Investigating the impact of memory reactivation on the successful forgetting of negative memories

Paula P. Brooks1,3, Justin Hulbert4, Arlene Lormestoire1, Maureen Ritchey3, & Kenneth Norman1,2

1Princeton Neuroscience Institute and 2Department of Psychology, Princeton University; 3Department of Psychology, Boston College; 4Department of Psychology, Bard College

Introduction

Hypothesis

Methodology

up to 5 s (ITI = 1 s)

Phase 1: Rating Scenes

6 s (ITI = 0.5 s)

Y or N? 1 2

CORRECT

< 4s

< 4s

2.5 s (ITI = 0.5 s)*Criterion test (70% learning criterion) not shown.

Phase 2: Study Face-Scene Pairs (incl. 3 Test-Feedback Cycles + Criterion Test)

2-hour break (± 30 min)

3 s

3 s (ITI = 1.5 – 4.5 s)

Phase 3: Think/No-Think Task

one-back image detection task

•negative scene•neutral scene•scrambled scene•object

Phase 4: Localizer Phase 5: Final Cued Recall Test

41+

41brown house on field…4 s

11 sbeep then ITI = 3s

N = 26 (target sample size = 50)Stopping an unwanted memory from coming to mind (i.e., memory suppression) might help to regulate negative memories. Memory suppression has commonly been studied using the think/no-think paradigm1.

Although multiple studies2 have demonstrated memory suppression effects, others have failed to replicate these findings3,4.

What causes the suppression of negative memories to succeed or fail?

DAY ONEDepression & anxiety questionnaires

DAY TWOStudy face-scene pairs

Think/no-think (TNT) memory suppression

Surprise memory test

• 1-7 days apart• Detailed schematic of

day two on the right

Relate memory reactivation during TNT task to subsequent memory performance (using Probabilistic Curve Induction and Testing)

Learning Phase TNT Phase Suppression Score =Baseline – No-Think

baseline no-think think

Perc

ent R

ecal

led

TNT Conditions

Anderson & Hanslmayr, 2014

We propose that differences in memory reactivation strength might lead to variability in suppression effects.

Nonmonotic plasticity hypothesis (NMPH)5,6

Some memories might be challenging to suppress because they are too strongly reactivated. In fact, individual variability (e.g., depression level) might impact memory reactivation and thus influence memory suppression effect.

Specifically, we predict an increase in memory reactivation as negative valence increases. We expect participants with more depression to reactivate negative stimuli more strongly than those with lower depression. In turn, this should affect subsequent memory performance.

Special thanks to Sam Nastase and Lizzie McDevitt for fMRI guidance.

[1] Anderson & Green, 2001; [2] Anderson & Hanslmayr, 2014; [3] Bulevich, et al., 2006; [4] Hertel & Mahan, 2008; [5] Detre et al., 2013; [6] Ritvo, Turk-Browne, & Norman, 2019; [7] Küpper et al., 2014

* Trials repeated 12x

Next Steps• Only use TNT trials that participants got correct in

the criterion test, in line with prior work7.• Take into account individual variability in

depression level, since this might impact the degree of memory reactivation.

• Relate memory reactivation during the TNT trials to memory performance during the surprise cued recall task ( ). We expect to see a relationship similar to the NMPH schematic.

3

1

23

Preliminary Results

1 Built logistic regression (L-2 regularized, penalty 𝜆 = 50) scene classifiers

2 Applied scene classifier TNT data

We looked at the response 3 and 4 TRs after cue onset (accounting for hemodynamic lag). As expected, there was a higher probability estimate for scenes for think versus no-think trials.

Surprisingly, there were no differences in scene prediction probability across negative valence level. However, any effect might be washed out by differences in depression level.

*used bilateral occipitotemporal mask*

Clas

sifie

r Lab

elne

gativ

e &

neut

ral s

cene

ssc

ram

bled

sc

enes

obje

cts

rest

negative scenes

scrambled scenes

objectsneutral scenes

rest0.0

0.2

0.4

0.6

0.8

1.0

Real Category Label

Confusion Matrix

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Pred

icti

on P

roba

bilit

y

no-think thinkTNT Trial Condition

p < 0.001

Average probability estimate when applying the scene classifier

high medium low high medium low0.0

0.1

0.2

0.3

0.4

0.5

0.6

Degree of Negative Valence Degree of Negative Valence

Pred

icti

on P

roba

bilit

y

No-think trials Think trials

We also examined the time course of the prediction probability relative to the cue onset of the think and no-think trials. We see a peak in activity 3 and 4 TRs after the cue onset.

Asterisks denote significance.0.30

0.35

0.40

0.45

0.50

0.55

0.60

-1 0 1 2 3 4 5 6

no-think trialsthink trials

*

*

*

Average probability estimate for scene classifier across time

Pred

icti

on P

roba

bilit

y

Time (in TR, relative to cue onset)