Beyond the challenge hypothesis: The emergence of the dual-hormone hypothesis and recommendations for future research

Highlights

• The dual-hormone hypothesis extends the challenge hypothesis.

• Testosterone × cortisol interactions predict status seeking behaviors.

• Social contextual and individual difference factors may alter these associations.

• We recommend methodological improvements and key next steps for research.

Abstract

The challenge hypothesis makes specific predictions about the association between testosterone and status-seeking behaviors, but the findings linking testosterone to these behaviors are often inconsistent. The dual-hormone hypothesis was developed to help explain these inconsistencies. Specifically, according to this hypothesis, testosterone's association with status-seeking behavior depends on levels of cortisol. Here, we (1) describe the dual-hormone hypothesis in relation to the challenge hypothesis; (2) review recent studies that tested the dual-hormone hypothesis as well as meta-scientific evidence of heterogeneous dual-hormone findings across studies; (3) discuss potential explanations for this heterogeneity, including methodological considerations, contextual factors, and individual differences; and (4) provide recommendations for new work aimed at testing and extending the dual-hormone hypothesis.

Introduction

The challenge hypothesis was originally developed to explain the relationship between testosterone, a product of the hypothalamic-pituitary-gonadal (HPG) axis, and aggression in seasonally breeding birds (Wingfield et al., 1990). The authors proposed a multifaceted role for testosterone, where baseline testosterone levels are involved in the development and maintenance of reproductive systems. Within the challenge hypothesis framework, competitive interactions during the breeding season induce surges in testosterone levels above this baseline, which in turn direct territorial aggression and inter-male competitive behavior for access to receptive females and territory. The challenge hypothesis authors also posited that elevated testosterone levels reduced expression of paternal care.

The challenge hypothesis was later adapted to explain testosterone's association with status-seeking behaviors more broadly in primates (including humans; Archer, 2006; Muller and Wrangham, 2004). However, evidence of a direct association between testosterone and status-relevant behavior in humans is often inconsistent: While testosterone correlates positively with status-seeking behaviors such as aggression (Archer, 2006) and competitiveness (Eisenegger et al., 2017) in some reports, there have also been several observations of weak or null associations (Archer et al., 2005; Apicella et al., 2011).

The dual-hormone hypothesis sprang from these inconsistent associations between testosterone and social behavior (Mehta and Josephs, 2010; Mehta and Prasad, 2015; Sarkar et al., 2019). Specifically, this hypothesis provided a framework for interpreting the relatively mixed findings on testosterone-behavior links in terms of a physiological moderator, the glucocorticoid steroid cortisol. Cortisol is a product of the hypothalamic-pituitary-adrenal (HPA) axis that is released in response to physical and psychological stress (Dickerson and Kemeny, 2004) and mediates an array of metabolic and physiological responses to stress (McEwen, 2019). Stress plays a fundamental role in animal competition and mating, which are relevant to social status (Sapolsky, 2005). Further, the HPA and HPG axes influence each other at multiple levels (Viau, 2002; Burnstein et al., 1995; Chen et al., 1997; Johnson et al., 1992; Rivier and Rivest, 1991; Smith et al., 1985; Tilbrook et al., 2000; see also Mehta and Josephs, 2010, and Grebe et al., 2019a for discussion of putative mechanisms), suggesting that testosterone and cortisol may interact to affect behavior. Theoretical frameworks that link endocrine activity to status seeking, such as the challenge hypothesis, should therefore also account for cortisol levels. However, the challenge hypothesis focuses entirely on testosterone and does not emphasize a role for cortisol. Thus, previous studies that have tested the challenge hypothesis in humans may have found weak or null results because they did not consider cortisol as a potential moderator of testosterone's association with behaviors linked to status.

According to the dual-hormone hypothesis, testosterone's association with status-relevant behavior depends on cortisol levels, specifically that the positive association between testosterone and status-seeking behavior is more robust when cortisol levels are low¹ (Mehta and Josephs, 2010), while this association is attenuated when cortisol levels are high. The statistical prediction of this hypothesis is a negative testosterone × cortisol interaction term, when higher scores on the outcome measure indicate increases in behaviors related to the pursuit of high status.

Mehta and Josephs (2010) formally proposed the dual-hormone hypothesis. Across two studies, testosterone and cortisol were measured in afternoon saliva samples in undergraduate students. In Study 1, men and women participated in a leadership task. Seven judges rated video recordings of these social interactions on dominant leadership behaviors, defined as an assertive and self-assured behavioral style (Anderson and Kilduff, 2009). The rated variables included assertiveness, confidence, decisiveness, anxiety (reverse scored), and being perceived as “leader-like”. Ratings of these dominance-related variables based on behavioral observations predict the attainment of higher status in social groups (Anderson and Kilduff, 2009). In Study 2 (males only), participants chose whether or not to re-enter a competition against the same opponent after winning or losing. Choosing to challenge an opponent to a re-match after losing can be considered a status-seeking behavior because it may enable advancement in the social hierarchy (Mehta and Josephs, 2006). Both studies found testosterone × cortisol interactions consistent with the dual-hormone hypothesis: Higher basal testosterone levels were related to more dominant leadership behaviors (e.g. assertiveness, confidence, being perceived as leader-like) and decisions to re-enter competitions against the same opponent after defeat, but only when cortisol levels were low. When cortisol levels were high, higher testosterone levels were unrelated to dominant leadership behaviors (Study 1) and were negatively related to decisions to re-enter competitions against the same opponent after defeat (Study 2).

Although Mehta and Josephs (2010) first named and delineated the dual-hormone hypothesis, testosterone × cortisol interactions had been observed earlier, in research on aggression and criminal violence in male delinquent adolescents (Dabbs et al., 1991; Popma et al., 2007); basal testosterone correlated positively with aggression and violence among low-cortisol individuals but not among high-cortisol individuals. In groups of delinquent adolescents, anti-social behaviors such as aggression and violence may be effective in increasing one's status by inducing fear and submissive behaviors in other group members (Daly and Wilson, 1988).

Since these initial observations, testosterone × cortisol interactions consistent with the dual-hormone hypothesis have been found to predict a wide array of behaviors that humans may employ in the acquisition or maintenance of social status (reviewed in Mehta and Prasad, 2015; Sarkar et al., 2019). Moreover, new evidence suggests that this dual-hormone interaction is related to the actual attainment of higher status in real-world hierarchies.

For example, in studies of collegiate and Olympic female athletes, higher basal testosterone levels were related to higher social status, including teammate perceptions of leadership, but only when cortisol levels were low; higher testosterone levels were either unrelated or negatively related to social status when cortisol levels were high (Edwards and Casto, 2013; Casto et al., 2019). In another study of male rugby players, a social network analysis showed a similar pattern of results: Higher basal testosterone levels were related to higher status in the social network (e.g. higher popularity scores) only among players with low-cortisol levels but not among players with high-cortisol levels (Ponzi et al., 2016). A fourth study recruited male executives who held leadership positions in the public sector, including federal government officials, senior military officers, and some private-sector managers whose work was tied to the public sector (Sherman et al., 2016). Social status was indexed by the number of subordinates over which the individual had authority (Sherman et al., 2016). In line with the dual hormone hypothesis, basal testosterone was positively related to social status (a higher number of subordinates) among individuals with low cortisol but not among individuals with high cortisol.

Taken together, these studies suggest that the interaction between testosterone and cortisol is related to behaviors implicated in the pursuit of status (e.g. assertive and confident behaviors; re-challenging an opponent to a re-match after facing a defeat) as well as actual attainment of status in social groups (e.g. being perceived as a leader and having authority over larger groups of subordinates; but see also Mazur et al., 2015).

Recently, a meta-analysis was conducted to estimate the effect size of testosterone × cortisol interactions across studies and test for potential moderators of the effect (Dekkers et al., 2019). This meta-analysis included thirty-three studies with forty-nine effects,² and found evidence of a small but significant testosterone × cortisol interaction consistent with the dual-hormone hypothesis, but with substantial heterogeneity in the direction and magnitude of effect sizes (Dekkers et al., 2019). Upon examining individual categories of behaviors, this meta-analysis found that the evidence in support of the dual-hormone hypothesis was stronger for explicit indices of status (e.g., actual status rankings within hierarchies) compared to specific categories of behaviors (e.g. aggression, psychopathy). There was also stronger support for the dual-hormone hypothesis in males than females. However, it is important to note that these were only directional patterns; as the authors point out, there was limited statistical power to detect statistically significant differences in these moderator analyses.

Another recent article examined the dual-hormone hypothesis in a set of unpublished “file-drawer” studies that primarily utilized self-reported measures of status-striving personality features (e.g., a self-reported scale of competitiveness; Grebe et al., 2019a). The results did not provide strong evidence for a testosterone × cortisol interaction in line with the dual-hormone hypothesis. The authors argued that this result, in conjunction with a p-curve analysis, suggests that the published literature is characterized by low statistical power. However, Bayesian analyses did not reveal strong evidence in favor of the null hypothesis either. Thus, the results do not clearly favor the interpretation that testosterone × cortisol interactions are unrelated to self-reported status-striving personality traits. Grebe and colleagues also highlight a pattern of imprecision regarding behaviors that are categorized as “status seeking,” as well as inconsistency in which hormone measures show significant T x C interactions (e.g., basal versus dynamic change measures). Each of these scenarios could reasonably lead to wide-ranging and subjectively interpreted outcomes.

Based on the findings from this recent meta-analysis and “file drawer” study (Dekkers et al., 2019; Grebe et al., 2019a), we draw the interim conclusion that although there is some promising support for the dual-hormone hypothesis, the evidence is somewhat tepid and inconsistent across studies. How should researchers approach these inconsistent results? There are several plausible explanations for the heterogeneous results across studies. As such, there are new opportunities for discovery regarding associations between these dual-hormone interactions, status-seeking behaviors and actual status attainment, as well as better understanding the causal pathways that explain these associations. In the subsequent sections, we build on these recent reports to discuss potential explanations for the heterogeneous results across studies and we offer recommendations for future research aimed at testing and extending the dual-hormone hypothesis in light of these explanations.