henhao
Sign in
NeurosciencebioRxiv
Preprint

Sex differences in exploration-exploitation strategies during home-cage decision making

Chantelle L. MurrelliD, Alex A. LegariaiD, Katie B McCullough, Andrew NwachaiD, Monsurat Ololade NasiruiD, Sebastian Alves Ferreira Dias, Rebecca ChaseiD, Mason R. BarrettiD, Matt GaidicaiD, Naoki HirataniiD, Meaghan C. CreediD, Joseph D. DoughertyiD, Susan E. MaloneyiD, Alexxai V. KravitziD

Abstract

The exploration-exploitation trade-off refers to the conflict between using known strategies that reliably yield reward (exploitation) and sampling uncertain options that might yield better outcomes (exploration). Dysregulation of this balance is implicated in neuropsychiatric disease, and while sex differences in this balance have been described, the biological bases remain unclear. To quantify sex differences in this trade-off, we tested mice (n=74 male, 62 female) on four home-cage based foraging tasks with an operant pellet dispensing device, Feeding Experimentation Device 3 (FED3). Mice completed the tasks continuously over multiple days and the tasks were their only source of food. Across multiple tasks, males showed higher win-stay behaviour than females, indicating greater exploitation of previously rewarded actions, an effect that was modest in size but highly significant. Power analyses revealed that >30 mice per sex were needed to detect these modest but significant sex differences with 80% power. No consistent sex differences were observed in pellet intake, suggesting that differences in exploitation did not reflect differences in hunger drive or demand for pellets. Exploitation is a more efficient strategy when environmental parameters are fixed, while exploration can be more advantageous when parameters such as reward locations are changing and uncertain. We tested this idea by re-running our mice in a probabilistic foraging task, where actions led to uncertain probabilities of reward. While males continued to show higher levels of win-stay behaviour on this task, this no longer led to increases in accuracy. Behavioural modelling also supported this framework, demonstrating that stronger win-stay behaviour was most advantageous in deterministic models, and less advantageous in probabilistic models. Together, our findings demonstrate that male and female mice have small but significant differences in their exploration-exploitation balance, which leads to more accurate foraging in certain, but not uncertain, environments.

PDFDOICodeData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
0
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (17 Sept 2026) · CC BY 4.0

In 136 mice on FED3 tasks, males showed slightly higher win-stay in all three choice tasks, with higher accuracy in a deterministic bandit but not a probabilistic one.

Earlier rodent studies conflict on which sex relies more on exploitation, possibly because of differing task designs or small, underpowered groups. Using closed-economy home-cage testing to reach 62 females and 74 males, this work measures the sex difference in win-stay across three choice tasks (plus a progressive-ratio motivation test), tests whether it helps under deterministic versus probabilistic reward, and estimates the group size needed to detect it.

Key results

  1. ShownMales returned to a just-rewarded nose-poke (win-stay) more often than females in all three choice tasks. In Bandit 100:0, staying after an unrewarded poke (lose-stay) did not differ by sex.

    Win-stay, M vs F: Bandit 100:0 82 ± 0.7% vs 77.1 ± 0.9% (p = 0.001); FR1 92.4 ± 0.5% vs 86.9 ± 1% (p < 0.001); Bandit 80:20 88.8 ± 0.6% vs 84.6 ± 0.7% (p < 0.001). Lose-stay in Bandit 100:0 53.1 ± 0.8% vs 52.5 ± 0.9% (p = 1.0). Previous-reward odds ratio in a 5-trial-back logistic regression 2.89 ± 0.14 (M) vs 2.24 ± 0.13 (F). · Results; Figure legends

  2. ShownMales were more accurate in the deterministic Bandit 100:0 and FR1 tasks, and this advantage held steady across days, so it did not reflect faster learning. It disappeared in the probabilistic Bandit 80:20 task.

    Accuracy, M vs F: Bandit 100:0 63.9 ± 0.6% vs 60.6 ± 0.5% (p = 0.001); FR1 74.5 ± 1% vs 66.7 ± 1% (p < 0.001); Bandit 80:20 59.1 ± 0.7% vs 59.1 ± 0.6% (p = 1.0). Mixed-model ANOVA over the 3 Bandit 100:0 days: Sex F[1,134]=16.5, p=0.0001; Day F[2,268]=157, p<0.0001; no Sex-by-Day interaction (F[2,268]=0.49, p=0.61). · Results

  3. ShownMeasures of food motivation did not differ by sex, which argues against motivation as the explanation for the win-stay difference. Progressive-ratio intake, breakpoints and demand-curve parameters were similar, and win-stay did not change between day and night.

    PR1, M vs F: pellets 118 ± 2 vs 117 ± 4; median breakpoint 9.5 ± 0.7 vs 9.1 ± 0.5; demand-curve alpha and slope p = 1.0. Bandit 100:0 pellets/day 179 ± 3 vs 176 ± 3 (p = 0.76). Male win-stay 82.3% (day) vs 81.9% (night), p = 1.0. Exception: males took more FR1 pellets (187 ± 3 vs 166 ± 3, p < 0.001). · Results; Demand curve modelling for the PR1 task

  4. ShownDetecting the win-stay sex difference with 80% power took about 30 mice per sex. Groups of 10 per sex, which the authors call common in behavioural neuroscience, had about 20% power.

    Win-stay data were resampled 1,000 times at each N from 4 to 60 per group: about 30 mice per group gave 80% power, and 10 per group gave 20%. An analytic power analysis with statsmodels solve_power gave largely overlapping curves. · Results; Power analyses

  5. SuggestedIn a simulated win-stay lose-shift agent, raising win-stay improved accuracy strongly under deterministic reward and less under probabilistic reward. The simulations overlapped the mouse data.

    Lose-shift was fixed at 0.5, lapse fitted to Bandit 100:0 data at 0.25, and win-stay varied from 0.5 to 1.0. There were 100 simulated mice per reward pair (100:0 to 60:40), 300 trials each. Agents overlapped individual mice in Bandit 100:0 and 80:20, with weaker win-stay/accuracy relationships for 80:20. Fit was described qualitatively; no fit statistic is reported. · Results; Behavioural modelling

Methods, data and reproducibility

Single-housed mice earned all food from an in-cage FED3 dispenser in four tasks over two weeks (Bandit 100:0 reversing every 20 pellets, fixed-ratio 1 (FR1), Bandit 80:20, resetting progressive ratio). Win-stay, accuracy and intake were compared by sex, alongside a win-stay lose-shift (WSLS) agent simulation, PCA and logistic-regression sex classification (71.3% accuracy), demand curves and power analyses.

  • C57BL/6J mice aged 8-16 weeks, single-housed; non-mutant littermates from 8 WUSMAC breeding lines, each tested on four FED3 home-cage tasks · mouse · n = 136 (62 female, 74 male)
  • Simulated win-stay lose-shift agents, 300 trials each, at reward pairs 100:0, 90:10, 80:20, 70:30 and 60:40 · n = 100 simulated mice per reward pair
Methods & tools
FED3 (Feeding Experimentation Device 3), Python 3.6.7, Pandas, Matplotlib, Seaborn, Pingouin, statsmodels (solve_power), win-stay lose-shift (WSLS) agent model, principal component analysis, logistic regression, behavioural economic demand-curve model
Code
https://github.com/KravitzLab/Murrell2026
Data & accessions
https://github.com/KravitzLab/Murrell2026
Software & environment
Python 3.6.7 with Pandas, Matplotlib, Pingouin and Seaborn; statsmodels for power analysis. The text gives no library versions or environment file.
Reproducibility gaps
Only the Python version is stated. Parental genotypes are in Table S1, which is not included in the text.
Limitations (6)
  • Each task ran for at most 3 days (on average 859 Bandit 100:0 and 611 Bandit 80:20 trials), so mice may not have fully learned the task structure. The simple WSLS model may also miss strategies that longer training could reveal. · stated by the authors
  • The authors read higher win-stay as exploitation, but it could instead reflect learning rate, repetitive behaviour or reward sensitivity. Exploration itself was not tested directly, so the data do not show that females are more exploratory. · stated by the authors
  • Estrous cycle was not tracked. The authors argue that the effect's consistency across tasks run over 2 weeks makes the cycle an unlikely main driver, but say dedicated experiments would be needed. · stated by the authors
  • The mice were non-mutant littermates from 8 WUSMAC lines, with a heterozygous parent and siblings carrying mutations linked to developmental disorders or schizophrenia. A lineage-by-sex interaction cannot be ruled out, though the accuracy difference went the same direction in 6 of 8 lines. · stated by the authors
  • All testing used single-housed C57BL/6J mice on specific FED3 tasks, so the results may not carry over to other strains, group housing or other task designs. · stated by the authors
  • The 71.3% sex-classification accuracy is reported without any described cross-validation or held-out set. Feature counts are also inconsistent (7 metrics per task, '14 features' in a legend, 8 selected), which makes the classifier hard to assess. · noted in this summary

Who should read:Read this if you design rodent bandit or reversal experiments and want effect sizes, group-size estimates, or open home-cage FED3 data on sex differences in win-stay strategy.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISChen '20Anversa '25Chen '20Kutlu '19Schuler '25Warner '19
82% semantic matchAnimal Behavior & Cognition
C. S. Chen, E. Knep, A. Han+2 more·bioRxiv·Dec 29, 2020