Spatial Workspace Symmetry Affects Generalization
of Sensorimotor Learning
Jens Tiggelbeck & Jochen Müsseler
Institute of Psychology
Department of Work and Cognitive Psychology
RWTH Aachen University, Germany
This Poster
Rotated visual
trajectories
Introduction
Introduction: Sensorimotor adaptation and its subsequent generalization is often conceptualized as formation
and updating of internal models aimed at compensating for the perceived incongruence between proximal and
distal action effects (Wolpert, et al., 1995). Recent findings by Wang and Müsseler (2012) suggest an influence of
action space symmetry as a determining factor: parallelism seems to promote parallel while symmetry results in
equally symmetric intramanual transfer.
Research Questions: Do the symmetry’s reference frame (egocentric vs. allocentric) and associated congruence
between proximal (motor) and distal (visual) action spaces influence adaptation to and transfer of rotational
visuomotor transformations?
Department Website
Adapted motor
trajectories
Target
CCW
(-30°)
CW
(+30°)
Left starting
position
Right starting
position
Fig. 2: General experimental setup and
apparatus for all reported experiments.
Fig. 1: Action space including visuomotor
rotation and required motor adaptation.
General method: Participants had to slide a cursor from one of two starting positions (balanced across subjects) towards a target (cf. Fig. 1) by performing short targeted flicking
motions (open loop) via a digitizer tablet and stylus. Each participant had to adapt to one of two possible visuomotor transformations: cursor trajectories were either visually rotated by
30° clockwise (CW) or counterclockwise (CCW). Intramanual transfer in the form of aftereffects was assessed at the untrained starting position without visual feedback.
Results: Generalization of adapFig. 3: Congruent proximal (bottom: tation at the untrained starting locadigitizer tablet) and distal (top: monitor) tion follows a symmetric pattern
action spaces resulting in an overlap of
(significant interaction between
egocentric and allocentric symmetry.
starting position and transformation).
However, overall reduced generalization is observed when outward
compensation is required, i.e. away from the egocentric midline.
Ex. 2: Allocentric Symmetry
After symmetric transfer could be
established under egocentric symmetry conditions, visual and motor
spaces are rotated 90° clockwise.
Thus, symmetry within the allocentric but not the egocentric reference frame is retained. While
adaptation should still be achieved,
diminished or even parallel generalization is expected to occur.
25
20
Angular Deviation in Degree
Visual and motor action spaces are
in alignment concerning allocentric
and egocentric symmetry. Based on
the results obtained by Wang and
Müsseler (2012) symmetrical transfer is expected, i.e. learned outward
and inward compensations should
be mirrored at the action space’s
midline during transfer trials.
a
Left x CW (outward compensation required = +°)
Right x CCW (outward compensation required = +°)
Left x CCW (inward compensation required = -°)
Right x CW (inward compensation required = -°)
b
CW
CCW
Left x CW
Right x CCW
Left x CCW
Right x CW
15
10
5
0
-5
-10
-15
-20
-25
-5
-1
6
12
18
24
30
36
42
48
54
60
Timeline / Baseline (negative numbers) and Experimental (positive numbers) Blocks
Left
Right
Starting Position in Adaptation Blocks
Fig. 4a: Adaptation and generalization performance across the entire experimental session. Each data point represents the mean of five consecutive
trials. Continuous lines refer to adaptation blocks with visual feedback. Dotted lines denote transfer blocks without visual feedback. Positive angular
values represent trajectories aimed away from the triangle’s central axis of symmetry, negative values refer to deviations towards the triangle’s midline.
Fig. 4b: Mean angular deviation across all transfer trials in each of the four experimental conditions. Error bars represent averaged intrapersonal standard
deviations of aiming errors. The gray dashed line refers to a clockwise (CW), the dashed and dotted line to a counterclockwise (CCW) transformation.
a
25
20
Angular Deviation in Degree
Ex. 1: Egocentric Symmetry
Left x CW (outward compensation required = +°)
Right x CCW (outward compensation required = +°)
Left x CCW (inward compensation required = -°)
Right x CW (inward compensation required = -°)
b
CW
CCW
Left x CW
Right x CCW
Left x CCW
Right x CW
15
10
5
0
-5
-10
-15
-20
-25
-5
-1
6
12
18
24
30
36
42
48
54
60
Left
Right
Results: The data suggests greatly
Timeline / Baseline (negative numbers) and Experimental (positive numbers) Blocks
Starting Position in Adaptation Blocks
reduced but qualitatively still observable symmetrical generalization Fig. 6a: Adaptation and generalization performance across the entire experimental session (see Fig. 4a for further explanations).
(interaction only by trend between Fig. 6b: Mean angular deviation across all transfer trials in each of the four experimental conditions (see Fig 4b for further explanations).
starting position and transformation). Note that the angular deviations from the unaltered ideal trajectories do not significantly differ from zero in any of the four experimental conditions.
Fig. 5: Congruent proximal (bottom:
digitizer tablet) and distal (top: monitor)
action spaces. Allocentric but not egocentric symmetry is maintained.
a
Ex. 3: Incongruence
Results: Data suggests a lack of
transfer (no a significant interaction
between starting position and
transformation and no aiming errors
significantly larger than zero).
Angular Deviation in Degree
20
Following the reduced symmetric
transfer in the allocentric condition,
the influence of an incongruency in
orientation between the visual and
motor action spaces is assessed.
Due to the increased uncertainty of
the overall spatial properties of the
action space, a lack of systematic
generalization is hypothesized.
Fig. 7: Incongruent proximal (bottom
right: digitizer tablet) and distal (top:
monitor) action spaces retain allocentric
and egocentric symmetry but do no
longer overlap.
25
Left x CW (outward compensation required = +°)
Right x CCW (outward compensation required = +°)
Left x CCW (inward compensation required = -°)
Right x CW (inward compensation required = -°)
b
CW
CCW
Left x CW
Right x CCW
Left x CCW
Right x CW
15
10
5
0
-5
-10
-15
-20
-25
-5
-1
6
12
18
24
30
36
42
48
54
Timeline / Baseline (negative numbers) and Experimental (positive numbers) Blocks
60
Left
Right
Starting Position in Adaptation Blocks
Fig. 8a: Adaptation and generalization performance across the entire experimental session (see Fig. 4a for further explanations).
Fig. 8b: Mean angular deviation across all transfer trials in each of the four experimental conditions (see Fig. 4b for further explanations).
Conclusion
Aiming accuracy following an adaptation to a visuomotor rotation of 30° CW or CCW is comparable (no significant differences in means and standard deviations of aiming accuracy)
across all three experiments. However, transfer performance is clearly influenced by manipulations of the underlying symmetry’s reference frame as (Ex. 1 vs. Ex. 2). These results are
in line with the findings of Wang and Müsseler (2012) and Taylor and Ivry (2013). However, egocentric action space symmetry does not seem to be the only influencing factor. Congruency between action spaces might also play a vital role (Ex. 3). Possible explanations pertain to different neuronal substrates for egocentic and allocentric motor representations (Chen, et
al., 2014) and to increased ambiguity of spatial action space properties leading to a conservative choice of the applicable internal model during transfer trials (Wang and Müsseler, 2012).
Contact:
Jens Tiggelbeck, jens.tiggelbeck@psych.rwth-aachen.de,
Institute of Psychology, RWTH Aachen University,
Jägerstr. 17-19, D-52066 Aachen, Germany
Telephone: +49 (0)241 / 80-96553
http://www.psych.rwth-aachen.de
Literature:
Chen, Y., Monaco, S., Byrne, P., Yan, X., Henriques, D. Y. P., & Crawford, J. D. (2014). Allocentric versus egocentric representation of remembered reach targets in human cortex. The Journal of
Neuroscience: The Official Journal of the Society for Neuroscience, 34(37), 12515–12526.
Taylor, J. A., & Ivry, R. B. (2013). Context-dependent generalization. Frontiers in Human Neuroscience, 7, 171.
Wang, L., & Müsseler, J. (2012). Generalization of visuomotor adaptation depends on the spatial characteristic of visual workspace. Experimental Brain Research, 223(3), 353–365.
Wolpert, D. M., Ghahramani, Z., & Jordan, M. I. (1995). An internal model for sensorimotor integration. Science, 269(5232), 1880–1882.
This research was funded by a grant from the DFG (Mu 1298/9)