Actual versus counterfactual fitness consequences of dispersal decisions in a cooperative breeder

In many species, dispersal from the natal group is crucial for reproduction. However, venturing into a new territory and integrating into a novel social environment is associated with risks and costs. In cooperatively breeding species, moreover, an individual’s prospects of future reproduction upon dispersal or upon remaining in the natal group might crucially depend on other group members’ concurrent decisions to disperse or to stay. We developed a methodology for evaluating how the actual decision of a potential disperser to join or not to join a dispersing individual or coalition affects its fitness in comparison with the fitness consequences it would have had if it had taken the reverse decision – to disperse instead of staying, or to stay instead of dispersing. We then examined 64 dispersal events by unisex coalitions of Arabian babblers who could not breed in their group of origin, and aimed to acquire breeding opportunities by joining another group. For each such dispersal event, we compared the fitness consequences for the members of the dispersing coalition as well as for their same-sex siblings who stayed, to the counterfactual consequences of taking the reverse decision for each of them in turn. Fitness consequences were assessed based on breeding success in the ensuing year as the leading criterion, and on social rank as a secondary criterion. We found that 69% of the dispersers and 38% of the individuals who stayed made fitness-enhancing decisions relative to the alternative they faced, and for an additional 10% of dispersers and 21% of those who stayed, their choice yielded fitness consequences on par with those of the alternative choice. These findings suggest that despite the risky and uncertain circumstances in which dispersal decisions are taken, most individuals make informed, fitness-enhancing dispersal choices, taking into account the concurrent choices of their groupmates.

In many species, dispersal from the natal group is crucial for reproduction. However, venturing into a new territory and integrating into a novel social environment is associated with risks and costs. In cooperatively breeding species, moreover, an individual's prospects of future reproduction upon dispersal or upon remaining in the natal group might crucially depend on other group members' concurrent decisions to disperse or to stay. We developed a methodology for evaluating how the actual decision of a potential disperser to join or not to join a dispersing individual or coalition affects its fitness in comparison with the fitness consequences it would have had if it had taken the reverse decision -to disperse instead of staying, or to stay instead of dispersing. We then examined 64 dispersal events by unisex coalitions of Arabian babblers who could not breed in their group of origin, and aimed to acquire breeding opportunities by joining another group. For each such dispersal event, we compared the fitness consequences for the members of the dispersing coalition as well as for their same-sex siblings who stayed, to the counterfactual consequences of taking the reverse decision for each of them in turn. Fitness consequences were assessed based on breeding success in the ensuing year as the leading criterion, and on social rank as a secondary criterion. We found that 69% of the dispersers and 38% of the individuals who stayed made fitness-enhancing decisions relative to the alternative they faced, and for an additional 10% of dispersers and 21% of those who stayed, their choice yielded fitness consequences on par with those of the alternative choice. These findings suggest that despite the risky and uncertain circumstances in which dispersal decisions are taken, most individuals make informed, fitness-enhancing dispersal choices, taking into account the concurrent choices of their groupmates.

INTRODUCTION
For many large animal populations, a basic ecological tenet is that individuals disperse away from their natal territory according to an ideal free distribution that equates everybody's fitness prospects (Fretwell 1969). However, in populations consisting of small social units, such equalisation may not be feasible when the higher ranked bar mating opportunities from the lower ranked. The result is a "despotic" distribution (Fretwell 1972) in which not everybody has the same chance to breed. In each patch or territory there may arise a queue for breeding (Kokko & Sutherland 1998), so a decision to disperse constitutes a "queue switching". Moreover, such queue switching is not a once-and-for-all decision, but can occur several times along an individual's lifetime. Each decision is then one out of numerous "career decisions" (Ens et al. 1995).
On the face of it, if an individual that cannot breed in their natal group manages to breed following dispersal, then dispersal might seem beneficial. Similarly, if an individual manages to breed without dispersing, then avoiding dispersal and the associated costs might seem to be a good choice. However, a good outcome might still be inferior to the outcome that could have ensued had the opposite choice been made; and a bad outcome may still be superior to the outcome that would have resulted from the opposite choice.
Such counterfactual outcomes are oftentimes hard to estimate, but sometimes such assessments are possible. For example, with an inventive experimental design Bonte et al. (2014) managed to compare the fecundity of two-spotted spider mites (Tetranychus urticaemites) that dispersed to a new territory to their fecundity had they not dispersed, by repatriating emigrants back to their natal territory, and with the reverse translocation they compared the breeding success of philopatric mites with the breeding success they would have had if they were to disperse to the new territory. In this example, the actual outcome of the "career decision" to disperse or not to disperse, as well as the counterfactual outcome that would have come about with the alternative decision, both depend on the match of qualities of an individual to a habitat.
However, when it comes to queues for breeding, outcomes depend not only on the qualities of the individual and the habitat but also on the decisions taken by the individual's peers. In the current work, we propose a novel methodology for comparing the fitness consequences of dispersal vs remaining in the group of origin in territorial cooperative breeders, and apply it to dispersal decisions in the Arabian babbler (Argya squamiceps), a cooperatively breeding songbird species that lives in territorial groups in arid areas along the Rift Valley. An ongoing long-term study of this species is conducted in the Shezaf Nature Reserve in southern Israel, in which individuals are habituated to the proximity of non-interfering human observers, who can document their complex social behaviour.
Within each group, a rank order exists separately among the males and among the females. Higher rank typically corresponds to older age, including earlier hatching within a given brood, but this order is sometimes overturned by aggressive-submissive interactions among group members. A higher rank implies priority of access to different resources, such as food, roosting positions, and mating opportunities. However, even when an individual reaches sexual maturity, around the age of 1 year, a high rank does not necessarily promise breeding potential, because no individual copulates with opposite-sex individuals who were present in the group when it hatched ("incest avoidance"). For example, when the dominant male disappears from the group, its highest-ranked son becomes the dominant male. However, he will not be able to breed in the group as long as the only females in the group are his mother and sisters. The same applies for females when the dominant female disappears from the group (Zahavi 1990). In these instances, a non-breeding group member gets a chance to breed only if it disperses and joins another group, or, alternatively, if a foreign individual of the opposite sex joins its group. Thus, as in other cooperative breeders like pied babblers (Nelson-Flower et al. 2012) and meerkats (Harrison et al. 2021), dispersal enables subordinates to breed while avoiding inbreeding.
Despite the potential for increased reproductive output, the dispersal process is associated with a number of costs. Firstly, during dispersal Arabian babblers often experience a loss of body mass (Ridley 2012). Further, in the Arabian babbler, dispersal attempts typically encounter resistance in the target group where individuals of the same sex there are ultimately pushed down in the social hierarchy and sometimes even chased away from the group by the newcomers. In addition, maintaining a breeding vacancy within the target group frequently depends on the ability to successfully compete against babblers from other groups who may also try to attain the same breeding position. Even when the invading coalition succeeds to join the target group, oftentimes only the higher ranked individuals within the joining coalition will breed there. Survival is lower among dispersers than among individuals, which do not disperse, possibly due to the conflicts involved in settling into the target group and into a new territory, which is initially less familiar (Zahavi 1989). To mitigate these costs, Arabian babbles often disperse in same-sex sibling coalitions, who try to join a target group together and attain a breeding status there (Ridley 2012).
Thus, when a coalition of same-sex non-breeding individuals is about to disperse, other sexually mature non-breeding individuals of the same sex in the group face a dilemma: if they join the dispersing coalition, they may have a higher chance to breed in the short run, but at the cost of reduced survival; if they stay, their chances to breed in the short run might be small, but they may get an automatic promotion in the social ranking due the dispersal of their higher-ranked peers, and consequently better access to mating whenever individuals of the opposite sex join the group in the future. In parallel, the members of the dispersing coalition face the diametrically opposed tradeoff: by aborting the dispersal attempt they may enjoy rank promotion and better survival chances at the original group, while forgoing a higher chance to breed in the short run.
The advantages of staying that dispersers forgo are what Bonte et al. (2012) labelled as the "opportunity costs" of dispersal. In addition to other well-documented costs of dispersal in cooperative breeders, like reduced survival, reduced body mass, or increased stress (Koenig & Dickinson 2016;Maag et al. 2019), such opportunity costs need to be taken into account when evaluating the overall cost of dispersal. Each potential disperser's "career decision" to disperse or to stay is a particular case of an "informed dispersal" decision (Clobert et al. 2009), where the information consists also of others' dispersal decisions.
Since relative social rank in the Arabian babbler is maintained among the dispersers and also among non-dispersers (Zahavi 1990), the data make it feasible to assess the rank of each potential disperser if it were to take the reverse decision -to stay instead of dispersing or vice versa, and thus its relative ranking among breeders and non-breeders in the ensuing year if it were to make the opposite choice. This then enables us to estimate its breeding success in the counterfactual scenario, and to compare it to the actual breeding success that did materialise.
Either decision -to stay or to join the coalition of dispersers -involves uncertainty, so it is unrealistic to expect that in retrospect, for all individuals the choice not taken was always worse than the choice that the individual did take. Rather, our main hypothesis is that (i) for a significant majority of the individuals that dispersed, dispersing was fitness enhancing as compared to staying, and also that (ii) for a significant majority of the individuals that stayed, staying was fitness enhancing as compared to dispersing. By fitness enhancing, we mean that the choice enabled the individual to breed in the ensuing year whereas under the reverse decision the individual wouldn't have gotten an opportunity to breed given the choices of its higherranked peers; or, if breeding success or lack thereof would have been similar under either decision, the actual decision led to a higher rank than under the reverse decision, thus providing a better head-start towards breeding opportunities a year later.

Field methods
The study was conducted from January 1991 to December 2008 in and around the Shezaf Nature Reserve (30°46'N, 35°15ʹE), approximately 30 km south of the Dead Sea, in southeastern Israel. The study area, general field methods in the long-term project on the Arabian babblers at this site, and aspects of the biology of Arabian babblers have been summarised previously (Zahavi 1989(Zahavi , 1990Ostreiher 2001Ostreiher , 2003. In brief, throughout this period, a wild population of approximately 35 groups with 2-20 individuals per group was monitored. The babblers at this site were habituated to human presence, enabling observers to follow and watch them from close proximity. Each individual was ringed, while in the nest or shortly after fledging, with a unique combination of four coloured rings. Researchers visited each group 1 to 3 times per week throughout this period, so in particular the age of all group members was known with ± 3 days accuracy. Every observation began by making a list of all group members present. In the particular study reported here, we monitored the entire breeding process daily in a subset of the groups. Observing a breeding process of a group meant visiting it daily from the first sign of nest building through copulations, up to the beginning of incubation (babblers usually start to incubate after laying the last egg), and again from hatching to fledging. We determined paternity of each egg by direct observation of the copulations and identification of the copulating females and males 24 hr before the egg was laid. We marked each egg immediately after the laying female left the nest, so we knew the laying order as well as the hatching order. We also fitted a coloured collar on the neck of each nestling immediately after hatching, ringed the nestlings when they became 10 days old, and removed the collars at the age of 12 days, 2 days before fledging (for more details see Ostreiher 1997). In this study, we define breeding success when the offspring of an individual successfully fledges.
Individuals become sexually mature when they are around 1 year old, but might have no breeding potential in their group due to "incest avoidance": opposite-sex Arabian babblers do not copulate with one another if one of them had already been a member of the group when the other hatched. The current study focuses on dispersal events by unisex coalitions of sexually mature babblers with no breeding potential in their group, and dispersal events by a single sexually mature individual with no breeding potential in which we observed the dispersing individual leaving the group without being evicted from it. We did not include in the analysis events in which we witnessed an individual being chased away from its group, or discovered a single individual when it was already away from its group, and therefore could not rule out the possibility that the individual had been evicted from its group. In each dispersal event considered, the other relevant participants were the remaining potential dispersers (if any) -those sexually mature babblers of the same sex with no current breeding potential in the group, who nevertheless stayed in the group and did not join those which departed. In all considered events the dispersers either joined an existing group or perished, and no dispersal event led to the formation of a new group.
Individual rank before dispersal was determined based on the individuals' age as the benchmark, modified when at least 10 observed aggressive-submissive interactions between group members indicated that an individual attained a higher rank than that of an older sibling. When the dispersers managed to join a target group, immediately after joining they maintained the rank order that they had in their group of origin, and all same-sex resident individuals of the target group (if any) were pushed down in the social hierarchy below the newcomers. In cases where the dispersers disappeared after dispersal, possibly after some period as floaters without a territory, our assumption was that they have most likely perished, and therefore also did not breed following dispersal. This study was based on observations only and no experimental manipulations were used.
In this study, we analyse the dispersal events for which we recorded both (i) the dispersal event itself and (ii) all the breeding processes in the target group and in the group of origin, with the methods described above, during the year following the dispersal event. We restricted attention in this study to breeding 1 year following the dispersal event, therefore containing one breeding season in which at most three breeding cycles take place, because breeding later in life is harder to associate with the quality of a particular "career decision" to disperse or not more than a full year earlier.

Assessing the decision to disperse or stay as compared to the alternative (counterfactual analysis)
Four regularities exist without exception in our dataset. First, the relative social hierarchy before dispersal among members of a unisex coalition of dispersers persisted among them immediately after joining a target group. Second, the relative social hierarchy before dispersal among potential dispersers that stayed persisted among them in the group of origin after the dispersers departed. Third, if in the target group a disperser bred during the year following dispersal, so too did all of its higher-ranked co-dispersers (if any). Fourth, if in the group of origin a potential disperser that stayed bred during the year following dispersal, so too did all the higher-ranked potential dispersers that stayed (if any). The combination of these four regularities enabled us to assess for each of these individuals what its rank and its breeding success would have been if it were to make the opposite choice in the dispersal event -to stay instead of dispersing, or to disperse instead of staying. An example is presented in Supplemental Data.
From the individual's perspective, the choices of its groupmates were not under its direct control; the individual directly controlled, and could have potentially reversed, only its own decision. Given its peers' decisions, a potential disperser improved its fitness if its decision to disperse/stay led to breeding the following year whereas the opposite decision would not have; in contrast, the decision worsened its fitness if the decision led to a loss of a breeding opportunity (Table 1). If either decision would lead to breeding or lack thereof, our secondary criterion was the individual's attained rank as compared to the rank that would be attained under the reverse decision (Table 2), because higher rank implies a better head-start in competition over breeding in later years. We present overall fitness The merit of the decision to disperse (a) and stay (b) in terms of rank.  Table 3: better fitness results from either improvement in breeding or same breeding outcomes and better rank as compared to the alternative choice; worse fitness results from loss of breeding opportunity or same breeding outcomes and worse rank as compared to the alternative; and similar fitness results from same breeding outcomes and same rank under both alternatives.
As an example, in Tables S1-S3 (in Supplemental Data) we detail how the individuals in Fig. 1 are classified into Tables 1-3.

Statistical analyses
To test whether the decisions potential dispersers took (staying or dispersing) improved their breeding success or rank, as compared to if they had taken the opposite decision, we fitted three cumulative logit link mixed models (ordinal models; Agresti 2007). With one of them we addressed the change in breeding success (model 1), with one we addressed the change in rank (model 2), and with the third we addressed the combined fitness effect of a decision to disperse or stay (model 3). In all three models, the response was coded as − 1, 0, or 1, denoting whether an individual did worse, the same, or better, respectively, as compared to the counterfactual decision. The coding of the response variables corresponded to the entries in Tables 1-3 (i.e., ↑ = 1; ~ = 0; ↓ = − 1). If individuals were to choose randomly between their actual and the counterfactual decision, one would expect the average outcome to be zero, and if individuals preferred the option optimising their breeding success and/or rank, one would expect the average response to be above zero. In all three models, we included fixed effects for the decision (disperse or stay), the sex of the individuals, and number of potential dispersers that stayed (for staying individuals) and dispersed (for dispersing individuals) (thereafter "coalition size"), to estimate the extent to which the changes in breeding success and/or rank differed according to these factors and control for their effects. We included coalition size as we reasoned that it could positively impact the outcomes due to an elevated competitive strength of larger coalitions or negatively impact the individual outcomes due to increased competition in larger coalitions. To rule out pseudo-replication we further included random intercepts effects for the group of origin and the target group (for the individuals who stayed in the original group we set the target group to the group of origin, and for individuals who dispersed but did not manage to join any target group and perished on their way we set the target group as "nil"), as well as the event ID. The random intercepts effect of event ID accounts for the mutual non-independence of the decision outcomes of different individuals involved in the same event. In the models, we nested event ID in the random effects of group of origin and target group. As each individual was present in the data with only one decision, we did not need to account for between individual variation in our models. To avoid an overconfident model and keep type I error rate at the nominal level of 0.05 we included the only theoretically identifiable random slope (Schielzeth & Forstmeier 2009;Barr et al. 2013), namely of that of decision (manually dummy coded and then centred to a mean of zero) within the group of origin. We did not include a parameter for the correlation among the random intercepts effect of group of origin and the random slope of decision as these were either estimated to be close to zero or unidentifiable. We fitted these models in R (version 4.2.1; R Core Team 2021) using the function clmm of the package ordinal (version 2019.12-10; Christensen 2019). Prior to fitting the model, we z-transformed coalition size to ease model convergence. We determined model stability by dropping individual events, groups of origin, and target groups one at a time, fitting the same model to each of the subsets, and comparing the range of model estimates with those obtained for the full data set. This revealed all models' coefficients to be of moderate to good stability (see Tables 5-7). For each model, we evaluated whether the proportional odds assumption was met by dichotomising the response at each possible value (i.e., ≥ 0 and ≥ 1) and fitting a logistic Generalised Linear Mixed Model (GLMM; Baayen 2008) with the same fixed and random effects as the ordinal models. We chose this option as, to our knowledge, this is the only option for ordinal mixed models with several random effects included. We then compared the estimates for the individual effects obtained from the logistic GLMM with those of the respective ordinal model. With the exception of the effect of sex in model 1, this revealed that the estimates did not vary much (Fig. S2 in Supplemental Data), indicating that the proportional odds assumption was not obviously violated. We obtained 95% confidence intervals of model estimates and fitted values by means of a parametric bootstrap (N = 1000). The sample considered with these models comprised 247 decisions made during 64 events involving 38 groups of origin and 57 target groups. In cases where 95% confidence intervals of fitted values were above (below) zero, we considered the individuals in the respective combination of levels of the factors (decision and sex) to have chosen significantly better (worse) than if they had taken the counterfactual decision. We did not conduct a full model comparison (Forstmeier & Schielzeth 2011) as we were mainly interested in controlling for the effects of sex and coalition size rather than testing their effects.
We also addressed the same questions by means of permutation tests. As their results were mainly identical to those of the models, we report the details of the permutation tests and their results in the appendix (Section C in Supplemental Data).

Dispersing, staying and attaining breeding success
Out of the potential dispersers in all 64 dispersal events, ca two-thirds dispersed and one-third stayed, and these proportions were roughly similar for females and males ( Table 4). The proportion of potential dispersers in these events who attained breeding success during the subsequent year was much higher for dispersers as compared to individuals who stayed, and again there was no strong difference between females and males regarding these proportions.

Fitness consequences of dispersing/staying (counterfactual analysis)
Considering breeding success, a majority of the 156 individuals who dispersed improved their breeding success compared to their breeding success had they stayed (58%; 65% of the dispersing males and 54% of the dispersing females), for an additional large fraction (40%; 35% of the dispersing males and 44% of the dispersing females) breeding success would have been unaffected by the dispersal, and only two individuals (females) did worse post dispersal compared to how they would have bred if they had stayed. In contrast, of the 91 individuals who stayed, only three did better in terms of their breeding success, but for 74% (57% of staying males and 82% of staying females) breeding success would not have been different if they had dispersed; the remaining 23% (40% of staying males and 15% of staying females) fared worse in terms of breeding than their assessed breeding success had they dispersed.
In the model of breeding success outcome (model 1) we found that individuals who dispersed on average clearly and significantly improved their breeding success as compared to if they had taken the opposite decision, with no obvious differences between males and females ( Fig. 1; Table 5). In contrast, individuals that stayed had slightly but significantly worse breeding success as compared to their breeding success had they dispersed, here too with no obvious differences between sexes. Coalition size did not have a significant effect on the outcome of the individuals' decision as compared to the opposite decision.
In terms of rank, of the 156 individuals who dispersed, 44% (46% of the dispersing males and 43% of dispersing females) achieved a higher rank than the rank they would have attained had they stayed, and 30% (males: 29%; females: 31%) achieved the same rank. Of the 91 who stayed, 40% (males: 33%; females: 43%) improved their rank in comparison to the rank they would have achieved had they joined the dispersing coalition, and 26% (males: 27%; females: 26%) attained the same rank as the rank they would have achieved had they dispersed. Table 5.
Results of model 1 with breeding success outcome being the response (estimates together with standard errors, 95% confidence interval limits, significance tests, and the min and max of estimates obtained when excluding events, groups of origin, and target groups one at a time). Z-transformed to a mean of zero and a standard deviation (SD) of one; mean and SD of the original coalition size were 2.709 and 1.050, respectively. The model of rank outcome (model 2) revealed that individuals that dispersed fared significantly better than if they had taken the opposite decision, but individuals that stayed did not significantly improve their rank as compared to if they had taken the opposite decision (Fig. 2). However, of the effects present in the model (decision, sex, and coalition size) only coalition size had a significant effect on the outcome whereby an increased coalition size correlated with a smaller gain in rank (as compared to when individuals had taken the opposite decision; Table 6).
Combining the comparisons of breeding success and rank, we found that the far majority of the 156 dispersing individuals (total: 69%; females: 67%; males: 72%) fared better than if they had stayed, and an additional small fraction fared just as well (total: 10%; females: 11%; males: 9%); the remaining fraction did worse. For the 91 individuals who stayed in their group of origin, less than half fared better than if they had dispersed (total: 38%; females: 41%; males: 33%), but an additional fraction fared just as well (total: 21%; females: 23%; males: 17%). The remaining large minority fared worse (total: 41%; females: 36%; males: 50%). Altogether, among all 247 potential dispersers in these 64 dispersal events, 58% made a fitness-enhancing decision given the choices of their peers, and for an additional 14% their choice was equally good in terms of breeding success and rank as the reverse decision would have been.
The model addressing the combined fitness consequences of a decision to disperse or stay (model 3) revealed that individuals that dispersed did clearly and significantly better than if they had stayed, with no obvious differences between sexes Fig. 2. -Change of rank as compared to what would have been achieved in case an individual had made the opposite decision (staying rather than dispersing or dispersing rather than staying), separately for females and males that stayed and dispersed. Values larger (smaller) than zero denote a rank higher (lower) as compared to if the individual had taken the opposite decision. Horizontal lines with error bars depict the fitted model and its 95% confidence limits (for coalition size being centred to a mean of zero). The area of the dots is proportionate to the number of decisions represented by the respective dot (range: 8-39).
( Fig. 3; Table 7). As was the case of the model with relative rank change as the response, the larger coalition size, the lower the benefits of the decision taken as compared to had individuals taken the opposite decision. In all three models the random intercepts effects of the event and the target group were estimated to contribute hardly anything to the respective response variable. However, the random intercepts effect of groups of origin and the random slopes effect of decision within group of origin were estimated to be of relatively high magnitude. This indicates that the respective response variables and also the effects of deciding to stay or to leave varied considerably between groups of origin (Section B in Supplemental Data).

DISCUSSION
In cooperative breeders, the outcome of major "career decisions" on the route to breeding crucially depends on peers' concurrent choices. This study provides a unique and rare window of opportunity for assessing the quality of such interlaced decisionmaking in 64 dispersal events of Arabian babbler same-sex coalitions of non-breeders, based on details about the involved individuals' life history both before and after the dispersal event.
Joining a target group often constitutes a "hostile takeover" of a breeding position, involving fights with existing group members who are eventually chased away from their group, and sometimes also with other foreign contenders for the same "prize". This may account for the fact that 81% of the dispersal events in our dataset were carried out by coalitions of more than one individual. In a study of Arabian babblers at the same study site, Ridley (2012) found that dispersing individuals lose body mass as a result of the dispersal process, and that birds dispersing as a group lose less body mass compared to birds dispersing alone. Dispersal by coalitions of individuals of one sex into breeding vacancies occurs regularly in other species of cooperative breeders (Koenig & Mumme 1987;Heinsohn et al. 2000;Sharp et al. 2008). Hence, this may be a strategy to improve their chances to breed and to lessen the effects of reduced survival following dispersal (Ridley Table 7. Results of model 3 with the combined merits of the decision being the response (estimates together with standard errors, 95% confidence interval limits, significance tests, and the min and max of estimates obtained when excluding events, groups of origin, and target groups one at a time). Z-transformed to a mean of zero and a standard deviation (SD) of one; mean and SD of the original coalition size were 2.709 and 1.050, respectively. 2012). Nevertheless, while partners in a dispersal event may help to join a group, too many partners may reduce one's chance to breed in the target group, especially for the lower-ranking individuals. This tradeoff was corroborated by our findings in models 2 and 3 that with increasing coalition size, social rank and combined fitness measures decreased. Altogether, our findings suggest that potential dispersers make informed decisions about whether to join a dispersal event whereby they take into account their same sex peers' decisions and assess their future prospects of reproduction given their peers' dispersal decisions. Exploring the implications of such informed decisionmaking in terms of population demography and gene flow is an open avenue for future studies.
Obviously, we cannot know the motives of the babblers, or the information that they used to make their decision. The outcomes of the impulse to breed and the desire to improve their rank are measurable, but their intensities potentially vary among individuals. In addition, at least two kinds of factors could have influenced their decision as well: those derived from their personality and those derived from their social relationships with other group mates. It is possible that certain individuals are more apprehensive or fearful than others and/or do not take risks. It is also possible that certain individuals decide to disperse or refrain from joining a dispersing coalition because another group mate chose to leave. These components of decisionmaking, which are so powerful among human beings, were concealed from our eyes. Additionally, we had no measurement in this study of body weight and condition, which could further affect dispersal decisions (Clobert et al. 2009).
In summary, we found that, given the choices of their groupmates, a significant majority of potential dispersers made fitness-enhancing or at least fitness-neutral decisions as compared to if they have made the opposite choice. These findings shed light on the underlying mechanisms behind dispersal, contributing to a more complete conceptual understanding and evaluation of the costs of dispersal in cooperative breeders. Furthermore, the counterfactual analysis, which we developed as a methodological framework, proved a useful tool when investigating such decisions of individuals in situations in which the consequences of an individual's decision might crucially depend on the decisions of its peers. ETHICAL STANDARD