1. Neuroscience
Download icon

Early life experience sets hard limits on motor learning as evidenced from artificial arm use

  1. Roni O Maimon-Mor  Is a corresponding author
  2. Hunter R Schone
  3. David Henderson Slater
  4. A Aldo Faisal
  5. Tamar R Makin
  1. WIN Centre, Nuffield Department of Clinical Neuroscience, University of Oxford, United Kingdom
  2. Institute of Cognitive Neuroscience, University College London, United Kingdom
  3. Laboratory of Brain & Cognition, NIMH, National Institutes of Health, United States
  4. Oxford Centre for Enablement, Nuffield Orthopaedic Centre, United Kingdom
  5. Departments of Bioengineering and of Computing, Imperial College London, United Kingdom
Research Article
Cite this article as: eLife 2021;10:e66320 doi: 10.7554/eLife.66320
5 figures, 2 tables and 2 additional files

Figures

Figure 1 with 1 supplement
Experimental design and main analyses.

(A) Left: An illustration of the robotic manipulandum device setup. Participants performed reaching movements while holding a robotic handle. A monitor displaying the task components was viewed via a mirror, such that participants did not have direct vision of their arm. Visual feedback was provided as a cursor depicting the current location of the arm. Right: A visualization of a single trial and the different terms used. In each trial, participants reached from the home position to a single visual target. The green line represents the participant’s arm trajectory. (B) Reaching trajectories to all targets from a randomly selected participant. The different colored lines are trajectories of individual reaching trials. (C) Reaching performance as measured by absolute errors for each group for each arm. Gray, blue, and red colors represent control, acquired, and congenital groups, respectively. Lighter colors represent intact/dominant-arm performance; darker colors represent artificial/nondominant-arms. We found a significant group effect (F(2,47)=13.81, p≤0.001, ηp2=0.37), with the congenital group making larger errors with their artificial arm compared to both acquired group’s artificial arm (t=−3.77, ptukey=0.001, Cohen’s-d=−1.39) and control group’s nondominant arm (t=−5.06, ptukey<0.001, Cohen’s-d=−1.705). Dotted lines connect errors between arms of individual participants. Artificial arm markers represent artificial arm types. (D) Relationship between age at first artificial arm use and artificial arm reaching errors in the congenital group. D – Dominant arm, ND – Nondominant arm, I – Intact arm, A – Artificial arm. ***p<0.001.

© 2010, Wilson et al. Panel A is reproduced from Figure 1 in Wilson et al., 2010, published under the Creative Commons Attribution 4.0 International Public License.

Figure 1—figure supplement 1
Intact hand errors and daily artificial arm use.

We found a significant correlation (r(39)=−0.41, p=0.008) between artificial arm daily use and intact-hand reaching errors. In this analysis, both artificial arm users’ groups (congenital and acquired groups) were analyzed together as we found no differences in intact-hand errors between the groups. Daily artificial arm use was quantified using questionnaires relating to both wear-time and functionality of use.

Figure 2 with 1 supplement
Exploring the source of increased reaching errors using additional analyses and tasks.

In all plots, gray, blue, and red represent the control, acquired, and congenital groups, respectively. (A) Left: rose plot density histogram of the distribution of bias angles across the groups, the larger the arc the more individuals from that groups had a bias within the arcs angle range. We found no significant differences in bias angle between the groups (Watson-Williams circular test: F(2,48)=1.95, p=0.15). Right: Error bias and noise results. No significant group differences were found for bias (F(2,47)=2.40, p=0.1, BFIncl=0.72). The congenital group shows significantly more motor noise than amputees and controls (F(2,47)=14.15, p<0.001, ηp2=0.38; post hoc significance levels are plotted). (B) Initial directional error results. The congenital group has larger directional error in the initial phase of reaching (F(2,47)=8.01, p<0.001, ηp2=0.26; post hoc significance levels are plotted). (C) 1D localization task results. Participants placed their residual limb or artificial arm inside an opaque tube and were asked to assess the location of the limb using their intact arm. We found no localization differences between the acquired and congenital groups in either condition (BF10<0.33 for both). The gray line next to the y-axis shows the mean ± s.e.m of control group’s nondominant hand localization errors. (D) 2D localization task results. Using the same apparatus, participants performed reaches to visual targets without receiving visual feedback during the reach. We found no group differences in absolute error (F(2,44)=0.71, p=0.5, BFIncl=0.33). (E) Relationship between artificial arm motor noise and age at first artificial arm use artificial arm in the congenital group. See Figure 2—figure supplement 1 for plots with individual participants’ data points. *p<0.05, **p<0.01, ***p<0.001.

© 2010, Wilson et al. Panel D is reproduced from Wilson et al., 2010, published under the Creative Commons Attribution 4.0 International Public License.

Figure 2—figure supplement 1
Plots with individual participants’ data points.

Scatter plots for data presented in Figure 2A–D.

Years of limbless experience before first artificial arm use in the acquired group.

(A) Relationship between years of limbless experience and (A) artificial arm reaching errors or (B) artificial arm motor noise in the acquired group.

Appendix 1—figure 1
Group values for Fitts law model fitting (r2, a, b).

A linear regression was fit for each participant’s reaches to obtain the Fitts law model’s parameters a and b. Parameters, as well as goodness-of-fit (r2), were compared across groups. We found no group differences in either goodness of fit (r2: p=0.84, BFIncl=0.167) or fitted parameters (a): p=0.31, BFIncl=0.347, (b: p=0.61, BFIncl=0.22) between groups, indicating artificial arms reaches follow Fitts’ laws and do not differ in their speed-accuracy trade-off strategy.

Author response image 1

Tables

Table 1
Demographic details of all participants.

Participant: ALD=individual with an acquired limb difference following amputation, CLD=individual with a congenital limb difference, CO=two-handed control; participants marked with an asterisk have valid data only for their intact-hand and were therefore excluded from most analyses. Y since amp=years since amputation. Gender: M=male, F=female. Amp side=side of limb loss or nondominant side: L=left, R=right. Amp level=level of limb loss: TR=trans-radial, TH=trans-humeral. Artificial arm type=preferred type of artificial arm: Cos=cosmetic, Mech=mechanical, Myo=myo-electric. Artificial arm wear time=typical number of hours artificial arm worn per week. PAL=functional ability with an artificial arm as determined by PAL questionnaire (0=minimum function, 1=maximum function). Usage score=Artificial arm usage score combining wear time and PAL. Age at first artificial arm use=Age at which individuals with a congenital limb difference were first introduced to an artificial arm. Years of limbless experience=Time after amputation at which amputees were first introduced to an artificial arm. Residual limb length=measured in cm.

ParticipantAgeY since ampGenderAmp sideAmp levelAmp causeArtificial arm typeArtificial arm wear timePALUsage scoreAge at first artificial arm useYears of limbless experienceResidual limb length
ALD015814MLTRTraumaMyo1190.51.870.515
ALD024616FLTRTraumaMyo560.590.490.514
ALD03*503FLTRTraumaMech770.440.4418
ALD04*5334MLTHTraumaMech480.2–1.40.33
ALD05211MRTRTraumaNone00.04–3.440.08334
ALD06*4218MRTRTraumaCos350.07–2.3316
ALD076121MLTRTraumaCos1050.672.210.12529.5
ALD086042MRTRTraumaMech87.50.280.053.59
ALD09*6537MRTHTraumaMech980.461.110.25
ALD10*4721MRTHTraumaCos840.30.031
ALD116812MLTRTraumaMech350.54–0.310.3321.5
ALD12495MRTRVascular diseaseCos420.590.10.520
ALD135729MLTRTraumaMech650.11–1.31118
ALD145333MLTRTraumaMyo980.430.980.6712
ALD15*2810FRTRTraumaMech20–3.5557.5
ALD162911MLTRTraumaNone00–3.61228.5
ALD174320MRTRTraumaMyo980.611.750.0838
ALD185512MLTRTraumaMyo980.651.920.533
ALD196117MLTRTraumaMech910.742.11118
ALD21303MLTRTraumaMyo490.590.290.5420.5
CLD0151FLTRCongenitalCos70.26–2.30.510
CLD0247MLTRCongenitalMech840.71.75113
CLD0345FLTRCongenitalMyo630.460.134.58
CLD04*26MLTRCongenitalMech60.13–2.880.2515
CLD05*55FLTRCongenitalCos1120.30.820.256
CLD0663MLTRCongenitalCos87.50.350.35210
CLD0735MLTRCongenitalCos560.28–0.84311
CLD0949MLTRCongenitalMyo910.571.39221
CLD1042MLTRCongenitalCos560.540.28210.5
CLD1166FRTRCongenitalCos420.35–0.9359
CLD1256FRTRCongenitalCos980.430.98311.5
CLD13*53MLTHCongenitalMech630.33–0.432
CLD1442MLTRCongenitalMech20.09–3.17412
CLD1555FLTRCongenitalMyo1050.652.12311.5
CLD1729MLTRCongenitalMyo700.460.33112
CLD1853FLTRCongenitalCos480.650.521.57
CLD2052FRTRCongenitalMyo32.50.26–1.580.311.5
CLD2132FRTRCongenitalMyo400.41–0.730.59
CLD2257MRTRCongenitalMech1260.692.88215.5
CLD2347FLTRCongenitalMyo840.892.56118.5
CLD2541MLTRCongenitalMyo1120.853.170.258
CO0128FL
CO0340FL
CO0459ML
CO0527FR
CO0661FL
CO0735ML
CO0834FL
CO0924ML
CO10*70ML
CO1124ML
CO1218FL
CO1367ML
CO1450MR
CO1551FL
CO1636FL
CO1741MR
CO1833MR
CO1945MR
CO2054MR
CO2153ML
Appendix 1—table 1
Frequentist and Bayesian analysis of model fitting reaches data to Fits’ Law.

Full statistical report of group comparisons of model’s parameters a and b as well as goodness-of-fit (r2) of the linear regression model. No differences were found across groups.



ANCOVA – r2 Artificial arm
Bayesian ANCOVA
Analysis of effects – r2 Artificial arm
FactorsSSdfMSFpEffectsP(incl)P(incl|data)BF incl
Group0.00320.0020.1750.84Group0.50.1430.167
r2 Intact0.07210.0727.5870.008r2 Intact0.50.8545.852
Residuals0.447470.01
ANCOVA – a Artificial armBayesian ANCOVA
Analysis of effects – a Artificial arm
FactorsSSdfMSFpEffectsP(incl)P(incl|data)BF incl
Group27,481.713213,740.8571.2110.307Group0.50.2580.347
a Intact158,782.4821158,782.48213.998<0.001a Intact0.50.98253.771
Residuals533,131.7734711,343.229
ANCOVA – b Artificial armBayesian ANCOVA
Analysis of effects – b Artificial arm
FactorsSSdfMSFpEffectsP(incl)P(incl|data)BF incl
Group1378.6062689.3030.4980.611Group0.50.1780.217
b Intact35,615.379135,615.37925.73< .001b Intact0.513856.606
Residuals65,056.014471384.171

Additional files

Supplementary file 1

Supplementary full statistical reports.

(a) Main analysis while controlling for artificial arm/nondominant arm side. Results of a follow-up ANCOVA analysis showing no effects of artificial arm side (L vs. R) on artificial arm reaching errors. Our main finding of a significant group effect was also unaffected by accounting for the side of the arm making the reaches. (b) Main analysis while controlling for residual limb length. Results of a follow-up ANCOVA analysis showing no effects of residual limb length on artificial arm reaching errors. Our main finding of a significant group effect was also unaffected by accounting for residual limb length. Note that this analysis only includes artificial arm users (congenital and acquired) as controls have a complete arm and therefore no residual limb length (c) Comparing artificial arm error noise while controlling for artificial arm bias. Results of a follow-up ANCOVA analysis showing that while there is a significant relationship between bias and noise, the group differences in error noise are independent of bias. (d) Analysis of reaching errors comparing the main task and the 2D localization task. Participants made overall larger errors in the 2D localization task compared to the main task which included visual feedback. (e) Analysis of movement time comparing the main task and the 2D localization task. Participants took longer to move to the target in the 2D localization task compared to the main task which included visual feedback.

https://cdn.elifesciences.org/articles/66320/elife-66320-supp1-v2.docx
Transparent reporting form
https://cdn.elifesciences.org/articles/66320/elife-66320-transrepform1-v2.docx

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)