Context The population of men who display persistent antisocial and violent behavior is heterogeneous. Callous-unemotional traits in childhood and psychopathic traits in adulthood characterize a distinct subgroup.

Objective To identify structural gray matter (GM) differences between persistent violent offenders who meet criteria for antisocial personality disorder and the syndrome of psychopathy (ASPD+P) and those meeting criteria only for ASPD (ASPD−P).

Design Cross-sectional case-control structural magnetic resonance imaging study.

Setting Inner-city probation services and neuroimaging research unit in London, England.

Participants Sixty-six men, including 17 violent offenders with ASPD+P, 27 violent offenders with ASPD−P, and 22 healthy nonoffenders participated in the study. Forensic clinicians assessed participants using the Structured Clinical Interview for DSM-IV and the Psychopathy Checklist–Revised.

Main Outcome Measures Gray matter volumes as assessed by structural magnetic resonance imaging and volumetric voxel-based morphometry analyses.

Results Offenders with ASPD+P displayed significantly reduced GM volumes bilaterally in the anterior rostral prefrontal cortex (Brodmann area 10) and temporal poles (Brodmann area 20/38) relative to offenders with ASPD−P and nonoffenders. These reductions were not attributable to substance use disorders. Offenders with ASPD−P exhibited GM volumes similar to the nonoffenders.

Conclusions Reduced GM volume within areas implicated in empathic processing, moral reasoning, and processing of prosocial emotions such as guilt and embarrassment may contribute to the profound abnormalities of social behavior observed in psychopathy. Evidence of robust structural brain differences between persistently violent men with and without psychopathy adds to the evidence that psychopathy represents a distinct phenotype. This knowledge may facilitate research into the etiology of persistent violent behavior.

Violence is a global public health problem.¹ Most violent crimes are committed by a small group of life-course-persistent male offenders, who meet diagnostic criteria for conduct disorder as children and antisocial personality disorder (ASPD) as adults.^2- 4 Behavioral genetic research suggests that such stable antisocial behavior is moderately heritable.^5,6

Significant clinical heterogeneity exists within this life-course-persistent offending group. Most are characterized by emotional lability, impulsivity, high levels of mood and anxiety disorders, and reactive aggression.^2,7- 10 However, a minority are characterized by deficient affective experience, typified by a lack of empathy and remorse, as well as persistent reactive and instrumental aggression.¹¹This subgroup meets diagnostic criteria for conduct disorder with callous-unemotional traits in childhood¹² and for the syndrome of psychopathy as defined by the Psychopathy Checklist–Revised (PCL-R)¹³ in adulthood (ASPD+P).^11,14 Men with ASPD+P begin offending earlier,¹⁵ engage in a broader range and greater density of offending behaviors,¹⁶ and respond less well to treatment programs in childhood^17,18 and adulthood¹⁹ compared with those with ASPD without psychopathy (ASPD−P). Behavioral genetic research suggests that conduct problems coupled with callous-unemotional traits are highly heritable,^20,21 while developmental studies support the stability of childhood psychopathic traits into adolescence^22,23and early adulthood.²⁴

Given the differences in patterns of antisocial and aggressive behavior, emotion processing, personality traits, and criminal offending between men with ASPD−P and men with ASPD+P, it is reasonable to hypothesize that the neurobiological mechanisms that initiate and maintain their aggressive behaviors differ. Reactive violence is purportedly underscored by dysfunction within the ventromedial prefrontal cortex (vmPFC). This region regulates emotional reactivity to perceived environmental threats or frustration in the absence of an expected reward and modulates behavior accordingly.^11,25- 27 By contrast, instrumental violence is hypothesized to be associated with abnormalities within both the vmPFC and amygdala. Deficits in aversive conditioning, reinforcement learning, and recognition of fearful facial expressions, which characterize children with conduct problems and callous-unemotional traits^28- 31 and adult psychopaths,^32- 36 are associated with dysfunction in both regions.

Evidence of differences in brain structure between offenders with ASPD−P and ASPD+P is limited.^37,38 Structural magnetic resonance imaging (sMRI) studies have typically contrasted participants with ASPD+P^39- 45 or offenders with ASPD who have not been assessed for psychopathy^46- 49 with nonoffenders. Among men with ASPD+P, those studies using whole-brain analyses have identified localized gray matter (GM) volume reductions in the frontopolar and orbitofrontal regions,^42,43 the anterior temporal cortex,⁴³ the superior temporal sulcus,^42,43 and the insula.⁴³ Studies using region-of-interest analyses have established localized GM volume reductions in the prefrontal cortex, particularly the frontopolar and orbitofrontal regions,^41,50- 53 the anterior temporal cortex,⁵² the superior temporal sulcus,⁵³ the posterior cingulate, the corpus callosum,⁴⁰ the striatum,⁴⁵ and the hippocampus.^39,54 Despite the hypothesized importance of amygdalar dysfunction among offenders with ASPD+P, only 2 studies have demonstrated reduced amygdala volumes.^44,53 Both employed targeted manual tracing techniques, while automated whole-brain methods have failed to replicate these differences. A validation study of the PCL-R demonstrated that cutoff scores for the syndrome of psychopathy varied in North America (30 out of 40) and Europe (25 out of 40).⁵⁵ In several of the structural imaging studies conducted in the United States, the criteria used to diagnose the syndrome of psychopathy were modified by lowering the cutoff score from 30 to 23.^{41,44,45,51- 53,55} The use of this lower cutoff score led to the inclusion of participants with ASPD who did not present with the syndrome of psychopathy as usually defined, making the findings difficult to interpret. Other structural imaging studies of antisocial men have either failed to report psychopathy ratings⁵⁶ or have used mixed groups of offenders with ASPD and psychopathy.^46,48,49,57

Furthermore, some studies of men with ASPD+P failed to exclude Axis I pathology such as schizophrenia spectrum diagnoses,^45,51,52,56which are independently associated with GM volume decrements in the areas of interest.^{45,51,52,56,58} Most studies have included offenders with comorbid substance use disorders (SUDs)^{39- 45,49- 53,56,57,59} that are also independently associated with GM volume decrements in regions of interest such as the orbitofrontal cortex.^48,60 These studies have attempted to control for the presence of SUDs either by including a group of control participants with similar histories of substance misuse (although significant differences between the groups have remained)^{39- 41,43- 45,51- 53,56} or by entering a measure of SUDs as a nuisance covariate in the analysis of brain structure.^{42,49,50,57,59}

Thus, the extant evidence describes a series of brain regions that may be structurally abnormal in ASPD and psychopathy, but it fails to clarify the abnormalities that characterize men with ASPD−P and those with ASPD+P. Only 2 studies have attempted a categorical distinction between small numbers of men with ASPD−P and ASPD+P to date and neither reported volumetric differences in GM.^49,57 One study used whole-brain analysis⁴⁹ and the other employed manual tracing of the hippocampus.⁵⁷ These results may represent a type 2 error owing to small sample sizes. Furthermore, the notionally ASPD−P group in these studies included participants who exceeded the PCL-R cutoff score for a diagnosis of psychopathy among Europeans,⁵⁵ which may have further minimized group differences.

Improved categorization within the broad ASPD phenotype has important implications for establishing the neurobiological underpinnings of the disorders and for developing optimal treatment approaches tailored to the distinct underlying emotional dysfunctions. Our study aimed to identify GM volume differences between violent offenders with ASPD+P and ASPD−P and a matched sample of nonoffenders using structural MRI together with fully-automated voxel-based morphometry (VBM) analyses. Voxel-based morphometry enables statistically principled between-group comparisons of GM volume, unconstrained by anatomical landmarks.⁶¹ Use of the VBM5 toolbox additionally enables robust cluster-based inferences to be made. Such automated techniques are not prone to the biases inherent in manual tracing methods, which may account, at least in part, for the excess significance bias reported in the structural imaging literature on brain volume abnormalities.⁶² The syndrome of psychopathy was assessed using the PCL-R and defined using the empirically derived European cutoff score of 25 to distinguish between violent offenders with ASPD−P and ASPD+P.⁵⁵ Participants with comorbid Axis I disorders other than SUDs were excluded. The proportions of offenders with ASPD+P and ASPD−P with past and current SUDs were similar. We studied only men because men commit most violent crimes, and most life-course-persistent offenders are male as are persons with ASPD with or without psychopathy.^3,11 Furthermore, some of the brain regions associated with ASPD and offending differ morphologically between men and women.⁵⁶ We hypothesized that the ASPD+P group would display reductions in GM volume in key frontal (vmPFC and frontal poles) and temporal (amygdala, anterior temporal cortex, and superior temporal sulcus) cortical areas relative to both the ASPD−P and healthy nonoffenders. We also hypothesized that the ASPD−P group would display GM volume decrements in the vmPFC compared with the nonoffenders. If established, such morphologic differences among offenders with ASPD would add to the mounting evidence of distinct phenotypes within this population.

The final sample included 66 men aged between 20 and 50 years, with English as a first language and a reading age of greater than 10 years. Participants had no history of neurological problems or head injury resulting in loss of consciousness for 1 hour or longer, no significant visual or hearing impairment, and no self-reported history of claustrophobia or contraindications to an MRI scan.

Violent offenders were recruited from the National Probation Service. A preliminary screening of potential participants was conducted by forensic psychologists to assess criminal, medical, and psychiatric history based on self-report and probation officer reports and files as well as to assess reading level. Offenders with a history of convictions for violent crimes (murder, rape, attempted murder, and grievous bodily harm), English as a first language, and a reading level of at least age 10 years were invited to participate in a diagnostic interview. If the interview indicated that they met DSM-IVcriteria for ASPD, with no lifetime history of major mental disorders (bipolar 1, bipolar 2, major depressive disorder, and psychotic symptomatology) or SUDs in the past month, they were invited to complete further tests. Following a PCL-R interview and a file review, offenders who obtained PCL-R scores of 25 or higher were assigned to the ASPD+P group, while those with scores of less than 25 were assigned to the ASPD−P group.

Healthy nonoffender participants were recruited using community websites and bulletin boards in local unemployment offices. None had been convicted of a criminal offense, met criteria for ASPD, or had a PCL-R score of 25 or higher.

Comparisons of the 3 groups of participants are presented in Table 1. There was a trend toward significant age differences between the 3 groups, with nonoffenders being significantly younger than the violent offenders with ASPD+P. There were significant differences in IQ and years of schooling, with nonoffenders differing from each of the offender groups but with no differences between the 2 offender groups. As intended, there were significant differences between all 3 groups on total and 4-facet PCL-R scores. None of the nonoffenders and small, but similar, proportions of the offenders with ASPD+P and ASPD−P met criteria for other personality disorders. The ASPD+P group had a significantly younger average age at first conviction for a violent offense than the offenders with ASPD−P and there was a trend suggesting that they had accumulated more convictions for violent crimes. The 3 groups differed in both total aggression scores and scores for reactive and proactive aggression. The offenders with ASPD+P obtained significantly higher scores for proactive (instrumental) aggression than the offenders with ASPD−P.

As presented in Table 2, significant differences emerged between the offenders and the nonoffenders with respect to lifetime diagnoses of abuse and/or dependence for alcohol, cannabis, and cocaine. Importantly, however, there were no significant differences between the ASPD−P and ASPD+P groups in the proportions with lifetime SUDs.

Forensic psychiatrists conducted diagnostic interviews with all the participants using the Structured Clinical Interview for DSM-IV I and II.⁶³ Both trained psychiatrists and forensic psychologists administered the PCL-R.⁶⁴ Psychopathy Checklist–Revised interviews were videotaped and a random 25% sample was rerated by a second trained psychologist. Intraclass correlation coefficient values for PCL-R total scores were acceptable (0.81). Total and scores for 4 facets were calculated.⁶⁵ All participants completed the Wechsler Adult Intelligence Scale⁶⁶ and the Reactive-Proactive Aggression Questionnaire⁶⁷ to assess levels of violent and aggressive behaviors. Criminal records for all participants were obtained from the Police National database.

This study was approved by the South London and Maudsley National Health Service and the Institute of Psychiatry research and ethics committees (reference 06/Q0706/87). Potential participants who met the inclusion and exclusion criteria were invited to take part in the study that included diagnostic interviews, an interview assessing maltreatment in childhood, completion of neuropsychological tests, and a structural and functional brain scan during a period of 4 days. After being fully informed about the study requirements and risks and having all of their questions answered, participants signed consent forms that included authorization for the research team to access their official criminal records. Participants were paid minimum hourly wage for their time, and they were strongly encouraged to desist from using substances 2 weeks prior to participation and during the period of testing. Each day on arrival at the laboratory, participants provided samples of urine and saliva, which showed that despite the request to refrain from substance use, some offenders' test results were positive for substances on the day of scanning. These individuals were assessed by the forensic psychologist and one of the authors (S.G.) to ensure that they were in a suitable state to enter the scanning environment and adhere to all safety requirements.

All participants were scanned in a 1.5T GE Signa Excite MRI scanner (General Electric Healthcare) using an 8-channel head coil for radio frequency detection at the Centre of Neuroimaging Sciences, Institute of Psychiatry. Participants were supine on the scanner bed, wearing headphones to reduce interference from scanner noise. Each participant underwent a 14-minute spoiled gradient-recalled echo scan, which is a structural scan that maximizes differences between tissue types. One hundred twenty-four slices of 1.6-mm thickness were collected with repetition time of 34 ms, echo time of 9 ms, a flip angle of 30°, and field of view of 20 cm. The acquisition matrix measured 256 × 192. Image preprocessing and analysis were carried out using Statistical Parametric Mapping software (SPM version 5.0, www.fil.ion.ucl.ac.uk/spm), running under Matlab 7.0.1 on a UNIX platform. All spoiled gradient-recalled echo images were initially checked for motion and hardware artifacts and reoriented to the intercommissural line to improve registration across the groups. Images were then analyzed using VBM for whole-brain comparisons.⁶¹ The data were processed using an iterative algorithm that incorporates tissue classification, bias correction, and normalization to achieve the model of best-fit.⁶⁸ Using the VBM5 toolbox (http://dbm.neuro.uni-jena.de/vbm/), a Markovian spatial prior was added to prior templates during segmentation. This incorporates spatial information (tissue class belonging) from neighboring voxels and enhances the accuracy of tissue classification to enable more sensitive cluster-based statistical inference. Accordingly, the VBM5 toolbox adjusts cluster size to correct for the nonstationary smoothness of the VBM data. The VBM5 toolbox also corrects for individual differences in total GM volume during the preprocessing stages. A modulation step is also inherent within the iterative process, which compensates for the effect of nonlinear spatial normalization and preserves the volume of GM within a voxel. Finally, all images were spatially smoothed using a Gaussian kernel of 8-mm full-width at half maximum to improve the signal to noise ratio and allow for inherent gyral variability across participants. This size of smoothing kernel was chosen to ensure that both the amygdala and the vmPFC could be identified.^69,70 Data for 2 offenders with ASPD−P and 3 healthy nonoffenders were excluded owing to poor image quality.

Comparisons of the 3 groups of participants on demographic and clinical variables were conducted using either 2-sample t tests or univariate analysis of variance with corresponding post hoc tests, or χ² and Fisher exact tests. One-tailed probabilities were employed to infer group differences predicted by a prior hypothesis. All analyses were carried out using the Statistical Package for Social Sciences version 15.

Group comparisons of GM volumes were performed using both voxel- and cluster-level inference within the framework of the general linear model using an absolute threshold of 0.2. Two covariates of no interest were included in the models: age, because of a difference at the trend level between the nonoffenders and offenders, and IQ, because it is associated with GM volume.⁷¹ Results from t tests modeled within an analysis of covariance design in SPM were corrected for nonstationary cluster extent using the VBM5 toolbox and assessed for significance using an initial cluster-defining threshold of z > 2.7 and a corrected cluster-significance threshold of P = .05 according to random field theory. Using the MarsBaR toolbox (http://marsbar.sourceforge.net//), volumetric values were extracted from significant regions for graphic presentation. In addition to the whole-brain analysis, we applied small volume correction in the a priori regions of interest using a threshold of P < .05 after false discovery rate⁷² correction for multiple comparisons. Based on previous findings, the WFU Pickatlas (Wake Forest University) was used to create multiple masks for small volume correction in the vmPFCs, the frontopolar cortex, the amygdala, the temporal poles, and the superior temporal gyri. Following previous examples in the literature,^73,74 sphere volumes with radii of 8 mm and 7 mm were also applied to the ventral striatum and the anterior insula, respectively.

As presented in Table 3, the offenders with ASPD+P demonstrated reduced GM volume in the bilateral anterior rostral medial prefrontal cortex, bilateral anterior temporal areas, and bilateral anterior insula when compared with the healthy nonoffenders. Frontal clusters were located medially with a peak in the anterior rostral area, extending posteriorly along the superior frontal gyrus (Brodmann are [BA] 9/10) and anteriorly toward the frontal poles. Within the temporal lobe, 2 similarly sized clusters extended from lateral temporal pole regions (BA 38) into the inferior temporal gyrus (BA 20). Bilateral clusters were also identified in the anterior insula, extending into the primary motor (BA 4) and premotor (BA 6) cortices. There were no areas where GM volumes were significantly increased in the ASPD+P group compared with nonoffenders. Region-of-interest analyses revealed no significant between-group differences in GM volume bilaterally in the vmPFC, the amygdala, the superior temporal gyrus, or the ventral striatum.

Neither VBM nor region-of-interest analyses detected areas in which GM volumes differed between the violent offenders with ASPD−P and the nonoffenders.

Among the violent offenders, those with ASPD+P compared with those with ASPD−P displayed significantly reduced GM volume bilaterally in the anterior rostral medial prefrontal and temporal pole regions. As presented in Figure 1 and Figure 2, prefrontal peak clusters were located medially, extending caudally from the superior to the medial frontal gyri (BA 9/10). As presented in Figure 3 and Figure 4, in the temporal region, both clusters extended from the temporal pole area into the inferior/medial temporal gyri. There were no areas where GM volumes were significantly increased in the ASPD+P group compared with the ASPD−P group. Region-of-interest analyses revealed no significant between-group differences in the vmPFC, the amygdala, the superior temporal gyrus, the anterior insula, or the ventral striatum.

Group comparisons were rerun after excluding participants who tested positive for drug misuse (mainly for tetrahydrocannabinol) on the scanning day, leaving 9 with ASPD+P and 19 with ASPD−P. The highly significant differences between the ASPD+P and ASPD−P groups in the bilateral anterior temporal cortex remained at the corrected level, while those in the bilateral anterior rostral medial prefrontal cortex remained significant at the uncorrected level.

The analyses were also rerun excluding participants with a comorbid personality disorder, leaving 12 with ASPD+P and 21 with ASPD−P. Again, the highly significant differences between the ASPD+P and ASPD−P groups in the bilateral anterior temporal cortex remained at the corrected level, while those in the bilateral anterior rostral medial prefrontal cortex remained significant at the uncorrected level.

To our knowledge, this study is the first to identify structural differences between violent offenders with ASPD+P and ASPD−P matched on age, IQ, and histories of SUDs, and who did not have comorbid major mental disorders. Using whole-brain analyses, our study demonstrated discrete areas of reduced GM volume in the bilateral anterior rostral medial prefrontal cortex (arMPFC) and the bilateral temporal poles among the violent offenders with ASPD+P compared with those with ASPD−P. The ASPD+P group also demonstrated GM volume reductions in the bilateral insulae as compared with nonoffenders.

Both the arMPFC and anterior temporal cortex independently feature in the assessment of current social stimuli using stored information. The medial prefrontal cortex monitors and coordinates action on complex social goals.⁷⁵The arMPFC engages in self-reflective processing,^75- 77 which not only facilitates other-person perception but also enables an emotional understanding of others' intentional acts.^78- 81 Recent computational modeling studies have demonstrated cognitive branching within the arMPFC, which enables 2 sets of competing information to be concurrently maintained, thus facilitating the simultaneous consideration of self and other perspectives.^82,83 The arMPFC is additionally engaged during fear appraisal through functionally dissociable cingulate pathways.^84- 87The temporal poles use stored social conceptual knowledge and contextual framing to facilitate understanding of social stimuli within a wider semantic and emotional context by way of multimodal sensory inputs.^88- 91 Temporal pole atrophy in such disorders as Kluver-Bucy syndrome and frontotemporal dementia^92,93 is associated with emotional blunting,⁹⁴ socially inappropriate behavior,⁹⁵ and diminished empathy.^96,97

The arMPFC and temporal poles work together to use stored knowledge representations to facilitate the understanding of emotional experience in others, while self-referential thinking and the use of conceptual knowledge extend the awareness of how we are perceived by others (our social reputation). These brain regions are thus central to the development of self-conscious emotions, such as guilt or embarrassment, which promote prosocial behavior and form the basis of moral learning.^98- 100 Both regions are routinely activated in moral reasoning tasks and in the recognition of moral violations.^101- 105 Based on this evidence, the structural volume decrements observed in these regions among the offenders with ASPD+P may contribute to the profound social impairments that characterize psychopathy. While the basic cognitive aspects of empathic processing (as assessed by first- and second-order mentalizing tasks)^106,107 appear to be intact in ASPD+P, the emotional aspects are clearly impaired, with diminished responsivity to both fear and distress in others.^31,32,35,108 Men with the syndrome of psychopathy fail to learn from their experience of punishment^29,109 and to experience self-conscious emotions such as guilt, remorse, or embarrassment,^32,51 which facilitate desistance from the use of inappropriate behaviors, most significantly aggression and violence.¹¹

In the present study, additional volume reductions were observed in the anterior insula when comparing the offenders with ASPD+P and nonoffenders. The insula acts as an interface between body-state representations and social/contextual information, facilitating the conscious experience of emotions¹¹⁰ and contributing to basic emotion processing,^111,112 decision making¹¹³ and empathic processing.^114,115 Among men with psychopathy, aversive conditioning impairments correlate with functional reductions in anterior insula activity.³² The structural differences observed in our study remained statistically significant after controlling for both IQ and histories of SUD and are consistent with findings from previous studies.⁴³ Moreover, the evidence that substance misuse can lead to GM atrophy in the insula is inconsistent.^116,117

No structural differences were observed between the offenders with ASPD+P and those with ASPD−P and the healthy nonoffenders in either the amygdala or the vmPFC. Abnormalities of amygdalar function have been observed in functional imaging investigations of affective processing among men with the syndrome of psychopathy.^32,33 However, only 2 manual tracing–based studies have demonstrated reductions in amygdalar volumes in offenders with ASPD+P.^44,53 Voxel-based morphometry is capable of detecting structural change in the amygdala, as evidenced in studies of people with schizophrenia and depression.^58,118 However, the results of our study cannot be interpreted to suggest that the amygdala is structurally intact in violent offenders with ASPD. Cortical volume within a given region is a product of 2 lower-order spatial properties—cortical thickness and surface area. Thus, an absence of differences in volume may result from marked alterations in opposing directions in these 2 dimensions (such that their product—cortical volume—does not indicate a group difference). Consequently, further investigations of cortical thickness and surface area are required. Similarly, there is limited evidence of structural abnormalities within the vmPFC among violent offenders with ASPD.^47,56 As discussed previously, these studies included participants with comorbid major mental disorders,^51,52,56 and 2 of the studies used manual tracing methods,^44,53 which may have introduced user bias. Either or both of these factors may have resulted in the observed differences in the vmPFC. Equally, multiple other factors impact vmPFC volume beyond categorical diagnosis, including, for example, the interaction between monoamine oxidase A genotype and drug misuse.¹¹⁹ Our strict exclusion of comorbid Axis I lifetime diagnoses (including significant mood disorders) may have significantly reduced the chances of observing differences in GM volume in the amygdala or vmPFC.¹²⁰ Nevertheless, abnormal function and connectivity of these areas may still occur in the context of intact structure.^36,121Consistent with this latter proposal, reduced fractional anisotropy in the right uncinate fasciculus (the primary white matter connection between vmPFC and the anterior temporal lobe) has recently been documented using diffusion tensor MRI in both psychopathic^122- 124and nonpsychopathic¹²³ men with ASPD.

Surprisingly, no significant differences in GM volume were observed in the comparisons between the offenders with ASPD-P and the nonoffenders despite differences in lifelong histories of antisocial and violent behavior, SUDs, and personality traits. Findings from our study suggest that violent offenders with ASPD−P do not present abnormalities of GM volume. Previous studies that reported such abnormalities among violent offenders⁴⁹ or men with ASPD^46,51,56may have included participants with the syndrome of psychopathy. Our study was limited to structural MRI measures of GM volumes. It is possible that alternate techniques for examining brain structure, such as cortical surface area and thickness measures and diffusion tensor MRI, will detect abnormalities in this subgroup of persistently violent men.

Several limitations of our study should be considered when interpreting the results. The study included only men. The violent offenders with ASPD who participated in the study, like most men with this disorder, had additional personality disorders and histories of SUDs.^125,126 However, the ASPD+P and ASPD−P groups were closely matched on age, IQ, and the proportions of comorbid personality disorders and SUDs. Consequently, the observed volumetric differences cannot be simply attributed to any of these factors. The strengths of the study include diagnoses made by trained clinicians, the inclusion of official criminal records, and daily objective checks during the testing period of alcohol and drug use. Furthermore, the MRI images were optimized for discerning morphological anomalies and analyzed using a fully automated, whole-brain technique.

The broader implications of our study relate to diagnostic classification systems and treatment approaches. There is currently robust evidence to suggest that life-course-persistent offending incorporates at least 2 distinct subgroups.¹²⁷ Our study has demonstrated that, in addition to divergence in the nature of aggressive behaviors, emotion processing, and personality traits, offenders with the syndrome of psychopathy display abnormalities of GM in key social brain structures. However, in contrast to the empirical literature, the DSM-5 proposes 1 antisocial personality disorder diagnosis that would merge these subgroups. For example, the criterion for pathological personality traits refers to antagonism as characterized by manipulativeness, callousness, deceitfulness, and hostility. While the 2 former traits are core characteristics of psychopathy, hostility, defined as persistent or frequent angry feelings, or irritability in response to minor slights or insults is characteristic of the larger group of persistently violent men who repeatedly engage in reactive aggression and show heightened threat responses. Failing to distinguish between these 2 subgroups may have an adverse impact on etiological research and efforts to develop intervention programs tailored to the differing behavioral and temperamental characteristics.¹²⁸ Among adult offenders, those with ASPD-P benefit from cognitive and behavioral rehabilitation programs showing reductions in criminal recidivism, while those with the syndrome of psychopathy do not.¹²⁸

In conclusion, to our knowledge, this is the first structural imaging study to compare 2 subtypes of violent offenders. All the offenders were characterized by lifelong histories of antisocial behavior, but they differed as to the personality traits of psychopathy. Furthermore, the 2 groups of violent offenders were matched on age, IQ, and lifelong SUDs. The violent offenders with ASPD+P were characterized by reduced GM volumes bilaterally in both the arMPFC and the temporal poles. These structural abnormalities may subserve the emotional dysfunctions that characterize psychopathy, underpinning deficits of empathic processing, moral reasoning, and the generation of prosocial behaviors. Taken together, the results of our study and those of previous studies of men with the syndrome of psychopathy,^43,53 as well as studies of boys with high levels of both conduct problems and callous-unemotional traits that identified abnormalities of GM concentration¹²⁹ and white matter volumes,¹³⁰suggest that psychopathy is a neurodevelopmental disorder characterized by structural abnormalities from a young age. Prospective, longitudinal investigations with repeated brain scans are required to test this hypothesis. Advancing understanding of the etiology of persistent violence and how to prevent and treat it will be facilitated by recognizing the accumulating evidence of distinct phenotypes.

Correspondence: Nigel Blackwood, MA, MD, MRCPsych, Forensic and Neurodevelopmental Sciences Department, Institute of Psychiatry, King's College London, PO Box 23, De Crespigny Park, Camberwell, SE5 8AF England (nigel.blackwood@kcl.ac.uk).

Submitted for Publication: October 10, 2011; final revision received January 6, 2012; accepted February 17, 2012.

Published Online: May 7, 2012. doi:10.1001/archgenpsychiatry.2012.222

Author Contributions: Drs Gregory and Blackwood take responsibility for the integrity of the data and the accuracy of the data analysis. All authors had full access to all the data in the study.