Effects of Testosterone Administration on Strategic Gambling in Poker Play

Testosterone has been associated with economically egoistic and materialistic behaviors, but -defensibly driven by reputable status seeking- also with economically fair, generous and cooperative behaviors. Problematically, social status and economic resources are inextricably intertwined in humans, thus testosterone’s primal motives are concealed. We critically addressed this issue by performing a placebo-controlled single-dose testosterone administration in young women, who played a game of bluff poker wherein concerns for status and resources collide. The profit-maximizing strategy in this game is to mislead the other players by bluffing randomly (independent of strength of the hand), thus also when holding very poor cards (cold bluffing). The profit-maximizing strategy also dictates the players in this poker game to never call the other players’ bluffs. For reputable-status seeking these materialistic strategies are disadvantageous; firstly, being caught cold bluffing damages one’s reputation by revealing deceptive intent, and secondly, not calling the other players’ bluffs signals submission in blindly tolerating deception. Here we show that testosterone administration in this game of bluff poker significantly reduces random bluffing, as well as cold bluffing, while significantly increasing calling. Our data suggest that testosterone in humans primarily motivates for reputable-status seeking, even when this elicits behaviors that are economically disadvantageous.

were using these contraceptives, but not during menstruation 1 . We exclusively recruited 23 women because the parameters (quantity and time course) for inducing neurophysiological 24 effects after a single sublingual administration of 0.5mg of testosterone are known in women 25 but not in men 2,3 . These parameters, that is, 0.5 mg testosterone with and measurements 4 26 hours after administration were established using vaginal pulse amplitude (VPA), a uniquely 27 non-habitual, centrally driven response evoked by erotic material 2 . To date, no method is 28 available to assess the time course of effects of testosterone in human males. Crucially, 29 behavioral effects after a 4-h delay has been successfully established in more than 25 studies, 30 addressing sexual, social, and emotional behaviors in young women (e.g. 4-10 ). 31 32 Zero-sum two-person poker game 33

Mood 67
The shortened version of the Profile of Mood States 20 was used to index possible effects of 68 testosterone on anger, tension-anxiety, fatigue, vigor, and depression. There were no 69 significant differences on any of these subscales between the testosterone and placebo 70 conditions (all p's > .05), in line with our previous observations 3 . This precludes any 71 alternative explanations of our effects in terms of secondary mood-generated changes 21 72 73 2D: 4D digit ratio 74 Individual digit ratios were measured from an image scan of the right-hand by taking the 75 length of the index and ring finger from the ventral proximal crease to the tip of the finger 76 using an Adobe Photoshop measurement tool 22 . An experienced rater who was blind to the 77 goals of the experiment, measured 2D: 4D twice, with a latency of several weeks. These two 78 measurements were highly correlated (r = .99, p < .0001). The mean value of the two 79 measurements was used for analysis (see Statistical analyses). Nederland BV) was used as a tracer following chromatographic verification of its purity. The 86 lower limit of detection was 10 pmol/L, and interassay variation was 16.1, 11.5, and 5.1% at 87 21, 100, and 230 pmol/L, respectively (n = 4, 5, 5). Samples of one subject were not available, 88 and sample of a second subject was contaminated on the basis of abnormal range. Main 89 analyses concern samples of 18 subjects in placebo condition. 90 91

Statistical analyses 92
For bluffing behaviour, we restrict our analysis to initial bets for hands < ⅔ according to 93 theoretical framework of the game 11 . We examined the impact of testosterone [binary 94 indicator for testosterone (1) or placebo (0)], hand (continuous variable ranging from 0 to 1), 95 and interaction term (testosterone x (hand -mean hand) on the probability of betting "high" 96 (= 1, betting "low" = 0) as the dependent variable.. To qualify the testosterone x hand 97 interaction observed for bluffing behaviour, the bluffing range was split into three equal-sized 98 parts (0 -.22; .23 -.44; .45 -.66) on the basis of a priori hypotheses concerning cold bluffing. 99 Crucially however, we objectively defined the hand ranges wherein the overall (i.e. placebo 100 4 and testosterone conditions combined) relationship between betting behaviour and hand 101 changed significantly. We applied a moving average to the logistic betting distribution (first 102 bets (B) only with betting "high" B=1 and betting "low" B=0), which was subsequently 103 transformed to its linear equivalent by applying a natural logarithm to the odds-ratio of high 104 bets (lnOR with lnOR(B)=ln(B /(1-B))). The moving average was also applied to the poker-105 hand data and its window was set to include all observations between hand(H)=H and 106 hand(H)=H+0.05, which ensured elimination of the occurrence of undefined lnOR values (i.e. 107 lnOR(0)=ln(0/1) and lnOR(1)=ln(1/0)). Next, we used a Bai-Perron multiple break point test 23 108 to identify significant (p < .05) changes in the coefficients of a linear regression of lnOR(B) 109 on hand(H). Two break points were identified that indicated a significant change in betting 110 behaviour on Hand=0.20 and Hand=0.36 (see Figure S1 for the associated regression lines 111 and coefficients). For the calling behaviours, we examined the impact of testosterone (1) or placebo (0)], hand 128 (continuous variable ranging from 0 to 1), and an interaction term (testosterone x (hand -129 mean hand) on the probability of betting "calling" ( = 1) as the dependent variable (0 130 = "folding"). We analysed individual differences (i.e. salivary testosterone levels, 2D: 4D 131 digit ratio and BAS-dominance scores) using random effects logistic models. In these models 132 "betting" was the dependent variable (0: betting low; 1: betting high) and hand, BAS scores 133 and hand x BAS scores as well as salivary T levels and hand x salivary T interaction terms 134 were predictor variables. On the basis of our previous studies in which we report effects of 135 testosterone on behaviour modulated by the right hand's second-to-fourth-digit ratio, a proxy 136 of prenatal hormonal priming 8,24,25 , we performed a full random effects logistic model with: 137 dependent variable (0: betting low; 1: betting high) and predictors: hand, testosterone/placebo 138 administration (1 or 0), digit ratio, two-way interaction terms (testosterone x (hand -mean 139 hand); digit ratio x hand; digit ratio x testosterone administration) and a three-way interaction 140 term (hand x T administration x digit ratio) as predictor variables. 141 142

Beliefs about the effects of testosterone on poker play 143
Most of the participants reported that they believed that testosterone administration: would: 1) 144 Increase a desire to gamble (90% reported that testosterone would increase a desire to gamble; 145 10% had no idea); 2) Increase a willingness to take risks (95% increase; 5% no idea); 3) 146 Increase bluffing 65% (35% no idea); 4) Increase calling (60% increase (40% no idea). 147 Importantly, there was thus not one participant who believed that testosterone administration 148 would decrease a desire to gamble, willingness to take risks, bluffing or calling. 149 6 150 151

Beliefs concerning condition 152
On the final day of the experiment participants were asked on what day they believed to have 153 received testosterone, that is, 1 st or 2 nd day or whether they had no idea. 10 guessed correctly 154 and 10 guessed wrong or had no idea. To test whether participants' correct belief of having 155 received testosterone in any way interacted with our effects of testosterone on bluffing and 156 calling, we ran the bluffing and calling analyses again with the binary "belief" indicator 157 denoting whether the participant guessed correctly (1) or not (0) as covariate in random 158 effects analyses. The reported bluffing and calling effects remained significant, and the 159 covariate "belief" did not predict bluffing (p = .225) nor calling behavior (p = .216). 160

161
Beliefs about the effects of testosterone on gambling, bluffing and calling 162 The majority of participants reported that they believed that testosterone administration 163 would: 1) increase a desire to gamble (90% reported that testosterone would increase a desire 164 to gamble; 10% had no idea); 2) increase a willingness to take risks (95% increase; 5% no 165 idea); 3) increase bluffing (65% more often place higher bets for hands < .50; 35% no idea); 166 and 4) increase calling (60% increase; 40% no idea). Specifically, no participants reported 167 that they expected testosterone administration to: 1) decrease a desire to gamble, 2) decrease a 168 willingness to take risks, 3) decrease bluffing, or 4) decrease calling. 169 170