بررسی فراتحلیلی از رابطه بین رفتار ضد اجتماعی و اقدامات عصب روانشناختی عملکرد اجرایی

Abstract Previous narrative reviews of the relation between antisocial behavior (ASB) and neuropsychological tests of executive functioning (EF) have raised numerous methodological concerns and produced equivocal conclusions. By using meta-analytic procedures, this study attempts to remedy many of these concerns and quantifies the relation between ASB and performance on six reasonably well validated measures of EF. Thirty-nine studies yielding a total of 4,589 participants were included in the analysis. Overall, antisocial groups performed .62 standard deviations worse on EF tests than comparison groups; this effect size is in the medium to large range. Significant variation within this effect size estimate was found, some of which was accounted for by differences in the operationalizations of ASB (e.g., psychopathy vs. criminality) and measures of EF. Evidence for the specificity of EF deficits relative to deficits on other neuropsychological tasks was inconsistent. Unresolved conceptual problems regarding the association between ASB and EF tests, including the problem of localizing EF tests to specific brain regions, are discussed. THE PAST DECADE has witnessed a heightened appreciation of the role of biological influences on antisocial behavior (ASB) Lykken 1995 and Raine 1993. Among the biological factors that have been found to be associated with ASB are genetic influences, pre- and perinatal complications (Raine, Brennan, & Mednick, 1994), psychophysiological abnormalities (Raine, 1997), and differences in neurotransmitter functioning (Berman, Kavoussi, & Coccaro, 1997). One major issue that has received heightened attention in recent years is the relation between ASB and both intellectual and neuropsychological functioning. Antisocial groups score approximately 8 points lower on intelligence tests than nonantisocial groups Heilbrun 1979, Heilbrun & Heilbrun 1985 and Henry & Moffitt 1997, although the reasons for this difference are unclear. In addition, numerous authors (e.g., Elliott 1978, Gorenstein 1982 and Raine 1997) have conjectured that deficits in the brain's executive functions (EFs) are an important correlate or risk factor for ASB. Nevertheless, the relation between ASB and EF remains controversial, largely because previous studies of this association have typically yielded contradictory results (see Kandel and Freed 1982 and Lilienfeld 1992, for reviews). The purpose of this article is to quantify the relation between ASB and EF using meta-analytic methods. Specifically, we combine effect sizes from individual studies into a mean estimate of the relation between ASB and performance on EF tests. Although the ultimate goal of research in this area may be to establish causal factors for ASB, that is not the primary purpose of this article. In all of the studies we reviewed, ASB was already present at the time of assessment. Consequently, it is not possible to determine whether EF is a precursor of ASB, a sequela of ASB, or a correlate that is associated with ASB via unidentified third variables.

.

Results Analysis of EF Tests Thirty-nine studies yielding a total of 4589 participants were included in the meta-analysis. Table 1 presents a summary of the studies and effect sizes for the EF measures. The effect sizes of each measure within a study were averaged to produce a single overall effect size. As a guide for the interpretation of effect sizes, Cohen (1988) considered effect sizes of .2 standard deviations to be small, effect sizes of about .5 standard deviations to be medium, and effect sizes of .8 standard deviations or higher to be large. A stem-and-leaf plot of the distribution of these effect sizes (see Rosenthal, 1991) is displayed in Figure 1. These combined effect sizes were averaged, yielding a grand mean effect size of .57 standard deviations difference between the ASB and comparison groups. Furthermore, 79% of the effect sizes were positive. The effect sizes were then weighted by their respective sample sizes and averaged. The grand mean effect size based on the weighted effect sizes was a .62 standard deviations difference between the ASB and comparison groups. This effect size is significantly different from zero (z = 18.60; p < .001) and is in the medium to large range. Table 1. Effect Sizes for Studies Included in Meta-Analysis Reference Operationalization Assessment Method Age SD Females (n) Minorities (n) n IQ SD Measure Effect Size Appellof (1985) Delinquent Criminal records 15.4 10 30 WCST-Cat 0.466 Normal controls 16.3 10 30 WCST-Per 0.4723 Porteus-Q errs 0.449 Word Fleuncy 0.5569 Aronowitz et al. (1994) CD DSM-III-R 15 1.65 6 14 WCST-Errors 0.3915 Clinical controls 15 1.65 2 6 TMT-A 1.2204 TMT-B 1.035 WCST-Cat 0.6198 Berman & Siegel (1976) Delinquent Legal adjudication 16.1 0.75 0 14 45 87.49 11.84 Category test 0.7358 Nondelinquent 16.1 0.75 0 14 45 101.78 11.47 TMT-A 0.9931 TMT-B 0.6943 Bihrle (1995) Criminal Criminal records 19.5 0 0 81 Word Fleuncy −0.2217 Noncriminal 19.5 0 0 108 Booklet 0.1415 Category test Deckel, Hesselbrock, & Bauer (1996) Delinquent DSM-III-R 23.3 1.8 0 0 34 Porteus TA score −0.3628 Nondelinquent 22.8 1.5 0 0 57 Devonshire, Howard, & Sellars (1988) Psychopathically disordered 1959 Mental Hygiene Act (G.B.) 22 Nelson WCST-Per −0.1991 Psychiatric controls 27 Nelson WCST-Cat 0.4266 Doctor & Winder (1954) Delinquents Criminal records 15.33 0 0 60 94 Porteus-Q errs 0.9424 Normal controls 15.33 60 94 Fooks & Thomas (1957) Delinquent/psychopath Criminal records and clinical judgment 14.85 25 50 95.9 Porteus-Q errs 1.0579 Normal controls 14.95 25 50 93.65 Giancola, Mezzich, & Tarter (1997) CDa DSM-III-R 15.8 1.5 40 20 40 8.58 2.7 Porteus-Q errs 0.3383 CP/PSUDa Criminal records 16.1 1.3 119 34 119 8.76 2.82 Stroop 0.3711 Normal controls 15.7 1.3 90 21 90 9.91 2.52 Go/No-go 0.3948 Gibbens (1958) Delinquents Criminal records 15 1.65 191 Porteus-Q errs 1.0375 Normal controls 52 Gillen & Hessebrock (1992) ASPD, alcoholic DSM-III 23.39 1.69 34 106.53 11.29 TMT-A 0.098 Non-ASPD, alcoholic 22.82 1.45 0 0 57 108.18 11.51 TMT-B 0.0646 Porteus-Q errs 0.6554 COWAT 0.0153 WCST-Per 0.0623 Gillstrom (1994) Psychopathy, inmates PCL-R and criminal records 32.24 9.18 0 17 102.56 12.65 Category test 0.3649 Nonpsychopathy, inmates 31.25 8.4 28 105.39 10.76 Gorenstein (1982) Psychopath CPI-So and RDC-ASPD 26.5 7.5 20 WCST-Per 1.2544 Nonpsychopathb 29.3 8.8 23 Stroop comp time 0.6729 Normal controlsb 19 18 Hare (1984) High psychopathy, inmates Hare RSAP 30.2 8.1 0 0 14 103.4 12.2 WCST-Per 0.0733 Medium psychopathy,b inmates 30.3 7.8 0 0 16 100.8 11.1 WCST-Cat 0 Low psychopathy,b inmates 34.7 9.9 0 0 16 103.6 9.4 Hart, Forth, & Hare (1990) High psychopathy, inmates PCL-R 30.8 8.6 22 9 2.2 COWAT −0.0444 Medium psychopathy, inmatesb 41 9.1 2.5 TMT-A −0.1127 Low psychopathy, inmatesb 27 9.1 2.1 TMT-B 0.4789 High psychopathy, inmates PCL-R 29.8 7.2 32 COWAT 0.1 Medium psychopathy, inmatesb 95 TMT-A 0.0152 Low psychopathy, inmatesb 40 TMT-B −0.0308 Hurt & Naglieri(1992) Delinquents Criminal records 15.67 0.58 0 0 30 Modified Stroop 0.8134 Normal controls 15.5 0.67 0 0 30 Langley (1989) Delinquent Court referred 15.6 1.2 0 118 155 74.01 15.66 TMT-A 0.4176 Normal controls 16.1 1.1 0 24 103 108.8 17.54 TMT-B 0.426 Word Fluency-Raw 0.9023 LaPierre, Braun, & Hodgins (1995) High psychopathy, inmates PCL 33.47 8.53 0 30 Porteus Q errs 1.417 Low psychopathy, inmates 33.47 8.59 0 30 WCST-Per 0.486 WCST-Cat Lilienfeld, Hess, & Rowland (1996) Psychopathy CPI-So 18.89 2.58 61 38 101 Porteus-Q errs 0.199 Porteus TA score Lueger & Gill (1990) CD DSM-III 15 1.22 0 21 WCST-Per 0.8301 Normal controls 16.2 1.05 0 20 WCST-Cat 0.9798 TMT-B 0.348 TMT-A 0.3778 Malloy, Noel, Rogers, Longabaugh, & Beattie (1989) ASPD, alcoholic DSM-III 32.1 7.3 36 77.1 12.6 Category test 0.0069 Clinical controls 40.2 12.7 146 78.1 12.3 TMT-B 0 Moffitt & Henry (1989) Delinquent, non-ADDa Self-report and adult nomination 13 44 89 Mazes-Q score 0.2338 ADD, nondelinquentb 13 1 14 TMT-B 0.2153 Normal controlsb 13 276 549 WCST-Per 0.2449 ADD, delinquenta 13 2 21 O'Keefe (1975) Delinquents Criminal records and court ordered 11.92 1.17 40 Porteus-Q errs −0.6179 Normal controls Porteus (1942) Delinquents Criminal records and clinic referred 14+ 100 100 Porteus-Q errs 1.4001 Normal controls 14+ 100 100 Delinquents Criminal records Clinic referred 14+ 0 50 Porteus-Q errs 1.8981 Normal controls 14+ 0 31 Delinquents Criminal records Clinic referred 14+ 0 100 Porteus-Q errs 1.373 Normal controls 14+ 0 100 Inmates Criminal records Clinic referred 14+ 0 100 Porteus-Q errs 1.8751 Normal controls 14+ 0 100 Porteus (1945) Delinquents Criminal records 15 1.65 50 Porteus-Q errs 1.1698 Normal controls 25 Inmates Criminal records 15 1.65 100 Porteus-Q errs 1.4087 Normal controls 50 Rirle (1993) High psychopathy PCL-R 34.48 7.61 0 26 43 TMT-B −0.0263 Medium psychopathyb 0 23 36 TMT-A 0.1032 Low psychopathyb 0 23 46 Schalling & Rosen (1968) High psychopathy Global ratings of Cleckley psychopathy 31.42 6.77 0 0 60 106.63 10.23 Porteus-no. of trials 0.6877 Inmates Porteus-Q errs 0.6377 Low psychopathy 27.65 5.81 0 23 109.35 13.02 Inmates Smith, Arnett, & Newman (1992) High psychopathy PCL-R and criminal records 25.88 4.2 0 0 37 96.73 11.56 COWAT 0.3912 Inmates Low psychopathy 26.03 4.38 0 0 32 97.42 8.8 Stroop time 0.196 Inmates TMT-B 0.0452 TMT-A −0.0763 Sobotowicz, Evans, & Laughlin (1987) Delinquents Criminal records 0 25 50 86.62 Category Test 0.7885 Nondelinquents 0 25 50 87.28 Sullivan (1992) CD DSM-III-R 15 0 11 104.9 13.9 Stroop no. of colors read −0.6141 Clinical controls 0 12 110.4 11.8 WCST-Per 0.0276 Normal controls 10 116.6 10.2 Book Category Test 0.3162 COWAT −0.0476 WCST-Cat −0.2766 Sutker, Moan, & Swanson (1972) Psychopaths MMPI 27.3 0 43 110 Porteus-Q errs −0.5529 Antisocial psychotics 26.6 27 103.6 Prison normals 27.6 24 109.9 Sutker, Moan, & Allain (1983) Psychopaths, inmates MMPI 29.11 7.75 0 44 116.39 15.53 WCST-Per −0.2266 Nonpsychopaths, inmates 29 4.95 0 14 119.62 11.64 Williams-Timo (1989) CD DSM-III-R 16 20 Category Test −1.2691 Clinical controls 20 Porteus-Q errs 0.6021 Wolff, Waber, Bauermeister, Cohen, & Ferber (1982) Delinquents Criminal records 15 56 Stroop 0.6172 Normal controls 0 48 101.9 0 48 109.9 TMT-A 0.8527 Yeudall, Fromm-Auch, & Davies (1982) Delinquent Criminal records 14.8 35 99 95.04 TMT-B 1.3468 Normal controls 14.5 18 46 118.98 Category test 0.6661 Word Fluency 0.7113 WCST-Cat: Wisconsin Card Sorting Test-Categories; WCST-Per: Wisconsin Card Sorting Test-Perservative Errors; Porteus-Q: Porteus Mazes Qualitative; CD: conduct disorder; DSM-III: Diagnostic and Statistical Manual of Metnal Disorders, third edition; TMT-A: Trail Making Test-Form A; TMT-B: Trail Making Test-Form B; Porteus TA: Porteus Mazes mental test age; ASPD: antisocial personality disorder; PCL-R: Psychopathology Checklist-Revised; CPI-So: California Psychological Inventory-Socialization scale; ADD: attention deficit disorder; MMPI: Million Multiphasic Personality Inventory. a Groups were combined into one antisocial group for the analyses. b Groups were combined into one comparison group for the analyses. Table options Stem-and-leaf plot of combined effect sizes (d̄) for each study Figure 1. Stem-and-leaf plot of combined effect sizes (d̄) for each study Figure options A test of homogeneity was performed using the weighted effect sizes. The results were statistically significant (Q = 303.21, p < .001) and indicate that the sample of effect sizes is heterogeneous. Therefore, the common weighted effect size does not appear to derive from a single underlying population. In an effort to reduce heterogeneity, effect sizes from each study were grouped by separate operationalizations of ASB and separate tests of EF, and weighted mean effect sizes and tests of homogeneity were computed for each operationalization. Table 2 displays the mean effect size, value of Q, and reduction in the value of Q associated with each operationalization and EF measure. Table 2. Mean Effect Sizes and Tests of Homogeneity for All Studies, Grouped by the Operationalizations of Antisocial Behavior and Measures of Executive and Nonexecutive Function Operationalization d d+ Q df ΔQ Δdf ASPD .10 .08*** .43 1 302.78*** 37 CD .40 .36*** 2.28 3 300.93*** 35 Psychopathy .29 .25*** 52.68*** 14 250.53*** 24 All clinical syndromes .27 .22*** 62.56*** 19 240.65*** 19 Criminality 1.09 .94*** 80.82*** 2 222.39*** 36 Delinquency .86 .78*** 119.03*** 12 184.18*** 26 All judicial status .91 .81*** 202.14*** 15 101.07*** 23 EF measure Category Tests .24 .37*** 37.70*** 8 265.51*** 30 Mazes .80 .74*** 254.07*** 18 49.14*** 20 Stroop Test .35 .43*** 12.78*** 5 290.43*** 33 Trails B .40 .33*** 46.45*** 12 256.76*** 26 WCST-Perseverative Errors .28 .24*** 19.82* 10 283.39*** 28 Word Fluency .26 .33*** 44.45*** 8 258.76*** 30 Non-EF measure Porteus TA scores .08 .02 10.40 2 292.81*** 36 Trails A .39 .34*** 46.49*** 13 256.72*** 25 WCST-Categories Achieved .39 .37*** 9.11 6 294.10*** 32 ASPD: antisocial personality disorder; CD: conduct disorder; EF: executive function; TA: Test Age score from Porteus Mazes; Trails A: Trail Making Test, Part A; Trails B: Trail Making Test, Part B; WCST: Wisconsin Card Sorting Test. * p < .05. *** p < .001. Table options As seen in Table 2, the weighted mean effect sizes across operationalization of ASB ranged from .08 to .94, and all were significantly different from 0 (all ps < .001). When operationalization of ASB was taken into account, there was a significant reduction in heterogeneity for each group. Nevertheless, the effect sizes were still significantly heterogeneous for all operationalizations except ASPD and CD. Table 2 reveals that the mean effect sizes for criminality and delinquency were in the large range, whereas the effect sizes for CD and psychopathy were in the small to medium range. The effect sizes for ASPD, although statistically significant, were negligible. Because the number of studies in each group was too small after grouping by operationalization, further subgrouping was not conducted. Across EF measures, the unweighted mean effect sizes ranged from .24 to .80 standard deviations difference between groups, and the weighted average effect sizes ranged from .24 to .74 standard deviation difference between groups (see Table 2). All of the weighted average effect sizes were significantly different from zero (all ps < .001). The largest effect size was found for Porteus Mazes Q score, which was in the large range, whereas the effect sizes for the other EF tasks were in the small to medium range. Nevertheless, all tests of homogeneity were significant, suggesting that all studies within each grouping do not share a common population effect size. Further subgrouping of the effect sizes for each measure could not be conducted because the number of effect sizes for each group was too small. Analysis of Non-EF Tests As displayed in Table 2, Trails A produced an unweighted average effect size of .39 standard deviations difference between ASB and comparison groups, which was almost identical to the amount of difference for Trails B (i.e., .40). Moreover, Trails A was associated with a weighted effect size of .34 standard deviations difference (z = 5.26, p < .001), which was almost identical to the weighted effect size for Trails B (.33). The mean unweighted effect size for categories achieved on the WCST was .39 standard deviations difference between groups, and the mean weighted effect size was .37 standard deviation difference. Similar to the results for Trails A, this weighted effect size was significantly different from zero (z = 3.23, p < .001). Finally, contrary to the previous two tests (i.e., Trails A, categories achieved on the WCST), the mean unweighted effect size for the Porteus TA was .08 standard deviations. The mean weighted effect size was .02 standard deviations for both groups, which was not significantly different from zero (z = .162). This figure is in marked contrast to the weighted effect size for the Porteus Q score, which is in the large range (see Table 2). The File Drawer Problem A comparison of the results of published (n = 33) versus unpublished (n = 5) studies revealed no significant difference in effect size magnitude, F(1, 37) = 2.69 (p = .11). This analysis suggests a slight, but nonsignificant, tendency for published studies to yield larger effect sizes than unpublished studies. The number of studies with null results that would be needed to bring the grand mean effect size down to a value at or below .2 (i.e., a small effect size; Cohen, 1988), which was calculated using the fail-safe N statistic, was 72. Analysis of Other Potential Moderators Both within and between group analyses yielded no significant correlations between effect sizes and age, sex, ethnicity, or IQ. Thus, the magnitude of effects for EF measures could not be attributed to individual differences in any of these four variables.