18. !18
Experiment 1: Train II categories with RB control
Conditions to rule out innate
difficulty differences
Procedural learning during
declarative control?
20. !20
Results are consistent with procedural learning
during declarative control
Parsed Training All II Training
Rotated impaired
relative to congruent
No innate
difficulty difference
21. !21
Parsed Training All II Training
Rotated impaired
relative to congruent
No innate
difficulty difference
Results are consistent with procedural learning
during declarative control
22. !22
Hard to rule out rules
75% correct!
If using rule 1
75% correct!
If using rule 2
75% correct!
If using rule 1
25% correct!
If using rule 2
Rule 1 Rule 2
Rule 1 Rule 2
23. !23
Hard to rule out rules
~75% correct!
If using rules
~50% correct!
If using rules
Rule 1 Rule 2
Rule 1 Rule 2
24. !24
Hard to rule out rules
75% correct!
If using rules
50% correct!
If using rules
Hard to say if results reflect procedural learning
or perseveration with rules
25. !25
How to rule out rules:
Turn off procedural learning during training
See if results hold up
34. !34
471.05/GGG21 - Savings in visuomotor adaptation depends on perturbation magnitude
!
J. R. MOREHEAD, S. QASIM, M. CROSSLEY, R. B. IVRY;
!
471. Voluntary Motor Control: Motor Learning II
Mon, Nov 11, 1:00 - 5:00 PM
771.17/KKK10 - A temporal-difference dopamine-dependent spiking network account of
instrumental contingency degradation
!
M. J. CROSSLEY, F. ASHBY;
!
771. Neural Mechanisms of Appetitive Behavior
Wed, Nov 13, 8:00 AM - 12:00 PM
842.19/VV19 - The difficulties of rapid switching between declarative and procedural
learning systems
!
J. L. ROEDER, M. J. CROSSLEY, G. CANTWELL, F. ASHBY;
!
842. Human Navigation and Spatial Representation
Wed, Nov 13, 1:00 - 5:00 PM
37. !37
Appropriate Control Conditions?
The issue is that we don’t know what size interference to
expect from a rotation during transfer with pure II training.
This is a fair point. We know that pos and neg aren’t different
from each other if left in isolation, but we can not rule out the
possibility that there is an innate difference in the size of the
rotation interference. However, Experiment 2 addresses this
since we can make the difference disappear with FB delay.
39. !39
1) Procedural and declarative systems compete for control
of motor resources, preventing trial-by-trial switching
between procedural and declarative strategies under normal
circumstances.
!
2) This competition can be reduced, and trial-by-trial
switching facilitated, by incorporating explicit cues to signal
which strategy is appropriate for a given stimulus.
!
3) Learning in the procedural system occurs even when the
declarative system is in control of behavior.
!
4) computational cognitive neuroscience model M1 to
striatal medium spiny neurons.
My abstract promised too much
40. No perfect terminology
• Procedural vs declarative
• Information-Integration vs rule-based
• Habitual vs goal-directed
• Model-free vs model-based
• Implicit vs explicit
!40