کنترل شناختی مکانیزم های عصبی زمینه ای مردان با رفتار ضد اجتماعی مادام العمر

Abstract Results of meta-analyses suggested subtle deficits in cognitive control among antisocial individuals. Because almost all studies focused on children with conduct problems or adult psychopaths, however, little is known about cognitive control mechanisms among the majority of persistent violent offenders who present an antisocial personality disorder (ASPD). The present study aimed to determine whether offenders with ASPD, relative to non-offenders, display dysfunction in the neural mechanisms underlying cognitive control and to assess the extent to which these dysfunctions are associated with psychopathic traits and trait impulsivity. Participants comprised 21 violent offenders and 23 non-offenders who underwent event-related functional magnetic resonance imaging while performing a non-verbal Stroop task. The offenders, relative to the non-offenders, exhibited reduced response time interference and a different pattern of conflict- and error-related activity in brain areas involved in cognitive control, attention, language, and emotion processing, that is, the anterior cingulate, dorsolateral prefrontal, superior temporal and postcentral cortices, putamen, thalamus, and amygdala. Moreover, between-group differences in behavioural and neural responses revealed associations with core features of psychopathy and attentional impulsivity. Thus, the results of the present study confirmed the hypothesis that offenders with ASPD display alterations in the neural mechanisms underlying cognitive control and that those alterations relate, at least in part, to personality characteristics.

Introduction Most violent offenders are men who have displayed a pattern of antisocial and aggressive behaviour since childhood (Hodgins, 1994 and Moffitt et al., 2002). They meet diagnostic criteria for conduct disorder (CD) before age 15 and for antisocial personality disorder (ASPD) in adulthood. Additionally, they present varying levels of the personality traits of psychopathy. Those with very high levels of these traits present the syndrome psychopathy as defined by the Psychopathy Checklist, Revised (PCL-R) (Hare, 2003). Offenders with ASPD and low levels of the personality traits of psychopathy have high levels of impulsivity (Swann et al., 2009), engage in reactive aggression, and show a hyperactive threat system (Blair, 2010), while those with high levels of psychopathic traits are less impulsive, engage in planned, premediated instrumental aggression (Glenn and Raine, 2009), and show a hypo-responsive limbic system (Blair et al., 1999). Two meta-analyses have demonstrated that impairments in executive function (EF) are associated with criminal offending (Morgan and Lilienfeld, 2000 and Ogilvie et al., 2011). The most recent meta-analysis (Ogilvie et al., 2011) showed, however, that there was considerable heterogeneity in EF across samples of offenders and suggested that the association may be due to impulsivity (Ogilvie et al., 2011). Interestingly, antisocial men with ASPD and low levels of psychopathic traits showed increased scores for attentional impulsivity (Swann et al., 2009), whereas those with high levels of psychopathic traits did not (Snowden and Gray, 2011). It is unclear how executive processes that are involved in implementing regulatory control become engaged (Cohen et al., 2000). One partial answer comes from conflict theory, which posits that monitoring of response conflict acts as a signal that engages control processes needed to overcome conflict and to perform effectively (Botvinick et al., 2001 and Van Veen and Carter, 2002b). For example, in the Stroop Colour-Word task, greater conflict is observed in incongruent (the word ‘red’ is printed in green ink) than congruent (the word ‘red’ is printed in red ink) trials (MacLeod, 1991), resulting in increased response times (RT) and error rates on the incongruent trials as compared with the congruent trials. Studies of antisocial males have been conducted using different versions of the classical Stroop test. Two meta-analyses reported low effect sizes (convergent point estimate of 0.35) for combined (RT and error) interference effects (Morgan and Lilienfeld, 2000 and Ogilvie et al., 2011), thereby suggesting subtle deficits in cognitive control. Most of the studies included in these meta-analyses, however, examined children/adolescents with aggressive/delinquent behaviour and/or CD (Wolff et al., 1982, Sullivan, 1992, Dery et al., 1999, Toupin et al., 2000, Golden and Golden, 2001, Kim et al., 2001, Olvera et al., 2005, Carroll et al., 2006 and Herba et al., 2006) and the small group of offenders, who in addition to persistent antisocial behaviour present very high levels of the personality traits of psychopathy and who thereby meet criteria for the syndrome of psychopathy as defined by the PCL-R (Gorenstein, 1982, Sutker et al., 1983, Smith et al., 1992, Hiatt et al., 2004, Blair et al., 2006, Dvorak-Bertsch et al., 2007, Gorenstein, 1982 and Pham et al., 2003). Consequently, it is not known whether impairments in cognitive control mechanisms are associated with a childhood onset of conduct problems and antisocial behaviour that persists through adulthood or with the personality traits of psychopathy. Interestingly, all studies of adult offenders with psychopathy except one (Pham et al., 2003) found that they performed similarly to non-psychopathic offenders with regard to both RT and error interference (Gorenstein, 1982, Sutker et al., 1983, Smith et al., 1992, Dolan and Anderson, 2002, Hiatt et al., 2004, Blair et al., 2006 and Dvorak-Bertsch et al., 2007). By contrast, results of studies of boys with CD, relative to typically developing boys, were inconsistent, with some studies reporting increased RT and/or error interference (Wolff et al., 1982, Dery et al., 1999, Golden and Golden, 2001 and Olvera et al., 2005), with some reporting reduced interference (Sullivan, 1992, Toupin et al., 2000, Olvera et al., 2005 and Carroll et al., 2006), and a few detecting no group differences (Kim et al., 2001 and Herba et al., 2006). Inconsistencies in results of these studies may result from the heterogeneity of samples, and/or might be due to the use of different versions of the Stroop task that vary in administration of stimuli and scoring of interference effects (for an overview, see Homack and Riccio, 2004). In particular, verbal responding might be problematic. Children with conduct problems and adult offenders show lower than average reading levels (Moffitt, 1990) and thereby may be less sensitive to written words than comparison subjects. This deficit would lead to increased RTs on a Stroop task, and might confound the measure of interference. An additional problem with the classical Stroop task concerns the frequency of stimuli in each condition. Stroop tasks usually require three types of responses, (1) reading a list of names of colours printed in black ink; (2) naming different colours; and (3) naming colours of the ink of names of colours (the word ‘red’ is printed in green ink, so the correct response is green). In contrast to computerised trial-by-trial versions commonly used in functional magnetic resonance imaging (fMRI), almost all studies of antisocial individuals used the classical Stroop task, such that neutral, congruent, and incongruent stimuli were not randomly assigned but administered in blocks. The frequency of incongruent trials, however, influences top-down mechanisms (Desimone and Duncan, 1995), and thereby affects conflict monitoring and anterior cingulate cortex (ACC) activation (Carter et al., 2000). There are a number of studies showing that the dorsal ACC is involved in the detection and monitoring of conflicts during Stroop and Stroop-like tasks (MacDonald et al., 2000, van Veen and Carter, 2002a, Ridderinkhof et al., 2004a, Nee et al., 2007 and Sohn et al., 2007). Further, the dorsal ACC (dACC) is thought to activate context representations following conflict (incongruent trials) or errors (errors being a high conflict state) in the dorsolateral prefrontal cortex (dlPFC), resulting in post-conflict adjustment, or post-error slowing (Botvinick et al., 2001, Ridderinkhof et al., 2004b and Kerns, 2006), and increased cognitive control (Carter and van Veen, 2007). When incongruent stimuli were presented frequently (or block-wise), however, individuals engaged top-down processes to overcome the pre-potent tendency to respond to the word (e.g., red) rather than the colour of the ink (e.g., green), resulting in both minimal response conflict and minimal activation in the dACC (Carter et al., 2000). To date there is only one study of antisocial individuals (Dvorak-Bertsch et al., 2007) that used a computerised trial-by-trial version of the Stroop task allowing for experimental manipulation of strategic top-down mechanisms and the analysis of behavioural adjustments. In this study, the Stroop task included mainly congruent trials in order to determine whether attentional deficits among offenders with psychopathy were specific to trials associated with ACC-mediated cognitive control. The psychopathic and non-psychopathic offenders performed equally well. However, this study, like almost all studies that tested offenders on the Stroop task, did not include a comparison group of non-offenders. Consequently, the role of the dACC in conflict monitoring among offenders, and most particularly among violent offenders, as compared with non-offenders remains unclear. The present study aimed to further characterise deficits in cognitive control presented by male violent offenders with a childhood onset and persistent pattern of antisocial behaviour. Violent offenders with ASPD were compared with non-offenders and matched for age, level of education, and lifetime histories of substance misuse, on brain activity assessed with fMRI while performing a modified non-verbal Stroop task. Since children with CD exhibit cognitive control dysfunction, and CD necessarily precedes ASPD, we hypothesised that offenders with ASPD would show increased RT and error interference, accompanied by reduced conflict- and error-related activity in the dACC and the dlPFC. Furthermore, we hypothesised that between-group differences in behavioural performance or neural responses would be explained by trait impulsivity rather than the personality traits of psychopathy.

3. Results 3.1. Performance on the Stroop task 3.1.1. Error interference A two [error type: number of errors on congruent trials, number of errors on incongruent trials] by two [group: offenders, non-offenders] ANOVA revealed a large and significant main effect for error type: F(1,41)=20.37, p<0.001] with error variance being equal for both groups (Levene test on congruent errors: F=0.99, p=0.326; and incongruent errors: F=3.08, p=0.151). There was no significant group×error type interaction (offenders: mean number of errors: congruent=1.94 (S.D.=2.7); incongruent=5.59 (S.D.=5.7); non-offenders: mean number of errors: congruent=2.04 (S.D.=2.0); incongruent=4.52 (S.D.=3.8) [F(1,41)=0.74, p=0.395]), nor a significant main effect of group on the mean number of errors [F(1,41)=0.27, p=0.608], suggesting similar error interference in both groups. 3.1.2. RT interference The two [trial type: RT on incongruent trials, RT on congruent trials] by two [group: offenders, non-offenders] ANOVA revealed a significant RT effect such that RT was shorter on congruent trials than on incongruent trials: F(1,41)=156.1, p<0.001). There was also a significant trial type×group interaction (F(1,41)=7.10, p=0.012)], indicating that offenders exhibited a significantly smaller Stroop effect (mean RT on congruent trials 659 ms (S.D.=56); mean RT incongruent trials 777 ms (S.D.=88) than did the non-offenders [RT on congruent trials 689 ms (S.D.=142); RT on incongruent trials 868 ms (S.D.=197). There was no significant group difference in overall RTs [F(1,41)=2.16, p=0.151]. 3.1.3. Behavioural adjustment effects As expected, the non-offenders exhibited robust, significant post-conflict adjustment (26 ms; t(22)=2.70, p=0.014) and post-error slowing (90 ms; t(22)=5.56, p<0.001). In contrast, offenders exhibited no significant post-conflict adjustment (2 ms; t(20)=0.20, p=0.838), although they did display significant post-error slowing (80 ms; t(20)=4.19, p<0.001). However, neither a two (previous trial type: congruent, incongruent) by two (current trial type: congruent, incongruent) by two (group: offenders, non-offenders) ANOVA, nor a two (responses after error trials, responses after correct trials) by two (group: offenders, non-offenders) ANOVA revealed significant interaction effects (p-values>0.05), suggesting that the behavioural adjustments of the offenders with ASPD and the non-offenders were similar. 3.2. Neural activity measured by fMRI 3.2.1. Conflict-related activity As presented in Table 2 and illustrated in Fig. 1, among all participants activations during incongruent minus congruent trials were significantly greater in the bilateral inferior parietal cortex (BA 7, 40), bilateral inferior and middle frontal cortex (BA 6, 8–10, 46, 47), thalamus and caudate, bilateral superior medial frontal cortex (extending into supplementary motor area and ACC; BA 8, 32), and parts of the cerebellum (extending into the middle temporal and fusiform gyrus, BA 21, 37). Table 2. Clusters of significant neural activation due to conflict- and error among all participants.a Cluster number Hemisphere Location (Brodmann areas) MNI coordinates (x, y, z) z Size (mm3) Conflict-related activity (incongruent–congruent) 1 Left Inferior parietal cortex (7,40) −36, −52, 40 6.41b 4527 2 Right Inferior/middle frontal cortex (6,8–10,46,47) 54, 18, −12 6.26b 4456 3 Right Inferior parietal cortex (7,40) 48, −50, 56 6.00b 2650 4 Left Inferior/middle frontal cortex (8,9,46,47) −34, −2, 64 5.90b 4165 5 Right/Left Cerebellum/middle temporal/fusiform gyrus (21,37) 32, −68, −30 5.86b 8947 −34, −68, −30 5.77b 6 Right/Left Caudate/thalamus −14, −6, 10 5.04b 794 20, −12, 22 4.75b 7 Left Middle/superior frontal cortex (10,11,47) −44, 50, −12 4.83b 436 8 Right/Left Superior medial frontal cortex/anterior cingulate cortex (8,32) 16, 10, 52 4.72 561 −4, 20, 48 4.63 Error-related activity (error – congruent) 1 Left Inferior frontal cortex, ext. into the superior temporal cortex and insula (13,38,47) −40, 20, −10 6.33 b 1424 2 Right/Left Superior (medial) frontal cortex / anterior cingulate cortex (6,8,9,24,32) 8, 20, 62 6.09 b 3649 −4, 26, 42 5.69b 3 Right Inferior parietal cortex 46, −52, 32 5.85b 1921 4 Right Inferior/middle frontal cortex, ext. into the superior temporal cortex (8,9,38,47) 38, 20, 48 5.63b 2643 5 Right Middle temporal cortex (21) 64, −32, −12 5.26b 380 6 Right Midbrain 6, −24, −24 4.34 567 7 Left Inferior parietal cortex (7,39,40) −64, −48, 30 4.16 993 a Activation was considered significant at p<0.001, uncorrected, for voxel-level difference and p<0.05, uncorrected, for cluster extent. b Peak voxel significant at p<0.05, corrected after Family Wise Error (FWE), at the voxel level and p<0.05, FWE corrected for cluster extent. Table options Brain regions showing significant main effects for (a) the contrast incongruent ... Fig. 1. Brain regions showing significant main effects for (a) the contrast incongruent – congruent and (b) error – congruent across all subjects and the whole brain superimposed onto a T1-weighted standard anatomic template (p<0.001, uncorrected at the voxel level; p<0.05 uncorrected at the cluster level). The significance levels (z-scores) of the activations are colour-coded as indicated in the scale bar. ⁎Numerals refer to cluster numbers in Table 2. Figure options As presented in Table 3 and Fig. 2, among the non-offenders compared with offenders, activations during incongruent minus congruent trials were greater in the left dACC (BA 24,32), and the left superior temporal cortex (including Wernicke׳s area; BA 22, 41,42), but not in the dlPFC. ROI analysis confirmed significant between-group differences in the left dACC (MNI: −6, 4, 33; z=4.22; and MNI: −10, 14, 37; z=3.19; k=225; pFWE<0.05), and not in the dlPFC. The parameter estimates indicate that the group-by-condition interaction in the left dACC was due to more pronounced activation in response to incongruent stimuli among the non-offenders. Accordingly, there was a significant positive correlation between dACC activity on incongruent minus congruent trials and RT interference (r=0.461, p<0.001). There were also a number of other regions in which activations passed height but not cluster-level thresholds: right supplementary motor area (BA 6), the pre/postcentral gyri of both hemispheres (BA 4,6,43), the left amygdala, and left inferior frontal cortex (including Broca׳s area and extending into the precentral gyrus; BA6, 9, 44). Except for one of these, all indicate greater activity among the non-offenders than offenders on incongruent as compared with congruent trials. The left amygdala was the only region which showed the opposite pattern, i.e., greater conflict-related activity in offenders relative to non-offenders. Table 3. Differences in brain activation between offenders and non-offenders.a Hemisphere Location (Brodmann areas) MNI coordinates (x, y, z) z Size (mm3) Conflict-related activity (non-offenders>offenders) Left Anterior cingulate cortex (24,32) −8, 4, 34 4.26 176 −10, 16, 38 3.22 Left Inferior frontal cortex (6,44) −52, −2, 22 3.89 b 45 Left Pre/postcentral gyrus (4,6) −44, −10, 42 3.84 b 64 Left Superior temporal cortex (22,41,42) −50, −26, 2 3.60 237 Right Pre/postcentral gyrus (43) 52, −8, 16 3.54 b 53 Right Precentral gyrus (6) 46, 6, 50 3.25 b 64 Conflict-related activity (Offenders>non-offenders) Left Amygdala −16, 0, −26 3.94 b 20 Error-related activity (non-offenders>Offenders) Left Thalamus; ventral anterior nucleus −12, −6, −8 4.50 79 Right Precentral gyrus (6) 40, −2, 50 4.47 87 Right Postcentral gyrus (43) 52, −12, 22 4.27 108 Left Postcentral gyrus (3,5) −28, −38, 62 3.89 b 59 Right Superior frontal cortex (6) 24, 0, 64 3.84 b 67 Right Amygdala/Hippocampus 30, −6, −12 3.74 b 37 Left Putamen −22, 12, 8 3.65 72 Left Anterior cingulate cortex (32) −10, 12, 34 3.55 82 Left Inferior/middle frontal cortex (46) −46, 42, 14 3.54 69 Error-related activity (offenders>non-offenders) None a Activation was considered significant at p<0.001, uncorrected, for voxel-level difference and p<0.05, uncorrected, for cluster extent. b Significant at the voxel level (p<0.001, uncorrected) but not at the cluster level. Table options Statistical parametric map (superimposed onto sections of a T1-weighted standard ... Fig. 2. Statistical parametric map (superimposed onto sections of a T1-weighted standard anatomic template) and parameter estimates (showing the size of the effect: % of global brain signal±S.D.) illustrating the significant group×condition interaction in the left dACC over congruent and incongruent trials (MNI coordinates: −8, 4, 34, and −10, 16, 38; pFWE-SVC<0.05) as derived from the whole brain analysis. The significance levels (z-scores) of the activations are colour-coded as indicated in the scale bar. Figure options 3.2.2. Error-related activity Among all participants, activations during error trials minus activations during right congruent trials were greater in the bilateral inferior frontal cortex (extending into the superior temporal gyrus and the insula on the left hemisphere; BA 13,38,47, and into the medial frontal and superior temporal gyrus on the right hemisphere; BA 8,9,38,47), superior medial frontal cortex (extending into the ACC; BA 6,8,9,24,32), bilateral inferior parietal cortex (BA 7,22,39,40), right middle temporal cortex (BA 21) and parts of the midbrain (see Fig. 1b, Table 2). As illustrated in Fig. 3, offenders with ASPD, relative to non-offenders, exhibited significantly reduced error-related activity in both the dACC and the dlPFC. Additional ROI analyses confirmed significant between-group differences for both the left dACC (MNI: −6, 15, 34; z=3.51; k=122; pFWE<0.05) and the left dlPFC (MNI: −44, 40, 15; z=3.49; k=95; pFWE<0.05). The parameter estimates illustrated in Fig. 3 show that the group-by-condition interactions in both regions were due to more pronounced activation in response to errors among non-offenders. As presented in Table 3, offenders, as compared with non-offenders, exhibited significantly reduced activity due to errors also in other regions including the left putamen, left thalamus, and the right postcentral gyrus (BA 43). Furthermore, differential activation of three regions that showed significantly reduced activation in offenders relative to non-offenders on the voxel level did not pass the significance threshold at the cluster level: the left postcentral gyrus (BA 3,5), the left superior frontal cortex (BA 6) and the right amygdala-hippocampus complex. Statistical parametric map (superimposed onto sections of a T1-weighted standard ... Fig. 3. Statistical parametric map (superimposed onto sections of a T1-weighted standard anatomic template) and parameter estimates (showing the size of the effect: % of global brain signal±S.D.) illustrating the significant group×condition interaction in (a) the left dACC (MNI coordinates: −10, 12, 34) and (b) the left dorsolateral PFC (MNI coordinates: −46, 42, 14) over congruent and error trials as derived from the whole brain analysis. The significance levels (z-scores) of the activations are colour-coded as indicated in the scale bar. Figure options 3.2.3. Impact of trait impulsivity and the personality traits of psychopathy Four regression models were calculated to assess the associations of attentional, motor, and non-planning impulsivity, and PCL factor 1 scores that assess the personality traits of psychopathy with: (1) RT interference; (2) differential activity of the left dACC due to conflict (incongruent–congruent) and (3) errors (error – congruent); and (4) differential activity of the left dlPFC due to errors (error – congruent). Scores for attentional (p=0.368), motor (p=0.567), and non-planning impulsivity (p=0.479) as well as PCL:SV factor 1 scores (p=0.703) were normally distributed across the whole sample as assessed by the Kolmogorov–Smirnov-Test. RT interference was significantly and negatively associated with scores for attentional impulsivity (beta=−0.407; p=0.013) in the whole sample, and within each group (offenders: r=−0.30, p<0.10; non-offenders: r=−0.41, p<0.05), indicating that the greater the attentional impulsivity the lower the RT interference. The same was true for the group difference in activity of the left dACC due to conflict (incongruent – congruent), which was significantly negatively associated with PCL factor 1 scores for the whole sample (beta −0.497; p=0.002) and within each group (offenders: r=−0.46, p<0.05; non-offenders: r=−0.29, p<0.10). The group difference in left dACC activity due to errors was not associated with either impulsivity or PCL factor 1 scores. Finally, the group difference in activity in the left dlPFC due to errors was significantly negatively associated with attentional impulsivity in the whole sample (beta −0.359, p=0.032), but this difference was no longer significant after taking account of multiple comparisons. Further, within each group, associations were not significant (offenders: r=−0.17, p>0.10; non-offenders: r=−0.22, p>0.10). 3.3. Post hoc analyses Given the above findings on the associations of attentional impulsivity with RT interference and group differences in activity in the left dlPFC due to error, and PCL factor 1 scores with group differences in activity in the left dACC due to conflict, ANCOVAs were calculated to determine if the group differences reported above remained after controlling for attentional impulsivity and PCL factor 1 scores. The significance of the group-by-condition interaction on RTs was reduced after controlling for attentional impulsivity but remained significant (F(1,41)=5.1, p=0.030). Similarly, the group differences in neural responses of the left dlPFC due to conflict after controlling for attentional impulsivity remained significant (x=−46, y=42, z=14; k=45; z=3.87). Group differences in the mean neural responses of the dACC due to conflict (x=−8, y=4, z=34; k=103; z=4.01) remained significant after controlling for PCL:SV factor 1 scores.