The Terrorist Radicalization Protocol 18 (TRAP-18) is tested on a sample of 80 people who were convicted for Islamist activities in Germany between 2006 and 2017. In the study, perpetrators of terrorist attacks will be compared to persons who have been convicted of propagandistic and financial terrorist support and of joining a terrorist organization abroad. Statistical analysis of the results shows that there are significant differences between terrorist perpetrators and persons convicted of nonviolent Islamist activities, both in the number of TRAP-18 items and in the proximal warning behaviors "pathway," "last resort," "energy burst," and "novel aggression." Subsequent ROC analyses underline both the specificity and sensitivity of the instrument. AUC values range from .83 to .90 for the four different models (TRAP-18 and the warning behavior typology as weighted and unweighted models). The highest discrimination between Islamist attackers and the non-attackers is achieved by the weighted warning behavior typology. The values for sensitivity (se = .80), specificity (sp = .93), positive predictive value (p+ = .80), and negative predictive value (p− = .93) are extremely promising.
Public Significance StatementThe study deals with the question of whether TRAP-18 is capable of distinguishing persons who have committed an Islamist act of violence from those who have taken on more supportive, nonviolent roles in the scene. The results show that this is the case. Specific behavioral patterns can be identified that are overly coincidentally common in violent terrorists. Likewise, terrorists show more warning behavior than those who have not committed a violent crime. Against this empirical background, TRAP-18 has proven to be an instrument with a high potential for successful early detection of Islamist-terrorist violence.