Lecture 11 – Psyco 350, A1 Winter, 2011

Psyco 350 Lec #11– Slide 1

Lecture 11 – Psyco 350, A1Winter, 2011

N. R. Brown


Outline

• Implicit Memory– Introduction

– Dissociating Implicit & Explicit Memory

– Transfer Appropriate Processing

– Implicit Learning

• Dual Process Models– Recognition & Remember/Know

– Process Dissociation Procedure• Direct Tests


Evidence for Implicit Memory on Indirect Tests

• Indirect tests typically use improved performance as the measure of implicit memory

• Priming = the improvement in performance on a subsequent occasion due to processing on a previous occasion


Priming Effects

Exp condition = a prior exposure to stimControl = “no” prior exposure to stim

Priming:• fragment, stem, anagram:

dv – % complete: exp > control• Perceptional Identification:

dv -- % correct: exp > control• Lexical Decision:

dv – RT: exp < control


Dissociating Implicit & Explicit Memory

Dissociation: One variable affects one task differently than it affects another

_______________________________________

Tulving, Schacter, & Stark (1982)

Design:

Delay X Test Type .

1 hr fragment completion (indirect)

1 week recognition (direct)



Phase 1 Intentional Study (96 words)

1 hr delay

Phase 2 Recognition48 words

Frag Completion

48 words1 week delay

Phase 3 Frag Completion

48 words Recognition

48 words



Results:• Reco w/ dealy• frag unaffected by delay

Delay causes a dissociation between reco & frag tests.

Implication:

Test tap different “forms” of memory.


(Double) Dissociating Implicit & Explicit Memory

Jacoby (1983)

Aims: Using same materials demonstrate:

• explicit memory w/ depth of processing

• implicit memory w/ perceptual similarity

Materials selected so that:• as depth of processing , perceptual similarity


Jacoby (1983)

Design: Encoding Task X Test .

(Antonym) Generation recognition

Read (antonym in context) perceptual ID (40 ms)

Read (target alone)


Jacoby(1983): Two perspective on Encoding Tasks

Predictions:• Recognition: deeper processing should produce

better performance.• Percp ID: priming should become stronger as study

and test materials become more similar.

Task example LOP perc similarity

Generate hot deep low

Read in context hot – COLD mid mid

Read no context COLD shallow high


Jacoby (1983): Results• w/out prior exposure

(control): perc ID = 60%• In all conditions:

– Perc ID > 60% – priming

_________________________

• Reco with Depth of Processing

• Perc ID perc similarity (Perc ID LoP)Evidence for 2 types of

memory


Jacoby (1983): Results• w/out prior exposure

(control): perc ID = 60%• In all conditions:

– Perc ID > 60% – priming

_________________________

• Reco with Depth of Processing

• Perc ID perc similarity (Perc ID LoP)Evidence for 2 types of

memory


Transfer Appropriate Process: Theory

Assumes:

• Performance depends of match between processing at study and processing at test.

Analogous to encoding specificity.

• Two-types of Processes– Data-driven (perceptual) – processing of

physical features.

– Conceptually-driven (semantic) – processing for meaning


Transfer Appropriate Processing: Two Types of Indirect Test

Data-driven

(Perceptual):• fragment completion• stem completion• anagram completion• lexical decision• perceptual

identification

Conceptually-driven

(Semantic):• word association

doctor ??• category-instance

generation“name a mammal”

• general knowledge“The capital of the US is

…?”


Transfer Appropriate Process: Theory

Processing-type & memory task typically confounded:

• Direct tests require/benefit from conceptual processing

• Indirect tests require/benefit from data-driven processing.

However, it is possible to unconfound test-type from process-type (e.g.):

• Fragment-cued, recall test – data-driven, direct• general knowledge test – conceptual, indirect


Transfer Appropriate Processing: Blaxton (1989)

• Goal to demonstrate;– data-driven processing can affect direct tests

– data-driven processing do not necessarily affect indirect tests

• Design:

TEST TYPE

Study Mode X Explicitness X Level .

visual direct data-driven

auditory indirect conceptually-driven


Blaxton (1989): Four Types of Memory Test

Data-driven Conceptually-driven

Direct Graphic-cuedRecall

Free Recall

Indirect FragmentCompletion

GeneralKnowledge

Target word: bashful• graphic-cued recall: looks like “bushful”• free recall• frag completion: b_sh_u_• General knowledge: “Name one of the 7 dwarfs”


Blaxton (1989): Competing Predictions


Direct no modality effect no modality effect

Indirect visual > auditory visual > auditory

Standard view: modality match should affect only indirect tests for both implicit tests: visual > auditory for both explicit test: visual = auditory


Competing Predictions


Direct visual > auditory no modality effect

Indirect visual > auditory no modality effect

TAP View: modality match should affect data-driven tasks only. priming depends on match between study/test processing match & not on test instructions: for both data-driven tests: visual > auditory for both conceptually-driven tests: visual = auditory


Blaxton (1989): Results

Priming Effect (v > a) for data-driven tasks only:

• indirect: frag completion• direct: graphemic-cued

recall

Not all indirect tests display priming effect.

• Gen Know (indirect, conceptual): v = A

graphemic-cued recall

fragment completion

free recall

General knowledge


Transfer Appropriate Processing: Conclusions

• It is the match between processing at study and processing at test that produces priming effects.

• Priming can affect performance on both indirect tests and direct tests.

• Implication: at least for direct tests, performance reflects both implicit and explicit memory

• Question: Is it also the true that indirect tests reflect both implicit and explicit memory?


Implicit Learning

Learning w/out intention or awareness• Covariation (required for use of availability/familiarity

in judgment)• Frequency-of-occurrence• Sequence Learning

– Fixed sequences

– Rule-based sequences


Fixed Sequence Learning: Nissen & Bullemer (1987)

Task: Press button under 1 of 4 lights

Sequence Type:

• Repeating– same pattern reports every across every 10 trials

• Random – lights presented at random (no sequence).

Details:

800 10-trial sequences (!)


On each trial a light goes on

Just press corresponding button


Nissen and Bullemer (1987): Results:

•Repeated: RT rapidly over blocks

• Random: RT little changed over block

•Repeat-condition Ps unable to report sequence

Conclusion: Sequence learning/use was unconscious/implicit


Rule governed

Another N & B (1987)Exp.

Subjects are sensitive to the presence of the sequence even when they deny knowing that there was a sequence

Violates rules


Rule-based Sequence Learning (Reber)

Artificial Grammars – General Approach

Use Artificial Grammar to define/generate “grammatical sequences”


Example of an Artificial Grammar


Aim: Can people learn “grammatical” rules w/out intention:

Two Groups:• Grammatical– (implicitly) learn sets of grammar-generated letter sequences• Random – learn sets of randomly generated letter sequences

Test: Grammatically judgment:

50% grammatical

50% ungrammatical

Reber (1967)


Results:• Phase-1Learning: grammatical faster than random• Phase-2 Grammatical-judgment task:

– Grammatical group: 79%– Random group: chance

• Grammatical group could not state the rules explicitly

Conclusion:• Grammatical group (implicit) learned the rules/grammar during Task 1•Rules: facilitated string learning; enabled classification

Reber (1967)


Approaches to Implicit Learning

• Rules (Reber)

• Instance Based (Brooks)– Encode examples/instances

– assess similarity between target and stored instances

• Fragment-based (Perruchet)– Learn string fragments (bigrams, trigrams)

– reject strings lacking learned fragments

Problem – knowledge might be explicit.


Dual-Process Accounts

General View – two sources of info analytic, explicit, controlled nonanalytic, implicit, automaticAssumption – “no process pure tasks”

Interpretation problem for Indirect Tests:• contaminationIssue generalizes to:• direct tests• judgments tasks


Recognition: Background

Recognition Task (Radvansky, pp. 55-56)

Process of list of items: STIM 1 …. STIMn

Test:“Was STIMx on prior list?”

STIMx on list “OLD”

STIMx not on list “NEW”


Dual-Process Account of Recognition: Mandler (1980)

Reco judgments based on 2 types of information:• Recollection:

– Judgment based on successful retrieval of information about the study episode

• Familiarity:– Judgment based on assessed familiarity

(fluency).• Evidence:

– Remember/Know Judgments– Process Dissociations


Remember/Know

Two memory processes

2 phenomenal experiences

• remember = successful recollect of details of prior episode

• know = high levels familiarity, in the absence of recollection

• remember R

• know A


Remember/Know: General Method

@ study: manipulate some factor likely to manipulate recollection

@ test: recognition

“yes” “no”

r/k


Remember/Know: An Example – Rajaram (1993)

Exp 1. Levels of Processing• R: semantic > rhyme; K: deep = shallow

Exp 2. Pictures vs words• R: picture > words; K: picture = word


Remember/Know (Radvansky, pp 307-308)

General Findings:

factors recollection, “remember”

LOP, repetition, short (vs long) delay

Problems:

• poor terminology


Remember/Know Instructions: Rajaram

(1993)


Remember/Know

General Findings:factors recollection, “remember”LOP, repetition, short (vs long) delay

Problems:• poor terminology• judgmental criteria• r/k as confidence judgment Converging Evidence:

Process dissociation studies


Process Dissociation; Jacoby (1991)

• Two Independent Process:

recollective (R)

automatic (A)

• Strategy: set processes in opposition

manipulate factor(s) affecting recollection

2 tests:

recollection yes (Inclusion)

recollection no (Exclusion)


Process Dissociation

2 tests:• recollection yes (Inclusion)• recollection no (Exclusion)

Goal: Compute values for R & A• Data:

Inclusion = R + A(1-R)

Exclusion = A(1-R)• Parameter Estimates

R = Inclusion – Exclusion

A = Exclusion / (1-R)


Process Dissociation

Evidence for the role of Dual-Processes in two classes of memory test

1. A Direct Test (recognition)

2. An Indirect Task (fragment completion)


Process Dissociation: Direct Test

• Read a list of words – List 1• Hear a list of words – List 2• Two recognition tests:

– Both tests include List 1, List 2 and novel words.

– Inclusion test: Respond “old” if word was on either list.

– Exclusion test: Respond “old” only if word was on List 2.


Inclusion test

• Inclusion test: Respond “old” if word was on either list.– Intentional (recollective) process will have a

certain probability of concluding “old” for List 1 words – R

– Automatic process will also have a certain probability of concluding “old” for List 1 words – A

– If either process concludes “old”, the subject will respond “old”

P(old) = R + A (1-R)


Inclusion Condition

List1Word

“OLDR”Recollected

NOT Recollected

High Familiarity “OLDA”

“New”Low

Familiarity

P(OLD) = P(OLDR) + P(OLDA)

R%

1-R%

1-A%

A%


Exclusion test

• Exclusion test: Respond “old” only if word was on List 2.– Subject will only respond “old” to List 1 words if

two things happen:• The automatic process responds “old” due to a

feeling of familiarity – A• The intentional process fails to recognise the

word (if it had, it would recall it was from List 1) – (1-R)

P(old ) = A(1-R)


Exclusion Condition

List1Word

“NEW”Recollected

NOT Recollected


“New”Low

Familiarity

P(OLD) = P(OLDA)

R%

1-R%

1-A%

A%


Dissociating the processesData:• Inclusion: P(old) = R + (1- R)• Exclusion: P(old) = A(1-R)


Inclusion Condition

List1Word

“OLDR”Recollected

NOT Recollected


“New”Low

Familiarity

P(OLD) = P(OLDR) + P(OLDA)

R%

1-R%

A%

1-A%


Exclusion Condition

List1Word

“NEW”Recollected

NOT Recollected


“New”Low

Familiarity

P(OLD) = P(OLDA)

R%

1-R%

1-A%

A%


Dissociating the processesData:• Inclusion: P(old) = R + (1- R)• Exclusion: P(old) = A(1-R)

Parameter Estimates• Inclusion – Exclusion = R• A = Exclusion / (1-R)


Jacoby (1991)

Materials:• List 1: READ words• List 2: HEAR wordsTests:• Inclusion

– List 1 “OLD”– List 2 “OLD”

• Exclusion– List 1 “NEW”– List 2 “OLD”


Jacoby (1991; Exp 3)Two recognition tests (% “OLD” for READ words):

• Inclusion test P(old) = 0.48• Exclusion test P(old) = 0.37*

– R = Inclusion – Exclusion = 0.11– A = Exclusion / (1-R) = 0.37 / 0.89 = 0.42

*in exclusion condition, “OLD” are errors


Jacoby (1991; Exp 3)Two recognition tests (% “OLD” for READ words):

• Inclusion test P(old) = 0.48• Exclusion test P(old) = 0.37*

– R = Inclusion – Exclusion = 0.11– A = Exclusion / (1-R) = 0.37 / 0.89 = 0.42

*in exclusion condition, “OLD” are errors


Jacoby (1991; Exps 2 & 3)Implication: When recollection is knocked out, P(OLD) in exclusion condition should equal A

Exclusion test w/ digit monitoring task (monitor for 3 odd digits in a row).

Expectation: Recollection eliminated by divided attention (digit task) – R = 0

Prediction: Exclusion = A(1-R) = 0.42 (1-0) = 0.42

Results: Exclusion w/ divided attention: Prob(Old) = 0.43

Lecture 11 – Psyco 350, A1 Winter, 2011

Documents

Lecture 11 – Psyco 350, A1 Winter, 2011