Psychopathy is a personality construct characterized by dysfunctional interpersonal, behavioral, and emotional patterns, including low empathy, callousness, self-centeredness, impulsiveness, and the exploitation of others (Gómez-Leal et al., 2021; Hare & Neumann, 2008; Megías et al., 2018; Pennington et al., 2015). While psychopathy has often been examined as a personality disorder, it can also be conceptualized as a set of dimensional personality traits that are continuously distributed in the general population (Edens et al., 2006; Sellbom & Drislane, 2021). The present study adopts this dimensional approach by examining psychopathic traits in a community sample.

Much of the literature has conceptualized psychopathy as a two-component construct, comprising affective/interpersonal traits (factor 1) and characteristics related to social deviance (factor 2) (Hare, 1991, 1993, 2003; Neumann et al., 2007). However, subsequent studies on clinical and subclinical groups have provided evidence supporting a four-component model (Hare & Neumann, 2008; Mahmut et al., 2011; Neal & Sellbom, 2012): factor 1 is “callous affect” (an individual’s lack of emotion, empathy, remorse, and guilt); factor 2 is “interpersonal manipulation” (achieving goals through superficial charm, egocentricity, and pathological lying); factor 3 is “erratic lifestyle” (behavioral tendencies and characteristics such as recklessness and impulsivity); and factor 4 is “criminal tendencies” (antisocial characteristics and engagement in criminal activity). For this research, we adopted the four-factor model of psychopathy proposed by Hare and Neumann (2008), as it provides a well-defined conceptual framework and has received considerable empirical support (Neumann et al., 2015).

These affective, interpersonal, and behavioral facets of psychopathy have been consistently linked to a range of maladaptive outcomes. Existing literature has demonstrated a positive correlation between psychopathic traits and behaviors such as aggression, criminal activity, deceit, emotional manipulation, risk-taking, and difficulty regulating emotions, among other actions with harmful societal consequences (Frick et al., 2003; Garofalo et al., 2018; Grieve & Mahar, 2010; Hampejs et al., 2025; Piatigorsky & Hinshaw, 2004; Stanwix & Walker, 2021; Walker et al., 2022).

Risk behavior refers to actions that carry a certain probability of significant losses, whether objective or subjective, for the individual (Yates & Stone, 1992). The consequences of such behaviors can seriously threaten both mental and physical health (Pellmar et al., 2002; World Health Organization, 2018). A key characteristic of risk behavior is its domain-specific nature. Depending on the context in which a risk decision is made, the cost-benefit analysis carried out by each individual can differ, meaning that risky tendencies may not remain stable across different contexts (Blais & Weber, 2006; König, 2021; Lozano et al., 2017; Sánchez-López et al., 2022a, 2022b; Weber et al., 2002). The risk literature suggests five domains to contextualize risk (see Weber et al., 2002): ethical, health/safety, financial, social, and recreational domain. The ethical domain includes behaviors related to the violation of moral values (for example, evading tax or passing off someone else’s work as your own); the health and safety domain relates to behaviors that may cause health problems (for example, failing to stop at a red traffic light or having unprotected sex); the financial domain is related to investments and gambling (for example, betting large sums of money on a game of poker or investing a considerable proportion of your salary in a new business); the social domain encompasses behaviors involving interaction with others (for example, disagreeing with a figure of authority or starting a new career in midlife); and the recreational domain involves risk-taking behaviors related to recreation and extreme sports (for example, bungee jumping or flying a light aircraft). Thus, risk-taking behaviors can vary according to the specific domain in which they occur; consequently, an individual may tend to take risks in financial matters while being cautious when it comes to health and safety issues.

Given the domain-specific nature of risk-taking, an important question is whether individuals’ risk perceptions change within these domains and, if so, how these changes influence their willingness to engage in risky behaviors. The literature on the relationship between risk perception and risk-taking presents somewhat mixed findings. Evidence supports the intuitive notion that engaging in risky behavior is influenced by perceived risk; the greater the perceived risk of a situation or event, the less likely an individual will engage in risky behavior (Brewer et al., 2007; Finucane et al., 2000; Weber et al., 2002). Given a specific situation, if there is a perceived high risk of significant loss (for example, when faced with the possibility of tax evasion, the individual perceives that there is a high probability they will end up paying a fine), the likelihood that such behavior will ultimately be carried out will decrease. Nonetheless, some studies suggest that the relationship between perception and risk-taking may not always be straightforward. Contextual factors, characteristics of the risk stimulus, levels of impulsivity, and sensitivity to rewards or punishment, may moderate this relationship, leading individuals to engage in risky behavior even when risk perception is high (Fryt et al., 2024; Maldonado et al., 2016; Maldonado et al., 2020; Megías-Robles et al., 2022; Megías et al., 2015; Mills et al., 2008; Reyna & Farley, 2006; Sitkin & Pablo, 1992; Weller & Tikir, 2011). Thus, a driver may perceive a high risk in exceeding the speed limit in a residential area or in running a red light, and yet end up doing so anyway, because the desire not to be late for work carries greater weight in the decision, because of certain personality characteristics, or because this type of behavior has become a habit driven by contextual cues.

Therefore, examining how risk perception and risk-taking vary across contexts and domains is critical to understanding the factors that drive these behaviors and their potential association with distinct personal and psychological traits. Given that individuals with higher levels of psychopathic traits tend to show higher impulsivity, greater reward sensitivity, less sensitivity to punishment, and may misinterpret contextual factors and risk-relevant cues (Hare & Neumann, 2008; Johnson et al., 2014; Mahmut et al., 2011; Neal & Sellbom, 2012; Włodarska et al., 2021), in the present study we sought to investigate the relationship between psychopathy and both risk perception and risk-taking.

In recent decades, the association between the factors that comprise psychopathy and the tendency to engage in risky behavior has been widely studied (e.g., Dean et al., 2013; Hosker-Field et al., 2016; Hunt et al., 2005; Maneiro et al., 2020; Stanwix & Walker, 2021). For instance, using self-report measures, Stanwix and Walker (2021) found a positive association between psychopathy and risk domains, including ethical, health/safety, financial, social, and recreational. Similarly, Hosker-Field et al. (2016) reported positive correlations between all four psychopathy factors and both perceived risk and the likelihood of engaging in risky behaviors in the domains of ethics, health/safety, finances, and recreation. In the social domain, however, risky behavior was only positively associated with the psychopathy factors of interpersonal manipulation and erratic lifestyle. In line with this, Yap and Berezina (2021) observed that the social risk domain showed a weaker correlation with psychopathy than the other risk domains. The positive association between psychopathy and risk behavior has also been confirmed through behavioral tasks in laboratory settings, such as the Balloon Analogue Risk Task (see Hunt et al., 2005).

Previous studies have employed a variable-centered approach to explain the relationship between psychopathy and risk-taking, offering broad insight into how these variables relate to one another, but without capturing how psychopathy factors may combine within individuals. In contrast, our research adopted a person-centered approach (Howard & Hoffman, 2018; Morin et al., 2018). Based on individuals’ scores on each of the four factors of psychopathy, we aimed to identify different profiles according to how these factors are configured. This approach categorizes individuals into profiles based on shared factors, distinguishing, for example, between those with high scores across all four psychopathy factors and those with high scores on only some of them. This would be the case, for instance, of a profile with high psychopathy scores but without pronounced criminal tendencies, as observed in previous studies (Cooke & Michie, 2001; Cooke et al., 2004). By studying the risk tendencies of each of these profiles, we seek to provide a deeper perspective on the relationship between psychopathy and risk behavior.

The first aim of this research was to group individuals through cluster analysis according to their scores on all four psychopathy factors. Along with the expected identification of two clusters, one with high and one with low scores on all psychopathy factors, we explored the possibility of identifying additional clusters based on the greater presence or absence of specific psychopathy factors. Once individuals were classified, the study’s second and primary aim was to examine differences in risk perception and risk-taking among the identified clusters across various risk domains. Based on the existing literature, we expected: (1) profiles characterized by high scores on all four psychopathy factors would show lower risk perception and higher risk-taking than profiles with low scores on all four factors; (2) given the domain-specific nature of risk-taking, these differences could vary across risk domains, with prior findings suggesting stronger associations in the ethical, health/safety, financial, and recreational risk domains; and (3) profiles characterized by different combinations of psychopathy factors would exhibit distinct patterns of risk perception and risk-taking.

A total of 371 young adults were included in the study sample. They were recruited through advertising on the University of Málaga, social media, and various online platforms. Participants who failed to respond appropriately to the control questions were excluded (see questionnaire description). The ages of the participants from 18 to 59 (M_age = 24.04, SD = 6.72), of whom 98 were men (26.42%).

All participants provided online consent to participate and were assured that their data would remain anonymous, and their treatment adhered to the guidelines outlined in the Helsinki Declaration (World Medical Association, 2008). The study protocol was approved by the ethics committee of the University of Málaga (approval number: CEUMA 144-2023-H), forming part of this research project (B1-2021_10).

In accordance with the guidelines of the Declaration of Helsinki, participants were informed of the confidentiality and anonymity of their data (World Medical Association, 2008).

Data were collected in a single session, during which participants utilized the LimeSurvey online platform (http://limesurvey.org) to complete a set of questionnaires assessing levels of psychopathy, risk perception, and risk-taking. The authors sent email invitations to grant access to these questionnaires. Participants dedicated approximately 20 minutes to completing the assessment. A detailed explanation of each scale is provided below.

To assess psychopathy, we used the SRP-III (Self-Report Psychopathy Scale-III; Mahmut et al., 2011) in its Spanish version (Gómez-Leal et al., 2021). It is a self-report questionnaire consisting of 34 items to evaluate each of the four factors of psychopathy (Interpersonal manipulation, Callous affect, Erratic lifestyle, and Criminal tendencies) through a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). The construct validity and reliability of this scale have been demonstrated across diverse populations (Garofalo et al., 2018). The questionnaire demonstrated adequate internal consistency in our sample (α = .87).

To assess risk perception and risk-taking, we utilized the DOSPERT-30 (Domain-Specific Risk-Taking Scale; Blais & Weber, 2006) in its Spanish version (Lozano et al., 2017). It is a self-report questionnaire of 30 items assessing risk perception and the likelihood of engaging in risk behaviors in five domains (Ethical, Health/safety, Financial, Social, and Recreational; 6 items per domain) using a 7-point Likert scale (1 = extremely unlikely, 7 = extremely likely). While the original scale includes three parallel subscales (Risk perception, Risk-taking, and Expected benefits of behavior), our study focused solely on the Risk perception and Risk-taking subscales. For each subscale, the questionnaire provides a score for each risk domain, calculated as the sum of the item scores within that domain (range = 6–42), and a Total score, calculated as the mean of the five risk-domain scores (range = 6-42). Internal consistency in our sample was adequate (α = .86 for risk perception and α = .81 for risk-taking).

To improve the reliability of the assessment, we added two control items to both scales. Participants were instructed to “choose alternative 3” for control item 1 and “choose alternative 4” for control item 2.

First, we computed descriptive statistics for the four factors and the total score of the SRP-III, as well as for the total and domain-specific scores of Risk perception and Risk-taking on the DOSPERT. Second, we conducted Pearson correlations between psychopathy factors and the DOSPERT scores to examine the relationships among the variables. Third, we performed a hierarchical cluster analysis using Ward’s method and squared Euclidean distance to create distinct profiles based on the scores of the participants across the four psychopathy factors. Raw scores were used for the cluster analysis because all SRP-III subscales were measured on the same 5-point Likert scale; standardization was therefore not required (Hair et al., 2019) . To determine the optimal number of clusters, we examined the dendrogram and agglomeration schedule. Additionally, we employed a discriminant analysis with the Wilks’ Lambda test to confirm the meaningful distinctions between the identified clusters. Finally, we conducted Analyses of Covariance (ANCOVA) for the total scores of Risk Perception and Risk-taking on the DOSPERT, and multivariate analyses of covariance (MANCOVA) for their corresponding domain scores, to examine differences among the identified clusters. Clusters were treated as independent variables, with age and gender as covariates. In cases where significant main effects were found, the Bonferroni method was used to perform the post-hoc comparisons at a 5% significance level. The statistical analyses mentioned were performed using SPSS 24.0 (IBM Corporation, Armonk, NY, USA).

The dendrogram and the agglomeration schedule supported a four-cluster solution as the best fit for the data. Figure 1 illustrates these four clusters and the psychopathy scores for each cluster. The first cluster showed scores below the mean across all factors and was labeled the Low psychopathy group (LPG; n = 199). The second cluster presented scores above the mean for the Erratic lifestyle factor, and scores close to the mean for the remaining factors, termed the Erratic lifestyle group (ELG; n = 60). The third cluster was characterized by scores above the mean for all factors except Criminal tendencies, where scores were below the mean. This cluster was labeled the High psychopathy group (without Criminal tendencies) (HPG; n = 51). Finally, the fourth cluster presented scores well above the mean across all factors and was labeled the High psychopathy with Criminal tendencies group (HPCTG; n = 61). A discriminant analysis (Wilks’ lambda test) revealed good discrimination among clusters (p < .001), with 85.7% of cases correctly classified.

Differences in Risk perception between the four identified psychopathy clusters were analyzed using ANCOVA for the total score of the DOSPERT and MANCOVA for the DOSPERT domains. Prior to conducting these analyses, assumptions of normality and homoscedasticity of residuals were evaluated. Visual inspection of Q-Q plots, together with skewness (ranging from -1.01 to 0.66) and kurtosis (ranging from -0.81 to 1.26) values, indicated that the residuals were normally distributed (Kline, 2023; Tabachnick & Fidell, 2013). Box’s M test and Levene’s tests were both non-significant (ps > .05), indicating that the assumptions of homogeneity of covariance matrices and error variances were met.

The ANCOVA revealed a significant main effect of Cluster, F(3, 365) = 6.50, p < .001. The MANCOVA also showed a significant multivariate effect of Cluster (Wilk’s lambda = .88, F(15, 996) = 2.93, p < .001. Follow-up univariate analyses of the MANCOVA indicated significant differences between clusters in the Ethical, F(3, 365) = 7.42, p < .001, Health/safety, F(3, 365) = 4.37, p < .001, Financial, F(3, 365) = 3.14, p < .05, and Recreational, F(3, 365) = 5.26, p < .001, risk domains. However, no significant differences were observed for the Social domain (p > .05).

Table 3 and Figure 2 show the Bonferroni post-hoc comparisons between clusters. These results revealed: a) across all significant domains and for Total Risk perception, the LPG showed higher Risk perception scores than the HPCTG (ps < .05); b) in the Ethical domain, the LPG showed higher Risk perception scores than the HPG (ps < .05); and c) in the Recreational domain, the LPG showed higher Risk perception scores than the ELG (ps < .01).

As with Risk perception, differences in Risk-taking behavior between the four identified psychopathy clusters were examined using an ANCOVA for the total score and a MANCOVA for the domain scores. Visual inspection of Q–Q plots and skewness (-0.97 to 0.95) and kurtosis (-0.92 to 1.07) values indicated that the residuals were normally distributed. The only exception was observed for kurtosis in the LPG group in the Social domain, which showed a value of 2.70; however, this value remained within an acceptable range for assuming normality (Kline, 2023). Box’s M test and Levene’s tests were both non-significant (ps > .05), confirming homogeneity of covariance matrices and error variances.

The ANCOVA showed a significant main effect of Cluster, F(3, 365) = 43.14, p < .001. The MANCOVA revealed a significant multivariate effect of Cluster (Wilk’s Lambda = .67, F(15, 996) = 10.50, p < .001. Follow-up univariate analyses of the MANCOVA for each risk domain showed significant differences between clusters for the Ethical, F(3, 365) = 32.35, p < .001, Health/safety, F(3, 365) = 27.01, p < .001, Financial, F(3, 365) = 6.60, p < .001, and Recreational, F(3, 365) = 16.67, p < .001, domains. The Social domain showed no significant differences (p > .05).

Table 4 and Figure 3 show the Bonferroni post-hoc comparisons between clusters. The results revealed the following differences: a) Total Risk-taking: the LPG showed lower scores than the HPG, ELG, and HPCTG (ps < .01). In addition, both the ELG and the HPG showed lower scores than the HPCTG (ps < .01); b) Ethical domain: the LPG showed lower scores than the HPG and HPCTG (ps < .01). The ELG also showed lower scores than the HPG and HPCTG (ps < .05), whereas the HPG showed lower scores than the HPCTG (ps < .05); c) Health/safety domain: the LPG showed lower scores than the HPG, ELG, and HPCTG (ps < .01). The ELG and the HPG showed lower scores than the HPCTG (ps < .01); d) Financial domain: the LPG showed lower scores than the HPCTG (ps < .01); and e) Recreational domain: the LPG showed lower scores than the ELG and HPCTG (ps < .01), whereas the HPG showed lower scores than the HPCTG (ps < .05).

The present study advances understanding of psychopathy by adopting a person-centered approach, which revealed distinct profiles reflecting individuals’ unique combinations of the four psychopathy factors assessed by the SRP-III. These profiles were differentially associated with risk perception and risk-taking across risk domains, suggesting that risk-taking depends not only on overall psychopathy severity, but on how psychopathic traits are configured within the person. Our findings extend previous literature by clarifying how psychopathy is associated with an increased propensity toward risk-taking, while also contributing to the ongoing debate regarding the role of criminal tendencies within the psychopathy construct. From an applied perspective, considering distinct psychopathy profiles, rather than the overall level of psychopathic traits, may provide a more informative framework for risk assessment in clinical and forensic settings. Although future research in clinical populations is needed, identifying distinct psychopathy profiles and their corresponding patterns of risk-taking may inform the development of more targeted prevention and intervention strategies tailored to the specific characteristics of each profile, rather than relying on general risk-reduction programs. Such an approach could contribute to reducing maladaptive risk-taking behaviors and their associated negative consequences at both the individual and societal levels.