fail to ask the placebo patients about side - effects or improvements, or the
patients might infer from the testers ’ behaviours that they are in the placebo
condition. The effect of experimenters ’ knowledge and expectations on the
210
Handbook of Psychology of Investigative Interviewing
people they test has long been established scientifi cally (Harris & Rosenthal,
1985 ). Conducting a line - up is functionally analogous to conducting an exper-
iment with human participants (Wells & Luus, 1990 ). How might a line - up
administrator who knows who is the suspect and who are fi llers infl uence the
witness? The possibilities are almost endless, but the reader should keep in
mind that we are not suggesting that these infl uences are intentional, nor are
we suggesting that the line - up administrator or the witness is necessarily aware
of these infl uences. Consider the dynamics of the administrator – witness inter-
action. The administrator is very aware that she or he has placed the suspect
in position 3 and that positions 1, 2, 4, 5 and 6 are mere fi llers who could
not have committed the offence. The witness now looks at the line and says,
‘ Well … number 2 ’ , then pauses. A natural and understandable reaction of the
line - up administrator at that point might be, ‘ Now … take your time, don ’ t
be too quick ’ or, ‘ Be sure that you look at all the pictures. ’ At that point, any
witness with a modicum of intelligence realizes that number 2 is not the
suspect and will move on to another photo. Suppose, on the other hand, the
witness says, ‘ Well … number 3 ’ . To number 3 the reaction of the administra-
tor is likely to be quite different – for example, ‘ Tell me about number 3. ’ It
is not uncommon for eyewitnesses to waver between two or more line - up
members. This conversational process shapes the witness
’ s behaviour away
from fi llers and towards the suspect. Notice, however, that it is not the wit-
ness ’ s memory that is guiding the process, but the beliefs of the line
- up
administrator.
There are various ways in which the double - blind administration could be
put into practice, though essentially what is required is that the line - up admin-
istrator operates completely independently of the personnel responsible for
line - up construction and is unable to cue the witness in any way. The latter
is, of course, more readily achieved when the line - up presentation is computer-
ized or otherwise automated, thereby removing any interaction between line -
up administrator and witness during the conduct of the line - up. Criticisms
sometimes levelled at a requirement for double - blind line - up administration
include the associated increased resource demands and insuffi cient fl exibility
to accommodate the sometimes immediate and pressing needs of police seeking
to conduct a line - up.
It is important to note that, even when double - blind line - up administration
is used, opportunities for line
- up administrator infl uence may still exist.
Douglass, Smith & Fraser - Thill (2005) showed that a combination of sequen-
tial line - up administration (i.e., presenting line - up members one at a time) and
multiple eyewitnesses can result in line - up administrator infl uence. Douglass
et al. required participant line - up administrators to test, in succession, two
witnesses. The line - up administered was a perpetrator - absent line - up. The fi rst
witness (a confederate) picked the fi fth line - up member presented (an innocent
foil), and did so either quickly and confi dently or slowly and with low confi -
dence. They found that the second witness, who had actually witnessed the
crime and was genuinely attempting an identifi cation, was more likely to rep-
Obtaining and Interpreting Eyewitness Identifi cation Test Evidence 211
licate the fi rst witness ’ s choice if the latter had been slow and unconfi dent.
Douglass et al. (2005) suggested that the unconfi dent confederate ’ s behaviour
may have suggested to the administrator that the identifi cation task was a dif-
fi cult one, leading the administrator to impart some subtle cues to the second
witness to assist with their (diffi cult) decision.
The obvious practical implication of the Douglass et al. (2005) fi ndings is
that, when there is more than one witness to a crime involved in an identifi ca-
tion test, the conduct of the test should be carried out by separate administra-
tors who are not only blind to the suspect ’ s identity but also to the outcome
of any previous line - up conducted. In a series of studies currently underway
in the fi rst author ’ s laboratory, the pattern of fi ndings detected by Douglass
et al. has not been replicated consistently. Nevertheless, this does not rule out
the possibility that line - up administrator infl uence could not occur in double -
blind line
- ups under at least some conditions. Accordingly, until further
research clarifi es this issue, the above practical recommendation remains a
sensible one.
One possible solution to both the resource issue (i.e., requiring an additional
person to administer the line - up) and the problem of successive administration
by the double - blind administrator is to computerize the line - up. In fact, many
eyewitness research labs ’ eyewitnesses have used computers for many years to
administer photographic line - ups for which, in effect, the computer administers
the line
- up, delivers instructions and collects the witness
’ s responses. The
American Judicature Society ’ s Institute of Forensic Science and Public Policy in
North Carolina headed the development of such a program (called the ‘ laptop
line - up procedure ’ ), which is being used in some police departments.
Unbiased line - up instructions
A highly infl uential interaction between the line
- up administrator and the
witness occurs at the time the line - up administrator instructs the witness just
prior to viewing the line - up. Unbiased instructions explicitly advise the witness
that the perpetrator may or may not be in the line - up. Biased instructions fail
to include the second element of these instructions, namely, that the perpetra-
tor may not be present. Given the expectations that witnesses are likely
to bring to the identifi cation test, it is possible that, for many witnesses,
the delivery of unbiased instructions leads to the fi rst inkling that, while
there may be a police suspect in the line - up, the suspect may not be the
perpetrator.
That this is likely to be the case is dramatically illustrated by the effects of
varying the line - up instructions on witness choosing rates. Failure to warn a
witness that the perpetrator may not be in the line - up signifi cantly increases the
likelihood that the witness will make a choice from a perpetrator - absent line - up
(Malpass & Devine, 1981 ; Steblay, 1997 ; Brewer & Wells, 2006 ), thereby
increasing the possibility of a damning misidentifi cation of an innocent suspect.
The impact of instructional bias on choosing is also apparent for
target - present
212
Handbook of Psychology of Investigative Interviewing
line - ups. Biased instructions produce fewer line
- up rejections, with the
increased choosing leading to a greater likelihood of target or foil identifi ca-
tions (Clark, 2005 ; Brewer & Wells, 2006 ). These patterns have been demon-
strated for both adult and child witnesses (Keast, Brewer & Wells, 2007 ).
Biased instructions can be communicated in a variety of ways. Interestingly,
the mere absence of a warning instruction that says that the perpetrator may
or may not be in the line - up is itself generally considered by eyewitness sci-
entists to be a biased instruction (although it is technically a bias resulting
from non - instruction or the failure to instruct). In fact, most research studies
compare the unbiased instruction to no instruction. But there are even higher
levels of bias than simply not instructing the witness. One type of explicitly
biased instruction places pressure on the witness to think that not choosing
someone is a bad thing and that they are expected to identify someone. For
example, ‘ I am going to show you a line - up. Choose the person whom you
saw commit the offence. ’ This instruction can be construed by the witness as
saying that neither the ‘ not sure ’ option nor the ‘ not there ’ option is accept-
able. Or consider the instruction ‘ Are you able to tell me which of these is
the guy you saw that night? ’ Notice how such an instruction not only implies
that the witness is expected to choose someone, but also seems to imply to
the witness that this is a test of whether the witness is able, in the sense of
‘ capable ’ or ‘ reliable ’ . In other words, if you cannot identify someone, then
you are not able or not reliable, and hence a ‘ bad witness ’ . Obviously, any
explicitly biased instructions like these have to be avoided because it is desir-
able for uncertain witnesses to say that they are unsure rather than guess, and
it is desirable for witnesses to indicate that the perpetrator is not there if, in
fact, that is the case.
From a line - up administrator ’ s perspective, there is clearly some temptation
to use biased instructions as this may well increase the likelihood that the
police suspect is identifi ed. However, the dramatic infl ation of false identifi ca-
tions which has been so consistently demonstrated highlights the likely costs
for the delivery of justice. In sum, witnesses should receive a very clear
warning that the perpetrator may not be in the line - up.
Police – w itness i nteractions at the
i dentifi cation t est: the d ecision
While the following may seem so obvious as to be not worth saying, we empha-
size the following crucial points. The faithful recording of each eyewitness ’ s
decision at the identifi cation test constitutes an important part of the preserva-
tion of evidence. Quite simply, the line - up administrator should clearly record
each witness ’ s exact response. This recording should clearly distinguish between
response options such as (a) the witness identifi ed a particular line - up member;
(b) the witness indicated that the perpetrator was not present in the line - up;
Obtaining and Interpreting Eyewitness Identifi cation Test Evidence 213
(c) the witness indicated that he or she was not sure enough to make an iden-
tifi cation; (d) the witness indicated that it could be number 4 or number 6; or
(e) the witness indicated that number 3 looks a lot like the perpetrator. Each of
these response options has different implications for assessments of the likeli-
hood that the police suspect in the line - up is or is not the perpetrator. Yet a
failure to record the exact response of each witness may, for example, lead to
some witnesses ’ decisions (e.g., responses (b) and (e)) not being preserved for
tendering in any subsequent trial, or perhaps to a ‘ transformation ’ of the wit-
nesses ’ responses between the identifi cation test and the trial (e.g., response (c)
may transform into response (a)). In sum, the failure to record carefully each
witness ’ s decision can have far - reaching practical implications for the overall
nature and quality of evidence that may be tendered at trial.
Police – w itness i nteractions at the i dentifi cation t est:
p ost - d ecision i nfl uences
After the witness has indicated his or her identifi cation decision (i.e., chosen
a line - up member, indicated that the perpetrator is not present or perhaps that
the witness is just not sure enough to make a decision), there are opportunities
for a whole new set of interactions that are now known to be of considerable
forensic relevance. Some of the most important of these are associated with
soliciting an expression of confi dence in the identifi cation decision from the
eyewitnesses. Here we consider issues such as why this type of information
may prove to be important in any particular case, what it suggests about the
likely accuracy of the identifi cation decision, how it should be collected in
order to maximize its informational value, and what use of this information
should be made in the courtroom. We also look at how these interactions can
infl uence other witness judgements about the witnessed event.
The r elationship b etween i dentifi cation
c onfi dence and a ccuracy
It is a common occurrence for people either to express spontaneously their
confi dence in the judgements that they have made or to be asked to do so.
Although most people probably do not accept that judgemental confi dence
necessarily equates with judgemental accuracy, the existence of at least a rea-
sonably close correspondence between confi dence and accuracy is likely to
align with people
’ s intuitions. It is not surprising, therefore, that police,
lawyers, judges and jurors are interested in knowing about witnesses ’ confi -
dence in their identifi cation decisions. Nor is it surprising to know that there
is ample evidence demonstrating that these groups fi nd an eyewitness ’ s con-
fi dence persuasive with respect to the likely accuracy of his or her testimony
214
Handbook of Psychology of Investigative Interviewing
(Cutler, Penrod & Stuve, 1988 ; Lindsay, Wells & O ’ Connor, 1989 ; Potter &
Brewer, 1999 ; Bradfi eld & Wells, 2000 ; Brewer & Burke, 2002 ).
If a witness ’ s confi dence in an identifi cation decision is likely to be inter-
preted as a strong pointer to identifi cation accuracy, it is important to consider
whether this interpretation is justifi ed. This issue has been a controversial one
in the psychology – law fi eld, with eyewitness researchers typically presenting
quite a different perspective from that which characterizes many people in the
criminal justice community. Specifi cally, many of the former group have main-
tained that confi dence in an identifi cation provides no useful guide to the
accuracy of that identifi cation. The focus here is not on the nature of these
differences, although it is worth noting that the specifi c approach used to
examine the relationship has made an important contribution to the different
<
br /> perspectives (see Brewer,
2006 ). Rather, we shall outline what we believe
(based on current knowledge) to be some reasonable generalizations about
the characteristics of the confi dence – accuracy (CA) relationship for eyewitness
identifi cation, and spell out precisely what the implications are for the inter-
pretation of witnesses
’ expressions of confi dence by police investigators,
lawyers, judges and jurors, and line
- up administrators
’ interactions with
witnesses.
Detailed examinations of the CA relationship (e.g., Sporer, Penrod, Read
& Cutler, 1995 ; Juslin, Olsson & Winman, 1996 ; Lindsay, Read & Sharma,
1998 ; Wells & Bradfi eld, 1998; 1999 ; Brewer, 2006 ; Brewer & Wells, 2006 ;
Keast, Brewer & Wells, 2007 ) suggest the following generalizations are appro-
priate. First, identifi cation confi dence expressed well after the identifi cation
(e.g., in court) should be considered uninformative (we return to consider this
issue in detail in the next section of this chapter). Second, witnesses who
express high confi dence immediately after making the identifi cation are by no
means guaranteed to have made an accurate decision; the CA relation is likely
to be characterized by some degree of overconfi dence, though there may be
exceptions. Third, for adult witnesses who made a positive identifi cation, CA
calibration data indicate that an immediately recorded confi dence estimate
does provide a guide to likely identifi cation accuracy. This conclusion does
not, however, hold for non - choosers; nor does it apply to identifi cations made
by children, at least for those in the 10 – 12 year age range.
There are several important implications of these fi ndings. First, the line - up
administrator should record the witness
’ s confi dence assessment imme-
diately after the identifi cation and, as will become clear shortly, this assess-
ment should be provided independently by the witness. Second, no matter
how confi dent the witness may be in the identifi cation decision, police inves-
tigators should not assume the identifi cation is accurate. Rather, a very
confi dent identifi cation made by an adult (but not a child) witness should
suggest to investigators that their suspect is at least a plausible one and a
Handbook of Psychology of Investigative Interviewing: Current Developments and Future Directions Page 41