The Personality Dispositions and Resting-state Neural Correlates Associated With Aggressive Children

Qingqing Li; Mingyue Xiao; Shiqing Song; Yufei Huang; Ximei Chen; Yong Liu; Hong Chen


Soc Cogn Affect Neurosci. 2020;15(9):1004-1016. 

In This Article



One hundred thirteen primary-school students (age ran- ge 9–12 years, mean age = 10.04 ± 0.9 years, 49 girls) participated in this study. All participants were from Chaoyang Primary School in Chongqing, China. All participants were right-handed, and none reported a history of psychiatric or neurological illnesses. From the 113 participants in the neuroimaging protocol, 36 were excluded because of missing data (n = 9) or excessive head motion (n = 27). Excessive head motion was defined as a maximal displacement > 2.5 mm or a maximal rotation > 2.5 degrees throughout the course of the scan, according to recent influential methodological rsFC study (Allen et al., 2011; Liao et al., 2018). These exclusions yielded a final sample of 77 participants (mean age = 10.17 ± 0.95 years, 42 girls). The independent-sample t-test was used to assess the effect of attrition (0 = excluded, 1 = remained), and no significant differences were found neither in aggression types nor in personality traits between those remained and those excluded (the Supplementary File Table 1, Table 2 and Table 3). All participants and their parents signed the informed consent document prior to the experiment and received an honorarium at the end of the study. Ethical approval of this study was granted by the Ethics Committee of the Southwest University, and all procedures were in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki).

Trait Aggression

The 34-item form of Buss and Warren's Aggression Questionnaire (BWAQ) was administered to evaluate participants' trait aggression (Buss and Perry, 1992; Maxwell, 2007). Each item was answered on a 5-point Likert scale ranging from 1 (extremely uncharacteristic of me) to 5 (extremely characteristic of me). The questionnaire measured five constructs related to aggression: physical aggression, verbal aggression, anger, hostility and indirect aggression. The Chinese version of BWAQ has been widely utilized and is reported to have good psychometric properties (Maxwell, 2008). We used this version and conducted a confirmatory factor analysis. The overall Cronbach's α with the present sample was 0.92, and the internal consistency estimates for the 5 subscales were 0.88, 0.60, 0.71, 0.73 and 0.66, respectively.

Personality Traits

Measurement of personality traits used the Big Five Questionnaire for Children (BFQ-C) (Barbaranelli et al., 2003 2008), consisting of 65 items that measure 5 personality traits, namely extraversion, agreeableness, conscientiousness, neuroticism and openness. The Chinese version of the BFQ-C has been widely used and demonstrates adequate reliability and validity (Zhou, 2015). Participants were required to rate each statement on a 3-point Likert-type scale of 1 (disagree) to 3 (agree). The overall Cronbach's α with the present sample was 0.88, and the internal consistency estimates for extraversion, agreeableness, conscientiousness, neuroticism and openness were 0.62, 0.80, 0.72, 0.81 and 0.79, respectively, which indicated good reliability.

rsfMRI Data Acquisition and Preprocessing

Image Acquisition. For each participant, an 8-min rsfMRI scan was acquired in a 3T Trio scanner (Siemens Medical, Erlangen, Germany). During scanning, each participant was asked to remain still and relaxed, not to open his/her eyes and not to think of anything deliberately. Foam pads and earplugs were employed to reduce head motion and scanning noise. We used a gradient echo planar imaging sequence to obtain the resting-state functional image, with the following parameters: repetition time = 2000 ms, echo time = 30 ms, slices = 33, slice thickness = 3.5 mm, resolution matrix = 64 × 64, flip angle = 90°, field of view = 224 × 224 mm2, slice gap = 1 mm and voxel size = 3.5 × 3.5 × 3.5 mm3. Each section contained 180 volumes. In addition, high-resolution T1-weighted structural images were acquired from all participants using the same scanner with a 3D spoiled gradient-recalled sequence, with the following parameters: repetition time = 2530 ms, echo time = 3.48 ms, filed of view = 256 × 256 mm2, flip angle = 7°, resolution matrix = 256 × 256, slices thickness = 1 mm and voxel size = 1 × 1 × 1 mm3. The latter images provided an anatomical reference for the functional scans.

Image Data Preprocessing. The Data Processing Assistant for rsfMRI (DPARSF; Yang et al., 2016) was used to preprocess the image data in MATLAB (The Math Works, Inc., Natick, MA, USA) platform. Preprocessing was conducted in the following stages. The first six images were discarded to allow for participant familiarization and fMRI signal stabilization. The remaining 174 images were corrected for temporal shifts between slices, realigned to the middle volume and unwrapped to correct for susceptibility-by-movement interaction. Then, each fMRI images were registered to their segmented high-resolution T1-weighted anatomical images; regressing nuisance variables included six head motion parameters, white matter signal and cerebral spinal fluid signal. Next, each image was normalized to the Montreal Neurological Institute (MNI) template with a resolution voxel size of 3 × 3 × 3 mm3 and smoothing with a 4-mm full width at half maximum of Gaussian kernel. Finally, linear detrending and band-pass filtering (0.01–0.1 Hz) to discard physiological noise drift from scanner instabilities and head motion. Moreover, the frame-wise displacement (FD) was calculated as a measure of head motion and was treated as a covariate in subsequent data analysis.

Statistical Analysis

fALFF-behavior Correlation Analysis. According to the methods proposed by Zou et al. (2008), the time courses for each voxel were first converted to the frequency domain, and then, the square root of the power spectrum was computed and averaged across the specified frequency range (0.01–0.1 Hz) in each voxel. The fALFF was then computed as the sum of the amplitudes across a low-frequency range (0.01–0.1 Hz) divided by the sum of the amplitudes across the entire frequency range (0–0.25 Hz). Subsequently, the normalized fALFF was obtained by dividing the fALFF of each voxel by the global mean fALFF value. Calculations were conducted using DPARSF software (Yang et al., 2016).

To identify the brain regions showing spontaneous brain activity related to aggression, we employed whole-brain correlation analyses between the scores for the aggression constructs and the fALFF values at each voxel in the brain, with sex, age and mean FD as controlling covariates. To determine statistical significance, the results were corrected for multiple comparisons using the Gaussian random field (GRF) program, and the threshold was set to a corrected cluster P < 0.05 (single voxel P < 0.005, cluster size > 40 voxels). These analyses were conducted using the DPABI software toolbox (, version 2.3) in MATLAB.

rsFC-behavior Correlation Analysis. We performed rsFC-behavior correlation analyses to investigate whether the clusters identified through the fALFF-behavior analyses interacted with other regions to explain aggression in children. To do so, seed regions were created using the clusters with a significant relation to aggression. For each participant, we first averaged the time series of all voxels in each seed. We then performed correlation analyses between the mean time series in each seed and that of other voxels in the brain, obtaining participant-level correlation maps. For standardization purposes, the correlation maps were normalized to z maps using Fisher's r-to-z transformation. In the group-level analyses, we conducted correlation analyses between the z maps and the aggression scores to detect any association between rsFC and aggression, with age and sex as controlling variables. For multiple comparisons correction, we used the GRF program with the threshold set to a corrected cluster P < 0.05 (single voxel P < 0.005, cluster size > 100 voxels). These analyses above were performed using DPABI software.

Moderation Analysis. To determine whether the personality traits of children affected aggression through resting-state brain activity or connectivity, we carried out moderation analysis using the PROCESS macro (Hayes, 2013) in SPSS 22.0. In the moderating model, we treated the fALFF and rsFC of brain regions as the moderating variables, agreeableness and neuroticism as the independent variables, and aggression as the dependent variable. The significance of the moderating effect was assessed using a bootstrapping method with 5000 iterations. If the 95% confidence interval (CI) did not contain zero, then the moderating effect was deemed significant.