Testosterone responses to competition predict future aggressive behaviour at a cost to reward in men

Summary

The extent to which trait factors (baseline testosterone concentrations, trait dominance) and state factors (change in social status, change in testosterone concentrations) would predict reactive aggression in a subsequent task that involved provocation was examined in 99 participants (39 men and 60 women). Participants first competed in same-sex dyads on a Number Tracing Task for which the outcome (win or loss) was rigged. After the competition, participants performed the Point Subtraction Aggression Paradigm (PSAP), a behavioural measure of reactive aggression against an opponent (actually a computer program). Trait dominance predicted baseline testosterone in men, but not women, and men made more aggressive responses than did women. Baseline testosterone concentrations did not predict aggressive behaviour in either men or women. Winners and losers did not differ in competition-induced change in testosterone. However, change in testosterone concentrations predicted aggressive responses in the PSAP for men in the loss condition, and aggressive responses were made at a cost to obtaining reward points. For men in the win condition, aggressive responses were predicted by an interaction between trait dominance and change in testosterone concentrations. These findings suggest that situational changes in testosterone concentrations modulate future aggressive behaviour in men.

Introduction

The World Health Organization has estimated that for every death due to physical aggression, another 20–40 youth require hospital treatment for an aggression-related injury (Mercy et al., 2002). The variety of ways in which aggressive behaviour is manifested (e.g., “road rage”, bullying, child abuse, domestic abuse and workplace violence) indicates the multifaceted nature of this behaviour. Despite the negative consequences of aggressive behaviour, the use (or threat) of aggression can be beneficial under certain conditions (e.g., athletic competition, self-defense, derogation of same-sex rivals and establishment of status hierarchies). Psychobiological investigations of the factors contributing to the expression of aggressive behaviour have identified many of the individual differences and situational factors that are associated with aggression, although most investigations in people have relied on self-report measures (see reviews by Anderson and Bushman, 2002, Bettencourt et al., 2006, Trainor and Nelson, 2007).

Dominance is a personality trait that involves the desire to seek control and/or influence over social situations, events, and relationships (Mehrabian, 1996). Although trait dominance is theoretically and empirically related to aggression, there have been few studies of the relationship between the two variables (Bettencourt et al., 2006). Individual differences in trait dominance predicted trait aggression as measured by self-report (Archer and Webb, 2006, Johnson et al., 2007), and men tend to score higher than women on self-report measures of trait dominance (Budaev, 1999, Costa et al., 2001) and on several measures of aggression (Archer, 2004). Given the empirical relationship between trait dominance and self-report aggression, it is plausible that trait dominance would also be related to behavioural aggression.

Testosterone is a biological factor of relevance to aggressive behaviour and to dominance in many species (reviewed in Simon and Lu, 2006, Trainor and Nelson, 2007). There have been reports of a positive association between self-report trait dominance and baseline testosterone concentrations (Cashdan, 1995, Grant and France, 2001, Sellers et al., 2007), although others have failed to replicate this finding (see Josephs et al., 2006, Stanton and Schultheiss, 2007). The relationship between baseline testosterone concentrations and various forms of aggressive behaviour is less evident in studies of people than in other animals (Book et al., 2001, Archer et al., 2005). The inconsistent findings for aggression may be due, in part, to the use of self-report measures as opposed to the direct measurement of aggressive behaviour (but see Pope et al., 2000, Klinesmith et al., 2006). Further, dynamic fluctuations in testosterone concentrations may be more related to aggressive behaviour than are baseline testosterone concentrations (Hermans et al., 2008). We recently found that baseline testosterone concentrations did not predict aggressive behaviour, but that aggressive behaviour was positively correlated with a rise in testosterone (Carré and McCormick, 2008). This result mirrors the findings of a study in non-human primates in which baseline testosterone concentrations did not predict aggressive behaviour, but aggressive behaviour was positively associated with a rise in testosterone concentrations (Ross et al., 2004).

Social interactions are known to modulate testosterone concentrations. For instance, winning competitive interactions (reviewed in Mazur and Booth, 1998, Archer, 2006, van Anders and Watson, 2006), good individual athletic performance (Edwards et al., 2006), the vicarious experience of victory and defeat (Bernhardt et al., 1998), and interactions with an attractive member of the opposite sex (Roney et al., 2003, Roney et al., 2007) all lead to changes in salivary testosterone concentrations. Dynamic shifts in testosterone concentrations have been proposed to influence future competitive and/or aggressive behaviours (Wingfield et al., 1990, Mazur, 1985, Mazur and Booth, 1998). A few studies have directly tested this hypothesis. For example, among losers (but not winners) of a competition, men whose testosterone concentrations had risen were more likely to choose to compete again, whereas men whose testosterone concentrations decreased chose the non-competitive option (Mehta and Josephs, 2006). We have also shown that changes in testosterone concentrations and aggressive behaviour during a competition predicted subsequent choice of a novel competitive task over a non-competitive task (Carré and McCormick, 2008). Furthermore, experimental studies have demonstrated that exogenous testosterone administrations increased cardiac responses to angry faces (van Honk et al., 2001), decreased fear-potentiated startle (Hermans et al., 2006a), increased visuospatial performance (Aleman et al., 2004), increased subcortical responses to angry faces (Hermans et al., 2008), decreased empathetic behaviour (Hermans et al., 2006b), and decreased conscious detection of angry faces (van Honk and Schutter, 2007). Although these studies support the idea that situational or experimental changes in testosterone concentrations are functionally related to future social behaviours, they do not speak to the issue of whether such changes in testosterone concentrations predict future aggressive behaviour.

Evidence from animal models suggests that the relationship between testosterone concentrations and future aggression is causal. A study of castrated male mice on low testosterone replacement found that those receiving a testosterone injection after a successful aggressive encounter were more aggressive in subsequent encounters compared to those that received a saline injection after a successful aggressive encounter (Trainor et al., 2004). One study has investigated the influence of a situation-specific change in salivary testosterone concentrations on future aggressive behaviour in people by comparing men who were given the opportunity to interact with a toy gun or a board game (Klinesmith et al., 2006). Men who interacted with the toy gun were more aggressive (as defined by the amount of hot sauce placed in another’s drink) compared to men who interacted with the board game. Importantly, the relationship between type of interaction and extent of aggressive behaviour was mediated by a rise in salivary testosterone concentrations, suggesting that testosterone was a factor influencing aggressive behaviour.

The studies above show relationships between either trait factors and aggressive behaviour or state factors and aggressive behaviour. The General Aggression Model (GAM) (Anderson and Bushman, 2002) posits that trait/personological factors (including personality traits, gender, attitudes and genetic predispositions) and state/situational factors (including features of the situation or environment such as the presentation of provocation, aggression cues, level of frustration and pain) influence various cognitive, emotional, metabolic and arousal mechanisms that mediate aggressive behaviour. However, studies of how trait and state factors interact to predict aggressive behaviour are lacking. We tested the hypothesis, derived from the literature reviewed above, that a competition-induced change in testosterone concentrations would predict subsequent aggressive behaviour as measured using the Point Subtraction Aggression Paradigm (PSAP). We included trait dominance as an individual difference variable and tested how this variable was associated with testosterone concentrations. Furthermore, based on previous self-report studies (Archer and Webb, 2006, Johnson et al., 2007), we predicted that trait dominance would be positively related to aggressive behaviour. Based on a few previous studies (Grant and France, 2001, Sellers et al., 2007), we also predicted that individual differences in baseline testosterone concentrations would be positively related to trait dominance. We included gender as a variable in our analyses because although men have higher concentrations of testosterone, are more physically aggressive (Archer, 2004), and have higher trait dominance scores (Budaev, 1999, Costa et al., 2001), the research literature is equivocal as to whether the relationships among these variables might differ for men and women (Dabbs and Hargrove, 1997, Mazur et al., 1997, Bateup et al., 2002, Kivlighan et al., 2005, Edwards et al., 2006, Josephs et al., 2006, Mehta et al., 2008).