Experimental design. (a) Pairs of female participants (dyads) attended on two separate days in a blinded, randomized, placebo-controlled cross-over design. Both dyad members received identical treatment order. (b) Participants had blood taken before treatment and testing. (c) During testing dyad members sat in the same room viewing separate monitors. In a 2-alternative forced choice, design gratings were presented at two intervals, one containing a target grating with increased contrast. Each participant initially responded without consultation, providing measures of individual decision-making (Sindiv). If they disagreed, a joint decision was requested, which provided a measure of collaborative decision-making (Scollective). (d) Example psychometric function for dyad 1 under placebo. Proportion of trials reported as second interval is plotted against target contrast difference. Highly sensitive observers give steep functions with large slope (S). Here individuals (Sindiv) are red and green, and the dyad (Scollective) blue.
Individuals derive a performance benefit from collaboration. The dyad's collaborative decisions were more sensitive (Scollective) than the individuals' decisions alone (Sindiv). Our metric for this performance benefit on the vertical axis is the difference between an individual's sensitivity and the cooperative sensitivity achieved by their dyad (Benefit of collaboration = Scollective− Sindiv). This benefit is attenuated by testosterone when collapsed across all 34 participants (Sindiv) and also when only the better (Smax) or worse (Smin) members of each dyad are included. All t-tests shown are paired. Error bars indicate s.e.m.
Testosterone disrupts collaboration by increasing the egocentricity of decision-making. Each member of the dyad announced the dyad's joint decision in half the trials where such a collaborative decision was required. The sensitivity of collaborative decision-making hinges on the distribution in weighting attributed to one's own and the other's opinions. For each participant, we measured this weighting by the ratio of times they agreed with themselves (egocentric decisions) to agreement with the other's opinion (allocentric decisions). An egocentric–allocentric ratio of 1 means that participants weight their own and the other's original judgement equally. On placebo, there is trend towards egocentricity bias (one-sample, t33 = 1.8, p < 0.1)—an egocentricity bias that becomes marked on testosterone (one-sample, t33 = 3.0, p = 0.005). We show a paired t-test for testosterone versus placebo (t33 = 2.4, p < 0.05). Error bars indicate s.e.m.