Fitness drivers of division of labor in vertebrates

Our model consists of a population characterized by overlapping generations residing in a habitat featuring a fixed finite number of breeding territories. These territories are monopolized by groups composed of a single breeder and a variable number of subordinate helpers, the number of which is shaped by population dynamics, with all group members capable of reproducing during their lifetime. Individuals in the population fall into one of three categories: (1)

breeders

that monopolize reproduction within a group; (2)

subordinates

that may perform different helping tasks, either in their natal group or in an outside group to which they dispersed; and (3)

floaters

that disperse from their natal group but remain not part of a group.

We assume complete reproductive skew such that subordinate group members are prevented from reproducing at a given timestep and may only help raise the offspring of the dominant breeders. Dispersers cannot reproduce without acquiring a territory (denoted here as floaters). Helpers are not sterile and may inherit the breeding position when the breeder dies. When this occurs, helpers and dispersers from other groups compete for the breeding position and win with a probability proportional to their dominance value (

Figure 1

;

Table 1

). We use the term dominance value to designate the competitiveness of an individual compared to other candidates in becoming a breeder, regardless of group membership. Dominance value increases as a function of age, serving as a proxy for resource holding potential (RHP), and decreases as a function of help provided, reflecting costs to body condition from performing working tasks (

Equation 2

).

Subordinate helpers in our model may engage in one of two distinct tasks with differing fitness costs: (1)

defensive tasks

(i.e. defense of the dominant breeders’ offspring at cost to their own survival); or (2)

work tasks

(i.e. tasks with an associated cost to their body condition and consequently to their competitiveness to become breeders in the future). Subordinate helpers can then evolve different workloads and exhibit task specialization that is either fixed throughout their lifetime or varies with their dominance value. The parameters

x

_h


and

y

_h


(

Table 1

) reflect how much the subordinates’ survival and dominance values decrease with increasing investment in defense or work tasks, respectively. Although we do not assume an inherent difference in body condition between individuals when born, variance in age and helping propensities can create differences in the fitness costs linked to task performance.

We assume net survival benefits from living in large groups (i.e. group augmentation), which occurs in many social species (

Rood, 1990

;

Taborsky et al., 2005

;

Bilde et al., 2007

;

Guindre-Parker and Rubenstein, 2020

). Subordinate members of a group may therefore gain both direct and indirect fitness benefits from helping to increase the productivity of breeders, if doing so results in an increase in group size. In the model, the benefit of alloparental care to the productivity of the breeder is greatest when group members provide overall similar workloads for both defensive and work tasks, since both are often needed for the survival of the offspring (

Rubenstein et al., 2016

;

Taborsky, 2016

). For example, in many cooperatively breeding birds, the primary reasons that breeders fail to produce offspring are (1) starvation, which is mitigated by the feeding of offspring, here considered as a work task, and (2) nest depredation, which is countered by defensive behavior (

Hatchwell, 1999

;

Rubenstein et al., 2016

). Consequently, both types of tasks are often necessary for successful reproduction, and focusing solely on one while neglecting the other is likely to result in lower reproductive success than if both tasks are performed by helpers within the group. To compare between the helper’s preferred task and the task needed to maximize group productivity, we created a benchmark model in which the breeder’s productivity is not constrained by the type of help provided. Higher productivity translates to larger group sizes, which enhances the survival prospects of all group members. Hence, subordinates have an incentive to develop division of labor. Subordinates that remain in their natal group may also gain improved fitness by indirectly passing on their genes through kin. Both group size and relatedness in our model are emergent properties of a series of demographic processes, including mortality, dispersal, and help-dependent breeder fertility.

To distinguish between the effects of group augmentation and kin selection on inclusive fitness, we created two additional models. In the first model, we removed the benefits of group size (group augmentation) by setting the parameter

x

_n


= 0 (

Table 1

). In the second model, we removed the effect of kin selection by simulating a cross-foster manipulation in which offspring just born are interchanged between different groups (see Materials and methods for additional details). To assess how habitat quality affects task specialization and the emergence of division of labor, we systematically varied habitat quality across simulations. Harsh environments are those that have high mortality (i.e.

m

= 0.3;

Table 1

), which also raises the turnover of the breeding positions. Lower values of

m

reflect an increased probability of individuals surviving another breeding cycle and, therefore, the chances of habitat saturation. The steps in the breeding cycle of the model are illustrated in

Figure 1

.

We first examined how both direct fitness benefits from group living and indirect fitness benefits from aiding kin influence the evolution of division of labor through task specialization across environments of varying quality. Our results suggest that voluntary division of labor involving tasks with different fitness costs is more likely to emerge initially because of direct fitness benefits (

Figure 2

; GA), but a combination of direct and indirect fitness benefits leads to higher rates and more stable forms of division of labor (

Figure 2

; KS+GA). It is important to note that, as depicted in

Figure 2

, intermediate values of task specialization indicate in all cases age/dominance-mediated task specialization (

γ

_R


≠ 0;

Table 1

) and never a lack of specialization (

γ

_R

=

0;

Table 1

); further details are shown in the next section: ‘

The role of dominance in task specialization

’.

In contrast to expectations from models of eusocial insects, our vertebrate model does not readily evolve division of labor when only kin selection is considered (see Appendix 2 for more details on further exploration on conditions that may favor the evolution of division of labor under only kin selection). Specifically, forms of help that impact survival rarely evolve under any environmental condition when only kin selection occurs (

Figure 2

; KS). This occurs because with kin competition for the breeding position and no survival benefits of group membership, the reduced incentives to remain in the natal group drive most subordinates to disperse and breed elsewhere (average dispersal rate between 88% in harsh environments and 93% in benign environments;

Table 2

). Adding a reaction norm of dispersal to dominance rank did not change the results (see Appendix 3). The few subordinates that remain in the natal territory help to obtain indirect benefits (average within-group relatedness between 0.36 in benign environments and 0.72 in harsh environments;

Table 2

). Even though individuals in our model cannot adjust their help according to their relatedness to the breeder, the high levels of relatedness that evolved under this paradigm indicate that allowing for targeted help would not qualitatively change the results. Survival costs of defensive tasks are then avoided, as mortality without group benefits is relatively high, while costs to dominance value are small for close relatives because of indirect fitness benefits and low conflict of interest (

Figure 2

; KS). Adjusting input parameters to reduce defense costs on survival or to increase survival probabilities and favor philopatry did not qualitatively change the results, except for extreme values of baseline survival (

Appendix 2—figures 1

and

2

). Only when (1) there is a tenfold increase in the productivity of the breeders per unit of help, and (2) division of labor is required for a productivity boost, do group members stay philopatric and coordinate labor to help raise kin in some circumstances (

Appendix 2—figure 4

). Hence, kin selection alone is unlikely to select for the evolution of defensive tasks and division of labor in vertebrates.

Conversely, under only direct fitness benefits (group augmentation), tasks associated with costs to the dominance value are costlier because the only path to obtain fitness benefits is via direct reproduction. Therefore, defensive tasks that do not affect the dominance value are favored (

Figure 2

and

Appendix 4—figure 1

; GA). Under this scenario, division of labor can evolve to maximize the reproductive output of the breeders (

Figure 2

; GA,

m

= 0.3,

x

_h


≤ 5), but also because it is the optimal strategy (

Figure 2

; GA,

m

= 0.3,

x

_h


= 7). In benign environments, however, no help evolves due to the lower incentives to increase survival and the steep competition for reproduction resulting from habitat saturation.

When subordinates can obtain both direct and indirect fitness, division of labor is more strongly selected in increasingly harsher environments to increase breeder productivity, a result mainly driven by an increase in direct fitness benefits (

Figure 2

; KS + GA). Conversely, when direct fitness benefits are low, division of labor primarily emerges in benign environments from a combination of both direct and indirect fitness benefits (

Appendix 4—figure 1

; KS + GA). Additionally, in moderate and harsh environments, work tasks with costs to the dominance value are more strongly selected than in benign environments due to high mortality increasing the breeder’s turnover, which reduces breeding competition (lower effective costs of work tasks), and the costs to survival (high costs of defensive tasks), while defense tasks are mainly promoted by direct fitness benefits and in more benign conditions (

Appendix 4—figure 1

). Hence, in more benign (and often highly productive) environments that lead to habitat saturation, helping likely evolved initially in family groups, and defensive tasks are favored because competition for the breeding position is lower under kin selection. In this scenario, kin selection acts not only by favoring subordinates in their natal group to increase the productivity of a related breeder (

Figure 2

; KS), but also by the increased survival of their siblings when augmenting the group size (

Figure 2

; KS + GA), regardless of whether they are related to the current breeder. In harsher environments, however, competition between unrelated individuals decreases in favor of coordinated cooperation to reduce mortality (

Figure 2

and

Appendix 4—figure 1

; GA). Together, these patterns indicate that direct fitness benefits are the primary force driving cooperative specialization in harsh environments, where increased breeder productivity boosts the survival of both helpers and their relatives (

Figure 2

). In contrast, in benign environments, cooperation arises mainly through kin-selected benefits, with subordinates contributing primarily by performing their preferred task of defending the breeder’s offspring (

Figure 2

).

In the previous section, we explored the role of direct and indirect fitness benefits in the evolution of division of labor via task specialization. However, determining whether parameter

γ

_R


(i.e. the direction and rate of change in task specialization with dominance) significantly differs from zero is crucial. Intermediate values of phenotypic task specialization may indicate division of labor among different age- or dominant-related groups (if

γ

_R


≠ 0), or lack of specialization in either task (if

γ

_R

=

0). Hence, for age-polyethism or temporal division of labor to arise in our model, individuals must switch between tasks based on their age-related dominance (probability of choosing a ‘defensive task’ versus a ‘work task’

T

≠ 0 or 1, and

γ

_R


≠ 0;

Table 1

). Across all scenarios, individuals are selected to perform tasks based on their probability of inheriting the territory (

Figure 3A

). Specifically, when subordinates prefer defensive tasks (

Figure 3A

,

m

< 0.2, circles), they engage in work tasks at a young age and increasingly undertake defensive tasks as their dominance rises (

Figure 3

,

m

< 0.2). Similarly, if their preference leans toward work tasks (

Figure 3A

,

m

≥ 0.2, circles), younger subordinate helpers and those of lower dominance focus more on defense initially, but switch to work tasks as they approach the breeding stage (

Figure 3

,

m

≥ 0.2). In essence, subordinate helpers allocate more time to tasks with lower fitness costs as they near a competitive dominance value, all without any active enforcement by other group members, resulting in age- or rank-polyethism. These results are consistent with the rise in division of labor, defined as within-individual consistency and between-individual differences in task choice (

Ulrich et al., 2018

).

We model a population consisting of a fixed number of breeding territories, each comprising a dominant breeder that reproduces asexually and monopolizes reproduction, as well as an undefined number of subordinates that queue for the breeding position. The number of subordinates is determined by the productivity of the breeders, the dispersal decisions of the group members, and immigration. Subordinates may help increase the fecundity of the breeder, so that the breeder’s fecundity depends on the cumulative level of help provided by the subordinates within the group, with an optimum increase when different helping tasks are performed to a similar degree.

Offspring inherit the genes for dispersal and helping tendencies from the breeders, potentially exhibiting slight variation in the presence of mutations (

Figure 1

, Step 1). Each genetic trait is controlled by a single locus and may take any real number. Mutations occur at each genetic locus at a frequency of

μ,

causing minor adjustments to the allele’s value inherited from the parent. This alteration involves the addition of a value drawn from a normal distribution characterized by a mean of 0 and a standard deviation of

σ

_μ


(

Table 1

).

After breeders reproduce, their offspring and the subordinates in the group may disperse independently from each other. Dispersal propensity is controlled by a gene with values ranging from 0 to 1, where 0 represents certain philopatry and 1 represents absolute dispersal (

Table 1

). In each breeding cycle, dispersers may migrate to another group to become subordinates or remain as floaters waiting for breeding opportunities, which is also controlled by the same genetic dispersal predisposition as subordinates (

Figure 1

, Step 2). Successful dispersers enter the breeding queue within the group based on their dominance value rather than starting at the bottom of the hierarchy. Subordinate group members, either natal individuals born in the group or immigrants to the group, express some level of help in the form of alloparental care, which can potentially evolve to zero (

Figure 1

, Step 3). To be conservative, we use the term help as synonymous with altruistic behavior, which increases the fitness of the beneficiary (i.e. breeder) with an immediate decrease in the actor’s fitness. Hence, cooperative behaviors such as defense from adult predators that incur immediate benefits to the actor as well as by-product mutualistic benefits to other group members, are not considered in this study. Dispersal propensity, alloparental care provisions, and the type of helping task they perform all depend on their genetic predisposition. They may perform tasks that reduce their body condition and, therefore, their chances of becoming breeders, such as feeding offspring or territory maintenance (hereafter called ‘work tasks’), or they may perform tasks that reduce their immediate survival such as predator defense (hereafter called ‘defensive tasks’) (

Heinsohn and Legge, 1999

). For simplification, we assume complete dichotomous, short-term non-cumulative fitness consequences between the different types of tasks (i.e. survival and dominance costs are only sustained during the breeding season in which they perform the task, and individuals only perform one of the tasks in the given cycle). We note that the names given for each type of task are meant to facilitate interpretation, and that in some species, some tasks classified as work tasks may have a higher immediate cost in survival than in the probability to breed and vice versa. However, helping tasks should be classified in empirical systems according to the primary fitness cost associated with each type of helping activity (e.g. if a work activity mostly negatively impacts survival in a given species, this task is described in this paper as defensive).

Individual survival depends on environmental quality, group membership, group size, and the amount of help provided during defensive tasks to the breeder’s offspring (

Figure 1

, Step 4). Environmental quality is expressed as a varying maximum survival probability, independent of social factors. Additional environmental effects, such as variation in mortality linked to dispersal, or on the probability of floaters finding a new group to start breeding, have been explored elsewhere (

García-Ruiz et al., 2022

).

If the breeder in a group dies, all subordinates in the group as well as a randomly drawn sample of floaters from outside the group compete for the breeding position (

Figure 1

, Step 5). The number of floaters bidding for an empty breeding spot is then proportional to the relative abundance of floaters with regard to the total number of breeding territories, to account for spatial viscosity and the probability of being accepted in a new group (

Table 1

). In this model, age influences the probability of attaining the breeding position. Older individuals, owing to their higher RHP, have a greater probability of becoming breeders, as observed in various species (

Parker, 1974

;

Shaw, 1986

;

Doolan and Macdonald, 1996

;

Brown et al., 1997

;

Buston, 2004

;

Bridge and Field, 2007

;

Kingma et al., 2011

;

Bang and Gadagkar, 2012

;

Taborsky, 2016

;

Unnikrishnan and Gadagkar, 2021

). Age in our model corresponds to the number of breeding cycles or seasons that an individual survives. Additionally, work tasks reduce an individual’s dominance value as a result of the impact on their body condition. The probability of filling an empty breeding position is then implemented as a lottery weighted by the dominance value of the candidates. Therefore, subordinates may stay in the natal group and queue to inherit the dominant position (

Woolfenden and Fitzpatrick, 1978

;

Russell and Rowley, 1993

;

Balshine-Earn et al., 1998

;

Dierkes et al., 2005

;

Bridge and Field, 2007

), but they may also disperse and breed or queue in another group (

Walters et al., 1988

;

Russell and Rowley, 1993

;

Zöttl et al., 2013a

). We ran simulations for 200,000 breeding cycles or until equilibrium was reached for all genetic traits, across 20 replicas to assess consistency.

The probability of choosing defensive versus work tasks

T

takes a logistic function with boundaries between 0 and 1 as given in

Equation 1

, in which individuals may adjust their helping task as they age and become more dominant:

where

γ

_R


modifies the strength and direction of the effect of dominance (

R

) on task choice, and

γ


₀

acts as the intercept. Therefore, when

T

approaches zero, individuals specialize in work tasks throughout their lives, and when

T

approaches one, they specialize in defensive tasks. If

γ

_R


≠ 0, given that

T

≠ 0 and

T

≠ 1, individuals show age-dependent task specialization.

In addition, the phenotypic expression of help (

H

) is regulated by another gene and equals the allelic value of gene

α

, which remains fixed throughout the helper’s life. Even though in nature individuals are likely to change the total amount of help given throughout their lives, we make this simplification to allow easier interpretation of how individuals may switch between task types as they age. How individuals in the model may adapt their level of help with age and social and environmental conditions has been described elsewhere (

García-Ruiz et al., 2022

). Hence, the gene

α

regulates the amount of help expressed, while the genes

γ

determine which specific helping tasks are performed at different time points in the breeding cycle. The fitness costs of helping are described in the following section.

Helping tasks are classified in our model according to their short-term fitness consequences. We assume that work tasks incur a more immediate cost to the body condition of helpers because of expenses in terms of time and energy. Work tasks reduce the dominance value

R

to inherit the breeding position in our model following

Equation 2

, where

γ

_h


is a parameter that influences the strength of the cost of help (

H

_work


) in the dominance value. We assume that dominance is influenced by age (

t

) as a proxy for RHP, with older individuals having a greater dominance value:

Note that the cost to an individual’s dominance value is influenced by their current age and the current specific task that they are performing, the cost of which is not cumulative over time. Therefore, an individual’s effective rank within the dominance hierarchy is, irrespective of whether they belong to a group or not, determined by their dominance value relative to other competitors for the breeding position, with higher dominance values corresponding to higher ranks in the hierarchy.

In contrast, we assume that defensive tasks have a more immediate fitness cost on survival (e.g. predator or space competitor defense) due to the high risk of injury. Therefore, subordinate helpers that defend have a survival probability (

S

_H


) that decreases with the amount of help provided (

H

_defense


) given by the logistic

Equation 3.1

, where

x

_h


is a parameter that influences the strength of the cost of help in survival:

In addition to defensive tasks, social and ecological factors also influence the survival of group members. The term

m

denotes the baseline mortality that determines how harsh the environment is. Higher values of

m

indicate higher overall baseline mortality for all individuals in the population, irrespective of social factors. For simplicity, we assume that the harshness of the environment only impacts survival, although other effects such as on reproductive output or the strength in which help improves fecundity are likely to have an effect in nature (

Rubenstein, 2011

). Group size (

N

) provides survival benefits, such as those offered by safety in numbers or increased resource defense potential, for both subordinates and breeders (

Equations 3.1 and 3.2

, respectively), where

x

_n


is a parameter that quantifies the effect size of the benefit of group size in survival, and

x


₀

is an intercept. Floaters sustain the highest mortality (

S

_F


,

Equation 3.3

) since they do not benefit from group protection:

To assess the rules governing task specialization and division of labor, we first outline a basic model where individuals evolve their preferred helping task. In this model, alloparental help increases breeder productivity regardless of how tasks are distributed among helpers. We then extend this framework to a second model in which the breeder’s reproductive outcome is maximized only when the group’s efforts are roughly balanced across the two types of tasks, thereby actively promoting division of labor. Thus, in the second scenario, the maximum effective contribution of cumulative group help for each task type (

H

_max


) that can enhance fecundity is given by

Equation 4

:

where

k

_m


is a parameter that relaxes the requirement for perfectly balanced task allocation, allowing fecundity to increase even if the contributions to defense and work are somehow uneven. The breeders then reproduce asexually. The total number of offspring produced (

K

) is contingent upon the breeders’ baseline fecundity (

k


₀

) and the cumulative helping effort of all subordinates. Following

Equation 5

, fecundity increases with total help but exhibits diminishing returns:

where

k

_h


is a parameter that scales the benefit of cumulative help on fecundity. To prevent exceeding the maximum effective help, any cumulative defense or work effort that surpasses

H

_max


is truncated at

H

_max


(i.e. if
$$



∑



i


=


1




n







H






w


o


r


k




i








>

H

_max


or
$$



∑



i


=


1




n







H






d


e


f


e


n


s


e




i








>

H

_max


). The breeder’s fecundity is then determined by a randomly selected value drawn from a Poisson distribution with a mean of

K

.

We include both direct and indirect fitness benefits of helping to assess the relative importance of each in the evolution of task specialization. Indirect fitness benefits from relatedness among group members (i.e. kin selection) emerge from demographic factors such as dispersal tendencies, mortality rates, and group dynamics, though they were not coded explicitly. To calculate the coefficient of relatedness between the breeder and subordinates in a group, we calculated the coefficient of a linear regression between the allelic values of the breeders and helpers of a neutral gene that changes exclusively by genetic drift (

Michod and Hamilton, 1980

;

Gardner et al., 2011

). Direct benefits may also be obtained by enhancing group size by helping to increase the breeder’s fecundity, as larger groups benefit from higher survival probabilities in our model (i.e. group augmentation).

To distinguish the effect of kin selection from group augmentation, we created two parallel models for comparison to the default model (group augmentation and kin selection acting independently), in which one of the fitness benefits was removed. In the ‘only group augmentation’ implementation, individuals just born (i.e. age = 1) that decide to stay in the natal territory as subordinates (

Figure 1

, Step 2) are shuffled into another group, thereby removing the influence of relatedness from the model. In the ‘only kin selection’ implementation, we removed the survival benefits linked to living in a group (

x

_n


= 0).

To determine whether variation in the cost of work tasks (

y

_h


) yielded qualitatively similar outcomes comparable to those observed when varying the cost of defense tasks (

x

_h


), we systematically varied

y

_h


while holding

x

_h


constant (

Appendix 1—figure 1

). As predicted, our results show that an increase in cost to rank promoted individuals to choose more defense versus work tasks. Our conclusions about the role of direct and indirect fitness benefits in shaping the evolution of division of labor remain unchanged.

To assess the robustness of our findings on the evolution of division of labor driven by kin selection independent of group survival benefits, we further examined the influence of survival, dispersal, relatedness, and breeding opportunities outside the natal group.

One factor contributing to defense not readily evolving under kin selection alone is that the cost to dominance is relatively small when competing against relatives. Another possible explanation is that without group benefits, mortality rates are too high for survival-costly helping tasks such as defense to evolve. In

Appendix 2—figure 1

, we examine the effects of increasing the baseline survival from

x


₀

= 1.5 (as in

Figure 2

) to

x


₀

= 4.5 (which simulates the addition of a group benefit of

x

_n


= 3 to the baseline survival) and

x


₀

= 10 (an extreme scenario that greatly reduces the survival cost of defense tasks; see

Equation 3.1

).

Our results indicate that even with high baseline survival, kin selection strongly favors work over defense tasks. However, in very harsh environments (

m

=0.3), where the survival costs of defense tasks are minimal (

x


₀

= 10 and

x

_h


= 3;

Appendix 2—table 1

), some degree of task specialization and division of labor emerges (

Appendix 2—figure 1

). Despite this, most individuals still choose to disperse and breed outside their natal territory when no group survival benefits exist (

Appendix 2—table 1

). This suggests that, without such benefits, offspring have little incentive to stay and help due to reduced reproductive opportunities and competition with kin. To test this, we eliminated the dispersal advantage by reducing the number of sampled groups by dispersers from two (

f

= 2) to one (

f

= 1), incentivizing individuals to queue for breeding in their natal territory (

Appendix 2—figure 2

).

Even when increased philopatry, resulting from reduced outbreeding opportunities for floaters, potentially enhances the indirect fitness benefits of helping and task partitioning, division of labor only evolves when survival costs are exceedingly low (

x


₀

= 10;

Appendix 2—figure 2

;

Appendix 2—table 2

). This suggests that while kin selection can promote the evolution of division of labor, it does so only under highly restrictive conditions with minimal survival costs. Therefore, in natural settings, other selective forces, such as direct fitness benefits, are more likely to drive the evolution of division of labor.

The strength of kin selection is likely more pronounced in our model than in natural populations due to the assumption of asexual reproduction. To assess whether the results remain valid with lower within-group relatedness, we shuffled half of the newly born individuals (i.e. age = 1) who choose to stay in their natal territory as subordinates, similar to the ‘

only group augmentation

’ implementation. The results indicate that a reduction in relatedness increases the probability of individuals favoring defense over work tasks, but the overall conclusions of our model remain unchanged (

Appendix 2—figure 3

).

So far, we have considered scenarios where alloparental care always increases breeder productivity, regardless of task partitioning. Because we assume parents provide parental care, reproduction is possible even without help, but productivity is maximized when helpers contributed similarly to both tasks, promoting rather than requiring division of labor. To explore whether kin selection alone can favor division of labor in situations where help boosts productivity

only

if all tasks are performed to a similar extent, we developed a new implementation of the model. In this implementation, the absence of task division prevents any productivity gains, thereby making division of labor obligatory. To achieve this, instead of relying on

H

_max


(

Equation 4

), we introduce an

H

_min

and compare it to the maximum values of the cumulative help for each task, allowing a proportional difference

k

_p


from the mean. Hence, if

then

K

=

k


₀

, where costs associated with help performance are still inflicted to the actor. The results of this implantation of the model show that help and division of labor, as well as philopatry, do evolve only under kin selection benefits only if the increase in productivity per unit of help is very high (

Appendix 2—figure 4

). Therefore, kin selection alone can promote division of labor only under restrictive conditions: when alloparental care produces a substantial increase in productivity and when both types of tasks are required in similar measure for successful offspring production and survival. These constraints suggest that grouping benefits are more likely to play a key role in the emergence of division of labor. In addition, due to the need for coordination between group members, division of labor only emerged when we assumed a philopatric starting point (

β

_init


= 0.5).