by witnesses. Indeed, logic dictates that essentially all of the putative ‘ loss ’ in
correct ‘ identifi cations ’ in sequential line - ups (47% vs. 35% in the studies con-
sidered here) is in the form of witnesses who guess the suspect in target - present
arrays ( ‘ guess ’ insofar as their memory is not strong enough to avoid a fi ller
identifi cation with the perpetrator is not present). Given that not more than
Table 15.2: Steblay et al . meta - analysis of
simultaneous presentations
Target
Target
Present
Absent
(N = 38)
(N = 38)
No Choice
28%
45%
Identifi cation
47%
–
Foil Choice
25%
55%
270
Handbook of Psychology of Investigative Interviewing
25% of witnesses seem capable of getting it right in both target - present and
target - absent arrays (the best estimate of doubly - reliable witnesses is 21% if
one also considers the 17% doubly - accurate rate in simultaneous arrays), we
might well conclude that the 47% ‘ correct identifi cation ’ rate in simultaneous
target - present arrays is mostly lucky guessing by witnesses (around 26% of the
47%) confronted with biased arrays; so, it appears that most of those witnesses
would pick a fi ller in a target - absent array, and, of course, those witnesses
would be most likely to pick the similar - looking designated suspect in a simi-
larly biased target - absent array rather than another fi ller.
In contrast, sequential arrays yield signifi cantly fewer lucky guesses (10% of
the 35% of ‘ correct identifi cations ’ in target - present arrays), substantially fewer
erroneous fi ller choices (41% vs. 55%) and fewer designated innocent suspect
choices (12% vs. 23%, although those results come from a somewhat different
set of studies). The overall ratio of correct identifi cations to innocent suspect
identifi cations may be far superior for sequential presentations (35/12 – about
3 : 1) than for simultaneous (47/23 – about 2 : 1).
Although some policy - makers may fret over the loss in accurate ‘ identifi ca-
tions ’ (we characterize them as lucky guesses) that can result from the sequen-
tial line - up, it should be noted that the odds of an identifi cation of the suspect
being accurate are substantially increased by the use of the sequential line - up
in spite of some loss of identifi cations. In addition, it seems clear that policy -
makers should not always favour a method of conducting line - ups merely
because it yields more ‘ hits ’ . Consider, for instance, a method in which wit-
nesses who claim that they do not recognize anyone are instructed to guess.
This guessing method would certainly yield more suspect identifi cations than
would a method that discouraged guessing, but it would also yield many more
innocent suspect identifi cations. Under most circumstances policy
- makers
surely would not want to encourage a method that relies on guessing and
generates more errors. We say ‘ most circumstances ’ because, as Penrod (2003)
has noted, one can imagine a system in which investigators were forced to
choose (prior to any identifi cation test) how they were going to employ eye-
witnesses. If they intended to use eyewitnesses and their identifi cations in court
as evidence against a defendant, they might be pressed to use stringent pro-
cedures which maximize the diagnostic value of identifi cations (that is, increase
the ratio of correct to incorrect suspect identifi cations). If, on the other hand,
investigators needed the witness to assist them in generating investigative clues,
they might wish to use procedures that encourage guessing and the generation
of cues, but would be forced to do so at the cost of losing the witness as a
source of ‘ identifi cation ’ evidence at trial.
Double - b lind vs. s ingle - b lind l ine - u p a dministration
Although the NIJ guidelines stopped short of recommending double - blind
line - up administration, in part due to concerns from law enforcement about
Recent Developments in Identifi cation Science and Practice
271
the feasibility of fi nding suffi cient administrators in small jurisdictions, the
AP - LS white paper recommended double - blind line - up administration as best
practice. What evidence is there that double
- blind line
- up administration
reduces mistaken identifi cations or wrongful conviction? At the time these
recommendations were made there was little direct evidence that the line - up
administrator ’ s knowledge of the suspect ’ s identity would affect the reliability
of an eyewitness ’ s identifi cation; however, there was initial support for the idea
that administrator feedback about whether the witness had identifi ed the
suspect infl uenced witness confi dence (Wells & Bradfi eld, 1998 ). Instead of
basing this recommendation on the results of eyewitness research, scholars
argued, based on the body of research on experimenter expectancy effects
(e.g., Rosenthal, 1976; 2002 ), that administrator knowledge of a suspect ’ s
identity would lead the administrator to communicate behavioural cues that
would infl uence witnesses ’ identifi cation decisions and their reported confi -
dence. Since that time, a signifi cant body of research has examined the role of
administrator feedback in the malleability of witness confi dence and a smaller
number of studies have examined the role of administrator knowledge on the
accuracy of witness identifi cations.
Confi dence m alleability
When a line - up administrator knows which line - up member is the suspect,
there is the potential for the administrator to provide witnesses with feedback
about whether they have identifi ed the suspect. If this feedback is provided
prior to the witnesses expressing their confi dence in the accuracy of their
identifi cation, it is likely that the feedback will infl uence their reported confi -
dence in predictable ways. In one of the earliest studies of this phenomenon,
pairs of witnesses viewed a staged theft (Luus & Wells, 1994 ). The witnesses
then separately viewed a line - up that did not contain the suspect; almost all
of them identifi ed a member of the line - up as the perpetrator. Subsequently,
eyewitnesses were given one of nine forms of feedback about the identifi cation
decision made by their co - witness and were interviewed by a confederate police
offi cer who recorded their confi dence in the accuracy of their identifi cation.
Witnesses who learned that their co - witness had identifi ed the same line - up
member were more confi dent in the accuracy of their identifi cation than were
witnesses who received no feedback. Witnesses who learned that their co -
witness had identifi ed a different line - up member or rejected the line - up were
less confi dent than witnesses who received no feedback.
Gary Wells and his colleagues have conducted a number of studies demon-
strating that feedback from a line
- up administrator infl uences witness
confi dence (Wells & Bradfi eld, 1998 ; Bradfi eld, Wells & Olson, 2002 ; Wells,
Olson & Charman, 2003 ). When line - up administrators provide confi rming
feedback (e.g., ‘ Good, you identifi ed the suspect ’ ), witnesses were much more
confi dent than witnesses who received disconfi rming or no feedback (Wells &
Bradfi eld, 1998 ). This effect holds even when witnesses are instructed that the
272
Handbook of Psychology of Investigative Interviewing
perpetrator may not be present in the line - up (Semmler, Brewer & Wells,
2004 ). Not only does confi rming feedback from the administrator increase
witness confi dence, it also causes witnesses to report that they had a better
view of the culprit and paid more attention to the culprit ’ s face when observ-
ing the event (Wells & Bradfi eld, 1998 ).
A meta - analysis of 20 tests of the post - identifi cation feedback effect on
witness confi dence, based on data from over 2,400 participants, demonstrated
that the effect is robust and relatively large (Douglass & Steblay, 2006 ). In
addition, confi rming feedback had a signifi cant impact on a variety of objective
and subjective measures of the witnesses ’ experience, including how good a
view they had of the perpetrator, their opportunity to view his or her face,
how much attention they paid, whether they had a good basis to make an
identifi cation, how easy it was to make an identifi cation, how quickly they
made an identifi cation, their willingness to testify, their memory for strangers
and the clarity of their memory. These fi ndings support the recommenda-
tions that witness confi dence be collected immediately after an identifi cation
before any feedback about the identifi cation decision is provided by the
administrator.
Investigator k nowledge e ffects on e yewitness a ccuracy
Despite the recommendation of the AP - LS subcommittee that line - ups be
conducted by a blind administrator, only a handful of studies have directly
examined the effects of administrator knowledge on the reliability of eyewit-
ness identifi cation decisions. These studies have produced mixed results in
terms of the conditions under which the effects of administrator knowledge
of the suspect ’ s identity are obtained (Phillips, McAuliff, Kovera & Cutler,
1999 ; Haw & Fisher, 2004 ; Greathouse & Kovera, 2009 ) and whether the
effect is obtained at all (Haw, Mitchell & Wells, 2003 ; Russano et al ., 2006 ).
The fi rst study to examine the infl uence of administrator knowledge empiri-
cally paired mock line - up administrators with mock witnesses who had previ-
ously viewed a live, staged crime involving two perpetrators (Phillips et al .,
1999 ). Phillips and colleagues manipulated whether the administrator knew
the identity of the suspect, the type of line - up presented (simultaneous vs.
sequential), as well as the presence of an observer during the line - up task.
When the line - up was presented sequentially and an experimenter - observer
was present, administrator knowledge infl uenced witnesses to choose an inno-
cent suspect.
Additional support for the effects of line - up administrator knowledge was
obtained in research manipulating the level of contact between administrators
and witnesses (Haw & Fisher, 2004 ). For some witnesses, administrators had
direct contact with them while administering the line - up. For others, admin-
istrators sat behind the witnesses out of their direct view, limiting their ability
to communicate cues to the suspect ’ s identity to the witnesses. When the
administrator was permitted high contact with the witness and presented a
Recent Developments in Identifi cation Science and Practice
273
target - absent simultaneous line - up, witnesses were more likely to identify the
innocent suspect. In contrast, other studies have failed to fi nd any infl uence
of administrator knowledge on witnesses ’ identifi cation decisions (Haw et al .,
2003 ; Russano et al ., 2006 ).
Given the variability in the infl uence of administrator knowledge on eyewit-
ness decisions across studies, it is likely that previously unidentifi ed variables
moderate its effects. Greathouse & Kovera (in press) hypothesized that factors
which lower the criteria that witnesses use to make a positive identifi cation,
such as the presence of biased line - up instructions or simultaneous line - up
presentation, may moderate the effects of administrator knowledge. If a witness
is presented with biased instructions (e.g., instructions that fail to remind the
witness that the perpetrator may not be in the line - up), they may cue unsure
witnesses to attend more to the investigator ’ s behaviour, allowing single - blind
administrators to wield more infl uence. To test this hypothesis, Greathouse &
Kovera manipulated whether the administrator had knowledge of the suspect ’ s
identity, the type of line - up (simultaneous vs. sequential), the presence of the
actual perpetrator in the line - up and the type of line - up instructions (biased
vs. unbiased). In an attempt to ensure that the fi ndings were not specifi c to a
particular perpetrator or line - up composition, witnesses viewed one of two
perpetrators committing a theft and then made an identifi cation from a target -
present or a target - absent line - up constructed for that perpetrator.
When the witnesses received biased instructions and simultaneous line - ups,
they were more likely to make suspect identifi cations in single - blind than in
double - blind line - ups ( ibid. ). The pattern of fi ller and suspect identifi cations
in both double - blind and single - blind conditions suggests that this increase in
mistaken identifi cations was the result of single - blind administrators infl uenc-
ing those who would have, under double - blind conditions, made fi ller identi-
fi cations to make suspect identifi cations instead. This fi nding could be
conceptualized as shifting unsure witnesses ’ guesses from fi ller identifi cations
in the double - blind condition to suspect identifi cations in the single - blind
condition. Line - up rejections did not signifi cantly increase or decrease as a
function of line - up administrator knowledge. Moreover, suspect identifi cations
were twice as diagnostic for double - blind administrations as they were for
single - blind administrations.
Thus, it appears from this research that double - blind line - up administration
would protect against mistaken identifi cations, especially when witnesses ’ cri-
teria for choosing are lowered; however, double - blind line - up administration
also appears to decrease correct identifi cations. Moreover, it is not clear that
even double - blind administration can protect against the administrator expec-
tancy effect when the administrator conducts line - ups with multiple witnesses.
There is evidence to suggest that the confi dence of the witness from the fi rst
line - up administration infl uences the administrator ’ s beliefs about the diffi culty
of the identifi cation task. As a result, double - blind administrators who admin-
istered a second line - up to a different witness were more likely to steer the
&nb
sp; second witness to the photo chosen by the fi rst witness when the fi rst witness
274
Handbook of Psychology of Investigative Interviewing
lacked confi dence in his or her identifi cation than when the witness was con-
fi dent in their decision (Douglass, Smith & Fraser - Thill, 2005 ). Although
double - blind line - up administration may not correct all problems associated
with administrator expectations and their infl uence on witness behaviour, there
is little evidence that it would produce harmful results. It is clear, however,
that much more research is needed to understand how the infl uence of admin-
istrator knowledge of the suspect ’ s identity operates in eyewitness identifi ca-
tion tasks.
Eyewitness e rror r ates in fi eld s tudies u sing a lternative
i dentifi cation p rocedures
The studies just reviewed were experiments – sometimes staged events, some-
times videotaped events in which the researchers knew the identity of the
targets. That knowledge permits us to ascertain precisely when a witness has
made an error. In studies of actual witnesses using police archives it is generally
not known whether the suspect is the actual perpetrator, but it is still possible
to gauge inaccurate identifi cation rates by looking at fi ller identifi cations. For
reasons that will be explained, the best archival studies come from the UK.
These studies include Slater ’ s (1994) examination of identifi cation attempts
by 843 British witnesses who viewed 302 suspects. Slater found suspect, fi ller
and no identifi cation rates of 36%, 22% and 42% respectively. Wright
&
McDaid (1996) examined identifi cation attempts of 1,561 British witnesses
who viewed 616 suspects in live line - ups and found suspect, fi ller and no
identifi cation rates of 39%, 20% and 41% respectively. Pike, Brace & Kynan
(2002) reported a 49% suspect identifi cation rate in 8,800 line - ups, a 39%
suspect identifi cation rate in a separate study of a video presentation system
(940 witnesses) and a 35% suspect identifi cation rate in 1,635 live line - ups (no
information was provided on fi ller identifi cation rates). Valentine, Pickering &
Darling (2003) reported the results of identifi cation attempts by 584 witnesses
Handbook of Psychology of Investigative Interviewing: Current Developments and Future Directions Page 52