The research team, led by Hejing Zhang and Yina Ma of Beijing Normal University, analyzed behavior from 528 participants engaged in a real-time offender-defender game. Participants were organized into six-person same-sex groups, with three-person offense teams attempting to extract resources from a three-person defense team. Over 24 rounds, each participant independently decided how many tokens to contribute to their group pool. The payoff structure was deliberately asymmetric: if the offensive group’s total contribution exceeded the defensive group’s total, the attackers retained their own uninvested tokens and equally shared the defenders’ remaining unspent tokens, while defenders earned zero for that round. If the defensive total matched or exceeded the offensive total, the attack failed and all participants retained only their own uninvested tokens. This design distilled offense and defense into their core strategic problems: acquiring resources under uncertainty versus protecting resources under threat.

To quantify the strategies participants used, the researchers computed two indices from the trial-by-trial contribution data. Behavioral variability, indexing strategic unpredictability, was defined as the intra-individual variance of contributions across the 24 rounds. Strategic tracking, indexing state estimation, was measured as the extent to which an individual’s contributions changed in response to the opponent group’s previous actions, calculated using a lagged regression model. The results revealed a robust complementary pattern: behavioral variability was positively associated with success in offensive roles (r = 0.361, p < 0.001) but negatively associated with success in defensive roles (r = -0.179, p = 0.003). In contrast, strategic tracking predicted higher success in defense (r = 0.277, p < 0.001) but was unrelated to success in offense.

The researchers went further, decomposing variability into its behavioral components to rule out a pure noise account. High variability was driven by a pulsed pattern: combining high non-zero contribution levels when engaging, interspersed with intermittent zero-contribution rounds. Both engagement intensity and zero-contribution rate were independently associated with offensive success. Zero-contribution rounds were systematically linked to lower defensive resistance, greater average spoils on successful offensive rounds, and greater resource-extraction efficiency. This pattern argues against random fluctuation and instead suggests a temporally pulsed mode of offensive engagement: larger active-round contributions were linked to successful attack, whereas intermittent disengagement was associated with lower defensive investment and more efficient resource capture when offense succeeded.

When the researchers examined whether these strategic optima aligned with sex-differentiated behavioral priors, a clear pattern emerged. Across roles, males showed significantly higher behavioral variability than females (20.580 versus 15.510; p < 0.001), and this effect remained significant after controlling for opponent average contribution. Decomposing the variability effect revealed that males showed higher non-zero contribution levels than females (9.686 versus 8.272; p < 0.001), and in covariate-adjusted sensitivity analyses, males also showed higher zero-contribution frequency. This suggests that the male strategy is characterized by pulsed, high-intensity engagement that generates unpredictability, rather than simple differences in effort.

Females, in contrast, exhibited a role-specific bias toward adaptive tracking. A significant sex-by-role interaction emerged for the tracking index (p < 0.001). In the defensive role, female defenders showed significantly stronger positive opponent-contingent tracking than male defenders (0.223 versus 0.120; p = 0.011). In the offensive role, female offenders showed more negative opponent-contingent coupling than male offenders (-0.200 versus -0.107; p = 0.021). This interaction remained significant after controlling for opponent average contribution. Thus, females showed stronger opponent-contingent adjustment than males in both roles, but this adjustment was expressed in opposite directions: positive tracking in defense and more negative coupling in offense.

These strategic biases translated into measurable performance asymmetries. Males outperformed females in offensive roles, achieving higher success rates (0.259 versus 0.189; p < 0.001) and making decisions closer to the theoretical optimum. In contrast, females outperformed males in defensive roles, achieving higher success rates (0.811 versus 0.741; p < 0.001) and earning substantially more (9.985 versus 8.485; p < 0.001). Crucially, process measures indicated that the female defensive advantage was underpinned by greater cognitive efficiency: female defenders not only made more rational decisions but also responded significantly faster than male defenders. These patterns remained robust in supplementary analyses controlling for the opponent behavioral environment.

The researchers also explored whether individual differences in salivary testosterone were related to these role-aligned behavioral patterns. In male offenders, higher log-transformed testosterone values were associated with greater behavioral variability both before and after the task, with post-task testosterone also covarying with offensive success. In female defenders but not female offenders, higher salivary testosterone values were associated with stronger strategic tracking. These findings are presented as preliminary associations within each sex rather than as tests of male-female differences in mean testosterone levels, and the authors caution that estimates from cotton-collected salivary immunoassays can depend on both the collection procedure and the analytical method, particularly at the lower testosterone concentrations typically observed in women. Replication using more precise serum-based measures will be needed to establish the robustness of these associations.

Why do the sexes diverge along these specific strategic lines? The authors interpret these patterns as average, role-congruent tendencies under the asymmetric demands of the task, rather than fixed sex-specific specializations. Males, who have historically faced higher variance in reproductive payoffs, appear to encode a high-volatility prior, enduring the costs of variable engagement to maximize the probability of high-stakes returns. Females, incurring higher costs of injury and parental investment, appear to encode a precision prior, prioritizing error minimization and sensitivity to social cues. This account also accommodates the possibility that female competition is active and strategic rather than passive: work on competitive gossip and informational competition emphasizes monitoring rivals, updating socially relevant information, and using such information to manage competitive threats. Such information-sensitive strategies may fit the defensive demands of the present task, where success depended on tracking opponent behavior and adjusting contributions accordingly.

Importantly, the data indicate that these sex-differentiated priors reflect functional competence rather than impulsivity. The male advantage in offense and the female advantage in defense were accompanied by higher decision rationality and greater economic efficiency, implying that the male tendency toward variability is a potentially functional tactic to capitalize on uncertainty, while the female tendency toward tracking reflects an efficient heuristic for threat management. The present data cannot determine whether these tendencies reflect evolved affordances, contemporary gendered socialization, prior competitive experience, or their interaction. A key question for future research is how these distinct priors interact in direct cross-sex competition: whether the high-variability offense of males overwhelms the tracking capacity of female defenders, or whether female vigilance effectively decodes and neutralizes that variability. The study nonetheless provides a compelling framework for understanding sex- and role-related variation in intergroup decision-making, revealing that human conflict relies on a dissociated strategy architecture in which unpredictability generation and opponent tracking serve as specialized tools for acquisition and protection.

Subject of Research: Behavioral science

Article Title: Offensive unpredictability and defensive tracking in intergroup conflict show sex-differentiated patterns

Article References: Zhang, H., Zhang, C., Widjaja, A., & Ma, Y. (2026). Offensive unpredictability and defensive tracking in intergroup conflict show sex-differentiated patterns. iScience, 29(11), Article 117887. https://doi.org/10.1016/j.isci.2026.117887

Image Credits: AI Generated

DOI: Not provided

Keywords: intergroup conflict, sex differences, unpredictability, strategic tracking, testosterone, game theory, evolutionary psychology, defense, offense, decision-making, Offensive, defensive