IMPLICIT MEMORY: A “HIDDEN WORLD?”

• Tasks and terms– “indirect” (vs. direct) memory tests: no

memory judgments; assess effects of prior exposure on• Fragment completion• Perceptual identification• Repetition and “feature” priming• Other decisions and actions

– Implicit (vs. explicit) memory: the memory systems and/ or processes that (largely) mediate performance in indirect memory tests

• Contrast to:– Incidental learning: no reference to

memory test during study– Implicit learning: of patterns or

correlations without intent or awareness

Anecdotal Examples of Implicit Memory

• Cases of “unconscious plagiarism”– George Harrison and the Chiffons– Freud’s “discovery” of universal

bisexuality, and Fliess’ reaction

• Use of expert knowledge– Peter Bonyhard: helped IBM develop

mag-resist disk drives, barred from working with competitor Seagate

• Implicit memory for traumatic events– Amnesia for rape on a brick path, but

words “brick” and “path” come to mind– Global amnesia, home is unfamiliar, but

“recently dreamed of that house”

• Implicit memory for words spoken during anesthesia– Kilstrohm & Schacter (1990)

THE SEARCH FOR DISSOCIATIONS

• Stochastic– Performance in IM and EM tasks given

same study is uncorrelated

• Functional– Weak: variable X influences one kind of

test, (not) the other• Levels of processing• Modality

– Strong: variable X has opposite effects on IM and EM tests• Read versus generate (Jacoby 83)

• Population– A functional dissociation where X is a

group factor (amnestics vs. controls)

• Reverse Association– X affects A and B the same, Y has

opposite effects on A and B, in same data set (Dunn & Kirsner, 1988)

A CAPSULE HISTORY of IMPLICIT MEMORY

• Late 19th century– Dissociations in the clinic (Dunn, 1845;

Claparede, 1889)– Savings without explicit memory

(Ebbinghaus, 1885)– Habit versus memory (James, 1890;

Bergson, 1911)

• 1970’s– Controlled studies of priming in

amnestics

• HM can learn motor skills• Amnestics show normal

fragment-completion priming (Warrington & Weiskrantz, 1970)

recogn fragment ID

Amnestics .42 .46

Controls .75 .45

• Demonstrations of implicit memory in normals– Jacoby & Dallas (1981):

• Depth affects recognition, not priming• Modality affects priming, not recog

– Tulving, Schacter & Stark (1982):

• much less forgetting for implicit tasks

– Jacoby (1983):• Opposite effects of context and

generation on implicit and explicit tasksNo context context generate

XXX-COLD HOT-COLD HOT-XXX

Jacoby, 1983

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

no context context generate

Study Task

Pro

ba

bil

ity

corr

ect

Identification

Recognition

• Demonstrations of implicit memory in normals (cont’d)

– Graf & Schacter (1987):

• Little interference with implicit tasks

Word pairs studied (AB)

RI: AB AD -- AB PI: AD AB -- AB

Control group learns CD

RI PICtl Exp Ctl Exp

Cued recall .55 .40 .67 .45Fragment

Completion .34 .32 .32 .35

THEORETICAL ACCOUNTS OF IMPLICIT MEMORY

• The activation view (Graf & Mandler, 1987)– IM as a subset of EM processes

• IM reflects activation of prior memories• EM requires integration / elaboration

– Problems:• Amnestics can learn new associations• Priming can last for months

• The systems view (Tulving, 1985; Schacter, 1987)– IM based on procedural system, EM on

declarative system• EM more advanced• Explains neuroanatomic dissociations

– Problems:• A system for every dissociation?• Lack of consensus about criteria

• The processing view (Roediger, Weldon & Challis, 1987)– Transfer-appropriate memory tests

• IM : data-driven processing• EM: conceptually-driven processing

– Dissociations can be TAP-based (Blaxton, 1989)

“generate” (vs. read) gives better memory for conceptually-driven tests free recall (EM) semantic cued recall (EM) Jeopardy question-answers (IM)and worse memory for data-driven tests fragment completion (IM) graphemically-cued recall (EM)

– Problems:• Fuzzy bounds of processes• Can become circular• Doesn’t handle amnestic data well

THE PROCESS-DISSOCIATION APPROACH (Jacoby, 1991)

• The problem of “process-impure tests”– Jacoby’s process-dissociation technique– Assumes indendent concious (C) and

unconscious (U) contributions to memory– To dissociate these:

two sets of items presented (e.g., some read, some heard)

inclusion task: recall allexclusion task: recall only heard items

p[corr]inclusion = p[C] + p[U] – p[U] x p[C]

= p[C] + p[U] x p[1-C]

p[corr]exclusion = p[U] x p[1-C]

so: p[C] = inclusion – exclusion then solve first equation for U

• Applying Process Dissociation: Jacoby, Toth & Yonelinas (1993)

study presentation

Read Heard Incl Excl Incl Excl

Full attn .61 .36 .47 .34

Divided .46 .46 .42 .46

Estimated contributions of C and U to memory:

C(conscious) U(automatic)

Full attn .25 .47

Divided .00 .46

Controversies about independence and other assumptions