Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The language of deceit: are there reliable verbal clues to deception in the interrogation context? Porter, Stephen B. 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1994-0598.pdf [ 1.42MB ]
JSON: 831-1.0087391.json
JSON-LD: 831-1.0087391-ld.json
RDF/XML (Pretty): 831-1.0087391-rdf.xml
RDF/JSON: 831-1.0087391-rdf.json
Turtle: 831-1.0087391-turtle.txt
N-Triples: 831-1.0087391-rdf-ntriples.txt
Original Record: 831-1.0087391-source.json
Full Text

Full Text

THE LANGUAGE OF DECEIT:ARE THERE RELIABLE VERBAL CLUES TO DECEPTIONIN THE INTERROGATION CONTEXT2bySTEPHEN B. PORTERB.Sc.(Hon.), Acadia University, 1992A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTSinTHE FACULTY OF GRADUATE STUDIES(Department of Psychology)THE UNIVERSITY OF BRITISH COLUMBIASeptember, 1994©. Stephen B. Porter, 1994We accept this thesis as conformingdardIn presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of_____________The University of British ColumbiaVancouver, CanadaDate______ _jqcfDE-6 (2188)11AbstractIn recent years, the need for enhanced methods ofcredibility assessment in criminal cases has becomeilluminated. Especially in cases of sexual assault, thewords of the accused and complainant are often the soleevidence available to police. Consequently, researchers andpractitioners have been searching for ways ofdifferentiating truthful and deceptive accounts, focussingmainly on witnesses and victims. With its recent history,however, assessment based on verbal clues has been somewhatmyopic and not well grounded in theory or integration. Thisthesis examined a general hypothesis, based upon conceptualinformation from a variety of perspectives, that reliableverbal indicators of deception exist in the interrogationsituation. Sixty undergraduates were recruited forparticipation in research addressing “securityeffectiveness.” Participants either committed a theft “totest the effectiveness of a new security guard” or carriedout a similar but innocuous task. They were then asked toprovide either: (1)a truthful alibi (2)a partiallydeceptive account (3)a completely false alibi or (4)atruthful confession regarding the theft to “an intervieweralso hired for the purpose of investigating thefts.” Toincrease motivation in the interview, honest and dishonestparticipants were offered a monetary incentive forconvincing the interrogator of their veracity. The accountsiiiwere then transcribed and examined for the presence ofeighteen language variables. A multivariate analysis ofvariance (MANOVA) revealed a profile of three variableswhich significantly differentiated the truthful anddeceptive accounts (amount of detail reported, coherence ofthe account, and admissions of lack of memory). Forexample, dishonest participants provided much less detail(although not fewer words) in relating an event thantruthful participants describing a similar event.Implications for credibility assessment in forensicinterrogations are discussed, emphasizing the need forestablishing external validity with eclectic researchstrategies.ivTABLE OF CONTENTSAbstract iiTable of Contents ivList of Tables . . . viList of Figures . viiAcknowledgment viiiChapter 1.INTRODUCTION . . . . . . . . . . . 1Chapter 2.PERSPECTIVES ON MEMORY ACCURACY . 92.1 The Undeutsch Hypothesis 92.2 RealityMonitoring 17Chapter 3.PERSPECTIVESONDECEPTION .. 193.1 Emotional - Motivational Approaches ..... 193.2 CognitiveApproaches . 23Chapter 4.THE PROPOSALS OF AVINOAM SAPIR . 324.1 Ratio of Component Elements 324.2 Unnecessary Connectors .... .. 334.3 UseofPronouns 34Chapter 5.SUMMARY OF THEORETICAL FRAMEWORK AND THEPRESENT STUDY. . . . . . . . . . . . . . . . . . . 36Chapter 6.METHOD ..... 386 . 1 Participants . . . . . 3 86.2 Procedure 386.3 Instructional Manipulation 396.4 Interviewing 416 .5 Transcribing . . . 436.6 Dependent Variables and ScoringProcedures . . . . 43Chapter 7.RESULTS . . 507.1 ReliabilityChecks ....7.2 Preliminary Analyses ....... 517.3 MultivariateAnalysis 547.4 Relative Discriminating Ability ofCriteria Showing Significant Effects .... 587.5 Followup Exploratory Analyses . 607.6 Effects of Interrogation Probes . 65Chapter 8.DISCUSSION 668.1 Limitations of the Current Study 798.2 Directions For Future Research ..... 84REFERENCES 87AppendixA: DebriefingForm 96Appendix B: Interrogation Protocol 99Appendix C: Standard Deviations For Table 1Dependent Variables 103VList of TablesTable 1: Mean Scores for the Four Conditions onAllDependentVariables. 53Table 2: Discriminant Analysis of CriteriaShowing Significant Effects forClassification of Truthful and DeceptiveAccounts . . . . . . . . . . • 61viList of FiguresFigure 1: Number of Details Reported AcrossConditions . 56Figure 2: Coherence Ratings (0 - 3) AcrossConditions 57Figure 3: Number of Admissions to Lack of MemoryAcross Conditions . . . . 59viiviiiAcknowledgmentThe author would like to express appreciation to theinterviewers and coders who contributed to this study. Inparticular, I would like to thank Deborah Tong, Barry Floyd,Karen Hausmann, Mark Suter, Monica Tymofievich, and SusanBault who devoted much time and effort to this research ininterviewing or coding data. I thank also my advisor, JohnYuille, for his support and excellent advice.The author of this thesis is supported by apostgraduate fellowship from the Natural Sciences andEngineering Research Council of Canada (NSERC). The presentresearch was funded directly by a grant from Division 41 ofthe 2merican Psychological Association (APA).DECEPTIVE LANGUAGEChapter 1INTRODUCTIONAnd the Lord said unto Cain, “Where is Abel thybrother?” And he said “I know not: Am I mybrother’s keeper?” — Genesis 5:9It did not take long for humans to employ language toattempt to circumvent penalty for their misdeeds. In thisfirst documented case of such a phenomenon, Cain’s verbaldeception is quickly ascertained by the Deity who imposes aharsh punishment for his murderous transgression.Unfortunately, human investigators do not possessomniscience and must seek alternative methods of credibilityassessment in criminal cases. The present thesis exploresone method whose complexities and potential utility to thisend have only recently been recognized — statement analysis.Indeed, the question of whether truthful and deceptiveaccounts can be reliably differentiated based on certainlanguage characteristics has been investigatedsystematically only in recent years (Landry & Brigham, 1992;Porter & Yuille, in press). Nonetheless, practicalapplications of statement analysis have been conducted byinvestigators since antiquity. Larsen (1926) contended thatthe earliest account of deception detection by a humanoccurs in the Bible when King Solomon is asked to decidewhich of two women claiming the same child is telling theDECEPTIVE LANGUAGE 2truth. He settled the dilemma by ordering that the child besevered into two pieces after which one woman renounced herclaim and the other maintained silence. By examining theselanguage behaviors, Solomon expeditiously determined theidentity of the genuine mother (the retractor) and resolvedthe dispute. Hocking, Bauchner, Kaminski, & Miller (1979)described a papyrus scroll from 900 B.C. which articulatesan Egyptian method of credibility assessment based in parton an analysis of the suspect’s language behaviors:The liar does not answer questions, or they areevasive answers; he speaks nonsense, rubs the bigtoe along the ground and shivers; his face isdiscoloured; he rubs the roots of his hair withhis fingers.More recently, Camps and Berger (1966) describedinvestigators’ utilization of verbal clues in detectingdeception in the 1938 London murder case involving JohnChristie. Police located eight women in various stages ofdismemberment and decomposition in the vicinity ofChristie’s residence. Most had been sexually assaulted atthe approximate time of death. Upon questioning, Christieeventually admitted responsibility for the deaths of all thewomen. An analysis of his statements, however, exhibited aselective presentation of information and other techniquesindicative of deception. The delusory elements of hisstatements became evident with systematic evaluation (mainlyby noting discrepancies between his testimony and physicalevidence). In retrospect, by comparing his statements todetailed scientific evidence available to the police, it wasDECEPTIVE LANGUAGE 3possible to assess their accuracy and completeness,revealing Christie’s deceptions. However, it is conceivablethat a more sophisticated system of statement analysis couldhave prevented several deaths; Christie had been questionedearlier in the investigation and released, after which otherwomen were killed and the wrong man was hanged for thecrimes (Camps & Berger, 1966).Documented cases of deception detection based on anexamination of the accused’s account are rare in relationto techniques founded on physiological attributes. InChina, an ancient tradition was to require a suspect to chewrice and spit it out; if the rice remained dry, the suspectwas deemed culpable in that a guilty conscience was thoughtto inhibit salivary secretions (Larsen, 1926). In medievalEngland, the detection of guilt was conducted by methodscollectively known as “trial by ordeal.” For example, ifthe accused were tossed into a river and failed toresurface, innocence was proclaimed. Alternatively, if thesuspect lived he/she was deemed guilty and consequentlydrowned or burned at the stake. Other trials by ordealincluded immersing the suspect’s arm in boiling water forlengthy periods or prolonged bleeding to elicit aconfession. These methods were founded on the premise thatan innocent person was able to endure the ordeal longer thana guilty criminal (Lykken, 1981).In modern times, increasingly sophisticated methods ofcredibility assessment based on physiology have emerged,DECEPTIVE LANGUAGE 4most notably the polygraph. Despite its renown, however,the polygraph has not gained unequivocal status as aneffective detector of deception. For example, a majorcriticism of the polygraph is the elevated rate of falsepositive decisions it yields (e.g., Barland & Raskin, 1976;Horvath, 1977; Kleinmutz & Szucko, 1984), running counter toa legal system designed to protect the rights of theinnocent (see lacono & Patrick, 1987). As well, guiltypeople can “fool” the polygraph under certain circumstances(Honts, Raskin, & Kircher, 1994).Other components of the detection arsenal includenarcoanalysis, hypnosis, and voice analysis (PsychologicalStress Evaluator). As in the case of the polygraph,however, serious shortcomings are associated with eachtechnique that limit their efficacy as investigative tools(e.g., see Gunn & Gudjonsson, 1988). Some will inevitablycontend that the experience or intuition of an investigatoris superior to any technique designed for credibilityassessment purposes. However, recent research has negatedthis assumption (e.g., DePaulo & Pfeifer, 1986; Kohnken,1987), indicating that law enforcement officers generallyperform at a similar level to untrained people. Ekman andO’Sullivan (1991) found that robbery investigators,psychiatrists, judges, professional polygraphers, collegestudents, and a special interest group who had receivedtraining in detecting deceit performed at chance levels indetermining the veracity of videotaped reports whereasDECEPTIVE LANGUAGE 5Secret Service agents did so slightly (but significantly)above chance level. Indeed, the effectiveness of intuitionin any assessment context is questionable (e.g., seecritiques of dangerousness assessments based on clinicians’intuition and experience by Cocozza & Steadman, 1978; Klein,1976; Werner, Rose, & Yesavage, 1983).The problems associated with traditional credibilityassessment methods indicate an exigency for the developmentof novel, innovative detection techniques. This isespecially salient if one considers the fact that often thesole evidence available to the police are the testimonialsof the accuser and the accused. For example, in cases ofadult and child sexual assault there is frequently little orno evidence to support or refute the allegation (e.g.,Yuille, 1988). Cases of this type have been increasing atan alarming rate (see Finkeihor, 1986) causing tremendousdifficulties for investigators and adjudicators inestablishing fact.Corresponding to the need for new detection techniques,a multitude of studies on the nonverbal correlates ofdeception have appeared in the literature (see Ekman, 1992;DePaulo, 1992) but comparatively very little research onverbal correlates. Yet, several researchers have arguedthat verbal behaviors are more reliable indices of deceptionthan nonverbal behaviors (e.g., Bauchner, Kaplan, & Miller,1980; deTurck & Miller, 1985; Hocking, Bauchner, Kaminski, &Miller, 1979; Zuckerman, DePaulo, & Rosenthal, 1981). InDECEPTIVE LANGUAGE 6the Zuckerman et al. (1981) study it was determined that theaccuracy of detecting deception from verbal content(transcript) was higher than from any other channel (facialexpressions, body language, etc.) spurring the authors toconclude that this evidence “is in direct contradiction tothe belief that nonverbal channels are more likely todisclose deception than are verbal cues” (Zuckerman &Driver, 1985, p. 130). Also, as noted by Gudjonsson (1992),many interrogators appear to have “blind faith” in nonverbalclues to deception, even though they are often unreliableindicators of deception (Ekman, 1992). McCornack (1992)lamented that a “primitive view” of deceptive interactionexists partly because of the myopic focus on nonverbal cuesassociated with deception and argued that researchers mustexamine the content of the messages to derive meaningfulprinciples of deception. In the past few years, a largelyunresearched but promising series of methods for determiningveracity based on an analysis of verbal clues in theaccounts of a suspect or witness have emerged. Clearly, thegenesis of a technology capable of establishing thetruthhood of a person’s statements would be welcomed byinvestigators of crime. Indeed, certain law enforcementagencies have expressed great interest in the development ofsuch a procedure (Lesce, 1990).In general, the statements of suspects elicited duringthe interrogation process are quite amenable to analysis.As described by Buckwalter (1983), a usual tactic ofDECEPTIVE LANGUAGE 7interrogators is to solicit the suspect’s free recallnarrative account of the criminal act being investigated orhis/her alibi. While the free recall account is beingprovided, few interruptions occur regardless of apparentfabrications in the story. This statement is audio- orvideo—taped and subsequently transcribed verbatim in writtenform. This unfragmented narrative account is ideal forstructural and content analyses for the purpose of deceptiondetection. It seems likely that in most interrogations anon-admitting guilty suspect will provide either a narrativerelating a false alibi or one relating a distorted versionof the circumstances or facts of the crime which isconducive to systematic analysis. On the other hand,statement analysis may have less applicability with denialstatements (e.g., “I didn’t kill anyone!”; “I know nothingabout this crime”), statements of memory loss (e.g., “I wasso drunk I just can’t remember what happened”), or obviouslycases in which the suspect refuses to make a statement(e.g., Moston & Stephenson, 1992, found that 16% of thecriminal suspects in their sample utilized this option).In summary, suspect interrogations represent a contextin which lying indubitably occurs with high frequency. Forcertain types of crimes, the words of the complainant andaccused are the only evidence available to police (e.g., adelayed report of sexual assault). Unfortunately, thevalidity of current methods of credibility assessment suchas the polygraph is questionable, pointing to the need forDECEPTIVE LANGUAGE 8novel approaches. The present thesis investigated thepotential utility of verbal clues in the detection ofdeception from the transcripts of interrogations with honestand deceptive suspects.Although the idea for this thesis was galvanized anddeveloped from applied issues relating to suspectinterrogations, the present work is based upon severaltheoretical approaches. Up to this point, there has beenlittle integration of the theoretical perspectives relevantto the development of statement analysis techniques. Threemajor theoretical approaches will be emphasized in theensuing three chapters: (1)Mexnory—based approachesdescribing how accurate and inaccurate memory reports can bedistinguished (2)Deception-based approaches describing howdeceptive and truthful statements can be distinguished(3)the theories of Avinoam Sapir which contain aspects ofthe first two approaches but are sufficiently unique andbroad in scope to warrant independent discussion.DECEPTIVE LANGUAGE 9Chapter 2PERSPECTIVES ON MEMORY ACCURiCY2.1 The Undeutsch Hypothesis. One of the founding figuresin the development of statement analysis was Undeutsch, aGerman psychologist. Undeutsch was instrumental in thegenesis of a systematic procedure for determining theveracity of child witness reports referred to as StatementReality Analysis (SRA) which later became known as StatementValidity Analysis (SVA). The theoretical basis for thistechnique, a proposition known as the “UndeutschHypothesis”, simply asserts that accounts based on memoryfor an actually experienced event differ from fictitiousaccounts in structure, content, and quality. This argumenthas been reiterated by psycholinguists who have argued thatwhen people shift from being truthful to deceptive theirlanguage behaviors are markedly altered (e.g., Dulaney,1982). For example, it is argued that recall accounts for atrue event contain a richness of detail not present ininaccurate or falsified accounts (Undeutsch, 1967).According to Arntzen (1983), a multitude of details in astatement is a good indication of credibility because it isdifficult for witnesses to intentionally (orunintentionally) embellish a false testimony with manydetails because the details do not exist in memory. Aswell, witnesses may not wish to provide a multitude ofdetails because of the increased difficulty of maintainingDECEPTIVE LANGUAGE 10the story’s consistency. As noted by Steller and Kohnken(1989), a memory report is considered detailed if itcontains the exact description of the place, vivid persondescriptions from different aspects, and a step-by-stepdescription of the sequence of events.The majority of research addressing the validity of theUndeutsch Hypothesis in credibility assessment has beenconducted with child witnesses. Statement Validity Analysis(SVA) for children’s statements has been formalized andempirically evaluated. SVA has proven useful in assessingthe credibility of child witnesses and victims. AlthoughSVA has been developed formally in the past decade, it hasbeen used for about forty years predominantly in Germany andScandinavia. During the past few years, an internationalgroup of researchers have developed and elaborated theprocedure (Undeutsch, Steller, Kohnken, Raskin, Esplin, &Yuille; see Raskin & Yuille, 1989; Yuille, 1988). The SVAprocedure consists of two stages: (1)criteria—based contentanalysis (2)examination of other evidence in the case.Content Analysis involves examining the statement forevidence of features associated with actual experiences.Nineteen criteria are used to evaluate the credibility ofthe content of the statement (e.g., quantity of details,logical structure, accounts of subjective mental state).Scientific evaluations have been conducted to measurethe efficacy of the content analysis component of SVA indetermining the veracity of children’s statements. YuilleDECEPTIVE LANGUAGE 11(1988) had 49 Grades 1 and 3 students prepare one false andone truthful story. The children were given two days toprepare their narratives. SA correctly classified 74.4% ofthe false stories and 90.9% of the true accounts. Steller,Wellershaus and Wolf (1988; cited in Steller & Boychuk,1992) conducted a study aimed at increasing ecologicalvalidity. Eighty-eight children from Grades 1 and 3 wereprovided a list of hypothetical experiences containingaspects of direct involvement, loss of control, and negativeemotional tone (e.g., having dental work done or beingattacked by an animal). They were asked to come up with atrue and false story about one of these events. The trueand false accounts were successfully distinguished by nineof the SVA criteria.Unfortunately, there has been a dearth of studiesaddressing the effectiveness of the SVA procedure indifferentiating genuine and fictitious recall accounts inthe case of adults. It is, of course, possible that SVA iscontext-dependent and its proper niche is solely withchildren’s accounts in cases of alleged abuse. Nonetheless,the two studies evaluating its utility in this regard havereported moderately high rates of accuracy in classifyingadult statements (Landry & Brigham, 1992; Zaparniuk, Yuille,& Taylor, 1993). Landry and Brigham (1992) recentlypresented one of the first empirical studies extending theuse of SVA to the accounts of adults. They studied adultsusing fourteen of the nineteen CBCA criteria. ParticipantsDECEPTIVE LANGUAGE 12judged the truthfulness of the videotaped statements oftwelve adults; six were true and six related an inventedtraumatic personal experience. For the trained subjects,ten of the fourteen criteria yielded significant differencesin the predicted direction between evaluations of truthfuland invented statements (reproduction of conversation,spontaneous corrections, and self—doubt were particularlyeffective) offering compelling evidence for SVA’s potentialas a tool for adult credibility assessment. On the otherhand, two of the criteria were significantly more prevalentin the deceptive accounts (logical structure and comments onanother’s mental state). Subjects trained in CBCAoutperformed the untrained subjects in detecting deceptionsuggesting that training in adult SVA may be useful inenhancing one’s ability to distinguish true and falseaccounts. In the Zaparniuk et al. (1993) study,participants were allocated to one of two groups; in onecondition participants viewed one of two videotaped eventsand then recalled the event while in the second conditionparticipants heard an audiotape of one of the events andthen recalled the event as if they had witnessed it. SVAdistinguished between the truthful and deceptive statementswith a “reasonably” high degree of accuracy (mean hit rate =76%).Despite its apparent utility with witnesses, bothchildren and adults, it is difficult to evaluate theefficacy of SVA with criminal suspects based on thisDECEPTIVE LANGUAGE 13research. The deceivers and truthtellers in the foregoingstudies were witnesses or victims, had not actuallyperpetrated an offense, and were provided little opportunityto prepare their stories prior to the interview.Additionally, the deceivers and truthtellers had littleimpetus for convincing the observer or interviewer of theirveracity. Hence, leakage (signs of deception “leaking” out)in their language behaviors due to arousal may have beenminimized. Thus, verbal characteristics of the accountsassociated specifically with the experience of committing anoffense, planning the lie, and being motivated to escapedetection likely did not emerge. It is clear that thepotential utility of SVA with criminal suspects has not beentested.Inconsistent evidence regarding the validity of theUndeutsch Hypothesis derives from empirical studiescomparing the amount of detail in true and fictitious memoryaccounts. The issue of whether truthful and deceptiveaccounts generally differ in length and degree of detail isunresolved. Some authors (e.g., deTurck & Miller, 1985;Knapp, Hart, & Dennis, 1974; Kraut, 1976; Mehrabian, 1971)have contended that people providing false statements aremore laconic than people delivering truthful reports (i.e.,their accounts are not as detailed) which appears compatiblewith the Undeutsch Hypothesis. Macdonald and Michaud (1987)argued that “the suspect who gives only brief answers isalmost certainly lying through concealment of information”DECEPTIVE LANGUAGE 14(p. 36). Correspondingly, Knapp et al. (1974) conducted anextensive examination of the language of deceit and foundthat deceivers used significantly fewer words, fewerdifferent words, fewer past tense verbs (was, were, etc.),more other references (they, them, etc.), fewer groupreferences (we, our, etc.), and fewer self—references (I,me, my, etc.). In contrast, Harrison, Halek, Raney, andFritz (1978) reported that deceivers in their study weresignificantly more garrulous in their descriptions of anevent than people providing truthful accounts, particularlywhen recounted in the presence of an interviewer.Similarly, Arcaro (1990) and Macdonald and Michaud (1987)asserted that an excessively detailed alibi is indicative ofdeception because an innocent person would not recallinsignificant activities coinciding with the time of theoffense. In this view, excessive detail is a clue to arehearsed response. Zuckerman & Driver (1985) conducted ameta-analysis examining twelve visual behaviors (e.g., pupildilation, smiling, postural shifts), six linguisticbehaviors (e.g., response length, latency, speech errors),and five verbal behaviors (e.g., self-references, irrelevantinformation, negative statements) that had been cited atleast twice in the literature as being indicative ofdeception. Response length was significantly negativelycorrelated with deception. They found that the behaviorsmost strongly associated with deception (p.<.OO1) includedpupil dilation and adaptors (visual), speech hesitationsDECEPTIVE LANGUAGE 15(paralinguistic), and immediacy (-) and negative statements(verbal). Other significant relationships between languagebehaviors and deception were speech errors, pitch,irrelevant information, and leveling (frequency ofovergeneralized statements indicated by terms such as“every”, “all”, “nobody”, “none”, etc.).One of the premier studies to examine clues todeception from multiple behavioral indices addressed thisissue. deTurck and Miller (1985) compared the videotapedaccounts of deceptive undergraduates who had been induced bya confederate to cheat on a task to win a prize forperformance to those who had been engaged in a task—relateddiscussion with the confederate. Dependent measuresincluded five nonverbal (number of blinks, adaptors,feet/leg gestures, eye contacts) and four verbal (number ofspeech errors, number of pauses, length of response latency,and message duration) cues to deception. The resultsindicated that six cues (adaptors, hand gestures, pauses,speech errors, response latency, and message duration[—])were significantly associated with deceptive accounts. Theauthors argued that these findings support the assertionthat verbal cues are more important than nonverbal cues indetecting deception. Stiff and Miller (1986), employing asimilar methodology to the deTurck et al. (1985) study,found that deceptive statements were predicted by “verbalcontent” (r=—.40, measured by how plausible, definite,clear, and concise the response was rated), statements ofDECEPTIVE LANGUAGE 16personal responsibility (r.21), self—references (r=.21),and number of words (r=—.21). In summary, the majority ofstudies addressing the account length/veracity relationshiphave found a positive correlation lending support toUndeutsch’s original claim.One of the first published formal applications ofstatement analysis to an actual crime considered theUndeutsch Hypothesis (i.e., the amount and coherence ofdetails) in evaluating credibility (Yuille & Cutshall,1989). These authors examined the 1972 case of JeffreyMacDonald whose wife and two children had been brutallymurdered. MacDonald was subsequently found guilty of thecrimes and imprisoned. MacDonald maintained that a group ofdrug-crazed hippies had entered his home and committed theoffenses. A woman named Helena Stoeckley corroborated hisaccount, claiming for years that she was one of theintruders of whom MacDonald spoke. Yuille and Cutshallexamined Stoeckley’s statements and concluded that “thelogical consistency of her statements, particularly therelationship of her account to the general pattern of themurders, the spontaneous nature of her description, and thewealth of details in her statements provide evidence whichsupports the credibility of Helena Stoeckley’s account” (p.188).DECEPTIVE LANGUAGE 172.2 Reality Monitoring. A seminal paper by Johnson andRaye (1981) described elaborately how memories for pastperceptions of actual events can be distinguished from pastacts of imagination, using a model termed “RealityMonitoring.” This model posits that perceptual memories areassociated with a higher level of external sensorialinformation (e.g., “It was a large cedar home”) whereasself—generated (imagined) memories are the result ofcreative cognitive processes and are thus associated withmore internally—generated cognitive details and subjectiveidiosyncratic information (“I remember seeing the home andthinking how lovely cedar homes are”).Schooler, Gerhard, & Loftus (1986) have demonstratedthat memory descriptions of genuine and imagined (suggested)events differ in ways congruous with the theory of realitymonitoring. Some participants were induced throughsuggestion to recall an object that had not actually beenwitnessed (employing Loftus’ postevent misinformationparadigm). Content analysis revealed that genuine memories(in written form) contained more references to attributes ofthe memory stimulus, less reference to cognitive processesoccurring during encoding, fewer “hedges” (e.g., “It seemsto me” or “I believe”), and fewer words. In a second study,Schooler, Clark, and Loftus (1988) employed more criticalmemory items and asked participants for verbal rather thanwritten accounts. Again, descriptions of genuine memoriescontained more sensory details, fewer references toDECEPTIVE LANGUAGE 18cognitive processes, fewer verbal hedges, and fewer self—references (i.e., pronoun “I”).Nonetheless, as pointed out by Leippe, Manion, &Romanczyk (1992), the memory stimuli employed by Schooler etal. (1986; 1988) were noncomplex (e.g., a stop sign).Leippe et al. (1992) suggested that even more frequent andricher distinguishing cues may have emerged in recallaccounts concerning a more complex event. Addressing thispossibility, Leippe et al. (1992) staged an interactive,elaborate event in which participants were given a bogusskin sensitivity test including touching and conversation.Participants were subsequently interviewed to gauge theirmemories for the event. Leippe et al. (1992) then dividedthe accounts into high and low accuracy and compared them ona number of measures. It was found that the accurateaccounts were delivered more confidently, contained lessinstances of verbal hedging, less admissions of memoryfailures (e.g., “I don’t know”), and were lengthier than theinaccurate accounts. As well, people judging the accuracyof the accounts generally overused confidence and underusedthe other cues in determining the accuracy of the accounts.The authors concluded from these results that “with furtherresearch and insight, it may be possible to train people tolook for them (cues) and so become better judges of memoryreports” (p.195).DECEPTIVE LANGUAGE 19Chapter 3PERSPECTIVES ON DECEPTION3.1 Emotional-Motivational Approaches.Inability to lie is far from being love of truth.Be on your guard. — Nietzsche’s ZarathustraPresumably, lying is a difficult undertaking for mostpeople especially when the consequences for a failed lie areof considerable magnitude. As noted by Simpson (1992), aliar acts self-consciously against a convention containing amoral dimension. Accordingly, emotional and motivationalfactors have been implicated as being critical indifferentiating truth and falsehood in the interrogationcontext (see Ben—Shakhar & Furedy, 1992 for a comprehensivereview of the various theoretical perspectives consideringthese factors in physiological deception detection).According to the theory of the “punishment mechanism”, fearof failure to deceive leads to a specific physiologicalresponse associated with the fear (Ben—Shakhar & Furedy,1992). Davis (1961) argued that the greater theconsequences of being detected, the greater thephysiological response during lying and, therefore, thegreater the likelihood of detection. In an interrogationsituation, a suspect with guilty knowledge will be expectedto exhibit high arousal due to this fear of detection,leading to discernible changes in verbal behaviors.Nonetheless, as pointed out by Inbau, Reid, and Buckley(1986), signs of nervousness during an interrogation in bothDECEPTIVE LANGUAGE 20guilty and innocent suspects can result from many factorsindependent of their level of veracity (e.g., concern thatpolice and family will discover other transgressions;concern for what will happen to them while in custody andduring the interrogation). However, these authors arguethat guilty suspects will experience higher levels ofanxiety than innocent suspects because of their knowledge ofthe crime.Other hypotheses relating deceptive behavior toemotional arousal include the “conditioned response theory”which postulates that lying produces autonomic reactionsbecause the related questions are associated with negativeexperiences (i.e., because past lying behaviors werefrequently linked with aversive stimuli) and the “conflicttheory” (see Kohnken, 1985) positing that observablepsychological reactions result from the conflicting needs tolie (situational demand) and tell the truth (societal orfamilial demand).It is evident that the motivation for being deceptiveis an important consideration in deception detection bystatement analysis. Ekman and Friesen (1969) asserted thatit is most difficult to lie successfully when theconsequences of deception detection are great. Morerecently, DePaulo and colleagues (e.g., DePaulo & Kirkendol,1989) have referred to this phenomenon as the “motivationalimpairment effect.” Physiological research has demonstratedan association between increased motivation and higher ratesDECEPTIVE LANGUAGE 21of detection (e.g., Gustafson & Orne, 1963; Elaad & BenShakhar, 1989). If this theory holds for observablebehaviors (including language behaviors), liars and truth-telling persons would be expected to behave differently to agreater degree when there are serious perceived consequencesfrom detection. Obviously, the impetus is high for a guiltysuspect to successfully deceive his/her inquisitor. Whenthe ramifications are this significant, deception clues, or“leakage”, will likely be emitted through speech, voice,body, and face (Ekman, 1992) despite the liar’s attempts toconceal emotional portrayals. The present study is foundedin part on the theory that emotional and motivationalfactors influence people to “make verbal mistakes that canprovide both leakage and deception clues” (Ekman, 1992,p.85).One aspect of Zuckerman et al.’s (1985) meta—analyticstudy concerned motivation for lying. Motivation wasconsidered substantial if participants had been promisedsome monetary reward for doing well on the deception task orif the task had been described as a measurement of a skill.Highly motivated deception was associated with responselength (—), speech rate (—), speech errors, speechhesitations, pitch, negative statements, self—references(—), and leveling more so than unmotivated deception. Thecriteria related to language behaviors proved quite usefulin differentiating truthful and deceptive statements,DECEPTIVE LANGUAGE 22particularly when motivation was high. The authors concludedthe following:It appears that although the content of the speechmay be quite controllable, its semantic structuremay be as leaky as the body and/or tone of voice.Consequently, the search for cues to deceptionmust focus on both verbal and nonverbal behaviors(Zuckerman et al., 1985, p. 144).Experimental studies have found that the experience oflying may result in an elevated frequency of speech errors(e.g., Kohnken, 1985), possibly related to the motivationalimpairment effect. Comber and Canter (1983) conducted astudy to determine whether malicious (hoax) and non—malicious (truthful) fire alarm calls could be distinguishedon the basis of their psycholinguistic characteristics.Eight malicious and eight non—malicious calls were comparedin terms of the number of speech errors of various kindsoccurring in the call (number of times the operator andcaller spoke simultaneously, number of filled pauses [“umm”,“ahh”] in the caller’s speech, number of questions asked bythe caller, and the number of false starts in the operator’sand caller’s speech). Although no single variablesignificantly differentiated malicious and non-maliciouscalls (probably due to low statistical power resulting fromthe sample size), the mean differences were in the predicteddirection for all variables leading the authors to concludethat “it was possible to differentiate the two types ofcalls based on the overall configuration of speech behaviorwithin the call” (p. 460).DECEPTIVE LANGUAGE 233.2 Cognitive Approaches.For words, like Nature, half revealAnd half conceal the soul within. -Aifred Lord TennysonThe interrogation may be viewed as a context in whichdeceptive information is employed as an adaptive response toa dangerous situation. As noted by McCornack (1992),researchers have recently argued that deceptiveness reflectsa functional adaptation to the demands of a complexcommunication situation. People are often confronted withinteractions in which the competing goals of conveyinginformation and minimizing the damage which might resultfrom the information are operating. Thus, people presentdeceptive messages in ways fostering control andmanipulation of dangerous information. Building on thispremise, McCornack (1992) formulated “InformationManipulation Theory” (IMT) postulating that deceptivemessages function deceptively because they covertly violateprinciples guiding conversational exchanges. IMT suggeststhat although deceptive messages are infinite in number, theinformation in them varies in systematic, identifiable ways.Given that conversational interactants possess expectationsregarding the quantity, quality, manner, and relevance ofthe information presented, it is likely that deceiversviolate these expectations in attempts to deceive thelistener. In the suspect interrogation, IMT would predictthat the information presented by a guilty suspect woulddiffer from that of an innocent suspect on these fourDECEPTIVE LPNGUAGE 24measures (quantity, quality, manner, and relevance) becauseof the “danger” which sensed by the deceiver. Deceptiveinteractants are likely to fail to fulfill the informationalrequirements of the interaction, provide information that isirrelevant to the central conversational topic, and provideinformation in a less orderly, coherent fashion with fewerattempts to avoid ambiguity and confusion.Others have also contended that information processingand attentional strategies are important considerations in adeception detection task. Waid & Orne (1981) have suggestedthat physiological detection may depend on how the suspectprocesses the information presented to him/her in theinterview. Suspects may avoid detection by ignoring thestimulus or responding mechanically to the stimulus without“deeply” processing it (Craik & Lockhart, 1972). It may bepossible for suspects to circumvent the problem of deceptionleakage by encoding information in the interview on asuperficial level. One interviewing strategy which may beeffective in counteracting the processing of interviewstimuli on a superficial level is the use of interrogativeprobes (e.g., “Come on. It is obvious that you are lying tome. What really happened?”). Such negative feedback maymake it very difficult for a suspect to encode and respondto information superficially (see Gudjonnson & Clark, 1986).It is possible that verbal cues to deceit may bedifferentially useful for credibility assessment purposes inDECEPTIVE LANGUAGE 25the context of a confrontive, negative interrogationsituation (cf. Stiff & Miller, 1986).The superficial—encoding strategy may be made moreeffective by planning its implementation prior to theinterview. As noted by DePaulo, Lanier, and Davis (1983),lies in the laboratory are usually characterized by lowlevels of cognitive planning (including the story to bepresented and strategies for avoiding detection), yieldingfindings that may not be generalizable to forensic contexts.Most experimental research on deceptive communications hasexamined the nature of the spontaneous, unprepared lie.However, a guilty person usually has the opportunity toprepare for an interrogation or courtroom testimony byinventing and rehearsing his/her story (Carpenter, 1990).He/she has anticipated lying and, as the critical questionsare asked, provides lies that are rehearsed and that he/sheis mentally prepared to tell. Additionally, as noted byBuckwalter (1983), fabrication and deception are predominantin the daily lives of many criminals for survival on thestreet and to avoid incarceration so they may be quitepracticed at and proficient at fabricating information. Indevising generalizable experimental research addressingstatement analysis it is apparent that this preparationfactor must be considered, although few studies have doneso.O’Hair, Cody, and McLaughlin (1981) investigateddeception of factual information for both prepared andDECEPTIVE LANGUAGE 26spontaneous lies. Truthful and deceptive accounts werecompared on paralinguistic measures and body leakage clues.Subjects who had anticipated lying showed shorter responselatencies and less postural shifting than truth tellers.Secondly, the prepared liars showed higher head-noddingrates, lower laugh/smile rates, and shorter messagedurations. This is consistent with Zuckerman et al. (1981)who argued that efforts to conceal deceptions by controlledrestraint of behavior can be counterproductive because thecontrolled behavior now appears overcontrolled andunnatural. Additionally, not all behavioral areas can becontrolled equally resulting in leakage from differentcommunication channels.Alonso-Quecuty (1990) showed subjects a film depictinga physical assault on a woman by a man in the context of anabortion dispute. All subjects provided both an accurateaccount of the event (in which the aggressor was convicted)and an account in which information was falsified (the manwas acquitted). Half of the subjects provided theiraccounts immediately following the film while the otherswere given ten minutes to prepare their accounts. Thefalsified accounts contained less words in the immediatestatements but more words in the delayed statements leadingthe author to conclude that “the length of the statementsmay be a good detector of deception when the individual hasdisposed of the necessary time to elaborate them.” Sincethis is generally the situation in criminal cases, theDECEPTIVE LANGUAGE 27length of the statement may be a useful consideration inforensic contexts. As well, Alonso-Quecuty’s (1990) findingmay help resolve the inconsistencies described earlierregarding the research addressing the accountlength/veracity relationship. Notably, in the AlonsoQuecuty (1990) study the delayed accounts were approximatelytwice as long as the immediate accounts for both thetruthful and deceptive statements, again illustrating theeffect of preparation on testimony.Kohnken (1985) has investigated the utility of theinformation processing approach in detecting the deceptionof eyewitnesses. He contended that providing a deceptiveeyewitness testimonial represents a more difficult cognitivetask than giving an honest account and that with cognitiveload symptom substitutions should appear. Based on previousstudies he concluded that cognitive load can result in anincrease in speech hesitations (cf. Macdonald & Michaud,1987, p. 36), a reduction of the informational value ofapplied words, and an increase in repetitions. To test hisinformation processing model, Kohnken devised an experimentin which subjects viewed a film depicting a theft. Prior tointerviews, participants were divided into a truthfulcondition (instructed to tell the truth about the film), adeception condition (instructed to implicate the wrongperson as the thief and offered an LP record if the lie wassuccessful), or a hearsay condition (subjects did not viewthe film but heard a verbal report about its contents andDECEPTIVE LMGUAGE 28were asked to convince the interviewer that they had viewedit; if successful they would receive an LP record).Thirteen dependent variables were examined including speechrate, filled pauses, self—reflections, self—corrections,type token ratio (TTR; see below), and mean word length.Results indicated that the groups could be reliablydistinguished by most criteria and a multiple discriminantequation was derived to efficiently differentiate the groups(the TTR exhibited the highest discriminating power). Thisprompted Kohnken (1985) to conclude that his hypothesis hadbeen accurate; telling lies is a more complex task thantelling the truth and this increased task difficulty isreflected in speech behavior. Thus, an informationprocessing approach appears to be appropriate ininvestigating speech and deception in witnesses. Thepresent work extends this approach to the deceptive languageof (mock) criminal suspects which has not yet been examined.Another cognitive approach to statement analysisconsiders the complexity of the verbal responses in theinterrogation context. Osgood (1960) argued that “behaviorunder increased drive, including encoding behavior, shouldbecome more stereotyped — the alternatives selected at allchoice points should tend to be the most familiar, the mostpracticed, and the most expected.” According to this view,a person providing testimony under greater motivation toappear credible should display lesser lexical diversity.Deviations in lexical diversity have proven useful inDECEPTIVE LANGUAGE 29identifying deceptions in the accounts of suspects. Lexicaldiversity is usually measured by calculating the number ofdistinct words (“types”) over the total number of words(“tokens”) in a statement, referred to as the “type—tokenratio” (TTR) in Forensic Psychology. For example, inLincoln’s “government of the people, by the people, for thepeople” the six types and ten tokens yield a TTR of .60whereas the statement “the boy left the store after hispurchase of milk” yields a .90 TTR (nine types, ten tokens).In other words, a high TTR suggests that a personcommunicating employs a broad vocabulary, whereas arelatively low one reflects a communicator’s preference forstereotypic language. The TTR has long been employed bylinguists as a descriptor of language behavior but has onlyrecently been applied to establishing veracity for lawenforcement purposes. Hollien (1990) posited that thehigher one’s motivation, the lower one’s TTR. Certainpsycholinguists have argued that deceivers show higher TTR5than those of people uttering truthful statements.Carpenter (1990) argued that the TTR is an excellentindex of an individual’s internal psychological condition.He found that when adults shift from being truthful to lyingthere are some clearly discernible changes in their TTR5.Carpenter provided several case studies in which thesuspect’s TTR had increased significantly from his/her meanTTR at points in the account later proven deceptive. AsCarpenter notes, knowledge of the crime allows guiltyDECEPTIVE LANGUAGE 30suspects to know which of their statements are false orself—incriminating and must be phrased with greater cautionduring interrogation or courtroom testimony. Although notdetectable to listeners, the additional time needed tophrase such statements results in a higher frequency ofcomparatively uncommon words occurring in particularcomponents of a statement. Similarly, when hidden knowledgeof events constrains a suspect at some points duringtestimony, the additional split-seconds needed to formulatethose sentences provides time for more different words,relatively uncommon to the individual, to be spoken.Carpenter noted that in criminal investigationssuspects attempt to portray credibility and thus many oftheir statements are truthful. It may be necessary tosystematically examine changes within an individual’sdiscourse to obtain meaningful information regardingveracity. Truthful statements should yield TTRs close tothe person’s mean TTR whereas significantly higher TTRs areindicative of deception.Thus far, there is insufficient evidence to concludethat the TTR has utility in detecting deception in criminalcases. Empirical studies have supported the assertion thatstatements uttered by people experiencing apprehension orcaution about possible adverse reactions from listeners havehigher TTR5 than those with no threat (e.g., Dulaney, 1982).Consequently, it is not known whether a high TTR in aninterrogation is necessarily indicative of deception; itDECEPTIVE LANGUAGE 31could result from anxiety unrelated to guilt (see Inbau,Reid, & Buckley, 1986). It is possible, as Carpenter hasasserted, that examining TTR variations within a suspect’sstatement could indicate deceptive aspects of the statement.Dulaney (1982) placed participants in a situation designedto elicit truthful and untruthful statements. By employingcomputer programs, he found that when falsifying informationthe participants used significantly fewer words, fewerunique words and larger TTRs, smaller perceptual—cognitiveactivity measures, and fewer past tense verb forms.Considerations of lexical diversity have proven to haveutilitarian value in several contexts. For example, thelexical diversity of different messages have been analyzedto determine whether they had a common source or to specifycommunication origin in a process referred to as“stylometry” (e.g., see Arens & Meadow, 1957; Gudjonsson,1992; Miron, 1984; Morton, 1978). Such methods have alsobeen used to successfully determine whether threats weregenuine or hoaxes. For example, it was successfullypredicted that the Symbionese Army, who kidnapped PattyHearst, would die in a suicidal confrontation with lawenforcement officers (Miron & Pasquale, 1978).DECEPTIVE LANGUAGE 32Chapter 4THE PROPOSALS OF AVINOAM SAPIRCertain principles of statement analysis have beenelucidated and formalized by Avinoam Sapir, an ex-Israelipolice polygrapher. Sapir teaches these concepts to lawenforcement personnel in a seminar entitled ScientificContent Analysis (SCAN). Although some criminalinvestigators have adapted Sapir’s concepts, there has beenno scientific assessment of their worth (Lesce, 1990). Itis not known what aspects of SCAN best distinguish true andfalse statements by suspects or the overall utility of thetechnique. Following are some of Sapir’s (generallyunsubstantiated) principles which were tested in the presentstudy:4.1 Ratio of Component Elements. Sapir argues that theratio of three main elements of a suspect’s statement can beindicative of deception. In general, statements arecomposed of three elements — an introduction, a centralissue, and a conclusion. According to Sapir, peopleintending to make a truthful statement will generallyrecount a lengthier conclusion than introduction whilepeople making a deceptive statement show the oppositepattern. Sapir argues that because humans tend to beegocentric they evaluate events primarily in terms of howthey are affected by them. Thus, truthful people are notDECEPTIVE LANGUAGE 33concerned with being perceived as credible but instead focussolely on the event’s personal impact. Witnesses to a fatalcar accident may state as much information about how thewreck halted traffic and how waiting for the ambulance madethem late for work as the wreck itself. In contrast,deceptive people are primarily concerned with appearingcredible. Sapir argues that they devote a great amount oftime describing how it was logical and rational for them tohave been where they were at the time of the accident.Since they did not personally experience the incident, thereis little affect present. Accordingly, they truncate theirconclusions. Sapir asserts that this factor singularlyidentifies deceptive statements with roughly 85% accuracy.4.2 Unnecessary Connectors. Unnecessary connectors includewords such as “started”, “continued”, “completed”, “after”,“afterwards”, “the next thing I remember”, “then”, and “andthen after that” which are employed by a speaker toselectively pass over incriminating information. Sapirargues that a person who is falsifying information willemploy a higher frequency of unnecessary connectors. Themost commonly used connector is “then after that” which isused to buy time for “mental processing”; a recent eventbeing honestly recalled should require little suchprocessing. Sapir asserts that connectors often indicatethat details have been deliberately edited. For example, “Ihad dinner and I went to the store” indicates that nothingDECEPTIVE LANGUAGE 34happened in the interim whereas “I had dinner and then afterthat I went to the store” is an intimation that something ofsignificance may have transpired in the hiatus. Anincreased frequency of the use of connectors should arousesuspicion and lead to further investigation into theparticular temporal segment being referred to.4.3 Use of Pronouns. According to Sapir, the use ofpronouns during speech also reveals a considerable amount ofsubtle information. Sapir argues that any deviation fromthe first person, past tense, or active voice in describingan event is significant in indicating deception. Thisperspective coincides with that of Macdonald and Michaud(1987) who asserted that a shift by a suspect from talkingin the first person to the third person indicates that thesuspect wants to distance himself from the interrogation andthe crime. Recollections from genuine memory usually beginwith the pronoun “I” whereas someone making a falsestatement is more likely to employ the pronoun “you”. Forexample, a person making an honest statement might remark “Icould see the blood dripping from the back of the truck”whereas a deceptive person would likely say “You could seethe blood dripping from the back of the truck.”Nonetheless, as discussed previously, Schooler, Clark, andLoftus (1988) have found that true memory accounts for anevent contain the pronoun “I” at a lower frequency thanfalse memory accounts (using the Loftus posteventDECEPTIVE LANGUAGE 35misinformation paradigm). They also found that false memoryaccounts included more “verbal hedges” (e.g., “It seems tome” or “I believe”) which was a replication of an earlierfinding (see Schooler, Gerhard, & Loftus, 1986), explicatedin the framework of Johnson and Ray’s (1981) theory ofreality monitoring.DECEPTIVE LANGUAGE 36Chapter 5SUMMARY OF THEORETICAL FRAMEWORK AND THE PRESENT STUDYIn the preceding discussion it has been argued that, inseeking principles of verbal deception in the interrogationcontext, it is necessary to consider theories from a varietyof perspectives. Approaches emphasizing memory accuracysuggest that systematic differences in several qualitativeand quantitative aspects of memory narratives can reliablyindicate their veracity (i.e., the Undeutsch Hypothesis,Reality Monitoring). Perspectives on deception considerdifferences in truthful and untruthful statements due toemotional, motivational, and cognitive factors unique to theinterrogation context. Finally, the assertions of an ex—Israeli polygraph operator and police officer named AvinoamSapir are discussed. Sapir argues that the ratio ofcomponent elements, the presence of unnecessary connectors,and deviations in pronoun usage in interrogation transcriptsare indicative of deception. Together, these approachesform an integrated model of statement analysis which can bedeveloped for use in certain criminal investigations.This thesis represents an empirical evaluation of themeasures discussed previously. In this investigation, theaccounts of truthful, partially deceptive, and completelydeceptive mock suspects were compared on eighteen verbalattributes derived from these approaches. The methodologyand dependent measures employed reflect a holistic approachDECEPTIVE LANGUAGE 37to statement analysis considering the emotional—motivationaland cognitive processes associated with criminalinterrogations.The major circumscribed objectives of this thesisresearch were: (1)to investigate the utility of the variousknown verbal correlates of deception for assessing thecredibility of criminal suspects (2)to investigate theutility of some previously unresearched statement analysisprinciples in assessing the credibility of criminalsuspects. Minor goals of this study included: (1)tointroduce a methodological variation on the traditional mocktheft paradigm aimed at maximizing experimental realism andecological validity (2) testing the hypothesis thatinterrogation probes will result in differential verbalbehavior by truthful and deceptive criminal suspects. Theultimate goal of this and future research is to develop amodel of credibility assessment for use with certaincriminal suspects based on verbal indicators of deception.DECEPTIVE LANGUAGE 38Chapter 6METHOD6.1 Participants. Sixty participants (N=60) were recruitedfrom the University of British Columbia undergraduatepopulation under the premise of participating in anexperiment addressing security effectiveness. The incentivefor participating was five dollars for a one hour sessionwith the “possibility of earning more.” Although it wasoriginally anticipated that equal numbers of males andfemales would be used, this proved impossible given thatmostly women responded to my advertisement (forty-fourwomen, sixteen men). The first language of all participantswas English.6.2 Procedure. Upon arriving at the laboratory, eachparticipant was asked to sign a consent form. Next, theywere falsely informed that the Psychology Department hadjust hired an un-uniformed security guard because of arecent rash of thefts in the building. Further, unbeknownstto the guard, our laboratory group had been contracted thetask of testing the effectiveness of this new measure. Inthree of the four conditions, the participants’ task was toproceed to a specified office in the Psychology Departmentand attempt to locate a one hundred dollar bill hidden in afolder on a bookshelf in the room. They were given tenminutes to obtain and pocket the cash, restore the room toDECEPTIVE LANGUAGE 39its former condition, and return to the laboratory. Whileacquiring the cash they were told to be as careful aspossible to escape detection from the guard. Participantsin Condition 1 were asked to obtain a folder (the samefolder used for the theft conditions but without the money)from the same location for a professor.Each participant was provided a key to the office anddirections to proceed to its location. It was reiteratedjust prior to their exit that they would have only tenminutes to return to the laboratory and to be very careful.Upon returning to the laboratory, participants gaveeither the money (Conditions 2, 3, and 4) or the folder(Condition 1) to the experimenter. Next, they were informedthat the second part of our contract with the securitycompany was to test the prowess of their interviewers atinterrogating suspects and establishing which are guilty andinnocent. Thus, all participants were told that they wouldbe interrogated by someone who was aware that he/she wouldbe questioning both truthful and deceptive suspects andwhose agenda was to establish the truth based solely on theinterview. Further, this interviewer had been informed thathe/she may hear some truthful confessions, falseconfessions, truthful denials, and false denials.6.3 Instructional Manipulation. The main experimentalmanipulation in this study involved the directions providedto participants preceding the interrogation. ParticipantsDECEPTIVE LANGUAGE 40were randomly assigned to one of four experimentalconditions (n=15 per group: 11 females, 4 males): 1.Truthful alibi condition 2. Partial deception condition 3.False alibi condition 4. Truthful confession condition. Allparticipants were given fifteen minutes to prepare theiraccounts before entering the interrogation room. Theparticipants from each condition were provided the followinginstructions:1. Truthful Alibi condition: These participants had beenassigned an innocuous task in the crime room (to obtain abrown folder for a professor). Prior to the interview theywere informed that they were suspects in the theft of a onehundred dollar bill and would be interrogated as such. Theywere told to simply relate the complete truth about anythingthat was asked by the interviewer (i.e., relate theirtruthful alibi). If they were able to convince theinterviewer that they were reporting truthfully they wouldreceive an extra five dollars from the experimenter.2. Partial Deception condition: These participants wereinstructed to provide the interviewer with an accountcontaining some truth and some falsehood. They were askedto relate a truthful account with the exception that thetheft of money was to be replaced by an unsuccessful attemptto retrieve “a folder” belonging to a specified professor.These participants were asked to deny any knowledge of moneyexisting in the room. Further, they were informed that ifDECEPTIVE LANGUAGE 41the interviewer believed their account they would receive anextra five dollars at the culmination of the interview.3. Complete Deception (false alibi) condition: Theseparticipants were asked to provide the interviewer with acompletely falsified account in the form of an alibi. Theywere instructed to create a plausible story accounting fortheir whereabouts and actions in the ten minute time periodduring which the money was stolen. Additionally, under nocircumstance should they admit their deception. Again,participants were told that if the interviewer believed thatthey were recounting a truthful account, they would receivean extra five dollars at the end of the interrogation.4. Truthful Confession condition: These participants wereasked to provide the interrogator with a completely truthfulaccount admitting their involvement in the theft ifrequested. Further, they were informed that theinterrogators were expecting some false confessions. Again,the monetary incentive for successfully convincing theinterviewer was offered.6.4 Interviewing. Each participant was initiallyquestioned in the fashion of the Step-Wise Interview (seeYuille, Hunter, Joffe, & Zaparniuk, 1993). Basically, thissemi—standardized interview procedure involved eliciting afree recall account in which the participant was asked tostate everything he/she could recall from the time he/sheexited the laboratory until the time he/she returned. ThisDECEPTIVE LANGUAGE 42was followed by a series of more directive questionsintended to clarify and elaborate the information providedin the free recall account. These questions consistedmainly of open questions (e.g., “You mentioned seeing adesk. Can you tell me more about that?”) but occasionalspecific questions (for a more complete description of theinterrogation protocol refer to Appendix B).Interrogation ProbesThe Step—Wise Interview was ensued by a series ofinterrogation probes designed to simulate a policeinterrogation and test participants’ responses to negativefeedback. This consisted of three statements expressingdisbelief by the interviewer: (1)”Do you have any knowledgeof the money that is missing from Room 2408?” (2)”Frankly,I’m a little skeptical about your explanation. Are youbeing completely honest?” (3)”Come on now, it’s reallyobvious that you are lying to me. Why don’t you just admitit?”. All interviews were video- and audiotaped.Five senior or graduated Psychology students acted asinterviewers in this study. They were trained in this semi-structured interview via reading materials (see Appendix B)and practice with critical feedback from the experimenterwho had previous training and experience with the technique;e.g., Porter, Yuille, & Bent, in press). They initiallyreceived two one and a half hour training sessions in whichthey observed a sample interview, received an audiotapecontaining the sample interview, and the handout detailingDECEPTIVE LANGUAGE 43the interview steps. Working in pairs, the interviewerssubsequently taped their practice interviews and providedthem to the experimenter who gave exhaustive feedback. Thisprocess continued until the quality of the interviews wasconsidered acceptably high.With the culmination of the interviews/interrogations,all participants were thanked and paid the monetary bonus.They were then completely debriefed regarding the purpose ofthe research and the reason for the experimenter’s deception(see Appendix A). They were also implored not to discussthe nature of this study with acquaintances.6.5 Transcribing. All interviews were transcribed verbatimfrom the audiotapes onto computer diskettes. All utterancesmade by the participants were recorded in the transcripts.In those cases where a participant’s speech was inaudible onthe audiotape, the videotaped interview was examined. Ifthe utterance in question was still considered ambiguous orinaudible by the transcriber, it was recorded as “inaudible”on the transcript. Finally, hard copies of all interviewswere generated.6.6 Dependent Variables and Scoring Proceduresa. Amount of Detail Reported: The accounts were subjected toan evaluation procedure based on the protocol developed byYuille and his colleagues (e.g., see Yuille & Cutshall,1986; Yuille & Porter, 1993). The reports were partitionedDECEPTIVE LANGUAGE 44into single units of information. All details were scoredas correct, incorrect, or unclassifiable. This was done bycomparing the details to the videotape of the event for eachparticipant. Each detail received one point if it containeda specific unit of information. For example, the sentence“I leaned on the beige desk” contains one action detail andtwo descriptive details. Because previous research withadults has typically found similar patterns with action-oriented and descriptive details (e.g., Yuille & Tollestrup,1990; Yuille, Tollestrup, Marxsen, & Porter, under review;Yuille, Davies, Gibling, Marxsen, & Porter, in press)possible differences were not tested in this study.Information that is vague or nonspecific was given one halfpoint (e.g., “The room was large”). A detail was consideredunclassifiable if it may have occurred but verification wasnot possible (e.g., “I felt a cool breeze”). Theinformation deemed unclassifiable was not be entered intothis analysis. This variable was frequency coded in lieu ofthe more subjective SVA rating process usually used inforensic practice.b. Laconic—Garrulous Continuum: The length of the accountswas measured by counting the number of words in eachaccount. For this variable, all filled pauses were deletedand a spelling check generated prior to a computer wordcount.DECEPTIVE LANGUAGE 45c. Lexical Diversity (Type-token ratio): An analysis oflexical diversity as measured by the TTR was conducted withthe use of a computer program developed for this purpose.d. Frequency of Verbal Hedges: This criterion was measuredusing Loftus’ definition of hedging (“I believe”, “It seemsto me”, “I think”, “I figure”). A computer search wasconducted for each of these phrases and the frequenciesrecorded.e. Frequency of Filled Pauses: The number of filled pauseswere tallied by recording the number of times theparticipant used one of the following sounds: “ummm”,“uhhh”, “welil”, “hnuum”, etc. The transcriber had beeninstructed to record filled pauses with the one of thesefour items which sounded most like the utterance. Thisallowed a computer search to later be conducted generatingfrequencies for each of these items.f. Frequency of Unnecessary Connectors: The frequency ofunnecessary connectors were tallied. These include thewords “after”, “afterwards”, “next”, “the next thing Iremember”, “then”, and “then after that”. These phrases hadto be employed during a description of a participant’sactions related to the event in question (i.e., from thetime the participant exited the lab until the time he/shereturned). Hence, a description of the initial interactionwith the experimenter using a “and then after that” wouldnot constitute an unnecessary connector. A computer searchwas generated for the six above phrases and an examinationDECEPTIVE LANGUAGE 46of the context of the utterance was made by the coder. Ifthe connector was employed in describing actions pertinentto the central event it was tallied.g. Frequency of Pronoun Usage Deviation: All deviations fromthe first person (e.g., “You could see a bunch of books onthe desk” instead of “I could see a bunch of books on thedesk”) and past tense (e.g., “I’m going into the office”) indescribing the event were recorded.h. Number of Self—References: All self—references weretallied (i.e., I, me, my, mine, myself) through a computersearch.i. Ratio of Component Elements: A coding system was devisedto trichotomize the accounts into three elements — anintroduction, a central issue, and a conclusion. Coding wasonly conducted on the initial, uninterrupted free recallaccount. The number of words employed by each participantto describe events up to the point of entering the “crimeroom” were counted and considered the “introduction” to thecentral issue. The number of words used to describe eventssubsequent to exiting this room (up to the culmination ofthe narrative) were counted and labelled the “conclusion”.The ratio of the number of words in the conclusion over thenumber of words in the introduction was calculated for eachparticipant. Data from participants in Condition 3 (falsealibi) were not coded for the present analysis because theevents described generally did not include the crime roomand were thus were not comparable to the other conditionsDECEPTIVE LPNGUAGE 47(recall that each of the other three groups were describingepisodes involving acquiring or attempting to acquire afolder from this office).Two raters with a recent extensive three-day trainingin SVA scoring by a recognized expert in the techniquejudged the presence of each of the following SVA criteriaand assigned a rating (0 - 3 [strongly absent - stronglypresent]) or tallied the number of pieces of evidence foreach criterion contained in the accounts. Where possiblethe more objective method of counting specific pieces ofevidence for each criterion was employed (note that thefollowing SVA criteria were not applicable to the event usedin the present study: reproduction of conversation,attributions of the perpetrator’s mental state, pardoningthe perpetrator, details characteristic of the offense,contextual embedding, descriptions of interactions,accurately reported details misunderstood, self—deprecation;for a more detailed explication see Steller, 1989; Steller &Kohnken, 1989; Yuille, 1990).j. Coherence: Basically, this criterion is present when thevarious parts of the account fit together in a coherent andconsistent fashion. A rating of 0 — 3 was assigned to eachaccount.k. Spontaneous Reproduction: The actual recounting of theevent should occur in a spontaneous, unstructured fashion.DECEPTIVE LANGUAGE 48Rigid structure would result in a low rating on thiscriterion (0 - 3).1. Unexpected Complications During the Incident: Thisrefers to the number of unexpected occurrences complicatingthe main task of the participant as reported in the recallaccount. This did not include being unable to locate thefolder (participants in Condition 2 had been instructed togive this falsehood).m. Presence of Unusual Details: This criterion refers tothe spontaneous reporting of details concerning the eventthat are realistic but unusual (e.g., reporting seeing abook by one of his/her favorite authors) to the coder. Thenumber of such details was tallied for each account.n. Presence of Peripheral Details: This refers to a reportcontaining details that are not directly related to thetarget event (e.g., temperature of the room). A rating ofthe 0—3 was assigned to each account from Conditions 1, 2,and 4. Condition 3 was omitted from this analysis becausepreliminary trials indicated that it would be very difficultto define details “peripheral” to the central event. Forexample, a few participants described simply wanderingthrough the building looking around as their alibi.o. Related External Associations: This criterion refers tospontaneous references to an incident related to the eventbut did not occur within the event (e.g., a participantmight report that the office reminded him/her of Dr.DECEPTIVE LANGUAGE 49Yuille’s office). The number of such details was talliedfor each account.p. Accounts of Subjective Mental State: This criterionrefers to the number of times a participant refers tohis/her thoughts, intentions, or emotions during the event.q. Spontaneous Corrections: This refers to the number oftimes the participant corrects him/herself in the course ofrecounting the event.r. Admitting Lack of Memory: This criterion refers to thenumber of times the participant admits that he/she cannotrecall an aspect of the event (generally in response todirective questions from the interviewer). This includesresponses to inquiries such as “Is there anything else youcan remember?”. Multiple admissions regarding the sameaspect of the event were tallied as a single admission.DECEPTIVE LANGUAGE 50Chapter 7RESULTSAll reported statistical analyses were conducted withSYSTAT 5.01 software.7.1 Reliability Checks. Inter-coder reliability checkswere conducted for all dependent variables. For variablescoded by computer tallying, test-retest reliability wasmeasured by conducting a second computer search (to controlfor any input errors in the initial search). For all othervariables, eight interviews (two from each condition) wererandomly selected for dual coding (see Kohnken, 1985). Forthe SVA variables two coders recently trained in SVA scoredeach of these interviews. Pearson’s product momentcorrelations were calculated for each variable followed bybetween—subjects t—tests to ensure that the averagefrequencies for the two coders did not differ (see Bordens &Abbott, 1991, p. 155). None of these t-tests showedsignificant mean differences between the two coders. ThePearson r reliability scores revealed that, with theexception of one criterion, inter—coder reliability wasacceptable (r > .80) as found in most previous studiesexamining SVA (e.g., Steller, 1989; Zaparniuk, Yuille, &Taylor, under review). As expected, all variables scoredvia computer—based coding showed 100% test—retestreliability (Variables b, c, d, e, f, h) including Variablef (Unnecessary Connectors) which contained a minor elementDECEPTIVE LANGUAGE 51of coder judgment. The 100% agreement for Variable q(Spontaneous Corrections) is misleading. As indicated inTable 1, the mean frequency of this variable in the accountsfrom all conditions was very low. A visual inspection ofthe data from the sample interviews used for the reliabilitycheck revealed that in all cases the coders found noevidence for this variable resulting in all “0” scores andthus perfect agreement. The one variable for whichagreement was not acceptable was Variable p (Accounts ofSubjective Mental State) showing r = Preliminary Analyses. Initially (prior to coding forthe other variables), it was necessary to compare the meanaccount lengths (Variable b) in the four conditions. Ifthere were no differences, coders could proceed with theprocess of examining frequencies for various variables(e.g., frequency of self—references). If there weredisparities, frequency coding would not have beenappropriate and an alternative approach would have beenselected (e.g., Likert scale). A one—way ANOVA wasconducted on Variable b. As Table 1 suggests (see Appendixfor corresponding standard deviations), no significantdifferences in the number of words reported were found amongthe four groups, (F[3, 56] = 0.264, 2>.05). Thus, thefrequency coding approach was validated.For Variables “i” (ratio of component elements) and “n”(presence of peripheral details) only three of the fourDECEPTIVE LANGUAGE 52conditions were examined. These data were not included inthe MANOVA. Instead, a one-way three-group ANOVA wasconducted for each of the two variables. As Table 1indicates no mean differences were detected for eitherVariable “i” (F[2,42] = 0.23, R>•05) or Variable “n” (F[2,42] = 1.02, R).05).DECEPTIVE LANGUAGE 53Table 1.Mean scores for the four conditions on all dependentvariables.EXPERIMENTAL CONDITIONDEPENDENTVARIABLE!A. PERSPECTIVES ON1 2 3 4Truthful Partial False TruthfulAlibi Deception Alibi ConfessionMEMORY ACCURACYDetalls*** SLJ.51 Jb.03 31.34 55.10Words 828.62 751.91 861.14 901.10Coherence** 1.73 0.80 0.87 1.93Sp. Reproduction 1.60 1.27 1.33 1.93Un. Complications 0.73 1.00 0.87 1.00Un. Details 1.47 1.20 1.00 1.73Pen. Details 1.00 1.53 — 1.13R.E. Associations 0.83 1.62 2.53 0.60S.M. State 1.27 2.13 1.13 1.33Spont. Corrections 0.33 0.40 0.20 0.60A.L. of Memory* 4.73 1.87 2.13 4.40B. PERSPECTIVES ON DECEPTIONVerbal Hedges 3.33 4.53 4.80Filled Pauses 21.15 23.90 27.62 27.33Self—references 70.70 61.91 83.32 83.73Type—token Ratio 0.28 0.29 0.29 0.28C. SAPIR’S CRITERIAUn. Connectors 6.73 5.73 5.93 6.07Pron. Deviations 7.07 9.33 8.67 6.33Comp. Elements 0.53 0.51 — 0.43‘Many criteria tall under multiple categories.*p<.O5**p<.OO1***p<. 0001DECEPTIVE LANGUAGE 547.3 Multivariate Analysis. The design of the experiment isa four group (levels of deception) between—subjects designwith eighteen dependent variables. With this number ofvariables, simple analyses of variances (ANOVAs) couldresult in an elevated Type 1 error rate. Consequently, asin previous SVA studies (e.g., Joffe, 1992; Landry &Brigham, 1992), a one—way multivariate analysis of variance(MANOVA) was conducted using the remaining fifteen dependentvariables. There was a significant main effect forcondition (Wilk’s laauda = 0.13; F[48,122] 2.51, 2<.0001).Subsequently, univariate F—tests were conducted for each ofthe criteria. These were succeeded by multiple comparisonsemploying the Tukey method in cases where significant groupdifferences were detected.Criteria Showing Significant Effect.Variable a (Amount of Detail Reported). Bartlett’stest for homogeneity of group variances indicatedhomogeneity (X[3] = 2.42, >.OS). As Figure 1 indicates,there was a highly significant main effect for condition onthe number of details reported, F(3,56) = 8.48, 2<.0001.Followup multiple comparisons revealed that participants inthe truthful alibi condition (Condition 1) reported moredetails (M = 50.5) than participants in the partialdeception condition (Condition 2; M = 36.0, Q<.05) and thefalse alibi condition (Condition 3; M = 31.3, R<.01).Similarly, participants in the truthful confession groupDECEPTIVE LANGUAGE 55(Condition 4) provided more details (M = 55.1) thanparticipants in Condition 2 (2<.O1) and Condition 3(2<.OO1). Mean scores from the two truthful conditions (1and 4) did not differ nor did mean scores from the deceptiveconditions (2 and 3).Variable j (Coherence). Bartlett’s test indicatedhomogeneity of group variances (X[3] = 0.71, R>•05)• Therewas a significant main effect for condition on ratings ofcoherence, F(3,56) = 6.22, <.0O1 as revealed in Figure 2.Subsequent Tukey comparisons revealed that accounts byCondition 1 participants (truthful alibis) were rated asbeing more than twice as coherent (M = 1.73) as Condition 2(partial deception) accounts (M = 0.80, <.05) and Condition3 (false alibi) accounts (M = 0.87, <.05). Similarly,reports given by Condition 4 (truthful confession)participants were rated as more than twice as coherent (M =1.93) as those provided by participants in both Condition 2(R<01) and Condition 3 (2<.05). Again, mean ratings forcoherence did not differ between the truthful conditions (1and 4) or the deceptive conditions (2 and 3).DECEPTIVE LANGUAGEFigure 1. Number of details reported across conditions.5655.160-50-40-30-20-100-50.536EXPERIMENTAL CONDITIONLii. Truthful AlibiLI 2. Partial Deception3. False AlibiR4. Truthful ConfessionCONDITION (1-4)DECEPTIVE LANGUAGE 570.5Figure 2. Coherence ratings across conditions.EXPERIMENTAL CONDITIONEli. Truthful AlibiEl 2. Partial Deception3. False Alibi4. Truthful Confession2.521 510CONDITION (1-4)DECEPTIVE LANGUAGE 58Variable r (Admitting Lack of Memory). Bartlett’s testindicated homogeneity of group variances (X[3] = 2.73,R>•05). There was a significant main effect for conditionon Variable r, F(3,56) = 3.00, R<•05 (see Figure 3). Tukeytests indicated that Condition 1 participants admitted tolacking memory for an aspect of the target event morefrequently (M = 4.73) than did Condition 2 participants (M =1.87, 2<.05). No other pairwise differences attainedstatistical significance although absolute mean differencesbetween the truthful groups and deceptive groups were large(see Table 1).7.4 Relative Discriminating Ability of Criteria ShowingSignificant Effects. A discriminant analysis was carriedout to assess the relative contribution of the threecriteria showing significant effects in the MANOVA todiscriminations between the accounts of truthful anduntruthful suspects. Because no differences were foundbetween the truthful groups (Conditions 1 and 4) or thedeceptive groups (Conditions 2 and 3) on any criteria thedata were pooled for the discriminant analysis to examinethe ability of the criteria to globally distinguish truthfuland deceptive reports.As expected, the initial multivariate test statisticwas highly significant, reiterating the initial MANOVA(Wilk’s lamda = 0.54; F[3,56] = 15.75, 2<.0001). Similarly,the three univariate F-tests were all highly significantDECEPTIVE LANGUAGE543210Number of admissions to lack of memory acrossconditions.59EXPERIMENTAL CONDITIONL1i. Truthful Alibi2. Partial DeceptionFalse AlibiI 4. Truthful ConfessionCONDITION (1-4)Figure 3.DECEPTIVE LANGUAGE 60(2<.001). The analysis using the three criteria showingsignificant effects as predictors of membership in thetruthful and deceptive conditions yielded a highlysignificant discriminant function (x2) = 34.56, 2<.0001).As can be seen from the canonical coefficients presented inTable 2, this discriminant function suggests that the threecriteria examined distinguished the truthful from deceptiveaccounts about equally.The discriminant function correctly classified 47 ofthe total sample of 60 (30 truthful and 30 deceptive)accounts according to group membership. Classificationsuccess was approximately equal for both groups.Specifically, 23 of the 30 (76.7%) truthful accounts and 24of the 30 (80.0%) deceptive accounts were correctlyclassified according to the discriminant function.7.5 Followup Exploratory Analyses. Although the MANOVAindicated that the truthful and deceptive groups differed onthree criteria, it was not clear at what point in theinterrogations these differences were emerging. Forexample, in the partially deceptive reports it was possiblethat the deception clues were emerging chiefly before thelying behavior occurred (typically near the end of the freerecall accounts) or following the initial lying. To answerthis question, 2 X 2 (truthful/deceptive condition X phaseof interrogation) between—within factorial analyses wereconducted using the data from Condition 1 (truthful alibi)Li0 Li‘-3 H LiI:-’C)Cl)Hi‘-3p)0CtI--0-•C)U)I-1U)CDbC)C)I-’11CDIIC)hiCDC)oCDhiP)HI-’H-HiI-Ik)C)0Q..HU)H-•C)CDHiH•U)iLiCDN-CDHiioPiCDQ..CtH-CthiCtci-C)H-HippHciII-’U)H-U)U)C)C)0I—’hiciH-I--H-U)U)CD00H-U)HHiU)H-0iohI.)ci-0CthiHiU)CDcici-C)—I-IHiH C-I.C)I-CDCtIII-i0U)C)CD0C)oCDF-iACtC)• 0<U)I—0CDH0I0’UiO’Ui0C)H0iI-’t.iCDC)HiOUiHi0i-HiC)C)CtU)CtCDCDci-HiU)HiCD C) CtDECEPTIVE LANGUAGE 62and Condition 2 (partial deception). The previous multiplecomparisons indicated that these were the only two groupswho differed significantly on all three criteria.Nonetheless, the data for the criterion Coherence were notamenable to the present exploration in that they were in theform of global ratings rather than by frequency and couldnot be dichotomized into interrogation stages. For theother two criteria (number of details and lack of memoryadmissions) the data were coded in terms of when in theinterrogation they had been elicited - during the freerecall phase or during the questioning phase, constitutingthe repeated—measures aspect of the present analysis.The first 2 X 2 factorial was conducted on the numberof details reported. As expected, the between-subjectscomponent replicated the results of the MANOVA with moredetails being reported by truthful participants (F[1,28] =4.48, <.O5) overall. As well, there was a main effect ofphase of interview (F[1,28] = 15.18, R<.001) with moredetails being reported overall during free recall (M =26.45) than directive questioning (M = 17.48). Nointeraction effect was found (F[1,28J = 0.215, R>•05) withtruthful and deceptive suspects showing similar reportingpatterns. Specifically, truthful suspects reported a meanof 30.27 details during free recall (60.0% of total details)and 20.23 (40.0%) during directive questioning. Similarly,deceptive suspects reported a mean of 22.63 details duringfree recall (60.6% of total details) compared to 14.73DECEPTIVE LANGUAGE 63details (39.4%) during directive questioning. Thus, thefinding that deceptive suspects report fewer details seemsto reflect a general trend occurring throughout theinterrogation rather than being associated with a particulartime in the interview or resulting from a particularquestioning format.The 2 X 2 analysis on the criterion “lack of memoryadmissions” revealed findings comparable to the firstanalysis. However, in this case the analysis on thebetween-subjects variable was not significant (F[1,28J =2.89, 2>.05, likely because of reduced statistical power.There was a highly significant main effect of phase ofinterview (F[1,28] = 23.17, 2<.0001) with more lack ofmemory admissions occurring during the directive questioningphase of the interrogation (M = 1.78) than free recall (M =0.67). No interaction effect was found (F[1,28] = 0.532,R>•05) with truthful and deceptive suspects showing verysimilar reporting patterns. Specifically, truthful suspectsadmitted lacking memory on a mean of 0.87 occasions duringfree recall (28.9% of total admissions) and 2.13 (71.1%)occasions during directive questioning. Deceptive suspectsadmitted lacking memory on a mean of 0.47 occasions duringfree recall (25.0% of total admissions) compared to 1.40admissions during directive questioning (75%). As withnumber of details reported, the finding that deceptivesuspects are less willing to admit lacking memory for someDECEPTIVE LANGUAGE 64aspect of the event appears to reflect a general trendoccurring throughout the interrogation.The final exploratory analysis was conducted onVariable f (frequency of unnecessary connectors). Althoughno effects of condition were found in the MANOVA on thiscriterion, it was possible that an effect may have beenmasked by analyzing the accounts globally. According toSapir, the unnecessary connector will surface morefrequently as a lie is being related rather than at otherpoints in an account. To address this possibility, asubsample of data from the accounts of suspects inConditions 1 (truthful alibi) and 2 (partial deception) wasexamined. Only the portions of the accounts relating eventsin the crime room itself (in the case of Condition 2 thiswould represent the deceptive portion of the accounts) wereexamined for the frequency of unnecessary connectors. Itseemed likely that unnecessary connectors would appear morefrequently by deceptive suspects when recounting eventsoccurring in the crime room. It was found that deceptivesuspects used a mean of 1.47 unnecessary connectors whenrelating events in the crime room compared to a mean of 0.93by honest suspects. Although this indicates a considerablyhigher rate of unnecessary connector usage in the deceptiveaccounts (approximately 60%), an independent-samples t-testdid not yield significance (t[28] = 1.11, >.O5).DECEPTIVE LANGUAGE 657.6 Effects of Interrogation Probes. A second one—wayMANOVA was conducted using only the information in theaccounts following the three interrogation probes to examinewhether such probes affect the behavior of honest anddeceptive suspects differentially. Only the elevendependent variables which had been frequency—coded (and thuscould be divided into pre- and post-probes) were included inthis analysis (a, b, d, e, f, g, h, 1, o, p, and r). TheMANOVA did not yield significant results (Wilk’s lamda =0.51, F[33,136] = 1.05, p>.O5) although univariate F-testsdid indicate a similar pattern of results to the inclusiveanalysis with regard to Variable a (Amount of Detail), F(3,56) = 3.19, 2C.05.DECEPTIVE LANGUAGE 66Chapter 8DISCUSS IONFrom error to error one discovers the entiretruth. - Sigmund FreudThis thesis represented the first empirical test of thegeneral hypothesis that systematic differences exist betweenthe accounts of honest and deceptive suspects in theinterrogation context. This hypothesis was founded onseveral theoretical approaches subsumed under one of:perspectives on memory accuracy, perspectives on deception,or the proposals of Avinoam Sapir. Employing a methodologyaimed at maximizing external validity, it was demonstratedthat certain verbal clues, chiefly derived from the SVAtechnique, clearly distinguish truthful and deceptivereports. Dishonest suspects engaging in verisimilitude gaveless detailed and less coherent accounts. Further, theywere much less likely to admit failure to recall some aspectof the target event. These three criteria successfullyclassified 76.7% of the truthful accounts and 80.0% of thedeceptive reports, each contributing about equally to thediscrimination. As well, the differences found for thesecriteria seem to reflect global reporting patterns unrelatedto the phase of the interrogation.These results corroborate the findings of other recentinvestigations looking at the utility of statement analysis(e.g., Joffe, 1992; Landry & Brigham, 1992). In the past,there has been a marked over—reliance on nonverbal clues toDECEPTIVE LANGUAGE 67deception by both forensic researchers and practitioners(McCornack, 1992), with a virtual disregard for the possibleutility of verbal clues. This is especially problematicgiven that in a considerable proportion of forensic casesthe sole evidence available to investigators is the words ofthe complainant and the accused. The situation has beenimproving with the advancement of Statement ValidityAnalysis (SVA) with children and some empirical research onverbal clues with adults in the mid-1980s, most notablystudies by Zuckerman (e.g., Zuckerman et al., 1981; 1985)and Miller and colleagues (DeTurck et al., 1985; 1986).Nonetheless, there have been few previous attempts tointegrate the various theoretical perspectives in the areaor to devise research with the issue of external validity inmind (Miller & Stiff, 1993).One theoretical bent considered in this study wasfounded on memory accuracy. More specifically, theUndeutsch Hypothesis states that accurate memory accounts(reports based on true experience) will differ qualitativelyand quantitatively from inaccurate memory accounts (reportsbased on events not actually experienced). RealityMonitoring posits that accurate (“perceptual”) memoryaccounts contain more external sensorial descriptiveinformation whereas inaccurate accounts contain morereferences to cognitive processes and subjective information(and thus more self—references, verbal “hedges”, andreferences to cognitive processes).DECEPTIVE LPNGUAGE 68In sum, the results of the present study buttress thegeneral Undeutsch Hypothesis but do not support the use ofthe Reality Monitoring model for credibility assessment withsuspects. One of the more intriguing and potentiallyconsequential findings yielded in this thesis lends credenceto Undeutsch’s claim that accounts based on true experiencewill contain a richness of detail not characteristic offictitious accounts, Specifically, participants relating acompletely false alibi provided many fewer details thantruthful participants describing an event within the sametime frame, consistent with the Zuckerman et al. (1985)meta—analysis. However, the Undeutsch Hypothesis fails toilluminate the foundation of a related finding; thepartially deceptive suspects also reported fewer detailsthan the both truthful groups. The striking aspect of thisdiscovery is that these groups were, with a minor exception(folder/money), relating exactly the same occurrence (whichhad actually been experienced). Obviously, this differenceis not associated with the memories themselves (assumingguilty suspects didn’t repress information!) but insteadwith the act of providing deceptive information to aninterrogator, Similarly, accounts by partially deceptivesuspects were rated as less coherent than those by truthfulsuspects. These findings are more congruous withperspectives on deception (discussed later).As in other recent explorations (e.g., Joffe, 1992),the theory of Reality Monitoring did not garner support inDECEPTIVE LANGUAGE 69this study. In accordance with this theory, it had beenhypothesized that statements by participants in the truthfulconditions would contain less references to cognitiveoperations, fewer verbal hedges (“I believe”, “I think”),and less self—references than those of the participantsproviding a false alibi. None of these criteria showedsignificant effects. On the other hand, an examination ofthe absolute mean differences indicates that potentialeffects may have been masked by sizable variances (and henceelevated Type 2 error rate). For example, deceptiveparticipants provided approximately twice as many referencesto subjective mental state as both truthful conditions, aspredicted by Reality Monitoring. As well, the five accountswith the highest frequency of verbal hedges were bydeceptive suspects. In particular, one suspect providing afalse alibi employed twenty-three hedges, approximatelytwice as many as any other suspect in the study. Again,further research is needed to clear up this ambiguity. Ofcourse, it is possible that the Johnson/Schooler variablesare indeed effective discriminators of genuine (externallygenerated) and false (internally generated) memories forwhich they were originally formulated. All truthful anddeceptive accounts in the present study were externallygenerated; similarly all dishonest suspects were aware oftheir deceptions. An interesting possibility is thatReality Monitoring may have utility in cases in which theDECEPTIVE LGUAGE 70suspect’s or witness’ memory is unintentionally erroneous orfalse (e.g., resulting from a previous suggestion).A second major theoretical perspective considered inthis thesis underscores the betrayal of deception inlanguage due to affective and cognitive factors. Emotional-motivational approaches posit that fear of detection resultsin arousal and associated verbal deception “leakage”.Cognitive approaches are founded on the tenet that deceptiveaccounts are distinctive quantitatively and qualitativelybecause of rehearsal, information processing strategies, andcognitive load.One of the classic indices of deception associated withaffective factors is a high frequency of pauses and filledpauses during speech (Ekman, 1992; Kohnken, 1985; Zuckermanet al., 1985). Ekman (1992) argued that pauses are indeedthe most common type of verbal deception clues.Nonetheless, as in other investigations (e.g., Stiff &Miller, 1986), the present study did not support the use ofthe pause as a criterion for credibility assessment.Truthful and deceptive suspects showed a very similar,generally high propensity for filled pause usage. Theexperience of being interrogated and motivation to convincethe interrogator of their veracity appears to have elicitedmany filled pauses in all suspects, honest or deceptive.Given this ubiquitous nature, the filled pause is probablynot an appropriate criterion for assessing credibility inthe interrogation context. This is consequential in lightDECEPTIVE LANGUAGE 71of previous research indicating that people rely heavily onthe filled pause in making judgments of veracity (Stiff &Miller, 1986).Fitting best within the framework of emotional-motivational theory are the revelations concerning thecoherence of the reports. Oliver Wendell Holmes onceobserved that “People who honestly mean to be true reallycontradict themselves much more rarely than those who try tobe consistent.” In the present investigation, it was foundthat both truthful groups provided more coherent, logicalaccounts than both deceptive groups. The principal evidenceto suggest affective underpinnings emerges in the lesscoherent reports of the partially deceptive suspects. Ifreduced coherence were related only to cognitive factors(e.g., difficulty in recalling and maintaining previouslyrehearsed false information), it would be expected incompletely untruthful reports (i.e., false alibis) but notin reports containing a very small proportion of deceptivestatements (i.e., partially deceptive group). Again, it islikely that heightened arousal from lying produced a generaltendency for relative incoherence across the accounts.The intriguing possibility that this criterion may beuseful in forensic settings with adult suspects has beenunveiled in this study. However, a strong word of cautionis demanded; forensic suspects or witnesses may very wellshow different coherence levels than the individuals in thistype of study. For example, as noted by Gudjonnson (1992) aDECEPTIVE LANGUAGE 72high proportion of innocent and guilty suspects are in astate of confusion, shock, and incoherence during thepolice interrogation for several reasons including thenature of the crime, intoxication, sleep deprivation, orpsychopathology. Similarly, a genuine adult victim and anadult feigning victimization may both provide relativelyincoherent accounts. Nonetheless, such suspects and victimsmay show a similar global pattern to the participants inthis thesis; that is, the reports of honest individuals mayshow relatively higher levels of logical structure thandishonest individuals. Further research is needed toevaluate this hypothesis. As well, previous investigations(e.g., Joffe, 1992) have indicated that coherence may not bea particularly effective discriminator of truth anddeception in children’s accounts. It is possible that inyounger children the experience of lying may produce lesspronounced affective reactions because of an under—developedmoral “conscience” against lying (Kohlberg, 1984; Lewis,1993).In general, cognitive approaches to deception arehighly functional in the interpretation of the presentresults. Enigmatically, the much less detailed accounts didnot contain fewer words giving credence to the proverb“Words should be weighed and not counted.” To restate, theliars spoke as much but the information communicated intheir speech was significantly less. A possible explanationfor this finding was formulated by Kohnken (1985) who arguedDECEPTIVE LANGUAGE 73that increased cognitive load should result in a reductionin the informational content of speech. In other words,deceptive suspects have the added burden of thinking aboutthe story they are reporting which detracts from the qualityof the information reported. Whereas deceptive suspectsrecall words previously rehearsed and worry about “accountmaintenance”, truthful suspects engage in the less complexcognitive task of recalling details from a memory image. Itis also conceivable that the deceivers reiterated detailsover the course of their accounts. The scoring procedureemployed involved coding repeated details as a singledetail. Details which had been rehearsed to “keep the storystraight” may have appeared repeatedly in these reports.Miller and Stiff (1993) argued that deceptive statementscontain fewer details because “unnecessary elaborationincreases the opportunity for contradiction and thelikelihood of detection” (p. 64). An alternative affect-based interpretation is that the deceptive subjects repeatedwords due to anxiety (as found by Beattie, 1981). However,this seems less plausible given that the lexical diversityscores did not differ between the truthful and deceptivegroups.Information processing theories received mixed supportin this study. McCornack’s Information Manipulation Theory(IMT), although somewhat nebulous in nature, seems to beapplicable to these findings. As predicted, deceptivesuspects failed to fulfill the informational requirements ofDECEPTIVE LANGUAGE 74the interaction (i.e., provided fewer details) and providedinformation in a less orderly, coherent fashion. It is notclear whether the details provided by deceptive suspectswere “less relevant” to the central issue being discussed(the theft) than honest suspects. Further research isrequired to test IMT’s utility in explicating credibilityassessment principles.Other cognitive approaches were not bolstered in thisstudy. Responses following negative feedback frominterrogators through the use of interrogation probes didnot appear to effectively discriminate truthful fromuntruthful suspects, contrary to the superficial—encodinghypothesis. Nonetheless, it is still possible that suchprobes might aid in credibility assessment with certaincriminal suspects. For example, the present author recentlyread the interrogation transcripts from an ongoing Americanhomicide case in which they were highly effective.Immediately following probes very similar to those employedin this study, the prime suspect drastically altered hisaccount of the event, from witnessing someone else shootingthe victim to admitting that he had pulled the triggeraccidentally. Lexical diversity was also unrelated todeception in this study. One explanation is that elevatedlexical diversity results from anxiety associated with theexperience of being interrogated rather than anxiety due toguilt or deception (Dulaney, 1982). On the other hand,systematic changes in lexical diversity within suspects’DECEPTIVE LANGUAGE 75accounts were not examined which might be a useful endeavorin the forensic context (Carpenter, 1990).The finding that suspects providing a partiallydeceptive account were less willing to admit lack of memorythan those providing a truthful alibi certainly requiresfurther investigation. It could be associated with thedepiction of confidence and certainty by deceiversattempting to appear credible. This seems a cogent strategygiven that a strong determinant of perceived credibility isconfidence (e.g., Leippe & Romanczyk, 1989; Wells, 1993).On the other hand, meta—analyses have indicated a very weakpositive relationship between accuracy and confidence inwitnesses (who are attempting to provide accurateinformation). The current study suggests that there may bea strong negative relationship between confidence andaccuracy in criminal suspects (who are being consciouslyinaccurate). Nonetheless, this explanation remains purelyspeculative in that the proposed confidence/lack of memoryacknowledgment relationship has never been established. Asecond possibility relates to the finding that deceptivesuspects provided less detailed accounts to interviewers.Lack of memory admissions typically occurred in response todirective questioning from interrogators rather than duringthe free recall stage of the interviews. The provision offewer details by deceptive suspects may have resulted infewer directive questions being posed by the interviewersand hence fewer opportunities for lack of memory admissions.DECEPTIVE LANGUAGE 76Adding to this quandary is the fact that suspectsrelating completely falsified accounts did not employ feweradmissions. Why was the effect evidenced in suspects lyingonly about the theft itself but not in suspects providingentirely deceptive information? One can only surmise thatfor the completely deceptive suspects lack of memoryadmissions functioned to facilitate account maintenance(i.e., it is easier for a suspect to maintain a story ifhe/she purports to recall less about it), whereas thepartially deceptive suspects were being more heavilyinfluenced by one of the factors discussed above. Futureresearch should address this mysterious finding.Sapir’s proposals concerning differences in pronoundeviation, component elements, and unnecessary connectorsdid not fare well in this study. However, their merits havenot been singly obliterated by this experiment. Forexample, both deceptive groups showed substantially (but notsignificantly) more pronoun deviations than the truthfulgroups. More research should address this interestingpossibility that attention to pronoun usage may aid incredibility assessment. Secondly, and more importantly inmy opinion, the unnecessary connector has not really had its“day in court” either. Variables like the unnecessaryconnector are extremely difficult to evaluate throughresearch simply because of the possibility of their rarity;it may be that an unnecessary connector surfaces inexceptional forensic cases but with momentous consequenceDECEPTIVE LANGUAGE 77when detected by an experienced investigator. Even in thepresent study, although mean differences were miniscule, aglaring unnecessary connector was employed by a particularsuspect (Condition 2) attempting to conceal informationrelating to the theft. During the free narrative account,he stated:I had to look for a folder that was supposed to bein a box but I couldn’t find it andthen, .uhhh..I looked around a bit more. Thenafter that I just went out of that office, madesure the door was locked again, went back the sameway...With little more than anecdotal evidence (Sapir has providedmany such examples), a conservative stance must be taken.It would be advisable to emphasize to law enforcementpersonnel that unnecessary connectors may, in certain cases,be indicative of the concealment of information, suggestinga topic requiring clarification or more in depth explorationbut certainly to never rely singularly on this variable.Three of the ten (30%) SVA variables investigated inthis thesis (coherence, sufficiency of detail, and admittinglack of memory) significantly discriminated the truthful anduntruthful accounts. This performance is encouraging andgives reason for the continued quest for reliable clues tocredibility in forensic contexts. It is especiallypromising given related findings with adult witnesses (e.g.,Brigham & Landry, 1992 found that 71% of the fourteen SVAvariables they examined were effective). Indeed, theperformance of SVA in this study may be consideredremarkable in that the methodology may not have beenDECEPTIVE LGUAGE 78conducive to the successful employment of SVA. Raskin andEsplin (1991) recommend that in designing generalizableexperimental research addressing SVA the event used shouldbe involving, have a component of lost control, and anegative tone. The present study lacked a component of lostcontrol with its intention to examine the utility of SVAwith suspects rather than victims or witnesses. Themethodology employed sought to create components ofinvolvement, anxiety during the commission of the “offense”,motivation to relate a deceptive or truthful accountsuccessfully, and realistic negative feedback in theinterrogation context. As earlier discussed, SVA wasoriginally developed for use in cases of child sexual abuse(Undeutsch, 1982) for which certain criteria may be context-dependent. Overall, this thesis and previous studies seemto paint a picture of SVA as a potent general credibilityassessment technique whose criteria are differentiallyuseful across contexts.One of the more striking serendipitous findings in thisstudy mirrored, to a lesser extent, Zimbardo’s classicprison studies from the early 1970’s (e.g., Zimbardo, 1972).As with the guards and inmates in his experiments, theinterrogators and suspects in this study appear to havebecome heavily enmeshed in their respective roles. Anexamination of the first couple of interviews showed theinterviewers becoming visibly aggravated with theparticipants when they suspected they were being lied toDECEPTIVE LANGUAGE 79(they were unaware of the participants’ veracity).Correspondingly, the participants reacted defensively andangrily to this negative feedback (e.g., one truthfulsubject stated “I brought the folder up and now you’reaccusing me of lying. What’s the problem with you? Jesus,people these days!”). It was necessary to reiterate to theinterviewers to treat all suspects with parity resisting theurge to evaluate the honesty of the suspects and to maintaina relatively polite, unpassioned tone. As well, themajority of suspects reported feeling very wary of eachperson encountered en route to the crime room with thesuspicion that he/she might have been the guard. Someparticipants reported feeling guilty but exhilaratedfollowing the commission of the theft. Although suchreports may lend support to the validity of this study, theyalso show how amazingly expeditious such a simulated role“was absorbing [the participants] as creatures of its ownreality” (Zinthardo, 1972, cited in Myers, 1990).8.1 Limitations of the Current Study. In thisinvestigation, an effort was made to overcome severalmethodological shortcomings evident in previous studiesaddressing credibility assessment. Specifically, thesimulated crime and interrogation were relatively involvingand anxiety—provoking and as analogous to the forensiccontext as ethically permissible. Both honest and deceptivesuspects were motivated to appear truthful and were suppliedDECEPTIVE LANGUAGE 80time to prepare their accounts. The intimated presence of aguard and motivation of the “administration’s interrogators”served to further increase the validity of this study. Asdescribed earlier, this validity seems to have been attainedgiven the comments of the participants during theinterrogation and debriefing phases.Despite these methodological strengths, others haveargued that the accuracy of credibility techniques can neverbe measured in the laboratory because the emotionalcharacteristics of the suspects and witnesses are toodissimilar from those involved in real-life investigations(Arntzen, 1983; lacono & Patrick, 1987). Although thisperspective has been criticized as revealing “a lack ofunderstanding of empirical research needs and methods”(Steller, 1989, p. 144), it certainly highlights a validconcern with generalizability.In considering the generalizability of this research,the issue of motivation becomes salient. It is clear thatthe monetary incentive offered participants to be convincingin an experimental setting cannot approximate the motivationpresent in forensic suspects — freedom. Not only is thesimulation problematic quantitatively but alsoqualitatively; in the present study a dynamic of negativereinforcement was created (i.e., “If you fail to convinceyour interrogator of your veracity you will not receiveextra money”) whereas in real life a punishment contingencyis more prevalent (i.e. “If you fail to convince yourDECEPTIVE LANGUAGE 81interrogator of your veracity you will go to jail”). It isnot clear if or how this motivational difference mightaffect language behavior during the interrogation.Intention to deceive must also be considered inexploring the issue of generalizability. In some cases,innocent suspects may come to believe that they committed acrime contrary to traditional legal doctrines (e.g., “Thepresumption upon which declarations are evidence is that noman would declare anything against himself unless it weretrue” - Judge Eyre in Hardy’s Trial, 1794, cited in Arcaro,1990). This was exemplified in the Christie case describedearlier; Timothy Evans became so convinced by police that hehad killed his wife that he eventually recalled the act.Subsequent to Evans’ execution, it was proven that Christiehad solely committed the homicide. Similarly, although notexamined in the present study, voluntary false confessionssometimes occur. This phenomenon was exemplified in thecase of serial murderer Henry Lee Lucas who, in the early1980’s, falsely confessed to several hundred murderssupposedly for the increased media attention he receivedwith each confession (Cox, 1991). It is not known whetherthe verbal clues examined in this study would have utilityin these types of cases. In this study, it was found thattruthful confessions tend to be quite detailed. It islikely that false confessions will be less detailed thantruthful confessions unless the false confessor has a greatdeal of knowledge about the crime scene from other sources.DECEPTIVE LANGUAGE 82Another aspect of this study which must be examined inconsidering external validity is the population employed.Undergraduates are certainly not representative of thecriminal class of society. For example, research hasconsistently demonstrated that middle-class youths showlower rates of criminal behavior than working-class andlower-class youths (e.g., Elliot & Huizinga, 1983; Wolfgang,Figlio, & Sellin, 1972). Evidently, undergraduates haveless experience being interrogated which would likely affecttheir behavior in this context.As well, undergraduates may show different intelligenceand verbal abilities than most criminals, affecting theirlanguage behaviors in general. The results in this thesismay not extend to psychopaths, who constitute 15-25% of thecriminal population (Hart & Hare, 1989). Exciting recentresearch by Hare and colleagues indicates that psychopathsprocess and communicate language on a different level thanmost people (Day & Wong, 1993; Hare & Jutai, 1988; Louth,Williamson, Alpert, & Hare, in preparation; Williamson,Harpur, & Hare, 1991). For example, the Louth et al. studyevinced that psychopaths speak with lower amplitude but witha faster rate than nonpsychopaths and do not distinguishbetween emotional and neutral words as did thenonpsychopaths. With a greater insensitivity to theemotional connotations of language, it is probable thatdissimilarities would also emerge in the variables examinedin the present study. Future research should address theDECEPTIVE LANGUAGE 83possibility that psychopaths show fewer verbal differenceswhen being honest and deceptive than nonpsychopaths becausethis insensitivity (cf. Raskin & Hare, 1978). It shouldalso be noted that pathological lying, deception, andmanipulation are some of the key diagnostic features of thepsychopath (e.g., Hare, Forth, & Hart, 1988) leading toexperiential adeptness in lying and corresponding behaviorsin the interrogation.Finally, cultural distinctions may also restrict theresults of this thesis which employed North AmericanEnglish-speaking participants. Aside from obvious cross-cultural language variations, cultural norms may alsodictate a suspect’s behavioral repertoire in theinterrogation context (e.g., see Vrij & Winkel, 1991).Similarly, sex differences were not examined in this thesisbecause of a female participation bias.One of the problems associated with the above factorswhich must be considered when evaluating generalizability isthe potential complexity of their effects; their presencemight increase the frequency of certain deception indicatorsand decrease the frequency of others. For example, it mightbe that a psychopath experienced with the interrogationprocess would recount a (rehearsed) coherent, detailed falsealibi, but frequently admit lacking memory for aspects ofthe event to avoid further probing by interrogators. It isevident that much research is needed to clarify thesepossible relationships.DECEPTIVE LM’GUAGE 848.2 Directions for Future Research. As noted by severalforensic researchers (e.g., Davies, 1990; Yuille, Davies,Gibling, Marxsen, & Porter, in press; Porter, Yuille, &Bent, in press; Yuille, 1993), only when we have convergingresults from diverse methods of inquiry should we generalizeto forensic contexts. The present research was intended asa preliminary examination of verbal clues to deception inwhich ground truth is established. Before intercalating myresults into the forensic interrogation context much furtherinquiry is essential, especially field studies. Cross—validation studies are crucial with this type of appliedresearch. In certain criminal cases it is possible todetermine with a fair degree of certainty whether a suspecthad been lying or telling the truth during an interrogation(e.g., in those cases in which the suspect subsequentlyconfesses or another suspect confesses). As observed byMiller and Stiff (1993) law enforcement agencies frequentlyrecord investigative interviews with of suspects, witnesses,and victims as well as pre-polygraph and polygraphexaminations, providing fertile ground for researchersinterested in deception. The results of the present studymust be applied to forensic populations to test theecological validity of this experimental paradigm and toexamine whether these findings would have utility in actualinvestigations. Importantly, in the one published fieldstudy applying statement analysis to an actual case,DECEPTIVE LANGUAGE 85quantity of details and coherence were two of main criteriaemployed in assessing the credibility of a suspect (Yuille &Cutshall, 1989).Unfortunately, the finding that dishonest suspects tendto give less detailed and coherent accounts in which thereare fewer admissions to lack of memory does not tell us howto apply this knowledge to a particular case. Within-subjects experimental designs are essential to examinewhether verbal characteristics change as a function ofveracity within individual suspects. If so, one way toutilize this knowledge would be to subtly ask a suspect torelate an event other than the crime being investigatedduring the interrogation (J.C. Yuille, personalcommunication, July 5, 1994). Patterns within this accountwould then be compared to those in the account relating tothe crime (e.g., an alibi). Similarly, there is a need forestablishing the base rates of any verbal characteristicsexamined for credibility assessment purposes (e.g., level ofdetail and coherence, frequency of lack of memoryacknowledgments) in the forensic context. As well,virtually no research has addressed the base rate of lyingin the criminal interrogation. This knowledge is importantfor practitioners of credibility assessment in any appliedcontext.As previously mentioned, verbal clues to deception maybe differentially useful across contexts. As well, theremay be a host of verbal clues to deception not examined inDECEPTIVE LANGUAGE 86this thesis. For example, after being asked if she had seenany evidence of a theft while retrieving the folder, oneparticipant responded “No..I didn’t see any envelope with ahundred dollar bill on it” expressing knowledge only aguilty suspect would possess. Although unclassifiable withthe set of criteria coded in this study, attention to suchverbal slips may obviously give clues to credibility.Another intriguing example from an actual case is reportedby Avner Less who interrogated the infamous S.S. officerAdolf Eichmann, captured in 1960 (Less, 1983). Less stated:“As time went on, I noticed that each time Eichmann saidNever! Never! Never, Herr Hauptmann!.’ or At no time! At notime!’ he was lying. That was always a cue for me to ask mycolleagues for additional material with which to probe thesensitive spot” (Less, 1983, p. xxi). The next formidablebut attainable step for researchers in credibilityassessment is to seek which verbal clues are effective withwhom and under which circumstances.DECEPTIVE LANGUAGE 87REFERENCESAlonso-Quecuty, M. (1990). Deception detection and realitymonitoring: A new answer to an old question? Paperpresented at the European Conference on Law andPsychology.Arcaro, G. (1990). Interrogations and confessions. InCriminal investigations and the formulation ofreasonable grounds (pp. 118-169). Toronto: EdmondMontgomery Ltd.Arens, R. & Meadow, A. (1957). Psycholinguistics and theconfession dilemma. Columbia Law Review, 56, 19—46.Arntzen, F. (1983). Psychologie der zeugenaussage: Systemder glaubwurdigkeitmerkmale. (2nd Ed.). Munich: Beck.Barland, G.H. & Raskin, D.C. (1976). Validity andreliability of polygraph examinations of criminalsuspects. (Report No. 76-1, Contract No. N1-99-0001).Washington, D.C.: National Institute of Justice, Dept.of Justice. In lacono, W.G. & Patrick, C.J. (1987).What psychologists should know about lie detection. InI.B. Weiner & A.K. Hess (Eds.) Handbook of forensicpsychology. New York: John Wiley & Sons.Bauchner, J.E., Kaplan, E.P., & Miller, G.R. (1980).Detecting deception: The relationship of availableinformation to judgmental accuracy in initialencounters. Human Communication Research, 6, 251—264.Beattie, G.W. (1981). A further investigation of thecognitive interference hypothesis of gaze patternsduring conversation. British Journal of SocialPsychology, 20, 243—248.Ben-Shakhar, G. & Furedy, J.J. (1990). Theories andapplications in the detection of deception. New York:Springer-Verlag.Bolinger, D. (1973). Truth is a linguistic question.Language, 49, 539-550.Bordens, K.S. & Abbott, B.B. (1991). Research design andmethods. London: Mayfield Publishing Co.Buckwalter, A. (1983). Interviews and interrogations.Boston: Butterworth Publishers.Camps, F. & Barber, R.(1966). The investigation of murder.London: Michael Joseph Ltd.DECEPTIVE LANGUAGE 88Carpenter, R. (1981). Stylistic analysis for law enforcementpurposes: a case study of a language variable as anindex of a suspect’s caution in phrasing answers.Communication Quarterly, 29, 32-39.Carpenter, R. (1990). The statistical profile of languagebehavior with Machiavellian intent or whileexperiencing caution and avoiding self—incrimination.Annals of the New York Academy of Sciences, 606, 5-16.Cocozza, J.J. & Steadman, H.J. (1978). Prediction inpsychiatry: An example of misplaced confidence inexperts. Social Problems, 25, 265-275.Comber, M. & Canter, D. (1983). Differentiation of maliciousand non—malicious fire—alarm calls usingmultidimensional scaling. Perceptual and Motor Skills,57, 460—462.Cox, M. (1991). The confessions of Henry Lee Lucas. NewYork: Pocket Books.Craik, F.M. & Lockhart, R.S. (1972). Levels of processing: Aframework for memory research. Journal of VerbalLearning and Verbal Behavior, 11, 671-684.Davies, G.D. (1990, Sept.). Influencing public policy oneyewitnessing: Problems and possibilities. Paperpresented at the Second European Conference on Law andPsychology, Nuremburg, Germany.DePaulo, B. (1992). Nonverbal behavior and self-presentation. Psychological Bulletin, 111, 203-243.DePaulo, B. & Jordan, A. (1982). Age changes in deceivingand detecting deceit. In R.S. Feldman (Ed.) Developmentof nonverbal behavior in children. New York: SpringerVerlag.DePaulo, B., Lanier, K., & Davis, T. (1983). Detecting thedeceit of the motivated liar. Journal of Personalityand Social Psychology, 45, 1096-1103.DePaulo, B. & Pfeifer, R. (1986). On-the-job experience andskill at detecting deception. Journal of Applied SocialPsychology, 16, 249—267.deTurck, M.A. & Miller, G.R. (1985). Deception and arousal:Isolating the behavioral correlates of deception. HumanCommunication Research, 12, 181-201.Dulaney, E.F. (1982). Changes in language behaviors asfunction of veracity. Human Communication Research, 9,75—82.DECEPTIVE LANGUAGE 89Ekman, P. (1992). Telling lies. New York: W.W. Norton & Co.Ekman, P. & Frieson, W. (1969). Nonverbal leakage and cluesto deception. Psychiatry, 32, 88-106.Ekman, P. & O’Sullivan, M. (1991). Who can catch a liar?American Psychologist, 46, 913—920.Elaad, E. & Ben-Shakhar, G. (1989). Effects of motivationlevel and verbal response type on psychophysiologicaldetection in the guilty knowledge test.Psychophysiology, 19, 203—218.Elliot, D.S. & Huizinga, D. (1983). Social class anddelinquent behavior in a national youth panel.Criminology, 21, 149-147Finkeihor, D. (1986). A sourcebook on child sexual abuse.Beverly Hills: SAGE Productions.Griffiths, G. (1992). Detecting deception throughpsycholinguistic analysis: a preliminary report.Unpublished manuscript. Advanced Forensic Technologies,Inc., Fayetteville, GA.Gudjonsson, G. (1992). The psychology of interrogations,confessions, and testimony. Chichester: John Wiley &Sons.Gunn, J. & Gudjonsson, G. (1988). Using the PsychologicalStress Evaluator in conditions of extreme stress.Psychological Medicine, 18, 235-238.Gustafson, L.A. & Orne, M.T. (1963). Effects of heightenedmotivation on the detection of deception. Journal ofApplied Psychology, 47, 408-411.Hare, R.D., Forth, A.E., & Hart, S.D. (1989). The psychopathas prototype for pathological lying and deception. InJ.C. Yuille’s (Ed.) Credibility Assessment (pp. 25-49).Dordrecht: Kiuwer Academic Publishers.Hare, R.D. & Jutai, J.W. (1988). Psychopathy and cerebralasymmetry in semantic processing. Personality andIndividual Differences, 9, 329—337.Harrison, A.A., Hwalek, M., Raney, D.F., & Fritz, J.G.(1978). Cues to deception in an interview situation.Social Psychology, 41, 156—161.Hart, S.D. & Hare, R.D. (1989). Discriminant validity of thePsychopathy Checklist in a forensic psychiatricDECEPTIVE LANGUAGE 90population. Psychological Assessment: A Journal ofConsulting and Clinical Psychology, 1, 211-218.Hocking, J., Bachner, J., Kaminski, E., & Miller, G. (1979).Detecting deceptive communication from verbal, visual,and paralinguistic cues. Human Communication Research,6, 33—46.Hollien, H. (1990). The acoustics of crime: The new scienceof forensic phonetics. New York: Plenum Press.fonts, C.R., Raskin, D.C., & Kircher, J.C. (1994). Mentaland physical countermeasures reduce the accuracy ofpolygraph tests. Journal of Applied Psychology, 17,252—259.Horvath, F.S. (1977). The effect of selected variables oninterpretation of polygraph records. Journal of AppliedPsychology, 62, 127—136.lacono, W.G. & Patrick, C.J. (1987). What psychologistsshould know about lie detection. In I.B. Weiner & A.K.Hess (Eds.) Handbook of forensic psychology. New York:John Wiley & Sons.Inbau, F.E., Reid, J.E., & Buckley, J.P. (1986). Criminalinterrogation and confessions, 3rd ed. Baltimore:Williams and Wilkins.Joffe, R.D. (1992). Criteria-based content analysis: anexperimental investigation with children. Unpublisheddoctoral dissertation. University of British Columbia.Johnson, M.K. & Raye, C.L. (1981). Reality monitoring.Psychological Bulletin, 88, 67-85.Klein, J.F. (1976). The dangerousness of dangerous offenderlegislation: Forensic folklore revisited. CanadianJournal of Criminology, 18, 109—122.Kleinmuntz, B. & Szucko, J. (1984). A field study of thefallibility of polygraphic lie detection. Nature, 308,449—450.Knapp, M. & Comadena, N. (1979). Telling it like it isn’t: Areview of theory and research on deceptivecommunication. Human Communication Research, 5, 270-285.Kohlberg, L. (1984). The psychology of moral development:The nature and validity of moral stages, Vol.2. NewYork: Harper & Row.DECEPTIVE LANGUAGE 91Kohnken, G. (1985). Speech and deception of eyewitnesses: Aninformation processing approach. In F.L. Denmark (ed.)Social/ecological psychology and the psychology ofwomen (pp. 117-139). North Holland: Elsevier SciencePublishers.Kraut, R.E. (1978). Verbal and nonverbal cues in theperception of lying. Journal of Personality and SocialPsychology, 36, 380—391.Landry, K. & Brigham, J. (1992). The effect of training inCriteria-Based Content Analysis on the ability todetect deception in adults. Law and Human Behavior, 16,663—676.Larson, J. (1926). The Berkeley lie detector and otherdeception tests. In R. Brown’s Legal psychology.Indianapolis: Bobbs-Merrill.Larson, J. (1932). Lying and its detection. Montclair, N.J.:Patterson Smith.Leippe, M.R., Manion, A.P., & Romanczyk, A. (1992).Eyewitness persuasion: How and how well do fact findersjudge the accuracy of adults’ and children’s memoryreports? Journal of Personality and Social Psychology,63, 181—197.Leippe, M.R. & Romanczyk, A. (1989). Reactions to childeyewitnesses: The influence of juror’s preconceptionsand witness behavior. Law and Human Behavior, 13, 103-132.Lesce, T. (1990). SCAN: Deception detection by scientificcontent analysis. Law and Order, 38, 1-2.Less, A.W. (1983). Introduction. In J. von Lang & C. Sibyll(eds.) Eichiuann interrogated (pp. v-xxii). New York:Farrar, Straus, & Giroux.Lewis, M. (1993). The development of deception. In M. Lewis& C. Saarni (Eds.) Lying and deception in everyday life(pp. 90—105). New York: Guilford press.Louth, S.M., Williamson, S., Alpert, M., & Hare, R.D. (Inpreparation). Can you identify a psychopath by the wayhe speaks?: Amplitude and rate differences in thespeech of male psychopaths.Lykken, D. (1981). A tremor in the blood. New York: McGrawHill.DECEPTIVE LANGUAGE 92Macdonald, J.M. & Michaud, D.L. (1987). The confession:Interrogation and criminal profiles for policeofficers. Denver: Apache Press.McCornack, S.A. (1992). Information Manipulation Theory.Communication Monographs, 59, 1—16.Mehrabian, A. (1971). Nonverbal betrayal of feeling. Journalof Experimental Research in Personality, 5, 64-73.Miller, G.R. & Stiff, J.B. (1993). Deceptive communication.Newbury Park: SAGE Publications.Miron, M. (1984). Content identification of communicationorigin. In Advances in forensic psychology andpsychiatry, 1, 113-146. Norwood, N.J.: Ablex PublishingCo.Miron, M. (1990). Psycholinguistics in the courtroom. Annalsof the New York Academy of Sciences, 606, 55-64.Miron, M. & Pasquale, T. (1978). Psycholinguistic analysesof coercive communications. Journal of PsycholinguisticResearch, 2, 95—120.Morton, A.Q. (1978). Literary detection. London: Bowker.Moston, S.J. & Stephenson, G.M. (1992). Predictors ofsuspect and interviewer behaviour during policequestioning. In F. Losel, D. Bender, & T. Bliesener(Eds.) Psychology and law: International perspectives.Berlin: Walter de Gruyter & Co.Myers, D.G. (1990). Social Psychology. New York: McGraw-HillPublishing Company.O’Hair, H., Cody, N., & McLaughlin, M. (1981). Preparedlies, spontaneous lies, Machiavellianism, and nonverbalcommunication. Human Communication Research, 7, 325—339.Osgood, C.E. (1960). Some effects of motivation on style ofencoding. In T. Sebeok’s (Ed.) Style in language.Cambridge: MIT Press.Philips, S.U., Steele, S., & Tanz, C. (Eds.). (1987).Language, gender & sex in comparative perspective.Cambridge: Cambridge University Press.Porter, S., Yuille, J.C., & Bent, A. (In press). Acomparison of the eyewitness accounts of deaf andhearing children. Child Abuse and Neglect.DECEPTIVE LANGUAGE 93Porter, S. & Yuille, J.C. (In press). Credibility assessmentof criminal suspects through statement analysis.Psycholoqy, Crime, and Law.Raskin, D. (1989). Polygraph techniques for the detection ofdeception. In D. Raskin’s (Ed.) Psychological methodsin criminal investigation and evidence. New York:Springer Publishing Co.Raskin, D.C. & Hare, R.D. (1978). Psychopathy and detectionof deception in a prison population. Psychophysiology,15, 126—136.Sapir, A. (1987). The LSI Course on Scientific ContentAnalysis (SCAN). Laboratory for ScientificInterrogation, Phoenix, AZ.Schooler, J.W., Clark, C.A., & Loftus, E.F. (1988). Knowingwhen memory is real. In M.M. Gruneberg, P.E. Morris, &R.N. Sykes (Eds.) Practical aspects of memory.Chichester: John Wiley & Sons.Schooler, J.W., Gerhard, D., & Loftus, E.F. (1986).Qualities of the unreal. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 12, 171-181.Simpson, D. (1992). Lying, liars, and language. Philosophyand Phenomenological Research, 52, 623-631.Steller, M. (1989). Recent developments in statementanalysis. In J.C. Yuille (Ed.), Credibility Assessment.Steller, M. & Boychuk, T. (1992). Children as witnesses insexual abuse cases: Investigative interview andassessment techniques. In H. Dent & R. Fun (Eds.)Children as witnesses. Chichester: John Wiley & Sons.Steller, M. & Kohnken, G. (1989). Criteria-based statementanalysis. In D.C. Raskin’s (Ed.) Psychological methodsin criminal investigation and evidence. New York:Springer.Stiff, J.B. & Miller, G.R. (1986). “Come to think of it...”:Interrogative probes, deceptive communication, anddeception detection. Human Communication Research, 12,339—357.Undeutsch, U. (1982). Statement reality analysis. In A.Trankell (Ed.), Reconstructing the past: The role ofpsychologists in criminal trials (pp. 27-56).Stockholm: P.A. Norsted & Sons.DECEPTIVE LANGUAGE 94Vrij, A. & Winkel, F.W. (1991). Cultural patterns in Dutchand Surinam nonverbal behavior: An analysis ofsimulated police/citizen encounters. Journal ofNonverbal Behavior, 15, 169-184Waid, W.M. & Orne, M.T. (1981). Cognitive, social, amdpersonality processes in the physiological detection ofdeception. In L. Berkowitz (Ed.) Advances inexperimental social psychology. New York: AcademicPress.Wells, G.L. (1993). What do we know about eyewitnessidentification? American Psychologist, 48, 553-571.Werner, P.D., Rose, T.L., & Yesavage, J.A. (1983).Reliability, accuracy, and decision-making strategy inclinical predictions of imminent dangerousness. Journalof Consulting and Clinical Psychology, 51, 815-825.Williamson, Harpur, & Hare (1991). Abnormal processing ofaffective words by psychopaths. Psychophysiology, 28,260—273.Wolfgang, M., Figlio, R.M., & Sellin, T. (1972). Delinquencyin a birth cohort. Chicago: University of ChicagoPress.Yuille, J.C. (1988). The systematic assessment of children’stestimony. Canadian Psychology, 29, 247-262.Yuille, J.C. (1989). Credibility assessment. TheNetherlands: Kiuwer Academic Publisher.Yuille, J.C. (1990). Statement analysis: An overview of thecontent criteria. Unpublished manuscript. University ofBritish Columbia.Yuille, J.C. (1993). We must study forensic eyewitnesses toknow about them. American Psychologist, 48, 572-573.Yuille, J.C. & Cutshall, J. (1989). Analysis of thestatements of victims, witnesses, and suspects. In J.C.Yuille (Ed.), Credibility assessment.Yuille, J.C., Hunter, R., Joffe, R., & Zaparniuk, J. (1993).Interviewing children in sexual abuse cases. In G.S.Goodman & B.L. Bottoms (Eds.), Child witnesses, childvictims: Understanding and improving children’stestimony. New York: Guilford Press.Yuille, J.C., Davies, G.D., Gibling, F., Marxsen, D., &Porter, S. (In press). Eyewitness memory of policetrainees for realistic role plays. Journal of AppliedPsychology.DECEPTIVE LANGUAGE 95Yuille, J.C. & Porter, S. (1993). Scoring procedure forevaluating eyewitness accounts. Unpublished manuscript.University of British Columbia.Yuille, J.C. & Tollestrup, P. A. (1990). Some effects ofalcohol on eyewitness memory. Journal of AppliedPsychology, 75, 1-6.Yuille, J.C., Tollestrup, P.A., Marxsen, D., & Porter, S.(Under review). Some effects of marijuana on eyewitnessmemory.Zaparniuk, J., Yuille, J.C. & Taylor, S. (1993). Assessingthe credibility of true and false statements.Unpublished manuscript. University of British Columbia.Zimbardo, P.G. (1972). Pathology of imprisonment.Transaction/Society, April, 4-8.Zuckerman, M. & Driver, R.E. (1985). Telling lies: Verbaland nonverbal correlates of deception. In A.W. Siegman& S. Feldstein (Eds.) Multichannel integration ofnonverbal behavior. Hilisdale, N.J.: Eribaum.Zuckerman, M., DePaulo, B.M., & Rosenthal, R. (1981). Verbaland nonverbal communication of deception. In L.Berkowitz (Ed.), Advances in experimental socialpsychology (Vol. 14), pp. 1-59. New York: AcademicPress.DECEPTIVE LANGUAGE 96Appendix ADEBRIEFINGThank you for your participation. Following is a briefdescription of the rationale behind the project in which youhave just participated:My area of research is the use of verbal clues for thepurpose of credibility assessment or “lie detection” withcriminal suspects. What is different about the language ofa person telling the truth and a person being deceptiveduring an interrogation? Although fieldwork is planned witha criminal suspect population, this analog study representsthe preliminary controlled research (in which we always knowwho is lying and who is not).In this study, participants were divided into fourconditions: truthful alibi, partial deception, false alibi,and truthful confession. All participants proceeded to thecrime room where most (except those in the truthful alibicondition) took a $100 bill. On returning to the lab, amonetary incentive was offered to motivate participants tolie successfully to the interviewer. The interviewers useda “Step—Wise” interview format as well as interrogationprobes (e.g., “It is obvious that you are lying to me. Whydon’t you just admit it?”).Eighteen variables relating to your statements in theinterview are being examined. For example, it ishypothesized that deceptive suspects will use more filledDECEPTIVE LANGUAGE 97pauses than honest suspects because of cognitive factorsassociated with lying under pressure.Some deceptions were used by the experimenter in thisstudy:1. There was no guard. There has not been a recent rash ofthefts in the building.2. This was not a “security effectiveness” study. It was astudy of the language behaviors of truthtellers and liars.3. It was not the agenda of the interviewers to establishthe truth. They were asked to obtain as much informationfrom all suspects and to treat all suspects the same. Itdid not matter whether the interviewer believed you or not.All participants received the extra $5.It was necessary to employ these deceptions to increaseexternal validity. The presence of the fictitious guard wasimplied to simulate the situation of an actual criminalcommitting a theft. The extra $5 was offered to increaseparticipants’ desire to successfully convince theinterviewer of their truthfulness (the typical criminalsuspect is highly motivated to do so). With the designemployed, I wanted to approximate a real interrogation asclosely as possible.Because experimenter deception is a crucial componentof this research it is extremely important that you notdiscuss this study with anyone until all subjects have beenrun (i.e, May, 1994). We appreciate your cooperation inthis regard.DECEPTIVE LANGUAGE 98Feel free to ask any questions to either Steve Porteror Dr. John C. Yuille. If you are interested, the resultsof this study will likely be available to you by the end ofSeptember, 1994.DECEPTIVE L?NGUAGE 99Appendix BINTERROGATION PROTOCOL1. INTRODUCTIONBefore you begin the interview, test the tape recorder.State your full name, the date, subject’s ID number and askthe subject to state his/her full name. Play back the tapeto ensure that the machine is picking up both of your voicesclearly. If during the interview, the subject movessignificantly or lowers his/her voice interrupt the account.Politely request that he/she resume the original position orvoice amplitude.2. FREE NARRATIVEIt is important that a dynamic is established in whichit is the subject who does most of the talking. State tothe subject “It is my job to determine which of the suspectsin this study committed the theft of a $100 bill from Room#---- downstairs. I’d like you to tell me everything youcan remember, no matter how trivial or irrelevant it mayseem.” Start from the time you exited this lab and keepgoing until the time you returned to the lab. I aminterested in determining your whereabouts and actions inthe last fifteen minutes.” Do not interrupt the subjectduring the free narrative, even if the subject starts todescribe irrelevant details of the event or contradictshim/herself. Allow the subject to go at his/her own paceand be patient when the subject pauses. If, however, itDECEPTIVE LANGUAGE 100seems that the subject is not going to continue the accountyou should attempt to restart the narrative. The bestmethod for doing this is simply to state “What happenednext?” or “You were saying that (here restate the last thingthe subject said). And then what happened?”.Make very brief notes regarding the content of thenarrative. These notes can consist simply of the noun andverb phrases mentioned by the subject (e.g., a desk). Whenthe subject has come to the end of the narrative it isimportant that you ask him/her if anything more can beremembered. If not proceed to the next step.3. OPEN QUESTIONINGOpen questions are simply requests for more informationabout the events, actions, and objects mentioned in the freenarrative. The easiest way to do this is to refer to yournotes taken earlier and asking if the subject recallsanything more about each of the noun and verb phrases (e.g.,“You mentioned some books. Can you tell me anything moreabout those?”). Before asking open questions it isimperative that you let the subject know that “No, I don’tremember” is a completely acceptable response. Never askquestions like “You told me you saw some books. Tell memore about those.” Similarly, your open questioning shouldnot be leading. Do not include information from othersources in these questions. If all previous suspectsmentioned seeing a picture on the wall it is easy to makeDECEPTIVE LANGUAGE 101the mistake of asking a subject who has not mentioned it ifhe/she can remember any more about the picture.4. INTERROGATION PROBE #1Following the open questions, express a puzzled lookand state “Frankly, I’m a little skeptical about your story.Are you being completely honest?”. When the subjectresponds positively (i.e., “Yes, of course”) say “Okay, butsomething doesn’t sound quite right.” If the subjectresponds negatively (i.e., “No”) ask what part of theaccount is not true. If an alternative response isencountered the interviewer should respond politely butstill convey suspicion (e.g., “Why? Are you accusing me ofsomething?”; Response — “No, I’m not accusing you ofanything. It’s just that I get the feeling that you are notbeing completely honest”).5. SPECIFIC QUESTIONINGIf, after asking the open questions, there are stillpoints from the subject’s account that require elaborationyou should now resort to specific questions (this isgenerally the case). It is especially important here toidentify any inconsistencies in the report and ask thesubject to clarify them. The best method is to simply pleadconfusion. You can say something like “I’m not certain Iunderstand everything. You said that you were in the officefor about five minutes but later said you spent seven oreight minutes looking for the folder in the office. Can youexplain this to me?”DECEPTIVE LANGUAGE 1026. INTERROGATION PROBE #2After pointing out any inconsistencies and hearing theexplanations you should ask the suspect directly “Do youhave any knowledge of the money which was taken from Room #————?“7. INTERROGATION PROBE #3After hearing the response to Interrogation Probe #2,state “Come on now, it’s really obvious that you are lyingto me. Why don’t you just admit it?”. Don’t say it in aparticularly hostile manner. Be relatively polite about it.After the subject provides a response to this statement,conclude the interview. Thank the subject for his/her timeand inform him/her that he/she can now leave the interviewroom, receive their money for participating, and becompletely debriefed in the next room.DECEPTIVE LANGUAGE 103Appendix CSTANDARD DEVIATIONS FOR TABLE 1 DEPENDENT VARIABLESEXPERIMENTAL CONDITION1 2 3 4Truthful Partial False TruthfulDEPENDENT Alibi Deception Alibi ConfessionVARIABLEDetails 18.90 13.41 13.57 13.91Words 351.00 575.87 535.44 386.39TTR 0.05 0.06 0.08 0.05Verbal Hedges 2.30 2.99 6.10 3.80Filled Pauses 11.52 25.29 20.80 14.33Un. Connectors 6.76 6.73 6.96 3.96Pron. Deviations 11.20 8.67 10.45 7.70Self—references 33.75 38.05 48.20 42.50Comp. Elements 0.24 0.32 — 0.22Coherence 0.96 0.86 0.99 0.79Sp. Reproduction 1.18 1.16 1.05 0.70Un. Complications 0.96 1.00 0.92 1.36Un. Details 0.72 0.93 0.68 0.94Pen. Details 0.83 0.78 — 0.88R.E. Associations 1.01 1.96 4.17 1.74S.M. State 1.91 2.23 1.46 1.68Spont. Corrections 0.62 0.91 0.41 0.74A.L. of Memory 2.76 1.89 1.80 3.48


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items