Eyewitness Reliability: Psychologists, Police and the Courts

1
(No Transcript)
2
How accurate are eyewitnesses?
spenrod_at_jjay.cuny.eduWebpage http//penrods.net
paper password rdp
3
Innocence Project

• As of September 2004
• -- 151 DNA exonerations
• -- eyewitness errors in 61 of first 70 cases
• -- Eyewitness error remains the single most
  important cause of wrongful imprisonment.
• http//www.innocenceproject.org/

4
(No Transcript)
5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
Gee, he looks just like his suspect sketch.
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
Are these errors exceptional?Do witnesses
generally perform well?How do we know?news
archives scientific experiments police
records
17

• Edwin Borchard (1932) Convicting the Innocent
• 65 Cases in 27 states and England
• Main cause mistaken identification
• Brandon and Davies (1973)Describe 70 cases of
  wrongful conviction due to inaccurate
  eyewitnesses

18
Huff Wrongful Conviction (1987)

• Searched 1,100 magazines and journals
• Found 500 cases of erroneous convictions
• Eyewitness error involved in nearly 60
• Rattner (1988)
• 205 wrongfully convicted defendants
• 52 due to inaccurate eyewitness testimony

19
Eyewitness Experiments

• Eyewitness/memory research is science
• Falsifiable hypotheses-Testable theories
• Most research based on true experiments
• Also field and simulation methodologies
• Peer reviewed often twice
• Quantity of research?
• 1,000s on memory500 on facial memory and
  eyewitness memory
• 95 of research post-1975

20
Growth in Eyewitness Research Published
Peer-Reviewed Articles
21
Fulero and Clark on Experiments
22
Buckhout (1975) N.Y. television station
broadcast a simulated purse-snatching

• Showed lineup of 6 menincluding perp TP
• A phone number was provided for viewers to call
  in and identify perpetrator
• 14.1 were correct (out of 2,000 viewers, 1,843
  were wrong)
• Performance no better than chancewitnesses were
  simply guessing and could have performed just as
  well without seeing the purse-snatching

23

• Performance in Staged crime experiments
• Haber Haber (2001) meta-analysis of 44
  studies
• Perpetrator Present
  Perpetrator Absent
•     Cor-   
  Foil     Miss      Cor- Foil 
•     rect ID rect ID
• Averages 47 28 25
  50 50

1 in 3 perp-present positive IDs is an erroneous
foil identification/bad guess 1 in 2 perp
present make an error 1 in 2 witnesses make a
bad guess in perp absent lineups
24
Field Experiments

• Brigham et al. (1982)
• 73 convenience store clerks, 2 hour delay
• Krafka and Penrod (1985)
• 85 convenience store clerks, 2 or 24 hour delay
• Platz and Hosch (1988)
• 86 convenience store clerks, 2 hour delay
• Pigott et al. (1990)
• 47 bank tellers, 4-5 hour delay

25
Identification Accuracy Rates in Field Experiments
Half of perp-present IDs are errors 58 of
perp-present witness make an error More than1/3
of witnesses in perp-absent guess wrong
26
Studies of Real Witnesses

• Problem generally cannot know if a suspect IS
  the perpetratorinstead, the focus is on suspect
  IDs

27
-- suspectsmix of guilty and innocent unclear --
9-person arrays -- more than 1/3 of IDs are
clearly bad guesses of foils -- some suspect
IDs must also be guesses min 2.5
28

• The informative databecause it includes foil
  IDs
• 1 in 3 IDs is a bad guess
• Some suspect IDs are guesses minimum of about
  5
• Some suspects are innocent
• Other data indicate showups produce false IDs at
  about doulbe the rate of fair lineups

29

• Guessing the suspect
• Six-person lineups 24 of witnesses chose one
  of five foils--about 5 per foil.
• In a fair array an equal percentage of witnesses
  should guessed/choose the suspect and the foils
  but suspect guesses look like memory-based
  positive IDs.
• Biased arrays.
• But, as we shall see there is good reason to
  think suspects are guessed more often than
  foils-two to three times more often-due to bias
  against suspects.

30
(No Transcript)
31
Just how bad are the different types of errors
and are lucky guesses good?
32
18 of 43 IDs were errors--21/39 Hits in TP 54
11/19 FA in TP 57--as in experiments 4 of 29
suspects identified were innocent A lot of
guessing going on (32 of 58 witnesses)the
subject of current research
33

• Overallin 58 Sacramento-area lineups--a lot of
  guessing and errors Penrod, 2003
• 32 of 58 witnesses guessed25 made errors
• --14 foil selections by burnt witnesses
• -- 4 of the 29 suspects identified were
  innocent and perp is on the streets
• -- 4 lucky-guess perp IDs
• -- 7 failures to ID in perp-present lineups
• -- 3 lucky-guess rejections in innocent
  suspect lineups
• 26 correct witnesses with good memories
• -- 21 memory-based perp IDs
• -- 5 memory-based rejections of innocent
  suspect lineups

34

• Police files and experimental studies indicate
• Substantial rates of guessing
• Substantial rates of foil innocent suspect IDs
• Poor performance raises question of whether IDs
  should be admissiblewould/do we admit results of
  lab tests which perform this poorly?
• Given the error rates, should IDs be admitted in
  the absence of a govt showing that
  scientifically valid methods were used to
  minimize the risk of error?

35
(No Transcript)
36
Diagnosticity

• How useful and harmful are the decisions made by
  witnesses?
• How much weight should we give to correct
  decisions versus different types of errors?
• What should we be trying to maximize and minimize?

37

• Standard Measures of Test Quality
• Sensitivity TP/(TPFN) the percentage of perps
  who are identified 25/39
• Specificity TN/(TNFP) the percentage of
  innocent suspects not identified 15/19
• Predictive Value Positive (P) TP/(TPFP) the
  probability that someone identified is the perp
  25/29
• Predictive Value Negative (P-) TN/(TNFN) the
  probability that someone not identified is
  innocent 15/29
• Diagnosticity Ratio TPFP 254 or 6.251

38
Table 1 Composite-Based Recognition Performance
as a Function of Preparation and Identification
Conditions Prior Research unweighted hit rate
averages 4 Studies View
View 50 1
Study View Memory 58
-22 vs V-V 10
Studies Memory View 28
None
Memory Memory none Kovera, Penrod,
Pappas Thill-1997 Memory Memory ??
Study Prep phase ID phase
hits Ellis, Shepherd, Davies (1975) Memory
View 13 Davies, Ellis, Shepherd
(1978) Memory View 16 Christie Ellis
(1981) Memory
View 23 Green Geiselman (1989) Study A
Memory View 33 Study
B Good composite Memory View 46
Poor composite Memory
View 19 Davies (1986) Good
composite Memory View 60
Poor composite Memory
View 19 Gibling Bennett (1994)
View View 54 Davies, Ellis,
Christie (1981) View View 31

Memory View 31 Christie, et al.
(1981) View View 34

Memory View 24 Cutler, Stocklein,
Penrod (1988) View View 80

View Memory 58
39

• Ten college freshmen prepared composites from
  memory and 50 viewers attempted to recognize the
  composites. The 10 preparers (two students from
  each of five schools) constructed five student
  and five faculty faces-- 100 composites in all.
• In the recognition phase, 50 students drawn from
  the same high schools as the preparers were told
  that they would be viewing composites that
  depicted students and faculty from local high
  schools, including their own 10 of 50 from own
  school.

40

• Viewers were unable to differentiate their
  classmates (old faces) from faces of students
  attending other schools (new faces)
• Viewers rarely generated a name for a composite
  (lt7).
• Variation in preparer familiarity with the
  target, preparer-assessed composite quality, and
  viewer familiarity with the target did not affect
  viewer recognition, viewer confidence, or name
  generation.
• Participants generated only three correct names
  for 500 possible old faces.

41
Table 1 Composite-Based Recognition Performance
as a Function of Preparation and Identification
Conditions Prior Research unweighted hit rate
averages 4 Studies View
View 50 1
Study View Memory 58
-22 vs V-V 10
Studies Memory View 28
None
Memory Memory none Kovera, Penrod,
Pappas Thill-1997 Memory
Memory 01 Study Prep phase ID phase
hits Ellis, Shepherd, Davies
(1975) Memory View 13 Davies, Ellis,
Shepherd (1978) Memory View 16 Christie
Ellis (1981) Memory
View 23 Green Geiselman (1989) Study A
Memory View 33 Study
B Good composite Memory View 46
Poor composite Memory
View 19 Davies (1986) Good
composite Memory View 60
Poor composite Memory
View 19 Gibling Bennett (1994)
View View 54 Davies, Ellis,
Christie (1981) View View 31

Memory View 31 Christie, et al.
(1981) View View 34

Memory View 24 Cutler, Stocklein,
Penrod (1988) View View 80

View Memory 58