Human Movement Science 44 (2015) 2231

Contents lists available at ScienceDirect

Human Movement Science

journal homepage: www.elsevier.com/locate/humov

Task-specificity of balance training

Louis-Solal Giboin , Markus Gruber, Andreas Kramer
Sensorimotor Performance Lab, Department of Sport Science, Universitt Konstanz, 78464 Konstanz, Germany

a r t i c l e

i n f o

Article history:
Received 5 May 2015
Revised 30 July 2015
Accepted 17 August 2015
Available online 25 August 2015
Keywords:
Sensorimotor training
Motor learning
Skill
Transfer
Postural control

a b s t r a c t
Despite much research on balance training, it is still unclear whether balance training leads
to highly task-specific adaptations or rather non-specific adaptations. Hence, in this study
we examined whether balance training increased performance only in the balance task
that was trained or also in non-trained tasks. Forty healthy participants (28 m 12 f,
25 4 years, 177 10 cm, 73 14 kg) were assigned to one of two training groups (TGs)
or a control group. Both TGs completed six sessions over 2 weeks, only the training device
differed. Before and after the training, performance in the trained task as well as in
additional untrained tasks was recorded. ANOVAs showed that each TG outperformed
the other groups only in the task they had trained (e.g., task trained by TG1: +225% in
TG1, only +41% and +30% in TG2 and control, group * time interaction, p < 0.001;
Untrained task 1: TG1 +48%, TG2 +48%, and control +30%, no significant interaction,
p = 0.72). In summary, 2 weeks of balance training resulted in highly task-specific effects,
no transfer even to very similar tasks was observed. Therefore, we recommend identifying
and training exactly those tasks that need improvement, and test the efficacy of training
programs using specific tests instead of general tests with limited functional relevance.
2015 Elsevier B.V. All rights reserved.

1. Introduction
Balance training has been recommended to improve performance in different sports (Zech et al., 2010), to prevent injuries
and accelerate the rehabilitation process (McGuine, Greene, Best, & Leverson, 2000; McKeon & Hertel, 2008), as well as to
prevent falls in at risk populations (Gauchard, Jeandel, Tessier, & Perrin, 1999).
However, it is still not clear whether balance training leads to highly task-specific adaptations, or more general
non-specific adaptations that can be transferred to other tasks. This is a fundamental question, as the answer will greatly
influence balance training recommendations as well as study design and testing procedures in future balance studies, and
can help to direct the investigation of the underlying physiological mechanisms of balance training adaptations. For instance,
if the effects of balance training were primarily task-specific, it would be mandatory to use task-specific tests in order to be
able to detect specific neurophysiological adaptations.
In the literature, balance is often treated as a general ability. In consequence, balance is usually assessed using generic
tests (e.g., one-leg stance) irrespective of training goals and type, and meta-analyses pool balance studies regardless of
the type of balance training used (Lubetzky-Vilnai & Kartin, 2010; Tofthagen, Visovsky, & Berry, 2012). This approach is supported by studies that report non-specific effects of balance training such as improved gait kinematics in stroke patients

Corresponding author at: FG Sportwissenschaft, Universitt Konstanz, 78457 Konstanz, Germany.

E-mail addresses: louis-solal.giboin@uni-konstanz.de (L.-S. Giboin), m.gruber@uni-konstanz.de (M. Gruber), andreas.kramer@uni-konstanz.de
(A. Kramer).
http://dx.doi.org/10.1016/j.humov.2015.08.012
0167-9457/ 2015 Elsevier B.V. All rights reserved.

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

23

(Yavuzer, Eser, Karakus, Karaoglan, & Stam, 2006), static stability in patients suffering from spinal cord injuries (Sayenko
et al., 2010), postural corrective responses to unexpected balance perturbation in healthy subjects (Sayenko et al., 2012),
or walking speed in older adults (Agmon, Perry, Phelan, Demiris, & Nguyen, 2011) and even on shuttle-run time and body
sway during different postural tasks in young recreationally active adults (Yaggie & Campbell, 2006). On the other hand, several studies indicate that the principle of task-specificity also applies to balance, postural control and gait rehabilitation in
old adults (Muehlbauer, Besemer, Wehrle, Gollhofer, & Granacher, 2012; Oddsson, Boissy, & Melzer, 2007), or that no effects
were found when the task trained was different from the task tested in old adults (Donath, Roth, Zahner, & Faude, 2015;
Kovacs, Sztruhar Jonasne, Karoczi, Korpos, & Gondos, 2013; McMurdo, Millar, & Daly, 2000) and in children (Donath et al.,
2013). Furthermore, several attempts have been made to divide balance into many subcategories that seem to have little
overlap. For example, the BESTest a clinical balance test battery developed by Horak, Wrisley, and Frank (2009) consists
of 36 tests in six categories, and it has been shown that patients with deficits in one category score poorly in that category
but not in the other categories, suggesting that there is no such thing as a general balance ability and that one should not
expect that training one balance task improves performance in a different balance task (Horak et al., 2009)).
However, none of the studies mentioned above can and were not intended to directly answer the question about
task-specificity of balance training adaptations, as they usually had only one training group and considerable differences
between the task that was trained and the tasks that were tested. To the best of our knowledge, the present study is the first
one that explicitly aimed to answer the question whether balance training leads to task-specific or rather general adaptations. We hypothesized that training one balance task would lead to performance improvements only in this task and not
in other balance tasks that had not been trained.
2. Methods
2.1. Study design
During a first set of tests (PRE), balance was assessed with two different balance devices, a Posturomed (Haider Bioswing,
GmbH Pullenreuth, Germany) and a custom-made tilt board (see Fig. 1), both used with two different directions of perturbation (antero-posterior AP, and medio-lateral ML) resulting in a total of four different tests. Participants were then divided
into three groups using a matching procedure picking the permutation with the least group mean differences in the performance in the four PRE tests via a Matlab script to ensure comparable group means: one control group that did not train
(C group, 12 subjects, 23 3 years, 175 8 cm, 71 10 kg), one group which trained only on the Posturomed in ML direction
(P-ML group, 14 subjects, 26 5 years, 176 12 cm, 70 14 kg) and one group which trained only with the tilt board in ML
direction (T-ML group, 14 subjects, 24 3 years, 178 11 cm, 75 16 kg). The C group was smaller than the other two groups
due to the drop-out of two subjects for reasons not related to the study (knee injury and scheduling problems, respectively).
After six training sessions, all participants were tested again on both devices in both directions (POST test).
2.2. Participants
The experiments were approved by the ethics committee of the University of Konstanz and were in accordance with the
latest revision of the Declaration of Helsinki. The 40 participants (12 female, 28 male, mean age 25 4 years, height
177 10 cm, body mass 73 14 kg) were mostly students with an average physical activity level of 5 h per week (assessed
via questionnaire). All participants were healthy with no history of leg injuries during the last 2 years or diseases associated
with balance impairments. All participants gave written informed consent before taking part in the study and were told not
to participate in any other balance training for the course of the study.
2.3. Balance devices
As shown in Fig. 1A D, the custom-made tilt board consisted of a wooden platform (24.5  24.5  1 cm) with a gripping
surface, mounted on a semicircular wooden block with a height of 6.5 cm. The aim was to bring the tilt board into a horizontal position and maintain it while standing on it with one leg. The Posturomed consisted of a movable platform that
is connected to a metallic structure by dampened pendulums (see Fig. 1E and F). The platform can be kept in a stable position
2.5 cm away from its centre via an electromagnet. From this starting position, the platform can be released by switching off
the electromagnet, thus inducing a damped oscillation with an initial amplitude of 5 cm. The subjects task was to reduce the
oscillations of the platform both the oscillations caused by the release of the platform and the oscillations caused by his
own movements as fast and completely as possible while standing on one leg. The two devices were chosen because they
are very common balance training and testing devices.
2.4. Balance tasks
During the PRE and POST tests, each participant had to complete four balance tasks: (i) on the tilt board with the
boards rotational axis aligned with the longitudinal axis of the foot, resulting in a medio-lateral direction of

24

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

Fig. 1. Illustration of the four balance tasks. (A) T-ML task, starting position. (B) T-ML task, equilibrium position. (C) T-AP task, starting position. (D) T-AP
task, equilibrium position. (E) P-ML task, the arrow indicates the initial direction of the platform after the release, it corresponds to the medio-lateral
direction of the subject. (F) P-AP task, the arrow indicates the initial direction of the platform after the release, it corresponds to the antero-posterior
direction of the subject.

perturbation (Fig. 1A and B, T-ML task); (ii) on the tilt board with the boards rotational axis perpendicular to the longitudinal axis of the foot, resulting in an antero-posterior direction of perturbation (Fig. 1C and D, T-AP task); (iii) on the
Posturomed with the longitudinal axis of the foot perpendicular to the release direction of the platform, resulting in a
medio-lateral direction of perturbation (Fig. 1E, P-ML task); (iv) on the Posturomed with the longitudinal axis of the foot
aligned with the release direction of the platform, resulting in an antero-posterior direction of perturbation (Fig. 1F, P-AP
task).
During PRE and POST tests as well as during the training, the subjects were using always the same leg (their self-selected
preferred leg for one-leg stance). The subjects performed the balance tasks during the tests and the training without shoes.
For the tilt board, the subject always had to start with the same side of the board on the floor (supinated starting
position). The subject had to put the foot of the preferred leg on the board, lift the other foot off the ground and then bring

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

25

the tilt board into a horizontal position and maintain this horizontal position for as long as he was able to during each trial. If
the subject lost equilibrium, he had to let the boards edge touch the ground before putting the free leg on the ground.
Then he had to lift his leg off again before trying again to put the tilt board into a horizontal position. The proper execution
was controlled by the experimenter and corrected if necessary.
For the Posturomed, the participant was standing on his preferred leg, with his hands on his hips and his gaze fixed on a
mark on the wall at eye level. For a trial to be valid, the free leg had to never touch the platform or the standing leg. Invalid
trials were not repeated and counted as 0 s out of 20 s of equilibrium. The latency of the release of the platform was always
randomised so that the subject could not anticipate the exact moment of release.
2.5. Rationale
The rationale of choosing the study design outlined above was as follows: if training one specific balance task results
in improvements only in this task and very similar tasks, the T-ML group should improve in the T-ML task but not the PML task and vice versa. Therefore, testing these two tasks would already be sufficient to confirm our hypothesis.
However, to further strengthen the study design, we included two additional tasks that none of the groups had trained.
The T-AP task was very similar to the T-ML task, only the direction of perturbation differed. Hence, the T-ML group
should not outperform the two other groups if balance training effects are highly specific. The P-AP task served a slightly
different purpose: although it seems to differ from the P-ML task in the same way as T-AP from T-ML, this is only true
for the initial perturbation; once the initial perturbation is compensated for, the P-AP and P-ML tasks are the same (see
Fig. 3). Therefore, the P-ML group should outperform the other groups in the second phase of the task (when the P-AP
task becomes the same as the P-ML task they trained), but not in the first phase (when the P-AP task is different from
the P-ML task).
2.6. Testing procedure
The PRE and POST tests consisted of assessing balance in each of the four tasks described above (T-AP, T-ML, P-AP and
P-ML). The order of the tasks tested was counter-balanced across participants i.e., each task was the first, second, third
and last task in the test order for an equal amount of times but was kept the same in the PRE and POST tests for each participant. For every task, there were 5 trials of 20 s each, separated by 20 s of rest. The first 2 trials served as familiarisation
trials. Only the last 3 trials (test trials) were used for performance assessment. Between each block of 5 trials, the subject
rested for 3 min.
2.7. Training
Subjects from the group P-ML trained only the P-ML task and the group T-ML trained only the T-ML task. For both groups
the training design was the same: 6 training sessions were completed over 2 weeks, with at least 1 day of rest between two
sessions. Each training session consisted of 20 trials, divided into four blocks of 5 trials each. Each trial lasted 20 s with 10 s of
rest in between. There was 1 min of rest in between blocks, so that the total time for each training session was approximately
15 min.
2.8. Data capturing and data processing
Four reflective markers were placed on the corners of the tilt board, so that the position and angle of the board
could be recorded and analysed with a motion capture system at 200 Hz (Vicon Nexus, 12 T-series T40s cameras).
The performance during a trial was evaluated by calculating the amount of time the tilt board was within a margin of 5 of the horizontal position (i.e., when the angle of the board with respect to the floor was between 5
and +5, see Fig. 2A). The performance for each task was defined as the mean of the three recorded trials (T-ML
and T-AP).
The movement of the platform of the Posturomed was recorded with a universal-mounted two-dimensional potentiometer (see Figs. 1EF and 2BC), and the resulting analogue signals were acquired with a data acquisition unit (Power1401 mk
II, Cambridge Electronic Design, Cambridge, UK; sampling frequency of 5000 Hz), controlled by the software Signal (Signal
5.08, Cambridge Electronic Design, Cambridge, UK). The traces of the x and y positions of the platform of the Posturomed
were analysed with Matlab (Matlab R2011a, Mathworks Inc., Natick, USA). The time at equilibrium during a trial was defined
as the sum of the time intervals that the platform deviation from its centre position was less than 0.2 cm in x direction and
less than 0.2 cm in y direction for at least 0.5 s (so that merely swinging through the neutral position while not in balance did
not count towards the time at equilibrium). The performances of the 3 trials of a subject were then averaged for each of the
two tasks (P-ML and P-AP). In addition, for the P-AP task the time to reach the first phase of equilibrium as defined above was
calculated and also averaged across the 3 trials.

26

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

Fig. 2. Exemplary raw data before and after the training. Raw data from one subject from the T-ML group demonstrating the performances for T-ML task (A)
and one subject from the P-ML group demonstrating the performances for the P-ML task (B and C) during PRE and POST tests. The thin line corresponds to
the performance of 1 trial of 20 s during the PRE test and the thick line corresponds to the performance of 1 trial of 20 s during the POST test. (A) Angle of the
tilt board with the subject balancing himself in medio-lateral direction (T-ML task). An angle of 0 means that the tilt board is horizontal. The dashed lines
(ordinate of +5 and 5) correspond to the upper and lower limit of the defined equilibrium range of the tilt board. (B and C) Displacement of the platform
of the Posturomed in the subjects medio-lateral direction (B) and in the antero-posterior direction (C).

2.9. Statistical analyses

To assess performance differences due to the training in the four tasks, analyses of variance (ANOVA) with repeated measures were performed for each of the four tasks (T-ML, T-AP, P-ML, and P-AP), using group (T-ML group, P-ML group and C
group) as an inter-subject factor and time (PRE and POST) as a repeated measure. The same analysis was applied to the time
to reach the first equilibrium in the P-AP task. To assess interaction effects between the four tasks and the group affiliation,
one ANOVA with repeated measures was performed on the performance changes (i.e., performance POST minus performance
PRE), using group (T-ML, P-ML and C group) as an inter-subject factor and task (T-ML, T-AP, P-ML and P-AP) as a repeated
measure. To follow up on significant differences, Bonferroni-corrected t-tests were used and Glasss delta was calculated as
an effect size for these pairwise comparisons.
All figures following a symbol represent standard deviations. Statistical tests were performed with SPSS (SPSS 19.0, IBM,
Armonk, USA).

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

27

Fig. 3. Similarities and differences between the P-ML and the P-AP task. (A and B) Displacement of the platform of the Posturomed in the subjects mediolateral direction (A) and the subjects antero-posterior direction (B) during the P-AP (thin lines) and P-ML (thick lines) tasks. The dashed line separates both
tasks into two phases: The first phase start until first equilibrium is reached is the specific part of each task (initial direction of perturbation differs),
whereas the second phase first equilibrium until the end is very similar for both tasks (perturbations in both directions are produced only by the subject
himself).

3. Results
As can be seen in Fig. 4A and Table 1, the T-ML group considerably improved their performance (i.e., the time at equilibrium in seconds, ranging from 0 s to 20 s) in the task they trained (T-ML) and the P-ML group also increased their performance in the task they trained (P-ML). The T-ML group did not improve their performance in the non-trained P-ML task
more than the control group and the same was true for the performance of the P-ML group in the non-trained T-ML task.
The analyses of variance underpin these observations: for the T-ML task as well as the P-ML task there was a significant
group x time interaction effect (F(2,37) = 27.1 and p < 0.001 for T-ML; F(2,37) = 12.5 and p < 0.001 for P-ML). In addition,
the ANOVA of the performance improvements (i.e., performance of POST test minus performance of PRE test, see Fig. 4B)
revealed a main effect of group (F(2,37) = 12.3, p < 0.001), a main effect of task (F(3,36) = 12.0, p < 0.001) and also a
group  task interaction (F(6,111) = 9.3, p < 0.001). The Bonferroni-corrected t-tests used to follow up these results further
showed that this interaction was mainly due to the performance improvements of the training groups (T-ML group and
P-ML group) in their specific training tasks (T-ML and P-ML, respectively), whereas they did not improve more than the
control group in the non-trained tasks: for the T-ML task, there was no difference between P-ML group and C group
(p = 0.527, D = 0.20) but there were significant differences between T-ML group and C group (p < 0.001, D = 2.90) as well
as between T-ML group and P-ML group (p < 0.001, D = 2.70); likewise for the P-ML task, there were differences between
P-ML and C group (p < 0.001, D = 1.39) and also between P-ML and T-ML group (p < 0.001, D = 1.07) but there was no
difference between C group and T-ML group (p = 0.33, D = 0.31).
The results of the T-AP task show that there was not even a transfer to a very similar task on the same device: all groups
showed small performance increases (main effect of time: F(1,38) = 15.0, p < 0.001), but there was no difference between
groups (main effect of group: F(2,37) = 0.1, p = 0.91) and no group  time interaction (F(2,37) = 0.3, p = 0.72).
The analysis of the time at equilibrium in the P-AP task (i.e., the phase where the P-AP task is essentially the same as the
P-ML task) reveals a significant group  time interaction (F(2,37) = 7.2, p = 0.002), and the post-hoc tests showed that this
was due to the superior performance of the P-ML group: there were differences between the C and P-ML group
(p = 0.002, D = 1.82) and also between the P-ML and T-ML group (p = 0.008, D = 1.60) but there was no difference between
the C and T-ML group (p = 0.65, D = 0.22). However, when analysing the time to reach the first equilibrium after the initial
perturbation, i.e., the first phase where the P-AP task differs from the P-ML task, there was no significant group  time interaction effect (F(2,37) = 2.6, p = 0.09), only a significant main effect of time (F(1,38) = 60.0, p < 0.001). The changes (POST
minus PRE) for the time to reach the first equilibrium for the P-AP task were as follows: the T-ML group reduced their time
on average by 6.5 8.7 s, the C group by 4.1 6.8 s and the P-ML group by 7.6 7.6 s.

28

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

Fig. 4. Results. Mean performance (i.e., time at equilibrium) before and after the six training sessions for the three groups: C (control, 12 subjects, no
training), T-ML (14 subjects, training with the tilt board in medio-lateral direction), P-ML (14 subjects, training with the Posturomed in medio-lateral
direction). The four tasks that were tested were the T-ML task (tilt board in medio-lateral direction), the T-AP task (tilt board in antero-posterior direction),
the P-ML task (Posturomed in medio-lateral direction) and the P-AP task (Posturomed in medio-lateral direction). The thin vertical lines represent the
standard deviations. (A) The columns correspond to the average performance, in seconds, of the 3 trials of all the subjects of the three groups for the PRE
tests (white columns) and for the POST tests (black columns). The * and *** symbols indicate a significant group  time interaction effect (p < 0.05 and
p < 0.001, respectively). (B) The columns represent the training effect for each task (i.e., time at equilibrium in the POST test minus time at equilibrium in the
PRE test), expressed in seconds. A *** symbol represents a significant effect with p < 0.001 and a ** symbol an effect with p < 0.01.

Table 1
Exact numbers for the pre and post tests. Exact numbers for the performance before and after the six training sessions for the three groups (T-ML tilt board
training group, P-ML Posturomed training group, and control group) in the four tasks: T-ML task (tilt board in medio-lateral direction), T-AP task (tilt board in
antero-posterior direction), P-ML task (Posturomed in medio-lateral direction) and P-AP task (Posturomed in medio-lateral direction). The numbers represent
the group means, and the numbers in brackets the standard deviations. All values are expressed in seconds (time at equilibrium, potentially ranging from 0 s to
20 s). The last column represents the time to reach the first equilibrium, also a value between 0 s and 20 s, but with 0 s being the best and 20 s the worst value.

T-ML group pre

T-ML group post
P-ML group pre
P-ML group post
Control pre
Control post

T-ML task

P-ML task

T-AP task

P-AP task

P-AP task time to reach EQ

2.8
9.2
2.8
4.0
2.5
3.3

3.1 (2.8)
7.6 (4.5)
4.1 (3.5)
14.9 (1.3)
4.3 (3.9)
6.9 (5.3)

3.8
5.6
3.8
4.9
3.8
5.6

2.7 (2.7)
6.2 (3.5)
2.9 (4.5)
11.9 (3.8)
2.8 (4.2)
5.5 (4.9)

13.7 (7.3)
7.2 (3.8)
11.9 (8.2)
4.3 (3.5)
12.4 (8.2)
9.1 (7.3)

(2.3)
(3.2)
(3.8)
(3.5)
(2.6)
(3.9)

(2.0)
(3.2)
(2.0)
(2.5)
(2.6)
(3.8)

4. Discussion
4.1. Task-specificity of the training effects
The basic question driving this study was whether there is a considerable transfer from training a specific balance task to
another balance task that was not trained. The results of this study suggest that the effects of balance training are highly

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

29

task-specific. After the training, the group that trained on the tilt board in ML direction performed significantly better at this
task than the other training group and the control group, but did not outperform the control group in the three non-trained
balance tasks. Similarly, the group that trained on the Posturomed in ML direction performed significantly better than the
other groups in this task, but only outperformed the control group in the second phase of the P-AP task, which was essentially the same task they had been training.
The group that trained on the tilt board in ML direction did not only lack considerable improvements on a different device
in the same direction of perturbation, but they did not even show higher improvements than the other groups in a task that
used the same device and was very similar except for the direction of perturbation (AP instead of ML). Likewise, the group
that trained on the Posturomed in ML direction did not outperform the other groups when the direction of perturbation was
changed in the first phase of the task. As expected, in the second phase of the P-AP task the P-ML group did outperform the
other groups because then the task was exactly the same as the one trained (see Fig. 3). The P-ML and P-AP tasks are good
examples of tasks that look like different tasks at first, but turn out to be identical for the most part upon closer inspection.
Thus, what could seem to be a transfer of skill between two different tasks can turn out to be simply a task-specific training
effect.
To our knowledge, this is the first clear demonstration that the effects of balance training can be very task-specific and
that it may not improve general balance ability, but rather only the specific skill that is trained. This raises two questions:
why are the effects of balance training so task-specific, even when using the same device or testing tasks with the same
direction of perturbation? And how is it possible that many studies show that balance training is associated with performance improvements in balance tasks that were not trained or even in different types of tasks?
4.2. Comparison to other studies
Several studies have demonstrated an effect of balance training on other balance tasks that were not trained or other
kinds of movements (Hirase, Inokuchi, Matsusaka, & Okita, 2014; Sayenko et al., 2010; Yavuzer et al., 2006). This difference
of results could be due to different reasons. First, the duration of the training in our study was not very long (6 training
sessions in 2 weeks). Thus, it is possible that a longer training duration might have influenced the results of our study,
although it seems unlikely that the general outcome would have been fundamentally different.
Second, with balance training one could improve fitness components other than balance, which could be the main reason
for the performance improvement measured. Indeed, maximal strength (Gusi et al., 2012), rate of force development (Gruber
& Gollhofer, 2004), proprioception (Emilio, Hita-Contreras, Jimenez-Lara, Latorre-Roman, & Martinez-Amat, 2014) and
psychological aspects (Wolf, Barnhart, Ellison, & Coogler, 1997), can be improved with balance training. The effect of the
improvement of different physical and psychological components on the improvement of balance is well documented
(Campbell et al., 1997; Gauchard et al., 1999; Lee & Park, 2013; Pamukoff et al., 2014; Wolfson et al., 1996). Thus, the performance improvement of non-trained tasks after balance training might be primarily induced by the improvement of other
components and not from the transfer of balance skills acquired through the training. This effect might be even stronger
when studying populations that present fitness deficits such as the elderly or patients.
Third, as shown in our study with the P-ML and the P-AP tasks, two tasks, while looking different, may actually be very
similar for the most part. Thus, what could seem to be a transfer of skill between two different tasks might actually be only
the use of the same skill in a functionally identical task.
4.3. Study limitations
Although the results of the present study seem to be quite unequivocal, there are some limitations that should be noted.
As mentioned before, the duration of our study was not very long (6 training sessions in 2 weeks). Thus, it could be possible
that the training duration was not long enough to develop skills that could transfer to other balance tasks. However, it should
be noted that some of the subjects in the present study did not improve their performance anymore after their third or fourth
training session, suggesting that a longer training duration would not have had greater effects.
Furthermore, one might argue that the four balance tasks examined in the present study all used one-legged stance,
potentially limiting the generalizability of the results, e.g., to two-legged stance or walking. While this possibility cannot
be totally excluded, it is unlikely that there is no transfer of skill in one balance category, while in other balance categories
there is substantial skill transfer. For example, Donath et al. (2013, 2015) investigated the effect of slackline training on
performance in a variety of balance tests (static and dynamic, one- and two-legged) and found that the improvements after
the training were limited to the slackline, with no significant transfer to non-trained balance tasks.
4.4. Possible physiological basis of balance training specificity
In man, balance training can induce specific corticospinal and spinal adaptations, which are observable only during the
postural perturbation (Beck et al., 2007; Schubert et al., 2008). This implies that training a balance task can elicit very specific
adaptations in the motor system (Schubert et al., 2008). It has been suggested that balance and postural control depend on a
multisensory system composed of several different subsystems subjected to a hierarchical control (Deliagina, Orlovsky,
Zelenin, & Beloozerova, 2006). We hypothesize that depending on the balance task executed, the different semi-

30

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

autonomous subsystems are more or less recruited, and that the control of those systems by supraspinal structures can be
different. Thus, training a specific task might induce adaptations only in the specifically required subsystems and in the parts
of hierarchically superior structures that coordinate these subsystems.
Moreover, the training specificity of AP and ML directions of perturbation could be explained by the sway-direction
dependent cortical activity modulation (Slobounov, Hallett, Cao, & Newell, 2008) and by the probability of different muscular
synergies used (Krishnamoorthy, Latash, Scholz, & Zatsiorsky, 2003; Torres-Oviedo & Ting, 2007), which let suppose the existence of different strategies employed by the nervous system for AP and ML posture perturbations. Indeed, after reviewing a
wide array of studies on balance control, Winter, Patla, and Frank (1990) concluded that The postural synergy enlisted by
the nervous system is task and perturbation specific. This conclusion was based on evidence from many studies that showed
very different strategies and muscle activation patterns when comparing maintenance of static unperturbed stance, static
posture control under the presence of perturbations, balance control during the voluntary execution of a movement and
balance control during movement in the presence of perturbation. Even within these categories, muscle activation patterns
changed when the direction of the perturbation changed or when the perturbation was applied to different parts of the body
(Winter et al., 1990).
All in all, it seems to be useful to view a balance task as a type of motor task that happens to use some parts of the postural
control system, but a priori, a considerable transfer of learning effects from one motor task to another one should not be
expected.
4.5. Consequences for training programs
If the effects of balance training are indeed mostly task-specific, the first consequence would be to avoid the general term
balance training, as it favours a view of balance as a general ability and overall transferability of training effects from one
balance task to another. Furthermore, the results of meta-analyses that pool together different forms of balance training
should be interpreted with caution, as different forms of training with different aims are hard to compare.
As far as training programs are concerned, the results of this study suggest that it would be useful to first identify the
specific task or tasks that should be improved and then implement these tasks into a specifically tailored training program.
For example, if the aim is fall prevention in the elderly, the most common situations with a high fall risk for this population
should be identified and then practiced. If the most common situations with increased fall risk cannot be identified, an alternative might be to design training programs in the broadest way possible, i.e., to include a plethora of balance tasks with lots
of variations to be prepared for a multitude of situations with a potentially higher fall risk. Another less time-consuming
option might be to first test for particular weaknesses and address these weaknesses with a suitable training program.
In any case, to verify the efficacy of training regimens, we recommend to include the tasks that were trained as part of the
tests used before and after the training period, as this will provide a much better impression of the relevant task-specific
improvements than using generic tests such as one-leg stance.
5. Conclusions
The balance training used in this study improved performance only in the specific task that was trained. There was no
transfer to different tasks even when the device used was the same or when the direction of perturbation was the same.
Therefore, we suggest to take into consideration that the effects of balance training can be highly task-specific, with
little if any transfer to other balance tasks. This has implications for the design of balance training programs: for instance,
in injury and fall prevention programs, it might be advisable to choose training tasks that are as close as possible to the tasks
imposing a high fall or injury risk. Furthermore, if the effects of balance training are indeed rather specific to the balance task
that is trained, the use of generic balance tests to assess specific balance training adaptations does not seem to be advisable.
Acknowledgements
The authors want to thank Nathan Stern and Markus Baumeister for their help with the data acquisition and training
supervision, and James Belsey for offering advice on an early draft of the manuscript.
References
Agmon, M., Perry, C. K., Phelan, E., Demiris, G., & Nguyen, H. Q. (2011). A pilot study of Wii Fit exergames to improve balance in older adults. Journal of
Geriatric Physical Therapy, 34, 161167.
Beck, S., Taube, W., Gruber, M., Amtage, F., Gollhofer, A., & Schubert, M. (2007). Task-specific changes in motor evoked potentials of lower limb muscles after
different training interventions. Brain Research, 1179, 5160.
Campbell, A. J., Robertson, M. C., Gardner, M. M., Norton, R. N., Tilyard, M. W., & Buchner, D. M. (1997). Randomised controlled trial of a general practice
programme of home based exercise to prevent falls in elderly women. British Medical Journal, 315, 10651069.
Deliagina, T. G., Orlovsky, G. N., Zelenin, P. V., & Beloozerova, I. N. (2006). Neural bases of postural control. Physiology (Bethesda), 21, 216225.
Donath, L., Roth, R., Rueegge, A., Groppa, M., Zahner, L., & Faude, O. (2013). Effects of slackline training on balance, jump performance & muscle activity in
young children. International Journal of Sports Medicine, 34, 10931098.
Donath, L., Roth, R., Zahner, L., & Faude, O. (2015). Slackline training and neuromuscular performance in seniors: A randomized controlled trial. Scandinavian
Journal of Medicine & Science in Sports.

L.-S. Giboin et al. / Human Movement Science 44 (2015) 2231

31

Emilio, E. J., Hita-Contreras, F., Jimenez-Lara, P. M., Latorre-Roman, P., & Martinez-Amat, A. (2014). The association of flexibility, balance, and lumbar
strength with balance ability: Risk of falls in older adults. Journal of Sports Science & Medicine, 13, 349357.
Gauchard, G. C., Jeandel, C., Tessier, A., & Perrin, P. P. (1999). Beneficial effect of proprioceptive physical activities on balance control in elderly human
subjects. Neuroscience Letters, 273, 8184.
Gruber, M., & Gollhofer, A. (2004). Impact of sensorimotor training on the rate of force development and neural activation. European Journal of Applied
Physiology, 92, 98105.
Gusi, N., Carmelo Adsuar, J., Corzo, H., Del Pozo-Cruz, B., Olivares, P. R., & Parraca, J. A. (2012). Balance training reduces fear of falling and improves dynamic
balance and isometric strength in institutionalised older people: A randomised trial. Journal of Physiotherapy, 58, 97104.
Hirase, T., Inokuchi, S., Matsusaka, N., & Okita, M. (2014). Effects of a balance training program using a foam rubber pad in community-based older adults: A
randomized controlled trial. Journal of Geriatric Physical Therapy.
Horak, F. B., Wrisley, D. M., & Frank, J. (2009). The Balance Evaluation Systems Test (BESTest) to differentiate balance deficits. Physical Therapy, 89, 484498.
Kovacs, E., Sztruhar Jonasne, I., Karoczi, C. K., Korpos, A., & Gondos, T. (2013). Effects of a multimodal exercise program on balance, functional mobility and
fall risk in older adults with cognitive impairment: A randomized controlled single-blind study. European Journal of Physical Rehabilitation Medicine, 49,
639648.
Krishnamoorthy, V., Latash, M. L., Scholz, J. P., & Zatsiorsky, V. M. (2003). Muscle synergies during shifts of the center of pressure by standing persons.
Experimental Brain Research, 152, 281292.
Lee, I. H., & Park, S. Y. (2013). Balance improvement by strength training for the elderly. Journal of Physical Therapy Science, 25, 15911593.
Lubetzky-Vilnai, A., & Kartin, D. (2010). The effect of balance training on balance performance in individuals poststroke: A systematic review. Journal of
Neurologic Physical Therapy, 34, 127137.
McGuine, T. A., Greene, J. J., Best, T., & Leverson, G. (2000). Balance as a predictor of ankle injuries in high school basketball players. Clinical Journal of Sport
Medicine, 10, 239244.
McKeon, P. O., & Hertel, J. (2008). Systematic review of postural control and lateral ankle instability, part II: Is balance training clinically effective? Journal of
Athletic Training, 43, 305315.
McMurdo, M. E., Millar, A. M., & Daly, F. (2000). A randomized controlled trial of fall prevention strategies in old peoples homes. Gerontology, 46, 8387.
Muehlbauer, T., Besemer, C., Wehrle, A., Gollhofer, A., & Granacher, U. (2012). Relationship between strength, power and balance performance in seniors.
Gerontology, 58, 504512.
Oddsson, L. I. E., Boissy, P., & Melzer, I. (2007). How to improve gait and balance function in elderly individualsCompliance with principles of training.
European Review of Aging and Physical Activity, 1523.
Pamukoff, D. N., Haakonssen, E. C., Zaccaria, J. A., Madigan, M. L., Miller, M. E., & Marsh, A. P. (2014). The effects of strength and power training on single-step
balance recovery in older adults: A preliminary study. Clinical Interventions in Aging, 9, 697704.
Sayenko, D. G., Alekhina, M. I., Masani, K., Vette, A. H., Obata, H., Popovic, M. R., et al (2010). Positive effect of balance training with visual feedback on
standing balance abilities in people with incomplete spinal cord injury. Spinal Cord, 48, 886893.
Sayenko, D. G., Masani, K., Vette, A. H., Alekhina, M. I., Popovic, M. R., & Nakazawa, K. (2012). Effects of balance training with visual feedback during
mechanically unperturbed standing on postural corrective responses. Gait Posture, 35, 339344.
Schubert, M., Beck, S., Taube, W., Amtage, F., Faist, M., & Gruber, M. (2008). Balance training and ballistic strength training are associated with task-specific
corticospinal adaptations. European Journal of Neuroscience, 27, 20072018.
Slobounov, S., Hallett, M., Cao, C., & Newell, K. (2008). Modulation of cortical activity as a result of voluntary postural sway direction: An EEG study.
Neuroscience Letters, 442, 309313.
Tofthagen, C., Visovsky, C., & Berry, D. L. (2012). Strength and balance training for adults with peripheral neuropathy and high risk of fall: Current evidence
and implications for future research. Oncology Nursing Forum, 39, E416E424.
Torres-Oviedo, G., & Ting, L. H. (2007). Muscle synergies characterizing human postural responses. Journal of Neurophysiology, 98, 21442156.
Winter, D. A., Patla, A. E., & Frank, J. S. (1990). Assessment of balance control in humans. Medical Progress Through Technology, 16, 3151.
Wolf, S. L., Barnhart, H. X., Ellison, G. L., & Coogler, C. E. (1997). The effect of Tai Chi Quan and computerized balance training on postural stability in older
subjects. Atlanta FICSIT Group. Frailty and Injuries: Cooperative Studies on Intervention Techniques. Physical Therapy, 77, 371381.
Wolfson, L., Whipple, R., Derby, C., Judge, J., King, M., Amerman, P., ... Smyers, D. (1996). Balance and strength training in older adults: Intervention gains and
Tai Chi maintenance. Journal of the American Geriatrics Society, 44, 498506.
Yaggie, J. A., & Campbell, B. M. (2006). Effects of balance training on selected skills. Journal of Strength & Conditioning Research, 20, 422428.
Yavuzer, G., Eser, F., Karakus, D., Karaoglan, B., & Stam, H. J. (2006). The effects of balance training on gait late after stroke: A randomized controlled trial.
Clinical Rehabilitation, 20, 960969.
Zech, A., Hubscher, M., Vogt, L., Banzer, W., Hansel, F., & Pfeifer, K. (2010). Balance training for neuromuscular control and performance enhancement:
A systematic review. Journal of Athletic Training, 45, 392403.