Group living is widespread across the animal kingdom and evolved convergently from an ancestral solitary state in different taxa (e.g. [1]). Many species remain solitary or are only seasonally social [2], showing that sociality is only beneficial when benefits outweigh the costs [3]. For example, in the social cliff swallow (Hirundo pyrrhonota), colony size is correlated with at least 13 different types of costs (e.g., parasitic infestation, brood parasitism) and at least 13 different types of benefits (e.g., predator-avoidance, information transfer, [4]). Thus, group size is an important trait that responds to cost-benefit regimes depending on a species, its ecological niche and life history [4, 5]. In fact, the size of animal aggregations can vary from small social groups below ten individuals like the prides of lions [6] to huge colonies with millions of seabirds or bats [7, 8]. Thus, a crucial step in any study is to distinguish between aggregations of individuals, due to external circumstances such as roost limitation, and “true” social groups with reciprocal relationships (which may be contained in larger aggregations) [9].

Sociality should be adaptive [3], we therefore expect fitness benefits of optimal group size resulting in prolonged survival, enhanced reproductive success or both. Life history theory predicts that animals should allocate their energy differently to individual reproduction or survival [10, 11]. As a general rule of thumb, small animals are short-lived and produce many offspring (e.g. rodents, r-strategists) while large animals are long-lived and have few offspring (e.g. elephants, K-strategists) [12]. Bats are an exception to this general pattern, being small but long-lived and producing relatively few offspring. However, while life history theory does not incorporate sociality, there are many studies linking group size with survival. Different parameters are used to investigate this, the two most common being maximum lifespan (or maximum longevity) and an averaged estimate for the survival of the group members. Comparative studies on birds and mammals did not find any correlation between maximum lifespan and group size [13–16]. The same is not true for the relationship between group size and survival. Group size is often positively correlated with survival in many taxa, including termites [17], social spiders [18], birds [19–21] and mammals [22, 23]. In all of these examples, social behaviours, such as predator avoidance, social thermoregulation or social foraging, lead to improved survival. However, there is a limit to the benefits of increasing group size. For instance in certain colonies of Neotropical spiders, survival of the colonies increased with colony size. But above a threshold in colony size (~15 individuals), survival of the colony decreased, presumably because of an increase in intra-colony competition [18]. In other species, such as the Seychelles warbler (adults) and a social spider (juveniles), there is even a strictly negative relationship between survival and group size, again probably due to competition for resources [18, 24]. Despite this decreased survival, increasing group size brings reproductive benefits in the Seychelles warbler. The reverse situation was observed in Neotropical spider, with survival benefits but reproductive costs with increasing group size leading to a trade-off situation and resulting in maximum fitness at intermediate size [25]. Finally, in some species, including wild dogs, juvenile rodents, primates or coatis, group size and survival are independent [26–29], interpreted to be a result of specific ecological conditions such as low competitor density and high food availability.

One important benefit of sociality is information transfer between individuals [30–32]. In a foraging context, animals can detect conspecifics present at food patches through “local enhancement” [33]. The number of animals at a food patch and the modality of the information they use (e.g., sound, vision, olfaction) can have crucial implications for their fitness. Many bird species rely on local enhancement through vision to detect conspecifics at a food patch (e.g. seabirds, vultures, ospreys and swallows) [34–37]. In an empirical test of recruitment of seabirds to food patches, adjusted estimates for average distance recruitment ranged from 4.9 to 11.3 km [35]. Therefore, vision, the most commonly used mode of information transfer during foraging leads to the attraction of individuals over long distances and is believed to have led to the evolution of bird colonies [38]. Echolocating bats, in contrast, cannot use vision during nocturnal foraging. Instead they benefit from information transfer by eavesdropping on changes in each others’ echolocation calls that indicate successful localization of a food source [39–44]. Compared to vision, the propagation distance of echolocation calls is very short due to rapid atmospheric attenuation [45]. For instance, maximum hearing distance of conspecifics was estimated at 54 m in M. molossus and 35–40 m in Noctilio albiventris [40, 41], however this is ten times the distance from which they can actively localize a prey item. The restriction to different modalities (e.g. vision vs. sound) therefore has direct implications for the foraging strategy. However, the relationship between social foraging, the composition of groups, survival, and group size remains poorly understood in bats despite the wide reliance on social information to locate resources and its effect on the evolution of group living in bats and other animals.

To test if and how social group size affects survival in bats, we studied Pallas’s mastiff bat Molossus molossus (Pallas, 1766), a species that forms stable social groups that roost and forage together [41]. The narrow-shaped wing morphology of M. molossus results in high energetic requirements within an open-air foraging niche [46]. As a result of this specialized morphology, this species depends on ephemeral insect swarms as their only food source that are only available at dawn and dusk, and therefore foraging time is restricted to short activity peaks of less than an hour [41]. These energetic and morphological limitations as well as the narrow foraging niche make increased foraging efficiency through the use of social information from conspecifics highly important. One might thus expect that a large number of foraging partners and correlated increase in social group size would be advantageous. However, theoretical work indicates optimal individual uptake in groups with a small number of signallers in a recruitment scenario [47]. Due to the short availability window of its resource and the modality of information transfer (acoustical), M. molossus must coordinate flight and eavesdrop on echolocating group members on the wing instead of recruiting. This quickly creates a complex system of signallers and receivers and thus a trade-off between benefits of improved indirect prey detection and costs of conspecific acoustic interference [48]. Thus, we hypothesized that there is an ideal group size for M. molossus, and that this group size should be small. Individual survival should then be highest in these ideal small groups. To test this we used two approaches: (a) We captured 14 social groups multiple times over several years. During a subset of this time period, (b) we also monitored four of these groups with automatic transponder readers to get a more precise temporal resolution of changes in group composition. We modelled the role of group size using these two data sets in two survival analyses based on the Cox proportional hazard and multistate mark-recapture models. We based our survival analysis on all adults present in a group because males and females are known to forage together [41]. However, we only analysed survival and lifespan for adult females because males probably spend time as bachelors before and possibly even after their presence in the groups, which then is probably only indicative of their harem tenure. The unique access to data from free-ranging socially foraging bats thus allowed us to test predictions from theoretical evolutionary models in a system of naturally behaving animals.

Our data suggest that even though aggregations of bats can consist of up to several million individuals, there is strong selection for small group size (≤25 adults) in Molossus molossus. There was little variation of group size and short median female survival of 1.8 years (corrected estimates from all recaptured data). Furthermore, group size was not correlated with longevity as calculated from recaptures and monthly survival as calculated from automated monitoring in a subset of groups (i.e. BaTLis).

Thus, along with our predictions, M. molossus forms small groups, composed of a stable core of adult females and one adult male in most groups (mean of 8.1 ± 5.1 adults). BaTLi data showed that all juveniles of both sexes dispersed from the natal group. Once they settled in a new group, adult females remained there longer than adult males (median of 0.8 vs. 0.6 years, maximum of 4.7 vs. 2.9 years). Thus, the cluster of adult females emerges as the stable and primary unit of M. molossus social organization. Stable clusters of females favour the evolution of a male strategy of female-defence polygamy [51, 52]. Where females live in social groups, one single male is commonly responsible for most or all mating, as shown for instance in primates, antelopes or bats [49, 53–55]. Our data suggest that maximum male tenure length (2.9 years) exceeds the sexual maturity of the sired daughters, driving natal dispersal of juvenile females to avoid inbreeding with their father [56]. This phenomenon is observed in few other species of mammals [57] but seemingly not uncommon in tropical bats living in stable groups [58, 59]. Natal dispersal of juvenile males may result from an eviction by the harem male (potentially their father) and/or an attempt to find mating opportunities [60, 61]. We are unable to estimate lifespan for male M. molossus because we do not know the time from natal dispersal to harem take-over as well as the fate of replaced harem males. We suppose males remain solitary or join a bachelor group before and after harem tenure as already observed in other tropical bats [49, 50] but the roost preferences of these males seem to differ from those of females as we did not find them during our exhaustive roost surveys in buildings. A challenge for future studies will be to follow males throughout their life cycle to obtain realistic lifespan estimates. Similar to other polygynous mammals, we expect to find reduced longevity in adult males resulting from intense male–male competition and weaker selection for longevity [62].

Group fidelity of adult females was very strong as shown by one to seven recaptures of 121 females in the roosts where they were initially marked as adults. In the roost we monitored the longest time, a female was recaptured after 4.7 years with an estimated lifespan of 5.6 years. We expect that maximum lifespan records might still increase slightly with a longer study period, but we are confident that our mean estimates of 1.2 and 1.8 years are representative for the species. This is unexpected, because bats are famous for exceptionally long lifespans relative to their small body mass (i.e. allometry of lifespan), the record being held by a 8-g insectivorous Myotis brandtii which was recaptured after 41 years in the wild. On average, lifespan of bats is around 3.5 times longer than in non-flying mammals after correcting for body mass [15]. Maximum lifespan for two other molossid species with a similar ecological niche as M. molossus is 12 and 13 years, respectively for the much larger European Tadarida teniotis (mean mass of 30 g) and the American T. brasiliensis (mean mass of 12.5 g similar to our study species) [63–65]. The higher longevity observed in these two species may result from decreased predation associated with cave roosting [15] and/or a broader foraging niche in terms of prey as well as temporal and spatial food availability. Despite its relevance for population dynamics, maximum lifespan represents only the upper limit of the survival curve. Better knowledge about average or median lifespans appears to be important, but often lacking life-history parameter in bat studies (but see [66]) for the better understanding of the pace and shape of the survival curve [67, 68]. Our data revealed a skewed survival curve for females, with median survival of 1.8 years and a maximum longevity of 5.6 years. Low values for median and maximum lifespan in M. molossus may result from life at the energetic edge due to a narrow foraging niche. This bat is an open-air forager, with long and narrow wings that result in high wing loading and high energetic costs of flight [69, 70]. This bat is also specialized to forage on insect swarms which are abundant when found but remain relatively unpredictable in space and time [39]. The species shows a bimodal activity pattern, after sunset and before sunrise [71, 72]. The predominant foraging activity occurs after sunset, sometimes for only a half an hour interval [71, 72]. This limited burst of activity probably result from a peak in insect density [70]. Because insect patches can be dispersed by wind and rain, bats sometimes entirely skip a night of foraging. To limit the risk of starvation, this species can maximize energy intake by socially foraging [41] and minimize energy investment by lowering metabolism when roosting [73]. However, due to the short foraging window, the unpredictability of the resource and the flight costs, these bats have a risk of starvation. Our data, suggesting that most females M. molossus only reproduce once or twice in their short lifetimes, is consistent with this hypothesis. Further investigation will be necessary to determine how the small percentage of longer-lived individuals contribute to the maintenance of the species, what causes these enormous variations in lifespans and how they are linked to the ecology of species. In addition, it will be important to find other factors influencing variation in female lifespan, such as foraging efficiency, and also following up on anecdotal reports about twinning as well as an additional smaller reproductive peak to better understand how stable populations of this species can persist.

Social groups of M. molossus are stable year-round, implying that benefits of sociality permanently outweigh the costs [3]. Foraging benefits via information transfer about ephemeral resources have been postulated as a major reason for coloniality in seabirds [74, 75]. And similarly, in M. molossus and other bats, the main benefit of group living is probably increased foraging efficiency through acoustic information transfer about ephemeral insect patches [39–41]. The daily availability of M. molossus’ food source is so short that information must be shared in the foraging arena on the wing, most likely via acoustic eavesdropping [40]. This means that groups must coordinate flight and filter relevant information from the echolocation calls of their social partners. Even though we do not understand yet how this works in detail, such a network of signallers and receivers quickly becomes very complex and confusing. Our results confirm our hypothesis that this should exert strong selection pressure on small group size because the 14 groups were ranging between one and 25 adults. This is in contrast to opportunistically eavesdropping species that do not emerge in coordinated flight and maintain group cohesion throughout their foraging period [76, 77].

While our expectations regarding small group size were confirmed, we did not find that individual lifespan was influenced by variation of group size. In both datasets (capture and BaTLi data), group size had no effect on lifespan or monthly survival despite group size being the most explanatory variable in the multistate mark recapture analysis. We postulate that selection on group size is so strong that the resulting variation, mainly caused by the brief appearance of pre-dispersing juveniles, is too small to have an effect on the adult females in the group. In fact, a closer look at previous work that found no or a negative relationship between group size and survival reveals consistently small group sizes in animals with complex social systems (≤33 individuals), e.g. 1–6 individuals in the Seychelles warbler [24], 2–12 adults in degus [29], 2–10 adult females in the coati [26], 2–17 individuals in the wild dog [28] and 6–33 individuals in the leaf monkey [27]. This suggests that selection for small group size in complex social systems may be fairly common. In other species, the limited variation of group size observed here (3–13 individuals) can still cause survival differences that can be detected. For example in the Seychelles’ warbler, survival is negatively correlated with group size (one to six individuals) [24].

The apparent absence of relationship between group size and survival in M. molossus could have technical and/or biological explanations. Our estimates of monthly survival for the three categories of group size were associated with narrow confidence intervals (respectively 0.85–0.97, 0.91–0.97 and 0.93–0.98). A dataset including more groups and individuals may lower confidence intervals and reveal survival differences that remained hidden so far. Alternatively, our results correspond to the reality and selection on ideal group size is so strong that the remaining low variation does not affect survival. In the framework of life history, the relationship between group size and reproduction remains to be investigated in future studies.

Bats are well known for roosting in large or even gigantic colonies, e.g. the closely related molossid species Tadarida brasiliensis that occurs in caves numbering up to tens of millions of individuals [7]. A confounding effect here may stem from the fact that many bat species are highly dependent on suitable, but limited roosts. Molossus molossus, too, may roost in such larger aggregations (300 individuals or more, [78]), composed of several social groups at other study sites where roosts are more limited. Several bat species, such as Myotis bechsteinii [79], Nyctalus lasiopterus [80] or Eptesicus fuscus [81], form fission–fusion societies. This allows animals to have access to the knowledge of a large pool of individuals but the daily subgroups are relatively small as a flexible reaction to social and environmental conditions. This may also mediate other detrimental effects of group living in bats such as parasite or disease transmission rates or competition [82, 83]. Similar societies are found in a great diversity of taxa, e.g. house sparrows [84], chimpanzees and leaf monkeys [85], or spotted hyenas [86]. However, most socially complex bat species form smaller and more stable social groups similar to M. molossus (≤25 adults), e.g. the molossid Tadarida pumila [87], other socially foraging bats [40, 49], tent-making bats [88–90] and other roost-making bats [50, 91]. We expect that with future research this will be a consistent pattern, especially in species that are ecologically dependent on information transfer about the location of ephemeral resources.

In summary, this study suggests strong selection for small groups (≤25 adults) in a socially foraging bat. Our results are in agreement with models of recruitment on ephemeral resources suggesting a small and stable range of signalers in the groups optimizes individual uptake [47]. In our in situ eavesdropping scenario, where every individual is a signaler and receiver at a same time, the same selection pressure seems to apply to optimize trade-off between foraging benefits from information transfer and acoustic confusion impairing prey detection performance [48]. Our alternative survival analyses based on free-ranging animals independently found no effect of group size on survival, a pattern found in few similar studies and potentially resulting from life at the energetic edge due to a highly specialized diet. We expect similar results for future research conducted on species dependent on information transfer and ephemeral resources.

The data sets supporting the results of this article are included within the article and its additional files.