Continuous visual feedback prevents from using new visuomotor mappings in a wrist rotation task Manuel Vidal
1,2
1
, Christine Mégard and Julien Barra 2
2
3
3
LPPA, Collège de France, List, CEA and Université Paris Descartes
[email protected]
· Normal trials: – No effect of the expected haptic angle and gain whatsoever on the initial haptic speed, suggesting a lack of anticipation of the visuomotor gain being used (Fig. a) – Manipulation with a gain of ½ was slower than with other gains [1], but this is the only significant difference (Fig. b) – No effect of the gain on the number of interceptions, suggesting the same level of precision across gains (Fig. c) · Blind trials: analysis of the used / trained gain ratio (Fig. d) – Different ratio then 1 for all gains but γ=2 [2] – Larger ratio than 1 for γ=1, which does not differ from that of γ=½ [3] 30
3
20 15 10
5 0 10°
Participants b)
1st block
2nd block
Total experimental time:
Break
3rd block
approx. 45 minutes
18 normal trials
18 normal trials
Angle reproduction
(passive)
(active)
Black screen
End
Angle presentation Reproduction: Normal trials: Blind trials:
Typical manipulation dynamics, automatic validation criteria and analyzed data
Data analyzed
1. 2. 3.
Bock O (1992) Adaptation of aimed arm movements to sensory-motor discordance: evidence for direction -independent gain control. Behav Brain Res 51:41–50 Gordon J, Ghilardi MF, Ghez C (1994) Accuracy of planar reaching movements. I. Independence of direction and extent variability. Exp Brain Res 99:97–111 Rossetti Y, Desmurget M, Prablanc C (1995) Vector coding of movement: vision, proprioception, or both? J Neurophysiol 74:457–463
Experiment
Start
End
1st block
2nd block
24 trials per block Total experimental time:
.............
7th block
1
20° Expected haptic angle
30°
Blind trials
160% 140%
approx. 45 minutes
Angle presentation
Angle reproduction
Feedback
(passive)
(active, in darkness)
(1000ms)
Black screen
End
4. 5. 6. 7.
Vindras P, Viviani P (2002) Altering the visuomotor gain. Evidence that motor plans deal with vector quantities. Exp Brain Res 147:280–295 Kawato M (1999) Internal models for motor control and trajectory planning, Cur Opin Neurobiol 9: 718–727 Jordan MI (1996) Computational aspects of motor control and learning. In: Heuer H, Keele SW (eds) Handbook of perception and action, vol 2: motor skills. Academic, London, 71–120 Flanagan JR, Rao A (1995) Trajectory adaptation to a nonlinear visuomotor transformation: evidence of motion planning in visually perceived space. J Neurophysiol 74:2174–2178
8.
Block
Condition
Gain
Rotation type
Post-trial feedback
1 2, 5 3, 6 4, 7
Control Learning Adaptation/Transfer test Extinction
=1 { 2/3 ; 3/2 } { 2/3 ; 3/2 } =1
yaw and pitch yaw only yaw and pitch yaw and pitch
yes yes no yes
Goodbody SJ, Wolpert M (1999) The effect of visuomotor displacements on arm movement paths. Exp Brain Res 127:213–223 9. Hinder M, Treslian J, Riek S, Carson R (2008) The contribution of visual feedback to visuomotor adaptation: How much and when? 1197: 123-134 10. Liu X, Scheidt R (2008) Contributions of Online Visual Feedback to the Learning and Generalization of Novel Finger Coordination Patterns. J Neurophysiol 99: 2546-2557 11. Bernier P-M, Chua R, Franks IM (2005) Is proprioception calibrated during visually guided movements? Exp Brain Res 167:292–296 12. Heuer h, Hegele M (2008) Constraints on visuomotor adaptation depend on
the type of visual feedback during practice. Exp Brain Res 185: 101-110 13. Choe C, Welch R (1974) Variables affecting the intermanual transfer and decay of prism adaptation. J Exp Psychol 102: 1076-1084 14. Imamizu H, Shimojo S (1995) The locus of visual - motor learning at the task or manipulator level: Implications from intermanual transfer. J Exp Psychol: Hum Percept Perform 21: 719 - 733 15. Sainburg R, Wang J (2002) Interlimb transfer of visuomotor rotations: independence of direction and final position information. Exp Brain Res 145: 437-447
0.8 Extinction Control Adaptation test
0.6
1.0 0.8 0.6
0.4
0.4
0.2
0.2
Extinction Control Transfer test
0.0
1
2
3
4
5
6
7
8
9
10
11
12
0
1
2
3
4
5
6
7
8
9
d)
2/3 Block (yaw rotations)
1.6
1.4
1.4
1.2
1.0 0.8 0.6 Extinction Control Adaptation test
0.8 0.6 0.4 Extinction Control Transfer test
0.2 0.0
0
1
2
3
4
5
6
7
8
9
10
11
12
0
1
2
Trial order
80% 60%
1/2
1
2
4
Trained gain
These findings show on the one hand that training the participants with a given visuomotor gain does alter the performance between gains with the continuous visual feedback (no anticipation or precision increase). On the other hand, when the visual feedback is suppressed, there is a rather good correction for these new gains, though a global overestimation is found even for the gain of 1 (cf. blind trials). This suggests that for such simple visuomotor tasks, an internal model of the new mapping is indeed constructed, though as long as the visual feedback remains available an automatic and possibly generic dynamic control process takes place, which does not rely on internal models. Learning blocks (yaw rotations)
2.0
Experiment 2 – Open-loop training
1.8 1.6
· Learning block (yaw): – Exponential fits for learning curves – Learning does occur for both gains: – Very rapidly for γ=3/2 (|λ|=9.5) – ~30 times slower for γ=2/3 (|λ|=0.28) · Control, Adaptation test and Extinction blocks (yaw): – Control: slow decrease of the simulated gain to adapt to the “natural” gain γ=1 1.2
1.0 0.8
12
1.0
0.0
100%
11
2/3 Block (pitch rotations)
f=-9.3+10.3*(1-exp(-3.83*x))
120%
10
Trial order (in block per type)
0.2
0% 30°
1.0
0.4
40%
20° Expected haptic angle
1.2
1.2
1.4
Start
Trial n
180%
0.1
1.2
b)
3
4
5
6
7
8
9
10
11
12
Trial order
· Adaptation and transfer vs. baselines (Fig. a & b): – ANOVAs on average performance of the last 6 trials of each condition – Simulated gains when trained with γ=3/2 (resp. γ=2/3) were higher (resp. lower) for adaptation (yaw) and transfer (pitch) trials than for both control and extinction trials [4] (resp. [5]), which did not differ a)
b)
Learned gain 3/2
1.6
1.6
Adaptation Transfer
1.4 1.2 1 0.8 0.6 0.4
Yaw axis
0.2
Pitch axis
0 Control
Test
Extinction
Learned gain 2/3
1.4 1.2 Transfer
1 0.8
Adaptation
0.6 0.4
Yaw axis Pitch axis
0.2 0 Control
Test
Extinction
These findings show that it is possible to construct an internal model with just post-manipulation visual feedback. The resulting visuomotor adaptation transfers to some extent to another type of rotations (from yaw to pitch). The induced gain for pitch rotations is partially but significantly updated, although contrarily to the yaw adaptation gain, this transferred gain decays rapidly (cancellation of the new mapping).
0.6 0.4
f=0.73+0.25*exp(-0.28*t-1) f=1.16+0.3*(1-exp(-9.53*t-1))
0.2 0.0
· Closed-loop trials: manipulation latency (∆t), initial haptic speed (ωinit) and number of intersections (see figure above) · Open-loop trials: signed final error and simulated gains References
Experiment 2 – Open-loop training Break
· Task: reorient the initially rotated teapot with a translucent clone of the teapot always presented in a canonic orientation · Closed-loop control: continuous visual feed-back is available during adjustments (exp. 1 except blind trials) · Open-loop control: no visual feedback during manipulation (black screen) with a final performance visual feedback provided (learning and extinction trials of exp. 2) or not (blind trials of exp. 1 and adaptation/transfer test trials of exp. 2)
continuous visual feedback, automatic validation no visual feedback (during or after manipulation), manual validation
Break
Protocols
0.2
10°
6 blind trials
Start
1
d)
0
Single gain value (visual/motor): i
Trial n
0.3
1.4
3/2 Block (pitch rotations)
Trial order
20%
End Break
Start
Block i
4th block
0.4
1.4
2
10°
1.6
0
Simulated gain
· Display: large screen (250cm x 180cm) at 120cm backprojection · Haptics: HaptionTM Virtuose force feedback robotic arm (6D34-45)
Experiment
End Break
distance, BarcoTM
Start Break
Material
Number of interceptions
Experiment 1 – Closed-loop training
1.6
f=1.1+1.9*exp(-1.7*x)
2
30°
Normal trials
1.8
0.0
0
Used/Trained gain
· 11 subjects (exp. 1) and 20 subjects (exp. 2)
20° Expected haptic angle
1.8
Simulated gain
25
Simulated gain
Material and Methods
Normal trials
c)
c)
3/2 Block (yaw rotations)
Simulated gain
Normal trials
a)
Simulated gain
a)
– For γ=3/2 (Fig. a): stable adaptation and fast extinction (|λ|=1.7) but ~17 times slower than learning – For γ=2/3 (Fig. b): stable adaptation and fast extinction (|λ|=3.8) but ~14 times faster than learning · Control, Transfer test and Extinction blocs (pitch): – For γ=3/2 or 2/3 (Fig. c & d): gradual decay of the adaptation leading to the absence of extinction (already back to baseline)
Simulated gain
Experiment 1 – Closed-loop training
Manipulation lattency (s)
subjects were asked to adjust the orientation of a virtual teapot using a force feedback robots’ arm. We manipulated the visual feedback to assess how it affects the mechanisms employed by the motor system to compensate for the imposed new wrists’ visuomotor gain. In a first experiment, a continuous visual feedback was provided to subjects during the learning of the new visuomotor gains while in the second experiment, only post-manipulation visual feedback was provided. In the latter, we also tested weather adaptation to yaw wrists rotations could transfer to pitch rotations, as suggested in earlier work [13-15].
and the visual space were the action’s feedback is available [5]. Two types of visual feedback for adaptation were investigated [6]: a continuous feedback of the hand position (closed-loop control) or a post-trial visual feedback (open-loop control). Contradictory results have been reported in the literature. Some found that only continuous visual feedback could induce adaptation [7-10], while others found that adaptation depended on post-trial feedbacks [11, 12]. In two experiments we examined the adaptation of wrists’ motor control to new visuomotor gain in rotation in a task where
Initial haptic speed (°/s)
Mechanisms involved in the production of visually guided reaching movements have been extensively investigated but little is known on the motor control of the wrist’s rotation. Previous studies found that the central nervous system plans the motor commands by processing independently amplitude and direction of the movement through feed forward mechanisms [1-4]. Changes in the feed forward motor commands that take the perturbation into account were interpreted as the updating or construction of internal models of new visuomotor mappings between the physical space where the motor action occurs
Results and discussion
Simulated gain
Introduction
0
Statistics 1. 2. 3. 4.
Tuckey HSD post-hoc: p=0.05 for γ=1, p
