Transcript Eyewitness limitations

Eyewitness testimony and its limitations
Why is eyewitness testimony so error-prone?
1. Poor view of events and their perpetrators.
2. May not appreciate events’ significance at the time (e.g., con-man).
3. Changes in the suspect's appearance (e.g. disguises).
3. Effects of witnesses’ stress/arousal:
“weapon focus” (e.g. Christianson and Loftus 1991).
4. Elapsed time since event was witnessed.
5. Effects of post-event information (e.g., witnesses’ own ruminations;
listening to other people's accounts of it; misinformation).
6. Effects of expectations and schemas (both on initial encoding and
subsequent memory).
7. Weak relationship between witnesses' accuracy and confidence.
Wells and Olson (2003):
Review factors affecting eyewitness performance.
System variables - under control of legal system (e.g.
interview procedures, lineup presentation modes, etc.)
Estimator variables - not under control of legal system
(e.g. sex and age of witnesses, lighting conditions at
time of crime, etc.)
Little evidence of effects of witness' sex and
intelligence.
Effects of age, event duration, stress/arousal, crossrace identifications.
Can we distinguish good witnesses from bad?
Bindemann, Brown, Koyas and Russ (2012):
Considerable individual variation in unfamiliar face
recognition ability (e.g. Burton, White and McNeill
2010). Could witnesses be tested on this?
Video of a staged theft followed by a lineup (either
TP or TA).
Followed by 1-in-10 face identification test (40 trials,
half TP).
Overall identification performance was poor. (TP
expt 1: 22% correct identifications; TP expt. 2: 18%).
For choosers, correlation between lineup and face
identification test performance (r = .70).
No good correlation for non-choosers (mixed bag?)
Darling, Martin, Hellmann and Memon (2009):
Measured individual differences in response to Navon letters.
Video of a staged bank robbery followed by a simultaneous (TP) lineup.
No. of participants
making correct lineup
identifications:
Strong global bias: 14/ 25
Weak global bias: 7 /25
High susceptibility to global interference when reporting small
letters was associated with better eyewitness identification
performance - individuals with strong "global" processing bias
might be better witnesses than ones with "local" bias.
Exposure time and delay:
MacLin, MacLin and Malpass (2001) review:
Increased exposure time usually improves recognition accuracy,
reduces false identifications.
Increased delay usually decreases recognition accuracy, increases
false identifications.
Bahrick, Bahrick and Wittlinger (1975):
Little effect of delay on familiar face recognition (e.g. schoolmates).
Loftus, Schooler, Boone and Kline (1987):
People overestimate duration of events, especially when stressed.
Read (1995):
Increased exposure time can decrease performance by increasing
witnesses' readiness to make false identifications (confuse
increased familiarity due to contextual information, with increased
familiarity from perceptual knowledge).
Effects of expectations and schemas:
Allport and Postman (1947):
Picture of a black man, and a white man holding knife.
Participants tended to recall that it was the black man
holding the knife.
Bartlett (1932):
"War of the Ghosts" - memory distortions stem from
attempts to make sense of events, relate them to
known facts, beliefs, etc.
Neisser (1981):
James Dean and Watergate testimony: memory for
gist, but inaccurate about temporal order of events,
who said what to whom, precisely what was said, etc.
Effects of expectations and schemas:
Bransford and Johnson (1972):
Comprehension rating
(1-7 scale, 1 = "poor")
Number of ideas recalled
(out of 18).
condition
no topic
topic after topic before
comprehension 2.29 (.22)
2.12 (.26)
4.50 (.49
recall
2.65 (.53)
5.83 (.49
2.82 (.60)
Effects
arousal:
of
stress
and
Yerkes-Dodson
"Law"
(1908):
inverted U-shaped relationship
between
stress/arousal
and
performance.
Easterbrook (1959): cue utilisation
hypothesis:
stress
narrows
attention to central items at
expense of peripheral ones.
Steblay (1992): "Weapon focus"
effect: decreased recognition due
to presence of a weapon.
Christianson (1992): effects of
stress are an interaction between
stress level and many other
factors.
Peters (1988):
Effects of stress on face recognition.
Memory for face of nurse and aide during immunisation.
Pulse rate higher for nurse than aide.
Aide identified better than nurse, from target-present
lineup.
Yuille and Cutshall (1986):
Naturalistic study of recall by 13 witnesses of a violent
crime. Their reports analysed for accuracy and also
compared with earlier police reports.
Accurate memories for events 4-5 months later.
Reported stress level at time of crime not significantly
related to subsequent recall. But higher-stress witnesses
also closer and more involved in the crime.
Odinot, Wolters and Koppen (2009):
Interviewed 14 witnesses to a supermarket robbery 3
months earlier.
Used CCTV recordings to assess memory accuracy.
84% of recalled information was correct.
Items provided in free recall were more accurate than
items produced by specific questions (90% vs 78%).
Large individual differences in accuracy (75% to 97%).
No effect of misinformation from TV re-enactment of the
crime.
Weak accuracy/confidence correlation (.38).
Higher levels of rumination were associated with higher
confidence levels (but not higher accuracy).
Higher levels of self-reported emotional impact were
associated wth greater accuracy.
Loftus, Loftus and Messo (1987):
Two filmed versions of an event in a restaurant.
Version A: a man pointed a gun at a cashier and she
handed him money.
Version B: the man gave her a cheque and she gave him
money.
Recorded eye-movements: version A subjects fixated on
the weapon more than version B subjects; also showed
poorer recall of other details and were less able to
identify the robber from a photo array.
Loftus and Burns (1982):
Violent and non-violent films of a "crime".
Subjects in violent version were less able to recall details
of the event. Memory impaired for details immediately
preceding the violent scene, and for details occurring up
to 2 mins earlier.
Are experiments on the effects of stress ecologically
valid?
Tollestrup, Turtle and Yuille (1994):
Lab research tells us about “unaffected witnesses” of
crimes but not victims.
Suspect identification rates by real-life witnesses –
Victims of robbery: 46%
Witnesses of robbery: 33%
Victims of fraud: 25%
Maybe stress enhances recognition?
Deffenbacher, Bornstein, Penrod and McGorty (2004):
Research confounds “defensive” and “orienting”
effects of stress, which either impede or enhance
eyewitness performance.
Real crimes likely to produce defensive responses,
Lab studies likely to produce orienting responses.
Consequence: lab studies may underestimate
impairment of eyewitness performance by stress.
Valentine and Mesout (2009):
Effects of stress on memory for London Dungeon actor.
High stress: 17% recognised him.
Low stress: 75% recognised him.
Morgan et al (2004):
509 U.S. Army survival school trainees.
Mock POW experience – sleep and food deprivation, then two
types of 45-min interrogation (high-stress and low-stress).
Identification of interrogators by live simultaneous lineups (15
members), and simultaneous or sequential mugshots (with/without
contextual cues): 8 members).
Overall, positive identifications of interrogators were higher in the
low-stress condition, and false positives were lower.
Large individual differences in ID performance:
42-45% unaffected by stress
42-50% worse when stressed
8-13% better when stressed.
No relationship between confidence and accuracy.
Overall, identification performance was poor, despite 30 min's
viewing.
Steblay (1992):
Meta-analysis of studies on "weapon focus".
Fairly reliable effect.
Pickel (1998, 1999):
Novelty can produce similar effects.
Effects of post-event misinformation:
Loftus, and Palmer (1974):
Confusion between originally-witnessed
information from post-event questioning.
Illustration from "Psychology: Themes and Variations, Sixth Edition, Briefer Version, Wayne Weiten"
event
and
Effects of post-event misinformation:
Loftus, Miller ands Burns (1978):
Confusion between originally-witnessed
information from post-event questions.
event
and
(a) Participants viewed slides of a
simulated car-pedestrian accident.
(b) Half the subjects got misleading
information about the accident while
the others got no misinformation.
(Told the car went past a "yield" sign
when actually it was a "stop" sign).
(c) All subjects tried to remember the
original accident: "stop" or "yield"?
75% of controls chose "stop", but
only 41% of misinformed subjects.
Influences of witnesses on each other "memory conformity":
Memon and Wright (1999):
1995 Oklahoma bombing
and hunt for "John Doe".
Wright, Self and Justice (2000):
Pairs were unaware they had seen different versions of a storybook
(present or absent "accomplice" to a crime). Discussion produced
conformity in pairs' responses.
Gabbert, Memon and Bull (2003):
Pairs saw different videos of a "theft".
60% of those who had not seen the crime came to believe it had
occurred;
30% who had seen it came to believe it had not occurred.
Paterson and Kemp (2006a)
Witnesses are more likely to recall misinformation
provided by a co-witness than from leading questions
or non-social sources such as the media.
Paterson and Kemp (2006b):
Survey of eyewitnesses - most reported there was at
least one other witness. Of these, 86% said they
discussed the event with their co-witness.
Paterson, Kemp and Forgas (2009):
Implies that witness testimonies should not be treated
as independent observations.
Paterson, Chapman and Kemp (2007):
77% of witnesses said they were acquainted with other
witnesses at the scene.
Hope, Ost, Gabbert, Healey and Lenton (2008):
Pairs saw two different versions of a video: one showed
a girl steal money from a wallet, other did not.
50% of friends, 50% of couples but only 17% of
strangers said they had seen the girl steal the money,
even though this was not in their version of the event.
Amongst stranger pairings, participants rated their
partner more positively were also more likely to accept
misinformation from them.
Conclusions:
Numerous factors affect eyewitness performance at
encoding and retrieval phases.
Encoding affected by stress, although difficult to
investigate experimentally.
Retrieval is affected by operation of standard memory
effects (schemas, trace decay, interference, etc.) plus
social factors (influence of questioner and other
witnesses) plus individual differences in witness
competence.
Exacerbated by own-race bias (ORB) in cases involving
cross-race identification.