Transcript The subjective vertical during roll rotation

The dynamic subjective vertical
Rens Vingerhoets
Supervisors: Ronald Kaptein & Jan van Gisbergen
Contents
• Introduction
-History
-Sensory signals involved in spatial perception
-Conceptual scheme
• Theories on spatial perception
-Static models and dynamic models
• Methods
• Results
-Static & Dynamic vs models
-Static vs Dynamic
-Body-position estimation experiment
• Conclusions
Introduction
We tested two different aspects of spatial orientation
under dynamic roll conditions:
• The ability to judge the roll tilt-angle of the long body axis. (Bodyposition estimation task)
• The ability to judge visual line orientations in space when tilted sideways. (Visual orientation task)
Introduction - History
So far, most subjective vertical (SV) experiments
have only been conducted during static tilt.
Error (degrees)
e.g. Anton van Beuzekom 2000
Tilt position (degrees)
Introduction - Sensors
Sensory signals involved in spatial orientation:
•
•
•
•
Otoliths
Semicircular canals
Visual cues
Somatosensory cues
Introduction - Sensors
The otoliths
• Consist of utriculus and sacculus
• Sensitive to acceleration and tilt (ambiguity)
Introduction - Sensors
The semi-circular canals
• Sensitive to angular acceleration
• High-pass filter
Introduction – Conceptual scheme
Head in space
HS
Line in space
LS
Canals
Utriculus
Eye
Sacculus
Retina
position
^
E
H
Internal model of head in space
Idiotropic
Egocentric vision
^
H
^
L
S
H
Spatial vision
^
L
S
^
L
R
Theories on spatial perception – Mittelstaedt
Mittelstaedt’s theory
• SVV gravity information only from otoliths
• Ambiguity problem is ignored
• Errors in fusion of utricular and saccular signals
• Idiotropic vector (M) to compensate for fusion error
• Idiotropic vector is task independent (M ≈ 0.5)
Theories on spatial perception – Mittelstaedt
Theories on spatial perception – Mittelstaedt
Persistence of Aubert-effect
Tilt angle = 111o
Theories on spatial perception – Reymond
Visuovestibular perception of self-motion modeled as a dynamic
optimization proces
The model constructs estimates , V, A en G, using a dynamic optimization process minimizing
a set of cost functions. Each cost function consists of measure and coherence constraints.
* Measure constraints express the relationship between a sensory signal and its
prediction, considering the internal model of a given sensor.
* Coherence constraints express the relationship between the estimations and the
physical laws binding the motion stimuli.
Theories on spatial perception – Reymond
Theories on spatial perception – Reymond
The model predicts correct estimations of G for the static experiment
But large errors for the dynamic experiment:
Real tilt position
Subjective tilt position
Theories on spatial perception – Angelaki & Merfeld
Central Orientation/Motion Estimator
w
w
g = (-w x g)dt
a
g+
S scc (s)
f
-
Body Dynamics
a oto
S oto (s)
Sensory Dynamics
+
-
ef
ew
kw
a
kf w
ka
kf
+
+
+
a scc
+ wˆ
ˆ x g)dt
ˆ
gˆ = (-w
g
gˆ - aˆ
+
ˆ
f
ea
S oto (s)
ˆ oto ˆ
a
a scc
ˆ
w
Model of Sensory
Dynamics
ˆ
w
gˆ
aˆ
a
ˆ
a
ˆ
ˆ
f
S scc (s)
Model of Body
Dynamics
+
-
ef= a x
Theories on spatial perception – Angelaki & Merfeld
This model predicts correct estimations of G for the static experiment
But errors for the dynamic experiment:
Theories on spatial perception – Dynamics, why?
Experimental evidence for dynamics
Udo de Haes & Schone, 1970
Question
Do dynamic processes play a role in subjective visual
vertical experiments?
1
Are static and dynamic responses similar?
2
Is there any experimental evidence for a phase
lag?
Methods
Setup: Vestibular Chair
Methods
Static luminous line experiment:
- Rotation alternating CW & CCW
- Speed 30 deg/s
- Measurements 30 seconds after stop
- Flash line method
- 3 subjects (age 23 - 59 )
Dynamic luminous line experiment:
- Rotation 720o, alternating CW & CCW
- Speed 30 deg/s
- Flash line method
- Measurements during rotation, approximately every 3 s
- 9 subjects (age 21 - 64 )
Verbal body-position estimation experiment
Results Static Luminous Line Experiment
Results – Static experiment
Clockwise
CW
CCW
Mittelstaedt
M = 0.70
M = 0.32
M = 0.53
Results – Static experiment
Summary static experiment:
Systematic errors (Aubert-effect)
All subjects show similar responses
Behaviour is ‘Mittelstaedt-like’ except near headdown
Models by Reymond and Merfeld cannot predict this
data (no errors predicted)
Results Dynamic Luminous Line Experiment
Results – Dynamic experiment
Data
Mittelstaedt
Reymond
Clockwise only
Static
M = 0.70
M = 0.32
M = 0.53
Dynamic
M = -0.01
M = 0.22
M = 2.21
Results – Dynamic experiment
Data first cycle clockwise
RV
RK
JG
Results – Dynamic experiment
Clockwise first cycle data from all subjects
Data
Mittelstaedt
Results – Dynamic experiment
Summary dynamic experiment vs static experiment:
Large intersubject differences in dynamic but not in
static experiment.
Dynamic responses differ from static responses
If an idiotropic vector exists, it’s task dependent
Results – Dynamic experiment
Comparing first and second cycle data
First cycle response
Second cycle response
Results – Dynamic experiment
Summary comparision first cycle and second cycle responses:
No significant differences
Consequently no experimental evidence for phase
lagging
Results – Dynamic experiment
Comparing clockwise and counterclockwise data
Clockwise response
Counterclockwise response
Results – Dynamic experiment
Summary comparision clockwise and counterclockwise responses:
Differences, but not simply a phase lag.
Experimental evidence for dynamics
Results Dynamic Body-Position estimation
Experiment
Results – Body-position experiment
Clockwise
Counterclockwise
Results – Body-position experiment
Clockwise
Counterclockwise
Merfeld clockwise
Merfeld
Counterclockwise
Merfeld might be useful for body-position estimation
predictions. Further research is necessary.
Conclusions
Summary of models and experimental results
Mittelstaedt Data
Test
Reymond
Merfeld
cycle 1 = cycle 2
No
Yes
Yes
Static: CW = CCW
Yes
Yes
No*
Dynamic CW = CCW No
Yes
No
Static = Dynamic
Yes
No
No
Conclusions
Conclusions
1
Dynamic aspects play a role during subjective vertical
experiments. But the dynamics are not the result of a lag
effect predicted by recent models based on canal-otolith
interaction.
2
The model by Mittelstaedt can provide good fits when the
magnitude of idiotropic vector can be varied. However, this
model cannot account for dynamic aspects.
Conclusion – Conceptual scheme
Head in space
HS
Line in space
LS
Canals
Utriculus
Eye
Sacculus
Retina
position
^
E
?
H
Internal model of head in space
Idiotropic
Egocentric vision
^
H
^
L
S
H
Spatial vision
^
L
S
^
L
R
Remaining questions
Remaining questions
•
•
The body-position estimation experiments indicates that the
model by Merfeld might be able to do correct predictions for
this paradigm.  More subjects
In a pilot experiment evidence is found that clockwise and
counterclockwise differences may be due to cognitive load.
Remaining questions
Tilt angle = 230o
No stimulus during rotation
Stimulus during rotation
Theories on spatial perception – Mittelstaedt
Decomposition of gravity-vector in otolith-fixed
coordinates
FY = g * sin ()
FZ = g * cos ()
Utricular signal: Y = U * Fy
Saccular signal: Z = S * Fz
Y/N
tan  =
Z/N M
N = Y2  Z2
Theories on spatial perception – Reymond
Example: estimation of angular velocity, 
Canals modeled as a high-pass filter:
fc = (Tc S) / (1 + Tc S)
(Laplace notation)
Internal model is assumed perfect. fc = fc
Visual system modeled as a low-pass filter: fv = 1/(1 + TcS)
Together with a perfect internal model fv = fv
Coherence constraint given by:
this can be rewritten as:
dG/dt = G x 
 ’ = [ dG/dt x G + ( · G)G ]/ ||G ||2
Cost function: J1= a1 || fc () – fc( )||2 + a2 || fv () – fv( )||2 +a3 || - ’||