Journal ofNeurology, Neurosurgery, and Psychiatry 1997;62:617-621 617 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.6.617 on 1 June 1997. Downloaded from Use of the Stroop phenomenon as a diagnostic tool for malingering

Alicia Osimani, Ariel Alon, Andrea Berger, Jacob M Abarbanel

Abstract situation in which the question of malingering Aims-To assess a computerised version arises, not only because of the medicolegal of the Stroop test for detection of malin- referral and the lack of objective symptoms, gering of cognitive deficit. but also due to many misconceptions and prej- Methods-Sixty subjects were assessed udices arising from studies in the early 1960s, using this test. Twenty had cognitive when the diagnosis of "compensation neuro- deficits due to of miscella- sis" became popular, establishing a mechanistic neous aetiologies. Ten were healthy, not criterion "no objective lesion, no brain dam- acquainted with the test, and were asked age". More recent literature has challenged to simulate cognitive impairment. many of these views. Even apparently mild Another 10 simulators were head injury may sometimes cause severe neu- students trained in the use of the test. robehavioural complications.4 5 The con- Twenty healthy subjects served as con- tention that patients become less symptomatic trols. Results were analysed for reaction after a legal claim is settled has also been stud- time, error percentage, and the Stroop ied by several authors and has not found sup- effect. port.5-7 In addition, no relation has been found Results-There was a significant differ- between the neurobehavioural syndrome and ence in reaction time among groups, social or psychological factors. showing a direct relation of age among In summary, many of the previous views control subjects, and also longer reaction tending to ignore the postconcussive syn- time in patients with brain damage than drome have proved false. This does not mean in controls. Controls and patients with that there are not many people who feign cog- brain damage showed a clear Stroop nitive inability to obtain compensation, but effect. Simulators had a significantly pro- the real frequency of malingering is not clear, longed reaction time, increased error ranging from 3% to 50% depending on the percentage, and inverted or absent Stroop studies.8 effect. This alteration of the Stroop effect Several techniques have been proposed to is never present in organic cognitive identify malingerers. The old clinical concept

deficits and seems to be a characteristic can also be applied in : symp- http://jnnp.bmj.com/ pattern of feigning, independently of toms must correspond to known syndromes. knowledge ofthe test. Amnestic patients, for instance, never show Conclusion-This technique is recom- impairment for all tasks, although lay mended as a valuable tool to detect people tend to think that memory is a unitary feigned cognitive impairment. process and therefore its impairment must result in failure of every mnemonic task. (7 Neurol Neurosurg Psychiatry 1997;62:617-621) Several studies on malingerers are based on this premise.9 10 on September 26, 2021 by guest. Protected copyright. Lack of consistency between the results of Keywords: malingering; Stroop effect; head injury; reaction time examination and the patient's everyday activi- ties is another important clue. Binder8 reports a Department of case of malingering who had a VIQ of 55 and Neurology, Kaplan The diagnostic and statistical manual-IV' PIQ of 59 but was able to give a detailed his- Hospital, Rehovot, Israel defines malingering as "the intentional pro- tory of his problems and drive his car safely A Osimani duction of false or grossly exaggerated physical around the city. Discrepancy among subtests J M Abarbanel or psychological symptoms, motivated by of a battery may also be an important clue.'I Department of external incentives" and notes that it should A battery to discriminate between simulated Behavioural Sciences, be University ofBen "strongly suspected" when there is a and genuine cognitive impairment would be Gurion ofthe Negev, medicolegal referral, a large discrepancy very useful. However, the increasing media Ber Sheva, Israel between subjective and objective findings, a interest on memory and cognitive problems A Osimani lack of cooperation with evaluation or treat- may enable intelligent simulators to fake cog- A Alon ment, or A Berger the antecedent of antisocial personal- nitive impairment and in some cases malinger- Correspondence to: ity disorder. Mild head injuries usually occur ers may be advised by professionals who are Dr Alicia Osimani, in the context of road traffic, military, or occu- well acquainted with these tests. Department of Neurology, Kaplan Hospital, POB 1, pational accidents. Neuropsychology consulta- Some authors have devised tests which are Rehovot, Israel. tion is usually required for medicolegal difficult to feign even when the subject knows Received 12 October 1995 purposes. The discrepancy between subjective the technique, such as the symptom validity and in final revised form 19 December 1996 and objective findings is the rule in mild head test'2 13 and the portland digit recognition test.8 Accepted 30 December 1996 injury,23 which is perhaps the most frequent We report the application of the Stroop test 618 Osimani, Alon, Berger, Abarbanel J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.6.617 on 1 June 1997. Downloaded from as a diagnostic tool for identifying malingerers. NORMAL CONTROLS It has the advantage that it can detect simula- Group I tion of cognitive impairment, even when the Ten healthy subjects, age range from 20 to 38 subject is familiar with the test. (mean 25) years, participated as controls for The Stroop test'4 is based on the finding groups IV, V, and VI. that for normal people, reading a word takes less time than naming the object or colour cor- Group II responding to that word. Basically, the test Ten healthy subjects, age range from 60 to 75 consists of presenting the names of different (mean 67) years, were controls for group III. colours written in coloured ink. Part of the To have a correct matching for age, educa- stimuli are presented in such a way that word tion, and cultural background, patients from and ink colour correspond that is, the word groups III and IV were asked to bring a "red" is written in red (congruent condition = healthy sibling, spouse, or close relative to be CC), and part of the stimuli consist of words tested as a control. written in a different colour that is, the word "red" is written in blue (incongruent condition PATIENTS WITH BRAIN DAMAGE = IC). As a baseline, a third situation (neutral Group III condition = NC) is often included, in which a Ten patients with diagnosis of mild to moder- meaningless string of letters in different ate cognitive impairment due to degenerative colours is presented. Subjects are asked to or vascular disease were recruited at random name the colour and ignore the written word. from a clinic. Their ages ranged In NC, the meaningless string of letters does from 60 to 75 (mean 68) years. not interfere with colour naming. In CC not only is there no conflict between the two Group IV simultaneous stimuli but there is facilitation, Ten patients with closed head injury were so that reaction time is usually shorter than in recruited from a neurosurgery clinic. Their NC. In IC, reaction time is the longest because ages ranged from 17 to 38 (mean 27) years. the automatic reading of the word interferes Time from injury ranged from six to 10 with the naming of the colour and the subject months and they were diagnosed as having must inhibit the automatic response. The mild cognitive deficits. They were aware of number of errors is also higher in IC. The participating in experimental research that had Stroop effect is the difference between reaction no value for any litigation process. time in IC and NC or CC. The Stroop effect Cognitive deficits were diagnosed by a certi- may be measured in terms of reaction time and fied behavioural neurologist, using general in terms of error percentage because this is clinical examination and a battery of neu- usually higher in IC than NC or CC. Previous ropsychology tests. In group IV, the main reports have shown that the inhibition effect deficits were in the sphere of concentration, as (longer reaction time for IC) is relatively large measured by serial 7's test; and memory, as and stable, whereas the facilitatory effect (RT measured by Rey's auditory verbal list difference between CC and NC) is small and and Rey's complex figure. In group III, the

may not reach significance.'5 deficits were varied depending on aetiology, http://jnnp.bmj.com/ In neuropsychology this technique has been but all patients had scores between 100 and applied in particular to the assessment of 120 using the Mattis dementia rating scale. frontal lobe function and closed head injury.16 2() Depressed patients have been SIMULATORS reported to show an increased Stroop effect.2 Group V Loss of the Stroop effect has never been Ten normal volunteers taken at random from described in patients with brain damage. hospital staff and from relatives of patients in The accidental finding of an inverted Stroop hospital participated as naive simulators. Ages on September 26, 2021 by guest. Protected copyright. effect (longer reaction time in CC than in IC) ranged from 23 to 35 (mean 25) years. They in three patients suspected of malingering were not acquainted with any neuropsychol- from the history and the results of other tests ogy test. led us to the present study. The objectives were: (1) to compare the results of patients Group VI with brain damage, in particular those with Ten psychology undergraduates, age range head injury, with those of healthy controls and from 23 to 33 (mean 25) years participated in simulators; (2) to assess the efficiency of the the experiment. They were well acquainted Stroop test as a tool for identifying malinger- with the Stroop test used in this study and had ing, even when malingerers are acquainted worked with it in different research projects. with the test and able to plan a feigned perfor- mance. METHODS Stimuli We employed four colours: blue, green, red, Material and methods and yellow. These colours are all four letter SUBJECTS words in Hebrew. For NC we used a repeated Sixty subjects participated in this study, con- Hebrew letter (vWWWW). There were three stituting six groups. In each group there was a types of stimuli: CC (for example, RED in red similar number of subjects of both sexes. All of ink); IC (BLUE in red ink) and NC (WWvWV in them had vision that was normal or corrected the different colours). There were four differ- to normal. ent CCs, four different NCs, and 12 different Use ofthe Stroop phenomenon as a diagnostic toolfor malingering 619 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.6.617 on 1 June 1997. Downloaded from ICs. The total number of stimuli was 216 and 100 000 dollars if you can fake the results and prove they were presented randomly during one ses- that you are cognitively impaired. Use your imagina- sion. tion on how to do it" Apparatus Group VI received the following instruc- Stimuli were presented on a 14 inch EGA tions: Olivetti colour monitor. Stimuli presentation "As you know, many people with mild head injury and data collection were controlled by an show some form of cognitive impairment mainly in Olivetti M-240 IBM compatible the sphere of information processing. You will be computer. given a Stroop test like the one you use in the labo- Vocal responses were recorded via a Gerbrand ratory. You must pretend that you had a mild head voice operated relay (model G1341T) inter- injury. You must devise a way to prove that you faced to the computer. have cognitive impairment, but you must keep in mind that the Stroop effect is never lost." Design and procedure The independent variables were: "group" (six This last stress was added so that they levels) and "congruency" (three levels, accord- would organise strategies, as it could be the ing to ink-word congruency condition: CC, case with malingerers who are trained for this NC, IC). NC was not included in the analysis test. for simplification purposes. The dependent Once the test was performed, simulators variables were: reaction time (measured in ms were interviewed and asked to explain the pro- from stimulus onset to the onset of the sub- cedure they had applied to feign the test. ject's vocal reaction) and error percentage (for For each subject in each condition median each subject, in each condition). reaction time in ms and the error percentage The experiment began with a central fixa- were computed. A separate analysis of vari- tion "plus" sign for 100 ms. Right after fixa- ance (ANOVA) was performed for mean reac- tion the stimulus appeared for 100 ms tion time and error percentage. In addition, followed by a fixed pattern of white symbols. planned comparisons were done between the The subject responded through the micro- following pairs of groups: young controls (I) phone whereas the fixed pattern was in the and simulators (V-VI); simulators (V-VI) and screen and the experimenter keyed the sub- patients with head injury (IV); young controls ject's response on four keys in the computer. (I) and older controls (II); older controls (II) The experimenter's pressure of the key deter- and demented patients (III). mined the end of the white fixed pattern and the appearance of the new stimulus on the screen. Results Subjects were run individually in a 15 POSTPROCEDURAL INTERVIEW minute session. A session began with a prac- Simulators used different strategies to feign tice block of 45 trials, composed of the same impairment. Some of them, mainly the conditions and proportion of trials as the untrained simulators, tried to increase the experimental block. number of errors, others, mainly those of

group VI, tried to increase reaction time, and http://jnnp.bmj.com/ Instructions some of them tried both. Some of the students The general instructions given to all groups tried to work as metronomes, without paying were: to the stimuli, in an attempt to "In the screen you will see words or strings of letters increase the number of errors. Most of them written in different colours. The words you will see stated that the original plan was to increase are: "blue", "green", "red", or "yellow". The strings reaction time and error percentage, but the speed of presentation of stimuli did not permit

of letters will be the Hebrew letter "shin" repeated on September 26, 2021 by guest. Protected copyright. several times. As soon as the coloured word appears them to manipulate the answers properly. Two you must say the colour you see, ignoring the writ- subjects from group V said that they were ten word or letters. Try to answer as quickly as pos- pulled by the dynamics of the test and could sible and refrain from making errors" not fake results. In addition to these instructions, group V ANALYSIS OF RESULTS received the following explanation: Reaction time analysis "As you probably know, people with head injury The table shows the means (SD) for each have difficulty in concentration, memory, and per- group. The main findings were the following: forming different mental tasks. When you perform (1) Age effects: young controls showed sig- this test, you must pretend that you have had a head nificantly shorter reaction time than older con- injury and the insurance company will pay you trols (F (1,54) = 3.8, P < 0-05). (2) Brain damage effects: comparison between young controls (I) and patients with Reaction times (mean (SD)) for each group head injury (IV) showed a significantly shorter Congruent Incongruent Neutral reaction time for group I (F (1,54) = 4-63, Mean (SD) Mean (SD) Mean (SD) P < 0-03). The difference between older con- Group I 388-0 (128-6) 504 7 (191-8) 437 0 (144-0) trols (II) and dementia (III) had the same Group II 581-0 (146-6) 790-1 (247-5) 653-3 (177-7) trend Group III 641-2 (84-9) 805-2 (161-2) 706-9 (112-4) but did not reach significance. Group IV 630-8 (435 9) 788-6 (390 3) 690-4 (462 9) (3) Simulators' reaction time: groups V and Group V 1016-8 (439-7) 905-9 (234-9) 798-7 (251-6) Group VI 1008-8 VI showed significantly longer reaction time (391-5) 827-9 (261-8) 773-5 (270 6) than young controls and patients with head 620 Osimani, Alon, Berger, Abarbanel J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.6.617 on 1 June 1997. Downloaded from the analysis of error percentage. Group V showed a trend to an inverted Stroop effect; group VI showed practically no difference ,t 1{ between CC and IC. !11,,l

Discussion It may be claimed that differences exist between the performance of malingering patients and simulators. However, the same method has been applied in many previous studies suggesting that the results obtained Group may be generalised.10 In the present study, on

Figu re 1 Stroop effect in ms (IC- CC) for each group. I the other hand, the participation of the stu- woung controls; II = older controls; III = dementia; dents in group VI was important because one IV == head injury; V = naive simulators; VI = trained of the aims was to show that training dlators. previous simu or knowledge of the test did not help in feigning the results. injutry (F (1,54) = 21-68, P < 0-01 and We shall not discuss the findings referring F (]1,54) = 4-71, P < 0 03 respectively). to the direct relation between age and reaction Fiigure 1 presents the results of the Stroop time seen in the control groups, or the longer effect (IC-CC), in reaction times, for each reaction time shown by patients with head groiup. The Stroop effect was significant for injury. These results have been repeatedly the six groups, as reflected by significant main reported and discussed."6 20122 However, the effects of the congruency variable in the six extremely long reaction time of simulators casees (F (1,54) = 4-88, P < 0 03; F (1,54) = compared with patients with head injury is 15-'7, P < 0-01; F (1,54) = 9-65, P < 0 03; F worthy of comment. After our initial three (1 ,554) = 8-93, P<0 04; F (1,54) = 4-41, patients we have studied two more subjects P <:0-04; F (1,54) = 11-7, P < 0-01). The with a high suspicion of malingering. Although samie pattern of Stroop effect was found for their number is too small to include them in the first four groups, and an inverted Stroop the study, in all cases we have found not only effect was found for simulators (V-VI), as loss or inversion of the Stroop effect but also refl4 ected in the significant interaction between an exaggerately prolonged reaction time, in groiup and congruency (F (5,54) = 9 55, the range of 2000 or 3000 ms. This reaction P <:0 01). The inversion of the Stroop effect is time is much longer than the one shown by maiinly due to the increased reaction time in our simulators. By contrast, two patients with CC hysterical amnesia showed completely normal results. Errm or analysis The most important finding in this study is Simiulators (V-VI) showed significantly more the inversion of the Stroop effect in simula-

totail errors than the other groups (I, II, III, tors, whereas in the other groups the Stroop http://jnnp.bmj.com/ IV) (F (1,54) = 9.7, P < 0 02). Group V effect was preserved or exaggerated. A pattern showed the highest error percentage. such as this has never been described before. EFigure 2 is a plot for error percentage for The results of this study confirm our chance eac]h congruency condition, for each group, findings in patients with a high suspicion of shopwing the Stroop effect as measured by malingering, who had very long reaction times errc)r percentage in the different groups. There and a markedly inverted Stroop effect. Naive was a significant interaction between simula- as well as trained simulators showed a pattern tors (V-VI) versus the other groups (I-IV) and incompatible with either normal subjects or on September 26, 2021 by guest. Protected copyright. con[gruency (F (1,54) = 15-3, p < 0-01). patients with brain damage. The situation cre- Furrther analysis yielded a significant Stroop ated in this experiment, by using subjects who effect for error percentage for group I, II, and were trained in the test, is extreme, and proba- III (F (1,53) = 4.74, P < 0 03; F (1,53) = bly not a common occurrence. However, 21- 4, P < 0-01; F (1,53) = 9.97, P < 0 02) questions always arise suggesting that many resj pectively. Group IV showed a similar trend, malingerers are "trained to feign". Our results wit]hout reaching significance. Simulators, on show that training does not help to feign cogni- the contrary, showed no significant effect in tive impairment in this kind of test. Three possible patterns of performance need to be considered: (1) the subject shows a Figure 2 Error 50 normal performance, which does not confirm percentage at each M Congruent condition or deny any diagnosis; (2) the subject shows congruency condition, for 40 -W Incongruent condition longer reaction time, or increased error per- each group. I = young centage, or both, and maintains the Stroop controls; II = older 30 controls; III = dementia; effect, which strongly suggests organic brain IV = head injury; V = 20 damage; and (3) the subject shows a very long naive simulators; VI = trained simulators. reaction time or increased error percentage, or 10 n, both, and the Stroop effect is lost or inverted, 11n-2MP|111 J 1 _ which strongly suggests malingering. In the 0 11 III IV V VI first case, although the result does not imply Group any diagnosis, normal reaction time indicates Use ofthe Stroop phenomenon as a diagnostic toolfor malingering 621 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.6.617 on 1 June 1997. Downloaded from that there is no serious brain damage, espe- Law and Human Behaviour 1988;12:57-78. 10 Mittenberg W, Azrin R, Millsaps C, Heilbronner R. cially when dealing with closed head injury, Identification of malingered head injury on the Wechsler which almost always causes a prolonged reac- memory scale-revised. J Clin Exp Neuropsychol 1992;14: tion These 119-24. time. results must be compared 11 Rawling PJ, Brooks DN. Simulation index: a method for with those of other tests (a very poor score in detecting fictitious errors on the WAIS-R and WMS. Neuropsychology 1991;4:223-38. tests of attention and concentration in the 12 Pankratz L Symptom validity testing and symptom retrain- presence of a normal Stroop test is suggestive ing: procedures for the assessment and treatment of func- tional sensory deficits. Jf Consult Clinical Psychol 1979;47: of malingering). 409-10. 13 Pankratz L. A new technique for the assessment and modi- fication of feigned memory deficit. Percept Mot SkiUs 1 American Psychiatric Association. Diagnostic and statistical 1983;57:367-72. manual IV. Washington, DC: American Psychiatric 14 Stroop JR. Studies of interference in serial verbal reactions. Association, 1994. J Exp Psychol 1935;18:643-62. 2 Dickmen S, McLean A, Temkin N. Neuropsychological 15 MacLeod CM. Half a century of research on the Stroop and psychosocial consequences of minor head injury. J7 effect: an integrative review. Psychol Bull 1991;109: Neurol Neurosurg Psychiatry 1986:1227-32. 163-203. 3 Levin HS, Amparo F, Eisenberg HM, Williams OH, High 16 Stuss DT, Benson DF. Neuropsychological studies of the WH, McArdle CB, Weiner RC. Magnetic resonance frontal lobes. Psychol Bull 1984;95:3-28. imaging and computerized tomography in relation to the 17 Stuss DT, Ely P, Hugenholtz H, Richard MT, Larochelle neurobehavioural sequelae of mild and moderate head S, Poirier CA, Bell I. Subtle neuropsychological deficits injuries. J7 Neurosurg 1987;66:706-13. in patients with good recovery after closed head injury. 4 Levin HS, Mattis S, Ruff RM, et al. Neurobehavioural out- Neurosurgery 1985;17:41-7. come of minor head injury: A three-center study. J 18 McLean A, Temkin NR, Dikmen S, Wyler AR. The behav- Neurosurg 1987;66:234-43. ioural sequelae of head injury. Jf Clin Neuropsychol 1983; 5 Binder LM. Persisting symptoms after mild head injury: a 5:361-76. review of the postconcussive syndrome. J Clin Exp 19 Chadwick 0, Rutter M, Brown G, Shaffer D, Traub M. A Neuropsychol 1986;8:323-46. prospective study of children with head injuries: II cogni- 6 McMordie WR. Twenty-year follow-up of the prevailing tive sequelae. Psychol Med 1981;11:49-62. opinion on the posttraumatic or postconcussional syn- 20 Van Zomeren AH. Reaction time and attention after closed drome. The Clinical Neuropsychologist 1988;2: 198-212. head injury. Lisse: Swets, 1981. 7 Resnick PJ. Malingering of posttraumatic disorders. In 21 Trichard C, Martinor JL, Alagille M, Masure MC, Hardy Rogers R, ed. Clinical assessment of malingering and decep- P, Ginestet D, Feline A. Time course of prefrontal lobe tion, New York, Guilford, 1988:84-107. dysfunction in severely depressed patients: a longitudinal 8 Binder LM. Malingering following minor head trauma. The neuropsychological study. Psychol Med 1995;25:79-85. Clinical Neuropsychologist 1990;4:25-36. 22 Kausler DH. Learning and memory in normal aging. New 9 Wiggins EC, Brandt J. The detection of simulated amnesia. York: Academic Press. 1994. http://jnnp.bmj.com/ on September 26, 2021 by guest. Protected copyright.