- Email: [email protected]
The detection of faked deficits on the Rey Auditory Verbal Learning Test: The effect of serial position
Adives Prided
ofCK.nical N..wo~hology, W. in the USA. All rights rcscaved.
6, pp. 81-88,1991 cm8ht
0887-6177/91$3.00 + .OO 0 1991 National Academy of NwopsychoIogy
The Detection of Faked Deficits on the Rey Auditory Verbal Learning Test: The Effect of Serial Position Larry C. Bernard Loyola Marymount
University
The ability of subjects to fake deficits on the Rey Auditory Verbal Learning Test (AVLT) was evaluated. Subjects were randomly assigned ro one of two malingering groups - one with a financial incentive (N = 30) and one without (N = 28) - or a control group (C) (N = 28), and a group of closed head inJ’ury patients (CHI) (N = 18) was matched on age, sex, and education level. The two malingering groups did not differ significantly and were combined into a single malingering group (M) (N = 57). There was a significant serial position (the pattern of “recency” and “primacy” effects in recall} by group interaction effect, which may be the most promising indicator of deliberate distortion. When only level of performance was examined, the M group couldfake believable deficits, but when the serial position effect was examined, it revealed that this was accomplished by suppression of recall from the first third of the word list (reducing the “primacy effect”). a recall pattern which did not occur in either the C or CHI group and may be an indication of malingering.
Memory complaints are commonplace in patients with neurological symptoms, in persons who wish to feign neurological symptoms, and in nonpatients alike (Brandt, 1988). This presents a problem in neuropsychological and forensic evaluations, particularly if maIingering is a consideration. In fact, Rogers (1988) contends that all psychological tests are vulnerable to faking. Heaton, Smith, Lehman, and Vogt (1978) demonstrated the ability of adult subjects to simulate malingering and to duplicate the overall level of impairment of head trauma patients, and Faust, Hart, and Guilmette (1988) demonstrated a similar capacity on the part of children to fake believable deficits on neuropsychological testing. This suggests that popular neuropsychological Requests for reprints should be sent to Larry C. Bernard, Psychology Department, Loyola Marymount University, LoyoIa Boulevard at West 80th Street, Los Angeles, CA 90045.
82
L. C. Bernard
tests should be examined to determine if the performance pattern in simulated malingering is detectable. A recent literature review revealed no studies of malingering on Rey’s Auditory Verbal Learning Test (AVLT, Rey, 1964), although it is a popular instrument among neuropsychologists (Lezak, 1985). The AVLT has lacked adequate psychometric analysis, but recent studies have added to the normative data that are available, including some for clinical populations (Bigler, Rosa, Schultz, Hall, & Harris, 1989; Mungas, 1983; Query & Megram, 1983; Rosenberg, Ryan, & Prifitera, 1984). Even so, more knowledge about the AVLT’s psychometric properties is needed, specifically in relation to faking bad. The AVLT is of particular interest because it allows for a measure of the serial position effect, and the implications of this effect for neuropsychological assessment have rarely been studied. Rundus (1971) first explored the serial effect in normal subjects, demonstrating that items at the beginning (primacy effect) and end (recency effect) of a word list are recalled better than those in the middle. The effect causes a U-shaped performance curve. Rundus theorized that these two seemingly contradictory effects are due to the separation of long-term memory (LTM) and short-term memory (STM) processes. He reasoned that the primacy effect occurs because words at the beginning of the list are rehearsed more often so they are more likely to be transferred to LTM, whereas the recency effect occurs because words at the end of the list are still available in STM. There is some evidence to support this theory (Fischler, Rundus, & Atkinson, 1970). Recently, Bigler et al. (1989) examined the serial position effect on AVLT recall in two clinical groups - closed head injury and dementia patients. Both “primacy” and “recency” effects were obtained, suggesting that the serial position effect may also be present in brain injury and dysfunction. [Quotation marks are used here because serial position effects across pooled trials of the AVLT may not represent STM and LTM processes as proposed by Rundus (1971).] However, the groups’ performance curves were shaped differently, which is of particular interest for assessment and diagnosis of these patient groups. The CHI patients produced a U-shaped curve similar to Rundus’ (1971) normals, whereas the dementia patients exhibited a more pronounced “primacy” than “recency” effect. The analysis of serial position effects on the AVLT provides an opportunity to evaluate memory functions that may have gone unexplored previously. If different clinical groups have characteristic performance patterns, and these can be demonstrated, this has differential implications for the type of underlying neuropathology. In addition, if malingerers do not produce known performance patterns characteristic of clinical groups, it would aid in the clinical determination of malingering. This study compares the performance of subjects attempting to fake memory deficits with a group of closed head injury (CHI) patients and controls on
Faked Deficits
83
the AVLT. CHI characteristically involves bilateral frontal and temporal injuries that impair memory abilities (Reitan & Wolfson, 1985), and a matched control group was thought to provide a reference point for comparing serial position effects on the AVLT. Heaton et al. (1978) did not investigate the AVLT, but based on their results and those of Faust et al. (1988), it was hypothesized that the level of performance of subjects attempting to fake bad would be significantly below controls and possibly as poor as CHI patients. It was also hypothesized that there would be a serial position effect similar to that found by Bigler et al. (1989), but that subjects attempting to fake bad would distort the shape of the curve. If distortion were found, this would introduce the possibility of a new approach to the clinical detection of malingering on the AVLT and also support the investigation of performance patterns rather than scores in future studies of malingering on neuropsychological tests. METHOD
Eighty-six undergraduate students, participating for partial course credit in a study of the effects of incentive versus nonincentive conditions on simulated malingering, were randomly assigned to either a control group (C) (N = 28), a malingering group given a financial incentive (N = 30), or a malingering group not given a financial incentive (ZV= 28). Univariate ANOVAs on the AVLT (Trials 1-5, Interference, Recall, and Recognition scores) revealed no significant differences between the incentive and nonincentive malingering groups, so they were combined for this study. The data were screened for multivariate analysis (Tabachnick & Fidell, 1989) and two cases were identified as multivariate outliers and dropped. With one subject dropped from each group, there remained 84 subjects: 27 in the C and 57 in the combined malingering (M) group. Subjects in the C group were instructed to perform to the best of their ability, while subjects in the M group (both incentive and nonincentive) were instructed that they had been in a car accident. They were told that they received “a head injury,” they had been initially “unconscious,” had been “hospitalized for several weeks,” now were discharged, but had “memory problems.” They were told to attempt to simulate “as serious memory problems as you can but in a believable way” that would keep the examiner (who was blind to the experimental hypothesis and to group membership) from knowing what they were doing. The CHI group was drawn from a large database of inpatients tested on discharge from a rehabilitation hospital between 1981 and 1987. All had sustained a closed head injury in a motor vehicle accident with conditions similar
L. C. Bernard
84
to the instructions given to the M group. The CHI patients had generalized brain injuries, primarily involving bilateral frontal and temporal lobes, and tissue damage was verified by multiple diagnostic procedures including a CT or MRI scan. A subgroup of patients was selected so that the CHI group would be equivalent to the C and M groups. Frequency distribution matching was used for age, and univariate ANOVA revealed that there was not a significant difference between the C (M = 19.1; SD = 1.3), M (M = 19.5; SD = 1.6), and CHI (M = 20.1; SD = 2.3) groups (F(2,99) = 1S6; p = .21). There also was not a significant difference in the distribution of males in the C (n = lo), M (n = 29), and U-II (n = 11) groups or of females in the C (n = 17), M (n = 28), and CHI (n = 7) groups (X2(2) = 2.68; p = .26). It was only possible to approximate a match for education level by insuring that the mean years of school completed by the CHI group (M = 12.7; SD = 1.4) was similar to the education level of the C and M groups (which was first year of college). The CHI group contained 18 subjects after matching. Testing took place during the week of discharge which was at a mean number of weeks postinjury of 15.1 (SD = 13.9). Of course, the mean WAIS-R IQ of the CHI group (78.8; SD = 13.3) was significantly below (F(2,96) = 73.4; p = .OOOl) the IQs of the C (107.4; SD = 6.3) and M (106.9; SD = 7.5) groups which were obtained before group assignment and were not significantly different from each other. Test Auditory Verbal Learning Test (Rey, 1964). The AVLT is a complex measure of several verbal memory functions including immediate recall, memory
TABLE 1 Rey Auditory Verbal Learning Test Performance of Controls (C), Malingerers (M), and Closed Head Injury (CHI) Subjects Group Test
AVLT Trial 1 2 3 4 5 Inter. Re4Xdl
C
M
CHI
M SD
M SD
M SD
7.6 (2.0)a 10.5 (2.3)a 12.0 (1.9)a 12.7 (1.4)a 13.4 (1.4)” 6.8 (1.9)a 12.0 (2.l)a 14.0 (l.op
6.1 (1.7)b 8.3 (2.4)b 9.4 (3.0)b 10.3 (3.4)b 10.8 (3.6)b 5.2 (1.9)b 8.3 (4.1)b 10.0 (4.3)b
4.3 (1.4)C 5.7 (1.7)C 6.4 (2.4)c 7.2 (3.3)c 7.7 (3.2)c 3.5 (2.1)C 5.3 (4.2)c 8.8 (5.8)b
F 18.27 24.00 24.34 18.75 18.35 16.08 18.39 11.54
P BOO1 BOO1 BOO1 BOO1 .OOOl .0001 BOO1 .ooOl
Inter. = Interference list; Recog. = Recognition Trial; the mean’s of groups with different superscripts differ significantly on univariate ANOVA and Tukey HSD analysis, df= 2.99 for all analyses.
Faked Deficits
85
span, learning over trials, and recall and recognition after distraction. The following scores were used: immediate recall after each Trial l-5; immediate recall of the distraction list (Interference); recall after interference (Recall); and recognition after interference (Recognition). Procedure The AVLT was a~~stered
according to Lezak (1985) as part of a longer test battery. Bigler et al.‘s (1989) procedure was used to test for “recency” and “primacy” effects on the AVLT. The 15 words in each trial were grouped into
85. P e r c e n t
80ISTO6560-
R e c a l 1
55so-
CHIC'
4540-
CHIh3
35-
\
30-
/
CHICZ
252015105oSummed Trials:
I
I
first third
middle third
primacy effect
I last third recency effect
FIGURE 1. “Primacy” and “recency” effects on the Auditory Verbal Learning Test in Control (C), Malingerer @I), and Closed Head Injury (CHI) groups. Significance is given for the first, middle, and Iast thirds of the word list within and between groups. Reading vertically (withim each of the three summed trials} groups with different superscript letters differ significantly @ c 301). Reading horizontally (across each group’s three trials) trials with different superscript numerals differ significantly (p e .OOl).
86
L. C. Bernard
the first third (l-5), middle third (6-lo), and last third (11-15). and summed for the five consecutive trials with the percent recalled calculated. In the absence of “primacy” and “recency” effects, performance would be equal across the first, middle, and last trials. If both a “primacy” and “recency” effect were present, then the U-shaped curve characteristic of serial learning would be present (Rundus, 1971).
RESULTS The scores of the C, M, and CHI groups on the AVLT are presented in Table 1. Although the groups did not begin or end at the same level, all three demonstrated equivalent improvement in learning with a positive learning curve from Trial 1 to Trial 5 (C Group mean performance: 7.6/15 to 13.4/15, gain of 43%; M Group mean performance: 6.1/15 to 10.8/15, gain of 43%; CHI Group mean performance: 4.3/15 to 7.7/15, gain of 44%). The percent recalled of the first, middle, and last portions of the word list pooled across the five trials are depicted in Figure 1, providing an analysis of “primacy” and “recency” effects. MANOVA for repeated measures was used for a profile analysis of the C, M, and CHI groups’ scores for serial position (first, middle, and last third of the word list). Following Wilks’ criterion, the profiles seen in Figure 1 deviated significantly from parallelism, F(4,192) = 4.109, p = .003. For the levels test, reliable differences were found among groups when scores were averaged over all serial positions, F(2,97) = 28.501, p = .OOOl,and when averaged over groups, the serial position scores differed significantly from flatness, F(2,194) = 34.512, p = .OOOl. Tukey HSD comparison of pairwise means between groups (designated by superscript letters in Figure 1) indicated that for the first, middle, and last third of the word list, all group means were significantly different with the exception of the C and M groups on the last third of the word list. Tukey HSD comparison of pairwise means within groups (designated by superscript numerals in Figure 1) indicated that for the M and CHI groups, the mean percent recalled was significantly different for each third of the word list, while for the C group, the percents recalled of the first and last third of the word list were not significantly different from each other, although both were significantly above the percent recalled from the middle of the word list. This pattern in the C group is similar to that of Rundus’ (1971) normals, although he did not use repeated trials. Therefore, as expected, the C group produced a U-shaped recall curve demonstrating “primacy” and “recency” effects, whereas neither the M nor the CHI groups did. In the M group, the percent recalled of the first third of the list was significantly below the percent recalled of the last third, while in the CHI group the situation was reversed.
Faked Deficits
87
DISCUSSION The CHI patients and the subjects faking bad did produce deficits in verbal memory as measured by the AVLT compared to the C group. The M group’s performance was significantly poorer than the C group, but it was significantly better than the CHI group. Subjects can fake bad on the AVLT, but unlike Heaton et al’s (1978) malingering subjects, the level of performance was not as poor as the CHI group. However, this discrepancy may result from the characteristics of different CHI patients. The CHI group in this study is representative of patients who require inpatient rehabilitation and who are at a relatively early point (although not the acute phase) in recovery. A different sample of CHI patients with less serious injuries and/or at a later point of recovery might perform at a level indistinguishable from the subjects faking bad. In fact, the level of performance of the M group is similar to that of Bigler et al.3 (1989) CHI group, suggesting that level of performance on the AVLT would not reliably differentiate malingering. Nevertheless, as predicted, the shape of the AVLT recall curve was distorted by subjects attempting to fake bad. This provides a tentative but promising indication that malingering may be detected on the AVLT if the serial position effect is taken into account. It also suggests that the performance pattern how scores are obtained - rather than the pattern of scores themselves, may be a good indicator of malingering on neuropsychological tests. Bigler et al. (1989) presented evidence that different clinical groups (CHI and dementia) produce different serial recall patterns. The present study suggests that there may be other serial recall patterns that may have diagnostic utility. The M group produced a significantly reduced “primacy” versus “recency” effect; their recall percent was significantly below that of the C group for the first third of the list but it was not significantly different from the C group for the last third of the list. The shape of the M group’s curve is also different from the CHI group in the present study and the CHI group in Bigler et al. (1989). These differences suggest that subjects who attempt to fake bad do so by suppressing recall of words from the first third of the list. According to Rundus (1971), absence of the primacy effect would suggest impairment of LTM (but not STM). However, the presentation of repeated trials on the AVLT is not consistent with Rundus’ original method, so conclusions from his theory about LTM versus STM may not be applicable. Any determination of faking bad must be approached cautiously for several reasons. First, it should be kept in mind that it is not the level of performance that is being compared (not the absolute percent recalled or the percent learned over trials) for the performance of malingerers may not be much different from some CHI patients with organic memory dysfunction. Second, there should be an explanation for these patterns that is consistent with what is known about neuropsychological functioning. One tentative explanation is that the mechanism underlying the performance of patients with verified
L. C. Bernard
88
memory problems secondary to CHI is different from subjects attempting to fake bad. The “primacy effect” appeared to be evident in the mean performance of the CHI group (in fact, in the present study, proactive interference appears to have caused a marked “primacy effect”), whereas when malingering subjects were consciously suppressing recall, the “primacy effect” was reduced to a level below the “recency effect.” Finally, these results, while intriguing, need to be replicated and explored further. In the meantime, the AVLT must be used with caution where malingering is suspected. Malingering is not a simple category, for it may include some people with genuine impairment who are also exaggerating their deficits. Therefore, malingering should not be determined on the basis of performance on a single test or procedure, instead a broad profile with psychometric data as one source should be developed (Rogers, 1988). Future research may be able to further delineate what pattern deliberate faking produces on tests of neuropsychological functioning, aiding the clinician who must confront the question of differential diagnosis.
REFERENCES Bigler, E. D., Rosa, L., Schultz, F., Hall, S., & Harris, J. (1989). Rey-Auditory Verbal Learning Test and Rey-Osterrieth Complex Figure Design performance in Alzheimer’s disease and closed head injury. Journal of Clinical Psychology, 45.277-280. Brandt, J. (1988). Malingered amnesia. In R. Rogers (Ed), Clinical assessment of malingering and deception (pp. 65-83). New York: Guilford Press. Faust, D., Hart, K., & Guilmette, T. J. (1988). Pediatric malingering: The capacity of children to fake believable deficits on neuropsychological testing. Journal of Consulting ana’ Clinical Psychology,
56,578-582.
Fischler, I., Rundus, D., & Atkinson, R. C. (1970). Effects of overt rehearsal processes on free recall. Psychonomic Science, 19, 249-350. Heaton, R. K., Smith, H. H., Lehman, R. A. W., & Vogt, A. T. (1978). Prospects for faking believable deficits on neuropsychological testing. Journal of Clinical and Consulting Psychology, 46.892-900.
Lezak, M. (1985). Neuropsychological assessment (2nd ed.). New York: Oxford University Press. Mungas, D. (1983). Differential clinical sensitivity of specific parameters of the Rey AuditoryVerbal Learning Test. Journal of Consulting and Clinical Psychology, 51, 848-855. Query, W. T., & Megram, J. (1983). Age-related norms for AVLT in a male patient population. Journal of Clinical Psychology, 39.136-138. Reitan, R. M., & Wolfson, D. (1985). Neuroanatomy and neuropathology: A clinical guide for neuropsychologists. Tucson, AZ Neuropsychology Press. Rey, A. (1964). L’examen clinique en psychologie. Paris: Presses Universitaires de France. Rogers, R. (1988). Current status of clinical methods. In R. Rogers (Ed.), Clinical assessment of malingering and deception (pp. 293-308). New York: Guilford Press. Rosenberg, S. J., Ryan, J. J., & Prititera, A. (1984). Rey Auditory-Verbal Learning Test performance of patients with and without memory impairments. Journal ofclinical Psychology, 40, 785-787.
Rundus, D. (1971). Analysis of rehearsal processes in free recall. Journal of Experimental Psychology, 89,63-72.
Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariate statistics (2nd ed.). New York: Harper & Row.
ofCK.nical N..wo~hology, W. in the USA. All rights rcscaved.
6, pp. 81-88,1991 cm8ht
0887-6177/91$3.00 + .OO 0 1991 National Academy of NwopsychoIogy
The Detection of Faked Deficits on the Rey Auditory Verbal Learning Test: The Effect of Serial Position Larry C. Bernard Loyola Marymount
University
The ability of subjects to fake deficits on the Rey Auditory Verbal Learning Test (AVLT) was evaluated. Subjects were randomly assigned ro one of two malingering groups - one with a financial incentive (N = 30) and one without (N = 28) - or a control group (C) (N = 28), and a group of closed head inJ’ury patients (CHI) (N = 18) was matched on age, sex, and education level. The two malingering groups did not differ significantly and were combined into a single malingering group (M) (N = 57). There was a significant serial position (the pattern of “recency” and “primacy” effects in recall} by group interaction effect, which may be the most promising indicator of deliberate distortion. When only level of performance was examined, the M group couldfake believable deficits, but when the serial position effect was examined, it revealed that this was accomplished by suppression of recall from the first third of the word list (reducing the “primacy effect”). a recall pattern which did not occur in either the C or CHI group and may be an indication of malingering.
Memory complaints are commonplace in patients with neurological symptoms, in persons who wish to feign neurological symptoms, and in nonpatients alike (Brandt, 1988). This presents a problem in neuropsychological and forensic evaluations, particularly if maIingering is a consideration. In fact, Rogers (1988) contends that all psychological tests are vulnerable to faking. Heaton, Smith, Lehman, and Vogt (1978) demonstrated the ability of adult subjects to simulate malingering and to duplicate the overall level of impairment of head trauma patients, and Faust, Hart, and Guilmette (1988) demonstrated a similar capacity on the part of children to fake believable deficits on neuropsychological testing. This suggests that popular neuropsychological Requests for reprints should be sent to Larry C. Bernard, Psychology Department, Loyola Marymount University, LoyoIa Boulevard at West 80th Street, Los Angeles, CA 90045.
82
L. C. Bernard
tests should be examined to determine if the performance pattern in simulated malingering is detectable. A recent literature review revealed no studies of malingering on Rey’s Auditory Verbal Learning Test (AVLT, Rey, 1964), although it is a popular instrument among neuropsychologists (Lezak, 1985). The AVLT has lacked adequate psychometric analysis, but recent studies have added to the normative data that are available, including some for clinical populations (Bigler, Rosa, Schultz, Hall, & Harris, 1989; Mungas, 1983; Query & Megram, 1983; Rosenberg, Ryan, & Prifitera, 1984). Even so, more knowledge about the AVLT’s psychometric properties is needed, specifically in relation to faking bad. The AVLT is of particular interest because it allows for a measure of the serial position effect, and the implications of this effect for neuropsychological assessment have rarely been studied. Rundus (1971) first explored the serial effect in normal subjects, demonstrating that items at the beginning (primacy effect) and end (recency effect) of a word list are recalled better than those in the middle. The effect causes a U-shaped performance curve. Rundus theorized that these two seemingly contradictory effects are due to the separation of long-term memory (LTM) and short-term memory (STM) processes. He reasoned that the primacy effect occurs because words at the beginning of the list are rehearsed more often so they are more likely to be transferred to LTM, whereas the recency effect occurs because words at the end of the list are still available in STM. There is some evidence to support this theory (Fischler, Rundus, & Atkinson, 1970). Recently, Bigler et al. (1989) examined the serial position effect on AVLT recall in two clinical groups - closed head injury and dementia patients. Both “primacy” and “recency” effects were obtained, suggesting that the serial position effect may also be present in brain injury and dysfunction. [Quotation marks are used here because serial position effects across pooled trials of the AVLT may not represent STM and LTM processes as proposed by Rundus (1971).] However, the groups’ performance curves were shaped differently, which is of particular interest for assessment and diagnosis of these patient groups. The CHI patients produced a U-shaped curve similar to Rundus’ (1971) normals, whereas the dementia patients exhibited a more pronounced “primacy” than “recency” effect. The analysis of serial position effects on the AVLT provides an opportunity to evaluate memory functions that may have gone unexplored previously. If different clinical groups have characteristic performance patterns, and these can be demonstrated, this has differential implications for the type of underlying neuropathology. In addition, if malingerers do not produce known performance patterns characteristic of clinical groups, it would aid in the clinical determination of malingering. This study compares the performance of subjects attempting to fake memory deficits with a group of closed head injury (CHI) patients and controls on
Faked Deficits
83
the AVLT. CHI characteristically involves bilateral frontal and temporal injuries that impair memory abilities (Reitan & Wolfson, 1985), and a matched control group was thought to provide a reference point for comparing serial position effects on the AVLT. Heaton et al. (1978) did not investigate the AVLT, but based on their results and those of Faust et al. (1988), it was hypothesized that the level of performance of subjects attempting to fake bad would be significantly below controls and possibly as poor as CHI patients. It was also hypothesized that there would be a serial position effect similar to that found by Bigler et al. (1989), but that subjects attempting to fake bad would distort the shape of the curve. If distortion were found, this would introduce the possibility of a new approach to the clinical detection of malingering on the AVLT and also support the investigation of performance patterns rather than scores in future studies of malingering on neuropsychological tests. METHOD
Eighty-six undergraduate students, participating for partial course credit in a study of the effects of incentive versus nonincentive conditions on simulated malingering, were randomly assigned to either a control group (C) (N = 28), a malingering group given a financial incentive (N = 30), or a malingering group not given a financial incentive (ZV= 28). Univariate ANOVAs on the AVLT (Trials 1-5, Interference, Recall, and Recognition scores) revealed no significant differences between the incentive and nonincentive malingering groups, so they were combined for this study. The data were screened for multivariate analysis (Tabachnick & Fidell, 1989) and two cases were identified as multivariate outliers and dropped. With one subject dropped from each group, there remained 84 subjects: 27 in the C and 57 in the combined malingering (M) group. Subjects in the C group were instructed to perform to the best of their ability, while subjects in the M group (both incentive and nonincentive) were instructed that they had been in a car accident. They were told that they received “a head injury,” they had been initially “unconscious,” had been “hospitalized for several weeks,” now were discharged, but had “memory problems.” They were told to attempt to simulate “as serious memory problems as you can but in a believable way” that would keep the examiner (who was blind to the experimental hypothesis and to group membership) from knowing what they were doing. The CHI group was drawn from a large database of inpatients tested on discharge from a rehabilitation hospital between 1981 and 1987. All had sustained a closed head injury in a motor vehicle accident with conditions similar
L. C. Bernard
84
to the instructions given to the M group. The CHI patients had generalized brain injuries, primarily involving bilateral frontal and temporal lobes, and tissue damage was verified by multiple diagnostic procedures including a CT or MRI scan. A subgroup of patients was selected so that the CHI group would be equivalent to the C and M groups. Frequency distribution matching was used for age, and univariate ANOVA revealed that there was not a significant difference between the C (M = 19.1; SD = 1.3), M (M = 19.5; SD = 1.6), and CHI (M = 20.1; SD = 2.3) groups (F(2,99) = 1S6; p = .21). There also was not a significant difference in the distribution of males in the C (n = lo), M (n = 29), and U-II (n = 11) groups or of females in the C (n = 17), M (n = 28), and CHI (n = 7) groups (X2(2) = 2.68; p = .26). It was only possible to approximate a match for education level by insuring that the mean years of school completed by the CHI group (M = 12.7; SD = 1.4) was similar to the education level of the C and M groups (which was first year of college). The CHI group contained 18 subjects after matching. Testing took place during the week of discharge which was at a mean number of weeks postinjury of 15.1 (SD = 13.9). Of course, the mean WAIS-R IQ of the CHI group (78.8; SD = 13.3) was significantly below (F(2,96) = 73.4; p = .OOOl) the IQs of the C (107.4; SD = 6.3) and M (106.9; SD = 7.5) groups which were obtained before group assignment and were not significantly different from each other. Test Auditory Verbal Learning Test (Rey, 1964). The AVLT is a complex measure of several verbal memory functions including immediate recall, memory
TABLE 1 Rey Auditory Verbal Learning Test Performance of Controls (C), Malingerers (M), and Closed Head Injury (CHI) Subjects Group Test
AVLT Trial 1 2 3 4 5 Inter. Re4Xdl
C
M
CHI
M SD
M SD
M SD
7.6 (2.0)a 10.5 (2.3)a 12.0 (1.9)a 12.7 (1.4)a 13.4 (1.4)” 6.8 (1.9)a 12.0 (2.l)a 14.0 (l.op
6.1 (1.7)b 8.3 (2.4)b 9.4 (3.0)b 10.3 (3.4)b 10.8 (3.6)b 5.2 (1.9)b 8.3 (4.1)b 10.0 (4.3)b
4.3 (1.4)C 5.7 (1.7)C 6.4 (2.4)c 7.2 (3.3)c 7.7 (3.2)c 3.5 (2.1)C 5.3 (4.2)c 8.8 (5.8)b
F 18.27 24.00 24.34 18.75 18.35 16.08 18.39 11.54
P BOO1 BOO1 BOO1 BOO1 .OOOl .0001 BOO1 .ooOl
Inter. = Interference list; Recog. = Recognition Trial; the mean’s of groups with different superscripts differ significantly on univariate ANOVA and Tukey HSD analysis, df= 2.99 for all analyses.
Faked Deficits
85
span, learning over trials, and recall and recognition after distraction. The following scores were used: immediate recall after each Trial l-5; immediate recall of the distraction list (Interference); recall after interference (Recall); and recognition after interference (Recognition). Procedure The AVLT was a~~stered
according to Lezak (1985) as part of a longer test battery. Bigler et al.‘s (1989) procedure was used to test for “recency” and “primacy” effects on the AVLT. The 15 words in each trial were grouped into
85. P e r c e n t
80ISTO6560-
R e c a l 1
55so-
CHIC'
4540-
CHIh3
35-
\
30-
/
CHICZ
252015105oSummed Trials:
I
I
first third
middle third
primacy effect
I last third recency effect
FIGURE 1. “Primacy” and “recency” effects on the Auditory Verbal Learning Test in Control (C), Malingerer @I), and Closed Head Injury (CHI) groups. Significance is given for the first, middle, and Iast thirds of the word list within and between groups. Reading vertically (withim each of the three summed trials} groups with different superscript letters differ significantly @ c 301). Reading horizontally (across each group’s three trials) trials with different superscript numerals differ significantly (p e .OOl).
86
L. C. Bernard
the first third (l-5), middle third (6-lo), and last third (11-15). and summed for the five consecutive trials with the percent recalled calculated. In the absence of “primacy” and “recency” effects, performance would be equal across the first, middle, and last trials. If both a “primacy” and “recency” effect were present, then the U-shaped curve characteristic of serial learning would be present (Rundus, 1971).
RESULTS The scores of the C, M, and CHI groups on the AVLT are presented in Table 1. Although the groups did not begin or end at the same level, all three demonstrated equivalent improvement in learning with a positive learning curve from Trial 1 to Trial 5 (C Group mean performance: 7.6/15 to 13.4/15, gain of 43%; M Group mean performance: 6.1/15 to 10.8/15, gain of 43%; CHI Group mean performance: 4.3/15 to 7.7/15, gain of 44%). The percent recalled of the first, middle, and last portions of the word list pooled across the five trials are depicted in Figure 1, providing an analysis of “primacy” and “recency” effects. MANOVA for repeated measures was used for a profile analysis of the C, M, and CHI groups’ scores for serial position (first, middle, and last third of the word list). Following Wilks’ criterion, the profiles seen in Figure 1 deviated significantly from parallelism, F(4,192) = 4.109, p = .003. For the levels test, reliable differences were found among groups when scores were averaged over all serial positions, F(2,97) = 28.501, p = .OOOl,and when averaged over groups, the serial position scores differed significantly from flatness, F(2,194) = 34.512, p = .OOOl. Tukey HSD comparison of pairwise means between groups (designated by superscript letters in Figure 1) indicated that for the first, middle, and last third of the word list, all group means were significantly different with the exception of the C and M groups on the last third of the word list. Tukey HSD comparison of pairwise means within groups (designated by superscript numerals in Figure 1) indicated that for the M and CHI groups, the mean percent recalled was significantly different for each third of the word list, while for the C group, the percents recalled of the first and last third of the word list were not significantly different from each other, although both were significantly above the percent recalled from the middle of the word list. This pattern in the C group is similar to that of Rundus’ (1971) normals, although he did not use repeated trials. Therefore, as expected, the C group produced a U-shaped recall curve demonstrating “primacy” and “recency” effects, whereas neither the M nor the CHI groups did. In the M group, the percent recalled of the first third of the list was significantly below the percent recalled of the last third, while in the CHI group the situation was reversed.
Faked Deficits
87
DISCUSSION The CHI patients and the subjects faking bad did produce deficits in verbal memory as measured by the AVLT compared to the C group. The M group’s performance was significantly poorer than the C group, but it was significantly better than the CHI group. Subjects can fake bad on the AVLT, but unlike Heaton et al’s (1978) malingering subjects, the level of performance was not as poor as the CHI group. However, this discrepancy may result from the characteristics of different CHI patients. The CHI group in this study is representative of patients who require inpatient rehabilitation and who are at a relatively early point (although not the acute phase) in recovery. A different sample of CHI patients with less serious injuries and/or at a later point of recovery might perform at a level indistinguishable from the subjects faking bad. In fact, the level of performance of the M group is similar to that of Bigler et al.3 (1989) CHI group, suggesting that level of performance on the AVLT would not reliably differentiate malingering. Nevertheless, as predicted, the shape of the AVLT recall curve was distorted by subjects attempting to fake bad. This provides a tentative but promising indication that malingering may be detected on the AVLT if the serial position effect is taken into account. It also suggests that the performance pattern how scores are obtained - rather than the pattern of scores themselves, may be a good indicator of malingering on neuropsychological tests. Bigler et al. (1989) presented evidence that different clinical groups (CHI and dementia) produce different serial recall patterns. The present study suggests that there may be other serial recall patterns that may have diagnostic utility. The M group produced a significantly reduced “primacy” versus “recency” effect; their recall percent was significantly below that of the C group for the first third of the list but it was not significantly different from the C group for the last third of the list. The shape of the M group’s curve is also different from the CHI group in the present study and the CHI group in Bigler et al. (1989). These differences suggest that subjects who attempt to fake bad do so by suppressing recall of words from the first third of the list. According to Rundus (1971), absence of the primacy effect would suggest impairment of LTM (but not STM). However, the presentation of repeated trials on the AVLT is not consistent with Rundus’ original method, so conclusions from his theory about LTM versus STM may not be applicable. Any determination of faking bad must be approached cautiously for several reasons. First, it should be kept in mind that it is not the level of performance that is being compared (not the absolute percent recalled or the percent learned over trials) for the performance of malingerers may not be much different from some CHI patients with organic memory dysfunction. Second, there should be an explanation for these patterns that is consistent with what is known about neuropsychological functioning. One tentative explanation is that the mechanism underlying the performance of patients with verified
L. C. Bernard
88
memory problems secondary to CHI is different from subjects attempting to fake bad. The “primacy effect” appeared to be evident in the mean performance of the CHI group (in fact, in the present study, proactive interference appears to have caused a marked “primacy effect”), whereas when malingering subjects were consciously suppressing recall, the “primacy effect” was reduced to a level below the “recency effect.” Finally, these results, while intriguing, need to be replicated and explored further. In the meantime, the AVLT must be used with caution where malingering is suspected. Malingering is not a simple category, for it may include some people with genuine impairment who are also exaggerating their deficits. Therefore, malingering should not be determined on the basis of performance on a single test or procedure, instead a broad profile with psychometric data as one source should be developed (Rogers, 1988). Future research may be able to further delineate what pattern deliberate faking produces on tests of neuropsychological functioning, aiding the clinician who must confront the question of differential diagnosis.
REFERENCES Bigler, E. D., Rosa, L., Schultz, F., Hall, S., & Harris, J. (1989). Rey-Auditory Verbal Learning Test and Rey-Osterrieth Complex Figure Design performance in Alzheimer’s disease and closed head injury. Journal of Clinical Psychology, 45.277-280. Brandt, J. (1988). Malingered amnesia. In R. Rogers (Ed), Clinical assessment of malingering and deception (pp. 65-83). New York: Guilford Press. Faust, D., Hart, K., & Guilmette, T. J. (1988). Pediatric malingering: The capacity of children to fake believable deficits on neuropsychological testing. Journal of Consulting ana’ Clinical Psychology,
56,578-582.
Fischler, I., Rundus, D., & Atkinson, R. C. (1970). Effects of overt rehearsal processes on free recall. Psychonomic Science, 19, 249-350. Heaton, R. K., Smith, H. H., Lehman, R. A. W., & Vogt, A. T. (1978). Prospects for faking believable deficits on neuropsychological testing. Journal of Clinical and Consulting Psychology, 46.892-900.
Lezak, M. (1985). Neuropsychological assessment (2nd ed.). New York: Oxford University Press. Mungas, D. (1983). Differential clinical sensitivity of specific parameters of the Rey AuditoryVerbal Learning Test. Journal of Consulting and Clinical Psychology, 51, 848-855. Query, W. T., & Megram, J. (1983). Age-related norms for AVLT in a male patient population. Journal of Clinical Psychology, 39.136-138. Reitan, R. M., & Wolfson, D. (1985). Neuroanatomy and neuropathology: A clinical guide for neuropsychologists. Tucson, AZ Neuropsychology Press. Rey, A. (1964). L’examen clinique en psychologie. Paris: Presses Universitaires de France. Rogers, R. (1988). Current status of clinical methods. In R. Rogers (Ed.), Clinical assessment of malingering and deception (pp. 293-308). New York: Guilford Press. Rosenberg, S. J., Ryan, J. J., & Prititera, A. (1984). Rey Auditory-Verbal Learning Test performance of patients with and without memory impairments. Journal ofclinical Psychology, 40, 785-787.
Rundus, D. (1971). Analysis of rehearsal processes in free recall. Journal of Experimental Psychology, 89,63-72.
Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariate statistics (2nd ed.). New York: Harper & Row.