Police Officers' Lie Detection Accuracy: Interrogating Freely Versus Observing Video

28
10.1177/1098611104264748 ARTICLE POLICE QUARTERLY (Vol. 7, No. 4, December 2004) Hartwig et al. / OFFICERS’LIE DETECTION ACCURACY POLICE OFFICERS’LIE DETECTION ACCURACY : INTERROGATING FREELY VERSUS OBSERVING VIDEO MARIA HARTWIG PÄR ANDERS GRANHAG LEIF A. STRÖMWALL Göteborg University ALDERT VRIJ University of Portsmouth The study investigated experienced police officers’ (N = 30) lie detection accuracy. Each police officer conducted an interrogation of a college student acting as a suspect either guilty or innocent of a mock crime and made a veracity judgment of the suspect. The police officers had the opportunity to conduct the interrogation in the manner of their own choice. The lie detection accuracy of these police officers was compared to that of police officers judg- ing videotaped versions of the interrogations. The police officers failed to detect deception better than chance. There was no difference in accuracy between police officers interrogating live and observing video. The interro- gators reported to rely on verbal content more when interrogating than when watching video. It seems as though police officers have a difficult time detect- ing deception not only in passive contexts but also in active ones. Keywords: deception detection; police officers; interrogation techniques Thanks are due to Karl Ask, Lina Bengtsson, Niklas Fransson, Fredrik Good, Anneli Larsson, Lina Leander, Malin Olsson, Viktor Prytz, and Johan Tengström for help with the data collection; to Detective Superintendent Royne Nilsson for giving us the opportunity to conduct this research; and to Dr. James Ost for helpful comments on an earlier version of this article. This research was funded by a grant from the Swedish Research Council given to the second author. Please send all correspondence to the first author at Department of Psychology, Göteborg University, P.O. Box 500, SE 405 30 Göteborg Sweden; e-mail: [email protected]. POLICE QUARTERLY Vol. 7 No. 4, December 2004 429–456 DOI: 10.1177/1098611104264748 © 2004 Sage Publications

Transcript of Police Officers' Lie Detection Accuracy: Interrogating Freely Versus Observing Video

10.1177/1098611104264748ARTICLEPOLICE QUARTERLY (Vol. 7, No. 4, December 2004)Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY

POLICE OFFICERS’ LIE DETECTION

ACCURACY: INTERROGATING FREELY

VERSUS OBSERVING VIDEO

MARIA HARTWIG

PÄR ANDERS GRANHAG

LEIF A. STRÖMWALLGöteborg University

ALDERT VRIJUniversity of Portsmouth

The study investigated experienced police officers’ (N = 30) lie detectionaccuracy. Each police officer conducted an interrogation of a college studentacting as a suspect either guilty or innocent of a mock crime and made averacity judgment of the suspect. The police officers had the opportunity toconduct the interrogation in the manner of their own choice. The lie detectionaccuracy of these police officers was compared to that of police officers judg-ing videotaped versions of the interrogations. The police officers failed todetect deception better than chance. There was no difference in accuracybetween police officers interrogating live and observing video. The interro-gators reported to rely on verbal content more when interrogating than whenwatching video. It seems as though police officers have a difficult time detect-ing deception not only in passive contexts but also in active ones.

Keywords: deception detection; police officers; interrogation techniques

Thanks are due to Karl Ask, Lina Bengtsson, Niklas Fransson, Fredrik Good, Anneli Larsson, LinaLeander, Malin Olsson, Viktor Prytz, and Johan Tengström for help with the data collection; to DetectiveSuperintendent Royne Nilsson for giving us the opportunity to conduct this research; and to Dr. JamesOst for helpful comments on an earlier version of this article. This research was funded by a grant fromthe Swedish Research Council given to the second author. Please send all correspondence to the firstauthor at Department of Psychology, Göteborg University, P.O. Box 500, SE 405 30 Göteborg Sweden;e-mail: [email protected].

POLICE QUARTERLY Vol. 7 No. 4, December 2004 429–456DOI: 10.1177/1098611104264748© 2004 Sage Publications

HUMAN LIE DETECTION ABILITY

Thirty years of research on deception detection ability has shown thatpeople generally are not very good at distinguishing between truthful anddeceptive statements. In most studies, accuracy rates fall below 60% (Kraut,1980; Vrij, 2000). This performance level is not impressive because anaccuracy rate of 50% is expected by chance alone. One commonly proposedexplanation is that there is lack of overlap between the cues that research hasshown to be associated with deception (objective cues to deception) and thecues that people believe to be associated with deception (subjective cues todeception) (Vrij, 2000).

Research has identified only a few objective cues to deception. Forexample, liars tend to have a higher-pitched voice (DePaulo et al., 2003),make longer pauses, fewer leg/foot and hand/arm movements than truthtellers (Sporer & Schwandt, 2002; Vrij, 2000). Research on subjective indi-cators of deception has shown that people associate deception with morespeech disturbances, a slower speech rate, longer and more frequent pauses,more gaze aversion, and increases in smiling and body movements (Vrij,2000).

Taken together, the results from research on objective and subjectivecues to deception thus indicate that there is a mismatch between cues actu-ally associated with deception and cues people associate with deception.This mismatch may partly account for people’s poor lie detectionperformance.

PROFESSIONAL LIE CATCHERS

Common sense suggests that professional lie catchers such as police offi-cers, police detectives, and customs officers ought to outperform lay peoplein terms of deception detection skills due to their more extended experienceof evaluating the veracity of others. A number of studies on presumed lieexperts’ ability to detect deception suggest that this common sense idea isincorrect (Vrij, 2000). Studies examining the lie detection ability of policeofficers (e.g., Ekman & O’Sullivan, 1991; Ekman, O’Sullivan, & Frank,1999; Köhnken, 1987; Vrij, 1993; Vrij & Graham, 1997; Vrij & Mann,2001a, 2001b) have found accuracy rates falling in the range of 45% to60%: in other words, similar to accuracy rates observed for lay people.However, in most of these studies, those lying are college students who par-ticipate in experimental laboratory research. Hence, liars attempt to deceive

430 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

mostly for the sake of the experiment, and, consequently, the stakes are notvery high. Using results from studies on police officers’ ability to detectlow-stake lies to draw conclusions about their ability to detect high-stakelies may be premature (Miller & Stiff, 1993).

Vrij and Mann (2001a, 2001b) addressed the shortcomings of previousstudies by letting police officers make veracity judgments of authentic tar-get materials. In the first study (Vrij & Mann, 2001a), police officers madeveracity judgments of video clips from an interrogation with a man who wassuspected and later convicted of murder. In this study, a mean accuracy rateslightly higher than usual was observed (64%). In the second study (Vrij &Mann, 2001b), police officers watched videotaped conferences of peoplewho asked the general public for help in finding out where their missing rel-atives were. Police investigations later revealed that all these people them-selves had murdered their relatives. In this study, the mean accuracy ratewas identical to chance level. So far, it thus seems unclear as to how the useof realistic, real-life target materials affects police officers’ lie detectionability.

It is important to make a distinction between realistic lies and realisticcontexts. In most studies on police officers’ lie detection accuracy, the liecatchers were put in a context not immediately applicable to real-life policeinterrogation situations. In real-life interrogations, the police officers cancontrol the number and type of questions being asked, ask follow-up ques-tions, and demand elaborations and clarifications. In other words, policeofficers conducting interrogations can employ different interrogation anddeception detection strategies, whereas watching a videotaped interroga-tion excludes these opportunities and leaves the lie catcher passive in thisrespect. The present study examined for the first time ever police officers’ability to detect deceit while given the opportunity to interrogate the“suspect.”

CONVERSATIONAL INVOLVEMENT

A few studies have examined differences in deception detection accuracybetween interactive and noninteractive contexts. The rationale behindempirically exploring the effect of conversational involvement on lie detec-tion ability is that research, in order to be ecologically valid, needs toacknowledge characteristics of real-life interactions. Indeed, real-lifedeceptive communications rarely take the passive form of most deceptiondetection experiments in which lie catchers are exposed to videotaped

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 431

interrogations instead of having the opportunity to conduct the interroga-tions. The general finding resulting from these studies is that less involvedlie catchers (observers) are at least as accurate and, in some cases, evenmore accurate in detecting lies than highly involved lie catchers (interroga-tors) (Buller, Strzyzewski, & Hunsaker, 1991; Burgoon, Buller, White,Afifi, & Buslig, 1999; Feeley & deTurck, 1997; Granhag & Strömwall,2001; Hartwig, Granhag, Strömwall, & Vrij, 2002). Moreover, interroga-tors have a more pronounced truth bias than observers, meaning that theytend to judge messages as truthful more often than deceptive. Furthermore,interrogators and observers attend to different types of information whenevaluating the target. Interrogators rely more on nonverbal behavior,whereas observers attend to verbal behavior (Buller et al., 1991; Feeley &deTurck, 1997; Gilbert & Krull, 1988; Granhag & Strömwall, 2001; Street,Mulac, & Wiemann, 1988). However, it should be noted that all these stud-ies have employed either completely or partially predefined interviewscripts. It is plausible to suggest that in a situation in which interrogators arefree to conduct the interrogation in the manner of their own choice, they toowill rely on verbal information when detecting deceit. When the interroga-tors choose the questions being posed themselves and when they can askfollow-up questions and demand more information on unclear issues, theverbal content of the answers should be of high interest to them.

Several complementary explanations have been proposed for the differ-ences in accuracy between interrogators and observers. One explanation isbased on the notion of the honesty effect; an assumption of truthfulness is apart of general conversation maxims. Hence, judges who are actively takingpart in a communication evaluate the conversation partner with greaterleniency than do passive observers, which will lead to a more pronouncedtruth bias (Granhag & Strömwall, 2001). However, it is plausible that thefinding that conversation partners exhibit a more pronounced truth bias willhold for lay people but not for police officers. It is rather improbable thatpolice officers will fall prey for general conversation maxims when interro-gating a suspect; they presumably encounter a different base rate of liarsand thus have a different frame of mind when facing the task of assessingveracity (Kassin & Fong, 1999). Indeed, studies on the deception detectionaccuracy of police officers have shown that their tendency to exhibit a truthbias is less pronounced and sometimes even lacking (Meissner & Kassin,2002; Porter, Woodworth, & Birt, 2000; Vrij, 1993; Vrij & Graham, 1997;Vrij & Mann, 2001a, 2001b).

432 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

It is important to note that the majority of studies investigating the effectof conversational involvement on deception detection accuracy have beenconducted from a non-legal perspective. Hence, generalizing results fromthese studies to a legal setting may be inappropriate.

Thus, to conduct a study of the effect of conversational involvement onlie detection ability with more relevance for legal contexts, we set experi-enced criminal investigators free to conduct an interrogation without prede-fined scripts and compared their accuracy to that of criminal investigatorswho made veracity judgments of the videotaped versions of the very sameinterrogations.

POLICE OFFICERS’ INTERROGATION TECHNIQUES

The police officers in the present study were all confronted with the sametask and received the same background information. This gives rise to thepossibility of making meaningful comparisons between the interrogationtechniques and strategies employed by the police officers.

As far as we know, this study is the first to explore police officers’interro-gation strategies under controlled conditions. Not only is experimental datalacking on the issue of questioning of suspects, but also relatively little isknown about how police interrogations are conducted in real-life cases(Milne & Bull, 1999). Research has mainly been carried out in the form offield studies (e.g., Baldwin, 1993; Moston & Engelberg, 1993; Pearse &Gudjonsson, 1999) or as case studies (Moston & Stephenson, 1993). One ofthe important findings that have emerged from real-life studies is that theprimary aim of the police seems to be to obtain a confession (Milne & Bull,1999; Mortimer & Shepherd, 1999; Vrij, 2003). For example, in one studyin which tape recordings from police interrogations were analyzed(Moston, Stephenson, & Williamson, 1992), it was found that the vastmajority of the interrogators spent most, if not all, time on accusing the sus-pect and asking for responses to those accusations instead of attempting toobtain an account of the event.

The findings indicating that police officers seem to focus on obtainingconfessions can be explained in many ways (for a discussion of this, seeVrij, 2003). One reason might be that a popular interrogation manual, Crim-inal Interrogation and Confessions (Inbau, Reid, & Buckley, 19861),describes an approach to bringing about confessions (the “nine-stepapproach”). These steps aim to create an environment that promotes confes-sion and suppresses any attempts of denial, for example, by highlighting the

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 433

positive consequences of confession while emphasizing the futility ofdenial. Several scholars have severely criticized the techniques proposed byInbau and colleagues, partly on the grounds that they may lead to false con-fessions (Gudjonsson, 2002; Kassin, 1997). Another problem with the tech-niques proposed by Inbau and colleagues is that they recommend relying onnonverbal cues to detect deception: for example, gaze aversion, posturalshifts, and placing hand over mouth. Critically, these particular cues are notidentified as reliable ones in the deception literature (DePaulo et al., 2003).Although there is a wide range of police interrogation manuals, the onewritten by Inbau and colleagues is doubtlessly the most influential(Gudjonsson, 2002; Vrij, 2003).

Scholars have proposed alternative ways of conducting interrogationsthan those provided in most police interrogation manuals (e.g., Milne &Bull, 1999; Sear & Williamson, 1999). In essence, it has been recom-mended that the focus of the interrogation should shift from obtaining con-fessions to obtaining information in search for the truth. This approach hasbeen labeled, among other things, ethical, inquisitorial, and information-gathering interrogation, but these labels contain the same central aspects(Holmberg & Christianson, 2002). The aim of this interrogation style is tocollect as much information as possible, to strive toward establishing a posi-tive atmosphere, and to keep an open mind throughout the interrogation(Gudjonsson, 2002; Moston & Stephenson, 1993).

Thus, there are suggestions both from professionals and scholars on howto conduct interrogations. Moreover, the proposed information-gatheringstyle may be difficult for police officers to translate into a practical interro-gation situation as it is not defined by any clear-cut guidelines. Takentogether and translated into the context of the present study, it would beplausible to suggest that the police officers under study will employ differ-ent interrogation strategies and tactics because there apparently are no clearand thoroughly tested instructions available to them on how to interrogatesuspects.

Research on real-life data has generated a number of other importantfindings on the characteristics of police interrogations; most of these find-ings have been generated through qualitative analyses of the interrogations.These findings indicate that the ways in which police interrogations areconducted suffer from limitations.

First, in a study by Baldwin (1992), it was found that only 64% of theinterrogations were judged to be conducted competently, which means thatmore than one third of the interrogations were deemed to have serious

434 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

limitations. These limitations were grouped under four headings: ineptitude(e.g., poor preparation, unfamiliarity with the evidence, limited socialskills), assuming guilt (e.g., meeting suspect with skepticism, posing lead-ing and repetitive questions), poor interview technique (e.g., interruptingthe suspect, preventing suspect from giving a free recall, reacting to aggres-sion provocation), and questionable or unprofessional conduct (e.g., adopt-ing an aggressive approach, using threats or inducements to obtain confes-sions). In addition, it has been found that interrogations on average are quitebrief but that there is a wide range in terms of duration. Williamson (1990)found that nearly half of the 1,000 interrogations analyzed were completedin 15 minutes. Results from a study by Baldwin (1993) were similar; 75% ofthe 600 interrogations analyzed were completed within half an hour, but theinterrogations varied in length between 1 minute and 7 hours. A study ana-lyzing 161 interrogations found that the mean duration was 22 minutes,with a range of 2 to 109 minutes (Pearse & Gudjonsson, 1996). It is indeeddoubtful whether such short interrogations make possible detailed and clearaccounts from the suspects.

One of the ambitions of this study was to explore the qualities of theinterrogations conducted by the police officers in our sample, and we thusposed the following research question: What are the characteristics of theinterrogation strategies of these experienced police officers, and what arethe differences and similarities between the interrogations?

HYPOTHESES

Hypothesis 1. Because the police officers had the opportunity to conduct the interroga-tion in the manner of their own choice, we predicted that both interrogators and ob-servers would achieve an accuracy rate significantly higher than chance. We also pre-dicted that the police officers would not show the truth bias found in previous studies,due to their more suspicious frame of mind (Burgoon, Buller, Ebesu, & Rockwell,1994; Kassin & Fong, 1999; Meissner & Kassin, 2002).

Hypothesis 2. We predicted that police officers watching video would perform signifi-cantly better than interrogators in terms of lie detection accuracy (Buller et al., 1991;Burgoon et al., 1999; Feeley & deTurck, 1997; Granhag & Strömwall, 2001).

Hypothesis 3. In line with previous research (e.g., Buller et al., 1991; Feeley & deTurck,1997; Gilbert & Krull, 1988; Granhag & Strömwall, 2001; Street et al., 1988), we pre-dicted that observers would report more use of verbal content than nonverbal behaviorwhen assessing veracity. Due to the fact that the interrogations were nonscripted, weadditionally predicted that interrogators too would rely on verbal content and thatthey would do so even more than observers, as the verbal content of the answers givenby the suspect probably is of particular interest for the interrogators because they con-

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 435

trol which questions are being posed. In addition, Vrij (2000) has pointed out thatwhen there is case-relevant information to verify (which is the case in the presentstudy), people have a tendency to rely more on verbal content than nonverbalbehavior.

Hypothesis 4. We expected to find that the interrogations would differ from each otherdue to the lack of clear-cut guidelines on how to interrogate suspects (Baldwin, 1992;Soukara, Bull, & Vrij, 2002).

METHOD

OVERVIEW

The experiment was conducted in two phases. In the first phase, collegestudents were instructed to either commit a mock crime or a noncriminalact. In the second phase, they were interrogated by police officers, and theseinterrogations were videotaped. Each police officer made two veracityjudgments: one after conducting an interrogation and one after seeing a vid-eotaped interrogation conducted by a fellow officer. Interrogators andobservers received identical background information.

PHASE 1

Participants. The student group was composed of 30 undergraduate psy-chology students (21 females and 9 men; age ranging from 18 to 53 years,M = 24.0) from Göteborg University. The students were informed that theywere going to be interrogated by police officers about an event and to eitherlie or tell the truth about that event. All students were guaranteed to receive amonetary compensation of 300 SEK (about $30) but could receive an addi-tional 200 SEK (about $20) if they managed to convince both the policeofficer whom they were interrogated by and the police officer judging thevideotaped version of their interrogation that they were telling the truthabout the event they had experienced.

Procedure. Students were instructed to come to the Department of Psy-chology in the morning on the day of the experiment. On arrival, they wererandomly divided into four groups; event 1 truth tellers, event 1 liars, event2 truth tellers, and event 2 liars (henceforth referred to as E1-T, E1-L, E2-Tand E2-L, respectively). The rationale behind using two events was that allpolice officers would make two judgments, and letting them judge theveracity of suspects talking about two different events would prevent com-

436 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

parisons between the stories. The participants in the E1-L and E2-L groupswere asked whether they had any objections against committing a mockcrime and lying about it during the interrogations and, if so, they couldchange groups, but no one did. Each group was placed in a room, where anexperimenter gave the participants further instructions. One experimenterwas responsible for each group and stayed in the room during the wholetime to ensure that no discussions of the event took place.

The participants in the E1-L group were told that they were going to buya package of drugs (a sealed envelope in fact containing flour) from a man (aconfederate) at a hot dog stand situated next to a newsstand. They wereinstructed to approach the seller and give him a sealed envelope with fakemoney in return for the drug package and return to the department as soon aspossible. Participants were told to show the envelope to the experimenterresponsible for their group to make sure that the exchange indeed had takenplace.

The participants in the E2-L group were told that they were going to sell apackage of drugs (a sealed envelope in fact containing candy) to a man (aconfederate) in an alley next to a baker’s shop. They were told to approachthe man and give him the package of drugs in exchange for an envelope withfake money and return to the department. They were told to show the enve-lope to the experimenter before entering the room and then to go back andsit in the room of their group without discussing the event.

The E1-T and E2-T participants received instructions to one at a timewalk to either the newsstand (E1-T, the same newsstand next to which par-ticipants in the E1-L group were instructed to buy drugs) or the baker’s shop(E2-T, the same baker’s shop next to which participants in the E2-L groupwere instructed to sell drugs), both a few minutes walk away from theDepartment of Psychology, without talking to anyone on the way. Theywere given a small sum of money to buy whatever they wished (a soft drink,a bun, etc.). The participants were asked, on return to the department, toshow the experimenter what they had bought and then to go back and sit inthe room of their group without discussing the event. The event took about 5to 10 minutes for each participant in all four groups.

After all participants had taken part in the events, they received furtherinstructions by their experimenter on the upcoming interrogations. Stu-dents in the E1-T and E2-T groups were instructed to simply tell the truthabout what had happened during the time they had been away from thedepartment. Students in the E1-L and E2-L groups were told to lie aboutwhat had happened during the time they had been away from the

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 437

department. They were semi-guided and told to distort the event in such away that the police officers were led to believe that they had simply visitedthe newsstand (E1-L) or the baker’s shop (E2-L). The students in these con-ditions were also instructed to keep the envelope containing either fakedrugs or fake money in their pocket during the interrogation. The reasonwhy the participants in the lying conditions were semi-guided was that inorder for the interrogation to be meaningful, we wanted the suspects’ storyto correspond with the information given to the police officers (see below).

All students were informed that the police officers by whom they wouldbe interrogated were aware of the fact that the students took part in an exper-iment and that no real crimes had been committed. The training programthat the police officers went through took place about 200 kilometers fromGothenburg. Students were taken there by bus, and during the bus trip thatlasted approximately 2 hours, they were instructed to prepare their story.

PHASE 2

Design. The experiment had a 2 (role: interrogator vs. observer) × 2(event: E1 vs. E2) × 2 (veracity: truthful vs. deceptive) mixed factorialdesign, with role as the within-subjects factor. All factors werecounterbalanced.

Participants. The group of police officers was composed of 30 criminalinvestigators from different parts of Sweden (11 females and 19 males; ageranging from 29 to 58 years, M = 44.6; years of experience as police officersM = 21.7 and years of experience working as a criminal investigator M =11.1). The police officers took part in a 2-week training program aboutinvestigative interviewing arranged by the Swedish national police and tookpart in the experiment as an exercise included in that program. Police offi-cers may enter the course after working for at least 5 years as a criminalinvestigator. Moreover, to be accepted to the course they need to pass a testthat focuses on their theoretical and judicial understanding of the interroga-tion procedure. In other words, the police officers who participated in thisstudy were all experienced and qualified criminal investigators.

Procedure. The police officers were randomly divided into two groups:one in which the police officers conducted the interrogations first (IF) andone in which the police officers judged the videotaped interrogations first

438 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

(VF). These two groups were seated in different rooms, where they weregiven further instructions. The police officers were informed that theywould perform two tasks: One was to conduct a short interrogation with astudent either guilty or innocent of taking part in a drug exchange (it wasstressed in the instructions that it was equally possible that the student waslying as telling the truth) and afterwards make a veracity judgment of thisstudent’s statement. The other was to watch a short, videotaped interroga-tion conducted by one of their colleagues with a student either guilty orinnocent of taking part in a drug exchange and afterwards make a veracityjudgment of this student’s statement. The order in which they would takepart in these tasks depended on whether they were part of the IF or VFgroup.

Each police officer made veracity judgments of one suspect who wasaccused of buying drugs at the hot dog stand (E1) and one who was accusedof selling drugs at the baker’s shop (E2). The police officers received twosets of instructions, one before conducting the interrogation and one beforejudging the videotaped interrogation. The first set of instructions was to beread on arrival in the waiting room and the other on returning from the firsttask. Each set of instructions contained written background information ofthe event in which the suspect to be judged had taken part and writteninstructions about the task. The police officers were also given a map of theplace where the crime had occurred. The police officers were free to con-duct the interrogation in whatever manner they considered appropriate.However, for practical reasons, they were instructed to interrogate for nomore than 10 minutes; after 10 minutes had passed, they would be remindedby a knock on the door that time was running out. If they had not ended theinterrogation one minute after the knock on the door, the experimenterwould enter the room and interrupt the interrogation (this happened in aboutone third of the interrogations). The interrogations took place in smallrooms with two chairs. The chair intended for the suspect was visible in thecamera, whereas the other was placed so that the interrogator would not beseen on tape.

The same written background information and map were given to thepolice officers before judging the videotaped interrogation. The instructionto the judgment of the videotaped interrogation was simply to watch and lis-ten to the suspect being interrogated and try to decide whether he or she wastelling the truth. The judgments of the videotaped materials took place insmall rooms with a TV screen and a chair.

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 439

As mentioned above, each police officer made veracity judgments of twosuspects, and it was stressed in the instructions that the two suspects thepolice officers were to judge had not taken part in the same event; hence,comparing stories would be meaningless.

POSTTEST QUESTIONNAIRES

Students. After being interrogated by a police officer, the studentsreturned to a waiting room and were asked to fill out a questionnaire. It con-tained questions about background characteristics (age and sex) and ratingsof their own truthfulness in their story on a 10-point scale from 1 (every-thing deceptive) to 10 (everything truthful).

Police officers. On completion of a task, the police officers returned tothe waiting room and filled out a questionnaire. It contained questions aboutage, sex, number of years of experience working as a police officer, andnumber of years of experience working as a criminal investigator. Theywere asked to rate on a 6-point scale whether the student lied, ranging from1 (no, definitely not) to 6 (yes, definitely). Moreover, the questionnaire con-tained a dichotomous truth/lie judgment, and the police officers were askedto indicate how confident they were in the dichotomous veracity judgmentthey had just given on a scale ranging from 50% to 100%. It was explainedthat 50% meant that they were completely unsure because they had tochoose between two answer alternatives, and 100% meant that they wereabsolutely convinced that their veracity judgment was correct. They werealso asked to rate at which point in time they decided on veracity on an 11-point scale, in which 0 = right in the beginning and 10 = after the interroga-tion. Furthermore, the police officers were asked to write down argumentsjustifying their veracity judgment (they were free to report as many argu-ments as they wished). Finally, they were asked to rate which type of infor-mation they used when making the veracity judgment on a 10-point scalefrom 1 (only nonverbal behavior) to 10 (only verbal content).

Categories of cues. For our analyses of the self-reported arguments of theveracity judgments, we used both verbal and nonverbal categories of cues,in line with Granhag and Strömwall (2000). Following are the seven catego-ries of verbal cues and a typical example of each (all the categories were bi-directional): completeness (“The statement was complete”), confidence(“The witness was confident in his statement”), consistency (“The state-

440 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

ment was consistent over time”), details (“The statement was rich indetail”), plausibility (“The statement seemed plausible”), rehearsed story(“The statement seemed rehearsed”), and statement in general (“The state-ment was clear”). In addition, we used the following four categories of non-verbal cues: body movements (“The witness made few trunk movements”),credibility in general (“The witness’s behavior was trustworthy”), gaze(“The witness looked away”), and nervousness (“The witness wasnervous”).

Interrater reliability. One individual coded all 60 verbalizations, and30% of the material (18 verbalizations) was also coded by a second coder.The interrater agreement for these codings was 79.0%. That is, 79.0% of thecues were coded identically between the two coders. The 10 discrepancieswere resolved in a discussion between the two coders.

INTERROGATION CHARACTERISTICS

A number of different characteristics were coded: (a) the number ofwords spoken by the student and the police officer; (b) the presence of freerecall; (c) the types of questions being asked, divided into the number ofopen-ended and closed-ended questions; (d) the number of times the inter-rogator interrupted the suspect; and (e) the number of questions that wererepeated as well as the total number of times such a repeated question wasposed. We chose to code these characteristics because we considered themto be a reflection of the strategy employed by the interrogator. For example,we consider many interruptions a sign of a dominant attitude toward thesuspect and many questions (in particular closed-ended and repeated) to beindicative of pressuring the suspect (Baldwin, 1993). To obtain a measureof the degree to which the police officers dominated the interrogations, wecalculated a “dominance index” (DI) by dividing the number of words spo-ken by the police officer with the number of words spoken by the student foreach interrogation. A DI smaller than 1 would thus indicate that the suspectspoke more than the interrogator, whereas a DI larger than 1 would indicatethat the interrogator spoke more than the suspect.

Due to technical issues, 6 of the 30 videotapes were of too poor soundquality and could thus not be coded. One individual coded all 24 interroga-tions (either the transcribed versions or the videotapes depending on whattype of characteristic was coded), and 20% was also coded by a second indi-vidual (although the number of words spoken by students and police

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 441

officers was counted using a computer and was hence not coded by thesecond coder).

Interrater reliabilities. For the types of questions being asked, theinterrater agreement was 76.8% for open-ended questions, 78.5% forclosed-ended questions, and 100% for presence of free recall. For interrup-tions, the interrater agreement was 71.4%. For the repeated questions, theanalyses of interrater agreement was divided into three stages; in the firststage, the frequency of questions that were repeated was scored by the twocoders; the correlation was r(5) = .98, p < .01. Second, for the repeatedquestions identified in stage 1, 67.0% was identical for the two coders.Third, for the questions that were identified as repeated by both coders inthe second stage, the number of times these questions were repeated wasscored by the two coders and the correlation was r(8) = .94, p < .001.

RESULTS

PRELIMINARY ANALYSES

Manipulation check. A t test was conducted to find out whether studentscomplied with the instructions to lie or tell the truth. Truth tellers rated theirtruthfulness as significantly higher than did liars (M = 9.33, M = 3.00,respectively; t[28] = –11.36, p < .001).

Order and event effects. A test for significance of difference betweenproportions revealed that there was no difference in terms of deceptiondetection accuracy between E1 and E2 (z = –1.54, p = .12). A test for signifi-cance of difference between proportions showed no difference in accuracylevels obtained by participants who interrogated first compared to thosewho first judged the videotaped statement (z = 0.12, p = .90). In other words,there were no order and event effects; hence, order and event conditionswere collapsed, and further analyses concerned only the role and veracityfactors.

VERACITY JUDGMENTS

Accuracy. Calculated from the dichotomous truth/lie judgment, bothinterrogators and observers obtained an accuracy level of 56.7%. Hence,

442 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

Hypothesis 2, stating that video observers would outperform interrogators,did not receive support. A binomial test showed that neither interrogatorsnor observers obtained an accuracy level higher than chance (both p’s =0.58). This means that the part of Hypothesis 1 stating that bothinterrogators and observers would perform better than chance failed toreceive support. To make possible use of the 6-point scale deception ratingas a dependent measure, the scale was recoded so that 6 = totally accurateand 1 = totally inaccurate. A paired samples t test showed no difference inaccuracy between interrogators and observers (M = 3.63 and M = 3.93,respectively; t[29] = –0.92, p = .37).

Truth/lie bias. Inspection of the distribution of truth/lie judgments madeby interrogators and observers reveals that interrogators made 53.3% truthjudgments and thus had a slight truth bias. Observers instead made 46.7%truth judgments and thus had a lie bias. Neither of these biases differed fromchance (both binomial p’s = .86). A McNemar test revealed that the differ-ence in the distribution of truth/lie judgment between interrogators andobservers was not significant (binomial p = .41, one-tailed). Thus, the partof Hypothesis 1 stating that the police officers would not exhibit a truth biasreceived support.

Point of time for veracity judgment. A paired samples t test showed thatinterrogators decided on veracity later than did observers (M = 7.68 and M =6.83, respectively; t[29] = 2.38, p < .05). Furthermore, observers whojudged the suspects to be lying reported deciding on veracity earlier thandid observers who judged the suspects to be telling the truth (M = 5.07 andM = 7.89, respectively; t[28] = –3.14, p < .01). Interrogators who judged thesuspects to be lying did not report deciding earlier than did interrogatorswho judged the suspects to be telling the truth (M = 7.58 and M = 7.77,respectively; t[28] = –0.22, p = .84).

CUES TO DECEPTION

Self-reported use of verbal and nonverbal information. As mentionedabove, the police officers were asked to rate on a 10-point scale which typeof information they used when assessing veracity (1 = only nonverbalbehavior and 10 = only verbal content). Both interrogators and observerswere on the higher end of the scale, indicating that they preferred use of ver-bal content (M = 6.35 and M = 5.23, respectively). A paired samples t test

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 443

revealed that police officers reported more use of verbal content when inter-rogating than when acting as observers; (t[29] = 2.35, p < .05). Thus,Hypothesis 3, stating that both interrogators and observers would rely moreon verbal content than nonverbal behavior, and that this tendency would beeven more pronounced for interrogators, received support.

Interrogators versus observers. Interrogators on average mentioned 2.13cues per person, and observers mentioned 1.70 cues per person. A pairedsamples t test revealed no significant difference in the number of cues givenfrom interrogators and observers (t[29] = 1.27, p = .21). Tests for signifi-cance of difference between proportions were conducted separately forinterrogators and observers to investigate whether more verbal than nonver-bal cues were mentioned. The tests revealed that neither interrogators (z =0.34, p = .73) nor observers (z = –.86, p = .39) mentioned more verbal thannonverbal cues. In addition, to see whether interrogators mentioned moreverbal cues than observers, a test for significance of difference between pro-portions was conducted. Interrogators mentioned 46.8% nonverbal cues,and observers reported 42.1% nonverbal cues. The test revealed no differ-ence in the proportion of nonverbal cues given by interrogators and observ-ers (z = 0.52, p = .60).

Distribution of frequencies of cues. Overall, many different types of cueswere reported to be used; hence, the number of cues in each category wasquite small. For interrogators, the most frequently reported cue for justify-ing the veracity judgments was credibility in general (24.2%). For observ-ers, on the other hand, the most frequently reported cue was body move-ments (19.3%). Percentages for all categories of cues are found in Table 1.

POLICE OFFICERS’ INTERROGATIONS

Number of words. The number of words spoken by the police officers inthe interrogation ranged from 270 to 952, with a mean of 565.2 words. Thenumber of words spoken by the student, on the other hand, ranged from 206to 1,256, with a mean of 617.8 words.

Number of open- and closed-ended questions. First, we investigatedwhether the police officers invited the student to give a free recall of theevent. It was found that 21 police officers did so, whereas 3 refrained fromasking for a free recall. Then, the number of questions were counted and

444 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

divided into closed-ended (i.e., “Did you buy a newspaper at the news-stand?”) and open-ended questions (i.e., “What did the person look like?”).The total number of questions ranged from 17 to 69, with an average of 38.4questions. The number of closed-ended questions ranged from 9 to 41 (M =23.0), and the number of open-ended questions ranged from 8 to 29 (M =15.4). A paired samples t test revealed that the police officers posedsignificantly more closed-ended than open-ended questions (t[23] = –4.31,p < .001).

Number of repeated questions. Two aspects were coded to investigate thenumber of repeated questions. First, for each interrogation, we coded thenumber of questions that were repeated; second, we coded the total numberof times such a repeated question was posed. The number of questions thatwere repeated ranged from 0 to 6, with an average of 2.50. The number oftimes such repeated questions were posed ranged from 0 to 13, with anaverage of 4.0.

Number of interruptions. The number of times the police officers inter-rupted the students while he or she was speaking was counted; the numberof times the student was interrupted ranged from 0 to 11, with a mean of 2.3interruptions.

Dominance Index. For each interrogation, the ratio between number ofwords spoken by the police officer and the number of words spoken by the

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 445

TABLE 1. Percentages for All Categories of Cues Used to Justify the Veracity Judgments

Cue Category Interrogators Observers

Verbal cuesCompleteness 1.6 3.5Confidence 8.1 3.5Consistency 4.8 5.3Details 9.7 1.8Plausibility 9.7 12.3Rehearsed story 6.5 3.5Statement, general 6.5 12.3

Nonverbal cuesBody movements 11.3 19.3Credibility, general 24.2 15.8Gaze 8.1 8.8Nervousness 9.7 14.0

student was calculated. The DI ranged between 0.33 and 2.40, with a meanof 1.05. For 12 of the interrogations, the DI was smaller than 1, meaning thatthe student spoke more than the police officer. In the other 12 interroga-tions, the DI exceeded 12, indicating that the police officer spoke more thanthe student. In Table 2, the mean and range of each interrogation character-istic are presented. As can be seen in the table, the range was large for eachinterrogation characteristic.

Relation between interrogation characteristics and lie detection accu-racy. To examine whether the interrogation characteristics had any predic-tive power on the accuracy of the police officers, a multiple regression wascarried out. We chose seven of the interrogations characteristics as predic-tors (number of words spoken by the student, number of words spoken bythe police officer, proportion of closed-ended questions, total number ofquestions, number of interruptions, total number of times a repeated ques-tion was posed, and the DI) because we considered those measures to mirrorcentral aspects of the interrogations. As a dependent measure, we used therecoded 6-point veracity degree scale. None of the predictors turned out beto be significant (all p’s > .05, adjusted R2= .01).

The effects of experience on interrogation characteristics. To investigatewhether the experience of working as a criminal investigator was related tointerrogation characteristics, we correlated the self-reported number ofyears of working as a criminal investigator with the eight interrogationcharacteristics correlated with accuracy above. One significant correlationbetween years of experience and number of open-ended questions (r(24) =

446 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

TABLE 2. Means, Minimum, and Maximum Values for the Interrogation Characteristics

Interrogation Characteristic Mean Minimum Maximum

Words—police officer 565.2 270.0 952.0Words—student 617.8 206.0 1265.0Closed-ended questions 23.0 9.0 41.0Open-ended questions 15.4 8.0 29.0Number of repeated questions 2.5 0.0 6.0Posing repeated questions 4.0 0.0 13.0Interruptions 2.3 0.0 11.0Dominance index 1.1 0.3 2.4

–.43, p < .05) emerged, indicating that less experienced criminal investiga-tors posed more open-ended questions than did more experienced ones.

DISCUSSION

The rationale behind this study was to investigate the lie detection accu-racy of police officers when given the opportunity to conduct an interroga-tion in the manner of their own choice.

The study revealed three major findings. First, in terms of lie detectionaccuracy, police officers failed to perform better than chance. Second, inter-rogators and observers were similar, both in terms of the obtained lie detec-tion accuracy and self-reported perception of the task (such as reports ofcues to deception). Third, the interrogations were very different, and therewas no systematic relation between the characteristics analyzed and liedetection accuracy. We will discuss each of these findings in closer detailbelow.

POLICE OFFICERS’ LIE DETECTION ABILITY AND DIFFERENCES

BETWEEN INTERROGATORS AND OBSERVERS

Accuracy. Contrary to our prediction, neither interrogators nor observersperformed better than chance. This finding is in line with results from stud-ies on lie detection accuracy of police officers using a noninteractive design.Thus, it seems as though the police officers did not benefit from having theopportunity to conduct the interrogation in the manner of their own choice.

One explanation commonly proposed for poor accuracy levels is lack ofmotivation to succeed in the task of detecting lies (Vrij, 2002). In this case,we do not consider this explanation to be plausible. It is quite likely that thepolice officers perceived the task as partially measuring a professional skillof relevance for them; moreover, the police officers knew that their perfor-mances were to be discussed and evaluated later on during the training inwhich they participated. Another similar explanation could be that the stu-dents were not motivated to be judged as truthful by the police officers; wedo not consider this explanation plausible either. The students were given afairly large monetary reward for convincing the police officers that they toldthe truth; moreover, they were informed that the interrogators were experi-enced criminal investigators, and indeed, analyses reveal that both lying andtruth-telling students rated their nervousness and level of motivation as high(Strömwall, Hartwig, & Granhag, 2003).

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 447

A more reasonable explanation for the poor lie detection performance isthat the experience of the police officers has not aided them in improvingtheir lie detection ability. Scholars from different domains have stressed theimportance of outcome feedback in order for improvement on a task tooccur (Allwood & Granhag, 1999; DePaulo & Pfeifer, 1986; Einhorn,1982; Granhag, Andersson, Strömwall, & Hartwig, 2004; Hartwig,Granhag, Strömwall, & Andersson, 2004; Vrij & Semin, 1996). The feed-back provided needs to be clear, immediate, and reliable to be helpful, and ithas been pointed out that such feedback often is lacking in the environmentof police officers (DePaulo & Pfeifer, 1986). Hogarth (2001) distinguishedbetween environments characterized by “kind” and “wicked” learningstructures. He argued that kind learning structures are those that make itpossible to learn the appropriate lessons from experience, whereas wickedlearning structures can cause stereotypical and wrongful beliefs to becemented rather than changed through experience. One of the crucial fea-tures of this distinction is that kind learning structures contain immediate,reliable, and systematic feedback, whereas wicked learning structures donot. It would be fair to say that the environment of police officers (e.g., thesituation in the interrogation room) has more qualities of a wicked learningstructure, and this can partly account for the poor lie detection performanceof the police officers in this study.

Yet another possible explanation for the poor accuracy levels is that thestories of liars and truth tellers probably were highly similar. Liars did notlie throughout the interrogation; rather, they altered their stories to excludethe part about selling or buying drugs. Indeed, research indicates that inreal-life cases, liars do not lie throughout the interrogation (Mann, 2001).Thus, in this sense, our study was probably realistic, but it made lie detec-tion difficult for the police officers.

Neither interrogators nor observers exhibited a significant truth bias,which confirms the second part of our first hypothesis. This finding mirrorsthe notion of police officers’ more suspicious frame of mind discussed inprevious research (Kassin & Fong, 1999).

Interrogators versus observers. There was no difference in lie detectionaccuracy between interrogators and observers. Importantly, this findingcontrasts the view expressed by police officers, who in a study by Strömwalland Granhag (2003), showed a strong tendency to believe that it is easier tocorrectly distinguish between truthful and deceptive statements when inter-

448 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

rogating, compared to when watching the videotaped version of the sameinterrogation. Moreover, the findings from this study differ from most stud-ies on conversational involvement and deception detection, which havefound that observers outperform interrogators in terms of accuracy indetecting lies (Buller et al., 1991; Burgoon et al., 1999; Feeley & deTurck,1997). We propose two complementary explanations for this discrepancy.

First, students, who acted as interrogators and observers in previousstudies, probably are not used to conducting interrogations. Differences inlie detection accuracy between interrogators and observers may exist onlyfor inexperienced interrogators, who plausibly are hampered more by highconversational involvement than are police officers. Police officers conductinterrogations on a daily basis and thus may be more trained in the dual taskof upholding the interrogation while attempting to detect deceit. In otherwords, it may be that the cognitive burden of interrogating is less heavy onthose with experience of this task.

Second, the interrogators may have failed in their strategies and did notmanage to interrogate in a way that elicited behavioral differences betweenliars and truth tellers. If this is the case, then it is no wonder that observerswatching these interrogations did not manage to perform better thanchance.

We found that observers who judged the students as deceptive decidedearlier than did observers who judged the students to be truthful. Moreover,we found that interrogators decided on veracity later than did observers.Detecting deception can be seen as a signal detection task; judges trigger onsignals of deceit and make a lie judgment after the occurrence of such sig-nals. Such subjective lie signals can occur in any phase of the interrogation;thus, the lie judgment can be made in any phase of the interrogation. On thecontrary, signals of truthfulness do not give rise to an instant truth judg-ment; it is rather the lack of lie signals that leads to a truth judgment. Truthjudgments are passed when judges can conclude that no lie signals havebeen exhibited, and such a conclusion can indeed be reached only in the endof the interrogation or when it is finished. However, this pattern of self-ratedpoint of time for veracity judgment was found only for observers and not forinterrogators. Speculatively, this can be explained by the fact that interroga-tors reported relying more on verbal cues than did observers. If one attendsto the verbal aspect of a statement when seeking to detect deceit, one proba-bly waits to decide on veracity until the end of the statement, as contradic-tions and lack of consistency—or indeed explanations for such

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 449

contradictions—may well occur in the end of a statement. If one refrainsfrom making a veracity judgment until the end of the statement, lie judg-ments are no more readily passed than truth judgments, which would be thecase when relying on nonverbal cues, according to the signal detectionnotion. This also accounts for the second finding that interrogators gener-ally decided on veracity later than did observers. Because observers whomade lie judgments decided earlier than observers who made truth judg-ments and there was no such difference for interrogators, the average timeneeded to make a veracity decision collapsed over truth and lie judgmentsbecomes lower for observers than for interrogators.

Self-reported cues to deception. In line with our third prediction, bothinterrogators and observers reported relying more on verbal than nonverbalcues, and interrogators reported doing so significantly more than observers.However, the tendency to report relying on verbal cues was far from pro-nounced, and moreover, inspection of the cues to deception reported by thepolice officers in this study reveals that neither interrogators nor observersreported significantly more verbal than nonverbal cues.

On the contrary, in the study investigating professionals’ beliefs aboutcues to deception (Strömwall & Granhag, 2003), police officers expressedthe belief that nonverbal cues are more reliable than verbal ones. This find-ing is mirrored by results from a study on the lie detection accuracy ofpolice officers (Vrij, Edward, & Bull, 2001), indicating that police officersrely more on nonverbal than verbal behavior when detecting deceit. Thereason for the discrepancy between our results and those in the two studiesmentioned above may be the fact that the police officers in the current studyhad the opportunity to control the questions being posed and ask follow-upquestions, which may have made the verbal information seem more relevantto them. This would also explain why interrogators reported relying on ver-bal behavior more than did observers. The verbal content was probably ofinterest to the observers because the questions were posed by a colleaguewho also had some insight into the case, but the verbal content probablyseems even more relevant for the interrogators because it was they whodecided what questions to pose. Interestingly, research indicates that crimi-nals, who probably have good insight into the psychology of deception, alsobelieve that verbal behavior is more reliable than nonverbal when seeking todetect deceit (Granhag et al., in press). This belief matches findings from arecent meta-analysis of objective cues to deception (DePaulo et al., 2003).

450 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

INTERROGATION CHARACTERISTICS

In line with our fourth prediction, the interrogations differed from eachother in some important aspects. There was a wide variation in the numberof words spoken, the number of questions asked, the number of interrup-tions, and the number of repeated questions. In some aspects, however, theinterrogations were quite similar, as, for example, in the request for freerecall and the distribution of open- and closed-ended questions (the vastmajority of the police officers posed more closed-ended questions). On thewhole, it seems as though the interrogations were more different than simi-lar to each other concerning the characteristics we examined. We proposetwo complementary explanations for this finding. First, the diversity interms of interrogation characteristics may be a sign that there are no clearguidelines on which the police officers can rely when conducting interroga-tions. There is but a meager amount of interrogation manuals (and no Swed-ish ones) written for practical use that propose scientifically based interro-gation techniques. In other words, there is not much literature that has theambition and/or potential to replace traditional interrogation manuals. Toimprove the interrogation skills of the police officers, such scientificallybased, practical interrogation manuals need to be developed. The secondexplanation is that behavioral differences on behalf of the suspect maypartly be responsible for the diversity in the interrogation characteristics.For example, a less talkative suspect may lead the interrogator to ask morequestions, both open- and closed-ended, and a more cooperative suspectmay lead to more words being spoken both by the police officers and thesuspect.

LIMITATIONS OF THE PRESENT STUDY

It could be that the stakes for the liars and truth tellers indeed were per-ceived as low. They were informed that their participation was completelyvoluntary and that they could leave the experiment at any time (although noone chose to do this). However, we do believe that the liars and truth tellersperceived the stakes to be high compared to studies in which students andnot police officers acted as interrogators. In this study, it was stressed thatthe liars and truth tellers would be interrogated by real-life, experiencedcriminal investigators. This ought to have increased the liars’ and truth tell-ers’ perception of the task as realistic.

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 451

CONCLUSIONS

The present study extends the literature by investigating the deceptiondetection accuracy of experienced police officers in a more ecologicallyvalid fashion. Despite professional experience and the opportunity to influ-ence the flow of information, the lie detection accuracy level obtained bythe police officers was mediocre. The results were thus in line with findingsfrom other studies investigating the lie detection accuracy of police officers.Moreover, the analyses of the interrogations indicate that the police officerslack shared and agreed-on guidelines about how to conduct interrogations.We know from previous research that police officers have wrongful and ste-reotypical beliefs about deception (Strömwall & Granhag, 2003). Perhapslack of knowledge about deception and poor interrogation techniques workin tandem to create the discouraging accuracy levels obtained in this study.It may be the case that to improve the lie detection accuracy of police offi-cers, one needs to give these professionals information about reliable cuesto deception as well as provide information about beneficial ways of inter-rogating. It is possible that interrogating in certain ways may increase thebehavioral differences between liars and truth tellers; identifying suchinterrogation techniques could be invaluable. We believe that a crucial stepfor future research is to explore this relationship between interrogationtechniques and deception detection ability.

NOTE

1. An updated version of Criminal Interrogation and Confessions was published in 2001(Inbau, Reid, Buckley, & Jayne, 2001). The section concerning the nine-step approach toeliciting confessions remains unchanged in the new version of the manual.

REFERENCES

Allwood, C. M., & Granhag, P. A. (1999). Feelings of confidence and the realism of confi-dence judgments in everyday life. In P. Juslin & H. Montgomery (Eds.), Judgment anddecision making: Neo-Brunswikian and process-tracing approaches (pp. 123-146).Mahwah, NJ: Lawrence Erlbaum.

Baldwin, J. (1992). Video taping of police interviews with suspects—an evaluation (PoliceResearch Series Paper 1). London: Home Office.

Baldwin, J. (1993). Police interview techniques: Establishing truth or proof? British Journalof Criminology, 33, 325-352.

452 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

Buller, D. B., Strzyzewski, K. D., & Hunsaker, F. G. (1991). Interpersonal deception: II. Theinferiority of conversational partners as deception detectors. Communication Mono-graphs, 58, 25-40.

Burgoon, J. K., Buller, D. B., Ebesu, A. S., & Rockwell, P. (1994). Interpersonal deception:V. Accuracy in deception detection. Communication Monographs, 61, 303-325.

Burgoon, J. K., Buller, D. B., White, C. H., Afifi, W., & Buslig, A. L. S. (1999). The role ofconversational involvement in deceptive interpersonal interactions. Personality andSocial Psychology Bulletin, 25, 669-685.

DePaulo, B. M., Lindsay, J. J., Malone, B. E., Muhlenbruck, L., Charlton, K., & Cooper, H.(2003). Cues to deception. Psychological Bulletin, 129, 74-118.

DePaulo, B. M., & Pfeifer, R. L. (1986). On-the-job experience and skill at detecting decep-tion. Journal of Applied Social Psychology, 16, 249-267.

Einhorn, H. J. (1982). Learning from experience and suboptimal rules in decision making. InD. Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncertainty: Heuristicsand biases (pp. 268-283). Cambridge, UK: Cambridge University Press.

Ekman, P., & O’Sullivan, M. (1991). Who can catch a liar? American Psychologist, 46, 913-920.

Ekman, P., O’Sullivan, M., & Frank, M. G. (1999). A few can catch a liar. Psychological Sci-ence, 10, 263-266.

Feeley, T. H., & deTurck, M. A. (1997, May). Perceptions of communications as seen by theactor and as seen by the observer: The case of lie detection. Paper presented at the Inter-national Communication Association Annual Conference, Montreal, Canada.

Gilbert, D. T., & Krull, D. S. (1988). Seeing less and knowing more: The benefits of percep-tual ignorance. Journal of Personality and Social Psychology, 54, 193-202.

Granhag, P. A., Andersson, L. O., Strömwall, L. A., & Hartwig, M. (2004). Imprisonedknowledge: Criminals’beliefs about deception. Legal and Criminological Psychology, 9,103-119.

Granhag, P. A., & Strömwall, L. A. (2000). Deception detection: Interrogators’ and observ-ers’ decoding of consecutive statements. The Journal of Psychology—Interdisciplinaryand Applied, 135, 603-620.

Gudjonsson, G. H. (2002). The psychology of interrogations and confessions: A handbook.Chichester, UK: Wiley.

Hartwig, M., Granhag, P. A., Strömwall, L. A., & Andersson, L. O. (2004). Suspiciousminds: Criminals’ ability to detect deception. Psychology, Crime and Law, 10, 83-95.

Hartwig, M., Granhag, P. A., Strömwall, L. A., & Vrij, A. (2002). Deception detection:Effects of conversational involvement and probing. Göteborg Psychological Reports,32 (1).

Hogarth, R. M. (2001). Educating intuition. Chicago: University of Chicago Press.Holmberg, U., & Christianson, S.-Å. (2002). Murderers’ and sexual offenders’ experiences

of police interviews and their inclination to admit or deny crimes. Behavioral Sciencesand the Law, 20, 31-45.

Inbau, F. E., Reid, J. E., & Buckley, J. P. (1986). Criminal interrogation and confessions. Bal-timore: Williams & Walkins.

Inbau, F. E., Reid, J. E., Buckley, J. P., & Jayne, B. C. (2001). Criminal interrogation andconfessions. Gaithersburg, MD: Aspen.

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 453

Kassin, S. M. (1997). The psychology of confession evidence. American Psychologist, 52,221-233.

Kassin, S. M., & Fong, C. T. (1999). “I’m innocent!”: Effects of training on judgments oftruth and deception in the interrogation room. Law and Human Behavior, 23, 499-516.

Köhnken, G. (1987). Training police officers to detect deceptive eyewitness statements.Does it work? Social Behaviour, 2, 1-17.

Kraut, R. E. (1980). Humans as lie detectors: Some second thoughts. Journal of Communica-tion, 30, 209-216.

Mann, S. A. (2001). Suspects, lies and videotape: An investigation into telling and detectinglies in police/suspect interviews. Unpublished doctoral dissertation, University ofPortsmouth, United Kingdom.

Meissner, C. A., & Kassin, S. M. (2002). “He’s guilty!”: Investigator bias in judgments oftruth and deception. Law and Human Behavior, 26, 469-480.

Miller, G. R., & Stiff, J. B. (1993). Deceptive communication. Newbury Park, CA: Sage.Milne, R., & Bull, R. (1999). Investigative interviewing: Psychology and practice.

Chichester, UK: Wiley.Mortimer, A., & Shepherd, E. (1999). Frames of mind: Schemata guiding cognition and con-

duct in the interviewing of suspected offenders. In A. Memon & R. Bull (Eds.), Hand-book of the psychology of interviewing (pp. 293-315). Chichester, UK: Wiley.

Moston, S., & Engelberg, T. (1993). Police questioning techniques in tape-recorded inter-views with criminal suspects. Policing and Society, 3, 223-237.

Moston, S., & Stephenson, G. M. (1993). The changing face of police interrogation. Journalof Community and Applied Social Psychology, 3, 101-115.

Moston, S., Stephenson, G. M., & Williamson, T. M. (1992). The effects of case characteris-tics on suspect behaviour during police questioning. British Journal of Criminology, 32,23-40.

Pearse, J., & Gudjonsson, G. H. (1996). Police interviewing techniques at two South Londonpolice stations. Psychology, Crime and Law, 3, 63-74.

Pearse, J., & Gudjonsson, G. H. (1999). Measuring influential police interviewing tactics: Afactor analytic approach. Legal and Criminological Psychology, 4, 221-238.

Porter, S., Woodworth, M., & Birt, A. R. (2000). Truth, lies and videotape: An investigationof the ability of federal parole officers to detect deception. Law and Human Behavior, 24,643-658.

Sear, L., & Williamson, T. (1999). British and American interrogation strategies. In D. Can-ter & L. Alison (Eds.), Interviewing and deception (pp. 67-81). Dartmouth, UK: Ashgate.

Soukara, S., Bull, R., & Vrij, A. (2002, September). Investigative interviewing of suspects:The real picture! Paper presented at the 12th European Conference on Psychology andLaw, Leuven, Belgium.

Sporer, S. L., & Schwandt, B. (2002, March). Nonverbal indicators of deception: A meta-analytic synthesis. Paper presented at the Biennial Meeting of the American Psychology-Law Society in Austin, Texas.

Street, R. L., Mulac, A., & Wiemann, J. M. (1988). Speech evaluation differences as a func-tion of perspective (participant versus observer) and presentational medium. HumanCommunication Research, 14, 333-363.

Strömwall, L. A., & Granhag, P. A. (2003). How to detect deception? Arresting the beliefs ofpolice officers, prosecutors and judges. Psychology, Crime and Law, 9, 19-36.

454 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)

Strömwall, L. A., Hartwig, M., & Granhag, P. A. (2003). To act truthfully: Nonverbal behav-ior during a police interrogation. Manuscript submitted for publication.

Vrij, A. (1993). Credibility judgments of detectives: The impact of nonverbal behavior,social skills and physical characteristics on impression formation. Journal of Social Psy-chology, 133, 601-611.

Vrij, A. (2000). Detecting lies and deceit: The psychology of lying and its implications forprofessional practice. Chichester, UK: Wiley.

Vrij, A. (2002). Telling and detecting lies. In N. Brace & H. Westcott (Eds.), Applying psy-chology (pp. 179-241). Milton Keynes, UK: The Open University.

Vrij, A. (2003). We will protect your wife and child, but only if you confess. In P. J. vanKoppen & S. D. Penrod (Eds.), Adversarial versus inquisitorial justice: Psychologicalperspectives on criminal justice systems (pp. 55-79). New York: Kluwer Academic.

Vrij, A., Edward, K., & Bull, R. (2001). Police officers’ability to detect deceit: The benefit ofindirect deception detection measures. Legal and Criminological Psychology, 6, 185-196.

Vrij, A., & Graham, S. (1997). Individual differences between liars and the ability to detectlies. Expert Evidence: The International Digest of Human Behaviour Science and Law, 5,144-148.

Vrij, A., & Mann, S. (2001a). Telling and detecting lies in a high-stake situation: The case ofa convicted murderer. Applied Cognitive Psychology, 15, 187-203.

Vrij, A., & Mann, S. (2001b). Who killed my relative? Police officers’ ability to detect real-life high-stake lies. Psychology, Crime, & Law, 7, 119-132.

Vrij, A., & Semin, G. R. (1996). Lie experts’beliefs about nonverbal indicators of deception.Journal of Nonverbal Behavior, 20, 56-80.

Williamson, T. M. (1990). Strategic changes in police interrogation: An examination ofpolice and suspect behaviour in the metropolitan police in order to determine the effect ofnew legislation, Technology and Organisational Policies. Unpublished doctoral disserta-tion, University of Kent, United Kingdom.

Maria Hartwig, B.Sc., Ph.D. candidate, is a member of the research unit forcriminal, legal, and investigative psychology at the Department of Psychol-ogy, Göteborg University, Sweden. Her research interests include deceptiondetection, interrogating suspects, and geographic profiling. Ms. Hartwigteaches forensic psychology.

Pär Anders Granhag is an associate professor at the Department of Psychol-ogy, Göteborg University, Sweden. He is since 10 years active in research onadults’ and children’s eyewitness testimony, deception, issues of reliability,and metacognition. He has published a number of scientific papers on thesetopics and is on the editorial board of the scientific journal Applied Cogni-tive Psychology. He is experienced in training the police in investigative in-terviewing and how to assess reliability.

Hartwig et al. / OFFICERS’ LIE DETECTION ACCURACY 455

Leif A. Strömwall, Ph.D., assistant professor, is a member of the research unitfor criminal, legal, and investigative psychology at the Department of Psy-chology, Göteborg University, Sweden. He has published studies on decep-tion detection, eyewitness testimony and memory, and realism in confidencejudgments. Dr. Strömwall teaches forensic psychology, general psychology,social psychology, and research methods and statistics.

Aldert Vrij is a professor of applied social psychology at the University ofPortsmouth, United Kingdom. His deception research focuses on (a) verbaland nonverbal indicators of deception, (b) people’s ability to detect lies, and(c) methods to improve people’s ability to detect deceit. He has publishedmore than 200 articles and book chapters and six books, including his recentbook Detecting Lies and Deceit: The Psychology of Lying and the Implica-tions for Professional Practice (2000, Wiley).

456 POLICE QUARTERLY (Vol. 7, No. 4, December 2004)