The Reactive Proactive Questionnaire (RPQ) was originally developed to assess reactive and proactive aggressive behavior in children. Nevertheless, some studies have used the RPQ in adults. This study examines the reliability of the RPQ within an adult sample by investigating whether reactive and proactive aggression can be distinguished at a variable‐ and person‐based level. Male adults from forensic samples (N = 237) and from the general population (N = 278) completed the RPQ questionnaire. Variable‐based approaches, including factor analyses, were conducted to verify the two‐factor model of the RPQ and to examine alternative factor solutions of the 23 items. Subsequently, a person‐based approach, i.e., Latent Class Analysis (LCA), was executed to identify homogeneous classes of subjects with similar profiles of aggression in the observed data. The RPQ proved to have sufficient internal consistency. Multiple‐factor models were examined, but the original two‐factor model was statistically and theoretically considered as most solid and in line with previous research. The multi‐level LCA identified three different classes of aggression severity (class 1 showed low aggressive behavior; class 2 subjects displayed modest aggression levels; and class 3 exhibited the highest level of aggressive behavior). In addition, class 1 and 2 showed more reactive than proactive aggression, whereas class 3 displayed comparable levels of reactive/proactive aggression. The RPQ appears to have clinical relevance for adult populations in the way that it can distinguish severity levels of aggression. Before the RPQ is implemented in adult populations, norm scores need to be developed. Aggr. Behav. 43:155–162, 2017. © 2016 Wiley Periodicals, Inc.