Introduction
Human aggression is a complex, heterogeneous, and multifactorial construct (Dorfman et al.
2014) with multiple motives and triggers (Wahlund and Kristiansson
2009), and thus has a wide range of different manifestations in the population (Dambacher et al.
2015a; Besteher et al.
2017). It is defined as any type of hostile, injurious, or destructive behavior (Siever
2008) toward other people or living beings that causes them physical or psychological harm (Anderson and Bushman
2002; Perach-Barzilay et al.
2013). The term aggression encompasses different varieties of behaviors, cognitions, and emotions that can range from mild irritability to overtly violent behavior (Rosell and Siever
2015; Blair
2016). In addition to the conventional classification into reactive/impulsive and proactive/instrumental aggression (Flanigan and Russo
2019), Dorfman et al. (
2014) distinguish between functional and pathological forms of aggression. Functional aggression is contextually appropriate and constrained by norms and rules (e.g., American football, martial arts), whereas pathological aggression is decontextualized, unstructured, and does not conform to rules (e.g., aggressive symptoms in antisocial personality disorder (APD) or Psychopathy) (Dorfman et al.
2014).
In its extreme form, aggressive behavior can lead to serious crimes with drastic physical and/or emotional consequences for victims and impose enormous costs on society (Dambacher et al.
2015a; Zhang et al.
2019). Therefore, deeper understanding of the correlates of aggressive behavior is crucial for the development of effective interventions for prevention and targeted treatment. In recent years, structural [voxel-based morphometry (VBM), surface-based morphometry (SBM)] and functional (functional magnet resonance imaging (fMRI), resting-state fMRI) neuroimaging techniques have emerged as powerful tools to study the neural basis of aggression (Bufkin and Luttrell
2005; Raschle et al.
2015). In particular, studies of the brain morphology in aggressive populations can provide evidence for causality and offer information about mechanisms that contribute to the maintenance of aggressive behavior (Schiffer et al.
2011). These studies have pointed to morphological differences underlying aggression in several brain structures, particularly the frontal and temporal lobes (Kumari et al.
2014; Bannon et al.
2015; Dambacher et al.
2015a; Peper et al.
2015; Smith et al.
2016), which are involved in emotion processing and behavior regulation (Rosell and Siever
2015; Leutgeb et al.
2016). In particular, the prefrontal cortex (PFC), especially the orbitofrontal cortex (OFC), the medial temporal cortex, the amygdala, the basal ganglia, and the anterior cingulate cortex (ACC) show structural and/or functional changes that appear to be strongly associated with aggressive behavior (Rosell and Siever
2015; Leutgeb et al.
2016).
Most previous studies on aggressive/violent behavior have been conducted with psychiatric patients, e.g., participants with schizophrenia (Hoptman and Antonius
2011; Fjellvang et al.
2018), borderline personality disorder (Bertsch et al.
2013), bipolar disorder (Soloff et al.
2014), APD (Hoptman
2003; Wahlund and Kristiansson
2009), or incarcerated/forensic participants with Psychopathy (Koenigs
2012; Johanson et al.
2020), and youths/adolescents with disruptive behavior disorder (Baker et al.
2015). Despite the extensive database, this approach presents several methodological challenges. A high proportion of clinical or incarcerated/forensic patients have comorbidities with other mental disorders, such as schizophrenia, organic brain syndrome (Schiltz et al.
2013), anxiety disorders, or various personality disorders, and especially substance abuse (Loeber et al.
2000; Palijan et al.
2010; Ermer et al.
2013). Alcohol abuse, for example, is associated with structural changes throughout the whole brain, most pronounced in the frontal and temporal cortex (Fortier et al.
2011). Cannabis abuse can lead to structural changes in the frontal and insular cortex (Lopez-Larson et al.
2011). Therefore, the results of studies examining these populations may be poorly interpreted due to confounding variables that are difficult to exclude. In addition, the selection of well-matched control subjects is difficult and critical. In previous studies, a common control group consisted of non-criminal, mentally healthy individuals (Wahlund and Kristiansson
2009). In addition to comorbid conditions, other important variables, such as age, education, or cognitive ability, may influence brain structure. Therefore, differences between groups may result from these confounding factors (Howner et al.
2012; Koenigs
2012). Another critical issue is the heterogeneity of current studies in terms of inclusion and exclusion criteria. Often, diagnoses are not clearly differentiated: some studies examine patients with psychopathy (Howner et al.
2012), APD (Kumari et al.
2014), or both diagnoses (Bertsch et al.
2013), while others examine violent individuals in general (Hofhansel et al.
2020). Some studies examine convicted offenders (Verdejo-Román et al.
2019), sometimes with recording of psychopathic traits (Ermer et al.
2012), while others examine un-convicted subjects with psychopathic traits (Pardini et al.
2014). Moreover, in several studies, the diagnostic definition of the included individuals is not entirely clear. Therefore, it is difficult to draw general conclusions from many of the current studies due to the unclear group characterizations (Raschle et al.
2015).
In contrast to the aforementioned studies in clinical/forensic samples, which must contend with multiple confounding factors, there is limited structural neuroimaging research in adult, mentally healthy, community-based samples with no history of incarceration that represent the lower end of the symptom continuum of aggression (Garvey et al.
2016; Besteher et al.
2017; Coccaro et al.
2018). In a morphometric study, Matsuo et al. (
2009) examined the relationship between gray matter (GM) volume of the ventromedial PFC and impulsivity in healthy participants. The GM volume of the right OFC was inversely correlated with non-planning impulsivity, whereas left OFC GM volume was inversely correlated with motor impulsivity (Matsuo et al.
2009). In another morphometric study, Matthies et al. (
2012) examined the relationship between amygdala volume and lifetime aggression in healthy female participants. They report a 16–18% reduction in amygdala volume in participants with higher aggression scores. Amygdala volume is significantly inversely related to trait aggression (Matthies et al.
2012). Besteher et al. (
2017) examined brain structural correlates of irritability in mentally healthy adults using VBM and SBM. They report significant positive correlations between GM volume and aggression/hostility in large clusters involving the bilateral ACC and OFC as well as the left lingual and postcentral gyri. For SBM measures, they describe positive correlations of aggression/hostility with cortical thickness in the bilateral precentral gyri and with the gyrification index (GI) in the left insula and superior temporal gyrus (Besteher et al.
2017). Finally, Coccaro et al. (
2018) used VBM to investigate the relationship between lifetime aggression and GM volume in healthy adult same-sex twins. They describe an inverse association between medial and lateral PFC GM volume and lifetime aggression (Coccaro et al.
2018). Common to all these studies is that they are purely correlative and thus do not include a well-matched control group with no propensity for aggressive behavior. Moreover, the samples in these studies consist of healthy individuals who are within the normal range of aggression/irritability and thus within the normal range of psychological functioning (Matsuo et al.
2009; Matthies et al.
2012; Besteher et al.
2017).
Whether healthy subjects with a propensity for aggressive behavior differ in their brain structure from control subjects without a history of aggressive behavior has not yet been explored. Therefore, in a first study, we investigated GM concentration differences in 21 male healthy martial artists compared to 26 male healthy controls and their association with aggressiveness using VBM (Breitschuh et al.
2018). We describe an interaction effect between group membership and aggressiveness in a cluster comprising the left temporal pole and the left inferior temporal gyrus. In martial artists, aggressiveness is inversely related to mean GM concentration in these brain regions, whereas the opposite pattern is observed in controls (Breitschuh et al.
2018). However, there were methodological limitations of this initial study that should be noted and require further thorough investigation. First, the experimental group of martial artists was heterogeneous in terms of the type(s) of sport (self-defense vs. full-contact sport) and the combat experience. Important potentially moderating variables were not examined, e.g., physical activity, psychopathic traits, early childhood trauma, etc. Finally, we did not perform surface measurements that have been shown to increase the accuracy of brain registration (Desai et al.
2005).
The present study examines brain morphological differences relevant to aggression in non-clinical/non-forensic samples. An experimental group prone to aggressive behavior is compared to a well-matched healthy comparison group, controlling for several demographic, behavioral, and psychiatric factors that commonly co-vary with aggression. For this purpose, differences in GM volume are examined using whole-brain VBM (Ashburner and Friston
2000; Good et al.
2001). Cortical thickness and GI differences are examined using SBM (Yotter et al.
2011; Dahnke et al.
2013) to investigate finer differences in brain morphology. In this study, we therefore test the hypothesis of a relationship between aggressiveness and structural brain changes in healthy community adults with different propensities for aggressive behavior.
Discussion
The aim of the present study was to investigate the relationship between brain morphology and aggressiveness in two healthy samples, one of which was prone to aggressive behavior in the context of its recreational activities (martial arts). Using high-resolution structural MRI, we identified group differences in GM volume in two frontal and one parietal brain clusters, as well as trend findings in GI variations in one frontal brain cluster.
Martial artists showed significantly increased mean GM volumes in two frontal (mainly bilateral superior frontal gyrus, bilateral ACC) and one parietal (mainly bilateral posterior cingulate gyrus, bilateral precuneus) brain clusters compared to control subjects. These anatomical regions are particularly involved in the processing and regulation of emotions, including anger and aggression (Dalwani et al.
2011; Besteher et al.
2017; Bogerts et al.
2017; Zhang et al.
2018) and/or are relevant for theory of mind and overlap with cortical regions relevant for empathic processes (Bernhardt and Singer
2012; Kanske et al.
2015; Schmidt et al.
2017). The superior frontal gyrus, located in the upper part of the PFC, is essential for social cognition, especially perspective taking (Rogers and De Brito
2016). The ACC, also a prefrontal structure, is part of the emotion regulation circuitry (Bogerts et al.
2017). Structural changes in this limbic region have been repeatedly associated with aggressive/antisocial behavior or psychopathic traits (Koenigs
2012; Rosell and Siever
2015; Smith et al.
2016; Raine
2019). The precuneus and the posterior cingulate gyrus are parietal structures that are both part of the default mode network. They are involved in moral/social cognition and empathic aspects of proactive aggression (Westlye et al.
2017; Zhu et al.
2019). In our post hoc analysis, we report positive correlations between GM volumes in the two prefrontal clusters and several aggression subscales, specifically mapping physical/reactive aspects of aggression. Consistent with our findings, Besteher et al. (
2017) reported a positive association between GM volumes in the left superior (medial) frontal as well as left anterior cingulate gyrus and aggression in a large healthy cohort. In addition, patients with conduct disorder (CD) had significantly higher GM volumes in the left precuneus, ACC, and superior frontal gyrus as compared to controls (Zhang et al.
2018). Schiffer et al. (
2013) found that men with schizophrenia and a history of CD had larger GM volumes in the left cuneus/precuneus and inferior parietal cortex compared to men with schizophrenia without CD. Aoki et al. (
2014), in a VBM study of antisocial behavior, reported an increase in GM volume in the right cingulate gyrus but also a decrease in GM volume in the left superior frontal gyrus. Some other studies report conflicting results regarding GM volume/concentration reductions associated with aggressiveness: GM volume reductions in the right superior frontal gyrus have been associated with antisocial behavior (Hofhansel et al.
2020), whereas a negative correlation between GM volume in the left superior medial frontal gyrus and hostility was found in violent patients with schizophrenia (Liu et al.
2020a). For the precuneus/posterior cingulate gyrus area, a reduction in GM volume has been described in subjects with (APD and) psychopathy (Bertsch et al.
2013; Contreras-Rodriguez et al.
2015) and in violent patients with schizophrenia (Kuroki et al.
2017), and a reduction in GM density has been found in male violent offenders with psychopathy (Boccardi et al.
2011).
In terms of gyrification, we report a trend for increased GI levels in martial artists compared to controls in one prefrontal cluster that includes the left lateral orbital frontal cortex and the left pars orbitalis. The pars orbitalis is the orbital division of the inferior frontal gyrus, which has been described to be involved in emotional empathy and mentalizing (Bernhardt and Singer
2012; Buades-Rotger et al.
2017). Consistent with our findings, Storvestre et al. (
2019) found increased folding patterns for the left lateral orbitofrontal gyrus in schizophrenic patients with a history of violence compared to those without a history of violence. Furthermore, Schoretsanitis et al. (
2019) highlighted the role of GM volume in the left inferior frontal gyri in schizophrenic patients with a history of aggression. However, the GI trend findings did not survive correction for multiple comparisons. In contrast to our strong VBM findings and trend differences in gyrification, we did not find robust effects related to cortical thickness. Although cortical thickness is often correlated with GM measures, there is not a complete overlap (Besteher et al.
2017).
There is ample evidence in the literature for reduced GM volumes in frontal regions, such as the OFC (Gansler et al.
2009; Matsuo et al.
2009), temporal regions, such as the superior temporal cortex (Müller et al.
2008), the (para-)hippocampus (Stevens and Haney-Caron
2012), the temporal pole (Bertsch et al.
2013) and especially the amygdala (Matthies et al.
2012; Pardini et al.
2014), as well as parietal regions (Tiihonen et al.
2008) in individuals with a propensity to violence or aggressive behavior. The lack of significant associations between aggressiveness and GM changes in other associated brain regions, as well as the inverse effects in our study (positive rather than inverse association effects), was not expected. It is conceivable that these contrasting results are due to the fundamental differences between clinical/forensic and community samples. Roberts and colleagues (Roberts et al.
2021) hypothesize that changes in amygdala volume associated with aggression, in particular, are due to comorbid affective or rule-violating symptoms that are not present in a population-based sample. Our study excluded subjects with a psychiatric or neurological disorder and/or psychotropic medical treatment and is therefore not directly comparable with previous studies in clinical samples with severe psychopathology or behavioral disturbances. A critical issue of previous studies is the confounding influence of a comorbid substance use disorder on brain morphology. It is well described that substance use disorder is associated with volumetric GM reductions (Kuroki et al.
2017; Mon et al.
2014), for example, in the OFC and the ventromedial PFC (Schiffer et al.
2011), leading to complications in the interpretation of volumetric changes (Liu et al.
2020). The same is true for the influence of intelligence on the brain structure (Goriounova and Mansvelder
2019). Not all of the above studies reported or controlled for potential confounders, such as age, handedness, IQ, TIV, history of childhood maltreatment, concussion, incarceration/ hospitalization, psychopathology, use of psychotropic medication, or substance use disorders (Pardini et al.
2014; Romero-Martínez et al.
2019). Another methodological difference is that most studies conducted hypothesis-driven region of interest analyses, whereas in the present study, we used a fairly conservative whole-brain analysis approach.
Thus, an important advantage of our nonclinical approach is the absence of some of the above-mentioned confounding variables (Pawliczek et al.
2013). Furthermore, we included only males to avoid gender-specific confounding effects (Dambacher et al.
2015). Using a health questionnaire, we recorded and controlled for a history of psychotropic drug or substance use and concussions. In addition, the groups did not differ with respect to age, IQ, handedness, psychopathic traits, early childhood stress, and smoking. Related to our specific sample of martial artists, there were also no differences in BMI and physical activity between our two groups.
Despite the advantages of including a well-characterized sample and controlling for various confounding factors, there are also some potential limitations that should be considered when interpreting our results. First, our study is based on a cohort of 61 participants. With a larger cohort, we may have been able to detect smaller effects, such as differences in cortical thickness. It should be noted that in our special sample of martial artists, only 30% of the volunteers screened met our strict inclusion criteria (most common exclusion criteria: MRI contraindications, such as tattoos, inappropriate kind of martial arts). The careful selection of participants in hard martial arts (in the sense of combat sports, such as boxing and MMA) was essential for our study paradigm, as previous research has shown significant differences in hostility and aggression between this group and traditional martial arts (Kuśnierz et al.
2014; Kostorz and Sas-Nowosielski
2021). Second, as with most previous MRI studies of aggressive behavior, our cross-sectional approach does not allow us to determine whether the reported structural differences are a cause or a consequence of the observed group differences. Third, aggressiveness in our two groups was measured by self-report questionnaires. This type of instruments can be influenced by subjectivity and social desirability (Vigil-Colet et al.
2012). Nevertheless, the BPAQ in particular is a commonly used instrument in this area of research and, in addition, some of these questionnaires used also provide scales for control/openness. Finally, critical aspects in the specific case of our martial arts sample are increased physical activity and the presence of concussions. Physical exercise and motor skill learning have been associated with changes in the regional brain morphology (Schlaffke et al.
2014), e.g., in athletes, such as dancers (Wei et al.
2011), golfers (Jäncke et al.
2009), or martial artists such as judokas (Jacini et al.
2009). However, we controlled for the physical activities of our two groups and found no group differences. It is well described that cumulative head trauma can lead to structural changes in the brain (Bernick et al.
2015) and is the leading cause of a neurodegenerative disease called chronic traumatic encephalopathy (CTE; Gardner et al.
2014; Tator
2014), which is associated with impulse control problems and aggressiveness, among other neuropsychiatric sequelae (Aaronson et al.
2020). Because participation in contact sports such as martial arts has been associated with CTE (Stern et al.
2011), we collected and controlled for a history of traumatic brain injury using a health questionnaire. In addition, macroscopic neuro-radiological evidence of CTE was assessed by blinded grading by two independent clinically experienced raters. No structural abnormalities were detected, including in brain structures typically associated with repetitive head trauma, e.g., thalamus and caudate (Bernick et al.
2015). In addition, no cognitive impairments or general brain morphological differences were found between the two groups.
Overall, this study shows robust group differences in GM volumes in two community samples that differ only in their propensity for aggressive behavior. Structural brain differences were found in prefrontal (superior frontal gyrus and ACC) and parietal (posterior cingulate gyrus and precuneus) regions, which are important brain areas for the processing and regulation of emotions, such as anger and aggression, as well as empathic processes. Follow-up studies with other and larger community samples prone to aggressive behavior with longitudinal and additional functional measures are planned. In addition, the use of other imaging techniques (e.g., diffusion tensor imaging) will be of interest to determine the interplay between the regions described.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.