Memory Consolidation: Transformation

Memory Consolidation: Transformation

CollaboratorsRebecca GomezAlmut HupbachOliver Hardt

EVENT ENCODE CONSOLIDATE FIXED MEMORY


REACTIVATION

SLEEP


REACTIVATION

RECONSOLIDATION

PROTEIN SYNTHESIS INHIBITORS

UNFIXED MEMORY


REACTIVATION

RECONSOLIDATION

NEW/ALTERED EXPERIENCE

TRANSFORMED MEMORY

EVENT ENCODE CONSOLIDATE TRANSFORMED MEMORY

REACTIVATION

SLEEP

Reactivation Transformation

Set 1 Set 2

Reactivation Transformation

Basic Design

Group DAY 1(Monday)

DAY 2(Wednesday)

DAY 3(Friday)

Reminder Learn Set 1 Same context, same experimenter, reminder questionLearn Set 2

Recall Set 1

No-Reminder Learn Set 1 Different context, different experimenter, no questionLearn Set 2

Recall Set 1

Interference Control

Learn Set 1 - Recall Set 1

Reminder Updating Effect

Hupbach, Gomez, Hardt & Nadel, Learning & Memory, 2007

N = 12 in all groups

% R

ecal

l

REMINDER NO REMINDER CONTROL

Set 1 Items

Set 2 Items

Immediate Recall: Reconsolidation or Retroactive Inhibition



Source Confusion?


N = 12 in both groups

Absence of intrusions from Set 1 into Set 2 -- effect is asymmetrical and not simple source confusion

REMIND NO REMIND

Set 1 Items

Set 2 Items

Persistence of Updated Memory: Delayed RecallMonday: learn set 1Wednesday: remind or not, learn set 21 or 2 weeks later: recall either Set 1 or Set 2

unpublished

The effect persists for at least 2 weeks, suggesting it is a real change in memory

Persistence: Delayed Reactivation1st Monday: learn set 1 2nd Monday: remind, learn set 23rd Monday: recall Set 1

Hupbach, Gomez, Hardt & Nadel, unpubl.

The effect can be elicited at least a week after initial storage of a memory

What Triggers Updating?Group DAY 1

(Monday)DAY 2 (Wednesday)

DAY 3 (Friday)

Context (C) Learn Set 1 Same context, new experimenter, no questionLearn Set 2

Recall Set 1

Experimenter (E) Learn Set 1 Different context, same experimenter, no questionLearn Set 2

Recall Set 1

Question (Q) Learn Set 1 Different context, new experimenter, reminder questionLearn Set 2

Recall: Set 1

Reminder Effects - One Cue

Hupbach, Gomez, Hardt & Nadel, 2008.


Reminder Effects -- Two Cues

Hupbach, Gomez, Hardt & Nadel, 2008.

N = 12 N = 8 N = 12

Context & Experimenter Context & Question Question & Experimenter

Set 1 Items

Set 2 Items

Imagining Context

Hupbach, Gomez, Hardt & Nadel, unpubl.

Set 1 Items

Set 2 Items

N = 12

Conclusions re Context

• Context reminder critical to updating and reconsolidation– Recall acts as another training trial, does not initiate update

• In episodic memory context determines whether – new memory formed

OR – old memory updated

• Similar result seen in place cell “remapping” studies

• BUT, a caveat re the role of context

Testing in a familiar context: 5 year-olds at home

Context alone fails to elicit updating in a familiar context

Testing in a familiar context: 5 year-olds at home

Context alone fails to elicit updating in a familiar context

5 year-olds tested in day-care setting

N = 11 N = 7

But, context alone does elicit updating in an unfamiliar context

REMIND NO REMIND

Set 1 Items

Set 2 Items

Bigger Picture

• Memory is not fixed - perhaps ever– Misinformation effect– Hindsight bias

• Adaptive nature of memory malleability• What then is “consolidation”?

– Not strengthening but transforming, assimilating

Sleep Transformation

Study effects of sleep by exposing infants to material prior to a nap and testing afterwards

Design

• Toddlers exposed to material 4 hrs before lab visit (48 15-month-olds)

• Conditions (exposure to an artificial language)– Nap– No Nap– Nap control

• A Minimitter actiwatch with computer driven chip attached to infant’s ankle and used to record body movements

Language 1 Language 2

S{ aXbcXd }

S { aXdcXb }

X x1, x2, ….xn

Nonadjacent dependency learning

15-month-olds Familiarized with one of two artificial languages

vot-kicey-jic, pel-wadim-rud……

Gómez (2002)Gómez & Maye (2005)

Variability manipulation

|X| = 12

X1

X2

X3

.

.

.

.

.X12

ac

bd

|X| = 24

X1

X2

X3

.

.

.

.

.

.

.

.

.

.

.

.X24

ac

bd

|X| = 3

X1

X2

X3

ac

bd

Gómez (2002); Gómez & Maye (2005)• Nonadjacent dependencies heard equally often in each condition • Difference between conditions was size of pool from which middle element drawn.

• Learners only track non-adjacencies when adjacent dependencies are sufficiently low, when |X| = 24.

Variability manipulation|X| = 12

X1

X2

X3

.

.

.

.

.X12

ac

bd

|X| = 24

X1

X2

X3

.

.

.

.

.

.

.

.

.

.

.

.X24

ac

bd

|X| = 3

X1

X2

X3

ac

bd

Nap control

Nap and No-nap groups

Familiarization

15-minute incidental exposure

vot wadim jic | pel kicey rud | vot coomo jic | vot kicey jic | pel gople rud | vot fengle jic | pel benez rud | pel wadim rud | vot loga jic | pel vamey rud | pel taspu rud | pel fengle rud | vot hiftam jic……….

pel wadim rud vs. pel wadim jic

Test

Head-turn preference procedure

Infant controls the amount of exposure on any given trial

Veridical memory vs. abstraction

• Infants could remember specific information about strings themselves

• Or, could acquire a rule focusing on relationship between nonadjacent pairs. If so, they should detect nonadjacent dependencies in novel strings.

• Responses scored according to veridical memory and abstraction.

Predictions

• Sleep (or delay) could change memory quantitatively or qualitatively – If delay is triggering factor, nap and no-nap infants will

perform identically – If sleep is trigger then performance should differ

between nap and no-nap conditions

Naps promote abstraction!

Gómez, Bootzin, & Nadel (2006)

Mea

n lo

okin

g tim

es (s

ec)

*

*

Veridical Memory

Abstraction: Difference conditional on first post-sleep trial

What are infants learning?

• Abstraction may take form of a greater weighting given to relationship between first and third words in strings

• This weighting translates into detection of nonadjacent dependencies in similar (but not identical) strings.

Additional questions1. Is the abstraction effect a transient one, caused

by infants being more rested after a nap, or can we observed it 24 hours later before their nap?

• 15-month-olds tested 24 hrs later in Nap condition

2. Is the abstraction effect dependent on an immediate nap, or like adults, will any sleep within a 24-hour period do?

• 15-month-olds tested 24 hrs later in No-nap condition

Is the effect dependent on infants being well-rested? No, it occurs 24-hours later before a new nap

Hupbach, Gómez, Bootzin, & Nadel (in press)

Mea

n lo

okin

g tim

e di

ffere

nces

(sec

)

Veridical Memory

Abstraction: Difference conditional on first post-sleep trial*

Is the effect dependent on an immediate nap? Yes. 24-hours later there was no retention in the No-nap condition

Hupbach, Gómez, Bootzin, & Nadel (in press)

Mea

n lo

okin

g tim

e di

ffere

nces

(sec

)

Veridical Memory

Abstraction: Difference conditional on first post-sleep trial*

Summary

• Infants in Nap and No-nap conditions were exposed to an artificial language 4 or 24 hours prior to test.

• No immediate nap: Infants retained veridical memory over a 4-hour delay but showed no retention over 24 hours.

• Immediate nap: Sleep facilitated abstraction 4 and 24 hours after exposure.

How could sleep make memory abstract?

1. Sensitivity to both specific and abstract information initially, but weight these differently before and after sleep

2. Forget specific details of stimulus with sleep3. Sleep protracts learning-dependent processing

necessary for extraction of general patterns (O’Reilly & Rudy, 2000; Wilson & McNaughton, 1994)

Implications

• Transformation occurring with sleep introduces flexibility– Infants detected general pattern in artificial language whether it

was instantiated exactly as before or not. • Abstraction is a crucial form of memory change for

developing learners who must retain key aspects of experience while generalizing to new information.

• Sleep appears to be instrumental in this process

Memory Consolidation: Transformation

Documents