From resource to female defence: the impact of roosting ecology on a bat's mating strategy

With their extraordinary species richness and diversity in ecological traits and social systems, bats are a promising taxon for testing socio-ecological hypotheses in order to get new insights into the evolution of animal social systems. Regarding its roosting habits, proboscis bats form an extreme by occupying sites which are usually completely exposed to daylight (e.g. tree trunks, vines or rocks). This is accompanied by morphological and behavioural adaptations to remain cryptic in exposed day roosts. With long-term behavioural observations and genetic parentage analyses of individually marked proboscis bats, we assessed its social dispersion and male mating strategy during day and night. Our results reveal nocturnal male territoriality—a strategy which most closely resembles a resource-defence polygyny that is frequent also in other tropical bats. Its contrasting clumped social dispersion during the day is likely to be the result of strong selection for crypsis in exposed roosts and is accompanied by direct female defence in addition to male territoriality. To the best of our knowledge, such contrasting male mating strategies within a single day–night cycle have not been described in a vertebrate species so far and illustrate a possible evolutionary trajectory from resource-defence to female-defence strategy by small ecologically driven evolutionary steps.

With their extraordinary species richness and diversity in ecological traits and social systems, bats are a promising taxon for testing socio-ecological hypotheses in order to get new insights into the evolution of animal social systems. Regarding its roosting habits, proboscis bats form an extreme by occupying sites which are usually completely exposed to daylight (e.g. tree trunks, vines or rocks). This is accompanied by morphological and behavioural adaptations to remain cryptic in exposed day roosts. With long-term behavioural observations and genetic parentage analyses of individually marked proboscis bats, we assessed its social dispersion and male mating strategy during day and night. Our results reveal nocturnal male territoriality-a strategy which most closely resembles a resource-defence polygyny that is frequent also in other tropical bats. Its contrasting clumped social dispersion during the day is likely to be the result of strong selection for crypsis in exposed roosts and is accompanied by direct female defence in addition to male territoriality. To the best of our knowledge, such contrasting male mating strategies within a single day-night cycle have not been described in a by occupying areas which are usually completely exposed to daylight and temporarily even direct sunlight ( [42]; L. G. and M.N. 2013, 2014, personal observations). Their camouflage coloration and cryptic behaviour (i.e. synchronized grooming and urinating among group members and rocking behaviour during gusts of wind to remain cryptic during motion) are interpreted as adaptation to these very exposed day roosts [40,42,[44][45][46]. Considering these strong morphological and behavioural adaptations to remain cryptic in exceptionally exposed roosts during the day, we also expect an impact on the social system of proboscis bats.
The very small insectivorous proboscis bat (3-4 g) is distributed in lowland rainforests from southern Mexico to southern Brazil [47]. It roosts on exposed parts of tree trunks, branches, vines or manmade structures in year-round stable social groups of up to 50 individuals with males and females at about equal numbers, which space themselves at 5-10 cm from each other in day roosts. Roost sites are situated in the immediate vicinity of rivers [42], upon which R. naso exclusively forages. Recently, it was shown that female proboscis bats habitually disperse from their natal colony at an early age of approximately two to four months, while at least half of the male colony offspring settles in the natal colony, where some of them reproduce [48]. Reproduction usually takes place within two distinct mating periods: one seasonal mating period (SMP) at the end of the rainy season (October/November) and one postpartum oestrus mating period (PEMP) during the parturition period (April/May) approximately five months after conception ( [48,49]; L.G. and M. N. 2013, 2014, personal observations). Based on behavioural observations in the day roost, the mating strategy of males has been reported to be one of direct female defence, probably with a dominance hierarchy among male group members [1,40,42,48]. With observations in the day roost, Nagy et al. [48] confirmed the dominance of one male in the group, but paternity analysis showed that six males successfully reproduced. Thus, previous day observations cannot fully explain the social structure and male mating success in proboscis bats.
In this study, we use long-term behavioural observations of individually marked proboscis bats in combination with genetic parentage analyses to assess social dispersion and male mating strategy during day and night. In contrast to the vast majority of other bat species, proboscis bats usually inhabit very exposed and well-lit structures [42]. Thus, we assume that the clumped roosting of mixed sex groups during the day is a derived trait and the result of selection for cryptic behaviour on exposed roost structures. At night, we hypothesize to still observe an ancestral strategy, namely that male proboscis bats establish themselves at preferred sites in their roost where they are territorial or dominant (see the electronic supplementary material, figure S2 for a sketch of this main hypothesis).

Field methods
The study was conducted between 2010 and 2014 in one colony (Cabina 5) at the La Selva field station of the Organization for Tropical Studies (Costa Rica, Province Heredia, 10°25 N/84°00 W). The study colony is located on the outside, under the extending roof (electronic supplementary material, figure S3) of an inhabited wooden station cabin, and thus the bats were well habituated to human presence. This roost site is known to have been inhabited by proboscis bats for at least 15 years. Mist nets (Ecotone ® monofilament, Gdynia, Poland) were used to capture the bats when emerging from their roost at dusk. To prevent bats from connecting the capturing event with a potential threat to their roost, mist nets were set several metres away from the roost. Bats were marked individually with coloured plastic bands on their forearms (AC Hughes ® Ltd., UK, size XCS). A small cut (around 3-4 mm) in the plagiopatagium ensured the correct fit of the bands to both forearms of the bat without moving and potentially hurting the bat's plagiopatagium. Females were banded with a unicoloured and numbered ring on the left forearm and with a bicoloured ring on the right forearm, whereas males were banded vice versa. A small tissue sample from the plagiopatagium or chiropatagium was taken (Stiefel ® biopsy punch, 4 mm Ø) of each bat for genetic analysis (the resulting hole healed completely within two to four weeks). Captured bats were sexed, weighed (16 g Pesola ® spring scale), and their age was determined (juvenile, subadult or adult; see [48] for details on age determination). The classes correspond to an approximate age of zero to four months for juveniles, 5-10 months for subadults and older than 10 months for adults. Details on numbers of banded and genetically sampled bats between 2006 and 2014 during this and a prior study by Nagy et al. [48] are provided in table 1. rsos.royalsocietypublishing.org R. Soc. open

Behavioural observations during day
We recorded all behavioural interactions among bats of the two social groups in the study colony (see results for a definition of groups) during 592:36 h (focal group one) and 532:46 h (focal group two; ad libitum sampling sensu Altmann [50] addition to focal site observations, the whole roost was constantly checked for behavioural interactions during an overall period of 24:32 h throughout 12 nights in October and November 2014. During that time, behavioural observations were carried out by constantly scanning the extending roof back and forth with an analogue night scope (Yukon 3 × 50 Exelon). If an interaction was observed, a photo was taken to determine the involved individuals.

Census analyses
Based on census observations, we calculated fidelity indices F sensu Heckel et al. [51]. First, based on the constant usage of the same areas of the roost by the same groups of bats, we calculated individual fidelity to a social group (F group ). F group was calculated during day and night and corresponds to the proportion a bat was observed in the area of a group in relation to its absolute presence in the roost. Second, individual fidelity of adult bats to a certain site in the roost (F site ) was calculated based on the number of observation events and corresponds to the proportion a bat was present at a certain site in the roost between its first and last day of observation. Only individuals that were adult during the respective observation period and present at least until the end of the period in which they had been banded were included in fidelity calculations (n = 22 females; n = 22 males). Bats were not included if they disappeared before the end of the period of their banding (n = 6 females and n = 4 males) because their disappearance was probably caused by disturbance. Individuals with a day roost fidelity below 0.5 during the respective period (n = 1 females and n = 6 males) were not considered as members of the social groups and were thus not included in the calculations. We also excluded individuals from calculations that were juvenile (n = 27 females and n = 31 males) or subadult (n = 20 females and n = 20 males) during the respective period.

Behavioural analyses
For this study, we defined three different behavioural interactions in the context of mating: copulations, copulation attempts rejected by females and female-defence actions performed by males. Copulations started with a male approaching a female from behind, subsequently mounting the female's back until their heads were almost at the same level. A copulation was considered to be successful if we observed the male flattening its interfemoral membrane, giving several short thrusts and finally tapping the female with his chin on her back and retracting from the female voluntarily. Copulation attempts also started with a male approaching a female from behind and mounting the female's back but were rejected by the female hitting the male with a wing and/or by flying/crawling away. The third category 'female-defence action' comprises different scenarios. A male was regarded to have performed a female-defence action, if he successfully chased away another male that was closely perching behind a female, approaching a female or attempting to mount a female. This defence was either achieved by quickly approaching the couple and/or hitting the male with its wing, prompting the male to crawl or fly away. In addition, a male that was perching closely behind a female or trying to mount a female and successfully defended his position against another male that attempted to take his position was also regarded to have performed a female-defence action.

Genetic analyses
Ethanol (80%) was used to preserve tissue samples, and the salt-chloroform procedure of Miller et al. [52] modified by Heckel et al. [51] was used for DNA isolation (for details on PCR conditions, see [53,54] We used the DNA Analyser 4300 (LI-COR ® ; Biosciences) and the SAGA GT (LI-COR ® ; Biosciences) allele scoring software to genotype a total of 156 individuals (n = 40 adults, n = 46 subadults, n = 70 juveniles) caught between 2012 and 2014 at the colony 'Cabina 5' and two other study colonies nearby at 10 highly polymorphic microsatellite loci [53,54]. All individuals were genotyped at least at eight loci, and genotypes were 99.7% complete. See electronic supplementary material, table S1 for allele numbers per locus, results of Hardy-Weinberg tests, null allele frequencies, and non-exclusion probabilities for the 10 microsatellite markers. A total of 354 individuals, which were caught in the study colony and 14 additional colonies in that area between 2005 and 2011 and genotyped by the authors in a prior study with same methods and protocols were included in the following parentage analyses [48].
Parentage analyses was performed with CERVUS v. 3.0 [55] for 87 potential colony offspring caught between May 2011 and November 2014, consisting of 58 bats caught as juveniles and 29 bats caught as subadults in the study colony Cabina 5. Maternity analyses were carried out with all 10 loci, while one x-linked gonosomal locus (Rn16) was left out for male-male paternity assignments [54]. Maternity of 49 R. naso caught as juveniles was determined by nursing observations and confirmed with genetic analysis. For six juvenile and 22 subadult individuals maternity was analysed entirely with CERVUS v. 3.0 [55]. All males caught as adults between 2006 and 2014 (n = 126) and resident males that matured prior to the respective mating period when potential colony offspring were conceived (n = 36) were treated as putative fathers for paternity assignment of pups with known (n = 59) and with unknown mothers (n = 28). Simulations were run with 100 000 cycles, a proportion of 80% sampled candidate fathers, an estimated genotyping error of 2.2% (estimated with CERVUS v. 3.0 and based on 22 mismatches between 80 known mother offspring pairs from the prior study by Nagy et al. [48] and this study), and a proportion of 14.5% candidate fathers that were relatives, related to the true father by r = 0. 5. Although on average we had sampled 98% of the individuals in the study colony, we attempted to account for possible extra-colony paternities by choosing a lower sampling rate of 80% candidate fathers. The percentage of relatives among candidate fathers was estimated by Nagy et al. [48] based on the results of the kinship analysis between adult males. Simulations were performed for two confidence levels (80% and 95%). One mismatch per parent-offspring dyad was accepted, thus two independent mismatches between an offspring and each of its parents to account for genotyping errors. Fifty-six parent pairs were assigned at 95% confidence, and one parent pair was assigned at 80% confidence. For two female subadult immigrants, only the mother was assigned at 95% confidence, and in one case where the mother was not genetically sampled only the father was assigned at 95% confidence.
In contrast to all other loci, the locus Sb85 showed significant evidence for null alleles (electronic supplementary material, table S1). Therefore, we analysed maternity with and without the locus Sb85 for the whole dataset from 2006 to 2014. With the locus Sb85, 145 mothers-offspring pairs were assigned at 95% confidence. The analysis without Sb85 resulted in 10 additionally assigned mother-offspring pairs and four differently assigned mother-offspring pairs. In 12 of these 14 differences, the locus Sb85 showed no evidence for null alleles since the candidate offspring was heterozygous; therefore, we used results from parentage analysis with all loci (see also [48]).

Other statistical analyses
Other statistical analyses besides parentage analyses were performed with R v. 3.2.1 [56]. The Shapiro-Wilk normality test was used to test for normal distribution. For examining the median difference of two non-normally distributed datasets, the two-tailed Mann-Whitney U-test with continuity correction was used. Medians are presented with interquartile range (IQR) in the form of first quartile to third quartile (i.e. IQR = Q 1 -Q 3 ). Whiskers in boxplots show Q 1 minus 1.5 times IQR or Q 3 plus 1.5 times IQR, respectively. Outliers are defined as values beneath or above the '1.5 cut-off'.

Group fidelity (F group )
We defined two social groups within the colony based on the observation of constant usage of the same sites within the roost by the same set of bats over each observation period (figure 1; group 1 in black, group 2 in dark grey). Members of group 1 usually roosted in sites 1-3 and members of group 2 usually occupied sites 4 and 5 during day and night (figure 1). See table 2 for details on group size, sex and age distribution. Our distinction of two social groups was supported by high fidelity indices for both female and male adults towards their groups during day (median F group females = 1.00, IQR: 1.00-1.00; median F group males = 1.00, IQR: 1.00-1.00) and night (median F group females = 1.00, IQR: 1.00-1.00; median F group males = 0.97, IQR: 0.85-0.99); though, at night individual group fidelity showed higher interindividual variation and lower values among males compared with females. This was consistent throughout all observation periods.

Site fidelity (F site )
In the following three paragraphs, the pattern of individual site fidelity in the study colony is

Site fidelity during afternoon (figure 1, rows 2 and 5)
Females and males of both groups still predominantly used the same sites as during the morning. However, the entire groups relocated to a second site (group 1 to site 1 and group 2 to site 5) around the corners of the house more often than in the morning, resulting in a lower fidelity to the preferred site in both groups (median max F site throughout all periods: females = 0. 66  Furthermore, for each of the four main sites we found only one male with very high site fidelity (figure 1, rows 3 and 6; max F site males during mating periods = 0.69-1.00; n = 13 territorial males). These males were rarely or never observed at multiple sites in the roost at night. This pattern was less distinct during NMP (max F site males during NMPs = 0.63-0.90; n = 4 territorial males). Subsequently, we refer to these males as territorial males. The other males-referred to as non-territorial males-switched among sites and used two to four different sites in the roost, therefore had lower site fidelity at each site (figure 1, rows 3 and 6; max F site = 0-0.6; n = 18 non-territorial males). For 'site 4' in PEMP and SMP 2013, we found two territorial males (ID217 and ID338). This might reflect a period of overlapping in a transition from one territorial male to another since after SMP 2013 the older male of the two disappeared, while the younger male remained territorial at the site until the end of the study. At 'site 3' only in PEMP 2011 and SMP 2014, a territorial male could be identified, as during other periods several males showed high fidelities to that site.
Throughout the study, we determined 13 different territorial males. During 2013 and 2014 with census data on four continuous mating periods, mean minimum tenure as territorial male was 2.4 mating periods (n = 8 territorial males; range: 1-4 mating periods). However, owing to the lack of information on start (n = 3), the end (n = 3) or both (n = 2) of the males' territorial status, actual average male tenure as territorial surely exceeds our estimates. 3.3

.1. During day
Observations during mating periods in 2013 and 2014 resulted in a total of 122 copulations, 763 copulation attempts rejected by females and 138 female-defence actions performed by males (see Material and methods for definition of female-defence actions). The same males which were determined as territorial at night (four to five different males per observation period and depending on the mating period) performed the majority of all rejected copulation attempts (66.7%, n = 513), copulations (86.8%, n = 105) and female-defence actions (88.4%, n = 122). Performances of the remaining actions were done by seven to nine of the 7-12 present non-territorial males. Individual proportion of performances in all categories per group and mating period was significantly higher for males which were determined as territorial at night than non-territorial males (figure 2; Mann-Whitney U-test: n non-territorials = 18, n territorials = 8; rejected copulation attempts: U = 2130.5, p < 0.001; copulations: U = 1928.5, p < 0.001; female-defence actions: U = 1870.5, p = 0.003). In addition, territorial males performed all daytime copulations at the site where they were also determined as being territorial at night (100%, n = 105) as well as the vast majority of rejected copulation attempts (91.8%, n = 456) and female-defence actions (82.6%, n = 114). In other words, if the group perched on site A, where male A was determined as territorial at night, male A was the most successful male in the group regarding the three mating-related categories during day. However, if the same group moved to site B, where male B was determined as rsos.royalsocietypublishing.org R. Soc. open  territorial at night, male B was the most successful male in the group during day and male A stopped performing. Note that three males were included as both non-territorial and territorial males in statistics, since they became territorial between 2013 and 2014.

At night
In SMP 2014, we conducted focal night observations of the five territories. We observed a total of 24 copulations, six rejected copulation attempts and no female-defence actions. All copulations and copulation attempts were performed by the five territorials within their own territories. Nine nonterritorial males performed none of these actions. In eight cases, females left their preferred roosting site to visit another territorial male of her group in his territory for copulation. Furthermore, we have evidence of one territorial male chasing another male out of his territory without any females present (n = 3 observations).

Parentage
We were able to genetically assign a mother to 53 juveniles and six subadults of 87 analysed potential colony offspring (n = 58 juveniles and n = 29 subadults). The 28 remaining individuals without assigned mother (n = 5 juveniles and n = 23 subadults) can be regarded as immigrants as all but one of the potential colony mothers were sampled and genotyped. In addition, 24 of the 28 individuals without assigned mother were also observed to immigrate into the colony as subadults (n = 20), or juveniles (n = 4), because offspring dispersal takes place very early at an age of approximately two to four months, when some bats are still classified as juveniles.
Of 59 offspring with a genetically assigned mother, 56 can be regarded as colony offspring as they were either observed to be nursed in the colony (n = 49) or the mother was at least present in the social group and season the pup was sampled (n = 7). The mothers of the three remaining offspring (subadult females) were constant members of a different nearby study roost. Thus, these subadult females can be regarded as immigrants. In addition to the 56 genetically identified colony offspring, one pup without an assigned mother was nursed by a female which was not genetically sampled. We

Siring success of territorial and non-territorial males
We possess data on male territoriality for fathers of 52 colony offspring from five different mating periods either gathered during time of conception (n = 26 offspring) or gathered four months after conception (i.e. during the following parturition period, n = 26 offspring). During these five mating periods, nine of nine territorial males sired 38 of the 52 colony offspring (73.1%), while eight of 21 non-territorial males sired the remaining 14 offspring (26.9%). Median individual reproductive success per mating period was significantly higher for territorial than non-territorial males (figure 3; non-territorial males = 0.00, IQR = 0.00-0.44; territorial males = 1.75, IQR = 1-2; Mann-Whitney U-test: n non-territorials = 21, n territorials = 9, U = 19.5, p < 0.001). Two males are included as non-territorial and territorial males in the statistics, as they became territorial between the mating periods when the sampled offspring was conceived.

Discussion
Our observational and genetic results reveal striking differences between day and night regarding social dispersion in the roost and the males' main mating strategy. At night, territorial males occupy a territory. Some territories permanently include females, whereas other territories are only occasionally visited by females. At four to five sites of the study roost, we found for each site only one male that showed very high site fidelity (max F site = 0.69-1.00). All other colony males were either absent at night or roosted at two to four different sites in the roost, resulting in low site fidelities per site (max F site = 0.00-0.63); we refer to these males as non-territorial males. All definitions of territoriality are variants of three main themes: defended area, site-specific dominance and exclusive area (reviewed in [57]). While the first two criteria are based on behavioural interactions, the latter can be seen as the outcome of behavioural interactions [58]. In general, several operational pitfalls can occur by using behavioural interactions to define territoriality [59], but the prevailing pitfalls with most bat species and many other taxa are high mobility, nocturnal activity and complex ultrasound vocal communication that may be difficult to relate to the behavioural context. At night, territoriality in R. naso comes closest to an exclusive area, suggested by the exclusively high site fidelity of only one male per site. This is also supported by the fact that only males with high site fidelity, i.e. territorial males, were observed copulating at night and they exclusively copulated at their territory. Furthermore, the evidence of one territorial male chasing another male out of his territory without any females present (n = 3 observations) seem to represent physical interactions among males (defence actions) and might be rare escalations of the very abundant acoustical interactions. Finally, the median proportional individual reproductive success within each social group was significantly higher for territorial males compared with non-territorial males. Thus, our behavioural and genetic results provide evidence that R. naso males follow a territorybased mating strategy. According to Emlen & Oring [5], such a strategy can be linked to either a resource defence or a lek system. In R. naso, territories are sites within the roost, which are used year-round by females and thus may be considered a crucial resource. Regardless of the question whether a particular site in the roost represents an important resource for females and thus gives males the possibility to guard and defend it, the high and long-term presence of territorial males at a single site in the roost (i.e. high site fidelity) may also indicate important male characteristics to females (e.g. assertiveness and efficient foraging). The observation that females left their preferred site and former mating partner (i.e. the territorial male at that site) to visit other territorial males at other territories for copulation supports the idea that female choice occurs in R. naso. Based on nocturnal observations and male mating success, in R. naso the prevailing social dispersion and mating system at night most closely resemble mating territories in a resource-defence polygyny (sensu Emlen & Oring [5]). Although we have never observed any female-defence actions at night, we cannot exclude the strategy of direct female defence by R. naso males at night. However, our observations show that direct female defence plays only a minor role in male mating strategy at night.
By contrast, during day R. naso lives in cohesive groups of multiple males and females. Entire groups occasionally moved between different sites within the roost. These sites correspond to the territory sites identified at night. During day and night, territorial males were most successful regarding copulations, and territorials performed by far most copulation attempts and female-defence actions but only within their own territory. During the day, non-territorial males attempted to copulate at any site, but were almost always interrupted by the territorial male of the respective territory (i.e. direct female defence). Territorial males rarely attempted copulation outside their territory. Thus, a male can be a dominant individual at one site (in his territory), but a non-dominant individual outside of his territory within the same social group. This resembles a rare form of territoriality called site-specific dominance (sensu Kaufmann [60]), where a territory is defined as an area in which one individual has priority of access to a resource over other individuals who have this privilege at other areas, achieved by social interaction. Hence, based on our diurnal observations R. naso groups can be classified as multi-male/multi-female groups with site-specific dominance relationships among male group members and direct female defence (female-defence polygyny sensu Emlen & Oring [5]).
To the best of our knowledge, such contrasting classifications regarding social dispersion and male mating strategy within a single day-night cycle have not been described yet for any mammal or vertebrate species. However, intraspecific variation in mammalian and vertebrate social systems among different populations or over longer time periods (seasons) is widespread. These include variation in spatial patterns (e.g. [61]), group size (e.g. [62,63]) and mating systems [17]. Several ecological variables have been identified to correlate with variation in social systems (resource abundance, competition for food, predation pressure, population density, habitat saturation; reviewed in [9,10]). For instance, striped mice (Rhabdomys pumilio) in the arid succulent Karoo live in groups of multiple male and female adults in one territory, whereas male and female individuals form solitary territories in resource-rich moist grasslands [61]. Larger groups are formed if predators are present as, for example, in eastern grey kangaroos (Macropus giganteus) when red foxes are present [62] or in musk ox when wolf densities are increased [64]. In Alaska moose (Alces alces gigas), group size positively correlated with distance from cover [65].
Colonies of S. bilineata consist of one to several territorial (harem) males defending roosting space in the day roost. Each harem includes one harem male and on average two to three females but no other adult males [42,74,75]. The common occurrence of a resource-defence strategy among bats including S. bilineata suggests that this strategy represents the plesiomorphic behaviour.
The vast majority of echolocating bats occupy concealed day roosts inside of trees, leaf tents, caves, rock crevices or at least hidden areas of these structures (reviewed in [22]). By contrast, proboscis bats roost on exposed parts of tree trunks, vines, rocks or man-made structures. This roosting behaviour offers them a great variety of roost sites (e.g. also close to their obligatory foraging sites above rivers), while many other bat species are likely to compete for limited concealed or hidden roost sites (e.g. [21,22,26,27,76]). Proboscis bats inhabit roosts that are usually completely exposed ( [42]; L.G. and M.N. 2010-2014, personal observations). This most probably results in a much higher predation risk in proboscis bats than in other bat species. As part of their camouflage, proboscis bats remain totally motionless during almost the whole day while rocking, grooming, stretching, urinating and behavioural interactions are almost exclusively restricted to gusts of wind [46]. This suggests that males are also restricted in their options to prevent other males from joining their group during the day. Interestingly, territorial males tolerated the presence of other males in their territory during day, although copulation attempts from other males during the day occurred. We assume that males only give up their camouflage to interact with other males if it is worth it-e.g. to prevent a copulation by another male (i.e. femaledefence action). Accordingly, predation pressure probably forces territorial males to tolerate the presence of additional males in their territory in order to maintain visual crypsis during the day, but requires territorial males to also guard and defend females directly. This highlights how an ecological factor like predation pressure may drive social dispersion and male mating strategies.
At present, we do not know how territoriality in R. naso is achieved or maintained. However, during approximately three weeks in August and September 2013 prior to the SMP in October, we observed a remarkable number of antagonistic male-male interactions during the day-a behaviour, which is usually only observed during mating periods. These interactions could be used to achieve or maintain territorial status and should be subject to future research. In addition, vocal communication might also play an important role in the mating strategies of males, but further studies are needed to verify this.

Conclusion
Our results on R. naso support the theory that the outstanding diversity of social systems in chiropterans-the second largest mammalian order-is strongly influenced by its astonishing diversity in roosting strategies [21]. By making use of very exposed day roosts, R. naso gained access to a large variety and number of potential day roosts that are usually not occupied by other bat species. Otherwise, access to day roost is frequently assumed to be a limiting factor for bats (e.g. [22,26,27,76]). At night, male R. naso adopt a presumably plesiomorphic territorial strategy, a behaviour that is frequent also in other tropical bats and most closely resembles a resource-defence polygyny. Its contrasting clumped social dispersion during the day is likely to be the result of strong selection for cryptic behaviour in the exposed roosts and requires direct defence of females in addition to male territoriality. However, with site-specific dominance, territorial males adopt a strategy that allows them to maintain primary access to females despite the need of crypsis during the day. Thus, our results on the social system of R. naso show how roosting ecology can shape the social dispersion and mating system of a species. Finally, the night-to-day transition from a rather classical territoriality with male resource defence to clumped mixed sex groups with site-specific dominance and direct female defence illustrates a possible evolutionary trajectory from a resource-defence mating strategy to a female-defence mating strategy by small ecologically driven evolutionary steps.
Ethics. All fieldwork was approved by the Costa Rican authorities (MINAET Ministerio del Ambiente, Energia y Telecomunicaciones and SINAC Sistema Nacional de Areas de Conservación) and was in compliance with the current laws of Costa Rica and Germany.