Cooperative breeding in a plural breeder: the vulturine guineafowl (Acryllium vulturinum)

Cooperative breeding is widely reported across the animal kingdom. In birds, it is hypothesised to be most common in altricial species (where chicks are dependent on parental care in the nest after hatching), with few described cases in precocial species (where chicks are more independent immediately after hatching). However, cooperative breeding may also be more difficult to detect in precocial species and therefore has been overlooked. In this study, we investigate whether vulturine guineafowl (Acryllium vulturinum)—which have precocial young—breed cooperatively and, if so, how care is distributed among group members. Using data collected from colour-banded individuals in one social group of vulturine guineafowl over three different breeding seasons, we found that multiple females can attempt to reproduce in the same breeding season. Broods had close adult associates, and most of these associates exhibited four distinct cooperative breeding behaviours: babysitting, within-group chick guarding, covering the chicks under the wings and calling the chicks to food. Further, we found that offspring care is significantly male-biased, that non-mother individuals provided most of the care each brood received, that breeding females differed in how much help they received, and that carers pay a foraging cost when providing care. Our results confirm that vulturine guineafowl are cooperative breeders, which they combine with an unusual plural-breeding social system. Our study also adds to growing evidence that cooperative breeding may be more widespread among species with precocial young than previously thought, thereby providing a counterpoint to the altriciality-cooperative breeding hypothesis.


INTRODUCTION 53
In some species of birds, mammals, fish and invertebrates more than two adults contribute 54 towards raising offspring, which is known as cooperative breeding. Typically, groups of 55 cooperatively breeding individuals consist of a breeding pair and their offspring from previous 56 breeding attempts that delayed dispersal and provide offspring care to their younger siblings 57 (Clutton-Brock, 2002). As cooperative breeding seems to be a suboptimal reproductive strategy 58 for such individuals, it has received considerable empirical (Koenig & Dickinson, 2016;Stacey 59 & Koenig, 1990) and theoretical (Emlen, 1982;Hatchwell & Komdeur, 2000;Shen et al., 2017) 60 attention over the past five decades. In birds, cooperative breeding is thought to be more 61 prominent among altricial species (Cockburn, 2006;B. Wang et al., 2017). A recent review 62 found that while 11% of species with altricial young breed cooperatively, cooperative breeding 63 is found in only 4% of species with precocial young (Scheiber et al., 2017). The relative paucity 64 of cases among species with precocial young has led to the hypothesis that there exists a link 65 between altriciality and cooperative breeding, because precocial chicks require less care (Ligon 66 & Burt, 2004; N. Wang & Kimball, 2016). Yet, early studies on species with precocial chicks, 67 such as the pukeko (Porphyrio melanotus), played an important role shaping the field of 68 cooperative breeding (Craig & Jamieson, 1990). The disparity in the presence of cooperative breeding between altricial and precocial species 74 could be because species with precocial offspring-which are more independent straight after 75 hatching-have less need for care. However, this logic may be flawed. Due to the high 76 predation risk to the eggs of ground-nesting birds (Thompson & Raveling, 1987), females of 77 ground-nesting precocial species typically have very high nest attendance rates, thus foregoing 78 feeding during incubation. For example, female ring-necked pheasants (Phasianus colchicus) 79 increasingly attend their nest throughout laying and, as a result, suffer a reduction in body mass 80 equating to approximately 19% (Breitenbach & Meyer, 1959). Given the substantial costs 81 associated with egg laying and incubation to breeding females, reduced post-hatching costs 82 could allow breeding females to recover from their reproductive investment and, thereby, 83 produce more chicks (the load lightening hypothesis) (Crick, 2008;Hatchwell, 1999). 84 Furthermore, given that chicks of most precocial species have lower survival probabilities than 85

Study species 152
Although they live in large groups with stable membership for most of the year, vulturine 153 guineafowl form pairs at the beginning of the breeding season (Papageorgiou et al., 2021), after 154 which pairs move separately from the remaining group and the male mate-guarding the female. 155 This is followed by the female laying-and independently incubating-a clutch of 13-15 eggs 156 in a scrape on the ground (Del Hoyo et al., 1994). (We note that our field observations indicate 157 that clutch sizes are smaller, and more variable, than the literature suggests, ranging from 7 to 158 12 eggs across 14 nests found). While the female incubates, the male re-joins the group, and 159 can re-pair with a new female. Vulturine guineafowl chicks are precocial, with the mother and 160 chicks typically re-joining the social group (or part of their social group) very soon after 161 hatching. As in other guineafowl species (Del Hoyo et al., 1994), vulturine guineafowl females 162 appear to receive no help during incubation. Although information about this remains 163 anecdotal, we have never observed evidence suggesting otherwise (e.g., via GPS tracking and 164 a few nests monitored using camera traps). Chicks are highly vulnerable to predation during 165 the first few weeks of life, meaning that they could benefit from protection offered by other 166 group members. 167

Study group and field data collection 168
Our study is focused on a long-term habituated social group living mostly within the fenced 169 area of the Mpala Research Centre (MRC), in Laikipia county, Kenya, and forms part of a long-170 term study population of vulturine guineafowl at the MRC. Mpala is characterised by semi-arid 171 savanna habitat with rainfall averaging between 500 and 600 mm per year, occurring 172 predominantly in two rainy seasons (Young et al., 2003). The natural vegetation is mainly 173 Acacia scrubland, and vulturine guineafowl specialise on the red soils that are dominated by 174 Acacia mellifera and Acacia etbaica. The study group was first colour-banded, and tracked with 175 GPS (He et al., 2022), in September 2016. Group size over the study period ranged from 30-35 176 adults. In the wet seasons spanning April to May 2019, April to May 2020, and October to 177 November 2020, we recorded breeding activity within the study group. This involved following 178 pairs that split from the main group, finding and monitoring the nests, and observing care 179 behaviours post-hatching. Pairs were defined as a female and an associated male who moved 180 together (that is, typically less than 5 m apart) and away from the group (that is, more than 20 181 m from other group members, but generally much further). Before considering two birds to be a pair, they had to be observed moving together for the whole day, but all pairs were recorded 183 even if they were paired only for a single day. By following pairs, we eventually located and 184 monitored the nests of all breeding females. When the clutch was expected to hatch (25 days 185 after the start of incubation), we searched for the female and chicks. 186

187
We also collected data on associations between group members and broods by recording group 188 composition data in the morning (0600-0930 hrs) and evening (1700-1900 hrs) for an average 189 of four days per week. Each time members of the group were encountered, the identity of every 190 adult bird present was recorded, as well as the total number of banded and unbanded adults, 191 and the number of chicks present from each of the broods. All birds that were present together 192 in an area (that is, within sight of each other and behaving cohesively) were considered to be 193 part of the same group, and given the same group identifier. Within these individuals, if some 194 were more spatially clustered (e.g., within two metres of each other but more than 10m from 195 other individuals), then these were recorded as a subgroup within the broader group (or as pairs, 196 if they remained distinctly separate for the entire day). 197

198
Although chicks could not initially be marked individually, broods (up to N=3 per season) were 199 generally identifiable because they hatched asynchronously, meaning that there was a clear size 200 difference between chicks from different broods. From these differences, we ascertained that 201 those chicks from the same brood remained together the vast majority of the time. However, 202 there were two exceptions. During two seasons, multiple broods merged after hatching at about 203 the same period. We therefore found a corresponding increase in overlap among the adults 204 associated with each brood over the second and third season, but we cannot know whether the 205 patterns we observed are common or not when broods are very close in age. Association data, 206 between adults and broods, were collected for each breeding season until chicks were four 207 months old. 208

209
In addition to collecting association data, we observed and recorded cooperative breeding 210 interactions via all-occurrence sampling (Altmann, 1974) over the three seasons. Specifically, 211 we found four types of cooperative breeding interactions to occur between adults and chicks. 212 The first is babysitting behaviour (BBS), whereby individual A stays more than 20 metres from 213 the rest of the group members with one or more chicks. The second is within-group chick 214 guarding behaviour (GRD), whereby an individual does not allow other adult group members 215 whereby an individual covers one or more chicks under its wings. The final cooperative 217 breeding behaviour is chick feeding (CFD) ( Figure 1B), whereby an individual performs soft 218 trills (a type of vocalisation) calling chicks to a food item. For each interaction, we recorded 219 the identity of the actor, the identity of the recipient brood, the event duration in the case of 220 cover events, and the number of chicks involved in each interaction. Because broods often 221 moved separately through the group's home range, we aimed to distribute our observation effort 222 evenly across broods over each season. We defined any individual that engages in one or more of the four cooperative breeding 229 behaviours described above as a 'carer'. These carers can include non-parents (e.g. older siblings 230 from previous years (Ligon & Ligon, 1978), unrelated individuals (Clutton-Brock, 2002) or 231 parents (both genetic and social). Due to a lack of data on paternity, we can only conclude 232 which female is the social mother. However, from our monitoring of the study group, we know 233 that many carers are the older (social) offspring of the female whose chicks they provided care 234 to, and that some individuals in our study engaged in cooperative breeding behaviours prior to 235 reproductive maturity (estimated age of maturity is 2 years (Del Hoyo et al., 1994), and males 236 only reach the size of adults after more than 12 months). Thus, for each bird in each season, we 237 provide three levels of evidence that point to some individuals being non-breeding helpers: (1) 238 individuals that provide care prior to being sexually mature, (2) individuals that provide care to 239 the brood of their social mother, and (3) individuals that provide care prior to ever having been 240 observed forming a pair themselves. 241 242 Finally, to quantify whether carers pay a cost by engaging in cooperative breeding behaviour, 243 we recorded videos during a period when the chicks were younger than five weeks (in the final season). From these videos, we were able to quantify the foraging activity by birds in each 245 recording session for the duration that an individual could be tracked in the frame without 246 moving out of the frame (due to occlusion or due to the movement of the person holding the 247 camera  We used a social network approach to identify the main carers for each group in each of the 258 three seasons. Using the group composition data, we calculated the rate of attendance of each 259 group member to each brood. This rate was calculated by dividing the number of subgroups 260 that the individual was observed in that comprised the brood by the total number of subgroups 261 that the individual was observed in, limited to group composition observations containing that 262 brood (because not all broods were observed in a given sampling day). For example, if an 263 individual was observed 10 times in different groups that contained brood A, of which 8 times 264 it was in a subgroup with brood A, the rate of attendance of the individual to brood A was 0.8. 265

266
To identify which individuals occurred with the brood more than expected by chance we used 267 a simple permutation test (Farine, 2017), which consisted of randomising the subgroups that 268 the focal chicks were contained in. The permutation test worked as follows: for each day, K 269 observations of the chicks were randomly allocated to subgroups observed on that day, where 270 K corresponds to the number of subgroups the chicks were originally observed in that day (thus, 271 all potential carers remained in the same subgroup, but the chicks were moved between 272 subgroups for the purpose of the permutation test). The rate of attendance for each individual 273 was re-calculated from these permuted data. The permutation test therefore maintained the 274 number of times each individual and each brood was observed, and the same number of total 275 groups and subgroups. This permutation procedure was repeated 1000 times, thereby generating 276 a distribution of the rate of attendance values for each individual. From this distribution, we 277 extracted the identities of the individuals whose observed rate of attendance was higher than 278 95% of the rates generated by the permutation procedure (i.e., significant at P≤0.05 using a one-279 tailed test). These individuals were recorded as having significantly higher attendance to that 280 brood than expected by chance. This process was repeated for all broods in all seasons. Finally, 281 to test whether males were disproportionately represented as carers, we conducted a two-sample 282 proportion test that compared the proportion of males to females among the carer and non-carer 283 categories (excluding mothers) in each of the three seasons. 284 285 We used the cooperative breeding interaction data to characterise the relative contribution of 286 each significant associate to the brood. From these data, we determined whether the mother 287 provided most of the care or not, and (if not) how help received varied between broods. First, 288 we counted the number of each interaction directed towards a focal females' chicks. Because 289 babysitting and within-group guarding behaviours were expressed relatively rarely, they were 290 combined into one category for analysis. Mothers clearly did not give the majority of the care 291 to their chicks, so we then tested whether some females received more help than others in each 292 of the three seasons. As we did not have an equal opportunity to observe each brood, and 293 difference can lead to spurious outcomes (Hoppitt & Farine, 2018), we used a two-sample test 294 for equality of proportions. We thus conducted pairwise contrasts of the proportion of total help 295 each mother gave to the chicks in her brood. This comparison was performed across the broods 296 within each of the three seasons. Significant effects mean that the difference in the proportion 297 of help given by the mother was significantly lower in one brood than the other. 298 299 Finally, we tested whether males gave overall more help than females. To do this, we counted 300 the number of chick-feeding interactions for each individual in each season as they were the 301 most prevalent compared to other interactions. We then constructed a generalised linear mixed 302 model with a Poisson error distribution, sex as the only predictor of the count of chick feeding 303 interactions per individual per season, and individual identity and season as nested random 304 effects. We excluded mothers from this analysis. While we could not explicitly account for 305 differences in the number of observations, in practice every individual had the same opportunity 306 to be observed helping each time we collected data. Thus, under random feeding, we would not 307 expect our model to produce a significant difference. 308 309

Do non-mother individuals pay a cost for caring? 310
Using the data extracted from the videos, we quantified the proportion of the time spent foraging 311 (response variable) and tested whether this was predicted by whether the focal individual was involved in the COV (wing-covering) behaviour (1) or not (0) (binary independent variable). 313 To do this, we used a general linear model with a binomial error distribution. Ideally, we would 314 have accounted for repeated observations of some individuals using mixed models, but we 315 unfortunately could not always identify the individual recorded, and on only three occasions 316 could we do so while an individual was covering the chicks. However, the data were relatively 317 well distributed amongst individuals (average 2.25 observations per individual, when the 318 identity was known). We thus do not expect repeated observations to have a major impact on 319 our conclusions.

345
Carers are predominantly male, and include non-breeders 346 There were four key insights from our analyses of significant associates (Figure 2). First, each 347 brood had a number of individuals that were observed with the brood more than expected by 348 chance, and not all individuals were consistently detected with a brood. Second, there was 349 relatively little overlap amongst the individuals that were significantly associated with each 350 brood, indicating that each brood had a distinct set of associates. Third, there was a significant 351 male bias among associates (Table 2)  Mothers also consistently provided a small proportion of the total care that their offspring 380 received. For example, in the first season, out of the 20 cover events recorded for YOBK's 381 chicks, the mother was found to cover chicks in one event for only one minute. She also 382 contributed only 42 of the 330 chick feeding events and one out of seven between group 383 guarding and within-group guarding events. However, not all females received the same 384 amount of help from carers. YOBK, a female who is known to have successfully bred before, 385 generally received more help in food provisioning and covering chicks under the wings than 386 the other females (see Supplementary Tables S1-S3), whereas there were fewer differences in 387 the help received among the other mothers (see Supplementary Tables S4-S9). YOBK received 388 more help with chick provisioning in each of the other two seasons, with the exception of one 389 female (GAGA) in the third season (see Supplementary Table S7). As cover interactions were 390 relatively rare, some comparisons could not be drawn effectively. There were no instances of 391 babysitting and guarding behaviours in the first season for two mothers as well as the third 392 season for all the mothers. 393 Overall, males provided significantly more caring interactions than females. Of the 2506 chick 395 feeding interactions given by non-breeding individuals that we recorded across the three 396 seasons, 2451 were given by males. This translates to an estimated 32.7 times more interactions 397 per individual per season for males relative to females (Table 3). 398 399

Non-mother individuals pay foraging costs while caring 406
Based on 27 observations of covering chicks under the wings, where cover duration data were 407 available, individuals performed this behaviour on average for 16 minutes (range=1-60 408 minutes). From the video data, we estimated that birds pecked on average 0.10 times per second 409 when they were not performing the cover behaviour (range=0-1.47) and pecked on average 410 0.01 times per second when they were performing the cover behaviour (range=0-0.01), a 411 significant difference ( In this study, we provide the first evidence for cooperative breeding in the vulturine guineafowl 424 (Acryllium vulturinum). Vulturine guineafowl groups contain multiple females (we observed 425 up to 8 breeding females out of 13 adult female group members) that can attempt to reproduce 426 within the same breeding season, meaning that they are also clearly plural breeders. This makes 427 them similar, in terms of social and reproductive system, to the sympatric superb starling 428 (Lamprotornis superbus) (Rubenstein, 2006). We found that carers provide the majority of the 429 care and pay a foraging cost when doing so. As in most avian cooperatively breeding birds, we 430 found that caring associations and caring interactions were very male biased. Overall, vulturine 431 guineafowl have the common hallmarks of plural cooperatively breeding species. 432 The predominant hypothesis explaining the evolution of cooperative breeding behaviour is 433 inclusive fitness (Hamilton, 1964). By helping, individuals can gain direct fitness benefits 434 through routes including parentage of offspring (Richardson et al., 2002), territory inheritance 435 (Kingma, 2017), group augmentation (Ligon & Ligon, 1978;Wright et al., 2010) or 'pay-to-436 stay' benefits (Wong & Balshine, 2011). In contrast to these direct routes, indirect fitness 437 benefits may be accrued by helping kin, and can be through enhanced offspring number 438 (Blackmore & Heinsohn, 2007) or survival (Hatchwell et al., 2004), as well as increased 439 parental survival (Downing et al., 2021) or reproductive rate (Russell, 2003). As male birds are 440 typically philopatric (Greenwood, 1980), non-breeding males (unlike the dispersing females) 441 are likely to be related to breeders and their offspring, and may thereby gain indirect fitness by 442 helping (Dickinson & Hatchwell, 2004). In precocial species where breeders invest 443 considerably in reproduction, additional offspring care from such group members could 444 reasonably generate indirect fitness benefits-such as greater offspring survival or enhanced 445 parental current or future reproductive investment-that outweigh the costs of helping. 446

447
Cooperatively breeding vulturine guineafowl conform to the hypothesised patterns of 448 cooperative breeding in birds. Carers are likely to increase the reproductive success of the 449 female or enhance the female's survival through reducing her need to care directly for the 450 chicks. Specifically, we found that the female only provides a small proportion of all the chick 451 caring interactions, which is likely to be important for her to regain the body condition she lost 452 through laying and attending to the nest. Further, as female vulturine guineafowl may breed 453 twice per year-as we found with several females being successful in both May 2020 and 454 November 2020 (Table 1)-additional caring may be key to females regaining body condition 455 in time for the next reproductive opportunity by reducing the inter-birth interval (Ridley & 456 Raihani, 2008). Thus, our study provides further support for a number of general hypotheses 457 surrounding cooperative breeding. 458 There is considerable variation in the structure of cooperatively breeding groups among species, 459 ranging from a single breeding pair with associated helpers to multiple breeding units, which offspring there are in a cooperative polyandrous group (Dietz & Baker, 1993;Goldizen, 1989). 467 However, in contrast to the golden lion tamarins, we found clear evidence for specialisation 468 among non-breeding group members in term of which female they helped-likely their 469 mothers. Plural breeding may be facilitated by breeding during a wet season, coinciding with a 470 temporary increase in resource abundance (both food and safe nesting sites) that allow many 471 individuals to reproduce at the same time. Our study also highlights the variance that can 472 emerge in terms of who is a successful breeder, with many clutches depredated during 473 incubation or early in the life of chicks. We ruled out dominance as a determinant for the 474 breeding success in females, because the structure of the female dominance hierarchy is not 475 very clear relative to that of males (Dehnen et al., 2022). However, more data are needed to 476 more explicitly evaluate the link between breeding and dominance, and whether factors such as 477 experience might drive variation in nesting rates and nest success among females. 478 In this study, we identified four juvenile-directed cooperative behaviours by non-parents, may not be immediately obvious to observers and typically require following groups at small 488 distances to make close observations. Doing so in our study was made substantially easier by 489 the focal social group being habituated to our presence and due to their home range being 490 centred within a fenced area that was safe to walk in. 491 The associates of each brood were predominantly male group members, and they typically 492 provided more care than the mother did. These results are consistent with those reported in 493 other cooperatively breeding Galliformes. One exception in our data was the second breeding 494 season, although this season was notable for having many failed nests, which could have 495 stimulated care from females. Why males attend to broods more than females in vulturine 496 guineafowl, given that subadult females can remain in their natal territory for several years 497 before dispersing (Klarevas-Irby et al., 2021), remains unknown, but this pattern has been 498 observed widely in other species (Green et al., 2016). Further, we found that while some 499 females were detected as significant associates of the brood, male carers provided nearly 98% 500 of all the chick feeding interactions (excluding those from the mothers). This is consistent with 501 the grey-crowned babbler (Pomatostomus temporalis), where only the number of male helpers 502 increased reproductive success (Blackmore & Heinsohn, 2007). 503 By studying the social network of individual carers and broods, we showed that carers appeared 504 to be generally brood specific. Among these, we found several lines of evidence supporting the 505 hypothesis that vulturine guineafowl carers include non-breeding males. This includes subadult 506 males not only associating significantly with their mother's brood, but also providing care 507 before they are fully grown (e.g., CT311, Figure 2). Others (e.g., CT317, a male that cared for 508 YOBK's brood in the second and third seasons) were part of the brood that their social mothers 509 cared for before being observed providing care themselves. Further, the significant associates 510 with each brood are relatively consistent over years. For example, YKOY was not detected as 511 a significant carer in the first season, when GOOP did not successfully breed, but was 512 significantly associated with her brood in the following two seasons. Thus, care is not given 513 randomly within the group, and a substantial proportion of the care that we observed came from 514 individual males that are unlikely to have any paternity with the brood. 515 A consistent pattern that emerged across all three seasons is that mothers did not provide the 516 majority of care to their chicks. The large amount of help given by non-mother is perhaps 517 unusual (Green et al., 2016). For example, in purple gallinules (Porphyrio martinicus), breeding 518 adults provide the majority of the care to chicks, which are sub precocial (Hunter, 1987), while 519 both male and female purple gallinules participate in incubation (Gross & Van Tyne, 1929). 520 One reason why female vulturine guineafowl receive so much help in raising their offspring 521 may be the high cost they pay during incubation-meaning that recovering their body condition 522 might compromise the amount of care they can provide to the current brood, which is then 523 offset by the care provided by other individuals. Further, care involves not only food 524 provisioning, vigilance against predators and agonistic behaviours against intruders (Clutton-525 Brock & Manser, 2016) but also learning so as to enhance foraging skills (Cant et al., 2016;526 Heinsohn, 1991 in meerkats (Suricata suricatta), helpers lose weight when they participate in cooperative 530 breeding activities, such as feeding the young (Russell, 2003). Similarly, in white-winged 531 choughs (Corcorax melanorhamphos), helpers lose weight when performing incubation, in 532 addition to costs they incur by choosing to remain in their natal territory (Heinsohn & 533 Cockburn, 1994 Figure S1. Amount of help given by the mother (red points) and helpers (black points) for the 777 YOBK brood in season 1. 778 779 780 Figure S2. Amount of help given by the mother (red points) and helpers (black points) for the 781 WOBY brood in season 1.