Content uploaded by Mathilde Noc
Author content
All content in this area was uploaded by Mathilde Noc on Nov 03, 2023
Content may be subject to copyright.
Vol.:(0123456789)
1 3
Journal of Police and Criminal Psychology
https://doi.org/10.1007/s11896-022-09543-5
False Confession inInnocent Suspects: ALook attheCognitive
Interview forSuspects
MathildeNoc1 · MagaliGinet1· NadineDeslauriers‑Varin2
Accepted: 15 August 2022
© The Author(s), under exclusive licence to Society for Police and Criminal Psychology 2022
Abstract
The Cognitive Interview for Suspects (CIS) is a recently designed information-gathering style interview method for interview-
ing suspects of crimes. Some components of this method (i.e., Mental context reinstatement, Report-everything inquiries)
should make it possible to collect a large quantity of correct information, and to limit the misinformation-inducing impact
of leading questions. However, certain other components (i.e., Reverse order, Challenge stages) are designed to increase
cognitive load among interviewees, and might therefore lead to false confessions in innocent suspects. Using a protocol
designed to provoke false confessions in innocent mock-suspects, the CIS was compared with another Information-Gathering
interview method (IGM) and an Accusatory Interview (AI) on these different aspects. Results showed a benefit of the CIS
in terms of information gathering and misinformation effects, compared to the two other interviews. Moreover, it did not
increase the risk of false confessions. These conclusions are of interest for field investigators, who could use this method
with suspects in real cases.
Keywords Cognitive interview for suspects· False confession· Information gathering· Misleading questions· Accusatory
interview
In criminal investigations, the suspect’s statement can be
crucial to the outcome of the judicial investigation. To assist
in the collection of such statements, certain models and
protocols for interviewing suspects have been created by
professionals and researchers working in the field of inves-
tigative psychology (Structured Interview, PEACE, Reid,
ProGREAI, etc.). Among these, the Cognitive Interviewing
for Suspects protocol (CIS; Geiselman2012) was designed
to facilitate the collection of a large amount of information
from individuals suspected of criminal actions.
The Cognitive Interview forSuspects
The CIS is inspired by a very similar protocol for witnesses
and victims of criminal events created in the 1980s: the Cog-
nitive Interview (CI; Geiselman etal. 1984; Geiselman and
Fisher2014). This protocol is based on the principles of
specific encoding and multiple access to memory, according
to which similarity between encoding and retrieval cues, par-
ticularly if they are derived from multisensory processing,
improves memory recall (Tulving2002). Indeed, the benefit
of the CI in terms of improving information gathering with-
out compromising accuracy has been widely highlighted in
numerous experimental and field studies with witnesses and
victims of crimes (e.g., Memon etal. 2010).
However, witnesses are assumed to produce “hon-
est” errors (e.g., caused by misperception, interference,
retrieval failures) while suspects tend to lie intentionally
(Yarmey2009). Thus, the objective of the CIS is not only
to increase the quantity and quality of gathered informa-
tion, but also to permit the detection of deception among
deceitful suspects. To do this, certain parts of the CIS
are intended to increase the interviewees’ cognitive load,
* Mathilde Noc
Mathilde.NOC@uca.fr
1 CNRS, Laboratoire de Psychologie Sociale et Cognitive,
Université Clermont Auvergne, 34 Avenue Carnot,
F-63000Clermont-Ferrand, France
2 École de Travail Social et de Criminologie, Université
Laval, Pavillon Charles-De Koninck, 1030, av. des
Sciences-Humaines, Local5444Québec, Canada
Journal of Police and Criminal Psychology
1 3
which in turn leads to difficulties in lying and the appear-
ance of lie indicators (Vrij etal. 2006).
The CIS protocol includes eight stages: (1) Rapport
building / Introduction: Development of a rapport with
the suspect while talking about neutral topics in order to
put them in a psychologically comfortable state. This stage
also serves as a baseline from which to observe subsequent
potentially deceptive behavior; (2) Narrative: Instruction
to report as many details as possible (Report everything
instruction) and to mentally recreate the context of the
crime (Mental context reinstatement instruction). This
makes it possible to gather more information; (3) Draw-
ing / sketch: Drawing of the criminal scene or a part of this
scene. This unexpected task is likely to lead to the emer-
gence of inconsistencies if the suspect is lying (Vrij etal.
2010); (4) Follow-up, open-ended questions: Questions
asked using an information-gathering approach rather
than a confrontational one; (5) Reverse order mnemonic:
Instruction to recall the criminal event backward, which
is a cognitively demanding task that improves deception
detection (Vrij etal. 2008); (6) Challenge: Gradual con-
frontation with evidence and incriminating contradic-
tions. The interviewer remains respectful and presents the
weakest evidence first. This procedure increases the like-
lihood of additional inconsistent statements for deceptive
individuals (Dando and Bull2011); (7) Review: Review
of the suspect’s statements and correction of any errors
or omissions; (8) Closure: Thanking of cooperative sus-
pects. To those who appear to have lied, interviewers may
“express that they feel disappointed and disrespected”
(Geiselman2012, p.5).
Only a few studies have tested the CIS protocol. The
first study showed that this method makes it possible to
discriminate between truthful and deceptive statements,
and that the accuracy of this distinction increases pro-
gressively as the interview proceeds (Geiselman2012).
Two subsequent studies confirmed the value of CIS for lie
detection, as more deceit indicators were found in decep-
tive statements than in truthful ones when participants
were interviewed with this method (Logue etal. 2015, for
verbal cues to deception; Frosina etal. 2018, for physi-
ological cues). Finally, a study by [name deleted to main-
tain the integrity of the review process] showed that when
field officers were CIS-trained, their interviews of guilty
suspects revealed a greater amount of correct information
than those of untrained officers using their usual inter-
view protocol (i.e., standard interviews). To the best of our
knowledge, this was the only study that compared the CIS,
in terms of information gathering, to another interview
protocol, conducted by untrained interviewers not using
“information-gathering focused techniques.” Although
this study showed a benefit of the CIS in terms of infor-
mation gathering, it appears that the CIS has never been
compared to any other protocol that focuses on informa-
tion gathering.
While the CIS made it possible to increase the quan-
tity of gathered details, another — as yet unassessed —
benefit could be an increase in the suspect’s resistance
to the interviewer’s suggestions (e.g., through leading
questions). Indeed, despite the administration of train-
ing designed to improve interviewers’ questioning skills,
interviews still contain a large number of questions that
are harmful to the collection of information, in particu-
lar leading questions (Bull and Soukara2010; Griffiths
and Milne2006; [name deleted to maintain the integrity
of the review process]). Leading questions can affect the
accuracy of the collected details and the interviewee’s
memory (Loftus and Palmer1974). In particular, questions
containing misleading suggestions (i.e., details suggested
by the interviewer, never mentioned by the interviewee,
and which are false) have the most negative impact on
the interviewee’s testimony and memory when compared
to yes/no leading questions (Sharman and Powell2012).
However, the CIS contains components that could coun-
ter the adverse effects of such questions. For instance,
the Mental context reinstatement and Report everything
mnemonics could build memory recalls that are more
resistant to misinformation (LaPaglia etal. 2014; Memon
etal. 2009; Milne and Bull2002). This phenomenon is
notably due to the fact that the Mental context reinstate-
ment instruction aims to help the interviewee generate a
large number of retrieval cues, by rethinking numerous
multisensory elements of the episodic memory. Indeed,
mentally or physically recreating contextual cues such as
temperature, brightness, and surrounding sounds helps
facilitate the retrieval of experienced memories (Mundorf
etal. 2021; Wheeler and Gabbert2017). Furthermore, the
Report-everything instruction usually results in a rich and
detailed narratives, allowing the interviewee to express
a large amount of correct details (Allwood etal. 2005),
which will themselves have a retrieval cue role. Thus,
the joint action of these two mnemonics should allow the
interviewee to have easier access to their memory, leaving
less room for misinterpretation of the information. Such a
benefit in terms of resistance to suggestion through leading
questions has already been shown with the witness/victim
CI protocol (Corson and Verrier2013; Geiselman etal.
1986; Memon etal. 2009). This positive impact, called
the “Geiselman effect,” is defined as “the resistance of
interviewees to the investigator-induced influences which
could lead to a distortion of the memory and testimony
gathered” (Geiselman etal. 1986, p. 32). Given the com-
mon theoretical basis shared by the protocol for witnesses/
victims and the protocol for suspects, it is possible that
the CIS also increases suspects’ resistance to suggestions
through misleading questions. Thus, it is likely that such
Journal of Police and Criminal Psychology
1 3
a Geiselman effect would also be present with a public of
suspects interviewed by the CIS protocol.
While some benefits can be expected from certain CIS
instructions, other components may raise questions. For
instance, the Reverse order and Challenge stages are sur-
prising and can increase cognitive load, and can also lead to
a feeling of pressure and anxiety (Vrij etal. 2008). Although
the effect of these components may be beneficial when inter-
viewing guilty suspects, questions can be raised about their
consequences with innocent interviewees. Indeed, a survey
conducted by Sigurdsson and Gudjonsson (1996) revealed
that the pressures and difficulties perceived by innocent
suspects during their interviews had, in some cases, led to
disastrous consequences such as false confessions. A study
by Otgaar etal. (2012) showed that increased cognitive load
during a recall task resulted in increased suggestibility of the
interviewees. Such observations were also proposed in the
study by Alm etal. (2019), where participants reporting high
cognitive loads also exhibited high levels of suggestibility.
Since suggestibility is one of the prevalent factors in the
occurrence of behaviors such as false confessions (Otgaar
etal. 2012), the purpose of this study is thus to explore the
fact that the CIS may or may not provoke a false confes-
sion phenomenon in innocent suspects assumed to be guilty.
This question could be of major importance in actual field
interviews.
False Confessions
False confessions are defined as “the fact that an individual
admits guilt for an act for which s/he is actually innocent”
(Kassin and Gudjonsson2004, p. 48), and this phenomenon
is the cause of wrongful convictions in more than 25% of
cases in which innocent suspects are incarcerated (West and
Meterko2016).
In order to study false confessions in innocent suspects,
Kassin and Kiechel (1996) created the “forbidden key proto-
col,” which has subsequently been used on numerous occa-
sions (see the meta-analysis by Stewart etal. 2018). In this
protocol, participants are dictated a list of letters to type in a
computer-based task. They are told that their typing speed is
to be evaluated and are then instructed not to press the “Alt”
key on the keyboard as this may cause the computer to crash
and the data to be lost. During the task, the program and
computer crash — being pre-programmed to do so as part
of the protocol —, and the participants are then accused of
having pressed the forbidden key. The objective of this pro-
tocol is to observe the percentage of participants who will
make compliant false confessions (i.e., sign a written state-
ment of guilt for the experimenter), internalize these confes-
sions (i.e., self-designate themselves as guilty when asked
by a confederate for details of the event), and confabulate
false information (i.e., add details highlighting their guilt,
when they are actually innocent) following this accusation.
The results of Kassin and Kiechel’s study showed that 69%
of participants signed a confession, 28% internalized their
guilt, and 9% confessed incriminating details, highlighting
this protocol’s effectiveness in inducing false confessions.
Several researchers have since used this protocol to test the
impact of interview techniques on the phenomenon of false
confession among innocent suspects. The results of a study
by Redlich and Goodman (2003) testing the impact of the
interviewer’s deceitful behavior showed that interviewees
produced more false confessions when the investigators
claimed to possess incriminating evidence than when they
did not use this technique. Furthermore, Klaver etal. (2008)
studied the influence of minimization: a technique used in
certain interview protocols (e.g., the Reid technique) which
consists of reassuring the suspects, finding explanations
and moral justifications for their actions, and minimizing
the legal consequences of their act if they confess their
guilt. The results of the study showed that interviewees pro-
duced more false confessions if they were questioned using
minimization techniques than when they were questioned
in a neutral manner. Furthermore, the study by Klaver etal.
(2008) highlighted the link between plausible guilt and false
confessions. Indeed, two different conditions were presented
to participants: in one (plausible guilt), participants were
told not to press the Alt key; in the other (implausible guilt),
participants were told not to press the Esc key, which is
further from the standard position of their hand. The results
showed that false confessions were more frequent the more
plausible the guilt. Similar results regarding the detrimental
role of uncertainty in the occurrence of false confessions
were subsequently presented in the meta-analysis by Stewart
etal. (2018).
Overall, the “forbidden key protocol” results were also
reported in a meta-analysis conducted by Meissner etal.
(2014). In particular, this work compared the impact of three
methods of questioning on the occurrence of false confes-
sions. The first type of questioning was direct questioning,
during which the interviewer directly asked the participants
whether they are guilty in a neutral, non-adversarial manner.
The second type of questioning was an accusatory method,
generally conducted as follows: (1) the experimenter instils
a sense of insecurity, isolation, and emotional pressure in the
participants; (2) the experimenter accuses the participants,
blames them, threatens them with the consequences of the
action, and prevents them from denying it; (3) the experi-
menter minimizes, is sympathetic, gains the participants’ con-
fidence by giving them moral justifications for their actions,
and finally minimizes the consequences of the event. This
interview method is inspired by the Reid technique (Inbau
etal. 2001), an adversarial suspect interview method widely
used in the USA and whose results in terms of the quality
Journal of Police and Criminal Psychology
1 3
of gathered information are debatable (Gudjonsson2003).
The third type of interview was an information-gathering
method, generally conducted as follows: (1) the experimenter
introduces the interview and its purposes, and develops rap-
port with the participant, (2) the participant freely recalls the
facts without the experimenter speaking, (3) the experimenter
asks questions, presenting them as a request for information,
not an accusation. This interview method is inspired by the
PEACE, an information-gathering interview model widely
used in the UK and whose benefits in terms of information
gathering have frequently been demonstrated (e.g., Walsh
and Bull2010). Analyses of the different studies addressed
in the meta-analysis by Meissner etal. (2014) showed that
only accusatory questioning led to an increase in the occur-
rence of false confessions when compared to the direct and
information-gathering methods.
The Present Study
The purpose of this study is therefore to gain an understanding
of the benefits, but also the possible risks, of the CIS. First,
we wanted to investigate its benefit in terms of information
gathering, since only one study has so far tested this effect
[name deleted to maintain the integrity of the review process].
Second, we wanted to assess whether this method could also
cause a Geiselman effect in suspect interviewees and reduce
the impact of misleading questions. Third, we wanted to fur-
ther analyze the risks of obtaining false confessions from inno-
cent suspects in an uncertain situation. To this end, and in line
with the meta-analysis conducted by Meissner etal. (2014), we
compared the CIS with two other interview methods currently
used in the field by professional interviewers: An accusatory
method and an information-gathering method, as described in
Meissner etal. (2014). The accusatory interview (AI), inspired
by the Reid technique and the accusatory methods studied
in the meta-analysis conducted by Meissner etal. (2014), is
composed of three stages: instilment of isolation, accusation,
and minimization. The information-gathering method (IGM),
inspired by the PEACE model and the information-gathering
interrogation methods also studied in this meta-analysis, is a
cooperative protocol, which involves stages of rapport building
and information gathering. Thus, like the CIS, the IGM aims
to improve the amount of correct information gathered and
therefore includes components such as the Rapport-Building
stage and Report everything mnemonic. However, the protocol
does not contain other mnemonics such as the Mental Context
Reinstatement instruction, which is one of the most effective
CI instructions (Dando etal. 2008). Moreover, the aim of the
IGM is not to obtain a confession or to enhance lie detection,
and it therefore does not include surprising or cognitively
demanding components such as the Drawing, Reverse order
and Challenge stages.
Thus, cooperative methods (i.e., CIS and IGM) should
provide more information than an accusatory interview (AI).
In addition, thanks to the different mnemonics contained in
this protocol, CIS is expected to gather even more informa-
tion than IGM, and also to reduce misleading information
reporting compared to the other two methods. Finally, AI
should lead to more false confessions (i.e., compliant and
internalized false confessions, confabulated incriminating
information) than cooperative methods. However, the use
of demanding tasks in the CIS could increase the number
of false confessions obtained when compared to the IGM.
Method
Participants andDesign
Interviewees
The participants in the initial sample were 129 undergradu-
ate students at [name deleted to maintain the integrity of
the review process] (115 women and 14 men). All of them
used French as their main language. Four participants were
excluded because of recording failures during the inter-
views. The final sample therefore consisted of 125 partici-
pants (111 women and 14 men), aged from 17 to 24years
(M = 18.86, SD = 1.16). They were recruited by voluntarily
registering via the laboratory bulletin board and all received
course credits for their participation. They were randomly
assigned to one of the three between-subject conditions of
our study (interview style: CIS vs. IGM vs. AI). An a priori
power analysis established using G*Power 3.1 (Guay etal.
2016) for omnibus tests on ANOVAs had indicated a nec-
essary sample size of 102 participants to allow an effect
size of f = 0.40 at a critical F = 3.09 (estimated from [name
deleted to maintain the integrity of the review process]),
with α = 0.05 and a 95% power.
Interviewers andConfederates
Three graduate women (aged 22, 24, and 24years) acted
as confederates and interviewers in the study. More specifi-
cally, each experimental run required one confederate and
one investigator and the roles of the three experimenters
were randomly rotated. The role of the accomplice was to
guide the participant through the experiment and the role of
the interviewer was to investigate what happened using the
CIS, IGM, or AI. The three experimenters learned a pre-
determined role which they played in the same way with all
the participants throughout the entire course of the experi-
ment. These roles are detailed in the Procedure section. The
experimenters received 8h of training in the interview meth-
ods, administered by the first two authors of this study. The
Journal of Police and Criminal Psychology
1 3
training included role-playing games followed by detailed
feedback.
Interviews
Interview Protocols
The participants were randomly interviewed with an IGM,
a CIS, or an AI.
In the IGM condition, the interviewer first presented the
interview as a means of collecting information about the
event, explained the purposes of the interview — among
which to highlight the involvement or non-involvement of
the participant —, and then built a rapport with the partici-
pants by talking about the university courses. The report-
everything instruction was then used to ask the participants
to give very detailed free accounts of every moment of the
event (i.e., the introduction of the experiment, the computer
task, the computer crash, and the confederate’s reaction).
The free account was followed by a series of questions,
and then a series of four misleading questions. The ques-
tions addressed to the participants in all the conditions are
described below. The interviewer then ended the interview
with a closure stage.
The CIS condition had the same basis as the IGM. In
addition to the interview presentation, explaining, and rap-
port building, the participants were also asked to recall the
environment of the event through the mental context rein-
statement instruction. Following their free recall, the par-
ticipants were asked to draw the computer as it was at the
time of the crash. After they had completed the drawing,
the participants were asked the series of questions. They
were then instructed to recall the event in reverse order. The
interviewer then introduced a challenge stage during which
she highlighted the inconsistencies in the participant’s state-
ment. Finally, the interviewer summarized the statement and
conducted the closure stage.
In the AI condition, the experimenter created a feeling
of isolation in the participants (isolation stage). She then
accused the participants of pressing the Alt key, presented
them with the consequences of this action and asked them
questions without allowing them to claim their innocence
(accusation stage). Finally, the interviewer minimized the
consequences of the crash, became sympathetic, and gained
the participants’ confidence by giving them moral justifica-
tions for their actions (minimization stage).
Questions
In order to assess the information gathering made possi-
ble by the questioning of the suspects in each of the three
interview types, the four pieces of information that the
interviewer needed to obtain were predefined: (1) whether
the experimenter warned the participant not to press the
Alt key; (2) the position of the Alt key on the keyboard;
(3) the participant’s reaction when the computer crashed;
(4) the content of the experimenter’s telephone conversa-
tion. In the CIS and IGM conditions, the questions were
funnel-structured, as recommended in the training material
for information-centered interview methods (Luke etal.
2016). For example, to obtain the first piece of information,
the interviewer first asked the open-ended question “Please
remind me of everything the experimenter said to you?”.
This was followed by the question “Please remind me of
the exact instructions?” and then “Where there any warning
or not?”, with the questions becoming increasingly specific
until the required information had been collected. In the AI
condition, the questions were asked in a closed-ended man-
ner (i.e., “Did the experimenter warn you not to press the Alt
key?”), as observed in standard interviews (Leahy-Harland
and Bull2017; Snook etal. 2012).
Thereafter, four pre-defined misleading questions were
asked in order to test the Geiselman effect and the misin-
formation reporting caused by the interviewer: (1) Did
you have to press all the keys on the keyboard?; (2) Did
the experimenter warn you that there were keys to avoid?;
(3) Did the experimenter tell you that the crash already
happened to a second-year student?; (4) After calling, did
the experimenter tell you that she was going to take you to
me?. In these questions, four misleading information were
included: (1) the fact that they were told to use all the keys
(including numbers, etc.), whereas they were asked only to
use the main “letters” part of the keyboard; (2) the fact that
the experimenter warned the participant to avoid multiple
(plural) keys, whereas they were told to only avoid one key;
(3) the mention of a second year student, who was actually
a first year student; (4) and the fact that the experimenter
mentioned the interviewer, whereas only the office was
mentioned.
Procedure
All the procedures and interrogation methods were
approved by the University Research and Ethical Com-
mittee (IRB00011540-2019–48). The procedure — which
was identical for participants across all conditions — was
inspired by Kassin and Kiechel’s (1996) “forbidden key”
protocol and took the following form: Participants arrived
at the University thinking they were going to participate in
a computer-based reaction-time task. An experimenter (con-
federate) welcomed them and led them into an experimen-
tal room where there was a table, a chair, and a computer.
After the participants signed the consent form, in which
they agreed to be audio-recorded during the experiment, the
accomplice gave them these instructions: “You are going to
do a dictation task on the computer. You will hear letters,
Journal of Police and Criminal Psychology
1 3
and you will have to type them as quickly as possible on
the keyboard. Be careful not to press the Alt key, which
can cause the computer to crash and the data to be lost”.
Following these instructions, the accomplice left the room,
indicating that she was waiting behind the door and that the
participant should alert her if a problem occurred. The par-
ticipants then began the dictation task. After the 125th letter,
the computer screen was pre-programmed to display a Win-
dows error message on a blue screen. The keyboard froze
and the participants left the room to inform the accomplice
of the problem. She came and read the error message, and
then phoned the “experiment supervisor” (the interviewer)
in the participant’s presence to ask what should be done.
After hanging up the phone, the accomplice told the partici-
pant to leave the computer as it was, and to follow her to the
supervisor’s office to fix the problem. The participants then
entered the supervisor’s office and were told that a report of
the incident had to be sent to the university’s IT department.
They were then asked about the event with a CIS, IGM, or
AI. After the interview, the participants were escorted back
to the experimental room by the accomplice. On the way, in
order to measure whether they had internalized their guilt
and added confabulated details, the accomplice asked them
(1) if they had been able to establish the cause of the com-
puter crash, (2) who was responsible, and (3) in what way
they were responsible. Finally, they received a debriefing
on the experiment. The purpose of this debriefing was to
explain the real purpose of the study to them and to ensure
their well-being before they left. They were then told that
they were absolutely not guilty, that all the participants
were having the same experience, and that the incident was
pre-programmed. The experimenter always had an audio-
recorder in her pocket to capture all the items of information
given by the participant.
Coding andMeasures
Information Recalled During theInterviews
The audio recordings were transcribed in order to analyze
the quantity and quality of the information contained in the
statements gathered during the interviews. The transcripts
were then rated on a term-by-term basis, and all the units of
information were classified on a shared coding grid. These
details were classified as correct (i.e., details that were cor-
rectly described), erroneous (i.e., details whose description
contained mistakes), or confabulated (i.e., details mentioned
even though they did not exist and did not happen). Details
were only scored the first time they were reported (i.e.,
repeated information was ignored). The quantity of correct,
erroneous, and confabulated details made it possible to com-
pare the accuracy rates of the statements gathered using the
CIS, IGM, and AI. The accuracy rates were calculated by
dividing the number of correct items of information by the
total amount of gathered information.
The units of information were also classified into one of
the following categories: action (e.g., walking in the cor-
ridor), object (e.g., the computer screen), location (e.g., the
experimentation room), person (e.g., the experimenter), time
(e.g., in the morning), and interaction (e.g., she said I had
to wait). The action, object, location, and person categories
are those usually analyzed in Cognitive Interview studies
(Geiselman2012; Stein and Memon2006). The time and
interaction categories were added in this study as they could
also be useful in a forensic investigation and because they
are relevant in the description of a personally experienced
event. Nineteen randomly selected interviews (15.2%) were
coded by a second coder in order to assess coding reliability.
The inter-rater agreement was very high, ICC = 0.881.
Acceptance ofSuggestions
In order to assess the misinformation effect induced by the
interviewer’s misleading questions, participants’ answers to
the four questions were analyzed on a dichotomous basis:
they were rated “1” if the answer contained erroneous or
confabulated information induced by the question, and “0” if
the participant contradicted the interviewer and did not give
erroneous information. Thus, each participant had a score
ranging from 0 to 4, with higher scores indicating a higher
acceptance of misinformation induced by the interviewer.
Number ofFalse Confessions
Three different types of false confessions were measured
as three dependant variables: Compliant false confessions
occurred when the participants admitted they had pressed
the Alt key (e.g., “I’m okay to say that I pressed Alt, but I'm
not sure”), Internalized false confessions occurred when the
participants actually thought they had pressed the Alt key
(e.g., “I pressed the Alt key and I'm sorry”), and confabu-
lated false confessions were observed when the participants
added details which did not actually occur to the event (e.g.,
“it was the bracelet on my left hand that pressed the Alt
key while my hand was above it”). All these types of false
confessions were coded on a dichotomous basis: they were
rated “1” if the participant made a false confession, and “0”
if they did not.
After theInterviews: Measures ofInternalization
andConfabulation
In order to add measures of internalization and confabula-
tion, the discussion between the confederate and the par-
ticipant on the way back to the experimental room after the
interview was also coded. These measures were made in the
Journal of Police and Criminal Psychology
1 3
same way as the analysis of statements and false confessions.
Thus, units of information were coded as correct, erroneous,
or confabulated, and classified as belonging to one of the
seven details categories. Compliant, internalized, confabu-
lated, and false confessions were coded “1” if they occurred,
“0” if they did not.
Results
Preliminary Analyses
Independent sample t tests were run in order to investigate
if the experimenter or interviewee’s gender had an impact
on the quantity of gathered information. Results showed that
there was no effect of the experimenter on the total number
of correct, erroneous, or confabulated details, ts < 1.47 and
ps > 0.144. However, a difference due to participants’ gender
was found on the total number of correct details, t(43) = 2.74,
p = 0.011, d = 1.15, with women providing more correct
details (M = 81.96, SD = 32.07, 95% CI [75.92; 87.99]) than
men (M = 66.38, SD = 17.99, 95% CI [55.97; 76.75]). Since
this effect is significant, the interviewee’s gender will be
included as a co-variate in the following analyses.
Interview Durations
A one-way ANOVA including the interview style as inde-
pendent variable was used to analyze the interview durations
across the different conditions. Results showed a significant
effect of the interview style, F (2, 123) = 45.45, p < 0.001,
d = 1.73. Post hoc analyses revealed that CIS interviews
(M = 19.52, SD = 4.93, 95% CI [17.98; 21.05]) were longer
(in minutes) than IGM interviews (M = 14.24, SD = 3.47,
95% CI [13.18; 15.31]), t (122) = 6.03, p < 0.001, which were
in turn longer than AI interviews (M = 10.93, SD = 3.51, 95%
CI [9.80; 12.05]), t (122) = 3.74, p < 0.001.
Information Gathered
Correct, Erroneous, andConfabulated Details, Accuracy
Rate
A one-way ANCOVA including the interview style as inde-
pendent variable and the interviewee’s gender as co-variate
was used to analyze the impact of interview style on the
total amount of correct details gathered. Results showed a
main effect of interview style, F (2, 123) = 101.10 p < 0.001,
d = 2.59. More specifically, post hoc analyses revealed
that CIS interviews (M = 111.09, SD = 22.33, 95% CI
[104.14; 118.06]) contained more correct details than IGM
interviews (M = 77.88, SD = 20.24, 95% CI [71.66; 84.11]),
t (122) = 8.04 p < 0.001, which in turn contained more cor-
rect details than AI (M = 50.28, SD = 13.02, 95% CI [46.11;
54.44]), t (122) = 6.60, p < 0.001. There was no difference for
the number of erroneous and confabulated details, Fs < 1.50
and ps > 0.226. The accuracy rates of the statements were
also calculated by dividing the number of correct details
by the total number of details. A one-way ANCOVA high-
lighted a main effect of interview style on accuracy, F (2,
123) = 8.83, p < 0.001, d = 0.76. Post hoc analyses revealed
that CIS-gathered statements (M = 97.12%, SD = 2.15, 95%
CI [96.45; 97.79]) were more accurate than IGM -gathered
statements (M = 96.04%, SD = 2.37, 95% CI [95.31; 96.77]),
t (122) = 1.96 p = 0.026, which were in turn more accurate
than AI-gathered statements (M = 94.76%, SD = 3.05, 95%
CI [93.79; 95.74]), t (122) = 2.30, p = 0.012. However, for
these three rates, the confidence intervals overlap.
Number ofCorrect Details/Time Ratio
Because the interview duration differed across conditions,
a number of correct details/total interview time ratio was
calculated for each participant. The results show the num-
ber of correct details gathered per minute of interview. A
one-way ANCOVA revealed a significant effect of inter-
view condition on the ratio, F (2, 123) = 4.47, p = 0.013,
d = 0.54, with post hoc analyses indicating that CIS
(M = 6.08, SD = 1.99, 95% CI [5.46; 6.69]) contained more
correct details per minute than AI (M = 4.96, SD = 1.60,
95% CI [4.45; 5.47]), t (122) = 2.95, p = 0.002. This was
also the case for the IGM interviews (M = 5.62, SD = 1.49,
95% CI [5.17; 6.08]), which contained more correct details
per minute than AI, t (122) = 1.78, p = 0.041. There was no
difference for the ratios between the CIS and IGM inter-
views, t (122) = 1.23, p = 0.112.
Moreover, the instructions given to the participants were
not the same across all the conditions. Since it can be time-
consuming to formulate these instructions, a number of
details/interviewee speaking time ratio was also calculated
for each participant. The results show the number of correct
details gathered per interviewee speaking minute. A one-way
ANCOVA revealed a significant effect of interview condi-
tion on this ratio, F (2, 123) = 4.27, p = 0.016, d = 0.53, with
post hoc analyses revealing that CIS (M = 12.74, SD = 10.41,
95% CI [9.49; 15.98]) elicited more correct details per inter-
viewee speaking minute than IGM (M = 10.08, SD = 3.03,
95% CI [9.15; 11.02]), t (122) = 1.86, p = 0.033, or AI
(M = 8.65, SD = 3.29, 95% CI [7.59; 9.71]), t (122) = 2.82,
p = 0.003. With regard to the quantity of details provided,
there was no difference between the ratios for IGM and AI,
t (122) = 0.99, p = 0.161.
Journal of Police and Criminal Psychology
1 3
Type ofDetails
A series of ANCOVAs with post hoc analyses were conducted
to analyze the impact of interview style on the number of cor-
rect action, object, location, person, time, and interaction details
gathered. Overall, results showed a main effect of interview
style on the number of correct action details, F (2, 123) = 64.83,
p < 0.001, d = 2.07, object details, F (2, 123) = 36.36, p < 0.001,
d = 1.55, location det ails, F (2, 123) = 39.17, p < 0.001, d = 1.61,
person-related details, F (2, 123) = 94.46, p < 0.001, d = 2.50,
time details, F (2, 123) = 22.74, p < 0.001, d = 1.23, and interac-
tion details, F (2, 123) = 34.99, p < 0.001, d = 1.52. For all these
categories, CIS-gathered statements contained more units of
information than IGM -gathered statements, which contained
more details than AI-gathered statements. Descriptive statistics
and post hoc analyses of these results are presented in Table1.
For all these categories, there was no difference between inter-
view styles in terms of the number of erroneous or confabulated
details, Fs < 2.18 and ps > 0.117.
Acceptance ofSuggestion
In order to analyze the impact of interview style on the
participant’s acceptance of suggestion, four misleading
questions were asked during the questioning stage of the
interviews. The number of reported items of misinformation
induced by the misleading questions was measured. A one-
way ANCOVA revealed a significant effect of the interview
condition on the number of reported items of misinforma-
tion, F (2, 123) = 4.95, p = 0.009, d = 0.57. Post hoc analyses
revealed that CIS interviews (M = 2.17, SD = 1.32, 95% CI
[1.75; 2.58]) contained less reported misinformation induced
by the leading questions than the IGM interviews (M = 2.95,
SD = 1.17, 95% CI [2.59; 3.31]), t (122) = − 2.82, p = 0.003.
This was also the case for AI (M = 2.40, SD = 1.36, 95% CI
[1.97; 2.83]), which contained less reported misinformation
than the IGM interviews, t (122) = 1.96, p = 0.026. There
was no difference between the CIS and AI in terms of the
number of wrong answers, t (122) = -0.82, p = 0.21.
False Confessions
Compliant, Internalized, andConfabulated False Confession
The overall compliant, internalized, and confabulated false
confession rates were calculated by dividing the number of
false confessions of each type by the total number of inter-
views. These three rates were 17.6% compliant, 3.2% inter-
nalized, and 1.6% confabulated false confessions among the
total number of interviews. The number of compliant false
confession was higher in the AI condition (n = 21, 52.5% of
Accusatory interviews) than in the CIS (n = 1, 2.4% of CIS)
or IGM (n = 0) conditions, χ2 = 49.49, p < 0.001, Cramer’s
V = 0.63.
In order to explore whether the confession occurred
during the beginning, the middle, or the end of the inter-
view, an “elapsed interview time” variable was created
by dividing the duration of the interview up to the point
at which the confession was made by the total duration of
the interview. The scores corresponding to the beginning,
middle, and end of the interview were included within
the intervals 0–33%, 33–67%, and 67–100%, respectively,
(for an explanation of the subdivision of the interview,
see Kelly etal. 2016). The one compliant false confes-
sion made by a CIS participant occurred at the beginning
of the interview (more specifically, when 6.19% of the
interview had elapsed, i.e., during the Free Recall stage).
The same phenomenon occurred for 1 of the 21 compliant
false confessions of the AI participants, since 9.62% of the
interview had elapsed when the participant confessed, i.e.,
during the Isolation stage. The remaining 20 AI partici-
pants who confessed did so at the end of the interviews,
during the Minimization stage. There was no difference
across conditions for the internalized or confabulated false
Table 1 Means, standard
deviations, contrast tests, and
size effects for the correct
gathered details
t1 corresponds to the contrast test for the difference between CIS and IGM, t2 corresponds to the contrast
test for the difference between IGM and AI
CIS IGM AI
Category M SD M SD t1pM SD t2pCohen’s d
Actions 39.24 9.92 25.67 8.29 7.55 < .001 18.35 6.11 4.02 < .001 2.07
Objects 22.86 6.49 15.60 5.63 5.86 < .001 12.08 4.83 2.82 .003 1.55
Locations 4.83 3.39 2.09 1.89 5.54 < .001 0.33 0.57 0.57 .001 1.61
Persons 11.64 2.46 7.53 2.83 7.88 < .001 4.23 1.76 1.76 < .001 2.50
Time 3.45 1.86 1.91 1.63 4.76 < .001 1.25 0.71 0.71 .024 1.23
Interactions 25.98 9.36 22.42 7.09 2.17 .016 12.25 5.77 5.77 < .001 1.52
Journal of Police and Criminal Psychology
1 3
confessions, χ2s < 3.9 and ps > 0.142. Results for false con-
fession rates are presented in Table2.
Discussion
The aim of this study was to analyze the impact of inter-
view methods on innocent suspects, and more specifically
with regard to three aspects: the information gathered, the
reported misinformation induced by the leading questions,
and the false confessions.
Information Gathered
Cooperative methods provided more details than the adver-
sarial method. In addition, and as expected, the CIS yielded
a greater amount of correct information than the AI and IGM
interviews. In particular, the CIS elicited more details about
actions, objects, locations, people, time, and interactions than
the other two types of interviews. In general, the differences
between AI and the two other protocols are in line with the
literature on interview methods, which has long shown the
parallel between the use of information-gathering style meth-
ods and the increase in the amount of correct information in
interviewee’ statements compared to accusatory and standard
interviews (Leahy-Harland and Bull2017). Furthermore, in
view of this predictable difference between the cooperative
interviews and the AI, our goal was specifically to test the dif-
ference between IGM and CIS in terms of benefit for quantity
of details. The results regarding the difference between the
two cooperative methods can be explained in terms of Milne
and Bull’s (2002) study, according to which the joint use of
the Report-everything and the Mental context reinstatement
mnemonics is a way to obtain a large amount of informa-
tion, compared to the report-everything mnemonic alone, as
used in IGM suspects’ interviews. In addition, the benefit
of the CIS in terms of information collected may be due to
other instructions, such as Reverse order. Indeed, a study by
Bensi etal. (2011) showed that when a CI (for witnesses and
victims) was conducted without the Reverse order, Mental
context reinstatement, or Report-everything instructions, the
amount of information collected was less than with the full
protocol. In addition, the analyses did not reveal any differ-
ence in the amount of erroneous and confabulated details
gathered with these three types of interviews. However,
despite a significant p-value for this analyse, it did not seem
that CIS statements had a higher accuracy rate than IGM
statements and AI — due to the overlapping confidence
intervals.
More simply, it could be that the observed benefit to the
CIS in terms of information gathering is only due to the fact
that the protocol itself is longer and contains more stages.
Thus, we conducted timing analyses to look more specifi-
cally at this possibility. First, the results showed that the
CIS was the longest interview condition and the AI was the
shortest. The first possible conclusion, therefore, is that the
greater amount of detail collected with the CIS can therefore
be explained by the simple fact that the interview was longer
and contained more stages. For this reason, we performed
a calculation of the amount of detail obtained in one min-
ute, which then allowed us to limit the confounding effect
of interview length. Analyses of this new variable showed
that CIS statements always contained more correct details
per minute than AI statements. Furthermore, calculating the
amount of detail obtained in one minute of interviewee talk
showed that CIS statements contained more correct detail
per minute than IGM and AI statements. These results sup-
port the beneficial impact of the different components and
mnemonics of the protocol which seem to improve memo-
ries retrieval, and to help interviewees recalling more perti-
nent details during the interview (Memon etal. 2010).
Misleading Questions
Findings related to the misleading questions showed a ben-
efit of the CIS on the acceptance of misleading information.
As expected, when compared to the IGM interviewees, the
CIS interviewees were less likely to accept and confirm the
incorrect information induced by the interviewer during the
questions stage. It seems that, by improving the memory
retrievals, the joint action of the Report-everything and Men-
tal context reinstatement instructions may be responsible for
the Geiselman effect found in the present study. Indeed, a
study conducted by Drohan-Jennings etal. (2010) showed
that interviewees were more resistant to false suggestions —
induced by misleading questions — when they were inter-
viewed using a Mental context reinstatement mnemonic com-
pared to a structured interview including no such instruction.
These observations, as in our study, could be due to the fact
that Mental Context Reinstatement, through its action of
Table 2 Number and percentage
of compliant, internalized, and
confabulated false confessions
by interview style condition
Compliant confession Internalization Confabulation
N % N % N %
CIS 1 2.4% 0 0% 0 0%
IGM 0 0% 0 0% 0 0%
AI 21 52.5% 4 10% 2 5%
Journal of Police and Criminal Psychology
1 3
improving memory retrieval, also improves source monitoring
(Drohan-Jennings etal. 2010; Johnson etal. 1993). Since mis-
leading questions tend to induce misinformation by introduc-
ing nonexistent or erroneous memory items, then this effect
would be countered by participants paying more attention to
the source of their memory and information given in the inter-
view. Further studies should test this possibility by directly
measuring the link between presence/absence of Mental Con-
text Reinstatement and source monitoring by measuring the
amount of detail that an interviewer managed to induce in the
interviewee’s memory and recall.
Furthermore, an unexpected phenomenon occurred in
the AI condition with regard to resistance to misleading
questions: the results do not show a difference in reported
misinformation between participants in the AI and CIS con-
ditions. In other words, the AI interviewees did not accept
the influence of the misleading questions more than the
CIS interviewees did. However, given that the presence of
a Geiselman effect may be due to the high level of memory
recollection, we have no reason to think that such an effect is
responsible for the resistance to suggestion in the AI condi-
tion. Thus, further studies should be conducted in order to
determine whether this phenomenon can be repeated and, if
it can, identify the reasons that might explain why a higher
level of misinformation is reported in IGM than in AI.
False Confessions
The final objective of this experiment was to test the impact of
the interview methods on false confessions. As we expected,
the number of compliant confessions was significantly higher
in the AI condition than in the cooperative interview condi-
tions. These results are consistent with the meta-analysis by
Meissner etal. (2014) and clearly demonstrate the negative
impact of accusatory methods on the behavior of innocent
suspects during an interview when compared to information-
gathering methods. Moreover, the analyses show that very
few of the suspects who confessed to being responsible for
the computer crash actually internalized this guilt or added
confabulated details. These very low internalization and con-
fabulation scores might indicate a social impact of the inter-
viewer on the interviewee. Indeed, the majority of participants
who produced false confessions reported doing so in order to
meet the interviewer’s requirements, but did not really think
they were responsible for the event. These observations high-
light the impact of the obedience-to-authority phenomenon in
the interaction between the interviewer and the interviewee,
which is considered one of the causes of false confessions
(Shawyer etal. 2009).
The analyses also showed that one false confession
occurred in the CIS condition. This confession was made
in the first part of the interview, during the Free recall
stage. Since the cognitively demanding tasks (i.e., Reverse
order and Challenge stages) had not yet been completed
when the participant confessed, these results seem to show
that the false confession was not due to the CIS protocol,
but could be the consequence of the interviewee’s faulty
memory. Indeed, a study by Zaragoza etal. (2001) showed
that even when given the opportunity to avoid confabu-
lations by verbally categorizing true or invented details,
individuals still created false memories about experienced
events, regardless of internal criteria such as their age.
These results are explained by models of memory and by
Johnson and Raye’s (1998) theory of Reality Monitoring,
according to which episodic memory for real events can
be interfered with by invented memories, the source of
which is erroneously attributed to the event in question.
In the present study, therefore, a false memory may have
been at the origin of the participant’s confession in the CIS
condition, and this may also have been the case for one of
the participants in the AI condition, who also confessed
early in the interview.
Interestingly, and surprisingly, the results for false confes-
sions are inconsistent with previous observations on mis-
information reporting. Indeed, AI interviewees produced
more compliant false confessions than CIS and IGM inter-
viewees, possibly due to an obedience-to-authority phenom-
enon. If obedience to authority is a possible explanation for
the occurrence of false confessions, this should also be the
case for the acceptance of the misinformation induced by
the interviewer. However, this was not the case in our study.
One possible reason for these results could relate to the stage
of the AI being conducted at the time of confession. While
the interviewees are asked misleading questions in the sec-
ond stage of the interview (i.e., Accusatory stage), the inter-
viewer does not ask for a confession until the third stage (i.e.,
Minimization). In this third part of the interview, the inter-
viewer behaves deceitfully towards the participant in order to
create a false sense of security and a trust relationship. This
change in the interviewer’s behavior may be responsible for
the occurrence of false confessions. Indeed, in the study by
Klaver etal. (2008), participants in the “maximization” con-
dition (i.e., accusatory interview, aggressiveness towards the
participant) made fewer false confessions than participants
in the minimization condition. These findings support the
idea that confession-driven protocols or deceitful techniques,
such as those used with suspects by untrained or accusa-
tory methods-trained interviewers (Inbau etal. 2001; [name
deleted to maintain the integrity of the review process]),
have a negative impact. However, it is important to note
that because the minimization stage took place in the third
and last part of the interview, the participants’ tiredness or
willingness to end the interview could also be responsible
for their false confessions. Further studies in which the mini-
mization stage takes place at the beginning of the interview
should be conducted in order to assess its effect.
Journal of Police and Criminal Psychology
1 3
Limitations
One of the main limitations of the study is related to the
protocol itself. Indeed, although Kassin and Kiechel’s (1996)
paradigm has been developed and its similarity to the real-
life situation of a wrongly accused innocent person has been
acknowledged, the perceived consequences of the forbidden
event are very different from those facing a real suspect. In
this paradigm, the risk of the participants suffering serious
negative consequences for being responsible for the com-
puter crash was low. This limitation was also pointed out by
Horselenberg etal. (2006), who conducted a study that var-
ied the severity of the consequences for participants within
the forbidden touch paradigm. The results of this study
showed that suspects who perceived the consequences as
being more serious tended to confess less than those who did
not perceive serious consequences following a confession. A
replication of this study, in which the perceived seriousness
of the consequences of the event are varied, should therefore
be conducted in order to increase the level of realism.
Another limitation of this study is the use of a “struc-
tured” AI protocol. Indeed, although the CIS and IGM
methods are step-by-step protocols that can be applied fol-
lowing training, this is not the case with the accusatory
interview, which is not a method per se. Indeed, the AI
“protocol” used in our study, as described in the Method
section, was inspired by field observations taken from the
scientific literature (Kassin and Gudjonsson2004; Leahy-
Harland and Bull2017; Meissner et al. 2014; Noc and
Ginet2020; Noc etal. 2022; Russano etal. 2005). In real-
ity, there is no “standard accusatory protocol,” since such
interviews are, by definition, conducted by investigators who
have not received any interview training. Thus, all accusatory
interviews will be different and unique, since the “techniques”
used by interviewers in this condition are dependent on each
individual’s personal experiences [name deleted to maintain
the integrity of the review process]. In the field, therefore, it
seems that the risk of false confessions should increase only
when untrained interviewers conduct interviews practiced
in a manner quite similar to the adversarial interview used
in the present study (i.e., would include the same steps and
techniques of accusation, minimization and maximization).
Replicating this study with untrained, IGM-trained, and
CIS-trained field officers would permit a more realistic view
of the interviewees’ reactions to each interview condition.
Furthermore, most of the mock-suspects recruited for our
study were young women studying at university. These results
cannot be generalized to the entire population of suspects,
who come from diverse social classes and are mostly men.
Indeed, women represent 5–9% of those accused of crimes
and thus facing interrogation situations (Campaniello2019).
However, the results of the study by Klaver etal. (2008) sug-
gest that women are more likely to confess than men in a high
plausibility condition. In addition, a study by Forrest etal.
(2006) also showed that women were more likely to internal-
ize false confession than men. It seems that these observations
may be related to the link between gender and suggestibility.
Indeed, a study by Polczyk and Pasek (2006) suggested that
women could be more suggestible than men, which could be
the source of more false confessions. Thus, this specific popu-
lation is of interest to be studied in light of the increased risk
of false confessions. Furthermore, in view of the elements
previously presented, we can hypothesize that men would not
be more likely to produce false confessions when interviewed
with a CIS — even if this hypothesis should also be verified.
Furthermore, the results of the present study showed a gender
effect on the amount of detail recalled: in general, women
recalled more details than men. Since the men were under-
represented in this study, it is of interest to replicate this study
with more men, especially since they are predominant in real
criminal cases.
Finally, although it appears that the adversarial aspect of
the AI was responsible for the number of false confessions
obtained, compared to the cooperative protocols, we initially
thought that the cognitive load caused by the CIS might be
the cause of such a risk. However, we currently do not know
whether the low number of false confessions in CIS is due to
the benefit of mnemonics, or whether the perceived cogni-
tive load and the suggestibility of the participants was in fact
low. In a future CIS study, a measure of participants’ per-
ceived cognitive load, for each component, would be needed.
In addition, measures of suggestibility could be combined
to further explore the observations made by Otgaar etal.
(2012) and Alm etal. (2019).
Conclusion
The results of the current study regarding false confessions
highlight the need to provide field professionals with train-
ing in protocols focused on information collection. Moreo-
ver, the benefits of the CIS in terms of information gathering
and reducing the acceptance of misleading questions make
the interest of validating this protocol scientifically and of
disseminating it in the justice field even clearer. Further
studies should be conducted with guilty suspects in order
to analyze the diagnostic value of the confessions collected
in these three conditions. Ultimately, studies should be con-
ducted with field investigators in order to test the benefit of
the CIS protocol when used by professionals in comparison
with standard usual practices.
Data Availability The datasets generated and analyzed the current study
are available on the web platform OSF.io with the https:// osf. io/ 7hcze/?
view_ only= 99c24 5b1a9 70403 e9657 ad6cf d3ee6 bc. [Link:https:// osf.
io/ qvyxf/? view_ only= c2108 c91c2 084e6 a8e98 5cfaa 3cc87 d4].
Journal of Police and Criminal Psychology
1 3
Declarations
Ethical Approval Ethical approval was waived by the local Ethics Com-
mittee of University in view of the retrospective nature of the study and
all the procedures being performed (IRB00011540-2019–48).
Competing Interests The authors declare no competing interests.
References
Allwood CM, Ask K, Granhag PA (2005) The cognitive interview:
effects on the realism in witnesses’ confidence in their free recall.
Psychol Crim Law 11(2):183–198
Alm C, Helmy Rehnberg N, Lindholm T (2019) Language and
eyewitness suggestibility. J Investig Psychol Offender Profil
16(3):201–212
Bensi L, Nori R, Gambetti E, Giusberti F (2011) The enhanced cogni-
tive interview: a study on the efficacy of shortened variants and
single techniques. J Cogn Psychol 23:311–321. https:// doi. org/ 10.
1080/ 20445 911. 2011. 497485
Bull R, Soukara S (2010) Four studies of what really happens in police
interviews. In G. D. Lassiter & C. A. Meissner (Eds.), Police
interrogations and false confessions: Current research, practice,
and policy recommendations. (pp 81–95). Washington, DC. Am
Psychol Assoc. https:// doi. org/ 10. 1037/ 12085- 005
Campaniello N (2019) Women in crime.IZA World Labor (105v2)
Corson Y, Verrier N (2013) Les faux souvenirs. De Boeck Université,
Paris, Bruxelles
Dando C, Wilcock R, Milne R (2008) The cognitive interview: the
efficacy of a modified mental reinstatement of context procedure
for frontline police investigators. Appl Cogn Psychol 23:138–147.
https:// doi. org/ 10. 1002/ acp. 145
Dando CJ, Bull R (2011) Maximising opportunities to detect verbal
deception: Training police officers to interview tactically. J Inves-
tig Psychol Offender Profiling 8(2):189–202
Drohan-Jennings DM, Roberts KP, Powell MB (2010) Mental context
reinstatement increases resistance to false suggestions after chil-
dren have experienced a repeated event. Psychiatr Psychol Law
17:594–606. https:// doi. org/ 10. 1080/ 13218 71100 37391 10
Forrest KD, Wadkins TA, Larson BA (2006) Suspect personality,
police interrogations, and false confessions: maybe it is not just
the situation. Personality Individ Differ 40(3):621–628
Frosina P, Logue A, Book T, Huizinga S, Amos S, Stark S (2018)
The effect of cognitive load on nonverbal behavior in the cogni-
tive interview for suspects. Personality Individ Differ 130:51–58.
https:// doi. org/ 10. 1016/j. paid. 2018. 03. 012
Geiselman RE (2012) The cognitive interview for suspects (CIS). Am
J Forensic Psyhol 30:1–12
Geiselman RE, Fisher RP (2014) Interviewing witnesses and victims.
In M. St-Yves (Ed.). Investigative interviewing: Handbook of best
practices. Toronto: Toronto, ON: Thomson Reuters Publishers
Geiselman RE, Fisher RP, Cohen G, Holland H, Surtes L (1986) Eyewit-
ness responses to leading and misleading questions under the cogni-
tive interview. J Police Sci Adm 14:31–39
Geiselman RE, Fisher RP, Firstenberg I, Hutton LA, Sullivan S, Avetissian
I, Prosk A (1984) Enhancement of eyewitness memory: an empiri-
cal evaluation of the cognitive interview. J Police Sci Adm 12:74–80
Griffiths A, Milne R (2006) Will it all end in tiers? Police interviews
with suspects in Britain. In T. A. Williamson (Ed.) Investigative
Interviewing: rights, research, regulation (167–189). Devon: Willan
Guay S, Gagnon J, Gélinas S, Cantinotti M (2016) Guide d’introduction
à la puissance statistique avec le logiciel G*Power 3
Gudjonsson GH (2003) The psychology of interrogations and confes-
sions: A handbook. Wiley, Chichester, England
Horselenberg R, Merckelbach H, Smeets T, Franssens D, Ygram Peters
G-J, Zeles G (2006) False confessions in the lab: do plausibility
and consequences matter? Psychol Crime Law 12:61–75. https://
doi. org/ 10. 1080/ 10683 10042 00030 3076
Inbau FE, Reid JE, Buckley JP, Jayne BC (2001) Criminal interrogation
and confessions, 4th edn. Aspen, Gaithersburg, MD
Johnson MK, Hashtroudi S, Lindsay DS (1993) Source monitoring.
Psychol Bull 114(1):3
Johnson MK, Raye CL (1998) False memories and confabulation.
Trends Cogn Sci 2:137–145
Kassin SM, Gudjonsson GH (2004) The psychology of confessions:
a review of the literature and issues. Psychol Sci Public Interest
5:35–69. https:// doi. org/ 10. 1111/j. 1529- 1006. 2004. 00016.x
Kassin SM, Kiechel KL (1996) The social psychology of false confes-
sions: compliance, internalization, and confabulation. Psychol Sci
7:125–128. https:// doi. org/ 10. 1111/j. 1467- 9280. 1996. tb003 44.x
Kelly CE, Miller JC, Redlich AD (2016) The dynamic nature of inter-
rogation. Law Hum Behav 40:295–309. https:// doi. org/ 10. 1037/
lhb00 00172
Klaver J, Lee Z, Rose VG (2008) Effects of personality, interrogation
techniques and plausibility in an experimental false confession
paradigm. Leg Criminol Psychol 13:71–88. https:// doi. org/ 10.
1348/ 13553 2507X 193051
LaPaglia JA, Wilford MM, Rivard JR, Chan JCK, Fisher RP (2014)
Misleading suggestions can alter later memory reports even fol-
lowing a cognitive interview. Appl Cogn Psychol 28(1):1–9.
https:// doi. org/ 10. 1002/ acp. 2950
Leahy-Harland S, Bull R (2017) Police strategies and suspect responses
in real-life serious crime interviews. J Police Crim Psychol
32(2):138–151. https:// doi. org/ 10. 1007/ s11896- 016- 9207-8
Loftus EF, Palmer JC (1974) Reconstruction of automobile destruction:
an example of the interaction between language and memory. J
Verbal Learn Verbal Behav 13:585–589
Logue M, Book AS, Frosina P, Huizinga T, Amos S (2015) Using real-
ity monitoring to improve deception detection in the context of the
cognitive interview for suspects. Law Hum Behav 39:360–367.
https:// doi. org/ 10. 1037/ lhb00 00127
Luke TJ, Hartwig M, Joseph E, Brimbal L, Chan G, Dawson E, Jordan
S, Donovan P, Granhag PA (2016) Training in the strategic use of
evidence technique: improving deception detection accuracy of
American law enforcement officers. J Police Crim Psychol 31:270–
278. https:// doi. org/ 10. 1007/ s11896- 015- 9187-0
Meissner CA, Redlich AD, Michael SW, Evans JR, Camilletti CR,
Bhatt S, Brandon S (2014) Accusatorial and information-
gathering interrogation methods and their effects on true and false
confessions: a meta-analytic review. J Exp Criminol 10:459–486.
https:// doi. org/ 10. 1007/ s11292- 014- 9207-6
Memon A, Meissner CA, Fraser J (2010) The cognitive interview:
a meta-analytic review and study space analysis of the past 25
years. Psychol Public Policy Law 16:340–372. https:// doi. org/ 10.
1037/ a0020 518
Memon A, Zaragoza M, Clifford BR, Kidd L (2009) Inoculation or
antidote? The effects of cognitive interview timing on false mem-
ory for forcibly fabricated events. Law Hum Behav 34:105–117.
https:// doi. org/ 10. 1007/ s10979- 008- 9172-6
Milne R, Bull R (2002) Back to basics: a componential analysis of the
original cognitive interview mnemonics with three age groups.
Appl Cogn Psychol 16:743–753. https:// doi. org/ 10. 1002/ acp. 825
Mundorf A, Lazarus LT, Uitvlugt MG, Healey MK (2021) A test
of retrieved context theory: dynamics of recall after incidental
encoding.J Exp Psychol: Learn Memory Cognition
Noc M, Ginet M (2020) Perceptions des pratiques en matière d’audition
de suspects: Une évaluation auprès des agents des douanes fran-
çaises. Criminologie 53(2):255–287
Journal of Police and Criminal Psychology
1 3
Noc M, Ginet M, Deslauriers‐Varin N (2022) The cognitive interview
for suspects: A test with customs officers. J Investig Psychol
Offender Profiling 1–18
Otgaar H, Alberts H, Cuppens L (2012) How cognitive resources alter
our perception of the past: ego depletion enhances the susceptibil-
ity to suggestion. Appl Cogn Psychol 26(1):159–163
Polczyk R, Pasek T (2006) Types of suggestibility: relationships among
compliance, indirect, and direct suggestibility. Int J Clin Exp
Hypn 54(4):392–415
Redlich AD, Goodman GS (2003) Taking responsibility for an act not
committed: influence of age and suggestibility. Law Hum Behav
27:141–156
Russano MB, Meissner CA, Narchet FM, Kassin SM (2005) Inves-
tigating true and false confessions within a novel experimental
paradigm. Psychol Sci 16(6):481–486
Sharman SJ, Powell MB (2012) A comparison of adult witnesses’ sug-
gestibility across various types of leading questions. Appl Cogn
Psychol 26:48–53. https:// doi. org/ 10. 1002/ acp. 1793
Shawyer A, Milne B, Bull R (2009) Investigative interviewing in the
UK. In T. Williamson, B. Milne, & R. Bull (Eds.), International
developments in investigative interviewing (pp 24–38). Devon,
England: Willan
Sigurdsson JF, Gudjonsson GH (1996) The psychological characteris-
tics of false confessors: a study among Icelandic prison inmates
and juvenile offenders. Personality Individ Differ 20:321–329.
https:// doi. org/ 10. 1016/ 0191- 8869(95) 00184-0
Stein LM, Memon A (2006) Testing the efficacy of the cognitive inter-
view in a developing country. Appl Cogn Psychol 20:597–605.
https:// doi. org/ 10. 1002/ acp. 1211
Stewart J, Woody W, Pulos S (2018) The prevalence of false confes-
sions in experimental laboratory simulations: a meta-analysis.
Behav Sci Law 36:12–31. https:// doi. org/ 10. 1002/ bsl. v36.1
Snook B, Luther K, Quinlan H, Milne R (2012) Let ’em talk!: a field
study of police questioning practices of suspects and accused per-
sons. Crim Justice Behav 39:1328–1339. https:// doi. org/ 10. 1177/
00938 54812 449216
Tulving E (2002) Episodic memory: from mind to brain. Annu Rev
Psychol 53:1–25
Vrij A, Mann S, Fisher R, Leal S, Milne B, Bull R (2008) Increasing
cognitive load to facilitate lie detection: the benefit of recalling
an event in reverse order. Law Hum Behav 32:253–265. https://
doi. org/ 10. 1007/ s10979- 007- 9103-y
Vrij A, Fisher R, Mann S, Leal S (2006) Detecting deception by manip-
ulating cognitive load. Cogn Sci 10:141–142
Vrij A, Leal S, Mann S, Warmelink L, Granhag PA, Fisher RP (2010)
Drawings as an innovative and successful lie detection tool. Appl
Cogn Psychol 24:587–594. https:// doi. org/ 10. 1002/ acp. 1627
Walsh DW, Bull RH (2010) What really is effective in interviews with
suspects? A study comparing interview skills against interview
outcomes. Leg Criminol Psychol 15:305–321. https:// doi. org/ 10.
1348/ 13553 2509X 463356
West E, Meterko V (2016) Innocence project: DNA exonerations,
1989–2014: Review of data and findings from the first 25 years.
Albany Law Rev 79:717–795
Wheeler RL, Gabbert F (2017) Using self-generated cues to facilitate
recall: a narrative review. Front Psychol 8:1830
Yarmey AD (2009) Truthfulness in witnesses’ and suspects’ reports.
In R. Bull, T. Valentine & T. Williamson (Eds.), Handbook of
Psychology of Investigative Interviewing: Current Developments
and Future Directions. Singapore: Wiley
Zaragoza MS, Payment KE, Ackil JK, Drivdahl SB, Beck M (2001)
Interviewing witnesses: forced confabulation and confirmatory
feedback increase false memories. Psychol Sci 12(4):73–77
Publisher's Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Springer Nature or its licensor holds exclusive rights to this article under
a publishing agreement with the author(s) or other rightsholder(s);
author self-archiving of the accepted manuscript version of this article
is solely governed by the terms of such publishing agreement and
applicable law.
