Full text of DOI:10.1348/135532500167976

Legal and Criminological Psychology (2000), 5, 57–70 Printed in Great Britain
Ó 2000 The British Psychological Society
57
V
The e ect of informing liars about
Criteria-Based Content Analysis on their
ability to deceive CBCA-raters
Aldert Vrij*, Wendy Kneller
Samantha Mann
and
Psychology Department, University of Portsmouth, UK
As soon as liars realize that evaluators use CBCA to assess the credibility of their
statements, it is possible that liars will gain knowledge of CBCA and try to
‘improve’ their statements in order to make an honest impression on CBCA-
judges. The present experiment investigated to what extent liars are capable of
doing this. In all, 45 participants were randomly allocated to one of the following
three conditions: a truth telling condition in which participants were asked to
recall a videotaped event which they had just seen; an uninformed deception
condition in which participants who had only been given guidelines about the
content of the videotaped event were asked to recall the event as though they had
seen the videotape; and an informed deception condition in which participants
received information about CBCA before they were asked to pretend that they had
seen the videotape. CBCA-raters scored the accounts and a comparison was made
between the total CBCA-scores of the three conditions. The study also examined
the extent to which CBCA-assessments could correctly classify truthful and
deceptive accounts, Ž rst by means of a discriminant analysis (with the total
CBCA-score as dependent variable) and secondly by asking a British CBCA-expert
to judge the veracity of the statements. The results indicated that liars are capable
of in uencing CBCA-assessments. First, the CBCA-scores of liars who were
informed about CBCA were similar to the CBCA-scores of truth tellers and
signiŽ cantly higher than the CBCA-scores of liars who were not informed about
CBCA. Secondly, the objective status of the participant (truth teller vs. informed
liar) could not be successfully predicted in a discriminant analysis on the basis of
total CBCA-scores. Thirdly, statements of the majority of informed liars were
assessed as truthful by a British CBCA-expert.
To date, Criteria-Based Content Analysis is probably the most widely used
instrument to assess the veracity of written statements. CBCA-assessments are
accepted as evidence in several courts in the United States and in courts in some
Western European countries, such as Germany and the Netherlands (Vrij, 1998,
2000).
CBCA was developed in order to evaluate statements from children who are
witnesses or alleged victims, most commonly of sexual abuse. Many authors still
*Requests for reprints should be addressed to Aldert Vrij, University of Portsmouth, Psychology Department,
King Henry Building, King Henry 1 Street, Portsmouth PO1 2DY, UK (e-mail: aldert.vrij@port.ac.uk).

58
Aldert Vrij et al.
describe CBCA as a technique solely developed to evaluate statements of children
in sexual o ence trials (Honts, 1994; Horowitz et al., 1997). Others, however,
V
advocate the use of the technique to evaluate the testimonies of adults who talk
about issues other than sexual abuse as well (Ko¨hnken, Schimossek, Aschermann,
& Ho¨fer, 1995; Porter & Yuille, 1996; Steller & Ko¨hnken, 1989).
In principle, it is possible that if people were to learn which methods
CBCA-evaluators use to assess the credibility of their statements, they could try to
‘improve’ their statements in order to obtain a truthful assessment from CBCA-
judges. One might wonder how likely it is that liars will seek information about
CBCA and train themselves to beat the technique. It probably depends both on the
liars’ motivation to avoid getting caught and on the likelihood of the technique
being used against them. Highly motivated liars, such as criminals, spies and
adulterous partners, will probably show more interest in methods of fooling lie
detectors than less motivated liars. Whether they will actually obtain information
about CBCA and actually train themselves to improve their statements will
probably depend on their perceptions of the likelihood of the technique being used
against them. It is therefore unlikely that adulterous partners will seek information
about CBCA as there is little chance of their partners ever using a CBCA-
assessment against them. On the other hand, criminals and spies might think that
they run a realistic risk of observers using CBCA and therefore might train
themselves to fool CBCA-judges.
The present experiment examines to what extent adult witnesses are capable of
fooling CBCA-judgments. Before outlining the background and procedure of the
experiment, the following brie y outlines CBCA, present results of previous
CBCA-experiments, and possible explanations for why CBCA discriminates
between truth tellers and liars.
Steller and Ko¨hnken (1989) compiled a list of 19 assessment criteria which can
be used for evaluating the veracity of a statement. The assessment of statements
based upon this list is called Content-Based Criteria Analysis, or CBCA. Trained
evaluators examine a written statement and judge the presence or absence of each
of the 19 criteria, usually on a 2-point scale where 0 is assigned if the criterion is
absent, and 1 if the criterion is present. Scores per evaluation may therefore vary
from 0 to 19. CBCA focuses on the presence of speciŽ c semantic content
characteristics. The underlying hypothesis is that a statement derived from an actual
memory of an experience di ers in content and quality from a statement based on
V
invention or fantasy, and that only a person who has actually experienced an event
is likely to incorporate certain types of content into a statement. In other words, the
presence of each criterion in the statement enhances the quality of the statement
and strengthens the hypothesis that the account is based on genuine personal
experience. The Appendix provides a brief description of each of the 19 criteria.
Vrij and Akehurst (1998) and Vrij (2000) give more detailed descriptions of the list
of criteria.
In a recent review, Vrij (2000) examined the accuracy of CBCA-assessments in
12 experimental CBCA-studies (Akehurst, 1997; Akehurst, Ko¨hnken, & Ho¨fer,
1995; Ho¨fer, Akehurst, & Metzger, 1996; Ko¨hnken et al., 1995; Landry & Brigham,
1992; Porter & Yuille, 1996; Ruby & Brigham, 1998; Sporer, 1997; Steller,

Informing liars about CBCA
59
Wellershaus, & Wolf, 1988; Vrij, Edward, Roberts, & Bull, 1999; Yuille, 1988;
Zaparniuk, Yuille, & Taylor, 1995). It was found that, on average, evaluators
correctly classiŽ ed 73% of the statements. A distinction between ‘detecting truths’
and ‘detecting lies’ showed that 76% of truthful statements and 70% of the
deceptive statements were correctly classiŽed. These Ž ndings indicate that by using
CBCA-assessments it is possible to classify truthful and deceptive statements above
the level of chance.
There are several reasons why truth tellers often obtain higher CBCA-scores than
liars (Vrij, 2000).
(1) When fabricating a story people do not have enough imagination to invent the
relevant characteristics. For example, a liar is often not creative enough to
introduce complications or to reproduce parts of the conversation, or to
describe their own mental state or someone else’s mental state.
(2) Alternatively, fabricators may be creative enough to incorporate these charac-
teristics in their stories, but do not realize that these characteristics are relevant
for credibility judgments and therefore, ultimately, do not include them into
their accounts.
(3) Alternatively, it is too diYcult to incorporate the criteria in a fabricated
statement. For instance, unstructured production (an important characteristic
of truthful accounts) is diYcult to incorporate in fabricated accounts. It is
much easier to recount a fabricated story in a chronological order.
(4) Finally, fabricators may not want to include certain characteristics (for example,
admitting lack of memory, or raising doubts about their own testimony)
because they believe that these characteristics will have a negative impact on
the credibility of their statement.
It is possible that on realizing which method evaluators use to assess the credibility
of their statements, liars will gain knowledge of CBCA and so may try to improve
their statements in order to make an honest impression on CBCA-judges. It is
possible that liars who are informed about CBCA will obtain higher CBCA-scores
than uninformed liars, because, as stated in reason (2), uninformed liars probably
do not realize that certain characteristics are relevant for credibility judgments and,
therefore, do not include these characteristics into their accounts.
In the present experiment, liars were asked to ‘recall’ a videotaped event that they
had not actually seen (instead a brief description of the event was given). Before
‘recalling’ the event, half of the liars were informed about nine CBCA-criteria,
namely criterion 2 (unstructured production), criterion 3 (quantity of details),
criterion 4 (contextual embedding), criterion 8 (unusual details), criterion 9
(super uous details), criterion 13 (attribution of perpetrator’s mental state),
criterion 14 (spontaneous corrections), criterion 15 (admitting lack of memory), and
criterion 16 (raising doubts about one’s own memory). It was expected that
informed liars would obtain higher CBCA-scores than uninformed liars, particularly
with regard to the coached nine criteria (Hypothesis 1).
Another group of participants was also included who saw the videotaped event
and recalled the event afterwards as a control. Comparisons between the liars
and truth tellers enabled examination of the extent to which informing liars of

60
Aldert Vrij et al.
CBCA-criteria actually improves the quality of their accounts. It was expected that
the informed liars’ accounts would be as good as the truth tellers’ accounts; there-
fore, no signiŽcant di erences between CBCA-scores of informed liars and truth
V
tellers was expected, but it was expected that the CBCA-scores of uninformed liars
would be signiŽ cantly lower than the CBCA-scores of truth tellers (Hypothesis 2).
The present experiment also examined to what extent CBCA-assessments can
correctly classify truthful and deceptive statements. This was investigated in two
di erent ways: Ž rst, by including the total CBCA-score in a discriminant analysis to
V
predict objective status of truth tellers (i.e. informed liars and uninformed liars);
and secondly, by asking a CBCA-expert to judge the veracity of the participants’
statements. Both methods should have obtained similar results, as, theoretically,
CBCA-experts calculate total CBCA-scores for each statement and then make
veracity judgments on the basis of these total CBCA-scores. However, in practice,
CBCA-experts do not always assess statements on the basis of total CBCA-scores
(Vrij, 2000). The main problem with judging the veracity of statements on the basis
of total CBCA-scores is that it does not take into account the fact that some criteria
may be more valuable in assessing truthfulness than others (Steller & Ko¨hnken,
1989). For example, the presence of accurately reported but misunderstood details
in a statement (criterion 10), such as a child who describes an adult’s sexual
behaviour but attributes it to a sneeze or to pain, is apparently more signiŽcant than
relatively simple descriptions of interactions (criterion 5) or contextual embedding
(criterion 4) (Steller & Ko¨hnken, 1989).
It is possible that CBCA-experts will become suspicious when they read
statements of informed liars. Liars might exaggerate when they deliberately include
CBCA-criteria into their accounts, and as a result some CBCA-criteria might be too
obviously present. For example, an informed liar may admit to a lack of memory so
often in a fabricated account that a CBCA-expert may become suspicious. It was
therefore expected that whilst total CBCA-scores would be ine ective in predicting
V
the objective status of truth tellers, uninformed liars and informed liars, it might be
possible for a CBCA-expert to accurately distinguish between the three groups
(Hypothesis 3).
Method
Participants
Undergraduates (N = 45; 33 females and 12 males) of the University of Portsmouth (UK) were
recruited under the premise of participating in an experiment about ‘recalling events’. The participants
varied in age from 19 to 46 years (M = 26.3 years). The Ž rst language of all participants was English.
Procedure
Participants were randomly assigned to the truth telling condition (N = 15), informed deception
conditon (N = 15) or uninformed deception condition (N = 15) and participated individually. Males
and females were equally distributed among the three conditions. Those who were assigned to the
truthful condition were told by the experimenter that this was a study to investigate memory and that
as part of this study they would be shown a video during which they should aim to try to remember

Informing liars about CBCA
61
as much as they could about what they would see. They were then shown a video of 2 min 3 s in
length. This videotaped event featured a colour presentation of a theft of a camera bag from a man
and woman, by a female stranger, while they were sitting on a park bench. The event was Ž lmed from
the perspective of a bystander, sitting directly opposite the couple in the park. In the video, a man and
a woman are sitting next to each other on a bench in a park. The woman is feeding pigeons. After a
short time, a woman walks along in front of the couple and sits at the end of the bench, next to the
grey camera bag. After she sits down, she gets a mobile ’phone out of her bag and appears to make
a phone call, all the time watching the couple and the grey camera bag. As the woman is making the
call, the Ž rst woman moves up closer to the man, leaving the grey bag where it is. The man gets a
newspaper out of his bag, and starts reading it. The Ž rst woman then starts reading the newspaper
with him. When the second woman has Ž nished her call, she puts the ’phone back into her bag, and
looks at the couple to see what they are doing. Noticing that they are reading the paper, she picks up
her things and the grey camera bag and walks o in the opposite direction from which she came. A
V
few moments later, the couple prepare to leave the bench. The man puts the newspaper away, and the
woman reaches for the grey camera bag. Upon Ž nding it missing, she stands up abruptly and tells the
man that her bag has been taken. She then marches o in the direction that the second woman took.
V
After the participants had seen the video, they were taken to another room where the interview
took place. The interviews were standardized (also across the three conditions). The interviewer was
blind as to which condition the participants were allocated to and said to each participant: ‘I would
now like you to tell me all you can remember about the event you have just seen.’ After the
interviewee had Ž nished their free recall the interviewer asked: ‘Is there anything else you can tell me
about the event?’ After the interview, participants were thanked for their participation and sent back
to the experimenter, who debriefed them.
Those who were assigned to the uninformed deception condition were told by the experimenter
that their task was to deceive the interviewer by giving a description of an event so convincing that
she will actually think that they really saw it on a videotape. The participants were then given a printed
copy with the following brief description of the event that had been shown to the truth tellers:
— A man and a woman are sitting on a park bench feeding pigeons
— By the side of the woman is a bag
— A second woman walks along and sits on the bench, next to the bag
— This woman then makes a ’phone call on a mobile ’phone, whilst the man and the Ž rst woman
read a newspaper
— While they are not looking, the second woman takes the bag and walks o
— A few moments later, the Ž rst woman notices that her bag has gone.
V
When the participant had read the description, the experimenter asked him/her to construct their
version of the event and to Ž ll in any appropriate details that were not mentioned in the description.
She also told participants not to feel rushed but take as much time as they required. When the
participant said he or she was ready, he or she was taken to the interviewer and the interview took
place (see truthful condition). After the interview, participants were thanked for their participation
and sent back to the experimenter for debrieŽng.
Those who were assigned to the informed deception condition received the same instructions and
information as those who were assigned to the uninformed deception condition. Additionally, after
they had read the brief description of the event, the following information was given:
‘To help with your task to deceive the interviewer, I am going to give you some additional
information which research has shown to be useful when trying to deceive another person.
First, you should try to include as many details as you can in your account of the event. For
example, what the people looked like or the clothes they were wearing [quantity of details:
criterion 3]. You will sound more convincing if you place the event in time and location, i.e.
when and where the event took place [contextual embedding: criterion 4]. Mentioning unusual
and super uous details can also help to deceive the interviewer. An unusual detail could be, for
example, a visible birthmark or scar on one of the persons involved in the event. An example
of a super uous detail could be perhaps seeing a child on a bike in the far distance [unusual and

62
Aldert Vrij et al.
super uous details: criteria 8 and 9]. If you have forgotten something which you think should
be mentioned, it is all right to go back to it at the end of your account [unstructured
production: criterion 2]. It also helps if you can describe the perpetrator’s feelings, thoughts or
motives during the incident—in this case, what the woman who stole the bag may have been
feeling or thinking [attribution of perpetrator’s mental state: criterion 13]. If you say something
which you subsequently feel to be wrong, it is all right to correct yourself [spontaneous
corrections: criterion 14]. It also helps to admit sometimes that you don’t remember some
things [admitting lack of memory: criterion 15]. Lastly, it helps to raise doubts about your
account of the event, e.g. expressing concern that some part of the statement seems incorrect
or unbelievable [raising doubts about one’s memory: criterion 16]. I will now leave you for a
few minutes with a printed copy of the information I have just given you concerning successful
deception and with a printed copy of the description of the event in order for you to construct
your version of the event. Please don’t feel rushed. Take as much time as you require.’
When participants said they were ready, they were taken to the interviewer and the interview took
place (see truth teller condition). After the interview, participants were thanked and sent back to the
experimenter, who debriefed them.
Transcribing and coding
All interviews were audiotaped (participants were told that their interviews would be audiotaped so
that the experimenter could listen to their statements later on) and were transcribed verbatim from the
audiotapes. Two independent raters with recent training1 in CBCA-scoring by a British CBCA-expert
judged the presence of each of 12 CBCA-criteria2 and assigned a rating of 0–1 (absent or present).
With regard to quantity of details (criterion 3) the number of details were counted as in Porter and
Yuille (1996). After all interviews had been scored it was decided by the two raters, on the basis of
the number of details that had been mentioned in the interviews, that this criterion was fulŽ lled if at
least 16 details were mentioned. The raters were blind to the experimental conditions and hypotheses
in this study.
Results
Inter-rater agreement
Inter-rater agreement was investigated by calculating the number of times each of
the two raters agreed on the presence or absence of a criterion in each of the 45
statements (so-called proportion agreement (Anson, Golding, & Guly, 1993)).
Overall, the inter-rater proportion agreement was very high, speciŽ cally .90. For
each criterion, inter-rater proportion agreements were high (between .87 and 1.00)
for 11 out of 12 criteria. A much lower inter-rater proportion agreement was
obtained for unstructured production (criterion 2), speciŽ cally .53.3 Given the high
overall proportion agreement between the two coders, it was decided to combine
the scores of both evaluators and to conduct statistical analyses on the mean scores
1 First, both raters read all the major papers about CBCA (they read most of the papers cited in the reference list
of this article). They were then trained in CBCA scoring by a British CBCA-expert. Both the trainee raters and the
expert rater evaluated several example transcripts (from a diVerent study), then the three raters compared their
results and feedback was given by the expert rater.
2 Given the fact that in this study statements of adults who witnessed an event were used, we thought that several
criteria would be inappropriate and were therefore ignored. These criteria were: reproduction of speech (criterion
6, because no conversation could be heard in the event); accurately reported details misunderstood (criterion 10);
related external associations (criterion 11); accounts of subjective mental state (criterion 12); self-deprecation
(criterion 17); pardoning the perpetrator (criteron 18); and details characteristic of the oVence (criterion 19).
3 Anson et al. (1993) obtained a similar, low proportion agreement (.57) for this criterion.

Informing liars about CBCA
63
Table 1.
Mean scores of CBCA criteria as a function of deception
Experimental condition
Uninformed Informed
Truthful
liars liars
Criteria
M
SD
M
SD
M
SD
General characteristics
1. Logical structure
2. Unstructured production*
3. Quantity of details*
.90a
.37a
.57b
(.20) 1.00a (.00) .93a (.18)
(.35) .20a (.25) .33a (.31)
(.50) .17a (.36) .53b (.48)
SpeciŽc contents
4. Contextual embedding*
5. Descriptions of interactions
7. Unexpected complications
8. Unusual details*
.73a
.13a
.00a
.37b
.00a
.40b
(.46) .83a (.36) .93a (.26)
(.35) .07a (.26) .10a (.28)
(.00) .00a (.00) .00a (.00)
(.40) .00a (.00) .03a (.13)
(.00) .00a (.00) .10a (.28)
(.47) .13a (.35) .50b (.50)
9. Super uous details*
13. Attribution of perpetrator’s mental state*
Motivation-related contents
14. Spontaneous corrections*
15. Admitting lack of memory*
.30b
.20a
(.46) .07a (.26) .10ab (.28)
(.14) .07a (.26) .13a (.35)
16. Raising doubts about one’s own testimony* .37ab (.48) .13a (.35) .47b (.44)
Total score
4.33b (1.73) 2.67a (1.10) 4.17b (.96)
*Coached criteria.
Mean scores with a diVerent superscript diVer signiŽ cantly from each other (p<.05).
of both evaluators. Therefore the scores regarding unstructured production should
be interpreted with caution owing to the lack of strong agreement between the two
evaluators.
Analyses of variance
In order to test Hypothesis 1 (informed liars will obtain higher CBCA-scores than
uninformed liars, particularly regarding the coached criteria) an ANOVA was
conducted with Liar (informed liars vs. uninformed liars) as the independent
variable and the total CBCA-score as the dependent variable. The ANOVA
g2
revealed a signiŽ cant e ect (F(1,28) = 15.93, p< .01,
= 36%). As predicted, the
V
CBCA-score for informed liars was signiŽ cantly higher (M = 4.17, SD = .96) than
the CBCA-score for uninformed liars (M = 2.67, SD = 1.10). Moreover, ANOVAs
were conducted comparing informed and uninformed liars on the individual
CBCA-criteria. Table 1 provides the mean scores and a summary of the ANOVA
results for each of the criteria.
As can be seen in Table 1 (columns 2 and 3), several signiŽ cant di erences
V
emerged between informed and uninformed liars. Informed liars included more
g2
details in their accounts (criterion 3; F(1,28) = 5.57, p< .05,
= 17%), made more

64
Aldert Vrij et al.
attributions of the perpetrator’s mental state (criterion 13; F(1,28) = 5.40, p< .05,
g2
= 16%), and raised more doubts about their own memories (criterion 16;
g2
F(1,28) = 5.22, p< .05,
= 16%) than uninformed liars. All of these criteria were
coached criteria. This supports Hypothesis 1.
Table 1 also gives insight into how often a criterion was included in a statement
(0.00 means that a criterion was totally absent and 1.00 means that a criterion was
always present). The Ž ndings show that logical structure (criterion 1) and
contextual embeddings (criterion 4) were very often present, whereas descriptions
of interactions (criterion 5) and super uous details were hardly ever present.
Unexpected complications (criterion 7) were never present.
In order to test Hypothesis 2 (truth tellers and informed liars will have similar
CBCA-scores, whereas CBCA-scores of uninformed liars will be signiŽ cantly lower
than the CBCA-scores of truth tellers) two ANOVAs were conducted, comparing
the total CBCA-score for truth tellers and uninformed liars (Ž rst ANOVA) and
comparing the total CBCA-score for truth tellers and informed liars (second
ANOVA). As predicted, uninformed liars obtained a signiŽ cantly lower CBCA-
score (M = 2.67, SD = 1.10) than truth tellers (M = 4.33, SD = 1.73; F(1,28) = 9.94,
g2
p< .01
= 26%), whereas no signiŽ cant di erence was found between the
V
CBCA-scores of truth tellers and informed liars (F(1,28) = .35, n.s.). These Ž ndings
support Hypothesis 2.
ANOVAs comparing truth tellers and informed liars on the individual CBCA-
criteria revealed only one signiŽ cant e ect. Truth tellers included more unusual
V
details in their accounts than informed liars (criterion 8, see Table 1, columns 1 and
g2
3; F(1,28) = 9.46, p< .01,
= 25%). Hence, for only one of the 12 individual
criteria did informed liars obtain a signiŽ cantly lower score than truth tellers. This
further underlines the e ectiveness of informing liars about CBCA to improve their
V
fabricated accounts.
Finally, ANOVAs were conducted to test di erences between truth tellers and
V
uninformed liars on the individual CBCA-criteria. Five e ects were signiŽ cant and
V
Table 1 provides a summary of these e ects (columns 1 and 2) Uninformed liars
V
scored signiŽcantly lower than truth tellers on criterion 3 (F(1,28) = 6.39, p<.05,
g2
g2
= 31%),
= 19%), criterion 8 (unusual details; F(1,28) = 12.64, p< .01,
criterion 13 (attribution of perpetrator’s mental state; F(1,28) = 3.09,
g2
p< .05,
p< .05,
= 10%), and criterion 14 (spontaneous corrections; F(1,28) = 2.98,
= 10%).
g2
Accuracy rates
Hypothesis 3 (CBCA-scores would be ine ective in predicting the objective status
V
of truth tellers, informed liars and uninformed liars, but a CBCA-expert might be
able to accurately distinguish between the three groups) was tested in two steps.
First, a discriminant analysis was conducted, using the total CBCA-scores to predict
the objective status of truth tellers, informed liars and uninformed liars.
The analysis yielded a highly signiŽ cant discriminant function (eigenvalue = .35,
k
v2
= .74, = 12.71, p< .01). Overall, 26 (58%) of the cases were correctly classiŽed.
A distinction into truth tellers, informed liars and uninformed liars shows that 53%

Informing liars about CBCA
65
Table 2.
Correct identiŽcation of true and false statements by experimental condition
Method of assessment
Experimental condition
Discriminant
Expert
Informed expert
Truth tellers
Uninformed liars
Informed liars
Overall
53%
80%
40%
58%
80%
60%
27%
56%
73%
60%
40%
58%
of the truth tellers, 80% of the uninformed liars and 40% of the informed liars were
correctly classiŽ ed (see also the Ž rst column of Table 2), whereas 13% of the
uninformed liars and 47% of the informed liars were classiŽ ed as truth tellers.4
Hence, the percentage of informed liars correctly classiŽed as liars (40%) was nearly
equal to the percentage of informed liars which were incorrectly classiŽ ed as truth
tellers (47%). This indicates that total CBA-scores do not successfully discriminate
between liars and truth tellers when liars are informed about CBCA, which is in line
with Hypothesis 3.
Secondly, a British CBCA-expert (blind to the experimental manipulations and to
the hypotheses in the experiment) was asked to judge the veracity of each
statement. The expert received training about CBCA by leading German CBCA-
experts and has completed a PhD thesis on CBCA-research. Results revealed (see
also column two of Table 2) that she correctly classiŽed 80% of the truth tellers as
truth tellers and 60% of the uninformed liars as liars. However, she only classiŽed
27% of the informed liars as liars. In other words, the majority of informed liars
managed to successfully deceive the CBCA-expert. After making her classiŽ cations,
the CBCA-expert was informed about the nature of the experiment and was told
that some liars had been informed about CBCA-criteria in order to deceive her. She
was then asked to judge the statements for a second time. The results of her second
assessment are presented in the third column of Table 2. Table 2 shows that this
time she correctly classiŽ ed 73% of truth tellers as truth tellers, 60% of the
uninformed liars as liars and 40% of the informed liars as liars. Hence, even after
being informed about the nature of the experiment, the majority of informed liars
managed to successfully deceive the CBCA-expert. Hypothesis 3, which stated that
it might be possible for a CBCA-expert to classify informed liars, is therefore
rejected.
A comparison between the assessments made by the computer (which were
based upon the total CBCA-scores) and the judgments made by the CBCA-expert
shows that the CBCA-expert was more accurate in classifying truth tellers, whereas
the computer was more accurate in classifying uninformed liars. Moreover, the
CBCA-expert showed a ‘truth bias’ (she was more accurate in classifying truths than
4Of the truth tellers, 47% were classiŽ ed as uninformed liars and 0% as informed liars. Of the uninformed liars,
13% were classiŽ ed as truth tellers and 7% as informed liars. Of the informed liars, 47% were classiŽ ed as truth
tellers and 13% as uninformed liars.

66
Aldert Vrij et al.
lies), whereas the computer showed a ‘lie bias’. Both computer and CBCA-expert
were inaccurate in classifying informed liars.
Discussion
The present experiment examined to what extent liars, who are informed about
CBCA, can produce statements which are comparable in terms of content and
quality with statements given by truth tellers. The results indicate that informed
liars are capable of doing this. First, the CBCA-scores of liars who were informed
about CBCA were similar to the CBCA-scores of truth tellers, and signiŽ cantly
higher than the CBCA-scores of liars who were not informed about CBCA.
Secondly, the objective status of the participant (truth teller vs. informed liar) could
not be successfully predicted in a discriminant analysis on the basis of total
CBCA-scores. Thirdly, statements of the majority of informed liars were assessed
as truthful by a British CBCA-expert.
In the present experiment, liars were informed about nine CBCA-criteria. Three
of these criteria, quantity of details (criterion 3), attribution of perpetrator’s mental
state (criterion 13) and raising doubts about one’s own memory (criterion 16) were
more often present in the statements of informed liars than in the statements of
uninformed liars.
One suggested reason why informing liars about unstructured production
(criterion 2), unusual details (criterion 8) and super uous details (criterion 9) failed
to have an e ect on informed liars is that it might be too diYcult to fabricate these
V
criteria. It is probably much easier to tell a fabricated story in a structured way
than in an unstructured way. Including unusual and super uous details requires a
great deal of imagination: more imagination than the liars in this experiment
demonstrated.
A possible explanation for why informing liars about contextual embeddings
(criterion 4) had no e ect is that this criterion was also found to be present in most
V
statements of truth tellers and uninformed liars. This is perhaps not surprising
given the salience of the location in the event. The whole event took place on a
park bench, and so it is likely that people would mention this. Informing liars about
this criterion may thus be more successful in situations where contextual
embeddings are less salient (and therefore less likely to be mentioned otherwise).
Although one might expect that informed liars would easily manage to
incorporate spontaneous corrections in their statements (criterion 14), they failed
to do so. The results revealed that spontaneous corrections were hardly ever
present in the accounts of informed liars. This suggests that making spontaneous
corrections is more diYcult than it would seem. A possible explanation is that
spontaneous corrections have little to do with the content of the recall but more to
do with the way the information is presented. Perhaps, informed liars were more
focused on the actual content of their recall than on the way they presented it.
The fact that informed liars did not admit lack of memory (criterion 15) more
often than uninformed liars (or truth tellers) might be owing to the interview
technique employed in this study. People usually only admit lack of memory when
a direct question has been put to them. By asking people to recall all they could

Informing liars about CBCA
67
remember about the event (as happened in this experiment), the opportunities for
the participants to admit lack of memory may have been restricted.
A comparison between the assessments made by the CBCA-expert and the
computer assessments based upon the total CBCA-scores showed that the
CBCA-expert was particularly accurate at detecting truths, better than could be
expected on the basis of the total CBCA-scores. This supports the experimenters’
suggestion that CBCA-experts do not form their impression merely on the basis of
total CBCA-scores but that some accounts with low CBCA-scores contain
particular characteristics which make CBCA-experts decide that the account was
truthful. The CBCA-expert was less accurate in detecting lies than could be
expected on the basis of the total CBCA-scores. This could be explained in a similar
way: The total CBCA-score is low, and the CBCA-expert should therefore decide
that the account was fabricated. However, some particular characteristics of these
accounts made her (wrongly) decide to judge the account as truthful.
The Ž nding that liars were capable of in uencing CBCA-judgments when they
were informed about CBCA may have implications for the use of CBCA-judgments
in courts. It clearly challenges the view of those who argue that the validity of
CBCA has been conclusively demonstrated (Honts, 1994) and supports those who
are sceptical of the issue (Lamb et al., 1997; Ruby & Brigham, 1997; Wells & Loftus,
1991). The present experimenters are, however, reluctant to draw deŽ nite conclu-
sions on people’s ability to in uence CBCA-judgments on the basis of the Ž ndings
of one single study, and therefore recommend further research.
Such research could investigate which types of liars are particularly good or bad
in incorporating CBCA-criteria into their fabricated accounts. For example, it
seems plausible to suggest that adults will be better in doing so than very young
children, as young children will not yet have developed the necessary cognitive
skills. Conducting studies with children as participants is important as, at present,
CBCA is used in courts exclusively on children. Also personality traits may have an
impact. For example, people high in Machiavellianism (people who manipulate
others to get what they want (Kashy & DePaulo, 1996)) may be better than their
counterparts in including CBCA-criteria into fabricated accounts because they are
particularly skilled and practised in in uencing other people.
Additionally, more CBCA-experts should be asked to assess the veracity of
statements of informed liars in order to determine whether the present Ž ndings
obtained with the use of one CBCA-expert can be generalized to CBCA-experts per
se. The experimenters believe that the CBCA-expert used in this study performed
well (she classiŽ ed the majority (70%) of truth tellers and uninformed liars
correctly), and therefore expect other CBCA-experts would also have diYculties in
correctly classifying statements of informed liars.
The informed liars were not told in advance that they had to repeat their recall
later on, nor were they asked to do this. This may have in uenced the Ž ndings. It
is expected that liars would be inclined to include fewer details in their fabricated
accounts if they expect that they may have to repeat their recall. While repeating
their stories, liars run the risk of forgetting to mention some details or of
contradicting themselves. One way of reducing these risks is to avoid providing
much information in the Ž rst account.

68
Aldert Vrij et al.
With regard to the ecological validity of the present study, probably the most
obvious criticism is that the participants were asked to describe an event they had
watched on a video rather than describing a live event in which they had actually
participated. There were two reasons why a videotaped event was chosen. First, the
experimenters wanted to create a highly controlled and standardized situation. They
believe that it is essential to test innovative ideas (such as informing liars about
CBCA-criteria) in highly controlled situations Ž rst, as the exact impact can only be
determined in such situations. Secondly, recent research has shown that watching
an event on a video or actually taking part in such an event results in similar
CBCA-scores (Akehurst, 1997; Akehurst et al., 1995). This suggests that a method
utilizing a videotaped event has a positive e ect on the standardization of the study
V
without compromising its ecological validity too much. Despite this, it is acknowl-
edged that CBCA-research should also include studies with higher ecological
validity than the present study.
References
Akehurst, L. (1997). Deception and its detection in children and adults via verbal and nonverbal cues. PhD thesis,
Psychology Department, University of Portsmouth.
Akehurst, L., Ko¨hnken, G., & Ho¨fer, E. (1995). The analysis and application of Statement Validity
Assessment. Paper presented at the 5th European Conference on Psychology and Law, Budapest,
Hungary.
Anson, D. A., Golding, S. L., & Gully, K. J. (1993). Child sexual abuse allegations: Reliability of
Criteria-Based Content Analysis. Law and Human Behavior, 17, 331–341.
Ho¨fer, E., Akehurst, L., & Metzger, G. (1996, August). Reality monitoring: A chance for further development
of CBCA? Paper presented at the Annual Meeting of the European Association on Psychology and
Law, Siena, Italy.
Honts, C. R. (1994). Assesing children’s credibility: ScientiŽc and legal issues in 1994. North Dakota
Law Review, 70, 879–903.
Horowitz, S. W., Lamb, M. E., Esplin, P. W., Boychuk, T. D., Krispin, O., & Reiter-Lavery, L. (1997).
Reliability of criteria-based content analysis of child witness statements. Legal and Criminological
Psychology, 2, 11–21.
Kashy, D. A., & DePaulo, B. M. (1996). Who lies? Journal of Personality and Social Psychology, 70,
1037–1051.
Ko¨hnken, G., Schimossek, E., Aschermann, E., & Ho¨fer, E. (1995). The cognitive interview and the
assessment of the credibility of adults’ statements. Journal of Applied Psychology, 80, 671–684.
Lamb, M. E., Sternberg, K. J., Esplin, P. W., Hershkowitz, I., Orbach, Y., & Hovav, M. (1997).
Criterion-based content analysis: A Ž eld validation study. Child Abuse and Neglect, 21, 255–264.
Landry, K., & Brigham, J. C. (1992). The e ect of training in Criteria-Based Content Analysis on the
V
ability to detect deception in adults. Law and Human Behavior, 16, 663–675.
Porter, S., & Yuille, J. C. (1996). The language of deceit: An investigation of the verbal cues to
deception in the interrogation context. Law and Human Behavior, 20, 443–459.
Ruby, C. L., & Brigham, J. C. (1997). The usefulness of the criteria-based content analysis technique
in distinguishing between truthful and fabricated accounts. Psychology, Public Policy and Law, 3,
705–737.
Ruby, C. L., & Brigham, J. C. (1998). Can Criteria-Based Content Analysis distinguish between true
and false statements of African-American speakers? Law and Human Behavior, 22, 369–388.
Sporer, S. L. (1997). The less travelled road to truth: Verbal cues in deception detection in accounts
of fabricated and self-experienced events. Applied Cognitive Psychology, 11, 373–397.
Steller, M., & Ko¨hnken, G. (1989). Criteria-Based Content Analysis. In D. C. Raskin (Ed.), Psychological
methods in criminal investigation and evidence (pp. 217–245). New York: Springer Verlag.

Informing liars about CBCA
69
Steller, M., Wellershaus, P., & Wolf, T. (1988, June). Empirical validation of Criteria-Based Content
Analysis. Paper presented at the NATO Advanced Study Institute on Credibility Assessment,
Maratea, Italy.
Vrij, A. (1998). De psychologie van de leugenaar: Liegen en voorgelogen worden op het werk, in de rechtszaal en thuis.
Lisse: Swets & Zeitlinger.
Vrij, A. (2000). Detecting lies and deceit. Chichester: Wiley.
Vrij, A., & Akehurst, L. (1998). Verbal communication and credibility: Statement Validity Assessment.
In A. Memon, A. Vrij, & R. Bull (Eds.), Psychology and law: Truthfulness, accuracy and credibility
(pp. 3–31). Maidenhead: McGraw-Hill.
Vrij, A., Edward, K., Roberts, K. P., & Bull, R. (1999). Detecting deceit via Criteria-Based Content Analysis,
Reality Monitoring and analyses of nonverbal behaviour. Manuscript submitted for publication.
Wells, G. L., & Loftus, E. F. (1991). Commentary: Is this child fabricating? Reactions to a new
assessment technique. In J. Doris (Ed.), The suggestibility of children’s recollections (pp. 168–171).
Washington DC: American Psychological Association.
Yuille, J. C. (1988). The systematic assessment of children’s testimony. Canadian Psychology, 29,
247–262.
Zaparniuk, J., Yuille, J. C., & Taylor, S. (1995). Assessing the credibility of true and false statements.
International Journal of Law and Psychiatry, 18, 343–352.
Received 21 July 1998; revised version received 4 February 1999
Appendix: Brief description of 19 CBCA criteria
(1) Logical structure. This is present if the statement essentially makes sense, that is, if the statement
is coherent and logical and the di erent segments Ž t together, i.e. di erent segments are not
V
V
inconsistent or discrepant.
(2) Unstructured production. This is present if the information is scattered throughout the statement
instead of mentioned in a structured, coherent and chronological order. The incoherent and
unorganized manner of presentation is, for instance, caused by digressions or spontaneous shifts
of focus.
(3) Quantity of details. This criterion requires that the statement must be rich in detail, that is, speciŽ c
descriptions of place, time, persons, objects and events should be present.
(4) Contextual embedding. This is present if the events are placed in time and location, and when the
actions are connected with other daily activities and/or customs.
(5) Descriptions of interactions. This criterion is fulŽ lled if the statement contains information about
interactions involving at least the accused and witness, and if this information consists of three
parts, i.e. an action of actor A leads to a reaction of actor B which leads to a reaction of actor
A again.
(6) Reproduction of speech. Reproduction of speech is present if speech is, or parts of the conversation
are, reported in orginal form and if the di erent speakers are recognizable in the reproduced
V
dialogues. This criterion is not satisŽ ed by a report about the content of a dialogue; it is only
satisŽ ed when there is a virtual replication of the utterances of a least one person.
(7) Unexpected complications during the incident. This criterion is present if there are elements
incorporated in the event which are somewhat unexpected.
(8) Unusual details. Unusual details refer to details of persons, objects or events which are unusual
and/or unique but meaningful in the context.
(9) Superuous details. These are present if the witness describes details in connection with the
allegations which are not essential for the accusation, such as a child who says that the adult tried
to get rid of the cat which entered the bedroom because he (the adult) is allergic to cats.
(10) Accurately reported details misunderstood. This criterion is fulŽ lled if the witness speaks of details that
lie beyond the horizon of the comprehension of the witness, for instance a child that describes
the adult’s sexual behaviour but attributes it to a sneeze or to pain.

70
Aldert Vrij et al.
(11) Related external associations. A related external association is present when the witness reports
conversation that is not part of the alleged o ences but refers to other similar events, for
V
instance: ‘He talked about the women he had done it with and the di erences between them.’
V
(12) Accounts of subjective mental state. This criterion is present when the witness describes feelings or
thoughts experienced at the time of the incident, as well as reports of cognitions, such as
thinking about how to escape while the event was in progress.
(13) Attribution of perpetrator’s mental state. This criterion is present if the witness describes perceptions
of the perpetrator’s feelings, thoughts or motives during the incident.
(14) Spontaneous corrections. This criterion is fulŽ lled if corrections are spontaneously o ered or
V
information is spontaneously added to material previously provided in the statement
(spontaneous means without any interference by the interviewer).
(15) Admitting lack of memory. This criterion is present if a witness admits lack of memory by saying
either ‘I don’t know’ or ‘I don’t remember’ or by giving a more extensive answer.
(16) Raising doubts about one’s own testimony. This criterion is present if the witness expresses concern
that some part of the statement seems incorrect or unbelievable.
(17) Self-deprecation. Self-deprecation is present if the witness mentions personally unfavourable,
self-incriminating details.
(18) Pardoning the perpetrator. Pardoning the perpetrator is present if the witness tends to favour the
alleged perpetrator in terms of making excuses for or failing to blame the alleged perpetrator.
(19) Details characteristic of the o ence. This criterion is present if a witness describes events in a manner
V
in which professionals know that certain crimes occur. For instance, the progression of events
in an incestuous relationship would be expected to di er from the dynamics surrounding a
V
single-incident stranger assault.