Resource sharing is sufficient for the emergence of division of labour

The study addresses how division of labour can originate in animal groups without pre-existing interindividual differences. Traditional response threshold models explain task specialisation by assuming stable differences in thresholds among individuals; however, these require pre-existing heterogeneity. The authors propose a model where identical individuals’ nutrition levels decline over time, and low nutrition triggers foraging. In isolation, individuals would alternate between foraging and nursing. In social groups, resource sharing frequently occurs, potentially interrupting this alternation and yielding specialisation. The research question is whether resource sharing alone is sufficient to generate division of labour in initially identical individuals, and how metabolic and dominance-related factors influence this process.

Division of labour is widespread across taxa, including eusocial insects, cooperative birds, cichlids, and lions. Response threshold models have historically been used to explain task specialisation, relying on interindividual differences in task thresholds influenced by genetics or development. Empirical and theoretical work has suggested links between nutrition and task performance in social insects, and that group size can affect specialisation, though evidence is mixed. Resource sharing is common in animal societies (e.g., trophallaxis in insects, food sharing in bats) and central in some microbial models, suggesting it might play a role in emergent specialisation. The present work builds on these insights by removing assumed threshold heterogeneity and focusing on nutrition-driven decision rules and resource exchange between individuals.

The authors developed an individual-based simulation model in continuous time representing a group of N individuals (default N=100; varied for group-size analyses). Each simulation runs for T=10,000 time steps. Individuals occupy one of two states: foraging or nursing. Nutrition dynamics: Each individual has a nutrition level n in [0, N_max] with N_max=100; all start at N_init=50. Metabolic rates reduce nutrition over time at m_for when foraging and m_nur when nursing (default m_for=1.0, m_nur=1.0; varied to test task-specific metabolic costs). Task choice: Individuals become more likely to forage as nutrition declines. A fixed threshold μ=50 determines the switch on average. Individuals perceive nutrition with noise: perceived n is drawn from a normal distribution centered on true n with standard deviation σ=1. Foraging and nursing durations: Foraging trips last t_for=5 time steps and return R=10 resources. After a trip, the individual decides to forage again or switch to nursing based on perceived n relative to μ. Nursing individuals that receive resources spend t_nur=5 time steps processing food, during which they cannot receive more; after processing, they reassess and either continue nursing or start foraging. Resource-sharing scenarios upon a forager’s return: (1) No sharing: forager keeps all resources. (2) Equal sharing: forager shares resources equally among itself and i nursing individuals (default i=1; sensitivity explored in Supplementary Material), so each receives R/(1+i). (3) Dominance-based sharing: each individual is assigned a fixed dominance value d sampled uniformly from [0,1] at initialisation. Resources are split among interacting individuals using a softmax over dominance values with parameter s (default s=1), giving individual k a share S_k = exp(d_k s) / sum_i exp(d_i s). Positive s biases resources toward more dominant individuals; s=0 reduces to equal sharing. (4) Nutrition-based sharing: dominance is dynamic and proportional to the individual’s current nutrition (d_k = n/N_max); resource allocation uses the same softmax as in (3) with s=1. If no eligible nursing recipients are available, the forager consumes all resources regardless of scenario. Quantifying division of labour: The division-of-labour metric D (Duarte et al.) measures within-individual task consistency adjusted by the baseline frequencies of tasks. D ranges from −1 (strict alternation) through 0 (random switching) to +1 (full specialisation). Metrics were computed over the last 10% of simulation time to avoid initialisation effects. Simulation design: The authors ran 20 replicate simulations for each sharing scenario; additional simulations varied metabolic rate differences, task durations, dominance parameters, and group size (for group-size figure, n=10 replicates per size). Implementation: Model in C++ (g++ 9.3.0); analysis and visualisation in R 4.1.0.

• Resource sharing is sufficient for division of labour: Without sharing, individuals alternate strictly between tasks (D = −1). With equal sharing between a returning forager and a nursing individual, division of labour emerges with D ≈ 0.6, indicating sustained task specialisation. Nutrition levels of nurses tend to be higher than foragers. - Metabolic rate differences reinforce specialisation: When nursing has slightly lower metabolic costs than foraging (e.g., 90% or 95% of foraging), division of labour strengthens and nutrition levels become bimodally distributed; if nursing is costlier than foraging, specialisation still occurs but is weaker. Shorter nursing duration relative to foraging similarly reinforces specialisation. - Dominance effects: Pre-assigned dominance-based sharing yields maximal division of labour (D = 1) and bimodal nutrition distributions. Highly dominant individuals tend to nurse, low-dominance individuals tend to forage, with intermediate dominance showing stochastic outcomes early on. - Nutrition-based dominance (no pre-existing differences): When dominance depends on current nutrition, initially identical individuals still achieve D = 1 with rapid divergence into bimodal nutrition states after small differences arise. Even altruistic variants (favoring low-nutrition recipients) yield lower but nonzero division of labour. - Group size: Division of labour emerges across group sizes; in very small groups, strength can depend on details such as parity (even vs odd), and can be particularly strong in pairs. - Replication details: 20 replicate simulations per sharing scenario; for group-size analysis, 10 replicates per size.

The findings demonstrate that resource sharing interrupts the default alternation between foraging and nursing driven by nutrition dynamics: fed nurses delay foraging, while foragers that share fail to fully replenish and thus reinitiate foraging sooner. These feedbacks create self-organised task specialisation among initially identical individuals, addressing the question of how division of labour can emerge without pre-existing differences. The work contrasts with response threshold models that assume stable individual threshold differences. It yields testable predictions: if specialisation stems from identical decision rules modulated by nutritional state, moving individuals between groups should not preserve roles once nutritional states are equalised, unlike predictions from stable-threshold models. The model highlights the amplifying roles of task-specific metabolic costs and dominance relations (including nutrition-based dominance) in strengthening division of labour, aligning with empirical links between nutrition, dominance, and task allocation in social insects. Resource sharing is proposed as a general mechanism potentially relevant beyond animals, including microbial systems and multicellular organisation, where shared resources can stabilise differentiated roles.

The study shows that resource sharing alone can generate strong, self-organised division of labour among identical individuals. Equal sharing is sufficient to produce sustained specialisation, and dominance or nutrition-based asymmetries can drive maximal specialisation (D = 1). Task-specific metabolic costs further reinforce differentiation, and effects are robust across group sizes. The results suggest a parsimonious, widely applicable mechanism for the emergence of division of labour in biological systems, with implications for understanding eusociality and multicellularity. Future work could experimentally test model predictions contrasting threshold heterogeneity versus nutrition-driven feedbacks, incorporate evolutionary dynamics to study how such non-evolved differentiation feeds into adaptive evolution, and explore more complex interaction networks, sharing topologies, and environmental variability.

• Evolutionary dynamics are not modeled; mechanisms (hunger-induced foraging, sharing rules) are fixed over simulations, limiting conclusions about the evolution of division of labour. - The model considers two tasks (foraging and nursing) with fixed durations; real systems may involve multiple tasks and variable durations. - Resource sharing interactions are simplified (e.g., i=1 equal-sharing partner by default); small-group outcomes can depend on details such as group parity and interaction structure. - Dominance-based sharing with pre-assigned dominance relies on initial interindividual differences; although addressed by the nutrition-based dominance variant, other sources of heterogeneity and dynamic social structures are not explored. - Nutritional assessment noise is simplified (Gaussian), and environmental/resource dynamics are not explicitly modeled beyond fixed returns per foraging trip.