Transcript Lecture_09
Lecture 9 – Psyco 350, B1 Fall, 2011 N. R. Brown Psyco 350 Lec #9– Slide 1 Outline • Forgetting – decay – retrieval failure – Interference – Inhibition (directed forgetting) • Interference In the Real-World – Hindsight Bias – Misinformation Effect Psyco 350 Lec #9– Slide 2 Ebbinghaus(1885): The 1st Forgetting Function • Main Findings: – AMOUNT of forgetting decreases w/ time • Interpretation: – orgetting driven by decay; information lost at a constant rate. Psyco 350 Lec #9– Slide 3 A Sample Decay Function 100 % Recalled 90 80 time start X finish 01 100 50 12 50 25 23 25 34 12.50 6.25 45 6.25 3.125 70 60 50 40 30 20 10 0 1 2 3 Study-Test Delay 4 5 • Rate of forgetting constant over time • Amount of forgotten/unit time with time Psyco 350 Lec #9– Slide 4 .50 12.50 Studying PI & RI • Classic studies: paired associate learning – study: cue-target word pairs (CUP-tree) – test: given cue, recall target (CUP-???) – manipulate presence, timing & similarity of additional targets Psyco 350 Lec #9– Slide 5 Studying PI & RI Design Proactive Retroactive List 1 A-B D-E A-B A-B List 2 Test A-C A-?C? A-C A-?C? A-C D-E A-?B? A-?B? Exp Control Exp Control General Findings: • • Cued Recall: Control > Experimental Similarity Effects: the more similar B is C, the more server the interference. Psyco 350 Lec #9– Slide 6 Yet Another Demo • • • • 4 list 8 words/list Study 23 s of social interaction Test – recall 8 words from prior list. Psyco 350 Lec #9– Slide 7 PI & RPI List 1 1 2 3 4 5 6 7 8 badger caribou wolverine otter rabbit gopher fox seal Psyco 350 Lec #9– Slide 8 List 2 goat skunk raccoon squirrel bear cougar elk mouse List 3 List 4 deer walrus mink beaver coyote woodchuck lynx moose Cleveland Paris Winnipeg Dallas Boston London Rome Halifax PI & RPI: Background Brown & Peterson Task Review • Task: learn triplet filled delay recall triplet • Finding: – recall drops off very rapidly w/ delay • Original Interpretation: – Forgetting caused by decay in STM – Forgetting indicates the rate of loss from STM • Alternative Interpretation (Keppel & Underwood): – Forgetting caused by PI from similar materials • Implication: PI should be reduced when new list differs from prior lists. Psyco 350 Lec #9– Slide 9 Release from PI: Wickens (1972) • Task: Standard Brown-Peterson Task • Procedure: – Trials 1 though 3: triples drawn from same semantic category – Trial 4: triple drawn from different category Psyco 350 Lec #9– Slide 10 Wickens (1972): Materials Psyco 350 Lec #9– Slide 11 Wickens (1972): Results • PI (recall ) across same-category trials. • when category changes, Recall Release from PI • RPI as similarity between initial category and new category • Finding generalize to realworld material (news stories) Psyco 350 Lec #9– Slide 12 Gunter, Berry, Clifford (1981): RPI w/ News Stories • Replicates Wickens with news stories. e.g., 3 sets of political stories 1 human interest story Psyco 350 Lec #9– Slide 13 RPI: Activation-Discrimination Interpretation Activation: • Concepts activated when accessed • Activation decays rapidly Retrieval: search some (cued) portion of memory for most active concepts. -------------------------------------------------------------------------PI: difficult to discriminate between many activated concepts. RPI: relatively easy to select active concepts among inactive ones. Psyco 350 Lec #9– Slide 14 RI -- Recent learning impedes recall of prior material Slamecka (1960) – a lab demonstration Materials: 20-word long sentences drawn from text books. Study: Sentence present 1 word/3 seconds Test: Verbatim recall Design: # Learning Trials X # Interpolated Trials 2 0 4 4 8 8 Psyco 350 Lec #9– Slide 15 Slamecka (1960): Results IMPORTANT: study-test delay constant across interpolation conditions • Recall w/ # learning trials (rehearsal effect) • Recall w/ # interpolated trials (RI) Psyco 350 Lec #9– Slide 16 RI in the Real World General Idea: New task-relevant information makes to difficult or impossible to recall or reconstruct prior beliefs, knowledge, responses A GOOD thing: knowledge revision (Friedman & Brown, 2000) And a BAD thing: Hindsight Bias Misinformation Effect. Psyco 350 Lec #9– Slide 17 Hindsight Bias: Typical 3-Phase Design Experimental Condition Phase 1: respond to a target question R1 • How many rhinos live in Africa? Phase 2: Learn the answer to target question. • There are 14,770 rhinos in Africa. Phase 3: Recall initial response (R1). Psyco 350 Lec #9– Slide 18 Hindsight Bias: Typical 3-Phase Design Control Condition Phase 1: respond to a target question R1 • How many rhinos live in Africa? Phase 2: Learn the answer to control question. • The per capita GDP of Guam is $14,770. Phase 3: Recall initial response (R1). Psyco 350 Lec #9 – Slide 19 Hindsight Bias Design A-B A-C A-?B? Experimental Condition A = # rhinos B = R1 A = # rhinos C = 14,770 A = # rhinos ?B? = R1 A-B D-E A-?B? Control Condition A = # rhinos B = R1 D = Guam GDP C = 14,770 A = # rhinos ?B? = R1 Psyco 350 Lec #9– Slide 20 Hindsight Bias(es) Recollection Bias: Correct recall of R1: Control > Experimental Reconstruction Bias (when R1 not recalled): In Exp Condition -- Phase 3 response shifted in direction of Phase 2 information In Control Condition – Phase 2 information has no affect on Phase 3 response Psyco 350 Lec #9– Slide 21 Hindsight Bias: 2 mechanisms for 2 biases Recollection Bias: • standard associate interference (Phase 2 answer competes w/ R1) Reconstruction Bias: • Phase 2 information cause a revision of underlying beliefs • When R1 not retrieved, answer reconstructed w/ revised information Psyco 350 Lec #9 – Slide 22 Misinformation Effect General Phenomenon: • memory for events distorted by exposure to inaccurate/misleading post-event information – benign aspect: post-event narration/discussion can alter autobiographical memories – forensic issue: post-event questioning can alter eyewitness testimony. Psyco 350 Lec #9 – Slide 23 Misinformation Effect: Basic Paradigm An event is witnessed (on tape) Post-event questioning used to introduce misinformation. • Correct Post-event Information – Did the repairman set down his hammer before taking the calculator? • Misleading Post-event Information – Did the repairman set down his screwdriver before taking the calculator? • Neutral – Did the repairman set down his tool before taking the calculator? Psyco 350 Lec #9– Slide 24 Misinformation Effect: Basic Paradigm Test: Recognition for details of original event 2IFC = two item forced choice Did you see a hammer or a screwdriver? Finding: % correct as a function of post-event info type: correct > neutral >> misleading Psyco 350 Lec #9 – Slide 25 Loftus, Burns, Miller (1978) Materials: – 30 slides; pedestrian being hit. – traffic sign (STOP) appears in 1 slide. Questioning: – “Did another car pass the red Datsun as it passed the STOP/YIELD sign?” • Delay: 20 min • Test: 2IFC picture recognition picture w/ STOP vs picture w/ YIELD Psyco 350 Lec #9 – Slide 26 Loftus, Burns, Miller (1978) • Results: – Accurate post-event info: 75% cor. – Misleading post-event info: 40% cor. • Loftus’ Interpretation: Memory change theory (knowledge revision) – misleading information replaces the original, which is permanently lost Psyco 350 Lec #9 – Slide 27 Misinformation Effect: Other Interpretations Memory Coexistence (RI) • Misleading information obscures original memory because it is more recent • Support: – Memory better when original context is reinstated – Memory better if people are warned of misleading information before test Psyco 350 Lec #9 – Slide 28 Misinformation Effect: Other Interpretations Source Monitoring Failure • Errors reflect a failure to identify the source – People remember information, but misremember where it came from • Information that people are mislead about is often that which they make source errors for Psyco 350 Lec #9 – Slide 29 Misinformation Effect: Other Interpretations Biased Guessing Account McCloskey & Zaragoza (1985) • Central Notions: – Target and Foil (misinformation) can coexist – Either or both can be forgotten – Magnitude of misinformation effect depends on: • Prob (Target recalled) • Prob (Foil recalled) • %(Foil selected over Target) Psyco 350 Lec #9 – Slide 30 Biased Guessing Account Magnitude of misinformation effect depends on: • Prob (Target recalled) • Prob (Foil recalled) • %(Foil selected over Target) Implication: If foil removed from reco test, then MISLED = CONTROL Reason: “remembered” misleading inform no longer competing with original info. Psyco 350 Lec #9 – Slide 31 Testing Biased Guessing Hypothesis Initial Event Post-Event Reco Test Control “Stop” --- “Stop” or “Yield Standard – Accurate “Stop” “Stop” “Stop” or “Yield” Standard -- Misleading “Stop” “Yield” “Stop” or “Yield” Modified – Misleading “Stop” “Yield” “Stop” or “Detour” • Introduce modified recognition test. • Predictions for recognition accuracy: – Biased Guessing: Modified Misleading = control – Memory Change: Modified Misleading < control • misleading info should memory for original info regardless of test Psyco 350 Lec #9 – Slide 32 Rationale for Biased Guessing Prediction Control Condition: “S” no “Y”; test: “S” or “Y” “S,” no “Y” no “S,” no “Y” Standard Test: “S” “Y”; test: “S” or “Y” “S,” no “Y” “S,” “Y” no “S,” “Y” no “S,” no “Y” Modified Test: “S” “Y”; test: “S” or “Z” “S,” no “Z” no “S,” no “Z” Psyco 350 Lec # 9 – Slide 33 McCloskey & Zaragoza (1985): Method Stims: – 79 slides of an office theft – 4 s / slide – 4 critical items: coffee jar, magazine, pop can, tool – Post-slide narrative • 735-words long • misinformation for 2 items; neutral for 2 items Psyco 350 Lec # 9 – Slide 34 McCloskey & Zaragoza (1985): Method Procedure: • view slides • 10 minute filler • read narrative • 10 minute filler • 36-item 2IFC recognition test: “The man slide the calculator beneath the ___ in his tool box” standard test: hammer vs screwdriver modified test: hammer vs wrench Psyco 350 Lec # 9 – Slide 35 McCloskey & Zaragoza (1985): Method Procedure: • view slides • 10 minute filler • read narrative • 10 minute filler • 36-item 2IFC recognition test: In Narrative “The man slide the calculator beneath the ___ in his tool box” standard test: hammer vs screwdriver modified test: hammer vs wrench Psyco 350 Lec # 9 – Slide 36 On Slide Never encountered McCloskey & Zaragoza (1985): Results Standard Misled Control 37% 72% 72% 75% hammer vs screwdriver Modified hammer vs wrench • Standard Test: – replicates Misinformation effect: Misled << Control • Modified Test: – consistent w/ Biased Guessing: Misled Control • access to original info unimpaired by post-event info. • Consistent w/ Coexistence & Source Monitoring Accounts Psyco 350 Lec # 9 – Slide 37 Misinformation w/ Modified Procedure: Belli (1992) Materials: – 44 slides (mother & child arguing) – 4 crit slides (coffeemaker, blender, toaster) – 500 word narrative w/ 2 misleading statements • 2IFC modified reco test Slide Narrative Test Control toaster --- toaster vs blender Mod Misinfo toaster coffemaker toaster vs blender Psyco 350 Lec # 9 – Slide 38 Misinformation w/ Modified Procedure: Belli (1992) Design – manipulates timing of misinformation view slides Exp1 5-min delay Exp3 5-day delay read narrative 10 min delay reco test Psyco 350 Lec # 9 – Slide 39 Belli (1992): Results Exp 1ns: 5-min delay Control 95% Mod Mis 95% Exp 3*: 5-day delay 80% 70% • W/ 5-min delay: Mod Misled = Control • consistent w/ Biased Guessing • W/ 5-day delay: Mod Misled < Control • (at lease) consistent w/ Coexistence & RI Psyco 350 Lec # 9 – Slide 40 Blocking Hypothesis: Belli’s Explanation Post-event information impairs access to original traces when: • original trace is weak • post-event information is strong Original trace Trace for Effect of mis-info misinfomation on modified test strong strong no M&Z weak weak no?? ?? weak strong yes Belli Psyco 350 Lec # 9 – Slide 41 Misinformation Effect: My Take Memory impairment, coexistence, & source monitoring errors are not mutually exclusive • As the work on Hindsight bias indicates, new information can: – modify existing information – coexist w/ existing information – block access to existing information • Biased guessing is a problem in 2IFC situation – particularly when target & misinformation are of equal strength • Nonetheless, bias to select foil in standard condition, indicates that post-event information is (some times) accepted as true & incorporated into event representation. Psyco 350 Lec # 9 Slide 42