano di Tecnologia
In previous works, we introduced a manifold reaching task
to study human behavior when facing target redundancy.
The pointing task was restricted to a very long target bar
without salient point. Despite the absence of imposed
endpoint, we found that subjects nevertheless make the
decision to reach to particular regions of the bar, depending
on the starting posture of the arm. The observed behavior
was shown to be in agreement with optimal arm movement
control. In this study, we now investigate the role of the
oculomotor system in resolving task indeterminacy and
decision making processes. Precisely, we wanted to verify if
a saccade could be endogenously induced by the arm motor
plan elaborated to reach the bar. Indeed, classical
paradigms did not allow to study such a question since an
imposed salient target point triggers a unique saccadic
reflex mechanism preceding the arm reaching. Using EOG
(electro-oculography) and 3D motion capture, we
investigated the behavior of subjects instructed to perform
horizontal reaching movements from a sitting posture. Three
types of stimulus were randomly presented to the subjects:
a single target point (stimulus S1), two simultaneous target
points (S2) and a horizontal target bar (S3). Two initial arm
23
Poster Sessions
Full Abstracts
postures were tested, starting from the left and right sides of the
workspace respectively. Before the stimulus appeared, subjects
had to look at a central cross displayed on a large vertical screen.
Exactly when the cross disappeared (tS) one of the stimuli was
presented on the screen and the subject had to reach to the
displayed target (whatever its type S1, S2 or S3). Preliminary
results indicate the existence of saccadic movements directed
toward the endpoint of the reach and starting before the hand
movement onset (tH). However, the latencies of
stimulus/eye/hand largely differed across the 3 conditions. For
stimulus S1, a saccade occurred 200 ms after tS whereas the
hand started approximately 500-600 ms after tS. The same eye
reaction time (tE) was observed when stimulus S2 was presented
to the subjects, except that in many cases one supplementary
saccade occurred just before or around tH. This often increased
tH, possibly underlining additional neural processes related to
decision-making. In contrast, for the bar stimulus (S3) the first
saccadic eye movement, which was directed toward the endpoint
of the reach, occurred significantly later than in the two previous
conditions: tE was greater than tS by more than 300 ms, but the
value of tH was not affected. Reaching to a bar was thus not
computationally more intensive than reaching to point but rather
induced a significant compression of the relative eye-hand latency
(tH-tE). Our results show consistent saccadic eye movements
toward predicted locations in space when no salient visual
stimulus was present. These locations overlapped the final hand
positions onto the bar. Moreover, the timing of eye and hand
displacements strongly suggests that eye movement reflect
endogenous orienting mechanisms that allocate attentional
resources using motor cues (e.g. arm motor plan). Since retinal
input cannot specify an unequivocal spatial goal, we speculate
that the oculomotor system would wait for a desired location
selected by different brain structures involved in decision making
and constrained by internal rewards. Thus, this study shows the
bi-directionality of the visuomotor processes and that vision can
be subordinated to the motor system.
D - Disorders of Motor Control
1-D-106 EMG-based visual-haptic biofeedback: A
promising tool to improve motor control in dystonia
Claudia Casellato1, Giovanni Zorzi2, Alessandra Pedrocchi1,
Giancarlo Ferrigno1, Nardo Nardocci2
1
Politecnico di Milano, 2Neurological Institute C. Besta
Dystonia is a syndrome characterized by excessive and sustained
muscle contraction causing twisting and repetitive movements,
abnormal postures, or both. New insights suggest that dystonic
motor impairments could also involve a deficit of sensory
processing actions. In this framework, biofeedback could be
useful, since it makes covert physiological processes more overt.
The present work proposes an innovative integrated set-up
providing EMG-based visual-haptic biofeedback during upper limb
movements, to test if augmented sensory feedbacks can induce
motor control improvement in patients affected by primary
dystonia. The set-up was composed of a haptic device and a
EMG device. The graphic interface was built up showing the
required task. The real-time control algorithm was developed in
Visual C , so as to synchronize the haptic position-force loop with
the EMG signals reading; the brachioradialis EMG values were
used to proportionally modify visual and haptic features of the
interface: the higher was the EMG level, the higher was the virtual
table friction and the background color proportionally moved from
green to red. We tested, on dystonic and healthy subjects, a
sequence of spiral tracking tasks, without and with biofeedback.
24
The subjects were asked to achieve the task naturally and,
with BF, keeping the color far from red as much as possible
and the surface friction as lowest as possible. Two-away
ANOVA test, where the within-subject factor was the
experimental condition (No-biofeedback / biofeedback ) and
the between-subject factor was the group (dystonic /
healthy), on the median Root Mean Square values of EMG,
outlined that the biofeedback had a dystonia-related
significant impact on the muscular control, inducing a
contraction decrease, correlated with the dystonia severity.
Thus, these tests pointed out the effectiveness of
biofeedback paradigms in gaining a better specific-muscle
voluntary control. The here-developed flexible tool show
promising potential for clinical applications and sensorimotor
rehabilitation.
E - Posture & Gait
1-E-67 Effects of low back pain and proprioception
disturbance on precision of trunk control
Nienke Willigenburg1, Idsart Kingma1, Jaap H van Dieen1
1
Research Institute MOVE
Introduction: Proprioception of the trunk appears to be
affected in low-back pain (LBP) patients. This might lead to
reduced precision in control of trunk posture and findings of
increased cocontraction could be explained as a
compensatory strategy to deal with this. AIM: To compare
precision of trunk postural control between LBP patients and
healthy controls and to study the effect of disturbing
proprioceptive information from the low back through
paraspinal muscle vibration on precision.
Methods: Eighteen subjects with a-specific LBP and 13
healthy controls with no (history of) LBP participated.
Subjects performed two tasks requiring high precision of
control over trunk orientation. In task 1, a slowly moving
medium-sized target was tracked over a spiral-shaped
trajectory by circular movements of the trunk, while
continuous visual feedback of target and actual trunk
orientation was provided. In task 2, subjects maintained a
self-chosen upright trunk posture within a static small or
large target. Visual feedback of trunk orientation was given
only when orientation was outside the target. In both
conditions subjects were instructed to stay within the target.
Both tasks were performed with and without paraspinal
muscle vibration. In task 1, the absolute distance to the
orientation corresponding with the center of the target was
used as an index of performance. In task 2, both the mean
distance to target center and the SD of the trunk orientation
were calculated. Surface-EMG was used to obtain the ratio
of antagonist over agonist EMG amplitudes as an index of
cocontraction. Mixed design ANOVAs were used for
statistical comparisons.
Results: In task1, errors were significantly higher in patients
and an interaction of group with condition, showed that
vibration degraded performance in healthy subjects only. In
task 2, performance was not significantly different between
groups and performance was degraded by vibration in both
groups. No indications for increased cocontraction in
patients were found.
Conclusion: Precision of trunk control was reduced in
patients and not affected by muscle vibration when constant
visual feedback (task 1) was present. In contrast, when
Poster Sessions
Full Abstracts
visual feedback was absent (task 2, large target) or intermittent
(small target), precision was not impaired and was affected by
vibration. These results suggest that patients reduce weighting of
proprioceptive information from the low back when visual
feedback is available, which leads to reduced precision. In
absence of visual feedback, proprioceptive information from the
low back appears to be used and precision is maintained
suggesting that the proprioceptive information itself is valid.
1-E-68 Persistence of motor-equivalent fluctuations
during quiet standing
Julius Verrel1, Didier Pradon1, Nicolas Vuillerme1
1
MPI for Human Development
According to current theories of motor control, the central nervous
system stabilizes performance by selectively suppressing taskrelevant variability (TRV) while allowing task-equivalent variability
(TEV). In line with such a control mechanism, TEV during quiet
standing has been observed to be larger than TRV (Hsu et al,
2007). In addition, TEV and TRV variability are expected to differ
in their temporal characteristics, with higher persistence of taskequivalent fluctuations.
This prediction was tested in the present study. Kinematics of
quiet standing (5 minutes) was measured in five healthy young
participants. Using the uncontrolled manifold analysis (Scholz &
Schöner, 1999), postural variability in sagittal joint angles was
decomposed into TEV and TRV with respect to four task
variables: center of mass (CoM) and head position, and trunk and
head orientation. Persistence of fluctuations within the two
variability components was analyzed using a multivariate
autocorrelation index, with time lags between 1 and 60 seconds.
The results show that (1) TEV exceeds TRV; (2) persistence is
higher for TEV than for TRV; (3) auto-correlation decreases with
increasing time lag; (4) this decrease is more pronounced for TRV
than for TEV. In addition, the relation between auto-correlation
and time lag differed between CoM position, head position and
trunk orientation on the one hand, and head orientation on the
other.
These findings confirm the hypothesis that TEV and TRV do not
only differ in amplitude but also in their temporal characteristics,
underlining the dynamic nature of whole-body postural control.
Differences in the temporal variability structure of different task
variables likely reflect functional differences between these
postural parameters (equilibrium versus visual orienting).
1-E-69
Swimming patterns of the Octopus vulgaris
Dimitris Tsakiris1, Asimina Kazakidi1, Michael Kuba2, Alex
Botvinnik2, Michael Sfakiotakis1, Tamar Gutnick2, Shlomi
Hanassy2, Guy Levy2, John A Ekaterinaris3, Tamar Flash4,
Binyamin Hochner2
1
Institute of Computer Science – FORTH, 2The Hebrew University
of Jerusalem, 3Institute of Applied and Computational
Mathematics – FORTH, 4Weizmann Institute of Science
Swimming is an important means of locomotion for the benthic
common octopus (Octopus vulgaris) whenever the animal needs
to cover distances at greater speed. Although swimming does not
appear to be the predilected mode of locomotion for octopuses in
captivity--possibly due to the limited available space and the large
amount of energy consumption required--, the behavior has been
previously observed in nature. In this study, we used six adult
animals (Octopus vulgaris, 200-400 grams of weight) kept in the
aquarium, in order to identify possible swimming mechanisms.
We define swimming as any type of locomotion of the octopus
that involves no attachment to the aquarium walls or other objects
and substrates. Using this definition, three different patterns of
swimming were observed: arm swimming, that involves
undulations of the arms in synchrony, with a power (closing)
and a recovery (opening) stroke of the arms; jet swimming,
in which the octopus uses the siphon, while the arms trail
tightly together, behind the mantle; and head-first swimming,
where the octopus swims with the head in front and the
arms trailing behind. This is the first time that these
swimming patterns have been recorded systematically in the
laboratory. Three-dimensional reconstructions of individual
or multiple arms during these movements show the
underlying mechanisms involved for each swimming pattern
and may assist in the design and control of robotic arm
prototypes.
1-E-70 Reflex control of treadmill walking in
subjects with spinal cord injury
Virginia Way Tong Chu1, Thomas G Hornby2, Brian D
Schmit3
1
Rehabilitation Institute of Chicago, 2University of Illinois at
Chicago, 3Marquette University
During walking, it is thought that reflex activity plays an
important role in the timing and magnitude of muscle activity
throughout the gait cycle. Movement of the body over the
stance leg imposes hip extension and limb loading during
gait. The effect of limb loading and hip proprioception during
stepping has been shown in decerebrated and spinal
animals and human infants. As an example, Yang and
colleagues (2000) introduced perturbations during different
phases of the gait cycle during infant stepping and saw that
the stance phase was shortened and swing phase
advanced when the hip was extended and load was
reduced. Since the myelination of the corticospinal tract do
not complete until the age of 2 (Yakovlev & Lecours, 1967),
it is reasonable to suggest that these changes in the gait
cycle in infant stepping are driven primarily by brainstem
and spinal cord circuitry. In this study, we examine the
contribution of reflex activity during walking in people with
spinal cord injury (SCI) to better understand the effect of
spastic reflexes on motor tasks such as walking.
We recruited 10 people with incomplete SCI and 5 healthy
control subjects to study the effects of hip angle
perturbations on walking. To study the effects of imposed
hip extension, the subjects walked on an instrumented splitbelt treadmill (Bertec). At certain random steps, one of the
treadmill belts accelerated briefly to increase hip extension
during mid to late stance. We observed that following a hip
extension perturbation, the swing phase was advanced and
the limb advanced farther in both subject groups. The
moment and EMG data revealed an increase in peak hip
and ankle moment and EMG activity in the thigh and shank
muscles as a result of the hip extension perturbation. In
controlling walking, the subjects appeared to regulate limb
position at heel strike. Hip angle at heel strike remained the
same regardless of the size of the perturbation and there
was a strong correlation between the stride length and hip
angle at toe off. The timing of the heel strike was also well
controlled. Subjects increased the swing speed such that in
control subjects, a further stride was taken in the same
amount of time, regardless of the size of the perturbation. In
the SCI subject group, the swing speed was increased, but
the time in the swing phase was still increased slightly,
possibly due to limitations in generating the forces needed
to further increase swing speed. This increased in swing
speed was only observed in the test leg in the control group,
but was observed in both legs (test and contralateral) in the
SCI subjects, resulting in an increase in time in double
stance phase. The limb position and timing at heel strike
25
Poster Sessions
Full Abstracts
has emerged to be an important control parameter in both subject
groups. Our results suggested evidence of both reflex activity and
cortical control of stepping in subjects with incomplete SCI.
1-E-71 Motor resonance during postural imbalance
observation
Thiago Lemos1, Banty Tia2, Ghislain Saunier1, Sebastian Hofle1,
Luís A Imbiriba1, Thierry Pozzo3, Claudia D Vargas1
1
Universidade Federal do Rio de Janeiro, 2Université de
Bourgogne, 3Université de Bourgogne and Italian Institute of
Technology
Neurophysiological as well as behavioral studies provided
evidences that the cerebral cortex may be involved in the coding
of motor aspects of postural control. In fact, postural control is a
complex motor behavior, in which both cognitive and affective
states are reflected. We hypothesized that motor representations
of postural control could be accessed during action observation,
trough a motor resonance mechanism - i.e., the motoric
translation of visual aspects of action. Motor resonance during
action observation paradigms has been studied via measurement
of a particular EEG oscillatory pattern, the so called Mu rhythm
(8-12 Hz frequency range) desynchronization.
Objective: the aim of this study was to investigate Mu rhythm
changes during the observation of postural imbalance situations.
Methods: four point-light display (PLD) animations were
constructed: biological quiet stance (QB); biological imbalance
stance (using an unstable board as support, IB); scrambled quiet
stance (QS); and scrambled imbalance stance (IS). Ten
individuals (males, 20-39 years) were asked to rest comfortably in
a sitting position, in a dimly light room, with a 19' LCD screen in
front of them. The experimental sessions comprised the random
presentation of the four PLD (3 s duration, 64 times each). A 128channels Geodesic system was used for EEG acquisition.
Analysis was performed in a region of interest comprising the
centro-parietal electrodes (CP3, CPz and CP4). EEGLAB tools
were used for data pre-processing, together with MATLAB script
for frequency analysis (Fast Fourier Transformation) in the 8-12
band frequency range. Frequency power changes were
calculated trough the logarithm transformation of power ratios
[quiet stance ratio = QB/QS; imbalance stance ratio = IB/IS;
expressed in arbitrary units (a.u.)]. One-sample t test was used to
search differences between log ratio and zero, since log ratio
lower than zero characterizes desynchronization, and ratios equal
to zero means no changes. Wilcoxon matched pairs test was
applied for testing differences between quiet stance and
imbalance ratios. Both tests were run assuming p < 0.05. Data
was presented as mean ± SE.
Results: significant difference from zero was found only for
imbalance ratio in the CP3 electrode (-0.39 ± 0.1 a.u.; p = 0.014);
there was also a difference between imbalance and quiet stance
ratio (0.17 ± 0.2 a.u.) for this electrode (p = 0.017). CPz and CP4
did not show significant Mu rhythm desynchronization (p > 0.05).
Conclusions: the observation of postural imbalance situations
promotes motor resonance as evidenced trough Mu rhythm
desynchronization within the left centro-parietal region. This result
suggests that motor representations of postural control are
accessed during the passive observation of imbalance situations.
Underscoring the role of cognitive states on balance control could
contribute to a higher understanding of the organization of
balance strategies.
1-E-72 Characterization of the nonlinearity in ankle
reflex stiffness dynamics
Kian Jalaleddini1, Ehsan Sobhani-Tehrani1, Robert E. Kearney1
26
1
McGill Univerisity
Joint stiffness defines the dynamic relation between the
position of the joint and the torque acting about it. Joint
stiffness plays an important role in the control of posture and
movement and thus its accurate identification is important to
understand the underlying mechanisms of postural and
movement control. Joint stiffness has two components:
intrinsic stiffness that is due to the mechanical properties of
the joint, active muscles and passive tissues; and reflex
stiffness which is due to the modulation of muscle activation
as a result of stretch reflex feedback through muscle
afferents. Decomposition of the recorded ankle joint torque
into its intrinsic and reflex subcomponents is a challenging
problem since they occur and change together.
To address this, our laboratory has developed a system
identification method to decompose ankle torque into
intrinsic and reflex torques. Using this we have
demonstrated that the reflex stiffness at the ankle joint is
uni-directionally sensitive to joint velocity. Moreover,
dorsiflexing displacements evoke large reflex responses
while plantarflexing displacements evoke little or no reflex
response. Therefore, reflex stiffness can be modelled as a
cascade of nonlinear static function and a linear dynamic
system, where the nonlinearity resembles a half wave
rectifier and the linear dynamics are low-pass in nature.
Subsequent experiments demonstrated that the reflex
dynamics, in particular the gain of the reflex pathway,
change dramatically as a function of the ankle operating
point defined by mean joint position and the level of
activation of the ankle muscles. However, the identification
methods were not able to distinguish whether there were
changes in the nonlinear element parameters associated
with changes in the operating point. This is important to
know since the reflex dynamics and static nonlinearity likely
reflect different physiological mechanisms.
This poster will present some recent methodological
developments made to improve the characterization of the
static nonlinear element of the reflex stiffness pathway.
These include:
1.
2.
3.
4.
Extensions to the identification algorithm to use a
subspace approach to estimate the parameters of
the linear and nonlinear elements in closed form.
Incorporating a spline representation for the
nonlinearity to improve the ability to model sharp
nonlinearities concisely. In addition, we have
developed objective methods of determining the
complexity of the representation.
Development of a novel perturbation waveform
whose probability distribution is designed to be
more informative in the characterization of the
shape of the nonlinearity.
Simulation studies demonstrate that with these
developments our methods accurately estimate
both the shape of the nonlinearity and linear
dynamics of the reflex stiffness. Moreover, pilot
experiments using these methods have shown that
the nonlinear-element does change significantly
with mean activation level. We will use these new
methods to systematically map out how the
nonlinear element changes as a function of the
operating point defined by activation level and
position.
Poster Sessions
Full Abstracts
1-E-73 Decoupling of upper and lower limb central
pattern generators during human crawling may highlight
the influence of cortical control on the upper limb
Michael MacLellan1, Yuri P Ivanenko1, Valentina La Scaleia1,
Francesco Lacquaniti1
1
Fondazione Santa Lucia
Generally, the motion of upper and lower limbs during upright
walking and adult hand-and-foot crawling is tightly coupled in a
one-to-one relationship. However, the investigation of cat
locomotion in which the fore and hindlimbs walk on separate
treadmills showed discrepancies: a one-to-one relationship
between limbs was not maintained when hindlimb treadmill speed
was manipulated (Akay et al. 2006). The purpose of the current
study is to use a crawling paradigm in healthy adults to examine
limb coupling patterns and suggest how upper and lower limb
central patter generators (CPGs) are controlled in a coordinated
manner. Ten healthy adults performed hands-and-feet crawling
on 2 separate treadmills where treadmill speed was manipulated
for the upper (1, 2 and 3 km/h) and lower (1, 2 and 3 km/h) limbs
totalling in 9 different speeds combinations. As well, 2 extreme
speed difference conditions were performed (upper limb treadmill
at 0.5 km/h, lower limb at 3 km/h and vice versa) in order to
encourage decoupling of the movements between the upper and
lower limbs. Full body 3D kinematics were recorded using a Vicon
motion capture system. When the treadmill speed for upper and
lower limbs was not matched, we observed in most cases a 1:1
relationship between the limbs. To achieve this 1:1 relationship,
increases in treadmill speed were accompanied with decreases in
stride time. Interestingly, relative stance and swing durations in
the lower limbs were adjusted in every treadmill speed
combination. In contrast, relative stance and swing durations in
the upper limb were only affected by changes of the upper limb
treadmill speed. Thus, the upper limb control seems to be more
independent than that of the lower limb. In extreme treadmill
speed differences, decoupling of upper and lower limbs was more
likely to occur when the upper limbs were at lower speeds (8/10
participants) as opposed to the lower limbs (2/10 participants).
Similarly, decoupling was always at lower speeds as in cats.
However, it occurred in humans in most cases for upper limbs
while decoupling in cats occurred in the hindlimb. Akay et al.
(2006) suggested that this decoupling in cats may be accounted
for by asymmetric neural pathways connecting the cervical and
lumbosacral spinal enlargements. The differences seen in
humans may suggest that the preference for the arms to
decouple may be due to a larger influence of cortical control on
the upper limb which allows for an overriding of spinal CPG
control. Supported by the Italian Ministry of Health, Italian Space
Agency (CRUSOE grant) and EU FP7-ICT program
(MINDWALKER and AMARSi grants).
1-E-74 Changes in the spinal motor output for stepping
during development from infant to adult
Yuri Ivanenko1, Nadia Dominici1, Germana Cappellini1, Ambrogio
Di Paolo2, Carlo Giannini3, Richard E Poppele4, Francesco
Lacquaniti2
1
Santa Lucia Foundation, 2University of Rome Tor Vergata,
Sant’Eugenio Hospital, 4University of Minnesota
3
Human stepping movement patterns emerge in utero and show
several milestones during development to independent walking.
Newborns that are adequately supported exhibit stepping
behavior that resembles walking, and the behavior can persist for
weeks. We examined the spinal locomotor output by mapping the
distribution of motoneuron activity in the lumbosacral spinal cord
during stepping in newborns, toddlers, preschoolers and adults.
Using both kinematic and electromyographic (EMG) recordings
from 24 bilateral leg muscles we reconstructed the patterns
of spinal segmental ventral root output and approximate
motoneuron activation in the lumbosacral spinal cord using
myotomal charts available for segmental innervation of
lower limb muscles in human neonates and adults. Newborn
stepping is characterized by an alternating bilateral motor
output with only two major components that are active at all
lumbosacral levels of the spinal cord. This pattern was
similar across different cycle durations of neonate stepping
and is consistent with a simpler organization of neuronal
networks. In toddlers, the step-related activity migrates to
the sacral cord segments, while the lumbar motoneurons
are separately activated at touch-down. In the adult, the
lumbar and sacral patterns become more dissociated with
shorter activation times. The results also showed that the
main features of the neonate stepping could not be
modelled by assuming crosstalk in 'mature' EMG recordings
or could not be reproduced by simply imitating the
appropriate biomechanical conditions. When we asked
adults to reproduce neonate stepping conditions (body
unloading, bent posture, slow speed) we found that the
differences between neonatal and mature patterns could not
be simply accounted for by biomechanical differences. We
conclude from these results that the development of human
locomotion from the neonate to the toddler involves a
reorganization of the spinal circuitry. The rostrocaudal
coactivation of motoneurons seen in the neonate is no
longer apparent in the toddler as the lumbar and sacral
motoneurons assume separate activation patterns. The
separation becomes more prominent with further
development with progressively shorter motoneuron
activations. The alternating spinal motor output in neonates
and its differential segmental maturation may suggest a
common central pattern generation circuitry development in
humans and other mammals. Supported by the Italian
Ministry of Health and EU FP7-ICT program
(MINDWALKER and AMARSi grants).
1-E-75 Error-signals driving locomotor adaptation:
Effects of peripheral nerve stimulation on ankle
control during walking in an elastic force field
Julia Choi1, Jesper Lundbye-Jensen1, Laurent J Bouyer2,
Jens Bo Nielsen1
1
University of Copenhagen, 2Université Laval
Proprioceptive and cutaneous feedback plays an important
role in modifying the ongoing motor pattern during
locomotion. In addition to feedback mechanisms, the
nervous system must also adapt movements based on
mismatch between intended and actual outcomes - this
reduces errors in the feedforward command for subsequent
movements. We hypothesize that cutaneous signals provide
a mechanism for monitoring contact forces during walking,
and that mismatch between expected and actual force
drives adaptation. Here we degraded sensory afferent input
using non-invasive repetitive nerve stimulation and
examined the effects on 1) cutaneous sensation, 2) ankle
proprioception and force production, and 3) ankle adaptation
during walking in a force field. We stimulated the superficial
peroneal (foot dorsum) and medial plantar nerves (foot
sole). Repetitive electrical nerve stimulation was
administered with a constant current stimulator that
delivered 1 ms pulses at 80 Hz. The intensity was set at a
level where subject perceived a 'tingling' sensation that
radiated towards the toes. EXPERIMENT 1: Touch
thresholds were measured using Von Frey monofilaments
on foot dorsum, and plantar side of great toe and heel.
Vibration threshold on great toe was determined using
27
Poster Sessions
Full Abstracts
Vibratron II. Our results suggest that 80 Hz electrical nerve
stimulation selectively worsened cutaneous sensation, without
affecting vibration sensation. EXPERIMENT 2: We tested the
effects of degraded cutaneous sensation on the accuracy and
precision of ankle force production. Subjects were seated and
instructed to maintain an isometric dorsiflexion force equivalent to
10% maximal voluntary contraction with the right foot attached to
an instrumented pedal. In addition, an active position-matching
task was used to test whether proprioception at the ankle joint
(above stimulation site) could be affected. We showed that
subjects were unable to precisely maintain a constant dorsiflexion
force with peroneal nerve stimulation without visual feedback.
This was not a general effect of repetitive stimulation because
subjects could maintain dorsiflexion force precisely with plantar
nerve stimulation. Proprioception at the ankle joint was also
unaffected. These results suggest that cutaneous feedback is
required for precise ankle force control. EXPERIMENT 3: Finally,
we tested the effects of degraded cutaneous feedback on walking
adaptation. Subjects were exposed to a position-dependent force
field produced by an elastic band attached between the back of
the shoe and the calf band of an ankle-foot orthosis, which
resisted ankle dorsiflexion. In the baseline period, subjects
walked with no external force at the ankle. In the adaptation
period, an elastic was attached and subjects were instructed to
'resist the force and walk normally' for about 30 strides. In the
post-adaptation period, the elastic force was removed to test for
after-effects. Each subject repeated the adaptation paradigm in a
randomly assigned sequence of conditions (peroneal nerve,
plantar nerve, no stimulation). Results indicate that adaptation of
ankle dorsiflexion and tibialis anterior activation during force field
walking was reduced with peroneal nerve stimulation, whereas
plantar nerve stimulation had no effects. In conclusion, we
showed that cutaneous feedback plays an important role in
driving human locomotor adaptation. Identifying salient neural
signals that drive adaptation could increase the efficiency of gait
rehabilitation.
F - Fundamentals of Motor Control
1-F-33 Effects of ageing on attentional cost and internal
models of proprioceptive control of movement
Matthieu P. Boisgontier1, Isabelle Olivier1, Vincent Nougier1
1
UJF-Grenoble 1 / CNRS / TIMC-IMAG UMR 5525
It is still not clear whether motor control based on proprioceptive
information is altered in older adults as compared to young ones.
Previous studies showed strong and unexplained discrepancies
and did not conclude about a presumed age-related alteration of
proprioceptive control. In the present study, it was hypothesized
that the attentional cost of proprioceptive control was altered with
ageing. It was also hypothesized this presumed age-related
alteration of proprioceptive control was associated to an alteration
of internal models. To test the attentional hypothesis, one group
of young adults and one group of older adults performed a joint
position matching task in single and dual-task paradigms with
different difficulty levels of the secondary task to assess
attentional costs associated to proprioceptive control in these
populations. To identify the central mechanisms involved in the
proprioceptive control, we proposed to assess the effects of
ageing on internal models. To this aim, one group of young adults
and one group of older adults performed the same ankle
matching task in two speed conditions (self-selected and fast).
Error, temporal, and kinematic variables were the dependent
variables used to assess the matching performance. Results
showed that proprioceptive control was as accurate and as
28
consistent in older as in young adults for a single
proprioceptive task at a self-selected speed. However,
performing a secondary cognitive task and increasing the
difficulty of this secondary task evidenced an increased
attentional cost of proprioceptive control in older adults as
compared to young ones. The results also demonstrated
that internal models of proprioceptive control were altered
with ageing. Behavioural expressions of these alterations
were dependent upon the considered condition of speed. In
the self-selected speed condition, this alteration was
expressed through an increased number of corrective submovements in older adults as compared to young ones. This
strategy enabled them to reach a level of end-point
performance comparable to the one of the young adults. In
the fast speed condition, older adults were no more able to
compensate for their impaired internal models through
additional corrective sub-movements and therefore
decreased their proprioceptive control performance.These
results advocated for an increased attentional cost for
proprioceptive control in older adults as compared to young
adults. The present age-related alterations of ankle
proprioceptive control also suggested a decreased
resistance to attentional and speed stressors in adult ageing
and supported the fact that proprioceptive control is involved
in the frailty syndrome, i.e., a decreased resistance to
stressors, which characterizes older adults.
1-F-34 Mechanisms of oscillatory drive in human
motor control
Tjeerd Boonstra1, Michael Breakspear1
1
University of New South Wales
Oscillatory activity in the neurophysiological drive to the
muscle plays an important role in corticospinal control of
muscle synergies. In pathological conditions or effortful
activity, this may be exaggerated and become evident as a
tremor. Under many circumstances, however, oscillatory
activity may not be directly noticeable, but can be discerned
using spectral analysis of electromyography (EMG) and
electroencephalography (EEG). In particular, many studies
use corticospinal (EEG-EMG) and intermuscular (EMGEMG) coherence to assess oscillatory input to the muscle.
Although many oscillatory components in the input to the
muscle have been identified, the physiological mechanisms
that bring about these oscillations remain to be identified.
The ongoing debate on EMG rectification as a
preprocessing step in assessing oscillatory activity
underlines our incomplete understanding. A number of
studies have addressed the rational of EMG rectification and
the effects it has on power and coherence spectra, but no
general agreement has been reached. The discussion
seems to divide experimental from computational
approaches: Most empirical studies use and promote
rectification of surface EMG, whereas simulation studies
generally argue against rectification. The elucidation of this
technical issue therefore directly contributes to our
understanding of the mechanisms underlying oscillatory
neuronal activity. We use an integrated computational and
experimental approach to reconcile experimental and
computational approaches on the issue of EMG rectification.
We study the mechanisms involved in the generation of
common oscillatory activity in a basic model of two MU
pools that incorporate key physiological properties.
Simulated data are compared to empirical data of
intermuscular coherence acquired from healthy subjects
during quiet standing.
Poster Sessions
Full Abstracts
In particular, we investigate the role of a heterogeneous motor
unit action potential (MAUP) population on the translation of
oscillatory activity into surface EMG. We show that a
heterogeneous MAUP distribution distorts the transmission of
common input and oscillatory components are only observed in
rectified EMG. In contrast, using a homogeneous MUAP
distribution oscillatory input is evident in both rectified and nonrectified EMG. The experimental data showed that intermuscular
coherence was mainly discernable in rectified EMG, hence
providing empirical support for a heterogeneous distribution of
MUAPs. These findings show that oscillatory components are
manifested as modulations of high-frequency EMG content and
implicate that the shape of MUAPs is an essential parameter to
reconcile experimental and computational approaches.
1-F-35 Linking motor and reward: The correlation
between nucleus accumbens and primary motor cortex
during goal directed reaching
Justin Sanchez1, Eric Pohlmeyer1, Babak Mahmoudi1, Shijia
Geng1, Noeline Prins1
1
University of Miami
A central role of the brain is to analyze the environment and make
decisions that lead to reward and survival. Access to hierarchical
levels of processing including sensory, motor, and reward
systems have enabled organisms to become more responsive to
their environments and ultimately shapes future motor behavior.
The use of motor and reward neural processing to carryout goaldirected behaviors has the effect of expanding capabilities for
responding to dynamical events in the environment. Our studies
seek to develop a deeper understanding of the simultaneous
neural coding among the nucleus accumbens (NAcc) and the
primary motor cortex (M1) during goal directed reaching tasks.
This includes characterization of the fine timing relationships and
incremental evaluative feedback from the NAcc and how it relates
to excitatory or inhibitory motor activation that is conducive or
non-conducive for obtaining reward. To study the relationship, we
implanted microwire arrays (50μm tungsten wires, 16 in each
target) into the hand-arm region of M1 and into the NAcc of
common marmoset monkeys (Callithrix jacchus). The animals
were trained to perform a reaching and grasping task. The
animals were required to place their reaching hand in a start
position. Upon doing this, they were presented with either a
desirable food item (worm) or a non-food item (wooden bead).
When presented a worm, the monkeys would reach and grasp the
food. Alternatively, when presented with a bead the animal would
not reach. During these experiments, arm position was tracked
and time synchronized with single neuron recordings from both
the cortical and subcortical targets. To characterize the coding in
both neural structures, all reaching and non-reaching behaviors
were time aligned and perievent time histograms were
constructed. In these experiments, we found that the time courses
of the neuromodulation in the M1 and NAcc as they are related to
reaching vs non-reaching are quite different. For example, in
rewarding trials, the M1 neurons ramp up their activity during the
reach. The firing of NAcc remains low. During non-rewarding
trials, the animal does not reach and there is a decrease in M1
activity. This decrease in M1 was concurrent with a persistent
increase in the NAcc activity signaling a shutdown in the motor
output. To test the temporal coding and representation of both M1
and NAcc during reaching in the paradigm described above, we
built linear classification models (Wiener Filter) for just the global
reach and non-reach trials in four experimental sessions. For
each session, half the trials were randomly selected and used to
train the model, which was tested by its classification of the other
50% of the trials. In each case, we scanned the performance
using between 0-400ms of data as the classifier input.
Interestingly, the M1 classification continued to improve as more
data from any arm movement was included in the trial
indicating a longer time scale. However, the NAcc classifier
accuracy appeared consistent throughout the entire time
period, suggesting the time course of the NAcc activity
provides short term feedback compared to M1.
1-F-36 Haptic illusions: A window into motor
control
Alexander Terekhov1, Vincent Hayward1
1
Université Pierre et Marie Curie
Haptic illusions: a window into motor control. Alexander V.
Terekhov and Vincent Hayward Dexterity and perceptual
proficiency requires knowledge about the properties of one's
own body. In one school, it is assumed that this knowledge
can be approximate and that the brain controls movement
through a cascade of stabilizing mechanisms. This view is
adopted in the equilibrium point hypothesis. Another school
insists on the necessity for the brain to utilize elaborate
internal models describing the dynamics of the body and of
the external world. It is only under this condition that optimal
mappings between sensory inputs and motor outputs can
arise. These two conflicting views are successful are
successful at explaining many experimental observations.
We suggest that the models used by the brain need to be
accurate, but only to a point. Attempting to characterize their
accuracy is an enticing and challenging task which cannot
be solved by motor control studies alone. Often, behaviors
can be explained by inaccuracies in the models or by the
particularities of the cost functions governing task execution.
Haptic perception provides an opportunity to gauge the
accuracy of internal models because it often depends on
motor activity involving multiple segments. For instance,
when experiencing shape by fingertip exploration, the brain
must combine information from multiple joints and account
for the geometry of the segments engaged in the
movements. Our ability to estimate shapes by touch
suggests that this process is quite accurate. We can
discriminate radii differences of about 1 mm in a stimulus of
60 mm in diameter. We found three haptic illusions related
to these questions. The first illusion occurs when exploring a
curved surface with a fingertip where different strategies for
fingertip orientation are employed. One can move as to
maintain the fingertip orientation fixed in space; or one can
keep the point of the contact stationary on the skin, thereby
causing the fingertip orientation to track the normal of the
surface. If the internal model of the body was accurate, then
the mode of exploration should not influence the percept of
curvature. Yet, experimental studies show that it is not the
case. Subjects tend to perceive the same surface to be
more or less curved depending on the mode of exploration.
The second illusion arises when exploring a concave
surface with a thin stick. Provided that the exploratory
movements involve rotations of the wrist, subjects tend to
perceive the same object to have a larger size when they
explore it with a short stick than when they explore it with a
longer stick. The illusion becomes weak, or even vanishes,
if the stick orientation remains the same during the
exploration. Finally, when an object is explored with a thick
stick, subjects size judgements depend on the stick
thickness. With a thicker stick, a convex object seem larger -- and a concave object seem smaller --- than with a thinner
stick. If the subjects correctly used available information
about the stick length and width this illusion would not occur.
In all these examples, the difference in the perceived radius
of the object was of the order of 10 mm for a stimulus
diameter of 60 mm. These effects suggests that in spite of
the considerable capacity of the haptic modality, the brain
29
Poster Sessions
Full Abstracts
does not always integrate information with the consequence an
internal model of the body or of the environment can be rather
imprecise.
1-F-37 Controlling output from the motor cortices:
Avoiding unintentional movement using the null space
Matthew Kaufman1, Mark M Churchland2, Stephen I Ryu3,
Krishna V Shenoy1
1
Stanford University, 2Columbia University, 3Palo Alto Medical
Foundation
Motor and premotor cortex are involved in the execution of
reaching movements, but are also active during preparation of
those movements. Given that the arm remains motionless during
preparation, we ask: why does preparatory activity not cause
movement? How is preparatory activity attenuated from premotor
cortex to motor cortex to the spinal cord to the muscles? One
possible mechanism is that activity is prevented from escaping
each area because output is suppressed by local inhibition. This
mechanism is used in the brainstem for saccade control. We have
previously searched for 'gating' by inhibitory interneurons in
primary motor (M1) and dorsal premotor cortex (PMd) of
monkeys. We could not find evidence for inhibitory gating in either
PMd (Kaufman et al., 2010, J Neurophys) or M1 (Kaufman et al.,
2010, SFN). Specifically, we did not find inhibitory neurons that
were highly active during motor preparation but quiet during
movement itself. We thus propose and test an alternate model.
Since there are many more neurons than muscles, there must be
many possible neural states for any given pattern of muscle
forces. We therefore predict that there will be 'output-null'
dimensions in neural space, such that changes in activity along
these dimensions do not alter muscle output. These output-null
dimensions could be exploited to permit motor preparation without
causing movement. Here, we attempt to identify output-null
dimensions using four neural datasets from two monkeys
performing a delayed-reach task. A target appeared while the
animal fixated. The monkeys were required to withhold movement
until a go cue, then reach to the target. Recordings were
performed in PMd and M1, using single moveable electrodes or
dual chronically-implanted 96-electrode 'Utah' arrays. We tested
whether some neural dimensions might be 'output-potent,'
identifiable because activity in these dimensions correlates with
muscle activity, while other dimensions might be output-null. We
predicted that neural activity would preferentially explore the
output-null dimensions during preparation, avoiding the outputpotent dimensions in order to avoid causing movement. To test
this prediction, we first reduced the dimensionality of both the
neural data (to six) and EMG data (to three) using PCA. We then
regressed the reduced-D EMG data onto the reduced-D neural
data, yielding the putative output-potent dimensions; the
remaining dimensions were taken as output-null. Finally, we
measured how far neural activity traveled in output-null
dimensions vs. output-potent dimensions. As predicted, we found
that preparatory activity travels mostly in output-null (non-musclelike) dimensions. This bias toward output-null dimensions was
substantial: during preparation the changes in neural activity were
2.4-9.8 times as large in the output-null dimensions as in the
output-potent dimensions (p < 0.05 in each dataset). This
mechanism therefore appears to explain much of how we can
prepare to move while holding still. We further tested whether a
similar mechanism might explain how M1 can display less
preparatory activity than PMd, even though the two are
connected. As predicted, PMd preparatory activity preferentially
lay in output-null dimensions with respect to M1, by a factor of
2.2-2.4 (p < 0.05 in both datasets). Selective use of output-null vs.
output-potent dimensions may thus constitute an important
mechanism for controlling communication between brain areas
and output to muscles.
30
1-F-38 Removing motion information reduces the
latency of manual movement corrections
Leonie Oostwoud Wijdenes1, Eli Brenner1, Jeroen Smeets1
1
VU University Amsterdam
If a target instantaneously changes its position while a hand
is moving towards it, short latency movement adjustments
occur so that the hand ends at the new target position.
Instantaneous changes in target position are perceived as if
the target moved from one position to the other. If the target
changes position during a saccade, the target motion is not
perceived, but subjects do adjust for the target
displacement. Previously reported response latencies
suggest that the latency to correct for a perceived target
jump may be faster than the latency to correct for target
jumps during saccades. We investigated whether the
perception of target motion decreases the latency of
movement corrections. Motion perception was disrupted by
introducing a 100 ms temporal gap between the two
successive target presentations. During the gap there was
no visible target. If motion information speeds up responses,
we expect longer response latencies when motion
information is removed by introducing a gap.
Subjects made horizontal pointing movements over
approximately 70 cm in a setup with back-projected targets
on a large screen. The target either jumped instantaneously
3 cm up or down (step condition) or disappeared and only
reappeared at the vertically shifted position 100 ms later
(gap condition). We determined the latency of responses to
the new target position from the difference between the
average vertical acceleration profiles to targets that jumped
up and down. The intensity of the response was defined as
the maximum of this difference in acceleration.
The results show that the latency to respond to the new
target position was 9 ms shorter after the gap. The
movement corrections in the gap condition were less
vigorous than those in the step condition, while there was no
difference in movement time. We conclude that movement
corrections are not guided by the perception of target
motion, but by a change in target position. The faster
response latencies in the gap condition might be explained
by a mechanism similar to the one causing the gap-effect in
initiating eye and hand movements. By our knowledge this
is the first demonstration of a gap-effect for fast online
movement corrections.
1-F-39 Individual Purkinje cell simple spike
discharge is full of errors
Laurentiu Popa1, Angela L Hewitt1, Timothy J Ebner1
1
University of Minnesota
Although the cerebellum has been hypothesized to play a
major role in error processing, evidence for cerebellar error
signals remains inconclusive. Most efforts aimed at
understanding cerebellar error encoding concentrated on
the complex spike discharge and yielded contradictory
results. A limited number of studies suggest that the simple
spike (SS) activity of Purkinje cells (PCs) could encode error
representations. The present study shows that SS discharge
encodes performance errors independently of kinematic
signals.
Two rhesus monkeys were trained to track a pseudorandomly moving target on a vertical screen using a two
joint manipulandum. The difficulty of the experimental
paradigm induced numerous excursions outside the target
that had to be corrected within 500 ms, resulting in a
Poster Sessions
Full Abstracts
behavior that required continuous processing of performance
errors. During tracking, we collected extracellular recordings from
120 PCs in the intermediate and lateral zones anterior to the
primary fissure.
Given that the movement goal is to maintain the cursor within the
target, cursor movement relative to the target provides well
defined signals related to the sensory consequences of motor
performance. Multiple linear regressions including temporal lags
from -500 ms to 500 ms were used to evaluate the PC simple
spike firing in relation to the following signals: cursor position
relative to the target center (XE, YE), cursor distance to the target
center (RE), relative direction of the cursor movement to the
direction of the target movement (DE), and the difference
between the cursor direction movement and the direction of the
error position vector (PDE). To insure that the error signals were
independent of the kinematic signals, the linear regressions were
performed on the residuals obtained after fitting a kinematic
model including position, velocity, and speed. In more than 95%
of the recorded cells, the SS discharge is modulated by each of
the error parameters. The error parameters resolve an amount of
firing variability comparable to that explained by the kinematic
model. The best R2 adjusted values for the error and kinematic
models obtained on the complementary residuals are highly
correlated, showing that kinematic and error signals are
integrated at the PC level. Analysis of the correlation profiles'
properties, show that more than 80% of the PCs have a bimodal
fit profile with two local maxima, one at predictive and one at
feedback lag values. Moreover, 66% of the bimodal timing profiles
show opposite modulations in the SS activity for the feedback and
predictive signals, possibly a substrate to generate a sensory
prediction error. Remarkably, individual kinematic signals,
independent of the error parameters, generate similar bimodal fit
profiles and opposite SS discharge modulations between
feedback and predictive representations. Therefore, SS discharge
encodes multiple error and kinematic parameters, each
parameter being represented as a combination of predictive and
feedback signals.
This study shows that the SS discharge of individual PCs
encodes motor errors that are integrated with kinematic signals.
The bimodal timing of the individual parameters found in the vast
majority of the PCs could provide the cellular substrate for the
computation of sensory prediction errors required by forward
internal model motor control theories.
Supported in part by NIH grants RO1 NS18338, T32 GM008244,
F30 NS071686.
1-F-40 Modulation of cerebellar function does not affect
the timing or magnitude of motor surround inhibition
Anna Sadnicka1, Panagiotis Kassavetis1, John C Rothwell1, Mark
J Edwards1
1
UCL Institute of Neurology
Background: Surround inhibition is a physiological mechanism to
focus neuronal activity in the nervous system. Initially described in
the sensory system it has been more recently demonstrated in
the motor system using transcranial magnetic stimulation (TMS)
over the hand region of the motor cortex around the onset of a
small voluntary movement of the index finger. At this time motor
evoked potentials (MEPs) from surrounding muscles (abductor
digiti minimi (ADM) and abductor policis brevis (APB)) are
reported to diminish in size. The neural mechanism underlying
motor surround inhibition (mSI) is unknown. We explored the role
of the cerebellum in the generation of mSI using transcranial
direct current stimulation of the cerebellum (cDC).
Method: 12 healthy subjects took part in a cross-over study in
which they received sham, anodal or cathodal cDC in a
randomised order, each session separated by a week. mSI
was tested before stimulation and at 0 min and 20 min after
cDC. The magnitude of mSI in ADM and APB was
calculated at 0, 50, 100 and 200 msec after the onset of
index finger movement (FDI was the active muscle) as a
normalised ratio compared to the mean rest MEP of each
muscle.
Results: We could reliably record mSI in ADM which was
found to be greatest at 0 msec and still present at 50 msec.
Repeated measures ANOVA demonstrated a significant
effect in ADM at time 0 and 50 msec (p<0.0001). Anodal
and cathodal cDC did not affect the timing or magnitude of
mSI observed in the ADM muscle (p = 0.31). Individual
variability of mSI measured on separate occasions is
characterised. We did not find evidence of the previously
reported mSI in the APB muscle.
Conclusions: Modulation of the cerebellum with cDC does
not affect mSI as assessed by TMS. Perhaps the
phenomenon of mSI is generated across the sensorimotor
network and transient weak modulation of one node of this
network is not powerful enough to disrupt the mechanism. In
addition we found no evidence that APB muscle is inhibited
using this paradigm, which is in contrast to previous
findings. Further experiments are needed to fully
characterise mSI and its adaptability as better
understanding of this mechanism may generate new
treatments in disorders of motor control such as dystonia
where impaired SI occurs.
1-F-41 Excitatory amino acid transporters
regionally sculpt the beam-like response in the
cerebellar cortex
Samuel Cramer1, Wangcai Gao1, Gang Chen1, Timothy J
Ebner1
1
University of Minnesota
The role of parallel fibers (PFs) in cerebellar physiology
remains controversial. Early anatomical and physiological
studies of the cerebellum inspired a model whereby PF
activation results in post-synaptic excitation of beams of
Purkinje cells (PCs), the "beam" hypothesis. However,
functional studies have yet to demonstrate activation of PCs
along beams of PFs. Instead, experiments suggest that the
ascending limb of the granule cell (GC) axon activates
patches of PCs, while PFs predominately modulate PC
activity. This alternative model is the "radial" hypothesis.
Proponents of this model postulate that stronger synapses
exist between the ascending limb of the GC axon and PCs,
though this is not uniformly supported by experimental
evidence. Others have proposed that inhibition by molecular
layer interneurons suppresses PF activation of on-beam
PCs, producing patch-like responses. Given that PFs are
central components in many theories of cerebellar function
and yet remain poorly understood, this study reexamined
the question of patches versus beams.
Employing Ca2 optical imaging in the mouse cerebellar
cortex in vivo, electrical stimulation of mossy fibers or GCs
evokes a beam-like response. Given the potential problems
with electrical stimulation, we also utilized a more selective
pharmacological approach to activate GCs. By
systematically microinjecting glutamate or NMDA in the
cerebellar cortex, we show that a maximal beam-like
response is evoked by these agonists in the superficial
granular layer.
Next we examined the responses to peripheral stimulation.
Remarkably, ipsilateral forepaw stimulation evokes a beam-
31
Poster Sessions
Full Abstracts
like response in Crus I. These results are in contrast to the
previously reported patch-like activation of Crus II evoked by
vibrissal pad stimulation. These regional differences in response
to sensory input prompted further investigation into the underlying
mechanism. The GABAA blocker, GABAzine, failed to convert
patch-like activity in Crus II to a beam-like response, suggesting
that molecular layer inhibition is not involved. However, the
excitatory amino acid transporter (EAAT) blocker, DL-TBOA,
rapidly converts patch-like activity in Crus II into a beam-like
response, suggesting that local extracellular glutamate
concentration determines the geometry of cerebellar activation in
response to GC input.
Finally, given its role as the main EAAT in PCs and its
parasagittal and folia dependent expression profile, we
investigated the role of EAAT isoform 4 (EAAT4) in determining
the response geometry in the cerebellar cortex. Using transgenic
mice that express GFP under control of the EAAT4 promoter, we
show that peripherally evoked patch-like responses in Crus II are
aligned between parasagittal bands of EAAT4.
This is the first study to demonstrate in vivo that beam-like
responses are evoked in the cerebellar cortex to peripheral,
mossy fiber, and GC stimulation. Furthermore, the results show
that the response to GC input is highly dependent on extracellular
glutamate and its local regulation by EAATs.
Supported in part by NIH grants T32 GM008244, T32 GM008471,
RO1 NS18338
1-F-42
Tool kinematics planning in humans
Tsuyoshi Ikegami1, Gowrishankar Ganesh1
1
NICT/ATR
While recent studies have shown that tool usage is not restricted
to human activities, the exceptional capability of man to learn
using a wide array of tools still distinguishes him from other
animals. During use of hand held tools, locating the tip of a tool
with respect to one's hand is the first and probably the most
fundamental step towards the use of the tool. Humans can
perform this localization immediately, such that while holding a
tool they can immediately plan new hand trajectories to
accommodate for the tool kinematics. How is this localization and
planning performed? To answer this question we asked subjects
to make arm movements to reach different targets with the tip of
'tools' of different size and orientation which they held in their
hand. The subjects were shown their hand position (with the tool)
before the movement start but were not given any feedback
during or after their movement on their performance. The target
was visible throughout the movement trial period. We looked at
the error distribution in the repeated reaching tasks to check for
the nature of planning and the coordinate system of planning
involved in the tool task. Our results show that humans are very
good at incorporating the tool orientation into their planning.
However, they consistently over compensate for the length of the
tool which we show as not to be due to an error in visual
perception but due to error in motor planning. Further the length
of arm movements were consistently under estimated in the
presence of a tool. This poster presents these results and
discusses the planning coordinates and planning structure which
can explain the observed results.
1-F-43 Intermittency in visual information acquisition in
continuous tracking task
Yasuyuki Inoue1, Yutaka Sakaguchi1
1
The University of Electro-Communications
In a visuo-manual tracking task, hand motion often precedes the
target motion, especially when the target motion is periodic with a
32
frequency greater than 0.5 Hz. This suggests that human
brain performs predictive or feed-forward motor control even
in a "tracking task." In order to continue accurate tracking,
however, we have to refer to visual information related to
target and hand motions for movement adjustment. Here, it
is unclear how the visual information operates in the motor
control: Do we refer to visual information continuously or
intermittently? If intermittently, when do we catch the
information? To answer these questions, we conducted a
behavioral experiment where visual information of target
was partially removed. We found that hand motion changed
when the target disappeared in the specific phase of
sinusoidal trajectory. In our experiment, green (target) and
red (marker) light spots were presented on a vertical screen
(located 2 m apart from the subject), where the marker
position reflected the hand position on a one-dimensional
sliding rail. The subject was asked to move the right hand so
that the red marker tracked the green target as accurately
as possible. The target moved sinusoidally in a vertical
direction and its frequency was 1.2 Hz. Amplitude of the
target movement was 20 cm on the screen, which
corresponded to 6.7 cm hand movement. In the control
condition, the target was visible throughout the trial. In the
vision removal condition, the target disappeared while it was
located within a specific phase range. The position of the
removal range was 0 (top), 90 (center), 180 (bottom) or 270
(center) degrees and its width was 60 or 90 degrees. The
marker was always visible regardless of the conditions. In
the control condition, the phase of hand motion was
generally advanced with that of target, but their phase
difference fluctuated periodically: The average phase
difference was the largest roughly around center points of
the sinusoids (i.e., where the target velocity was highest),
whereas it was nearly zero around the turning points (i.e.,
where the target motion reversed). This confirms that our
brain performs the tracking task in a predictive manner, but
suggests that the predictive motor control was performed in
a non-stationary manner. In the vision removal condition, on
the other hand, this tendency remarkably changed, and it
depended on the phase of the vision removal. Especially,
removal around one turning point (top/bottom) increased the
variance of phase difference in the other turning point
(bottom/top). This suggests that out brain refers to visual
information intermittently synchronized with the motion
phase.
1-F-44 Preparing to grasp pleasant and unpleasant
objects
Laura Oliveira1, Luis Aureliano Imbiriba2, Maitê Mello
Russo1, Erika de Carvalho Rodrigues3, Anaelli Aparecida
Nogueira-Campos1, Mirtes Garcia Pereira4, Eliane Volchan1,
Cláudia Domingues Vargas1
1
IBCCF/UFRJ, 2EEFD/UFRJ, 3UNISUAM, 4UFF
Prediction of the shape and size of objects is incorporated in
the planning of actions toward them. Here, we examine if
the valence of a to-be-grasped object also exerts an
influence on movement planning. Readiness potential, an
event-related marker of motor preparation, was recorded by
means of electroencephalogram before grasping pleasant,
neutral and unpleasant objects (chocolate, clip papers and
cockroach for e.g.). Items used were balanced in weight and
placed inside transparent cups to prompt a similar grip
among trials. To appraise significant differences between
the mean values of readiness potential amplitude, statistical
analysis was performed using a three-way repeated
measures ANOVA with channel (C3, C4 and Cz), wave
segment (Early readiness potential and Late readiness
Poster Sessions
Full Abstracts
potential) and category (pleasant, unpleasant and neutral) as
factors. A main effect was found for category (p<0.01). Post-hoc
analyses revealed that the mean readiness potential amplitude for
the unpleasant condition was significantly higher than that of the
neutral condition. Furthermore, the amplitude was significantly
lower for the pleasant condition than for the neutral condition. No
other main effect or interaction was significant. Thus, we show for
the first time that the valence of the object to which an action is
directed affects the sensorimotor cortex activity preceding its
grasping. Smaller readiness potential amplitudes preceding the
interaction with pleasant objects could imply in the recruitment of
pre-set motor repertoires, whereas higher amplitudes found for
unpleasant objects could emerge from a discrepancy between the
required action and the object's aversiveness. Theories of motor
control propose that planning involves the prediction of motor
commands outcomes. Our results indicate that motor planning
also encodes the valence of an object with which one is about to
interact.
1-F-45
Do humans prefer to see their grasping points?
Dimitris Voudouris1, Eli Brenner1, Jeroen BJ Smeets1
1
VU University Amsterdam
In order to grasp an object the digits need to be placed at suitable
positions on its surface. The selection of such grasping points
depends on several factors. Here we examined whether being
able to see one of the selected grasping points is such a factor.
Subjects grasped large cylinders or oriented blocks that would
normally be grasped with the thumb visible and the index finger
occluded by the object in question. An opaque screen that hid the
thumb's usual grasping point was used to examine whether
people would choose a grip that was oriented differently to
maintain vision of the thumb's grasping point. A transparent
screen served as control. Occluding the thumb's grasping point
made subjects move more carefully (e.g. larger grip aperture) and
choose slightly different grip orientation. However, the change in
grip orientation was much too small to keep the thumb visible. We
conclude that humans do not particularly aim for visible grasping
points.
1-F-46 Cortical silent period duration and its
implications for surround inhibition of a hand muscle
Sahana Kukke1, Brach Poston1, Rainer Paine1, Sophia Francis2,
Mark Hallett1
1
National Institutes of Health, 2University of Maryland
Surround inhibition is a well-known mechanism in sensory system
physiology whereby an activated neuron decreases the activity of
immediately adjacent neurons to sharpen the localization of
stimulus information (Blakemore et al., 1970). This process
appears to be a fundamental pattern of neural organization
because it operates in various forms in practically every sensory
system (Nabet and Pinter, 1991). In the motor system, neural
processes analogous to surround inhibition have also been
identified (Hallett and Khoshbin, 1980) in which excitatory drive is
focused to muscles responsible for a given movement (agonist
muscles) while suppressed in muscles not relevant to the
movement (surround muscles). However, the specific
physiological mechanisms responsible for the generation of
surround inhibition in healthy human subjects are currently
unknown. The purpose of this study was to determine the
contribution of GABAB receptor mediated intracortical inhibition
as assessed by the cortical silent period (CSP) to the generation
of surround inhibition in the motor system. Eight healthy righthanded adults (5 women and 3 men, 29.8 ± 9 years) gave
consent to participate in this transcranial magnetic stimulation
(TMS) experiment. The task was to isometrically contract the
abductor digiti minimi (ADM, surround muscle) muscle at a
constant force level, and perform a concurrent index finger
flexion movement (agonist contraction) upon hearing a go
tone. TMS was applied to the contralateral primary motor
cortex at four times before and during the index finger
movement. In response to the TMS pulse, the motor evoked
potential (MEP) amplitude and the CSP duration of the ADM
muscle were measured. The significance of the timing of
TMS on the outcomes was tested using the non-parametric
Friedman test (ADM MEP), and a repeated measures oneway ANOVA (log-transformed ADM CSP duration). Post hoc
testing with Bonferroni adjustment for ADM MEP and with
the Dunnett adjustment for ADM CSP duration was used to
compare each of the three testing times after the go tone to
the testing point before the go tone. Results of this study
confirmed the presence of surround inhibition; the ADM
MEP was decreased during index finger movement initiation
compared to the ADM MEP prior to the go tone (p =
0.0141). In the presence of surround inhibition, an increase
in ADM CSP duration would implicate GABAB receptor
mediated intracortical inhibition as a potential mechanism of
surround inhibition. On the contrary, our results
demonstrated the CSP duration of the ADM decreased
during index finger movement initiation compared to the
CSP duration prior to the go tone (p = 0.0004). Therefore,
our findings indicate that GABAB receptor mediated
intracortical inhibition as measured by the duration of the
CSP does not contribute to the generation of surround
inhibition in ADM. Similar to previous studies (Beck and
Hallett, 2011), these results were able to exclude the
possible contribution of a specific cortical pathway to
surround inhibition, but unable to identify the pathway
responsible for the phenomenon. Future work is planned to
examine other cortical inhibitory and excitatory pathways
that may be responsible for surround inhibition in the human
motor system.
1-F-47 Grip force control strategies during slow
load variations
Thibault Giard1, Jean-Louis Thonnard1, Philippe Lefèvre1
1
University College London
The grip force (GF) adaptation strategies used to
manipulate an object during fast load perturbations (for
instance, caused by collisions) have already been described
in the literature (Serrien et al. 1999, White et al. 2011).
However, very few studies, have dealt with slow load
variations. In this respect, it has been demonstrated that
there is an important difference in grip force control between
(i) the manipulation of an object in presence of a tangential
torque and (ii) the manipulation of an object without
tangential torque (Crevecoeur et al. 2011). The main aims of
this study are to characterize the grip force control
strategies during slow load variations as well as to bring
new clues to better understand the results about the
compensation for torques during object manipulation. In
order to fulfill these objectives, the GF response to slow
increments in the tangential force (TF) and to slow
increments in tangential torque (TT) were tested separately
and compared. The difference in strategy between the two
conditions and the reaction time were analyzed.
A robotic arm has been used to produce a linear variation of
the load. Two 3-dimensional force and torque sensors have
been adapted on the end effector of the arm which was held
by the subjects (n = 6) using the precision grip. They were
blindfolded and were required to keep the end effector
vertically and to avoid any slip. The force was applied in
three different directions: upward (TF variation), forward (TT
variation) and backward (TT variation). Three durations
33
Poster Sessions
Full Abstracts
were used to reach the maximal force (1.5N in rotation and 9N in
translation): 10, 20 or 30 seconds. Each subject performed 4
repetitions for each condition.
Two distinct GF control strategies were observed : a "smooth
strategy" and a "stepwise strategy". The first one was
characterized by a smooth GF variation scaled to the load
variation. In the second one, important and sudden increases of
GF (steps) were present and formed a discontinuous staircase
increase of GF. In the TF condition, the average number of steps
during one trial was 0.63 while in the TT condition it corresponded
to 1.74. This significant difference (p-value = 1.7x10-¹¹) means
that the "smooth strategy" and the "stepwise strategy" were
preferentially used in the TF condition and the TT condition
respectively.
The mean reaction times were respectively 7% and 33% of the
total duration of the force increase in the TF condition and the TT
condition (p-value = 2.7x10⁻³³). It means that the subjects were
able to detect the force variation during the translations faster
than during the rotations. This slower detection in the TT
condition might arise because of the type of event triggering the
grip force update and might explain the difference of strategy
used in each condition.
This study highlights the main grip force control strategies used
during slow load variations : a "smooth strategy" and a "stepwise
strategy". Moreover, it confirms the presence of a difference in
grip force control when a tangential torque is present. Indeed, in
presence of a tangential torque, both the strategy used and the
reaction time, were significantly different from the condition
without torque.
1-F-48 Tactile detection of slip: Fine characterization of
skin deformation during the onset of slip
Benoit Delhaye1, Philippe Lefèvre2, Jean-Louis Thonnard1
1
Institute of Neuroscience - Université catholique de Louvain,
Institute of Information and Communication Technologies,
Electronics and Applied Mathematics
2
In daily life, we manipulate objects that have different physical
properties (mass, surface, shape). Most of the time, we succeed
in this task and prevent the objects dropping from the hand. This
is explained by the fine coordination between the load force (LF),
the tangential force due to the object weight and inertia, and the
grip force (GF), the normal force applied to the objects to
counteract the LF. This fine coordination is possible with the help
of the cutaneous feedback from the fingertips mechanoreceptors.
These respond to the stresses and strains created by the
deformation of the skin during object grasping and manipulation.
Measuring the fingertip skin deformation before and during
slipping may thus contribute to better understand how we can
prevent the objects from slipping during their manipulation.
In this study, we used an imaging system to record the fingertip
deformation during tangential sliding movements of the fingertips
on a glass. The constraints were applied passively to the fingertip
of subjects (n=4) in six different directions : front-rear translations,
left-right translations and clockwise-counterclockwise rotations. A
robotic system controlled the normal force applied to the fingertip
(three normal forces were tested : 0.5, 1 and 2 N) and the
tangential speed (three speeds were tested: 5, 10 and 20 mm/s in
translation and 20, 40 and 80 °/s in rotation). The contact forces
and torques were recorded. Each condition was repeated three
times. Using computer vision techniques, the contact area was
extracted from the images and the relative velocity field between
the glass and the fingertip contact area was computed during the
whole sliding phase. An ellipse was fitted to the contact area
during the whole slip in order to measure the displacement (for
translations) and the tilt angle (for rotations) of the contact area.
34
Results showed that the skin deforms in a complex way
from a full stuck state to a full slipping contact. During this
transition, while the tangential force was monotonically
increasing, the skin progressively slipped on the plate,
beginning from the border of the contact area and ending at
the central zone, and forming a velocity gradient from the
border to the center. This behavior was observed for both
translational and rotational movements. This transition from
a full stuck state to a full slip was accompanied by a
reduction of the contact area, which then stayed constant
during full slip. The ratio between reduced area (during full
slip) and full area (before the onset of the slip) is 0.70 ± 0.14
(mean ± std) for front-rear translations, 0.65 ± 0.15 for leftright translations and 0.90 ± 0.12 for rotations. The skin
displacement, due to the stiffness of the fingertip, was
measured by computing the displacement of the center of
contact during the transition phase. This displacement was
significantly larger in lateral movements (3.65 ± 1.49 mm) in
comparison with front-rear movements (2.58 ± 1.01 mm).
The direction of rotation did not affect the absolute value of
the tilt angle significantly.
1-F-49 Observing lifting errors modulates corticospinal excitability
Gavin Buckingham1, Jeremy D Wong1, Paul L Gribble1,
Melvyn A Goodale1
1
The University of Western Ontario
Using Transcranial Magnetic Stimulation (TMS) to induce
Motor Evoked Potentials (MEPs) during passive
observation, it has been demonstrated that merely watching
an action modulates cortico-spinal excitability. This action
observation system is often interpreted as a low-level
mechanism to prepare the observer to perform the actions
they are passively viewing. Recently, it has been
demonstrated that the sensorimotor system encodes the
fingertip force requirements of an observed lift - watching
lifts of a heavy object elicits a larger MEP than observing
lifts of a light object (Alaerts, Swinnen & Wenderoth, 2010,
European Journal of Neuroscience). However, the feedforward way in which humans parameterize their fingertip
forces means that errors are often made when lifting
objects. Almost no research has examined the way in which
the sensorimotor system responds to the passive
observation of common overestimation- and
underestimation-style lifting errors. To examine corticospinal modulation in response to the observation of errors,
we applied single-pulse TMS to the left primary motor cortex
of individuals while they passively watched others lift up
large and small cubes of equal mass. In half of these
videos, the actors made the errors that lifters normally make
when confronted with these stimuli (i.e., overestimating the
weight of the large cube and underestimating the weight of
the small cube), whereas in the other half of the videos, the
actors lifted the cubes with the appropriate (and identical)
forces. When observing error-free performance, individuals'
cortio-spinal excitability was tailored to the size of the cubes
being lifted: watching lifts of the large cube elicited
significantly larger MEPs than watching lifts of the
(identically-weighted) small cube (p<.001). Thus, when the
lifting kinematics were equivalent, cortico-spinal excitability
was driven by object size. This size-related modulation
appeared to be abolished when individuals observed the
underestimation- and overestimation-style lifts of these
same cubes: observing underestimation-style lifts of the
small cube elicited an MEP of a similar magnitude to that
seen when observing overestimation-style lifts of the large
cube (p=.58). The elimination of this size-related modulation
Poster Sessions
Full Abstracts
may be due to (1) the kinematics of the error lifts giving the visual
impression that the cubes have different weights (i.e., that the
large object appeared to be lighter than the small object), and/or
(2) an automatic sensorimotor response to engage more
appropriate grip and lift forces in the observer (i.e., preparing the
observer to pre-empt the errors they have just observed). These
suggestions are discussed in relation to the improvements seen
in actual lifting behaviour after observing object lifting errors
(Buckingham et al., 21st annual meeting of the Society for the
Neural Control of Movement, 2011).
1-F-50 Asymmetrical Modulation of corticospinal
excitability as a function of emotional sounds along with
ear laterality
Naeem Komeilipoor1, Fabio Pizzolato1, Andreas Daffertshofer2,
Paola Cesari1
1
University of Verona, 2VU University Amsterdam
Emotional sounds are essential to our survival as they give hue to
the mental pictures of events, thus enable us to decipher the
surroundings and form effective behavioral responses.
Transcrainial Magnetic Stimulation (TMS) studies have evidenced
the link between action readiness and emotional processing,
reporting an increased activity in the motor cortex during
emotional experiences. However, the majority of these studies
employed preferentially visual and neglected audio stimuli.
Moreover, data were usually recorded from just one hemisphere
at the time. Hence, in the present study, we employed TMS to
clarify whether non-verbal emotionally-characterized sounds
processed separately by the left and right ear, would modulate
motor evoked potential (MEP), recorded from hand muscle in an
asymmetrical fashion. Overall, we found that the motor cortex
significantly increased its activity in response to unpleasant as
compared to pleasant and neutral sounds. Interestingly, the motor
system excitability was asymmetrically modulated as a function of
the stimulus valence; unpleasant stimuli resulted in a significant
higher facilitation of motor potentials evoked in the left
hemisphere, while pleasant stimuli yielded a higher activation in
right one. Furthermore, TMS induced higher MEPs when listening
to unpleasant sounds with the Left ear as compared to the Right
ear, which complements previous findings indicating the left ear
as selectively reactive to emotional contents. The increment of
the left ipsilateral motor system excitability for the unpleasant
sounds was presided over the dominant hand, suggests the
presence of a biological preference for a direct motor-auditory
pathway depicted to the processing of threatening auditory
stimuli. This system might have developed to allow for faster
movements revealing a more efficient way for a "fight or flight"
reaction. Taken together, our findings contribute to the
evolutionary perspective on the role of emotions on action
readiness.
1-F-51 A neural basis for hand muscle synergy in
primate spinal cord
Kazuhiko Seki1, Tomohiko Takei1
1
National Institute of Neuroscience
Grasping is a highly complex movement, which requires the
control of 27 hand and arm muscles. Electromyographic (EMG)
studies have shown that activities of the hand muscles during
grasping can be explained with a combination of few basic
components ("muscle synergy") that could be used by the CNS
for the grasping control. However, the neural mechanism that is
responsible for forming the muscle synergy is still unknown. Here
we explored how the spinal premotor interneurons (PreM-INs) are
involved in the forming of the muscle synergy. Twenty-three
PreM-INs were recorded from lower cervical segments (C6-T1) in
two macaque monkeys performing a precision grip task. EMG
activities were simultaneously recorded from 12 finger and
wrist muscles, and muscle synergies were extracted from
them by using non-negative matrix factorization that
decompose EMGs into predefined number of synergy
weights and their temporal activities [1]. Coefficient of
determinations indicated that a linear combination of three
synergies account for the > 85% of variance of original
EMGs. The distribution of the weights of each synergy was
consistent within each monkey (R2=0.89±0.17). The postspike effects of PreM-INs on muscles were quantified by
spike-triggered averaging (STA) of rectified EMGs. PreMINs showed divergent effects in more than one muscles
(2.1±1.5 muscles), forming the muscle fields [2].
Comparison between the spatial distribution of synergy
weight and muscle field of PreM-INs indicated that the
muscle field of PreM-IN was preferentially matched with one
of the extracted synergies (preferred synergy) and a
similarity between neuron's muscle field and weight of
preferred synergy was significantly higher than those
expected from chance level. Moreover, the temporal activity
of PreM-INs had a significantly higher correlation with the
preferred synergy than that with the non-preferred
synergies. These results suggest that the muscle synergy in
primate grasping can be formed by the system that involves
the spinal PreM-INs. [1] Cheung, V.C.K., et al., Stability of
muscle synergies for voluntary actions after cortical stroke in
humans. Proc Natl Acad Sci USA, 2009. 106(46): p. 195638. [2] Takei, T. and K. Seki, Spinal Interneurons Facilitate
Coactivation of Hand Muscles during a Precision Grip Task
in Monkeys. J Neurosci, 2010. 30(50): p. 17041-50.
1-F-52 A neural correlate of arousal in the
subthalamic nucleus facilitates force production
Anam Anzak1, Alek Pogosyan1, Huiling Tan1, Thomas
Foltynie2, Patricia Limousin2, Ludvic Zrinzo2, Marwan Hariz2,
Keyoumars Ashkan3, Wesley Thevathasan1, Marko
Bogdanovic1, Alexander Green1, Tipu Aziz1, Peter Brown1
1
University of Oxford, 2Institute of Neurology, 3Kings College
Hospital
Loud auditory stimuli have been shown to increase force
and rate of development of force in maximal hand grips in
both healthy subjects and patients with Parkinson's disease
(PD) (Anzak et al, 2011. Eur J Neurosci; 34(1) p.124-32).
We reasoned that phasic arousal could be the underlying
mechanism driving this phenomenon. To this end,
electrophysiological evidence of a neural correlate of
arousal, present in motor cortico-subcortical circuitry, and
scaling with force improvements, was sought.
Simultaneous local field potentials (LFPs) were recorded
from surface EEG and bilateral deep brain stimulation (DBS)
electrodes implanted in the subthalamic nuclei (STN) of 7
patients with PD, whilst off and on dopaminergic medication.
Patients gripped a force dynamometer as quickly and
strongly as possible in response to a visual cue, which was
accompanied by an auditory tone at one of five different
sound levels. Levels were selected at random and ranged
from very quiet to very loud.
An evoked potential, which increased in amplitude in
response to louder auditory cues, was found to be focally
generated in the STN. Its latency was very short (peak
between 50-100ms from cue onset). Force and rate of
development of force (averaged over 0-100ms from cue
onset) also increased with louder cues. The amplitude of all
three variables was similar whether patients were off or on
dopaminergic medication. The peak amplitude of the evoked
potential correlated with the average initial force
35
Poster Sessions
Full Abstracts
(spearman's ρ=0.359, P=0.002) and rate of development of force
(ρ=0.307, P=0.010) across sound levels.
1-F-54 Artificially evoking physiological finger
tremor suggests mechanical origin
The evoked potential recorded in the STN was similar in
morphology and latency to the P50 - an evoked potential normally
recorded in surface EEG, and believed to be a correlate of
ascending arousal originating from the cholinergic reticular
activating system (Reese et al, 1995. Prog Neurobiol; 47(2)
p.105-33). The independence of our STN P50 from L-DOPA state
is in line with such an origin. This non-dopaminergic system may
lend itself for therapeutic manipulation in PD, either
pharmacologically or through mimicry of the evoked potential with
a more targeted and specific form of DBS.
Carlijn Vernooij1, Raymond F Reynolds1, Martin D Lakie1
1-F-53 Dynamic control of lower extremities declines
with aging
Emily Lawrence1, Veronica Stern1, Mark Lyle1, Carolee Winstein1,
Philip Requejo2, Francisco J Valero-Cuevas1
1
University of Southern California, 2Rancho Los Amigos National
Rehabilitation Center
The ability to control ground reaction forces in the lower
extremities is necessary for balance, locomotion and fall
prevention. We examined the question of why older adults fall
more than younger populations by assessing dynamic
coordination of the lower extremities independently of muscle
strength. We used the Lower Extremity Coordination (LEC) test-an extension of the Strength--‐Dexterity (S--‐D) test known to
quantify dynamic finger control-- to measure dynamic control of
the lower extremities. [1,2] The test consists of compressing a
slender spring prone to buckling using the foot with the goal to
sustain the highest force possible, where compression makes the
spring increasingly more unstable. Compression force was
measured a 1--‐axis load cell (Transducer Techniques, Temecula,
CA) and data were sampled at 2000Hz, processed, and displayed
as real--‐time feedback to the subject with custom MATLAB (The
Mathworks, Natick, MA) software. To minimize the effects of leg
strength and mitigate fatigue, we chose spring parameters (i.e.
stiffness and slenderness) such that spring instability occurred at
low forces. The largest sustained compression force is
representative of the maximal sensorimotor ability to stabilize the
leg in contact with the unstable surface at submaximal force
levels. Nine young adults (25--‐35) and 15 older adults (55--‐95),
all physically fit and active in fitness regimens or recreational
sports, took part in this multi--‐site study. The average maximal
compression forces for the best 3 out of 25 trials from each
subject were lower for the older adults (111.32±10.3N vs.
124.93±9.3N; p<0.03). The maximal compression force recorded
for the older adult population (average weight = 750N) [3] was
approximately 15% of their body weight compared to 20% for the
young adult population (average weight = 620N). This shows that
the known propensity for falls in the older adults is also
associated with reduced sensorimotor ability to stabilize the leg
during dynamic tasks--and mirrors recent findings [1] that young
women, who are at higher risk of non--‐contact ACL tears, have
lower scores than age--‐matched young men. This novel
detection of lower sensorimotor processing capabilities for leg
control at low forces in older adults provides new insight into the
potential neuromechanical contributors to falls, and opens new
avenues to investigate their underlying mechanisms and clinical
countermeasures. References: 1. Mark A. Lyle, Francisco J.
Valero--‐Cuevas, Robert J. Gregor, Christopher M. Powers, The
lower extremity dexterity test as a measure of leg dynamical
capability, J Neurophys, In Review. 2. Valero-Cuevas FJ, Smaby
N, Venkadesan M, Peterson M and Wright T., The strengthdexterity test as a measure of dynamic pinch performance. J
Biomech. 2003 36(2): 265-270. 3. Health Calculator and Charts,
2011 Steven B. Halls Professional Corporation, August 2011
<www.halls.md> 
36
1
University of Birmingham
Introduction
People cannot move their fingers smoothly or even hold
them perfectly stationary; some degree of jerkiness is
always present. Researchers studying what causes this
physiological tremor can be roughly divided into two groups:
those arguing for neural or spinal oscillations and those
arguing for a mechanical resonance. This last view is
strengthened by a recent paper demonstrating a decrease
in the frequency of hand tremor post-movement
unaccompanied by corresponding changes in EMG. We
studied the mechanical resonance of the finger and
associated muscles of healthy subjects by artificially evoking
various sizes of tremor using white noise torques or
electrical muscular stimulations as an input. Using this novel
way of studying tremor, we could examine its amplitude and
frequency over different movement sizes while bypassing
the nervous system.
Method
Acceleration of the relaxed middle finger of eight healthy
subjects was measured while we artificially evoked tremor
by electrically stimulating the extensor digitorum communis
muscle or by applying torques directly to the finger with a
torque motor. We used six different intensities of white noise
signals for both stimuli and torques to study tremor over a
wide range of movement sizes. Intensities were individually
determined based on acceleration response, which ranged
from hardly visible to a response of ~7.5 m/s2. In addition,
we included a condition in which subjects were asked to
actively hold their finger in a middle position or track a very
low frequency sine wave. When not stimulating the muscle,
we also measured EMG of the extensor muscle. For each
trial, we calculated the FFT of the white noise input, the
measured acceleration and, where applicable, the rectified
EMG. In addition, the input-to-acceleration gain and the
EMG-to-acceleration gain were calculated.
Results
The white noise input signals all showed an evenly
distributed frequency spectrum, which implies similar
spectra for acceleration and gain. The frequency spectra of
the input-to-acceleration gains resulting from both input
modalities showed a peak at ~20 Hz for low intensities.
Higher electrical stimulation intensities decreased this peak
to ~8 Hz while higher torque intensities led to a broad
spectrum between 8 and 20 Hz with a minor peak at 8 Hz.
The EMG signals measured while applying torques had a
flat frequency spectrum. Actively holding the finger in a
middle position resulted in an EMG-to-acceleration gain
peaked at ~20 Hz while the tracking task resulted in a peak
at ~8 Hz. EMG spectra in these conditions were fairly flat
with a slight peak around 12 Hz.
Conclusion
The presence of an acceleration frequency peak with a
white noise input suggests that it is not caused by central
oscillations. It seems likely that a resonance is involved. The
fact that there are two distinct frequencies suggests that
there may be two modes of resonance. We speculate that
the high frequency mode involves the finger and tendon and
the low frequency mode involves the muscle. We have
shown that during movement wrist tremor frequency
Poster Sessions
Full Abstracts
decreases. Finger tremor shows a reduction that is even bigger
but more complicated. Controlling the finger in the transition from
static to movement must therefore include big computational
problems.
1-F-55 Stochastic resonance improves sensorimotor
performance of the finger
Rumyana Kristeva1, Ignacio Mendez-Balbuena1, Elias Manjarrez1,
Juergen Schulte-Moenting2, Frank Huethe1, Joshua A Tapia3,
Marie-Claude Hepp-Reymond4
1
Univ. Freiburg Dept. Neurology, 2Univ. Freiburg Inst. Biomed.
Biometry, 3BUAP Instituto de Fisiologia, 4Inst. Neuroinf., Univ.
Zürich and ETH Zürich
Several studies about noise-enhanced balance control in humans
support the hypothesis that stochastic resonance can enhance
the detection and transmission in sensorimotor system during a
motor task. The purpose of the present study was to extend these
findings in a simpler and controlled task. We explored whether a
particular level of a mechanical noise (0-15 Hz) applied on the
index finger can improve the performance during compensation
for a static force generated by a manipulandum. The finger
position was displayed on a monitor as a small white point in the
center of a green circle. We considered a good performance
when the subjects exhibited a low deviation from the center of this
circle and when the performance had less variation over time.
Several levels of mechanical noise were applied on the
manipulandum. By means of the mean variation in the position we
observed an inverted U-like graph of the performance versus the
input noise level in all subjects (8/8). We compared the
performance between zero noise (ZN), optimal noise (ON) and
high noise (HN). The mean variation was significantly lower
during ON than during ZN or HN. The findings suggest that the
application of a tactile-proprioceptive noise can improve the
stability in sensorimotor performance via stochastic resonance.
Possible explanations of this motor improvement can be a
suppression of the physiological tremor and/or increased
sensitivity of the muscle spindle afferents.
1-F-56
model
Movement automatic guidance by direct internal
Valerie Gaveau1, Olivier Sillan1, Calude Prablanc1
1
INSERM
Movements planed in peripheral vision are inaccurate, but
processes of correction occur during motor execution and guide
the movement towards its goal (Goodale et al, 1986, Pelisson et
al 1986, Prablanc & Martin 1992). This guidance is largely nonconscious, automatic and pre-emptive to voluntary control (Pisella
et al 2000). In this study we want to test the hypothesis that this
automatic guidance of the movement depends on a comparison
between a predictive sensory feedback signal and an updated
signal of the target position, rather than a comparison between
the simultaneous visual information of the target and the hand.
This prediction is the output signal of a "direct" internal model
designed before movement execution and supplied both by initial
information on the visual and kinesthetic position of the hand and
by proprioceptivo-motor commands. Thus, under normal
conditions, the predicted hand position and the target visual
feedback signals integrated during the execution of the movement
are consistent. In order to test the hypothesis of a closed
relationship between automatic guidance of the hand and the
output of the internal model of the same effector, we conducted a
first experiment in which the internal model was biased prior to
movement (by giving a visual distortion signal about the hand
position); this misestimated position would lead to a distorted
internal model representation of the hand. Then, at movement
onset, normal vision of the hand was restored and we assessed
whether the system is able or not to produce appropriate
automatic corrections. Two hypotheses were put forward: 1)
if the automatic guidance of the movement is based on
internal model output, then guidance will be skewed
because the predictor was flawed; 2) if automatic guidance
and internal model are independent (e.g. guidance depend
on comparison of the simultaneous visual information of
target and hand), then guidance will be appropriate. To
further test our predictions, we conducted a second
experience in which internal model was not biased but we
introduced a visual feedback discordance (the target will be
displaced at movement onset). No comparison of the
simultaneous visual information of target and hand was
possible. Two alternative predictions follow: 1) if the error is
automatically corrected in the absence of hand visual
feedback and without lengthening movement duration,
automatic correction must be attributed to the direct action
of the internal model; 2) if the error is not corrected, the
error processing requires a visual comparison between
target and hand position. Our results demonstrate a tight
link between automatic guidance of the hand and its internal
representation: the automatic guidance is strongly
dependent upon the integrity of the internal model of hand
position.
G - Theoretical & Computational Motor Control
1-G-76 Methods for the study of modularity in
muscle activities: A unifying model for extracting
spatial and temporal modules from EMG signals
Bastien Berret1, Ioannis Delis1, Stefano Panzeri1, Thierry
Pozzo1
1
Istituto Italiano di Tecnologia
Motor control has long been hypothesized to be facilitated
by the presence of modularity in the motor system, i.e., the
brain may presumably generate a wide repertoire of
movements by combining a small number of pre-existing
modules. The nature of these modules has been examined
from different viewpoints. In recent years, several studies
have focused on modularity in muscle space, aiming to
identify invariant patterns of muscle activity across a variety
of motor behaviors. To do so, a number of efficient
algorithms that perform dimensionality reduction on
electromyography (EMG) signals have been developed.
However, different models make different assumptions. The
previously used models differ substantially in a) the invariant
quantities that are assumed to be encoded in the central
nervous system (CNS), b) the degree of compression of
EMG data and c) their ability to capture motor variability
across tasks and/or repetitions of the same task (i.e. what
we call an episode). In the present study, we propose a new
model giving a low-dimensional yet flexible and relevant
representation of motor patterns. Our model is moreover
compatible with most of the existing ones. We assume the
existence of both spatial and temporal modules as the
building blocks of motor control, in agreement with a huge
amount of literature. Hence in our model any single muscle
pattern can be expressed as a double linear combination of
such modules. Low-dimensionality results from the fact that
only a finite number of scalar parameters (combining N
spatial modules with P temporal modules) are assumed to
be controlled by high-level centers in order to generate an
adequate motor pattern and achieve a given task at hand.
Flexibility results from the double summation that allows
37
Poster Sessions
Full Abstracts
mixing temporal modules with any of the spatial modules and
vice-versa. Relevance is because a number of neurophysiological
studies have reported evidence for each type of module
separately and that a large repertoire of tasks can indeed be
performed reliably on a single-trial basis. The unifying aspect of
the model is due to the fact that the spatial modules correspond
to the so-called synchronous or time-invariant synergies and the
temporal modules to the so-called muscle activation patterns,
premotor drives or motor primitives. Furthermore, it can be shown
that time-varying synergies, i.e. spatiotemporal muscle patterns,
can actually be built from the spatial and temporal modules we
assume, suggesting that they may constitute more primal building
blocks of the CNS. We thus developed a specific algorithm based
on an episode-based tri-factor non-negative matrix factorization
method. It extracts simultaneously spatial and temporal modules
from rectified EMG signals. The proof of convergence follows
simply from existing theorems related to non-negative matrix
factorization. The method has been successful when applied to
synthetic data, proving its effectiveness. When applied to real
EMG data recorded during goal-directed reaching movements, its
performance was compared to other classical models. Results
show that our model provides a good trade-off between lowdimensionality and biological plausibility in terms of single-trial
task decoding scores.
1-G-77 Methods for the study of modularity in muscle
activities: A single-trial task-decoding metric to evaluate
muscle synergy models
Ioannis Delis1, Bastien Berret1, Thierry Pozzo1, Stefano Panzeri1
1
Istituto Italiano di Tecnologia
Muscle synergies, i.e., invariant coordinated activations of groups
of muscles, have been proposed as the building blocks with which
the central nervous system (CNS) constructs the patterns of
muscle activity utilized for executing movements. In particular, the
muscle synergy hypothesis states that the CNS performs a motor
task by recruiting a certain number of synergies and combining
them in the appropriate way. Several efficient algorithms that
extract synergies via dimensionality reduction techniques have
been developed in recent years. Yet, little is known about the
extent to which the combination of those synergies can describe
each individual task in a set given the variability in the data. Here
we conceive and develop a novel computational framework to
address this question. The procedure, which is based on singletrial task decoding from electromyography (EMG) recordings,
quantifies how well the tasks can be identified based on a small
given number of muscle synergies and determines the minimal
set of synergies that describes all task-related variability of the
activity of multiple muscles. We validate this method on plausibly
simulated datasets, and we illustrate its merits on EMG data
recorded during goal-directed reaching movements. We show
that, unlike previous analysis methods, our algorithm succeeds in
determining correctly and robustly the set of synergies needed to
capture all task related variations of EMGs. We finally show how
this method can be useful for comparing and/or validating
different classes of muscle synergy models.
1-G-78 Safety margins and variability in a redundant
object manipulation task
Christopher Hasson1, Dagmar Sternad1
1
Northeastern University
The human body is fundamentally redundant, with many more
neurons, muscles, and joints than are necessary to perform most
actions. Redundancy also exists at the task-level when there are
multiple ways to reach a spatial goal in a given time, such as
when placing a cup of coffee on a coaster. Although many
movement strategies may lead to the same goal, some
38
accommodate task execution variability better than others
through larger safety margins. How safety margins are
shaped by redundancy and execution variability is
addressed in a task that modeled transporting a cup of
coffee. Eighteen subjects used a manipulandum to transport
a virtual cup containing a ball ("coffee") to a target without
losing the ball. Nine subjects were asked to complete the
cup transit in a comfortable target time of two seconds (a
redundant task by instruction, with infinite solutions), and
nine were asked to transport the cup in minimum time (a
non-redundant task with one explicit cost). Three
hypotheses were tested: H1) in the minimum-time task
subjects converge to a single optimal strategy, while in the
target-time task they choose different strategies. H2) in the
minimum-time task subjects decrease safety margins to
optimize movement time, but in the target-time task they
increase safety margins. H3) in both tasks subjects
modulate safety margins according to their execution
variability. The safety margin was defined as the ball energy
relative to the ball's escape energy. Execution variability
was quantified by the inter-trial standard deviation of the
total energy of ball and cup. Results showed that both
groups developed individualized strategies with practice
(counter H1). The minimum-time group decreased their
safety margins, while the target-time group increased safety
margins (H2). In the target-time group changes in safety
margins were correlated with changes in execution
variability: smaller variability decreases over practice had
larger safety margin increases (H3). In contrast, in the
minimum-time group such a relation was observed only at
the end of practice, not across practice. These results show
that when learning a redundant object manipulation task
subjects increase their safety margins and shape their
movements in accordance with their changing variability.
1-G-79 Can we learn what the brain optimizes?
Alexander Terekhov1, Vincent Hayward2
1
Université Pierre et Marie Curie (Paris 6), 2Institut des
Systèmes Intelligents et de Robotique, UPMC – CNRS
If we ask someone to reach her or his nose with the index
finger, we do not expect her or him to wrap the arm beneath
the knee, although this option is clearly feasible. Somehow,
we find certain movements to be more appropriate than
others. When there is something we want or need to do, we
prefer to do it in a specific way. This observation suggests
that the brain associates a cost with every movement and
tends to minimize this cost during motor behavior. The
question of what is being minimized by the brain has been
examined for the last 50 years after having being first raised
by Nubar and Contini (1961), who suggested a candidate
function made of a weighted sum of squared moments.
Since then, numerous candidate cost functions were
proposed for the whole spectrum of motor activities and at
different levels of analysis, ranging from the activation of the
individual motor units within a single muscle to the whole
body trajectory control during walking. The question at hand
is typically addressed in terms of static optimization and/or
optimal control problems in deterministic as well as in
stochastic settings. It is somewhat disheartening that almost
identical motor behaviors can result from different cost
functions, each able to fit experimental data equally well.
With this in mind, we ask the question of whether it is
possible, at least hypothetically, to say for sure what the
brain optimizes. Imagine for the sake of argument that we
could measure the activation of every single
mechanoreceptor and the stretch of every single muscle
fiber in as many motor tasks as needed. Would this
Poster Sessions
Full Abstracts
knowledge allow us to determine the cost function(s) used by the
brain? If the answer is negative, then such result would
immediately cast serious doubts on any attempt to find what the
brain optimizes. On the other hand, a positive answer would
encourage researchers to investigate the question in simpler
tasks. The problem of finding a cost function from experimental
data is called "inverse optimization". The central question
addressed in the proposed presentation will be to discuss
whether an inverse optimization problem can have a unique
solution. We will formulate a theorem of uniqueness for a class of
inverse optimization problem relevant to human motor control.
This theorem provides conditions on experimental data whose
satisfaction guarantees that the cost function can be determined
unambiguously. The limits of applicability of the theorem will be
also discussed. To illustrate the discussion we will consider the
inverse optimization approach to describe behavior during
prehension and multiple finger pressing tasks. We will provide
examples of how the violation of the conditions of the theorem
induces non-uniqueness in solving the problem. We will also
report the recent results obtained for grasping. In particular, we
will show how the cost function determined from the experimental
data for the four-finger grasp can explain the experimental data
for the three-finger grasp. It will be concluded that yes,
theoretically, cost functions optimized by the brain can be found.
The question of their meaning, however, must be left to the
researcher.
1-G-80 Implementation of stochastic feedback control
and Bayesian nonlinear filtering in spiking neuron
populations
Atiyeh Ghoreyshi1, Terence Sanger1
1
University of Southern California
States of things in nature, such as the position or velocity of an
object, or the angle of one's elbow, are usually continuous in time
and in space. Their neural representations in the brain, however,
are not continuous either in time or in space. External variables
are coded in the brain by spikes in populations of neurons. Thus,
there is a limitation in both spatial and temporal resolutions of
neural representations. Besides, we often encounter "unfamiliar"
or "uncertain" situations, in which we have to observe, decide,
and navigate. What makes our brain so highly efficient and
accurate in communicating with and navigating through various
environments, and yet, most elaborate robots with ample
resources fail when put in a fairly noisy environment? This makes
one wonder if our traditional structure of thinking about estimation
and control as such is too "rigid" in handling uncertainty or
stochasticity, which is an intrinsic property of neural behavior and
response. A new framework for dealing with such problems has
been developing recently(1). In this framework, states of a system
are considered random variables represented by their probability
density functions (pdfs) rather than as explicit deterministic
variables, and all estimation and control operations are performed
on state pdfs. The Fokker-Planck equation is used for the forward
propagation of pdfs in time, the Zakai/Kushner's type equations
for incorporating observations into the pdf temporal propagation,
and proper control parameters are calculated given specific
control operators and cost functions. This framework is powerful
and efficient in handling nonlinear behavior, uncertainty, and
ambiguity in the system or environment(1). In this work, we
demonstrate how computations on pdfs can be implemented in
populations of spiking neurons and what can be achieved doing
so. Our neural network includes a receptor layer that provides
state observations; a sensory cortex layer that, through nonlinear
Bayesian filtering, computes internal state estimates; and a motor
cortex layer that calculates motoneuron commands via stochastic
feedback control calculations involving the cost function, current
internal state estimates, and control operators. For the first time,
we have combined a spike-based stochastic estimation
algorithm with a spike-based controller. We show that this
combination provides efficient and smooth control in the
presence of state and parameter uncertainty, and can
provide on-line optimization of trajectories with respect to a
known cost function. The real-time calculation of control
variables and desired trajectories greatly reduces the
required computational resources. We demonstrate realtime tracking simulation results, where the actuator
accurately follows a time-varying trajectory based on a given
cost/utility function. We also demonstrate how our
implementation predicts behavioral motor control
experimental results, such as Trommershauser's
experiments(2). These examples suggest the plausibility of
this approach both in theory and in practice, in terms of
understanding and utilizing the underlying mechanisms of
human motor control. (1) T. D. Sanger (2011). "Distributed
Control of Uncertain Systems Using Superpositions of
Linear Operators". Neural Computation, 23 (8), 1911-1934.
(2) J. Trommershäuser, L. T. Maloney, & M. S. Landy
(2003). "Statistical decision theory and trade-offs in the
control of motor response". Spatial vision, 16(3-4), 255-275.
1-G-81 Learning rate modulation vs. dimensionality
reduction as a mechanism for structural learning
Alkis Hadjiosif1, Maurice A Smith1
1
Harvard University
The rate in which we learn to adapt to new environments
can be greatly affected by our previous experiences with
similar environments. Studying point-to-point reaching
movements subjected to visuomotor (VM) transformations,
Braun et al. (2009) found that exposing subjects to a series
of randomly varying visuomotor rotations could facilitate
learning subsequent visuomotor rotations but not
visuomotor shears. The authors described this as an
example of structural learning: repeated exposure to
transformations of the same class - there, rotations of
varying angles - reveals the underlying structure of the
class, specifically facilitating adaptation to that class. The
idea put forward was that transformations within a single
class represent a low dimensional manifold within the high
dimensional space of all possible transformations that could
be encountered, and that learning the structure of the class
(i.e. the location of the low dimensional manifold) could
make the process of within-class motor adaptation
considerably more efficient by reducing the dimensionality of
the learning problem. An alternative to this idea is that the
rate of learning is modulated - facilitated or inhibited - by the
statistical features of the experienced environment. Here we
tested two differential predictions of these competing
hypotheses: First, the statistics of the environment would
predict that negatively correlated versus positively
correlated switching between different VM rotation angles
should produce different changes in the rate of learning for
VMs. In contrast, dimensionality reduction would predict that
both should facilitate VM rotation learning because the
environment would be limited to a single class of
transformations in both cases. We thus compared
adaptation rates before and after exposure to a VM rotation
structure in two conditions: one characterized by a strong
correlation between the imposed transformation from one
trial to the next (PC condition) and another one in which the
transformation in one trial was negatively correlated to the
next (NC condition). The PC condition showed a substantial
upregulation of rotation learning rates (mean±st. error
across subjects: 0.129±0.014 vs. 0.314±0.066, p<0.015),
whereas the NC condition showed no difference in learning
39
Poster Sessions
Full Abstracts
rates for rotation before and after the inconsistent structure
(0.080±0.014 vs. 0.084 ±0.011, p>0.7). This suggests that the
statistical properties of the exposure to a structure rather than the
mere presence of exposure to the structure determines whether
learning rate facilitation can occur. If so, a prediction would be
that the dimensionality reduction associated with exposure to a
structure should not play key role in learning rate facilitation. We
then looked at the ability of an environmental exposure to
facilitate learning rate increases for transformations outside of the
experienced structure for the PC condition. We found that when
transformations consisting of various combinations of gain
changes and rotations were studied, transformations away from
the learned structure containing non-zero gain changes showed
similar facilitation of learning rates compared to pure rotation
transformation (p>0.42 between all cases). This shows that
learning rate facilitation was not limited to the learned structure.
Together these findings suggest that environmental statistics
rather than dimensionality reduction accounts for the structural
learning of visuomotor transformations.
1-G-82 Augmentation of perceived visual error improves
control and enhances retention of a discrete task
Dagmar Sternad1, Meghan E Huber1, Anastasia Kyvelidou1
1
Northeastern University
Previous studies show that the control of movement, in particular
its accuracy and variability, benefit from learning under erroraugmented conditions. The present study examined how the
enhancement of perceived visual error can increase accuracy and
decrease variability in motor performance. Importantly, we show
how this improved performance persists for five days.
Using a virtual throwing task, skittles, subjects manipulated a
lever arm with a single-joint elbow movement and released a
virtual ball that traversed a concentric force field to hit a target.
The ball's trajectory and hitting success were fully determined by
the release angle and velocity at ball release. The task is
redundant as different combinations of release angle and velocity
lead to identical results and the set of successful solutions
describes a nonlinear manifold. Previous work showed that
subjects' variability aligns with the direction of the solution
manifold and remains non-zero even in skilled performers that
reached a plateau in result accuracy and variability. We tested
whether 1) skilled subjects could further improve their strategy
when the perceived visual error is augmented and whether 2) this
improvement persists when augmentation is removed. To
manipulate perception of visual error the threshold of a success
signal about the target center was decreased.
Twelve subjects trained under normal task conditions until their
skill performance reached a plateau (session 1, 3 days); subjects
trained with the augmented visual error perception (session 2, 3
days); augmentation was removed (session 3, 5 days). Two
hypotheses were tested: 1) Skilled subjects decrease their
variability and increase their accuracy in session 2 as compared
to session 1 in order to compensate for the augmentation. 2)
Subjects return to their initial more variable and less accurate
strategy in session 3 when task constraints were relaxed by
removing the augmentation.
Results showed that subjects improved their accuracy and
reduced their internal variability to compensate for the tighter task
constraints. In contrast to hypothesis 2, however, subjects
maintained their low variability and high accuracy for 5 days when
the success threshold was reestablished to its initial value. These
results suggest that subjects are sensitive to their variability and
reduce it in the presence of tighter task demands. Importantly, the
effects of increased accuracy and reduced variability persist even
after removal of this enhanced visual error perception. These
results have implications for training and rehabilitation as signal
40
detection and control processes in the sensorimotor system
may be enhanced with augmented visual error perception.
1-G-83 Interaction of sensory uncertainty and
motor variability during reaching: A simulation study
Gregory Apker1, Christopher Buneo1
1
Arizona State University
Reaching movements are inherently variable, a result of
noise in sensorimotor processing associated with the
planning and execution of limb movements. In this context,
planning related noise represents variability in the encoding
of reaching parameters associated with specifying a motor
plan, dependent in part on the anisotropic nature of sensory
uncertainty. Execution noise describes variability in the
processes associated with generating the appropriate motor
output, and manifests largely as reaching variability along
the movement vector. During natural reaching, these noise
processes interact throughout movement to influence
patterns of variability. Growing behavioral evidence
suggests that the brain coordinates sensorimotor processes
to minimize this variability, which has fostered the belief that
the brain exploits the characteristics of sensory and motor
noise to optimize reaching performance. To evaluate this
claim, we developed a feedback control model augmented
with a Kalman filter to assess the influence of anisotropic
sensory uncertainty (planning noise) on endpoint control of
reaching. Multiple simulations were performed with distinct
characteristics of feedback variability: Zero feedback noise,
isotropic feedback noise, and noise representative of known
visual and proprioceptive feedback uncertainty. In addition,
the model was developed to integrate multiple feedback
inputs to assess the effects of multimodal feedback control
on movement variability, and thus additional simulations
were performed to evaluate the effect of optimal feedback
integration on endpoint variability. Simulated reaching
performance with isotropic feedback noise yielded patterns
of variable error similar to those expected of pure execution
noise (zero feedback noise). On the other hand, anisotropic
feedback noise significantly affected predicted endpoint
variability. Specifically, the orientation of 95% confidence
ellipses calculated for simulated endpoints indicated a
combined effect of feedback variability and execution noise.
Conversely, calculated aspect ratios of these ellipses
reflected a strong influence of the characteristics of
feedback noise. Multimodal feedback control also produced
distinct patterns of endpoint variability. In fact, predicted
performance in this condition reflected a unique combination
of feedback and execution noise. In all cases, total error
volume was consistent with a ?near optimal? combination of
planning and execution noise (Faisal and Wolpert, 2009).
These results suggest that under optimal sensorimotor
control endpoint variability arises as a result of the
interaction of sensory/planning noise with execution noise.
Further, we found that this interaction is significantly
affected by both the size and the anisotropic nature of
sensory feedback uncertainty.
1-G-84 Synaptic changes in strength at the
hippocampal formation during the acquisition of
classical eyeblink conditioning in behaving rabbits
Alejandro Carretero1, Renny Pacheco1, Jose Maria
Delgado1, Agnès Gruart1
1
División de Neurociencia
Despite of the high number of works focused in elucidating
the role of the hippocampus in learning and memory
processes, this point is still a matter of discussion. In this
Poster Sessions
Full Abstracts
work we show data of hippocampal recordings from conscious
rabbits during the classical eyeblink conditioning. Briefly, four
groups of rabbits were used, one of them for baseline recordings,
one for pseudoconditioning, and the other two for classical
eyeblink conditioning using delay or trace paradigms. The
conditioned stimulus (CS) consisted of a tone, whilst the
unconditioned stimulus (US) consisted of an air puff. In addition,
two more groups of rabbits were prepared. In one of them,
electrophysiological recordings of hippocampal field EPSPs
(fEPSPs) were carried out whitout applying any conditioning
stimulus, to see the putative effects of enviromental influences at
the different hippocampal synapses (Baseline); and, in the other,
a pseudoconditioning paradigm was used (CS and US stimuli
presented at random) to check if uncoupled stimuli evoked the
same synaptic changes than in their coupled presentation. A pair
of electrical pulses were applied to the perforant path (PP), or
CA3, during the CS and before the US in the case of delay
paradigm, or during the CS-US interval for trace paradigm, in the
CS for pseudoconditioning, and in a fixed interval of time for
baseline. The slope of evoked fEPSPs did not change across
baseline sessions. In contrast, an increase, or a decrease,
tendency was observed in conditioning sessions depending on
the synapse and on the paradigm used. Surprisingly, the
pseudoconditioning protocol provoked changes in the PP to DG,
CA3, CA1 and in the CA3 to CA1 and cCA1 synapses, similar in
magnitude to these obtained with the trace paradigm, but with
relevant differences in some nodal points across the acquisition
curve. These results show that a constant context do not evoke
significant changes in strength at the hippocampal synapses, in
contrast with changes evoked by relevant but uncoupled stimuli,
and those evoked because of the associative learning proccess.
1-G-85 Correlates of desirability in the primary motor
cortex of primates
Joseph Francis1, Brandi Marsh1, Marcello DiStasio1, Aditya
Tarigoppula1
1
SUNY Downstate Medical School
Reward-modulated neural activity is an important component of
conditioned behavior, motor planning, and plasticity, with ample
evidence of its influence on behavior and physiology. Signals of
reward conditions are observed across the motor system,
interacting with systems governing action selection, trajectory
planning, and motivation to generate motor efforts. These signals
also offer the possibility of use as performance feedback to
unsupervised controllers whose goal is to maximize reward. Such
a controller is much more flexible in its ability to choose
component actions that achieve larger goals than one trained with
a supervised learning algorithm. Though this type of controller
can quite reasonably be hypothesized to exist in the primate
brain, here we propose its use in-silico in a brain-machine
interface (BMI) paradigm. A BMI operating under the control of a
reinforcement-learning (RL) agent requires defined rewards that
the agent's goal is to maximize. We suspect that brain-derived
signals of states' relative desirabilities may be able to serve as
useful rewards to an RL agent that has control over movements
of a robotic system. We will present results indicating the
presence of reward signals in the ensemble activity of neurons
recorded in the primary motor cortex of macaques performing a
delayed center-out reaching task, where a color cue informs the
animal during the delay period if it will be a rewarded trial or not.
We pruned all of our data before further analysis so that there
were no significant differences in kinematics between these two
types of movements. By examining mean spiking activity of M1
neurons over pre- and post-movement time periods in principle
component space, we demonstrate a clear decision boundary
separating rewarded and non-rewarded movements. We also
demonstrate the feasibility of using functional near-infrared
spectroscopy (fNIRS) to access hemodynamic signals of
state desirabilities, a source of complementary reward
signals for an RL algorithm, which we have found in both the
monkeys and humans. This work establishes a set of
signals that are available in commonly-used BMI systems
(motor system neural spike recordings and non-invasively
recorded hemodynamic activity) whose properties we aim to
define further, allowing for higher-rate feedback of richer
information to RL motor prosthetic interfaces. These signals
are of known biological importance, and exploring their utility
in BMI research will also contribute to our understanding of
the natural process of motor planning.
1-G-86 A context-dependent process mediates
decay in motor adaptation
James Ingram1, J Randall Flanagan2, Daniel M Wolpert1
1
University of Cambridge, 2Queen's University
Motor learning has been extensively studied using novel
force-fields which perturb the arm during reaching
movements and induce adaptation of the motor commands.
This adaptation is context-specific, being confined to the
movement direction in which the perturbation is
experienced, with limited generalization to novel directions.
Several state-space models have been developed to
capture this pattern of adaptation, modeling how errors
experienced in one movement direction lead to the
progressive acquisition of a context-specific representation
of the perturbing dynamics. Recently, we reported a similar
pattern of context-specific adaptation to the familiar
dynamics of everyday objects. In our task, subjects rotate a
virtual hammer-like tool around the grasp point. In this case,
adaptation involves learning the parameters of the dynamics
(such as the mass), allowing subjects to keep the grasp
point stationary during the rotation. When the object is
presented at different visual orientations, adaptation is
confined to the local orientation at which the dynamics are
experienced, with limited generalization to novel
orientations. Previously, we developed a multiple-context
state-space model, with a generalization function tuned to
visual object orientation, which reproduced this contextdependent behavior. Existing state-space models of motor
adaptation typically include two update terms which
describe how the adaptation state changes from one trial to
the next. The first term, the retention factor, determines how
much of the adaptation state from the current trial is carried
over to the next trial. Its value is always less than one, such
that the adaptation state tends to decay passively from one
trial to the next. The second term, the learning rate,
determines how much of the error from the current trial is
used to update the adaptation state on the next trial. In
order to capture context-dependent behavior, multiple
adaptation states are modeled. In this case, the error is
weighted by a context-dependent tuning function, such that
errors have the greatest influence on the state associated
with the current context, decreasing progressively for
contexts which are further removed. Thus, the contextdependent behavior of previous models has been
implemented by applying a contextual tuning function to the
learning rate whereas the passive trial-by-trial decay
associated with the retention factor has been assumed to be
context independent. Using our virtual object manipulation
task, we tested this assumption that the decay associated
with the retention factor is context independent. Subjects
adapted to the dynamics of the object at a single orientation
and then performed multiple blocks of error-clamp trials at
one of five probe orientations, including the original training
orientation and four novel orientations. After each probe
41
Poster Sessions
Full Abstracts
block, subjects were re-exposed to the dynamics at the training
orientation. Because kinematic errors are essentially eliminated
on error-clamp trials, any de-adaptation induced by probe blocks
must be the result of passive trial-by-trial decay. Contrary to the
assumption of all previous models, we found that de-adaptation
was greatest for probe blocks presented at the training
orientation, decreasing progressively as the relative probe
orientation increased. These results show that context-dependent
processes, traditionally thought to apply only to error-driven
adaptation, also mediate decay in motor adaptation.
1-G-87 Evolution of grip and object representations in
the MI-PMv circuit: A neural trajectory analysis
1
1
2
Carlos Vargas-Irwin , Lachlan L Franquemont , Michael J Black ,
John P Donoghue1
1
Brown University Neuroscience Department, 2Max Planck
Institute for Intelligent Systems
Dimensionality reduction applied to neural ensemble data has led
to the concept of a 'neural trajectory': a low-dimensional
representation of the evolution of the network state over time (for
example, see Afshar et al. 2011). We have developed a novel
algorithm to generate neural trajectories based on spike train
similarity. The algorithm combines spike train similarity metrics
(Victor & Purpura, 1996) with stochastic neighbor embedding
using t-distributions (van der Maaten and Hinton, 2008) to map
spiking patterns into a low-dimensional 'similarity space'. In this
space, network activity at a given time can be represented by a
single point, and distances between the points denote similarities
between ensemble firing patterns. In our approach the low
dimensional space is defined in terms of relative similarity
measures between a given spike train and a set of baseline spike
trains recorded at relevant times during the task. This method
allows us to use a set of key spike trains as landmarks that define
the neural space in a task-relevant manner. Time-evolving neural
activity can then be tracked in relation to these known points. We
have applied this technique to track the evolving states of
neuronal ensembles in PMv and MI through different phases of a
cued grasping with instructed delay (CGID) task. In this task, two
different objects are presented using a turntable. Each object can
be lifted using two different grip strategies (including power grip,
which is common to both objects). By illuminating the object for
one second before providing the grip instruction, we can separate
neural responses associated with object presentation from those
associated with choosing a grip strategy. The resulting neural
trajectories for PMv display a basic re-organization of the neural
space reflecting the information available to the monkey at
different time points: During the object presentation phase, neural
trajectories associated with the same object grow closer together.
We interpret this pattern as the activation of combinations of grip
affordances associated with the target object. Once the grip cue
is provided, the neural trajectories re-arrange themselves,
transitioning from an object-based to a grip-based organization
(with similar grips producing more closely matched activity
patterns). We interpret this pattern as the selection of a particular
grip strategy from the possible grip affordances invoked by the
visual presentation of the object. The PMv neural trajectories can
be contrasted with similar plots derived from MI data. In MI,
responses to the visual presentation of the object are reduced.
The neural trajectories from this area therefore tend to remain undifferentiated until the grip strategy is chosen, without going
through the object-related clustering observed in PMv. There is
still grip related information during the delay period, but it tends to
emerge at a later stage than PMv. This suggests that information
transfer between PMv and MI mainly happens only after a grip
strategy has been selected from among the possible grip
affordances evoked by the target object. Our results also show
that similar grips aimed at different objects are represented by
42
statistically different network states. This observation could
be explained by a partial activation of the competing grip
affordances associated with the target object.
1-G-88 Trial-by-trial assessment of separate
learning processes during motor adaptation
Jordan Taylor1, Melissa A Burney1, Jacob L Wilson1, John
W Krakauer2, Richard B Ivry1
1
University of California, Berkeley, 2Johns Hopkins
University
Visuomotor rotations have commonly been employed to
assess trial-by-trial adaptation of the motor system. Recent
work suggests that multiple learning processes operate in
visuomotor adaptation paradigms. In addition to error-based
processes associated with adaptation of a visuomotor map,
participants may employ strategies to facilitate task
performance. Recent studies show that in such conditions,
error-based learning continues to operate, even though this
may lead to a deterioration in task accuracy. Modeling work
suggests that this phenomenon arises because adaptation
processes use an error signal defined by the difference
between expected and actual sensory feedback, whereas
strategic processes are modified with information related to
measures of task accuracy. The issue of strategy change
has been generally neglected in the motor learning
literature. Insight into this problem may benefit by
considering work in decision-making, and in particular, the
contribution of exploratory behavior to maximize reward.
Strategy change and exploration both involve sampling the
environment to uncover the relationship between a potential
action and its outcome.
In the current study, we systematically manipulated
information in the visual display to make obvious the
relationship between visuomotor rotations and potential
strategies that could offset the rotation. Participants learned
to make 10 cm reaching movements to visual targets with a
45° visuomotor rotation. In a control condition, a singletarget was presented, similar to displays used in most
adaptation studies. For the other two conditions, the
workspace included visual landmarks -- circles that
surrounded the target location, and the spacing between the
landmarks was either sparse or dense. In all three
conditions, vision of the limb was occluded and only
endpoint feedback was provided in the form of a cursor
when the hand passed 10 cm. No explicit instructions were
given about the rotation or strategies that might facilitate
learning. While the presence of landmarks did not appear to
affect the final state of adaptation (similar aftereffect
magnitude), there were notable differences during training.
In the two conditions with landmarks, performance gains
were significantly faster. Moreover, the trial-by-trial
variability differed between groups, with variance being
largest for the groups with landmarks, suggestive of an
exploratory strategic process. This strategic process was
sensitive to the spacing of the visual landmarks, with sparse
spacing associated with high trial-by-trial variance and
dense spacing associated with low trial-by-trial variance.
These results suggest that the time course of the strategic
process is influenced by the training environment, offering a
method to manipulate strategy-use directly during training.
We will discuss how trial-by-trial strategy adjustment can be
observed independently from prediction error-based
learning and incorporated into models of visuomotor
adaptation.
Poster Sessions
Full Abstracts
SESSION 2
A - Adaptation & Plasticity in Motor Control
2-A-1
Cerebellar rTMS disrupts fast learning motor
adaptation process
Robert Hardwick1, Jon S Kennedy1, Chris Miall1
1
University of Birmingham, UK
Smith, Ghazizadeh and Shadmehr (2006: PLoS Biology) propose
that two interacting processes contribute to short term adaptation
of reaching movements; a 'fast process' that is quick to respond
to error but has poor retention, and a 'slow process' that has a
weak response to error but strong retention. While behavioural
evidence indicates that the fast process shares resources with the
declarative memory system (Keisler and Shadmehr, 2010:
Journal of Neuroscience), our understanding of which brain
structures contribute to these processes is still relatively limited.
Here we present data from an rTMS experiment indicating that
the cerebellum is a key contributor to the fast process.
Participants interacted with a robotic arm manipulandum to
control the movements of an on screen cursor, performing point
to point reaching movements from a near start position to a
distant target. After receiving 10 minutes of rTMS over either the
vertex (in a control condition) or the right lateral cerebellum,
participants performed reaching movements under differing curl
fields induced by the robot. In accordance with previously
established protocols (Smith, Ghazizadeh and Shadmehr, 2006:
Keisler and Shadmehr, 2010), participants were first exposed to a
long block of 120 clockwise curl field trials, after which they
completed a brief second block of 15 counterclockwise curl field
trials. Curl field adaptation and after effects were assessed using
error clamp trials; the robot generated a virtual wall that only
allowed movement in a straight line from the start position to the
target location, and the forces that participants produced against
the wall were measured to determine the extent of their
adaptation. Single and dual state models were fit to the error
clamp trials collected from both the vertex control group and
cerebellar stimulation group. The Akaike information critereon
corrected (AICc) was calculated for the two models based upon
chi-squared best fits for each participant. These data were
subjected to repeated measures analysis for each model type
(single or dual state) for the two groups (vertex and cerebellum).
This analysis revealed that while the use of a dual state model
was justifiable for the vertex control group, a single state model
was better able to explain performance of the cerebellar
stimulation group. Further analysis indicated that cerebellar
stimulation specifically disrupted the activity of the fast process
that normally underlies motor adaptation. These data indicate that
the cerebellum is a key contributor to the fast process that
underlies normal motor adaptation, and the experimental
approach utilised here highlights the ability of neurostimulation
techniques to identify brain structures that contribute to modelled
neural processes.
2-A-2
Transfer of motor learning between the two arms
Robert van Beers1, Eli Brenner1, Jeroen B Smeets1
1
VU University Amsterdam
When we learn a movement task with one hand, does this
learning transfer to the other hand? Several studies have
addressed this question by examining how learning to make
reaching movements in the presence of a perturbation with one
hand transfers to the other hand. The results depend on the
perturbation used. Whereas transfer is generally large for
visuomotor perturbations (e.g., Imamizu and Shimojo 1995),
transfer is at most partial for dynamic perturbations (e.g.,
Criscimagna-Hemminger et al. 2003; Galea et al. 2007).
These studies also found conflicting results about the
coordinates (intrinsic vs. extrinsic) in which transfer
occurred. This suggests that the amount and coordinate
system of transfer are specific for the perturbation used. The
aims of the present study were to determine whether there
is also intermanual transfer that occurs independent of
perturbations, and if so, whether this occurs in intrinsic or
extrinsic coordinates. We therefore modified the paradigm
developed by van Beers (2009) to study intermanual
transfer in the absence of perturbations. Ten subjects made
reaching movements from starting positions to visual
targets. They could not see their hand during the
movements, but received visual feedback about the
movement endpoint immediately after each movement. The
experiment consisted of blocks of 50 movements in which
they alternated between movements with the right and the
left hand. The start and target location for each hand were
the same for all 25 movements in a block. To determine
whether transfer occurred in intrinsic or extrinsic
coordinates, we tested two target layouts. In the Symmetric
condition, start and target locations for the two hands were
mirror images of each other relative to the body midline. In
the Parallel condition, start and target locations were the
same for both hands, except for a slight lateral offset. We
analyzed the autocorrelation of the movement endpoints to
infer whether an error made by one hand led to an
adjustment of the planning of the next movement of the
other hand. The lag 1 autocorrelation expresses the
statistical relationship between the endpoints of one hand
and the consecutive endpoints of the other hand. In the
Symmetric condition, we found negative lag 1
autocorrelations for both the extent and the direction of the
movements. In the Parallel condition, the lag 1
autocorrelation of the movement extent was also negative,
but that of the movement direction was zero. Since a
negative lag 1 autocorrelation indicates transfer of error
correction between the hands, these results provide
evidence for intermanual transfer of learning in intrinsic
coordinates. The negative autocorrelation for only the
movement extent in the Parallel condition is consistent with
this as the target movements for the two hands in this
condition had the same amplitude but different directions.
We conclude that there is error-driven intermanual transfer
of motor learning in the absence of perturbations, and this
transfer occurs in intrinsic coordinates.
References Criscimagna-Hemminger SE, Donchin O,
Gazzaniga MS, Shadmehr R (2003) J Neurophysiol 89, 168176 Galea JM, Miall RC, Woolley DG (2007) Exp Brain Res
182: 267-273 Imamizu H, Shimojo S (1995) J Exp Psychol
Hum Percept Perform 21, 719-733 van Beers RJ (2009)
Neuron 63, 406-417
2-A-3
Sensori-motor adaptation to a novel force
field does not transfer to the non-exposed arm in
aged adults and a proprioceptively-deafferented
patient
Fabrice Sarlegna1, Marvin Dufrenne2, Lionel Bringoux2,
Jean-Louis Vercher1, Christophe Bourdin2
1
CNRS & Aix-Marseille University, 2Aix-Marseille University
Everyday, human adults display a surprising ability to adapt
to multiple changes, in the environment or in their own
sensorimotor system, which may perturb the efficiency of
the behaviour. In the present study, we assessed the roles
43
Poster Sessions
Full Abstracts
of visual and proprioceptive signals to adapt to a change in the
dynamic properties of the upper limb. It is well established that
proprioception is an important source of information for the
maintenance of an adapted control of limb dynamics (Ghez &
Sainburg 1995; Sober & Sabes 2005). However, it remains
unclear whether proprioception is crucial for the generalization
and the transfer of newly learned patterns of coordination. Thus,
the starting question of our study was "Does interlimb transfer of
force-field adaptation depend on intact proprioception?". We
addressed this issue by testing the ability of older adults, whose
proprioception is known to be impaired relative to young adults
(Goble et al. 2009), and a deafferented patient who has been
deprived of limb proprioception for years (Cooke et al. 1985). All
subjects were asked to seat on a platform and reach with their
right, dominant hand toward visual targets with full vision. The
platform could rotate and thus create a novel force field (Lackner
& DiZio 1994) as the Coriolis force acted on the moving arm. We
mainly replicated the procedure described by DiZio & Lackner
(1995) and tested the left hand performance in the normal force
field before (pre-test) and after (post-test) the right hand was
exposed to the altered force field for 90 reaching movements
(platform rotation = 120°/s). Our preliminary findings extend those
reported by DiZio & Lackner (1995) by showing that in young
adults, sensori-motor adaptation transfers to the non-exposed
arm, supporting a great body of literature. Thus, in the present
study, young adults saw, and felt, their first reaching movements
with the right hand being perturbed by the altered force field.
Young adults rapidly adapted and restored a straight hand path to
the target despite the platform rotation. Once the rotation
stopped, young adults had to reach with the left hand. Here, initial
movement direction was shifted with respect to that in the pre-test
(before rotation). In contrast, we observed no such interlimb
transfer when the sixty-year old deafferented patient was tested,
even though she was able to rapidly adapt to the altered force
field with vision (see also Sarlegna et al. 2010). Moreover, no
significant interlimb transfer was observed in healthy, sixty-year
old subjects (who also were able to adapt to the altered force
field). In other words, there was no shift in initial movement
direction of the left hand between the pre- and the post-test for
any of the sixty-year old subjects. Before elaborating on the
factors that could explain the observed difference in interlimb
transfer between young and older adults and the possible role of
proprioception, we need to determine whether the lack of shift in
left hand movement direction represents a failure to transfer
sensorimotor adaptation due to ageing or, in contrast, whether the
absence of after-effects highlight the expertise of older adults to
use contextual cues to avoid performing "maladapted" reaching
arm movements in the normal force field.
2-A-4
More realignment for imposed than for naturally
occurring biases
Jeroen Smeets1, Katinka van der Kooij1, Robert J van Beers1,
Willemijn D Schot1, Eli Brenner1
1
VU University
Does the nervous system make lasting corrections for intersensory mismatches? Conflicting answers to this question have
been given. Research imposing a sensory mismatch has provided
evidence that the nervous system realigns the senses, reducing
the mismatch. At the same time, research exploiting natural intersensory biases provided evidence that the nervous system does
not realign the senses. It is unclear whether this difference is due
to a difference in experimental approach or whether corrections to
natural and imposed biases are different. Here, we directly
compare how the nervous system corrects for natural biases and
imposed mismatches. Subjects moved a hand-held cube to virtual
cubes appearing at random locations in 3D space. We alternated
test blocks where subjects moved in complete darkness with
44
feedback blocks where we rendered a cube based on the
position of the hand-held cube. The first test block allowed
us to measure natural biases, whereas subsequent test
blocks allowed us to measure realignment to feedback. In
feedback blocks, we imposed an eye-centered rotation of
plus or minus five degrees on the visual feedback, creating
a mismatch between vision and proprioception. We either
provided feedback during the movement (continuous
feedback) or after the movement had ended (terminal
feedback). In this paradigm, endpoint errors are caused by a
combination of natural biases and the imposed rotation.
Taking advantage of the imposed rotations (-5, 5) canceling
each other, we could decompose errors into a component in
the direction of a subject's natural bias and a component in
the direction of the imposed rotation. We found that there
was much more realignment for the imposed mismatch than
for the natural biases. This difference in realignment was
found with terminal as well as with continuous feedback.
Thus, the nervous system corrects differently for imposed
and natural mismatches.
2-A-5
Transcranial direct-current stimulation
(tDCS) over somatosensory cortex modulates
synaptic mechanisms involved in classical eyeblink
conditioning in rabbits
Javier Márquez-Ruiz1, Rocío Leal-Campanario1, Raudel
Sánchez-Campusano1, Claudia Ammann1, Behnam MolaeeArdekani2, Fabrice Wendling2, Giulio Ruffini3, Agnès Gruart1,
José María Delgado-García1
1
University Pablo de Olavide, 2INSERM, U642 - Université
de Rennes, 3Starlab Barcelona SL
Transcranial direct-current stimulation (tDCS) is a noninvasive brain stimulation technique that has been
successfully applied for modulation of cortical excitability.
Although the effects of weak direct-current stimulation on
the excitability of the central nervous system were reported
decades ago in acute animal experiments, it is only during
recent years that its clinical use in humans has been
promoted. TDCS is capable to induce changes of neuronal
membrane potentials in a polarity-dependent way. When
tDCS is long enough synaptically-driven after-effects are
induced. Since mechanisms underlying these effects are
largely unknown, the development of experimental animal
models to test the immediate and after-effects induced by
tDCS at different cortical areas and its implications in
complex cerebral processes is compellingly needed. In
order to address these objectives, we determined in a first
series of experiments whether simultaneous tDCS applied
to the somatosensory cortex could modify the
characteristics of local field potentials (LFPs) evoked in the
vibrissa S1 area of alert behaving animals by air-puff
stimulation of the contralateral whisker pad or electrical
stimulation of the ipsilateral ventroposterior medial (VPM)
thalamic nucleus. Here we show in behaving rabbits that
tDCS applied over the somatosensory cortex modulates
cerebral cortical processes consequential to localized
stimulation of the whisker pad or of the corresponding area
of the ventroposterior medial (VPM) thalamic nucleus.
Longer stimulation periods indicate that post-stimulation
effects were only observed in the somatosensory cortex
after cathodal tDCS. In a second series of experiments, we
checked whether simultaneous tDCS could also modulate
the acquisition of a well-known model of associative
learning, namely the classical conditioning of eyelid
responses, when stimulation of whisker pad was used as
conditioned stimulus. Consistently with the polarity-specific
reported effects, the acquisition of a classical eyeblink
Poster Sessions
Full Abstracts
conditioning was potentiated or depressed by the application of
anodal or cathodal tDCS respectively, when stimulation of
whisker pad was used as conditioned stimulus, suggesting that
tDCS modulated the sensory perception process necessary for
associative learning. Finally, we also studied the putative
mechanisms underlying immediate and after-effect of tDCS
observed in the somatosensory cortex. Pairs of pulses applied to
the thalamic VPM nucleus (mediating sensory input) during
anodal and cathodal tDCS suggest that tDCS modifies
thalamocortical synapses at presynaptic sites. In addition, we
show that blocking activation of adenosine A1 receptors prevents
the long-term depression (LTD) evoked in the somatosensory
cortex after cathodal tDCS. In conclusion, results reported in this
study confirm earlier studies in humans regarding the effects of
tDCS on cerebral cortex, highlight the potential of this technique
to modulate associative learning, and demonstrate the
participation of adenosine A1 receptors in its selective actions on
cortical circuits.
2-A-6
tDCS modulates adaptation strategies in a
myoelectric-controlled interface task
Claire Schofield1, Kianoush Nazarpour1, Andrew Jackson1
1
Newcastle University
Myoelectric-controlled interfaces (MCIs) provide a unique
opportunity to study mechanisms of motor learning and
adaptation as they allow the manipulation of visuomotor
mappings at the level of individual muscles. We investigated
behavioural responses to perturbation of an MCI mapping which
was redundant and therefore permitted different adaptation
strategies. We also examined whether we could influence
subjects' strategies using transcranial direct current stimulation
(tDCS) to modulate the excitability of primary motor cortex (M1).
In an MCI task, electromyogram signals from multiple muscles
are assigned a direction of action (DoA) in order to control the
movement of a cursor in 2D space. Subjects quickly acquire feedforward control of the cursor to reach targets that appear in a
centre-out design. Subjects exploit task redundancy by
distributing effort across multiple muscles, resulting in broad
muscle tuning functions that are peaked at the DoA. After training
subjects on a bimanual MCI mapping using four muscles in each
hand, we introduced a visuomotor perturbation consisting of a 90°
rotation of the DoAs for half of the muscles in each hand. From
the tuning functions we examined how the preferred direction
(PD) of rotated and un-rotated muscles changed in response to
movement errors caused by the perturbation.
A range of strategies could compensate for a local perturbation to
the DoAs of only some muscles. At one extreme, global re-aiming
would rotate the PD of all muscles by 45° to compensate for the
average rotational error introduced by the perturbation. However,
since the muscle PDs are then no longer aligned to the DoAs,
greater effort is required to reach targets. The optimal solution is
local re-mapping, whereby the PD of rotated muscles is shifted by
90°, while the un-rotated muscles remain unchanged.
During the initial perturbation, subjects received unilateral anodal
or cathodal tDCS over M1 (24 subjects, 8 each for anodal,
cathodal and no stimulation). Anodal stimulation resulted in
increased use of the suboptimal, global re-aiming strategy in the
hand contralateral to stimulation. By contrast, cathodal stimulation
caused a greater tendency towards optimal, local re-mapping and
improved recovery of the original mapping when the perturbation
was reversed.
Previously, anodal tDCS has been found to be beneficial for
motor learning. In contrast, in our experiments we found instead
that cathodal tDCS enhanced the acquisition of an optimal
adaptation strategy. The difference may be that adaptation to
local MCI perturbations requires subjects to identify the
changes at the muscle level that cause global movement
errors. We suggest that increased excitability in M1
produces an incorrect solution to this 'credit assignment'
problem, because suboptimal adaptation in muscles that do
not contribute to errors is enhanced. As such, while tDCS
may enhance adaptation in some circumstances, care must
be taken in generalising these results to more complex
situations where tDCS may cause the motor system to
converge on suboptimal behaviours.
2-A-7
Modular learning of different kinematic
perturbations: A model-based approach
Laura Patanѐ1, Francesco Nori1, Bastien Berret1,
Alessandra Sciutti1, Giulio Sandini1
1
Istituto Italiano di Tecnologia
Reaching, apparently one of the simplest human behaviors,
is actually the result of a quite complex process. In fact,
when we reach for an object our central nervous system
needs to transform sensory signals into the proper muscle
activations to perform the task. This process is often thought
to rely on internal models, i.e. neural representations of the
underlying sensorimotor transformations. A question still
under debate is how the nervous system builds and adapts
these sensorimotor maps and, in particular here, we
investigate whether or not the CNS exploits of the modular
structure of the forward kinematic internal model. Assuming
that modularity may represent an interesting trade-off
between a-priori information and adaptability/flexibility within
the central nervous system, the present study explores the
presence of kinematic modularity in the context of planar
reaching movements. More in detail, we tested the
prediction that modularity should imply a faster learning rate
for perturbations compatible with the existing modules than
perturbations that are incompatible. To investigate this
question we immersed 11 subjects on a virtual reality
environment where arm movements were recorded by a
manipulandum and remapped onto a screen. We confronted
human subjects with two different kinematic perturbations of
comparable difficulty: one compatible with the natural
kinematic modules (or intra-modular) and one incompatible
(extra-modular). We observed that human subjects adapt
faster to intra-modular perturbations, thus providing
evidence in favor of the adoption of a modular strategy by
the central nervous system. Our modularity analysis relies
on a model that allows to predict how subjects behave when
they only have partial knowledge of the remapping they
have been exposed to. We first validated a model (relative)
that allows us to predict how a subject moves when he
assumes a certain model but he is exposed to a remapping
(i.e. different model). A relative model, based on the visual
input (i.e. the relative positions of visual starting point and
target on the screen), has been shown to be much better
than an absolute (totally proprioceptive) model, confirming
previous results on trajectory adaptation to a nonlinear
visuomotor transformation (see J. Flanagan and A. K. Rao,
1995) for our experimental protocol. Then, exploiting this
knowledge, we started to address the problem of modular
learning in order to answer the following question: when we
learn the extra-modular mapping, do we try to exploit our
existing modules or not? This makes a clear prediction for
two different learning strategies: the error-based strategy,
that would just consist of attempting to reduce the error trial
after trial, and the modular-learning approach, that would
predict that subjects first try to perform the new task with
pre-existing modules, before learning the necessary new
ones.
45
Poster Sessions
Full Abstracts
2-A-8
Inter-manual transfer in a neuromotor interface:
An age-controlled study
a desired trajectory, when interleaved with active movement
trials, improve motor learning?
Sabine Gretenkord1, Kianoush Nazarpour1, Andrew Jackson1,
Janet Eyre1, Sara Graziadio1
Subjects were instructed to reproduce both the time-varying
position and velocity of novel hand trajectories in a 2D
horizontal plane, while grasping the handle of an InMotion2
robot arm. Subjects underwent 3 days of training, 90
movement trials per day. During each training session
subjects were shown the target trajectory at regular
intervals, interleaved every 3 trials. For one group of
subjects, these interleaved demonstration trials consisted of
visual training alone. That is, a cursor was presented that
traveled along a visual representation of the target
trajectory. A second group of subjects received visual and
proprioceptive training simultaneously. This group was
presented with the same visual stimulus but in addition their
passive limb was moved through the target trajectory by the
robot using servo control. The robot synchronized the
passive motion of the subject's hand with the motion of the
visual cursor.
1
Newcastle University
'Inter-manual transfer' (IMT) describes an improvement in
performance in a unimanual motor task as a result of contralateral
hand training. Previous imaging studies have observed increased
bilateral cortical activation in the ageing population. This
'hemispheric asymmetry reduction in older adults' (HAROLD) may
be advantageous when motor learning is to be transferred from
one limb to another and thus be beneficial for stroke
rehabilitation. We investigated 'inter-manual transfer' in an agecontrolled study with two aims: first, to clarify the existence and
magnitude of IMT in older adults, and second, to explore the
hypothesis that IMT may be more pronounced with greater age.
We studied right-handed subjects from two age groups: young
adults (10 subjects aged 20-35, mean: 28.4, range 26-32) and
older adults (13 subjects aged over 55, mean: 69.9, range: 5582). Subjects made repeated movements of a myoelectric cursor
towards a target, receiving scores that reflected the distance from
the target during the hold period. Cursor position was determined
by smoothed, rectified electromyogram (EMG) from two intrinsic
hand muscles. Controlling muscles were either in the right or left
hands acting orthogonally in the task space. The study was
conducted in a crossover design: we assessed left hand
performance before and after a period of either right hand
training, or a resting period of similar length during which a video
was shown. We analysed the scores and also the distance to the
target at the beginning of the hold period. In the young adult
group, the right hand training led to a 39% increase in left hand
score, in the older adult group the percentage increase was
higher (72%). Analysis of the distance parameter showed that the
right hand training improved the left hand performance by 28% in
the young adults group and by 35% in the older adults group.
Cursor control with right hand improved with training in both
groups comparably in terms of both score (44% in group 20-35
and 42% in group 55+) and distance (41% in group 20-35 and
19% in group 55+). In addition, we found a correlation between
the right hand training and inter-manual transfer in the young
adults group. However, this correlation was not found in the older
adults group. Our results revealed that inter-manual transfer
occurs in both age groups. The 55+ age group showed an even
higher relative improvement than the young adults group. We
showed that training of the dominant hand can improve motor
function in the non-dominant hand in an age-controlled
experiment. This effect might prove beneficial for stroke
rehabilitation, as the training of the paretic hand is often a tiring
and frustrating experience, and in the most severe cases
impossible. We therefore speculate that training of the non-paretic
hand could improve motor function in the paretic hand.
2-A-9
Proprioceptive training on a desired trajectory
improves motor learning
Jeremy Wong1, Dinant A Kistemaker1, Paul L Gribble1
1
University of Western Ontario
Recent work has investigated the link between motor learning and
sensory function in arm movement control. Force-field learning
results in reliable changes to sensed limb position (Ostry et al.
2010). Learning to reach to visual targets quickly and accurately,
in the absence of a force-field, results in improvements in
proprioceptive acuity (Wong et al., 2011). These findings are
consistent with the idea that reducing motor error through learning
may result in more precise estimates of limb position. Here, we
sought to investigate the reverse: does proprioceptive training on
46
After each active movement trial the difference between the
subject's movement and the target trajectory was measured.
Data suggest that passive proprioceptive training on the
desired trajectory results in benefits to motor learning, both
in terms of both position and timing error. Surprisingly, an
additional group of subjects who were asked to actively
move during demonstration trials, with robot assistance, did
not show the same benefits to reduction in position error.
These results support the notion that passive proprioceptive
training improves motor learning. Here we also report data
on motor learning retention, measured 2 months post
training.
References Ostry DJ, Darainy M, Mattar AA, Wong J,
Gribble PL. Somatosensory plasticity and motor learning. J
Neurosci. 2010 Apr 14;30(15):5384-93. Wong JD, Wilson
ET, Gribble PL. Spatially Selective Enhancement of
Proprioceptive Acuity Following Motor Learning. J
Neurophysiol. 2011 May;105(5):2512-21.
2-A-10 Capacity of LTP-like plasticity is essential
for motor learning
Gabriela Cantarero1, Rebecca O'Malley1, Pablo A Celnik1
1
Johns Hopkins Medical Institution
Plasticity of synaptic connections in the primary motor
cortex (M1) is thought to play an essential role in learning
and memory. In humans, training a motor skill results in an
occlusion of (i.e. reduced capacity for) LTP-like plasticity,
where the magnitude of occlusion is proportional to the
retention of the motor skill. Previous work has speculated
that occlusion is essential for consolidating a previously
learned motor task; however, whether the capacity to
undergo LTP-like changes is essential for learning a
subsequent novel motor task remains unknown.
We have previously shown that training 2 motor skills in an
interleaved practice order (IPO) results in superior learning
retention and occlusion of LTP-like capacity. In contrast,
training 2 motor skills in a blocked practice order (BPO)
results in impaired retention of learning and normal LTP-like
aftereffects (i.e. lack of occlusion). We hypothesized that the
capacity to undergo LTP-like changes following learning in a
BPO would result in a renewed ability to learn and retain a
novel 3rd motor skill, whereas occlusion following learning in
an IPO schedule would result in a subsequent impairment to
learn and retain a novel 3rd motor skill. We compared
training and retention of a naïve 3rd motor task after
subjects trained 2 motor skills in either a BPO or IPO
Poster Sessions
Full Abstracts
schedules. We found that after training in a BPO subjects were
capable of learning and retaining a third naïve motor skill,
whereas subjects who trained first in an IPO were then impaired
in their retention of the 3rd novel motor skill. These results
suggest that the capacity to undergo LTP-like changes prior to
training is essential for learning a new task, without it retention is
impaired.
hand grasping function, relies on structural and functional
connectivity in both ipsilesional and contralesional
parietofrontal pathways involved in visuomotor information
processing. Extant integrity of this structural network may
serve as a predictor of response to longitudinal therapeutic
interventions geared towards training SMR in the lesioned
brain.
2-A-11 Parietofrontal visuomotor pathway integrity after
stroke determines volitional sensorimotor rhythm (SMR)
modulation skill
2-A-12 Can older adults learn new dynamics?
Ethan Buch1, Amir Modir Shanechi2, Alissa D Fourkas1, Cornelia
Weber3, Niels Birbaumer3, Leonardo G Cohen1
1
National Institute of Neurological Disorders & Stroke, NIH,
Washington University School of Medicine, 3University of
Tübingen
2
Rehabilitation strategies that actively engage the extant motor
system are more successful at alleviating motor deficits following
stroke (Wang et al., 2010). Currently available interventions
however, such as constraint-induced movement therapy (CIMT)
or active forms of robotic-assisted upper limb therapy, exclude
patients with severe motor deficits due to their minimum
requirements for residual motor function. Alternative paradigms
based on motor imagery, the conscious rehearsal of egocentric
motor actions without overt motor output, have been proposed as
a means of actively engaging the motor system in these patients
(Zimmermann-Schlatter et al., 2008). Longitudinal brain-computer
interface (BCI) training based on motor imagery of affected hand
grasping results in improved control of ipsilesional sensorimotor
rhythm (SMR) modulation after chronic stroke (Buch et al., 2008).
However, the mechanisms underlying a patient's ability to use
motor imagery to successfully modulate SMR are not known.
Here, we investigated the impact of individual patient's lesion
pathology on functional and structural network integrity related to
this volitional skill.
Magnetoencephalography (MEG) data was acquired during BCI
training. Functional connectivity between each MEG sensor was
estimated, and used to derive a functional network graph. A
structural network model was also constructed, and local
estimates of extralesional gray and white matter microstructure
determined from T1-weighted and diffusion-weighted MRI data.
We used a graph theoretical approach to assess how properties
of distributed network architecture related to skill acquisition
across our patient group.
We found that inter-individual variability in patients' lesion
characteristics resulted in differential changes to both functional
and structural network characteristics that were related to SMR
modulation skill. Patients displaying greater functional global costefficiency, a measure of overall information integration throughout
the distributed functional network, achieved greater levels of skill
(p<0.05). An analysis of lesion damage to structural network
connectivity revealed that changes in nodal and edge
betweenness centrality, a measure that characterizes the
importance of a brain region or fiber pathway for integrating
visuomotor information between frontal and parietal cortical
regions and related thalamic nuclei, correlated with skill for both
the ipsilesional M1 (p<0.05) and superior longitudinal fascicle
(p<0.01), respectively. Finally, white matter microstructure
integrity in regions of the contralesional SLF adjacent to primary
sensorimotor and posterior parietal cortex, as well as co-localized
regions of grey matter volume, positively correlated with SMR
modulation skill (p<0.05, FDR corrected).
These collective results suggest that volitional modulation of
ipsilesional SMR acquired through BCI training, which can be
used to control a mechanical hand orthosis supporting affected
Alaa Ahmed1, Helen J Huang1
1
University of Colorado
Older adults tend to move more slowly and with greater
muscle coactivation (1). These age-related changes may
indicate a reduced ability to learn new dynamics and/or
adapt to a changing dynamic environment. The purpose of
this study was to compare learning of novel dynamics in
older adults with learning in younger adults. We used the
well-studied motor learning task of reaching in a viscous curl
force field while holding the handle of a robotic arm (2). We
hypothesized that older adults would learn less well than
younger adults.
We tested 12 older adults (73.4 ± 5.5 yrs) and 15 younger
adults (23.8 ± 4.7 yrs) as they performed goal-directed
reaching movements. Seated subjects controlled a robot
handle to move a cursor from a home circle to a target circle
20 cm away. The cursor, home, and target circles were
displayed on a computer monitor suspended vertically in
front of the subjects at eye-level. Subjects performed blocks
of 200 null trials (no forces, Null 1), 250 force trials (Force
1), another 250 force trials (Force 2) and 200 null trials (Null
2). During the force trials, the robot applied a perturbing
force perpendicular and proportional to the hand velocity
(gain = 20 Ns/m), creating a curl force field. Also, one in five
trials was a catch trial during which anticipatory force was
measured (i.e. a force channel).
We quantified learning using movement error and a learning
index based on anticipatory force. Movement error was
quantified as the maximum perpendicular deviation of the
handle from a straight line path between the home and
target circles. The learning index was equal to ratio of the
estimated gain, b, to the actual gain of 20 Ns/m. The
estimated gain, b, was obtained from a least-squares fit of
the perpendicular force profile to the ideal force profile
based on the hand velocity (Fx = b*Vy).
When just reaching with no robot forces (Null 1), older and
younger adults had similar movement errors (p = 0.313).
Upon first experiencing the force field, both groups had
large movement errors. Younger adults were able to rapidly
reduce movement error from 8.78 ± 0.51 cm to 1.71 ± 0.09
cm by late Force 1 whereas older adults only reduced
movement error from 7.98 ± 0.40 cm to 2.47 ± 0.22 cm. The
movement errors of the older adults at late Force 1 were
significantly greater than the younger adults (p = 0.002). As
learning continued in Force 2, both groups reduced
movement errors further, but older adults continued to have
larger movement errors than the younger adults (older =
2.28 ± 0.23 cm, younger = 1.36 ± 0.11 cm, p < 0.001). The
learning index revealed that by late Force 1, older adults
had learned 51.8 ± 6.4% of the actual gain which was
significantly less than the learning observed in younger
adults (74.5 ± 4.2% of the gain, p = 0.005). By the end of
the learning period, late Force 2, both groups had higher
learning indices but older adults still learned significantly
less (older = 64.0 ± 4.8 %, younger = 83.8 ± 2.3%, p <
0.001). The results demonstrate that older adults can learn
novel dynamics, but learn less well compared to younger
47
Poster Sessions
Full Abstracts
adults. This suggests that age-related changes in movement
control may be a strategy to compensate for impaired learning.
Supported in part by NIH 5T32AG000279.
1. Seidler-Dobrin, He, & Stelmach, Motor Control (1998)
2. Shadmehr, Mussa-Ivaldi, J Neurosci (1994)
2-A-13 The role of the cerebellum in force-field
adaptation
Damien Pastor1, Adrian Haith2, Yves Rossetti1, Reza Shadmehr2,
Jacinta O'Shea3
1
Lyon Neuroscience Research Center INSERM U1028, 2Johns
Hopkins University School of Medicine, 3FMRIB, University of
Oxford, John Radcliffe Hospital
Patient lesion data have demonstrated that the cerebellum is a
critical structure for the adaptation of reaching movements. Here
we used cerebellar Transcranial Direct Current Stimulation
(TDCS) in neurologically intact individuals and tested the effect of
increasing (anodal) and decreasing (cathodal) cerebellar
excitability on learning and retention of force field adaptation with
the right hand. Stimulation of the right cerebellum was applied
during adaptation (2mA, 25min) and altered motor learning
symmetrically: anodal TDCS caused faster learning, while
cathodal TDCS slowed learning. Both error reduction curves and
the development of after-effects were significantly altered by
TDCS. Specifically, stimulation altered the magnitude of trial-bytrial movement plan updating based on error feedback from the
previous trial. Anodal TDCS caused greater proportional error
updating, whereas cathodal TDCS reduced the update
magnitude. After 24 hours, in the sham condition, subjects
retained a motor memory of the acquired adaptation. This was
more labile in the cathodal condition and significantly more stable
in the anodal condition. These results suggest that the cerebellum
drives motor adaptation by a process of trial-by-trial error
updating and that the cerebellum is an important structure for
initiating the development of long-term motor memories.
2-A-14 Neurophysiological constraints for biological
plausibility of machine learning
Martin Nilsson1, Fredrik Bengtsson2, Carl-Fredrik Ekerot2, Henrik
Jörntell2
1
Swedish Institute of Computer Science (SICS), 2Lund University
It is soon 70 years since the first paper on a mathematical model
of a neuron was published (McCulloch and Pitts, 1943). This
paper sparked an explosive interest in neural networks and
machine learning. Although this research was originally inspired
by neurophysiology, it gradually evolved into a research field in its
own right, not necessarily limited by requirements of biological
plausibility. Presently, the field has become highly sophisticated,
and has generated techniques for machine learning and adaptive
systems pervasive in engineering and industrial applications.
Also neurophysiology research has made great strides forward,
but along different lines. At this point, it is interesting to ask what
results in machine learning can contribute to a better
understanding of biology. For instance, at least in vivo, direct
forms of back-propagating multi-layer learning appear to be
unlikely. Further, each neuron seems to have only a single output,
which implies that if operating as an adaptive filter (Widrow and
Hoff, 1962), it can either output reconstructed input or error, but
not both. On top of this, all neurons display more or less
continuous probability distributions of interspike intervals, with a
wide spectrum of variances. How are the complex adaptive
behaviours observed possible under such severe conditions? Any
tentative principle of adaptivity must allow neurons to bootstrap
48
automatic sensorimotor association, i.e. connect sensor
input to appropriate motor output; to implement stable
feedback control of limbs; and to construct synergies, the
combination of elementary muscle movements into a lowerdimensional subspace of aggregated movements.
We discuss these issues in the context of central nervous
system communication and control of movement, with an
emphasis on the spinal cord, the cerebellum, and the precerebellar structures, which have served as sources of
inspiration for some of the historically most popular adaptive
control paradigms (Marr, 1969; Albus, 1971). A micro-circuit
perspective ("bottom-up") is used rather than a
psychophysics one ("top-down"), and the preliminary
conclusion is that, yes, adaptive control by the central
nervous system is indeed possible, but only barely so.
Nature is ingenious.
This presentation describes modeling research by MN,
based on neurophysiological in vivo experiments by FB,
CFE, and HJ.
References: McCulloch,W.S. and Pitts,W.: A logical calculus
of the ideas immanent in nervous activity. Bull. Math.
Biohys., vol. 5, pp. 115-133. 1943; Widrow,B., and
Hoff,M.E., Jr.: Adaptive Switching Circuits. IRE WESCON
Conv. Rec., Pt. 4, pp. 96-104. 1962; Marr, D.: Vision. W.H.
Freeman. 1969; Albus, J.: Theory of cerebellar function.
Mathematical Biosciences, Volume 10, Numbers 1/2,
February 1971, pp. 25-61. 1971.
2-A-15 Dual processes in prism adaptation Evidence from computational modelling and
transcranial direct current stimulation
Jon Kennedy1, Damien Pastor2, Valerie Gaveau2, Matthieu
Kandel2, Yves Rossetti2, Chris Miall1, Jacinta O'Shea3
1
University of Birmingham, 2University of Lyon, 3University of
Oxford
The shift in perceived visual locations caused by wearing
wedge prism goggles induces adaptation that consists of
multiple empirically dissociable components (Redding,
Rossetti, & Wallace, 2005; Pisella et al., 2004). These
include changes in visual and proprioceptive perception,
and strategic changes in pointing behaviour relative to those
percepts. Perceptual changes generalize across entire
stimulus dimensions, whereas strategic changes show a
typical associative generalization gradient (Bedford, 1993).
Perceptual changes also generalize across time (i.e. they
persist), whereas strategic changes are less durably
retained (Redding & Wallace, 1993). Smith, Ghazizadeh &
Shadmehr (2006) have shown that a model with two states,
one which learns and forgets quickly, and one which learns
and forgets slowly, can explain memory savings and other
learning phenomena in a force field adaptation procedure.
Such a model has also proved effective in explaining
saccadic adaptation phenomena (Ethier, Zee, & Shadmehr,
2008). Can such a model successfully explain performance
in a prism adaptation task? We used a 10 degree wedge
prism adaptation design with blocks of open loop pointing (a
standard method of measuring aftereffects of adaptation in
prism procedures; Redding et al., 2005) interleaved
throughout closed-loop pointing adaptation, and found that a
two-state model was a significantly better fit to the data
(after adjusting for model complexity, using the chi-squarebased small-sample-corrected Akaike Information Criterion,
AICc; Burnham & Anderson, 2002) than a one-state model,
across 25 healthy participants. To reflect what is known
about the empirically dissociable components of the overall
Poster Sessions
Full Abstracts
adaptive response in prism adaptation, we also tested on this
data set a model with two qualitatively distinct learning modules,
one which learnt from a perceptual error signal (the prism
displacement, less the current perceptual adaptation), and one
which learnt from a behavioural error signal (performance error;
the visual distance of the end-point of the movement from the
target). This model, with the same number of parameters as the
Smith et al. (2006) two-state model, provided a significantly better
fit to the data. With the same procedure, and ten new healthy
participants, cathodal, anodal, or sham (repeated measures)
transcranial Direct Current Stimulation (tDCS) was applied to the
cerebellum prior to exposure to the wedge prisms. Marked
disruption of retention was observed in the cathodal tDCS
condition, compared to the anodal and sham tDCS conditions, as
manifested by participants showing less adaptation to the prisminduced shift during exposure to it, with more rapid loss of
aftereffects during interleaved open-loop pointing blocks, and
more rapid return to pre-adaptation performance during wash-out.
The dual state model's ability to account for the effects of tDCS
are presented and discussed.
2-A-17 A shared mechanism underlying the use of
visual and proprioceptive information in reaching
and grasping
2-A-16 Changes of phase synchrony in motor learning
during whole versus parts practice
Karen Bourns1, Francisco Colino1, Darian Cheng1, Keith
Brewster1, Brendan Cameron1, Gordon Binsted1
Pablo Burgos1
1
1
Universidad de Chile
Pablo Burgos, Christian Arellano, Pedro Maldonado. Program of
Physiology and Biophysics, ICBM, Faculty of Medicine, University
of Chile, Santiago, Chile. ([email protected])
Objectives: The proposal for a modular organization for motor
learning (MOSAIC model), reported evidence in tasks of
kinematics and kinetics adaptation first with respect to the ability
of the nervous system to learn parts of a motor task and then
integrate these parts (composition), and second the ability to use
parts of a motor task after learning it in an integrated manner
(decomposition). The learning of new sensorimotor internal
models explain these adaptations and would be a possible
explanation for a refinement in sensorimotor networks when we
practice, what could be evidenced with the decline of some
measure of long-distance synchronization during learning. The
aim of this study is to report changes in phase locking value
(PLV) in electroencephalographic signals (EEG) in both the
composition condition (practiced in parts) and the condition of
decomposition (total practice) during motor learning of a
continuous task of visuomotor and kinetics adaptation.
Methods: Using a computer and an analog gamepad, 10 people
learned a continuous visuomotor task (video game), which was to
advance by different routes without colliding with stationary and
moving obstacles. People had to learn to move with his left hand
the character without colliding with obstacles, but the gamepad
axes had rotated 90 degrees clockwise (visuomotor adaptation),
and also had to learn to shoot various bullets with his right hand
to destroy obstacles, but the gamepad lever had an elastic
resistance higher than usual (adaptation kinetics). One group (n =
5) practiced 4 days, 30 trials of 90 seconds in an integrated
manner (visuomotor kinetic adaptation), and another group (n =
5) practiced in parts (15 trial of an adaptation and then 15 trials of
the other). On the fifth day all were evaluated in the integrated
task. We recorded the performance in the game, the movements
on the joystick and continued activity of 32-channel EEG
(Biosemi).
Results: Learning is evident in both groups, under the conditions
of decomposition and composition, with significant improvements
in performance in visuomotor and kinetic adaptations.
Furthermore, between day 1 and day 5 in the variables of
distance traveled, number of crashes, and number of objects
destroyed. The best average performance in the task of
evaluation of the day 5 was for the whole group for both
kinetic and kinematic adaptation. The phase synchrony
decreases between day 1 and day 5 for all subjects
between occipital and frontal electrodes in Beta band. No
significant differences between groups were found.
Conclusions: Our results are consistent with other studies
is proposing to lower the EEG phase synchrony as a global
mechanism and a marker of learning. The findings are also
consistent with studies that suggest a modular organization
of learning and motor control,MOSAIC model. The practice
of a motor task generates sensorimotor memories or
internal models that allow you to refine the use of largescale cortical networks. Probably the greatest activation for
stages of motor consolidation in striatal and cerebellar
networks, and premotor cortex, reported in other studies,
could explain the decline in large-scale synchrony of EEG.
University of British Columbia Okanagan
Little is known about adaptation to acute onset vision loss.
Based on the seminal work of Pascual-Leone (see: Merabet
& Pascual-Leone, 2010) significant cortical plasticity is
known to take place during prolonged visual deprivation.
Here, we use a two-experiment study to investigate
adaptations in processing visual and proprioceptive
information in acute onset vision loss. Experiment 1: Visual
input was removed for 2-hr to simulate acute onset shortterm vision loss. Subjects underwent visual deprivation on
two separate occasions; adaptations in grasping behaviour
were assessed in one session, and adaptations in cerebral
blood flow velocity (CBFv) during the other. Behavioural:
Subjects performed 80 trials of a reaching and grasping task
pre- and post-deprivation (160 trials). Proprioceptive control
(No-Vision) was used for the first 40 trials, followed by 40
trials using visual control (Vision). Vision was removed using
PLATO goggles (Translucent Technologies, Inc.); goggles
were occluded for the duration of No-Vision trials. In all trials
subjects grasped a circular target in response to an auditory
tone. Prior to the initiation of each trial, the subject's arm
was passively moved to the target location and returned to
the start position by an experimenter. Kinematic
measurements (e.g. limb position, grip aperture) were
obtained using 3 infrared markers (thumb, forefinger, and
wrist) and an Optotrak Certus (Northern Digital, Inc.). CBFv:
Using transcranial Doppler ultrasound (TCD), CBFv was
measured at the posterior cerebral artery (PCA) and middle
cerebral artery (MCA) pre- and post-deprivation, and at 30minute intervals during, while subjects performed a standard
neurovascular-coupling task (see: Aaslid, 1982; Ainslie &
Duffin, 2009). Results of Experiment 1 show significant
modulation of kinematic behaviour and regional CBFv of the
MCA and PCA following 2-hr of acute vision loss.
Experiment 2: Visual input was removed for 8-hr to simulate
acute onset long-term vision loss. Pre- and post-deprivation
subjects performed the same grasping task as Experiment
1. In addition to the grasping task, subjects completed an
oddball detection task pre- and post-deprivation. SSEPs &
VEPs were recorded in response to tactile and visual
stimuli. Both tasks involve detecting stimulus onset to either
the right or left index finger. For one block, stimulus
probability was 0.8 right (non-target stimuli) and 0.2 left
(oddball stimuli), and opposite in the second block. During
the 8-hr deprivation, subjects performed normal daily
49
Poster Sessions
Full Abstracts
activities with the assistance of an experimenter. Each subject
also participated in a control day, where a pre- and post-test was
performed around an 8-hr period of normal daily activities with full
vision. Modulation of kinematic behaviour and variations in ERP
amplitudes and latencies observed in Experiment 2 will be
compared to the results of Experiment 1 and discussed in relation
to current theories in cortical plasticity. It seems a possible shared
physiological or attentional mechanism may underlie processing
of visual and proprioceptive information for action.
2-A-18 Changes in muscle synergies during adaptation
to novel visuomotor transformations
1
1
2
Denise Berger , Reinhard Gentner , Timothy Edmunds , Dinesh
K Pai2, Andrea d'Avella1
1
IRCCS Fondazione Santa Lucia, 2University of British Columbia
A long standing hypothesis is that the CNS generates motor
output via flexible combinations of muscle synergies. A modular
architecture would allow for efficient and fast adaptation to
visuomotor transformations, i.e., changes in the mapping between
motor output and visual feedback, because it reduces the number
of parameters to learn. We previously reported that during
isometric reaching in a virtual environment ("virtual surgery")
adapting to a novel visuomotor transformation that is incompatible
with the set of muscle synergies required by the task is more
difficult than adapting to a compatible transformation. In contrast
to previous studies of muscle synergies reporting lowdimensionality in the motor output, we provided evidence for
modularity from testing the prediction that a truly modular
controller cannot easily adapt to perturbations which are
incompatible with the modules. Here we present new data and
results on the changes in muscle synergy structure during
adaptation to virtual surgeries. Subjects sat with their hand and
forearm in a splint attached to a force transducer positioned on a
desktop in front of them. The subjects' view of their hand was
occluded by a LCD monitor displaying a virtual scene with a
desktop and a spherical cursor matching the position of the hand.
The task required moving the cursor to reach a target in one of
eight directions in the horizontal plane. Cursor displacement was
computed either from the recorded forces (force control) or the
forces estimated with a linear EMG-to-force mapping of the EMGs
recorded from several shoulder and elbow muscles (EMG
control). Data collected from an initial block in force control were
used to estimate the EMG-to-force matrix used for EMG control in
the rest of the experiment and to extract time-invariant muscle
synergies using a non-negative matrix factorization algorithm.
EMG control allowed to perform virtual surgeries by altering the
virtual forces generated by the muscles simulating complex rearrangement of the tendons. All surgeries were constructed by
introducing a rotation in muscle space and did not affect the
forces that could be generated by activation of individual muscles.
However, only incompatible surgeries remapped forces
associated with the synergies along a single direction in force
space, thus affecting the force generation capability of a
synergistic controller. With respect to our previous experimental
protocol, we decreased the trial duration and increased the total
number of trials in each surgery phase. We found that subjects
with practice significantly increased the number of successful
trials during the compatible surgery but not during the
incompatible one. The average error in the initial movement
direction decreased during both surgeries. However, the learning
rate during the compatible surgery was significantly higher and
led to significant less errors at the end of the perturbation. We
then investigated the changes in synergy structure by computing
the error in the reconstruction of the muscle patterns of each
experimental block with the synergies extracted before any
surgery. We found that muscle patterns could be captured by the
initial synergies throughout all blocks in baseline and during the
50
compatible surgery. On the contrary, the reconstruction
error decreased significantly during the incompatible
perturbation, suggesting that changes of synergy structure
occurred during the incompatible surgery.
2-A-19 Bihemispheric transcranial direct current
stimulation enhances skill learning and transfer to
the untrained hand
Jörn Diedrichsen1, Sheena Waters-Metenier1, Masud
Husain1, Tobias Wiestler1, Jörn Diedrichsen1
1
University College London
Transcranial direct current stimulation (tDCS) to primary
motor cortex (M1) has been shown to facilitate motor skill
learning in the contralateral hand (Reis at al. 2009),
putatively by increasing neural plasticity in the tissue
underlying the anode. Particularly impressive benefits are
exhibited when tDCS is implemented using a bihemispheric
montage, with the cathode placed over the ipsilateral M1
(Vines et al. 2008). However, it has not been established
whether (a) there is an adverse impact on learning
processes in the hemisphere beneath the cathode or (b)
tDCS-induced plasticity has a negative effect on the
performance of other non-trained skills. To explore these
two issues, we investigated the influence of bihemispheric
tDCS on two motor learning tasks, including (1) the
'sequential finger task' (SFT), which involved the rapid
production of sequences of five isometric finger presses and
(2) the 'configural finger task' (CFT), which required
simultaneous presses of certain sets of fingers while
immobilising other fingers (i.e. movements analogous to
those executed in generating a chord on a piano or guitar).
We aimed to (a) investigate the effects of bihemispheric
tDCS on the learning of these motor skill tasks, (b)
determine if there is a negative effect of tDCS on the
transfer of skill from the trained hand to the untrained hand,
and (c) investigate whether there is a link between tDCScoupled training and attenuated performance in untrained
motor skill modalities. Forty-four young, healthy, righthanded participants were pseudo-randomly assigned to one
of four groups (SFT-tDCS, SFT-sham, CFT-tDCS, and CFTsham). Baseline skill in the two motor tasks was assessed in
all subjects. Subsequently, they completed four days of lefthand training in either the SFT or CFT, with real or sham
stimulation. During tDCS, which was executed in a doubleblind fashion, the anode was placed above the contralateral,
right M1, and the cathode over the ipsilateral, left M1. In
each training day, participants practiced the same four
sequences or configurations for an hour, and tDCS was
administered for the first 25 min. Training was followed by
three testing sessions without tDCS, at 5, 12, and 33 days
after completion of training. For both the SFMT and CFMT,
the movement times were measured, while accuracy
demands were held constant. After training, the movement
times required to execute sequences and configurations
were, respectively, 40% and 32% faster in individuals who
received anodal tDCS relative to sham. This advantage was
preserved even four weeks after the final tDCS training
session. Importantly, across both the SFT and CFT,
individuals who received tDCS additionally exhibited
enhanced training-induced transfer of skill to the right
(untrained) hand (i.e. the hand associated with the cathodemodulated ipsilateral M1). Finally, tDCS-coupled training in
one task was not associated with impairment in
performance in untrained motor tasks and, in some cases,
actually was linked to improved motor performance. These
results add to the emerging literature that supports a
powerful translational capacity of tDCS for the treatment of
Poster Sessions
Full Abstracts
motor deficits incurred as a result of stroke or traumatic injury.
Given the unexpected finding of an enhanced transfer of skill to
the untrained hand, it will subsequently be important to establish
the neural underpinnings of such facilitation via functional and
structural MRI.
2-A-20 Dominant arm is not always dominant: A hidden
excellence of non-dominant arm in adapting to dynamical
environments resulting from contralateral arm movement
Atsushi Yokoi1, Masaya Hirashima1, Daichi Nozaki1
may be related to the functional specialization of each arm
during bimanual movement (e.g., prime actor and
supporting role for dominant and non-dominant arm,
respectively).
2-A-21 Partial learning of object dynamics based on
fingertip forces in the absence of kinematic errors
Frederic Danion1, Jonathan S Diamond2, Randy J
Flanagan2
1
Universite de la Mediterranee, 2Queens University
1
Graduate School of Education, The University of Tokyo
It is believed that a greater control ability of dominant arm makes
it play a leading role in bimanual movements like opening a bottle
of wine. Here we challenge this widespread notion from the
viewpoint of ability of motor adaptation. Specifically, we show that
during bimanual movement the non-dominant left arm has a
greater ability in adapting to dynamical environments resulting
from the motion of the right arm than vice versa.
In the first experiment, right-handed subjects trained to perform
simultaneous bimanual reaching movement in forward direction (0
degree) under the presence of a velocity-dependent force field
imposed on either left or right arm (n = 8 for each experiment).
After the 80 trials of training session, the degree of adaptation
was assessed using error clamp trials interleaved in every other
trial. The movement directions of the untrained arm were
randomly selected from 8 directions (0, 45, 90, 135, 180, 225,
270, 315 deg) to evaluate how the degree of motor adaptation
changed as the movement direction of the untrained arm became
distant from the training direction (0 degree). It should be noted
that the reach direction for the trained arm was maintained the
same direction (i.e., forward direction) throughout the
experiments.
Consistent with our recent study (Yokoi et al., J Neurosci 2011),
regardless of the arm trained, we observed a Gaussian-like
generalization pattern that was modulated with the reach direction
of the untrained arm. A novel finding in this study was that the
modulation amplitude of the generalization function was
significantly larger when the training was performed with the left
arm (Left arm: 48.4±10.3 %; Right arm: 36.0±10.2 %; p = 0.015,
one-sided t-test), indicating the dominance of the right arm in
terms of the greater influence on the motor learning of the
opposite arm. However, the greater influence that the left arm
received could be interpreted as its greater flexible ability to the
dynamical environments changing with the right arm's movement.
To test this prediction, the second experiment was designed so
that the same forward reaches of the left/right arm were exposed
to conflicting (clockwise: CW and counter-clockwise: CCW) force
fields associated with different movement directions of
contralateral arm [i.e., (0 deg: CW), (90 deg: CCW), (180 deg:
CW), (270 deg: CCW)]. The order of the 4 movement
configurations was randomized, and the error clamp trials were
randomly interleaved in one of 7 trials. A state-space model
(Yokoi et al., J Neurosci 2011) whose parameters were estimated
using the results of the first experiment predicted the excellent
adaptation ability of the left arm in this task. As predicted from this
simulation, the non-dominant left arm outperformed the dominant
right arm; the amount of the aftereffects averaged across 4
movement directions of the untrained arm was significantly
greater for the right-handed subjects trained with the left arm
(n=4) than those trained with the right arm (n=4) [F(1,6) = 6.096,
p < 0.05].
These results suggest that the dominant arm is not always
superior to the non-dominant arm. The non-dominant arm has the
hidden superior ability specialized in adapting to mechanical
interaction resulting from the contralateral arm's movement. This
Humans can learn to perform accurate movements when
grasping objects with a variety of novel dynamics and it is
thought that this involves acquiring an internal model of the
dynamics of the object. Kinematic errors (i.e., discrepancies
between predicted and actual movement trajectories) are
considered central for such adaptation (Smith et al. 2006;
Shadmehr et al. 2010; Wolpert et al. 2011). However, when
moving a grasped object with novel dynamics, participants
also experience kinetic errors in the form of mismatches
between actual and predicted load forces acting on the hand
or fingertips. Here we investigated the contribution of such
kinetic errors to the learning of object dynamics under
conditions in which significant kinematic errors were
removed. Participants were asked to transport an object
with novel dynamics (a rotary velocity-dependent force field)
between targets in a horizontal plane. They grasped the
object with the tips of the index finger and thumb contacting
vertically aligned horizontal surfaces located on the top and
bottom, respectively. Sensors measured the forces applied
by the digits. The object was attached to a lightweight
robotic device that generated load forces in the horizontal
plane. A second robotic device was attached to a cuff
around the wrist and could generate a force channel that
constrained the movement of the hand to a straight line. The
experimental group of participants completed a set of
movements in the force channel followed by a second set
with the channel removed. The control group only
completed a set of movements without the channel. We
reasoned that if load forces at the fingertips are sufficient to
enable learning of object dynamics, then participants in the
experimental group should adapt their movements (and grip
force) more quickly to the load (when the channel is
removed) than naïve participants in the control group. When
participants in the experimental group transported the object
in the channel, they learned to modulate grip force in
anticipation of load force which suggests that they acquired
a component of the object's dynamics. However, when
these participants subsequently transported the object with
the channel removed, they exhibited substantial kinematic
errors that were similar to those exhibited by the participants
in the control group who did not previously experience the
object in the channel. Nevertheless, participants who
previously experienced the object in the channel continued
to effectively modulate grip force when the channel was
removed. These results indicate that in the absence of
kinematic errors, participants can learn a component of the
object's dynamics, used to control grip force, based on
kinetic errors. However, this knowledge is apparently not
accessible or usable for controlling the movement trajectory
when the channel was removed.
2-A-22 Cerebellar modulation of human associative
plasticity
Masashi Hamada1, Nagako Murase1, Anna Sadnicka1,
Joseph M Galea1, Mark J Edwards1, John C Rothwell1
1
UCL Institute of Neurology
51
Poster Sessions
Full Abstracts
Objective: To test whether the response to paired associative
stimulation (PAS) could be influenced by modifying the excitability
of cerebellum. Background; In the PAS protocol, the relative
timing of cortex and peripheral stimuli is usually between 21.5 and
25 ms. Considering the onset of the N20 response, an interval of
21.5 ms implies a relatively direct sensory input to cortex, whilst
the 25ms timing allows for more indirect sensory pathways to be
involved. We hypothesized that concurrent cerebellar transcranial
direct current stimulation (cDC, 2 mA), which can modify
cerebellar excitability (Galea et al., 2009), might modulate the
response to PAS, particularly at longer PAS intervals given that
some somatosensory input may reach motor cortex via a
cerebellar projection.
Methods: Subjects were twelve healthy volunteers. We tested
PAS at 25 ms (PAS25) simultaneously with anodal, cathodal, or
sham cDC. We also tested PAS21.5 with anodal or sham cDC.
Motor evoked potential (MEP) was measured before and after the
intervention.Results; We found that plasticity induced by PAS25
was blocked by concurrent anodal or cathodal cDC. In addition,
the effect was timing specific since plasticity induced by PAS21.5
was not blocked by anodal cDC.
Conclusions: This study provides evidence that the cerebellum
is involved in PAS-induced plasticity in a timing-specific manner.
It has been generally accepted that PAS at short intervals (21.5 or
N20 latency) and PAS25 share similar mechanisms in terms of
induction of human associative plasticity. Instead, the present
results provide evidence that PAS21.5 and PAS25 have different
characteristics with important implications for research which
uses PAS to investigate the pathophysiology of neurological
disorders, or the effects of behavioural learning.
by comparing with an unpaired t-test the mean response at
sites 0cm, 1cm, 2cm and 3cm from the centre of the grid (i.e
the contralateral hotspot) for ipsilateral and contralateral
muscles.
Results: Bilateral MEP's were seen in all 20 subjects. Of
note there was no significant difference between athletes
and controls for contralateral muscle representation. This is
important as it establishes that peripheral factors such as
muscle size do not explain observed differences. In
ipsilateral muscle the mean response at 1 cm from the
hotspot was 246%+/-21(mean +/-SEM) for athletes and
107+/-11 for controls (p=<0.001), at 2 cm 266 +/-17 for
athletes and 108 +/-7 for controls (p=<0.001) and at 3cm
219 +/-14 for athletes and 75+/-15 for controls (p=<0.001).
Discussion: We show that athletes have a cortical region
surrounding the contralateral muscle representation that is
able to produce ipsilateral responses. Athletes drive
plasticity to increase the motor cortex map of ipsilateral
muscles when bilateral axial muscle activation is important.
This knowledge provides the athlete with a basis for
alternative training strategies, and provides the neurologist
with a therapeutic target in the intact hemisphere after brain
injury.
2-A-24 Role of reuniens-medial prefrontal cortex
and reuniens-hippocampal CA1 pathways in
associative learning
Juan Carlos López Ramos1, Lyndell Eleore1, Rafael GuerraNarbona1, José María Delgado-García1
1
2-A-23 Cortical plasticity in elite athletes: Lessons for
brain injury rehabilitation
1
1
1
Damon Hoad , Lekshmi Desikan , Poppy Flanagan , Sam
Shribman1, Paul Hammond1, Richard Greenwood1, John
Rothwell1
1
Institute of Neurology
Introduction: Muscles that habitually perform a bilateral
activation receive a stronger bilateral corticospinal innervation.
Athletes perfecting control of a movement that involves muscles
acting bilaterally may alter the cortical representation of that
muscle. With training a single cortical origin may provide a
stronger bilateral control. Transcranial Magnetic Stimulation
(TMS) provides a non-invasive means of mapping the cortical
representation and characterizing corticospinal control of a
muscle. Using TMS we demonstrate a pattern of cortical
reorganization whereby ipsilateral projections from one
hemisphere are strengthened in elite athletes, having implications
both for training athletes and treating brain injury patients.
Methods: Canoe-polo players who had accumulated over 10,000
hours of training and represented their country in world-cup
competition were compared to healthy controls. EMG was
recorded from abdominal External Obliques, ipsilateral and
contralateral to the side of cortical stimulation. Background
activation at 10% of Maximum Voluntary Contraction was
produced. Neuronavigation guided the TMS coil to points on a
scalp grid 1cm apart, centred on the site of optimal stimulation
(Hotspot) for the contralateral muscle. Each point in the
surrounding cortex was then stimulated at 120% of the active
motor threshold (aMT).
Analysis: EMG was rectified and Motor Evoked Potentials (MEP)
were averaged for each grid point. MEPs were normalized to
mean+1S.D. of background activity. Normalised responses for
each grid coordinate expressed as mean %Background Activity
were pooled for the two groups. The pooled maps were analysed
52
Universidad Pablo de Olavide
We studied here the interactions between short- and longterm plastic changes taking place during the acquisition of a
classical eyeblink conditioning and following high-frequency
stimulation (HFS) of the reuniens nucleus in behaving mice.
For classical eyeblink conditioning, five-month-old animals
were implanted with two pairs of electrodes in the upper
eyelid of the left eye, one of them aimed at the supraorbital
nerve for the presentation of electrical stimuli, and the other
in the ipsilateral orbicularis oculi muscle to record its EMG
activity. Additional bipolar stimulating electrodes aimed at
the reuniens nucleus and a monopolar one aimed at
hipoacampal CA1 or at infralimbic area of the prefrontal
cortex were implanted to study the involvement of that
pathways in cognitive processes. Synaptic changes in
strength were studied at the reuniens-medial prefrontal
cortex (mPFC) and the reuniens-CA1 synapses.
Input/output curves and a paired-pulse study enabled
determining the functional capabilities of the two synapses
and the optimal intensities to be applied at the reuniens
nucleus during classical eyeblink conditioning and during
long-term potentiation (LTP) evoked by HFS of the nucleus.
Animals were conditioned using a trace paradigm, with a
tone as conditioned stimulus (CS) and an electric shock to
the trigeminal nerve as unconditioned stimulus (US). A
single pulse was presented to the reuniens nucleus to evoke
field EPSPs (fEPSPs) in the mPFC and the CA1 area during
the CS-US interval. No significant changes in synaptic
strength were observed at the reuniens-mPFC and
reuniens-CA1 synapses during the acquisition of the eyelid
conditioned responses (CRs). However, two successive
HFS sessions carried out during the first two conditioning
days significantly decreased the percentage of CRs, but
without evoking any concomitant LTP at the indicated
synaptic sites. HFS of the reuniens nucleus also prevented
the proper acquisition of an object discrimination task. A
subsequent study revealed that HFS of the reuniens
Poster Sessions
Full Abstracts
nucleus evoked a significant decrease of paired-pulse facilitation
lasting for up to three days. Thus, reuniens nucleus projections to
prefrontal cortex and hippocampal circuits could participate in the
acquisition of associative learning tasks by the modulation of
selective presynaptic mechanisms.
William Lytton1,3, Samuel A. Neymotin1, George L.
Chadderdon1, Cliff C. Kerr1,2 Joseph T. Francis1
2-A-25 Analysis in individual differences in learning
curves reveals greater variability for short time-scale
motor adaptation
Few attempts have been made to model learning of
sensory-motor control using spiking neural units. We
trained a 2-degree-of-freedom virtual arm to reach for a
target using a spiking-neuron model of motor cortex that
maps proprioceptive representations of limb position to
motor commands and undergoes learning based on
reinforcement mechanisms suggested by the dopaminergic
reward system. A 2-layer model of layer 5 motor cortex
(M1) passed motor commands to the virtual arm and
received proprioceptive position information from it. The
reinforcement algorithm trained synapses of M1 using
reward (punishment) signals based on visual perception of
decreasing (increasing) distance of the virtual hand from the
target. Output M1 units were partially driven by noise,
creating stochastic movements that were shaped to achieve
desired outcomes.
Yohsuke Miyamoto1, Maurice A Smith1
1
Harvard University
Individual differences in motor learning are widely acknowledged
but not well understood. Some of these differences can be
attributed to variations in attentional state, genetic predisposition,
or white matter anatomy. However, previous work on individual
differences has generally focused on overall learning ability,
leaving unexplored individual differences in the progression of
learning during the course of training. Here we examine the
temporal structure of inter-individual differences in the shape of
learning curves in a motor adaptation task. This may give insight
into understanding the mechanisms that underlie individual
differences in motor learning ability. We examined learning curves
from 194 subjects during exposure to a novel force field
environment. Subjects performed rapid 10cm point-to-point
reaching movements, and learned to produce forces that
counteracted a velocity-dependent curl force field with an
amplitude of 15N/(m/s). In order to identify the structure of the
variability in these learning curves, we performed principal
components analysis on the force field adaptation data. This datadriven analysis examines the covariance of learning levels, with
each trial treated as a separate dimension, and each subject
providing separate observation of each dimension. This analysis
identifies the set of orthogonal axes (the principal components)
which most efficiently explain the variance of the data. The first
principal component (PC1) represents the shape along which
individual learning curves differ most. Surprisingly, we found that
this single principal component, PC1, by itself accounted for
greater than 37% of the total variability in individual learning
curves - more than triple the amount accounted for by the next
largest principal component (PC2). This indicates the presence of
a single dominant temporal structure that characterizes a large
fraction of the individual differences in motor adaptation that we
observed. We also found this principal component to have an
interesting shape. If the main source of variability in learning
curves were signal-dependent noise the variability would echo the
shape of the learning curve itself. Instead, however, we found that
the shape of the first principal component rapidly increases to a
maximum within the first few trials, and then gradually decreases
a small amount over the remainder of the training period.
Strikingly, the shape of this principal component resembles the
fast learning process in the two-state model described by Smith et
al 2006. This study showed that two adaptive processes with
different time scales underlie motor adaptation. The slow process
responds weakly to error but retains information well, whereas the
fast process responds strongly but has poor retention. A
regression analysis reveals the shape of PC1 is remarkably well
explained by a linear combination of the hypothesized shapes of
the fast and slow processes (R^2=0.995). Notably, the we find
that the contribution of the fast process is significantly greater
than that of the slow process, (p<0.01) suggesting that individuals
differ mostly in terms of their capacity for short time-scale learning
and are more stereotyped in their capacity for long time-scale
learning.
2-A-26 Reinforcement learning of 2-joint virtual arm
reaching in motor cortex
1
SUNY Downstate Medical Center, 2School of Physics,
University of Sydney, 3Kings County Hospital
The virtual arm consisted of a shoulder joint, upper arm,
elbow joint, and forearm. The upper- and forearm were each
controlled by a pair of flexor/extensor muscles. These
muscles received rotational commands from 192 output
cells of the M1 model, while the M1 model received input
from muscle-specific groups of sensory cells, each of which
were tuned to fire over a range of muscle lengths. The M1
model had 384 excitatory and 192 inhibitory event-based
integrate-and-fire neurons, with AMPA/NMDA and GABA
synapses. Excitatory and inhibitory units were
interconnected probabilistically. Plasticity was enabled in the
feedforward connections between input and output
excitatory units. Poisson noise was added to the output
units for driving stochastic movements. The reinforcement
learning (RL) algorithm used eligibility traces for synaptic
credit/blame assignment, and a global signal (+1=reward, 1=punishment) corresponding to dopaminergic
bursting/dipping. Eligibility traces were spike-timingdependent, with pre-before-post spiking required. Reward
(punishment) was delivered when the distance between the
hand and target decreased (increased).
RL learning occurred over 100 training sessions with the
arm starting at 15 different initial positions. Each subsession consisted of 15 s of RL training from a specific
starting position. After training, the network was tested for its
ability to reach the arm to target from each starting position,
over the course of a 15 s trial. Compared to the naive
network, the network post-training was able to reach the
target from all starting positions. This was most clearly
pronounced when the arm started at a large distance from
the target. After reaching the target, the hand tended to
oscillate around the target. Learning was most effective
when recurrent connectivity in the output units was turned
off or at low levels. Best overall performance was achieved
with no recurrent connectivity and moderate maximal
weights.
Although learning typically increased average synaptic
weight gains in the input-to-output M1 connections, there
were frequent reductions in weights as well. Our model
predicts that optimal motor performance is sensitive to
perturbations in both strength and density of recurrent
connectivity within motor cortex and that therefore the wiring
of recurrent connectivity during development might be
carefully regulated.
53
Poster Sessions
Full Abstracts
Acknowledgments: research supported by DARPA grant N6600110-C-2008
2-A-27 The influence of prior experience and symbolic
cueing on human path integration
Stefan Glasauer1,2,3, Paul Maier1,2, Frederike H. Petzschner1,3
1
Institute for Clinical Neurosciences, 2Bernstein Center for
Computational Neuroscience, 3Integrated Research and
Treatment Center for Vertigo, Ludwig-Maximilian University
Munich, Germany
Actions based on current sensory input are frequently shaped by
prior experience. As we have shown, human visual path
integration behaviour in a simple homing path integration task can
be described as the optimized result of an optimal probabilistic
combination of the current sensory input and short-term prior
experience gathered over the preceding trials (Petzschner and
Glasauer, J Neurosci 31, 2011). The observed trial-to-trial
modifications in performance are explained by a rapidly adapting
experience-dependent prior that does not depend on error
feedback. Here, we investigate whether additional information
given by symbolic verbal cues can influence the path integration
behaviour and how the observed results can be explained by
probabilistic modelling.
Subjects were asked to produce and reproduce travelled
distances in a virtual environment in three different conditions. (1)
In the ‘two-range’ condition distances were drawn from two
sample ranges (first 5-13m, then 11-19 m or vice versa), which
were overlapping for two distances (11, 13m). (2) In the ‘onerange’ condition, the same distances were presented but in
randomized order. (3) In the ‘cued’ condition distances were
presented in the exact same order as in the ‘one-range’ condition,
but each trial was preceded by the verbal cueing (two cues, “The
following distance will be short” or “… will be long”). The verbal
cue was always valid and corresponded to the two ranges of
distances of the ‘two-range’ condition. No further information on
the meaning of ‘short’ and ‘long’ was provided. Subjects received
no feedback on their performance in any condition.
Our results show that subjects are able to utilize the additional
symbolic information given in the ’cued’ condition to modify their
estimate of self-displacement. In the ‘cued’ condition, even
though the order of sample distances was exactly the same as in
‘one-range’, the reproduction of overlapping distances
significantly depended on the cueing. In contrast, distance
reproduction in the ‘one-range’ condition showed a uniform
behaviour. As expected from our previous work distance
reproduction for the ‘two-range’ condition depended significantly
on the sample range.
To explain our results, we propose a model of distance estimation
by iterative Bayesian inference based on our previous work that
optimally combines 1) the current noisy sensory input, 2) a prior
expectation of the presented distance adaptively adjusted within
each trial, and 3) the discrete symbolic cues, the calibration of
which is learned over trials. We conclude that the probabilistic
modelling approach to understanding ‘cognitive’ influence, such
as prior experience or symbolic cues, on action production can
lead to simple but powerful models, which can explain a whole
range of previous psychophysical findings.
Acknowledgements: supported by the BMBF (grants IFB
01EO0901 and BCCN 01GQ0440)
2-B-33 The negative BOLD homunculus: Different
contributions of negative and positive BOLD to the
somatotopic representation in motor homunculi in
humans
Noa Zeharia1, Tamar Flash2, Amir Amedi1
1
Hebrew University of Jerusalem, 2Weizmann Institute of
Science
One of the most important attributes encoded in motor
homunculi is somatotopy - encoding movements of different
effectors by different neurons and the organization of the
representation in a gradual spatial pattern. However,
mapping the motor somatotopic representation in a noninvasive, accurate and detailed manner and across cortical
and sub-cortical homunculi in the human brain encountered
some difficulties, due e.g. to the extensive overlap between
the representations of different body parts.
In contrast to motor somatotopic mapping, fMRI retinotopic
mapping is usually achieved using a periodic design and
phase-locking analytic approaches. This form of analysis,
suitable for mapping gradually shifting representations, is
considered the classical means for defining early visual
areas. We applied this method to the motor system, using
periodic and event-related experiments incorporating
bilateral/axial movements of 20 organs from tongue to toes.
We report detailed mototopic imaging maps in various
homunculi.
In addition, we further addressed the issue of somatotopic
representation inspecting positive and negative BOLD
signals. A crucial attribute in movement encoding is the
adequate balance between suppressing unwanted muscle
activations and activating relevant ones. We studied the
different contributions of positive and negative BOLD to
movement encoding across M1 and the SMA. In addition to
the positive BOLD, significant negative BOLD was detected
in M1, but not in SMA. Its spatial pattern was neither
homotopic (i.e. negative BOLD in the same location as the
positive BOLD, only in the ipsilateral M1), random, nor
converging for various organs. Rather, it was organized
somatotopically across the entire homunculus, in an inverse
manner to the positive BOLD, thus creating a "negative
BOLD homunculus". The neuronal source of negative BOLD
is still debated; M1 provides a unique system in which
different regions receive their blood supply from different
arteries, ruling out "blood stealing" explanations. In addition,
MVPA demonstrated that positive BOLD in M1 and SMA
and negative BOLD in M1 contain somatotopic information,
enabling prediction of the moving organ from within and
outside its somatotopic location.
We conclude that fast and accurate motor somatotopic
representations can be constructed using phase-locked
analysis methods, and that negative BOLD plays a role in
somatotopic representation in M1, but not in the SMA,
perhaps due to the different roles of these two area in
movement suppression.
This study was supported by the EU FP7 TANGO.
2-B-34 Eye movement characteristics during ball
catching
Benedetta Cesqui1, Francesco Lacquaniti1, Andrea d'Avella1
1
B - Integrative Control of Movement
54
IRCSS Santa Lucia Fundation
To successfully catch a fast ball it is important to keep the
eyes on it. Tracking a moving target improves the ability to
predict its future positions and helps directing the hand to
Poster Sessions
Full Abstracts
the interception point within the available time interval. How the
CNS integrates visual information, an efferent copy of the
oculomotor command, and prior knowledge of the target
dynamics (e.g. gravitational acceleration) to control interceptive
movements is an important and open question. We have recently
reported large differences in catching hand kinematics among
individuals with similar performance level, suggesting that the
CNS might adopt different but equally successful control
strategies. Here we investigated the characteristics of eye
movements during catching of fast flying balls and their
relationship to the observed inter-individual kinematic differences.
Eight subjects with different catching success rates were
instructed to catch a ball projected from a distance of 6 m by an
actuated launching apparatus with different flight characteristics
(four flight times, ranging from 0.55 to 0.85 s and two arrival
heights). The spatial position of the ball and of several markers
placed on the subject head and body were recorded with a motion
capture system. The orientation of the eyes in the head was
recorded using a head-mounted video-oculography system. A
dedicated calibration procedure was developed to estimate gaze
direction in space using head and eye tracking data. To minimize
inaccuracies due to small movements of the eye tracker with
respect to the head, we also corrected for rotational drifts on a
trial-by-trial basis. Overall, we were able to track gaze direction
with a mean angular error of less than 1°. In all subjects, we
observed a similar sequence of eye movements starting with
catch-up saccades followed by smooth pursuit up to ~120 ms
prior to interception. The duration of the catch-up phase, the
duration of the pursuit phase, the number of the saccades, and
the amplitude of the initial saccade increased with flight time and
arrival height. However, most eye movements parameters varied
across subjects. Moreover, in each condition, individual difference
in eye movement parameters were not clearly related to the
individual performance levels suggesting that non-oculomotor
factors might underlie the observed inter-individual kinematic
differences.
2-B-35 Rapid updating of the time-course of the
visuomotor reflex gain to task demands
Michael Dimitriou1, David W Franklin1, Daniel M Wolpert1
1
University of Cambridge
Previous studies have shown that visual information about the
location of the hand is incorporated into continuous online motor
control. Specifically, perturbations of the visual location of the
hand during reaching movements induce appropriate rapid
feedback responses. We examine the time-course of the
modulation of the visuomotor feedback gain during reaching.
Subjects held a robotic manipulanum and received visual
feedback through a veridical display which obscured vision of the
hand. To measure the feedback gain we perturbed the visual
location of the hand perpendicular to the principal axis of
movement and measured the induced response. We measured
the gain at different locations within the movements while
subjects reached to either a 'near' or a 'far' target. We show that
the visuo-motor gain shows a systematic modulation over the
time-course of the reach with the gain peaking around the middle
of the movement and dropping rapidly as the target is
approached. There were differences in gain for the same location
of the hand for reaches to the near and far targets. In a second
experiment, the target suddenly jumped from the 'near' location to
the 'far' location and vice versa during the movement. We
measured the reflex gain simultaneously with the perturbations of
the target position and show that the gain was updated towards
that associated with the new position of the target. We conclude
that feedback responses to perturbations of the visual location of
the hand are determined by the distance to the target and are
flexibly adjusted to accommodate online modifications to
changes in target location.
2-B-36 Nonlinear interactions between visuomotor
responses to hand and target motion: Evidence
against the difference vector model
Sae Franklin1, Alexandra Reichenbach2, Jörn Diedrichsen2,
David W Franklin1
1
University of Cambridge, 2University College London
Goal-directed reaching movements are guided by visual and
proprioceptive feedback from the target and from the hand.
The classical view on feedback control is that the brain
extracts estimates of target and hand positions from a visual
scene, and then calculates a difference vector representing
the distance between them. This difference vector is then
used by the motor system to produce the necessary motor
commands for any required online correction. However, this
intuitive theory has never been supported by direct
experimental evidence. An alternative view would be that
these two feedback systems (responding to estimated errors
or disturbances in hand and target location) are
independent, only producing combined responses at the
level of the muscle activity. Here we directly examine the
theoretical predictions by simultaneously studying the rapid
visuomotor responses to hand and target motion, elicited by
unpredictable shifts in visual hand and target position.
Specifically, the difference vector theory predicts similar
sized responses to any combination of hand and target
motion that produces identical sized difference vectors. On
the other hand, independence of the two feedback
mechanisms would predict that the response to combined
hand and target motion would be equal to the summation of
the responses to hand and target motion separately.
Because the feedback responses to hand and target motion
exhibit a non-linear saturation with respect to the amplitude
of perturbation, these two theories make different
predictions. Subjects made reaching movements to a target
while grasping a robotic manipulandum. Visual feedback of
both the hand position (cursor) and target position was
provided in the plane of movement using a virtual reality
setup. On random trials (probe trials), the cursor, the target,
or the combination of cursor and target were laterally
displaced visually for 250 ms before returning to the actual
hand or target position. The single displacements were
either 1.5 or 3 cm while the combined displacements were 3
cm. On all probe trials, the hand trajectory was constrained
by a mechanical channel produced by the robot in order to
measure the force resulting from the corrective response.
We found that the response to the combined hand and
target motion did not agree with either the difference vector
theory or that of independent feedback systems but
produced responses in between these two predictions.
Overall, these results provide evidence against the simple
difference vector model, and suggest that the two visual
feedback systems of hand and target motion elicit partially
independent responses that are integrated in later stages of
the sensory-motor system.
2-B-37 Cortical activity differentiates automatic and
controlled processes in a speeded response
switching task
Douglas Cheyne1, Paul Ferrari1, James A Cheyne2
1
Hospital for Sick Children, 2University of Waterloo
Human action involves a combination of controlled and
automatic behavior. In choice response tasks requiring fast,
yet accurate responding, errors are thought to arise
55
Poster Sessions
Full Abstracts
because of a failure in executive control, due either to brief lapses
in attention or inadequate inhibition (or modification) of automatic
by controlled processes. In either case, temporal constraints in
rapid serial tasks may preclude timely intervention by controlled
process. Neuroimaging can help disentangle brain networks
involved in these two aspects of motor control. We combined
MEG recordings with spatial filtering techniques to examine
cortical activity during a response switching variant of the
Sustained Attention to Response Task. Subjects made a default
button press to randomly presented digits 1 to 9 (duration = 250
ms, ISI = 1150 ms) with the exception of an infrequent target digit
'3', which required switching the response finger or hand on 20%
of the trials. Twelve right-handed subjects performed both
bimanual (right to left, or left to right hand) and unimanual (right
index or right middle finger) versions of the task (1500 trials per
condition). Whole-head 151 channel MEG was recorded at 625
samples/s. Beamforming source reconstruction was used to
examine the time course of oscillatory brain activity time-locked to
movement onset and co-registered with structural MRI for group
averaging. Errors were frequent (28-30%) and error RTs were
significantly shorter (290-320 ms) than correct switch RTs (447480 ms) or non-target (default) RTs (337-373 ms) across all tasks
(p < 0. 01 for all comparisons).
We observed differences in oscillatory brain activity associated
with correct switch responses compared to errors. Theta band (48 Hz) activity was observed in the right middle frontal gyrus for
correct switches, beginning after cue onset and peaking before
movement onset in all tasks, i.e., independent of both side of
movement and direction of switching. Interestingly, this activity
significantly decreased in 'fast' compared to 'slow' switches
(shortest versus longest 1/3 RTs). Theta oscillations were also
observed on error trials bilaterally in the medial frontal and
anterior cingulate cortex, peaking 150 ms after movement onset,
likely reflecting error processing. As expected, suppression of
beta band (15-25 Hz) activity was observed in the motor cortex
preceding all movements, but began prior to cue onset, indicating
automatic motor preparation beginning immediately after the
previous response. Notably, in the bimanual task this beta
suppression was bilateral for correct trials, but lateralized
contralateral to the default hand in error trials, suggesting a failure
to prepare the (ipsilateral) switch hand in advance. These results
suggest that speeded choice response tasks may involve an
automatic, yet biased preparation for movement that predicts
errors even prior to cue presentation. In contrast, controlled
processes involved in inhibition of the prepotent default
movement and/or selection of the alternate response are reflected
by non-effector specific theta activity in right frontal brain regions,
which does not begin until after response cue processing. These
separate brain networks likely work in parallel under task
demands of fast responding, and may underlie concomitant
subjective experiences of unintended errors due to incorrect
preparation, or reduced sense of 'effort' during fast correct
responses.
Supported by NSERC (RGPIN 184018-09).
2-B-38 Spatiotemporal characteristics of muscle
patterns during ball catching
Mattia D' Andola1, Benedetta Cesqui1, Alessandro Portone1,
Francesco Lacquaniti1, Andrea d'Avella1
1
Santa Lucia Foundation
Intercepting a moving target requires accurate visuomotor control
of the effector. For rapidly moving targets, visual information must
be combined with prior knowledge of target motion characteristics
to overcome visuomotor delays and predict the right interception
position and timing. What strategy the CNS adopts to perform
such challenging computation is an open question. We have
56
recently shown that the right time and place of the collision
is not univocally specified by the CNS for interception under
loose task constraints: for a given target motion different but
equally successful solutions can be adopted by different
subjects. To gain insights on the control of interception and
the role of individual factors, we investigated the pattern of
arm muscle activity during catching of fast balls flying in
three dimensional space. We recorded and analyzed arm
kinematics and surface EMGs from 16 arm muscles in 6
subjects catching lightweight balls projected from a distance
of 6 m by an actuated launching apparatus with three mean
flight times (550, 650, 750 ms) and two mean arrival heights
(below and above the shoulder height). We identified
spatiotemporal characteristics of the muscle patterns of
individual subjects that were invariant across ball flight
conditions by decomposing the EMG waveforms, smoothed
and averaged over repetitions in the same condition, into
combinations of time-varying muscle synergies, coordinated
recruitment of groups of muscles with specific activation
profiles. In addition, the timing and amplitude coefficients of
the time-varying synergies captured the changes in the
patterns across conditions. We found that two synergies
explained on average 73% of the data variation (range 6683%). Consistently across subjects and conditions, the
timing of one synergy was aligned to the ball launch and the
timing of the other synergy was aligned to the time of ball
impact with the hand. The spatiotemporal organization of
both synergies showed similar features in all subjects, with a
strong activation of elbow and shoulder flexor muscles in the
onset-related synergy and a strong activation of elbow and
shoulder extensor muscles in the impact synergy. However,
the relative level of activation of elbow extensors and flexors
in the impact synergy varied across subjects. In particular,
subjects who stopped the hand at the impact point showed
a higher level of co-contraction than subjects who kept
moving the hand after the impact. Moreover, there were
differences in the onset of the launch synergy related to the
different individual timing of the initial response. These
results suggest that the control of interceptive movements
involves an initial component in response to the visual
stimulus and a predictive component, timed to the impact,
driving the hand to the interception point.
2-B-39 Catching something we don't see. How can
that be done?
Gianfranco Bosco1, Delle Monache2, Francesco Lacquaniti2
1
University of Rome Tor Vergata, 2University of Rome Tor
Vergata / IRCCS Santa Lucia Foundation
Prediction is a fundamental aspect of the control of fast
interceptive actions, since sensory-motor delays can be
source of spatial/temporal errors and potential instability.
Prediction of the future motion of a visual target (i.e., visual
extrapolation) is even more critical when objects in the
foreground occlude the target, making moment-to-moment
visual feedback transiently unavailable. Manifold sources of
information may contribute to this predictive process. Visual
memory about the target motion before the occlusion, for
example, is known to represent a major source of
information but there is also evidence that implicit
knowledge about invariant features of the environment, like
gravity, may contribute. In order to get further insight on how
different information may be integrated into a predictive
representation of the target trajectory for the control of the
interceptive action, we manipulated specific features of a
computer-generated visual environment representing the flyball play of the baseball game. Twenty-four subjects were
instructed to intercept the fly-ball trajectories by moving
Poster Sessions
Full Abstracts
interactively a cursor on the screen with a computer mouse and
by pressing the mouse button to indicate the interception time. In
two experimental sessions, about 30 days apart, fly-ball
trajectories were either fully visible or occluded for three possible
time intervals (750, 1000 and 1250 ms) before the ball reached
the interception point. Natural ball motion imposed by earth
gravity could be perturbed, in separate trials, with the effects of
either weightlessness (0g) or enhanced gravity (2g) at times such
that, for occluded trajectories, 500 ms of perturbed motion were
visible before ball disappearance. Moreover, separate groups of
subjects performed in reversed order the experimental sessions
with either fully visible or occluded trajectories in order to examine
the possibility that previous visual experience with the non-natural
trajectories could also contribute to the interception of the invisible
targets. These groups of subjects showed, in fact, significantly
different patterns of interceptive responses across ball laws of
motion. When intercepting occluded targets, subjects without prior
visual experience of the perturbed trajectories showed responses
very consistent with an expectation of the effects of natural
gravity. On average, these subjects estimated correctly the
responses to 1g trajectories, timed earlier and underestimated the
0g trajectories while they showed late and overestimated
responses to the 2g trajectories. Conversely, subjects with
previous visual experience of the perturbed trajectories showed
systematic anticipation and underestimation of the 0g trajectories,
but rather similar interception errors for both 1g and 2g
trajectories. This suggests that implicit knowledge of the
perturbed motion, developed during the first session with fully
visible targets, contributed significantly to the interception of
occluded targets by reducing the degree to which subjects relied
on presupposed effects of gravity on the ball motion. In
conclusion, predictive estimates for the interception of invisible
targets reflect a-priori knowledge of the visual environment based
mainly on an internal model of the effects of earth gravity,
although implicit representations of arbitrary features noncongruent with a natural setting, which may be developed through
visual experience, might also contribute.
2-B-40 A model of visuomotor coordination and
submovements in three dimensional object interception
Sang Hoon Yeo1, Martin Lesmana1, Debanga R Neog1, Dinesh K
Pai1
1
University of British Columbia
We propose a generative model of three dimensional object
interception that includes many observed features of the human
sensorimotor system. The sensory part simulates a simplified
model of active vision with uncertainty. We first simulate noisy
visual stimuli from a thrown target based on a probabilistic model
of foveal and peripheral vision. The acquired stimuli are then used
to estimate the position and velocity of the target using a
Bayesian framework with an internal model. For given estimates
of the state and error covariance of the target, the corresponding
eye movement is planned online. Our model implements both
saccades and smooth pursuit, and the decision to switch between
them. The motor part of the model implements a hand movement
controller. The hand trajectory is synthesized by combining
submovements with typical bell-shaped velocity profiles. The
physical parameters for submovements are determined based on
the current estimates of the target and the corresponding gaze
behavior. We propose that the eye and hand movements share
the same motor program, by which a submovement of the hand is
triggered in synchronization with the gaze shifts following sensory
events. Also, the direction and amplitude of each submovement
are determined by the current gaze behavior, matched to open
loop and closed loop phases of hand movement control. This
framework enables us to efficiently simulate coordinated and
realistic visuomotor behavior in object interception. We also
describe preliminary data measuring eye and body
movements during ball catching which appear to be
consistent with the model.
2-B-41 Firing pattern of spinal interneurons
mediating a variety of segmental reflex pathway in
awake, behaving monkey : A new hypothesis
GeeHee Kim1, Tomohiko Takei1, Kazuhiko Seki1
1
National institute of Neuroscience
Organization of spinal reflex pathway in vertebrate is well
established in the anaesthetized or decerebrated animals,
but there are no direct evidence showing their function
during voluntary movement. To address this issue, we
identified spinal interneurons (INs) involving segmental
reflex pathways and examined their activities in three
monkeys performing a wrist flexion and extension task. A
tungsten microelectrode was used to record the activity of
INs (C4-T1). Electromyographic activities (EMGs) were
recorded from wrist flexor and extensor muscles and nerve
cuff electrode was implanted to the deep radial nerve (DR).
The INs that responded within a segmental latency of 1ms
to the DR electrical stimulation were identified as first-order
INs (FO-INs) from DR afferent. Outputs of these INs to
muscles were identified by measuring the postspike (Psp)
effects from the result of spike-triggered averaging of EMGs.
In total, 78 INs were identified as the FO-INs from DR, and
32 INs of them showed significant Psp effects to one or
more muscles. Next, we analyzed activity of the INs
mediating each spinal reflex pathway during the task.
Significant modulation of the IN's activity was evaluated by
comparing the average firing rate during movement period
with that of rest period. Most of the FO-INs (29/32) exhibited
significant modulation during flexion and/or extension
movement. This result strongly suggests that spinal INs
mediating segmental reflex contributes to control muscle
activities during voluntary movement. Next, the input-output
pattern of FO-INs and their activity during voluntary
movement was compared in detail. Their input-output
relations were categorized into 6 patterns: A) Psp facilitation
(PspF) to extensor (n=18), B) Psp suppression (PspS) to
extensor (n=3), C) PspF to flexor (n=3), D) PspS to flexor
(n=4), E) PspF both to extensor and flexor (n=1), F) PspF to
extensor & PspS to flexor (n=3). Among these FO-INs, taskrelevant firing pattern was observed in the four classes of
INs (A,B,C, and D). For example, FO-INs with PspF to
extensor (autogenic excitation: A) showed activity
exclusively during extension trials (15/18), especially, the
activity significantly increased throughout the extension
torque compared to the FO-INs without PspF in any tested
muscles (p<0.01). This result suggests that disynaptic,
excitatory reflex pathway that mediating autogenic
facilitation of extensor muscle is involved in the
maintenance of static muscle force during voluntary
movements. FO-INs with PspS to extensor (autogenic
inhibition: B) significantly activated during flexion and
showed significant suppression of firing rate during
extension (2/3). This result suggests that disynaptic,
inhibitory reflex pathway that mediating autogenic
suppression of extensor muscle is involved in the facilitation
of agonistic and suppression of antagonistic muscle force.
Further, FO-INs with PspF to flexor (reciprocal excitation: C)
showed significant facilitation of firing rate during both
flexion and extension (2/3), and FO-INs with PspS to flexor
(reciprocal inhibition: D) also bi-directionally activated (2/4).
This result may suggest that disynaptic, reciprocal reflex
pathway is involved in the control of joint stiffness
irrespective to movement direction. In conclusion, we
57
Poster Sessions
Full Abstracts
propose that spinal reflex mediating proprioceptive input
contribute to form a pattern of muscle activity of both agonist and
antagonist in a task-relevant way.
2-B-42 Analysis of spinal interneuron circuitry in lowlevel motor control of cat forepaw
Henrik Jörntell1, Fredrik Bengtsson1, Pontus Geborek1, Anton
Spanne1
1
Lund University
The cervical enlargement of the spinal cord represents a massive
piece of neural tissue primarily devoted to the low level motor
control of the muscles of the arm and hand. In primates as well as
in felines, corticospinal effects mediated over the spinal premotor
circuitry can have a substantial role in the motor control of armhand muscles. This circuitry has the anatomical prerequisites for
managing synergy control and is directly, powerfully and
dynamically updated by peripheral sensory feedback from skin
and muscles afferents. It consists of different types of last order
interneurons and the network they form together with each other
and other indirectly connected premotor neurons. We have
performed in vivo whole cell patch clamp recordings from these
neurons in the cat cervical spinal cord in an attempt to
systematize its neuron types with respect to synaptology of
corticospinal and peripheral inputs, spike firing dynamics and
anatomical location. Based on these data we design models of
the synaptic connectivity and dynamics of the spinal premotor
circuitry. Our aim is to characterize its principles of operation
relevant for synergy control, utilization of biomechanics and object
interactions.
2-B-43 Agency assignment for automatic responses in
reaching movements
Alexandra Reichenbach1, David W Franklin2, Jörn Diedrichsen1
1
2
University College London, University of Cambridge
Both cursor and target displacements elicited rapid
corrective responses. However, we found systematic
differences in the responses to cursor and target
displacements that could not be explained by the perceptual
differences between these stimuli. In experiment 1, the
corrective motor responses indicated an equal or higher
sensitivity of the motor system to cursor displacements than
the perceptual system. With 4 distractors, the motor
responses clearly differentiated between cursor and
distractor displacements, while perceptual judgements were
at chance. In contrast, perceptual sensitivity to target
displacements was always higher than motor sensitivity. In
experiment 2, the perceptual judgements of cursor
displacements improved with additional movement-related
information, resulting in similar performance in perceptual
and movement tasks. We found no such benefits for target
displacements. Furthermore, only the motor system showed
clear erroneous responses to target distractor
displacements, but not to cursor distractor displacements.
In sum, the tracking mechanisms for hand and target exhibit
distinct characteristics when challenged with distracting
visual events. This suggests that the filtering is achieved by
two different mechanisms. We suggest that hand tracking is
mediated by an assignment mechanisms, which binds
together the motor plan and its visual consequences. This
mechanism provides accurate and interference resistant
motor performance, and improves perceptual judgements.
In contrast, the mechanism that tracks the movement goal is
independent of movement-related processes.
2-B-44 Cervical kinematics and neuromuscular
activity of whiplash patients in the early stage and 6
months after trauma
Pierre-Paul Vidal1, Jennyfer Lecompte1, Baptiste Sandoz2,
Sophie Bancho3, Danping Wang1, Sebastien Laporte2
1
Reaching in a crowded visual scene constitutes a difficult
challenge for the visuo-motor system. To achieve accurate online
control, the visuo-motor system has to track visual information
from two sources: the target of the reach and the position of the
hands. The visuo-motor system reacts rapidly and appropriately
to changes in target or hand position, demonstrating that it
accomplishes both tasks efficiently. In everyday situations with a
rich visual scene, information arising from distracting, irrelevant
objects has to be filtered out, increasing the demands on tracking.
We hypothesise that tracking of hand and target is achieved by
two separate mechanisms. To investigate this, we examined how
well human participants can achieve their reaching goal in a
crowded visual scene despite changes in target or hand position.
To control for perceptual differences between the tasks
(stationary targets vs. moving cursors), we added a perceptual
task with identical visual stimuli.
Participants performed bimanual reaching movements to two
visual targets while grasping robotic manipulanda. Two cursors
rendered on a screen above the hands indicated the positions of
the hands. For each hand, we presented 1-4 distracting visual
objects, resembling either the targets or the visual cursors. During
the reach, either one of the movement relevant objects (cursor or
target) or one of the distractors was displaced laterally while the
involuntary corrective response was measured. In the control
task, participants indicated which of the objects had been
displaced. In experiment 1, the perceptual and reaching tasks
were performed separately. In experiment 2, participants did the
perceptual task directly after each reaching trial, such that one
could use the efferent or afferent information from the movement
for the perceptual judgement.
58
CNRS, 2Arts et Metiers ParisTech, 3Institut pour la
Recherche sur la Moelle épinière et l’Encéphale
Whiplash injury and subsequent disorders have been
extensively investigated in the literature. Even if most
individuals recover within a few weeks of injury, a significant
proportion (14-42%) will develop persistent ongoing pain up
to two years post injury (Sterling 2004). The processes that
may underlie whiplash-associated disorders (WAD), the
heterogeneous of pain and also the maintenance of
symptoms in those who do not recover are not well
understood. Chronic WAD displayed mostly reduced ranges
of motion (ROM) and larger head repositioning error (Feipel
et al. 1999), as well as altered cervical motion patterns
during active movement compared to asymptomatic
subjects (Feipel et al. 1999, Woodhouse et al. 2010).
Recent data has also demonstrated structural and
neuromechanical muscle changes in patients with chronic
WAD, which could be associated, in some way, with the
development of chronic pain following whiplash injury.
Indeed, Elliot et al. (2006) found muscle fatty infiltrates on
magnetic resonance imaging in the neck extensor muscles
of patients with chronic WAD. In addition, Falla et al. (2004)
have shown neuromuscular alteration of the superficial neck
muscles under low load and delayed onset during shoulder
flexion. So, it seems that neuromuscular evaluation
combined with classical mechanical tests could be relevant
prognostic indicators for WAD and transition to chronicity.
Therefore, this study was designed to improve kinematic,
proprioceptive, postural response outcomes and associated
muscle efficiency of the head neck segment in WAD at two
stages of recovery compared to healthy control. Eight
patients (age 20-50 years) with acute WAD were followed
Poster Sessions
Full Abstracts
and assessed from within 8-to-15 days of injury and 6 months
post injury. They were recruited via local hospital emergency
departments as they all complaining about neck pain in the first
48 hours after the collision, and classified as grade I or II
according to the Quebec Task Force (Sterling 2006). Eight
healthy subjects, physically active, with no concern of specific
neck training or history of whiplash injury or neck/back pain, were
also recruited. Full body kinematics was recorded with an active
3D motion analysis system (Coda-motion). Surface EMG (Delsys)
of sternocleidomastoid, paraspinal, anterior deltoid and trapezius
muscles was assessed bilaterally. Cervical principal and coupled
movements as well as muscle EMG patterns were measured in
flexion/extension, axial rotation and side bending, as well as
repositioning in neutral position after axial rotation and extension.
Furthermore, frequency analyses of the head during simple
trajectory tasks and jerk indexes were quantified. Results are
being processed; differences are expected in particular for
muscles strategies between each group. Elliott et al, Spine,
31:847-855, 2006 Falla et al, Man Ther, 9:125-33, 2004 Feipel et
al, Int Orthop, 23:205-9, 1999 Sterling, Man Ther, 9:60-70, 2004
Woodhouse et al, Exp Brain Res, 201:261-70, 2010
2-B-45 Why does picture naming take longer than word
reading? Contribution of motor processes
Nicole Malfait1, Stéphanie Ries2, Thierry Legou1, Boris Burle1, F.Xavier Alario1
1
CNRS & Aix-Marseille Université, 2University of California at
Berkeley
Since the 19th Century, it has been known that response
latencies are longer for naming pictures than for reading words.
While several interpretations have been proposed, a common
general assumption is that this difference stems from cognitive
word-selection processes and not from motor articulatory
processes. Here we show that, contrary to this widely held view,
articulatory processes are also affected by the task performed. To
demonstrate this we used a procedure that has not been used
before in research on language processing: response-latency
fractionating. Along with vocal onsets, we recorded the
electromyographic (EMG) activity of facial muscles while
participants named pictures or read words aloud. On the bases of
these measures, we were able to fractionate the verbal response
latencies into two types of time intervals: pre-motor times (from
stimulus presentation to EMG onset), mostly reflecting cognitive
processes, and motor times (from EMG onset to vocal onset),
related to motor execution processes. We show that pre-motor
and motor times are both longer in picture naming than in
reading, although articulation is already initiated in the latter
measure. Future studies based on this new approach should
bring valuable clues for a better understanding of the relation
between cognitive and motor processes involved in speech
production.
2-B-46 Training at different movement speeds: Transfer
to different speeds or directions of movement
Ulrike Hammerbeck1, Nada Yousif2, Joern Diedrichsen2, John C
Rothwell1
1
Institute of Neurology, UCL, 2Institute of Cognitive Neuroscience,
UCL
Introduction: Rehabilitation after stroke focusses on relearning
movement skills and its transfer to related tasks. In a clinical
setting this usually involves intensive, task-specific practice,
which is aimed at improving the accuracy of movement. Speed of
movement is rarely emphasised in rehabilitation even though
there is a well-recognised trade-off between movement accuracy
and speed. In this study we investigated in healthy adults the
effect of training at either fast or slow movement speed, on i)
endpoint accuracy at the trained speed, ii) transfer
characteristics to different movement speeds and iii) transfer
in space to an untrained direction of movement.
Methods: Eighteen healthy adults (mean age: 30.4 /-8.30
SD years, 7 males) attended for 5 consecutive days
performing horizontal arm reaching movements in a robotic
manipulandum. The starting point of the reach was ~15cm
anterior to the sternum in the midline; there were 2 possible
targets 25cm from the start. One target was immediately
straight ahead (0 deg) the other was at 45 deg clockwise.
Points from 1-5 were given for landing within circles of
decreasing (5 to 1 cm) diameter centred on the target.
Visual feedback of the cursor location was removed
immediately following the onset of movement and was
restored at termination. On day 1 and day 5 we plotted the
speed-accuracy trade-off (SAT) by establishing endpoint
accuracy for 4 enforced movement times (MT: 300, 500,
700, 900ms) at 0 and 45 degrees. On days 2, 3 and 4,
participants were randomly assigned to either the fast
(300ms) or slow (900ms) training group during which they
performed 630 reaches per day attempting to improve their
endpoint accuracy to the 0 deg target.
Results: After training both groups showed improvements in
accuracy for most speed conditions. The 300ms group
showed the largest gains in endpoint accuracy for the speed
that they trained on. When the error was partitioned into
perpendicular (directional) error and parallel (distance) error
both groups improved in their perpendicular accuracy mostly
at the trained MT and this transferred to all MTs. The
parallel accuracy also demonstrated training specific
improvements but the fast group additionally showed a
significantly greater improvement at 300ms (p=0.005 and
post-hoc t-test p=0.034) which accounted for the effects on
overall accuracy. Further analyses suggested that much of
this improvement in parallel error in the fast training group
was due to reduced trial-to-trial variability in the maximal
movement speed. The accuracy gains, due to this reduction
in speed variability, generalized well to the 45 deg target,
while the decrease in the perpendicular component of the
endpoint error was more specific to the trained training
direction.
Conclusion: Training at different movement speeds leads
to differential improvement in different subcomponents of
the task, which only partially generalizes to new movement
speeds. Training at fast training shows better transfer
characteristics than slow training, to new movement
direction.
2-B-47 Role of the rostral medial prefrontal cortex
during the associative learning in behaving rabbits
Rocio Leal Campanario1, José María Delgado García1,
Agnès Gruart1
1
Pablo de Olavide University
We have studied the role of the rostral medial prefrontal
cortex (mPFC) in the classical conditioning of eyelid
responses in alert behaving rabbits. The rostral mPFC was
identified by its afferent projections from the medial half of
the thalamic medio-dorsal nuclear complex, and by the firing
rate synchronization of mPFC neurons evoked by the
stimulation of this thalamic nucleus. Classical conditioning
consisted of a delay paradigm using a 370-ms tone as the
conditioned stimulus (CS) and a 100-ms air puff directed to
the left cornea as the unconditioned stimulus (US). During
classical eyeblink conditioning sessions, the firing rate of
recorded single unit activity of mPFC neurons increases
during the CS-US intervals in simultaneity with the presence
59
Poster Sessions
Full Abstracts
of CRs. Electrical train stimulation of the contralateral rostral
mPFC produced a significant inhibition of air puff-evoked blinks.
The same train stimulation of the rostral mPFC presented at the
CS-US interval for 10 successive conditioning sessions
significantly reduced the generation of conditioned responses
(CRs) compared with values reached by control animals.
Interestingly, the percentage of CRs reached almost control
values when train stimulation of the rostral mPFC was removed
from the 5th conditioning session on. The electrical stimulation of
the rostral mPFC in well-conditioned animals decreased the
percentage of CRs. The stimulation of the rostral mPFC also
modified the kinematics (latency, amplitude, and velocity) of
evoked CRs. In contrast, lidocaine injections in the mPFC during
the 2nd conditioning session increased the expected rate of CRs,
and modify the kinematics of both reflex and CRs. Thus, local
administration of lidocaine produced opposite effects to that
evoked by mPFC stimulation. In conclussion, the rostral mPFC
seems to be a potent inhibitor of reflex and learned motor
responses. In addition, these results seem to point out the need
of the right activation of the mPFC to execute the acquired motor
response in an appropriate and timed way.
2-B-48 Anosmin-1-over-expression on in vivo
hippocampal long-term potentiation and postnatal
neurogenesis
V. Murcia-Belmonte1, D. García-González1, P.F. Esteban1, A.
Gruart2, J.M. Delgado-García2, F. de Castro1
1
Grupo de Neurobiología del Desarrollo-GNDe, Hospital Nacional
de Parapléjicos, 2Division of Neurosciences, Pablo de Olavide
University
Anosmin-1 is an extracellular matrix glycoprotein encoded by the
KAL1 gene, which is responsible for the X-linked form of
Kallmann syndrome in humans. During CNS development,
anosmin-1 participates in the migration of immortalised GnRH
neurons, neuronal precursors, as well as of oligodendrocyte
precursors. It also has a role in axon guidance, neurite outgrowth
and the formation of axon collaterals from neurons of the olfactory
system and proyection neurons of the cerebellum. Although to
date, the mechanism of action of anosmin-1 it is not completely
understood, it has been proposed that anosmin-1 could regulate
the signalling of FGF2/FGFR1 and other receptors and ECM
molecules.
We have generated a transgenic mouse line that over-expresses
anosmin-1 in which the histological examination of adult did not
reveal seeming neuroanatomical alterations compared to control
mice. So far, we have studied the effect of anosmin-1 overexpression in adult neurogenesis and the distribution of several
neuronal markers in the hippocampus: although no significant
differences in BrdU incorporation were observed, homozigous
anosmin-1-over-expressing mice showed a significant decrease
in the number of doublecortin+-neurons in the hippocampal
dentate gyrus, and significantly more parvabalmin+-neurons in
the CA3 layer.
We monitored the physiological properties of the CA3-CA1
synapse in adult mice in vivo under specific experimental
conditions, which included input/output curves, paired-pulse
facilitation, and electrically-evoked LTP. Field excitatory
postsynaptic potential (fEPSPs) evoked at the CA1 area by
stimulation of Schaffer collaterals in anosmin-1 transgenic mice
for input/output curves and the paired-pulse test presented values
similar to those reached by controls. Induction of long-term
potentiation (LTP) in alert-behaving mice showed that anosmin-1
over-expression prevented evoking an LTP response. Altogether,
our data suggest that anosmin-1 is involved in synaptic plasticity
in the adult mice.
60
This research was supported by grants from the Spanish
Ministerio de Economía y Competitividad-MINECO
(SAF2009-07842), Gobierno de Castilla-La Mancha
(PI2009/29; MOV-2007-JI/19 and MOV-2011-JI/11) Spain,
and ‘Fundación Eugenio Rodríguez Pascual’ to FdC and by
MINECO (BFU2008-00899, BFU2008-03390) and Junta de
Andalucía (BIO-122, CVI-02487 and P07-CVI-02686 )
grants to JMDG and AGM. VMB and DGG are hired
currently under SAF2009-07842, and FdCS and PFE by
SESCAM.
D - Disorders of Motor Control
2-D-81 Clinical assessment tool development:
Measuring cognitive-motor integration in healthy
aging and early Alzheimer's disease
Kara Hawkins1, Jeya Thayaparan2, Adriana Bida3, Lauren E
Sergio1
1
York University, 2York Central Hospital, 3Southlake
Regional Health Centre
The central hypothesis guiding our research is that different
types of visuomotor compatibility are processed in separate,
but overlapping, parietofrontal networks, and that these
separate networks are affected differently in healthy aging
versus disease. Generally when reaching for an object in
the environment the visual stimulus and its required motor
action are in alignment. However, the evolution of the
capacity for tool-use in primates has resulted in situations
where the correspondence between vision and action is not
direct. A common example is the use of a computer mouse
to move a cursor on a monitor, which involves a decoupling,
or dissociation, between the spatial location of the visual
target and the spatial location of the movement goal. Such
learned sensory- and cognitive-motor associations underlie
much of our everyday activities (including driving), yet the
basic cortical mechanisms responsible for these behaviours
remain unknown. Following damage to the cerebral cortex
(e.g. neurodegeneration, stroke, TBI), these complex
visuomotor transformations may become impaired, and the
pattern of impairment may provide insight into the
underlying neural mechanisms.
While the ability to directly interact with objects does not
appear to be impaired in early Alzheimer's disease (AD)
relative to healthy aging, performance decrements are
observed when early AD patients perform motor control
tasks under conditions in which direct visual feedback is not
provided¹. We propose that measuring visuomotor
integration under conditions that place demands on visualspatial and cognitive-motor processing may provide an
effective behavioural means for the early detection of
underlying Alzheimer's-type neuropathology. Currently, we
are evaluating the validity, practicality, and, over time,
predictability of using such a measure in clinical populations.
To this end, patients clinically suspected of amnestic mild
cognitive impairment (aMCI) or early AD are tested on six
visuomotor transformation tasks: vertical (direct visual
feedback), vertical memory, vertical rotated cursor
feedback, horizontal, horizontal memory, and horizontal
rotated cursor feedback, using a dual touch-screen tablet.
Preliminary results have revealed that aMCI patients and,
unexpectedly, participants with a maternal family history of
AD exhibit fragmented velocity profiles and trajectory
deviations during the horizontal tasks, resulting in larger
endpoint errors, path lengths, and total movement times
Poster Sessions
Full Abstracts
when compared with healthy controls. aMCI patients also
produced more 'direction reversal' and 'leave too early' errors in
conditions with rotated cursor feedback and memory delays,
respectively. These impairments may reflect early neuropathology
disrupting the intricate reciprocal communication between parietal
and frontal brain areas required to successfully prepare, execute,
and update complex reaching behaviours. Recent studies using
diffusion tensor imaging (DTI) in MCI and AD have documented
disruption to the white matter tracts forming these parietal-frontal
connections², as well as projections from hippocampal regions to
parietal association areas³. Future DTI work in our lab will
investigate the neural correlates of observed decrements in
performance on our cognitive-motor integration tasks in preclinical
AD populations. 1. Ghilardi et al. (2000) Brain Res 876: 112-123,
2. Bosch et al. (2012) Neurobiol 3. Villain et al. (2008) J Neurosci
28: 6174-6181
2-D-82 Progressive resistance exercise improves
bradykinesia and muscle activation patterns in
Parkinson's disease
Fabian David1, Julie A Robichaud1, Sue E Leurgans2, David E
Vaillancourt3, Cynthia Poon1, Wendy M Kohrt4, Cynthia L
Comella2, Daniel M Corcos1
1
Universtiy of Illinois at Chicago, 2Rush University Medical Center,
University of Florida, 4University of Colorado School of Medicine
3
Reduced movement speed and impaired EMG activation patterns
are consistent findings in patients with Parkinson's disease (PD).
Previously, we demonstrated that antiparkinsonian medication
and deep brain stimulation improve elbow movement speed and
EMG activation patterns. In this poster we report on the efficacy
of progressive resistance exercise (PRE) to improve elbow
movement speed and EMG activation patterns in PD. We tested
the hypothesis that engaging in 24 months of PRE in PD will
provide significant improvement in elbow movement speed and
muscle activation patterns when compared to Fitness Counts
(FC), a non-progressive exercise program recommended by the
National Parkinson's Disease Foundation. The basis for this
hypothesis is that a progressive/dynamic exercise intervention
would be better than a non-progressive/static exercise
intervention. We conducted a randomized controlled trial
comparing the effect of PRE with FC on elbow movement speed
and muscle activation patterns. Patients with PD were matched
on sex and off-medication UPDRS-III score and randomized to
the PRE or FC group. Patients were assessed by raters who were
blinded to group assignment. Both exercise interventions were
supervised and patients exercised between 60 to 90 minutes, 2
days/week for 24 months. The 24 patients in the PRE group
performed a progressive weight lifting program. The 24 patients in
the FC group participated in a non-progressive flexibility, balance
and strengthening program. At baseline, 6, 12, 18, and 24 months
of the intervention, elbow kinematics and EMG activity were
recorded while non-medicated patients performed point-to-point
movements at maximal speed on a light-weight instrumented
manipulandum. Fifty-eight percent of the subjects were male,
average age was 58.9 years, and average disease duration was
6.5 years. Forty-eight patients completed the 6-month evaluation
and 38 patients completed the 24-month evaluations. The group
by time interaction was significant for the following variables:
Elbow peak velocity (p = 0.02), magnitude of the first agonist
EMG normalized to burst duration (p = 0.013), and number of
agonist bursts prior to peak velocity (p = 0.003). Post-hoc
analysis at the study end-point of 24 months revealed the
following: First, the PRE group increased peak velocity by 30o/s
more than the FC group (p < 0.0001); second, the PRE group
increased the magnitude of the first agonist EMG normalized to
burst duration more than the FC group (p = 0.009); and third, the
PRE group decreased the number of agonist bursts prior to peak
velocity more than the FC group (p = 0.001). In conclusion,
PRE brought about significant gains in upper limb
movement velocity and therefore had a significant benefit on
bradykinesia. In addition, PRE significantly increased the
magnitude of the first agonist EMG normalized to burst
duration and reduced the number of agonist bursts. These
findings indicate that PRE may modify the cortical output to
skeletal muscle in PD.
2-D-83 Nonlinear summation of evoked forces and
EMG using intraspinal microstimulation trains in the
macaque monkey
Jonas Zimmermann1, Andrew Jackson1
1
Newcastle University
Cervical intraspinal microstimulation (cISMS) is a potential
approach for restoring upper-limb function in patients
paralysed by stroke or spinal cord injury. In a recent study,
we controlled functional arm and hand movements by
delivering cISMS trains to individual electrodes in nonhuman
primates. Here we investigate the effect of delivering
stimulus trains at two sites simultaneously. We were
interested in the extent to which spatio-temporal interactions
mediated by intraspinal circuits would produce a motor
output that differed from the linear combination of individual
cISMS responses.
Four female macaque monkeys were terminally
anesthetized. Following a cervical laminectomy, we inserted
tungsten microwires (50 μm diameter) and floating
microelectrode arrays (MicroProbes) to depths of 3-5 mm
into the cervical enlargement. Pairs of electrodes were
stimulated with interleaved trains (0.5 s long, 50 Hz per
channel) delivered at different relative phase. Stimulation
currents were set to evoke a visible motor response and
were typically in the range of 20-200 μA. We recorded
electromyogram (EMG) responses from 12-16 muscles of
the arm and hand, and isometric force trajectories with a 6axis force/torque transducer attached to the hand.
We collected data from 68 pairs of electrodes. We compiled
average force trajectories over the last 250 ms of the
stimulation period. EMG recordings were rectified and
integrated over the whole stimulation period after baseline
removal. We compared forces and EMG generated by
interleaved trains against a prediction generated under the
null hypothesis of linear summation of the responses to
stimulation of individual electrodes. 27 (40%) pairs showed
no supra-linear interaction in any recorded muscle. For the
remaining pairs, supra-linear summation was seen in 6
muscles on average. Sub-linear summation was found less
frequently: 31 (45%) pairs showed no sub-linear effects,
while the remaining pairs showed sub-linear effects in 3
muscles on average.
Non-linear interactions often occurred only for particular
relative phase shifts between stimulus trains. The proportion
of muscles showing supra-linear summation was largest for
0, 2.5, and 5 ms phase shifts (19%, 16%, and 14%
respectively). Sub-linear summation was most prevalent for
opposite phase shifts (-5, -10, -2.5 ms; 9%, 8%, and 8%,
respectively).
Finally, we investigated whether these non-linear muscle
responses had an effect on the evoked force trajectories. In
15 (22%) pairs there was a significant change of direction
for at least one phase shift, 14 (20%) showed changes in
direction and magnitude and 6 (10%) showed changes in
force magnitude only. The remaining 33 (50%) of pairs did
61
Poster Sessions
Full Abstracts
not show any deviation from the force trajectories predicted by
linear summation.
In summary, spatio-temporal interactions between the neural
elements activated by cISMS produce a range of supra- and sublinear motor responses. This is inconsistent with the notion of
independent spinal cord modules activated by individual
electrodes. cISMS control strategies will need to reflect the
distributed nature of stimulus-evoked activity, but careful timing of
stimuli delivered at multiple sites may extend the repertoire of
possible upper-limb movements that can be restored by spinal
cord stimulation.
2-D-84 Sham-controlled study of transcranial direct
current stimulation (TDCS) for dystonia in children
Scott Young1, Matteo Bertucco1, Terence D Sanger1
1
University of Southern California
Childhood dystonia is a movement disorder in which excess
muscle activity interferes with voluntary movement. Most
treatments have limited success, so there is a need to investigate
alternate non-invasive approaches for the control of dystonia in
children. Research has shown that dystonia is associated with
increased excitability of the motor cortex. This suggests the
possibility of reducing dystonic symptoms through a reduction in
cortical excitability. In a previous study, our laboratory found
evidence that cathodal transcranial direct current stimulation
(TDCS) might reduce overflow muscle activity in a subset of
participants. We have completed the present sham-controlled
study to verify those earlier findings. Participants were stimulated
with sham or cathodal TDCS of primary motor cortex. To
determine the effects of stimulation, we compared measures of
muscle control and stretch reflex prior to stimulation with the
same measures following stimulation. We investigated control of
intrinsic hand muscles by measuring tracking error and overflow
in an EMG tracking task. We measured stretch reflex activity in
response to a mechanical perturbation of the first dorsal
interosseous muscle. We tested 15 children (ages 4 to 18) with
various causes of dystonia. Initial results from 8 participants have
shown a significant decrease in overflow activity, but no
significant difference between sham and real stimulation. There
was no significant effect of stimulation on tracking error. We will
present further analysis of these results as well as analysis of the
effects of TDCS on partipants' stretch reflex. This is one of the
first studies on TDCS for children with dystonia. Further studies
will focus on increasing the effect size through stimulation on
multiple consecutive days. While TDCS still appears to hold
promise for reduction of dystonic symptoms in some cases, the
findings from this controlled study do not support a significant
effect.
2-D-85 Vibro-tactile biofeedback for neuromuscular
rehabilitation in children with dystonia
Francesca Lunardini1, Serena Maggioni1, Claudia Casellato1,
Matteo Bertucco2, Alessandra Laura Giulia Pedrocchi1, Terence D
Sanger2
1
Politecnico di Milano, 2University of Southern California
Dystonia is a movement disorder in which involuntary sustained
or intermittent muscle contractions cause twisting and repetitive
movements, abnormal postures or both. Recently, some studies
support the possibility that some symptoms of dystonia may result
from a lack of sufficient sensory feedback about motor actions.
Biofeedback is a method that could improve patients' movement
performances by providing additional sensory information about a
particular physiological process. In this study dystonic children
receive vibro-tactile biofeedback proportional to the activation of a
target muscle, to test whether an augmented sensory feedback
62
could allow patients to be more aware of their muscular
activation and thus to more properly control the muscle
contraction. To investigate the effectiveness of this
technique, dystonic and healthy children are asked to
perform an eight-figure writing and a back and forth tracking
task using a stylus pen. They are required to do the same
task first without and then with biofeedback. During
movement execution kinematic data are recorded through a
magnetic motion-capture system by placing eight sensors
on the bone landmarks of the moving arm. In addition, the
electrical muscle activity is recorded from eight upper-limb
muscles. The 2-D coordinates of the stylus pen tip on a
touch-screen tablet are used to measure the end-effector
kinematics. For the session with the biofeedback a vibrating
motor is attached to a target muscle, assessed individually
for each patient. This integrated set-up has been developed
and tested. The data analysis aims to quantify the
movement behavior, outlining any differences within-subject
(with and without biofeedback) and between-groups (healthy
and dystonic). Kinematics and EMGs are processed both in
time and in frequency domains. Preliminary results on
healthy subjects have shown repeatable frequency aspects
in joint kinematics (the vertical and the horizontal frequency
components are in an exact ratio of two). These rhythmic
properties allow for a kinematic-EMG coupled analysis,
which associates each muscle activity to one or both the
kinematic components, thus defining functional motor
synergies. The results of this study will be useful for testing
the effects of short-term vibro-tactile biofeedback on the
kinematics and muscular parameters. A promising outcome
could be the starting point for a prolonged use of this
technique during daily-life activities.
2-D-86 Kinematics analysis of constrained reaching
movements in children with dystonia
Matteo Bertucco1, An Chu1, Terence D Sanger1
1
University of Southern California
Childhood dystonia is defined as a movement disorder in
which involuntary sustained or intermittent muscle
contractions cause twisting and repetitive movements,
abnormal postures, or both, where bradykinesia and
hyperkinesia jointly or separately occur. The clinical
consequences of dystonia often lead to severe
communication impairments to interact with the outside
world, and the cognitive and social development may,
therefore, be secondarily affected. Nowadays, it is a
common practice that these children communicate through
the use of assistance devices. These communication
devices are characterized by physically constrained features
such as the number, size and location of the buttons on the
screen and rarely taken in consideration the perceived cost
of hitting the unwanted button. Thus, as these children
become dependent upon their sensory-motor abilities to
interact with touch screen devices, through deficits that may
hamper the skilled use of their upper limbs, their social
interaction abilities may be further significantly limited. The
aim of the study is to quantify the influence of bradikinesia
and hyperkinesia on assisted communication devices.
Healthy and dystonic children are asked to point at a touchscreen with different targets widths and spacing between
the correct and erroneous target. The kinematic data are
recorded by using a magnetic tracking device and eight
sensors are attached at the bone landmarks of the reaching
arm joints. As we expected, the preliminary data have
shown significantly slower movements, higher variability and
lower optimal trade-off between speed and accuracy of the
pointing movement in children with dystonia. In addition we
Poster Sessions
Full Abstracts
will discuss the results of the frequency domain and principal
components analysis of the movement trajectories to distinguish
the motor synergies that characterize children with dystonia. The
results will be used to optimize the choices made in device
features and user interface design in order to maximize the
augmented communication rate with assisted devices based on
specific motor impairments in childhood dystonia.
2-D-87 Development of a new brain ischemia model to
induce motor deficit in the non-human primate: Anterior
choroidal artery occlusion
1
Sandra Milena Puentes Martinez , Kazuhiko Seki
1
1
National Institute of Neurosciences
Non-human primate models of stroke can offer to brain ischemia
research several advantages in comparison with less evolved
animals models because of the similar ergonomics and hand
dexterity which put them closer to the human condition. In the
marmoset, a model using the occlusion of the middle cerebral
artery (MCAO) is available. Although this procedure can induce
motor impairment, the compromised brain volume is very wide
affecting additional brain functions hampering the evaluation of
the pure motor damage. In contrast, lacunar infarctions affecting
the internal capsule are generated often by the obstruction of a
single artery: the anterior choroidal artery (AChA); this lesion
induces in most cases motor dysfunctions without cognitive
impairment in human patients (Likitjaroen et al. J Neurol. 2012).
The purpose of our investigation is to develop a brain ischemia
model for marmosets by the direct occlusion and transection of
the proximal AChA, and to compare the anatomicaly
compromised zones and the neurological outcome with the
already developed MCAO model. Adult marmosets were
distributed in 3 groups to underwent MCAO (n=2), anterior
choroidal artery occlusion AChAO (n=2) or sham operation (n=2).
Using a frontal approach, the first portion of the MCA or the AChA
were transected. After surgery animals were hand-fed and nursed
until required. From day 1 to 10 days after surgery, animals were
assessed by the Overt abnormal signs scale which is used to
assess the neurological outcome of marmosets. We evidenced
that the recovery of animals after AChAO was faster than the
MCAO; moreover the neurological score (Virley et al. JCBFM.
2004) which evaluate the motor statement of upper and lower
limbs of marmosets showed an important neurological deficit for
both groups after surgery, which was recovered briefly, rising
similar scores at day 10th. In order to evaluate the anatomically
compromised area, MRI analysis was performed at day 4th after
surgery. For MCAO group, we found a wide compromised area
including the cortex, basal ganglia and part of the internal
capsule. A complete transection of the left MCA was confirmed by
the AngioMR. For the AChAO group, we found a lesion
compromising the internal capsule and partially the basal ganglia.
The AngioMR showed that the MCA and the internal carotid
artery were intact. Animals were sacrificed at day 11 and
histology performed. Nuclear markers (Nissl) and Myelin staining
were performed in serial sets. For MCAO a wide area
compromising the cortex, basal ganglia and internal capsule was
found, with diffuse infiltrating cells on the affected parenchyma.
Myelin staining also showed degeneration of the internal capsule.
The AChAO showed a focal injury corresponding with the MRI
findings, sectioning completely the internal capsule, with dense
cell infiltration and compromising briefly the basal ganglia. In
conclusion, we succeed to develop an effective technique to
induce a lacunar infarction in the marmoset internal capsule.
Comparing both procedures, the AChAO becomes a more
accurate technique to induce motor deficit since it transected the
internal capsule without damage broadly the surrounding
structures. Detailed behavioral assessment is required to
evaluate the differences in the motor function and cognition
between the two techniques. This technique also will be
used in the macaque monkey to evaluate in detail the hand
dexterity impairment after internal capsule infarction.
2-D-88 Spasticity emerges when abnormal firing
thresholds are introduced into the spine neuronal
network emulated on hardware
C. Minos Niu1, Sirish K Nandyala1, Won Joon Sohn1,
Terence D Sanger1
1
University of Southern California
Our overarching goal is to reveal the organizing
mechanisms underlying movement disorders believed to
cause hypertonia in childhood. In this study we focus on the
disorder of spasticity, defined in clinic as increases in joint
resistance to externally imposed movement with a velocitydependent threshold. This characteristic suggests that
spasticity is caused by a low firing threshold of the
motoneurons that are synaptically connected to the dynamic
spindle fibers (primary afferent, group Ia). As such, it makes
these motoneurons more likely to fire and therefore provide
more resistance due to imposed joint velocity. In order to
emulate this possibility, we previously developed a scalable,
general-purpose neural emulation platform by leveraging the
recently available Field Programmable Gate Array (FPGA)
technology. The platform allowed us to emulate -- in 365x
real-time -- a motor nervous system consisting of 1,024
motoneurons, 1,024 sensory afferent neurons (both using
Izhikevich models), proprioceptors (Loeb's model) and Hilltype muscles. Here in this study we indiscriminately reduced
the firing threshold of the motoneuron pool innervated by Ia
afferent spindle fibers. We will demonstrate that a symptom
compatible with the above definition of spasticity will
naturally and robustly emerge. Our next goal is to
investigate whether the severity of spasticity in individual
patients can be differentiated solely based on a threshold
mechanism. This will allow us to investigate changes in
spasticity over time during development in children. Our
results are consistent with current models of the
development of spasticity and show that these models are
sufficient to yield symptoms consistent with clinical findings.
2-D-89 Kinematic investigations of reaching
slowness in hemiparetic stroke patients
Agnes Roby-Brami1, Laurence Mandon2, Johanna V
Robertson2, Raphael Zory2, Djamel Bensmail2
1
University Pierre et Marie Curie, CNRS, 2Hôpital Raymond
Poincaré
It is well-known that reaching movements made by
hemiparetic stroke patients are slow and irregular. However,
the relationship between the impairments following stroke
(weakness, spasticity, disruption of coordination) and
slowness of movement remains poorly understood. To
investigate the factors limiting velocity in hemiparetic
patients, we recorded aiming movements directed to two
targets (at respectively 60% and 90% of anatomical arm
length) in 6 hemiparetic patients and examined the ability of
the patients to increase aiming velocity under the instruction
to move as fast as possible. Two experimental conditions
were compared: unconstrained condition and trunk blocked
condition. The reaching movements were recorded
(MotionAnalysis) and the following kinematic variables
measured: amplitude and smoothness of the hand velocity
profile, amplitude and velocity of elbow extension and
amplitude of acromion displacement. Force production was
recorded using an isostatic dynamometer (ConTrex). The
isokinetic dynamometer was also used to impose passive
63
Poster Sessions
Full Abstracts
elbow extensions at a controlled speed thanks (6°/s and 30180°/s by 30°/s steps) and an index of spasticity was computed
as the slope of the maximum torque versus angle plot. The
results confirmed the reduced aiming velocity of hemiparetic
patients when they point at their preferred velocity. In addition, we
demonstrated that they were able to increase aiming velocity
when instructed to do so. The effect of the instruction was
independently added to the effect of target distance. Elbow
extension velocity was linearly related to aiming velocity. The
amount of elbow extension increased with target distance but
decreased when the instruction was to move faster. The
displacement of the acromion increased with target distance, but
only moderately with the instruction to move faster. The trunk
restraint limited acromion movement as expected and increased
elbow extension, in some patients trunk restraint induced a
limitation of the amplitude of the hand movement. The slope of
the relationship between hand velocity and elbow rotation velocity
was increased in the trunk blocked condition. Preliminary
correlation analysis showed that the maximum hand velocity
obtained for each patient (in the condition "fast condition/far
target/trunk blocked") was inversely correlated with the spasticity
index. The amount of acromion displacement in the trunk free
condition was inversely correlated with maximum flexion strength.
Maximum elbow extension velocity was correlated with maximum
concentric extension strength. In conclusion, these results show
that the patients are able to voluntarily increase movement
velocity, at the risk of reducing elbow extension. This underlines
the importance of limiting compensations through trunk restraint
in these conditions. It is not possible in this preliminary study to
determine if slowness is related to flexor spasticity and/or to
weakness. The correlations should be taken with care because of
the limited number of patients included. Further studies are
planned in a larger number of patients and following Botulinum
Toxin therapy.
2-D-90 Quantitative assessment of driving performance
in Parkinsons disease (PD): Deficient coordination of
gaze-hand-feet-control with a deactivated sub-thalamic
nucleus stimulator
Wolfgang H. Zangemeister1, Lea Maintz1, Carsten Buhmann1
accident- errors. According to the severity code of the
errors, they were graded from 1=small to 4=critical. The
error amount was generated from the quantity and the
quality of the errors. Compared to approximately normal
driving (Average error (AE): 10.4 -7), PD Patients with STN
On drove worse (AE: 20.9 -6); however, PD Patients with
STN Off, medication ON demonstrated a lot more traffic
errors (AE: 24,6 -5); PD Patients in a total Off-condition
(STN Off, med. OFF) showed the highest number of traffic
errors (AE: 27,1 -8).
Comparing the driving time, PD Patients with STN stimulator
ON needed on average 258 sec to accomplish the specified
driving route. This can be considered the upper limit of the
normal value (Average: 231 -30). PD Patients in a total Offcondition needed a mean time of 278 sec to accomplish the
route.- Integration of steering movements over total time of
driving showed a three times reduced power in the PD´s
with STN OFF [off=2.56, on=7.56].
A similar effect was obtained in the Lomb periodograms,
and also in Wavelet analyses of steering movements. They
demonstrated that driving frequency power was similar for
PD ON and normal subjects with frequency peaks around
0.03, 0.05 and 0.09 . For PD with STN OFF in comparison,
frequency power was highly significantly lowered with
frequency peaks around 0.0025, 0.01 and 0.025.- The basic
frequencies of the parametric [three sine wave]
reconstruction of PD´s in STN OFF condition were around
0.2 , 0.4 and 0.9 Hz, whereas normal subjects and PD´s in
STN ON condition showed 0.3, 0.7 and 1.4 Hz respectively.
Furthermore, Gaze and steering movements were closely
correlated.
From a clinical point of view these data demonstrate the
loss of fine control of gaze-arm-foot coordination in the STN
OFF condition as well as the special role of the narrowing of
"attention" as a single, most important part of the disease
progress.
2-D-91 Decreased saccade velocity in
spinocerebellar ataxia 6
1
John Anderson1, Peka C Savayan2
Driving a car is an essential everyday coordination task that may
be severely limited in idiopathic parkinson´s disease (PD) with
narrowed attention, prolonged sensory-motor latencies, tremor,
rigidity, slowness and hypometria of the patients´ movements.
During smooth pursuit vision the target is fixed on to the fovea as
the eyes track the slowly moving target up to velocities of 30°/sec,
i.e. dynamic as compared to static fixation.
1
Universität Hamburg
We recorded and analyzed driving performance of 20 PD
patients´ (mean age 63.6 , 4 females) and 20 age matched
healthy subjects using an infrared system (Gazetracker) that
allowed free head-eye movements within a driving simulator
recording, as well as steering, indicator and accelerator/ brake
signals. PD symptoms were graded according to: UPDRS pt.1-4,
Hoehn & Yahr, MMST and Demtect. Average duration of PD
symptoms was 5 years (stdev. - 0.8y.). All patients were on
dopaminergic medication and were treated with an implanted
STN stimulator (sub-thalamic nucleus stimulator). After some
basic oculomotor checks, and 5 minutes of practicing with the
simulator driving system, patients had to drive through an
unknown realistic course, which lasted about 5 minutes (Bessier
Software: 3d driving school), with STN stimulation On, after a 30
min. pause with STN stim. Off, and after another 30 min. rest,
under high L-Dopa medication.
Typical driving errors were Velocity-, gap selection-, lane keeping, brake/accelerator-, indicator-, traffic sign-, near accident/
64
University of Minnesota, 2Minneapolis VA Health Care
System
Spinocerebellar ataxia 6 (SCA6) is a neurodegenerative
disease with clinical manifestations indicating cerebellar
dysfunction. It is characterized by progressive limb and gait
ataxia, dysarthria, and nystagmus along with cerebellar
atrophy. A decrease in saccade velocity has been described
in some SCA6 patients who are moderately to severely
affected by the disease and also in some presymptomatic
subjects (Chirstova et al., Arch Neurol 2008; 4:530-536),
although the decrease is not to the extent that occurs in
other SCA subtypes, e.g., SCA2 (Federighi et al., Brain
2011; 134:879-891). Although decreases in saccade
velocity are well documented in patients with brainstem
lesions (Ramat et al., Brain 2007; 130:10-35.), velocity
changes have been reported in patients with signs of
cerebellar dysfunction (Kumar et al., Ann N Y Acad Sci
2005; 1039:404-416; Zivotofsky et al., Brain Res 2006;
1093:135-140), in animal studies with inactivation of the
fastigial nucleus (Goffart et al., J Neurophysiol 2004;
92(6):3351-3367), and in modeling studies of cerebellar superior colliculus - brainstem reticular formation
interactions (Pelisson et al., Prog Brain Res 2003; 142:6989). These studies indicate that there can be a change in
the velocity profile or in the timing of the acceleration and
deceleration phases of saccades. The aim of this study was
Poster Sessions
Full Abstracts
to identify and characterize changes in the saccade velocity
profile and time course in SCA6 in order to gain further insight
into the pathophysiology of SCA6.
Data were collected from seven normal control subjects, four
presymptomatic subjects, and five ataxic patients. The horizontal
and vertical components of eye movements were recorded
monocularly using the magnetic search coil technique. The
saccade target was displaced 5, 10, 20, and 39 degrees from one
side of the center in the horizontal and vertical directions. Matlab
programs were used to digitally filter the sampled data,
differentiate the eye position signal, calculate the amplitude,
velocity, duration, and time of peak velocity for saccades, and
calculate confidence ellipses for two-dimensional plots.
The results showed that there were changes in the velocity profile
in those cases were the peak velocity was less than normal. This
could be due to abnormal signals from the cerebellum to the
burst/ominpause network in the brainstem in SCA6.
2-D-92 Developmental dyspraxia in children with motor
disorders
Stefanie Bodison1, Terence Sanger1
1
University of Southern California
Developmental dyspraxia is a childhood disorder that affects the
child's ability to learn complex motor actions. Developmental
dyspraxia is reported to exist in approximately 6% of the
population and negatively impacts the child's attainment of gross
motor skills, fine motor skills, self-care skills and the play skills
needed to effectively participate in age-appropriate activities with
peers. While studies have suggested that the underlying
mechanisms associated with developmental dyspraxia may be
the result of problems in sensory processing, motor planning or
limb kinetic deficiencies, few if any have compared these
underlying mechanisms across diagnostic groups. Recent
research has suggested that children with Autism Spectrum
Disorder (ASD), Developmental Coordination Disorder (DCD) and
Dystonia may have some level of delopmental dyspraxia, but it
was suspected by this research group that the underlying
mechanisms of the dyspraxia may vary among the three
diagnostic groups. This research study explored whether a newly
developed assessment of developmental dyspraxia (the Test of
Hand Gestures) could provide insight into the varying underlying
mechanisms of dyspraxia and if there are any group differences
in developmental dyspraxia among children ages 2-21 years with
ASD, DCD and Dystonia. The results to be presented include
normative data on the Test of Hand Gestures and preliminary
data on the Test of Hand Gestures across the three diagnostic
groups described.
F - Fundamentals of Motor Control
2-F-28 Investigating motor learning with the ETH Pattus
- a robotic approach to studying the neural control of
forelimb movements in rodents
Bogdan Vigaru1, Olivier Lambercy1, Maximilian Schubring-Giese2,
Jonas Hosp2, Melanie Schneider1, Andreas Luft2 , Roger Gassert1
1
ETH Zurich, 2University of Zurich
Building on the recent insights robotic manipulanda have given
into the neural mechanisms underlying sensorimotor learning and
control in human and non-human primates, we developed ETH
Pattus, a robotic approach to investigating planar reaching and
pronosupination movements in rats. These movements
correspond to the widely used training paradigm of grasping
a food pellet and moving it to the mouth. The compact,
highly transparent three-degree-of-freedom manipulandum
can render forces up to 2N to guide or perturb rat forelimb
movements, allows to automate the time-consuming training
periods, and offers the possibility to quantitatively assess
endpoint kinematics and kinetics of rat movements.
Preliminary experiments with 10 healthy rats provided
evidence for the ability of animals to learn how to touch and
interact with the robotic manipulandum by performing a
10mm pulling movement in a null field as well as a guiding
and a perturbing force field. Kinematic and kinetic endpoint
data collected during interaction suggest that rats modify
their behaviour in response to the force fields applied during
pulling movements. Training in a haptic tunnel resulted in a
significant reduction of the integrated straight line error,
which persisted after force field removal. An inter-session
decrease in interaction force within the haptic tunnel force
field further suggests that rats progressively produced
trajectories that were closer to the straight line. Similar to
human studies, velocity profiles in rat movements exhibited
bell-shaped profiles, which could indicate comparable motor
control and optimization strategies. These promising results
open up new research avenues for future investigations of
motor learning stages in healthy animals as well as in stroke
models, where measurements of endpoint kinematics and
kinetics could reveal the changes in movement patterns
between pre- and post-lesion conditions.
2-F-29 Mental rotation of hand movements in
congenitally blind subjects
Maitê Mello Russo1, Luis Aureliano Imbiriba2, Laura Alice
Santos de Oliveira1, Erika de Carvalho Rodrigues3, Claudia
Domingues Vargas1
1
IBCCF/UFR, 2EEFD/UFRJ, 3UNISUAM
It has been suggested that mental simulation of an action
corresponds to any mental state involving action contents
where brain activity seems to mimic or simulate the same
aspects of movements, except for the absence of any overt
motor behavior. While mentally simulating one's own
movement induces the use of primarily motor resources,
about the motor imagery of another person's movement
activates visual rather than motor processes. Although
blindness has been shown to affect visual representation in
the brain, much less is known about how higher order action
representations are affected by complete visual loss. How
does early blindness affect perspective taking? Are blind
subjects able to represent spatial visuomotor actions
performed by another person? Ten congenitally blind (mean
age 25.8 (standard deviation ± 4.1) years) and nineteen
sighted adults (mean age 23.6 (± 2.2) years), all righthanded, were instructed to perform a laterality judgment
which induced the implicit simulation of their own hand
posture (first-person perspective - 1P) or of another person's
hand (third-person perspective - 3P) located in front of their
face. They were asked to respond as quickly as possible
queries about the location of a single finger on the imagined
hand after the end of the instruction (Sirigu and Duhamel,
2001). Each instruction was presented through headphones
and the subject's verbal response was digitally recorded for
offline analysis. Response time (RT), the interval between
the end of the instruction and the subject's verbal response,
and error rates were calculated for each perspective. Twoway repeated measures ANOVA was employed to compare
the RT and the number of errors, with imagery perspective
(1P and 3P) as a within-subject factor and group (blind or
sighted) as a between-subjects factor. RT of wrong
65
Poster Sessions
Full Abstracts
responses was not included in the analysis. Statistically
significant differences were found for RT and number of errors
between groups (F(1, 27)=6.72; p=0.015 and F(1, 27)=9.64;
p=0.004, respectively) and perspectives (F(1, 27)=29.69;
p=0.00001 and F(1, 27)=6.89; p=0.014, respectively). In all
cases, higher values were observed for the blind group and the
3P. No significant interaction was found for any parameter. Taken
together, these results suggest that congenitally blind subjects
keep the capacity to perform spatial visuomotor imagery of the
hand, although presenting a worse performance when compared
to sighted subjects.
2-F-30 Muscle specific modulation of corticospinal
excitability depends on valence of the object to be
grasped
Anaelli Aparecida Nogueira-Campos1, Laura Alice Santos de
Oliveira1, Valeria Della-Maggiore1, Paula de Oliveira Esteves1,
Érika de Carvalho Rodrigues2, Cláudia Domingues Vargas1
1
Instituto de Biofísica Carlos Chagas Filho / UFRJ, 2UNISUAM
Daily in life humans are faced with situations where they need to
interact with emotional objects. Can the valence of objects
influence the brain activity preceding their grasping? The aim of
this study was to examine the corticospinal excitability during
motor preparation to grasp emotionally-laden objects. Ten
participants were instructed to perform two conditions: to grasp
(action block) or just to observe (no-action-block) objects with
different valences (pleasant, unpleasant and neutral). Four blocks
(2 action and 2 no-action) containing 42 objects each were
randomly presented. Objects used were balanced in weight and
placed inside transparent cups to prompt a similar grip among
trials. Participants sat on a comfortable chair in front of a table
where the emotionally loaded objects were presented on a sliding
slab by an experimenter sitting behind a black curtain. At the
beginning of each trial the left arm of the participant laid with the
palm facing down over the table (initial position). Trials began with
the object presentation. Then, a go signal was turned on 3s later
indicating that the participant should grasp the object.
Transcranial magnetic stimulation (TMS) was applied over their
primary motor cortex randomly at 500 or 250ms before the go
signal. Motor evoked potentials (MEP) were measured by
recording the electromyographic signal from first dorsalis
interosseous (FDI) and abductor digiti minimi (ADM) muscles.
MEP amplitude for pleasant and unpleasant trials was normalized
relative to neutral trials, creating an unpleasant/neutral and
pleasant/neutral ratio. Tree-way ANOVA revealed a condition X
valence X muscle interaction (p<0.05). A 2-way ANOVA was
applied to find out what muscle was driving this interaction. We
found a significant condition X valence interaction (p<0.05) for
FDI, which is more directly enrolled in the grasping movement.
The MEP amplitude was larger for unpleasant/neutral compared
to the pleasant/neutral conditions. There was no significant effect
for ADM and no valence effect in the no-action block (p>0.05). In
conclusion, we showed here that the preparatory activity
preceding the grasping of an object, as measured by corticospinal
excitability, is affected by its emotional value. More specifically,
our findings show a muscle specific modulation of corticospinal
excitability depending on the valence of the object to be grasped
during motor preparation.
2-F-31 Directional tuning of arm muscle activation in
isometric force generation and its prediction by flexible
and synergistic models
Daniele Borzelli1, Andrea D'Avella1, Reinhard Gentner1, Timothy
Edmunds2, Dinesh K Pai2
1
2
Santa Lucia Foundation, University of British Columbia
66
A long standing question in neuroscience is how the CNS
coordinates the large number of degrees-of-freedom of the
musculoskeletal system to achieve a variety of goals. One
possibility is that CNS optimizes task specific muscle
activations according to some performance criterion, such
as effort or accuracy, without imposing any constraint on the
muscle patterns, i.e. flexibly. Alternatively, to overcome the
complexity inherent in a system with many non-linear,
compliant actuators acting on multiple skeletal segments,
the CNS might reduce the number of degrees-of-freedom it
has to control directly by combining a small number of
muscle synergies. Evidence for muscle synergies comes
from the observation of low-dimensionality in the muscle
patterns. However, whether muscle synergies are a
simplifying control strategy actually employed by the CNS or
they only represent a parsimonious description of the
regularities in the motor output generated by a nonsynergistic controller and due to specific task constraints is
an open and debated issue. A recent study (Kutch et al.
2008) has argued against muscle synergies in the
generation of planar isometric forces on the basis of the
comparison of the directional dependence of the covariance
of the force fluctuations due to signal-dependent noise
observed experimentally with the dependence predicted by
either a synergistic model of muscle activation or a model
assuming flexible activation of individual muscles. Here we
compared the directional tuning of the activation of several
muscles acting on the shoulder and elbow joints with those
predicted by flexible and synergistic models. We recorded
EMGs and forces while five subjects generated planar
isometric target forces in 16 directions and 3 different
magnitude levels (0.1, 0.2, and 0.3 of their mean maximum
voluntary force). We estimated the end-point isometric force
generated by each muscle by multiple linear regression of
the recorded forces and EMGs. We identified muscle
synergies underlying the generation of isometric hand forces
using a non-negative matrix decomposition algorithm.
Finally, for each target, we predicted muscle activations
assuming minimization of the sum of square of the either
muscle or synergy activations. We found that the synergistic
model predicted the experimental directional tuning of
muscle activation better than the flexible model. For all
subjects, the mean squared prediction error was lower with
the synergistic model. This result supports the hypothesis
that the CNS employs combinations of muscle synergies to
efficiently, even if not optimally, select the muscle activity
patterns required to achieve a goal. However, further work is
required to validate the specific approximations and
assumptions introduced in the models. In particular, the
result depends on the estimation of subject-specific EMG-toForce matrices which was performed with multiple linear
regression. Such estimation could be improved by
constraining the regression with a priori information on the
musculoskeletal geometry and physiological characteristics
of the muscles.
2-F-32 Context dependent changes in cue
responses prior to movement in primate ventral
premotor cortex
Lachlan L. Franquemont1, Carlos E Vargas-Irwin1, Michael J
Black2, John P Donoghue1
1
Brown University, 2Max Plank Institute for Intelligent
Systems
Neurons in primate ventral premotor cortex (PMv) are
preferentially active during specific types of grasping
movements. These neurons may also fire selectively in
response to the visual presentation of specific objects. The
Poster Sessions
Full Abstracts
visually evoked responses of PMv neurons have been interpreted
as contributing to the transformation of the visual attributes of an
object into a motor plan suitable for grasping it. This
transformation need not represent a one-to-one mapping, since it
is possible to grasp a given object in different ways. We examined
the interaction between object and grip-related neural activity with
a grasping task where each object could be grasped with multiple
grip types. We compared two experimental conditions: one where
the grip type is specified prior to the visual presentation of the
object (G-O), and one where an object is presented before
specifying the grip type (O-G). We also examined an additional
control condition similar to O-G that included a delay period
without any stimuli. This 'memory' period was followed by a
second presentation of the object, allowing us to compare the
object-evoked response in two different contexts (with and without
a previously cued grip strategy). Neural activity was recorded in a
macaque monkey (Macaca Mulatta) using a 96-microelectrode
array chronically implanted in PMv on the exposed cortex behind
the arcuate genu. About 30 well-isolated single units (with
SNR>1.5) were recorded in each session. For each session the
monkey alternated between performing blocks of the O-G task
and either the G-O or memory task. Note that all tasks required
the same motor responses (lifting one of two objects from the
same location using the prescribed grip). We found that neural
activity evoked by the presentation of the target object varied
according to whether it was presented before (O-G condition) or
after the grip cue (G-O condition). On average 68.4% of all
selective neurons (30.0% of all well isolated units) displayed
different firing rates depending on the epoch of object
presentation (Kruskal-Wallis test, p<0.01). Similarly, neural
activity following the grip cue was different depending on the
temporal order of the cues, with an average of 63.6% of selective
neurons (27.2% of all well isolated units) displaying statistically
different responses. Of all well isolated units that displayed
selectivity for grip or object type in, or between, epochs, 97.5%
displayed context dependent selectivity. An SVM classifier
utilizing the neural population was used to classify epoch as
either a grip cue and object presentation. It correctly classified
73.6+/-0.24% of object epochs, and 75.7+/-0.24% of grip epochs
(95% confidence boundary at 60.2%) using the first second of
neural activity after the object presentation or grip cue. The two
object presentations in the memory condition (before and after the
grip instruction) also elicited different responses. On average,
85.0% of all selective cells (45.9% of well isolated) neurons
responded differently during the two object presentations. These
findings show that object-evoked responses in PMv can be
influenced by ongoing motor plans for grasping. This suggest that
PMv represents an evolving combination of object and grip
related information rather than a strict mapping from one to the
other.
2-F-49 Reach and grasp encoding in macaque parietal
area AIP
Hans Scherberger1, Sebastian J Lehmann1
1
German Primate Center
The anterior intraparietal cortex (AIP) in the macaque brain is
strongly involved in the visuo-motor transformation of hand
grasping actions. However, the influence of target location during
these reach-to-grasp movements have not been analyzed
systematically. Here, we varied the eye and target position in
space to study the influence of these signals on the grasp
encoding in AIP. Macaques monkeys performed a delayed reachto-grasp task where they grasped a target with one of two grip
types while the target and the eye fixation position were
systematically varied (total of 26 task conditions; randomly
interleaved). Eye position was monitored with an optical eye
tracker. Grasp trials consisted of four epochs (fixation, cue,
memory, and movement). We recorded 353 single units in
AIP (207 in monkey P, 146 in monkey S). Of those, 39.9%
(n=141) were highly significantly tuned for grip-type and
71.1% (n=142) for the target position in space in at least
one of the epochs 'cue', 'planning' or 'movement' (2-wayANOVA, p<0.01). Grasp-related activity increased towards
movement execution, whereas reach-related information
was rather constant throughout the task. A substantial
number of cells encoded reaching information in retinotopic
coordinates, i.e. relative to the gaze position. Furthermore,
we simulated the decoding of grasp-type and position from
this dataset separately for each epoch using a maximum
likelihood approach. Decoding performance was 78% (81%
for the second animal) for grasp-type in the cue epoch, 78%
(87%) in the planning, and 98% (99%) in the movement
epoch, whereas for the different spatial conditions
performance was at 61% (59%) for fixation, 91% (83%) for
cue, 92% (81%) for planning, and 85% (76%) for the
movement epoch (chance level 7.7%). These findings
indicate that AIP encodes reach-related information in
spatial and retinotopic coordinates in addition to the grasp
type.
2-F-50 Role of motor cortex in skill learning
depends on learning strategy
Risa Kawai1, Tim Markman1, Bence P Ölveczky1
1
Harvard University
Much of our behavioral repertoire, whether it is hitting a
tennis serve or playing the piano, requires learning and
executing complex motor sequences with precision. Motor
cortex (MC) is thought to be integral for learning such motor
skills, but its specific role remains to be understood. Studies
in both rodents and primates have shown some behaviors to
be severely affected by MC lesions, while others are largely
spared. Here we propose that the role of MC depends on
how the motor skill is first acquired. We distinguish two
fundamentally different learning processes: (i) trial-and-error
learning in which the animal relies on differential
reinforcement of internally generated exploratory actions for
shaping motor output, and (ii) instructive learning, in which
external sensory cues inform the the timing and quality of
the behavior. To test the involvement of MC, we
manipulated its function by way of lesions and temporary
inactivations in rats trained to press a lever with their
forepaw in a precise temporal sequence. To explore the
dependence of learning strategy, the same behavior was
trained either through trial-and-error (i) or with external cues
instructing the tap times (ii). In both cases, animals
developed highly precise and stereotyped motor behaviors,
as measured by the intertap intervals and paw trajectories
extracted from high speed movies. Preliminary results show
that rats learning the task through trial-and-error were
unaffected by MC lesions when tested a week after lesions,
while animals learning through instruction showed no
memory of the acquired behavior post lesion. Irrespective of
which learning strategy was used, reversible
pharmacological inactivations of MC severely disrupted the
acquired lever press sequence. However, the
pharmacological dose required to induce an effect in rats
that had learned through instruction was more than an order
magnitude smaller than that required for rats trained through
trial-and-error. These results suggest that motor sequences
acquired through trial-and-error learning are generated by
sub-cortical circuits that are only transiently affected by
removal of MC, while motor sequences learned with
instructive cues are MC dependent. Furthermore, bilateral
MC lesions in naive animals did not significant impair
67
Poster Sessions
Full Abstracts
learning in the trial-and-error paradigm, but showed severe
deficits in the instructive learning paradigm. These results
suggest that the function of MC in the acquisition and production
of motor skills depends on the mode of learning.
2-F-51 The influence of movement speed on variances
of target tracking arm movements using a computer
mouse
Jozsef Laczko1, Bence Borbely2, Gabor Fazekas3, Jozsef Takacs2
2-F-52 How octopuses coordinate their eight
flexible arms while crawling
Guy Levy1, Tamar Flash2, Benny Hochner1
1
Hebrew University of Jerusalem, 2Weizmann Institute of
Science
Crawling is a typical way of locomotion of octopuses on the
seabed or along rocks in shallow waters or outside of the
water.
1
Pazmany Peter Catholic University and Semmelweis University,
Pazmany Peter Catholic University, 3National Institute for
Medical Rehabilitation
2
We study a visuo-motor task performed by 10 healthy individuals.
The subject sat on a chair in front of a computer desk and faced a
monitor and saw a moving target on the screen. The instruction
was to follow the moving target with the mouse pointer as
accurately as possible. The motion of the mouse pointer was
controlled by hand movements of the participants using a specific
mouse device on a digitizer tablet. There were 2 path and 2
speed conditions for the moving target: circle- and square-shaped
paths and normal and fast speeds based on a speed calibration.
Subjects performed 10 trials under each condition while
coordinates of the arm's anatomical points were recorded using
an ultrasound based movement analyzer (ZEBRIS). Arm
configuration was defined by 7 rotational angles: 3 in the shoulder
(shoulder frontal, shoulder sagittal and upper arm rotation), 2 in
the elbow (elbow flexion and lower arm rotation) and 2 in the wrist
(wrist elevation and wrist azimuth). Tracking precision, hand
position- and arm configuration- variances were computed for
each condition.
Moving on square path the increased speed induces significantly
less accurate pursuing of the target than in the case of circle path.
Neither the shape nor the speed brings significant differences in
hand position variances, although hand position variance is
affected more by the movement speed than by the shape of the
path. Higher speed led to increased hand position variances. For
higher speed the variance of arm configurations increased at a
smaller rate than hand position variances. The structure of arm
configuration variances was investigated using the uncontrolled
manifold (UCM) approach to discern why arm configuration
variances changed relatively less while hand position variances
increased with speed. We assume that the structure of arm
configuration variances has changed with movement speed in
such way that compensated (within the UCM) variance that did
not affect mean hand position variance decreased while
uncompensated (outside the UCM) variance increased. Thus
movement speed affected total arm configuration variance at a
smaller rate in comparison to hand position variance since the
latter does not depend on the compensated component of arm
configuration variance and the drop of this component is not
reflected in hand position variance. Consequently hand stability
weakened at larger degree than arm configuration stability when
the target moved faster. The increment of the uncompensated
component of arm configuration variance as a consequence of
increased speed was smaller for circle than for square tracking.
This may explain the more accurate tracking when the target
moved on circle path.
We conclude that the subjects rather paid attention to control
whole arm movements while watched the target on the screen
and moved the mouse on the tablet. For this kind of visou-motor
coordination - that is quite common in using computers - the
participants tried to reduce mainly arm configuration changes
instead of focusing on hand position stability as movement speed
increased.
68
The first stage of this study show that octopuses use their
arms while crawling for pushing the body by alongation with
rhythmical stereotypical steps composed of anchoring
several neighboring suckers to the substrate followed by
arm elongation to push the body in the direction opposite to
the elongating arm.
The second stage of the study show how the octopus
coordinates its arms when crawling. Mature octopuses were
videotaped from underneath while crawling in shallow
waters and chosen sections of interest were stored as single
images. Arms that participated in pushing and additional
points of interest were labeled manually on consecutive
images for the kinematic analysis. The data was used to test
our hypothesis suggesting that the crawling direction is a
simple vectorial combination of the pushing arms and
changing the direction of crawling is done simply by
activating a suitable set of arms rather than by rotating the
body. Following the octopus morphology, the arms are
described as spreading around the body symmetrically with
a 45º between each adjacent couple of arms. We assume
that each arm has a fixed direction of pushing by elongation
relative to the body. Using this assumption together with the
information regarding the identity of the pushing arms at
each moment, we were able to reconstruct, in a simple toy
model, the direction of crawling using only the simple
vectorial combination of the participating arms. We confirm
our hypothesis by showing an excellent match between the
reconstructed and the actual crawling path.
We suggest that as seen in other octopus movements,
control simplification is achieved by using simple
stereotypical movements generated at the level of the
peripheral neuromuscular system of the arm (rhythmical
suckers adherence and length changing) while the
participating arms are coordinated by commands from the
central brain.
Supported by the European Commission 7th Framework
(FP7), Project OCTOPUS
2-F-53 Functional characterization of the
cholinergic motor innervation in the special
neuromuscular system of the octopus arm
Nir Nesher1, Naomi Feinstein1, Lili Englister1, Finkel Eran1,
Benny Hochner1
1
Hebrew University of Jerusalem
The octopus arm is an outstanding example of an efficient
skeleton-free motor system. The arm is composed almost
entirely of muscle cells, whose constant volume constraint
creates a muscular-hydrostat. Previously we showed that
the obliquely striated muscle cells are small (1ý~0.01mm),
isopotential, lack Na but have a fast Ca 2 spike and appear
to be innervated by three excitatory cholinergic synaptic
inputs (Matzner et al 2000, Rokni and Hochner 2002). Here
we ask if these unique properties involve special
neuromuscular junction. Using rhodamine conjugated alphabungarotoxin ( ¿-BGT) to label enzymatically dissociated
Poster Sessions
Full Abstracts
muscle cells and transverse arm sections. ¿-BGT labeling reveal
a loclized bulk of AchR were localized, eye-shaped, at the center
of the cell; close to the nucleus. The labeled area coincide with
membrane invaginations which were clearly visible also in
transverse sections. We next characterized the physiological and
spatial properties of AChRs. 10 mM ACh was pressure injected
through a micropipette at specific locations. In accordance with
the ¿-BGT labeling, the strongest and fastest depolarization
responses were obtained at the cell center. In contrast to other
muscular AChR, these responses were not desensitized. Similar
to nerve evoked EPSP in-vivo, the ACh induced current/voltage
relationship indicated a linear relationship with a reversal potential
close to zero membrane potential. These results confirmed that
ACh is the neuromuscular transmitter in the octopus arm and the
results fit our previous deduction, based on the electrotonic
compactness of the muscle cells, that a single neuromuscular
junction is sufficient to control the membrane potential of the
muscle cell, with no need for the poly-terminal innervation
common in other invertebrates.
Anastasia Sylaidi1, Aldo Faisal1
Support: European Commission 7th Framework (FP7), Project
OCTOPUS and Smith Family Laboratory the Hebrew university
Here we employ an empirical approach to examine three
possible hierarchical architectures of human motor learning
in object manipulation tasks: (a) The first suggests that task
dynamics are principally learned and represented in an
intrinsic joint-based frame of reference. Based on this
model, task dynamics are learned as a single motor
representation for all possible joint configurations but for a
single object position. Separate task representations are
thus used for different object positions. Conversely, (b) the
second scheme predicts that learning is realised with regard
to an extrinsic object-based frame of reference, whereas (c)
the third scheme points to the independent and equally
weighted representation of learned task dynamics both in an
intrinsic and extrinsic frame of reference. In the last case
separate task representations are used for each different
joint- or object configuration.
2-F-54
When emulation becomes reciprocity
Luisa Sartori1
1
Università degli studi di Padova
It is well known that perceiving another's body movements
activates corresponding motor representations in an observer's
brain. It is nevertheless true that in many situations simply
imitating another's actions would not be an effective or
appropriate response, as successful interaction often requires
complementary rather than emulative movements. At what point,
then, does the automatic tendency to 'mirror' become the
inclination to carry out an appropriate, complementary action? In
the present study, single-pulse transcranial magnetic stimulation
(TMS) was used to evaluate the shift from emulation to reciprocity
in observers' corticospinal activity. The effects of single-pulse
TMS on the muscle-specificity of motor evoked potentials (MEP)
during action observation were measured in participants at time
intervals matched to specific kinematic events taking place during
a model's movements. MEP were recorded from the abductor
digiti minimi (ADM) and the first dorsal interosseus (FDI) muscles
of the right hand while participants observed video-clips
portraying a sequence of movements which, in some cases,
elicited a spontaneous complementary response. We found that
observation of such double step action characterized by a final
complementary request engendered a shift from simulation to
reciprocity in the participants' corticospinal activity. When an
observed action calls for a non-identical complementary action,
an interplay between the automatic tendency to resonate with
what is observed and to implicitly prepare for the complementary
action do emerge. A variation in motor evoked potentials (MEP)
was noted early, even before the complementary request became
explicit. Control conditions in which participants observed the
same action sequence but in a context not implying a
complementary request were included. As corroborated by
kinematic analysis, our findings demonstrate that observers are
attuned to advance motor information implicit in observed
movements and use it to anticipate a future course of action and
to prepare for an appropriate, complementary gesture. Implicit
complementary requests might have the ability to prime nonidentical responses. Modulation of corticospinal excitability
appears to be a reliable, indirect measure of the automatic human
tendency to interact, regardless of whether there will be an
effective interaction.
2-F-55 Hierarchical representations of object
manipulation tasks
1
Imperial College London
Motor tasks are encoded by the brain at a symbolic level
(Badre, D., 2008), but are ultimately executed at the level of
muscle activations. This illustrates that the central nervous
system (CNS) has to solve a problem of hierarchical motor
control. In this context motor control problems can be
defined with regard to an intrinsic frame of reference based
on the body's actuators and sensors (joints and muscles)
and/or an extrinsic frame of reference related to task
context, environmental settings and object properties. These
two frames of reference have been indirectly suggested or
explicitly examined in the past (Shadmehr, R., Mussa-Ivaldi,
F., 1994; Ahmed, A., 2008; Ingram, J. 2010). However, it
remains unclear how motor coordination is linked to one or
the other depending on task conditions and whether the
frames can be explained as parts of a hierarchical structure
underlying motor learning.
In order to test the three representation systems as
candidate mechanisms of motor control we used a 3D
virtual reality setup with body motion tracking. Human
subjects were asked to match a rotational task to a given
accuracy while holding a half full 1 L bottle of water at
varying orientations relative to the hand. In contrast to
previous studies, which primarily made use of tools with no
internal degrees of freedom or static dynamics (Shadmehr,
R., Mussa-Ivaldi, F., 1994), we selected naturalistic object
dynamics to facilitate motor tasks encountered in real-life
conditions.
We measured subjects' performance in terms of the pivot
point displacement and examined how learning is
generalised to multiple test orientations after exposure to a
single training orientation of the object. An inverse
experimental protocol was subsequently implemented, in
which subjects were asked to perform object rotations while
the object's orientation remained constant relative to the
hand, whereas joint coordinates varied across different
experimental sessions. Our results revealed a significant
difference between pivot point displacement in training and
testing phases for both experiments (Wilcoxon test,
p<0.0125). This difference demonstrates a poor
generalization capability of learned task dynamics to novel
task contexts. It thus provides supporting evidence for our
third scheme of hierarchical motor control in which different
task representations are employed for different joint and
object configurations. As such, our work introduces a novel,
inclusive formalisation for capturing learning during the
interaction with objects of complex internal dynamics.
69
Poster Sessions
Full Abstracts
2-F-56 Expectation about movement error and its
influence on real-time reach control
Darian Cheng1, Brendan Cameron1, Gordon Binsted1
1
University of British Columbia
When error arises during an unfolding movement, our
sensorimotor system rapidly corrects for it. These online
corrections generally occur without our awareness, and they are
largely resistant to our attempts to override them, though there is
evidence that they can be partially suppressed under the right
conditions (Cameron et al. 2009; Streimer et al. 2010; McIntosh et
al. 2010). In the present study, we further examined the limits on
the 'automaticity' of online corrections. Specifically, we tested
whether expectation about upcoming movement error could be
used to enhance online responding. Participants made targeted
movements with a stylus on a graphics tablet. They were given a
pre-cue at the start of each trial that indicated whether the target
was likely to jump on that trial (80% probability) or unlikely to jump
on that trial (20% probability). The target jump (up or down) was
orthogonal to the primary reach vector, and no information was
provided by the pre-cue about the direction of the jump. When a
jump occurred, it was coincident with the onset of the movement.
In one condition (Pro-point), participants were instructed to
correct their movement to the target's jump location. In the other
condition (Anti-point), participants were instructed to correct their
movement to the location opposite the jump location. In the Propoint condition we examined the latency of the online correction
to the target, while in the Anti-point condition we examined the
latencies of both the initial (unintended) correction towards the
target and the secondary (intended) correction away from the
target. In the Pro-point condition we did not find clear evidence of
a faster correction when participants were expecting a target
jump. In the Anti-point condition, however, we observed 1) a
slower and smaller initial correction when participants were
expecting a jump and 2) a faster secondary correction when
participants were expecting a jump. That is, the anticipation of a
pending jump appeared to allow some suppression of the
automatic response to the target in addition to a speeding of the
voluntary response.
2-F-57
Thalamo-cortical network in locomotion
Irina Beloozerova1, Vladimir Marlinski1, Mikhail G Sirota1
1
Barrow Neurological Institute
Locomotion is one of the most essential and frequently used
behaviors. The neural mechanism that determines the order of
muscular contractions and the coordination of limb movements
during locomotion resides in the spinal cord. The spinal
mechanism, however, lacks the distant information about the
outside world, in which locomotion takes place and thus has to be
adapted to, and lacks information about the purpose of
locomotion. The motor centers of the brain adapt locomotion to
the peculiarities of the environment and to the needs of a subject.
The motor thalamo-cortical network plays a central role in this
adaptation.
The ventro-lateral thalamus (VL), a part of "motor thalamus",
receives locomotion-related information from the spinal cord both
directly and indirectly via the cerebellum. It receives information
about the outside world from the cerebellum and also likely from
the cortex. The VL merges these two streams of information and
conveys the resulting signal for locomotion adaptation up to the
motor cortex (MC) for delivery to the spinal cord. This signal from
the VL is, however, gated by the reticular nucleus of thalamus
(RE), a layer of inhibitory neurons on the rostro-lateral border of
the thalamus. The RE is under direct influence from the MC.
70
In this study in freely walking cats, we analyzed single
neuron activity in the VL, RE, and MC during two locomotion
tasks. One of them was a walk on a flat surface, a task that
can be accomplished by the spinal cord alone, while another
was stepping on cross-pieces of a horizontal ladder, a task
for which vision is essential and which thus requires the
activity of thalamo-cortical network in order to be successful.
Based on this analysis, we defined five modes of integration
of locomotor-based and vision-based information in the VL;
we uncovered distinction in the thalamic and cortical
controls over the shoulder, elbow, and wrist; and revealed
striking differences in signals that are channeled through
different cortical efferent systems during locomotion. Using
this information we propose a novel model of how thalamocortical network contributes to visual guidance of
locomotion.
2-F-58 Limb movement amplitude systematically
influences temporal order judgments
Robert Hermosillo1, Paul van Donkelaar1
1
University of British Columbia – Okanagan
When generating limb movements, motor planning signals
affect the perception of somatosensory signals to allow for
differentiation between self- and externally-imposed
cutaneous sensations. In the present study, we examine
whether limb movements can systematically influence
temporal order judgments based on the relative position of
the limbs and amplitude of the movement. We have
previously shown that planning to cross the arms increases
error rates for temporal order judgments (TOJs) of
cutaneous stimuli delivered to the tip of each index finger.
To further investigate how movement planning influences
TOJ decisions, we performed an experiment in which
participants moved their limbs symmetrically or
asymmetrically. In particular, we investigated whether or not
the amplitude of the movement affected the accuracy in the
TOJ task. Results showed that when the hand generating
the larger amplitude movements was stimulated first, error
rates in the TOJ task were larger relative to when the hand
generating the shorter amplitude movements was stimulated
first. By contrast, we observed no difference between error
rates for either hand during symmetrical movements. Taken
together, this implies that some aspect of planning
movements which vary in amplitude interacts with the
decision-making processing associated with judgments of
temporal order.
2-F-59 Properties of force fields in the primate arm
induced by intraspinal microstimulation
Hiroaki Yaguchi1, David P Kowalski2, Tomohiko Takei1,
Kazuhiko Seki1
1
National Institute of Neuroscience, 2Drexel University
Isometric force fields output by the limb have been
evaluated in the frog, rat, and cat, and the concept of the
motor primitive has been postulated as the underlying
fundamental mechanism in these systems. However, the
force field generated by intraspinal microstimulation of the
cervical cord has yet to be studied in the monkey, and as
such it is not known if the primate spinal cord still follows the
same principle for movement control. Toward this end, two
macaque monkeys were chronically implanted at the C5-7
level with floating microelectrode arrays, each consisting of
12 electrode shanks arranged in a three by four grid with an
intershank spacing of 2.5mm and a uniform penetration
depth of 3mm (Monkey U) or 4mm (Monkey E). These
arrays were specifically chosen for their usability as a
Poster Sessions
Full Abstracts
chronically implanted device, and have previously been used as a
stimulating electrode array in acute experiments for individual
muscle recruitment. EMG recording electrodes were implanted in
the left shoulder (N=3) and arm muscles (N=5-6) . Under
atropine-medetomidine (Dormitor®) anesthesia, the monkey was
placed in a supine position and the arm moved throughout its
normal range of motion lateral to the body on a 4 or 8cm grid. A
200µs bipolar stimulation train was applied at 40Hz for 500ms to
either one or two electrode sites. Different force fields were
evoked which, in some cases, contained an equilibrium point or
line where the force output was zero with restorative forces being
generated at other positions, while others were uniformly directed
as would be the case for direct recruitment of motor pools or
passing motor fibers. Further, each stimulation site evoked
activity in particular sets of muscles, or synergies, which were in
some cases dependent on the configuration of the arm.
Simultaneously stimulating two sites mainly led to a winner-takeall or non-linear response at an electrode depth corresponding to
Lamina VII (3mm from cord surface), while an electrode depth
corresponding to Lamina IX or the anterior funiculus (4mm) lead
to a linear summation in most cases. In some cases, stimulation
at one site inhibited the ability of the second site to evoke
movement even after stimulation of the first site had ceased,
evidence that a long-term pattern of interneuronal activity was
being evoked. We also found that the finger or arm movement
was evoked for more than five months, and evoked force fields
were stable over that time on all but one channel. These results
suggest the feasibility of long-term stimulation of the primate
spinal cord through implanted microelectrode array for restoring
upper arm movement in a patient with spinal cord injury. In
conclusion, although the level of dexterity in the primate upper
limb is defined mainly by descending control from the brain, the
underlying motor primitives are still present and can be recruited
in a similar manner to what has been previously shown in lower
animals.
2-F-60 Evidence for distinct posture and movement
states at the neural population level
Nicholas Sachs1, Christian Ethier1, Rachel M Cassidy1, David P
Bontrager1, Zachary A Wright1, Lee E Miller1
1
Northwestern University
One of the fundamental questions in motor behavior is whether
the contrasting goals of limb posture and movement are realized
though distinct specialized processes or are simply the result of a
single robust control system. Single unit recording and human
psychophysics studies have addressed this question, but there
appears to be a lack of analysis at the neural population level to
support either theory. We analyzed ensemble neural activity
simultaneously recorded from tens of neurons in the primary
motor cortex of monkeys during sequences of point-to-point
reaches separated by stable hold periods. Attempts to model the
relationship between neural discharge and limb endpoint
kinematics with a linear filter or Wiener cascade nonlinearity
across the entire dataset were moderately successful, however
we discovered two sub-regions within which the relationship
between neural discharge and limb endpoint velocity was very
nearly locally linear. These two sub-regions were distinguishable
by limb endpoint velocity, comprising lower velocities associated
with limb stabilization and higher velocities associated with
movements between targets, respectively. These findings support
the existence of distinct posture and movement states at the
neural population level. Furthermore, we could identify these
states accurately using a linear classifier trained on neural firing
rates, where ground truth posture and movement classes were
distinguished based on a threshold applied to limb endpoint
speed. Investigation of individual firing rates within the neural
population indicated that some neurons had a tendency to
increase their mean firing rates in the movement state
relative to those in the posture state, while others
demonstrated the opposite effect. This suggests the
presence of different classes of neurons that may be
consistent with the description of M-class and S∆ neurons
presented by Humphrey and Reed (1983). Algorithms for
brain-machine interfaces (BMIs) typically have not
distinguished between postural and movement intents when
decoding neural activity. In order to test the potential role of
movement state detection in BMI applications we devised a
dual-state control paradigm that calculated cursor velocity
with a decoder based on a mixture of linear filters that were
trained exclusively on neural data during either posture or
movement periods from the reaching experiment described
above. During online testing the mixture weights assigned to
the movement and posture filters were calculated according
to the following equations: w(movement)=1/(1 e^(-4(d-2.5)))
w(posture)=1-w(movement) where 'd' represented the
distance (in cm) between the center of the cursor and center
of the target. This was based on the assumption that the
subject would attempt to stabilize the cursor near the target,
but would want to move the cursor more freely when moving
toward new targets, potentially engaging its motor cortical
neurons with patterns similar to those present in natural
posture and movement states. We compared the
performance of monkeys using this dual-state decoder to
control cursor movement with the use of a single linear filter.
Monkeys consistently performed better with the dual-state
decoding method than with the linear filter, suggesting that
the use of algorithms that respect differences between
posture and movement states in the neural space,
consistent with those observed during normal behavior, may
be beneficial when designing BMIs.
2-F-61 Dissociating anticipatory control of digit
positions and force in the primary motor cortex
Marco Davare1, Qiushi Fu2, Jason Choi2, Marco Santello2
1
Institute of Neurology, 2Arizona State University
When manipulating objects, visual cues and past
experience enable us to prepare an accurate hand shaping
and to adjust fingertip forces according to the object intrinsic
properties. How anticipatory control of digit positions and
force is processed in the primary motor cortex (M1) is still
debated. In a total of eight subjects, we used transcranial
magnetic stimulation (TMS) to quantify changes in
corticospinal excitability (CSE) during tasks involving
different hand shaping and force control. The object
consisted of an inverted T-shaped manipulandum with
force/torque sensors to measure digit placement and force
development. In the first task, subjects had to place their
thumb and index finger on pre-defined points on the object
and exert a force equal to 10% of their maximum voluntary
contraction (MVC, 'place+squeeze' condition) or simply
place their fingers on the same points on the object without
exerting force ('place'). In control conditions, we also asked
subjects to perform the task in a reversed order
('squeeze+place') or to simply squeeze the object using a
10% MVC force level ('squeeze'). In the second task,
subjects had to grip and lift the object using a precision grip
at constrained or self-chosen digit locations. The object
center of mass was shifted to the right so that to minimize
object roll during lift, subjects had to anticipate precisely
either digit forces or digit forces and positioning on the
object, respectively. TMS was delivered during movement
preparation (the TMS click was the 'go' signal) and motor
evoked potentials (MEP) where measured in the first dorsal
interosseus and abductor pollicis brevis muscles. We found
71
Poster Sessions
Full Abstracts
that the peak-to-peak MEP amplitudes were larger in the 'place'
compared to the 'place+squeeze' condition (on average 46 and
27% increase in the FDI and APB, respectively). MEP amplitudes
were also larger the 'squeeze+place' compared to the 'squeeze'
condition (41 and 36% increase in the FDI and APB,
respectively). The fact that a significant difference in CSE was
found between 'place' vs. 'place+squeeze' conditions indicates
the involvement of different neural resources even though the first
event (digit positioning) was identical in the two tasks. Moreover,
the increase in CSE was not specific to digit placement being the
first event per se as 'squeeze+place' also elicited larger MEPs
than 'squeeze'. Together, these results indicate that cortical
excitability during grasp planning is not merely related to adding
more elements to a task sequence as one-event task elicited
larger MEPs than a two-event task. Finally, we found larger MEPs
(16 and 21% increase in the FDI and APB, respectively) using
self-chosen rather than constrained digit locations, which
suggests that planning the coordination of digit placement and
forces required for unconstrained grasping and manipulation
relies on different neural mechanisms than force planning
associated with constrained grasping. Our preliminary results
suggest that the corticospinal system attributes different neural
resources to the anticipatory control of finger positioning and
force scaling. It is possible that the visuomotor circuits related to
the control of hand shaping and digit contact distribution have a
larger gain than the sensorimotor pathways related to the
anticipatory control of force.
2-F-62 Development of spatial and temporal bimanual
coordination during childhood
Betteco de Boer1, Lieke Peper1, Peter J Beek1
1
VU University
The effects of development on bimanual spatial and temporal
coordination were examined by comparing task performance
among four age groups: 6/7 years, 10/11 years, 14/15 years, and
young adults. During development across these ages, myelination
of the corpus callosum may induce changes in interlimb coupling,
along with many other relevant changes in neuronal sites and
networks.
Spatial coupling was assessed using a bimanual line-circle
drawing task. The effect of coupling was assessed by comparing
performance in drawing the two different shapes simultaneously
(i.e., line-circle) to bimanual drawing of the same shapes (i.e.,
line-line and circle-circle) and unimanual drawing of a single
shape (i.e., line or circle). Temporal coupling was assessed by
studying the stabilizing contributions of interlimb interactions
related to movement planning, error correction, and reflex
interactions. For this purpose, participants performed rhythmic
flexion-extension movements of the wrist in a horizontal plane. By
comparing several tasks in which these interactions related to
planning, correction, and reflexes are involved to a different
extent, the stabilizing contribution of each of these interactions to
the temporal coordination pattern was assessed.
Temporal coupling strength and coordinative stability increased
with age, at similar rates for in-phase and antiphase coordination.
Age-related changes in underlying interlimb interactions differed
for the three interaction sources: the relative contributions of
planning and reflex interactions to the achieved stability did not
change with age, whereas interactions involving error corrections
improved. Spatial drawing performance increased with age, as
evidenced by improved circularity, decreased variability, and
improved smoothness. With increasing age, participants were
better in drawing two different shapes simultaneously, as
evidenced by a smaller deterioration in the circularity index of
circle drawing for adults than children when drawing line and
circle together relative to drawing two circles or one circle.
72
Thus, temporal coupling changes were observed gradually
across age 6 to adulthood, whereas spatial coupling
changes occurred mainly after age 14/15. This difference in
the development of temporal and spatial coupling
corresponds to the anterior-posterior direction of corpus
callosum myelination as reported in the literature.
Furthermore, development of in-phase and antiphase
temporal coordination was not different across age groups,
indicating that improved inhibition of mirror activity is not the
main effector of improvements in bimanual coordination.
2-F-63 Neuronal correlates of running speed in the
dorsal striatum
Pavel Rueda-Orozco1, David Robbe1
1
Institut d'Investigacions Biomèdiques August Pi i Sunyer
(IDIBAPS) (NIF:Q5856414G)
Dysfunction of the basal ganglia results in severe motor
disabilities but how this subcortical network contributes to
normal motor control is still poorly understood. We designed
a new behavioral paradigm to study how the striatum
contributes to the control of the locomotion dynamics. Rats
were trained to run on automatized treadmill equipped on
one side of a salient photo-detector and a reward delivery
port. Breaking the beam of the photo-detector stops the
treadmill and triggers the delivery of a drop of sucrose. Rats
are progressively trained to run for at least 7 seconds before
stopping the treadmill. Early breaks of the photo-detector
beam result in a 15 seconds-long running penalty and
absence of sucrose delivery. Rats succeeded in this task by
learning to perform very stereotyped runs during which they
modulate their speed in a timely manner. Following training,
rats were implanted with tetrode arrays in the dorsolateral
part of striatum. We found that the spiking activity of about
70% of the recorded cells was strongly modulated during
the task. Specifically, while the animals ran, cells fired
across successive trials in a reliable order. Remarkably a
large fraction of those cells had their firing rate sharply
correlated with running speed and those correlations could
not be accounted by spatial or temporal bias on the firing
rate. Our results suggest that dorsolateral striatum plays a
central role in the control of the kinematic of well-learned
movements.
G - Theoretical & Computational Motor Control
2-G-64 Gravity-compensating muscular torque
explains biases in perceived arm-movement extent
Nienke Debats1, Jeroen Smeets1, Robert J van Beers1
1
VU University Amsterdam
In daily life, humans seem to adequately control their arm
movements. This suggests that the nervous system is quite
accurate in sensing the limbs' position and movements.
However, several examples of kinesthetic biases prove
otherwise. One example of such a bias is the haptic radialtangential illusion: if a human perceiver traces the outline of
a rectangular shape in the horizontal plane, the length of the
radial segment is substantially overestimated relative to the
length of the tangential segment. For a rectangle to feel
square, the radial segment has to be shorter by about 13%
to 20%. This illusion thus indicates substantial biases in the
perceived extent of arm movements in the horizontal plane.
At present, the underlying cause of these biases is unclear.
Poster Sessions
Full Abstracts
Recently, it was proposed that the radial-tangential illusion is
related to gravity (Debats et al., 2010) When keeping your
unsupported arm horizontal at shoulder level, a certain torque is
needed to counteract gravity from pulling the arm down. The
magnitude of this torque depends on the position of the arm's
center of mass, and thus on its orientation in the horizontal plane.
When making an outward or inward radial arm movement in the
horizontal plane, the magnitude of this torque increases or
decreases, respectively. The difference in torque between the
start and endpoint of the movement is what we refer to as
∆Torque. The present study tests the ∆Torque-hypothesis
(Debats et al., 2010), that is, whether the magnitude of ∆Torque
affects the strength of the illusion.
Blindfolded participants (n = 18) performed a two-alternative
forced-choice perceptual task in which they traced an L-shaped
figure with the index finger of their right hand. They verbally
indicate whether the radial segment of the L-figure was either
longer or shorter than the tangential segment. The L-figure was
presented at shoulder height in the horizontal plane. It had a fixed
tangential segment length (20 cm), and eleven possible radial
segment lengths (13.5 cm to 18.5 cm with steps of 0.5 cm). Each
size was presented ten times. From the psychometrical functions
we obtained the point of subjective equality, which indicates the
relative overestimation of the radial segment and thus the
strength of the illusion. We increased the magnitude of ∆Torque
by adding mass (0.5 kg) to the participants' wrists. There were
two control conditions: one in which no extra mass was placed,
and one in which the extra mass was attached to the participants'
elbow. In this latter condition, the absolute amount of torque
needed to counteract gravity was increased, while ∆Torque was
not.
between sensory and motor areas of the brain of
anaesthetized rats and a simulated point mass moving in a
viscous medium. This can be achieved by programming the
interface for generating control policies in the form of force
fields acting on the controlled external device. We
implemented this behavior through a sensory interface
which maps the state of the device into a pattern of
electrical stimuli, and a motor interface which translates the
recorded neural activity into a force vector. The spiking
activity recorded from multiple units is converted into a twodimensional force vector by using Principal Component
Analysis (PCA).
However, a limitation of our previous work is that the
mathematical model for sensory and motor maps
implemented so far required the desired force field to be
invertible. This limited severely the range of control
problems that the interface can afford to address. Moreover,
the algorithms for extracting information from neural activity
that we used in this previous work were not yet able to
extract an optimal amount of information and this limited the
control that neurons could exert on the dynamical system
interacting with the brain.
Debats, N.B., Kingma, I., Beek, P.J. & Smeets, J.B.J (2010).
"Muscular torque can explain biases in haptic length perception: a
model study on the radial-tangential illusion". Lecture Notes in
Computer Science 6192, 392-397.
Here we present novel techniques that can overcome these
limitations and we compare the performances of the
resulting algorithms with respect to the original one on
simulated data. In particular we study the effect of
substituting PCA with Isomap, a nonlinear dimensionality
reduction technique able to find nonlinear degrees of
freedom. We also study a new version of the algorithm in
which the sensory interface is set such that the electrical
stimuli encoding neighbouring portion of space will elicit
more similar spike trains and electrical stimuli that are
encoding more distant portions of space will elicit more
diverse spike trains. We achieve this by computing
distances in terms of spike train metrics between stimulusevoked activities of recorded neurons, and by projecting
them onto the sensory domain by Multi Dimensional
Scaling. We then associate each stimulus to the center of
mass of its evoked spike trains and partition the space into
sensory regions with the nearest neighbour algorithm. The
motor interface then decodes the evoked activity by
computing the most likely stimulus eliciting it. The force that
is associated with the portion of space corresponding to the
decoded stimulus is then applied to the dynamical system.
2-G-65 Algorithms for shaping the dynamics of a
bidirectional neural interface
2-G-66 Transitions between rhythmic and discrete
performance in unimanual movements
Marianna Semprini1, François D Szymanski1, Francesco Grussu2,
Ferdinando A Mussa-Ivaldi3, Stefano Panzeri1, Alessandro Vato1
Hamal Marino1, Neville Hogan2, Marcos Duarte3, Steven
Charles4, Lauren DiPietro2, Dagmar Sternad5
We found no difference in illusion strength between the two
control conditions and a 7% increase in strength with an
increased ∆Torque. These results demonstrate that an increase
in ∆Torque is sufficient to alter the strength of the radial-tangential
illusion, whereas an increase in muscular torque alone is not. This
suggests that the perceived extent of arm movements is
influenced by the changing amount of torque that is needed to
keep the arm lifted during the movement.
1
Istituto Italiano di Tecnologia, 2University of Genoa,
Northwestern University
3
Brain Machine Interfaces (BMIs) offer a promising route towards
the recovery of motor functions in patients paralyzed by stroke or
spinal cord injury. While BMIs conventionally use the decoded
neural activity recorded in the motor cortex in order to control
actuators, bidirectional BMIs do also make use of
microstimulation to directly provide the brain with information
about the outside world. The establishment of a two-way
communication between brain and devices is crucial for restoring
motor functions after paralysis.
Our group has recently developed a novel computational and
experimental framework called dynamic BMI (dBMI), with the goal
of reproducing some of the mechanisms by which the spinal cord
operates the translation of cortical commands into motor
behavior. The emulation of spinal cord operations is achieved by
establishing and shaping the bidirectional communication
1
Scuola Superiore Sant'Anna, 2Massachusetts Institute of
Technology, 3University of Sao Paulo, 4Brigham Young
University, 5Northeastern University
While everyday actions flexibly combine rhythmic and
discrete movements, motor control research has largely
studied these movement types in isolation. Our previous
work provided some arguments and evidence from brain
imaging data that rhythmic and discrete movements may
form two distinct primitives. The present experiment used
behavioral data in a parametric scaling paradigm to test
whether abrupt transitions between the two types of
movement can be elicited. Abrupt transitions would further
support the hypothesis that these are two separate control
primitives. Ten subjects performed planar two-joint forearm
movements w on a low-friction horizontal surface, with their
wrist-joint immobilized, moving their arm between two large
targets (i.e., with a minimal end-point accuracy constraint).
73
Poster Sessions
Full Abstracts
The back-and-forth movements were paced by a metronome:
Starting with 20 sounds at 2 sec intervals, the metronome
intervals decreased uniformly by 36 ms each over the next 50
intervals to reach an interval of 200 ms. After a plateau with 20
sounds at 200 ms intervals the intervals lengthened again by the
same 36 ms over the next 50 intervals; after return to the initial 2
sec this interval was presented for a final 20 sounds. The
instruction to subjects was to end each movement by remaining
at rest in the target position for a duration (dwell time) equal to the
movement time. The experiment was performed under two
metronome conditions: each sound was of the same (brief)
duration or the duration of the metronome sound matched the
instructed dwell time. Results showed that subjects responded
imperfectly to the changes in interval. Despite the systematic and
predictable pattern of metronome intervals, subjects did not
anticipate the constant change correctly and consistently lagged
behind. Although dwell time was instructed to be equal to
movement time, it decreased substantially below that value--as
expected, subjects were unable to sustain discrete movements at
the fastest pace--and dwell time reduced to zero on average at a
metronome interval of 940 ms. These results were observed even
if information about dwell time was perceptually enhanced by the
duration of the metronome sound. Remarkably, the transition
between movements with non-zero dwell time and movement with
zero dwell time was abrupt. Acceleration profiles showed
discontinuous changes, with different subjects exhibiting different
transition points. Velocity profiles showed non-uniform changes in
symmetry and discrete relative phase between position and
acceleration peaks profile showed discontinuous changes.
Overall, results were consistent with the hypothesis that discrete
and rhythmic movements form two distinct movement types.
Support: NIH: R01 HD045639, NSF: BCS-0450218
2-G-67 Modeling gaze-dependent errors when reaching
to visual and proprioceptive targets
Joost Dessing1, Masahiro Kokubu1, Armin Abadeh1, Patrick A
Byrne1, J. Douglas Crawford1
1
York University
Reaching movements in the dark overshoot memorized visual
targets relative to the gaze direction held just before reach onset,
even if a saccade intervenes between target presentation and
reach onset. Traditionally, these errors have been suggested to
reflect misestimates of target position, arising when transforming
visual target position into body centered coordinates. Recently,
however, we showed that these gaze dependent reach errors
were entirely suppressed with online visual feedback of the hand.
This was most parsimoniously explained by a model in which
gaze dependent biases arise within the transformation of
proprioceptive hand position into gaze centered coordinates;
when vision of the hand is continuously available the brain would
mostly rely on accurate, visual hand position, thus eliminating the
reach errors. Here, we expand on these observations. We
included movement planning in multiple reference frames into our
model and allowed for all possible transformation errors, as well
as (visual and proprioceptive) workspace-dependent errors. To
constrain the model parameters, we will record pointing errors in
paradigms involving multiple target and fixation directions in a
variety of tasks (continuous visual targets or proprioceptive
targets [with or without online visual feedback]). Preliminary
findings for proprioceptive targets showed that gaze direction
influenced the reach errors (largely undershoots), while this effect
increased with online visual feedback for some subjects but
decreased for others. To account for these and other
observations, we will fit our model to the reach errors of individual
subjects. This will reveal the contribution of different biases within
the visuomotor transformation to the observed reaching
behaviour, thus pinpointing the relative role of target and hand-
74
related transformation biases. The individual variations in
the preliminary data for instance suggest that, besides
proprioception-to-vision transformation biases, some
subjects also show evidence of vision-to-proprioception
transformation biases.
2-G-68 Behavioral insights into neural mechanisms
of movement planning: Continuous and abrupt
updating of a motor plan following changes in task
goals
David Huberdeau1, Adrian M Haith1, John Krakauer1
1
Johns Hopkins University
Producing the commands to successfully reach a movement
goal requires a period of preparation prior to movement
onset, evidenced by the so-called "reaction time advantage"
whereby reaction time is lower if movement goals are known
in advance. Neurophysiological studies have revealed clear
neural correlates of movement preparation in motor and
premotor cortex. However, the underlying representations
and computations occurring during the period prior to
movement execution are not yet clearly understood. One
hypothesis is that the each potential movement corresponds
to a unique state of the motor cortex and the goal of the
preparatory period is to guide the motor cortical activity
towards the state that represents the best movement for the
present task (Churchland, 2010). An alternative hypothesis
is that multiple potential movements can be simultaneously
represented, with the preparatory period constituting a
period of competitive selection between potential
movements (Cisek, 2011). Here we sought to address this
question behaviorally by forcing subjects to generate
movements at intermediate stages of planning. We used a
timed response paradigm (Ghez, 1997) in which subjects
were trained to launch center-out reaching movements at a
pre-specified time. One of eight targets equally distributed
around a circle was visible for 1s prior to movement onset,
allowing ample preparation time for the movement. On a
subset of trials (15%), we jumped the target by 45° or 90° to
the left or right at a random time prior to movement onset.
This allowed us to vary the time permitted for movement
replanning. Subsequent movement kinematics were
measured and the initial launch angle computed. Fourteen
subjects participated in our experiment. We predicted that, if
movement preparation occurs through a continuous
adjustment of a single control policy, then at intermediate
stages of re-planning following a target jump, we would
observe initial movement directions that were intermediate
between the initial and jumped targets. Alternatively, if
multiple movements can be planned simultaneously, there
should be an abrupt switch from movements towards the
original target to movements aimed towards the new target
location as re-planning time increases. For 45° jumps, we
found that launch angle varied continuously as a function of
re-planning time. By contrast, for 90° jumps, launch angles
tended to clustered around either the initial or final target
directions, with no substantial intermediate behaviors. The
emergence of intermediate movements for incompletely replanned movements when changes in task goals were small
(45°), lends credence to the hypothesis that movement
planning occurs through continuous specification of a single
movement, rather than through competition between
multiple potential goals. However, the abrupt switching we
observed in movement kinematics for large changes in task
goals (90°) supports the notion that two distinct movement
plans can be held simultaneously. Churchland et.al. (2010)
Cortical preparatory activity: Representation of movement or
first cog in a dynamical machine? Neuron 68(3):387-400.
Poster Sessions
Full Abstracts
Cisek et.al. (2011) Neural correlates of biased competition in
premotor cortex. J Neuro Sci. 31(19):7083-88. Ghez, et. al.
(1997) Discrete and continuous planning of hand movements and
isometric force trajectories. Experimental Brain Research
115(2):217-33.
2-G-69 Heterogeneous attractor modules for motor
planning in macaque premotor cortex
Maurizio Mattia1, Stefano Ferraina2, Pierpaolo Pani2, Giovanni
Mirabella2, Stefania Costa2, Paolo Del Giudice1
1
Istituto Superiore di Sanità, 2Sapienza University
Cognitive functions like motor planning rely on the concerted
activity of multiple neuronal assemblies underlying still elusive
computational strategies. Here we show that in macaque
monkeys performing a reaching countermanding task, motor
plans coded in dorsal premotor cortex (PMd) can be detected as
multi-unit activity (MUA) patterns resulting from a network of
cortical modules14. We found sudden and stereotyped MUA
sharp transitions (STs), signalling an increase or a decrease in
the population firing, as a late reaction to target appearance, and
predicting at single trial level forthcoming actions. Occurrence of
such STs was observed even when movement was successfully
cancelled after a stop signal or during delay epochs of delayed
reaching task, excluding that they are the mere substrate of the
motor execution. We developed an attractor multi-modular
network of spiking neurons which accounted for STs and
predicted a peculiar modulation of high-frequency Fourier
components of unfiltered local field potentials (LFP). In vivo
observations confirmed such theoretical framework, providing a
strong evidence that local synaptic reverberation was in action,
making neuronal modules bistable. Our results demonstrate that
motor plans mature in PMd as an emergent cooperative
representation driven by a subset of "active" modules with strong
self-excitation, capable to amplify subthreshold input and to
recruit other "passive" modules with weaker synaptic local
feedback like in a chain reaction ending up in a stereotyped
distributed representation. What emerges is a general
computational machinery composed of a web of bistable modules
acting as interacting "flip-flops", as prescribed by the longstanding theory of associative networks.
2-G-70 The traveling salesman problem in human motor
tasks
Jakob Uecker1, Aldo Faisal1
1
Imperial College London
Finding the shortest route that connects a number of points is
known as the traveling salesman problem (TSP), which is known
for the exponential complexity of its computation. Previous
studies of human performance on the TSP in various cognitive
tasks (MacGregor et al., 1996; Vickers et al., 2001; Best, 2005)
have reported that subjects routinely solve problems of sizes up
to 20 points nearly optimally within a few minutes, with several
subjects finding the actual optimal solution. This is considerably
better than the performance of algorithms on present day
computers and arguably achieved with far less raw computing
power. Most of these studies have allowed subjects ample time to
evaluate and alter their solutions until satisfied (e.g. providing
pencil and eraser) and all have treated the problem as purely
combinatorial with the goal of optimizing Euclidean distance. In
this study, we present a novel experimental paradigm for testing
human performance in a visual motor TSP reaching task not
previously used in studies with human subjects. Specifically, we
set tight limits on motor planning and movement time to observe
how planning time scales with an increasing number of points.
Subjects had to solve a visually presented TSP instance in a
virtual reality environment by moving their dominant hand to touch
all targets as quickly as possible. Targets were generated at
random positions within a circle of 42cm radius. Each
subject (n=10) solved 800 trials, with each trial presenting 5,
10, 15 or 20 targets. Subjects received visual performance
feedback via a screen message after each trial -- positive if
they completed it in time smaller than 0.4 seconds per target
and negative otherwise. We find that subjects behave
optimally or near optimally for N=5 targets and that
performance decays sub-linearly for larger N. Performances
on the standard measure for optimization (total Euclidean
distance) are on average worse than those reported for nonmovement tasks in other studies, which can be explained by
the fact that the allowed time per trial differs by one to three
orders of magnitude as compared to those studies.
However, we find that those movement decisions of human
subjects which are suboptimal with respect to minimizing
distance have a predictable pattern. Based on this pattern
we propose that the Euclidean distance cost function may
not accurately capture the characteristics of good
performances in conditions of high density targets where the
optimal path features frequent changes of movement
direction that are atypical of human reaching movements.
Under these conditions, humans appear to minimize the
time needed for task completion by taking into account the
dynamic mechanical state of their arm: we find that
deviations from the distance-wise optimal movement
trajectory exhibit a consistent preference for straight-line
motion over sharp turns. We propose an alternative to the
Euclidean distance cost function and model the dynamics of
our experiment to evaluate both the performance benefits of
such a cost function and its accuracy in predicting subjects'
movements.
2-G-71 Towards the metabolic basis of the cost
function in human motor control
Scott Taylor1, Aldo A Faisal1
1
Imperial College London
Optimal feedback control theory - which derives control
policies from the minimisation of a cost function - has been
very successful in explaining human movements in a
principled manner (Todorov & Jordan, 2002; Diedrichsen &
Shadmehr, 2010). However, little is known about its
neuronal implementation or the cost functions used by the
motor system (Scott, 2004). Here we take a top-down
approach, by conducting reaching tasks and analyzing the
data. Crucially, we solve the inverse optimal control problem
in order to infer the cost function underlying human motor
coordination for planar reaching tasks and relate it to a
trade-off of two biophysically motivated cost functions:
signal-dependent motor noise (Harris & Wolpert, 1998) and,
as novelty, the metabolic cost of muscle activation.
Mammalian muscle can be modelled as populations of fast
and slow muscle fibres, which have characteristic metabolic
efficiencies (Reggiani et al. 2000) and activation regimes
(Tansey et al. 1996). At low muscle force levels slow muscle
fibres are predominantly activated with metabolically less
efficient fast fibres being activated only at higher force
levels. While the relationship between metabolic energy
consumption and force production in a single fibre is linear,
considering a population of muscle fibres allows the
derivation of a quadratic cost function dependent on the
ratio of fast to slow muscle fibres. Crucially, this fibre ratio
varies over a muscular system in a manner not related to
the size of motor pools (the leading explanation for
variations in signal dependent noise (Hamilton et al. 2004)),
suggesting a difference between optimal behaviours in
75
Poster Sessions
Full Abstracts
signal-dependent noise and metabolic efficiency derived cost
functions.
We model the human arm as a 6-muscle 2-joint system operating
under non-linear optimal feedback control (Todorov & Li, 2005).
The cost function is defined as a weighted sum of two biologically
motivated costs: signal dependent motor noise and metabolic
cost. These are evolutionarily sensible costs that maximise the
fitness of an organism's movements, minimizing task-relevant
variability and maximising energy efficiency. Simulated annealing
optimisation of the cost function was used to identify an optimal fit
with experimentally measured human behaviour. This allowed a
quantitative description of the cost function implemented in the
human motor system.
We have analysed experimental data collected in two motor
tasks: a dynamic centre-out reaching task and a static force
production task. In the dynamic task, subjects performed planar
reaching movements from the centre of the workspace to a target
in one of 8 directions. In the static task, subjects were asked to
produce a number of target forces as accurately as possible for
10 seconds. The force supplied by the subject was defined as the
sum of forces acting upon two perpendicular force sensors: one
at the wrist, and one just below the elbow. We use bootstrapping
to estimate the distribution of the cost function parameters. This
statistical technique shows that the observed changes in human
behaviour are indeed the result of interpolating between signal
dependent noise and metabolic cost. Greater understanding of
the costs minimised by the human motor system will allow us to
better describe the neuronal implementation of motor control
strategies. This work was supported by BBSRC.
p of 0.1 and 0.9. This created very different worst case
loads - a 9-fold difference wrt the b=0 baseline (b = 3σ vs
σ/3, as in our paradigm negative values of b actually
required less GF than b=0). Thus worst case scenario GF
control would predict changes from the baseline GF that
were 9 times stronger for the first environment, however,
variability-measure based control would predict equal GF
responses. We found essentially equal responses for the
two cases (2.33±0.38 vs. 2.36±0.33 Ns/m above baseline)
indicating variability-measure based control. We next
investigated the nature of the variability measure used by
the motor system by comparing GFs across 4 environments
with the same μ and σ but different high-order statistics: the
two skewed bimodal ones, a Gaussian and a balanced
bimodal (b of {-σ, σ}). Interestingly, we found that GFs were
the smallest for the environments with the largest high-order
moments of variability, and that a measure of variability
based on a moment of order 1.3 best explained the GF data
across groups; significantly better (p<0.02) than a statistic of
order 2 (corresponding to σ) or any higher order statistic.
Note that a measure of variability with order<2 can be
viewed as a more robust estimator of variability than σ
because of lower sensitivity to extremal conditions. Our
results indicate that environmental variability determines GF
control based on a robust measure of variability that
discounts higher order statistical moments.
2-G-73 Tapping along with ADAM: Synchronizing
with an adaptive and anticipatory virtual partner
Maria C van der Steen1, Merle T Fairhurst1, Peter E Keller1
2-G-72 Mechanisms for variability estimation in the
motor system
1
Maurice Smith1, Alkis Hadjiosif1
Interacting with a virtual partner (VP) is a relatively novel but
promising approach to investigate different aspects of
human coordination (Kelso et al., 2009; Repp & Keller,
2008). Interpersonal coordination in dynamic environments
relies on sensorimotor synchronization (SMS). SMS is the
temporal coordination of self-generated motor rhythms with
external rhythmical events (Repp, 2005). Temporal
adaptation and anticipation mechanisms seem to be
involved with SMS (Keller, 2008). Adaptation mechanisms
enable humans to modify the timing of their actions online
when synchronizing with external event sequences.
Temporal error correction processes, such as phase and
period correction, influence the timing of upcoming
movements and therefore facilitate the maintenance of
synchrony (Repp, 2005). Anticipatory mechanisms concern
predictions about the unfolding external event sequence
with which the action is to be coordinated. It is suggested
that these anticipatory processes are related to online action
simulation and internal models (Keller, 2008). The current
study explores the role of adaptation and anticipatory
mechanisms in a SMS task between human participants
and a VP. The VP is based on an ADaptation and
Anticipation Model, ADAM. ADAM is created in Simulink®
and combines an established formal model of adaptation
(phase and period correction) (Repp & Keller 2008) with an
anticipatory internal model. During a paced finger tapping
task, healthy participants interacted with the VP. The VP
produced an auditory pacing signal, and parametrically
adjusted the timing of this signal based on the human
participant's timing so as to reduce asynchronies (mismatch
between VP tone and human's action). Participants were
instructed to synchronize with the VP and to maintain a
stable tempo. Task conditions included different
combinations of adaptive and anticipatory processes, which
were modeled within ADAM to simulate human-like
1
Harvard University
In order to skillfully manipulate an object, for example a wine
glass, one must not only compensate for its dynamics but also
maintain a stable grasp avoiding slippage. The minimum grip
force (GF) required to prevent slip is a function of both the glass's
weight & the friction coefficient between its surface & our fingers.
A safety margin above this minimal GF is often maintained,
allowing us to maintain a grasp robust to unexpected
perturbations, misestimation of load forces, or errors in GF
production. Here we hypothesize that this safety margin is
determined by the uncertainty about environmental dynamics: if
the motor system is certain about its load force estimate it can
afford to keep the safety margin low without risking slip, but a
large safety margin would be beneficial if uncertainty is high. To
test this hypothesis we designed a task where participants
performed point-to-point reaching movements while grasping a
small object with dynamics that changed randomly from one trial
to the next. These dynamics were characterized by a viscous
force-field of strength b, chosen from a Gaussian PDF with
standard deviation (σ). We found that GFs increased when σ was
higher (p<0.001); in fact GFs were twice as sensitive to σ as to
the mean of b. This sensitivity of GFs to environmental variability
could result from the motor system adjusting GFs under a worstcase scenario: trying to avoid slip against the highest loads in
recent memory. Alternatively, GFs could be directly programmed
based on a measure of environmental variability in combination
with an estimate of the mean environment. To distinguish
between these possibilities, we compared GF adaptation in two
environments with different PDFs for b. These PDFs were
oppositely skewed but shared the same μ & σ, and had equal
values for the amplitudes of all higher order moments about the
mean. In the first, b took on values of -σ/3 and 3σ with p of 0.9
and 0.1, whereas in the other b took on values of -3σ and σ/3 with
76
Max Planck Institute for Human Cognitive and Brain
Sciences
Poster Sessions
Full Abstracts
behavior. Different weightings of the two mechanisms within the
model were tested to probe subsequent effects on participants'
tapping behavior. Variables of interest include tapping accuracy
(mean absolute asynchronies) and tapping variability (standard
deviation of asynchronies). The relative importance of these two
mechanisms will be discussed in terms of the variable effect on
SMS.
2-G-74 Coordinate dependence and distribution
dependence of blind source separation for motor
synergies: Robust separation behavior of ICA across
transforms
Corey Hart1, Simon Giszter1
1
Drexel University College of Medicine
There has been some recent controversy over the extent to which
different variance-based and other statistical decomposition
techniques can detect fundamental motor synergies or primitives.
For example, are techniques applied to EMG activity capable of
uniquely isolating functional related groups of muscles. Hogan
and Sternad have argued that often these relationships are
merely a consequence of an arbitrary choice of scale and
coordinates for motor data. In a related argument, it is also
possible to maintain that even information-based algorithms such
as independent component analysis (ICA) can be made to fail by
applying a highly nonlinear mapping to the data such that the
resultant transformed data are very nearly normal. ICA will fail for
the transformed Gaussian data, as ICA, by definition, cannot be
applied to Gaussian distributed data sets. We set out to
determine just how fragile ICA is and how much of a problem
such re-mappings are for a typical implementation of ICA (i.e.,
Makeig-Sejnowski's infomax ICA algorithm). This study relied on
artificial mixture data generated in the MATLAB ? programming
environment. Two kinds of data were generated. Case 1: Data
were generated from mixtures of ten source channels of 10000
points each, drawn from Gaussian pseudo-random data
generated within MATLAB. Random mixing matrices were
multiplied by the source channels, resulting in a 10x10000 matrix
of mixtures. These mixtures were then either (a) given to the ICA
algorithm for identification of the mixing matrix, or (b) scaled by an
exponential function and then supplied to the ICA algorithm. Case
2: Data were generated from sources that were initially nonnormal (created by first taking Gaussian values and then applying
the exponential function to them prior to mixing, yielding a
10x10000 source matrix with a more-or-less log-normal
distribution in data of sources. Data were then mixed as above
and given to the ICA algorithm for factoring (c). To examine the
effect of "warping" the initially non-Gaussian Case 2 data back
toward a normal distribution, then we used the fmincon() function
in MATLAB to find an optimal polynomial function that minimized
both kurtosis and skewness of the warped mixture data (resulting
in a more or less Gaussian transform of the mixed data). In these
simulations ICA unmixing was surprisingly much better than
anticipated, even for very small deviations from true Gaussian in
Case 1 'normal data'. Although ICA will in principle fail for truly
Gaussian distributed data, almost any real world example will
have small deviations from true normality. Furthermore, any
invertible transform used in an attempt to rescale the mixed data
in such a way as to make it normal (as in Case 2), only 'crushes'
the extra information that ICA relies on into smaller and smaller
bins in the analysis. This does not destroy it, and this information
can still be successfully resolved with sufficient computing power
and time, allowing identification and unmixing of the original
information-based groups. We conclude ICA is significantly more
robust and immune to the coordinate-dependence concerns
voiced by Hogan and Sternad than many other techniques in use,
and thus less subject to this critique.
2-G-75 Influence of arm velocity on haptic localization
Femke Maij1, W. Pieter Medendorp2, Alan M Wing1
1
University of Birmingham, 2Donders Institue for Brain,
Cognition and Behaviour, Radboud University Nijmegen
Different sensory signals are integrated in our brain to
construct a perceptual representation of the world. Here we
study how the brain integrates tactile and proprioceptive
signals by investigating systematic spatial errors that people
make when localizing briefly presented tactile stimuli on
their index finger while making arm movements. We
hypothesize that spatial localization errors originate from
uncertainty about the time of the tactile stimulus with respect
to the movement. Subjects moved their extended right arm
horizontally through the air. At different times before, during
and after the movement they felt a tactile stimulus. After the
arm movement subjects indicated where they had perceived
the tactile stimulus in space. Varying the velocity of the arm
movements resulted in different patterns of localization
errors. We explain our results with the hypothesized
temporal uncertainty model.
2-G-76 Co-articulate of straight movements with an
artificial neural network
Andre Lemme1, Yaron Meirovitch2, Tamar Flash2, Jochen
Steil1
1
Research Institute for Cognition and Robotics, 2Weizmann
Institute of Science
In our daily lives, we perform tasks that require complex
movements such as drawing or handwriting. It has been
suggested that complex movements are composed of
simpler movement primitives, which represent basic
movement units or simple strokes, characterized by (e.g.,in
the case of reaching movements) bell-shaped-velocity
profiles[1].The notion of co-articulation suggests the
possibility that with extensive training a new movement
primitive can emerge [2]. The process may start with simple
straight movements which are eventually substituted with
more complex curved primitives.
The contribution of the present work is in proposing a
flexible learning setup that is able to generate biologically
plausible movements based on co-articulation of simple
movement primitives.
Methods: The used learning module is a high-dimensional,
single-layer feed-forward network. The weights are
initialized randomly and only the output weights are
obtained by linear regression, called Extreme Learning
Machine (ELM) [3].The ELM represents the movement
primitive as a mapping from position to velocity. To generate
a movement the velocity is integrated over time to
reproduce the learned basic (curved)strokes. We
qualitatively compare the learned new primitives to velocity
profiles from human data and to the predictions of the
Minimum Jerk Model[4], which plans a trajectory starting
from a given starting point to a given end-point through a
via-point. The constraint for planning the trajectory is to be
as smooth as possible (minimum jerk). This module was
also used in[2] and showed to provide a good approximation
of the movements emerging from human co-articulation, but
no learning algorithm was proposed in that study.
Learning setup: The algorithm starts from a set of straight
movements each having a bell shaped velocity profile and
form a new curved movement primitive. The task setup has
three targets (ABC), which need to be connected. The
algorithm repeatedly connects the three points with straight
strokes.The learning paradigm requires a certain amount of
77
Poster Sessions
Full Abstracts
variance in the data. Therefore Gaussian noise was added to the
position of target B. These recorded trajectories were
subsequently used as training data for the ELM.
Further more we implement an iterative learning process, in which
the learning module creates its own training data to further evolve
the primitive. During these iterations we follow the trajectories as
the learning progresses and compare this progress to the one
observed in human subjects. The result of our work is a learning
mechanism that simulates the co-articulation process observed in
human subjects with quantitatively similar velocity profiles.
Acknowledgment: Many thanks to Ronen Sosnik who provided
the human data to this work. This work was supported by the
European Communitys FP7 under grant agreement No 248311 AMARSi
Ref:[1]W.Abend, E.Bizzi,and P.Morasso. Human arm trajectory
formation. Brain: a journal of neurology, June 1982. [2]R.Sosnik,
T.Flash, B.Hauptmann, and A.Karni.The acquisition and
implementation of the smoothness maximization motion strategy
is dependent on spatial accuracy demands. Experimental Brain
Research, 2007. [3]Guang-Bin Huang and Chee-Kheong Siew.
Extreme learning machine:Theory and applications.
Neurocomputing,2006.[4]T.Flash, E.Henis, R.Inzelberg, and
A.D.Korczyn. Timing and sequencing of human arm
trajectories:Normal and abnormal motor behaviour. Human
Movement Science,1992
2-G-77 Task-dependent structure of neuronal variability
during abstract BMI control
Kianoush Nazarpour1, Tom M Hall1, Andrew Jackson1
1
Newcastle University
Primates can volitionally modulate neuronal activity according to
arbitrary reward rules and learn to control Brain-Machine
Interfaces (BMIs) with randomized decoders. We have previously
shown that in redundant myoelectric-control tasks, subjects
improve accuracy by buffering trial-to-trial variability into taskirrelevant dimensions of the muscle space. However it is not
known whether variability in the neural space can also be shaped
in a task-dependent manner during control of a redundant BMI.
We therefore investigated the structure of neural noise covariation
in two macaque monkeys performing a simple brain control task
involving two neurons, C1 and C2. Over separate sessions, 81
pairs of neurons were sampled from chronically-implanted
electrode arrays in M1 and PMv. The one dimensional position of
a cursor was determined by the instantaneous firing rate of C1
alone or the sum (C1 C2) or difference (C1-C2) of the firing rates.
The monkeys' task was to hold the cursor for 500ms in targets
that appeared in one of four different locations.
We calculated trial-to-trial variability of the firing rate trajectory in
neural space. Trajectories for each trial were aligned to the first
time that cursor entered the target. To remove any structure that
this alignment imposed on the trajectory distribution, we
characterised the evolution of neural variability relative to the
point in neural space at which the cursor entered target.
Variability was calculated along dimensions of positive (C1 C2)
and negative (C1-C2) covariation and defined an index of
covariation (IoC) to quantify correlation structure. A zero IoC
indicated a uniform pattern of variability while the limiting values
of 1/-1 imply the variability is entirely constrained to axes of
positive/negative covariation in the neural space. IoC was
evaluated up to 500ms after the alignment point.
We found an effect of task condition on neural IoC. Overall, the
distribution of IoC tended towards positive values, consistent with
a general pattern of positive noise correlation between neurons.
During the C1-C2 control block, IoC was increased further
78
suggesting greatest neuronal variability in the dimension of
positive correlation (the task-irrelevant dimension). In
contrast, IoC was, on average, slightly negative in the C1
C2 control condition. IoC during control by C1 alone fell
between that of C1-C2 and C1 C2 conditions. This
modulation of IoC suggests that trial-to-trial variability was to
some extent shaped to reflect the task-relevant and irrelevant dimensions of the neural space.
It remains to be seen whether the task-dependent structure
in neuronal variability emerges from divergence in feedforward inputs or feed-back correction of errors. In either
case, our results suggest that redundancy in abstract BMIs
can be exploited to improve accuracy of control by buffering
variability into task-irrelevant dimensions of the neural
space.
2-G-78 A common motor optimisation principle in
healthy subjects and parkinsonian patients
Pierre Baraduc1, Stéphane Thobois2, Jing Gan2, Emmanuel
Broussolle2, Michel Desmurget1
1
CNRS / U. Lyon, 2Lyon Pierre Wertheimer Neurological
Hospital
Recent research on Parkinson's Disease (PD) has
emphasized that parkinsonian movement, although
bradykinetic, shares many attributes with healthy behavior.
This observation has led to the suggestion that bradykinesia
in Parkinson's Disease could be mostly due to a reduction in
motor motivation. This hypothesis can be investigated in the
framework of the optimal control theory which models
account for many characteristics of healthy movement, while
providing a link between the motor behavior and a
cost/benefit tradeoff.
Here we studied 14 PD patients with bilateral subthalamic
nucleus (STN) stimulation and 16 age-matched healthy
controls and asked whether reaching movements are
subserved by similar execution rules in these two groups,
but with a different allocation of effort. We show that a single
optimal control model accounts for the reaching movements
of healthy subjects and PD patients, whatever the condition
of STN stimulation (On or Off). The choice of movement
speed was explained in all subjects by the existence of a
preset dynamic range for the motor signals. This range was
idiosyncratic and applied to all movements irrespective of
their amplitude. In PD patients this dynamic range was
abnormally narrow and correlated with bradykinesia. STN
deep brain stimulation widened this range in all patients, but
did not restore it to a normal value.
These results, consistent with the motor motivation
hypothesis, suggest that optimization of motor effort is the
main determinant of movement production in both healthy
and PD subjects.
2-G-79 Arm-EMG control for assistive lower limb
exoskeleton
Francesca Sylos Labini1, V La Scaleia1, M J MacLellan1, T
Hoellinger2, K Seetharaman2, M Petieau2, A Bengoetxea2, G
Cheron2, Y P Ivanenko1
1
Santa Lucia Foundation, 2Université libre de Bruxelles
Success in locomotor rehabilitation programs can be
improved with the use of brain-computer interfaces.
Although a wealth of research has demonstrated that
locomotion is largely controlled by spinal mechanisms, the
brain is of utmost importance in monitoring or shaping
locomotor patterns in humans. In addition, there is also a
Poster Sessions
Full Abstracts
tight coordination between the upper and lower limbs, which can
also be useful in controlling locomotion. The current study
critically investigates different approaches for using upper limb
electromyogram (EMG) that are applicable to this field of
controlling assistive lower limb exoskeletons. The control may be
based on the spatiotemporal aspects of arm muscle activity or
using dynamic recurrent neural networks (DRNNs). Taking into
account the high involvement of some shoulder muscles (e.g.,
deltoid) in most locomotor-related movements in humans we took
advantage of this natural coordination for the control of leg
kinematics. It is important to emphasize that these muscles are
naturally involved in locomotion, in contrast to distal (wrist or
finger) muscles that are more cortically controlled during
locomotion. Once properly parameterized, DRNN may acquire
independent knowledge patterns (attractor states) representing
EMG-kinematics identification. The expected final output
corresponds to the kinematics-related signals of the intent
movement that will be exploited as an input to the exoskeleton
controller combined with virtual reality training of individuals with
reduced mobility, such as spinal cord injury patients. We
investigated different experimental conditions for the DRNN
procedure in healthy subjects when using this approach. In
conditions of prominent arm muscle activity (such as during fast
walking speeds or enhanced arm swinging) a high
correspondence between original and predicted leg kinematics
was observed. The current work is aimed at integration of DRNN
with virtual reality locomotor training and exoskeleton supervisor
controller. The aim of the clinical validation for the arm-EMG
control approaches would be to evaluate effectiveness, usability
and comfort of this control for patients. Supported by EU FP7-ICT
program (MINDWALKER grant).
These new methods make it possible to construct optimal
control models for more complex behaviors than what is
currently possible. How well such models will agree with
experimental data remains to be seen, and indeed there are
many different ways to construct them which will yield
different predictions. The work presented here aims to
enable the construction of such models.
This work is a summary of two recent technical papers
whose PDFs can be found on our website:
Synthesis and stabilization of complex behaviors through
online trajectory optimization
Tassa Y, Erez T and Todorov E (2012), under review
Discovery of complex behaviors through contact-invariant
optimization
Mordatch, Popovic and Todorov (2012), to appear in ACM
SIGGRAPH
2-G-80 Automatic synthesis of complex behaviors with
optimal control
Emo Todorov1, Igor Mordatch1, Yuval Tassa1, Tom Erez1, Zoran
Popovic1
1
Computer Science and Engineering, University of Washington
In this video poster, we will show animations of complex motor
behaviors synthesized automatically using new optimal control
methods, as well as explain how these methods work. The
behaviors include getting up an arbitrary pose on the ground,
walking, hopping, swimming, kicking, climbing, hand-stands, and
cooperative actions. The synthesis methods fall in two categories.
The first is online trajectory optimization or model-predictive
control (MPC). The idea is to re-optimize the movement trajectory
at every step of the estimation-control loop, up to some time
horizon (in our case about half a second), execute only the
beginning portion of the trajectory, and repeat the re-optimization
at the next time step (say 10 msec later). This approach has been
used extensively in domains such as chemical process control
where the dynamics are sufficiently slow and smooth to make
online optimization possible. We have now developed a number
of algorithmic improvements, making it possible to apply MPC to
biomechanical systems. The second method is based on the
realization that most movements performed on land are made for
the purpose of establishing contact with the environment, and
exerting contact force. This suggests that contact events should
not be treated as side-effects of multi-joint kinematics and
dynamics, but rather as decision variables that the controller can
reason about directly. We have developed a method in which the
optimizer can explicitly specify the desired contact events, using
continuous decision variables that facilitate optimization, and at
the same time optimize the movement trajectory in a way
consistent with the specified contact events. This makes it
possible to optimize movement trajectories with a large sequence
of contact events which are discovered automatically.
79
Poster Session Floor Plans
Venetian Ballroom A & B & Foyer
Food
Foyer
9 10 11
12 13 14 15 16 17 18 19 20 21 22
44 43 42 41 40 39 38 37 36 35 34
33 32 31 30 29 28 27 26 25 24 23
1
Entrance
89
105
90
91
103
92
102
93
Exhibit Tables
104
3
4
5
6
7
8
Food
106
2
Venetian Ballroom
A&B
To Foyer
45 46 47 48 49 50 51 52 53 54 55
56 57 58 59 60 61 62 63 64 65 66
88 87 86 85 84 83 82 81 80 79 78
77 76 75 74 73 72 71 70 69 68 67
Coffee
Food
Venetian Ballroom C, D & E
THEMES
94
Poster Layout
100
95
Session 1
99
96
98
97
101
All meeting sessions will be held here.
A - Adaptation & Plasticity in Motor Control
B - Integrative Control of Movement
C - Control of Eye & Head Movement
D - Disorders of Motor Control
E - Posture & Gait
F - Fundamentals of Motor Control
G - Theoretical & Computational Motor Control
Tuesday, April 24 & Wednesday, April 25
Venetian Ballroom A & B and Foyer
Foyer
Food
Foyer
9 10
11 12 13 14 15 16 17 18 19 20
40 39 38 37 36 35 34 33 32 31
30 29 28 27 26 25 24 23 22 21
1
Entrance
2
3
4
5
6
7
Food
Venetian Ballroom
A&B
To Foyer
41 42 43 44 45 46 47 48 49 50
51 52 53 54 55 56 57 58 59 60
80 79 78 77 76 75 74 73 72 71
70 69 68 67 66 65 64 63 62 61
81
92
91
82
90
83
Exhibit Tables
8
Coffee
Food
Venetian Ballroom C, D & E
84
Poster Layout
88
85
Session 2
87
86
89
Friday, April 27 & Saturday, April 28
Venetian Ballroom A & B and Foyer
Foyer
All meeting sessions will be held here.
THEMES
A - Adaptation & Plasticity in Motor Control
B - Integrative Control of Movement
D - Disorders of Motor Control
F - Fundamentals of Motor Control
G - Theoretical & Computational Motor Control
The 22nd Annual NCM Meeting
thanks our sponsors, supporters and donors
`