At the Scene of the Crime: An Examination of the External Validity of Studies on Eyewitness Identifi

1
At the Scene of the Crime An Examination of the
External Validity of Studies on Eyewitness
Identification AccuracyHeather D. Flowe and
Ebbe B. EbbesenDepartment of PsychologyUniversit
y of California, San Diego
2
Acknowledgments
Professor Ebbe B. Ebbesen

• Michael Corley
• Paul Mertens
• Grant Schrader
• And a small army of 199s

3
Mistaken Identification

• Mistaken eyewitness ID often cited as 1 reason
  for erroneous convictions.
• The advent of DNA technology has allowed for
  conclusively determining whether a suspect was
  falsely identified.

4
Evidence for Widely Held Belief

• Innocent people on death row
• Governor Ryan, of IL, left office asking that all
  cases for death row inmates be re-examined on
  grounds that some might be innocent.
• How many more cases of wrongful convictions have
  to occur before we can all agree that this system
  in Illinois is broken? Gov. Ryan
• Innocence Project www.innocenceproject.org
• DNA and falsely convicted innocents
• 158 convicted people have been exonerated as of
  today
• Eyewitness ID is the primary cause
• Among 1st 70 cases, the most common cause of
  false conviction was incorrect eyewitness ID (61
  of cases)
• Experts who testify, testify for defense over 94
  of time (Kassin, et al., 2001)

5
We Can and Should Use Science to Fix the Problem
Two Solutions

• Expert testimony in court
• Part of problem must be with jurors who do not
  understand how human memory works
• Improve reliability of procedures used to collect
  eyewitness evidence
• Design new procedures, based on science of human
  memory, that will reduce error rate

6
US Supreme Court Daubert

• Daubert was about
• Whether Dauberts birth defects were caused by
  Bendectin
• What types of (scientific) evidence (e.g.,
  test-tube, animal studies, pharmacological
  toxicology studies, human epidemiological/statisti
  cal studies) could be admitted?
• Decision Trial judge must apply a two part test
  to determine whether to admit testimony by a
  scientific expert
• Will testimony be about valid scientific
  knowledge?
• But what constitutes scientific knowledge?
• Will testimony help the jury understand fact(s)
  at issue?
• Is the danger of unfair prejudice (too much
  influence by expert), confusion (area is too
  complex), and misleading of the jury (doesnt
  present all sides) outweighed by the probative
  value of the testimony?
• Probative value Improves chances of correct
  decision.

7
Survey of Memory Experts

• Percent of experts agreeing reliable enough to
  testify Kassin, et al., 2001

8
Some common conclusions by experts about the
nature of eyewitness memory

• Stress High stress decreases accuracy of memory
• Confidence Accuracy of identifications unrelated
  to witness confidence in them
• Post-event Much of what witnesses remember is
  the result of source misattribution and/or
  post-event memory effects
• Cross-race Other-race identifications worse than
  same-race identifications
• Weapon Presence of a weapon decreases witness
  reliability
• Retention interval Over time the accuracy of
  witness memory drops off rapidly and then levels
  off
• Identification procedures have major problems
• Showups are unfair
• Lineups are filled with biases and suggestive
  features
• Witnesses use relative rather than absolute
  decision strategies in simultaneous lineups

9
APLS Recommendations

• Eyewitnesses are prone to making relative
  judgments during lineup identifications.

10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
Relative v. Absolute Decision Strategy (Wells,
Lindsay, others)

• Relative Decision Strategy
• Witness compares all members of lineup to each
  other and then chooses the one most similar to
  the culprit (most familiar?)
• Absolute Decision Strategy
• Witness compares presented alternative to memory
  of culprit and decides whether that alternative
  is actually the culprit

14
Policy Changes

• New Jersey first state to adopt sequential
  procedures
• California legislature currently considering
  whether to switch to sequential lineups
• Is the decision to switch premature?

15
Common Sources of Evidence for Theoretical and
Applied Claims

• Face memory studies (lab only)
• Multiple faces per subject
• No crime/same event (view faces)
• ID by Yes/no recognition task
• Event Memory Studies (lab and field)
• Same single-face crime-event for all subjects
• ID by Lineup (mostly simultaneous)
  identification
• Target present/Target absent lineups
• Archival studies (actual crimes/criminals)
• Different single-face crime-event per witness
• ID by name, location, show-up, lineup, in court

16
Can We Generalize?

• Face validity of the published literature seems
  high
• Illegal activity
• There is a witness
• A lineup test is administered
• The witness makes an identification

17
Can We Generalize?

• But does the literature fall short in capturing
  characteristics of
• Witnesses
• Culprits
• Other aspects of the critical event, such as
• Duration of exposure to culprit
• Presence of bystanders
• Retention interval
• Lineup Test
• Criteria used to select distractors
• Criteria used to obtain innocent suspect

18
Can We Generalize?

• The particulars of the people and contexts
  studied in the lab are important because
• Variables may interact
• The levels of factors as they occur naturally
  might be correlated
• Eg, Stress and Accuracy
• Standard lab finding More stress decrease in
  accuracy
• However, in actual crimes, some witnesses might
  be more stressed than others, but this doesnt
  necessarily mean they have worse memories as a
  results. Stressed witnesses were probably more
  involved in the event, and thus encoded it better
  than less stressed witnesses.

19
Meta-Analysis of ID StudiesOverview

• How well do laboratory studies capture real world
  eyewitness conditions?

20
MethodMeta-Analysis of Literature

• 185 experiments from 92 peer-reviewed published
  papers involving over 28,000 participants were
  coded.
• Across studies, 48 involved a staged live event,
  and 52 involved a pre-recorded event
• Crime Type Theftgtnoncrime situationgt robberygt
  assaultgtvandalismgtrapegtmurder

21
Method Analysis of Prosecutors Files

• 721 cases (robbery, rape, and assault) randomly
  selected from prosecutor files
• 65 of the cases sampled were accepted for
  prosecution
• The testimony of 1319 eyewitnesses is represented

22
ResultsMost Common Manipulations
23
ResultsWitness Characteristics

• Only 12 of laboratory experiments reported age
  of participants.
• Range 2-94 years, mode 18-22
• Archival witnesses
• 17 under 18
• 81 between 18 and 65
• 2 over 65

24
ResultsWitness Characteristics

• 13 of real world witnesses under the influence
• Of these, 45 gave a description of culprit to
  the police
• 22 were give opportunity to identify suspect
• Only 3 published papers examined the effects of
  alcohol on eyewitness memory.

25
ResultsCulprit Characteristics

• In the archives, the defendants were aged 29
  years (range 19-79 years, median28 years)
• In 85 of the laboratory simulations, the
  culprit was a college student

26
ResultsCulprit Characteristics

• Distinctive physical features noted on the arrest
  report for 33 of the defendants
• None of the papers in the literature reported
  whether their culprits had such features.

27
ResultsWitness Characteristics
28
ResultsGeneral Viewing Conditions

• Very little information presented in the
  articles. For instance
• Only 10 provided information regarding the
  distance between the participant and the culprit
• No mention ever made of lighting (except for in
  two articles in which the variable was
  manipulated)

29
ResultsWeapon Exposure

• In the real world cases, 59 of witnesses exposed
  to violence. About 1/3 interacted with culprit
  prior to violence.
• Effect of timing of violence on memory for the
  culprit not studied in any of the published
  experiments.
• Only 2 published papers examined the effect of
  violence on accuracy.

30
ResultsWeapon Exposure
31
ResultsCritical Event Exposure Duration
32
ResultsPerpetrator Exposure Duration
33
ResultsDistribution of Lineup Test Type

• 58 of archival IDs were cross-race

34
ResultsRetention Interval
35
Conclusions

• Experimental designs contain too few factors
• Range of values of independent variables
  unrepresentative
• Interactions among estimator and system variables
  not considered

36
Aggregation Over What Sources of Variation Are
We Claiming to Generalize?

• Some sources of variation
• Crimes
• Criminals (faces, bodies, voices, etc.)
• Witnesses
• Observational conditions
• Test procedures
• Tasks
• Responses