Internet trolling (“trolling”) is an antisocial online behaviour characterised by posting inflammatory, provocative comments with the intention of upsetting others (Buckels et al., 2014). As experiencing trolling can have significant psychological impact (Coles & West, 2016), management and prevention of this online behaviour is critical. Understanding individual differences associated with trolling, such as gender and personality traits, can inform development of appropriate interventions (March, 2019).

Although women engage in other antisocial online behaviour more than men (e.g., intimate partner cyberstalking; March et al., 2022), men are more likely to perpetrate trolling (Buckels et al., 2014). Possibly, this is due to their more antisocial use of social media (Howard et al., 2019), their higher dominance and competitiveness (March & Steele, 2020), and viewing trolling as having a functional purpose (Brubaker et al., 2021). People with higher subclinical psychopathy and sadism also perpetrate more trolling, likely due to their lower empathy and a deceitful interpersonal style (i.e., psychopathy; March & Steele, 2020), and their enjoyment of hurting others (i.e., sadism; March, 2019). People with lower agreeableness, lower conscientiousness (Buckels et al., 2014), and higher extraversion (Zezulka & Seigfried-Spellar, 2016) also perpetrate more trolling – although extraversion results are mixed. Here, trolling may be a product of increased antagonism (i.e., low agreeableness), increased impulsivity (i.e., low conscientiousness), assertiveness and a need to establish social status (i.e., higher extraversion; Zezulka & Seigfried-Spellar, 2016).

In the current study, we aimed to replicate these findings in a representative sample of self-reported “trolls”. As previous studies have largely assessed tendencies to troll (Zezulka & Seigfrield-Spellar, 2016) and agreement with trolling behaviours (March, 2019) in community samples, the validity of results may be questionable. Although someone might agree with trolling behaviours (e.g., the GAIT; Buckels et al., 2014), this agreement may not translate to actual behaviour. We predicted that men would be more likely than women to troll, and that people with higher sadism, higher psychopathy, lower agreeableness, lower conscientiousness, and higher extraversion, would perpetrate more trolling.

Method

After receiving institution ethical approval, participants were recruited to complete a voluntary, anonymous online questionnaire via social media (e.g., Facebook, Reddit, and Instagram) advertisements. After providing informed consent, participants (N = 663) were presented with an operational definition (see supplementary material) of trolling, followed by the inclusion question, “do you troll other people/groups online?” Only those who answered yes (N = 266; 40.12%) could commence the questionnaire. The questionnaire took approximately 20–30 min to complete and included an attention check. We also included an open-ended question to screen for inattention and assessed start/finish times for outliers.

The final sample (N = 163; 50.3% women) completed all measures (See Table 1) and satisfied attention checks. Participants were aged 18–62 years (Mage = 27.36, SD = 8.78), predominantly heterosexual (75.5%) and Australian (63.8%). An a priori power calculation indicated a minimum sample size of 153 (effect size = 0.15, alpha = 0.05, 95% power) required for statistical power. The study was a correlational, cross-sectional design, and based on research recommendations we controlled for socially desirable responding (see March, 2019).

Table 1 Summary of Online Questionnaire Measures
Full size table

Results

Assumptions of linearity, homoscedasticity, normality, and multicollinearity were screened and met. Table 2 presents total and gendered descriptive and bivariate correlations. To test the hypothesis, a 2-Step Hierarchical Multiple Regression Analysis was run (see Table 3). The total model explained 55.5% of variance in trolling, R²=0.56, F(7,149) = 26.59, p < .001, with a large effect size of ƒ²=1.27.

Table 2 Total and Gendered Descriptives for Gender, Social Desirability Sadism, Psychopathy, Extraversion, and Agreeableness, and Criterion of Trolling
Full size table
Table 3 Coefficients for Control of Social Desirability and Predictors of Gender, Extraversion, Agreeableness, Conscientiousness, Sadism, and Psychopathy Predicting Trolling
Full size table

Discussion

In this brief report, we aimed to evidence the utility of gender and personality to predict trolling in a sample of self-reported trolls. Corroborating past findings of general population samples (Buckels et al., 2014; Craker & March, 2016; Sest & March, 2017), men were more likely than women to troll, and trolling was predicted by lower empathy and a deceitful interpersonal style (i.e., higher psychopathy), enjoyment harming others (i.e., higher sadism), and antagonism (i.e., lower agreeableness).

Results also demonstrate some key differences in general population samples and representative samples. Conscientiousness did not predict trolling, a finding contrary to the hypothesis and to general population samples (Buckels et al., 2014). However, as trolling is deliberate (Sest & March, 2017), and can even include strategic coordination (Etudo et al., 2019), it follows that trolls may not display low conscientiousness. Further, extraversion did not predict trolling, a finding also contrary to previous research – though extraversion findings have been mixed (see Buckels et al., 2014; Gylfason et al., 2021; and Zezulka & Seigfried-Spellar, 2016), possibly due to the different measures employed to assess extraversion. Still, extraversion shares an inverse relationship with forms of deception (McArthur et al., 2022), and as deception is a key characteristic of trolling (Hardaker, 2010), the nonsignificant extraversion finding is logical. However, it is worth noting that extraversion was a significant predictor until sadism and psychopathy were entered, indicating either (1) sadism and psychopathy capture the variance explained by extraversion, or (2) the strength of sadism and psychopathy rendered extraversion nonsignificant.

We conclude that general population samples may somewhat, but not entirely, capture the psychological profile of trolls. Interventions seeking to target trolling behaviours should consider the deliberate and strategic nature of trolling; the troll may not be a reactionary, impulsive aggressor (i.e., low conscientiousness). Although the modest sample size of the current study somewhat limits generalisability, there was adequate statistical power and a large effect size (ƒ² = 1.27) was large. Future research exploring trolling could seek to apply theoretical frameworks to understand this antisocial use of social media, such as Uses and Gratifications theory (see Geary et al., 2021). Further, as trolls are often diverse in their purpose and goals (Sanfilippo et al., 2017), future research could focus on examining the individual differences of trolls who vary by goals, strategies, and platform.