Handbook of Psychology of Investigative Interviewing: Current Developments and Future Directions
Page 51
each fi ller violated the participant ’ s description on at least one major feature
(e.g., the fi ller had black hair but the participant described the thief as having
blond hair). In the match - to - suspect condition, experimenters chose the fi llers
that most closely resembled the suspect. In the match - to - description condi-
tion, researchers discarded all the fi llers that failed to match the witnesses ’
descriptions. From those photos that remained, the experimenters chose those
that least resembled the suspect. Witnesses made more correct identifi cations
from line - ups constructed using the match - to - description strategy than from
the match - to - suspect strategy and made fewer mistaken identifi cations when
line - up fi llers were matched to description rather than mismatched. The match -
to - description strategy also produced fewer fi ller identifi cations and more
Recent Developments in Identifi cation Science and Practice
265
correct rejections. Other research, however, has not yielded an advantage for
match - to - description (see, e.g., Tunnicliff & Clark, 2000 ; Darling, Valentine
& Memon, 2008 ).
Line - u p i nstructions to the w itness
One recommendation that was made in both the AP - LS white paper and the
NIJ guidelines was that witnesses should be instructed that the person who
committed the crime may or may not be in the line - up. The purpose of this
instruction is to reduce the pressure that witnesses may otherwise feel to
choose someone from the line - up, even when they are not sure whether that
person is the culprit. The AP - LS guidelines included an additional instruction
to witnesses that the line
- up administrator does not know which line
- up
member is the suspect. This instruction is necessarily absent from the NIJ
guidelines, given that they do not include a recommendation that line - ups be
conducted by administrators blind to the suspect ’ s identity. We know of no
published empirical research testing whether this instruction infl uences the
accuracy of witnesses ’ identifi cations; however, there is a fairly substantial lit-
erature addressing the effects of warning the witness that the culprit may not
be present in the line - up.
A meta - analysis of studies that manipulated whether line - up instructions
were biased (e.g., that implied that the witness ’ s task was to identify the per-
petrator from the line - up) or unbiased (e.g., that reminded witnesses that the
culprit might not be in the line - up) showed that participants who received
unbiased instructions were more likely than participants who received biased
instructions to reject a line - up if the perpetrator was absent (Steblay, 1997 ).
There was no effect of instruction type on the rate of correct identifi cations.
It was the case, however, that more people failed to make a choice from a
target - present line - up when the instructions reminded witnesses that the per-
petrator might not be present than when the instructions contained such
reminder.
These fi ndings – that a reminder that the culprit might not be in the line - up
decreases mistaken identifi cations while leaving the rate of correct identifi ca-
tions unchanged
– led some police departments to adopt the instruction.
However, a recent reanalysis of the data from the Steblay meta - analysis by
Steve Clark (2005) suggested that this asymmetry in the infl uence of biased
instructions in target - absent vs. target - present line - ups may not exist. First, he
argued that if one assumes that biased instructions increase false identifi cations
by shifting downward the criterion that witnesses use to make a choice of a
line - up member, then that criterion shift should also produce an increase in
correct identifi cations. However, it is more diffi cult to demonstrate the infl u-
ence of biased instructions in target - present line - ups than target - absent line -
ups because increasing guessing through biased line - up instructions would
produce asymmetrical effects in the two types of line - ups. In target - absent
line - ups, for every new identifi cation produced by guessing, there will be an
266
Handbook of Psychology of Investigative Interviewing
equivalent drop in the correct rejection rate. In target - present line - ups, only
a fraction (1/k, where k = the number of line - up members) of the identifi ca-
tions produced by guessing, namely the identifi cations of the target, will result
in an increase in the correct identifi cation rate.
Next, Clark scrutinized the studies including in the original Steblay meta -
analysis (1997). Steblay reported that six studies found that biased instructions
increased correct identifi cations, six studies found that they decreased correct
identifi cations and three studies found no effect of biased instructions. Clark ’ s
reanalysis of these studies demonstrated methodological problems (e.g., ceiling
effects in the unbiased instruction conditions) in several studies and he argued
for a different categorization of some conditions into either biased or unbiased
instructional conditions in other studies. The results of this reanalysis question
Steblay ’ s fi nding that biased instructions have a negligible effect on correct
identifi cations. It appears that the size of the biased instruction effect in target -
present line - ups is infl uenced by the correct identifi cation rate in unbiased
instruction conditions. When that rate is high, ceiling effects prevent the biased
instructions from signifi cantly increasing correct identifi cations. Further
research should resolve the issue.
In addition to recommending that witnesses be instructed that the perpe-
trator may not be present in the line - up, the NIJ guidelines recommended
that police include an instruction as part of their standard eyewitness identifi -
cation procedure about the possibility that perpetrators have changed their
appearance (Technical Working Group for Eyewitness Evidence, 1999 ). The
guidelines provide the specifi c instruction to witnesses: ‘ Individuals depicted
in line - up photos may not appear exactly as they did on the date of the inci-
dent because features such as head and facial hair are subject to change ’ ( ibid .,
p. 32). Charman & Wells (2007) questioned the guidelines ’ recommendation
of the appearance - change instruction without theoretical and empirical evi-
dence for its use. Among the possible hypotheses, they posited that the
appearance - change instruction may be problematic if witnesses infer that per-
petrators may no longer look like their memory of them and, therefore, reduce
their criterion level for identifi cation. In the only empirical examination of the
effects of appearance - change instruction (as opposed to studies of the effects
of appearance change itself), each participant witness was presented with
four line - ups, two of which were target - absent and two target - present. All
participants were instructed that the perpetrator may or may not be in the
line - up, and half of the participants additionally received the appearance
-
change instruction.
The appearance - change instruction increased fi ller identifi cations in target -
present line - ups and target - absent line - ups, but did not signifi cantly increase
correct identifi cations in target - present line - ups. Choosing rates were signifi -
cantly higher in the appearance - change instruction conditions, supporting the
hypothesis that the instruction results in witnesses lowering their response
criterion. Response latency to make an identifi cation decision was signifi cantly
longer for witnesses receiving the instruction, perhaps due to the lower crite-
Recent Developments in Identifi cation Science and Practice
267
rion increasing the number of photos witnesses needed to consider before
making their decision (Charman & Wells, 2007 ). Although it would be pre-
mature to make policy recommendations based on a single study, the results
of this initial test of the appearance - change instruction suggest that the instruc-
tion may not have its intended effect. What is not yet clear is whether the
presence of the appearance - change instruction interacts with the actual appear-
ance change of the perpetrator such that the instruction increases correct
identifi cations of perpetrators with changed appearance, but does not increase
mistaken identifi cations when the perpetrator is not present in the line - ups.
Although the Charman & Wells study included four perpetrators who varied
in the extent to which their appearance changed between the event and the
subsequent line - up, the extent of the appearance change differed among the
perpetrators, making the results diffi cult to interpret. Thus, what is needed are
studies that cleanly manipulate appearance change and the presence of the
appearance - change instruction in a single study.
Simultaneous vs. s equential l ine - u p p resentation
In a simultaneous line - up, witnesses view all line - up members at the same time.
In the sequential line - up, most often tested in psychological research, witnesses
view each line
- up member one at a time, making a yes/no decision after
viewing that member as to whether he or she is the culprit (Lindsay & Wells,
1985 ; Lindsay, Lea & Fulford, 1991 ). If the witness indicates that the line - up
member is not the culprit, the administrator presents the next line - up member.
If the witness indicates that the line - up member is the perpetrator, the pre-
sentation ends. In contrast, the NIJ description of a sequential line - up proce-
dure allows the administrator to continue showing the witness more line - up
members even after the witness has made an identifi cation, if it is consistent
with departmental procedures to do so.
Although both the NIJ guidelines and the AP - LS white paper described the
sequential line - up procedure, neither document recommended the procedure
over the use of the simultaneous procedure despite research suggesting the
superiority of the sequential procedure. A meta - analysis of 25 studies compar-
ing the effects of simultaneous and sequential line - up presentation on eyewit-
ness accuracy, combining the data from over 4,000 witnesses, concluded that
sequential line - ups reliably and substantially reduce mistaken identifi cations in
target - absent line - ups (Steblay, Dysart, Fulero & Lindsay, 2001 ) – a conclu-
sion reaffi rmed in the much larger meta - analysis by Steblay & Dysart (2008a) .
This effect has been dubbed the sequential superiority effect. Although this
drop in mistaken identifi cations is laudable, there is also a smaller reduction
in correct identifi cations, which concerns prosecutors and law enforcement
offi cials.
There continues to be debate over why sequential line - ups produce fewer
mistaken identifi cations than simultaneous line
- ups. Some have argued
that simultaneous line - up presentation encourages witnesses to make relative
268
Handbook of Psychology of Investigative Interviewing
judgements among the line - up members and results in witnesses choosing the
line - up member who looks most like the perpetrator (Lindsay & Wells, 1985 ;
Gronlund, 2004 ). Assuming that, in target - present line - ups, the perpetrator
will most closely resemble the witness ’ s memory of the perpetrator, this relative
judgement process should produce correct identifi cations when the target is
present in the line - up. For target - absent line - ups, if an innocent suspect most
closely resembles the perpetrator, perhaps because he became a suspect because
of his physical resemblance to the perpetrator, one should expect an increase
in false identifi cations. It is thought that sequential line - ups reduce these rela-
tive judgements, encouraging witnesses to make absolute judgements of simi-
larity between each line - up member and their memory of the perpetrator.
Other scholars have recently argued that rather than encouraging witnesses to
shift from a relative to an absolute judgement process, sequential line - ups raise
the criterion threshold that witnesses must cross before they are willing to
make a choice from a line - up (Meissner, Tredoux, Parker & MacLin, 2005 ;
Flowe & Ebbesen, 2007 ). Using a signal detection approach, Meissner and
his colleagues demonstrated that indeed simultaneous presentation produces
a lower criterion for choosing a line - up member than sequential presentation
does.
In part because of this uncertainty about the psychological processes under-
lying the sequential superiority effect, some scholars have questioned the
wisdom of making public policy recommendations to adopt the sequential
line - up presentation as a preferred method of obtaining eyewitness identifi ca-
tions (McQuiston - Surrett, Malpass & Tredoux, 2006 ). These authors also
argue that it is premature to recommend sequential line - ups because:
1. the Steblay meta - analysis contained many unpublished studies that do not
meet the admissibility criterion of peer review promulgated in the Daubert
v Merrell Dow Pharmaceuticals (1993) Supreme Court decision;
2. the body of research on which the meta - analysis rests is relatively small and
therefore prone to unreliability; and
3. the results supporting the sequential superiority effect were obtained pri-
marily from a single laboratory.
What is clear from this analysis is that more research is needed to understand
what variables (e.g., culprit - suspect similarity, line - up construction method,
double - vs. single - blind line - up administration) moderate the effects of simul-
taneous vs. sequential line
- ups on identifi cation decisions. The new meta
-
analysis by Steblay and Dysart offers rebuttals to all of these points.
One likely advantage of the sequential line - up is often overlooked by com-
mentators and those who argue about the relative merits of the two proce-
dures. The rates at which
‘ designated innocent
’ suspect fi llers are chosen
relative to the other fi llers in experiments with target - absent conditions clearly
indicates that the arrays used in those studies – like the arrays used in the real
>
world – are biased against the designated innocent suspect and therefore likely
Recent Developments in Identifi cation Science and Practice
269
to be even more biased against the actual suspect. The respective numbers
from Steblay & Dysart (2008a) for simultaneous arrays was 23% for the des-
ignated fi ller vs. an average of about 7% for other fi llers (a ratio of 3 : 1), which
is on a par with the results from other studies of line - up bias that we noted
earlier. In sequential arrays, Steblay & Dysart report that designated fi llers were
picked by 12% of witnesses vs. 7% for other fi llers – a ratio of less than 2 : 1.
These results may, indeed, arise because the sequential format makes it more
diffi cult for witnesses to merely compare members of the line - up and pick the
‘ most similar ’ face.
Another way to look at bias in sequential and simultaneous arrays is to
consider what the changes in choosing patterns are like when moving from
target - present to target - absent arrays in simultaneous and sequential presenta-
tions. Here we focus on Steblay & Dysart ’ s (2008b) results from 38 studies
with complete data on identifi cations under both present/absent and proce-
dure manipulations. Results from these studies indicate that 30% of the 47%
(nearly two - thirds) of witnesses who seemingly correctly ‘ identifi ed ’ the suspect
in simultaneous target - present arrays would have guessed a fi ller in the target -
absent condition, and a further 17% would have been correct in both types of
arrays (see Table 15.2 ). For sequential arrays the respective numbers were 35%
correct ‘ identifi cations ’ , 24% fi llers and 41% no choice in target - present condi-
tions and 34% fi ller choices and 66% no choice in target - absent conditions.
For target - absent arrays these numbers indicate that 10% of the 35% (less than
one third) of witnesses who seemingly correctly ‘ identifi ed ’ the suspect in
target - present arrays would have guessed a fi ller in the target - absent condition,
but 25% would have correctly rejected the line - up; those 25% appear to have
a reliable memory vs. 17% in simultaneous arrays.
Sequential presentation thus appears to do more than simply reduce choos-
ing by witnesses; it appears to be differentially effective at suppressing guessing