Improving Its Probative Value
Gary L. Wells,1Amina Memon,2and Steven D. Penrod3
1Iowa State University;2University of Aberdeen, Aberdeen, Scotland; and3John Jay College of Criminal Justice
SUMMARY—The criminal justice system relies heavily on
eyewitnesses to determine the facts surrounding criminal
events. Eyewitnesses may identify culprits, recall conver-
sations, or remember other details. An eyewitness who has
no motive to lie is a powerful form of evidence for jurors,
especially if the eyewitness appears to be highly confident
about his or her recollection. In the absence of definitive
proof to the contrary, the eyewitness’s account is generally
accepted by police, prosecutors, judges, and juries.
However, the faith the legal system places in eyewit-
nesses has been shaken recently by the advent of forensic
DNA testing. Given the right set of circumstances, forensic
DNA testing can prove that a person who was convicted of
a crime is, in fact, innocent. Analyses of DNA exoneration
cases since 1992 reveal that mistaken eyewitness identifi-
cation was involved in the vast majority of these convic-
tions, accounting for more convictions of innocent people
than all other factors combined. We review the latest fig-
ures on these DNA exonerations and explain why these
cases can only be a small fraction of the mistaken identi-
fications that are occurring.
Decades before the advent of forensic DNA testing,
psychologists were questioning the validity of eyewitness
reports. Hugo Mu ¨nsterberg’s writings in the early part of
the 20th century made a strong case for the involvement
of psychological science in helping the legal system
understand the vagaries of eyewitness testimony. But it
was not until the mid- to late 1970s that psychologists be-
gan to conduct programmatic experiments aimed at
understanding the extent of error and the variables that
govern error when eyewitnesses give accounts of crimes
by psychologists that contained strong warnings to the le-
gal system that eyewitness evidence was being overvalued
by the justice system in the sense that its impact on triers
of fact (e.g., juries) exceeded its probative (legal-proof)
value. Another message of the research was that the
validity of eyewitness reports depends a great deal on the
procedures that are used to obtain those reports and that
the legal system was not using the best procedures.
Although defense attorneys seized on this nascent
research as a tool for the defense, it was largely ignored
or ridiculed by prosecutors, judges, and police until the
mid 1990s, when forensic DNA testing began to uncover
cases of convictions of innocent persons on the basis of
mistaken eyewitness accounts. Recently, a number of
jurisdictions in the United States have implemented
procedural reforms based on psychological research, but
psychological science has yet to have its fullest possible
influence on how the justice system collects and interprets
The psychological processes leading to eyewitness error
represent a confluence of memory and social-influence
variables that interact in complex ways. These processes
lend themselves to study using experimental methods.
Psychological science is in a strong position to help the
criminal justice system understand eyewitness accounts of
variables that affect eyewitness accuracy fall into what
researchers call system variables, which are variables
that the criminal justice system has control over, such as
how eyewitnesses are instructed before they view a
lineup and methods of interviewing eyewitnesses. We
review a number of system variables and describe how
psychological scientists have translated them into pro-
accounts. We also review estimator variables, variables
that affect eyewitness accuracy but over which the system
has no control, such as cross-race versus within-race
We describe some concerns regarding external validity
and generalization that naturally arise when moving from
the laboratory to the real world. These include issues of
base rates, multicollinearity, selection effects, subject
populations, and psychological realism. For each of these
concerns, we briefly note ways in which both theory and
field data help make the case for generalization.
Direct correspondence to Gary L. Wells, Psychology Department,
Iowa State University, Ames, IA 50011; e-mail: firstname.lastname@example.org.
PSYCHOLOGICAL SCIENCE IN THE PUBLIC INTEREST
Volume 7—Number 2
Copyright r 2006 Association for Psychological Science
he was convictedin March 1985 for the 1984 sexualassault and
slaying of a 9-year-old girl in Maryland (State of Maryland v.
Kirk N. Bloodsworth, 1984). Five eyewitnesses identified
Bloodsworth at trial. Later that month, a judge sentenced him to
death. He spent 2 years on death row before he received a new
other suspects. This time he received a life sentence. Bloods-
not until 1993 that he was released from prison on the basis of
the little girl’s underwear. Bloodsworth was lucky that the
underwear had been preserved, becauseearlier(pre-DNA)tests
had indicated nothing of value on the underwear. But what kind
of luck is being convicted of a murder you did not commit? His
he was innocent. And despite his release from prison, some
people, including one of the original prosecutors, continued to
believe that Bloodsworth may have been the murderer. The
really was the murderer, they reasoned, and the tiny speck of
semen came from someone other than the murderer—perhaps
someone who had access to the little girl’s dresser drawer, for
instance.Bloodsworth went onwith hislife,confidentinhisown
innocencebuthaving tolivewiththeoccasionaldoubtraised by
2003, DNA testing got a hit on the actual murderer, Kimberly
Shay Ruffner. Nineteen years after Kirk Bloodsworth was sen-
tenced to death, the proof was finally there: He had had nothing
to do with the sexual assault and slaying of the young girl.
The case of Kirk Bloodsworth illustrates several problems
with eyewitness evidence. First, it illustrates the fallacy of as-
suming that inter-witness agreement is necessarily strong evi-
dence of accuracy. Many factors can lead to inter-witness
agreement, such as interaction among the witnesses in which
they share information. In general, factors that lead one eye-
witness to make a particular error will lead others to make the
same error. Second, the Bloodsworth case illustrates the pro-
found level of proof required for exonerating evidence to trump
eyewitness identification evidence. Even when the semen was
proved not to match Bloodsworth’s DNA, many people were
unwilling to believe he was innocent. It was necessary to prove
that someone else had committed the murder. Third, the
Bloodsworth case illustrates that mistaken identification is a
dual problem: Not only might an innocent person be convicted
but the guilty party remains free to reoffend.
The role of scientific psychology in the problem of eyewitness
evidence is a profound one. With few exceptions, the legal
system has not conducted research on eyewitness evidence, has
theory regarding how memory works. The scientific study of
eyewitnesses ispurely the domain of psychology.When the U.S.
Department of Justice finally wrote guidelines on eyewitness
evidence in 1999, the only scientific studies cited were those
published by psychologists in psychology journals. Today, psy-
of eyewitness reports. The credibility of scientific psychology
has risen immensely in the legal system recently, largely be-
cause psychologists were already ‘‘blowing the whistle’’ on
eyewitness evidence well before forensic DNA testing began
uncovering mistaken identifications in the 1990s. In effect,
psychologists were able to use experiments to identify eyewit-
ness problems long before the legal system was smacked in the
face with DNA exonerations.
A primary purpose of this article is to describe empirical
evidence supporting the proposition that some of the problems
with eyewitness evidence can be addressed by improving the
way the evidence is collected and preserved. We discuss how
eyewitnesses are interviewed, how lineups are conducted, and
why procedures can have a strong impact on the resulting pro-
bative value of eyewitness testimony. These variables are called
system variables, because they are under the control of the
justice system (Wells, 1978). The importance of system vari-
ables that can reduce eyewitness error has become increasingly
apparent in light of the proven inadequacies of traditional
safeguards against eyewitness mistakes,such as the presence of
counsel at lineups and the opportunity to present motions to
suppress suggestive procedures (Stinson, Devenport, Cutler, &
Kravitz, 1996, 1997). But even if the system reaches a point at
which it makes perfect use of system variables, eyewitness
errors attributable to other factors will remain. Thus, it is
important to review these other (non-system-controlled) factors
eyewitness literature. Instead, we focus on practices, proced-
ures, and research that address the most common threats to
eyewitness reliability. Although the bulk of the scientific and
legal literature we cite has a North American origin, the inter-
national research community has made extremely important
contributions. In fact, historically, it was Europeans who played
the much greater role in the study of eyewitness memory.1
We begin with a brief history of psychology’s attempt to help
the legal system on the eyewitness issue. Then we describe the
role these exonerations have played in giving scientific psy-
chology a stronger voice in the legal system’s policies and pro-
cedures involving eyewitness evidence. We then give an
overview of the standard methods used in eyewitness research,
followed byselected findingsonestimatorandsystemvariables.
1We are fortunate to have Siegfried Sporer, a strong European contributor to
the empirical literature on eyewitness issues, write the editorial preceding this
monograph (see p. i). Sporer places our report in a broader historical and in-
Volume 7—Number 2
A BRIEF HISTORY OF EYEWITNESS PSYCHOLOGY
In his book La Suggestibilite ´, Alfred Binet (1900) argued for the
creation of a practical science of testimony based on his ob-
servations about the effects of suggestion. Binet was the first to
report that suggestive questioning influenced responses. But it
how eyewitnesses were questioned makes a great deal of dif-
ference.Louis William Sternwaspublishingandeditingstudies
of eyewitness testimony as early as 1904 (Stern, 1904). In the
United States, Guy Montrose Whipple published a number of
articles in Psychological Bulletin on eyewitness testimony
(Whipple, 1909, 1910, 1911, 1912). But it was Hugo Mu ¨nster-
berg’s (1908) book On the Witness Stand and his injection of
himself into the legal system that had a more lasting impact in
the United States.
Mu ¨nsterberg wasrecruited byWilliamJamesin1892tocome
to Harvard to run the university’s psychological laboratory.
Mu ¨nsterberg was very much a public figure and he appeared
frequently in the popular press. He also was a somewhat con-
troversial figure at Harvard, presumably because his colleagues
did not see a great deal of merit in applying psychology. His
lectures and writings were extremely perceptive and well rea-
soned, albeit rather short on data by modern standards. His
prescience is evident in such matters as his claim that eyewit-
ness certainty has a tenuous relation to accuracy and that while
jurors might understand forgetting, they are not likely to
understand that a witness can remember the wrong thing.
Although Mu ¨nsterberg maintained a certain prominence in
psychology, his impact on the legal system was muted dramat-
ically by the skilled counterargumentation of one of the greatest
minds in American jurisprudence, John Henry Wigmore. Par-
ticularly problematic for Mu ¨nsterberg was a law review article
by Wigmore (1909) that challenged Mu ¨nsterberg’s (1908)
overstatements about the ability of psychology to help the legal
system. Wigmore was especially effective in arguing that psy-
chology did not yet have ready tools for handling the problem of
evaluating eyewitness accounts, as Mu ¨nsterberg had claimed.
For the most part, Wigmore won the argument, at least from the
perspective of the legal system.
to 1960s. Some important work was done in the 1930s by Burtt
(1931) and Stern (1939). The 1940s produced some important
studies by Snee and Lush (1941) on question effects and by
Allport and Postman (1947) on person-to-person information
the effects of personal prejudice on perception in the 1950s,
there was little discussion of the relevance of this to the legal
system and to eyewitnesses in general. There are differing
overgeneralizations by psychologists that failed to meet the
needs and standards of the courtroom.
The Modern Era of Eyewitness Research
More than any other individual’s work, it was Elizabeth Loftus’s
elegant experiments on postevent information that gave rise to
the modern era of eyewitness research. Loftus managed to show
in their natural settings, could be used in rigorous scientific
had practical utility for understanding eyewitness error. By
publishing her work in prestigious scientific psychology jour-
nals in the mid- and late 1970s—journals such as Cognitive
Journal of Experimental Psychology: Human Learning and
Memory—Loftus legitimized the study of eyewitnesses in the
minds of psychological scientists. Her book Eyewitness Testi-
mony (Loftus, 1979) remains one of the best known psychology
books almost three decades after it was released. Like
Mu ¨nsterberg, Loftus was criticized for some of her claims (e.g.,
McCloskey & Egeth, 1983), but, unlike Mu ¨nsterberg, she
helped spawn a new generation of researchers who have care-
fully and strategically builtan empirical literature that the legal
system must contend with.
While Loftus was focusing on memory for events and the
malleability of memory, Robert Buckhout at Brooklyn College
was focusing on memory for people. Buckhout was more con-
cerned with mistaken identification from lineups than with
memory for objects. Although Buckhout wrote a highly visible
article in Scientific American reviewing research on eyewitness
reliability (Buckhout, 1974), he was not otherwise particularly
journals. He did, however, create his own ‘‘in house’’ outlet
called Social Action and the Law. Buckhout often used dramatic
means to get his point across. For example, he got a New York
a six-person lineup. Of the 2,145 viewers who called in, nearly
his experiments in better journals but chose not to spend the
time and effort required to go through the rigorous review pro-
cess.Still, Buckhout influenced many younger researchers, who
took up the issue of mistaken identification. At about the same
time, eyewitness research activity was growing in the United
Kingdom, prompted by the investigation of the Devlin Com-
mittee (Devlin, 1976; see also Bull & Clifford, 1976; Clifford &
Bull, 1978; Davies, Ellis, & Shepherd, 1978; Ellis, Davies, &
One of the organizing themes that emerged from the 1970s
was the distinction between system variables and estimator
variables (Wells 1978). The argument was that some of the
variables that affect the accuracy of eyewitness reports were
under the control (or potentially under the control) of the justice
system (system variables) while others were not (estimator
variables). For example, how eyewitnesses are interviewed by
police and how eyewitnesses are instructed prior to viewing a
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
lineup are system variables, because they can be controlled by
the system that is collecting the eyewitness evidence. Other
or stress experienced by the witness during the event—cannot
can be controlled and manipulated in experiments, but only
system variables can be controlled in actual cases. Variables
can be controlled in experiments) are called estimator variables
because the best that eyewitness psychology can do is help es-
timate their impact in a given case.
addressed the primary argument that Wigmore used in his
devastating criticism of Mu ¨nsterberg—namely, that psychology
had no practical recommendations for dealing with the eyewit-
ness problem. Loftus’s main findings fit nicely into the system-
variable framework. For instance, if certain types of questions
(leading questions) result in eyewitnesses incorporating infor-
mation into their later reports regarding matters they did not
witness, then psychology could devise practical ways to avoid
this problem. Likewise, if certain instructions to eyewitnesses
prior to viewing a lineup reduce the chances of mistaken iden-
tification, then psychology could advise on the best ways to in-
Throughout the 1970s and 1980s, eyewitness research
was largely ignored by the criminal justice system. The big
exception was criminal defense lawyers. Defense lawyers were
quick to recognize the potential for psychology to help them
convince juries that eyewitness memory was not to be trusted,
and they saw expert testimony as the mechanism to do this. The
battle topermit expert testimony on eyewitness issues, however,
was and is a contentious one. Expert testimony has been both
permitted and denied in nearly every state in the United States,
depending on the discretion of the trial judge. Prosecutors
generally use four arguments against the admission of expert
testimony on eyewitness issues. One argument is that the eye-
witness literature is not sufficiently mature or precise to be
considered scientific. Today, this argument almost never pre-
vails. However, the three other arguments continue to prevent
expert testimony on eyewitness issues in many jurisdictions.
One is that such testimony invades the province of the jury,
because it is the jury that must decide the credibility of wit-
of common sense and that juries already know these things from
their everyday experience. Yet another argument is that the
prejudicial value of expert testimony regarding eyewitnesses
outweighs its probative value. This argument assumes that
eyewitness experts can make juries more dubious of the eye-
witness than they ought tobe. Itis not the purpose of the current
monograph to argue the merits of expert testimony. We simply
note that expert testimony for the defense was, until recently,
virtually the only way the legal system acknowledged the
scientific study of eyewitnesses.
Forensic DNA Testing: An Awakening of the Legal System
Much has changed in the past few years, but not because of any
Rather, the advent of forensic DNA testing has changed the way
the legal system views eyewitness evidence. Previous studies of
the conviction of innocent people had shown that mistaken
eyewitness identification was implicated in the majority of
1957; Huff, Rattner, & Sagarin, 1986). But it was the develop-
ment of forensic DNA testing in the 1990s that permitted de-
finitive cases of the conviction of innocent people in the United
Neufeld, cofounders of the Innocence Project in New York City,
took the lead and are still the central figures in facilitating the
use of forensic DNA to test claims of innocence by people who
were convicted by juries. Scheck and Neufeld were quick to see
the pattern: Eyewitness-identification error was at the heart of
the evidence used to convict the vast majority of these innocent
people. Press accounts of these exonerations caught the atten-
tion of U.S. Attorney General Janet Reno, and an early report
commissioned by Reno revealed that 26 of the first 28 exoner-
ations were cases of mistaken eyewitness identification (Con-
that 36 of the first 40 DNA exonerations were mistaken-identi-
fication cases (Wells, Small, Penrod, Malpass, Fulero, &
Brimacombe, 1998). Scheck, Neufeld, and Dwyer (2000) re-
ported that 52 of the first 62 DNA exonerations were mistaken-
identification cases. As of this writing, there have been more
than 180 definitive DNA exonerations; the proportion that in-
75% or more. The Innocence Project in New York maintains an
up-to-date Web site, www.innocenceproject.org, that catalogues
these DNA exonerations, and there are now innocence pro-
jects worldwide (http://forejustice.org/wc/wrongful_conviction_
the results of eyewitness experiments in psychology were mere
academic exercises, games played with people’s memories that
would not apply to real witnesses and real crimes. At the very
least, the DNA exonerations have proved that eyewitnesses can
be absolutely positive and yet absolutely mistaken, just as was
found in the experiments. But do 180-plus cases of mistaken
identification prove anything? If these cases were the total, then
it might be argued that this is a rather small fraction of con-
involved murder, robbery, and other offenses, but sexual assault
is the common feature.
It is not that sexual assault witnesses are especially poor
eyewitnesses. In fact, they might be the very best at identifying
their attackers, because they tend to get longer, closer views of
Volume 7—Number 2
casesforwhichbiologicallyrichDNAtraces were left behind by
the perpetrator in the form of semen. (In 2004, nearly 95,000
sexual assaults were reported, with a 43% clearance rate. For
violent_crime/index.html.) Stranger-rape cases, in which iden-
about 30,000 cases of stranger-rape come to the attention of the
police each year.In contrasttosexualassault cases,only a small
fraction of murders (more than 16,000 reported in 2004) and
almost no robberies (more than 400,000 reported in 2004) or
aggravated assaults (more than 850,000 reported in 2004) result
in biologically rich trace evidence being left behind. What can
drive-by shooting use to prove that the eyewitness identification
was mistaken? Thus, these 180-plus DNA exonerations repre-
sent a small proportion of the crimes for which eyewitness
identification evidence has been used to convict people. Fur-
thermore, only a fraction of old sexual assault convictions can
now be tested, because the evidence was never collected, was
destroyed. All in all, the 180 (and growing) DNA exonerations
canonly bea small fractionofthe totalnumberofcasesinwhich
people have been convicted because they were mistakenly
identified by eyewitnesses.
We will not venture an estimate of the number of people in
prison who are innocent victims of mistaken eyewitness iden-
able to do to help prevent these mistakes from occurring in the
future. This is where we have seen some promising progress
recently. Janet Reno’s appointment of a working group to de-
because the group included five eyewitness researchers. Reno
system both in recognizing the eyewitness problem and in de-
veloping solutions for it. An account of this process, which
yielded the first set of U.S. national guidelines on eyewitness
evidence, has been published elsewhere (Wells, Malpass,
Lindsay, Fisher, Turtle, & Fulero, 2000). Since the publication
recommendations and have gone well beyond the guide to in-
clude procedural changes recommended by eyewitness scien-
tists. These jurisdictions include the states of New Jersey, North
Carolina, and Wisconsin, as well as the cities of Boston and
Minneapolis, among others (Wells, 2006).
Despite these encouraging reforms, it is estimated that
only about 10% of the U.S. population reside in reformed
jurisdictions (Wells, 2006). Will these system-variable im-
provements continue by increasing numbers of jurisdictions
in the years to come? Only time will tell. In the following sec-
and we note how the eyewitness-research area must continue to
develop to ensure that the evolving relationship between the
legal system and psychological science will be a fruitful and
COMMON METHODS USED IN EYEWITNESS
The experimental method has dominated the eyewitness liter-
ature, and most of the experiments are lab based. Lab-based
experimental methods for studying eyewitness issues have
methods is that they are proficient at establishing cause–effect
relations. This is especially important for research on system
variables, because one needs to know in fact whether a par-
ticularsystem manipulationis expected tocause better orworse
performance. In the real world, many variables can operate at
the same time and in interaction with one another. Multicol-
linearity can be quite a problem in archival/field research,
because it can be very difficult to sort out which (correlated)
of variables that is possible in experimental research can
research. For example, experiments on stress during witnessing
have shown, quite compellingly, that stress interferes with the
ability of eyewitnesses to identify a central person in a stressful
& McGorty, 2004). However, when Yuille and Cutshall (1986)
those who reported higher stress had better memories for details
than did those who reported lower stress. Why the different re-
sults? In the experimental setting, stresswas manipulated while
other factors were held constant; in the actual shooting, those
who were closer to the incident reported higher levels of stress
(presumably because of their proximity) but also had a better
view. Thus, in the actual case, stress and view covaried.
estimator variables—that is, there may be limits to generalizing
from experiments to actual cases. One reason is that levels of
estimator variables in experiments are fixed and not necessarily
fully representative of the values observed in actual cases. In
addition, it is not possible to include all interesting and plau-
sible interactions among variables in any single experiment
(or even in a modest number of experiments). Clearly, general-
izations to actual cases are best undertaken on the basis of a
substantial body of experimental research conducted across a
wide variety of conditions and employing a wide variety of
variables. Nevertheless, the literature is largely based on ex-
learn about cause and effect. Furthermore, ‘‘ground truth’’ (the
actual facts of the witnessed event) is readily established in
experiments, because the witnessed events are creations of the
experimenters. That kind of ground truth is difficult, if not im-
possible, to establish when analyzing actual cases.
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
The ecological validity of witnessed events (when examined at
the surface level) varies considerably across experiments. Some
eyewitness experiments simply show slides to participant wit-
have been elaborate ruses in which calls are made to ‘‘police’’
(actually confederates of the experimenter) and participants are
shown lineupswhilestillbelievingthatwhattheywitnessed was
real (e.g., Luus & Wells, 1994; Wells & Murray, 1983). Perhaps
the most common witnessed event used by researchers is the
crimes has undoubtedly made them less common in the litera-
ture in recent years, but the success of video crime experiments
method manages to capture the elements that are important for
process is incidental in the sense that the participant witnesses
do not know when they watch the video that the study concerns
eyewitness memory. Instead, researchers commonly tell them
that they are going to have to form impressions or make judg-
ments about the people or scenes. Only later are they informed
that the study concerns eyewitness memory.
In lineup experiments, the participant witnesses are usually
tested with photo lineups rather than with live lineups. Again,
the savings in cost and time are factors, but the use of photo
lineups in experimentsparallels their usein actual cases. Inthe
United Kingdom, there has been a move toward the use of video
lineups (Pike, Kemp, Towell, & Phillips, 1997; Valentine &
Heaton, 1999). Although some jurisdictions (such as New York)
still use live lineups, most jurisdictions in the United States use
often than not they are preceded by a photo lineup, and the live
lineup is merely a confirmatory tool. Thus, the prevalence of
photo lineups in experiments reflects what is happening in ac-
tual criminal investigations.
It is standard practice in experiments to use lineups in which
the actual perpetrator is present in the lineup for some partici-
pant witnesses and not present for others. The not-present
lineups (target-absent or perpetrator-absent lineups) are crit-
ically important for eyewitness-identification studies that are
designed to examine accuracy. Target-absent lineups simulate
the real-world situation in which police have focused their
suspicion on an innocent suspect. The standard procedure in
lineup experiments is to create a target-absent lineup by re-
placing the target with another person who fits the target’s de-
scription and leaving the fillers (the innocent distracters or foils
in the lineup) the same.
Participant witnesses in experiments typically take the per-
spective of a bystander rather than a victim. However, some
experiments have examined possible differences between by-
stander eyewitnesses and victim eyewitnesses and have found
no significant differences (Hosch & Cooper, 1982; Hosch,
Leippe, Marchioni, & Cooper, 1984).
Participant witnesses in experiments are typically college
students. The reliance on this population has been criticized,
especially by prosecutors. However, many experiments have
included other populations, such as young children, adults, and
the elderly. Importantly, when differences are found, the results
suggestible and more accurate as eyewitnesses overall than are
either children or the elderly (Cutler & Penrod, 1995; Searcy,
Bartlett, & Memon, 1999). Presumably this is due to the higher
education level, intelligence, memory ability, visual acuity,
alertness, and general health of college students relative to the
general population. Thus, if anything, heavy reliance on college
student subject populations for eyewitness research may paint
an unrealistically rosy picture of eyewitness abilities.
Within the basic eyewitness-experiment paradigm, manipu-
lations are embedded and their effects are observed. For ex-
ample, an experiment focusing on system variables might have
everyone view the same simulated crime and then randomly
assign some participant witnesses to receive a postevent sug-
gestion or randomly assign some to receive a particular pre-
lineup instruction. In an experiment focusing on estimator
variables, participants might be randomly assigned to view a
crime in which the perpetrator is of a different race or the same
of eyewitnesses have become more common in recent years. A
major drawback to archival studies is the inability to establish
archival analyses. Archival analyses have proven to be par-
ticularly informative with regard to lineups. A lineup that is
properly constructed includes only one suspect (who might or
might not be the perpetrator); the other people in the lineup are
innocent fillers who would not be charged with the crime if they
were identified by the eyewitness. Thus, when an eyewitness
selects a filler in an actual lineup, it is immediately classifiable
as an error. It is not the type of error that could send an innocent
that), but it is an identification error nevertheless.
Archival analyses of filler identifications have yielded very
interesting results. Wright and McDaid (1996) analyzed 1,561
lineup outcomes in London and found filler-identification rates
of 19.9%. These data are similar to the 21% filler identification
rate reported by Slater (1994) in a study of 843 lineups con-
ducted by the Metropolitan Police in London. Behrman and
Davey (2001) reported that 24% of identifications from live
lineups inSacramento,California, wereidentificationsoffillers.
Valentine, Pickering, and Darling (2003) analyzed 119 lineups
in the greater London area and found that 21.6% of the
Volume 7—Number 2
eyewitnesses identified fillers. In these four studies of actual
eyewitnesses to serious crimes, filler identifications constituted
approximately one third of all positive identifications. These
archival results represent a very important complement to the
experimental studies of eyewitnesses for several reasons. First,
with rates obtained in experiments (Ebbeson & Flowe, n.d.;
Steblay, Dysart, Fulero, & Lindsay, 2001). Second, these ar-
chival results address a common criticism of experiments—
namely, that participant witnesses in experiments are not as
cautious as actual crime witnesses are, because the conse-
quences of a mistaken identification in an experiment are not
serious. But the witnesses in the archival studies were actual
witnesses to crimes and yet mistakenly identified fillers in one
third of their positive identifications. Third, the filler-identifi-
cation rates in the archival studies permit us to make conser-
vative estimates of the risk that an innocent suspect would face
in these lineups. For example, with five fillers in each lineup
(six-person lineup minus the suspect) and a 20% filler-identi-
fication rate, the risk to any given filler is 4%. If an innocent
suspect has the same risk as a filler, the estimated risk to an
innocent suspect is 4%.
These estimates of the risk to an innocent suspect are con-
servative for two reasons. First, lineups rarely yield equal dis-
tributions of error because the innocent suspect will commonly
stand out for any number of reasons, including the selection of
fillers that bear a poor resemblance to the description of the
perpetrator given by the witness (Valentine & Heaton, 1999;
Brigham, Meissner, & Wasserman, 1999). Second, when the
actual perpetrator is not in the lineup (i.e., the suspect is in-
nocent), the rates of filler identification increase (see Wells &
Olson, 2002). Assuming that the perpetrator was present in a
large proportion of the lineups in these archival studies, the
filler-identification rates underestimate the expected error rate
for any given lineup in which the perpetrator is absent.
Archival studies also permit analyses that examine results as
a function of different levels of critical variables. For example,
Wright and McDaid (1996) found that the filler-identification
rate was 20.8% for violent crimes and 17.6% for nonviolent
rate was 15.9% when a weapon was present and 23.7%
when there was no weapon. The latter result seems peculiar in
light of the experimental results indicating a deleterious effect
for the presence of a weapon (see meta-analysis by Steblay,
1992)—but in the weapons-effect section later in this mono-
graph, we note that archival data are subject to ‘‘selection ef-
Another interesting archival finding does not concern eye-
witnesses per se but has a powerful bearing on expected rates of
mistaken identification in the courtroom: Archival studies in-
90% of cases (Cole, 1986). Let us assume that 80% plead guilty
(the argument is stronger at 90%). We might assume that no
mistakenly identified (innocent) suspects plead guilty and
that all the guilty pleas are from guilty suspects. (In no sense do
we intend for this assumption to be interpreted as a denial
of the important work of Kassin & Gudjonsson, 2004, and other
false-confession researchers, who have clearly made a com-
pelling case that innocent people plead guilty.) Even if we
90% of the innocent suspects and only 20% of the guilty sus-
pects will go to trial. Assume further that a mere 4% of suspects
identified from a lineup are innocent and 96% are guilty. If we
not go to trial, only 20% of the 96% (19.2% of the guilty) will go
to trial, whereas 90% of the 4% (3.6% of the innocent suspects)
will go to trial. Thus, at the trial level, 16% of the defendants
(3.6% of the 22.8% going to trial) will be cases of mistaken
identification. Charman and Wells (2006) called this the
‘‘pleading effect’’; it illustrates how the mistaken-identification
rate can be expected to be higher at the trial level than at the
lineup level (see Fig. 1).
not under the control of the justice system, they are important to
central to our understanding of when and why eyewitnesses are
most likely to make errors. Informing police, prosecutors,
judges, and juries about the conditions that can affect the ac-
curacy of an eyewitness account is important. Second, our
at least at the extreme, dependent on levels of the estimator
variables. Consider a case in which a victim eyewitness is ab-
ducted and held for 48 hours by an unmasked perpetrator; the
witness has repeated viewings of the perpetrator, lighting is
good, and so on. We have every reason to believe that this wit-
ness has a deep and lasting memory of the perpetrator’s face.
Then, within hours of being released, the eyewitness views a
lineup. Under these conditions, we would not expect system
variables to have much impact. For instance, a lineup that is
biased against an innocent suspect is not likely to lead this
eyewitness to choose the innocent person, because her memory
is too strong to be influenced by lineup bias. On the other hand,
The effects on identification accuracy of a large number of
estimator variables—witness, crime, and perpetrator charac-
teristics—have been investigated by psychologists. Here we
significant research attention and achieved high levels of con-
Kassin, Tubb, Hosch, & Memon,2001) or have been the subject
of interesting recent research.
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
Meissner and Brigham (2001a) published the most recent broad
review of research on the problems associated with what has
sometimes been called other-race or cross-race identification
impairment or own-race bias (ORB). Meissner and Brigham
analyzed data from 39 research articles, with 91 independent
samples involving nearly 5,000 participant witnesses. They
examined measures of correct identification and false-alarm
rates, as well as aggregate measures of discrimination accuracy
and response criterion. They reported that the chance of a mis-
taken identification is 1.56 times greater in other-race than in
likely to correctly identify a previously viewed own-race face as
they were to identify an other-race face. Participants were more
than 2.2 times as likely to accurately categorize own-race faces
as new versus previously viewed as they were to accurately
categorize other-race faces. Meissner and Brigham explored the
and found that such contact played only a small role in ORB,
accounting for just 2% of the variability across participants (see
also Wright, Boyd, & Tredoux, 2003). They also found that the
amount of viewing time available to witnesses significantly in-
fluenced ORB; specifically, false alarms to other-race faces in-
creased when study time was limited.
Recent research by Pezdek, Blandon-Gitlin, and Moore
(2003) examined cross-race impairment in kindergarten
children, third graders, and young adults who viewed black and
white target faces and a day later were tested with a six-person
lineup. These researchers observed the usual cross-race
cross-race identification was less accurate than own-race
Despite the importance of knowledge about the effects of stress
on witnesses, researchers cannot simulate violent crimes and
pose a threat to the well-being of experimental subjects. Re-
induce stress, including the use of violent versus nonviolent
videotaped crimes. Increased violence in videotaped re-
enactments of crimes has been shown to lead to decrements in
both identification accuracy and eyewitness recall (Clifford &
Hollin, 1981; Clifford & Scott, 1978), but this finding is not
universal (Cutler, Penrod, & Martens, 1987a).
of stress effect studies. The meta-analysis was conducted on 27
tests of the effects of heightened stress on identification accur-
They found that high levels of stress negatively affected both
types of memory. The effect of stress was notably larger for tar-
get-present than for target-absent lineups—that is, stress par-
Fig. 1. The‘‘pleadingeffect’’(Charman&Wells,2006).Assumingfirstthat96%ofsuspectsidentifiedfrom
alineup areguiltyand 4%areinnocent,if 80%of theguiltysuspectsand10% ofthe innocent suspectsplead
is, cases of mistaken identification.
Volume 7—Number 2
considerably larger for eyewitness-identification studies that
simulated eyewitness conditions (e.g., staged crimes) than for
These effects are well illustrated in a study by Morgan et al.
(2004) that examined the eyewitness capabilities of more than
500 active-duty militarypersonnel enrolledin asurvival-school
program (see Table 1). After 12 hours of confinement in a mock
prisoner-of-war camp, participants experienced both a high-
stress interrogation with real physical confrontation and a
low-stress interrogation without physical confrontation. Both
interrogations were 40 minutes long; they were conducted by
different persons. Adayafter releasefrom the camp, andhaving
recovered from food and sleep deprivation, the participants
viewed a 15-person live lineup, a 16-person photo spread, or a
sequential presentation of photos of up to 16 persons. Regard-
less of the testing method, as Table 1 shows, memory accuracy
for the high-stress interrogator was much lower overall than for
the low-stress interrogator.
perpetrator’s weapon during the courseof acrime.Itisexpected
that the attention the eyewitness focuses on the weapon will
reduce his or her ability to later recall details about the per-
petrator or to recognize the perpetrator. Researchers have as-
establish the parameters of weapon-focus effects on perception
and memory; these efforts were reviewed in a meta-analysis by
Steblay (1992). The review included 19 studies with a total
sample of 2,082 participants. The weapon-focus effect on
identifications was statistically significant but reflected a mod-
est impairment; the effect on description accuracy was larger.
Research by Mitchell, Livosky, and Mather (1998); Pickel
(1998, 1999); and Shaw and Skolnick (1999) indicates that any
surprising object can draw attention away from the perpetrator
and that novelty, rather than threat, may be the critical ingre-
dient in the effect.
Researchers have tried to detect weapon-focus effects in field
studies, and the results are somewhat conflicting. Tollestrup,
the rate of suspect identification and obtained data consistent
with laboratory findings. But Valentine et al. (2003) did not
find a weapon-focus effect in their study of 640 attempts by
eyewitnesses to identify the alleged target in 314 lineups.
Of course, as noted earlier, in nonexperimental studies it is
difficult to control for variables that might obscure a weapon-
focus effect. For example, in the study by Valentine et al., the
primary outcome variable is suspect choices rather than per-
petrator choices (i.e., witness identifications are intended to
determine whether suspects are perpetrators)—whereas in ex-
perimental research the identity of the perpetrator is known to
Field research can also suffer from selection effects that can
obscure the effects of variables of interest. For example, a true
weapon-focus effect could be obscured if witnesses to crimes
involving weapons believe that their memory is weak and are
therefore less inclined to attend lineups. The result could be a
reduction in the number of weapon-focus-impaired witnesses
presented with lineups and thus a reduced number of cases of
As mentioned earlier,a selection effectmight actuallyreduce
our concern about the potential impact of weapon focus on
eyewitness performance. On the other hand, it is conceivable
that more intensive police investigations of weapon-present
for weapon-present witnesses, with the result that the apparent
performance of weapon-present witnesses is improved even
though their memories are impaired. If investigations of all
crimes were similarly intense, a weapon-focus effect might
emerge. One might also imagine that the police are more mo-
tivated to ‘‘help’’ weapon-present witnesses identify perpetra-
help might take the form of suggestive instructions to witnesses
and suggestive lineups.
Common sense tells us that the amount of time available for
viewing a perpetrator is positively associated with the witness’s
ability to subsequently identify him or her. A meta-analysis by
Shapiro and Penrod (1986) showed that the linear trend for
exposure time was associated with improved performance. The
effects of exposure time were illustrated in a study by Memon,
Hope, and Bull (2003) in which mock witnesses viewed a real-
istic videotaped crime in which the target/perpetrator was
visible for 12 versus 45 seconds. Witnesses were tested with
target-present and target-absent arrays 40 minutes later. The
proportion of correct identifications in target-present arrays and
Percentages of Accurate and Mistaken Identifications From
Study of Eyewitness Identification Under High Versus Low Stress
High stress Low stress
Correct identifications [target-present]
Live lineup method
Sequential photo method
Mistaken identifications [target-absent]
Live lineup method
Sequential photo method
Note. Source: Morgan et al. (2004).
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
correct rejections in target-absent arrays increased substan-
tially when exposure time increased from 12 seconds to 45
seconds (from 32% to 90% for correct identifications and from
15% to 59% for correct rejections), although mistaken identi-
fications in target-absent arrays remained high even with longer
exposure (85% at 12 seconds and 41% at 45 seconds).
It is common for people to don disguises before engaging in
criminal acts. Full-face masks, stockings, hats, and hoods can
be quite effective in diminishing the facial-feature cues ne-
cessary for recognition (Cutler, Penrod, & Martens, 1987a,
1987b; McKelvie, 1988; Patterson & Baddeley, 1977). For ex-
ample, Cutler et al. (1987b) had participants view a videotaped
liquor store robbery and later attempt an identification from a
videotaped lineup. In half of the robberies, the robber wore a
knit pullover cap that covered his hair and hairline. In the other
half, he did not wear a hat. The robber was less accurately
identified when he was disguised: 45% of the participants
identified the robber in the lineup test if he wore no hat during
the robbery; only 27% identified him if he wore a hat during the
Shapiro and Penrod, in their 1986 meta-analysis, coded ex-
periments for whether or not faces were changed between the
initial viewing and recognition phases. Transformations in-
cluded changes in facial hair and deliberate disguises, such as
masks or hats. Nontransformed faces were more accurately
recognized (effect size d 5 1.05; 75% vs. 54%) and less
Not all disguises or changes in appearance work. Yarmey
woman viewed for 15 seconds in naturalistic circumstances,
regardless of whether or not she wore a baseball cap and dark
sunglasses. There was, however, an interaction involving dis-
(involvingmentalrehearsal ofthe encounter)madesignificantly
more correct rejections in the no-disguise condition than in the
Common sense tells us that memory declines over time. Can we
expecteyewitness-identificationaccuracy todecline asthetime
betweenthecrime andtheidentificationtestincreases? Shapiro
and Penrod (1986) included retention interval in their meta-
grouped into long versus short time delays (the exact manipu-
lationdependedon thestudy),longer delays led tofewercorrect
identifications (d 5 .43; 51% vs. 61%) and more false identi-
fications (d 5 .33; 32% vs. 24%). Across all the studies exam-
ined in that meta-analysis (including those that did not directly
manipulate retention interval), retention interval also proved an
important determinant of correct identifications (r 5 ?.11, p <
.05), although there was no significant relationship with false
Read, Yuille, and Tollestrup (1992, Experiment 1) tested
person lineup; they found that alcohol intoxication while wit-
nessing the event was associated with a lower rate of correct
identifications when the level of arousal (manipulated by vary-
ing the participants’ perceptions of the probability of getting
caught stealing an item from an office) was low during the event.
False identification rates were the same for intoxicated and
sober participants. Of course, after one week the participants
effect of intoxication at viewing and identification would be.
Dysart, Lindsay, MacDonald, and Wicke (2002) note that the
popular belief is that intoxicated witnesses are less accurate
than sober witnesses. However, one theory concerning ‘‘alcohol
myopia’’ (Steele & Josephs, 1990) predicts an interaction be-
tween blood-alcohol level and identification procedures in
which witnesses who were intoxicated at encoding will be less
accurate only in target-absent conditions. The theory suggests
that, compared with intoxicated witnesses, sober witnesses will
encode more information/cues about the perpetrator, which will
facilitate correct rejections in target-absent procedures. In-
toxicated witnesses are likely to encode only salient cues, and
erroneous identifications will result where more subtle cues
would have indicated that the suspect was not the target. On the
other hand, using salient cues will be effective for intoxicated
witnesses when the target is present.
Dysart et al. (2002) examined the effect of alcohol con-
sumption on identification accuracy using ‘‘showups,’’ a pro-
cedure in which the witness is shown the suspect alone, without
any fillers. A showup is the identification procedure most likely
to be used by police with intoxicated witnesses. As predicted,
the researchers found that in the target-present showup condi-
tion, blood-alcohol level was not significantly related to correct
identification; however, in the target-absent condition, higher
blood-alcohol levels were associated with a higher likelihood
(52%) of a false identification than were lower blood-alcohol
System variables (variables that can be controlled in actual
cases) tend to center on factors that come into play after the
witnessed event has passed. At that point, the legal system has
some control over a number of important variables, but not
necessarily all variables. For instance, first responders at a
crime scene can separate eyewitnesses so they do not influence
each other, but some interactions could have already occurred
before the arrival of investigators. Similarly, although investi-
Volume 7—Number 2
gators have total control over how a lineup is conducted, some
example, when an eyewitness spontaneously identifies someone
on the street as the perpetrator of an earlier crime.
System variables tendtobedivided intotwobroadcategories.
One category is interviewing eyewitnesses, a process that
generally involves recall memory. The other category is the
identification of suspects, a process that generally involves
recognition memory. It is important to note that neither inter-
to be purely a memory process. Social influence can be a huge
factor in both.
The case of James Newsome, a man who served 15 years for a
murder he did not commit, is an extreme example of an eye-
witness making a positive identification from a lineup, even
though his memory told him that the man he identified was not
the man who committed the murder. After Newsome was proved
innocent and the actual perpetrator was found through physical
evidence, eyewitness Anthony Rounds came forward and de-
scribedhowChicago policehadforced himtoidentifyNewsome
from the lineup, even though he knew that Newsome was not the
administrators told Rounds whom to identify; when he resisted,
their intimidating insistence led him to identify Newsome and
yielded strong evidence to support Rounds’s claim, and a jury
awarded damages to Newsome; the finding was upheld by the
U.S. Seventh Circuit Court of Appeals (Newsome v. McCabe
et al., 2002).
Although this is an extreme example, it illustrates how ex-
traneous external variables can influence eyewitness testimony
without operating through memory mechanisms. Under other
circumstances, social-influence variables are thought to actu-
as ‘‘What kind of hat was the gunman wearing?’’ when the
gunman had no hat could lead an eyewitness to develop a
scientists concern themselves with both social-influence vari-
ables and memory variables.
Research on interviewing eyewitnesses dates back to the early
children in France, and William Stern (1904) initiated eyewit-
ness research on interrogation in Germany. Snee and Lush
(1941) wrote a short empirical article on the use of interrogatory
versus narrative methods of interviewing eyewitnesses. Modern
Elizabeth Loftus, who used the method of asking questions of
eyewitnesses to implant misleading information (e.g., Loftus &
Palmer, 1974). This line of research paved the way for experi-
mental studies of the effects of explicit and subtle forms of
misinformation imparted during questioning of adult and child
witnesses (for reviews see Bruck & Ceci, 1999; Loftus, 2005;
Wright & Loftus, 1998). This work led to important theoretical
advances in our understanding of the mechanisms underlying
eyewitness suggestibility in interviews. Examples include the
source-monitoring framework (Lindsay & Johnson, 1989;
Mitchell & Johnson, 2000; Poole & Lindsay, 2001); fuzzy-trace
theory (Brainerd & Reyna, 1998; Memon, Hope, Bartlett, &
1998);retrieval-induced forgetting (MacLeod, 2002);the role of
metacognition (Koriat, Goldsmith, & Pansky, 2000); and the
social-influence approach (Echterhoff, Hirst, & Hussy, 2005;
Drivdahl, & Beck, 2001).
In this monograph, we do not discuss the mechanisms re-
sponsible for distortions in information retrieved in eyewitness
interviews. Instead, we use one example of a procedure that
probative value of eyewitness evidence. This example shows
how researchers have attempted to wrap their knowledge about
memory and social influence into a set of procedures for inter-
viewing eyewitnesses. It is also the most developed and exten-
sively researched procedural package for gathering detailed
reports from cooperative eyewitnesses. (Readers who are inter-
ested in other approaches to interviewing eyewitnesses, in-
cluding interviews designed to detect deception, should refer to
reviews by Granhag & Stromwell, 2004; Memon & Bull, 1999;
Poole & Lamb, 1998; and Vrij, 2000.)
The Cognitive Interview
The cognitive interview (CI) was initially developed by the
psychologists R. Edward Geiselman (University of California,
Los Angeles) and Ronald P. Fisher (Florida International Uni-
versity) in the early 1980s (Geiselman et al., 1984; Geiselman,
Fisher, MacKinnon, & Holland, 1985) and has resulted in more
than two decades of research. Two main forces drove the de-
velopment of the CI. The first was a request from police officers
and legal professionals to improve the practices of police in-
terviewers when gathering information from eyewitnesses.
Analysis of the techniques used by untrained police officers in
Florida (Fisher, Geiselman, & Raymond, 1987) suggested that
there existed some fundamental problems in the conduct of
interviews, leading to ineffective communication and poor
memory performance. The ‘‘standard police interview’’ was
characterized by constant interruptions, excessive use of a
predetermined list of questions with an expectation that wit-
nesses could provide answers, and questions that were timed
inappropriately. For example, if the witness was describing one
of the perpetrators, the officer might switch the line of ques-
tioning to the actions of another perpetrator. Interestingly, the
same problems were identified in studies of the typical police
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
The CI in its present form represents the alliance of two fields
of study: communication and cognition. The social-psychologi-
cal concerns of managing a face-to-face interaction and com-
municating effectively with a witness were integrated with what
psychologists knew about the way people remember things. The
social aspects areembodiedinwhat isreferredtoasastructured
interview, which consists of a phased procedure (free report
followedbyopen-ended questions) andincorporates techniques
to facilitate communication. These techniques include rapport
building, which is designed to increase the transfer of control
from the interviewer to the witness, and the use of a questioning
strategy guided by the witness’s own free report rather than one
that is based on a predefined protocol. The structured interview
resembles the recommended procedure for conducting investi-
gative interviews with witnesses and victims in many countries
(see Poole & Lamb, 1998; Westcott, Davies, & Bull, 2002).
The original version of the CI was presented as a set of four
specific cognitive techniques for improving eyewitness recall.
Following a series of laboratory simulations and field research,
the procedure was revised in 1992 (Fisher & Geiselman, 1992).
The version of the CI that has subsequently evolved focuses
heavily on communication techniques and social dynamics and
is a procedure in which the cognitive and communication
components work in tandem. Here we will focus primarily on
research and practice relating to the revised CI (also referred to
in the literature as the ‘‘enhanced’’ CI). For a summary of the
revised CI procedure, see Table 2.
The revised CI comprises several phases during which the
interviewer engages with and establishes rapport with the wit-
ness, asks the witness to provide a narrative account of the
witnessed event, and then probes with questions relating to
interviewer interrupts as little as possible, allows the witness to
dictate the subject matter and sequence of questioning, and
aims of the CI is to facilitate the exchange of information be-
tween the witness and interviewer through effective communi-
The first task of the interviewer is to build rapport with the
witness. This rapport serves two functions. First it puts the
witness at ease, minimizing the discomfort and distress some-
times associated with sharing an intimate or fearful experience
with a stranger. Second, there is some evidence that building
rapport with open-endedquestions can increase the accuracy of
a child witness’s report (Roberts, Lamb, & Sternberg, 2004). An
important component of rapport building in the revised CI is for
Revised Cognitive Interview Procedure
Step 1. Build rapport
(a) Personalize the interview
details as possible but not to guess or fabricate.
(b) Transfer control to the witness
possible. Listen actively to what he or she has to say. Allow for pauses.
Step 2. Recreate the context of the original event and ask the witness to report in detail.
To reinstate context, invite the witness to close his or her eyes and place himself or herself back at the scene.
Step 3. Open-ended narration
(i) Request narrative description
Ask the witness to give a narrative account of the event in his or her own words. If clarification is required, use open-ended questions. Do not
interrupt the narration to ask questions, although prompts such as ‘‘tell more’’ may be used. Avoid judgmental comments and closed (yes/no)
(ii) Focused retrieval
This is not a technique but a general guideline to follow to help the witness concentrate on what he or she is describing by
? using open-ended questions
? allowing for long pauses
? not interrupting the witness when he or she is speaking
(iii) Extensive retrieval
Encourage the witnessto search throughhis or hermemory moreextensively byasking him or herto report details from a number of different
perspectives and in different chronological orders.
(iv) Witness-compatible questioning
Time the questions appropriately so they are compatible with the witness’s retrieval pattern rather than adhering to a protocol.
Step 4. Closure
Be sure to leave time to brief the witness and let him or her know what might happen next. Exchange contact information and encourage the
witness to get in touch if he or she remembers additional details.
Note. Adapted from Fisher and Geiselman (1992).
Volume 7—Number 2
Table 2 for details).
The ‘‘cognitive’’ part of the CI relies on two theoretical prin-
an overlap between the encoded information and the retrieval
cue. Reinstatement of the original encoding context increases
the accessibility of stored information (Tulving & Thomson’s
encoding specificity hypothesis, 1973). Second, multiple trace
theory (Bower, 1967)—which proposes that memories are made
up of networks of associations rather than discrete and uncon-
nected incidents—states that a memory can be cued by several
means and that information not accessible with one technique
may be accessible with another.
Having established rapport with the witness, the interviewer
instructs the witness to mentally reconstruct the physical and
personal contexts that existed at the time of the crime. Inter-
viewers can help witnesses by asking them to form an image or
impression of the environmental aspects of the original scene
(e.g., the location of objects in a room); to comment on any
emotional reactions and feelings (e.g., surprise, anger) at the
time; and to describe any sounds, smells, and physical condi-
tions (e.g., hot, humid, smoky) that were present. Occasionallya
witness can be taken back to the scene of the crime. Once the
witness has mentally reconstructed the context, the interviewer
asks him or her to provide a detailed account of the event
(the free narrative). To extend retrieval, the witness is asked to
report all details, including partial or incomplete memories.
To minimize editing, Fisher and Geiselman (1992) advised in-
terviewers to instruct witnesses to report everything that comes
to mind, even if it is trivial or out of chronological order. In
addition to facilitating the recall of additional information, this
techniquemay yield informationthatcan bevaluable inpiecing
together details from different witnesses to the same crime.
Roberts and Higham (2002) obtained ratings of the forensic
relevance ofdetails elicitedwiththeCIbyaskingpolice officers
andprosecutorstoratethe relevanceofeach detailtoacriminal
investigation/court proceeding. At best, only 50% of the
information the CI elicited was deemed relevant by forensic
experts. Most of the correct, forensically relevant details ap-
peared in the free-narrative account (cf. Memon, Wark, Bull, &
Ko ¨hnken, 1997).
interviewer can probe for details using open-ended questions
and, when appropriate, can ask follow-up questions to clarify
what the witness has said. It is imperative that interviewers
remind witnesses that if they are unsure of an answer to a
question, they should say so and not guess. Appropriate se-
quencing of the interviewer’s questions (referred to as inter-
have a unique mental representation of the event, depending on
the details or aspects of the event he or she attended to and the
order in which events unfolded for him or her (Fisher &
Schreiber, in press). The interviewer should be guided by the
interviewee’s pattern of recall rather than adhering to a rigid
protocol or predetermined checklist. For example, if an inter-
viewee is describing a suspect’s face, this indicates that the
mental image of the perpetrator’s face is currently active and
details about the face are accessible (Pecher, Zeelenberg, &
Barsalou, 2003). At this point, the interviewer should ask
questions relating to the suspect’s appearance and not switch to
another topic, such as the suspect’s car.
In a CI, the witness is encouraged to focus or concentrate on
mental images of the various parts of the event, such as the
suspect’s face (Fisher & Geiselman, 1992). The interviewer
an image and then describe it in as much detail as possible.
Bekerian and Dennett (1997) refer to this focus on specific
features as ‘‘molecular imaging,’’ as compared to the general
‘‘molar’’ approach, which emphasizes reinstating environmental
context. To effectively engage the interviewee in focused re-
appropriate points to allow the interviewee time to create an
image and respond (Memon, 2006). Unfortunately, the use of
imagery can produce increases in errors and increased use of
inferences in eyewitnesses’ spoken reports (Bekerian & Den-
nett, 1997; for a discussion, see Stevenage & Memon, 1997).
Alternative retrieval cues can be used to access an event. For
example, witnesses can be asked to recall an event in different
temporal order or from different perspectives. Some researchers
have found that witnesses can recall additional details if they
or the middle or if they describe its most memorable aspect
(Fisher & Geiselman, 1992; Geiselman & Callot, 1990). How-
ever, in other studies, no additional details have surfaced when
the witness recalls the event for a second time, in a different
order (Memon, Wark, Bull, et al., 1997). One of the most con-
troversial components of the original CI was that witnesses were
asked to ‘‘recall’’ an event from the perspective of another wit-
ness or from another location at the scene. The instruction to
change perspective typically does not yield additional details
and can increase errors, particularly if witnesses do not under-
stand what the interviewer wants them to do (Boon & Noon,
1994; Memon, Cronin, Eaves, & Bull, 1993). Fisher, Brennan,
and McCauley (2002) suggest that changing perspectives could
be potentially valuable for highly traumatized witnesses who
might find it too stressful to describe the event from their own
perspective. However, forensic investigators are uncomfortable
with the instruction to change perspective, presumably because
it could invite witnesses to speculate (Kebbell, Milne, & Wag-
Evaluation of the CI
The CI has been examined in approximately 65 studies to date.
A meta-analysis of 53 studies found a median increase of 34%
in the amount of correct information generated in the CI
as compared with a different interview model (Ko ¨hnken,
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
Milne, Memon, & Bull, 1999). There was also an increase in
incorrect details; we will return to this later. With the exception
of two field studies, all the studies have tested volunteer wit-
nesses (typically college students) in the laboratory. Witnesses
observe either a live event or a videotape of a simulated crime.
After a short delay (typically hours or days), the witnesses par-
ticipate in a face-to-face interview.The witnesses receive either
the CI or a control interview. The control is either a standard
police interview or a structured interview that incorporates the
phased approach referred to earlier. The interviews are tape
and incorrect statements. The accuracy of the reported state-
ments is high and comparable for both types of interview.
Gu ¨nter Ko ¨hnken and his colleagues in Germany (Ko ¨hnken,
Schimmossek, Aschermann, & Ho ¨fer, 1995; Ko ¨hnken, Thurer,
of the CI over the structured interview. In their studies, the
structured-interview group received training in basic commu-
nication skills that was comparable in quality and length to the
CIgroup’straining. Thetraining includedinstruction onrapport
study, both interviewees and interviewers were non-psychology
students with no prior experience in investigative interviewing.
The to-be-remembered event was a videotape showing a blood
donation. Participants were tested a week after viewing the
videotape. Each interviewer conducted one interview (n 5 30).
The CI significantly increased the amount of correctly recalled
information over thestructured interview without increasingthe
number of errors and confabulated (made-up) details. In a
subsequent study with adult participants, a small increase in
confabulated details was also noted (Ko ¨hnken et al., 1995).
Memon and colleagues (Memon, Wark, Holley, Bull, &
Ko ¨hnken, 1997) directly examined whether the CI advantage
was due to the use of the communication components of the
revisedCI(rapportbuilding, transfer ofcontrol, and elements of
reinstatement, imagery, reverse order, and reporting in detail).
As in the Ko ¨hnken research, cognitive and structured inter-
viewers received similar training, and each group was led to
believe it was using the superior interview technique. A third
group of interviewers served as the control and was not trained.
Both trained groups elicited more correct information than the
untrained group did. However, this was offset by the fact that
they also produced a significantly higher number of errors and
confabulations than the untrained group. These findings are
important in themselves but also raise the question of what is an
appropriate control group. The cognitive interviews produce
more correct details than do interviews conducted by an un-
trained group of interviewers. However, a structured interview
with some of the communication components of the CI built in
can also yield increases in correct recall. The increase in errors
that occasionally occurs could be somewhat problematic (for a
discussion, see Memon & Stevenage, 1996; Memon, 2006).
Some have argued that the production of incorrect as well as
correct information suggests that the CI may be affecting report
criteria (Memon & Higham, 1999; Roberts & Higham, 2002).
Others argue that there is no suggestion that witnesses should
lower their output criteria to produce unsure responses and in-
terviewers should instruct witnesses not to guess or fabricate
details (Fisher et al., 2002). It is important to note that accuracy
rates typically do not differ between the CI and comparison
young children, the elderly, and people who are intellectually
impaired—has also been examined. Given that the primary aim
of the CI is to increase the amount of information retrieved, it
may be the most effective procedure to use with young children,
because children tend not to provide as much information as
adults do. The results are somewhat mixed. The CI has been
found to increase the amount of correct information recalled by
children aged 7 to 11 years when the comparison group was a
standard (untrained) group (Saywitz, Geiselman, & Bornstein,
1992). When the comparison is a structured interview, the CI
increases correct information but can also increase errors in 8-
to 9-year-olds (Memon, Wark, Bull, et al., 1997; Milne, Bull,
Memon, & Ko ¨hnken, 1995).
More recently, Akehurst, Milne, and Ko ¨hnken (2003) exam-
8 to 9 years and 11 to 12 years after a 6-day delay. Children
days later. The CI led to an increase in correct recall as com-
pared with a structured interview, with no increase in errors.
There were no interactions involving age group or delay. As to
the suitability of the CI for younger children, Holliday (2003a)
reported thatamodifiedversionofthe revisedCIcould increase
recall) phase of the interview in 4- to 5-year-olds as compared
with a structured interview. In a later study with 4- and 8-year-
olds, Holliday (2003b) found that a CI given after postevent
misinformation reduced children’s reporting of misinformation
in the interview (for a similar finding with 8- to 9-year-olds, see
Memon, Wark, Bull, et al., 1997).
There is some evidence to suggest the CI can aid the
recall of adults (Milne, Clare, & Bull, 1999) and children
(Milne & Bull, 2006) with mild learning disabilities, although
further research is required with this population, using larger
sample sizes and people with a broader range of learning
To date, there have only been two published studies of the
efficacy ofthe CIwhenthe witnesses are olderadults. Melloand
Fisher (1996) found the CI led to similar increases in correct
recall when the participants were older adults (mean age 5 72
years) but Searcy, Bartlett, Swanson, and Memon (2001) found
no differences in correct identification (recognition) of a target
when witnesses aged 62–79 years were interviewed using a
procedure resembling the CI.
Volume 7—Number 2
The failure to find an effect of CI on recognition (in this case,
identification of a target) is consistent with earlier studies. In
four separate studies, Fisher, Quigley, Brock, Chin, and Cutler
(1990) found no advantage of the CI in recognition, but it did
elicit better descriptions of the target as compared with a no-
instruction control. Gwyer and Clifford (1997) compared the
revised version of the CI with a structured interview and again
found no reliable effects on recognition performance in target-
present conditions but a reduction in false identifications in
target-absent conditions in their short (48-hour) delay group
(cf. Yarmey, 2004). This finding did not generalize to the long
(96-hour) delay group.
These findings come as no surprise. The literature indicates
that environmental manipulations of context are not effective in
a recognition test when alternative cues are available. Accord-
ing to Smith and Vela (2001), this is because the influence of
contextual cues will be reduced or will be outshone when there
are strong retrieval cues present at the time of the memory test.
This is referred to as the outshining hypothesis. For instance, in
a recognition test in which a copy of the item to be remembered
is provided, this item serves as a retrieval cue, and contextual
cues are rendered ineffective. When the task is to recall an item
of information in the absence of a specific retrieval cue, the
reinstatement of context should guide memory (Smith, 1994).
However, as pointed out by Fisher and Schreiber (in press), the
outshining hypothesis leads to the prediction that experimental
in target-absent conditions.
with the revised CI are more likely to make correct rejections
and whether the effect of a CI in an identification situation will
vary as a function of retention interval (Gwyer & Clifford, 1997)
and other relevant system and estimator variables.
Police officers complain that eyewitnesses seldom provide suf-
ficient information (Kebbell & Milne, 1998). The CI has proved
to be a prime system variable in that a full and accurate eye-
The question is, what impact has the CI had on interviewing
United States, and it does not appear to have had a substantial
impact on the methods police officers use to interview witnesses
(Fisher & Schreiber, in press). Nevertheless, personnel from
technique. These organizations include the FBI; the National
Transportation Safety Board; the Department of Homeland Se-
curity; the Federal Department of Law Enforcement; and the
receive such training in the near future. The training has varied
across states and differs between federal and state training
academies. Fisher and Schreiber (in press) note that federal
investigators receive 18 hours of training in interviewing, in-
cluding techniques for interrogating suspects and nonpsycho-
logical topics such as the legal aspects of interviewing.
In England and Wales, the CI was introduced in a booklet to
every police officer as part of the national investigative inter-
examples of police training in the CI, with input to the trainers
from researchers, the training is typically limited to the de-
tective ranks or isonly providedin aminimal, introductory form
to junior officers (see Milne & Bull, 2006). A survey of police
officers (Kebbell, Milne, & Wagstaff, 1999) suggested that
relatively few officers used the full CI in practice. Training
programs have also been developed in other European coun-
tries, as well as in Australia, New Zealand, and Israel (Fisher,
2005). The efficacy of the CI has also recently been demon-
strated in Brazil (Stein & Memon, in press), with the aim of in-
troducing it to the Brazilian police and judiciary in the near
Given the extensive research on the CI and the light it has
shed on faulty interviewing practices, have police interviews
improved in the 20 years since the CI was first introduced? In a
recent analysis of police interview techniques, Fisher and
Schreiber (in press) asked 23 Miami detectives experienced in
investigations of robbery, sexual assault, homicide, and internal
affairs to tape record their witness interviews. Analysis of these
interviews revealed techniques and behaviors similar to those
identified 20 years earlier. This was particularly disappointing
by Fisher and Geiselman to disseminate their findings to prac-
titioners and to implement training programs.
The picture is just as bleak across the Atlantic. Clarke and
Milne (2001) conducted a national evaluation of investigative
interviewing training (the Planning, Engage, Account, Closure,
Evaluation—or PEACE—model) in England and Wales to see
if it had improved workplace practice. The PEACE model pro-
vides two ways ofobtainingan interviewee’s account: the CI and
conversation management. The latter involves asking witnesses
to give their own account of events; the interviewer then selects
logicalsequence.ClarkeandMilne (2001)foundlittle evidence
of any cognitive interviewing taking place. Most officers seem
preoccupied with getting a statement from the witness and
use of interrogative techniques or suspect interviews at the ex-
pense of gathering information from cooperative witnesses
(Milne & Bull, 2006).
Resources need to be directed toward training in witness-
interviewing practices. Milne and Bull (2006) argue that this
will involve procedural changes in collecting evidence in the
United Kingdom, such as electronic recording of all witness
interviews to maintain an accurate record of the original ac-
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
count, assessment of training and supervision of witness, and
victim interviews to ensure that appropriate techniques are
used. With respect to the United States, R.P. Fisher (personal
communication, March 28, 2006) has noted that nonpolice
groups, such as engineers, have displayed a willingness to use
CI in investigations, suggesting that perhaps those with an
academic background or a motivation to use investigative
techniques to arrive at solutions find it easier to understand the
theory behind the CI. Following this line of reasoning, perhaps
police officers with specialist skills (homicide, child protection)
might benefit more from training in the CI. However, those who
are specialists may already have an established protocol for
interviewing and thus be less willing to adopt new techniques
(Memon, Milne, Holley, Bull, & Ko ¨hnken, 1994).
We advocate a two-tiered approach to training. First, there
is a need for more extensive training programs on witness-
interviewing techniques for new police officers. Training and
examples of how faulty witness testimony can contribute to
miscarriages of justice might also prove fruitful (see Savage &
Milne, in press). The monitoring and assessment of witness in-
terviews (e.g., recording) is essential. A second approach is to
present trainees with a simpler, more accessible version of the
to encourage wider use.
Identifying Criminal Suspects
The identification of a criminal suspect can be the most im-
portant eyewitness evidence that is presented at a trial. This is
especially true when the eyewitness claims to have seen the
suspect commit the criminal act. In that case, the eyewitness-
identification testimony is direct evidence of guilt in the sense
that the accuracy of the identification has a one-to-one rela-
tionship to the ultimate issue of whether the suspect committed
the crime. In other situations, eyewitness identification evi-
dence may be circumstantial—for instance, if the eyewitness
evidence are needed to complete the inference that the person
who was seen is the same person as the one who committed the
crime. Regardless of whether the identification is direct or cir-
cumstantial, those who observe identification testimony (for
example, jurors) are likely to accept it as accurate if the eye-
witness is confident and consistent (e.g., Berman & Cutler,
1996; Bradfield & Wells, 2000; Brigham & Bothwell, 1983;
Cutler, Penrod, & Stuve, 1988; Lindsay, Lim, Mirando, & Cully,
1986; Lindsay, Wells, & O’Connor, 1989; Lindsay, Wells, &
Rumpel, 1981; Maas, Brigham, & West, 1985; Wells & Leippe,
1981; Wells, Lindsay, & Ferguson, 1979).
A primary method for obtaining identifications of criminal
suspects is the use of the lineup. Lineups can be either live, as
of the first and third authors, most lineups in the United States
are conducted using photographs. At its simplest level, a lineup
involves placing a suspect among distracters (called fillers) and
asking the eyewitness if he or she can identify the target. The
lineup is more complex than it at first appears. Understanding
how mistaken identifications can occur with lineups and what
kinds of system improvements can be made to prevent mistakes
and their possible outcome distributions.
Lineup Structure. Regardless of whether there is more than one
culprit, or target, a lineup should contain only one suspect, with
the remaining members being known-innocent fillers (Wells &
or might not be the target (i.e., might or might not be the actual
culprit). Hence, we will refer to two possible states of truth: (a)
the suspect is the target, and (b) the suspect is not the target.
Because there is only one suspect per lineup, these two states of
truth are equivalent to target-present and target-absent lineups.
incorrect rejection (making no identification), and (b) the
an innocent suspect in a target-present lineup. The only time an
eyewitness can mistakenly identify an innocent suspect is in a
target-absent lineup. Target-absent lineups can also result in
filler-identification errors, but these errors would not result in
charges being brought against an innocent person. We reserve
an innocent suspect; the identification of anyone other than the
suspect is called filler identification. Thus, the structural
properties of a lineup produce the set of possible outcomes
shown in Table 3. In an experiment, participant witnesses are
the real-world fact of an unknown probability that the police are
focusing on an innocent suspect. The proportion of target-
present and target-absent lineups (the target-present base rate)
is commonly 50/50 for experiments, but Bayesian statistics
base rates for any given experiment (see Wells & Lindsay,1980;
Wells & Olson, 2002; Wells & Turtle, 1986).
Typical Outcome Distributions. As would be expected from
better-than-chance performance, experiments typically show
that accurate identifications are more likely than inaccurate
identifications and that true rejections are more likely than are
false rejections (Clark, 2003; Wells & Lindsay, 1980; Wells &
Olson, 2002). Notice, however, that there are two types of filler
identifications. Filler identification Type 2 is a ‘‘miss’’ in the
theeyewitnesspickedsomeone else.Filleridentification Type1
is an accurate rejection in the sense that the suspect is innocent
and the eyewitness did not pick him or her. In general, experi-
mentsshowthatType2filler identificationsare more likelythan
Volume 7—Number 2
are Type 1 filler identifications (Wells & Olson, 2002). This
makes sense and fits well with the concept of relative judgments
(Wells, 1984), in which it is presumed that eyewitnesses tend to
is absent, the chances increase that one of the fillers will be
perceived as looking like the target. Usually, eyewitness-
identification performance is calculated by the extent to which
accurate identifications exceed mistaken identifications and
true rejections exceed false rejections. However, the rate of
mistaken identifications can be decreased without increasing
correct rejections by shifting identifications to fillers in the
target-absent lineup. This is a key to understanding how careful
selection of fillers for lineups can reduce mistaken identifica-
tions even if it does not reduce the propensityof eyewitnesses to
attempt identifications from target-absent lineups.
Target Removal Without Replacement. The relative-judgment
conceptualization (Wells, 1984) has permeated the literature
on lineups. It simply states that eyewitnesses have a tendency
to identify a person from the lineup who looks most like their
memory of the target relative to the other lineup members. As
long as the actual target is in the lineup, the relative-
is not in the lineup, problems ensue, because there will always
be someone who looks more like the target than the other lineup
members. Various results have been interpreted as supporting
the relative-judgment conceptualization, but the removal-
without-replacement (RWR) effect is the best evidence in sup-
port of the relative-judgment conceptualization.
eyewitnesses viewed either a six-person lineup that contained
the target or a five-person lineup in which the target was re-
moved and not replaced with anyone. In both conditions, the
(see following section on pre-lineup instructions). When the
targetwaspresent,54%pickedthe target,21%selected noone,
and 25% selected fillers. Wells reasoned that if the 54% rep-
resented true recognition rather than a relative judgment, re-
moval of the target should result in the 54% joining the 21% in
picking no one. When the target was removed, however, only
32% selected no one, and 68% selected fillers.Thus, among the
54% selecting the target when the target was present, it is es-
of the fillers in the absence of the target. Recent data show the
RWR effect to be robust across a variety of conditions, and the
magnitude of the effect appears to be greater when memory is
weaker (Clark & Davey, 2005; MacLin, Wells, & Phelan, 2004).
There remains some debate about the psychological processes
underlying the RWR effect. Ebbessen and Flowe (n.d.), for in-
stance, suggest that it could simply represent a downward cri-
that accrues to an innocent suspect when the actual target is not
The effect also further illuminates the problem of filler
selections that we noted earlier in the discussion of archival
studies using police files, in which one third of positive
identifications by witnesses were identifications of innocent
fillers. In the American archival study, Behrman and Davey
(2001) found that nearly a quarter of witnesses selected a filler
(and 50% selected the suspect). Thus, the average filler was
selected by 5% of witnesses—what might be termed ‘‘bad
good enough to avoid errors; Penrod, 2003). Of course, in a
perfectly fair array, one would have to assume that at least
another 5% of witnesses would ‘‘guess’’ the suspect. These
selections might be characterized in various ways: Steblay et al.
(2001) called them ‘‘calculated guesses’’ and Penrod called
them ‘‘lucky guesses.’’
As we discuss later, there are reasons to believe that many
lineups are not fair and that calculated/lucky guesses produce
many suspect identifications that look like ‘‘hits’’ but are really
the product of biased arrays and witness guessing. Steblay et al.
(2001) reported, for instance, that in studies of target-absent
simultaneous arrays in which a filler similar to the suspect was
designated the ‘‘innocent suspect,’’ that person was picked by
27% of witnesses (across all studies, one of the six fillers—in-
cluding the suspect—was picked by 51% of witnesses). One
might expect that in a fair lineup the innocent filler would be
selected by 8.5% (51%/6) of witnesses instead of 27%. The
much higher rate of suspect identification suggests that the
witnesses had some memory for the appearance of the missing
target but not enough of a memory to avoid mistakenly identi-
fying an innocent person.
Possible Outcomes From a Lineup
State of truth
Response of Eyewitness
Identification of suspect Identification of fillerNo identification
Suspect not target
Suspect is target
Filler identification type 1
Filler identification type 2
picked. Source: Charman & Wells (2006).
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
Pre-Lineup Instructions. One of the first and most fundamental
lineup system variables to be tested empirically was the in-
struction (or warning) to eyewitnesses that the target might or
target-absent lineups and target-present lineups; they either
be present or gave no instruction. When participants viewed a
target-present lineup, the instruction had little effect on the
distribution of responses. When they viewed a target-absent
lineup, however, the instruction reduced choosing rates dra-
the chances of both mistaken identifications and filler identifi-
cations, has been replicated extensively (see meta-analysis by
Steblay, 1997). A more recent meta-analysis indicates that ac-
curate identification rates in target-present lineups might be
slightly harmed by the instruction, but the decline in accurate
identifications when the target is present is much smaller than
the decline in mistaken identifications when the target is absent
Selection of Fillers. The characteristics of the fillers used
in a lineup have a strong influence on the chances that
an innocent suspect will be identified in a target-absent lineup.
In general, if the innocent suspect fits the description
of the target and the fillers do not, the innocent suspect
is likely to be mistakenly identified. The first empirical dem-
about the optimal criteria for selecting fillers. Two primary
strategies for selecting fillers have been advocated. One is to
select fillers who resemble the suspect. Luus and Wells (1991)
argued against this strategy because it has no ‘‘stopping point’’
and also because it risks creating a lineup of clones, which
would reduce accurate identification rates for target-present
lineups. Wells, Rydell, and Seelau (1993) found that selecting
fillers on the basis of the description given by the eyewitnesses
managed to protect the innocent suspect in target-absent
lineups without harming accurate identification rates in target-
present lineups. On the other hand, selecting fillers on the basis
of their resemblance to the suspect harmed hit rates with no
additional protection for the innocent suspect in target-absent
Wogalter, Marwitz, and Leonard (1992) presented another
argument against selecting fillers on the basis of their
resemblance to the suspect: The ‘‘backfire effect’’ refers to the
idea that, somewhat ironically, the suspect might stand out if he
or she was the basis for selecting the fillers in the lineup, be-
lineup. Clark and Tunnicliff (2001) reported evidence for the
backfire effect. However, eyewitnesses’ descriptions of the tar-
get are often sparse and sometimes do not even match the
Meissner, Sporer, & Schooler, in press; Sporer, 1996, in press).
The general recommendation for selecting fillers for lineups has
been to use the eyewitness’s description of the target and to take
any additional measures needed to make sure that the suspect
does not stand out in the lineup (Wells et al., 1998).
Along with these strategies for selecting fillers, various
techniques to assess lineup fairness by using ‘‘mock witnesses’’
have been developed. The task of a mock witness is to examine
the lineup and try to discern which person is the suspect. From
this mock-witness paradigm, various metrics have been devel-
(Malpass & Lindsay, 1999). In lab studies, the mock-witness
paradigm appears to be sensitive to lineup bias and is relatively
robust across variations in lineup procedure (e.g., simultaneous
vs. sequential procedures; see McQuiston & Malpass, 2002).
Studies of photo arrays and lineups from actual cases using the
mock-witness method reveal that arrays are frequently biased
against suspects, who are picked more than twice as often
(relative to the fillers) as one would expect by chance alone
(Brigham et al., 1999; Valentine & Heaton, 1999; Wells &
Lineup Size. A common practice in the United States is to use
five or six persons (a suspect plus four or five fillers) in a live
lineup and six or eight photos in a photo lineup. For purposes of
this discussion of lineup size, we will assume that each lineup
member is viable in the sense that the fillers are selected to fit
the description and in other ways do not make the suspect stand
out. Given a set of properly selected lineup fillers, there is no
reason to believe that an innocent suspect has a greater chance
than any of the fillers to be identified by an eyewitness. Hence,
eyewitness researchers have adopted the assumption that the
chances of a mistaken identification are (1/N) ? p(I), where N
is the number of lineup members and p(I) is the probability
that an eyewitness will make an identification (see Doob &
Kirshenbaum, 1973; Wells, Leippe, & Ostrom, 1979). Note that
increasing lineup size reduces the chances of a mistaken
identification in a negatively decelerating fashion (i.e., each
additional lineup member reduces the chances of a mistaken
identification less than the previous addition did). Because of
this negative deceleration, the addition of persons to the lineup
brings diminishing returns. Thus, adding six additional mem-
bers to a six-person lineup reduces the chances of mistaken
identification from 16.7% to 8.3% (i.e., among those making an
identification). But, adding six members to a 12-person
lineup reduces the chances of mistaken identification from
8.3% to 5.5%.
If reducing the chances of a mistaken identification were the
only consideration, increasing the size of the lineup to a very
high number is a good idea, even with diminishing returns. But
the formula speaks only to mistaken identifications from target-
from target-present lineups. The idea of a system variable im-
provement for lineups is to reduce the chances of a mistaken
identification without harming the chances of an accurate
Volume 7—Number 2
identification. Thus, the critical question is what happens to
accurate identifications as a function of increasing lineup size?
The eyewitness-identification literature has not derived a pre-
cise function relating lineup size to accurate-identification
rates. Levi (2002) reported no drop in accurate-identification
rates when lineup sizes were increased from 10 to 40 persons.
to view up to about 300 photos with little reduction in the
chances of an accurate identification (Ellis, Shepherd, Flin,
Shepherd, & Davies, 1989; Lindsay, Nosworthy, Martin &
Martynuck, 1994). These results are consistent with the general
observation that identifications of the target from target-present
lineups are not as sensitive to lineup variations as mistaken
identifications from target-absent lineups are (Charman &
Wells, 2006). For example, the ‘‘might or might not be present’’
instructions have little effect on accurate identifications from
target-present lineups but appreciably reduce identifications
filler-biased lineup has little effect on accurate identifications
from target-present lineups but increases mistaken identifica-
tions from target-absent lineups (Wells, Rydell, & Seelau,
1993). Also, suggestive influences from lineup administrators
appear to have little effect when eyewitnesses view a target-
present lineup but have a strong effect when the eyewitnesses
view a target-absent lineup (Haw & Fisher, 2004). More sys-
tematic research isneededbeforeitwillbepossibletoconclude
that lineup sizes can easily be raised to 20 or more persons
without harming accurate identification rates, but there appears
to be great promise in the simple idea of increasing the nominal
size of lineups.
Double-Blind Lineups. Police conducting a lineup has been
likened to psychologists conducting an experiment (Wells &
Luus, 1990). One element of this rich analogy is the idea of the
double-blind lineup (Wells, 1988). Normally, a lineup is con-
knows which person is the suspect and which people are merely
for lineups. Experiments have shown that when the lineup ad-
ministrator is led to believe that a particular lineup member
(randomly selected) is the suspect, the chances that the eye-
witness will identify that person are increased (Haw & Fisher,
when the eyewitness selects the person whom the lineup ad-
ministrator was led to believe is the target, the eyewitness ex-
presses higher levels of confidence in the identification
(Garrioch & Brimacombe, 2001).
The idea of the double-blind lineup is straightforward: The
person who administers the lineup should not be aware of which
lineup member is the suspect and which members are fillers
(Wellsetal., 1998).Thisrecommendation doesnot presume any
intention or awareness on the part of the lineup administrator
to influence the eyewitness. Some police jurisdictions might be
concerned about manpower issues involved in using an inde-
pendent lineup administrator. Because most lineups in the
is to have a laptop computer administer the lineup, thereby ef-
fectively eliminating any possible influence from the lineup
administrator (for a description of such a program, see MacLin,
Zimmerman, & Malpass, 2005).
Sequential Lineups. An alternative to the traditional police
lineup, the sequential lineup, was introduced in the mid-1980s
which all members are presented to the eyewitness at once
members to the eyewitness one at a time. The eyewitness is told
that he or she will view a number of people—the number is not
specified.The witness makesa decision on each lineup member
theory behind the sequential lineup is that it prevents eyewit-
nesses from relying on relative judgments, in which one lineup
member is compared with another and the one most similar to
the target is picked. Although the eyewitness can compare the
lineup member currently being viewed with those already seen,
there is a chance that a lineup member yet to come might look
even more similar to the target. The initial results indicated
support for a sequential-superiority effect in which identifica-
tions from target-absent lineups diminished while identifica-
tions of the target from target-present lineups remained largely
Years of additional experiments culminated in a meta-
(Steblay et al., 2001). The meta-analysis supported the original
observation of lower mistaken identification rates for the se-
quential than for the simultaneous lineup; however, there was
also a reduction inaccurate identificationsof the target from the
target-present lineups. In general, the sequential procedure
appears to result in fewer identification attempts overall com-
procedure reduced mistaken identifications at a greater rate
than it did accurate identifications, this shift in performance is
consistent with a criterion shift in which eyewitnesses set a
higher criterion for identification with the sequential than with
the simultaneous procedure (Meissner, Tredoux, Parker, &
shift away from relative judgments.
Recall that the RWR effect indicates that some proportion of
accurate identifications appears to result from relative judg-
ments rather than true recognition. Thus, a shift away from
relative judgments is likely to result in fewer accurate identifi-
cations as well as fewer mistaken identifications. An argument
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
to a higher criterion for making a positive identification or to a
reduced reliance on relative judgments) is desirable, as mis-
taken identification is the primary cause of convictions of the
innocent. The trade-off of accurate and mistaken identifications
ultimately is a decision for policymakers, not scientists. How-
ever, Steblay et al. (2001) and Penrod (2003) argue that any
losses of accurate identifications that result from reduced reli-
ance on relative judgments are merely reductions in lucky or
A recent field experiment involving actual lineups conducted
in Illinois (Illinois Pilot Program, 2006) has been touted as a
comparison of the sequential lineup and the traditional simul-
yielded fewer filler identifications and more suspect identifica-
tions than did the sequential lineups. In fact, however, this two-
condition experiment actually confounded several variables.
Perhaps the most important confound was that the simultaneous
lineups were never conducted using double-blind procedures
but were always conducted by the case detectives. The se-
quential lineups, in contrast, were always conducted using the
double-blind method. Thus, the low filler rate obtained in the
simultaneous lineups could have been the result of not using
be noted that the double-blind sequential-lineup data in the
Illinois Pilot Program conformed quite well with data obtained
using the double-blind sequential procedure in the Hennepin
County (Minnesota) pilot project (about 8% filler identification
rates; see Klobuchar,Steblay,& Caligiuri, in press). In contrast,
the very low filler rate reported in the Illinois Pilot Program
using the nonblind simultaneous procedure (about 3%) is an
extreme outlier from the approximately 20% rate found in other
jurisdictions with simultaneous lineups (see Behrman & Davey,
2001; Slater, 1994; Valentine et al., 2003; Wright & McDaid,
1996). The profoundly low filler-identification rate for simul-
taneous lineups reported in the Illinois PilotProgram suggests a
suppression of filler identifications and/or a reluctance to report
filler identifications by the nonblind lineup administrators.
Thus, we are reluctant at this time to consider the Illinois Pilot
Program to be an interpretable test of the simultaneous versus
the use of sketch artists or composite faces. Little systematic
fairly large sample would be required to reach generalizable
conclusions. Considerable research exists, however, on com-
crime investigators in place of sketch artists. Composite pro-
duction systems create faces by selecting features (e.g., nose,
eyes, chin,head shape,hair,mouth, brows,ears) and combining
them into a face. One of the original systems, Identi-Kit, used
line drawings of facial features on transparencies. An accom-
panying booklet displayed all the possible features, and the
eyewitness selected features that were then overlaid on each
same system, except that the features were black-and-white
photos of actual facial features instead of line drawings.
In recent years, computer software programs have replaced
transparency-based composite systems. Examples of such soft-
ware are E-Fit, Evo-FIT, CD-Fit, and Mac a Mug (Frowd et al.,
enforcement agencies (Cote, 1998). FACES includes 361 hair
selections, 63 head shapes, 42 forehead lines, 410 sets of eye-
brows, 514 sets of eyes, 593 noses, 561 sets of lips, 416 jaw
shapes, 145 moustaches, 152 beards, 33 goatees, 127 sets of
chin lines. In each feature category, a selection button permits
the user to view subsets of the feature that meet a particular
description. For instance, eyes are divided into the subsets
narrow, deep set, overhanging lids, heavy lids, average blue or
green, almond-shaped blue or green, average brown, almond-
narrow, average with round base, average with broad base,
average pointed, hooked nostrils not showing, hooked nostrils
showing, slightly flared nostrils, very flared nostrils, round
(bulbous), average large, wide base with nostrils showing, and
wide base with nostrilsnot showing. In addition, controls permit
the features to be moved up or down and closer or farther apart,
and to be made larger or smaller. The features are displayed on
side. When a feature is clicked, it appears on the face. To make
changes—for example, in the eyes—one simply clicks a dif-
face: The eyewitness constructs a face by selecting features and
assembling them. Numerous face-recognition researchers have
noted that this method may conflict with the natural way faces
are encoded in memory—namely, in a holistic manner (e.g.,
Tanaka & Farah, 1993; Tanaka & Sengco, 1997; Wells & Hry-
ciw, 1984). Research experiments generally indicate that com-
(e.g., Bruce, Ness, Hancock, Newman, & Rarity, 2002; Ellis,
Davies, & Shephard, 1978; Kovera, Penrod, Pappas, & Thill,
1997). The research by Kovera et al. illustrates the difficulty of
generating a composite that resembles the intended target. The
researchers used a set of 50 composite images of the faces of
high-school classmates and faculty created by former students.
Other graduates of the same schools judged the composites’
rating the faces’ familiarity and their own confidence in that
Volume 7—Number 2
assessment and, where possible, giving names. Ratings of fa-
miliarity and confidence did not differentiate significantly be-
tween the known and unknown faces, and only 3 of the 167
names offered for the composites proved to be correct! Ratings
and their assessments of the quality of their composites were
unrelated to identification accuracy on any measure. The re-
searchersconcludedthat‘‘thefindings. . .raisedoubtsaboutthe
likelihood that composites prepared under field conditions will
yield a pinpointed identification of a perpetrator by individuals
who know the perpetrator’’ (Kovera et al., 1997, p. 245).
Although early research using the Identi-Kit and Photo-Fit
systems themselves(e.g.,toofew choicesoffeatures; Ellisetal.,
1978), thereisan emerging consensusthatpeople simplydonot
have good memories for isolated facial features and that any
system that requires parts-to-whole-face recall will be severely
limited. Furthermore, research suggests that having eyewit-
nesses build a composite face can damage memory for the ori-
ginal face and make the witnesses less able to recognize the
original target face in a later lineup (Wells, Charman, & Olson,
2005). Similar effects have been observed for giving verbal
descriptions of faces, a phenomenon called the verbal over-
shadowing effect (originally demonstrated by Schooler &
Engstler-Schooler, 1990; and see meta-analysis by Meissner &
Recent research has produced some encouraging results for
cases in which multiple eyewitnesses independently produce
composites. In such cases, morphing the individual composites
any individual composite (Bruce et al., 2002; Hasel & Wells, in
press). But even the morph of four individual composites does
not produce a dramatic likeness of the original face. Hasel and
Wells reported that the ability to pick the original target from
sets of four alternative faces was 35% for individual composites
and 48% for four-composite morphs (chance 5 25%).
Postdiction variables are neither system nor estimator variables
in the traditional sense, because they are not presumed to
measurable products that correlate with the accuracy of eye-
witnesses in a noncausal manner. The most researched of these
is the confidence (certainty) of the eyewitness. Another post-
diction variable is response latency—specifically, how long the
variable that we review here is self-reported decision process.
The confidence an eyewitness expresses in his or her identifi-
cation is one of the most researched questions in the study of
eyewitnesses. First, there is a strong intuitive appeal to the idea
thatconfidenceandaccuracy shouldbecloselyrelated. Second,
courtshave explicitlyendorsed the idea that the reliabilityofan
eyewitness should be gauged at least in part by the person’s
confidence, a tenet advocated by the U.S. Supreme Court
(Manson v. Braithwaite, 1977). Third, even in the absence of
instructions to pay attention to eyewitness confidence, partici-
pant jurors rely heavily on the confidence of the eyewitness in
Bradfield & Wells, 2000; Fox & Walters, 1986; Lindsay et al.,
1986; Lindsay et al., 1989; Lindsay et al., 1981; Wells, Fergu-
son, & Lindsay, 1981; Wells et al., 1979).
Initially, eyewitness researchers focused on the relationship
between eyewitness-identification confidence and eyewitness-
identification accuracy (Wells & Murray, 1984). This was a
useful starting point, but it is now clear that the relationship
between confidence and accuracy varies greatly as a function of
many other factors. For instance, it depends, in part, on how
similar the mistakenly identified person is to the actual target
(Lindsay, 1986). The confidence–accuracy relationship is gen-
erally higher when memory strength is stronger rather than
weaker (Deffenbacher, 1980); when it is calculated only among
those who make an identification rather than among both those
who make an identification and those who do not (Sporer, Pen-
rod, Read, & Cutler, 1995); and when it is calculated across
witnesses under different viewing conditions rather than among
In their meta-analysis of 30 studies involving a total of 4,036
participant witnesses, Sporer et al. (1995) estimated that the
confidence–accuracy correlation among choosers could be as
high as 1.41. Wells, Olson, & Charman (2002) note that a .41
point-biserial correlation (a correlation between a two-level
variable and a continuous variable) between confidence and
accuracy in eyewitness identification is less than the point-
biserial correlation between height and gender in humans.
Nevertheless, under conditions of uncertainty, a postdiction
variable that has a .41 correlation to a criterion variable is not
something that should be ignored. In fact, the American Psy-
chology-Law Society’s white paper on lineups endorses the idea
of making a clear record of the confidence of an eyewitness that
triers-of-fact might later use (Wells et al., 1998).
Accuracy of Highly Confident Witnesses
Though confidence–accuracy correlations are sometimes rela-
tively high, most research yields relatively low correlations.
Attempts have been made to increase the correlation through
accountability, context reinstatement, and other thought ma-
nipulations, but none has been successful, and such measures
commonly have the reverse effect of harming the confidence–
accuracy relationship (Robinson & Johnson, 1998). Some have
argued that despite the generally weak confidence–accuracy
correlation, accuracy may be very high among the most confi-
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
dent witnesses. One analytic method that addresses this ques-
tion uses calibration methods that measure peoples’ confidence
on a percentage scale (zero, 10%, 20%, 30%, and so on) and
then clumps people together at different levels of confidence to
assess their accuracy (see Brewer, Keast, & Rishworth, 2002;
& Penrod, 1989; Juslin, Olsson, & Winman, 1996; Weber &
Brewer, 2003, 2004).
CutlerandPenrodfoundwitness overconfidence of10to20%
confidence levels indicated). Juslin et al. (1996) found that
confidence scores were roughly comparable to accuracy scores;
in particular, in a 95% confidence group, judgments were 85 to
90% accurate (the exact numbers are not reported—numbers
are estimated from figures). Although these numbers look
promising, even inthe 95%confidence group there appear tobe
10 to 15% errors; errors are much higher—with greater over-
confidence—at lower confidence levels.
Other researchers have found less promising results. Though
the published numbers are slightly ambiguous, it appears that
and Spaulding (1982) were 85% correct. Brewer et al. (2002)
of their identifications (95% certain) were about 70 to 75%
correct—that is, high error rates and substantial overconfi-
dence.Ina1987studybyFleet, Brigham, andBothwell, 75%of
subjects who rated themselves as extremely confident were ac-
curate. Brigham (1990) found a 74% accuracy rate for the top
27%mostconfidentwitnesses.Bornstein andZickafoose (1999)
reported that they found overconfidence in both general-know-
were correlated. The latter finding suggests that confidence has
an individual-difference component that can be independent of
the task. Research by Perfect and Hollins (1996) suggests that
poor confidence–accuracy relationships are at least partly at-
tributable to people’s lack of insight regarding their general
abilities in the eyewitness domain.
The general point is that these results are consistent with
other measures of the confidence–accuracy relationship.
Even the calibration approach does not uniformly support the
notion that confidence is a highly reliable indicator of accuracy.
Error rates can be high among even the most confident
witnesses. Furthermore, these numbers presume that the crim-
inal justice system would skim off only the most confident wit-
nesses and that none of those witnesses would have had their
confidence artificially boosted.
The Problem Grows Worse
Imagine that prosecutors are skimming only the most confident
witnesses; there is no artificial confidence-boosting among the
witnesses; and we have reliable measures of confidence, not the
vague verbal reports currently obtained by police. Among these
highly confident witnesses, the results above indicate that 20 to
30% could be in error. But even if the error rate is only 10% for
these highly selected and most confident witnesses, they will all
appear highly confident to jurors—so confidence cannot help
the jurors figure out which witnesses have made errors. Indeed,
the simple correlation between confidence and accuracy for
these witnesses will be much worse than among all witnesses,
because there is very little variability in confidence and maybe
nousefulvariance.Though it istempting toconclude that jurors
might be entitled to assume a fairly high base rate of accuracy
among these highly confident witnesses (even if confidence
cannot aid them in differentiating accurate and inaccurate
witnesses), the pleading effect discussed earlier suggests that it
would not be safe to conclude that the accuracy rate is fairly
high; indeed, the accuracy rate could be fairly low, because the
guilty defendants facing confident witnesses have already
pleaded guilty. In short, the research results and logic call into
question the notion that witness confidencecan beofsignificant
assistance to jurors.
Even if the research showed that eyewitness-identification
confidence and accuracy are related at a level that could have
practical utility, this conclusion would come with another huge
caveat. Wells and Bradfield (1998) showed that giving con-
firming feedback to eyewitnesses who had made mistaken
identifications (e.g., ‘‘Good, you identified the suspect’’) pro-
duces profound distortions in their retrospective judgments,
including their recollections of how confident they were when
they made their identification, how good a view they had when
they witnessed the event, and how much attention they devoted
to the target’s face during the event.
The idea that eyewitness confidence can be driven by vari-
ables that are independent of accuracy has theoretical roots in
Leippe’s (1980) early analysis of the problem, but the fact that
other testimony-relevant variables (such as self-reports of at-
tention and view) are also malleable is a startling revelation.
There are numerous replications of this phenomenon, known as
2002; Dixon & Memon, 2005; Hafstad, Memon, & Logie, 2004;
Neuschatz et al., 2005; Semmler, Brewer, & Wells, 2004; Wells
& Bradfield, 1998, 1999a; Wells, Olson, & Charman, 2003; also
identification feedback effect occurs even if the feedback is
delayed for 48 hours (Wells et al., 2003). The effect occurs for
et al., 2004), and the effect occurs for both the elderly (Neus-
chatz et al., 2005) and young children (Hafsted et al., 2004).
Importantly, the confidence-inflating effect of confirming feed-
back is greater for eyewitnesses who have made a mistaken
identification than for those who have made an accurate iden-
tification; as a result, confirmatory post-identification feedback
harms the accuracy–confidence relationship (Bradfield et al.,
2002). Furthermore, a recent experiment showed that the post-
identification feedback effect occurs for actual eyewitnesses to
real crimes (Wright & Skagerberg, in press).
Volume 7—Number 2
The post-identification feedback effect is of considerable
practical import, because it is a common practice for lineup
administrators to give eyewitnesses feedback about their iden-
tifications. When an eyewitness has received some form of
feedback before being asked about his or her confidence in the
identification, the confidence statement is contaminated. Eye-
witnesses tend to believe that the feedback did not affect them;
however, those who report that the feedback did not affect their
response to the retrospective confidence question are never-
theless affected just as much as are the smaller portion of wit-
nesses who report that it might have affected them (Wells &
Bradfield, 1998). Fortunately, if the eyewitness is asked to in-
tends to inoculate the eyewitness against post-identification
feedback effects (Wells & Bradfield, 1999a). The need for im-
mediate measures of confidence is further indicated by the fact
that repeated questioning, expenditure of effort over time, and
to inflate eyewitness confidence even when accuracy is held
constant (Shaw & McClure, 1996; Shaw & Zerr, 2003; Shaw,
Zerr, & Woythaler, 2001). Clearly, the most pristine measure of
witness confidence is one collected from the witness at the time
ofidentification and beforethe contaminating influence ofthese
An intriguing phenomenon that appears to be related to the
post-identification feedback effect is visual hindsight bias.
Harley, Carlsen, and Loftus (2004) presented participants with
photographs of familiar faces that were severely degraded
(blurred) but gradually resolved to full clarity. After the identity
of the face became apparent, participants predicted the level of
blur that would permit a naı ¨ve observer to identify the face.
Participants who had already learned the identity of the face
consistently predicted that a naı ¨ve participant would be able to
identify the face at levels of blur that were in fact too severe for
think that objectively poor viewing conditions are nevertheless
sufficient for accurate identification. This ‘‘saw it all along’’ ef-
fect could be an important component of the propensity for
eyewitnesses to have retrospective overconfidence in their
Another interesting postdictor of eyewitness accuracy is the
response latency of the eyewitness in making a lineup identifi-
cation. We use ‘‘response latency’’ rather than ‘‘decision time,’’
because the former term incorporates both decisional and motor
components (Weber, Brewer, Wells, Semmler, & Keast, 2004).
The effect was first documented by Sporer (1992); considerable
data have accumulated showing that witnesses who make ac-
make inaccurate identifications (Dunning & Perretta, 2002;
Dunning & Stern, 1994; Smith, Lindsay, & Pryke, 2000; Smith,
Lindsay, Pryke, & Dysart, 2001; Sporer, 1993, 1994; Weber
et al., 2004). Sporer (1992) suggested that this occurs because
comparisons made to the target involve a large number of
common features between memory and the stimulus face,
thereby permitting a very fast decisionin recognizing the target.
Comparisons to an innocent lineup member, on the other hand,
involve fewer common features between memory and the stim-
ulus, thereby resulting in a slower decision. The potential
practical value of the negative relation between response la-
tency and identification accuracy is considerable because, un-
like confidence, response latency is a performance variable
rather than a self-report. And, unlike confidence, response la-
tency can be measured without the eyewitness’s awareness.
Furthermore, response latency and confidence are not fully re-
For response latency to be useful at the level of evaluating an
individual eyewitness, however, some criteria have to be set for
‘‘fast’’ and ‘‘slow.’’ How are police, prosecutors, judges, and
juries to know whether a given result (e.g., response latency of
20 seconds) was fast or slow and thus should be considered
accurate or inaccurate? Dunning and Perretta (2002) ap-
proached this problem by repeatedly selecting different re-
sponse latencies, examining the percentages correct above and
below each response latency, and calculating the obtained chi-
square values for each response latency. The response latency
that produced the greatest value was then considered to be the
best rule for deciding on the decision criterion. Using this ap-
proach, Dunning and Perretta found that a response latency of
10 to 12 seconds worked best across four different data sets.
Furthermore, the 10–12-second response latency was highly
discriminating—those who responded before the 10–12-second
who took longer than 10–12 seconds had a probability of ac-
curacy of approximately 50%. Dunning and Perretta called this
the ‘‘10–12 second rule.’’ The consistency of the 10–12-second
response latency data sets fits nicely with Dunning and Stern’s
(1994) notion of automatic versus deliberative processing in
eyewitness identification. They argued that automatic decision
processes (which are fast) are likely to be characteristic of ac-
curate eyewitnesses, while deliberative processes (which are
slower) ought to be more characteristic of inaccurate eyewit-
nesses. Furthermore, because automatic processes tend to be
uninfluenced by decision context, the speed of accurate iden-
tifications ought to be relatively stable across situations—
hence, the 10–12-second rule was proposed to be stable across
various circumstances and conditions.
More recent research, however, has shown that the 10–12-
second rule is not stable across variations in witnessing and
lineup conditions. Weber et al. (2004) found that the maximally
discriminating time ranged from 5 seconds to 29 seconds across
variations in conditions. Furthermore, eyewitnesses who re-
sponded faster than the optimal time boundaries did not show
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
in the 50 to 60% range rather than the 90% range found by
Dunning and Perretta (2002). Although the 10–12-second rule
does not appear to be stable, the fact that accurate identifica-
tions are made faster than inaccurate identifications is itself a
very reliable finding.
Self-Reported Decision Processes
Another potential postdictor of eyewitness-identification ac-
identifications tend to arise from making relative judgments in
decide who looks most like the target; Wells argued that an
absolute judgment (comparing the lineup member to memory)
would be more likely to be accurate. Consistent with this as-
sumption, Stern and Dunning (1994) found that eyewitnesses
who agreed with the statement ‘‘I compared the photos [in the
lineup] to each other to narrow the choices’’ were more likely to
have made a mistaken identification than were those who en-
dorsed the statement ‘‘I just recognized him, I cannot explain
why’’ or those who said the photo ‘‘popped out.’’ Similar results
have been reported by Smith et al. (2000), Smith et al. (2001),
Dunning and Stern (1994), and Lindsay and Bellinger (1999).
One of the problems with self-reported decision processes is
that, like eyewitness confidence, they are subject to distortion.
For instance, confirmatory post-identification feedback leads
eyewitnesses to be more likely to recall that the lineup photo
‘‘popped out’’ and less likely to report having made a relative
judgment (Wells & Bradfield, 1998). Furthermore, if eyewit-
nesses thought these kinds of self-reports would be used to as-
their answers accordingly.
Overall, it appears that postdiction has not been highly suc-
cessful for eyewitness identification. Indicators of confidence
measured at the time of the identification may have some di-
agnostic value with regard to accuracy, but feedback, prose-
be better to use procedures that help prevent mistaken identi-
fications from occurring in the first place than to try to detect
errors after the fact.
PROGRESS AND PROSPECTS
Eyewitness science has made considerable progress in recent
years in getting a number of jurisdictions in the United States to
improve their identification procedures and undertake training
in the cognitive interview. The state of New Jersey, for instance,
has adopted an entire package of reforms for how it conducts
lineups. These reforms are based explicitly on the eyewitness
literature and include the adoption of recommendations for
selecting lineup fillers, instructing eyewitnesses before the
lineup, using double-blind lineup administrators, using the
sequential procedure, and obtaining a confidence statement
from the eyewitness before external factors can influence the
person’s confidence. Other jurisdictions—including the states
of Wisconsin and North Carolina and the cities of Boston and
Minneapolis—have alsoadopted thesereforms. Ineach ofthese
jurisdictions, eyewitness scientists played a central role in ex-
plaining the literature and helping translate the findings into
practical reforms of eyewitness-identification procedures.
In many jurisdictions, eyewitness researchers have become
involved in training police investigators in eyewitness-identifi-
cation procedures or training the trainers. Increasingly, eye-
witness researchers are targeting some of their writings toward
law enforcement journals to more directly make the research
findings accessible to law enforcement (e.g., Turtle, Lindsay, &
Wells, 2003). Jury simulations have shown that mock jurors
respond more favorably to eyewitness-identification testimony
when it was obtained using these packages of reformed pro-
cedures than when procedures deviate from these reforms
(Lampinen, Judges, Odegard, & Hamilton, 2005). This is an
unusualimpact foralaboratory-basedpsychologicalscience. In
the years ahead, it is expected that these reforms will become
deeply ingrained in the legal system.
Despite this progress, we believe that research has only
scratched the surface of ways to help the legal system improve
the accuracy of eyewitness accounts. Thus far, almost all im-
provements to lineup procedure have been designed to reduce
the chances that an innocent suspect will be identified without
reducing identifications of the target. It has been more difficult
identify the target in target-present lineups. Both research ex-
periments and archival analyses of actual lineups suggest that
eyewitnesses fail to identify the target about 50% of the time.
This does not necessarily mean that the target walks away; in
some cases, other evidence is sufficient to charge or convict the
person. Nevertheless, there is room to improve these hit rates. It
seems likely that some failures toidentify the target from target-
present lineups are due at least in part to changes in the target’s
appearance. Specifically, the appearance of the target when the
eyewitness viewed the crime represents a moment in time. The
successful inincreasing accuracy;infact, they seem toincrease
errors (Charman & Wells, in press).
It could be argued that research has been profoundly con-
servative in its approach to the eyewitness-identification prob-
lem. Specifically, researchers have tended to operate within the
confines of the traditional lineup, in which a suspect is placed
among fillers and the eyewitness makes a verbal identification.
But what if the lineup had never existed and the legal system
Volume 7—Number 2
turned to psychology to determine how information could
obtaining detailed reports from witnesses—such as the cogni-
tive interview—do not appear to aid identification, but the
quality of witness descriptions could be improved though
innovative questioning procedures. This is an area in which
research is sparse, despite the potential to study the effective-
ness of various types of retrieval cues in eliciting descriptions
(Sporer, in press). The focus on target identification has also
a specific system or estimator variable on lineup performance,
instead of exploring relevant interactions. For example,
is the weapon-focus effect more pronounced when a witness has
a shorter exposure to the target, when the retention interval
is longer, and when the witness is making a cross-race
identification? Operating from scratch, it seems likely that
modern psychology would have developed radically different
ideas. For instance, brain-activity measures, eye movements,
rapid displays of faces, reaction times, and other methods for
studying memory might have been developed instead of the
traditional lineup. Once we step outside the confines of the
traditional lineup, it is possible to imagine a future science of
eyewitness evidence that isradicallydifferent from the methods
age and delay on recall in a cognitive or structured interview.
Psychology, Crime and Law, 9, 97–107.
Allport, G., & Postman, L. (1947). The psychology of rumor. New York:
Ayers, M.S., & Reder, L.M. (1998). A theoretical review of the misin-
formation effect: Predictions from an activation-based model.
Psychonomic Bulletin and Review, 5, 1–21.
Bekerian, D.A., & Dennett, J.L. (1997). Imagery effects in spoken and
written recall. In D. Payne & F. Conrad (Eds.), Intersections in
basic and applied memory research (pp. 175–191). Mahwah, NJ:
Behrman, B.W., & Davey, S.L. (2001). Eyewitness identification in
actual criminal cases: An archival analysis. Law and Human
Behavior, 25, 475–491.
Berman, G.L., & Cutler, B.L. (1996). Effects of inconsistencies in
eyewitness testimony on mock-juror decision making. Journal of
Applied Psychology, 81, 170–177.
Berresheim, A., & Weber, A. (2003). Die Strukturierte Zeugenver-
nehmung und Ihre Wirksamkeit [Structured witness interviewing
and its effectiveness]. Kriminalistik, 57, 757–771.
Binet, A. (1900). La suggestibilite ´ [On suggestibility]. Paris: Schleicher.
Boon, J., & Noon, E. (1994). Changing perspectives in cognitive in-
terviewing. Psychology, Crime and Law, 1, 59–69.
Borchard, E. (1932). Convicting the innocent: Errors of criminal justice.
New Haven, CT: Yale University Press.
it’’: The stability of the confidence-accuracy relationship
across domains. Journal of Experimental Psychology: Applied, 5,
Bower, G. (1967). A multicomponent theory of memory trace. In K.W.
Spence & J.T. Spence (Eds.), The psychology of learning
and motivation (Vol. 1, pp. 229–325). New York: Academic
Bradfield, A.L., & Wells, G.L. (2000). The perceived validity of eye-
witness identification testimony: A test of the five Biggers criteria.
Law and Human Behavior, 24, 581–594.
of confirming feedback on the relation between eyewitness cer-
tainty and identification accuracy. Journal of Applied Psychology,
Brainerd, C.J., & Reyna, V.F. (1998). Fuzzy-trace theory and children’s
false memories. Journal of Experimental Child Psychology, 71,
Brewer, N., Keast, A., & Rishworth, A. (2002). The confidence-
accuracy relationship in eyewitness identification: The effects
of reflection and disconfirmation on correlation and calibration.
Journal of Experimental Psychology: Applied, 8, 46–58.
In N. Brewer & K.D. Williams (Eds.), Psychology and law: An
empirical perspective (pp. 177–221). New York: Guilford.
in eyewitness identification: Effects of lineup instructions, filler
similarity and target-absent base rates. Journal of Experimental
Psychology: Applied, 12, 11–30.
Brigham, J.C. (1990). Target person distinctiveness and attractiveness
as moderator variables in the confidence-accuracy relationship in
eyewitness identifications. Basic &Applied Social Psychology, 11,
Brigham, J.C., & Bothwell, R.K. (1983). The ability of prospective ju-
rors to estimate the accuracy of eyewitness identifications. Law
and Human Behavior, 7, 19–30.
Brigham, J.C., Maas, A., Snyder, L.D., & Spaulding, K. (1982). Ac-
curacy of eyewitness identifications in a field setting. Journal of
Personality and Social Psychology, 42, 673–681.
Brigham, J.C., Meissner, C.A., & Wasserman, A.W. (1999). Applied
issues in the construction and expert assessment of photo lineups.
Applied Cognitive Psychology, 13, S73–S92.
Bruce, V., Ness, H., Hancock, P.J.B., Newman, C., & Rarity, J. (2002).
Four heads are better than one: Combining face composites yields
improvements infacelikeness. Journalof AppliedPsychology,87,
Bruck, M.,&Ceci,S.J.(1999).The suggestibilityofchildren’smemory.
Annual Review of Psychology, 50, 419–439.
Buckhout, R. (1974). Eyewitness testimony. Scientific American, 231,
Buckhout, R. (1980). Nearly 2,000 witnesses can be wrong. Bulletin of
the Psychonomic Society, 4, 191–192.
Bull, R., & Clifford, B.R. (1976). Identification: The Devlin Report.
New Scientist, 70, 307–308.
Charman, S.D., & Wells, G.L. (2006). Applied lineup theory. In R.C.L.
Lindsay, D.F. Ross, J.D. Read, & M.P. Toglia (Eds.), Handbook of
Charman, S.D., & Wells, G.L. (in press). Eyewitness lineups: Is the
appearance-change instruction a good idea? Law and Human
Clarke, C., & Milne, R. (2001). National Evaluation of the PEACE
(Report No. PRAS/149). London: Home Office.
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
Clark, S.E. (2003). A memory and decision model for eyewitness
identification. Applied Cognitive Psychology, 17, 629–654.
Clark, S.E. (2005). A re-examination of the effects of biased lineup
instructions in eyewitness identification. Law and Human Be-
havior, 29, 395–424.
Clark, S.E., & Davey, S.L. (2005). The target-to-fillers shift in simul-
taneous and sequential lineups. Law and Human Behavior, 29,
Clark, S.E., & Tunnicliff, J.L. (2001). Selecting lineup fillers in eye-
witness identification: Experimental control and real-world
simulation. Law and Human Behavior, 25, 199–216.
London: Routledge & Kegan Paul.
the number of perpetrators on eyewitness memory. Journal of
Applied Psychology, 66, 364–370.
Clifford, B.R., & Scott, J. (1978). Individual and situational factors in
Cole, G.F. (1986). The American system of criminal justice (4th ed.).
Monterey, CA: Brooks/Cole.
Connors, E.,Lundregan, T.,Miller,N.,&McEwan,T.(1996).Convicted
by juries, exonerated by science: Case studies in the use of DNA
evidence to establish innocence after trial. Alexandria, VA: Na-
tional Institute of Justice.
Cote, P. (1998). FACES: The Ultimate Composite Picture (Version 3.0)
[Computer software]. Saint Hubert, Quebec, Canada: Interquest.
the relation between eyewitness identification accuracy and con-
fidence. Journal of Applied Psychology, 74, 650–653.
Cutler, B.L., & Penrod, S.D. (1995). Mistaken identification: The eye-
witness,psychology,and thelaw.New York:Cambridge University
Cutler, B.L., Penrod, S.D., & Martens, T.K. (1987a). Improving the
reliability of eyewitness identifications: Putting context into con-
text. Journal of Applied Psychology, 72, 629–637.
Cutler, B.L., Penrod, S.D., & Martens, T.K. (1987b). The reliability of
eyewitness identifications: The role of system and estimator
variables. Law and Human Behavior, 11, 223–258.
eyewitness identification cases. Law and Human Behavior, 12,
Davies, G.M., Ellis, H.D., & Shepherd, J.W. (1978). Face recognition
accuracy as a function of mode of representation. Journal of Ap-
plied Psychology, 63, 180–187.
Davis, M., McMahon, M., & Greenwood, K. (2004). The efficacy of the
mnemonic components of the cognitive interview: Towards a
shortened variant for time-critical investigations. Applied Cogni-
tive Psychology, 19, 75–93.
Deffenbacher, K. (1980). Eyewitness accuracy and confidence: Can we
Deffenbacher, K.A., Bornstein, B.H., Penrod, S.D., & McGorty, E.K.
(2004). A meta-analytic review of the effects of high stress
on eyewitness memory. Law and Human Behavior, 28,
Devlin, L. (1976). Report to the Secretary of State for the Home De-
partment of the Departmental Committee on Evidence of Identifi-
Dixon, S., & Memon, A. (2005). The effect of post-identification feed-
back on the recall of crime and perpetrator details. Applied Cog-
nitive Psychology, 19, 935–951.
remembering. Journal of Police Science and Administration, 1,
Douglass, A.B., & Steblay, N. (in press). Memory distortion in eyewit-
nesses: A meta-analysis of the post-identification feedback effect.
Applied Cognitive Psychology.
Dunning, D., & Perretta, S. (2002). Automaticity and eyewitness ac-
curacy: A 10–12 second rule for distinguishing accurate from
inaccurate positive identifications.Journal ofApplied Psychology,
Dunning, D., & Stern, L.B. (1994). Distinguishing accurate from in-
accurate identifications via inquiries about decision processes.
Journal of Personality and Social Psychology, 67, 818–835.
Dysart, J.E., Lindsay, R.C.L., MacDonald, T.K., & Wicke, C. (2002).
The intoxicated witness: Effects of alcohol on identification
accuracy from showups. Journal of Applied Psychology, 87,
What do we really know? Retrieved October 12, 2006, from http://
Echterhoff, G., Hirst, W., & Hussy, W. (2005). How eyewitnesses resist
misinformation: Social postwarnings and the monitoring of mem-
ory characteristics. Memory & Cognition, 30, 770–782.
Ellis, H.D., Davies, G.M., & Shepherd, J.W. (1977). Experimental
studies of face identification. Journal of Criminal Defense, 3,
Ellis, H.D., Davies, G.M., & Shepherd, J.W. (1978). A critical exami-
nation of the Photo-fit system for recalling faces. Ergonomics, 21,
Ellis, H.D., Shepherd, J.W., Flin, R.H., Shepherd, J., & Davies, G.M.
(1989). Identification from a computer-driven retrieval system
compared with traditional mugshot album search: A new tool for
police investigations. Ergonomics, 32, 167–177.
Fisher, R.P. (2005). The cognitive interview: Question & answers. Ash-
burn, Virginia: The National Transportation Safety Board Acad-
emy. Retrieved June 13, 2006, from http://www.ntsb.gov/
Fisher, R.P., Brennan, K.H., & McCauley, M.R. (2002). The cognitive
interview method to enhance eyewitness recall. In M. Eisen, G.
Goodman, & J. Quas (Eds.), Memory and suggestibility in the
forensic interview (pp. 265–286). Mahwah, NJ: Erlbaum.
Fisher, R.P., & Geiselman, R.E. (1992). Memory enhancing techniques
for investigative interviewing: The cognitive interview. Springfield
IL: Charles C. Thomas.
Fisher, R.P., Geiselman, R.E., & Raymond, D.S. (1987). Critical
analysis of police interviewing techniques. Journal of Police Sci-
ence and Administration, 15, 177–185.
Fisher, R.P., Quigley, K.L., Brock, P., Chin, E., & Cutler, B.L. (1990,
March). The effectiveness of the cognitive interview in description
and identification tasks. Paper presented at the American Psy-
chology-Law Society, Williamsburg, VA.
Fisher, R.P., & Schreiber, N. (in press). Interviewing protocols to im-
prove eyewitness memory. In M. Toglia, R. Lindsay, D. Ross, &
J. Reed (Eds.), The handbook of eyewitness psychology: Vol. 1.
Memory for events. Mahwah, NJ: Erlbaum.
Fleet, M.L., Brigham, J.C., & Bothwell, R.K. (1987). The confidence-
accuracy relationship: The effects of confidence assessment and
choosing. Journal of Applied Social Psychology, 17, 171–187.
expert testimony and eyewitness confidence upon mock juror
judgment. Law and Human Behavior, 10, 215–228.
Volume 7—Number 2
Frank, J., & Frank, B. (1957). Not guilty. London: Gallanez.
Frowd, C.D., Carson, D., Ness, H., McQuiston, D., Richardson, J.,
Baldwin, H., & Hancock, P.J.B. (2005). Contemporary composite
techniques: The impact of a forensically relevant target delay.
Legal and Criminological Psychology, 10, 63–81.
Gabbert, F., Memon, A., & Wright, D.B. (in press). Memory conformity:
Disentangling the steps towards influence during a discussion.
Psychonomic Bulletin and Review.
Garrioch, L., & Brimacombe, C.A.E. (2001). Lineup administrators’
expectations: Their impact on eyewitness confidence. Law and
Human Behavior, 25, 299–314.
Geiselman, R.E., & Callot, R. (1990). Reverse versus forward recall of
script-based texts. Applied Cognitive Psychology, 4, 141–144.
Geiselman, R.E., Fisher, R.P., Firstenberg, I., Hutton, L.A., Sullivan,
S.J., Avetissian, I.V., & Prosk, A.L. (1984). Enhancement of
eyewitness memory: An empirical evaluation of the cognitive
interview. Journal of Police Science and Administration, 12,
Geiselman, R.E., Fisher, R.P., MacKinnon, D.P., & Holland, H.L.
(1985). Eyewitness memory enhancement in the police interview.
Journal of Applied Psychology, 27, 358–418.
George, R.C. (1991). A field evaluation of the cognitive interview. Un-
published master’s thesis, Polytechnic of East London.
Granhag, PA., & Stromwall, L. (2004). The detection of deception in
forensic contexts. Cambridge, England: Cambridge University
on recall, identification, confidence and the confidence/accuracy
relationship. Applied Cognitive Psychology, 11, 121–145.
Hafstad, G.S., Memon, A., & Logie, R. (2004). Post-identification
feedback, confidence, and recollections of witnessing condi-
tions in child witnesses. Applied Cognitive Psychology, 18,
Harley, E.M., Carlsen, K.A., & Loftus, G.R. (2004). The ‘‘saw-it-all-
along’’ effect: Demonstrations of visual hindsight bias. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 30,
Harris, M.J., & Rosenthal, R. (1985). Mediation of interpersonal ex-
pectancy effects: 31 meta-analyses. Psychological Bulletin, 97,
Hastorf, A.H., & Cantrill, H. (1954). They saw a game: A case study.
Journal of Abnormal and Social Psychology, 49, 129–134.
Hasel, L., & Wells, G.L. (in press). Catching the bad guy: Morphing
composite faces helps. Law and Human Behavior.
Haw, R.M., & Fisher, R.P. (2004). Effects of administrator-witness
contact on eyewitness identification accuracy. Journal of Applied
Psychology, 89, 1106–1112.
Holliday, R.E. (2003a). Reducing misinformation effects in children
with cognitive interviews: Dissociating recollection and familiar-
ity. Child Development, 74, 728–751.
Holliday, R.E. (2003b). The effect of a prior cognitive interview on
children’s acceptance of misinformation. Applied Cognitive Psy-
chology, 17, 443–457.
Hosch, H.M., & Cooper, D.S. (1982). Victimization as a determinant
of eyewitness accuracy. Journal of Applied Psychology, 67,
Hosch, H.M., Leippe, M.R., Marchioni, P.M., & Cooper, D.S. (1984).
Victimization, self-monitoring, and eyewitness identification.
Journal of Applied Psychology, 69, 280–288.
Huff, R., Rattner, A., & Sagarin, E. (1986). Guilty until proven inno-
cent. Crime and Delinquency, 32, 518–544.
Illinois Pilot Program on Double-Blind Sequential Procedures. (2006).
Retrieved March 21, 2006, from http://www.chicagopolice.org/
Juslin, P., Olsson, N., & Winman, A. (1996). Calibration and
diagnosticity of confidence in eyewitness identification: Com-
ments on what can and cannot be inferred from a low confidence-
accuracy correlation. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 5, 1304–1316.
Kassin, S.M., Tubb, V.A., Hosch, H.M., & Memon, A. (2001). On the
‘‘general acceptance’’ of eyewitness testimony research. American
Psychologist, 56, 405–416.
Kassin, S.M., & Gudjonsson, G.H. (2004). The psychology of confes-
sions: A review of the literature and issues. Psychological Science
in the Public Interest, 5, 33–67.
Kebbell, M., & Milne, R. (1998). Police officers’ perception of eye-
witness factors in forensic investigations. Journal of Social Psy-
chology, 138, 323–330.
Kebbell, M.R., Milne, R., & Wagstaff, G. (1999). The cognitive inter-
Law, 5, 101–115.
Kebbell, M.R., Milne, R., & Wagstaff, G. (2001). The cognitive inter-
view in forensic investigations: A review. In G.B. Traverson & L.
Bagnoli (Eds.), Psychology and law in a changing world: New
trends in theory, research and practice (pp. 185–198). London:
Klobuchar, A., Steblay, N., & Caligiuri, H. (in press). Improving eye-
witness identifications: Hennepin County’s blind sequential
lineup project. Cardozo Public Law, Policy, and Ethics Journal.
on the effects of the cognitive interview. Psychology, Crime and
Law, 5, 3–27.
Ko ¨hnken, G., Schimmossek, E., Aschermann, E., & Ho ¨fer, E. (1995).
The cognitive interview and the assessment of the credibility
of adults’ statements. Journal of Applied Psychology, 80,
Ko ¨hnken, G., Thurer, C., & Zorberbier, D. (1994). The cognitive
interview: Are interviewers’ memories enhanced too? Applied
Cognitive Psychology, 8, 13–24.
memory accuracy. Annual Review of Psychology, 51, 481–537.
Kovera, M.B., Penrod, S.D., Pappas, C., & Thill, D.L. (1997). Identi-
fication of computer-generated facial composites. Journal of
Applied Psychology, 82, 235–246.
Lampinen, J.M., Judges, D.P., Odegard, T.N., & Hamilton, S. (2005).
The reactions of mock jurors to the Department of Justice
Guidelines for the Collection and Preservation of Eye-
witness Evidence. Basic and Applied Social Psychology, 27,
Leippe, M.R. (1980). Effect of integrative memorial and cognitive
processes on the correspondence of eyewitness accuracy and
confidence. Law and Human Behavior, 4, 261–274.
Levi, A.M. (2002). Lineup size, the modified sequential lineup, and the
sequential lineup. Cognitive Technology, 7, 39–46.
Lindsay, D.S., & Johnson, M.K. (1989). The eyewitness suggestibility
effect and memory for source. Memory and Cognition, 17,
Lindsay, R.C.L. (1986). Confidence and accuracy in eyewitness iden-
tification from lineups. Law and Human Behavior, 10, 229–239.
Lindsay, R.C.L., & Bellinger, K. (1999). Alternatives to the sequential
lineup: The importance of controlling the pictures. Journal of
Applied Psychology, 84, 315–321.
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
Lindsay, R.C.L., Lim, R., Mirando, L., & Cully, D. (1986). Mock juror
evaluations of eyewitness testimony: A test of metamemory hy-
potheses. Journal of Applied Social Psychology, 16, 447–459.
Lindsay, R.C.L., Martin, R., & Webber, L. (1994). Default values in
eyewitness descriptions: A problem for the match-to-description
lineup filler selection strategy. Law and Human Behavior, 18,
Lindsay, R.C.L., Nosworthy, G.J., Martin, R., & Martynuck, C. (1994).
Finding suspects in mugshots. Journal of Applied Psychology, 79,
Lindsay, R.C.L., & Wells, G.L. (1980). What price justice? Exploring
the relationship between lineup fairness and identification ac-
curacy. Law and Human Behavior, 4, 303–314.
Lindsay, R.C.L., & Wells, G.L. (1985). Improving eyewitness identifi-
cation from lineups: Simultaneous versus sequential lineup pre-
sentations. Journal of Applied Psychology, 70, 556–564.
Lindsay, R.C.L., Wells, G.L., & O’Connor, F. (1989). Mock juror belief
of accurate and inaccurate eyewitnesses: A replication. Law and
Human Behavior, 13, 333–340.
Lindsay, R.C.L., Wells, G.L., & Rumpel, C. (1981). Can people detect
Journal of Applied Psychology, 66, 79–89.
Loftus, E.F. (1979). Eyewitness testimony. Cambridge, MA: Harvard
Loftus, E.F. (2005). A 30-year investigation of the malleability of
memory. Learning and Memory, 12, 361–366.
Loftus, E.F., & Palmer, J.C. (1974). Reconstruction of an automobile
destruction: An example of the interaction between language
and memory. Journal of Verbal Learning and Verbal Behavior, 13,
Luus, C.A.E., & Wells, G.L. (1991). Eyewitness identification and the
selection of distracters for lineups. Law and Human Behavior, 15,
Luus, C.A.E., & Wells, G.L. (1994). The malleability of eyewitness
confidence: Co-witness and perseverance effects. Journal of Ap-
plied Psychology, 79, 714–724.
Maas, A., Brigham, J., & West, S.G. (1985). Testifying on eyewitness
reliability: Expert advice is not always persuasive. Journal of
Applied Social Psychology, 15, 207–229.
MacLeod, M. (2002). Retrieval-induced forgetting in eyewitness
memory: Forgetting as a consequence of remembering. Applied
Cognitive Psychology, 16, 135–149.
MacLin, O.H., Zimmerman, L.A., & Malpass, R.S. (2005). PC_Eye-
witness and the sequential superiority effect: Computer-based
lineup administration. Law and Human Behavior, 29, 303–321.
MacLin, O., Wells, G.L., & Phelan, C. (2004). [Target removal-without-
replacement in lineups]. Unpublished raw data.
Malpass, R.S., & Devine, P.G. (1981). Eyewitness identification:
Lineup instructions and the absence of the offender. Journal of
Applied Psychology, 66, 482–489.
Malpass, R.S., & Lindsay, R.C.L. (1999). Measuring lineup fairness.
Applied Cognitive Psychology, 13(Suppl. 1), S1–S8.
Manson v. Braithwaite, 432 U.S. 98 (1977).
a psychologist tell a jury? American Psychologist, 38, 550–563.
McQuiston, D.E., & Malpass, R.S. (2002). Validity of the mock witness
McKelvie, S.J. (1988). The role of spectacles in facial memory:
A replication and extension. Perceptual and Motor Skills, 66,
Meissner, C.A., & Brigham, J.C. (2001a). Thirty years of investigating
the own-race bias in memory for faces: A meta-analytic review.
Psychology, Public Policy, and Law, 7, 3–35.
Meissner, C.A, & Brigham, J.C. (2001b). A meta-analysis of the verbal
overshadowing effect in face identification. Applied Cognitive
Psychology, 15, 603–616.
Meissner, C.A., Sporer, S.L., & Schooler, J.W. (in press). Person
descriptions as eyewitness evidence. In R.C.L. Lindsay,
psychology: Memory for people (pp. 1–34). Mahwah, NJ:
Meissner, C.A., Tredoux, C.J., Parker, J.F., & MacLin, O.H. (2005).
Eyewitness decisions in simultaneous and sequential lineups.
Memory and Cognition, 33, 783–792.
Mello, E.W., & Fisher, R.P. (1996). Enhancing older adult eyewitness
memory with the cognitive interview. Applied Cognitive Psychol-
ogy, 10, 403–418.
Memon, A. (2006). The cognitive interview. In O. Hargie (Ed.), A
Memon, A., & Bull, R. (Eds.). (1999). Handbook of the psychology of
interviewing. Chichester, England: Wiley.
Memon, A., Cronin, O´., Eaves, R., & Bull, R. (1993). The cognitive
interview and child interviewees. In G.M. Stephenson & N.K.
Clark (Eds.), Children, evidence and procedure: Issues in crimino-
logical & legal psychology. Division of Criminological and Legal
Memon, A., & Higham, P. (1999). A review of the cognitive interview.
Psychology, Crime and Law, 5, 177–196.
Memon, A., Hope, L., Bartlett, J., & Bull, R. (2002). Eyewitness
recognition errors: The effects of mugshot viewing and choos-
ing in young and old adults. Memory and Cognition, 30,
Memon, A., Hope, L., & Bull, R. (2003). Exposure duration: Effects on
eyewitness accuracy and confidence. British Journal of Psychol-
ogy, 94, 339–354.
Memon, A., Milne, R., Holley, A., Bull, R., & Ko ¨hnken, G. (1994).
Towards understanding the effects of interviewer training in
evaluating the cognitive interview. Applied Cognitive Psychology,
Memon, A., & Stevenage, S.V. (1996). Interviewing witnesses: What
works and what doesn’t? Psycholoquy 7(6), Witness Memory (1).
Retrieved October 20, 2006, from http://psycprints.ecs.soton.
Memon, A., Wark, L., Bull, R., & Ko ¨hnken, G. (1997). Isolating the
effects of the cognitive interview techniques. British Journal of
Psychology, 88, 179–198.
Memon, A., Wark, L., Holley, A., Bull, R., & Ko ¨hnken, G. (1997).
Eyewitness performance in cognitive and structured interviews.
Memory, 5, 639–655.
Milne, R., & Bull, R. (2006). Interviewing victims of crime, including
children andpeople withintellectual difficulties. InM.R.Kebbell
& G.M. Davies (Eds.), Practical psychology for forensic investi-
gations. Chichester, England: Wiley.
Milne, R., Bull, R., Memon, A., & Ko ¨hnken, G. (1995). The cognitive
interview and suggestibility. In G.M. Stephenson & N.K. Clark
(Eds.), Criminal behaviour: Perceptions, attributions and ration-
ality. Division of Criminological and Legal Psychology (Occa-
sional Paper No. 22). Leicester, England: British Psychological
Volume 7—Number 2
Milne, R., Clare, I.C.H., & Bull, R. (1999). Using the cognitive inter-
and Law, 5, 81–100.
Mitchell, K.J., & Johnson, M. (2000). Source monitoring: Attributing
mental experiences. In E. Tulving & F.I.M. Craik (Eds.), The
Oxford handbook of memory (pp. 179–195). New York: Oxford
Mitchell, K.J., Livosky, M., & Mather, M. (1998). The weapon focus
effect revisited: The role of novelty. Legal and Criminological
Psychology, 3, 287–303.
Morgan, C.A., Hazlett, G., Doran, A., Garrett, S., Hoyt, G., Thomas, P.,
Baronoski, M., & Southwick, S.M. (2004).Accuracy of eyewitness
stress. International Journal of Law and Psychiatry, 27, 265–279.
Mu ¨nsterberg, H. (1908). On the witness stand. New York: Doubleday.
Newsome v. McCabe et al., 256 F.3d 747 (2002).
Neuschatz, J.S., Preston, E.L., Burkett, A.D., Toglia, M.R., Lampinen,
J.M., Neuschatz, J.S., Fairless, A.H., Lawson, D.S., Powers, R.A.,
& Goodsell, C.A. (2005). The effects of post-identification feed-
back and age on retrospective eyewitness memory. Applied
Cognitive Psychology, 19, 435–453.
Patterson, K.E., & Baddeley, A.D. (1977). When face recognition fails.
Journal of Experimental Psychology: Human Learning and
Memory, 3, 406–417.
Pecher, D., Zeelenberg, R., & Barsalou, L.W. (2003). Verifying con-
ceptual properties in different modalities produces switching
costs. Psychological Science, 14, 119–124.
Penrod, S. (2003, Spring). Eyewitness identification evidence: How
well are witnesses and police performing? Criminal Justice, 54,
Perfect, T.J., & Hollins, T.S. (1996). Predictive feeling of knowing
judgments and postdictive confidence judgments in eyewitness
Pezdek, K., Blandon-Gitlin, I., & Moore, C. (2003). Children’s face
recognition memory: More evidence for the cross-race effect.
Journal of Applied Psychology, 88, 760–763.
Phillips, M.R., McAuliff, B.D., Kovera, M.B., & Cutler, B.L. (1999).
Double-blind photoarray administration as a safeguard against
investigator bias. Journal of Applied Psychology, 84, 940–951.
Pickel, K.L. (1998). Unusualness and threat as possible causes of
‘‘weapon focus.’’Memory, 6, 277–295.
Pickel, K.L. (1999). The influence of context on the ‘‘weapon focus’’
effect. Law and Human Behavior, 23, 299–311.
Pike, G.E., Kemp, R.I., Towell, N.A., & Phillips, K.C. (1997). Recog-
nizing moving faces: The relative contribution of motion
and perspective view information. Visual Cognition, 4, 409–437.
Poole, D., & Lamb, M. (1998). Investigative interviews with children: A
guide for helping professionals. Washington, DC: American Psy-
Poole, D., & Lindsay, D.S. (2001). Children’s eyewitness reports after
exposure to misinformation from parents. Journal of Experimental
Psychology: Applied, 7, 27–50.
Read, J.D., Vokey, J.R., & Hammersley, R. (1990). Changing photos of
faces: Effects of exposure duration and photo similarity on rec-
ognition and the accuracy-confidence relationship. Journal of
Experimental Psychology: Learning, Memory and Cognition, 16,
Read, J.D., Yuille, J.C., & Tollestrup, P. (1992). Recollections of a
robbery: Effects of arousal and alcohol upon recall and person
identification. Law and Human Behavior, 16, 425–446.
Roberts, W.T., & Higham, P.A. (2002). Selecting accurate statements
from the cognitive interview using confidence ratings. Journal of
Experimental Psychology: Applied, 8, 33–43.
Roberts, K., Lamb, M., & Sternberg, K. (2004). The effects of rapport
building style on children’s reports of a staged event. Applied
Cognitive Psychology, 18, 189–202.
Robinson, M.D., & Johnson, J.T. (1998). How not to enhance the con-
fidence-accuracy relation: The detrimental effects of attention
to the identification process. Law and Human Behavior, 22,
Savage, S., & Milne, R. (in press). Miscarriages of justice—the role of
the investigative process. In T. Newburn, T. Williamson, & A.
Wright (Eds.), Handbook of criminal investigation. Cullompton,
Saywitz, K.J., Geiselman, R.E., & Bornstein, G.K. (1992). Effects of
cognitive interviewing and practice on children’s recall perform-
ance. Journal of Applied Psychology, 77, 744–756.
Scheck, B., Neufeld, P., & Dwyer, J. (2000). Actual innocence. New
York: Random House.
of visual memories: Some things are better left unsaid. Cognitive
Psychology, 22, 36–71.
Searcy, J.H., Bartlett, J.C., & Memon, A. (1999). Age differences in
accuracy and choosing rates on face recognition and eyewitness
identification tasks. Memory and Cognition, 27, 538–552.
Searcy, J.H., Bartlett, J.C., Swanson, K., & Memon, A. (2001). Aging
and lineup performance at long retention intervals: Effects of
meta-memory and context reinstatement. Journal of Applied Psy-
chology, 86, 207–214.
Semmler, C., Brewer, N., & Wells, G.L. (2004). Effects of postidentifi-
cation feedback on eyewitness identification and nonidentifica-
tion. Journal of Applied Psychology, 89, 334–346.
studies. Psychological Bulletin, 100, 139–156.
Shaw, J.S., III, & McClure, K.A. (1996). Repeated postevent ques-
tioning can lead to elevated levels of eyewitness confidence. Law
and Human Behavior, 20, 629–654.
Shaw, J.S., III, & Skolnick, P. (1999). Weapon focus and gender dif-
of Applied Social Psychology, 29, 2328–2341.
may be associated with increases in eyewitness confidence. Law
and Human Behavior, 27, 315–329.
Shaw, J.S., III, Zerr, T.K., & Woythaler, K.A. (2001). Public eyewitness
confidence ratings can differ from those held privately. Law and
Human Behavior, 25, 141–154.
Police Research Group, Home Office.
In P. Morris & M. Grunerberg (Eds.), Aspects of memory: Theor-
etical aspects (2nd ed., pp. 169–195). New York: Routledge Press.
Smith, S.M., Lindsay, R.C.L., & Pryke, S. (2000). Postdictors of eye-
witness errors: Can false identification be diagnosed? Journal of
Applied Psychology, 85, 542–550.
Smith, S.M., Lindsay, R.C.L., Pryke, S., & Dysart, J.E. (2001). Post-
Smith, S.M., & Vela, E. (2001). Environmental context-dependent
memory: A review and meta-analysis. Psychonomic Bulletin and
Review, 8, 203–220.
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod
Snee, T., & Lush, D. (1941). Interaction of the narrative and inter-
rogatory methods of obtaining testimony. Journal of Psychology,
Sporer, S.L. (1982). A brief history of the psychology of testimony.
Current Psychological Reviews, 2, 323–339.
Sporer, S.L. (1992). Post-dicting eyewitness identification accuracy:
Confidence, decision-times and person descriptions among
choosers and nonchoosers. European Journal of Social Psycholo-
gy, 20, 157–180.
Sporer, S.L. (1993). Eyewitness identification accuracy, confidence,
and decision times in simultaneous and sequential lineups.
Journal of Applied Psychology, 78, 22–33.
Sporer, S.L. (1994). Decision-times and eyewitness identification ac-
curacy in simultaneous and sequential lineups. In D.F. Ross, J.D.
Read, & M.P. Toglia (Eds.), Adult eyewitness testimony: Current
trends and developments (pp. 300–327). New York: Cambridge
Sporer, S.L. (1996). Psychological aspects of person descriptions. In S.
Sporer, R. Malpass, & G. Ko ¨hnken (Eds.), Psychological issues in
eyewitness identification (pp. 53–86). Mahwah, NJ: Erlbaum.
Sporer, S.L. (in press). Person descriptions as retrieval cues: Do they
really help? Psychology, Crime and Law.
Sporer, S., Penrod, S., Read, D., & Cutler, B.L. (1995). Choosing,
confidence, and accuracy: A meta-analysis of the confidence-
accuracy relation in eyewitness identification studies. Psycho-
logical Bulletin, 118, 315–327.
State of Maryland v. Kirk N. Bloodsworth, 84-CR-3138 (Baltimore
Steblay, N.M. (1992). A meta-analytic review of the weapon focus
effect. Law and Human Behavior, 16, 413–424.
Steblay, N.M. (1997). Social influence in eyewitness recall: A meta-
analytic review of lineup instruction effects. Law and Human
Behavior, 21, 283–298.
Steblay, N.M., Dysart, J., Fulero, S., & Lindsay, R.C.L. (2001). Eye-
witness accuracy rates in sequential and simultaneous lineup
presentations: A meta-analytic comparison. Law and Human
Behavior, 25, 459–474.
Steele, C.M., & Josephs, R.A. (1990). Alcohol myopia: Its prized and
dangerous effects. American Psychologist, 45, 921–933.
Stein, L., & Memon, A. (in press). Testing the efficacy of the cognitive
interview in a developing country. Applied Cognitive Psychology.
Stern, L.B., & Dunning, D. (1994). Distinguishing accurate from in-
accurate eyewitness identifications: A reality monitoring ap-
proach. In D.F. Ross, J.D. Read, & M.P. Toglia (Eds.), Adult
eyewitness testimony: Current trends and developments (pp. 273–
299). New York: Cambridge University Press.
Stern, L.W. (1904). Die Aussage als geistige Leistung und als
Verho ¨rsprodukt [The statement as a mental achievement and
product of interrogation]. Beitra ¨ge zur Psychologie der Aussage, 3,
Stern, L.W. (1939). The psychology of testimony. Journal of Abnormal
and Social Psychology, 34, 3–20.
cognitive interview. Reply to Roberts. Psycholoquy, 8(3), Witness
Memory (16). Retrieved October 20, 2006, from http://psyc-
Stinson, V., Devenport, J.L., Cutler, B.L., & Kravitz, D.A. (1996).
How effective is the presence-of-counsel safeguard? Attorney
perceptions of suggestiveness, fairness and correctability
of biased lineup procedures. Journal of Applied Psychology, 81,
Stinson, V., Devenport, J.L., Cutler, B.L., & Kravitz, D.A. (1997). How
of suggestiveness and fairness of biased lineup procedures.
Journal of Applied Psychology, 82, 211–220.
Tanaka, J.W., & Farah, M.J. (1993). Parts and wholes in face recogni-
tion. Quarterly Journal of Experimental Psychology A, 46, 225–
face recognition. Memory and Cognition, 25, 583–592.
Tollestrup, P., Turtle, J., & Yuille, J. (1994). Actual victims and wit-
nesses to robbery and fraud: An archival analysis. In D.F. Ross,
J.D. Read, & M.P. Toglia (Eds.), Adult eyewitness testimony: Cur-
Tulving, E., & Thomson, D.M. (1973). Encoding specificity and
retrieval processes in episodic memory. Psychological Review, 80,
Turtle, J.W., Lindsay, R.C.L., & Wells, G.L. (2003). Best practice rec-
ommendations for eyewitness evidence procedures: New ideas for
the oldest way to solve a case. Canadian Journal of Police and
Security Services, 1, 5–18.
Valentine, T., & Heaton, P. (1999). An evaluation of the fairness of
police line-ups and video identifications. Applied Cognitive Psy-
chology, 13, S59–S72.
Valentine, T., Pickering, A., & Darling, S. (2003). Characteristics of
eyewitness identification that predict the outcome of real lineups.
Applied Cognitive Psychology, 17, 969–993.
Vrij, A. (2000). Detecting lies and deceit: The psychology of lying
Weber, N., & Brewer, N. (2003). The effect of judgment type and con-
fidence scale on confidence-accuracy calibration in face recog-
nition. Journal of Applied Psychology, 88, 490–499.
Weber, N., & Brewer, N. (2004). Confidence-accuracy calibration in
absolute and relative face recognition judgments. Journal of Ex-
perimental Psychology: Applied, 10, 156–172.
Weber, N., Brewer, N., Wells, G.L., Semmler, C., & Keast, A. (2004).
Eyewitness identification accuracy and response latency: The
unruly 10–12 second rule. Journal of Experimental Psychology:
Applied, 10, 139–147.
Wells, G.L. (1978). Applied eyewitness testimony research: System
variables and estimator variables. Journal of Personality and
Social Psychology, 36, 1546–1557.
Wells, G.L. (1984). The psychology of lineup identifications. Journal of
Applied Social Psychology, 14, 89–103.
Wells, G.L. (1988). Eyewitness identification: A system handbook. To-
ronto: Carswell Legal Publications.
Wells, G.L. (1993). What do we know about eyewitness identification?
American Psychologist, 48, 553–571.
Wells, G.L. (2006). Eyewitness identification: Systemic reforms. Wis-
consin Law Review, 2006, 615–643.
Wells, G.L., & Bradfield, A.L. (1998). ‘‘Good, you identified the
suspect’’: Feedback to eyewitnesses distorts their reports of
the witnessing experience. Journal of Applied Psychology, 83,
Wells, G.L., & Bradfield, A.L. (1999a). Distortions in eyewitnesses’
recollections: Can the postidentification feedback effect be mod-
erated? Psychological Science, 10, 138–144.
Wells, G.L., & Bradfield, A.L. (1999b). Measuring the goodness of
lineups: Parameter estimation, question effects, and limits to the
Volume 7—Number 2
Wells, G.L., Charman, S.D., & Olson, E.A. (2005). Building face Download full-text
composites can harm lineup identification performance. Journal
of Experimental Psychology: Applied, 11, 147–157.
of Applied Psychology, 66, 688–696.
Wells, G.L., & Hryciw, B. (1984). Memory for faces: Encoding and
retrieval operations. Memory and Cognition, 12, 338–344.
Wells, G.L., & Leippe, M.R. (1981). How do triers of fact infer the
accuracy of eyewitness identifications? Memory for peripheral
detail can be misleading. Journal of Applied Psychology, 66, 682–
Wells, G.L., Leippe, M.R., & Ostrom, T.M. (1979). Guidelines for em-
pirically assessing the fairness of a lineup. Law and Human
Behavior, 3, 285–293.
Wells, G.L., & Lindsay, R.C.L. (1980). On estimating the diagnosticity
Wells, G.L., Lindsay, R.C.L., & Ferguson, T.J. (1979). Accuracy, con-
fidence, and juror perceptions in eyewitness identification. Jour-
nal of Applied Psychology, 64, 440–448.
Wells, G.L., & Luus, E. (1990). Police lineups as experiments: Social
methodology as a framework for properly conducted lineups.
Personality and Social Psychology Bulletin, 16, 106–117.
Fulero, S. (2000). From the lab to the police station: A successful
application of eyewitness research. American Psychologist, 55,
Neil vs. Biggers criteria for judging eyewitness identification
accuracy? Journal of Applied Psychology, 68, 347–362.
Wells, G.L., & Murray, D.M. (1984). Eyewitness confidence. In G.L.
Wells & E.F. Loftus (Eds.), Eyewitness testimony: Psychological
perspectives (pp. 155–170). New York: Cambridge University
Wells, G.L., & Olson, E. (2002). Eyewitness identification: Information
gain from incriminating and exonerating behaviors. Journal of
Experimental Psychology: Applied, 8, 155–167.
confidence. Current Directions in Psychological Science, 11, 151–
Wells, G.L., Olson, E., & Charman, S. (2003). Distorted retrospective
eyewitness reports as functions of feedback and delay. Journal of
Experimental Psychology: Applied, 9, 42–52.
Wells, G.L., Rydell, S.M., & Seelau, E.P. (1993). On the selection of
distractors for eyewitness lineups. Journal of Applied Psychology,
Wells, G.L., Small, M., Penrod, S., Malpass, R.S., Fulero, S.M., &
Brimacombe, C.A.E. (1998). Eyewitness identification proced-
ures: Recommendations for lineups and photospreads. Law and
Human Behavior, 22, 603–647.
Wells, G.L., & Turtle, J.W. (1986). Eyewitness identification: The im-
portance of lineup models. Psychological Bulletin, 99, 320–329.
Westcott, H., Davies, G., & Bull, R. (2002). Children’s testimony: A
handbook of psychological research and forensic practice. Chi-
chester, England: Wiley.
Whipple, G.M. (1909). The obtaining of information: Psychology of
observation and report. Psychological Bulletin, 8, 419–421.
Whipple, G.M. (1910). Recent literature on the psychology of testi-
mony. Psychological Bulletin, 8, 365–368.
Whipple, G.M. (1911). The psychology of testimony. Psychological
Bulletin, 9, 307–309.
Whipple, G.M. (1912). Psychology of testimony and report. Psycho-
logical Bulletin, 10, 264–269.
Wigmore, J.H. (1909). Professor Mu ¨nsterburg and the psychology of
testimony. Illinois Law Review, 3, 399–445.
Wogalter, M.S., Marwitz, D.B., & Leonard, D.C. (1992). Suggestiveness
in photospread lineups: Similarity induces distinctiveness. Ap-
plied Cognitive Psychology, 6, 443–453.
Wright, D.B., Boyd, C.E., & Tredoux, C.G. (2003). Inter-racial contact
and the own-race bias for face recognition in South Africa and
England. Applied Cognitive Psychology, 17, 365–373.
Wright, D.B., & Loftus, E.F. (1998). How misinformation alters
memories. Journal of Experimental Child Psychology, 71, 155–
Wright, D.B., & McDaid, A.T. (1996). Comparing system and estimator
variables using data from real lineups. Applied Cognitive Psy-
chology, 10, 75–84.
Wright, D.B., & Skagerberg, E.M. (in press). Post-identification feed-
back affects real eyewitnesses. Psychological Science.
experiment. Psychology, Crime and Law, 10, 53–68.
Yuille, J., & Cutshall, J. (1986). A case study of eyewitnesses’ memory
of a crime. Journal of Applied Psychology, 71, 291–301.
Zaragoza, M.S., Payment, K.E., Ackil, J.K., Drivdahl, S.B., & Beck, M.
(2001). Interviewing witnesses: Forced confabulation and
confirmatory feedback increase false memories. Psychological
Science, 12, 473–477.
Volume 7—Number 2
Gary L. Wells, Amina Memon, and Steven D. Penrod