1. Stan Van Pelt and W. Pieter Medendorp
Gaze-Centered Updating of Remembered Visual Space During Active Whole-Body Translations
Stan Van Pelt and W. Pieter Medendorp
J. Neurophysiol 97: , 2007
Journal Club January

2. Research on rotational (eye-) movements suggest that various cortical and sub-cortical structures update the gaze-centred coordinates of remembered visual stimuli to maintain an accurate representation of visual space and further to produce motor plans
Translation is a challenge for the visual system as motion parallax shifts position of objects in front or behind fixation point to opposite side of the retina

3. They propose two prediction models to determine the reference frame
Authors aim to investigate which reference frame is used as a computational basis during translations
They propose two prediction models to determine the reference frame
Gaze-dependent model
Gaze-independent model

4. 1 1 2 2 3 3 Observer on one side, fixating FP
Target flashed in front or behind FP
Observer translates, fixating FP
Observer reaches to the memorised target 2 2 3 3

5. Reversed retinal position for targets Tf and Tn
Must estimate translation correctly to update memorised targets correctly
If not there will be reach errors
Basic assumption
Observers generally misestimate the amount of self-motion when translating

6. Gaze-dependent prediction
Simulation of motion parallax involved in the update
Reach errors in opposite directions for targets Tf and Tn

7. Gaze-independent prediction
Parallax geometry plays no role if body-fixed reference frame
Reach errors in same direction for targets Tf and Tn

8. Methods Main experiment and 3 control experiments
12 observers in main experiment, 5 in controls
Optotrak 3020 tracking of head and index finger
Eyelink II Eyetracker
2 conditions
Stationary task (160 trials)
Translation task (160 trials)
Catch trials (32 (x2?) trials)
On two separate occasions
Both times: Half of translation trials, then half of stationary

9. Methods Observer on one side, fixating FP
Target flash in front or behind FP (4 positions)
Observer translates sideways
Reach to memorised target
Tracked position of observer and finger tip
Recording of horizontal body position, gaze vergence, gaze version, reaching
This is a figure describing the translation task. The stationary task was basically the same.
The observer was placed on one side. When the trial starts (signal?) he should look at FP. After 1.5 s a target was flashed for 0.5 sec in 1 out of 4 positions. Target off, and observer moved to the other side. Auditory signal to indicate time for reaching. Signal for when to stop reaching. As I understand it, observer reach AT the remembered position

10. Data analysis
Relationship between the two models by Model 2 regression (NV and ME)
Slope and confidence limits estimated by Bootstrap method
Using stationary task as a measure of errors from perception and motor effects – assuming equal contribution in both tasks
2-D vectorial analysis for interaction between initial target position, translational motion and reach response for the two reference frames
Exclusion of trials
Saccades or not keeping fixation within 3 deg from FP
Reaching or stepping too early

11. Main findings
Reach responses showed parallax-sensitive updating errors
Errors reversed in lateral direction for targets presented at opposite depths from FP
Errors increased with larger depth from FP

12. Reach error = translation error – mean stationary error for each target distance
Random points from each target distance compared to random points from another target distance: T1/T4 and T2/T3
Reach responses showed parallax-sensitive updating errors
Errors reversed in lateral direction for targets presented at opposite depths from FP
Larger errors in outermost pos

13. Combinations of T1/T4 and T2/T3 according to their values
Best fit line through data
Favour gaze-dependent

14. table

15. Predicted updating error E
Position before translation Ti
(estimated average response stationary)
Position after translation Tf
Actual translational motion (Tf -Ti)
Reach response R
Predicted updating error E
Gaze-dependent closer match
Better prediction 9/12
Higher correlation coeff (p<0.01)
Authors suggest that this model can account for systematic errors in the data (gaze-ind does not)
R – Ti = a(Tf – Ti ) +b
Coordinate axes
Gaze-dependent model
With and orthog to gaze line
Gaze-independent model
With and orthog to shoulder line
table

16. Control experiments 5 observers Horizontal reaching errors
Conclusion remains
No feed-back from hand
Fixate FP-position all the time
FP disappear, but fixate location until reaching

17. Authors’ conclusions
According to their quantitative geometrical analyses, they claim that updating translational errors were better described in gaze-centred than in gaze-independent coordinates
Authors conclude that spatial updating for translational motion operates in gaze-centred coordinates

18. Considerations from authors
3/12 observers did not favour the model
The authors admit that they base their models on very simple geometry, and that the brain might have a far more complex visual representation of space
They have focussed on the central representation of body translation as an underlying mechanism for errors, and there may be many other sources to errors in updates

19. And now… US! Are the models valid?
Are there fundamental problems in the experimental setup?
Is this a sensible way of processing these data?
Do we believe in the results?

20. Horizontal data

21. Vectorial analysis