t . , . . . . . s , . . THESIS WMV This is to certify that the dissertation entitled DETECTION OF DECEPTION IN NEUROPSYCHOLOGICAL TESTING: A MULTI-METHOD APPROACH presented by David Dean Cordry has been accepted towards fulfillment of the requirements for I”h . D. degree in m Datefirzl 25: QDOA MS U is an Affirmative Action/Equal Opportunity Institution 0‘ 1 2771 LIBRARY Michigan State University PLACE "q RETURN BOX to remove this checkout from your record. 0 AVG“) FINES return on or before date due. MAY BE RECALLED Wlth earlier due date if requested. , DATE DUE DATE DUE ' DATE DUE f \ 6/01 c:lClRC/DateDue.p65-p. 15 DETECTION OF DECEPTION IN NEUROPSYCHOLOGICAL TESTING: A MULTI-METHoD APPROACH By David Dean COrdry A DISSERTATION Submitted to . Michigan State Univeyrsity 1n . . . partlal fulflllment of the requ1rements for the degree c)f DOCTOR OF PHILOSOPHY DEPARTMENT OF PSYCHOLOGY 2002 ABSTRACT DETECTION OF DECEPTION IN NEUROPSYCHOLOGICAL TESTING: A MULTI—METHOD APPROACH BY David Dean Cordry An experiment to determine the utility of incorporating several methOdS to detect simulated malingering of coqnitive prOblems was conducted. Patterns Of performance, on neuropsyczhological tests, specifically a magnitude of error strategy, as well as responses on a self-repOrt Ineasure and nonverbal behaviors during a . . ere Clinical inteerew were examined. College students W here assigned to either a control or an eXperimental group v1 they were a sked to simulate having a head injury for the purpose of attempting to win a $5 reward. When all variables Were Used to predict group membership, the discriminant function achieved sensitivity, or a true positive rate for malingering of 92% and a specificity, or true negative rate for controls of 100%, with an overall Results are discussed Q lassification accuracy of 95%. Within the context of the benefits of using a multi—method approach to detect feigning. To those who hear the sound of one hand clapping. iii ACKNOWLEDGMENTS I would like to thank many people who made this Particular pursuit of erudition possible. I would like to express my gratitude for my mentor and chairperson: Professor Abeles for all Of his guidance and support in writing of this dissertation’ as well as his advocacy on my behalf. Furthermore: I WOUld like t0 thank Professor Levine for all of his assistance with the statistical analyses and insiqhtf‘ll Comments into the design of this research I would also like to express my gratitude tO llent ProfesSOrS Winder and Thornton who have both been exce ‘ ' ' 'n comments and clinical supeerSOrS in addition to prOVldl g criticisms that helped guide this work. I WOUL d like to especially thank Angela McBride for all of her aSSistance in this project, not to mention all of the hours of Monopoly Junior that kept us both sang. Camilla Williams is also deserVing Of thanks for her assistance in this project. Several friends and COlleagUeS helped make this possible, PrOViding support and friendShip that sustained me through this effort. I would like to thank my family for providing me with the means to pursue this opportunitY- Their emOtional SUpport and encouragement throughout my life has made many things possible for me. iv I would like to thank my wife, Meg for all Of her love and support through this endeavor. Without her by my side, none of my accomplishments would have been possible. Finally, I would like to thank my son, Maxim for coming into my life and giving me the motivation to complete this task, showing me what things truly matter, and filling my life with more joy than I would have thought possible. TABLE OF CONTENTS LIST OF TABLES .................................................................................................. vii LIST OF FIGURES .................................................................. V111 INTRODUCTION .............................................................................. 1 Detection of Malingering: Patterns of Performance on Neuropsychological Tests ..................... 3 Personality Measures in the Detection of Neuropsychological Malingering ----------- 13 Clinical Interview in Detecting Deception; Verbal and Nonverbal Indicators of Deception ...... 23 Aims of the Current Study ............................. 33 Hypothesis 1 ............................................................................ 41 Hypothesis 2 .................................................................................. 42 Hypothesis 3 ...................................................................................................... 42 Hypothesis 4 ...................................................................................................... 43 Hypothesis 5 ...................................................................................................... 44 METHODS .................................................................................................................................... 45 participants ...................................................................................................... 45 Procedure ............................................................................................................... A6 Measures ................................................................................................................. . 50 RESULTS ............................................................................................................. 56 Hypothesis 1 ..................................................................................................... . 58 Hypothesis 2 ..................................................................................................... . Hypothesis 3 ..................................................................................................... , 59 Hypothesis 4 ..................................................................................................... . 6O Hypothesis 5 ............................................................................ 61 66 DISCUSSION .................................................................................................. Limitations of the current study .......................................... 97; Directions for further research ............................................. 94 REFERENCES .................................................................................................. 96 APPEND :[X ,, . .................. .............................. 104 I hstruction Set A .................... -- ..................... 105 I hstruction Set B ............................................................ 106 P reg—Screening Interview ...................... 107 Q linical Interview (Set A) ............................................................ 108 C linical Interview (Set B) ......................... 109 vi L IST OF TABLES Table 1. Means, Standard Deviations, and Ranges for Experimental and Control Group8 ................................................ Table 2. V'ratio, StandsardiError, Upper and Lower Bound of Confidence Interval Table 3. Means and Standard Deviations for Experimental and Control Groups Table 4. 2X2 Within-Subjects ANOVA for Adaptors 000000000 Table 5. 2X2 Within-Subjects ANOVA for Foot/Leg Movements I. a. OOOQOOIOOIOODOIIO '00 ......UOIOOIOCIOIOOIOOI0.0..0.IOCOOOIOOOO.IIIIOOIOOOIOCO0.0. 0. 0.0.0.0.... IIOOOOOOIIIIOOOOIOOI ooooooooo Table 6. 2X2 Within-SubjecfistNOVA for Voice Frequency ...................................................................................................................... Table 7. Canonical DiscrdJninant Function Coefficients Table 8. Discriminant Cfliissification Results OOOOOOOOOOOOOOOOOO Table 9. Correlations .......... ccccccccccccc o oooooooooooooooooooooooooooooooooooooooooo 00000 ----- .............. Table 10, Pooled WithinsGroups Correlations Bet Variedxles and Canonical Values ‘ween oooooooooooooo .0. ooooooooooooooooooooooo ccccccc ooooooooooooooo vii 56 57 62 63 64 65 67 68 69 7O IJIST OF FIGURES Figure 1“. .Adaptors Obselfved During Baseline and (Hinical Interview ....................................... m 63 Figure 2. East/Leg Movements Observed During Enseline and Clinical.lnterview ............................................................ 64 Figure 3. Voice FreequencyiDuring Baseline and Clinical Interview ........................................................ 66 viii Detection of Deception in Neuropsychological Testing: A Multi—method Approach I NTRODUCT I ON Deception is used to some degree in all forms of social interactions. The various motivations for successfully deceiving another person include maintaining appropriate politeness, gaining another’s approval, and attempts to deceive in order to achieve some financial reward, to name a few. The problem of deception also exists in the area of neuropsychological testing. Psychologists rely primarily on self-disclosure from their Clients in order to make an appropriate diagnosis and begin the course of treatment. However, when psychologists are asked to evaluate clients to answer questions of impairment that have financial or legal incentives for their Clients, naive assumptions about the genuineness of their clients, efforts may reflect a form of self-deception. The increasing use of neuropsychological test data to determine neuropsychological impairment in a legal Setting may in fluence a client’s approach to the testing procedure in order to achieve some type of financial reward (Lees- Haley, 1990) . The possibility of contamination of test result S by monetary or other incentives has lead to an increase in development of procedures to detect attempts to malinger—that is, deliberate Simulation or exaggeration of a disability—with varying degrees of success (Lees—Haley and Fox, 1990; Bernard, 1991; Faust, 1995). When one considers a diagnosis of malingering, 1t 15 important to note that malingering is not a dichotomous construct (Sweet, 1999) . Individuals may perform in a manner consistent with malingering on some tasks, while performing at their actual ability level on others. It is important to consider that performance on any particular neuropsychological test can not serve as conclusive evidence of malingering, just as performance on some tests that is not consistent with malingering does not rule out malingering on other measures. Furthermore, estimates of the base rate of neuropsychological malingering in actual Clinical populations have ranged from 7.5-15% (Trueblood & SChmidt, 1993) to 18—33% (Binder, 1993) suggesting that diagnosis of malingering occurs infrequently. The aim of the current investigation is to evaluate the 51: rategies used by students instructed to malinger. Specif jcaler do analog malingerers produce a different patteJ:I1 of performance when attempting to simulate a head injury , than do students asked to perform to the best of their ability, or the performance of actual head injured patients demonstrated in previous research? Detection of Malingering: Patterns of Performance on Neuropsychological Tests One approach to the detection of exaggerated performance on neuropsychological tests involves the analysis of differences in performance on commonly used tests between malingerers, controls and impaired subjects. Being able to explain the procedures for distinguishing between a malingered performance and an impaired performance is a task likely to be performed by the clinical neuropsychologist during testimony in a litigated case (Iver-son, Franzen & MCCracken, 1991) . The use of analyzing performances on existing and commonly used tests to differentiate a malingered performance represents an advantage over adding additional tests and procedures to an already cumbersome neuropsychological test battery (Bernard, McGrath, & Houston, 1993). The majority of the research literature that has been done regarding the detection of malingering on neurop SyChological tests has focused on exaggeration of memory deficits, rather than on faking of specific disorders such as learning disabilities, or focal brain lesions (Nies & Sweet, 1994). Memory complaints are among the mo st common complaints to be investigated by Clinical neuropsychologists, and are an area for additional investigation in regard to ability to detect malingered performances (Bernard, 1990). Malingered memory defiCitS and complaints may occur more often in clients motivated to receive some type of compensation because they are presumably easy to demonstrate, yet hard to detect as exaggerated (Nies & Sweet). A study by Brandt, Rubinsky and Lassen (1985) found that college students given instructions to fake a memory deficit did not differ in their performance on a free recall task from a head trauma group or a group of Huntington’s patients; although they performed significantly below a Comtrol group. However, on a recognition trial, only the malingerers scored below chance levels on a forced-choice task. Although the authors of this Study suggested that recognition might distinguish malingerers from normals or Clinical groups, they cautiOned that a substantial portion of naive malingerers performed at or above chance levels. A performance on a forcedi choice task below chance levels is suggestive of maling Qring according to binomial theory; that is, if a subjeQ 1: guesses randomly on a task where one must Choose between two alternatives approximately 50% accuracy should be achieved (Guilmette, Hart, Giuliano, & Leininger, 1994). However, since malingering represents a conscious attempt to simulate impairment, a malingering subject would be expected to monitor performance and choose the incorrect response when they know the correct one and thus perform below the level of chance (Iverson, Franzen, & McCracken, 1994). lverson, Franzen and McCracken (1991) used a 21-item word list to examine patterns of performance exhibited by college students instructed to fake memory impairment as compared to patterns demonstrated by patients with documented memory impairment. The investigators evaluated the performance of all participants on bOth a free recal1 task, as well as on a forced-choice recognition task, where SUbjects were given word pairs and asked to decide between the Word from the original list and a distractor. They found that although there was no significant difference between the free recall performance of the malingerers and the memory-impaired participants, on the recognition task malinggrers performed significantly more poorly than the memory\impaired participants. Furthermore, when cutoff pOintS were selected in order to Classify the group member ship of participants, they were able to correctly identi :Ey 100% of malingerers using a cut off score that fell in the confidence level of chance performance on the recognition task, while ODly misidenti fying one memory impai red participant. When more Conservative cutoff points were chosen, the false positive rate was reduced to 06, ' ' d to although the identification rate for malingerers drOPPe The pattern of performanCe demonstrated by participants instructed to fake memory deficits on forced" choice recognition measures may be indicative of their level. of motivation to appear impaired, Binder, Villanueva, Howieson, and Moore (1993 ) examined the . . aflma patterns of performance of patients with mild head ’6‘ with differing levels of financial incentive on a - . Is recognition memory task. In their study, the investigato divided a group of patients with a history Of mild head injury into a group that was seeking financial o Ompensation for the 11: symptoms and a group With definite ev' ldence of brain dysfunction that were not seeking financial Compensation. The group seeking financial compensation was fUrther diVided based on their performance on a task measuring motivation (Portland Digit Recognition Test; PDRT) - Binder and his colleagues found that the mild head trauma group With low motivation, as measured by Poor PDRT performance, demonstrated significantly worse scores on the rec"O‘B‘Dition trial Of the Rey Auditory Verbal Learning Test (RAVL'I‘) than did any other group. The ir results suggest that: analysis of performance on recognition measures may be useful in discriminating groups of patients who are exaggerating deficits and patients With actual memory impa i rment . Bernard (1990) found that Using stepwise discriminant function analyses, he was able to achieve a 77% accuracy rate in prediCtj—ng participants group membership as either a control or a Simulated malingerer. His Study found that . , :98 performance on the recognition trial of the RAVLT and were recall performance on the Complex Figure Task (Cfrfl the primary indicators in his analyses that detected . . . . . . . . tiO“ malingering PerthipantS- USing the discriminant funC from this study, Bernard, Houston, and Natoli (1993) achieved an overall accuracy rate of 86% in th . Gir h ' . subseqUent study, Furthermore, t ey did not mlSidentify any control subjects as malingerers and had a 75% true positi ve rate in identifying malingerers. Bernard (1991) extended his analysis of patterns Of performance on neuropsychological tests to detect malinggring to include an analysis of the serial position effectS during free recall on the RAVLT. He hYPotheSlzed that malingerers might produce a characteristic pattern Of free recall in their attempts to exaggerate memory deficits that would discriminate them from cont. r015. Bernard foUnd that a grOUp of college StUdentS given instructions to simulate a memory deficit recalled significantly fewer ' the words from the first third of the word list than from last third. Bernard interpreted the absence of a U—shaped recall curve, which was demonstrated by controls, as indicative of the intentional suppression of items from the first third Of the word list by malingering participants. Furthermore, SUbjects with closed head injuries, having documented brain injury demonstrated a recall patter“ d a e opposite of the malingerers; that is, they demonstrat third significantly greater level of recall from the first of the iist than from the last third. This finding suggests that analysis of the pattern of free recall may be of use in differentiating malingerers from pat. lents with actual neuropsychological deficits. The use of the serial position Of free recall h as Only been reported in two literature studies and has had incons istent Slipport in its use for distinguishing betWEen malingQrers and control participants. Bernard, HouStorl and Natoli (1993) found that simulated malingerers demonstrated a U—sh aped curve on a free recall task similar to that Of Contro :Ls, Suggesting that the use of serial position patterns of performance on free recall may not be a solid indicator of exaggerated deficits. Likewise, IVerson/ Franzen and McCracken (1991) were not able to differentiate patients between malingerers, contrOlS Or memory disordered . . . . ee on the basis of serial position Curves produced during fr - rch recall on a verbal memory task; however, their reSea protocol did not involve repeated presentation of a word list- These findings suggest that their exists a considerable amount 0f variability in the performance of participants asked to malinger. Furthermore, Wiggins and Brandt (1988) found that participants instructed to malinger displayed serial position curves Similar to normals on a verbal memory . . . . . O Wiggins and Brandt asked their partiCipants to reSPOnd t test items as thou h they had memor roblems . g y p ' but did “Qt include any simulation of secondary gains. Int erestingly, they did find that the malingering grOUP performed Significantly below a group of memory—disordered patients on a recognition measure but not on a free-recall measure. Many Studies of malingering Of memory defiCits have relied on the analysis of one memory measure - However, in clinic;- El practice several measures of memory may be admini stered in addition to tests specifically deSigned to detect, malingering- Iverson and Franzen (1996) utilized a battery Of five obj ective assessment measures to enhance their accuracy in detecting experimental malingerers. They established cutoff scores for which there were no false positives and compared these scores for their individual ’ ' ' ilar to accuracy in correctly cla551fylng participants- Sim previous studies: they fourld the forced--choiCe recognition task was effective in correctly identifying 69% Of experimental malingerers. By combining all of the cutoff scores from the memory measures 115861, and USing deficierlt performance on any One of the measure 8 used, resulted in the most accurate classification rate of 92.5%. This Ct finding supports the use of a battery of tests to date malingered Performances . Trueblood and Schmidt (1993) used a below Chance performance on a forced-choice test (Symptom Validity Test? SVT) to detect malingerers from a group of patients referred for Henropsychological evaluation. In additiOn they identified patients who demonstrated questionable Validi ty on neuropsychological procedures. These resear: Qhers matched participants with malingered and Questi gnable performances with controls and used a variety 0f neubopsychologieal procedures to discriminate between malinggred performances and performances of head injured patients. They established cutoff scores that ' ' ' - . - - - 'ti dlscrlmlnatEd malingerers while minimiZing false p051 Veg! X0 and found that the California Verbal Learning Test (CVLT/ Recognition score was one Of the measures that performed W811 as a validity indicator, identifying 9 Of 16 probable to identify 13 malingerers, They found that they were able of 16 probable malingerers when they used several measures and found more than two teSt performances falling below the malingering CUtoff . The utilitfil of the CVLT in identifying malingering in participants with more specific instructions On faking port! amnesia has been also been demonstrated (Coleman, Rap Millis, Ricker, & Farchione, 1998). Coleman and her colleagues found that analog malingerers who were COac d on how to fake memory prOblems, performed Similarly to hea in'ur atients e- ' ' . 3 ed p on fre recall indices, howerr’ on more subtle measures Of learning such as recognition 811C1.Sllape of learning: they performed similarly to na‘i've mal . ingerers Coleman et al. used logistic regression and Were able 1: O aChieVe a Specificity rate Of 85°2% and sensitivity (to malingering) Of 90.5% by lnClUding recognition The CVL'I‘ has also been discriminability in their analyses. found 1:0 be useful in discriminating between minor head from pat ients Wi th injury patients with poor motivation severe head injuries (Mil lis, Putnam, Adams & Rickerr 1995\ - Discriminant function analySiS Of several CVLT XX variables correctly ClaSSifiEd 91% Of the patients With minor head injuries according to Millis and his colleagues. They found that the CVLT demonstrated both a high degree of sensi tivity in detecting poor motivation and a high degree t5 With of specificity in correctly identifying patien documented severe head injuries. Another Strategy for detecting malingering involves the analysis Of magnitude of errors. Martin, Franzen, and Grey (1998) had 10 independent raters assign a probability of selection value to each answer choice on a multiple, tic“ Oi choice recognition task, which followed the preSeDta ax a story that participants were told to remember (Logic Memory; Wecshler Memory Scale-Revised) . Martir1 and hi5 collea es avera S . . . gu ged the election probability Values and were able to achieve a .85 inter—rater agreemQht - They found that students given instructions to fake m emory impairment and patients who were identified as having questionable motivation were Significantly Imore likely t 0 Select: low probability responses than either controls or patien 1:5 with documented Closed—head injuries. The lnc30rp©ration of a magnitude of error approach provided additi Qnal Classification accuracy beyond that: found when jUSt the total recognition memory score was USEd- Furthermore, certain items were only selected by the l7. malingering and questionable mOtiVatiOh group, SuggeStihg that the magnitude of error apprOach may result in «critical items” that provide additional indications Of This strategy is also exaggerated memory impai rment . ied to advantageous in that It can potentially be appl— neuropsychological tests that are COInInonly used during asses smentr once probability Values are assigned- Personality Measures in the Detection of Neuropsychologica Malingering ’2 The Minnesota Multiphasic Personality Inventory mail (MMPI—Z; Butcher, Dahlstrom, Graham, Tellegen, 6‘ gae 00159 1989) is one Of the most widely used measures in the C of neuropsychOlOgical assessments (Lezak, 1995) Assessment of emotional adjustment and personal . lty is a routine part of neuropsychological evaluations Furthermore, the MMPI—Z CONS.lStS Of a number of items that describe neurological symptoms that are associated with head injury. When head injuries are assessed as Part Of a litiga tiOn prooess, the forensic nature of these evalua tions raises the possibility of response diStOrtiQn and e): aggeration on all measures, including the M'MPI’Z. One adVantage of using the MMPI—Z in a forensic conteXt 13 X3 the extensive body Of research on the Validity Scales incluCied in the measure - IVIalingering of symptoms for the purpose of secondary ance OH gain can occur as either deliberately poor perform . tic neuropsychological measures, GXaggerat ion of soma complaints, or as a combination of bOth (Larrabee' 1998) . Although no specific closed head injury profile has been established, Certain trends of performance have been demonstrated (Berry et al., 1995). Generally, patients . . 1e5 with closed head injuries produce elevated Scores on sea lwafls F, l, 2, 3, 7 and 8, although these scores are not a elevated to Clinical levels (Alfano, Neilson, pafliak' FinlaYSOnr 1992? Gass & Russell, 1991; Diamond, Barth 8‘ Zillmer, 1988; Gass, 1991; Beery et al., 1995) The ‘ specificity of WEI-2 profiles for closed head injured patients imPrOVeS When the presence Of litigation and other factors influencing motiVation are taken into account and Specific disSimulation scales are used (Larrabee, 1998) S Qveral WPI ——2 Scales have demonStrated utility in detect jng diSSimulation of psychiatric symptoms, and have of def :‘icits in a neuropsychological context. The F (Frequ ency) scale is a validity scale on the MMPI—Z that consists of answers that are infrequently endorsed in the Hi scored direction by either disturbed individuals or normals (Lees—Haley, 1991). Persons who demonstrate elevated F scores typically admit to a wide range of psychological problems. The Fb scale (back F; Butcher et al., 1989) was constructed in a similar manner as the F scale, only uslng items taken from the second half of the MMPi—Z. The Dissimulation Index, or F — K (Gough, 1950) was also developed as a method for detecting malingering of psychiatric syndromes. Obvious and Subtle subscales (0 ' S; Wiener, 1948) were also developed as a means of . l detecting malingering based on the number of items common y kUOWn to indicate psychopathology Versus the items that repitesent more subtle indications of pathOlOgY- The Dissimulation Scale (D5; Gough, 1954) was originally constructed of 74 items from the original MMPI' thought to assess inaccurate stereotypes Of neurotic pSyChOpathOlOgy and has been updated using 58 items on the WWI—2 (D8 2’; I Berry, et al., 1995) . Another validity scale that has b Sen described by Arbisi and Ben—Porath (1993) as the F e Psychopathology (Fp) scale consists of 27 items rarely endors ed by psychiatric patients. Lees—Haley! EngliSh and Glenn (1991) presented an additional scale, the Fake Bad Scale (FBS) Specifically designed to detect malingering among personal injury litigants. 15 In a comparison of MMpI_2 SCOres of personal injUry claimants who were suspected of malingering and litigants not su5pected of malingering, Lees—Haley (1991) found that suspected malingerers produced both higher F scores and higher F - K scores. However, the Utility of the MMPl _ 2 for detecting malingering in neurOpSychological testing. was not specifically addressed in this study, as patients with other types of claims, such as Post-traumatic Stress Disorder were also included in the sample. Analog malingerers have been found to Produce Similar overall levels of deficits, although their performance Patterns are distinguishable (Heston, Smith, Lehman, & Vogt, 1978) , Heaton and his colleagues found that participants instructed to fake a head injury produced Significantly higher Scores on the F scale as Well as on Cll'nl'CZal scales 1, 3, 6, 7' and 8 than non-litigating h sad- injured participants. Furthermore, they found that he ad- injUred patients and the malingering participants perfOr Ined poorly on different neurOpSYChOlOgical measures. Heaton O. and hi 8. colleagUes were able to correctly classify 1006 of their participants using discriminant function analySiS Of the neuropsychological tests alone. In a separate discriminant function analysis using the MMPI SCOres, 94 O of their participants were correctly claSSifiedr SUQQGSting 16 that the MMPI may be Of use in detecting malingering When used in conjunction With other neurOpsychological measures. Berry et al. (1995) compared mp1 -2 validity scales of normal controls, partiCipants asked to feign head injury and given a description of closed head injury symptoms. litigating head injury patients, and non-litigating head injury patients. Berry et al., used the F, F — K, Fb, Ds and Fp Scales in their analyses and found that the analog malingerers were distinguishable from the normal controls on all the included scales. Furthermore, litigating heald injUIy patients were distinguished from non-litigating patients on the validity scales, and produced a simila‘: Pattern of elevated scores on the clinical scales as the ana log malingerers . The presence of elevations on clinical Seales 1 and 3 are Often seen in neuropSyChOlOgical settings because the pain, paresthesias, and malaise (Gass, 1991)° However only one F scale item is included on these scales, which has led some investigators to question reliance on the F scale to detect malingered performances (Larrabee, 1998) The PBS was designed to detect a mixture of faking good and faking bad response styles Seen in personal injury Claimants (Lees-Haley, English, & Glenn, 1991). Lees-Haley 17 et al. hypothesized this combination of response styles, as representing several goals of a malingerer. These goals include an attempt by malingering persons to appear honest, appear pSyChOiogically normal eXCept as influenced by the injury, and avoid admitting pre~existing psychopathologYI while presenting a degree Of Complaints that suggests a believable degree of disabilitY- Larrabee (1998) found that the standard validity scales F, F- K, Fb, and Fp were insensitive to detection of exaggerated somatic complaints in patients Who performed below suggested cutoff scores of neuropsychological tests Oi malingering. In addition to an elevated FBS scale, Lalfrabee found that medically and neurologically normal litigants demonstrated elevations on scale 1 and 3 greater than a group of chronic pain patients and a group of . . 1’dead injured patients, The FBS demonstrated Good sensitivity deteCting 91% of patients CODSidered to be malingerih I based on other test performances- Millis, Putnam and Adams (1995) reported Similar findings in their investigation of mild head injury patients involved in litigation, scoring below chance, in(heating probable malingering on a neurOpSYChOlOgical test. Although the litigants in their study produced significant elevations over patients with moderate to 18 severe head injuries on Validity Scales F, Fb, and Fp, and F33, the FBS was found to have the best diagnostic efficiencY- Furthermore, Significant elevations on the Clinical scales 1, 2: 3, 7 and 8 were found in the litigating head injury group. Slick, Hopp, Strauss, and speiiacy (1996) found that scores on the F, E‘ — K, and F35 were Significantly correlated with performance on the Victoria Symptom Validity Test (VSVT) , a test specifically designed to detect malingering; however, the PBS was the most strongly associated with neuropsychological test per fiormance . Schretlen and Arkowitz (1990) examined the strategies employed by prison inmates to fake either mental retardation or mental disorders. The performance of the inmates was compared tO clinical groups Of mentally retarded persons and psychiatric inpatients, and discriminant function analysis was used to predict group membership. Schretlen and Arkowitz found that both tYpes Of fakers (mental retardation and psychiatric disorders) demons trated significantly higher F and F _ K values than did “0 h—faking groups. Furthermore, the scores Of both faking groups on neuropsychOlogical tests of malingering were also significantly worse than non-faking groups. Their findings supported the hypothesis that the use of a 19 battery of test to detect malingering increased diagnostic accuracy over the use of a single measure, as they were able to correctly CiaSSifY 95% Of participants using discriminant function analysis of multiple variables. The influence of litigation upon MMPI-Z profiles of minor head injury patients and moderate/severe head injury patients was investigated by Youngjohn, Davis and Wolf (1997). Youngjohn et al., found that in a group of 30 moderate or severe head injured patients referred to their clinic 18 were in ongoing litigation; however, in a group Of 30 mild head injured patients all were in ongoing litigation. The investigators found that the litigatihg severe head injury group produced significant Elevations 0n the MMPI-Z clinical scales 1, 3, and 8 relatiVe to the non— iitigating severe head injury group. Mild head injured patiel‘lts produced elevated scores on Clinical Scales 1, 2 3' and 7 relative to both the litigating and non-liti gating Severe head injury group. Youngjohn et al., noted the apparefit paradoxical effect that litigation had on the profil es of less severely injured patients; that is, patients with less severe injuries produced the most signif icantly elevated profiles on the MMPI‘Z- The authors suggested that although litigation may have some effect On mild head injury patients. causing them to endorse more 20 pathology, they hypothesized those patients with Significant emotional difficulties or psychopathology may be more likely to pursue financial compensation. Interestingly, they did not find any difference on the standard validity scales of L, F, or K between the different groups . However 1 their lack of findings may be due in part to the inconsistent utility of the F scale in detecting exaggerated profiles as well as their relatively small sample (N=30). The incorporation of other validity scales, particularly the FBS may have been appropriate for their participants, as it was desigTIEd to detect mal ingering in personal injury claimants. Other investigators have also found cognitive and Psychological malingering to produce independent reSanse patterns on the MMPI-Z and neuropsychological tests of maiinCJering (Greiffensteinl Gola, 8‘ Baker, 1995) - Grief fenstein et al., diVided participants who were referred for neuropsychologicai assessment into three groups , a traumatic brain injury group, a probable maiingering Group: and a persistent post-concussion group The Probable malingerers and post—concussion group were diSti—rlguished by the postcCOncussive group having returned to Previous employment and not being involved in third— party lawsuits. That is, the persistent Poet‘Concussion Zl group was involved in lawsuits to recover medical expenses, while the probable malingerers were seeking jury awards through third-party lawsuits. Using a known—groups method, Greiffenstein et al., found that domain specific measures were generally more sensitive to noncompliance than were MMPI—Z measures. The researchers found that the validity scales F, F — K, and O — S were more closely associated With a psychiatric malingering factor than a r1eurOpSychological malingering factor . However, they did not incorporate the FBS in their analyses, which has shown some utility for detecting exaggeration of deficits in per sonal injury litigants. The influence of litigation Upon standard cognitive and personality measures has been a relatively consistent finding (Binder & Rohlingr 1996). In a meta‘analytic reView of studies done on personal injury litigants, Binder and Rohling found a moderate effect size of 0.47 for head trauma. Of particular interest was the confirmati 0 Of Previous findings of the paradoxical effect of llthation that abnormality and disability increased with less 8 evere injuries. 22 clinical. Interview in Detecting Deception; Verbal and Nonverbal Indicators of Deception The clinical interview is another standard feature of neuropsychological assessment, during which the clinician gathers information regarding the patient's complaints, premorbid functioning, and details of the incident in question. The clinical interview provides an opportunity for the neuropsychologist to compare the patient's presenting complaints with known findings from neuropathological examinations (i.e., electroencephalogram“ computed tomOgraphy scan) to evaluate the presence of inconsistencies (Franzen, Iverson, & McCracken, 1990) . Altkioug‘n some authors have questioned the Utility of the interview for detecting malingering Of PSYChlatric diSOrders empirical evaluation of the usefulness of the inter‘dew in detecting neuropsychological malingering has not been undertaken (Ziskln' 1984; Franzen, IVerson, McCra Cken) . Research in social psychology laboratories has invest igated nonverbal behaViors that are associated With lying; however, the application of this data has yet to be SYStematically applied to the detection of malingering in ClinlC‘:.al psycholOgy contexts. Although structured interViews have sthn some promise for their utility in 23 detecting different malingering strategies, the ability of clinical psychologists to be trained in detecting nOHVerbal indicators of lying and detecting false or exaggerated statements in a clinical interview has yet to be examined (Rogers, Gillis, Bagby and Monteiro, 199M. Deception occurs when an individual attempts to convince another person tO believe Something that the deceiver considers false (Zuckerman & Driver, 1985), Folk Wisdom purports that nonverbal means of COInmunicating are of fundamental importance in detecting deception primarily because it is assumed that nonverbal behaviors are largely inV01untary and hence more diffiCUlt tO fake than are verbal messages. However, the deceiver may more easily control some nonverbal channels of communication than others. Ekman and Friesen (1974) found that relative to the face, the body tends to be a better source Of Cues that . 0 0 o _ . ' deception lS occurring. Furthermore, an lnd1Vldual 8 tone and frequency of voice, provides additional cues that a person's communication is deceptive (Zuckerman, Amidon, Bishop & Pomerantz, 1982) . The literature in the area of deception detection suggests that rather than a dichotomy of verbal versus nonverbal indications of deception, a hierarchy of communication channels with differing degrees 24 of controllability exist for examining the veracity of a speaker (Zuckerman & Driver, 1985). Reviews of the literature regarding the detection Of deception suggest that na'i've judges of deceiverS achieve accuracy rates ranging from '45 to ~60 (Zuckerman, DePaulo, & Rosenthal, 1981). Detection accuracy rates were improved when either bodily cues or Verbal Cues, such as tone and frequency of voice: were used rather than facial expressions, lending support to a hierarchy of communication channels model (Zuckerman et al., 1981). In explaining the behavioral cues associated With deceptive messages, two models emerged that appear t0 have that most relevance for the application of this area of research to detection of. malingering on neuropsychological tests Physiological arousal has received some support as a mechanism through which behaviors indicating deception are produced (Zuckerman & Driver, 1985). Autonomic responSe such as skin resistance and skin conductance differ When an individual is telling the truth or is being deceitful (Lykken, 1974; Waid & Orne, 1981). Furthermore, certain behaviors are considered to be associated with deception specific arousal, such as pupil dilations, voice frequency, eyeblinks, speech errors and speech hesitations (Zuckerman & Driver, 1985; DeTurck & Miller, 1985). 25 Cognitive factors also have a potential role in th 9 production of behaviors that indicate deception. Wh Erl . . an indiVidual is required to lie On an experimental t ask, he or she must monitor several Channels Qt commuhic at- lQn wh. ile trying to convince a judge that their message Depending Upon the type of decreption quuir ed of the this task can become qutL ‘te Complex and participant, Such an increasg . in Cognitive taskS cognitively taxing. i may lead to the production of such bghavior l indicators 0 a deception as speech errors and pauses, longer response i9853° latency, and pupil dilation (Zuckermar1 & Driver, Some behavioral indices, such as pupil dilatiQn may be accounted for by more than one theoretical model and therefore the underlying mechanism for their 10er ucti during deceptive communication may not be diSce rnabl (Zuckerman & Driver, 1985) ' A meta-analytic review Of the literature 0 11 COrt\ of deception found that eight verbal and nonverbal e behaviors distinguished reliably between lie telling and truth telling (Zuckerman & Driver, 1981)- Of primary relevance to the current study were the following identi fied behaviors: adaptors (behaviors unrelated to verbaL content; e- g., scratChing), speech hesitations and The effect sizes for these indicators VOlCe frequency. "2.6 ranged from d=.38 for adaptors, 07:54 for Speech . . Wh‘ hesitations , and d=.68 for VOlCe frequency lCh rep]: eSent small to medium effect sizes (ZUCkerman & Driver, 19 81) Although, negative statements (i.e,, frequent dist) aragin g statements) and irrelevant lnfcrmation were ident- lfied as the analysis of indicators of deception, the s peCific content of deceptive communiCation ”la de by neuropsychological malingerers is beyond the ope of the SC current investigation. DeTurck and Miller (1985) Se 90“ 5 identified hand gestures, leg/feet gQStureS/ and re latency as additional indicators that significantly contributed to the identification Of deceptive communication. DeTurck and Miller (1985) found that perSOnS instrUcted to deceive demonstrated Significantly g re . . s t Sympathetic arousal than did non—deCElVers. FUrther er they :found that adaptors/1land gestures: speech erro ‘th l" S I and reSpOhse latencies increased in frequency and intensit among deceivers relative to non—deceivers. DeTurek and Miller also found that aroused deceivers could be dietihguished from aroused truth tellers on the basis of the Six verbal and nonverbal behaviors they identified, 2'] suggesting that these behaviors are specific to arousal induced by deception rather than being unique to general arousal. One concern regarding the application of nonverbal and verbal behaviors to the detection of malingering is the process of training observers to reliably identify these behaviors and make judgments. In a recent study by Kassin and Fong (1999) college students were given training in verbal and nonverbal behaviors for identifying deceptive communications; however, the accuracy of their judgments were no better than untrained controls, despite their having more confidence in their judgments. However, DeTurck, Harszlak, Bodhorn and Texter (1990) found that training in nonverbal indicators of deception significantly enhanced the accuracy of social perceivers in detecting lies. Furthermore, psychologists specifically interested in deception were more accurate in identifying lying than were other groups of psychologists (Ekman, O'Sullivan, & Frank, 1999). One factor that contributed to the different findings was differences in the training. Kassin and Fong provided portions of police interrogation training, while the other studies provided training focused on identifying specific behaviors. 28 Accurate judgment of communication as being either truthful or deceptive is also affected by familiarity with the deceiver (Brandt, Miller, & Hocking, 1980). DeTurck and Miller (1985) suggest that in order for a judge to perceive another person’s verbal and nonverbal behaviors and make an accurate decision about the presence of deception, exposure to the deceiver's unaroused truthful communication is necessary to provide a baseline against which subsequent behavior is evaluated. The lack of exposure to a previous baseline may account for some of the findings of poor judgment accuracy. Although a baseline of unaroused and truthful communication may be more difficult to obtain in clinical practice of evaluating persons who may be malingering, it seems possible that an initial period of interaction prior to the beginning of the evaluation may be slightly less arousing than the actual examination process. In other words, until a patient is asked specific questions regarding their disability, less deception specific arousal should be present. The motivation of participants to lie successfully is another factor that has an impact on how easily they are detected. Many studies in this area have utilized monetary awards for successful performance in being deceptive (Zuckerman, DePaulo, & Rosenthal, 1981). In their meta- 29 analytic examination of research on lying with varying degrees of motivation, Zuckerman and Driver (1985) found that most visual behaviors (i.e., adaptors and blinking) showed a decrease in frequency and/or intensity when participants were highly motivated (offered a monetary reward) versus low motivated participants (no financial incentive). Furthermore, high motivation was associated with shorter response lengths, slower rates of speech, and increases in voice frequency than was low motivation. In other words, nonverbal behaviors were differentially impacted by changes in motivation to deceive. If conditions of high motivation result in greater arousal than low motivation conditions, deceivers might be expected to engage in more eye blinks, more adaptors, demonstrate more speech hesitations, and increases in voice frequency. However, Zuckerman and Driver (1985) found that only voice frequency and eye blinks demonstrated the expected pattern. Zuckerman and Driver suggested that when deceivers are highly motivated they might expend more effort at controlling their behavior, which results in a reduction of some indicators of deception, such as body movements and length of communication. Another possibility is that simply increasing the reward for successful 30 deception does not necessarily increase arousal during the task; rather, arousal may also be affected by possible punishments if caught. DeTurck and Miller (1985) suggested that high motivation, in the form of monetary awards, might produce enough justification to reduce dissonance arousal and thus diminish behaviors associated with deception specific arousal. However, in the examination of malingerers, high motivation in the form of jury awards or reduced responsibility is of particular interest. Furthermore, DePaulo, Lanier and Davis (1983) found that when motivation was manipulated, highly motivated participants were more accurately judged from their nonverbal behavior, whereas low motivation participants were more accurately judged from their verbal behavior. Given the difficulty in approximating the high motivation of actual malingerers to succeed at deceit for a rewarding payoff and to avoid the costly consequences of being caught, analysis of both verbal and nonverbal aspects of behavior are of particular importance for simulated malingering studies. It is crucial to the investigation of verbal and nonverbal behaviors of deception to evaluate several ' ' ' ’ ' ' ntrol channels of communication, as deceiver s ability to co certain aspects of behavior is variable (Ekman, O’Sullivan, 31 Friesen, Scherer, 1991). Ekman et al., found that when several behaviors were combined (voice frequency and smiles) the hit rate for detecting deception was 86.4%, suggesting that while some deceivers may be able to control some nonverbal behaviors when lying, other indicators are apparent. When considering malingering in a forensic neuropsychological setting, it is important to remember that the details of a supposed disability communicated during an interview by a person who is malingering, probably does not represent a completely false statement. That is, an individual who has sustained a mild head injury is likely aware of the symptoms they have experienced and in reporting them to the evaluating neuropsychologist may exaggerate them without manufacturing untrue symptoms; thus, it may be particularly difficult to detect malingering through behavior analysis. However, malingering requires both deliberateness and the pursuit of an external reward to be diagnosed (Sweet, 1999). It is the deliberate attempt to deceive in addition to the possibility of receiving a reward that should contribute to the increase in arousal felt by persons intending to 32 malinger, which should result in an increase of the behaviors that have been shown to be effective in detecti deception. Aims of the Current Study The purpose of the current study is to investigate the application of multiple techniques to the detection of deception on neuropsychological assessment. Of particular interest is whether or not evaluation Of Verbal and nonverbal indices of deception are detectable during a simulated clinical interview with participants instructed to believably fake the neuropsychological seqUelae of head injuries. The first stage of this study involves the appLiCation of a magnitude of error strategy to detect malingering The magnitude of error strategy is lbased on the theo . Simulators respond in a characteriStically different In: that than do impaired patients or normal controls (Rogers, liner Harrei, & Liff, 1993). SPeCificallyz the magnitude Of error strategy prediCtS that since malingerers are aware Of the correct answers on some neuropsychological teSts, they intentionally give inaccurate responses in order to appear more impaired. The analysis of the qualitative differenCeS in patterns of responding among participants instructed to 33 fake versus those that are instructed to do their best: should reveal that fakers are more likely to provide u h incorrect answers that are grossly incorrect. Altho g , \ ' ' re near some research has found malingerers to provide mo ' ' bserved in miss” errors, apprOXimate answers are also 0 honest responders suggesting that grossly incorrect responses may improve discrimination more than ‘near misses” (Rogers, et al.) . The application of this technique is best suited to neuropsychological tests that have fC31—”Ced‘Choice answer8 such as the recognition trial 0f the call'fornia Verbal Learning Test (CVLT) and Raven's Progressive Matrices (RPM). The first step in this approaoh inVOi-Ves the determination of probabilities of selection for the it ems contained in the measure- Martin, Franzen and Grey (19 accomplished this on an adapted recognition trial f0 98) the Wechsler Memory scale— Revised, Logical Memory subteS by haVing graduate students, and faculty rate the prObabil- lty of being selected for each item on a multiple—choiCe recognition trial. After averaging across each rater th ey arrived at probability values and were able to achieve inter‘rater reliability Of '85- 34 One advantage of the magnitude of error aPPrOaCh 15 that it has the potential for generating POSSible “critical items” which may be useful in identifying POSSible malingerers in a clinical setting. Furthermore, it inCOrporates already used measures and does not add significantly to the length Of the 1’1europ'syolfmiogc.131l test battery. It is hypothesized that participants instructed to malinger will select more low-prObability items than will controls on the CVL'I‘ and Raven' S Progressive Matrices, and the use of the magnitude of erro r approach will offer additional classification accuracy beyond that found with a calculation of recognition hits alone. The application of research from social DSYCholoqy regarding deceptive communication is also of interest in evaluating the utility 0f the magnitUde of error appr Cach Speci fically, it has been hYPOtheSlZed that the proCS s 8 Of intentionally deceiving another can be cognitively challenging (Zuckerman & Driver, 1985). In the current study this seems particularly relevant, as participants will be asked to convincingly simulate a brain injury and be given specific information about the symptoms associated 35 that the cognitively taxing nature of deception can lead to the prOduction of SUCh behavioral indicators as longer response latency. When participants are attempting to simulate deficitsp they must monitor their performance to ensure that they are appearing impaired, while still seeming believable, which should be a cognitively complex task - The magnitude of error strategy would predict that fa kers would be more likely to deliberately select grossl y incorrect responses. The additional decision making proce S8 of the individual attempting to fake should lead to increased response latencies due to the additional cognitive reSOurces used to monitor their performance. Therefore, it is hypothesized that by recording responses Of participants, SignifiCant increases in response latency Will be demonstrated b 5’ th . e malingering groups, relative to controls particularl On forc ed—choice test i terns - It is also one of the aims Of this StUdy for investigating the utility of the ”4131-2 for differenti ‘ g bEtWQen controls, and indiVidualS given instructions to fake specific cognitive dYSfunction. Some investigators have foqu that faking on the MMPI—Z and on neuropSYChological tests represent different kinds of attempts to deceive (GreiffenStein' Gola, 8‘ Baker, 1995) 36 However, certain scales such as the FBS have not been routinely included in the analysis of the MMpI—z, which may have contributed to the discrepancies regarding the MTMPI— 2: 5 utility in detecting neuropsychological malingering in the literature. The FBS has shown Some promise for use as a means of detecting exaggerated performances partiCM—axly with personal injury litigants (Lees ‘Haley, English, & Glenn, 1991; Larrabee, 1998). Furthermore, other researchers have found that participants instructed to fake mental retardation also demonstrated severe emotional disturbance (Schretlen, & Arkowitzr 1990) Suggesting that some malingerers do not differentiate betWeen cognitiVe and emotional impairment Wei 1 - In the current study, performance Of the two grOups on the MMPI—Z is of partiCUIar interest. The Faking Ba Q SCal . . , e clas sification accuracy 0f partiCipantS given instru Ct . ions to fake a head injury, The items Selected for this are primarily associated Wi th somatic complaints that are endorsed by patients with minor head injuries, suspected to be exaggerating impairment. The FBS also consists of it ems that provide information about the vehemence With Which participants attempt to portray themselves as being honest 37 It is also of interest for the current study to determine the extent to which verbal and nonverbal . w information that is conveyed in a mock clinioal intelfVleto is useful in classifying indiViduals who are attempilrig i simulate a head injurY- In Clinical situations indl‘udua S seeking evaluations of head injuries have potential rewards to gain, In the case of head injuries, the reward is often in the form of excuse from work or financial awards by juries. The clinical interview is a standard part of most neuITOpsychological evaluations and may provide the trained observer additional cues to the possibility Of symptom exaggeration, through the observation of behaviors associated with deception~ The application Of the analysis and observation of behaViorS associated with deceptign to actual clinical populations is beYOnd the SCOpe of the current investigation. Furthermore, it is importan»C o realize that malingering in clinical populations exists on a Continuum of persons deliberately trying to fake an injury to receive monetary awards, to individuals Who a e less aware of their reduced motivation to perform at the best of their ability on neuropsychological evaluations. Iodividu is who are less aware of their reduced motivation a I to perform poorly, may experience less arousal While 38 ”l discussing symptoms with a neuropsychologist; hence, these ssociated individuals may not display the behavioral signs a with deception. However, just as with any currently used - - - . . - ot measure of malingering, just because an indiVidual does n score below a cut—off scolfe does nOt mean that he is “Qt malingering, nor does the presence Of a single questionable performance indicate that he 15° Observation and analysis of behaviors associated with deception during the clinical interview may be useful in detecting certain approaches to malingering, Specifically persons intentionally attempting to fake may be detected in this way. It is the aim of this investigation to determine if ratings of nonverbal behaviors by trained observers are able to accurately claSSifY individuals instructed t. . . . O feign cognitive impairment and lndWldualS instructed to De . . rfo to the best of their ablllty° rm The analog malingerers in the current investi . . . . gatio will be offered a financ1al incentive for producing successful deceptive performance in order to Simulat e the incentive that head injured individuals might roceiv a 311ry. Some have questioned the effect offering f' lnancial incentives has on deceptive performances (DeTurck and Miller, 1985) . It is the Specific aim of this Study to attempt to replicate real world situations in Which 39 malingering may occur, by Offering a financial reward for successfully appearing impaired Without being detected as faking and by trying to simulate an actual neuropsychologcial evaluation. Furthermore: the possibility of receiving a SUbStantial financial incentive has been demonstrated to result in increased autonomic arousal, which should result in an increase in behaviors indicative of deception (GUStafson ‘5‘ Orne, 1963). Detection of deception through observation of verbal and nonverbal behaviors can be imp3-”C>‘ved by the provision Of a baseline observation of an individual’s behaviors while less aroused (Brandt, Miller, & Hocking, 1980). Although someone attempting to malinger 0“ a neuropsychOlOgical evaluation is likely to EXPerience autonomic arousai throughout the investigation: it is likely that the . lei/e1 Of arousal will increase during Specific questions as duririg the interview in regard to SYInptomS and the ked individual's experience Of their supposed Cognitive deficits. In order to investigate this hYPothesis, participants in the current investigation Will be Video and audio taped during a brief screening interview prior to their participation in a mock neuropsychological examination. It is hypothesized that participants instruCted to ma]_ inger will demonstrate increased Voice 40 frequency, response latencies, adaptors, and foot/leg movements, during a mock clinical interview than during their initial screening- Additionally, it is hypothesized that the malingering group will produCe Significant . ' . 0 increases in the aforementioned bEhavlors relative t controls. The hypotheses are: Hypothesis 1. Participants in the thalingering group will select more low-probability 3-th than W111 control Participants on the CVLT and Raven’ 8 Progressive Matrices. Previous research by Martin, Franzen and Grey (1998) has found that malingerers were more likely to select low probability items than controls on a forced‘Choice memory task. Other experimenters have suggested that the low- prObability item method may be applied to Other neuropsychological testS- SUpport for this hYpothes. s WOUld demonstrate the utility Of the lOW—prObability approach on commonly used neuropsychological measure memory and of nonverbal reasoning. Low-probability ite ms have not been investigated on the CVLT or RPM: twO Co mmOnly used neuropsychological test. Furthermore, low‘probability items, if effective in detecting malingering, can lead to the generation of “Critical1 items” that can qUiCkly alert 41 / the examiner to the possibility that faking is occurring and additional scrutiny is needed. Hypothesis 2. Participants in the malingering group will demonstrate longer response latencies on the recognition trial of the CVLT and on Raven's progressive MatrieeS relative to control participants. Research in the area of nonverbal indicators Of deception has demonstrated that longer response latencies are typically seen during deceptive comrmmication (ZUCkerman & Driver,1981; DeTurck & Miller 1985) . The additional cognitive load of selecting the correct answer: monitoring if enough items have been missed, and making a decision as to whether or not to giVe an incOIrrect a‘f‘swer and Which incorrect answer to give ShOUld reqUire additional time for a response in participants atterflf)ting to fake a brain injury. The present StUdY is unique in that it involves applying analysis Of nonverbal behaviors to evaluating malingering on neuropsychological tests. fiYPOthesis 3. Participants instructed to fake a bee. injUry will produce higher scores on the PBS of the Ma: relative to normal control-5° 1‘2 The Fake Bad Scale (FBS) has been found to be a uSeful measure in detecting malingered performance on neuropsychological tests (Slick, Hopp, Strauss, & Spellacy, 1996) Incorporating the MMPI-Z into a test battery Provides a more aaipplfoxj-mate testing Situation and allows for an analysis Of the self-report styles Of malingerers 42 l versus normal controls. Furthermore, the use of self“ report measures in conjunction with other measures of functioning should lead to an overall improvement in detection accuracy, especially When those self—report measures are designed specifically to detect a reSPOnse style consistent with malingering. Participants instructed to malinger Will-1nd Hypothesi s 4 . to demonstrate an increase in voice frequency, adaptorsf foot/leg movements during a clinical interview relativeng their demonstration of these behaViors during a screen!- interview and demonstrate a greater increase in these bFINN-tors than control participants - The application of nonverbal indicators of deception found in social psychology literature to the Clinical interview represents a novel approach to detecting malingering. Voice frequency, adapters and foot movements have been found to have moderate effect SiZes in detecting deception (Zuckerman 5, Driver,1981; DeTurck & Miller 1985) Additionally, the use of a baseline evaluation should d a d to the ability of these variables to discriminate malingered performances (Brandt, Miller, & Hocking, 1980) 43 Hypothesis 5. The incorporation of a multiple’methOd approach (e.g. I test performance, self-report: nonverbal behaViors) to detecting malingering should lead to improved accuracy of classification than the use of any PartiCUIar dependent variable. Furthermore, the number of 10“" probability items selected on the CVLT and Raven’s Progressive Matrices will offer additional classifi accuracy in a discriminant function analysis be)!0nd found by raw scores on CVL'I' recognition or total number correct on the Raven’s Progressive Matrices. cation that o ' S The use of multiple measures and detection Strategle has Offered the greatest promise in improving detection rates Of malingering (Rogers, Harrell, & Liff, 1993; N185 Sweet, 1994; Iverson & Franzen, 1996 ) . Malingering represents a cognitively complex taSkr making success]£01 dissimulation across several avenues Of analYSis particularly difficult to achieve. The inClusion of 39 analysis of the magnitude of error strategy has not been applied to the CVLT or RPM; however, for this approach to have an additional utility over simply examining total scores, it should provide classification accuracy be .ybnd the total scores on these measures. The hypotheses of this experiment were analyzed through independent samples t—tests, Within Sllbjects ANOVA, and discriminant function analysis. The dependent variables that were analyzed through one-tailed t-tests fOr the two groups are number Of low-probability items seleCted on the CVLT recccgnition and RPM, response latencies on the 44 CVLT recognition and RPM, and F88 scores on the WPI'Z’ Changes in average voice frequency from the screening interview to the clinical interview, change in adaptors from the first to second interview, and change in foot/leg movements from the first to Second interview were analyzed by separate 2X2 within subjects ANOVAS, A step—wise discriminant function analysis was conducted to determine the rate at which participants are classified using the . . - '11 best discriminating variables, into either the malinqerl (3 group or the control group- ME THODS 1*gti c ip ant s 100 undergraduate students were recruited to participate in the study through the human sum-acts pool at MiChigan State University and received partial course 10 participants did credit for their partiCipation' n 01: complete the neuropsychological exam and were subsequently removed from the analy’sis- 6 participants were removed from the analeis for endorsing a history Of SignifiCant head injury and/or participation in litigation regarding a n accident injury. 30 participants were assigned to the 45 control group and 54 subjects were assigned to the experimental group. Participants were assigned On a variable assignment schedule. No other data were trimmed- The average age of participants was 21.4 years (SD _—..— (SD = 0-5)’ 3.8). The mean education level was 14.2 years The participants consisted of 79% CaucaSianS: American, 4% Hispanic and 1% mixed; 60% of partiCipants , . . e5 were female and 40% were male. No significant differenc were found between control and experimental groups on ethnicity, gender, age, or education variables, We: Informed consent was eXplalned to participants, consisted of informing them of the nature of the instruments to be used, the fact that they WOUld be alldio and videotaped during parts Of the experiment, and risk 8 associated with participating‘ Participants were not informed about the purpose Of the audio/Video taping prOCedure prior to participation’ Participants were the n assigned to one of two group5° Participants received a s et Of instructions that explained how they were eXpected to perform on the administered tests. Control group 46 participants were instructed to do their best, answer hOHGStlY and tOld that this is an evaluation of tests 0f cognitive and emotional functioning. Participants assigned to the head injury malingering group (Experimental Group) were given a scenario that describes them being in an automobile accident, where another driver was at fault. Included in this Scenario was a brief informational statement about typical symptoms of mild head injuries and these participants were instructed . - 't5 to try and produce the most severe and believable def“;l without making it obvious that they were faking. d we Participants in the eXperimental grOUP were also info‘L that if they succeeded in appearing impaired but welte “Qt detected as faking through statistical analysis they would be awarded $5. The Scenario also included additional information regarding the potential consequences if this was a real-world situation and they were deteCted as fakin (G.g- not receiving the reward, being fined); however, 9 participants were informed that no attempts to SimUlate aCtual punishments would OCCUIo After receiving their instruction set Participants completed the MMPI-Z. participants were scheduled for the second phase of the experiment in which they completed a neuropsychological evaluation lasting approximately 90 47 minutes. Upon arrival for the neuropsychological exam portion Of the experiment, informed consent was discussed again. The participants were then told that the experimenter would be asking them some questions to calibrate the computer. They were told that they shOuld . . PI-Z l(Jnore the instructions they were giVen prior to the MM while answering these questions and answer honestlY° This was done in an attempt to establish a baseline of VOiee frequency, adaptors and foot/leg movements. After the . fits baseline interview, the baseline interview the partielpa . to were given their instruction sets again and Were to1d read over them and given time to prepare for the neurOpsychological testing- After 10 minutes the examiner returned, asked i f the participant had any questions regarding the instructions, and asked the participant to explain what they were being aSked to do to ensure that the participants understood the giVen instructions. Another experimenter Was then Sent in to begin the neuropsychological exam, and participants Were told that this experimenter was unaware of their group aSSignment, and the eXperimental participants were told to imagine that this examiner was evaluating them as part Of their attempt to receive financial compensation for their injury. The second examiner then came in and administered 48 The the neurOpSYChOlOgical exam and clinical interview. neurOpsychological exam consisted of administration of the CVLT, RPM and a clinical interview. The CVLT was administered first, and during the dElay the RPM and clinical interview were conducted. The clinical interView conSisted of questions regarding thel . . . . . - s lhjury and it’s impact for experimental group parthlpant 1:01 and several benign questions about summer plans for Cent participants, Following the conclusion of the test battery, the participants were given a brief questionnal tO assess if they followed instIUCtiOhS and how they They were thee attempted to deceive the experimenter. debriefed about the nature of the eXperimeht , Parti were informed about the variables of intereet to the experimenter and the reason for the audio/Video tapi 179 procedure. Participants were informed that they no identifying information would be associated with thei r aUdio/video taped performance. Participants in the experimental group were informed that they WOUld be paid for their performance if they were not detected as faking by Statistical analysis. Participants were given course credi t , 49 Measures ' . 1 ' , Kramer’ a California Verbal Learning Test (CVLT, D6 .15 Kaplan! & Ober, 1987) verbal e The CVLT is a learning task that assess he . dla ' imme ' CVLT es the learning and memory. The measur ptesente . - no": la :11th short delay, cued, long de y and recoq ° ' of l6 wordS {com information, The task COnSlStS Of a list five . for four Semantic categories that 15 presented orally . word immediate—recall trials (LiSt A)- A second 16’ltem d ,cue ' . ' t:ed once. Free and ChategoIY list (List B) is presen Yree recall trials follow a free recall trial for List Bo {e , , a and cued recall trials as well as a recogng tio ‘rlal llow‘ing a 20-minute delay. Test presented f0 \~retXBSt reliability for the CVLT was '59 Wlth the “9 tmatiVe sample . l and further analySeS of the normative sampl 4 d lit—h if r aVe Q Coefficient alpha Of '7 an sp a Eliabil teda lt Criterion-related Vallidlty has been demonstrated as 13 ~63. correlation with the WeChSLer Memory Scale of .65 its I ( Freeland, Kramer & Kaplan' 1988) . 0811's The CVLT recognition trial is Composed Of it em 8 ' ' - f six categories. Correct items or hltS, List B it I“om em S th . . . . . at Sham):e a category with LJLSt A items, List B items th t a are not categorically similar to List A, items from neith Sr 118‘: bUt categorically related, neither liSt Similar 50 phonemically, and unrelated items from neithe r 11-553 mfg? pSYChOlOgists at a large mental health clinic: having familiarity with the CVLT rated items from thQSe categorie: e as having low, medium or high prObability of being select by a normal adult taking the CVLT. Inter—rater reliablllty was assessed and correlation Coefficient of, 3A was _ . ed achieved across the 44 recognition items. The “Eyela‘i . _ - 1 items that did not appear On either list were CODSlStent y . . a rated as low—probability items and each item was assigned score Of 1 in the calculation of the low—probability items score- b. Raven's Progressive Matrices (RPM; Ravel—1 960) , 1 RPM is a multiple-Choice paper and pe n t at Q11 test th Consists of a series of visual matching and problems. RPM was designed to assess abili nonverbal constructs to SOlVe COmplex probl f e 0U!) presents items with an increasing level of 1) IGq - . lffiC S Lllres the selection Of the correct ChoiQ Q e presented Choices. TGSt‘reteSt reliability Of bee . . the 8 r1 found to be .80, while studies of COHCUrr RPM Gut demonstrate a correlation of .70 with CO “a1 32 intelligence (Burke, 1985) ~ 0f 51 Add't' nally the possible responses for eacfijtelfi l 10 I .555 Cm91091 the first 3 subtestS Of the RPM were rated by PSY 'gh h]. . . either .1. -th the an aas to the probability ( the fami iar wi 'n d l . lecte _ ch ltem would be Se _ 1. medium or low) that ea . divldua 1n . - ed . . . st to a non—impallt aMinistration Of the te melat‘ion a a CO . - - as assessed an Inter—rater reliability w 5 on sets '9 36 item COfoiCi nt of 80 was achi ved across the e . I n . a . iven low pr b bility items by the various raters were (3 o a e. . Scot score 1 in the calculation Of low—probability 1tem5 - {id The responses that fell OH average in between low a scote mecilum probability of being selected were aSSigned Of 0 5 in the calculation of the low—probabi Jity items Score. 0 Minnesota Multiph?=lSic Personali ty Inverl.t orig- aham Tellegen, & KaQ 2 (WP]-2. Butcher, Dahlstrom, Gr ’ IerGr , The MMPI~2 contains 567 items that arg . ’ 13$ as “true”, “false” or left blank. The MMPI\2 administered using only 370 items, which Contains the Clinical scales and the L, F, and K Validit Participants were administered the shortened VerSioijles' MMPI~2 due to time COnSttaintS' The MMPI‘Z was esigxff the aid in psychiatric diagnOSis. This inventory is Widel:d to used in neuropsychological evaluations and Contai HS SeVeral 52 _ “ scales to assess the validity of a generated Profjje Test—retest reliability of the mere? (Lezak, 1995). .92 and scales has been reported as ranging from .58 to to ~87 internal consistency values ranging from 34 89) 19 . t al.: according to normative sample data (Butcher pafit's . {tiCl Of Particular interest to this study was the Pa & score on the Fake Bad scale (FBS) (Lees—HaleYI $1an ' Convergent validity Q f the FBS has heeln Glenn, 1991). Sci 9 ~41 0 with other meas‘1r eStablfished by a correlation of sCored malingering (Slick et al., 1996)- Only the F35 was during this experiment, and no Clinical seales wete Galen}. ated . d - Voice Frequency The complete utterances of participantS were recorded Clinical interview using a microphone and The digitized signal was analYZed for funclaIn 6% shtaj p in Hz using Prevaricator V 1-0 SOftware. Fr ego - en 9 werQ averaged for the baseline and Clinical 1' 03) I) 0909, Ute the means were compared to determine change 8 . Cor Reliability analysis of VOice frequency achiev ed coefficient alpha of .78~ 53 e. Response Latency . he darlflg t ' ' rded rt: ic1pants were reco Responses of Pa 'n recognition trial of the CVLT and RPM u51 w and laptop computer. Response latency mine time elapsing between the end Of the exa . , onse- and the beginning of the Partlcxpant' s res? end oi reCiSe were recorded on a computer program and the 9 f the ' ' ' n o the eXaminer's question and preClse beginnl. g . , . Dds. partic ' nt’s response were measured in milliseCO lpa ° ' lit analys is revealed a Response latency reliabil— Y be Rm“. Coeiticient alpha of .72 for the CVLT and . 30 for t Split—half reliability for the RPM was -92- f~ Adaptors: Foot/Leg Movements Videotapes (without sound) of partlmpehts were made ' d clinical inte during both the basellne an I: iewS . Frequency of adaptors (e.g., scratching, Or 's bein said 9 behaviors unrelated to what 1 g ) were c boom?) 0 9 . ' to the e ' Q tWO independent Judges blind eXp rlmgntal 0, S06 J/ . d d f e- purpose. These ridges also CO e requency 0f foo 1912 t/ mo e for t ‘91; ul 0’ ' wha ' 1 Vements. DeciSlon r S constituteE e9 . ada included any self—groomlng 01: hand movements that ' ptor ' n t bl lnvOlve another object (e-G-v tapplng t e a 8’ Opening w bottle) Any eXpreSSive hand geswre was not Co Sid . ered I an W continuous were adaptor. Adaptors that ere Counted as One 54 _—--_-— movement until it stopped- If the same movement began again after a stop it was counted as an adaptOr- . any DeCision rules for foot/leg movements lnvolved e. movement of the feet or legs that were obselcvalb one . nted as COHtlnuOus movement of the fOOt or leg were Con e movement movement until the movement Stepped - it t“ ' its started again it was counted Separately, POSV—U‘Cal Shl that involved movement of the legs Were also counted- Inter‘Zra’cer reliability coefficients of 92 were achieved by the raters for both adaptors and foot/leg movements. 55 RESULTS t d a ' 17 Table 1 presents mean scores, litY Scores ranges corresponding to calculated low-probabl and RPM) I . L (i.e., infrequently selected items on the CV LT FBS scores from the MMPI'2I response laten‘fl fioO’tIieq and RPM, changes in adaptors (AAdathrsl. kmoice “73 from movements (AFoot/Leg), and VOiCe frequency baseline to clinical interView. m__ \ (3011131018 xperlme' ntal X Iow- ZObeiJ-ity ean (SD) 1.3 (1.17) 6.25 (4. 69} Range 0 — 4.5 0 _ 19 3 \Mean (SD) 12.77 (4.27) 28.78 (7.41) \Eange i 5 " 25 9 x 40 Thesponse atengy ean (SD) 1.09 (.33) 2.25 (. 9 Range .70 - 2.21 1,1 r 6%) Response ‘5 atency ean (SD) F 7.43 (3.15) 9.34 (6 Range 3.68 - 17-35 3.06 — 43 ‘ 0) A(inpucars * Mean (SD) ’2 (4'9) 4.57 (8 ‘ 88 Range {—17'5 — 4'5 —8.5 \ ‘ 0) Feet/leg * Mean (SD) /‘1-9 (5-53) 4-84 (8‘ 20 Range ”12.5 "' 12.5 -12 \Q4) VOice Hz ** Mean (SD) ~79 (1.4) .48 (1. S 36 Range '2-35 ' 5'38 -4.23 _ l) .99 * . . Positive scores reflect an increase 1“ observed 1) baSeline to Clinical interview e haVior ** . ' ' . 8 Positive scores reflect an increase in vaice fre gilen Cy f to!” r to Clinical interview. 9122 Table 1, Means, Standard DeViationS, an bee For Experimental and Control GrOUpS Ranges eline Given the apparent differences between Stan deviations between the control and experimental groups additional analyses were conducted using the v rat' lo, Wh' lch 56 is a comparison ratio of standard deviations - The V ratio . . . . d is calculated by leldlng the eXperimental grOLIp Standar roup- deviation by the standard deviation of the Control g ated greater In every case the experimental group demonstr Variability on the measure5° Significance Oi . . ls with an was determined by calculating confidence intefqa andard error rate of 5%. The null hypothesis is that the St ' . tiO deVlatlons are equal in the populat: i on; hence the V ra ShOUId be equal to l. The difference in standard deviations between groups is signif icant if a V ratio of is not contained in the confidence interval Se UB LB Sig, Low—Probability \ 0.47 3.69 1.8 4.01 S 3 Pas [0.30 2.321.1 1 74 S 5 CVLT Response 0,5l 4.01 1.9 3.00 s Latency 9 Response 0.33 2.541.2 1 90 s Latency 6 A Adaptors 0.21 1'64 0-8 1 22 ns 1 A Peer/leg V 0.25 2 04 1.0 1.53 S 1 AVOice Hz 0.13 1.44 0.7 1.08 ns 2 Table 2. V ratio, Standard Erro Upper and Lower Bound of Cohfidence In, terval Low-probability itemS, FBS scores, and r es p01) latencies all demonstrated significant differ Se e es - l Variance between the control and experimental 11 gr Ollps at th e 57 p < ~05 level. The nonverbal variables of adeptofi, and voice frequency were not significant, while foot/189 1 - . interva movements was just outSide of the 95% Confldence . - ' . . ariances and represents a Significant difference in V between the control and eXPerimental groups° . . 0“ will HYPOthesis 1. Participants in the malingerlnq 9‘ 5 select more low-probability items than W111 centre - . . ices. Participants on the CVLT and Raven’ s progressnte Mat! . Dd In order to evaluate differences between control a eXperiInental groups on number Of low—probability items selected, a one-tailed, independent samples t-test “a . . re conducted. Low—probability scores Qn the CVLT and RPM we summed and this score was used in th - hYpothesis was supported. LOW-prObability . {mind 1 terns We}:6 to be statistically significant between cont I01 and -O 01. predicted, participants in the malingering experimental groups, t (82) = ‘5-68, p < As mo . g”but: re low-probability items on the CVLT recQ 8% 9131' ti 5’0; 01? ed ts 1' 6Q. Control group. The strength Of the rElatiOn I) b}? a] Ship b S and RPM sets A, B, and C than did participa 11 91“on and selection Of low—probability item 8 a nwas .53 indicating a strong effec‘: 5128- Th e as ' ' ' estimated by the d statistic was 1.45 fOr low size probability items . 58 Hypothesisl- partiCJ-Pants 1n them-lingering gronp will E—e—Efionstrate longer response lafencles on the recognition trial of the CVLT and onReven 5 Progressive Matrices relative to contT-‘O1 partiCipants, In order to evaluate dlfferenees between control and experimental groups on response latency on the CVLT l and RPM trials A, B, and C, Recognition tria independent 5 were conducted - samples t—test This hypothesis was partially supported. Group differences were found to statistically Significant for response latency on the CVLT, t (82) = —6.20, p < '001' AS prediCtEdr participants in the malingering group tOOk longer to respond to items the CVLT Recognition trial. The strength Of the relationship between group and CVLT responsg latency as indexed by “was -56 indicating a strong effect size. The effect size as estimated by the d statistic: Was 1 57 ‘ for CVLT response latency. Contrary to predictions, the effect for exper- l , . . . . entel group did not achieve Significance for RPM response latency, t (82) = -l.63, p > .05. On average partie. pants in the malingering group demonstrated longer res‘ponse latencies on the RPM, but this difference was not signifi cant. The strength of the relationship between group and RPM response latency as indexed by n was .17 indicat ing a small effect size. The effect size as 59 estimated by the d statlStiC was -40 for RPM response latency. Further analyses were done With the RPM response . . b ' ' latency llSll'lg luference pro ablllty evaluation. Inference probability is a statistlcal tecihnique that is used to - ' 'hood 3 Type II determine the likell error has been committed when a directional hypotheSls is made. Inference . . e oint biseral . probability uses th P Correlation as the mean and calculates the area under the CUrve betWeen a Z score of zero and the correlation found in the Sample. This area is added to the area under the curve above the correlation . bil 't that a T made in to result in a proba 1 Y ype II e3: ID]: was the decision to fail to reject the null hpr thesis. This technique indicated a 95% likelihood that a TYpe II error was Committed by failing to reject the null hYpOth S . . 813, The true correlation of response latency on the RP A1 . . a group was positive, but was likely nOt cathred du nd . e to insufficient power. wthesis 3. Participants instructed to fake a hea injury will produce higher scores on the PBS of the ‘3 relative to normal controls. I"MPI‘Z In Order to evaluate differences between control an d experimental groups on the FBS on the MMPI—Z, a One-tailed independent samples t—test was conducted. This hypothesis was SUDEOrted. The overall effect for experimental group 60 (:8 for FBS, t (82) ._ achieved significan ‘ ‘10- 86, p < - 00]. As predicted, PartiClpants 1n the malingering group endorsed more items on the PBS than did participants in the h control group. The strengt of the relationship between group and FBS as indexed by 11 Was ~77 indicating a large effect size. The effect Size as estimated by the d statistic was 2.65 for the FBS - Hypothesis 4. Participahts instructed to Maui . demonstrate an increase In VOILCE'e. frequency adnger Wlllnd foot/leg movements during a Cllnlcal interviewaptjeriv: to their demonstratiOn of these behaviors durin re :eening interview and demonstrate a greater 9 a 5° . . increase in these behaviors than control Patti-(3113311118 - in order to evaluate differences betWeeh control and experimental groups on voice frequency, adab tors and toot and leg movements, 2x2 within subjects ANOVAg were conducted. Participants were not randomly 51831 n . . g ed: which limits interpretations of results. Means and St an Qfird deviations for observations of behaviors for baSel 1' interview are reported below in Table 3. Q and 61 Con 1:1013 Experixnexm ”"fii ' 6.65“~— 3:231? ( 4 . 0 8 ) ' ' a? Bali-Sit“ <4 . a 4) ' 8 . 8 0 ‘ Ba;::;ne ( 1 . 64) 10 - Clinical Mean ( 7 267 Adapt (SD)’__ (4- 64) . 2) Clinical Mean 10 . 4 0 Foot (SD) (6.62) Clinical Mean 9 . 3'7 Freq- (SD) (1.13) Table 3° Means and Standa rd Dev - iati For EXperlmental and control G Ons This hypothesis was partially SUPPOrted. A significant main effect was found for adaptors, which was and not predicted, F (1:82) = 4.09: P < .05. 1 . a EK‘perlment control groups demonstrated differences on Qbserved adaptors during the baseline measurement, with participants in the malingering group demonstrating fewer adapto rs. The strength of the main effect of adaptors as indexed .19 indicating a small effect size. 3’77was A significant interaction was also found fOr ad sptors (baseline/clinical) X group (malinger/control) , F (1,82) :- 26.15, p < .001. AS predicted, participant S l the malingering group demonstrated a greater increase in adaptors during a clinical interview than did Control partiCipants. The strength of the interaction as indexed 62 by nwas .48 indicating a large effect size. ANOVA re it 511 S for adaptors are sumarlzed 1n Table 4 and Figure 1 below Source adaptors Adaptors * Group .—_—— Error (Adaptors) 1306. 852 Table 4. 2X2 Within-Subjects ANOVA fOr Adapt ors Adaptors / Base\'\ne Clinical Figure 1. Adaptors Observed During Baseline and Clinical Interview The main effect for foot/leg movements was not significant, F (1,82) = 2.76, p = .10. The strength Of the main effect for foot/leg movements as indexed by leas l ' 7 indicating a small effect size. A significant interacti on of was found for foot/leg movements (baseline/Clinical) x 63 group (malinger/ContrOl)’ F (1'82) = 14-90, p < .001, AS predicted, participants in the malingering group demonstrated a greateI increaSe in this behavior from their . . ols - . . . . baseline relatlve to contr during a clinical interView. The strength of the interaCtlon as indexed by nwas -39 indicating a moderate effeCt Slze- ANOVA results for foot/leg movements are sumIHari zed in Table 5 and Figure 2 below. ‘ Baseline C|inica| Figure 2. Foot/Leg Movements Observed During Baseline and Clinical Interview 64 A significant ma in effeCt was found for Voice frequency, whiCh “as net prediCtEd' F(lr82) = 14.12, p < .001. Both eXPer imental and Control groups demonstrated Significant increases in voice frequency during the clinical interView Compared to their baseline levels. The strength of the Main effect for voice frequency as indexed by n was -38 indicating a moderate effect size. Contrary to predictions, an interaction between frequency (baseliHE/Clinical) X group (malinger/control) did “Qt achieve Significance, F (1,82) = .83, p = .36 . On average participants in the malingering group demonstrated a smaller increase in frequency from baseline to their clinical interview than did contrOlS. but this difference was not significant The strength of the interaction as indexed by T1 Was 09 indicating a small effect size. ANOVA reSUlts f o . . . v - frequency are summarized in Table 6 and Figure 3 b Olce e1 Source 88 df Mean F Sig. Square Voice rgquency 15.57 1.00 15.57 14.12 \ Ow. 0.00 Voice rr’equency 0.92 1.00 0.92 0.83 W .* Group \ Error (Voice 90.41 82.00 1.10 [Freq-net: g1, ) Table 6. 2x2 Within-Subjects ANOVA for Voice Frequency 65 /l_ m ' \, \IOIce Frequency Baseline Clinical Figure 3. Voice Frequency Du: 1 Baseline and Clinical Interviegg Hjmothesis 5. The incorporation of a multiple—methc’d approach (e.g., test performance, Self- behaviors) to detecting malingering sho - uld lead to accuracy of classification than the use of any par dependent variable. Furthermore, the number of 10W“ probability itetns selected on the CVLT and Raven's Progressive Matrices will offer additional Classifi cation accuracy in a discriminant function analySi s b found by raw scores on CVLT recognition or tot correct on the Raven's Progressive Matrices r1331 r epob t , non"? “Proved tic‘llar eyond that al number This hypothesis was partially supported_ A St S o . discriminant function analysis was Conducted to de D Wlse ' ' . . Ermine the rate at which part1c1pants are c1a551f1ed into ithe r the malingering group or the control group. The grou p variable degrees of freedom value for this two—group d esi 9n is one; so one discriminant function was calculated. Th e single discriminant function was significantly aSsociated with group membership [X2<4> = 97.99, p < -0011 and the canoniQal Correlation coefficient related to this function 66 was r = .84. Standardized discriminant function coefficients for the four variables included in the analyses following the step-Wise procedure are presented in Table 7. /Standardi “dam Discriminant Function Coefficients LOWPROB FBS CVLT Response Latency A Adaptors Table 7. canonical Discriminant Function COeffiCients The classification accuracy Of the discriminant function is shown in Table 8 below. The diS’CLij'LnaU’lt function achieved sensitivity, or a true positive rate for malingering of 92% and a specificity, or tru e negatiVe t for controls of 100%, with an overall Classificat. ra e l On accuracy of 95%. Furthermore, the use of the dis Cr . . - - lmin function results in a decrease in error rate of 86g ant compared to decisions of group membership based on s a”T1101 . . . e Size alone. These findings support the first part of thi s hypothesis in that use of several measures in a discriminant function analysis lead to increased accuracy in prediction of group membership. 67 Predicted Group Membership Total 61100? control experimeatal Count control 3 O O 30 XPerimenta l 4 50 54 3 control 1 OO O 100 xperinlenta 1 7 - 4 l 9 2 . 5 9 100 Table 8. Discriminant Class-liication In order to examine the hypothesis that Results lOW’ probability items would add additional Classification 1'1 O COIEEC‘ on the REL“! t are e11 re independently into the discriminant fUncti 'elded a On and Y1 Q classitication rate Of 880- When low~prOb were added to the discriminant function t1) Contribute to the ClaSSification accuracy as thus the second part Of this hypothesis wéx ability it 633’ (1113 Ilcgt prEdi Qted' I Shots L1 b Analyses of correlations between Variablg % reveal ported I Q . Significant correlations Of r = _- 91 (RPIvI t t Q 0 al) . S , 55 (Recognition hits) Wlth low-Probabili t . Dd r \ y items \ \ suggesting that a significant prOportion tha low—probability score is accounted fo Variables° Point biseral correlations are incl cc) lumn 1 of Table 9, and correlations bet“,Gen I; are also reported in Table 9. 68 Uded in "‘94 GROUP Low PBS CVLT RPM 1:308 31‘ R1. A“ A A GROUP 1 leg Hz 3" Raw 3"?- cos \1 Low 0.53” 1 It!- PROB \fi as 0.77** 0.43** 1 CVLT 0.56** o.41** 0.41”” 1 RL _—/——‘ RPM 0,13 0.11 0.15 0.48””r RL 1 A 6** * O 49** 0 17 0.4 0.28 0. pt 32** l A * 7** 0.40** 0.25 0.3 0.24* -0 Foot/ '03 0-18 1 leg A —o 10 -o F —0- 11W o . . . ' -04 -O. - VOlce T 1 0°17 1 RPM V 0 591”: _0_91** —-0.49** -0,38** ms -0.24* ‘0 30** ' 0.14 1 33006 -.0 60** —o-55** ~0~52** -0.50** 4.2715 —o.23* ‘0 an '34” 0'32*:]0.54** 1 t" Correlation is significant at the 0.01 level (2 ta'l d _ 1 e ). \* Wforrxelation is signj_ficant at the 0.0513531‘TETE33IEE) { \ \ \l Table 9 . Correlations The intercorrelations between the variables of interest in this study revealed significant correlations ranging from .22 to .91. The relatively large correlations between variables likely contributed to the final dis criminant function using only 4 variables (FBs, CVLT reSponse latency, low-Probability items, and Change in adaptors) Ollt of 7 entered in. Although the variables 69 account for different: methods of examining approach malinger-mg, the overlap in Variance accounted for SS to the discriminating power Of Some Of the variables cl:L imited the Step-wise Procedure“ When all of the variablesLlJ-fllng entered together, the contribution Of each variabl are e more apparent. In Table 10 pooled within_groups becomes Correlations between discriminating Variables and standardized canonical discriminant functions are displayed. This matrix provides another way to study th usefIllness of each variable in the discriminant functione when all variables are entered simultaneously Th - e classification accuracy of the discriminant functi OD remains the same. Matrix Structure Function 1'38 0 . 7 27 W o . 4 15 Latency Low-Probability 0 . 3 8 1 ‘A Adaptor, 0.343 W 0-263 RPM Response 0 . l 0 9 Latency Table 10. POOled Within—Groups Correlations Between Variables and Canonical Values 70 DISCUSSION The reliance of neurop5YChology on self~report of symptoms and best effort on administered tasks demand 5 that . . , . - ’C ' clinicians be Vigil-ant to he lnfluence of secondary gain du ‘ on the data they collect rlng assessment of patient 8. u . O . Although multlple reVleV"S f malingering research exi St ccraCken, 1990; Rogers I (Franze on & M n, Ivers Harrell & Liff, 1993; Nies & Sweet, 1994) no specific method for detecting dissimulation or exaggeration Of Symptoms has emerged as the most effective approach. The use of multiple methods to detect malingering provides the investigator With more evidence from WhiCh to make a more accurate decision about. the likelihood that a patient is faking. The purpose of the present study was to evaluate several methods for detecting deception and determine the utility of Using a combined approach to improve specificity and sensitivity rates. Specifically, this study examined 5 of performance on neuropsychological tests, self— pattern report Of symptoms on the MMPI‘2, and various nonverbal behaviors during a simulated clinical interview. The use d in high of several methods for detecting deception resulte classification rates in the present study, with va 71 fl” form each of the inVeStigated methods Contributing tQ ClaSsif‘lcatiOn accuracy. The variability Of performance by participants in th e3"‘Perimental grout) compared to the Control group was 9 Significant. The differences betWeen instructional set given to the experimental and control groups may account for part of this difference‘ The eXperimental part- . . l s lClpc‘ints were lhformed abOUt severa ymptOIns that may accompany head injury, but were also informed that most Persons with brain injury do not exhibit all of these Symptoms and were thus given considerable license to simulate neuropsychological deficits. Furthermore, Similar variability has been demonstrated in other research with analog malingerers and suspected malingerers. The increased variability of performance on neuropsyChOlOgicel and self—report measures by analog malingerers compared to normal controls is reflective of the differences in standard deviations seen in actual head injured patients attempting to exaggerate differences and controls (Ju & Varney, 2000; Berry et al., 1995)- Interestingly, nonverbal indices of deception did not demonstrate the same discrepancy pattern between mal ingering participants and controls. Although foot/leg movements did demonstrate a significant difference in V 72 ratio, the v ratio of 1.0 (null hypothesis) f l est OUtSide of the upper bound of the confidenCe . t in erVa l (1.01). The pattern 0f varlabllity 0“ foot and leg m OVeIne nts e was made Clearer when th y Were examined at b l‘ ase lrue u during the clinical interView' For control p t Hts, more Variability of performance Was seen during th e baseline observation of foot/leg movements Wh'l I l e partici - h x eri pants in t e e p mental group demon Strated l ess mOVement and less variability at baseline ' This patte rn was reversed during the clinical Observation of fOOt/leg movements. The variability Seen in nonverbal behaviors may be due in part to mEthOdOlochal limitations Of the current study. The baseline evaluation does not represent a baseline evaluation in a pure sense because it takes place after the participants have been assigned to groups. Although participants were tOld that the baseline questions were not part of the experiment and for calibration purposes, it is possible that the experimental group participants attempted to control their behaviors due to suspicion that this would affect future decisions about their veracity. ConttOl participants may have been more anxious during the initial baseline evaluation, knowing that they were being recorded 73 r1 . ‘ . or to the Cllnlcal evaluatlon ° The relative laCk Of variability on certain n (JrIVErb . . . a the Se indices by malingerers as they do their performanc es On neuropSyChOlegical and Self‘report measures Pr research on nonverbal indices Of de . , ' ception has fOuDd a SlIHllar pattern of variability beCWeen intervie wees tellirug the truth and those being deCeptiv e e (V 1.32 adaptors,- v = 1.12 1 deTuer & Miller . foot/ eg) ( I 1985) - Mallngerers use a var' ches in their a iety of approa ttempt to demonstrate their impairment, which underscores the need to utilize information from several channels in making determinations of: the presence of maL ingering. Hypothesis 1. Participants in the malingering group will select more low-Pr0bability items than will control participants on the CVLT and Raven’s Progressive Matrices. The low~probability items on the California Verbal Learning Test (CVLT) and Raven’s Progressive Matrices (RPM) were more likely to be selected by participants given instructions to malinger than by control participants. However, low—probability items did not contribute to the - . . - b -tion hits dis criminant function a ove and beyond the recogni SCQ re on the CVLT or total correct score on the RPM- In one of the irlitial studies of the method Martin, Franzen 74 and Grey (1998) found that a magnitude of error cflippt‘oach ‘ o recognition Portlons Of the WeChsler Memory Scal R n e~ ev - Ild Lo ic iSEd Visual Memory WM) a , .g 5,11 Memory (I‘M) Contributed significantly to discrimination beYond recognition raw score Values. Although the preSent study did find th . - ' e 118% Of low‘probablllty items to have potential utility in detecting malingered Performances, the clinical util. this approach is called into question since it did lty Of not add to discriminating power beyond scores provided during routine administration (i.e., rECOgnition hits and total score) . Further examination Of the ClaSSifiCation rate of low- Probability items in the present study reveals that when low-probability score is USGd by itself, a 78% correct classification rate is achieved with a 10% false positive and 27% false negative rate. In comparison, when recognition hits are entered independently, an 81% correct classification rate is achieved with a 13% false pOSitiVG and 22% falSe negative rate. Furthermore, the use of total score on the Raven's Progressive Matrices (RPM) results in only a 60% classification rate, yielding 40% false positive and 38% false negative rates. Essentially, the present study found that selection of low-probability items and raw scores on recognition measures perform similarly in 75 discriminating between malingerers and con: 1.015, with recognition measures providing Slightly bet: ter . ' ' l ' discrimination- Addltlona eVldence for the similarity , . lfl . in - . these variables 15 found the point blseral correlatio HS .—- r - . . CorrElatiOns between low p Obablllty ltemS and the total M scores were r = -.91 (RP tOtal) and r = -.55 ach.is based on the theory that malingerers have a difficult time accurately monitoring their performance and do not make more mistakes as the test items increase in difficulty (Gudjonsson & Shackleton, 1986). Although cross Validation for this approach appears promising, it has not been validated on a clinical sample of probable malingerers (McKinzey, podd, Krehbiel & Raven, 1999) , aflothesis 3 - Participants instructed to fake a he (1 injury will produce higher scores on the PBS of t}, a relative to normal controls. 9 MMPI-z Participants in the malingering grOUp demonst higher scores on the Faking Bad Scale (FBs) . ‘ 2 than did controls as predicted. The FBS was added t O the battery in the current study in order to make the t e o . Sting S ituation as close as poss1ble to real-world Conditio HS Well as to provide an additional avenue (1.9. O f symptomS) for detecting the possibility of malingeri mg. The FBS CODSiStS Of items that capture the particular 81 respense set of malingerers to report a variety Of non- Speflfic somatic Symptoms While also presenting themselves as being hODESt and forthright (Lees—Haley, English, & Glenn, 1991) . In the present study FBS represented the largeSt effect size and When the PBS was used independently in a discriminant funCtlon analYSis it yielded a correct classification rate Of 909'- It Can not be determined from the present study how wall the PBS performs with clinical populations of malingerers as Opposed to analog malingerers; however, previous r6BSearch has found the PBS to provide similar classification rates in clinical Donders & Millis, 2001) - HypotheSiSA- Participants instructed to malinger will Ejeinonstrate an increase in veice. freqaency, adapters, and f ot/leg movements during a clinical interview rel , o - tion of these behaviors d - atlve to their demonstra Bring a screen interview and demonstrate 3.91.393th increase in 1:}, ing behaviors than control partlclpants. 989 The application of research from social psychol 0g 0 n deception seems to have potential for being applied t o ring in clinical psychOlOgy. detecting malinge Behaviors Snob as adaptors and foot and leg movements were found t o bQ Sensitive to deceptive communication in the present Study. As predicted, observed adaptors increased for participants in the malingering group but not for the 82 COntrol group. Additionally, foot and leg movements Showed the Same predicted pattern Of increase from baseline to clinical interview for the eXperimental group but not for the control group - One finding that was not predicted was group differences at baseline on adaptors and foot/leg movements. Control participants demonstrated more adaptors and foot/leg movements at baseline than did experimental participants. Both groups were e> ' .C 1 t1) 1] ' th ' Stle to CllIl] a r :1: sed GSear Q measures I) Y ' have. determination is llkely to ° . ' Ormlll . he discriminant f dlllg how we a SS b egar I 1. t". Grates -' 'ured patients, ' 1’8 from head in] malingere How ever, previOuS 87 . . . M Studies involving Clinical samples have SUpported this approach (Rogers, Harreil, & Liff, 1993; IVerSon & Franzen, 1996- Martens! ponders 8‘ M11118, 2001) . The compleijty Of malingering successfully requires an individual to monitor their performance across mUItiple domains to appear believable and aVOid detection. BY including Self—report and nonverbal information Observed during a clinical inteerew in Colljuncluon With analysis of Patterns Of performance on neurODSyChOlOgiCal instruments, malingering becomes increaSingly difflcult to achieve and easier to detect. The performance Of the four indiVid als Who were malingering but were not detected by thQ discrimi nafit function is of particular interest. Oerall st memo ry performance on the CVLT was below 1 percentil e ' rh percentile when compared to their same 51 ed p O the 8 Get their performance on the Recognition trial rangg , Yet to 15 out of a possible 16. All four particl' rom 12 ' b ' in for s approacmng the taSk by be g 99”“ and east. re orted they had difficulty paying attention. Two of th as if e Earticipants did not demonstrate any increase nOnve ' r hehaviors, While the other two did demonstrate an . bai ' a but not of sufficient Size to be detected. 8e 11: Classification rates achieved us1ng multiple In chOdS were acceptable it is likely that despite the d jfficujty most participants had in maintaining their dece it across domains it can be done- More importantly, no control I participants were incorrectly identified as malingerers. Despite fairly high rates Of individual measures in differentiating between malingering and control participants in the present StUdY, researcr1 With clinical Populations has not supported thS use Of any particular mea ure as definitive in identifying exaggerated deficits- s U f the magnitude Of error Strategy 011 the CVLT and RPM se 0 . ' Of 7 $35- whilQ Yielded ClaSSi flcatlon rate 8 he response latency of the CVLT resulted an aSsification Ia . . - ‘cal opulations have Studies With clinl p Quad Sim—i lat 1 ssitication rates using the CVLT perfo c a rmanCe patterns et a1 Q 000), ti (Baker Donders, & Thompson, 2000; Sweet I The F135 produced an individual Classific 909 in the present stUdY; hOWEVFEC, recent resea rate of actual head~injured patients found that the FBs with ' tif im rob bl ‘Vas able to SUffiCientlY lden y p a e respons not ' Dr . 1 dependent 0f the CVLT and Vlce Versa (Martens Ofiles n O , . . ane . __ robablllty items or PBS 338, & Millls’ 2001" Low p SQores e t hi her false positive Qred independently lead to g rates (130 and - his finding reinforce %, IESPeCthQly)’ T 8 th importance 89 Of- using mUltiple measurQS and data sources to enha DCE’ . . 'n re _ deClSlon accuracy 1 gard to malingering .. . ion of The incorporat nOnVerbal behaviors observed , . - interview - - . durlng a CllnlCal Showed some promise 1n aiding . . . l . 1n detectlon accuracy though the use of indices such as f0 increased adaptors and Ot/leg mOVementS reSUlted in relatively high classification rates when used independently (77% and 71%” resPSQtiVely) ' Use of these indices alone resulted 1“ unacceptable false positive error 0 9 Ct. l ' rates (56a and 63o, respe 1V9 Y) - Whllg Use of Observations during cllnlcal lnterview mgy rep sent a rich re source of data for determining effort, thEs vations e ObseI lack sufficient specificity to be used 1 d ependerltiY- Limitations of the Current Study One of the primary limitations of th e Cuer k clinical comparison grou the lac of a p. Althoh t Study is analog malingering groups is useful in Etermi Hi internal validity of detection measures, it doe 9 the e . ' I] fiddress external validity. One of the paradox Qt . - Of QpprOaCh is that participants are aSked to Comp this . 13’ by :Eaking, while actual malingerers fake their pe When asked to COmpLy (Nles & Sweet, 1994) . Flirt hermore . _ . _ , , t use of pOSltive incentlves lS eaSler to achiev he Q than 90 ‘ negative Consequences using this paradigm due to 9th' 1 1C8 . - However i conSlderatiOnS° ' n real—world settings Serious malingering ranging from disCODtinuation of medical benefits to fines of Possible imprisonment (Rogers & Cruise, 1998). Roger5 and Crul'se found that participants who were given negative incentives produced more focused SYIIIptom Presentations’ suggesting that empE13518 on possible negative outcomes may improve fel gning perfOfmances in . . . rthermore . ' 1 not suf ~ ' nt present study was like y flcle QtiVation to Larger rewa rds or - erfo rmance - _ ch improve p QlternativeS Su le rizes may produce sufficient . . as raft P Otlvatio‘fl wlthout raising additional ethical concerns, One solution to the limitations of th e 31mg design is analeis of convergence with findi {atlon ngs Although use groups comparison deSlgnS° of 6 k1) r01” know~ - the sco e of t Comparison was beyond p he C11rrent W13‘9rOUps findings using established prOCEdures (USe Of Q (Jay, b . LT ere Similar to those found 1n Other knOWn‘grouio8 and F88) . 98 (Baker: Donders, & Thompson' 2000' SWGEt et a1 earCh Raley et al.; Larrabeer 1998; Martens, DOHderS 2001). Nonverbal indices of deception have Ye examined in a known—groups design, and although h t ey 91 demonstrated some Utility in th Udy, cl inical use’hllnes S has sophisticated in their attempts to malinger than actual head injured patients (Haines & NOrris 200].) ' ° Haines a nd Norris found that student malingerers were onl Y identified at a rate of 26% when compared tQ lhjured Patients, whereas patients With history of h Gad-injury asked to simulate Symptoms were detected Vvith 100% accuracy. However, JU and Varney (2000) found that head injured patients could avoid detection Q . . their attempts to malinger on the Portland Digit Re cogniti 0n T ll as non-clinical simulators, eSt as we AlthOUgh StudQ appefir to be a difficult group to detect as malingering I they typical clinical populations On Several d \ emograa variables and further research using Sa N. Closely approximate likely referrals fOr f Q mot e evaluation of head injury is needed_ Q Methodological limitations of the CUrr em; . . . St Warrant discussion. The partiCipants Were a . udy al 831 SO . , . gned t groups and given instIUCtions about how to a 0 Dr ' - . . ' OaCh th hask in the initial testing seSSion, Partici e pants th took the MMPI—Z and were scheduled fOr a foil en Ow_u P 92 appointment to complete the neuropsyChOlog" lea} tes ts When , , ned f partiCipantS retur gr the Second SeSsi On they were ions asked several quest to establish a has eline Of - . Alt: . nonverbal behaVlOrs hough they were told that this 1: Of . evaluation was not par the experiment, the Instructions . ' V they were prEVlOUSlY 93' en may have had an impact on their performance, thus limiting the conclusions that Can be drawn from changes Observed during the Clinic5:11 interview- Another limitation Of the “Ghverbal Clata is the differences in time for the interviev‘“ Although efforts were made to equalize the time Spent on the interview for . tal 93:01.1 s on ontrol and experimen p , experimental group interv1ews were Slight t ' . 13" although no signlficantly longer. This llkEly COntti u ted tQ the P . artlcipa control group were encouraged to elaborate th {fits in the increase in nonverbal behaviors seen. during the interview to equalize the amount of anSWerS time between groups and this may also have had tervjew Additionally, the lack of a true random as *3 effect. limits the interpretations from the Current stuqlgnment Quilthough no differences were observed On demOgba:1;. Variables for the ContrOl or experimental grOUIb, ti: $tatistical analyses done in this study are bag d e . on the assumptiOn of COmplete random aSSignment to gr Cups. 93 IIIIIIIIIIIIIIIIIIIIIIIlFIIII”IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII""“ . r ‘ for Further Research Directions . lini(:a] . uring c . . g ballavlor‘s d lq I ‘ GChn ECtl e . ' tection h potential for improvlng de e ' has t interv1ews ' (31in§]° (I f maling accura y O " Ward, 0) berts and.Bull (200 R0 ' tion and found t t of Verbal communlca . n en . f Zlng CO - v d detectlon O for analY f these metths lmpro e tion 0 . - rpora er alone- that lnCO se Of nonverbal Obs Vatlons be u deception te ' clude research ShOuld at mbt to ln future Furthermore, l I l a for findings. contribUte research can Cory Labora O improxfifig ' ' however, ' erinq detectlon, maling caution 1171181; I e $xercised O 0 act . . . clinical pr ice. A . . lying findings into ltlonal in app re sophisticated and 8e Q tudies using mo (1 analog 5 Us: . \fi‘tiVe evaluation of VOlCe freqUSIicy may ' ent for equipm .on regarding the utlllty Of adapt, ' ti e informa mor SOCial pSyChOlOgY reSearch to th ique from techn b ratory settings that more le Also, la 0 lm. tea WO 1d eliminate poten lal confounding noise 1] Q raneous t StUd Xt ' he presen y. bles experienced in t ia Var . ti e g1. \IE]. I . ' ns Manlpu a O EmphaSis On negative consequences may alter the performance of persons trying to be deceptive, and attempts to more closely approximate the forensic setting would improve external validity of findings - 95 REFERENCES Alfano, D., Neilson, P., Paniak, C., & Finlayson, M- (1992) . The MMPI and closed-head injury. The Clinical Neuropsychologist, 6, 134-142. Baker, R., Donders, J., & Thompson, E. (2000). Assessment of incomplete effort with the California Verbal Learning Test. Applied Neuropsychology, 7 (2) , lll-ll4- - e Bernard, L. C. (1990). Prospects for faking believazl memory deficits on neuropsychological tests and theus: and incentives in simulation research. Journal W mental Neuropsychology, 12 (5), 715—728. Bernard, L. C. (1991). The detection of fakédThe deficits on the Rey Auditory Verbal Learning Test. effect of serial pOSitiOD. Archives Of CliniC//al mpsychology 6' 81—88- ’1’.- Bernard, L. C., Houston, W., & Natoli, Lo upotential Malingering on neuropsychological memory tests: 49 . . . 0931' objective 1ndiczators. Journal of Clinical Psycho1 Kl) , 45—53 . Bernard, L. C., McGrath, M. J., & Hons ton, W. (1993) . Discriminating between simulated malingering and closed head injury on the Wechsler Memory Scale-Revised. ' of Clinical Neuropsychology, 8, 539-551. M Berry, D. T. R., Wetter, M. w., Youngj ohn, J Gass, C. 3., Lamb, D. G., Franzen, M. D., MacInneS‘ R 'I & Buchholz, D. (1995). Over-reporting of closed\’ W, injury symptoms on the MMPI-Z. Psychological ASSe ead ' (4), 517-523. Ssment, 7 Binder, L. (1993) . Assessment of malingerin mi 1d head trauma with the Portland Digit Recoqnitioafter . n T W1 of Clinical and Experimental Neuropsychology’ fgt. 1 '7 0—182. ' Binder, L. M., & ROhling, M. L. (1996). Money ma tters: A meta-analytic review of the effects of financial incentives on recovery after closed-head injury, Wurnal of Psychiatry, 153, 7-10. 96 Binder, L. M., Villanueva, M. R., Howieson D Moore, R. T. (1993). The Rey AVLT recognition $11,551.33}. measures motivational impairment after mild head trauma. Archives of clinical Neuropsychology, 8, 137—147. Brandt, J., Rubinsky, E., & Lassen, G. (1985). Uncovering malingered amnesia. Annals of the New York Academy of Sciences, 444, 502—503. Brandt, D. R., Miller, G. R., & Hocking, J. E. (1980) . The truth—deception attribution: Effects of familiarity on the ability of observers to detect deception. Human Wicationfiesearch, 6, 99—110, . S: Burke, H.R. (1985). Raven's Progressive Matilge More on norms, reliability, and validity. W Qical Psychology, 41, 231-235. .R.l Butcher, J. N., Dahlstrom, W.G., Graham: J Tellegen, A., 5, Kaemmer, B. (1989). W sic o . . ' ha administratMscoring the Minnesota Multlp it‘l oi PersonaW. Minneapolis: UniVe Minnesota Press . Coleman, R. D., Rapport, L. J., Millis, s. R". on 3. R., s. Farchione, T. J. (1998). Effects of coaching . detection Oi malingering on the California Verbal Learning Test. ggurnal of Clinical and Experimental Neuropsychology, 20 (2), 201-210. Delis, D. C., Kramer, J. H., Kaplan, E. , & (1987). The California Verbal Learning Test, Nobel“, B. A cholo i 1 c ' SW ° Psy 9 ca orporation. York. Delis, D.C., Freeland, J°I Kramer, J- H-, (1988) . Integrating clinical asessment with Co 5.5191311, E neuroscience: Construct validity of the Californ. lVe ‘ Learning Test. Journal of Consulting and Clinicala Verbal \ gigygggiogy, 56, 123-130. DePaulo, B. M., Lanier, K., & Davis, T. (1983), IDetecting the deceit of the motivated liar. Journal of EEEEEEEElEty and Social Psychology, 45, 1096—1103. DeTurck, M. A., Harszlak, J. J., Bodhorn, D. J., & 'Iexter, L. A. (1990) . The effects of training social 97 perceivers to detect deception from behavioral cues Communication Quarterly, 38 (2), 189-199. ' DeTurck, M. A., & Miller, G. R. (1985). Deception and arousal; Isolating the behavioral correlates of deception. Human Comunw Research, 12 (2), 181—201. Diamond, R., Barth, J., & Zillmer, E. (1988) . Emotional correlates of mild closed head injury: 0f the MMPI. International Journal ow Ne&Ifl/chology, 10, 35-40- The r0l€e . . O) - DiCarlo, M.A., Gfeller, J.D., & Oliveri, M.V. (200 Effects of coaching on detecting feigned cognitive ' ' - linical impairment with the Category Test. Archives of C Wsychoiogy, 15 (5), 399—413. Ekman, pol O’Sullivan, M., p. 999)- , & Frank, M, G- (1 263 few can catch a liar. O (3) ' Psycholoqu 266. & Scherer' Ekman, P- , O’Sullivan, M., Friesen, w, V” in (1991). Invited article: Face, voice, and. DOS” l5 (2): detecting deCeit. J_ournal of Nonverbal W 125-135. Ema“: P -. & Friesen, w. v. (1974). Detecting deception from the body or face. Journal - of Personality 929130019)— Psychology, ;_9_, 288—298. v Faust, D. (1995) . The detection of deceptiQ Neurologic Clinics, 13 (2), 255—265. I3. Gass, C. (1991). MMPI-Z interpretation an head injury: A correction factor. Psychological Closed giggessment, 3, 27—31. fi“‘\\ Gass, C. & Russell, E. (1991), closed head trauma patients: MMPI profiles of complaints . Impact of neurological Journal Of Clinical PSYChOlOgYI 47, 253 2 x ‘ 6O Gough, H. G. (1950). The F minus K dissimulatiOn index for the MMPI. Journal of Consulting PSYcholOgy 14 408‘413o H 98 Gough, H. G., (1954), about neuroticism. Some common mi sconcept i 0125 287—292. Journal of Consulting Psycholo y 18 Greiffenstein, M. F o, GOla, To I Baker, We J- (1995) 0 MMPI~2 validity scales versus domain specific measures in detection of factitious traumatic brain injury. The Clinical Neuropsyghologist, 9 (3), 230-240. Gudjonsson, G., & Shackleton, H. Of Scores on Raven’s Progressive Matrices during ‘faklng bag, and ‘non-faking' performance. gritish JW Ell—3&1 Psychology, 25, 35—41, (1986) . The pattern Guilmette, T. J., Hart, K. J., Lelninger, B.E. (1994). Giuiano, A. Jr 5‘ . Detecting simulated memo? impairment; Comparison of the Rey fifteen-item Hi Scock forced’czhoice procedure . Ne_\L1ropsychol o g M o the W (3) , 283—294. Haines, M. E., & Norris, M. P. (2001) , mance 0“ student and patient simulated malingerers' p fined standard neurc>psychological measures to detect is" 15 cognitive 99f icits. The Clinical Neurops ycholog K2) , Vii-3.82 . Heaton, R. K., Lehman, R. A., Voqt, A' T. K1978) - Prospects for faking believable deficits OD neuropsychological testing. Journal of COnsulti Clinical Psychology, 46 (5), 892—900. Smith, H. H. , ng and Iverson, G. L., & Franzen, M. D. (1996) , multiple objective memory procedures to detect 8 .3ng malingering. Journal of Clinical and Experimentlmulated Real-Lopsychology, 18 (1), 38—51. a Iverson, G. L., Franzen, M-D-r & MCCraCken I L. (1994). Application of a forced-choice memory pro . designed to detect experimental malingering. Archfedure wcal Neuropsychology, 9 (5), 437-450. W Iverson, G. L., Franzen, M-Dor & MCCl’aCken, L. M (1991). . Evaluation of an objective assessment teChniqu for the detection of malingered memory deficits, Law aid Human Behavior, 15 i6), 667-678. —“‘-— 99 Ju, D” &'Varfmy, N.IL (2000). patients simulate malingering? (4) I 201‘207. Can head in ' ' .3L1ry AppliedW Kassin, S. M., & Fong,<3. T. (1999). “I’Hliumnocentl’V : Effects of training on judgments of truth and deception in the interrogation room. Law and Human behavior, 23 (5), 499-516. Killgore, W. D., DellaPietra, L. (2000). Using the WMS‘III to detect malingering: Empirical validation Of the Rarely Missed Index (RMI) . Journal of Clinical and E&EMental Neuropsychology, 22 (6), 761-771. . the Larrabee, G. J. (1998). Somatic malingerlng on - l . linlCa MMPI and MMPI—Z in personal injury litigants. W Ne&pS}/chologistl 12 (2), 179—188. K scores Lees-Haley, P. R. (1991). MMPI—Z F and F ’ , an 0f personal injury malingerers in vocational . W neuropsychological and emotional distress clalm5° glrnal of W Psychology, 9 (3), 5‘13- . W- J. Lees-Haley, P. R., English, L. T., & Glenn’ 501131 (I991). Ix take bad scale on the MMPI - 2 for Per 210 injury claimants. gsychological Reports, 68, 203’ ' Lees-Haley, P. R. (1990). Contamination of neuropsychological testing by litigation. Foren . Reports, 3, 421-426. ————~\\§g§ Lees-Haley, P, R., a Fox, D. D. (1990) _ Neuropsychological false positives in litigatiO making test findings. Perceptual and MOtor Skil : Trail 1379-1382. 1 8’ (70), Lezak, M.D. (1995) . Neuropsychological ass ed.) . New York: Oxford University Press.W (3m Lykken, D. T. (1974). Psychology and the lie 01 industry. American Psychologist, 29, 725-739, eteCtor Martens, M., Donders, J., & Millis, S. R. (2001) Evaluation of invalid response sets after traumatic head injury. Journal of Forensic Neuropsychology, 2 (1), 1‘18. 100 Martin, R. C., Franzen, M. D., & Grey, S. (1998) Magnitude of error as a strategy to detect feigned memory impairment. The Clinical Neuropsychologist, L2 (1), 84-91, McKinzey, R. K., Podd, M. H., Krehbiel, M. A., & Raven, J, (1999) . Detection of malingering on Raven’s Standard Progressive Matrices: A cross-validation. w @gnal of Clinical Psychology, 38, 435-439. Millis, S. R., Putnam, S. H., 5. Adams, K. M. (1995: March), _I\I_europsychological malingering and t Md new indicators. he MMPI -2; )K/ al Paper presented at the 30t Ann” Symposium on Recent Developments in the use of the MMPII MMPI- 2, and MMPI — A, St. Petersburg Beach, FL- Rickerv Millis, s. R., Putnam, s. H., Adams, K. Mus: in the J: H. (1995) - The California Verbal Learning Teica detection of incomplete effort in neuropsycholic363,a’llo evaluation. gigglogical Assessment, 7 (4): . - l Nies, K- J. & Sweet, J. J. (1994), . crl‘ilca Neuropsychological assessment and malingerinq' e5 of review of Past and present strategies. Archlv Clinical Neuropsychology, 9, (6), 501-552 L. (1998)° Pachana, N. A., Boone, K. Bu & Ganzell, - False positive errors on selecte d tests 0 :f malingering- MJournal of Forensic Psychology, 16 (2) 17-25. Guide to Using the CO London: H. K. Lewis. loured Raven, J. C. (1960). Progressive Matrices. Rogers, R., Cruise, K. R. (1998). AssessmQ malingering with simulation designs: Threats to “t Of validity. Law and Human Behavior, 22 (3), 273‘2gegternal Rogers, R., Harrel, E. H., & Liff, C. D. Feigning neuropsychological I pairment: (19 A or“ - 93) - . . , . . lthal of methodological and clinical conSlderatlons, Clin-eview Psychology Review, 13., 255-274. w Rogers, R., Gillis, J. R., Bagby, R. M., & MOHtier E. (1991). 0, Detection of malingering on the Structured Interview of Reported Symptoms (SIRS): A Stilldy of COaChed and uncoached simulators. Psychological Assessment, 3 (4) 673’677. I 101 & Arkowitz, H. (1990) - A psychological Schretlen, D. , test, battery to detect prison inmates who fake inSanity or mental retardation. Behavioral Sciences and the Law, 8, 75-84. & Spellacy, F- .33 Slick, D. J., HOpp, G., Strauss, E., . Victoria Symptom Validity Test: EffiCiency for H996). detecting feigned memory impairment and relationship to neurOpsychological tests and MMPI — 2 Validity Scales. 8 JOurnal of Clinical and Experimental Neuro s cholOGYr l Numanr B” Sattlberger, E., & Nies, K. J. (2000). y versus (6), 911—922. J., Wolfe, P., Sweet, J. ROSenfeld, J, P., Clingerman, 5., Further investigation of traumatic brain injur Ding inSllfficient: effort with the California Verbal 8821: 105’ Test. Archives of Clinical Neuropsychology, 1/5 ( I 113. entia1 Sweet, J- J. (1999). Malingering: DiffieI 85)- diagnosis. In J. J. Sweet (Ed.). Forensic 255’2 Neuropsychology: Fundamentals and Practice WP. Lisse: Sweets & Zeitlinger. s — ros J. J. (1996 ) , 1300th6 C ting Ten‘nula, W. N., & Sweet, validation oi the Booklet Category Test in deteC malingered traumatic brain injury. The C linical WhOlOgist, 10, 104-116. Trueblood, W., & Schmidt, M. other validity considerations in the neuropsya1 evaluation of mild head injury. Journal of C11 Experimental Neuropsychology, 15 (4), 578-590, Roberts, (1993). Dialillgering and 0,logical lCal and K. P., & BUl Vrij, A., Edward, K., (2000) . Detecting deceit via analysis of Verba , R nonverbal behavior. Journal of Nonverbal BehaViojnd 239—263. ’ 24 (4), (1981). Cognitive, SOc' lal Waid, W. M., & Orne, M. T. and personality processes in the physiological datecti On Of mental deception. In L. Berkowitz (Ed.) . Advances in Experi Academic PIESS. social Psychology (Vol 14.) New York: fl D. N. (1948). wiener, Subtle and ObVious keys for the Journal of Consulting Psychology, 12, 43—47. MMPI. 102 & Brandt, J. (1988)- The detection of Wiggins, E’ C" Law and Human Behavior, 12 (1) , 57—78 - simu lated amnesia . J. R., Davis, D., & Wolf, I. (1997). Head Paradoxical severity effects and Psychological Assessment, 9 Youngjohn, injury and the MMPI-Z: the influence of litigation. (3) , 177-184. (1984). Malingering of psycholOgical Ziskin, J. 39—49. disorders. Behavioral Sciences and they, . ‘ S: Zuckerman, M- & Driver, R. E. (1985). Telling lle Verbal and nonverbal correlates of deception. In Phations thegman and S. Feldstein (Eds.), MultichanneW wmverbal Behavior. New Jersey: Hillsdale. & Amidon, M. D., Bishop, S. E" 'n the Zuckerman, Mu 'ce l d POInerantz, S- D. (1982) . Face and tone of Vol ality an Communication of deception- W Mial Psychology, 43, 347—357. “981‘. l, ' B. M. & Ros tha n L- I @171 1 logy Zuckerman, M. , DePaulo, . n. Verbal and no nverbal communication of dec eptl? a1 PSYChO Berkowitz (Ed -) , Advances in Experimental soc). (Vol. 14) - New York: Academic Press. 103 APPENDIX 104 IN S'I'RUCTION SET A Participate ts and . The ex . lnv perimen ' a p::ves several 02min which You a , are tosp ltems on the {e We are trying to Vglidatetessome neurOpsyiiformFO 1:he be:S YOU will be administered. You the it QlOglca]. test of Your abilities on all 0 the NO Ghee?ls Will be diffi: You will be a inistered. Some of YOU]: be 5 expected to anUlt while some may be easy for You. items h::° On the persOIslweF all items correctly. CNS: do will be ,estly as they allty inventor lease answer the will ldentified apply to 31' Val' be kept COnf' by a subject Y0u~ Your respo Sheet We ldating Some tidenFial. We number and your r appreciate eSt 1tems, somare Only interested in eSGarC-h . YOUr Willin e p1 K1 tO project gness ease respond tram Y- anSWer ' One Of t to . . - any questions he ex Partialpate in our that youpsrlmenters Wi11 be happy ave at the conclus t h 105 INSTRUCTION SET B . ipate . The experiment in which you are going to-éfftgsts, involves several commonly used neuropSYChOlC?gl personality inventory and a brief clinical .1 the purposes of this experiment you are to hat was n Sus“Cained a head injury in a car accident fault. Initially You had some memory a prOblems, as Well as headaches whiCh r e after the accident. Although these Symptoms hilt agai‘flst imagine that you have decided to pursue a laws 0 the other driver and hope to increase the 8128 d fake claim by faking brain damage. You are ‘10 try all comIIlon symptoms of head injury in the boost real), iner- fashion in order to prevent detection by e exam COmmon symptoms of a mild head injury include: 0 Memory problems Attention and concentration prob lemS A general slowing Of mental abilities Difficulty With problem solving tatigue, sensitivity to light and noise! irritabillty o Nausea, headaches, dizziness If you are able to demonstrate imp . '0 detection by the examiner, you will be lrment and aVOl However, if you are determined by the kaarded $5. faking, this could result in the loss Q emi settlement ($5) and in a real—world sit]S Yo her to be result in other consequences Slich as a L1ati Ur Court penalty. SEve On DVD] . in the first phase of this experi re Q H d likely complete a personality inventory. plement y urt questions as if you are pursuing litigasé 1):“ inqury and are trying to appear impaira 1011 f0” {\ while trying to seem believable, In ti: by thb Q <3 ipate in Our conclusion of this eXperiment you Will be given a feW seCOn tests of memory and problem-solving abiltePrOps 912.3 interViewed by the examiner. During then-:les ych be asked questions by the examiner regardinterv-ewbe gleel injury has impacted your school or Work s3: how thgou experienced, and your recollection Of tge acp§0ms Youhe will be better able to deceive the examiner ildent- 26% thought about your aHSWers ahead of time. f you have11 We appreciate your willingness to partic research project. One of the experimenters w to answer'any questions that you have at the the experlment. 106 PRE-SCREENING INTERVIEW Part of the ted to the ew is NOT few brief Note: This screening intervi , experiment and this should be COmmunlCal subject. This sshould be explaine S iildeo CJIJeStions to calibrate the audiO and Vl iment- used in the actua expat equipment to be . tar? . . 1‘1 80 1- HOW many experiments have you partlclPated far? . , en 50 2. What has your experience of the experlmen’CS e in? ticipated What kinds of experiments have you par 0 I E major 3 . What 18 your major at MSU? How did you ChOOSe tha or what about it appeals to you? 4. Have you ever had a head injury befob happened? S? If 50! what 5. Have you ever been involved in any k. the claimant or defendant)? 113d Of 1 . tj . gdtlon ( as 107 CLINICAL INTERVIEw (Instruction Set A) ' ns and I am going to ask you some questJ-O record your answers 1. Why did you choose to particip experiment? 2. Why did you choose to attend MSU? 3 . What has been your most interesting Class and why? 4. What are your plans for the summer? 108 “CLINICAL INTERVIEW” (Instruction Set B) . e to 381‘ I have a few questions I would 111‘ you regarding your accident . 1- Tell me what happened in the accident“ a . d as leDCe 2- Please describe the symptoms you have expeI result of your injury. - ' 'uIY 3 . In what ways have you nothed the effects of Your 1“] in your school or work? 4 . What would you like to find out as e testin ? r g esult 109