Accessible Published by De Gruyter August 7, 2021

Pathways to radicalization in adolescence: The development of ideological beliefs, acceptance of violence, and extremist behavior

Thomas Bliesener, Carl Philipp Schröder and Lena Lehmann


This paper examines the link between attitudes, the acceptance of violence and the performance of extremist acts. It is tested, if different ideologies of extremism develop commonalities in their positions, attitudes and structures. Data from a school survey of more than 6,700 9th grade students from Germany serves as the empirical basis. The results show among other things that different extremist ideologies (right-wing, left-wing, Islamism) covary on the level of attitudes, acceptance of violence and extremist behavior. In a further step of analysis, the theoretically based paths of initially extremist attitudes, then an additional approval of violence and finally the commission of extremist acts was examined approximately in a person-oriented approach. Except for one path, the analyses of radicalization patterns confirm a model that commences on the cognitive level, followed by the acceptance of deviant means to reach ideological goals and finally leading to extremist acts.

1 Introduction

Among the most important developmental tasks of adolescence (Dreher & Dreher 1985; Havighurst 1972) are the development of a personal value system and ethical awareness as a guide for one's behavior. Adolescence also involves the beginning of the formation of political opinions (Beelmann 2020). In a manner that is typical of adolescents, individuals initially experiment with different political positions, which favors both the adoption of extreme political positions and short-term changes in basic attitudes (Alwin & Krosnick 1991; Rekker et al. 2015). Extreme political positions and values are also regarded as an early characteristic of individual radicalization. However, within theoretical explanations of radicalization, these are neither regarded as a sufficient condition nor the origin of radicalization processes. Models for describing and explaining the ontogenetic development of radicalization trajectories toward extremist behavior typically assume the experience of injustice to be a triggering condition. Moghaddam (2005) thus identifies the experience of relative deprivation, in the form of discrimination or unjustified disadvantage, which generates an effort to counteract the perceived injustice, as the origin of radicalization. If democratic processes do not lead to successful change, aversive emotions such as anger, frustration, and aggression emerge. This in turn leads to self-distancing from prevailing norms, values, and conventions and to a personal moral commitment to counteract the perceived injustices. Finally, this commitment is translated into behavior, and extremist or terrorist acts are committed. Similarly, Wiktorovicz’s (2005) phase model views the experience of economic, political, cultural, or personal disadvantage and discrimination as a trigger for a cognitive opening and search for new ideas and world views. The search for meaning and political and/or religious ideals paves the way to radical groups that offer solutions to such problematic situations. Through appropriate indoctrination, a change in values occurs and the seeker is ideologically socialized by the radical group (Silke 2008). Although a distinction is made between political and religious extremism, ethnic, nationalist and regional motives can also be found in the latter in addition to religious ones.

It is important to emphasize that not all those affected by disadvantage or discrimination are radicalized. The vast majority of those affected by structural disadvantages, restrictions, or discrimination do not show any radicalization toward extremist behavior and do not even indicate radicalization-related thoughts. Additionally, not everyone who adheres to radical ideas, attitudes, and values expresses this outwardly through extremist behavior (Borum 2017). In the following, extremist behaviors are understood as acts that serve a violent enforcement of normative systems, which differ from existing normative (political/religious) principles. These principles are legal (e.g., constitutional), political (e.g., free democratic order), humanitarian (e.g., general human rights), or social norms (e.g. culturally fixed standards, as well as the concept of inequality of members of different social groups). Radicalization is a developmental process towards fundamental extremist attitudes, ideas, positions or behaviors (Borum 2017; Beelmann 2019)

However, the triggers of radicalization described above are not sufficient conditions for such a development. Sageman (2004) has referred to this as a ‘problem of specificity’. Given radicalization biographies that emerged in environments of affluence, good social integration, and favorable personal prospects (e.g., some of the RAF terrorists, Osama bin Laden), the question arises as to the extent to which the aforementioned triggering conditions are prerequisites for radicalization. Instead, we may question to what extent individual experiences, predispositions, as well as social and situational triggers, which are not that dissimilar, may initiate and promote radicalization (Gill, Horgan & Deckert 2014; McGilloway, Ghosh & Bhui 2015).

With regard to their generalizability to different contexts and ideologies, the existing models have so far proven to not be particularly empirically robust (Böckler & Zick 2015). The studies available to date indicate that radicalization trajectories can be characterized by different »radicalization paths« (Herding 2013). The linearity of the phase progressions within the models, however, has not yet been sufficiently confirmed (Beelmann 2019). Nevertheless, relatively consistent findings with regard to the risk factors of the individual stages have emerged (LaFree & Schwarzenbach in this issue; see also Walther 2014).

There is also relative agreement that incipient radicalization is reflected at two levels: first, at an emotional-cognitive level, i.e., attitudes, values, ideological positions, and second, at the behavioral level, i.e., the person's visible or recognizable expressions and actions. As far as theoretical model assumptions or empirical findings on the individual-ontogenetic development are available (Simi, Sporer & Bubolz 2016), these assume an initial cognitive-emotional radicalization, in which prejudice structures (e.g., devaluation of members of a foreign group) and ideological positions (e.g., ethnocentric or religious values) are first developed. Only then does radicalization occur at the behavioral level via extremist acts (e.g., discrimination against and aggression toward foreign group members, active support for extremist groups, perpetration of politically-ideologically motivated crimes; see Borum 2011; Dzhekova et al. 2017; McCauley & Moskalenko 2008). Although this assumed developmental sequence from attitude to behavior seems relatively plausible and can also be justified in general theoretical terms (Fishbein & Ajzen 1975), the available empirical tests of this developmental sequence are based on retrospective surveys of already radicalized individuals who were interviewed about their entry into an extremist scene (e.g., Bjørgo 2011; Schils & Verhage 2017). However, especially for the field of norm-defying behavior, reverse developmental trajectories can also be justified and demonstrated. According to Sykes and Matza (1957), arguments for legitimation and justification are acquired within group social interactions, which neutralize negative effects following the occurrence of a norm violation and thereby contribute to the maintenance of deviant behavior. Therefore, the subsequent acquisition of an extremist pattern of justification (cognitive level) following an antisocial act or extremist offense (behavioral level), which was initially exhibited for other motives, would be conceivable (Ribeaud & Eisner 2010). Other motives, particularly in the area of juvenile delinquency, are peer influence (encouragement by the group, prospect of gaining status, imitation effects) and misinterpretation of the statements of others as provocation or threat (Crick & Dodge 1994; Lösel, Bliesener & Bender 2007).

The previous remarks show that longitudinal studies, with regard to individual (ontogenetic) developmental trajectories, are desirable. However, these are still rare within the field of radicalization research. Thus, some biographical and cross-sectional studies on risk and protective factors of radicalization and extremism exist to date (e.g. Borum 2014; Emmelkamp et al. 2020; Lösel et al. 2018; Nivette, Eisner & Ribeaud 2017; Wolfowicz et al. 2019). Nevertheless, these hardly allow robust conclusions to be drawn about developmental trajectories. Whilst previous studies of individual biographies, using mostly retrospective observation of the development of radicalized individuals, examine the course of development, they do not allow for any reliable conclusions regarding the specificity of individually identified characteristics or processes for the course of radicalization to be drawn. Albeit also cross-sectional in nature, the study presented below attempts to specify and analyze different individual trajectories of incipient radicalization among young people. In doing so, the question of how the process of radicalization proceeds on an individual-ontogenetic level will be examined. It is assumed that a radicalization process is often initiated but only rarely leads to extremist and seriously violent behavior. As in the staircase model of Moghaddam (2005) or in the pyramid model of McCauley and Moskalenko (2008; 2011; 2017), the number of radicalized individuals at each level should decrease more or less steadily and the proportion of radicalized individuals should decrease substantially (Wikström & Bouhana 2016). In addition, a model test, which examines the occurrence of individual development trajectories that are consistent or incompatible with the theoretical model concepts, is carried out.

2 Procedure

2.1 Sample

The following analysis is based on data from an online survey of 9th grade students at general education schools in 11 of 16 German federal states in which the state school authorities approved the survey. The sample design included oversampling of cities with suspected high extremist potential in order to reach a sufficiently large number of relevant participants, as the original goal of the project was, amongst others, to analyze Islamist radicalization. Locations were selected e.g., from reports by the Federal Office for the Protection of the Constitution, as well as police and media reports with regard to their Islamism, right-wing and left-wing extremism problems. Here, we assumed a higher prevalence of extremist ideological beliefs. In the selected cities, all schools except Special Educational Needs and vocational schools were contacted. Rural, as well as urban, towns and cities were included in the sampling process. A total of 1,091 general education schools in 55 German cities were contacted and asked to participate in the study. 208 schools (19.1 %) agreed to participate, and 6,863 students from 432 9th grade classes took part in the study. This corresponds to a participation rate of 65.0 % of the students in the participating school classes. Prior to subsequent analysis, data cleaning was performed. To remove cases with high proportions of omissions due to fast clicking, an index for the average processing time of a questionnaire page was used (Leiner 2016). Implausible responses or noticeable response patterns were checked manually. Following this, 6,715 cases remained in the data pool. Of these participants, 90 % finished the questionnaire. The share of missing values due to item-nonresponse or dropout for the relevant items are between 6.1 % and 27.4 %. These numbers indicate a possible premature end of the survey due to a lack of time or decreasing motivation or ability to participate in the course of the survey. This may possibly have increased error variance due to carelessness in responding. The completion rate was checked for each page of the survey. No conspicuous rates could be found for any page. This indicates no systematic premature ending due to single items. Missing data were not imputed. The proportion of female students in the sample was 52.6 %. 28.2 % of the respondents indicated an affiliation with the Protestant Church, 18.9 % with the Catholic Church, and 3.4 % with other Christian demoninations. The percentage of students with a Muslim religious affiliation was 14.8 %, other religious affiliation was 3.4 %, and 31.3 % indicated no denominational affiliation. 57.0 % attended a Gymnasium (students reach A-Level standard), 24.2 % a Gesamtschule (comprehensive school), 18.8 % a Realschule (intermediate school), Integrierte Haupt- und Realschule or Oberschule (combination of lower secondary and intermediate schools).

2.2 Method

The survey was conducted anonymously and usually in class; students completed the survey individually on a PC or a suitable mobile device. The survey was led by trained test supervisors. Prior to the survey, the voluntary nature of participation and the possibility of skipping questions was pointed out. It was also pointed out that no disadvantages would result from non-participation or from non-completion of individual questions. In addition to information on sociodemographic data, leisure time behaviors, as well as family and friends, the online questionnaire contained questions on the experience of one's own identity and position in a group and in society. The questionnaire also contained questions on experiences of exclusion and victimization, personal radical attitudes and values, the willingness to use violence, as well as one's own extremist behaviors in relation to the ideologies of right-wing extremism, left-wing extremism, and Islamist-Jihadist extremism. The constructs relevant for this analysis were recorded as follows:

Ideological attitudes: Ideological attitudes with a right-wing orientation were collected using 14 items. The items stem from different scales used to record the approval of a dictatorship (»We should have a leader who rules Germany with a strong hand for the good of everyone«), trivialization of National Socialism (»National Socialism also had its good sides«), xenophobia (»There are too many foreigners living in Germany«), antisemitism (»Jews have too much influence in the world«), chauvinism (»We should finally have the courage to hold a strong national feeling again«), and social Darwinism (»There is valuable and unworthy life«) (Decker & Brähler 2006; 2020; Heitmeyer 2002; Küpper, Krause & Zick 2019; Maresch & Bliesener 2015). For the collection of ideological attitudes with a left-wing orientation, five items referring to anti-capitalism (»Nowadays people are exploited by the rich and powerful«), anti-fascism (»Right-wing parties and comradeships should be banned«), anti-repression (»The police make life difficult for us instead of helping us«), and abolition of the state (»We can only be truly free once the entire state is abolished«), which also stem from different studies, were used (Baier & Pfeiffer 2011; Bergmann et al. 2017; Deutz-Schroeder & Schroeder 2016). Ideological attitudes with an Islamist orientation were collected using seven items: consent relating to the IS (»It is a good thing when people leave for Syria to join ISIS«), religious ideology (»The Islamic Laws of the Sharia are much better than the German laws«), or Western discrimination of Islam (»Muslims are oppressed in the Western world«). These items also stem from previous studies (Frindte et al. 2011; Heitmeyer, Müller & Schröder 1997; Brettfeld & Wetzels 2007).

Acceptance of violence: Established items or scales were also used to measure the acceptance of violence. Constructs developed by the authors were also sometimes used, such as for measuring right-wing violence, which was captured using six items, (e.g., »Violence must be used to show refugees that they are not welcome«) (see Goede, Schröder & Lehmann 2020). The endorsement of left-wing violence was also captured with the use of six items (e.g., »It is right to burn the luxury cars of the rich«) (see Bergmann et al. 2017). The endorsement of Islamic violence was collected using eight items (e.g., »The soldiers of the IS are not terrorists but freedom fighters«) (see Goede et al. 2020), which have already been proven to be effective.[1]

General affinity for violence: In addition to the ideologically influenced attitudes, the adolescents’ general affinity for violence was also recorded using four items (e.g., »Violence is simply part of having a little fun«, »If I have to show what I'm made of, I would also use violence«) (Bergmann et al. 2017).

All items on extremist attitudes, as well as on ideological and general approval of violence, could be answered on a 5-point Likert scale (1 = »strongly disagree« to 5 = »strongly agree«). The items were aggregated by averaging. The scale parameters are shown in Table 1.

Table 1:

Scale and item statistics.

# of items Mean Standard


Right-wing attitudes 14 2.15 .78 .90
Acceptance of right-wing violence 6 1.78 .79 .86
Right-wing behavior 4 10.8 %b
Left-wing attitudes 5 2.26 .73 .73
Acceptance of left-wing violence 6 1.87 .72 .81
Left-wing behavior 5 9.1 %b
Islamist attitudesa 7 2.89 .98 .85
Acceptance of Islamist violencea 8 1.74 .76 .84
Islamist behaviora 3 8.9 %b
General propensity to violence 4 1.68 .85 .90
General delinquency 9 1.35 1.55 .74

a Only respondents with Islamic religious affiliation.

b Rate of respondents that have committed at least one extremist act from the respective ideological area.

Extremist behavior: Regarding the implementation at the behavioral level, four different right-wing behaviors (e.g. »I have painted or sprayed swastikas or a slogan like ›foreigners out‹ on a wall or a public lavatory«), five leftwing behaviors (e.g. »I have used bottles, stones, or similar items to act against the police at a demonstration«) and three Islamist behaviors (e.g. »I have insulted, sworn at, or said mean words to somebody because they did not belong to my religion«) were surveyed (Bergmann et al. 2017; Goede et al. 2020). The data on extremist behaviors were then dichotomized (no behavior vs. at least one behavior committed so far). Table 1 shows that between 8.9 and 10.8 % of the respondents have committed at least one extremist act from the respective ideological area.[2]

In all three ideologies, verbal attacks on members of rejected groups outweigh extremist behaviour, with 9.1 % of right-wing attacks against foreigners, 9.1 % of left-wing attacks against right-wingers and 5.2 % of Islamist attacks against non-believers. Physical attacks against foreigners (2.0 %), against police officers (1.6 %) or against people who are not of the Islamic faith (1.0 %) are much less frequent.

General delinquency: Finally, criminal behaviors without an ideological component were also surveyed. The prevalence rates for nine different delinquent acts (e.g., property damage, shoplifting, burglary, fare evasion, drug dealing) were dichotomized (never committed vs. committed at least once) and added up to a sum index (see Tab. 1).

3 Results

3.1 Interrelationships of radicalization characteristics

Table 2 shows the intercorrelations of each of the three attitudes, acceptance, and behavior scales.[3] First, it is evident that almost all correlations are positive. The highest correlations occured within the individual ideologies. Ideological attitudes and acceptance of violence correlated very highly in the right-wing domain (.71), but also showed a clear correspondence in the left-wing (.52) and Islamist (.52) domains. This is surprising since a thematization of violence had been excluded in the selection of the attitude items. The correlations between the acceptance of violence and the engagement in extremist behavior were also significant but turned out to be lower (right-wing: .35, left-wing: .21, Islamist: .16).

However, it is also noticeable that the three dimensions of attitudes, approval of violence, and behavior correlated positively across the three ideologies. On the attitudinal level, right-wing and left-wing political positions corresponded to .29, left-wing and Islamist attitudes even to .35. The correlations across the ideologies were even higher in the approval of violence dimension (r = .38 to .46). At the behavioral level, on the other hand, the correlations between ideological influences were somewhat weaker (r = .09 to .24), yet also positive. This means that ideological attitudes, the ideologically influenced acceptance of violence, and ideologically influenced behaviors were by no means mutually exclusive between the ideologies. In fact, they also corresponded across the different ideologies, and in some cases this correspondence was considerable.

Table 2:

Intercorrelations of scales on attitudes, acceptance of violence, and extremist behavior

Attitudes Acceptance of violence Behavior
– Right-wing
– Left-wing .29***
– Islamist .22*** .35***
Acceptance of violence
– Right-wing .71*** .33*** .25***
– Left-wing .29*** .52*** .33*** .41***
– Islamist .40*** .29*** .52*** .46*** .38***
– Right-wing .34*** .14*** .09*** .35*** .12*** .11***
– Left-wing .13*** .18*** .04ns .18*** .21*** .14*** .24***
– Islamist .14*** .11*** .10*** .23*** .21*** .16*** .23*** .09***
General propensity to violence .39*** .32*** .23*** .48*** .39*** .28*** .28*** .28*** .23***
General delinquency .05*** .17*** -.02ns .06*** .11*** -.02ns .23*** .33*** .08**

a R-W: Right-Wing; L-W: Left-Wing; ISL: Islamist

Note: N = 6715, except for correlations with Islamist attitudes and behaviors (n = 975). *** p≤ .001; ** p ≤ .01; ns p ≥ .05.

The correlations with the general affinity for violence were also relatively high. Here, the ideological attitudes correlated with .23 to .39. The correlations with the ideologically influenced acceptance of violence were each slightly higher. The correlations with extremist behaviors were again somewhat lower (.23 to .28).

In the case of extremist attitudes and approval of violence, the correlations with the behavioral level of general delinquency were low or disappeared completely.[4] Substantial correlations were, however, found within the extremist behavioral level. General delinquency corresponded substantially with extremist behaviors in the extreme right and left (.23 and .33, respectively), but less so in the Islamist domain (.08).

3.2 Pathways to radicalization according to the model

Beyond the correlations between radicalization levels and ideologies, the following person-centered approach attempted to verify the sequence of radicalization characteristics according to the theoretical assumptions. For this purpose, the attitude and violence acceptance variables were also initially dichotomized.[5]

The models presented above (Moghaddam 2005; Wiktorovicz 2005) assume that radicalization begins at the cognitive level with the adoption of ideologies or problematic attitudes and, via the endorsement of violence to achieve extremist goals, ultimately leads to one's own extremist behaviors. Developmental paths toward extremist behavior would thus be dependent upon a previously displayed extremist attitude and approval of violence (see Fig. 1). In line with the model, it would also be possible for a radicalization process to be interrupted and only lead to the formation of a radical attitude along with the possibility of subsequent approval of extremist violence, or for the process to be terminated after the initial development of a radical attitude. Finally, a course of development ‑ free from any radicalization processes ‑ without any problematic attitudes toward radical ideas, without approval of violence and without corresponding behavior is also possible. In contrast, developmental paths that do not conform to the model would include those in which extremist behavior is exhibited without prior attitude development and/or without prior approval of violence or in which isolated approval of violence occurs. Figure 1 shows these model-compliant (solid lines) and non-model-compliant (dashed lines) developmental paths.

Figure 1 
Paths to radicalization according to the model and coding scheme. Solid, bold lines show developmental paths conform to theoretical models, dashed lines theoretically non-conform paths.

Figure 1

Paths to radicalization according to the model and coding scheme. Solid, bold lines show developmental paths conform to theoretical models, dashed lines theoretically non-conform paths.

The frequencies of occurrence regarding the combinations of characteristics resulting from the different paths are shown in Table 3. In addition to the descriptive analysis, configural frequency analyses (CFA see Lienert 1969; von Eye 1990) were conducted. The CFA is a multivariate method for typological research to detect patterns in the data that occur significantly more or less often than expected by chance (Schrepp 2006). When a combination of characteristics in a cross-classification contains more observations than expected, this combination constitutes a CFA type. When there are fewer cases than expected, this cell is said to constitute a CFA antitype (von Eye 2002). Such a person-oriented approach has already proven successful in analyzing the developmental trajectories of problematic behavior in adolescence (see Lösel & Stemmler, 2012; Stemmler & Lösel, 2012).

Table 3:

Observed frequencies (fo), expected frequencies (fe), and CFA types (T) resp. antitypes (A) of paths to radicalization.




Patha fo fe Type/

fo fe Type/

fo fe Type/

111 301 39.1 T 217 40.2 T 20 5.6 T
110 648 322.8 T 760 386.1 T 99 56.9 T
101 188 174.4 132 140.1 19 21.2
100 840 1440.8 A 802 1344.6 A 162 216.3 A
011 39 93.6 A 65 101.71 A 11 12.5
010 241 773.5 A 462 976.0 A 73 128.0 A
001 197 417.9 A 222 354.1 A 37 47.7
000 4261 3452.9 T 4082 3399.2 T 554 486.8 T
Missclass.: 9.9 % 13.1 % 14.4 %
Total N 6715 6742 975

a Path coding: Attitudes/acceptance of violence/behaviors (0=no, 1=yes). Theoretically conform paths bold.

b Each test was performed with 1st order CFA. Lehmacher test with Bonferoni adjusted alpha (.00625) was used. For all types/antitypes p ≤ .001

Table 3 lists the observed and the expected frequencies of occurrence of the characteristic combinations resulting from the possible path courses. In the table, the combinations that did conform to the model are shown in bold and highlighted in grey, while those that did not conform to the model are shown in normal font. Table 3 first shows that the combination ‘000’, which marks the absence of a problematic attitude, an endorsement of violence, and any extremist acts, greatly predominated in the observed frequency in each case. This means that the vast majority of students surveyed did not show any radicalization characteristics. However, the observed frequency in each case also outweighed the expected frequency that would stochastically occur if the three characteristics were independent, meaning that the combination formed a CFA type in each case. Although significantly less frequently represented, the two combinations ‘111’ and ‘110’, which mark the presence of a problematic attitude and an approval of violence with and without extremist behaviors, respectively, also formed one type each. In contrast, the non-model-compliant paths ‘001’, ‘010’, ‘011’, and ‘101’ each appeared less frequently. ‘001’, ‘010’, and ‘011’ were classified as antitypes in the domains of right-wing and left-wing radicalization domain. The non-model-compliant path ‘101’, on the other hand, occurs with a frequency that corresponds to the stochastic expected value and thus consistently forms neither a type nor an antitype. In the Islamist radicalization domain only ‘010’ was classified as an antitype. However, with reference to the latter case, the smaller sample size may contribute to the fact that the deviations from the expected value did not reach significance here.

A noticeable exception to the pattern determining which model-compliant paths or combinations represent types, however, can be seen in the combination ‘100’. This combination only included a problematic attitude, yet without further radicalization characteristics. Contrary to expectations, this model-compliant path also formed an antitype within every ideological orientation, i.e., it occured less frequently than would be stochastically expected. Despite this unexpected finding, the overall agreement with the model assumptions was relatively good. When calculating the proportion of cell occupations that did not conform to the model, relatively homogeneous rates of only 9.9 % to 14.4 % of misclassifications for the three radicalization sub-categories were found.

4 Discussion

The present analysis is based on a survey of more than 6,700 9th grade students at general education schools in selected cities and regions in Germany where higher levels of radicalization are known or suspected by authorities or the media. Based on this background alone, the reported rates of students who have indicated the perpetration of radicalized behaviors should not be viewed as generalizable prevalence estimates. In addition, instruments that can capture specific yet comparable trait expressions across the different ideologies are not yet available.[6] Although this study also relies on established scales and items for recording data, the issue of alignment with different content of the attitudes, acceptances, and behaviors surveyed, meaning that the items used exhibit psychometrically different item difficulties, remains. The present strategy of creating a common definition of radicalization, which encompasses all ideologies studied, appears to be a step in the right direction. However, it seems urgently necessary for future studies to address the measurement-theoretical problems using sufficiently large representative samples to enable the development of equivalent measurement scales and thus the determination of a comparable severity of radicalization. For the present analysis, however, selective sampling has had the desired effect of allowing the study of developmental trajectories with advanced radicalization in noteworthy numbers.

The findings reveal somewhat unexpected notable correspondences between the individual radicalization characteristics. Not only did the respective ideologically problematic attitudes correlate positively with one another, but the approval of violence to achieve ideologically determined goals as well as the perpetration of ideologically influenced behaviors also covary across the radicalization sub-categories. This contradicts the assumption that ideologically distinct groups, in which one's own positions are affirmed but contrary ideologies are strictly rejected, form among the young people surveyed. The available data show that young people with right-wing or left-wing extremist attitudes tend to have similar attitudes and goals.

An explanation for this overlap could lie in general response tendencies towards the extremes or also towards the middle. In order to test the possibility of artificial correlations, we separated the respective initial distributions of attitudes or approval of violence at the quartiles and tested the correlations once only for the outer two quartiles (to test tendencies towards extreme answers), and another time only for the two middle quartiles (to test answer tendencies towards the middle). In all cases, positive correlations were found here as well, so that the influence of general response tendencies can be excluded.

Such similarities may also result from commonalities in individual goals, which, for example, dictate that the state and police must be abolished as regulatory powers, that the form of government must be changed, and that dissenters must be rejected. A new aspect within the discussion of such commonalities is that, according to the present findings, this convergence in extremes also seems to apply to the representatives of Islamism among the young people surveyed. This finding is, however, consistent with the observation that (strategic) alliances are forming between different ideological extremes.[7]

However, it is also plausible that the finding of high overlap between ideologies is due to the relatively young age of the sample. As models of political opinion formation (Beelmann 2018; Rekker et al. 2015) suggest, patterns of political attitudes are first differentiated and then acquire a certain direction during the course of ontogenetic development. The sample we studied is still at the beginning of this process, with most students not yet having reached the age of 16. It should thus be examined whether these ideological overlaps tend to decrease in older samples.[8]

The analysis of the radicalization process produced overall confirmatory findings for a model that assumes radicalization to commence development at the emotional-cognitive level, in which the first stage involves the adoption of ideologically influenced attitudes. This is followed by the second stage, which sees the development of an attitude that also approves of significant rule violations in order to achieve ideological goals. The behavioral level is reached at the third stage, in which significant rule violations are also committed. This process can be terminated at any stage or can alternatively continue until extremist behavior is exhibited. In contrast, this model does not predict the formation of extremist behavior without initial preparation at the emotional-cognitive level. The descriptive analysis of the survey data initially revealed that the characteristic combinations which correspond to the model-compliant developmental pathways are significantly more prevalent than the non-model-compliant trait combinations. The proportion of non-model-compliant combinations ranged from only 9.9 % to 14.4 %. Subsequent configural frequency analyses conducted for each radicalization ideology also confirmed the model assumptions. Combinations of characteristics marking model-compliant paths occurred significantly more frequently than would be expected from a stochastic model, with one exception in each case, and were classified as types. In contrast, combinations that did not conform to the model were classified as antitypes in seven out of twelve cases, i.e., they occurred significantly less frequently.

A consistent exception to the model assumptions, however, was the path ‘100’. This path marks an early phase or a discontinuation of the radicalization process, which sees the development of radical attitudes without further radicalization taking place (Talaska, Fiske & Chaiken, 2008). This path should be relatively common, especially in a young sample. However, this is not the case. One possible explanation for this underrepresentation could refer to a particular response tendency in which respondents who indicate problematic attitudes are also very likely to concurrently admit to approving of violence. This hypothesis is supported by the high correlations between problematic attitudes and approval of violence across the different radicalization ideologies, as well as between problematic attitudes and general propensity to violence (see Tab. 2). In contrast, the correlations between attitudes and behaviors are significantly lower. However, the findings on path ‘101’ argue against this hypothesis; this path does not conform to the model and should therefore be underrepresented. The described bias, which prevents the separate occurrence of attitudes and approval of violence, should additionally reduce the occurrence of such a combination. However, the observed frequencies of this path in all three domains of extremisms are in line with stochastic expectations, which thereby speaks against such a bias.

Although the theoretical entry path ‘100’ did not occur with the expected frequency, the present findings nevertheless support a theoretical radicalization model that assumes radicalization to commence at the cognitive-emotional level, prior to the advocation of violence, and the final translation into radicalized behaviors. Alternative models that, for example, firstly assume a positive attitude toward violence and subsequent attitude formation[9] consistently show poorer characteristic values regarding classifications and more ambiguous patterns of types and antitypes.

This model of gradual radicalization, beginning from the development of basic radical attitudes through to the approval of violence and finally, to the behavioral level, is also supported by findings regarding the correlations between the various ideologically influenced attitude patterns and behaviors with a general approval of violence and general delinquency. Here, the correlation patterns indicate a general attitude of protest, which also approves the use of violence, among the surveyed students, rather than an initial ideologically directed radicalization. The generally lower correspondences between the attitude scales (including general and specific approval of violence) on the one hand, and extremist or general delinquent behaviors on the other, indicate that there is a transition between the attitudinal and behavioral levels that is not typically followed. In other words, a general or specific attitude of protest and approval of violence does not inevitably or even regularly transition into transgressive behaviors. This thesis is also supported by the fact that path ‘110’ is clearly more strongly occupied in all three ideologies than path ‘111’.

4.1 Limitations

The choice of appropriate cutpoints, from which point a characteristic expression is to be considered problematic and should thus be included in the analysis, has so far constituted an unsolvable problem. Thus far, so-called consensus solutions, which use a theoretical scale center as a cutpoints or agreement with at least individual items including problematic content as a criterion, have been established within research (e.g., Decker et al. 2020). For the behavioral level of the present analysis, the latter procedure was also chosen. For the dichotomization of characteristic expressions at the attitude level, different percentiles of the distribution (70th and 80th percentiles, respectively) were used as cutpoints in the absence of available standards. The percentiles chosen are not theoretically justifiable and may appear arbitrary. However, other cutpoints were used to corroborate the findings, and each yielded very similar results in relation to types and antitypes. Based on this background, the present findings can be described as robust.

In addition to the problems of criteria selection described above, the study has other limitations that should be considered when interpreting the findings and drawing conclusions for research and practice.

This study examines radicalization trajectories among students in the 9th grade. This age group, in the last year of compulsory education in Germany, allows for the inclusion of all school types and thus enables a broad view of an age group that is in the early stages of its engagement with political and social issues (Greve 2007; Eckstein & Noack 2018) Based on the study’s funding framework and data protection reasons, only a cross-sectional study could be conducted. Developmental trajectories could thus only be approximately examined via combinations of characteristics.

The study is based solely on self-reports by the surveyed students. Self-reports are subject to biases due to socially desirable response tendencies. In addition, response tendencies which result in contrasting self-portrayals may also play a role. Although intensive data cleaning can rule out distortions due to noticeable response tendencies, milder cases of under- or overreporting may have gone undetected.

As mentioned, the sample studied is selectively composed, which means that the relevant radicalization characteristics are likely to be overrepresented compared to the general population of students. For this reason alone, the proportions of respondents who meet individual radicalization characteristics cannot be used as an estimation of prevalence.

Another aim of the present study was to analyze the radicalization processes involved in three radicalization ideologies (right-wing, left-wing, Islamist). The investigation of different ideologies of radicalization cannot be conducted without addressing different attitude objectives, motives, patterns of behavior, etc. This results in items which differ in terms of content, which can only be aggregated into similar, but not identical, scales. Nevertheless, this procedure provides important indications of intersections as well as distinctions between different radicalization ideologies and related phenomena.[10]

4.2 Conclusions and outlook

Whilst considering the methodological limitations, the present study provides clear indications of initial radicalization trajectories among some of the ninth-grade students at general education schools. The correspondences of the individual radicalization characteristics of different radicalization ideologies also indicate that incipient radicalization in different directions (right-wing, left-wing, Islamist) does not lead to an increasing disassociation between attitudes, the formation of violence, and the performance of radical acts, but that these radicalization characteristics also show considerable areas of overlap. The present analysis thus supports the thesis that different ideologies of extremism develop commonalities in their positions, attitudes and structures. For preventative purposes, this means that for the age group studied here, programs and measures with the same or similar content for political education and democracy promotion are promising in relation to all three ideologies. However, the central role of the approval of violence to achieve one's own ideological goals for all three ideologies of radicalization also suggests that democratically legitimate forms of achieving political goals should be conveyed to this age group and that related exploration should be made possible. This also includes the appropriate handling of minority votes within democratic processes in order to avoid experiences of exclusion and the resulting recourse to radical and violent solutions.


Alwin, D.F. & Krosnick, J.A. (1991). Aging, cohorts, and the stability of sociopolitical orientations over the life span. American Journal of Sociology, 97, 169-195. Search in Google Scholar

Baier, D. & Pfeiffer, C. (2011). Jugendliche als Opfer und Täter von Gewalt in Berlin. Hannover: KFN. Search in Google Scholar

Beelmann, A. (2018). Entwicklungsorientierte Kriminalprävention: Wissenschaftliche Fundierung und Ergebnisse der Evaluation, in: M. Walsh, B. Pniewski, M. Kober und A. Armborst (Hrsg.), Evidenzorientierte Kriminalprävention in Deutschland: Ein Leitfaden für Politik und Praxis, 387-406. Wiesbaden. Search in Google Scholar

Beelmann, A. (2019). Grundlagen eines entwicklungsorientierten Modells der Radikalisierung, in: E. Marks und C. Heinzelmann (Hrsg.), Prävention & Demokratieförderung: Gutachterliche Stellungnahmen zum 24. Deutschen Präventionstag, 181-209. Godesberg. Search in Google Scholar

Beelmann, A. (2020). A social-developmental model of radicalization: A systematic integration of existing theories and empirical research. International Journal of Conflict and Violence, 14, 1–14. Search in Google Scholar

Bergmann, M.C., Baier, D., Rehbein, F. & Mößle, T. (2017). Jugendliche in Niedersachsen: Ergebnisse des Niedersachsensurveys 2013 und 2015. Hannover: KFN. Search in Google Scholar

Bjørgo, T. (2011). Dreams and disillusionment: Engagement in and disengagement from racist groups. Crime, Law and Social Change, 55, 277-285. Search in Google Scholar

Böckler, N. & Zick, A. (2015). Wie gestalten sich Radikalisierungsprozesse im Vorfeld jihadistisch-terroristischer Gewalt? Perspektiven aus der Forschung. In: D. Molthagen (Hrsg.), Handlungsempfehlungen zur Auseinandersetzung mit islamistischem Extremismus und Islamfeindlichkeit: Arbeitsergebnisse eines Expertengremiums der Friedrich-Ebert-Stiftung. Berlin. Search in Google Scholar

Borum, R. (2011). Radicalization into violent extremism: A review of social science theories. Journal of Strategic Security, 4, 7. Search in Google Scholar

Borum, R. (2014). Psychological vulnerabilities and propensities for involvement in violent extremism. Behavioral Sciences and the Law, 32(3), 286–305. Search in Google Scholar

Borum, R. (2017). The etiology of radicalization. In: G. Lafree & J.D. Freilich (ed.), The Handbook of the Criminology of Terrorism, 17 ff. Chichester. Search in Google Scholar

Brettfeld, K. & Wetzels, P. (2007). Muslime in Deutschland: Integration, Integrationsbarrieren, Religion sowie Einstellungen zu Demokratie, Rechtsstaat und politisch-religiös motivierter Gewalt. Ergebnisse von Befragungen im Rahmen einer multizentrischen Studie in städtischen Lebensräumen. Hannover: KFN. Search in Google Scholar

Crick, N.R. & Dodge, K.A. (1994). A review and reformulation of social information-processing mechanisms in children’s social adjustment. Psychological Bulletin, 115, 74-101. Search in Google Scholar

Decker, O. & Brähler, E. (2006). Vom Rand zur Mitte: Rechtsextreme Einstellung und ihre Einflussfaktoren in Deutschland. Berlin. Search in Google Scholar

Decker, O. & Brähler, E. (2020) (Hrsg.). Autoritäre Dynamiken: Alte Ressentimens – neue Radikalität. Leipziger Autoritarismus Studie 2020. Giessen. Search in Google Scholar

Decker, O., Kiess, J., Schuler, J., Handke, B., Pickel, G. & Brähler, E. (2020). Die Leipziger Autoritarismus Studie 2020: Methode, Ergebnisse und Langzeitverlauf. In: O. Decker & E. Brähler (Hrsg.), Autoritäre Dynamiken: Alte Ressentiments – neue Radikalität. Leipziger Autoritarismus Studie 2020, 27-89. Giessen. Search in Google Scholar

Deutz-Schroeder, M. & Schroeder, K. (2016). Linksextreme Einstellungen und Feindbilder: Befragungen, Statistiken und Analysen. Frankfurt am Main. Search in Google Scholar

Dreher, E. & Dreher, M. (1985). Wahrnehmung und Bewältigung von Entwicklungsaufgaben im Jugendalter: Fragen, Ergebnisse und Hypothesen zum Konzept einer Entwicklungs- und Pädagogischen Psychologie des Jugendalters. In: R. Oerter (Hrsg.), Lebensbewältigung im Jugendalter: Edition Psychologie, 30-61. Weinheim. Search in Google Scholar

Dzhekova, R., Mancheva, M., Stoynova, N. & Anagnostou, D. (2017). Monitoring Radicalisation: A Framework for Risk Indicators. Sofia. Search in Google Scholar

Eckstein, K. & Noack, P. (2018). Politische Sozialisation. In: B. Gniewosz & P.F. Titzmann (Hrsg.), Handbuch Jugend: Psychologische Sichtweisen auf Veränderungen in der Adoleszenz, 371-387. Stuttgart. Search in Google Scholar

Emmelkamp, J., Asscher, J. J., Wissink, I. B. & Stams, G. J. J. M. (2020). Risk factors for (violent) radicalization in juveniles: A multilevel meta-analysis. Aggression and Violent Behavior, 55. Advance online publication. Search in Google Scholar

Eye, A. von (1990). Introduction to Configural Frequency Analysis: the Search for Types and Antitypes in Cross-Classifications. Cambridge, UK. Search in Google Scholar

Eye, A. von (2002). Configural Frequency Analysis: Methods, Models and Applications. Mahwah, NY. Search in Google Scholar

Fishbein, M. & Ajzen, I. (1975). Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research. Reading, MA. Search in Google Scholar

Frindte, W., Boehnke, K., Kreikenbom, H. & Wagner, W. (2011). Lebenswelten junger Muslime in Deutschland: Ein sozial- und medienwissenschaftliches System zur Analyse, Bewertung und Prävention islamistischer Radikalisierungsprozesse junger Menschen in Deutschland. Berlin. Search in Google Scholar

Gill, P., Horgan, J. & Deckert, P. (2014). Bombing alone: Tracing the motivations and antecedent behaviors of lone-actor terrorists. Journal of Forensic Sciences, 59, 425-435. Search in Google Scholar

Goede, L.-R., Schröder, C.P. & Lehmann, L. (2020). Perspektiven von Jugendlichen: Ergebnisse einer Befragung zu den Themen Politik, Religion und Gemeinschaft im Rahmen des Projektes »Radikalisierung im digitalen Zeitalter (RadigZ)«. Hannover: KFN. Search in Google Scholar

Greve, W. (2007). Selbst und Identität im Lebenslauf. In: J. Brandstädter & U. Lindenberger (Hrsg.), Entwicklungspsychologie der Lebensspanne, 305-336. Stuttgart. Search in Google Scholar

Havighurst, R.J. (1972). Developmental Tasks and Education. 3rd ed. New York. Search in Google Scholar

Heitmeyer, W. (2002). Gruppenbezogene Menschenfeindlichkeit: Die theoretische Konzeption und erste empirische Ergebnisse. In: W. Heitmeyer (Hrsg.), Deutsche Zustände: Folge 1, 15-34. Frankfurt am Main. Search in Google Scholar

Heitmeyer, W., Müller, J. & Schröder, H. (1997). Verlockender Fundamentalismus: Türkische Jugendliche in Deutschland. Frankfurt am Main. Search in Google Scholar

Herding, M. (2013). Forschungslandschaft und zentrale Befunde zu radikalem Islam im Jugendalter. In: M. Herding (Hrsg.), Radikaler Islam im Jugendalter: Erscheinungsformen, Ursachen und Kontexte, 21-39. Leipzig. Search in Google Scholar

Küpper, B., Krause, D. & Zick, A. (2019). Rechtsextreme Einstellungen in Deutschland 2002-2018/19. In: A. Zick, B. Küpper & W. Berghan (Hrsg.), Verlorene Mitte – feindselige Zustände: Rechtsextreme Einstellungen in Deutschland 2018/19, 117-146. Bonn. Search in Google Scholar

Leiner, D. J. (2019). Too fast, too straight, too weird: Non-reactive indicators for meaningless data in Internet surveys. Survey Research Methods, 13(3). 229-248. Search in Google Scholar

Lienert, G.A. (1969). Testaufbau und Testanalyse. München. Search in Google Scholar

Lösel, F., Bliesener, T. & Bender, D. (2007). Social information processing, experiences of aggression in social contexts, and aggressive behavior in adolescents. Criminal Justice and Behavior, 34, 330-347. Search in Google Scholar

Lösel, F., King, S., Bender, D. & Jugl, I. (2018). Protective factors against extremism and violent radicalization: A systematic review of research. International Journal of Developmental Science, 12(1-2), 89–102. Search in Google Scholar

Lösel, F. & Stemmler, M. (2012). Continuity and patterns of externalizing and internalizing behavior problems in girls: A variable- and person-oriented study from preschool to youth age. Psychological Test and Assessment Modeling, 54(3), 307–319. Search in Google Scholar

Maresch, P. & Bliesener, T. (2015). Regionalanalyse zu Rechtsextremismus in Schleswig-Holstein: Abschlussbericht. Kiel. Search in Google Scholar

McCauley, C. & Moskalenko, S. (2008). Mechanisms of political radicalization: Pathways toward terrorism. Terrorism and Political Violence, 20, 415-433. Search in Google Scholar

McCauley, C. & Moskalenko, S. (2011). Friction: How Radicalization Happens to Them and Us. Oxford. Search in Google Scholar

McCauley, C. & Moskalenko, S. (2017). Understanding political radicalization: The two-pyramids model. American Psychologist, 72, 205-216. Search in Google Scholar

McGilloway, A., Ghosh, P. & Bhui, K. (2015). A systematic review of pathways to and processes associated with radicalization and extremism amongst muslims in western societies. International Review of Psychiatry, 27, 39-50. Search in Google Scholar

Moghaddam, F.M. (2005). The staircase to terrorism: A psychological exploration. American Psychologist, 60, 161-169. Search in Google Scholar

Nivette, A., Eisner, M. & Ribeaud, D. (2017). Developmental predictors of violent extremist attitudes: A test of general strain theory. Journal of Research in Crime and Delinquency, 54, 755-790. Search in Google Scholar

Rekker, R., Keijsers, L., Branje, S. & Meeus, W. (2015). Political attitudes in adolescence and emerging adulthood: Developmental changes in mean level, polarization, rank-order stability, and correlates. Journal of Adolescence, 41, 136-147. Search in Google Scholar

RIAS Berlin. Antisemitische Vorfälle 2019: Ein Bericht der Recherche- und Informationsstelle Antisemitismus Berlin. Berlin 2020. (05.02.2021). Search in Google Scholar

Ribeaud, D. & Eisner, M. (2010). Are moral disengagement, neutralization techniques, and self-serving cognitive distortions the same? Developing a unified scale of moral neutralization of aggression. International Journal of Conflict and Violence, 4, 298–315. Search in Google Scholar

Sageman, M. (2004). Understanding Terror Networks. Philadelphia. Search in Google Scholar

Schils, N. & Verhage, A. (2017). Understanding how and why young people enter radical or violent extremist groups. International Journal of Conflict and Violence, 11, 1-17. Search in Google Scholar

Schrepp, M. (2006). The use of configural frequency analysis for explorative data analysis. The British Journal of Mathematical and Statistical Psychology, 59, 59-73. Search in Google Scholar

Silke, A. (2008). Holy warriors: Exploring the psychological processes of jihadi radicalization. European Journal of Criminology, 5, 99-123. Search in Google Scholar

Simi, P., Sporer, K. & Bubolz, B.F. (2016). Narratives of childhood adversity and adolescent misconduct as precursors to violent extremism. Journal of Research in Crime and Delinquency, 53, 536-563. Search in Google Scholar

Sykes, G.M. & Matza, D. (1957). Techniques of neutralization: A theory of delinquency. American Sociological Review, 22, 664. Search in Google Scholar

Stemmler, M. & Lösel, F. (2012). The stability of externalizing behavior in boys from preschool age to adolescence: A person-oriented analysis. Psychological Test and Assessment Modeling, 54(2), 195–207. Search in Google Scholar

Talaska, C.A., Fiske, S.T., & Chaiken, S. (2008). Legitimating racial discrimination: Emotions, not beliefs, best predict discriminaltion in a meta-analysis. Social Justice Research, 21, 263-396. Search in Google Scholar

Walther, E. (2014). Wie gefährlich ist die Gruppe? Eine sozialpsychologische Perspektive kriminalitätsbezogener Radikalisierung. Zeitschrift für Internationale Strafrechtsdogmatik, 393-401. Search in Google Scholar

Wikström, P.-O.H. & Bouhana, N. (2016). Analyzing radicalization and terrorism. In: G. Lafree & J.D. Freilich (ed.), Handbook of the Criminology of Terrorism, 175-186. Chichester, West Sussex. Search in Google Scholar

Wiktorowicz, Q. (2005). Radical Islam Rising: Muslim Extremism in the West. Lanham, MD. Search in Google Scholar

Wolfowicz, M., Litmanovitz, Y., Weisburd, D. & Hasisi, B. (2019). A field-wide systematic review and meta-analysis of putative risk and protective factors for radicalization outcomes. Journal of Quantitative Criminology, 36, 407–447. Search in Google Scholar

Published Online: 2021-08-07
Published in Print: 2021-09-14

© 2021 Walter de Gruyter GmbH, Berlin/Boston