You are on page 1of 9

Law and Human Behavior 2012, Vol. 36, No.

1, 68 76

2011 American Psychological Association 0147-7307/11/$12.00 DOI: 10.1037/h0093965

Does the Truth Come Out in the Writing? SCAN as a Lie Detection Tool
Galit Nahari
Bar-Ilan University

Aldert Vrij
University of Portsmouth

Ronald P. Fisher
Florida International University
We tested the accuracy of Scientific Content Analysis (SCAN), a verbal lie detection tool that is used world-wide by federal law enforcement and military agencies. Sixty-one participants were requested to write down the truth, an outright lie or a concealment lie about activities they had just completed. The statements were coded with SCAN and with another verbal lie detection tool, Reality Monitoring (RM). RM discriminated significantly between truth tellers and outright liars and between truth tellers and concealment liars, whereas SCAN did not discriminate between truth tellers and either kind of liar. Implications of the findings for the suitability of SCAN as a lie detection tool are discussed.

Scientific Content Analysis (SCAN), developed by Avinoam Sapir, a former polygraph examiner in the Israeli police, is a verbal veracity assessment tool. It is used in countries such as Australia, Belgium, Canada, Israel, Mexico, UK, US, the Netherlands, Qatar, Singapore, and South Africa (Vrij, 2008). It is used by federal law enforcement (including the FBI), military agencies (including the US Army Military Intelligence), secret services (including the CIA) and other types of investigators (including social workers, lawyers, fire investigators and the American Society for Industrial Security) (Bockstaele, 2008; www.lsiscan.co.il). http://www .lsiscan.com/id29.htm provides a full list of past participants of SCAN courses. According to (the American version of) the SCAN website (www.lsiscan.com) SCAN courses are mostly given in the US and Canada (on a weekly basis in those countries). In addition, online courses are also available. Our impression is that SCAN is used widely. The second author gave a seminar on lie detection during the Second International Investigative Interviewing Conference, held in July 2006 at the University of Portsmouth (UK). That seminar was attended by about 100 criminal investigators from many countries including Canada, UK, US, and the Netherlands. When asked which lie detection tool, if any, they mainly use, the most frequent answer, from investigators on both sides of the Atlantic, was SCAN. It was reported to be used even more frequently than the well known Reid technique in lie detection. In the SCAN procedure, the examinee is requested to write a detailed description of all his/her activities during a critical period of time in such a way that a reader without background information can determine what actually happened. The handwritten statement is then analyzed by a SCAN expert on the basis of a list of predetermined

This article was published Online First January 21, 2011. Galit Nahari, Department of Criminology, Bar-Ilan University, Ramat Gan, Israel; Aldert Vrij, University of Portsmouth, Portsmouth, UK; Ronald P. Fisher, Florida International University, Miami, FL, USA. Correspondence concerning this article should be addressed to Galit Nahari, Department of Criminology, Bar-Ilan University, Ramat Gan 52900, Israel, E-mail: naharig@mail.biu.ac.il 68

criteria. It is thought that some SCAN criteria are more likely to occur in truthful statements than in deceptive statements, whereas other criteria are more likely to occur in deceptive statements than in truthful statements (Sapir, 1987/2000). We know of four published studies that examined the validity of SCAN as a lie detection tool (Bachenko, Fitzpatrick, & Schonwetter, 2008; Driscoll, 1994; Porter & Yuille, 1996; Smith, 2001). Both Driscoll (1994) and Smith (2001) carried out field studies and analyzed statements of actual suspects. They reported high accuracy rates, around 80%. However, Smith, who compared the performances of experienced SCAN users with raters not trained in SCAN, found no difference in accuracy between the two groups. Smith also noticed that different experts used different SCAN criteria to justify their decision of whether or not a statement was deceptive. In other words, there was a lack of consistency in the application of SCAN amongst SCAN users. Both studies, however, were limited in that the ground truth, the actual veracity of the statements, could not be established. This makes interpreting the findings problematic. Porter and Yuille (1996) carried out a laboratory SCAN experiment in which participants did or did not commit a mock theft and were interviewed about this alleged crime. They examined three SCAN criteria: Structure of the statement; missing information; and first person singular, past tense. Truthful and deceptive statements did not differ from each other regarding these three criteria. However, this experiment was also limited. First, it tested only a small number (three) of SCAN criteria. Second, as participants gave oral statements that were subsequently transcribed, the procedure did not follow one of the basic principles of SCAN, the request to write down a statement. In a fourth study, some SCAN criteria including verb tense were examined (Bachenko et al., 2008). However, the results for the individual criteria were not discussed. In sum, all four SCAN studies published to date were limited and, therefore, not much is known about the accuracy of the SCAN tool, despite the fact that it is used by many agencies. In the present experiment, we examined whether SCAN could discriminate between true and false statements. We compared SCAN with Reality Monitoring (RM), another verbal veracity assessment tool. Al-

69
though not used by professionals in the field, RMs validity has been widely tested and it appears to discriminate truth tellers and liars well above chance level (Masip, Sporer, Garrido, & Herrero, 2005; Vrij, 2008). The underlying assumption of SCAN is that true statements, derived from memory of an actual experienced event, differ in content and quality from statements that are based on invention or fantasy (Smith, 2001). This is similar to the assumption underlying RM. Unlike SCAN, RM provides a theoretical underpinning for why specific differences between truth telling and lying will occur. RM originally refers to the assessment of ones own memory and describes the processes people use to decide for themselves whether a given memory for an event was based on experience or imagination. According to the reality monitoring theory (Johnson, Foley, Suengas, & Raye, 1988; Johnson & Raye, 1981), memories of experienced events are externally derived and originate in perceptual experiences. They are characterized by spatial and temporal contextual attributes (e.g., where and when the event took place), sensory attributes (e.g., what the person felt, smelt, heard or saw when the event took place), and affective attributes (e.g., expressions of emotions). In contrast, memories for imagined events are internally derived and originate in self-generated thought or imagination. They are characterized by cognitive operations, such as thoughts and reasons, which probably helped to generate these memories. Empirical evidence shows that people also use these attributes to judge whether other peoples memories are externally or internally derived (Johnson, 2006; Johnson, Bush, & Mitchell, 1998; Johnson & Suengas, 1989). Based on these findings, it has been argued that RM can be applied to credibility judgments (Sporer, 1997, 2004; Vrij, 2008). A statement is considered true when it is based on externally derived memories and false when it is based on internally derived memories (Sporer, 1997, 2004). Although SCAN and RM share the same underlying assumption, that truthful statements differ in quality from deceptive statements, each tool takes entirely different sets of criteria into account to assess the veracity of statements. This deviation may reflect a difference in the theoretical underpinnings of the two tools. RM focuses on the clarity, realism, and reconstructability of the statements and examines perceptual, spatial, temporal and affective information, and cognitive operations (Sporer, 1997, 2004). The SCAN list of criteria is more extensive and includes a variety of cues such as denial of allegations, out of sequence information and change in language (see, Vrij, 2008). See Table 1 for definitions of the 13 SCAN and eight RM criteria. A careful examination of both lists of criteria gives the impression that SCAN and RM are oriented to different types of lie: concealments (SCAN) and outright lies (RM), respectively. To demonstrate the distinction between concealment and outright lies (DePaulo, Kashy, Kirkendol, Wyer, & Epstein, 1996) consider the following hypothetical example. Dan, a teenager, wandered around the city with his friends on Friday night. During that night, among other activities, he committed an act of vandalism against a school building. The police have reasons to believe that Dan committed this act of vandalism and invites him to the police station. During the interview, Dan could use two alternative strategies. He could tell the truth about being with his friends, and even could describe truthfully all their activities during that night, but omitting the act of vandalism. Alternatively, he could deny that he went out on that Friday night, and could make up a story pretending that he did something else. If Dan chooses the first strategy he provides a concealment lie, and if he chooses the second strategy, he provides an outright lie. RMs primary concern is assessing the richness in detail, and the vividness of the statement. Outright lies are imaginations and should therefore be less detailed and vivid than truthful accounts. Concealed liars, in contrast, tell the truth about what happened but leave out details that may incriminate them. Therefore, concealment lies include descriptions of activities, conversations, and events that the examinee actually has experienced. RM thus may have difficulty in discriminating between truths and concealed lies. We therefore predicted that RM should be able to discriminate between truths and outright lies, (as was previously evidenced; Masip et al., 2005; Vrij, 2008), but should not be able to discriminate between truths and concealed lies (Hypothesis 1). The SCAN literature does not mention an underlying theme for the list of criteria. Examining the criteria listed in Table 1 gives the impression to be a motivational approach in which liars attempt strategically to select the exact words that reflect their knowledge but hide their guilt. For example, at least four (out of 13) criteria explicitly deal with what the examinee did not say or concealed. Within the missing information criterion, SCAN experts look for words or phrases such as finally, later on, some time after, which can imply that some information is left out. If someone says I started to cook, and some time after I found my father lying on the dining room floor, one could wonder what happened after the examinee started to cook and before he found his father on the floor. Another SCAN criterion is objective and subjective time. Here, the proportions of the actual durations of the activities (objective time) are compared with the number of words that the examinee used to describe these activities (subjective time). When the subjective time that was allocated to an activity is shorter than the objective time, it may imply that the examinee is attempting to conceal information regarding that activity. Concealments can be further reflected by the absence of social introductions in the statement. This criterion relates to how the people who were involved in the activity are introduced. If an examinee writes we and it is not clear who we are, the examinee might be concealing something regarding the identity of the other person. Finally, extraneous information could indicate concealment. This criterion relates to whether the examinee reports information that does not seem relevant. If so, it could indicate that the examinee is trying to hide important information. One could thus expect SCAN to be able to distinguish truths from both outright and concealment lies (Hypothesis 2). In the experiment we also asked participants to report whether they had developed a strategy to appear convincing and, if so, to describe their strategy. Research into the strategies suspects use is emerging only recently, after DePaulo et al.s (2003) plea to carry out such research (Colwell, HiscockAnisman, Memon, Woods, & Michlik, 2006; Hartwig, Granhag, & Strmwall, 2007; Hartwig, Granhag, Strmwall, & Doering, 2010; Strmwall, Hartwig, & Granhag, 2006; Vrij et al., 2009). Yet, as Vrij, Granhag and Porter (in press) argued, for successful lie detection it is important to take differences in strategies between liars and truth tellers into account and to develop interview protocols that exploit these differences.

70
Table 1 Definitions of the SCAN and RM Criteria
Criterion SCAN Spontaneous corrections First person singular past tense Location of emotions Change in language Definition The presence of corrections in the statement, such as crossing out what has been written. Although explanations and additions are allowed, examinees are explicitly instructed not to cross anything out. A failure to follow this instruction could indicate deceit. The format in which a statement is written. It is thought that truthful statements are written in the first person singular, past tense because the writer describes an event that has taken place (I saw the smoke coming out of the window). Deviations from this norm should raise suspicion. The position of the emotions in the statement. Just before the climax of the story (liars), or throughout the story (truth tellers), and after the climax of the story (truth tellers). Change of terminology or vocabulary in the statement. A change in language indicates that something has altered in the mind of the writer. For example, if an examinee refers in his statement to all conversations he had as conversations except one conversation which he describes as a discussion, it is likely that he perceived this conversation differently from the other conversations. The examinee is vague about certain elements in the statement (I believe . . ., I think . . ., kind of . . .) or reports that s/he cannot remember something. Lack of memory indicates deceit. Does the statement recount the events in chronological order. A deviation of the chronological order indicates deceit. The inclusion of words that indicate that some information has been left out, such as sometime after, finally, later on, and shortly thereafter. Inclusion of those words indicates deceit. Does the statement contain information that does not seem relevant. The presence may indicate deceit because examinees could include extraneous information to hide more important information. Objective time refers to the actual duration of events described in the statement, whereas subjective time refers to the amount of words spent to describe these events. It is thought that in a truthful statement the objective and subjective time will correspond with each other, unlike in a deceptive statement. For example, if someone devotes five lines in a statement to describe a 30-minute period, and then three lines to describe a subsequent two- hour period, the objective and subjective time do not correspond and this may indicate deceit. How are the persons described in the statement introduced. They can be unambiguous (e.g., My wife Lisa . . .) or ambiguous (e.g., We went outside) without saying who we are. Ambiguous introductions indicate a lie. The use of pronouns (I, my, he, his, they, their etc.) in the statement. Pronouns signal commitment, responsibility and possession. Omitting pronouns (Left the house rather than I left the house) suggests reluctance of the writer to commit him/herself to the described action and thus indicates deceit. The balance of the statement. It is thought that in a truthful statement the first 20% is used to describe activities leading up to the event, the next 50% to describe the actual event, and the final 30% to discuss what happened after the event. Thus, a ten-line statement is thought to comprise two lines to introduce the event, five lines to describe the event, and three lines about the aftermath. The more unbalanced a statement, the greater the probability that the statement is deceptive. Does the examinee directly deny the allegation in the statement by stating I did not . . . . Denials indicate truth. The statement is clear and vivid. The story is realistic and makes sense. The event can be reconstructed with the information given. Are there emotions described in the statement. Information about what the examinee felt, smelt, heard or saw when the event took place. Information about locations (It happened in a park) or the spatial arrangement of people and/or objects (The man was sitting to the left of his wife). Information about when the event happened (It was early in the morning) or explicitly describes a sequence of events (When the visitor heard all that noise, he became nervous and left). Thoughts and reasonings (e.g., I must have locked the door, because it was locked when I came home).

Lack of conviction and memory Out of sequence Missing information Extraneous information Objective and subjective time

Social introductions Pronouns Structure of statement

Denial of allegation RM Clarity Realism Reconstructability of the story Emotions Perceptual information Spatial information Temporal information Cognitive operations

Method Participants
A total of 61 Bar-Ilan undergraduate students (54 females and 7 males) participated in the experiment for course credits. Their mean age was 23.20 (SD 2.17 years).

Procedure
Participants arrived individually at the laboratory at a predetermined time. The experimenter told them that the experiment dealt with examining the efficiency of certain tools to detect lies. All

participants signed a consent form indicating that participation was voluntary and that they could withdraw from the experiment at any time without penalty. Participants were then assigned randomly to one of three experimental conditions: (i) concealment liars (N 20): mock criminals who concealed their criminal activity but truly reported their non-criminal activities, (ii) outright liars (N 20): mock criminals who concealed both their criminal and noncriminal activities and thus gave a completely false statement; and (iii) innocents (N 21): truth tellers who gave a completely truthful account of their activities. The experimenter read aloud the instructions. The instructions for guilty participants (concealment and outright liars) were as

71
follows: You are requested to commit a mock theft as part of an experiment about deception. The mock theft is about stealing a statistics exam from a mail box in the Criminology department. First, please go to the Social Science library (First floor of this building), look for the Prison Journal volume No. 77 (1997). Get the article that appears on pp. 125134 and write down the article authors names. You will be asked about it later. After the library assignment, you should come back to the department, and steal the exam. The exam is in mail box No. 50 (Prof Aharonovitz mail box). The mailbox is locked, and you can find the key under the mail box of the secretariat. At the bottom of the secretariat mailbox, you will find the key, which is attached with cellophane tape. Take the key, open Prof. Aharonovitzs mailbox, find the exam and hide it on you. Make sure that you leave the mailbox in the same condition as you found it, so that Prof. Aharonovitz will not notice that someone broke into his mailbox. Finally, re-attach the key with the cellophane tape to the bottom of the secretariat mailbox. When you take the key from the bottom of the secretariat mailbox, please make sure that you keep the cellophane tape in good condition so that you will be able to use it again. Please make sure that nobody witnesses you committing this crime. Also, do not tell anybody that you took the exam. You should be back here 30 min from now. In the remaining time after you steal the exam, do your normal business. Make sure that you carry out more than one activity and do not stay at one place for more than 10 min. Please carry out as many activities as you can, such as buying a coffee, having a conversation with a friend, visiting the book shop, etc. Please do not do anything that you may later be unable or uncomfortable to talk about. Please return in exactly 30 min from now (the experimenter indicated the exact time) to the lab for the second stage of the study. The instructions for the innocent participants (truth tellers) were the same, excluding the theft of the exam. The assistant repeated the instructions to make sure that everything was clear, and asked the participants whether they had any questions. When the participants indicated that they had understood the instructions, they were sent on their mission. Upon returning to the lab after exactly 30 min, the experimenter told the participant: While you were out a statistics exam had been stolen from a mailbox in the Criminology department. You are one of the suspects in this theft. In a short while, you will be interviewed about it. The interviewer will ask you to write down a statement about what you were doing during the last 30 min. She does not know whether you are guilty. In case you stole the exam, do not admit this under any circumstances. At this point the experimenter suggested a different strategy to the participants in the guilty conditions (concealment liars and outright liars). To the concealment liars she said: The best strategy will be to give a truthful statement about everything you did during the last 30 min, except stealing the exam. Therefore, report only your non-criminal activities and skip the act of stealing. To the outright liars she said: The best strategy will be to give a complete falsely statement about what you did during the last 30 min. Therefore, avoid reporting any of your activities, both criminal and non-criminal, and give a complete falsely report by making-up acts that you did not do in this period of time. No strategy was suggested to the innocent participants. The experimenter continued by saying to all participants: While giving your statement, act as convincing as possible. The interviewer will use your statement to determine whether you are guilty or innocent. If the interviewer believes that you are innocent, you will receive your course credit and be allowed to leave. If the interviewer believes that you are guilty, you will be asked to write a second statement about what actually occurred during the last 30 min. You will be interviewed in 10 min, and the statement you will be asked to write down is your only chance to convince the interviewer that you are innocent. Please use these 10 min to prepare the statement that you are going to write. Finally, the experimenter reminded the participants that they should not confess to the crime, and she repeated the suggested strategies to the concealment and outright liars. The participants were then left alone in the lab, giving them the opportunity to prepare for the interview and plan a strategy. After 10 min they were asked to complete a pre-interview questionnaire. The questionnaire included three main questions: (i) How would you describe your level of preparation for writing the statement? The participants rated their answer to this question on three 7-point scales: (1) Shallow to (7) Thorough; (1) Insufficient to (7) Sufficient; (1) Poor to (7) Good. The average score on these three scales was used as the Quality of Preparation variable ( .85); (ii) Have you developed a strategy to appear convincing? Participants who answered this question negatively were asked to explain why not, and participants who answered it positively were asked to describe their strategy; (iii) To what extent are you motivated to appear convincing in your statement? Participants rated their answer to this question on a 7-point scale (1not at all to 7 completely). After completing the questionnaire, participants were brought to the interview room. The interviewer (blind to the participants condition) sat opposite the participant and gave an instruction that follows the standard instructions given by SCAN users in the field: An exam has been stolen from a mail box in the Criminology department within the last hour. You are one of the suspects in this theft. Please write down in as much detail as possible what you were doing during the 30 min that you were out. Start your statement from the time you left the lab and end it at the time you re-entered the lab. Be sure that you mention all details, all activities, all people you have met, all conversations that took place, etc. Give as much information as you can, including information that seems irrelevant. I will only be asking this single question. You will have this one opportunity to give me as much information as you can. While writing your statement, you may explain or add details as much as you want, but do not cross out anything that you have already written. After writing their statements, participants were asked to complete a post-interview questionnaire. They were asked to rate on 7-point scales (1not at all to 7 completely): (i) To what extent they were motivated to appear convincing in their statement; (ii) To what extent they found it difficult to appear convincing in their statement; (iii) To what extent they thought they were convincing in their statement; and (iv) To what extent they were concerned about being asked to write an additional statement about what actually had occurred. In addition, the participants reported whether they (v) gave any truthful details in their statement; (vi) any false details in their statement, and (vii) concealed any details from their statement. Finally, they were asked to indicate what

72
percentage of their statement they thought was true. Before filling out the post-interview questionnaire, participants were told that the experiment had ended, that their answers would not influence the assessment of their statements as false or true, and that they were requested to answer the questions truthfully. Upon completing the questionnaire, all participants were debriefed and given their course credits. None of the participants was requested to write a second statement. The coders received a written list of the criteria with detailed definitions and examples. The trainer (one of the authors) went over each criterion, and explained the definitions and examples. She answered all questions about the criterion and made sure that all coders understood the criterion before she moved on to the next one. This training took 90 min. The coders carried out their assessments individually at their homes. They took with them the written explanations and were asked to read them again before and during the coding. Each coder assessed 20 21 statements, and each statement was assessed by two coders.

Coding of the Statements with SCAN and RM


Six independent coders assessed the presence of SCAN and RM criteria. The coders were not informed that these criteria were used to determine true from false statements. Instead, they were told that participants were requested to carry out some activities outside the lab for 30 min and to provide a detailed statement about these activities. The coders received a list of 13 SCAN and 8 RM criteria which were mixed together. The SCAN criteria were taken from Vrij (2008) and the RM criteria from Sporer (2004). In his list of SCAN criteria, Vrij (2008) included the criteria that are most emphasized in workshops on the technique (Driscoll, 1994) or are used in research (Porter & Yuille, 1996; Smith, 2001). The coders were requested to rate the presence of seven criteria on 10-point scales (1not at all to 10 very much) (e.g., to what extent is the statement realistic?). They were asked to mark the frequency of occurrence of 13 other criteria in the statement (e.g., the frequency of occurrence of mentioning emotions in the statement). The remaining criterion refers to where emotions appeared in the statement (a SCAN criterion). The coders indicated whether or not (0 no 1yes) emotions appeared (i) at the beginning of the statement; (ii) throughout the statement; or (iii) at the end of the statement. This SCAN criterion thus resulted in three dependent variables. Table 2 Inter-Rater Agreement for the SCAN and RM Criteria
Criterion Clarity Realism Reconstructability of the story Emotions Perceptual information Spatial information Temporal information Cognitive operations Spontaneous corrections First person singular past tense Location of emotions Beginning Throughout End Chang in language Lack of conviction and memory Out of sequence Missing information Extraneous information Objective and subjective time Social introductions Pronouns Structure of statement Denial of allegation Method RM RM RM RM RM RM RM RM SCAN SCAN SCAN

Inter-Rater Reliability
Variance was low for several criteria. For example, emotion was rated as absent in 84% of the statements. Consequently, Pearson correlations were not appropriate for assessing the inter-rater reliability. We think this is not problematic as the important issue is whether or not the coders agreed that the criterion occurred rather than whether or not they agreed on its frequency of occurrence. Thus, inter-rater reliability was calculated by determining for each criterion the percentage of agreement about its presence in the statement. We dichotomized the original data set: present (1) or not present (0). Criteria that were originally measured on 10-point scales were recorded as present when the original rating fell in the 6-to-10 range, and as not present when the original rating fell in the 1-to-5 range. Criteria that were originally measured by counting the number of appearances in the statement were recorded as present when they appeared at least once. Table 2 presents the agreement percent between coders. High agreement was achieved for the RM criteria (ranging between 71% and 98%), except cognitive operations which achieved only 44% of agreement. High agreement was achieved also for SCAN criteria (ranging from 70% to 98%), although low agreement occurred for four

Cue for Truth Truth Truth Truth Truth Truth Truth Lie Lie Truth Lie Truth Truth Lie Lie Lie Lie Lie Truth Lie Lie Truth Truth

Coding method Rating Rating Rating Counting Counting Counting Counting Counting Counting Rating

% of agreement 87 98 93 84 74 80 71 44 66 89 97 87 90 72 74 98 61 84 70 56 64 70 80

SCAN SCAN SCAN SCAN SCAN SCAN SCAN SCAN SCAN SCAN

Counting Counting Counting Counting Counting Rating Counting Counting Rating Counting

73
criteria: Spontaneous corrections (66%), missing information (61%), pronouns (64%), and social introduction (56%). Since agreement between the coders was not perfect, all disagreements between the two coders were relegated to a third coder who made the final decision regarding the presence of the criterion. The third coder was blind to the veracity status of the statements. We calculated total scores for RM and SCAN, by using the dichotomized variables. A score of 0 was assigned when the criterion was absent, and a score of 1 was assigned when the criterion was present. Criteria that indicate lies (e.g., cognitive operations, spontaneous corrections, missing information) were re-coded; thus a score of 1 was assigned when the criterion was absent, and a score of 0 was assigned when the criterion was present. The total score for each method (RM and SCAN) was the sum of all criteria that were included in the method. Therefore, the total score for RM could range from 0 to 8, and the total score for SCAN from 0 to 15 (As reported earlier, location of emotions was divided into three sub-variables, thus the SCAN total score could attain 15 points). According to the SCAN and RM assumptions, the higher the score, the more likely it is that the statement is truthful.1 three experimental conditions regarding motivation or concerns about writing a second statement, all Fs (2, 58) 2.44.

Pre-Interviewing Preparation
The self-reported quality of preparation for the interview was M 4.69 (SD 1.49). There were no differences among the three conditions, F 1. Most of the participants planned a strategy (63.9%, N 39; 13 innocents, 12 concealment liars and 14 outright liars). Innocents reported that they planned to provide details that are easy to verify and to mention a potential witness for their alibi (N 9), to be as detailed as they could (N 7), or to convince the interviewer that they have no criminal record and that they are honest people (N 2). Concealment liars planned to be very detailed in reporting their non-criminal activities (N 8), and tried to inflate their non-criminal activities to compensate for leaving out their criminal activities (N 4). Outright liars thought that they should mention only activities that are difficult to verify (N 5), thought of evidence they could provide for activities they did not do (N 3), planned to report activities they carried out another day (N 2), thought about what they reasonably could have done in 30 min (N 2), and tried to convince themselves about their own lies, so that they would appear like innocents (N 2). Participants who did not develop a strategy said they did not need a strategy because they were innocent (N 8); because they believe it would be easy to convince the interviewer since s/he can (in case of innocents) or cannot (in case of liars) verify their statement (N 4); or because planning may interrupt them from being natural and from altering their statements if necessary (N 4).

Results Manipulation Check


Significant differences emerged between experimental conditions regarding the percentages of truthful details participants said they reported in the statements, F(2, 58) 99.37, MSE 387.27, 2 .77, p .001. Scheffe tests revealed that innocents reported more truthful details (M 99.52% of the statement, SD 1.50, 95% CI [98.84, 100.21]) than concealment liars (M 85.90% of the statement, SD 18.56, 95% CI [77.21, 94.95]), who reported more truthful details than outright liars (M 18.15% of the statement, SD 28.90, 95% CI [4.62, 31.68]). This indicates that the participants followed the experimenters instructions. All innocent participants mentioned that they provided truthful details and none of them mentioned that they provided false details in their statements. Further, 90.5% (N 19) of the innocents said that they did not conceal any information. All outright liars mentioned that they provided false details and concealed information. Sixty percent (N 12) of the outright liars said that they did not provide any true details, and most outright liars who said that they provided true details did this only occasionally (mentioning that between 1% and 10% of their statement was true). All concealment liars said that they provided true details. Thirty five percent (N 7) of them also mentioned that they provided false details.

Number of Words
The length of the statements differed between the three experimental conditions, F(2, 58) 9.87, MSE 4731.23, 2 .25, p .001). Scheffe tests revealed that statements of innocents were significantly longer (M 218.43 words, SD 91.35, 95% CI [176.84, 260.01]) than statements of concealment liars (M 144.20 words, SD 52.77, 95% CI [119.50, 168.90]) and outright liars (M 129.80 words, SD 53.60, 95% CI [104.71, 154.89]). The length of statements of concealment and outright liars did not differ significantly from each other.

Discriminating Between Groups by RM and SCAN


RM total score. In order to test the validity of RM criteria in detecting outright and concealment lies, we first conducted an
After completing their coding, the coders were asked to rate on 10-point scales how difficult or easy they found it to assess each of the criteria (1 very difficult to assess and 10 very easy to assess). The coders found most of the criteria relatively easy to apply (about 57% (N 12) of the criteria were rated as 9 or 10). The criteria they found the most difficult to assess were the SCAN criteria: social introductions, pronouns, and structure of the statement (Ms are 6.83, 7.50 and 7.83, respectively). The correlation between the difficulty in applying the criterion (Table 3) and the inter-rater percentage agreement regarding that criterion (Table 2) was (Spearman) r .50, p .05). That is, the more the coders perceived the criteria as easy to apply, the greater was the agreement between coders.
1

Participants Motivation
The motivation of the participants to be convincing was measured both before and after providing the statement. The reported motivation was high both prior to interviewing (M 6.20, SD .99) and after interviewing (M 6.26, SD 1.05). These two motivation scores were highly correlated with each other (r .92, p .001). Further, twenty eight percent (N 17) of the participants were concerned that they might be asked to write a second statement (ratings of 6 or 7). There was no difference between the

74
ANOVA with RM total score as the dependent variable, and group (innocents, outright or concealment liars) as a factor. This analysis revealed a large and statistically significant Group effect, F(2, 58) 4.33, MSE .51, 2 .13, p .05. Scheffe tests revealed that, as predicted in Hypothesis 1, innocents (M 6.71, SD .56, 95% CI [6.46, 6.97]) obtained higher RM total scores than outright liars (M 6.15, SD .75, 95% CI [5.80, 6.50]), representing a large effect size, d .92. However, contrary to our prediction, innocents also obtained higher RM total scores than concealment liars (M 6.15, SD .81, d .82, 95% CI [5.77, 6.53]). We further tested the validity of RM criteria in discriminating between groups. We ran three different discriminant analyses to distinguish between (i) outright liars and innocents; (ii) concealment liars and innocents; and (iii) outright and concealment liars. The objective group-belonging was the classifying variable and the RM total score was the predictor. As Table 3 shows, a significant discriminant function was found for distinguishing between outright liars and innocents, 2(1) 6.82, Wilks Lambda .84, p .01. The function correctly identified 75.0% of the outright liars and 66.7% of the innocents, resulting in the total accuracy of 70.7%. Discriminating concealment liars from innocents was also significant, 2(1) 6.15, Wilks Lambda .85, p .05, but achieved a lower total accuracy rate (63.4%). Rate of correctly classifying innocents was identical as in the previous function (66.7%), but the classification of the concealment liars was lower, 60.0%. Determining outright liars from concealment liars was not possible, 2(1) 0.00, ns. SCAN total score. Similar to RM, we conducted an ANOVA with Group (innocents, outright or concealment liars) as a factor and SCAN total score as a dependent variable to test the validity of SCAN total score in detecting outright and concealment lies. The effect was not significant, F 1, and the SCAN total scores of all three groups were highly similar (innocents: M 9.57, SD 1.29, 95% CI [8.99, 10.16]; outright liars: M 9.55, SD 1.36, 95% CI [8.92, 10.18]; concealment liars: M 9.10, SD 1.65, 95% CI [8.55, 9.65]; all ds for differences between means ranged from .01 to .32). Hypothesis 2 is therefore rejected. We further ran three discriminant analyses for SCAN, similar to what we had done with RM. None of the three discriminant functions were significant (2 ranged from .03 to 1.46). Scan users do not use total SCAN scores. Instead, they examine the individual criteria. We carried out two MANOVAs to accommodate this procedure. In the first MANOVA we compared truth tellers with outright liars with the 15 SCAN criteria as dependent variables. At a multivariate level the analysis was not significant (F 1). At a univariate level only one of the fifteen variables showed a significant effect (F[1, 39] 4.99, p .05). In the second MANOVA we compared truth tellers with concealment liars, and again, the 15 SCAN criteria were the dependent variables. The multivariate analysis was not significant (F 1), neither were any of the univariate effects (all Fs 1.72, all ps .20). RM SCAN. The correlation between RM total score andSCAN total score (r .08) was not significant, supporting our observation that they measure different aspects in written statements. Although SCAN by itself could not discriminate between the three experimental groups, we tested whether it contributed to the accuracy rate when added to RM. For this purpose, we ran three more discriminant analyses, where the objective groupbelonging was the classifying variable and RM and SCAN total scores were the two predictors. We used thestepwisemethod. In discriminatinginnocents fromoutright liars or concealment liars only RM total score emerged in the function.In other words, the discriminant functions obtained in these analyses were identical to the discriminant functions described in the RM section, and SCAN did not contribute anything beyond the RM score.

Discussion
In the present experiment, we failed to find support for SCAN, as the total SCAN score did not differ between truth tellers and outright liars or between truth tellers and concealment liars. However, SCAN users do not examine total SCAN scores; they examine the individual SCAN criteria instead. Examining individual criteria is problematic as it easily leads to a lack of standardization with different SCAN users examining and relying upon different criteria. This is exactly what Smith (2001) found in her field study. Even when we examined the individual criteria, however, SCAN did not receive empirical support. Only one SCAN criterion distinguished between truth tellers and outright liars and none of the criteria distinguished between truth tellers and concealment liars. Only one out of thirty univariate tests (3.33%) yielded a significant result, a percentage that could be expected by chance alone. SCAN users may argue that our experimental design was inadequate for SCAN differences to occur between truth tellers and liars. This would mean that SCAN cannot be used in all situations.

Table 3 Accuracy Rates From the SCAN and RM Tools in Different Types of Lies
Innocents versus outright Innocents versus concealment Outright versus concealment

Innocents (%) Outright (%) Total (%) Innocents (%) Concealment (%) Total (%) Outright (%) Concealment (%) Total (%) RM Hit rate SCAN Hit rate RM and SCAN Hit rate 66.7 52.4 66.7 75.0 65.0 75.0 70.7 58.5 70.7 66.7 52.4 66.7 60.0 45.0 60.0 63.4 48.8 63.4 65.0 55.0 60.0

Note. Discriminant analysis for RM and RM & SCAN among outright versus concealment could not be obtained (2(1) 0.00). Accuracy rates for significant discriminant functions appear in bold.

75
SCAN users should then make clear under which circumstances SCAN can be used and under which circumstances it cannot be used. At present, SCAN users do not discuss this issue. Alternatively, it could be that we did not conduct the SCAN procedure properly. Such a criticism is unlikely, as the SCAN procedure is very straightforward. It just invites the examinee to write down his/her activities during a critical period of time in enough detail so that a reader without background information about the examinees activities could fully grasp the statement (Adams, 1996; Sapir, 1987/2000). Finally, someone could argue that the null findings were the result of a lack of statistical power. However, one should note that the largest effect size in the SCAN analyses (between innocents and concealment liars) was still small, d .32. Admittedly, drawing conclusions from null effect is problematic and from a statistical standpoint we cannot conclude that SCAN is not effective. However, we are still left with the conclusion that SCAN has no empirical support to date. The poor performance of SCAN is further highlighted by the results for RM, the alternative veracity tool we examined. RM did discriminate between truth tellers and outright liars as well as between truth tellers and concealment liars. The findings comparing truth tellers and outright liars support the growing body of literature showing that RM can be used for lie detection purposes (Masip et al., 2005; Vrij, 2008). RM being able to discriminate between truth tellers and concealment liars is worth discussing, as it was the first time that such a comparison was made in the RM deception literature. We were skeptical that RM would effectively discriminate between truth tellers and concealment liars. In their statements, concealment liars do not include much information that is deceptive, as their lie mainly consists of hiding vital information. We argued that RM would have difficulty in spotting concealments as the RM method examines what the examinee has written (mostly truthful) rather than what the examinee has concealed (a lie). The findings concerning the length of the statement (e.g., number of words) and the strategies the examinees said they would use in their concealment lies enables us to explain this finding. In order to be successful in concealing information, the examinee needs to compensate the hidden information by elaborating more thoroughly when describing the information they present. The concealment liars did not do this. The statements of concealment liars were significantly shorter than the statements of truth tellers. In addition, only 4 out of 20 concealment liars reported as a strategy that they would try to inflate their noncriminal activities to compensate for leaving out their criminal activities. As a result, the truthful information that concealment liars reported was less detailed than the information truth tellers reported, and this enabled RM to discriminate between the two types of statement. SCAN skeptics may argue that we were perhaps too optimistic in expecting differences between truth tellers and liars, given the lack of an underlying theoretical rationale for SCAN. Linguists, in particular, have criticized SCAN for that reason. For example, Shuy (1993) is concerned that SCAN does not take into account socio-cultural variation in language use, and Heydon (2008) believes that SCAN strengthens the mythology of interviewing by promoting certain views and beliefs as commonsensical. SCAN skeptics may even go one step further: Why bother examining a tool that has no theoretical foundation? We disagree with such skeptics. First, although SCAN users do not give a theoretical foundation for their tool, they may do so in the future. In that respect, SCAN could be similar to Criteria-Based Content Analysis (CBCA), an alternative verbal lie detection tool. CBCA was presented (in English) in 1988 by Khnken and Steller, but the first publication in English about its theoretical underpinning appeared eight years later (Khnken, 1996). Second, we think that it is important to examine SCAN since it seems to be used in practice on a wide scale. The self-reported strategies of truth tellers and liars revealed more noteworthy findings. First, 8 out of 21 truth tellers did not develop a strategy because, as they reported, they were innocent. This supports the view that truth tellers are less likely than liars to take their credibility for granted (DePaulo et al., 2003; Kassin, Appleby, & Torkildson-Perillo, 2010; Vrij et al., in press). Truth tellers typically assume that their innocence shines through (Granhag, Strmwall & Hartwig, 2007; Kassin et al., 2010; Vrij, Mann, & Fisher, 2006), which could be explained with the illusion of transparency (Gilovich, Savitsky, & Medvec, 1998), the belief that ones inner feelings will manifest themselves on the outside, and belief in a just world (Lerner, 1980), the belief that people will get what they deserve, and deserve what they get. Second, the most frequently cited strategy for truth tellers (9 out of 21) was to provide details that can be checked, whereas the most frequently cited strategy for outright liars (5 out of 20) was to provide details that cannot be checked. Truth tellers strategy makes sense: The best way for them to convince the interviewer that they are telling the truth is to provide evidence that can be corroborated. Liars, of course, cannot do this and need to avoid providing evidence that can be corroborated. This differential strategy of truth tellers and liars may result in a verbal cue of deceit that has not been examined in the verbal lie detection literature so far: the use of corroborated cues. That is, to what extent do examinees include information in their statements that, in theory, can be checked? The hypothesis is that truth tellers include more corroborated cues in their statements than liars. This hypothesis is worth examining. In sum, no support for the use of SCAN was found in the experiment, whereas an alternative verbal lie detection tool, RM, was successful in discriminating between truth tellers and liars. Those findings do not justify the use of SCAN for lie detection purposes in real life practice. The results further revealed a narrative strategy that is worth examining: The use of corroborated cues.

References
Adams, S. H. (1996). Statement analysis: What do suspects words really reveal? FBI Law Enforcement Bulletin, October, 1220. Retrieved from http://www.fbi.gov/stats-services/publications/law-enforcementbulletin/leb Bachenko, J., Fitzpatrick, E., & Schonwetter. M. (2008). Verification and Implementation of Language-Based Deception Indicators in Civil and Criminal Narratives. Proceedings of the 22nd international conference on computational linguistics (Coling 2008), pp. 41 48. Bockstaele, M. (2008). Scientific Content Analysis (SCAN). Een nutting instrument by verhoren? In L. Smets & A. Vrij (Eds.), Het analyseren van de geloofwaardigheid van verhoren (pp. 105156). Brussels: Politeia. Colwell, K., Hiscock-Anisman, C., Memon, A., Woods, D., & Michlik, P. M. (2006). Strategies of impression management among deceivers and truth tellers: How liars attempt to convince. American Journal of Forensic Psychology, 24, 3138.

76
DePaulo, B. M., Kashy, D. A., Kirkendol, S. E., Wyer, M. M., & Epstein, J. A. (1996). Lying in everyday life. Journal of Personality and Social Psychology, 70, 979 995. Retrieved from http://www.apa.org/pubs/ journals/psp/ DePaulo, B. M., Lindsay, J. L., Malone, B. E., Muhlenbruck, L., Charlton, K., & Cooper, H. (2003). Cues to deception. Psychological Bulletin, 129, 74 118. doi:10.1037/0033-2909.129.1.74 Driscoll, L. N. (1994). Validity assessment of written statements from suspects in criminal investigations using the SCAN technique. Police Studies, 17, 77 88. Gilovich, T., Savitsky, K., & Medvec, V. H. (1998). The illusion of transparency: Biased assessments of others ability to read ones emotional states. Journal of Personality and Social Psychology, 75, 332 346. Retrieved from http://www.apa.org/pubs/journals/psp/ Granhag, P. A., Strmwall, L. A., & Hartwig, M. (2007). The SUE technique: The way to interview to detect deception. Forensic Update, 88, January, 2529. Retrieved from http://www.bps.org.uk/dfp/ psychologists/forensic-update.cfm Hartwig, M., Granhag, P. A., & Strmwall, L. (2007). Guilty and innocent suspects strategies during interrogations. Psychology, Crime, & Law, 13, 213227. doi:10.1080/10683160600750264 Hartwig, M., Granhag, P. A., Strmwall, L, & Doering, N. (2010). Impression and information management: On the strategic self-regulation of innocent and guilty suspects. The Open Criminology Journal, 3, 10 16 (special issue on deception research). doi:10.2174/ 1874917801003010010 Heydon, G. (2008). The art of deception: Myths about lie detection in written statements. In L. Smets & A. Vrij (Eds.), Het analyseren van de geloofwaardigheid van verhoren (pp. 171182). Brussels: Politeia. Johnson, M. K. (2006). Memory and reality. American Psychologist, 61, 760 771. doi:10.1037/0003-066X.61.8.760 Johnson, M. K., Bush, J. G., & Mitchell, K. J. (1998). Interpersonal reality monitoring: Judging the sources of other peoples memories. Social Cognition, 16, 199 224. Retrieved from http://wwwguilford.com/ cgi-bin/cartscript.cgi?pagepr/jnco.htm&dirperiodicals/per_psych &cart_id Johnson, M. K., Foley, M. A., Suengas, A. G., & Raye, C. L. (1988). Phenomenal characteristics of memories for perceived and imagined autobiographical events. Journal of Experimental Psychology: General, 117, 371376. Retrieved from http://www.apa.org/pubs/journals/xge/ Johnson, M. K., & Raye, C. L. (1981). Reality monitoring. Psychological Review, 88, 67 85. Retrieved from http://www.apa.org/pubs/journals/ rev/index.aspx Johnson, M. K., & Suengas, A. G. (1989). Reality monitoring judgments of other peoples memories. Bulletin of the Psycho-nomic Society, 27, 107110. Kassin, S. M., Appleby, S. C., & Torkildson-Perillo, J. (2010). Interviewing suspects: Practice, science, and future directions. Legal and Criminological Psychology, 15, 39 56. doi:10.1348/135532509X449361 Khnken, G. (1996). Social psychology and the law. In G. R. Semin & K. Fiedler (Eds.), Applied social psychology (pp. 257282). London, Great Britain: Sage Publications. Khnken, G., & Steller, M. (1988). The evaluation of the credibility of child witness statements in German procedural system. In G. Davies & J. Drinkwater (Eds.), The child witness: Do the courts abuse children? (Issues in Criminological and Legal Psychology, no. 13) (pp. 37 45). Leicester, United Kingdom: British Psychological Society. Lerner, M. J. (1980). The belief in a just world: A fundamental delusion. New York: Plenum. Masip, J., Sporer, S. L., Garrido, E., & Herrero, C. (2005). The detection of deception with the reality monitoring approach: A review of the empirical evidence. Psychology, Crime & Law, 11, 99 122. doi: 10.1080/10683160410001726356 Porter, S., & Yuille, J. C. (1996). The language ofdeceit: An investigation of the verbal clues to deception in the interrogation context. Law and Human Behavior, 20, 443 459. doi:10.1007/BF01498980 Sapir, A. (1987/2000). The LSI course on scientific content analysis (SCAN) Phoenix, AZ: Laboratory for scientific interrogation. Shuy, R. (1993). Language crimes: The use and abuse of language evidence in the courtroom. Cambridge, MA: Blackwell. Smith, N. (2001). Reading between the lines: An evaluation of the scientific content analysis technique (SCAN). London: UK Home Office, Research, Development and Statistics Directorate. 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, 373397. doi:10.1002/ (SICI)1099-0720(199710)11:5373:AID-CP461[3.0.CO;2-0 Sporer, S. L. (2004). Reality monitoring and detection of deception. In P. A. Granhag & L. A. Stromwall (Eds.), The detection of deception in forensic contexts (pp. 64 102). UK: Cambridge university press. A., Hartwig, M., & Granhag, P. A. (2006). To act truthfully: Nonverbal behaviour and strategies during a police interrogation. Psychology, Crime, & Law, 12, 207219. doi:10.1080/10683160512331331328 Vrij, A. (2008). Detecting lies and deceit: Pitfalls and opportunities. Chichester, England: Wiley. Vrij, A., Granhag, P. A., & Porter, S. B. (in press). Pitfalls and opportunities in nonverbal and verbal lie detection. Psychological Science in the Public Interest. Retrieved from http://www.wiley.com/bw/journal .asp?ref1529-1006 Vrij, A., Leal, S., Granhag, P. A., Mann, S., Fisher, R. P., Hillman, J., et al. (2009). Outsmarting the liars: The benefit of asking unanticipated questions. Law and Human Behavior, 33, 159 166. doi:10.1007/ s10979-008-9143-y Vrij, A., Mann, S., & Fisher, R. (2006). Information-gathering vs accusatory interview style: Individual differences in respondents experiences. Personality and Individual Differences, 41, 589 599. doi:10.1016/ j.paid.2006.02.014

You might also like