Unconscious influences on Unconscious influences on conscious forgettingconscious forgetting

Lili SahakyanLeilani B. Goodmon

Presentation for Festschrift in honor of Do glas L NelsonPresentation for Festschrift in honor of Douglas L. NelsonMarch, 2008

Standard List Method Paradigmg

List 1 List 1List 1

Forget

List 1

Remember

List 2

Forget

List 2

Remember

Remember

List 2 List 2

RememberRemember Remember

M TMemory Test

Typical Findingsyp g

List 1 List 1List 1

Forget

List 1

Remember

List 2

Forget

List 2

Remember

Remember

List 2 List 2

RememberRemember Remember

M TMemory Test

Context Change as a Mechanism of Directed Context Change as a Mechanism of Directed ggForgettingForgetting

F group R group

List 1 List 1

F - group R - group

List 1

Forget

List 1

Remember

List 2 List 2

Recall Recall

Associative Structure of Associative Structure of PLANETPLANET

VENUS PLUTO

MARS

PLANETUNIVERSE MOON

EARTHSATURN

SPACESATURN

STARSTAR

Memory Test = ExtraMemory Test = Extra--List CuingList Cuing

Studied TargetsPLANET

WINPOLITECORKCORK

.

.

.Test Cues

SPACE – ?SPACE ?CONTEST – ?MANNERS – ?STOPPER ?STOPPER – ?

Effects of PreEffects of Pre--existing Associationsexisting Associations

Most effects of pre-existing associations can be grouped by the types of connections they involve:by the types of connections they involve:

target-to-associates linksg

associates-to-associates links

associates-to-target links

Large Set Size (17)Large Set Size (17)party

wall

wedding Xmas

pretty

DECORATIONdesign

colornice

wall

house

DECORATIONornament

treecreate

cake

balloon

armyflower holida

ydirty

detergent

clotheswork

dirtyt

Small Set Size (8)Small Set Size (8) soap baskt

LAUNDRY wash

( )( )et

clean

TargetTarget--toto--Cue StrengthCue StrengthA1

A18 A3A21

A5

A4

A

A16

A17

TARGETTARGET A6

A5

A7A13

A14

A15

A8A10A12

13

A11

A99

TargetTarget--toto--Cue StrengthCue StrengthA1

A18 A3A2A1

A18 A3A2

A5

A1

A4

A

A16

A172A1

A5

A4

A

A16

A172

A5

A7

TARGETTARGET A6

A13

A14

A15 TARGETTARGET A6

A5

A7A13

A14

A15

better test cue

A8A10A12

13

A11

A9

A8A10A12

13

A11

A999

Target-to-Cue Strength EffectTarget to Cue Strength EffectAssociates that are more strongly activated by the target during study are more successful in retrieving the target atmore successful in retrieving the target at test (compared to weaker associates)

High connectivityHigh connectivityRUDE

KIND

NASTY

GENEROUS

NICE

GENEROUS POLITE

L ti itL ti itMANNERS

SWEET SCREW

Low connectivityLow connectivity

BOARDCHAMPAGNE

WINEFLOATCORK

BOTTLESTOPPER

High resonanceHigh resonanceLOTTERY

ACHIEVEMENTLOTTERY

MONEY

Low resonanceLow resonanceLOSE

VICTORY

WIN

SCREWCONQUER

BOARDCHAMPAGNE

WINEFLOATCORK

BOTTLESTOPPER

Context moderates these effects…Context moderates these effects…

Changing physical or mental context between the study and test reduces the influence of pre-existingand test reduces the influence of pre-existing associations (e.g., Nelson et al., 1998; Nelson & Goodmon, 2002; Nelson, Goodmon & Akirmak, 2007; Nelson, Goodmon, & Ceo, 2007)

Implications for Current StudiesImplications for Current StudiesIf directed forgetting involves a context change, we should observe a reduction of effects of associative variables in the Forget groupForget group

Exp 1: Set Size (T-A links)

E 2 B k d hExp 2: Backward strength (T-A links)

Exp 3: Connectivity (A-A links)Exp 3: Connectivity (A-A links)

Exp 4: Resonance (A-T links)

General MethodGeneral Method

8 items of one level of associative variable

P t ti t 4 /it8 items of another level of associative variable

Presentation rate: 4 s/item

Controlled Associative Variables

RememberForget Set sizeConnectivity

8 items of one level of

ResonanceForward strength

Backward strength N b d t th fassociative variable

8 items of another level of associative variable

Number and strength of indirect connections

Target frequencyConcreteness

Test: Extra-list Cued RecallConcreteness

Set Size EffectSet Size Effect60

.50

.60

all

small setlarge set

30

.40

List

1 R

eca

.20

.30

Prop

ortio

n L

.00

.10

P

.00Remember Forget

Set size: F(1, 70) = 43.12, p < .001Cue: F(1, 70) = 10.72, p < .01Set size x Cue : F(1, 70) = 5.65, p < .05

Connectivity EffectConnectivity Effect60

.50

.60

all

high connectivitylow connectivity

30

.40

List

1 R

eca

.20

.30

Prop

ortio

n

.00

.10

P

Remember Forget

Connectivity: F(1, 70) = 101.24, p < .001Cue: F(1, 70) = 10.37, p < .01Connectivity x Cue : F(1, 70) = 7.66, p < .01

Backward TargetBackward Target--toto--Cue Strength EffectCue Strength Effect60

.50

.60

ll

high strengthlow strength

30

.40

List

1 R

ecal

.20

.30

ropo

rtion

L

.10

Pr

.00Remember Forget

Backward Strength: F(1, 70) = 18.97, p < .001Cue: F(1, 70) = 11.86, p < .001Backward Strength x Cue : F(1, 70) = 4.46, p < .05

Resonance EffectResonance Effect50

.40

.50

all

high resonancelow resonance

.30

List

1 R

eca

.20

Prop

ortio

n

.00

.10

P

Remember Forget

Resonance: F(1, 70) = 8.07, p < .01Cue: F(1, 70) = 9.89, p < .01Resonance x Cue : F(1, 70) = 3.38, p = .07

Explaining the effects of context disruptionExplaining the effects of context disruptionInteractive cuing assumption the probability of target recovery is jointly determined by an extra-list cue and context cues

Pr( T | Q, C) = Pr (T | Q) * (t)–c ,

where t is the length of the retention interval, and c is a context loss parameter that varies with the procedure used to induce the context loss It must be estimated from the dataloss. It must be estimated from the data

Pr( T | Q, C) = Pr (T | Q) * C ,

where C represents the degree of contextual loss induced by experimental manipulation (i.e., forget instruction).

at theEstimatingEstimating CC parameterparameter

at theBecause the model predicts the recall rate of individual cue-target pairings, the data was scored at the item level to estimate C.

Pr( T | Q, C) = Pr (T | Q) * C

We regressed the observed recall of individual cue-target pairs on the predicted recall rate provided by PIER modelp p p y

ved

reca

ll

The slope of the best fitting regression line estimates C withPIER predictionO

bser

v

The slope of the best-fitting regression line estimates C, with lower values of C imply greater change of context

Remember = (.59)x + .17

Forget = (.31)x + .16

Remember = (.58)x + .19

Forget = ( 29) + 19Forget = (.29)x + .19

Remember = (.59)x + .11( )

Forget = (.35)x + .11

Remember = ( 59)x + 12Remember = (.59)x + .12

Forget = (.34)x + .11

PIER Model PredictionsPIER Model Predictions

The model predicts that each link in the associative network should be affected to the same degree by contextual changeshould be affected to the same degree by contextual change

… directed forgetting should have a similar impact on recall regardless of the manipulated implicit variable

C estimates should be approximately the same for the set…. C estimates should be approximately the same for the set size, connectivity, resonance, or backward strength studies.

Backward T Q strengthBackward T-Q strengthF(1, 30) = 21.35, p < .001, R2=.42

S t iSet sizeF(1, 30) = 33.52, p < .001, R2=.55

ConnectivityConnectivityF(1, 30) = 28.55, p < .001, R2=.51

ResonanceResonanceF(1, 30) = 37.85, p < .001, R2=.58

Backward T-Q strengthF(1, 30) = 5.06, p < .05, R2=.15

Set sizeF(1, 30) = 3.95, p = .06, R2=.13

ConnectivityF(1, 30) = 11.44, p < .01, R2=.29

ResonanceF(1, 30) = 4.81, p < .05, R2=.15

CONCLUSIONS

Di t d f tti b bt i d ith t li t d llDirected forgetting can be obtained with extra-list cued recall

Directed forgetting produces effects similar to thoseDirected forgetting produces effects similar to those observed in studies of context disruption – it reduces the influence of pre-existing associations on memoryp g y

Model fits revealed that all the links in the associative network were affected to the same extent by directed forgetting across the studies

Conscious attempts to forget impact not only consciously accessible information but also information that is implicitlyaccessible information but also information that is implicitly activated and yet is not consciously reportable

THANK YOUTHANK YOU

APPENDIXAPPENDIX

Th PIER M d l P iThe PIER Model: Processing Implicit &Explicit Representations

The Target Activation ProcessThe Target Activation ProcessThe Cue – Target Intersection Process

Target Activation EquationTarget Activation Equation

STi( )= S T,T( )+ SAi,T( )n

∑⎡

⎣⎢ ⎤

⎦⎥ +n

∑ S T,Aj( )+ S Ai,Aj( )n

∑⎡

⎣⎢ ⎤

⎦⎥i( ) ,( ) i,( )

i=1∑

⎣ ⎢ ⎦ ⎥ j=1∑ , j( ) i, j( )

i=1∑

⎣ ⎢ ⎦⎥

Uses summation rule to add up the activation from the following sources:

target self strength (nominally 1.0)associates to targetassociates-to-targettarget-to-associatesassociates-to-associates

Computing Target Activation EquationComputing Target Activation EquationTARGET A1 A2 A3 A4 A5 A6 A7 A8

TARGET 1.00 .25 .23 .09 .08 .06 .05 .04 .03A 03 02 29A1 .03 .02 .29A2A3 .01 .12 .22 .03A 01A4 .01A5 .53A6 .08 .24AA7A8 .08 .16 .12 .16 .28

Net Strength1 12 79 53 29 08 97 36 04 031.12 .79 .53 .29 .08 .97 .36 .04 .03

TARGET’s Activated State

[(1.12)+(.79)+(.53)+(.29)+(.08)+(.97)+(.36)+(.04)+(.03)] = 4.10

Cue-Target Intersection StrengthTakes into account the associative information activated by the cue and by the target – overlap of two implicit

S Qq ,TI( )= SqtS It( )+ SqqS Iq( )representations

where q ≠ t

Qq , I( ) qt t( ) qq q( )

Sqt is forward cue-to-target strength from free association normsS(I ) is implicit activation strength of the targetS(It) is implicit activation strength of the targetSqq is cue-to-cue self strength (nominally, 1.0)S(Iq) is implicit activation strength of the cueS(Iq) is implicit activation strength of the cue

Probability of Target Recovery on immediate test

Cue-Target Intersection

Pr TI |Q q( )=S Q q,T( )

S Q T( ) A An∑

n∑

( )S Q q,T( )+ Aq + Att

∑q∑

Cue-Target Intersection CompetingCue Target Intersection

Competing Associates of Cue

Competing Associates of Target

Retrieval Matrix using ARetrieval Matrix using A88 as a test cue on as a test cue on an immediate testan immediate testan immediate testan immediate test

C T t T t CUE C tit C titCue-Target Intersection

Target CUE A8

Competitors (target)

Competitors (cue)

Target 4.10 .0338 08C 08 1 00

Q-T Intersection Strength = (.08 * 4.10) + (1.00 * .03) = 37

.38 .08Cue .08 1.00

.37

.37Target Recovery = .47=[(.37) + (.38 + .08)]

Selected papers from Sahakyan Lab: Sahakyan, L. & Delaney, P. F. (2005). Directed forgetting in recognition testing and incidental learning: Support for a two-factor account. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(4), 789-801. Sahakyan, L. (2004). Destructive effects of “forget” instructions. Psychonomic Bulletin & Review, 11(3), 555-559. Sahakyan, L. & Kelley, C.M. (2002). A contextual change account of the directed forgetting effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28(6), 1064-1072. Sahakyan, L. & Delaney, P. F. (2003). Can encoding differences explain the benefits of directed forgetting in the list method paradigm? Journal of Memory and Language, 48(1), 195-206. Sahakyan, L. & Goodmon, L. B. (2007). The influence of directional associations on

directed forgetting and interference. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33(6), 1035-1049.

Key Papers Cited in the Presentation File:

Nelson, D. L., & McEvoy, C. L. (2005). Implicitly activated memories: The missing links of remembering. In C. Izawa & N. Ohta (Eds.), Human Learning and Memory: Advances in Theory and Application: The 4th Tsukuba International Conference on Memory (pp. 177-198). Mahwah, NJ: Erlbaum.

Nelson, D.L., McKinney, V.M., Gee, N.R., & Janczura, G.A. (1998). Interpreting the influence of implcitly activated memories on recall and recognition. Psychological Review, 105, 299-324.

Nelson, D. L., & Goodmon, L.B. (2003). Disrupting attention: The need for retrieval cues in working memory theories. Memory & Cognition, 31, 65-76.

Nelson, D. L., Goodmon, L.B., & Akirmak, U. (2007). Implicitly activated memories are associated to general context cues. Memory & Cognition, 35, 1878-1891.

Nelson, D. L., Goodmon, L.B., & Ceo, D. (2007). How does delayed testing

reduce effects of implicit memory: Context infusion or cuing with context? Memory & Cognition, 35, 1014-1023.