Learning to Learn: Environmental Consistency Modulates Motor Adaptation Rates    L. Nicolas Gonzalez Castro*1,2, Matthew Hemphill*1, Maurice Smith1.    The human motor system has the ability to adapt to different physical environments. This motor adaptation acts to  reduce future motor errors, however this ability depends on the consistency of the environment. Here we explore  whether  human  subjects  can  not  only  adapt  their  motor  output  but  also  adapt  the  rate  at  which  this  adaptation  occurs. This would allow environment‐specific adaptation that could improve motor performance beyond the level  possible with  a static  learning  rate.  In  a highly  inconsistent  environment, the  motor  system  would  benefit from  a  low  learning  rate  in  order  to  avoid  responding  to  current  disturbances  that  do  not  predict  future  disturbances.  However,  in  a  highly  consistent  environment,  the  motor  system  would  benefit  from  learning  at  a  higher  rate  because  disturbances  experienced  on  one  trial  should  be  highly  predictive  of  future  disturbances  –  and  thus  any  learning associated with the current disturbance is likely to improve future performance.  We tested this prediction  by  exposing  subjects  to  different  learning  environments  and  measuring  the  learning  rates  associated  with  each  environment.   

In  these  environments  the  consistency  of  force‐field  application  was  systematically  manipulated  by  varying  the  duration  of  repeated  blocks  during  which  the  force‐field  was  activated  (and  occasionally  inverted).  The  environments are diagrammed in Figure 1. In the first environment (P/N1; 3 subjects) subjects were exposed to a  single positive force‐field trial followed by a negative force‐field trial, followed by 7‐9 washout (null field) trials. In  the second environment (P1; 12 subjects) subjects were exposed to a single positive force‐field trial followed by 7‐9  washout (null field) trials. And in the third and fourth environments (P7 and P20; 12 subjects each), subjects were  exposed to 7 and 20 force‐field trials, respectively, followed by 14‐16 and 27‐29 washout trials, respectively.    

The  single‐trial  learning  rates  associated  with  these  environments  were  measured  by  inserting  error‐clamp  (EC)  trials  before  and  after  the  first  force‐field  (FF)  trial  in  a  random  subset  (44%)  of  the  learning  blocks.    These  measurement  triplets  (EC‐FF‐EC)  are  diagrammed  as  dashed  black  vertical  lines  Figure  1.  The  difference  between  the  lateral  force  profiles  during  the  post‐FF  and  pre‐FF  error‐clamp  trials  was  used  to  compute  the  learning  rate.  Hand  paths  as  well  as  mean  angular  errors  assessed  at  the  peak  speed  point  are  shown  for  the  initial  force‐field  trials  in  each  environment  in  Figure  2.  The  errors  induced  on  these  trials  did  not  vary  significantly  between  environments (one‐way ANOVA, F(3,17) = 1.32, p = 0.28).    

Despite the similarity between the motor errors caused by these force‐field trials, the learning rates associated with  these trials were significantly different between environments (one‐way ANOVA, F(3,17) = 7.1 , p<0.003 ) and varied  by more than a factor of ten. Figure 3 shows the change in lateral force between the post‐FF and pre‐FF produced  by subjects in the different learning environments and Figure 4 shows the associated learning rates (i.e., the fraction  of the force‐field  compensated  on  the  post‐FF  trial).  These  learning‐induced  changes  in  lateral force  after single‐ trial  FF  exposure  reached  levels  as  high  as  0.40  of  the  force‐field  magnitude  during  the  latter  part  of  the  P20  environment  which  had  an  average  learning  rate  of  0.28.  In  contrast,  learning  rates  averaged  0.17  in  the  P7  environment, 0.062 in the P1 environment, and 0.026 in the P/N1 environment.   

Although the learning rates observed after the first encounter with the learning stimuli early in the experiments are  not  significantly  different  (one‐way  ANOVA,  F(3,17)  =  2.0,  p  =  0.15),    the  last  learning  rate  measured  in  each  environment shows clear statistical differences (one‐way ANOVA, F(3,17) = 16.7, p < 0.0001).  As shown in Figure 5  learning  rates  in  the  two  more  consistent  environments  (P7  &  P20)  appeared  to  increase  with  subsequent  exposures  to  the  stimuli.    This  indicates  that  the  increases  in  learning  rates  associated  with  these  environments  progress slowly and gradually.  These results show that learning rates are readily modulated by the statistics of the  environment and that these modulations appear to reflect both increases and decreases from baseline adaptation  rates. Further work is needed to delineate the mechanisms that control these modulations.    * These authors contributed equally to this work.    1. Harvard School of Engineering and Applied Sciences, Cambridge, MA  2. Harvard‐MIT Division of Health Sciences and Technology, Cambridge, MA 



2  Mean Trajectory During First Trial of Learning Stimulus

Learning Stimuli 1

P/N1 P1 P7 P20 P/N1 Peak Speed P1 Peak Speed P7 Peak Speed P20 Peak Speed

0

0

0

-0.01

P/N1

Testing

-1

20

40

60

80

100

120

140 -0.02

1 0

-0.03

20

40

60

80

100

120

140

1 0 Testing

-1 0

P7 20

40

60

80

100

120

140

-0.04

Mean Angular Error @ Peak Speed 20

-0.05

18 16

Angular Error (deg)

0

Vertical Displacement (m)

P1

Testing

-1

-0.06

-0.07

1 0 Testing

-1 0

-0.08

P20 20

40

60

80

100

120

14 12 10 8 6 4 2

140

-0.09

Trial Number

P/N1

P1

P7

P20

Learning Stimulus -0.1 -0.02

-0.01

0

0.01

0.02

0.03

Horizontal Displacement (m)







Mean Learning Rates for the Different Learning Stimuli

Learning‐induced Changes in Force Profiles

Evolution of Learning Rates 0.6

1.8

P/N1 P1 P7 P20

1.6 1.4

0.4

Late Exposure

0.4

1 0.8 0.6

Learning Rate

0.3

Learning Rate

Force (N)

1.2

P/N1 P1 P7 P20

0.5

All Trials

0.2

0.3 0.2 0.1

0.1 0.4 0

0.2

0 -0.1

0 -0.2

-0.1 0

200

400

600

Time (ms)

800

1000

1200

P/N1

P1

Learning Stimulus

P7

P20

-0.2

0

2

4

6

8

10

12

Measurement

14

16

18

20

Modulation of Learning Rate Based on the Features ...

... appear to reflect both increases and decreases from baseline adaptation rates. Further work is needed to delineate the mechanisms that control these modulations. * These authors contributed equally to this work. 1. Harvard School of Engineering and Applied Sciences, Cambridge, MA. 2. Harvard-MIT Division of Health ...

103KB Sizes 1 Downloads 88 Views

Recommend Documents

INVESTIGATIONS ON EXEMPLAR-BASED FEATURES FOR SPEECH ...
from mobile and video sharing speech applications. How- ever, most ... average distance between the hypothesis X and the k-nearest .... the hypothesis in the lattice with the lowest edit cost. ... scribed in Section 2 for development and tuning.

Object Tracking based on Features and Structures
appearance and structure. II. GRAPH MODEL. Graph models offer high representational power and are an elegant way to represent various kinds of information.

The Beat Goes On: Rhythmic Modulation of Cortical ...
A good illustration of the type of information conveyed by tagging frequencies .... each trial, 20 brief (50 ms) stimuli were presented serially at a constant rate of 1 per 840 ms ..... is consistent with a greater effect on sensorimotor than posteri

Sparse Distributed Learning Based on Diffusion Adaptation
results illustrate the advantage of the proposed filters for sparse data recovery. ... tive radio [45], and spectrum estimation in wireless sensor net- works [46].

Batch Mode Reinforcement Learning based on the ...
We give in Figure 1 an illustration of one such artificial trajectory. ..... 50 values computed by the MFMC estimator are concisely represented by a boxplot.

Exemplar-Based Sparse Representation Features ...
in LVCSR systems and applying them on TIMIT to establish a new baseline. We then .... making it difficult to compare probabilities across frames. Thus, to date SVMs ...... His conversational biometrics based security patent was recognized by.

Kurdish Teacher's Perspectives on Play-based Learning in the Early ...
... degree within any other framework. Page 3 of 109. Kurdish Teacher's Perspectives on Play-based Learning in the Early Years - Nazeera Salih Mohammed.pdf.

Kurdish Teacher's Perspectives on Play-based Learning in the Early ...
Master of Arts in Professional Development ... Fulfilment of the requirements for the degree of. Master of Arts in Professional Development ... Teacher's Perspectives on Play-based Learning in the Early Years - Nazeera Salih Mohammed.pdf.