University of Groningen
Individual differences in learning rate are reflected in feedback related brain processes van den Berg, Berry
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Publication date: 2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
van den Berg, B. (2020). Individual differences in learning rate are reflected in feedback related brain processes.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Data
Modelled data Data
I
NDIVIDUAL
DIFFERENCES
IN
LEARNING
RATE
ARE
REFLECTED
IN
FEEDBACK
RELATED
BRAIN
PROCESSES
VIRTUAL Cognitive Neuroscience Society 2020
berry.van.den.berg@rug.nl
Here, participants chose on each trial either a face or a house, which was
fol-lowed by receiving either a zero (0) gain (+) or a loss (-) of different
magni-tudes (0:8)
On each set of 20 trials either the face or house was the set-winner and was more likely to yield net gains.
Summary
Feedback processing was marked
by amplitude modulations induced by both magnitude and valence in the earlyl atency range with distinct topo-graphical effects. Specifically,
va-lence showed a classical negative polarity feedback related negativity (FRN). Magnitude showed a frontal postive deflection for larger
out-comes
Participants learned over the course of 20 trials to choose the stimulus that yielded
higher net-gains. There was substantial variability in how well participants were able
to do so.
ERP amplitudes in the late latency range were modulated by feedbackn-1. High learn-ing rates were characterized by an LPC
that was stronger modulated by previous feedback information as opposed to low learning rates.
The ability to use and integrate feedback information over time is key to our ability to learn and decision making. Although it is fairly well established how the brain processes outcomes on a single trial, it is less well studied how
these processes depend on encountered information on previous trials.
In sum, this study provides a
novel and important set of
findings providing more
in-sight into how the brain
dy-namically integrates feedback
information over multiple trials
to guide decision making in an
uncertain world.
+ + + + ++8
duration (ms)
1000-1500 until resp.
(max 1200)
300
700-900
500
Fixation Fixation Feedback Highlight choice Choice Cue2000-2500
Berry
van den Berg, Timothy Sondej, Celina Pütz, Marty G Woldorff & Monicque M Lorist
University of Groningen; Duke University
P
ROCESSING
OF
FEEDBACK
I
NTEGRATION
OF
F
EEDBACK
ACROSS
TRIALS
[-7; -8] 0 [-5; -6] [-3; -4] [-1; -2] [+1; +2] [+3; +4] [+5; +6] [+7; +8] feedback -2 2 6 10 14 18 -200 0 200 400 600 800 -2 2 6 10 14 18 time (ms) amplitude ( μV)H
OW
DO
WE
USE
FEEDBACK
?
0.00 0.05 0.10individual learning rate β trial
trial number 0.5 0.6 0.7 0.8 5 10 15 20 p(choosing set−winner)
learning rate < median learning rate > median average
250 to 350
Loss minus Gain
250 to 350 250 to 350 Gain: [+5:+8] minus Gain [+1:+4] Loss: [-8:-5] minus Loss [-4:-1] amplitude (μV) -3 3 500 to 600
Loss minus Gain
Gain: [+5:+8] minus Gain [+1:+4] Loss: [-8:-5] minus Loss[-4:-1] 500 to 600 500 to 600 amplitude (μV) -3 3 feedbackn 7.5 10.0 12.5 15.0 17.5 −8 −6 −4 −2 0 +2 +4 +6 +8 feedback n feedbackn-1 2 amplitude (µV) 14 −8 −6 −4 −20 +2 +4 +6 +8 −8−6−4−2 0 +2+4+6+8 4 amplitude (µV)20 −8 −6 −4 −20 +2 +4 +6 +8 −8−6−4−2 0 +2+4+6+8 feedbackn-1 feedback n amplitude (µ V) amplitude (µ V) feedbackn 6 8 10 12 −8 −6 −4 −2 0 +2 +4 +6 +8 −8−6 −4 −20 +2 +4 +6 +8 −8−6−4−2 0 +2+4+6+8 −8 −6 −4 −20 +2 +4 +6 +8 −8−6−4−2 0 +2+4+6+8 −8−6−4−2 0 2 4 6 8 −8−6−4−2 0 2 4 6 8 −8 −6 −4 −20 +2 +4 +6 +8 feedback n
learning rate < medianamplitude (μV)
3 20 Data feedback n feedbackn-1 Modelled data feedbackn-1
learning rate > median
amplitude (μV)m 3 20 −8−6−4−2 0 +2+4+6+8 −8−6−4−2 0 +2+4+6+8 −8 −6 −4 −20 +2 +4 +6 +8 −8−6−4−2 0 +2+4+6+8 −8−6−4−2 0 +2+4+6+8 −8 −6 −4 −20 +2 +4 +6 +8 Late latency range
Early latency range
Processes in the late latency range
(500-600ms) were modulated by both cur-rent feedback contents, and also by the
feedback on the previous trial, indicating an integrative role. Strikingly, this integration
was even further modulated by the individu-al participants’ learning rate. As such, the processes that are marked by the LPC sub-serve a dynamic updating role that is highly susceptible to prior information.
Feedback processing was characterized by amplitudes in the early latency range
(250-350ms) being modulated by the magni-tude and valence on the current trial. In this early time range we found minimal influence of the feedback of the previous trial,
sug-gesting a feedback registration mechanism, that is not modulated by prior information
(i.e. expectation).
Modelled data
Both magnitude and valence modu-lated amplitudes in the later latency range. These modulations had si-miliar scalp topographies (sugges-tive of a modulation of the Late Pos-tive Complex [LPC]), suggesting a similiar neuro-cognitive process by both factors is involved in this later time period. feedback onset Early processing (250-350ms) Late processing(500-600ms) feedbackn-1
learning rate < median learning rate > median
ERP amplitudes in the early latency range were slightly modulated by feedbackn-1 but not learning rate.