Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

APPLIED COGNITIVE PSYCHOLOGY

Appl. Cognit. Psychol. 16: 287300 (2002) Published online in Wiley InterScience 11 February 2002 (www.interscience.wiley.com) DOI: 10.1002/acp.788

Interviewing Techniques and the Assessment of Statement Credibility


KEVIN COLWELL1*, CHERYL K. HISCOCK1 and AMINA MEMON2
1

Sam Houston State University, USA 2 University of Aberdeen, UK

SUMMARY The objective of this study is the development of an applicable comprehensive questioning and statement analysis procedure. One hundred and thirty-six male residents of the Wynne Unit of the Texas Department of Criminal Justice Institutional Division (TDCJ ID) witnessed a staged theft, and provided testimony. Interview formats followed semi-standardized scripts derived from Structured Interview, Inferential Interview and Cognitive Interview techniques. Participants were randomly assigned to one of six experimental conditions in a 2(honesty of reporting) 3(interview technique) matrix. Results found a 62% classication accuracy for the Structured Interview, 68% for the Cognitive Interview and 82% for the Inferential Interview, when comparisons were made between treatment phases of each interview condition. When responses given to each interview segment were analysed, 83% of the Structured Interview transcripts were correctly identied, as were 91% of the Inferential and 92.9% of the Cognitive Interview statements. The desire to prevaricate while escaping detection produced statements which were systematically different from honest reporting, as seen in the variables of coherence, response length, type-token ratio, and verbal hedges. These results indicate a potential forensic utility for strategies which attempt to detect deception through a combination of qualitative and quantitative statement characteristics, and underscore the need for systematic, question-by-question analysis of eyewitness statements. Copyright # 2002 John Wiley & Sons, Ltd.

The criminal justice system places a great deal of faith in eyewitness testimony despite psychological research which suggests eyewitness reports may be inaccurate (Cutler and Penrod, 1995; Loftus and Ketcham, 1991; Memon et al., 1998; Sporer et al., 1996; Wright and Davies, in press). Furthermore, the degree of condence presented by the witness contributes to his or her believability (Cutler et al., 1988; Penrod and Cutler, 1995) indicating that the demeanuor of the witness has as much to do with perceived credibility as the testimony he or she supplies. As eyewitness testimony is the single most effective predictor of case outcome, it is incumbent upon the criminal justice community to develop valid and reliable techniques to assess the credibility of such statements. Such a system would provide an objective means to corroborate statements made by witnesses, allowing for more accurate, and possibly more efcient adjudication. This point becomes especially salient when one considers the relatively poor ability of both professionals and lay persons in detecting deception (Vrij et al., 1997).
*Correspondence to: Kevin Colwell, Department of Psychology, Sam Houston State University, Huntsville, TX 77341, USA. E-mail: psy_kwc@shsu.edu

Copyright # 2002 John Wiley & Sons, Ltd.

288

K. Colwell et al.

It has been consistently observed that individuals fail to accurately discriminate honest from dishonest statements (Bond et al., 1985). This has been demonstrated with many professionals including police ofcers, judges, psychiatrists, university students, and agents from several government law enforcement branches (Ekman and OSullivan 1991; DePaulo and Pfeifer, 1986; see Vrij and Akehurst, 1998 for a recent review). There are three prominent hypotheses that may explain our inability to detect deceit. First, it has been suggested (Ekman and Friesen, 1974, 1969; Ekman et al., 1991) that observers are typically unable to detect motivated deceit because prevaricators are aware of which portion(s) of their statements are untrue, and are therefore able to monitor and inhibit obvious behavioral cues. The remaining behavioral correlates of deception are usually so subtle that they are only detectable via technical analysis (such as Ekmans Facial Action Coding System; Ekman et al., 1988). Second, McCornack and Parks (1985) have proposed a truthful judging bias: Individuals are generally more likely to judge a statement as truthful than deceptive. Finally, DePaulo et al., (1984) posit that an inadequate amount of feedback concerning the veracity of a statement makes it difcult for individuals to hone their lie detecting skills. All three explanations underscore the need for a systematic approach to lie detection. The statements of the alleged victim and perpetrator are often the only evidence available to authorities a state of affairs which impels the development of systematic approaches to discriminate honest reporting from fabrication. Several means of assessment have been proposed, encompassing motivational, cognitive, and memory based approaches to deception. Predominant among the memory-based approaches is Statement Validity Analysis (SVA). SVA is based on the Undeutsch hypothesis, which holds that memory for a real event will differ from fabrication in structure, content, and quality. This has received some empirical support, and has been used for court-mandated assessments of childrens testimony in East and West Germany and Sweden for over 30 years (Farr and Yuille, 1988; Arntzen, 1982, 1983; Trankell, 1972; Undeutsch, 1967, 1982). The emotionalmotivational perspective to understanding deception assumes that lying is a difcult task, with considerable anxiety produced by failure. As this anxiety builds, the likelihood of detection increases due to increased physiological arousal. It is contended that it is most difcult to lie successfully when the consequences of detection are great, later known as the motivational impairment effect (DePaulo et al., 1988). In accord with this idea, highly motivated deception has been associated with shorter response length, a slower rate of speech, and more speech errors; paralinguistic behaviors known as a leakage. Congruent with this theory is the nding by Comber and Canter (1983) that malicious (fabricated) re-alarm phone calls can be discriminated from non-malicious (true) re-alarms due to a higher incidence of aahs and uhms (lled pauses), and a higher rate of speech errors in the fabricated accounts. Cognitive approaches to deception assume that interrogation is a setting in which deception is an adaptive response to situational demands of conveying information and minimizing damage. Effective communication requires that those involved in conversation are responding as accurately and completely as possible. Information Manipulation Theory (IMT) (McCornack, 1992) postulates that deception violates these covert rules of conversational exchange, and this leads to deceptive information being different in quantity, quality, manner, and relevance from truthful information. Therefore, one engaged in deception can take advantage of these implicit assumptions to control the release of information which could potentially lead to their being detected. This theory is used to account for lies of omission as well as commission.
Copyright # 2002 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 16: 287300 (2002)

Statement credibility

289

Highly motivated deception is hypothesized to result in noticeably different patterns of speech (Ruby and Brigham, 1997). It has been contended (Osgood, 1960) that people who lie tend to become more stereotypical in their responses, which leads to less lexical diversity as measured by the type-token ratio (TTR). The TTR is the number of distinct words, types, divided by the total number of words, tokens, in a statement. For example, the sentence, One small step for man, one giant leap for mankind has a TTR of 0.80 eight non-repetitive words divided by ten total. Low anxiety will hypothetically lead to higher TTR and high anxiety will lead to lower TTR (Porter and Yuille, 1995; Carpenter, 1990). The former arises from an attempt at impression management, the latter stems from the motivational impairment effect. It is suggested that deceptive persons are aware of their prevarications, and therefore are able to carefully phrase the apocryphal portions of their statements (Vrij and Akehurst, 1998).

INVESTIGATIVE INTERVIEWING The goals of an effective interview are (a) minimize the trauma of investigation, (b) maximize the information obtained about the event(s), (c) minimize the contamination of the memory trace by the interview, and (d) maintain the integrity of the investigative process (Yuille et al., 1993). Interview techniques which proportedly meet the above goals will be compared to assess the relative efcacy of each in the assessment of statement credibility. These include the Cognitive Interview (Geiselman and Fisher, 1997), the Structured Interview (Yuille et al., 1993; Yuille and Cutshall, 1989; Farr and Yuille, 1988), and a new technique, the Inferential Interview. The core of the Structured Interview is an organization designed to maximize recall and minimize contamination.1 The steps of the interview begin with the most open, least leading forms of questioning, and then progress to more specic questioning as circumstances require. The initial goal is to provide every opportunity to give a free narrative account before more specic questions are used. After the free narrative is provided, the participant is prompted to elaborate on details mentioned in the narrative through the introduction of open-ended questions. Specic probing questions are then used to obtain clarication and exhaust detail recall. The last phase of interviewing requires the participant to recount the entire event. The Cognitive Interview (CI) technique (Fisher and Geiselman, 1992) has also demonstrated promise as a method of effectively meeting the above goals (e.g. Geiselman et al., 1995). The CI follows the same principles as the SI, but is based on four mnemonic strategies derived from pooling research on retrieval pathways, encoding specicity, and schema-related recall (Bekerian and Dennett, 1993; Geiselman and Fisher, 1997; Memon, 1997; Memon and Koehnken, 1992; For detailed reviews, see Memon and Stevenage, 1996; Memon and Higham, in press; or Koehnken et al., in press). The incorporation of mnemonics allows for maximal recall by exhausting potential retrieval cues and providing
1 Note that the term Structured Interview has been applied to a variety of interview formats all with the aim of structuring the witnesses account using good questioning techniques. Koehnken et al. (1995) and Memon et al. (1997) used the Structured Interview as a control group for the Cognitive Interview (see Memon and Stevenage, 1996 for a description). The present study followed the Structured Interview format advocated by Yuille and colleagues.

Copyright # 2002 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 16: 287300 (2002)

290

K. Colwell et al.

multiple opportunities to access encoded information (See Method section for a description of the procedure). Because lying is a cognitively taxing task, a usual strategy of prevarication is to mentally prepare and rehearse a ctitious account of an event. In this case, the prevaricator employs supercial-encoding to respond to questioning, i.e. responses are based on the well-rehearsed script rather than objective reality. This allows for the use of lower levels of cognitive processing, which in turn minimizes leakage (Porter and Yuille, 1995). A prevaricator responding to inferential questioning is engaged in unrehearsed lying. This necessitates a deeper level of cognitive processing than does rehearsed lying, and results in increased anxiety and cognitive demand, thereby evoking more leakage. So, inferential questioning at the outset of an interview section is proposed to counter supercial encoding, promote deeper level data-driven processing, and subsequently thwart efforts to prevaricate by interjecting improvisational material into an otherwise well-rehearsed script. To meet the above goal, a third interviewing condition contains short answer inferential questions inserted between each CI mnemonic section. This Inferential Interview is an elaboration of the CI technique which we propose as a possible statement analysis tool. The underlying rationale is that the mnemonics, which exhaust detail recall, coupled with the inferential questions, which promote deeper (rather than supercial) processing, will necessitate unrehearsed lying for those who prevaricate and result in discernibly different patterns of speech. Conversely, these same manipulations should have little or no effect on the speech patterns of those who are truthful. The objective of this study was the preliminary development of an applicable comprehensive questioning and statement analysis procedure which facilitates complete and accurate recall, minimizes contamination and allows for the detection of deception through systematic analysis of verbal content. Attention is given to interviewing strategy as well as psycholinguistic characteristics of the resulting statements. This study takes external validity into consideration by addressing a forensically relevant population offenders within a maximum-security Texas prison. A diverse population of incarcerated offenders had not been used to date in verbal credibility assessment research. Also, a staged theft using live actors was used as the stimulus material. Much of the previous research of this nature has used video recordings or written descriptions for this purpose. The measures assessed included: the type-token ratio (TTR a measure of lexical diversity calculated by dividing the number of unique words in a statement by the total number of words in the statement), the frequency of verbal hedges (dened as verbal techniques which allow for additional processing time that the witness can use to formulate a lie), coherence (Does the statement make sense? Is it in violation of the laws of nature? Is there a logical consistency within the statement, or does the statement contradict itself.), and response length (the number of words in a statement). It is hypothesized that the cognitive demands associated with being deceptive will produce statements which differ signicantly from true accounts. Specically, fabrication should be associated with less coherence, shorter responses, more verbal hedges, and an increase in TTR.

METHOD One hundred and thirty-six randomly selected residents of the Wynne Unit of the Texas Department of Criminal Justice Institutional Division (TDCJ ID) witnessed a staged theft, and provided testimony. Participation was voluntary, following written informed
Copyright # 2002 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 16: 287300 (2002)

Statement credibility

291

consent, and participants were able to terminate at any time. Security concerns limited the sample to general population offenders. This population was drawn is composed of 43.7% Caucasian, 36.3% African-American, 13% Hispanic, and 6% other (primarily Asian) males. This study utilized a 2(honesty of reporting) 3(interview technique) analogue experimental design. A staged theft was presented to participants on their unit of assignment. They were then instructed to report honestly, or to distort testimony to exculpate the perpetrator (honest or dishonest reporting). Interviews were conducted in one of three conditions (SI, CI, or II) to elicit accounts of the event. The theft was staged using live actors, two of whom posed as research participants and one as a researcher. Groups of approximately thirty participants were seated in education rooms in the prison unit. After informed consent had been obtained, the two confederates entered the room and began conversation with the researcher. After some conversation, the researcher left the room, and one of the confederates stole several items off of her desk. He made an excuse and left. When the researcher returned, she noticed that items were missing, and enlisted the help of the remaining participant in contacting the campus police. The entire scripted scene lasted approximately 3 minutes, and was repeated for each group of participants. Individual interviews were conducted by trained interviewers. Interview formats followed scripted versions of SI, CI, or II. Interviewer training consisted of a one-hour lecture concerning memory and cognition and two hours spent describing the specic techniques. A sample videotape of each technique followed the didactic sessions. Interviewers were then divided into pairs and role-played each of the techniques. All interviewers were trained in each technique. Finally, a pilot study was conducted on campus at Sam Houston State University. This provided in-vivo experience with each technique. The script employed in each interview was contingent upon experimental condition. The CI and II conditions differed only in the presence or absence of an inferential questioning component. These were yes or no questions such as Was there a crime committed? All interviews began with standard demographics questions followed by two baseline recall tasks. Structured Interview The SI began with a free narrative, followed by specic open-ended probing questions for clarication. The free narrative was elicited by the request Please describe, in as much detail as possible, everything you remember happening in the event that you witnessed. The probe questions used the same structure, but concerned specic actors in the scene. Thus, a typical probe request would be Please describe, in as much detail as possible, everything you remember about the woman behind the desk. Interviewers were trained to only use terminology and details which was rst introduced by the participant, in order avoid introduction of new material into the report. Three specic probes were used, one for each confederate. Finally, each participant was asked to describe the entire scene again. Cognitive Interview This interview began with the same questions as the SI. Following the free narrative, the participants were asked to describe the entire event after mentally reinstating the sights,
Copyright # 2002 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 16: 287300 (2002)

292

K. Colwell et al.

sounds, thoughts, and any other items they could remember from the time of the event (Fisher and Geiselman, 1992). Participants were then asked to describe the event from the perspective of someone else in the room. Lastly, they were required to recount the entire event backwards. Inferential Interview Participants in the Inferential Interview conditions were questioned in the same manner as those in the CI conditions, with the exception of nine yes or no questions. These were inserted between the questions of the CI technique in groups of three. Typical inferential questions were Was this an act of violence? and Did anyone intend to harm anyone else? Participants were randomly assigned to one of six experimental conditions in a 2(honesty of reporting) 3(interview technique) matrix, with the conditions as follows: 1 Inferential Interview, honest reporting; 2 Inferential Interview, dishonest reporting; 3 Cognitive Interview, honest reporting; 4 Cognitive Interview, dishonest reporting; 5 Structured Interview, honest reporting; 6 Structured Interview, dishonest reporting. Participants in the honest reporting conditions were simply asked to report everything as completely as possible. Those in the dishonest conditions were told to distort testimony to preclude conviction of the perpetrator. In order to operationalize the dependent measures, the following denitions were used: (1) Coherence does the statement make sense, not contradict itself, and not violate the laws of nature? If so, then it is coherent. (2) Response length total number of words in a statement, (3) Type-Token Ratio unique words divided by total words in a statement. (4) Verbal Hedges verbal techniques used to buy additional processing time. These fall into four major categories: (1) Unnecessary connectors Uhs. . .Ums. . .ands. . .ors. . . etc. (2) Repeated questions when the participant repeats all or part of the experimenters question before responding, e.g. Tell you what I saw just now in the room? or In reverse order? (3) Claimed lack of memory followed by a description of the event, e.g. I really cant remember (or couldnt see, or wasnt paying attention) but I saw this guy come in with an attitude. (4) Repeated information which is not used to clarify, e.g. He left, he left the, he left the room with her rings. Participants witnessed a staged theft. After the scene, numbered instruction packets, previously randomized, were distributed. These packets determined interview and bias conditions. Interviews were conducted after a short delay. Interviews were recorded, transcribed, and analysed for patterns of speech in an attempt to differentiate among experimental conditions by statement content i.e. to ascertain statement credibility through systematic variations in verbal behaviour. Scoring of response length and the Type-Token Ratio was computer generated through Wordscan, software developed for this purpose. Coherence and frequency of verbal hedges were scored blindly by the rst author and four undergraduate research assistants. Raters participated in three one-hour sessions during which transcripts containing allegations of sexual abuse were scored as a group. After the nal session, all subsequent rating rules were created by group consensus. Test retest reliability for response length and Type-Token ratio was 1.00. To address the reliability of Coherence and the Frequency of Verbal Hedges, all the raters were required
Copyright # 2002 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 16: 287300 (2002)

Statement credibility

293

to score a subset of twenty-eight transcripts. Proportion agreement for presence or absence of coherence was 0.83. Scoring of verbal hedges was acceptable, but somewhat less reliable, r 0.67, using a Pearson ProductMoment correlation coefcient.

RESULTS The data were analysed using a multivariate analysis of variance. Honesty of reporting and interview conditions were used as factors. Dependent measures were the Type-Token Ratio of the treatment phase (TTRT) for each statement, coherence (COH), response length (RESP), and the frequency of verbal hedges (VH). This analysis indicated signicant differences between experimental conditions for honesty of reporting; F(4, 116) 5.66, p<0.05 (see footnote 2). No signicant differences emerged as a function of interviewing technique, and no signicant interactions were present. Univariate analyses revealed signicant effects for COH, TTRT, and RESP. The F-values and level of signicance for each dependent measure are reported in Table 2. A discriminant function was derived from the entire sample, with dependent measures being coherence, TTRT, response length, and the frequency of verbal hedges. This allowed the prediction of group membership for each case as a function of the dependent measures i.e. statements were classied as true or false based upon group means and variances. As a whole, 68.18% of the statements were accurately classied as true or fabricated on the bases of these four variables, with 76% of the true transcripts (sensitivity proportion of true transcripts labelled as credible) and 59.7% of the fabricated transcripts correctly classied (specicity proportion of dishonest statements labelled as non-credible). (See Table 3) To compare the utility of each interviewing strategy in the assessment of statement credibility, separate discriminant functions, based upon the subsamples represented within each interviewing condition, were used to classify the statements elicited by each technique. The SI allowed for an overall hit rate of 62%, correctly classifying 66.7% of the true and 57.75% of the false statements. The CI effectively discriminated 67.74% of

Table 1. Group means by interview for dependent measures during treatment phase True Interview Inferential Cognitive Structured COH 1.13 1.0 1.05 VH 0.0288 0.0237 0.0135 TTR 0.524 0.419 0.511 RESP 510 796 856 COH 1.36 1.28 1.37 VH 0.233 0.0627 0.0280 False TTR 0.61 0.56 0.67 RESP 326 412 405

Table 2. Univariate F ratios and level of signicance for dependent measures for the main effectof truth Variable Coherence Response length TTR Verbal hedges
Copyright # 2002 John Wiley & Sons, Ltd.

F ratio 10.022 9.391 10.804 2.870

Signicance 0.002 0.003 0.001 0.093


Appl. Cognit. Psychol. 16: 287300 (2002)

294

K. Colwell et al.

Table 3. Overall classication rates, treatment phase taken as a whole Actual group TRUE FALSE Number of cases 65 67 Predicted true 50 76.9% 27 40.3% Predicted false 15 23.1% 40 59.7%

Note: Four cases were not included in this analysis. Overall accuracy 68.18%.

Table 4. Classication rates for each interview condition, across entire treatment phase Actual group TRUE FALSE Overall accuracy 62.00% Cognitive interview TRUE 8 3 72% 27.3% FALSE 20 7 13 35.0% 65.0% Nine cases were excluded from this analysis due to missing data. Overall accuracy 67.74% TRUE FALSE 26 18 Inferential interview 23 88.5% 5 27.8% 3 11.5% 13 72.2% 11 Number 24 26 Predicted true Structured interview 16 66.7% 11 42.3% Predicted false 8 33.3% 15 57.7%

Six cases were excluded from this analysis due to missing data. Overall accuracy 81.82%

the true from fabricated statements; 72.7% of the true accounts and 65% of the false accounts were accurately classied. The II condition yielded an overall hit rate of 81.82%: 88% of the true cases and 72.2% of the false cases were correctly classied in this condition. (See Table 4.) One purpose of this study was to increase our ability to detect deception through the application of semi-standardized interview scripts which facilitate question-by-question analysis of resulting statements. Traditional interviewing techniques are sufciently amorphous to preclude this type of investigation. The techniques used here, however, were designed with this in mind. Initially, 112 participants from the original sample were analysed in an interview segment by honesty of reporting by dependent measures MANOVA (16, 95) F 3.381, p<0.01. Transcripts which did not include complete responses to all questions were excluded from further analysis. Subsequent univariate analyses (ANOVAs) revealed signicant main effects for truth in the variables coherence, response length (free recall), verbal hedges (free recall), and verbal hedges (contextual embedding). A discriminant function was used to classify statements as true or false, with an overall hit rate of 80.4%. Fifty of fty-eight true transcripts were accurately classied
Copyright # 2002 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 16: 287300 (2002)

Statement credibility

295

Table 5. Classication results for each interview using dependent measures by questioning phase Actual group TRUE FALSE Overall accuracy 83.00% Cognitive interview TRUE FALSE Overall accuracy 92.90% TRUE FALSE Overall accuracy 91.9% 23 14 Inferential interview 22 95.7% 2 14.3% 1 4.3% 12 85.7% 11 17 10 90.9% 1 5.9% 1 9.1% 16 94.1% Number 24 23 Predicted true Structured interview 20 83.3% 4 17.4% Predicted false 4 16.7% 19 82.6%

(sensitivity 86.82%), as were 40 of 54 false transcripts (specicity 74.1%). (See Table 5.) Further elucidation of the effects of question-by-question analysis was obtained by comparing the accuracy of predictions made from a discriminant derived from the interview as a whole to the predictive accuracy of a discriminant which considers each question separately. The procedure used follows one employed by Sporer (1997). Accuracy of the predictions made by each function was coded as zero (inaccurate) or one (accurate). A repeated measures MANOVA was then conducted on these accuracy scores. Results indicated a signicant increase in predictive ability as a result of more complex analysis; F(1, 94) 8.898, p<0.05. To more accurately assess the credibility of statements it is necessary to derive discriminant functions from the subsamples comprising each interview technique. This is because the demands of the various techniques generate noticeably (though not statistically signicant) different effects on the dependent measures. A discriminant function was derived from the subsample present in each interview condition, and group membership (honesty of reporting) was predicted using this function. Structured Interview This procedure was the least effective of the three evaluated here. Again, this difference is not statistically signicant, but this correlation is important for practical application. Twenty of twenty-four transcripts were correctly predicted as true (sensitivity 83.3%). Nineteen of twenty-three statements were accurately labelled as false (specicity 82.6%) Overall, 83% of the transcripts in this condition were accurately classied. This was noticeably lower than the classication rates of the other interviewing strategies, but this is still a considerable improvement over chance (50%). There were signicant differences at the 0.05 level for the variables of coherence, response
Copyright # 2002 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 16: 287300 (2002)

296

K. Colwell et al.

length during free-recall, and the frequency of verbal hedges when describing the research assistant. In this condition, shorter response lengths, an increase in TTR, and an increase in the frequency of verbal hedges were associated with prevarication.

Inferential Interview The performance of this interviewing strategy was somewhat improved over the SI and comparable to the CI, with an overall discrimination rate of 91.9%. Twenty-two of twentythree true statements were predicted as such (sensitivity 95.7%). Twelve of fourteen false statements were grouped as apocryphal (specicity 85.7%). In this condition, deception was associated with shorter responses, increased verbal hedging, an increase in TTR during the free-recall, reinstatement of context, and changing perspective mnemonics, and a decreased TTR during reverse order recall.

Cognitive Interview The CI generated the best results of any strategy employed in the current investigation. Ten of eleven true statements were accurately classied (sensitivity 90.9%). Sixteen of seventeen false statements were properly labeled as such (specicity 94.1%). Cumulatively, 92.9% of the statements in this condition were accurately classied. Prevarication in this condition was correlated with shorter responses, a change in TTR, and increased verbal hedging.

DISCUSSION The purpose of this investigation was to take preliminary steps in the development of an interviewing and statement analysis procedure which can be used to assess statements given by adult eyewitnesses. To accomplish this, memory-based and cognitive-motivational indices of credibility were combined and used in conjunction with current investigative interviewing techniques. The data obtained in this study are promising; revealing consistent trends associated with deception across interviewing conditions. The novel introduction of semi-scripted interview formats must be further investigated. This approach offers the benet of allowing for more complex statistical models to predict credibility, but this gain is negated when participants fail to respond to all the questions.

Detection of deception One purpose of this study was the demonstration of psycholinguistic variables which effectively discriminate truthful from fabricated eyewitness testimony. To this end, it was shown that truthful accounts are more coherent, are typically longer than dishonest accounts, and contain less verbal hedging. Interestingly, the fabricated accounts exhibited greater lexical diversity. The act of deception apparently increased the verbal performance of the participants in the dishonest reporting conditions. It is possible that mild anxiety improved performance through impression management on the part of the deceptive witnesses. As others have noted, (Carpenter, 1990; Vrij and Akehurst 1998) those who lie
Copyright # 2002 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 16: 287300 (2002)

Statement credibility

297

are aware which portion(s) of their statements are false or incriminating, and consequently phrase those more carefully.2 Interviewing techniques and credibility assessment To more properly address the relationship between prevarication and statement characteristics, the responses to each questioning phase of the interviews were analyzed. These results generated a statistically signicant increase in our ability to discriminate true from fabricated accounts. As a whole, 84% of the data set was accurately classied in this manner excluding those statements which did not have a legibly transcribed response for every question posed. The SI (83% classication accuracy) was somewhat less efcacious in making discriminations(though not signicantly so), when compared to the II(91.9%), and CI(92.9%) techniques. The performance of the latter two techniques was essentially the same, which stands to reason, given the minor differences between the two strategies. Such an increase in discriminative ability as a result of question-byquestion analysis cannot be ignored, and future investigations will need to address this as a possible means of increasing our ability to detect deception in forensically applicable settings. Comparisons to other statement analysis techniques Porter and Yuille (1995) obtained an overall classication rate of 78.3%, using the criteria of details, coherence, and admissions. Koehnken et al. (1995) were able to effectively discriminate 93% of truthful from fabricated accounts, using nineteen content-criteria, thirteen of which were original SVA variables. In the Koehnken study the performances of the Cognitive and Structured Interviews were compared. Classication rates were 100% for the Cognitive and 93% for the Structured Interview. The classication rates of 83% (SI), 92.9% (CI), and 91.9% (II) are slightly lower than those obtained by Koehnken et al. Criticisms of existing statement analysis techniques include: (1) that the statistics used are not appropriate for the data being considered, (2) adult memory is sufciently complex to allow for successful deception when memory-based criteria are used, (3) current techniques generate low intercoder reliability, (4) the generally high rate of false-positives generated by these techniques, and (5) analog studies do not produce sufcient levels of motivation. Many of these points were considered in the current investigation. With the exception of coherence, the data generated in the present study are continuous, ratio level frequency data, and appropriate for the analyses conducted. The currently proposed system of assessment focuses on qualitative characteristics that emerge from examination of a statement as a whole (coherence), as well as quantitative examination of motivation-related criteria (Type-Token Ratio, verbal hedges, and response length), as they arise in response to each individual question that comprises a statement. When individual responses were analyzed, only seven false-positives arose from a set of ftyfour fabricated statements. This gives a false-positive rate of 13%. A low rate of
2 The possibility that the difference evidenced in TTRT could be a by-product of the shorter responses in the dishonest reporting condition was considered. Due to this, the data were re-analysed excluding all cases in which the response length for the treatment phase was less than 100 words. The results of this analysis were similar to those reported above. 3 It should be noted that the baseline measures for all dependent variables were compared across experimental conditions. There were no signicant differences evident, indicating that the differences obtained in the treatment conditions are a result of instructions to distort testimony.

Copyright # 2002 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 16: 287300 (2002)

298

K. Colwell et al.

false-positives is extremely important in any potential forensic application, because of the often serious consequences of erroneous case disposition. Ruby and Brigham (1997), in a recent review of SVA, cautioned about the application of these techniques in the legal arena, due to doubts that surround the ideographic application of statement analysis techniques that are derived from nomothetic analog data. Future experiments must address the utility of this system of assessment with highly motivated participants, and consider statements given in actual forensic interviews. CONCLUSIONS The performance of the Cognitive and Inferential Interviews gives strong indication of their potential as applied investigative interviewing and statement analysis techniques, and this becomes especially salient when one considers the population used in this study. Statement analysis using a semi-structured interview format allows for examination of each discrete questioning phase, in addition to the evaluation of a statement as a whole, produces a more complex statistical model from which to make predictions, and thereby increases the amount of information available for assessment decisions. Question-byquestion analyses signicantly increased predictive accuracy, and may facilitate the evolution of statement analysis techniques from checklist formats to standardized instruments. Overall, the quantitative examination of emotional/motivational statement characteristics, coupled with semi-structured interviewing and question-by-question analysis, may enhance the ability of current statement analysis techniques. These considerations must be taken into account in future research. REFERENCES
Arntzen F. 1982. Die Situation der Forensischen Aussagenpsychologie in der BundesrepublikDeutschland Forensic psychology in West-Germany: State of the art. In Reconstructing the Past: The role of psychologists in criminal trials, Trankell A (ed.). Kluwer: Deventer; 107120. Arntzen F. 1983. Psychologie der Zeugenaussage: Systematik der Glaubwrdigkeitsmerkmale u (Psychology of witness statements). C. H. Beck: Munich. Bekerian DA, Dennett JL. 1993. The cognitive interview: Reviving the issues. Applied Cognitive Psychology 7: 275298. Bond CF Jr, Kahler KN, Paolicelli LM. 1985.The miscommunication of deception: An adaptive perspective. Journal of Experimental Social Psychology 21: 331345. Bower G. 1967. A multicomponent theory of memory trace. In The Psychology of Learning and Motivation (Vol 1), Spence KW, Spence JT (eds). Academic Press: New York. Boychuk T. 1991. Criteria-based Content Analysis of Childrens Statements about Sexual Abuse: A Field-Based Validation Study. Unpublished doctoral dissertation, Arizona State University. Carpenter R. 1990. The statistical prole of language behavior with Machiavellian intent or while experiencing caution or avoiding self-incrimination. Annals of the New York Academy of Sciences 606: 516. Cutler BL, Penrod SD. 1995. Mistaken Identication: The eyewitness, psychology and the law. Cambridge University Press: New York. Cutler BL, Penrod SD, Stuve TE. 1988. Juror decision making in eyewitness identication cases. Law and Human Behavior 12: 4155. DePaulo BM, Pfeifer RL. 1986. On-the-job experience and skill at detecting deception. Journal of Applied Social Psychology 16: 249267. DePaulo BM, Kirkendol SE, Tang J, OBrien TP. 1988. The motivational impairment effect in the communication of deception: Replications and extensions. Journal of Nonverbal Behavior 12(3): 177202.
Copyright # 2002 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 16: 287300 (2002)

Statement credibility

299

DePaulo BM, LeMay CS, Epstein JA. 1991. Effects of importance of success and expectations for success on effectiveness at deceiving. Personality and Social Psychology Bulletin 17(1): 1424. DePaulo BM, Stone JI, Lassiter GD. 1984. Deceiving and detecting deceit. In The Self and Social Life, Schlenker BR (ed.). McGraw-Hill: New York; 323370. Ekman P. 1988. Lying and nonverbal behavior: Theoretical issues and new ndings. Journal of Nonverbal Behavior 12(3): 163175. Ekman P, Friesen WV. 1969. Nonverbal leakage and clues to deception. Psychiatry 32: 88106. Ekman P, Friesen WV. 1974. Detecting deception from the body or the face. Journal of Personality and Social Psychology 29: 288298. Ekman P, Friesen WV, OSullivan M. 1988. Smiles when lying. Journal of Personality and Social Psychology 54(3): 414420. Ekman P, OSullivan M. 1991. Who can catch a liar? American Psychologist 46(9): 913920. Ekman P, OSullivan M, Friesen WV, Scherer KR. 1991. Invited article: Face, voice, and body in detecting deceit. Journal of Nonverbal Behavior 15(2): 125135. Elliot R, Farrington B, Manheimer H. 1988. Eyewitness credible and discredible. Journal of Applied Social Psychology 18: 14111422. Farr VL, Yuille JC. 1988. Investigating and assessing the credibility of childrens allegations of sexual abuse: Issues for professionals. Preventing Sexual Abuse 1(1): 812. Fisher RP, Geiselman RE. 1992. Memory Enhancing Techniques for Investigative Interviewing: The Cognitive Interview. Charles C. Thomas: Springeld, IL. Geiselman RE, Fisher RP, MacKinnon DP, Holland HL. 1985. Eyewitness memory enhancement in the police interview: cognitive retrieval mnemonics versus hypnosis, Journal of Applied Psychology 70: 401412. Geiselman RE, Fisher RP. 1997. Ten years of cognitive interviewing. In Intersections in Basic and Applied Memory Research, Payne D, Conrad F (eds). Lawrence Erlbaum: Mahwah, NJ. Hollien H. 1990. The Acoustics of Crime: The New Science of Forensic Phonetics. Plenum: New York. Koehnken G, Schimossek E, Aschermann E, Hofer E. 1995. The Cognitive Interview and the assessment of the credibility of adults statements. Journal of Applied Psychology 80: 671684. Koehnken G, Milne R, Memon A, Bull R. in press. A meta-analysis on the effects of the Cognitive Interview. Psychology, Crime, and Law (special issue) Loftus EF, Ketcham K. 1991. Witness for the Defense: The Accused, the Eyewitness and the Expert who Puts the Memory on Trial. St Martins Press: New York. McCornack SA. 1992. Information manipulation theory. Communication Monographs 59(1): 116. McCornack SA, Levine TR, Solowczuk KA, Torres HI. 1992. When the alteration of information is viewed as deception: An empirical test of information manipulation theory. Communication Monographs 59(1): 1729. McCornack SA, Parks MR. 1985. Deception detection and relationship development: The other side of trust. In Communication Yearbook 9, McLaughlin ML (ed.). Sage: Beverly Hills, CA. Memon A. 1997. The cognitive interview. In A Handbook of Communication Skills, Hargie O (ed.). Routledge: London. Memon A, Koehnken G. 1992. Helping interviewees to remember more: The Cognitive Interview. Expert Evidence 1(2): 3948. Memon A, Wark L, Holley A, Bull R, Koehnken G. 1997. Eyewitness performance in Cognitive and Structured Interviews. Memory 5: 63955. Memon A, Vrij A, Bull R. 1998. Psychology and Law: Truthfulness, Accuracy and Credibility. McGraw-Hill: Maidenhead. Memon A, Stevenage SV. 1996. Interviewing witnesses: what works and what doesnt? Psychology 7(6): psyc.96.7.06.witness memory.1.memon Memon A, Higham P. in press. A review of the cognitive interview. Psychology, Crime, and Law. Osgood CE. 1960. Some effects of motivation on style of encoding. In Style in Language, Sebeok T (ed.). MIT Press: Cambridge, MA; 293306. Penrod SD, Cutler B. 1995. Witness condence and witness accuracy: Assessing their forensic relation. Psychology, Public Policy and Law 1: 817845.

Copyright # 2002 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 16: 287300 (2002)

300

K. Colwell et al.

Porter S, Yuille JC. 1995. Credibility assessment of criminal suspects through statement analysis. Psychology, Crime, and Law 1: 319331. Ruby CL, Brigham JC. 1997. The usefulness of the criteria based analysis technique in distinguishing between truthful and false allegations: A critical review. Psychology, Public Policy and Law 3: 705737. Sporer SL. 1997. The less travelled road to truth: verbal cues in deception detection in accounts of fabricated and self-experienced events. Applied Cognitive Psychology 11(5): 373397. Sporer SL, Malpass RS, Koehnken G. 1996. Psychological Issues in Eyewitness Identication. Lawrence Erlbaum: Mahwah, NJ. Trankell A. 1972. Reliability of Evidence. Beckmans: Stockholm. Tulving E. 1974. Cue-dependent forgetting. American Scientist 62: 7482. Tulving E, Thomson DM. 1973. Encoding specicity and retrieval processes in episodic memory. Psychological Review 80: 353370. Undeutsch U. 1967. Beurteilung der Glaubhaftigkeit von Aussagen (Evaluation of statement credibility. In Handbuch der Psychologie, Vol. 11: Forensische Psychologie, Undeutsch U (ed.). Hogrefe: Gttingen; 26181. o Undeutsch U. 1982. Statement reality analysis. In Reconstructing the Past: The Role of Psychologists in Criminal Trials, Trankell A (ed.). Kluwer: Deventer; 2756. Vrij A, Akehurst L. 1998. Verbal communication and credibility: Statement Validity Assessment. In Psychology and Law: Truthfulness, Accuracy and Credibility, Memon A, Vrij A, Bull R (eds). McGraw-Hill: Maidenhead. Wells G, Lindsay RCL, Ferguson TJ. 1979. Accuracy, condence, and juror perceptions in eyewitness identication. Journal of Applied Psychology 64: 440448. Yuille JC. 1988. The systematic assessment of childrens testimony. Canadian Psychology 29: 247262. Yuille JC, Cutshall JL. 1986. A case study of eyewitness memory of a crime. Journal of Applied Psychology, 71(2): 291301. Yuille JC, Cutshall J. 1989. Analyses of the statements of victims, witnesses and suspects. In Credibility Assessment, Yuille JC (ed.). Kluwer Academic Press: Dordrecht. Yuille JC, Farr VL. 1987. Statement validity analysis: A systematic approach to the assessment of childrens allegations of sexual abuse. British Columbia Psychologist Fall: 1927. Yuille JC, Hunter R, Joffe R, Zaparniuk J. 1993. Interviewing children in sexual abuse cases. In Child Victims, Child Witnesses: Understanding and Improving Childrens Testimony, Goodman G, Bottoms B (eds). The Guilford Press: New York; 95115. Zaparniuk J, Yuille JC, Taylor S. 1995. Assessing the credibility of true and false statements. International Journal of Law and Psychiatry 18: 343352. Zuckerman M, Driver RE. 1985. Telling lies: verbal and nonverbal correlates of deception. In Multichannel Integration of Nonverbal Behavior, Seigman AW, Feldstein S (eds). Lawrence Erlbaum: Hillside, NJ.

Copyright # 2002 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 16: 287300 (2002)

You might also like