Dr. Anne Cybenko University of Dayton Research Institute.

75
Eyewitness Memory Dr. Anne Cybenko University of Dayton Research Institute

description

 There have been 289 exonerations in the US ◦  Over 70% of those have been due at least in part to a mistaken identification  It’s a unique situation that tests human memory and decision making

Transcript of Dr. Anne Cybenko University of Dayton Research Institute.

Page 1: Dr. Anne Cybenko University of Dayton Research Institute.

Eyewitness MemoryDr. Anne Cybenko

University of Dayton Research Institute

Page 2: Dr. Anne Cybenko University of Dayton Research Institute.

“The Innocents” Taryn Simon

Page 3: Dr. Anne Cybenko University of Dayton Research Institute.

There have been 289 exonerations in the US◦ www.innocenceproject.org

Over 70% of those have been due at least in part to a mistaken identification

It’s a unique situation that tests human memory and decision making

Why do People Care?

Page 4: Dr. Anne Cybenko University of Dayton Research Institute.

Ronald Cotton

CORRECT IDENTIFICATION

FALSEIDENTIFICATION

Picking Cotton

Page 5: Dr. Anne Cybenko University of Dayton Research Institute.

This guy commits a crime and someone witnesses it.

The Basics: What Happens when there’s a witness to a crime?

Page 6: Dr. Anne Cybenko University of Dayton Research Institute.

Police get a description from the witness

The Basics: What Happens when there’s a witness to a crime?

He was about 6’ tall, brown hair, brown eyes, no facial hair…

Page 7: Dr. Anne Cybenko University of Dayton Research Institute.

Police find a suspect who may be: Guilty or Innocent

The Basics: What Happens when there’s a witness to a crime?

Page 8: Dr. Anne Cybenko University of Dayton Research Institute.

Witness is shown a lineup that is eitherPerpetrator Present

Perpetrator Absent

The Basics: What Happens when there’s a witness to a crime?

Page 9: Dr. Anne Cybenko University of Dayton Research Institute.

Perpetrator Present

Perpetrator Absent

Picks the suspect CORRECT IDENTIFICATIONBad guy goes to trial

FALSE IDENTIFICATIONInnocent guy goes to trial

Picks a foil FOIL PICKBad guy may go free

FOIL PICKInnocent guy may go free

Rejects the lineup

FALSE REJECTIONBad guy may go free

CORRECT REJECTIONInnocent guy may go free

The Basics: Possible Outcomes of a Lineup

Type of Lineup

Witn

ess’s

Dec

ision

Page 10: Dr. Anne Cybenko University of Dayton Research Institute.

Mostly undergraduate participants Participants watch a mock crime (video or

live) Participants usually give a description Filler task Presented with a lineup Make an Identification

How is it studied?

Page 11: Dr. Anne Cybenko University of Dayton Research Institute.

Witness interview techniques Child interview techniques Showups Foil selection Lineup Presentation Lineup Instructions Multi-modality lineups Double-blind lineups

What Exactly do People Study

Page 12: Dr. Anne Cybenko University of Dayton Research Institute.

Simultaneous or Sequential Lineups Accuracy and Response Time Verbal Overshadowing Effect Interference Theory

Topics for today

Page 13: Dr. Anne Cybenko University of Dayton Research Institute.

Sequential vs. Simultaneous Lineups

Page 14: Dr. Anne Cybenko University of Dayton Research Institute.

I am going to show you six photographs. Please look at all six photographs before making any comment.

The person who committed the crime may or may not be among those shown in the photographs you are about to see. If you recognize any of the persons in the photographs as the suspect, go back and pick out the person you recognize.

If you recognize any of the persons please do not ask me whether your choice was ‘right’ or ‘wrong’, as I am prohibited by law from telling you.

Page 15: Dr. Anne Cybenko University of Dayton Research Institute.

SAN BERNARDINO COUNTY SHERIFF PHOTO LINE-UP

1 2 3

4 5 6

Page 16: Dr. Anne Cybenko University of Dayton Research Institute.

I am going to show you some photographs.

The person who committed the crime may or may not be among those shown in the photographs you are about to see. If you recognize any of the persons in the photographs as the suspect, please tell me and we will stop the lineup.

If you recognize any of the persons please do not ask me whether your choice was ‘right’ or ‘wrong’, as I am prohibited by law from telling you.

Page 17: Dr. Anne Cybenko University of Dayton Research Institute.

SAN BERNARDINO COUNTY SHERIFF PHOTO LINE-UP

Page 18: Dr. Anne Cybenko University of Dayton Research Institute.

SAN BERNARDINO COUNTY SHERIFF PHOTO LINE-UP

Page 19: Dr. Anne Cybenko University of Dayton Research Institute.

SAN BERNARDINO COUNTY SHERIFF PHOTO LINE-UP

Page 20: Dr. Anne Cybenko University of Dayton Research Institute.

SAN BERNARDINO COUNTY SHERIFF PHOTO LINE-UP

Page 21: Dr. Anne Cybenko University of Dayton Research Institute.

SAN BERNARDINO COUNTY SHERIFF PHOTO LINE-UP

Page 22: Dr. Anne Cybenko University of Dayton Research Institute.

SAN BERNARDINO COUNTY SHERIFF PHOTO LINE-UP

Page 23: Dr. Anne Cybenko University of Dayton Research Institute.

Decision PP Simultaneo

us

PP Sequential

PA Simultaneo

us

PA Sequential

Suspect ID .58 .50 .43 .17Foil ID .12 .02 .15 .18Reject .30 .48 .42 .65

Sequential vs. Simultaneous Lineups

Lindsay & Wells (1985)

Page 24: Dr. Anne Cybenko University of Dayton Research Institute.

Results depend on:◦ Similarity of the innocent suspect to the

perpetrator ◦ Innocent suspect position in the sequential lineup

(Clark & Davey, 2001)

Sequential vs. Simultaneous Lineups

Page 25: Dr. Anne Cybenko University of Dayton Research Institute.

Decision

PPSimultaneo

us

PPSequenti

al#2

PPSequenti

al#4

PASimultaneo

us

PASequential

#2

PASequential

#4

Suspect .417 .292 .667 .313 .292 .458

Foil .438 .500 .167 .354 .208 .334

Reject .146 .308 .167 .333 .500 .208

Innocent suspect is medium similarity to perpetrator

Decision

PPSimultaneo

us

PPSequenti

al#2

PPSequenti

al#4

PASimultaneo

us

PASequential

#2

PASequential

#4

Suspect .250 .292 .625 .417 .417 .250

Foil .542 .583 .208 .312 .208 .542

Reject .208 .125 .167 .271 .375 .208

Innocent suspect is high similarity to perpetrator

Page 26: Dr. Anne Cybenko University of Dayton Research Institute.

The Mecklenburg Report (2006)

Sequential vs. Simultaneous Lineups

Simultaneous (N = 319)

Sequential(N = 229)

Suspect .599 .45Foil .028 .092Reject .376 .472

Page 27: Dr. Anne Cybenko University of Dayton Research Institute.

Absolute vs relative decision making

Sequential vs. Simultaneous Lineups

Page 28: Dr. Anne Cybenko University of Dayton Research Institute.

WITNESS model (Clark, 2003) to generate predictions

Perpetrator is represented as a vector of features, P.

Memory of the perpetrator (M):◦ Feature j of P is stored correctly in M with probability a.◦ Feature j of P is stored incorrectly in M with probability 1 – a.

Innocent suspect and foils:◦ Feature j matches the perpetrator with probability s◦ Feature j mismatches the perpetrator with probability 1 – s.

Match each lineup member to memory (M).◦ m(L1,M), m(L2,M), … , m(L6,M) (for six person lineup)

Apply decision rule to make a decision.

Model-Based Predictions

Page 29: Dr. Anne Cybenko University of Dayton Research Institute.

◦ Best Above Criterion (BAC): A witness identifies the best matching lineup member if the match to memory is above criterion. (Absolute judgment)

◦ Relative Difference Model (RD): Best matching lineup member is identified if the difference between the best match and the next-best match is above a difference criterion. (Relative judgment)

WITNESS model

Page 30: Dr. Anne Cybenko University of Dayton Research Institute.

Dunning & Perretta (2002) Four eyewitness studies using a

combination of videotaped and live crimes. Measured response time and accuracy

Accuracy and Response Time

Page 31: Dr. Anne Cybenko University of Dayton Research Institute.

Witnesses making their ID’s faster than 10-12 seconds were nearly 90% accurate

Witnesses that took longer than 12 seconds were 50% accurate

Automatic vs Deliberative processes

Accuracy and Response Time

Study N Accurate Inaccurate t1 201 38.4 63.9 -3.07**2 96 11.1 15.7 -2.46*3 41 13.2 23.5 -2.78*4 50 13.3 36.4 -4.26**

* p<.05 ** p<.01Study 1 participants had think aloud task

Page 32: Dr. Anne Cybenko University of Dayton Research Institute.

Optimal response time changes with:◦ An increase in retention interval (Brewer et al., 2006)

◦ An increase in the size of the lineup (Brewer et al., 2006)

◦ Confidence of the witness (Weber et al., 2004)

◦ Age (Weber et al., 2004)

Accuracy and Response Time

Page 33: Dr. Anne Cybenko University of Dayton Research Institute.

Verbalizing the appearance of previously seen visual stimuli impaired subsequent recognition performance

Verbal Overshadowing Effect

Page 34: Dr. Anne Cybenko University of Dayton Research Institute.

Experiment 1◦ Does describing a perpetrator affect ones ability

to identify him?

Schooler & Engstler-Schooler (1990)

Description NoneCorrect ID .38 .64Foil ID .37 .22Rejection .25 .14

Page 35: Dr. Anne Cybenko University of Dayton Research Institute.

Experiment 3◦ Do these results apply to all stimuli that are

difficult to describe?

Description

Visualization

None

Correct ID .33 .64 .73Foil ID .67 .36 .27

Schooler & Engstler-Schooler (1990)

Page 36: Dr. Anne Cybenko University of Dayton Research Institute.

Experiment 5◦ Does this pattern of results change if there’s a

limited amount of time to ID the suspect?

Description NoneUnlimited Recognition .50 .805 Second Recognition .73 .76

Schooler & Engstler-Schooler (1990)

Page 37: Dr. Anne Cybenko University of Dayton Research Institute.

Meta-analysis on 29 published and unpublished verbal overshadowing studies

Zr = -.12, a small but significant negative effect of verbal overshadowing

Meissner & Brigham (2001)

Page 38: Dr. Anne Cybenko University of Dayton Research Institute.

Long delay between description and ID (Finger & Pezdek, 1999)

More than 1 trial (Fallshore & Schooler, 1995; Houser et al., 1997; Melcher & Schooler, 1995)

Following re-presentation of visual stimulus(Schooler et al., 1996)

When a cognitive interview is used(Finger & Pezdek, 1999; Meissner et al., 2001)

Studies that don’t find verbal overshadowing

Page 39: Dr. Anne Cybenko University of Dayton Research Institute.

Availability assumption◦ Visual representation remains available in

memory despite the temporary verbal impairment Modality mismatch hypothesis

◦ Competing representations in memory from different modalities

Possible Mechanisms

Page 40: Dr. Anne Cybenko University of Dayton Research Institute.

40

Proactive interference: When information that is presented BEFORE a target event interferes with the memory for that target event

Retroactive interference: When information that is presented AFTER a target event interferes with the memory for that target event

Retroactive interference is stronger than proactive interference (Schemeidler, 1939; McGeoch and Underwood,1943; Melton and Von Lacrum, 1941)

Types of Interference

Page 41: Dr. Anne Cybenko University of Dayton Research Institute.

41

Changed-trace hypothesis: Outside information interferes with target information because it changes the memory trace of the target information.

Multiple-trace hypothesis: Outside information interferes with target information because it creates a separate memory trace that is recalled instead of the target trace.

Hypotheses of Interference

Page 42: Dr. Anne Cybenko University of Dayton Research Institute.

42

Argued that the Loftus et al. results could have been due to 2 other issues:◦ Original stimulus never encoded◦ Remembered both, trusted/responded with the

one from the narrative

McCloskey & Zaragoza (1985)

Page 43: Dr. Anne Cybenko University of Dayton Research Institute.

43

Participants were presented with slides depicting a maintenance man enter an office, fix a chair, then steal $20.

4 Critical items within the slides ◦ tool (hammer, wrench, or screwdriver)◦ soda can (Coke, 7-up, or Sunkist)◦ coffee jar (Folgers, Maxwell House, or Nescafe)◦ magazine (Glamour, Vogue, or Mademoiselle)

Participants presented with written account of what they just saw. This written account contained misinformation for 2 of the items.

McCloskey & Zaragoza (1985)

Page 44: Dr. Anne Cybenko University of Dayton Research Institute.

44

Two types of tests: ◦ Traditional Recognition Test: Target vs. Interfering

Saw Coke, read about 7-up, Test: Coke or 7-up◦ Modified Recognition Test: Target vs. Novel

Saw Coke, read about 7-up, Test: Coke or Sunkist Traditional Recognition test – Picked 7-up Modified Recognition test – Picked Coke Participants could not have picked Coke in

the modified test if reading about 7-up changed their memory trace

McCloskey & Zaragoza (1985)

Page 45: Dr. Anne Cybenko University of Dayton Research Institute.

45

Replicated McCloskey and Zaragoza (1985) with 2 changes

Gave either incorrect or neutral information in the recap of the slides (“7-up” or “soda” )

Test was in the form of open ended questions: “The key to the desk was next to a ____ can?”

participants performed equally well on the recall tests

If the misinformation really changed the memory trace, participants should have performed considerably worse on the recall test for the misinformation items.

Zaragoza, McCloskey, and Jamis (1987)

Page 46: Dr. Anne Cybenko University of Dayton Research Institute.

46

Replicated McCloskey and Zaragoza’s (1985) experiment Testing was in the form of Yes/No statements: “Below the magazine

rack there was a copy of Vogue magazine.” Tversky and Tuchin asked about the critical item (the one from the

slides), the misinformation item (the one from the summary of the slides), and the novel item (the one that did not appear at all)

Belli asked about the critical item, and the novel item. Misleading postevent information reduced the "Yes" responses to

the question about the original item. T & T Subjects were equally good at rejecting the novel item T & T Similar number of subjects responded yes to original item as

responded yes to the misleading item. Belli - misled subjects were better than control subjects at rejecting

the novel item Belli concluded memory trace was changed, T & T did not.

Tversky & Tuchin (1989)Belli (1989)

Page 47: Dr. Anne Cybenko University of Dayton Research Institute.

47

McCloskey and Zaragoza’s results could have been a product of their stimuli having too many unique discernable features such that even if the trace was altered, some of those features would remain unchanged.

Used stimuli that were very similar to each other.

Conducted 10 Experiments

Chandler (1989 & 1991)

Page 48: Dr. Anne Cybenko University of Dayton Research Institute.

48

Nature photographs Cut in thirds

Chandler - Stimuli

A A’ A”

Page 49: Dr. Anne Cybenko University of Dayton Research Institute.

49

Variables between experiments: ◦ Proactive or Retroactive interference◦ Presentation time◦ Retention interval◦ Number of stimuli

Chandler’s Methodology

Page 50: Dr. Anne Cybenko University of Dayton Research Institute.

50

Chandler’s Methodology (Retroactive Interference)

Target List Interfering List

Retention

Test

Page 51: Dr. Anne Cybenko University of Dayton Research Institute.

51

Found Retroactive interference when:◦ Presentation time was long(7.5 and 10 seconds)◦ Delay was short (3 – 15 minutes)

Did not find Retroactive Interference when◦ Presentation time was short (4 seconds)◦ Lots of stimuli (140 not 48)◦ Delay was long (48 hours)

Did not find significant evidence of proactive interference under any conditions

Chandler’s Results

Page 52: Dr. Anne Cybenko University of Dayton Research Institute.

52

Participants tested in groups of 1-8 All stimuli presented on overhead Responses were made on paper

Experiments 1-6 Method

Page 53: Dr. Anne Cybenko University of Dayton Research Institute.

53

Experiment 1 was conducted to gather evidence for/against item specific retroactive and proactive interference in a modified recognition test.

Replicate Chandler’s results

Experiments 1 & 2- Purpose

Page 54: Dr. Anne Cybenko University of Dayton Research Institute.

54

Participants: 60 Undergraduates Materials: 48 Color Photographs of Nature

Scenes. Each scene was divided into thirds (A, A’, A”)

Experiments 1&2 - Method

Page 55: Dr. Anne Cybenko University of Dayton Research Institute.

55

Experiments 1&2 - Stimuli

A A”A’

Page 56: Dr. Anne Cybenko University of Dayton Research Institute.

56

Experiments 1&2- Stimuli

A A”A’

Page 57: Dr. Anne Cybenko University of Dayton Research Institute.

57

Experiments 1 & 2 - Method Experiment 1

Math Phase Study Phase Interfering Phase Math Phase Test Phase

Interfering Phase Study Phase Math Phase Math Phase Test Phase

Experiment 2

Page 58: Dr. Anne Cybenko University of Dayton Research Institute.

58

Experimental Accuracy: M = 68.7, SD = 11.1

Control Accuracy: M = 75.8, SD = 8.8 t (31) = 4.682, p < .001, r = -.644 Significant evidence of retroactive

interference

Experiment 1 - Results

Page 59: Dr. Anne Cybenko University of Dayton Research Institute.

59

Experimental Accuracy: M = 71.88, SD = 9.65

Control Accuracy: M=72.77, SD = 10.43 t (27) = .486, p = .631, r = -.093 No evidence of proactive interference

Experiment 2 - Results

Page 60: Dr. Anne Cybenko University of Dayton Research Institute.

60

See if the results of Chandler (1989, 1991) and Experiments 1&2 are found when the stimuli are words instead of images

Find support for/against the changed-trace and multiple-trace hypotheses

Experiments 3&4 - Purpose

Page 61: Dr. Anne Cybenko University of Dayton Research Institute.

61

Participants: 60 Undergraduates Materials: 144 words from Van Overschelde,

Rawson, and Dunlosky’s (2004) updated category norms. The 18 most common words from 8 categories(animals, colors, body parts, fruits, sports, clothing, countries, and car models) were used.

Experiments 3&4 - Method

Page 62: Dr. Anne Cybenko University of Dayton Research Institute.

62

Experiments 3 & 4 - MethodExperiment 3

Math Phase Study Phase Interfering Phase Math Phase Test Phase

Interfering Phase Study Phase Math Phase Math Phase Test Phase

Experiment 4

Page 63: Dr. Anne Cybenko University of Dayton Research Institute.

63

Experimental Accuracy: M = 80.51, SD = 11.30

Control Accuracy: M = 85.62, SD = 10.86 t (28) = 2.139, p = .013, r = -.375 Evidence of retroactive interference was

present

Experiment 3 - Results

Page 64: Dr. Anne Cybenko University of Dayton Research Institute.

64

Experimental Accuracy: M = 76.14, SD = 12.49

Control Accuracy: M = 82.33, SD = 9.94 t (30) = 2.66, p = .013, r =- .437. Evidence of proactive interference was

present

64

Experiment 4 - Results

Page 65: Dr. Anne Cybenko University of Dayton Research Institute.

65

Find evidence for/against the Threshold Theory of Interference

Experiments 5&6 - Purpose

Page 66: Dr. Anne Cybenko University of Dayton Research Institute.

66

Participants: 39 Undergraduates Materials: Same as Experiments 1 & 2

(nature pictures)

Experiments 5&6 - Method

Page 67: Dr. Anne Cybenko University of Dayton Research Institute.

67

Experiments 5 & 6 - MethodExperiment 5

Math Phase Study Phase Interfering Phase (4 or 1) Math Phase Test Phase

Interfering Phase Study Phase Math Phase Math Phase Test Phase

Experiment 6

Page 68: Dr. Anne Cybenko University of Dayton Research Institute.

68

Experimental (4) Accuracy: M = 67.78, SD = 12.33

Experimental (1) Accuracy: M = 72.25, SD = 12.66

Control Accuracy: M = 76.79, SD = 14.35 F(2,34) = 8.534, p = .001 Tukey test revealed that all groups are

significantly different from each other.

Experiment 5 - Results

Page 69: Dr. Anne Cybenko University of Dayton Research Institute.

69

Experimental (4) Accuracy: M = 68.79, SD = 12.79

Experimental (1) Accuracy: M = 72.95, SD = 10.70

Control Accuracy: M = 75.05, SD = 10.80 F(2,34) = 8.534, p = .001 Tukey and paired samples t-tests:

◦ Control = Experimental (1)◦ Control > Experimental (4)◦ Experimental (1) > Experimental (4)

Experiment 6 - Results

Page 70: Dr. Anne Cybenko University of Dayton Research Institute.

70

Contrasts

Predictions:◦ Control = Experimental (1)◦ Control > Experimental (4)◦ Experimental (1) > Experimental (4)

t (20) = 3.16415, p = .047, r =- .578

Experiment 6 - Results

Page 71: Dr. Anne Cybenko University of Dayton Research Institute.

71

Retroactive interference is found when the stimuli are pictures and words and when there is a single or multiple presentation of interfering information.

Proactive interference is only found when there are multiple presentations of interfering information.

Summary of Results

Page 72: Dr. Anne Cybenko University of Dayton Research Institute.

72

An interfering stimulus only interferes with a target stimulus IF the difference in strength between the memory trace of the interfering stimulus and the memory trace of the target stimulus crosses a threshold.

Threshold Theory of Interference

Page 73: Dr. Anne Cybenko University of Dayton Research Institute.

73

Support for a modified version of multiple trace hypothesis

Threshold Theory of Interference is supported◦ Evidence of proactive interference found with

multiple presentations of interfering stimuli◦ Evidence of retroactive interference ◦ No evidence of retroactive interference after 48

hours (Chandler, 1991)

Discussion

Page 74: Dr. Anne Cybenko University of Dayton Research Institute.

74

What determines the strength of the memory trace?

If Experiments 5 & 6 had more than 4 repetitions, would they cause even more interference?

Future Directions

Page 75: Dr. Anne Cybenko University of Dayton Research Institute.

75

Thank you!