This is a case-control study investigating radicalised adolescents involved in terrorist activities in France and comparing their occurrence of psychiatric diagnosis, their empathic abilities and their suicidal tendencies with adolescents in France convicted of common crimes. The list of all adolescents convicted of AMT in France was obtained by special authorisation from the French Department of Justice in 2018. We obtained ethical authorisation to assess the AMT group from the Ethical Committee (Comité de Protection des Personnes) of Strasbourg Hospital (CPP-Est IV in January 2019) and to assess the common delinquency group from the Ethical Committee (Comité de Protection des Personnes) of the Université Aix-Marseille (n° 2016-09-11-1 on November 10, 2016). Each participant and their parents received a briefing note, and they signed a written consent form to participate in the study. The management of individual and protected data has been reported to the “Commission Nationale Informatique et Liberté” (French for “National Commission on Informatics and Liberty”).

For the group of radicalised adolescents, we kept three principal inclusion criteria: (1) prosecution for “criminal association to commit terrorism” (AMT); (2) a minority at the moment we started our study in January 2018; and (3) being able to speak, read, and write French. Concerning “radicalisation,” we wanted to choose the most objective extreme commitment criteria. Being prosecuted for “criminal association to commit terrorism” is a charge supported by evidence that confirms links with one or more individuals whose ideology preaches the use of violence. When the recruitment of adolescents prosecuted for AMT began, in January 2018, the Youth Judicial Protection Service (“Protection Judiciaire de la Jeunesse” in French or “PJJ”) informed us that 34 subjects met our inclusion criteria in French territory. Three of them were released during the investigation and were thus excluded. Sixteen refused to participate in the study. Therefore, we were able to include and collect complete data from 15 subjects. Even if it does not seem like many, it represents half of a marginal population of French adolescents involved in AMT prosecution in January 2018. We attempted to achieve representativeness in the sample studied by comparing specific objective characteristics from subjects included and from those who refused to participate in the study (e.g., age, gender, sibling composition, proportion of school years repeated, intervention of social services or care pathway). We did not find significant differences between the subjects included and those who refused to participate in the study, which allows us to assert that the sample tested for the study is representative of this population. We met the adolescents where they were detained, in jail or at the closed educational centre (CEC) (“Centre Educatif Fermé” in French) or educational units in an open setting. Consultations were always made by psychologists, child psychiatrists or clinical research associates.

Regarding the group of adolescents with a profile of “common” delinquent, the PJJ identified 18 CECs through the country to ensure sample representativeness (the list is available in the Supplementary Material). The study was proposed to all adolescents between 13 and 17 years old who were placed in these CECs by a judge as an alternative to detention for criminal acts. Five of the 18 CECs contacted did not have enough residents to participate in the study, and four were not able to do so. Therefore, recruitment was carried out in the remaining nine CECs. A total of 113 adolescents agreed to participate in the study, but complete data were not available for 12 of them. In total, we selected a group of 101 adolescents placed in CEC.

To analyse the group of radicalised adolescents, we started creating a specific grid for the study. Given the low amount of data in the scientific literature about radicalised individual characteristics in France, this grid was part of an exploratory approach. Various characteristics (e.g., sociocultural, life history, psychopathology, behaviour, speech, commitment reasons, etc.), identified in some studies on this topic (10, 14, 15) were first collected to create this grid (available in Supplementary Material). It was completed by semistructured interviews with participants and, thereafter, in collaboration with assigned educators or with additional information collected from PJJ's educational files. The grid, specific to the radicalisation phenomenon, was not completed with adolescents placed in CECs for common delinquency. So we do not present the comparative results between the two populations.

Afterwards, we evaluated several sociodemographic characteristics for both groups using a specific questionnaire that shows, for example, the family situation of the subject, his life history, and his criminal record. We also evaluated the presence of negative childhood events through the “Adverse Childhood Experience” (ACE) questionnaire. Previous research using this scale showed links between childhood maltreatment and public health problems in adults (16).

Categorical psychiatric diagnoses were mainly evaluated using MINI-KID 2 according to DSM-V and CIM-10 criteria. This semistructured interview is aimed at children and adolescents, but it does not evaluate the presence of psychotic disorders, which is why we added the subscale of the MINI for adults, which is more related to these disorders. Finally, we evaluated the presence of a borderline personality disorder (BPD) using the Ab-DIB (Abbreviated-Diagnostic Interview for Borderline) (17). For intellectual functioning, we evaluated fluid intelligence with Raven Matrices from the WAIS-IV evaluation scale.

Several psychological dimensions were explored. The severity of depression was evaluated using the Adolescent Depression Rating Scale (ADRS). The suicidal risk was evaluated using the Columbia evaluation scale for suicidal risk gravity (C-SSRS), created by Posner et al. using suicidal behaviour definitions from “The Columbia History Form.” Distress was evaluated using the Beck Hopelessness Scale (BHS). Conceived by Beck, Weissman, Lester and Trexler (18), this scale evaluates the intensity of people's pessimism and their negative perception of the future. This questionnaire has been examined among the clinical population (19–24). This scale has predictive value for suicide attempts. Indeed, a high score indicates a better correlation of suicidal intention than the severity of clinical depression. For practical reasons, this scale was not used with adolescents placed on CEC, and for this reason, we could not perform a comparison with the AMT sample.

In addition to exploring depressive and suicidal dimensions, we also evaluated several psychological dimensions that have been associated with psychopathology and/or deviant behaviours. We used the Reasons for Living Inventory for Adolescents (RFL-A). This tool was developed and approved by Osman et al. (25). It evaluates beliefs that lead adolescents to want to live and not to resort to suicidal actions. We also evaluated their sense of empathy using Bryant's child and adolescent empathy scale (26), translated into French in 2008 (27), and their emotional attachment through the inventory of parent and peer attachment (IPPA) from Armsden and Greenberg (28) [French translation by Touch and Sigel, 2006, (29)]. We also evaluated the level of impulsiveness and hostility of these adolescents using a section from the Eysenck I7 questionnaire for children and adolescents (30) and the adolescent version of the Buss-Durkee Hostility Inventory (31). Self-esteem was evaluated with the Rosenberg Self-Esteem Scale (RSES), translated into French by Vallieres and Vallerand (32). Last, coping was evaluated using the Adolescent Coping Scale (ACS). Created and approved in Australia by Frydenberg and Lewis, this scale evaluates specific behaviours used to deal with a situation or to solve a problem. ACS results allow us to evaluate the recurrence of using coping strategies, divided into productive coping strategies, coping strategies that require external support, unproductive coping strategies, and less effective strategies to deal with and solve difficulties (33).

Statistical analyses were performed using R software 4.0.2 by resorting to two-tailed tests with a level of significance fixed at 5%. First, we described the variables in AMT and CEF samples. Quantitative variable distributions were summarised using the mean, standard deviation, and min-max. Alternatively, we used the median, q1-q3 and min-max, depending on the shape of the distribution. Qualitative variable distributions were summarised using the number and percentage of occurrences. Second, we compared the variables between the AMT and CEC samples. Depending on the validity of the assumptions, quantitative variables were either compared using Welch t-tests or Wilcoxon rank-sum tests. Similarly, qualitative variables were either compared using chi-squared tests without continuity correction or Fisher's exact tests.

The AMT group was composed of six girls and nine boys, with an average age of 17 years. In terms of family environment, 5 (33.3%) of them had poor socio-economic backgrounds, 12 (80%) of them went through either a bereavement or a sudden separation during their life, seven (46.7%) of them expressed that they suffered a sort of negligence during childhood, nine (60%) of them notified social care interventions, and five (33.3%) of them confirmed that they experienced family socioeconomic insecurity. Last, four (26.7%) of them reported that household members' had been imprisoned. Academically, only five (33.3%) of these adolescents had enough difficulties to repeat a year, which is a common practise in France. Otherwise, they had unremarkable results on the fluid intelligence tests (Matrices); all had average intelligence, 11 (73.3%) adolescents were in the core and four (26.7%) in the low core. Surprisingly, given their having been prosecuted, a large majority of them expressed a desire to work in personal care jobs (childcare assistant, nurse, specialised educator or doctor).

Table 1 summarises the religious sphere, origins and migration status, and views about marginalisation or political ideology for each AMT individual and his/her family. Most adolescents were born in France mostly from migrants or mixed parents. The sample included three (20%) adolescents who converted to Islam, while the others came from families with established Muslim belief traditions. When we met these adolescents, they all assured us they were believers and Muslims. Even if none of them had faithfully practised any religion before the ideological adherence, seven (46.7%) of them affirm that they now do. Additionally, the ACS results show that resorting to spirituality is the most frequent coping strategy used to face difficulties they encounter in their everyday life. Regarding feelings of marginalisation, we distinguished whether the adolescent and/or his/her family expressed feelings of being marginalised or not, and whether they did not share similar views. We found that one third of AMT adolescents (all females) expressed no such feelings or political discourse for themselves or their families, one third had such feelings that contrasted with their own families, and one third shared these feeling with their families.

Concerning their medical history, six (40%) of them were scheduled to participate in psychological follow-up, and three of these were currently under psychiatric care. The psychiatric examinations did not show axis 1 disorders for most of them (N = 10, 66.7%). For the other five adolescents, we found comorbid associations with the following diagnoses: borderline personality disorder (BPD), generalised anxiety disorders, conduct disorder (CD) and depression (details are given in Table 2). Suicidal thoughts were observed for only two adolescents, and one of them had attempted suicide. Last, the empathy scale results are at the core of what is expected for adolescents of their age.

Last, their radical commitment was mostly developed on the internet, as only four (26.7%) adolescents had had their first physical contact with individuals professing radical ideology. Most of them mentioned one person they especially believed to be linked with their ideological adherence. Among evident reasons for their ideological adherence, we mostly observed altruistic motivations such as helping the Syrian people (14, 93.3%), reaching an ideal society (13, 86.7%), feelings of injustice or discrimination (11, 73.3%) or running away from their living environment (9, 60%). More rarely, we found attraction to armed struggle (6, 40%), a search for “adrenaline” (5, 33.3%), seeking a clearer distinction between the roles of women and men (5, 33.3%), and fear of hell or a desire to reach heaven (4, 26.7%). All these adolescents joined an ideology that advocates the use of violence in the name of Islam. Finally, 7 (46.7%) of them accomplished plans they first put into words within the ideological group (try to join Daesh: N = 4, 26.7%; unfinished knife attack, because they changed their mind at the last minute: N = 2, 13.3%; and knife attack against a police officer: N = 1, 6.7%).

Table 2 summarises the main characteristics of radicalised adolescents (AMT group) vs. “common” delinquent adolescents (CEC group). From a sociodemographic perspective, radicalised adolescents were older, and the proportion of girls was significantly higher than that in the CEC group. The loss of a biological parent, caused by divorce or abandonment, is less common in radicalised adolescents than in CEC adolescents, as well as the imprisonment of household members.

Concerning educational background, repeating a grade is significantly lower among the radicalised adolescents than among those in CECs. Qualitative results show that these repeats are often due to academic disruptions resulting from the radicalisation process. In addition, radicalised adolescents have a much higher cognitive level than CEC adolescents; 20% of the latter present fluid intelligence corresponding to a mild intellectual disability.

Psychiatric assessment results revealed more recurrent disorders in adolescents detained at a CEC, with two significant differences: (1) the lack of mental disorder was much more common in the AMT group (67% vs. 9%) and (2) there were far fewer conduct disorders (CD) in the AMT group (20% vs. 81.5%). Thus, CEC adolescents present much more CD and, in general terms, more psychiatric comorbidities than radicalised adolescents (see details in Table 1). We also noticed a particularly weak suicidal symptomatology proportion in both samples. However, the number of hospitalisations due to a crisis is slightly higher in the AMT group.

The important presence of CD in CEC adolescents is consistent with their preceding indictments, since CEC placement is given to repeat offenders whose acts partially correspond to the definition of CD. These CDs are almost non-existent in radicalised adolescents (except for the AMT indictment). This observation also seems consistent with the hostility level (BDHI) towards others, which is clearly greater in CEC adolescents, similar to the obtained results on the impulsiveness scale (Eysenck). Bryant's empathy scale results also go in this direction, as radicalised adolescents obtain much higher empathy scores than CEC adolescents. They also present more balanced attachment results, as we notice much higher levels of alienation to attachment figures among the CECF adolescents (Table 2).

Coping strategies are also very different in both groups. On the one hand, radicalised adolescents have more resources with which to face their difficulties than do CEC adolescents, as they use much more productive coping strategies. They also more often turn to spirituality, to seeking social action, and to unproductive coping strategies that require a greater capacity for insight (self-blame, worry, seek to belong) in comparison to CEC adolescents. On the other hand, adolescents in CECs willingly turn to unproductive coping strategies, in particular to release their internal stress and reduce tension through physical activity or drug/alcohol abuse.

The main limitation of this study is the disparity between AMT and CEC samples. However, it seems important to remember that this phenomenon is extremely marginal, making our AMT group a representative sample (15 of 31 adolescents). Additionally, we took care to verify that the other 16 adolescents did not present significant differences in terms of age, gender, sibling composition, the proportion of school years repeated, the intervention by social services or care pathway. Despite this representativity, a statistical comparison was affected by the disparity between the AMT and CEC groups in terms of number of subjects and age (AMT group is older), but above all, in terms of gender variability, as female subjects represent 40% of the AMT sample but only 5% of the CEF sample. Beyond the observation that the radicalisation phenomenon concerns more young girls (11) than the delinquency phenomenon (42), it seems necessary to consider it a limit for statistical comparison. Another limitation concerns the quick transformation of the radicalisation phenomenon at the time we consider these results (3, 13). Indeed, all 15 AMT prosecuted adolescents examined in this study clearly joined Daesh ideology between 2014 and 2017. Most of them were prosecuted because of conversations with Daesh members (on the internet or in person) that led some of them to try to commit violent acts in France or to travel to the Iraqi-Syrian war zone. Daesh defeat and the reinforcement of French security service operations now restrict the possibility of carrying out such acts, whether due to the drastic reduction of Daesh recruitment campaigns on the internet or to the recent impossibility of joining a territory that is less and less controlled by the jihadist organisation. These contextual elements, decisive for the evolution of radicalisation, raise questions about the reproducibility of our study.