Content uploaded by Gary L. Wells
Author content
All content in this area was uploaded by Gary L. Wells on Oct 20, 2014
Content may be subject to copyright.
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
10.1146/annurev.psych.54.101601.145028
Annu. Rev. Psychol. 2003. 54:277–95
doi: 10.1146/annurev.psych.54.101601.145028
Copyright
c
° 2003 by Annual Reviews. All rights reserved
First published online as a Review in Advance on August 27, 2002
EYEWITNESS TESTIMONY
Gary L. Wells and Elizabeth A. Olson
Psychology Department, Iowa State University, Ames, Iowa 50011;
e-mail: glwells@iastate.edu; vorndran@iastate.edu
Key Words witness, testimony, lineup
■ Abstract The criminal justice system relies heavily on eyewitness identification
for investigating and prosecuting crimes. Psychology has built the only scientific liter-
ature on eyewitness identification and has warned the justice system of problems with
eyewitness identification evidence. Recent DNA exoneration cases have corroborated
the warningsofeyewitness identification researchers by showingthatmistakeneyewit-
ness identification was the largest single factor contributing to the conviction of these
innocent people. We review major developments in the experimental literature con-
cerning the way that various factors relate to the accuracy of eyewitness identification.
These factors include characteristics of the witness, characteristics of the witnessed
event, characteristics of testimony, lineup content, lineup instructions, and methods of
testing. Problems with the literature are noted with respect to both the relative paucity
of theory and the scarcity of base-rate information from actual cases.
CONTENTS
COVERAGE OF THIS REVIEW ......................................... 278
BASIC CONCEPTS ................................................... 278
ESTIMATOR VARIABLES ............................................. 280
Characteristics of the Witness .......................................... 280
Characteristics of the Event ............................................ 281
Characteristics of Testimony ........................................... 283
Lay Observers’ Judgments of Accuracy .................................. 284
SYSTEM VARIABLES ................................................. 285
Culprit-present Versus Culprit-absent Lineups ............................. 286
Instructions ........................................................ 286
Lineup Content ..................................................... 287
Lineup Presentation Method ........................................... 288
Behavioral Influence: The Need for Double-Blind Testing ................... 289
Base Rates as System Variables ......................................... 289
PROBLEMS AND PROSPECTS ......................................... 290
Eyewitnesses are critical in solving crimes, and sometimes eyewitness testimony
is the only evidence available for determining the identity of the culprit. Psycho-
logical researchers who began programs in the 1970s, however, have consistently
0066-4308/03/0203-0277$14.00 277
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
278 WELLS
¥
OLSON
articulated concerns about the accuracy of eyewitness identification. Using var-
ious methodologies, such as filmed events and live staged crimes, eyewitness
researchers have noted that mistaken identification rates can be surprisingly high
andthateyewitnessesoftenexpresscertaintywhentheymistakenlyselectsomeone
from a lineup. Although their findings were quite compelling to the researchers
themselves, it was not until the late 1990s that criminal justice personnel began
takingtheresearch seriously.Thischange inattitudeaboutthepsychological litera-
ture on eyewitness identification arose primarily from the development of forensic
DNA tests in the 1990s. More than 100 people who were convicted prior to the
advent of forensic DNA have now been exonerated by DNA tests, and more than
75% of these people were victims of mistaken eyewitness identification (Wells
et al. 1998, Scheck et al. 2000). The apparent prescience of the psychological
literature regarding problems with eyewitness identification has created a rising
prominence of eyewitness identification research in the criminal justice system
(Wells et al. 2000).
Because most crimes do not include DNA-rich biological traces, reliance on
eyewitness identification for solving crimes has not been significantly diminished
by the development of forensic DNA tests. Interestingly, research on eyewitness
reliability has been done only by psychologists—primarily cognitive and social
psychologists—and the psychological literature represents the only source of em-
pirical data on eyewitness identification. The vast criminal justice system itself has
never conducted an experiment on eyewitness identification.
COVERAGE OF THIS REVIEW
No review of the eyewitnessidentification literature haspreviouslyappeared in the
Annual Review of Psychology. Therefore, we include here references to articles
fromthe 1970sand1980s thatwe think especiallycritical to thedevelopmentofthe
literature, but we primarily emphasize more recent developments. Also, because
the eyewitness identification literature has become so vast, we are necessarily se-
lective in our citations and coverage. Readers should note that this review focuses
on eyewitness identification rather than on eyewitness testimony in general. Eye-
witnesses commonly testify about many things, such as which hand a gunman
used, the color of a car, or recollections of a conversation, but these event memo-
ries are outside the scope of this review. The large literature on child eyewitnesses,
suggestibility, and recovery of repressed memories is not reviewed here.
BASIC CONCEPTS
The eyewitness identification literature has developed a number of definitions and
concepts that require explanation. A lineup is a procedure in which a criminal
suspect (or a picture of the suspect) is placed among other people (or pictures of
other people) and shown to an eyewitness to see if the witness will identify the
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 279
suspect as the culprit in question. The term suspect should not be confused with
the term culprit. A suspect might or might not be the culprit (a suspect is suspected
of being the culprit). Fillers are people in the lineup who are not suspects. Fillers,
sometimes called foils or distractors, are known-innocent members of the lineup.
Therefore, the identification of a filler would not result in charges being brought
against the filler. A culprit-absent lineup is one in which an innocent suspect is
embeddedamongfillersandaculprit-presentlineupisoneinwhichaguiltysuspect
(culprit) is embedded among fillers. The primary literature sometimes calls these
target-present and target-absent lineups.
A simultaneous lineup is one in which all lineup members are presented to
the eyewitness at once and is the most common lineup procedure in use by law
enforcement. A sequential lineup, on the other hand, is one in which the witness
is shown only one person at a time but with the expectation that there are several
lineup members to be shown.
A lineup’s functional size is the number of lineup members who are “viable”
choices for the eyewitness. For example, if the eyewitness described the culprit as
being a tall male with dark hair and the suspect is the only lineup member who
is tall with dark hair, then the lineup’s functional size would be 1.0 even if there
were 10 fillers. Functional size was introduced as a specific measure (Wells et al.
1979), and competing measures have been proposed, such as Malpass’s (1981)
“effective size.” Today functional size is used generically to mean the number of
lineup members who fit the eyewitness’s description of the culprit.
Mock witnesses are people who did not actually witness thecrime butare asked
to pick a person from the lineup based on the eyewitness’s verbal description of
the culprit. Mock witnesses are used to test the functional size of the lineup.
The diagnosticity of suspect identification is the ratio of accurate identification
rate with a culprit-present lineupto the inaccurate identification rate with a culprit-
absent lineup. The diagnosticity of “not there” is the ratio of “not there” response
rates with culprit-absent lineups to “not there” response rates with culprit-present
lineups. The diagnosticity of filler identifications is the ratio of filler identification
rates with culprit-absent lineups to filler identification rates with culprit-present
lineups.
Among variables that affect eyewitness identification accuracy, a system vari-
able is one that is (or could be) under control of the criminal justice system, while
an estimator variable isone that is not.System variablesinclude instructionsgiven
to eyewitnesses prior to viewing a lineup and the functional size of a lineup. Esti-
mator variables include lighting conditions at the time of witnessing and whether
the witness and culprit are of the same or of different races.
The distinction between estimator and system variables has assumed great sig-
nificance in the eyewitness identification literature since it was introduced in the
late 1970s (Wells 1978). In large part, the prominence of this distinction attests to
the applied nature of the eyewitness identification literature. Whereas the devel-
opment of a literature on estimator variables permits some degree of postdiction
that might be useful for assessing the chances of mistaken identification after the
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
280 WELLS
¥
OLSON
fact, the development of a system variable literature permits specification of how
eyewitness identification errors might be prevented in the first place.
ESTIMATOR VARIABLES
Estimator variables can be sorted into four broad categories: characteristics of the
witness,characteristicsoftheevent,characteristicsofthetestimony,andabilitiesof
the testimony evaluators to discriminate between accurate and inaccurate witness
testimony.
Characteristics of the Witness
Are members of certain groups better eyewitnesses than those of others? The
empirical evidence is not overwhelming. For example, there is no clear evidence
that males and females differ significantly overallin ability to identify people from
lineups. A meta-analysis byShapiro & Penrod(1986) indicated that females might
beslightly morelikelytomake accurateidentifications butalsoslightlymore likely
to make mistaken identifications than are males (due to females being more likely
toattemp anidentification),thereby yieldinganoverallequivalent diagnosticityfor
males and females. Although males and females might take an interest in different
aspects of a scene and thereby remember somewhat different details (e.g., Powers
et al. 1979), overall abilities of males and females in eyewitness identification
appear to be largely indistinguishable (but see Brigham & Barkowitz 1978, Shaw
& Skolnick 1999).
The age of the eyewitness, on the other hand, has been consistently linked to
eyewitness identification performance, with very young children and the elderly
performing significantly worse than younger adults. The eyewitness identification
errors of young children and the elderly are highly patterned: When the lineup
contains the actual culprit, young children and the elderly perform nearly as well
as young adults in identifying the culprit, but when the lineup does not contain
the culprit the young children and the elderly commit mistaken identifications at a
higher rate than do young adults (see the meta-analysis on children versus adults
by Pozzulo & Lindsay 1998).
There is little evidence that intelligence is related to eyewitness identification
performance. Although an early study by Howells (1938) indicated a significant
relation between face recognition accuracy and intelligence, later studies have
shown no relation (e.g., Brown et al. 1977). A word of caution is in order here,
however, because Howells’s sample of witnesses included a much greater range
of intelligence at the low end than have later studies. At the low extremes of
intelligence, a pattern similar to that found with children seems likely, namely a
high rate of mistaken identifications in response to culprit-absent lineups.
The race of the eyewitness has been examined extensively. Although no con-
sistent overall differences attributable to race have emerged, the evidence is now
quite clear that people are better able to recognize faces of their own race or
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 281
ethnic group than faces of another race or ethnic group. A recent meta-analysis
by Meissner & Brigham (2001) shows that this effect is robust across more than
25 years of research.
Little published research relates personality characteristics to eyewitness iden-
tification accuracy. Hosch et al. (1984) found that high self-monitors (individuals
who adapt their behavior to cues regarding what is socially appropriate) are more
susceptible to biased lineup procedures than are low self-monitors, and Hosch &
Platz (1984) found a relation between self-monitoring and correct identifications.
Also, a meta-analysis by Shapiro & Penrod (1986) indicated that individuals high
in chronic trait anxiety (a general attitude of apprehension) made fewer mistaken
identifications than individuals low in chronic trait anxiety. Their meta-analysis
also indicated that field independents (those with a perceptual tendency to dif-
ferentiate parts of a visual field from the whole) made fewer accurate identifi-
cations (but equal mistaken identifications) than did field dependents. However,
little research has been directed at the role of personality in eyewitness identifi-
cation, and no strong theory relating personality to eyewitness identification has
emerged.
Characteristics of the Event
A variety of factors affect the ability of an eyewitness to identify the culprit at
a later time, including the amount of time the culprit is in view, the lighting
conditions, whether the culprit wears a disguise, the distinctiveness of the culprit’s
appearance, the presence or absence of a weapon, and the timing of knowledge
that one is witnessing a crime.
Distinctive facesare much more likely to be accurately recognized than nondis-
tinctive faces (e.g., Light et al. 1979). Faces that are highly attractive or highly
unattractive are easier to recognize than are faces that are average in attractive-
ness (e.g., Fleishman et al. 1976), but what makes a face distinctive is not entirely
clear. Because the arithmetic mean (averaged at the pixel level) of several faces
(a prototype) is judged to be more attractive than the individual faces that were
averaged (see Langois & Roggman 1990), the distinctiveness-recognition relation
is probably not due to a simple deviation from the arithmetic mean of individual
facial features.
Simple disguises, even those as minor as covering the hair, result in significant
impairmentofeyewitnessidentification(Cutleretal.1987).Sunglasses alsoimpair
identification, although the degree of impairment can be reduced by having the
targets wear sunglasses at the time of the recognition test (Hockley et al. 1999).
Photos of criminal suspects used in police lineups are sometimes severalyears old.
Changes in appearance that occur naturally over time and changes that are made
intentionally by suspects can have quite strong effects on recognition. Read et al.
(1990) found that photos of the same people taken two years apart were less likely
to be recognized as the same people when their appearance had naturally changed
(via aging, facial hair) than when their appearance had remained largely the same.
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
282 WELLS
¥
OLSON
Clearly, at the extreme of low light levels there is a point at which a face cannot
be perceived well enough to be recognized later. Surprisingly, however, we know
of no experiments that have measured the light levels required for the encoding of
faces. We encourage researchers to address this question.
As would be expected, the amount of time a culprit’s face is in view affects
the chances that the eyewitness can identify the person later (Ellis et al. 1977).
However, this relationship depends less critically on the eyewitness’s opportunity
to viewper se and more on the amount and type of attentionthat the witnessdirects
at the culprit. Given equal exposure time to a face, people are more likely to be
able to recognize that face later if they make abstract inferences about it (e.g., is
this person honest?) than if they make physical judgments (e.g., does this person
have a large or small nose?). Presumably, this effect occurs because the abstract
inferences require holistic processing of the face whereas the physical judgments
require feature processing (Wells & Hryciw 1984).
In general, the amount of time a culprit’s face is in view is not as critical for
eyewitness identification accuracy as the type or amount of attention given by
the witness. For example, Leippe et al. (1978) exposed unsuspecting people to a
staged theft of a package. Some were led to believe that the package contained a
valuable item and some were led to believe that the package contained a trivial
item. In addition, some learned of the value of the item in the package before the
theft and some only learned the value after the thief had fled. Although all had the
same opportunity to view the thief, the witnesses who knew the value of the item
beforehand were significantly more accurate at identification than the other three
groups. Observers often do not realize that they have witnessed a crime until after
the culprit has fled. Although they might have had significant opportunity to view
the culprit, they might have had little reason to attend closely.
One factor that can signal to eyewitnesses that a crime is occurring is the pres-
ence of a weapon. Unfortunately, learning that one is an eyewitness to a crime
via the culprit’s display of a weapon might not make the person a better eye-
witness. A number of studies have been directed at the question of the so-called
weapon-focus effect. A meta-analysis of these studies indicates that the presence
of a weapon reduces the chances that the eyewitness can identify the holder of
the weapon (Steblay 1992). Loftus et al. (1987) monitored eyewitnesses’s eye
movements and found that weapons draw visual attention away from other things
such as the culprit’s face. Complicating the issue somewhat is the fact that the
presence of weapons or other types of threatening stimuli can cause arousal, fear,
and emotional stress. The effects of such stress on memory are still being debated.
Some research shows that increased levels of violence in filmed events reduces
eyewitness identification accuracy (e.g., Clifford & Hollin 1981) whereas other
research has failed to find this effect(e.g., Cutler et al. 1987). Deffenbacher (1983)
suggested that the effect is likely to follow the Yerkes-Dodson Law where only
veryhigh and verylowlevelsofarousalwill impair memory. Christianson’s (1992)
reviewofthe evidencerelatingemotionalstressto memory suggeststhatemotional
events receive preferential processing; emotional response causes a narrowing of
attention (as suggested by Easterbrook 1959) with loss of peripheral details.
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 283
Characteristics of Testimony
Considerable interest and research have been directed at the question of whether
there are characteristics of aneyewitness’s testimonythat could be used to postdict
whether the witness made an accurate or false identification. The bulk of this re-
search has focused on the certainty (confidence) of the eyewitness. Although early
research suggested that the certainty an eyewitness expresses in an identification
is largely unrelated to the accuracy of the identification, current analyses suggest
a more hopeful but also more complex view of the certainty-accuracy relation.
Although any given experiment might show a statistically nonsignificant relation
between certainty and accuracy, meta-analyses of the literature show a reliable
correlation. Several moderators of the strength of the relation have been identi-
fied. One important moderator is the overall accuracy of the eyewitnesses. When
accuracy is low (e.g., from poor witnessing conditions), the certainty-accuracy
relationship suffers (Bothwell et al. 1987). Later meta-analyses indicate that the
certainty-accuracy relation is stronger if the analysis is restricted to those mak-
ing an identification (choosers only) than if it also includes witnesses who make
correct and false rejections (Sporer et al. 1995). In fact, using a weighted average
of effect sizes for choosers only, Sporer et al. reported a 0.37 certainty-accuracy
correlation across 30 studies. More recent work indicates that directing eyewit-
nesses to reflect on their encoding and test conditions or asking them to entertain
hypotheses regarding why their identification might have been mistaken can im-
prove the relation between accuracy and certainty, especially when this relation
is calculated using calibration methods rather than the point-biserial correlation
(Brewer et al. 2002).
Although the 0.37 correlation estimate for the certainty-accuracy relation is
more optimistic than theearly estimates, recentstudies suggest the literature might
be overestimating the utility of eyewitness certainty in actual cases. In a series of
experiments, eyewitness certainty was shown to be highly malleable among eye-
witnesses who had made mistaken identifications (Wells & Bradfield 1998,1999).
After making mistaken identifications, some eyewitnesses were given confirming
feedback by the lineup administrator (“Good, you identified the suspect”) whereas
others were given no feedback about their identification. This feedback served to
distort the eyewitnesses’ recollections of the certainty they had in their identifica-
tions. Those given confirming feedback recalled having been very certain in their
identification compared to those given no confirming feedback. This certainty-
inflation effect is greater for eyewitnesses who make mistaken identifications than
it is for those who make accurate identifications, resulting in a significant loss in
the certainty-accuracy relation (Bradfield et al. 2002). In actual cases, it is com-
mon for lineup administrators (usually the detectivein the case) to give confirming
feedbackto eyewitnesses,therebyinflatingthecertainty of the eyewitnessandcon-
founding the certainty-accuracy relation. Even if the lineup administrator refrains
from giving the witness confirming feedback, the witness is likely to make con-
firming inferences from later events (e.g., an indictment of the identified person).
Another real-world factor that can muddle the meaning of eyewitness certainty is
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
284 WELLS
¥
OLSON
repeated testing. Shaw and his colleagues (Shaw 1996, Shaw & McClure 1996)
have shownthatrepeatedquestioning ofeyewitnesses onamatter about whichthey
were inaccurate servesto inflate their certainty that they were accurate. Hence, it is
unclear whether the .37 correlation between certainty and accuracy revealed in the
Sporer et al. meta-analysis of experiments can be directly applied to actual cases
in which there are other influences that inflate the certainty of eyewitnesses.
An even more promising indicator of eyewitness accuracy is the speed with
which the eyewitness makes an identification from a lineup. Several studies have
now found that witnesses who make accurate identifications from a lineup reach
their decision faster than do witnesses who make mistaken identifications
(Dunning & Perretta 2002; Dunning & Stern 1994; Robinson et al. 1997; Smith
et al. 2000; Sporer 1992, 1993, 1994). In an impressive set of results, Dunning &
Perretta found that those who made their decision in less than 10–12 seconds were
nearly 90% accurate in their identifications from a lineup whereas those taking
longer were approximately 50% correct. The 10–12-second rule was developed
post hoc to produce the best separation of accurate and inaccurate witnesses, so
some caution is called for with regard to how well the 10–12-second rule works
in other situations; but the general relation between accuracy and speed of iden-
tification has received support in several studies. In addition, the idea that faster
identifications are more likely to be accurate than are slower identifications makes
good theoretical sense. It has long been theorized that mistaken identifications re-
sult from a deliberated judgment in which witnesses compare one lineup member
to another and use inferences and elimination strategies to decide which person
must be the culprit whereas accurate identifications result from a more automatic
recognition process that does not require comparisons of one lineup member to
another (Wells 1984a).
Lay Observers’ Judgments of Accuracy
Observers (e.g., jurors) have little ability to make correct discriminations between
accurate and inaccurate eyewitness identification testimony. Severalmethods have
been used to assess the adequacy of people’s judgments about eyewitness iden-
tification accuracy. Surveys, for example, show poor agreement (often less than
50%) between the answers that lay people give about variables affecting eyewit-
ness identification accuracyand the answers researchersscore correct based on the
empirical literature (e.g., Deffenbacher & Loftus 1982, McConkey & Roche 1989,
Noon & Hollin 1987). Another approach has been to use “prediction” studies in
which eyewitness identification experiments are described and people are asked
to predict the results. The results of these studies show a tendency to overestimate
eyewitness identification accuracy and a failure to correctly predict interactions
between variables (e.g., Brigham & Bothwell 1983, Wells 1984b).
A third approach is to cross-examine eyewitnesses to staged crimes and to
ask subject-jurors to determine whether witnesses made accurate or mistaken
identifications. In a series of experiments using this methodology, subject-jurors
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 285
have shown little or no ability to make such discriminations (Lindsay et al. 1989,
Lindsay et al. 1981, Wells et al. 1981, Wells & Leippe 1981, Wells et al. 1979).
Because observers’ belief rates exceeded eyewitnesses’ accuracy rates, these
studies are commonly cited as evidence that people are overbelieving of eye-
witnesses. However, this pattern of overbelief is restricted primarily to poorer
witnessing conditions; when witnessing conditions were good, belief rates and
eyewitness identification accuracy rates were more similar. In addition, mock
jurors sometimes underbelieved the eyewitnesses who had quite low levels of
certainty.
SYSTEM VARIABLES
System variables are those that affect the accuracy of eyewitness identifications
and over which the criminal justice system has (or can have) control. In general,
these tend to be lineup test factors, such as how witnesses are instructed prior to
viewing a lineup or how the lineup is structured. The distinction between system
variablesandestimatorvariablesisconsequential inseveralrespects.Whereasesti-
mator variables can at best increase the probability that the criminal justice system
can sort accurate from inaccurate eyewitness identifications, system variables can
help prevent inaccurate identifications from occurring in the first place. Consider,
for instance, the idea that jurors tend to overbelieve eyewitness identification tes-
timony. Although expert testimony about eyewitness identification might manage
to reduce jurors’ tendencies to overestimate eyewitnesses’ accuracy, the system
variable approach might enable eyewitness identification accuracy to match the
level of jurors’ beliefs (Seelau & Wells 1995).
Theprocedureusedbycrime investigators conducting a lineup has been likened
to that of researchers conducting an experiment (Wells & Luus 1990). Crime
investigatorsbeginwithahypothesis(that thesuspect isthe culprit),create adesign
for testing the hypothesis (embed the suspect among fillers), carry out a procedure
(e.g., provide pre-lineup instructions and present the group to an eyewitness),
observeand record the eyewitness’sbehavior(witnessdecision), and then interpret
and revise their hypothesis (whether the suspect is the culprit). All the types of
things that can go wrong with an experiment to cause misleading results can also
go wrong with a lineup. For instance, the instructions might bias the witness, the
hypothesis might be prematurely leaked, the design might be flawed, the behavior
mightbemisinterpreted,confirmationbiasesmightbeoperating,andsoon.Indeed,
a great deal of the research literature on system variable eyewitness identification
could be construed as the extension of sound experimental methodology to the
design and procedure of police lineups.
Most system variable research in eyewitness identification can be placed into
four categories: instructions, content, presentation method, and behavioral influ-
ence. Before reviewing these system variables, however, it is important to under-
stand the role played by the presence versus absence of the culprit in the lineup
and the concept of a relative-judgment decision process.
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
286 WELLS
¥
OLSON
Culprit-present Versus Culprit-absent Lineups
A lineup might or might not include the actual culprit. If police investigators have
unknowinglyfocused on an innocent person as their suspect and place that suspect
inthelineup,thentheeyewitness(es)willendupviewingalineupforwhichtheonly
correctansweris“notthere.”Researchrepeatedlyshowsthatculprit-absentlineups
present great problems for eyewitnesses. The same eyewitnesses who identified
an innocent person from a culprit-absent lineup might otherwise have been able to
identifythe actual culprit from a culprit-present lineup (Wells1984a).In one study,
for example, 54% of eyewitnesses were able to identify the actual culprit from a
6-person culprit-present lineup and 21% made no identification. When the culprit
was removed without replacement (making it a 5-person culprit-absent lineup),
however, the rate of no identification rose only to 32%, with the other 68% of
the eyewitnesses who saw this lineup mistakenly identifying someone from the 5
remaining members of the lineup (Wells 1993).
A theoretical view that has been used heavily in the eyewitness identification
literature is that eyewitnesses tend to use a relative-judgment decision process in
making identifications from a lineup (Wells 1984a). The relative-judgment con-
ceptualization states that an eyewitness tends to select a person from a lineup who
most resembles the eyewitness’s memory of the culprit relative to the other lineup
members. Although the relative-judgment decision process permits eyewitnesses
to do a reasonable job of identifying the culprit from a culprit-present lineup,when
eyewitnesses view a culprit-absent lineup there will likely be one lineup member
who looks more like the culprit than the others.
An alternative explanation of the errors witnesses make with culprit-absent
lineups is that eyewitnesses tend to have lax criteria of resemblance; under culprit-
absent circumstances, innocent lineup members easily meet these undemanding
criteria (Ebbesen & Flowe 2002). Experimental data have not yet favored one
of these interpretations over the other. Recent mathematical modeling of lineup
data by Clark may help to refine our understanding of the roles of both relative
judgments and criterion setting (Clark 2002).
Instructions
A variable shown repeatedly to have considerable impact on eyewitness identifi-
cations from lineups is the pre-lineup instruction given to eyewitnesses. Malpass
& Devine (1981) were the first to demonstrate that the ratio of accurate to inaccu-
rate identifications is strongly affected by whether or not eyewitnesses have been
instructed (warned) prior to viewing the lineup that the culprit might or might
not be in the lineup. A meta-analysis of the eyewitness identification literature on
pre-lineup instructions reveals that the loss of accurate identifications from such
instructions is minimal whereas the reduction of mistaken identifications is con-
siderable (Steblay 1997). Steblay’s meta-analysis showed that the presence of the
“might or might not be present” instruction (compared to no instruction) reduced
mistaken identification rates in culprit-absent lineups by 41.6% whereas accurate
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 287
identification rates in culprit-present lineups were reduced by only 1.9%. Based
on this compelling research, the U.S. Department of Justice included this type of
instruction in its first set of national guidelines for law enforcement on the collec-
tion of eyewitness evidence (Technical Working Group for Eyewitness Evidence
1999).
Lineup Content
When police have a suspect and decide to conduct a lineup, nonsuspect (filler)
members of the lineup must be chosen. The importance of the selection of fillers
as a system variable was demonstrated early, and it remains one of the primary
active issues in the eyewitnessidentification literature. Ideally, lineup fillers would
be chosen so that an innocent suspect is not mistakenly identified merely from
“standing out,” and so that a culprit does not escape identification merely from
blending in. The first experimental demonstration of the importance of filler selec-
tion showed what can happen when this idea is not achieved. When fillers did not
at all resemble the culprit, eyewitnesses tended to mistakenly identify an innocent
suspect who resembled the culprit; when the suspect was the culprit, however,
the manipulation of fillers had little effect on the rate of accurate identifications
(Lindsay & Wells 1980).
Although the issue of lineup fillers seems simple at first glance, it is in fact
complex. In the early demonstrations, researchers used their knowledge of the
culprit’s identity to select fillers. In actual cases, of course, the identity of the
culprit is not known. Using the suspect as a proxy for the culprit will havedifferent
effects on rates of accurate and mistaken identification depending on whether the
suspect is the culprit or an innocent person. Accordingly, selecting fillers who
are highly similar to the suspect can help protect the innocent suspect in a culprit-
absentlineup,butcanalsoreduceaccurateidentificationsinaculprit-presentlineup
(Luus & Wells 1991). Another line of research has shown that using the suspect
as the reference point to select fillers can create a “backfire effect” in which an
innocent suspect, being the origin or central tendency of the lineup, actually has
an increased chance of being identified as the culprit (Clark & Tunnicliff 2001,
Navon 1992, Wogalter et al. 1992).
An alternative to the strategy of selecting fillers based on their resemblance to
the suspect is to select fillers based on their fit to the verbal description the eyewit-
ness had given of the culprit. This fit-to-description strategy has several practical
advantages(see Wellsetal.1994) andhasworkedwell insomeexperiments(Juslin
et al. 1996, Wells et al. 1993). However, biases against the innocent suspect can
remain with the fit-to-description method when the description is especially sparse
or when the innocent suspect happens to show a high resemblance to the culprit
(Clark & Tunnicliff 2001, Lindsay et al. 1994). In actual cases, high resemblance
between the innocent suspect and the culprit can occur by chance or it can occur
whenever the innocent person became a suspect because she or he resembled a
composite or a security video image of the culprit.
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
288 WELLS
¥
OLSON
Lineup Presentation Method
Many alternatives to the traditional lineup have been proposed and tested, and
future research will likely focus on solving the lineup system variable problems.
The first proposed alternative to the traditional lineup was the blank lineup control
method (Wells 1984a). A blank lineup is one that contains only fillers (no suspect).
The eyewitness is first shown the blank lineup under the belief that this is the only
lineup to be shown. The identification of someone froma blank lineup is known to
bean error (because the lineup members are allfillers), and witnesseswho make an
identification from a blank lineup can thereby be discarded. Witnesses who do not
make an identification from the blank lineup can then be shown the actual lineup
(which contains a suspect). Data indicate that eyewitnesses who do not make an
identification from the blank lineup are much more reliable on the second (actual)
lineup than are those who were not screened with the blank lineup method. In
effect, the blank lineup method is analogous to the use of a control condition in
a within-subjects design and could be used in actual cases. In general, however,
crime investigators have not liked the idea of the blank lineup control method on
grounds that it “tricks” the eyewitness and could sever the eyewitness’s trust in
investigators.
Another proposed alternative to the traditional lineup procedure, and the best
known of these alternatives, is the sequential lineup (Lindsay & Wells 1985). Un-
likethe traditional lineup inwhich the lineup membersare shownto theeyewitness
simultaneously, the sequential lineup shows the eyewitness only one lineup mem-
ber at a time and requires the eyewitness to make a decision (“Is this person the
culprit or not?”) prior to viewing the next lineup member. The most powerful ver-
sion of the sequential procedure is one in which the eyewitness does not knowhow
many lineup members are to be viewed. In theory, the sequential lineup procedure
prevents eyewitnesses from selecting the person who looks most like the culprit
relative to the other lineup members, a process called relative-judgment decision
(see above) (Wells 1984a). To the extent that relative judgments are operating,
eyewitnesses will have difficulty with culprit-absent lineups because by definition
someone in the lineup resembles the culprit more closely than the other lineup
members do. Unlike the simultaneous lineup, the sequential lineup prevents eye-
witnesses from making a relative-judgment decision because at any point in the
sequence a lineup member who has not yet been viewed may turn out to resemble
the culprit more than anyperson viewed thus far.Eyewitnessesmust compare each
member of the sequential lineup to their memory of the culprit and thus make a
more “absolute judgment” about identity. A recent meta-analysis of 25 studies
comparing simultaneous and sequential lineups showed that the sequential lineup
reduced the chances of mistaken identifications in culprit-absent lineups by nearly
one half (Steblay et al. 2001). Unfortunately, the sequential technique was also as-
sociated with a reduction in accurate identification rates in culprit-present lineups.
Although this reduction was not as great as that in mistaken identifications, it was
nevertheless statistically reliable. The pattern of these results has led Ebbesen &
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 289
Flowe (2002) to speculate that the sequential lineup raises the criteria for making
a positive identification rather than changing the process from relative to absolute
judgments.
Another alternative to the traditional lineup is the elimination lineup, a proce-
dure in which the witness’s task is to eliminate all but one lineup member and then
make a separate decision as to whether that person is the culprit or not (Pozzulo
& Lindsay 1999). Although the elimination lineup does not seem to work well
with adults, it seems to eliminate some of the problems young children have with
lineups.
Behavioral Influence: The Need for Double-Blind Testing
One of the ways that the justice system itself can influence eyewitness identifica-
tion evidence is through the behaviors of the person who administers the lineup
(Wells 1993). Commonly, the person who administers a lineup is the case de-
tective who, of course, knows which member of the lineup is the suspect and
which members are fillers. The need for double-blind testing is well established
in the behavioral sciences (Rosenthal 1976) but is largely unknown or unheeded
in criminal investigation procedures and forensic science (Risinger et al. 2002).
Lineup administrators could inadvertently communicate their knowledge about
which lineup member is the suspect and which members are merely fillers to the
eyewitness through various verbal and nonverbal means. Phillips et al. (1999) ma-
nipulated lineup administrators’ assumptions about the identity of the culprit and
found that this manipulation affected the choices that eyewitnesses made from the
lineup, especially when a sequential lineup procedure was used. In addition to
influencing eyewitnesses’ choice of particular lineup members, the person admin-
isteringthe lineup cancauseother problems. Wells&Bradfield(1998, 1999) found
that post-identification suggestions to eyewitnesses from lineup administrators led
mistaken eyewitnesses to develop high levels of false certainty that they had made
an accurate identification. The problem of influence from the lineup administrator
is easily fixed by having lineups administered by someone who does not know
which lineup member is the suspect and which ones are fillers (Wells et al. 1998).
Base Rates as System Variables
Base rates can be considered system variables in some cases. The important base
rate in eyewitness identification is the base rate for the culprit being present versus
absent in a lineup. Most mistaken identifications occur when the culprit is not in
the lineup. Although the relation between the culprit-absent versus culprit-present
base rate and the chances of mistaken identification has been established and
modeled mathematically (Wells & Lindsay 1980, Wells & Turtle 1986), the case
was only recently made for treating this base rate as a system variable (Wells
& Olson 2002). Previously, this base rate was treated as a fixed (albeit largely
unknown) variable in actual cases. In fact, however, no laws or rules determine
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
290 WELLS
¥
OLSON
when a suspect is placed in a lineup and, therefore, this base rate varies as a
function of the decisions crime investigators make when conducting a lineup.
Consider, for instance, two police departments, a lax-criterion department and
a strict-criterion department. In the lax-criterion department, investigators will
place a suspect in a lineup for the slightest of reasons (e.g., a mere hunch) whereas
the strict-criterion department requires certain evidence against a person (e.g.,
possession of stolen goods) before placing that person in a lineup. These two
departments will, over the long run, have different base rates for culprit-present
and culprit-absent lineups. Suppose, for example, that over a run of 1000 lineups
the lax-criterion department shows 500 culprit-absent lineups and 500 culprit-
present lineups whereas the strict-criterion department shows only 100 culprit-
absent lineups and 900 culprit-present lineups. Given equivalent eyewitnesses in
both of these departments, mistaken identifications of suspects will be nine times
as likely in the lax-criterion department than in the strict-criterion department.
(These surprising differences in the chances that an identification will be mistaken
are simple derivations from Bayes’ theorem.) Although the justice system has not
yet done so, it could control the culprit-present versus culprit-absent base rate by
requiring “probable cause” before placing someone in a lineup (Wells & Olson
2002). The base rate for culprit-present and culprit-absent lineups might be the
most powerful system variable affecting the chances of mistaken identification.
PROBLEMS AND PROSPECTS
In spite of the successful application of the eyewitness identification literature,
significant work has yet to be done. The eyewitness identification literature has
been driven much less by theoretical frameworks than by practical perspectives.
Two problems are related to this state of affairs. One is that the premium on appli-
cation and forensic relevance reduces the interplay and sharing of ideas between
eyewitness identification researchers and their counterparts in basic areas of psy-
chology, especially cognitive and social psychology. In addition, the experimental
eyewitness identification literature is likely never to be complete enough to cover
every possible situation that arises in actual cases; hence, better theory is needed
to generalize this body of literature and to fill in gaps regarding what is likely to
happen under various conditions.
A second concern is that while laboratory data on eyewitness identification
are extensive, some key forms of real-world data are lacking. Certain estimable
rates of eyewitnessidentification behaviorand lineup conditions from actual cases
could assist the design and interpretation of laboratory work. For instance, there
have been no empirical estimates of the base rate for culprit-present versusculprit-
absent lineups in actual cases. Although it is difficult to establish the ground truth
(actual guilt or innocence) needed for precise estimates of this base rate in actual
cases, methods exist for estimating upper limits (see Wells & Olson 2002). In
addition, although the identification of a suspect from a lineup usually cannot be
definitively classifiedasanaccurate or mistakenidentification in anactualcase, the
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 291
identification of a filler is a known error in actual cases and the rate at which these
known errors occur can be informative. Two estimates of the filler identification
rates in actual cases have been published. Wright & McDaid (1996) reported a rate
of about 20% and Behrman & Davey (2001) reported a rate of 24%. One problem
in collecting filler identification data from real cases is that police records often
do not distinguish between eyewitnesses who make identifications of a filler and
those who make no identification, which can result in a serious underestimation
of the rate of filler identifications (Tollestrup et al. 1994). Another problem is that
filler identification records from actualcases often lack anindication of the level of
eyewitness certainty. These problems can be avoided by scripting data collection
with police departments.
Actual case data of these types (e.g., base rates, filler identification rates, eye-
witness certainty on known errors) can supplement the laboratory literature on
eyewitness identification in two important ways. First, actual case data can be
compared to laboratory data to see if the general rates of certain behaviors (e.g.,
nonidentification responses) are similar. Second, the rates for certain conditions
in actual cases (e.g., rates of culprit-present versus culprit-absent lineups) are crit-
ical for Bayesian estimations of posterior probabilities that cannot themselves be
derived from experiments.
Eyewitness identification research is likely to continue to focus on system vari-
ablesfor the foreseeable future because of the way system variablescan be mapped
onto the problem of improving eyewitness identification accuracy in actual cases.
At the same time, estimator variables might be re-emerging with new promise for
postdictionforthreereasons.First, conditions are being found in which eyewitness
certainty might be more closely related to eyewitness identification accuracy than
oncethought, especially when externalinfluences on eyewitnesscertainty are min-
imized. Second, new postdiction variables, such as decision time, are emerging.
Third, Bayesian analyses are being used to show that some eyewitness responses
to lineups, such as filler identifications, have postdiction value in exonerations.
Each of these represent potentially superior estimator variables because they can
be more precisely measured in actual cases than can some of the more traditional
estimator variables (such as stress or arousal). In any case, there is little evidence
thateyewitness identification researchis veering awayfrom its appliedorientation,
especially in the face of recent successes in affecting legal policies and practices
(Wells 2001, Wells et al. 2000).
The Annual Review of Psychology is online at http://psych.annualreviews.org
LITERATURE CITED
Behrman BW, Davey SL. 2001. Eyewitness
identification in actual criminal cases: an
archival analysis. Law Hum. Behav. 25:475–
91
Bothwell RK, Deffenbacher KA, Brigham JC.
1987.Correlationofeyewitnessaccuracyand
confidence: optimality hypothesis revisited.
J. Appl. Psychol. 72:691–95
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
292 WELLS
¥
OLSON
Bradfield AL, Wells GL, Olson EA. 2002. The
damaging effect of confirming feedback on
therelationbetweeneyewitnesscertaintyand
identification accuracy. J. Appl. Psychol. 87:
112–20
Brewer N, Keast A, Rishworth A. 2002. Im-
proving the confidence-accuracy relation in
eyewitness identification:evidencefrom cor-
relation and calibration. J. Exp. Psychol.:
Appl. 8:44–56
Brigham JC, Barkowitz P. 1978. Do “They all
look alike?” The effect of race, sex, experi-
ence, and attitudes on the ability to recognize
faces. J. Appl. Soc. Psychol. 8:306–18
Brigham JC, Bothwell RK. 1983. The ability of
prospective jurors to estimate the accuracyof
eyewitness identifications.Law Hum. Behav.
7:19–30
Brown E, Deffenbacher K, Sturgill W. 1977.
Memory for faces and the circumstances
of the encounter. J. Appl. Psychol. 62:311–
18
Christianson SA. 1992. Emotional stress and
eyewitness memory: a critical review. Psy-
chol. Bull. 112:284–309
Clark SE. 2002. A memory and decision model
for eyewitness identification. Appl. Cogn.
Psychol. In press
Clark SE, Tunnicliff JL. 2001. Selecting lineup
foilsineyewitness identification: experimen-
tal control and real-world simulation. Law
Hum. Behav. 25:199–216
Clifford BR, Hollin CR. 1981. Effects of the
type of incident and the number of perpetra-
torson eyewitness memory. J.Appl. Psychol.
66:364–70
Cutler BL, Penrod SD, Martens TK. 1987. The
reliability of eyewitness identification: the
role of system and estimator variables. Law
Hum. Behav. 11:233–58
DeffenbacherK. 1983. The influence of arousal
on reliability of testimony. In Evaluating
Witness Evidence: Recent Psychological Re-
search and New Perspectives, ed. SMA
Lloyd-Bostock, BR Clifford, pp. 235–51.
Chichester, Engl.: Wiley
Deffenbacher K, Loftus EF. 1982. Do jurors
share a common understanding concern-
ing eyewitness behavior? Law Hum. Behav.
6:15–30
Dunning D, Perretta S. 2002. Automaticity and
eyewitness accuracy: a 10- to-12 second rule
for distinguishing accurate from inaccurate
positive identifications. J. Appl. Psychol. In
press
Dunning D, Stern LB. 1994. Distinguishing ac-
curate from inaccurate identifications via in-
quiriesabout decisionprocesses.J.Pers.Soc.
Psychol. 67:818–35
Easterbrook JA. 1959. The effectof emotion on
cueutilizationand the organizationof behav-
ior. Psychol. Rev. 66:183–201
EbbesenE,FloweS.2002.Simultaneousversus
sequential lineups: what do we really know?
Law Hum. Behav. In press
Ellis HD, Davies GM, Shepherd JW. 1977. Ex-
perimental studies of face identification. J.
Crim. Def. 3:219–34
Fleishman JJ, Buckley ML, Klosinsky MJ,
SmithN,TuckB.1976.Judgedattractiveness
in recognition memory of women’s faces.
Percept. Mot. Skills 43:709–10
Hockley WE, Hemsworth DH, Consoli A.
1999. Shades of the mirror effect: recogni-
tion of faces with and without sunglasses.
Mem. Cogn. 27(1):128–38
HoschHM,LeippeMR,MarchioniPM,Cooper
DS. 1984. Victimization, self-monitoring,
and eyewitness identification. J. Appl. Psy-
chol. 69:280–88
HoschHM, Platz SJ.1984.Self-monitoringand
eyewitness accuracy.Personal.Soc. Psychol.
Bull. 10:283–89
Howells TH. 1938. A study of ability to recog-
nize faces. J. Abnorm. Psychol. 33:124–27
JuslinP,Olson N, WinmanA.1996.Calibration
anddiagnosticityofconfidenceineyewitness
identification: comments on what can and
cannot be inferred from a low confidence-
accuracy correlation. J. Exp.Psychol. Learn.
Mem. Cogn. 5:1304–16
Langois JH, Roggman LA. 1990. Attractive
faces are only average. Psychol. Sci. 1:115–
21
Leippe MR, Wells GL, Ostrom TM. 1978.
Crime seriousness as a determinant of
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 293
accuracy in eyewitness identification. J.
Appl. Psychol. 63:345–51
Light LL, Kayra-Stuart F, Hollander S. 1979.
Recognition memory for typical and unusual
faces. J. Exp. Psychol.: Hum. Learn. 5:212–
28
Lindsay RCL, Martin R, Webber L. 1994. De-
fault values in eyewitness descriptions: a
problem for the match-to-description lineup
foil selection strategy. Law Hum. Behav. 18:
527–41
Lindsay RCL, Wells GL. 1980. What price
justice? Exploring the relationship between
lineup fairness and identification accuracy.
Law Hum. Behav. 4:303–14
Lindsay RCL, Wells GL. 1985. Improving eye-
witness identification from lineups: simulta-
neousversussequential lineup presentations.
J. Appl. Psychol. 70:556–64
Lindsay RCL, Wells GL, O’Connor F. 1989.
Mock juror belief of accurate and inaccurate
eyewitnesses: a replication. Law Hum. Be-
hav. 13:333–40
Lindsay RCL, Wells GL, Rumpel C. 1981. Can
people detect eyewitness identification accu-
racy within and between situations? J. Appl.
Psychol. 66:79–89
Loftus EF, Loftus GR, Messo J. 1987. Some
facts about “weapon focus.” Law Hum. Be-
hav. 11:55–62
Luus CAE, Wells GL. 1991. Eyewitness iden-
tification and the selection of distracters
for lineups. Law Hum. Behav. 15:43–
57
MalpassRS. 1981. Effective size and defendant
biasineyewitnessidentificationlineups.Law
Hum. Behav. 5:299–309
Malpass RS, Devine PG. 1981. Eyewitness
identification: lineup instructions and the ab-
sence of the offender. J. Appl. Psychol. 66:
482–89
McConkey KM, Roche SM. 1989. Knowledge
of eyewitness memory. Aust. Psychol. 24:
377–84
Meissner C, Brigham JC. 2001. Twenty years
ofinvestigating the own-racebias in memory
for faces: a meta-analytic review. Psychol.
Public Policy Law 7:3–35
Navon D. 1992. Selection of foils by similar-
ity to suspect is likely to misfire. Law Hum.
Behav. 16:575–93
Noon E, Hollin CR. 1987. Lay knowledge of
eyewitness behaviour: a British survey. Appl.
Cogn. Psychol. 1:143–53
Phillips MR, McAuliff BD, Kovera MB, Cutler
BL. 1999. Double-blind photoarray admin-
istration as a safeguard against investigator
bias. J. Appl. Psychol. 84:940–51
Powers PA, Andriks JL, Loftus EF. 1979. Eye-
witness accounts of males and females. J.
Appl. Psychol. 64:339–47
Pozzulo JD, Lindsay RCL. 1998. Identification
accuracy of children versus adults: a meta-
analysis. Law Hum. Behav. 22:549–70
Pozzulo JD, Lindsay RCL. 1999. Elimination
lineups: an improved identification proce-
durefor childeyewitnesses.J.Appl. Psychol.
84:167–76
Read JD, Vokey JR, Hammersley R. 1990.
Changingphotosoffaces:effectsofexposure
duration and photo similarity on recognition
and the accuracy-confidence relationship. J.
Exp. Psychol.: Learn. Mem. Cogn. 16:870–
82
RisingerMD,SaksMJ, Thompson WC, Rosen-
thal R. 2002. The Daubert/Kumho implica-
tions of observer effects in forensic science:
hidden problems of expectation and sugges-
tion. Calif. Law Rev. 90:1–56
Robinson MD, Johnson JT, Herndon F. 1997.
Reaction time and assessments of cognitive
effort as predictors of eyewitness memory
accuracy and confidence. J. Appl. Psychol.
82:416–25
Rosenthal R. 1976. Experimenter Effects in Be-
havioral Research. New York: Irvington
Ross DF, Read JD, Toglia MP, eds. 1994. Adult
Eyewitness Testimony: Current Trends and
Developments. New York: Cambridge Univ.
Press
Scheck B, Neufeld P, Dwyer J. 2000. Actual
Innocence. New York: Random House
Seelau SM, Wells GL. 1995. Applied eyewit-
ness research: the other mission. Law Hum.
Behav. 19:317–22
Shapiro PN, Penrod SD. 1986. Meta-analysis
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
294 WELLS
¥
OLSON
of racial identification studies. Psychol. Bull.
100:139–56
Shaw JI, Skolnick P. 1999. Weapon focus and
gender differences in eyewitness accuracy:
arousal versus salience. J. Appl. Soc. Psy-
chol. 29:2328–41
Shaw JS III. 1996. Increasesin eyewitnesscon-
fidenceresultingfromposteventquestioning.
J. Exp. Psychol.: Appl. 2:126–46
Shaw JS III, McClure KA. 1996. Repeated
postevent questioning can lead to elevated
levels of eyewitness confidence. Law Hum.
Behav. 20:629–54
Smith SM, Lindsay RCL, Pryke S. 2000. Post-
dictors of eyewitness errors: Can false iden-
tification be diagnosed? J. Appl. Psychol.
85:542–50
Sporer SL. 1992. Post-dicting eyewitness accu-
racy: confidence, decision times and person
descriptions of choosers and non-choosers.
Eur. J. Soc. Psychol. 22:157–80
Sporer SL. 1993. Eyewitness identification ac-
curacy, confidence, and decision times in si-
multaneous and sequential lineups. J. Appl.
Psychol. 78:22–33
Sporer SL. 1994. Decision times and eyewit-
ness identification accuracy in simultaneous
and sequential lineups. See Ross et al. 1994,
pp. 300–27
Sporer SL, Penrod SD, Read JD, Cutler BL.
1995. Choosing, confidence, and accuracy:
a meta-analysis of the confidence-accuracy
relation in eyewitness identification studies.
Psychol. Bull. 118:315–27
Steblay NM. 1992. A meta-analytic review of
the weapon focus effect. Law Hum. Behav.
16:413–24
Steblay NM. 1997. Social influence in eye-
witness recall: a meta-analytic review of
lineup instruction effects. Law Hum. Behav.
21:283–98
Steblay NM, Dysart J, Fulero S, Lindsay RCL.
2001. Eyewitness accuracy rates in sequen-
tial and simultaneous lineup presentations: a
meta-analyticcomparison. Law Hum. Behav.
25:459–74
Tech.Work.GroupEyewitnessEvidence.1999.
Eyewitness Evidence: A Guide for Law En-
forcement. Washington, DC: US Dep. Jus-
tice, Off. Justice Programs
Tollestrup PA, Turtle JW, Yuille JC. 1994.
Actual witnesses to robbery and fraud: an
archival analysis. See Ross et al. 1994, pp.
144–62
Wells GL. 1978. Applied eyewitness testi-
mony research: system variables and estima-
tor variables.J.Pers. Soc. Psychol. 36:1546–
57
Wells GL. 1984a. The psychology of lineup
identifications. J. Appl. Soc. Psychol. 14:89–
103
Wells GL. 1984b. How adequate is human in-
tuition for judging eyewitness testimony? In
Eyewitness Testimony: Psychological Per-
spectives, ed. GL Wells, EF Loftus, pp. 256–
72. New York: Cambridge Univ. Press
Wells GL. 1993. What do we know about eye-
witnessidentification?Am.Psychol.48:553–
71
Wells GL. 2001. Police lineups: data, the-
ory, and policy. Psychol. Public Policy Law
7:791–801
Wells GL, Bradfield AL. 1998. “Good, you
identified the suspect:” Feedback to eyewit-
nesses distorts their reports of the witnessing
experience. J. Appl. Psychol. 83:360–76
Wells GL, Bradfield AL. 1999. Distortions in
eyewitnesses’ recollections: Can the post-
identification feedback effect be moderated?
Psychol. Sci. 10:138–44
Wells GL, Ferguson TJ, Lindsay RCL. 1981.
Thetractabilityof eyewitnessconfidenceand
its implication for triers of fact. J. Appl. Psy-
chol. 66:688–96
Wells GL, Hryciw B. 1984. Memory for faces:
encoding and retrieval operations. Mem.
Cogn. 12:338–44
Wells GL, Leippe MR. 1981. How do triers of
fact infer the accuracy of eyewitness identi-
fications? Memory for peripheral detail can
be misleading. J. Appl. Psychol. 66:682–
87
Wells GL, Leippe MR, Ostrom TM. 1979.
Guidelines for empirically assessing the fair-
ness of a lineup. Law Hum. Behav. 3:285–
93
15 Nov 2002 17:30 AR AR178-PS54-11.tex AR178-PS54-11.SGM LaTeX2e(2002/01/18) P1: IBC
EYEWITNESS IDENTIFICATION 295
Wells GL, Lindsay RCL. 1980. On estimating
the diagnosticity of eyewitness nonidentifi-
cations. Psychol. Bull. 88:776–84
Wells GL, Lindsay RCL, Ferguson TJ. 1979.
Accuracy, confidence, and juror perceptions
ineyewitnessidentification.J.Appl.Psychol.
64:440–48
Wells GL, Luus CAE. 1990. Police lineups as
experiments:Socialmethodologyas aframe-
work for properly conducted lineups. Per-
sonal. Soc. Psychol. Bull. 16:106–17
Wells GL, Malpass RS, Lindsay RCL, Fisher
RP,TurtleJW,FuleroS.2000.Fromthelabto
thepolicestation:Asuccessful applicationof
eyewitness research. Am. Psychol. 55:581–
98
WellsGL,OlsonEA.2002.Eyewitnessidentifi-
cation: Information gain from incriminating
and exonerating behaviors. J. Exp. Psychol.:
Appl. In press
Wells GL, Rydell SM, Seelau EP. 1993. On the
selection of distractors for eyewitness line-
ups. J. Appl. Psychol. 78:835–44
Wells GL, Seelau EP, Rydell SM, Luus CAE.
1994. Recommendations for properly con-
ducted lineup identification tasks. See Ross
et al. 1994, pp. 223–44
Wells GL, Small M, Penrod SD, Malpass
RS, Fulero SM, Brimacombe CAE. 1998.
Eyewitness identification procedures: rec-
ommendations for lineups and photospreads.
Law Hum. Behav. 22:603–7
Wells GL, Turtle JW. 1986. Eyewitness iden-
tification: the importance of lineup models.
Psychol. Bull. 99:320–29
WogalterMS,Marwitz DB, Leonard DC. 1992.
Suggestiveness inphotospread lineups: Sim-
ilarity induces distinctiveness. Appl. Cogn.
Psychol. 6:443–53
Wright DB, McDaid AT. 1996. Comparing sys-
tem and estimator variables using data from
real lineups. Appl. Cogn. Psychol. 10:75–84
