bination of the two techniques used in the PDI is evaluated both in the labo-
ratory and in the fi eld, before any recommendation that the PDI should be
used by the police. First, we examined the effect of the PDI protocol on com-
pleteness and accuracy of person description produced in the laboratory, in
comparison to a control interview made up of two successive standard instruc-
tions and recall. This control interview began with a standard instruction
( ‘ Please describe the person you saw in as much detail as possible ’ ). After this
fi rst description, the experimenter told the participant to describe he same
target person again with the same standard instruction ( ‘ Now, describe the
same person again in as much detail as possible ’ ). Forty participants (all uni-
versity students) were randomly assigned to one of two conditions, and were
tested individually. They were shown a video of a target person performing
routine activities. For external validity, we used 20 videos (same actions, scene,
settings and view) with a different actor in each. No information about the
aim of the experiment was given (incidental encoding). After watching the
A Method to Enhance Person Description
249
video, participants were given a 5 - minute fi ller task and subsequently admin-
istered one of the two protocols.
Each interview was audio - taped and transcribed. All the descriptors reported
by participants for the same target person were listed and distributed into
several categories (facial elements, general components, clothes and subjec-
tive/attitudinal/behavioural details). Subjective details are those that refer to
ambiguous qualities of face shape or personality/occupational impressions
(Meissner et al ., 2001 ). Twenty students independently coded all the descrip-
tive details recalled as either correct or incorrect.
Results showed that the descriptions obtained by the PDI contained sig-
nifi cantly more correct information about general features ( M = 4.00), facial
features ( M = 9.25), clothing ( M = 5.15) and subjective/attitudinal/behav-
ioural information ( M = 2.45) than those obtained by the control interview
(respectively M = 2.60, M = 3.50, M = 3.10 and M = 2.45), without an
accompanying increase in errors, except for facial components (respectively
M = 4.25 and M = 2.45 for PDI and control interview).
Ultimately, the PDI yielded more complete descriptions in comparison to
a control interview. Nevertheless, before this method is employed in the fi eld,
it must be tested in collaboration with police offi cers because each professional,
in their everyday job, could have some damaging practice which may decrease
the effi ciency of a new tool. There are also some informal institutional matters
which are very diffi cult to detect in laboratory.
In our second experiment, we examined the effect of both the PDI protocol
and the standard French police interview for person description (SFPIPD) on
completeness and accuracy of person description produced in the fi eld. When
the SFPIPD is used, police offi cers usually begin with an open - ended instruc-
tion ( ‘ Give me all the details you can about the criminal ’ ), and carry on with
many questions about physical traits unreported by the witness or the victim.
The main procedural difference from the previous experiment was that all
the interviews were conducted by police offi cers experienced and specializing
in criminal investigation. All the interviews took place at a police station, in a
police offi cer ’ s offi ce. Police offi cers were trained in the PDI protocol by an
experimenter for approximately 20 minutes.
Twenty - eight university students were randomly assigned and were tested
individually. They were shown one of the same videos as before. No informa-
tion about the aim of the experiment was given. Consequently, they were not
warned that their memory would be tested. After a 5 - minute fi ller task, an
experimenter told the participant that, if they agreed to take part, the next
part of the study would take place in a police station. If the participant refused,
their participation in the study ended. If they accepted, then the experimenter
drove the participant to the police station. Immediately, a police offi cer would
come and instruct the participant to accompany him to his offi ce. Then, the
interview would begin.
Results showed that the descriptions obtained by the PDI contained sig-
nifi cantly more correct information about general features ( M = 5.93), facial
250
Handbook of Psychology of Investigative Interviewing
components ( M = 9.43) and subjective information ( M = 3.29) than those
obtained by the SFPIPD (respectively M = 3.50, M = 2.07, and M = 1.07),
without an accompanying increase in errors. The accuracy for these particular
elements was signifi cantly higher for the PDI than for the SFPIPD, respectively
90% and 71% for general features, 82% and 41% for facial components, and
96% and 56% for subjective elements. However, we did not observe any quali-
tative or quantitative difference between clothing descriptors reported by PDI
or SFPIPD. Considering the total number of descriptors, we observed that
the PDI obtained a 135% increase of correct information (respectively M = 9.14
and M = 21.50 for SFPIPD and PDI), with a 34% decrease in errors (respec-
tively M = 7.57 and M = 5.00 for SFPIPD and PDI), and these signifi cant
differences led to more accurate descriptions (respectively 56% and 82% for
SFPIPD and PDI).
Other results of interest to police offi cers concerned the proportion of wit-
nesses whose description included each type of descriptor, which was equal to
or higher for PDI compared to the SFPIPD (see Table 14.4 ). The PDI descrip-
tions more frequently included ethnicity, forehead, chin, subjective informa-
tion or face shape descriptors than those obtained by means of a SFPIPD.
More important for the effectiveness of the investigative process is the signifi -
cantly higher accuracy rate obtained with the PDI, particularly concerning
Table 14.4: Occurrence and accuracy of descriptors
Descriptors
Occurrence (%)
Accuracy (%)
SFPIPD
PDI
SFPIPD
PDI
Gender
100
100
100
100
Height
100
100
44
80 *
Age
93
100
41
81 *
Ethnicity
79
100 *
91
93
Build
71
93
82
94
Hair color
100
93
61
76
Hair length
100
100
35
75 *
Forehead
0
29 *
–
67
Eyes
100
100
40
55
Nose
43
93 *
17
81 *
Mouth
43
100 *
17
93 *
Chin
0
79 *
–
86
Face shape
21
57 *
25
77 *
Other descriptors a
29
64 *
75
93
Notes
‘ – ’ = unavailable or incalculable data.
a. Ears, attitudes, etc.
* Signifi cant difference at p < 0.05.
SFPIPD = standard French police interview for person description; PDI = person description
interview.
A Method to Enhance Person Description
251
height or age, hair length, nose, mouth and face shape. Height and age are
important descriptors used for a mug shot search, while the other descriptors
are highly relevant for the construction of a facial composite.
Thus, police offi cers, when using the PDI, obtained more complete and
more accurate descriptions of a target person than with their standard protocol
interview. The superiority of the PDI is specifi cally prevalent in the most useful
descriptive element categories, namely general features, facial descriptors (ideal
for a mugshot search) and personality traits. For these categories, the SFPIPD
not only provided fewer correct elements, but resulted in numerous errors.
There is a potential explanation for this lower accuracy rate. The performance
of police offi cers in this study resulted in completeness of the descriptions that
were largely and signifi cantly greater than those observed in archival study
(Demarchi, 2003 ). There is a general ethos of evaluation in their everyday
practice and they would have wanted to show that their standard protocols
were effective ones. However, because they did not have any methods to
increase the completeness of recalls, they must have used numerous closed
questions which led to a decrease in the quality of the information reported
(Stern,
1902 ; Borst,
1904 ; Whipple,
1909, 1913
; Cady,
1924 ; Goulding,
1971 ; Dent & Stephenson, 1979 ).
The poorer performance obtained with the standard methods does not
imply necessarily that the descriptions provided are less useful than those
obtained with a PDI. For police offi cers, an effi cient method must not only
improve completeness and quality of perpetrators
’ descriptions, but also
increase suspect detection and identifi cation. As a result, there is a need to
evaluate the effi ciency of both the PDI and the SFPIPD in their relative ability
to detect the target person. So we asked 60 participants to match target
persons and their corresponding descriptions among a sample group of similar
people (see Christie & Ellis, 1981 , for a similar experimental procedure). The
material consisted of descriptions obtained in the second experiment, and a
set of 33 photographs of similar individuals (matched for sex, age, height,
ethnicity, build, hair), which included the 20 target persons. Participants were
provided with seven descriptions obtained from only one type of interview
(SFPIPD or PDI), and their task was to pair all the descriptions with the cor-
responding people from the panel of photographs. Results showed that descrip-
tions provided by the PDI led to a 50% increase in correct matches in comparison
to SFPIPD. Nevertheless, the absolute matching rate (i.e., total number of
correct matches divided by total number of matches) was low (0.12 for the
FSPIPD and 0.18 to for the PDI).
To explain the superiority of the PDI in this detection task, we investigated
the signifi cant links between completeness, accuracy and correct matches. If
PDI increases the detection frequency of a previously described person, it is
fi rst and foremost because it leads to more complete recall whilst keeping
constant the overall quality. If the opposite phenomenon is observed with the
SFPIPD, it is because this method decreases the recall accuracy as soon as the
number of reported descriptors increases.
252
Handbook of Psychology of Investigative Interviewing
Conclusions and f uture d irections
By using adapted instructions based on an analysis of spontaneous cognitive
processes, with just 20 minutes ’ training, police offi cers can obtain descriptions
which contain twice as many correct descriptors and half the number of errors
than a standard police interview. These fi ndings are in contrast to previous
studies, which found that encouraging people to generate more complete
descriptions, particularly facial descriptions, resulted in a greater proportion
of inaccurate details. These quantitative and qualitative improvements are
immediately translated into an improved ability to identify a potential suspect.
From a research point of view, this fi nding showed that the effi ciency of
a physical description should not be measured only in terms of qualitative
and quantitative aspects, but also through more concrete and ecological
measures, such as the calculation of a matching score between the target
person and its corresponding description among a defi nite population of
similar individuals.
More important for professionals is the ease of learning and using the PDI
protocol. The PDI training in France lasts approximately 20 minutes as in our
second study: 15 minutes for theoretical and empirical concepts and 5 minutes
of attempts to obtain correct recitation of the two instructions. In France, the
PDI training is now incorporated into standard and CI training for experienced
police offi cers. However, the PDI is not yet taught to novice police offi cers or
investigators.
Police offi cers can use the PDI protocol in isolation or to augment a stan-
dard or a cognitive interview. Preliminary research suggests that when a CI
framework is used, the best time to give the witness the PDI instructions is
after the last witness ’ s free recall which follows the ‘ change perspective ’ instruc-
tion, and before the questioning phase. This order appears to produce the
most effective performance. Further research will address this issue.
Some issues that have implications for professionals need to be further
explored. First, the impact of delay on completeness and accuracy of descrip-
tions obtained by the PDI needs to be explored. Second, the ability of particu-
lar sub - populations (e.g. children, older people, learning disabled) to produce
better descriptions with aid of the PDI should be investigated. Third, research-
ers must also investigate the possible interaction of the PDI with some poten-
tial adverse factors affecting person description, for example the
verbal
overshadowing effect (VOE; Schooler
& Engstler
- Schooler,
1990 ; Meissner
et al ., 2001 ). The VOE is defi ned as impairment in recognition performance
when people are required to provide a verbal description of a complex stimulus,
such as the fac
e of a perpetrator. Finger & Pezdek (1999) showed that the
VOE does not appear if there is a suffi cient delay (about 24 minutes) between
description and recognition from line
- ups or a photographs. Nevertheless,
identifi cation from a line - up can only be used after a suspect has been appre-
hended. Witnesses usually do not know the criminal and the investigation
A Method to Enhance Person Description
253
often includes an examination of mugshots. The latter would be selected from
available databases using the description given by the witness. According to
the French police investigative procedure, police must ask witnesses for a
detailed description of a perpetrator and immediately present them with an
album of photographs of suspects known to have committed a crime in the
local area and whose descriptions fi t the witness ’ s account. Whether verbal
overshadowing will occur when such procedures are used remains to be estab-
lished. Demarchi, Py, Parain & Groud - Tan (2006) found that recognition
performance from a mugshot search following a PDI or a SFPIPD was similar
to the performance of a control group who did not describe the face prior to
viewing the mugshots.
In conclusion, the PDI protocol, which allows the witness to recall pertin-
ent information by using relevant instructions, promotes the completeness
of a description of a person in the absence of social pressure or suggestive
prompts for unrecalled information. Such a technique elicits more accu-
rate descriptors and reduces the likelihood of contaminating subsequent
attempts at verbal recall and perceptual identifi cation from a line
- up or a
mugshot fi le.
References
Boon , J. C. W. & Noon , E. ( 1994 ). Changing perspectives in cognitive interviewing .
Psychology, Crime, and Law , 1 , 59 – 69 .
Borst ,
M. (
1904 ).
Recherches exp
é rimentales sur l
’ é ducabilit é et la fi d é lit é du
t é moignage [Researches on educability and reliability of testimonies] . Archives de
Psychologie , III.
Bower , G. H. & Karlin , B. M. ( 1974 ). Depth of processing, pictures of faces and
recognition memory . Journal of Experimental Psychology , 103 , 751 – 757 .
Brown , C. & Lloyd - Jones , T. J. ( 2003 ). Verbal overshadowing of multiple face and
car recognition: Effects of within
- versus across
Handbook of Psychology of Investigative Interviewing: Current Developments and Future Directions Page 48
