University of Groningen Better together Groenewoud, Frank

Better together

Groenewoud, Frank

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from

it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date:

2018

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Groenewoud, F. (2018). Better together: Cooperative breeding under environmental heterogeneity.

University of Groningen.

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

Chapter 6

Experimentally induced

anti-predator responses are

sex specific and mediated

by social and environmental

factors in a cooperatively

breeding passerine

Frank Groenewoud, Sjouke A. Kingma, Kat Bebbington, David S. Richardson & Jan Komdeur

ABSTRACT

Nest predation is a common cause of reproductive failure for many bird species, and vari-ous anti-predator defense behaviors have evolved to reduce the risk of nest predation. How-ever, trade-offs between current reproductive duties and future reproduction often limit the parent’s ability to respond to nest predation risk. Individual responses to experimen-tally increased nest predation risk can give insights into these underlying trade-offs. Here, we investigate the social and ecological factors that underlie these trade-offs by experimen-tally manipulating the risk of nest predation using taxidermic mounts in the cooperative breeding Seychelles warbler (Acrocephalus sechellensis). Our results show that dominant fe-males alarm called more often when they confront a nest predator model alone than when they do so with a partner, and that individuals that confront a predator together attacked more than those that did so alone. Dominant males increased their anti-predator defense by spending more time nest guarding after a presentation with a nest predator, compared to a non-predator control, but no such effect was found for females, who did not increase the time spent incubating. In contrast to incubation by females, nest guarding responses by dominant males depended on the presence of other group members and food availability. These results show that while female investment in incubation is always high and not de-pendent on social and ecological conditions, males have a lower initial investment, which allows them to respond to sudden changes in nest predation risk.

INTRODUCTION

Predation risk is an important factor driving changes in life-history and behavior in many animals (Barbosa & Castellanos 2005; Caro 2005; Creel & Christianson 2008). In birds, nest predation is one of the most common causes of nest failure, and is therefore one of the key drivers in the evolution of avian breeding biology (Ricklefs 1969b; Martin 1995). For example, individuals can vary nest site location or clutch size according to predation risk (e.g. Martin 1995; Eggers et al. 2006; Dillon, Conway & Skelhorn 2018), and parents might visit the nest less often when nest predation threat is high (e.g. Ghalambor & Martin 1999; Fontaine & Martin 2006; Ghalambor, Peluc & Martin 2013). If anti-predator behavior is cost-ly then individuals experiencing different levels of nest predation risk should adjust their investment in such a behavior accordingly (Lima 2009). However, investment is often con-strained by trade-offs between current and future reproduction, or because investment in the current breeding attempt, e.g. through anti-predator behaviors, precludes investment into others important activities, such as obtaining additional matings (Trivers 1972; Stearns 1989). Experimental studies are necessary to determine which conditions shape anti-pred-ator responses, and the trade-offs underlying anti-predanti-pred-ator responses, but such studies are scarce (Lima 2009). Here, we experimentally increased nest predation risk in a cooperative-ly breeding passerine to provide insights into the social and environmental factors that shape the costs of anti-predator responses on an ecological time-scale.

Increased nest attendance or vigilance is a common response to increased nest predation risk, and can improve predator detection (Montgomerie & Weatherhead 1988; Caro 2005) and nesting success (Komdeur & Kats 1999). Such behavior is also hypothesized to be costly, because individuals are unable to simultaneously invest in other activities, such as forag-ing (Komdeur & Kats 1999; Duncan Rastogi, Zanette & Clinchy 2006). Therefore individu-als face a trade-off between investing in their current brood by increasing vigilance and reducing nest predation risk, or investing in self-maintenance and, thereby, potential fu-ture reproduction (Stearns 1989). However, the costs of increased investment in vigilance are not necessarily the same for all individuals. For instance, individuals in areas with high food availability may be better able to increase their food uptake after sustained periods of investment, and therefore suffer fewer costs of nest defense compared to individuals from lower quality areas (Duncan Rastogi et al. 2006). Similarly, the costs and benefits of increased anti-predator behavior can also differ between males and females, in species with bi-parental care (Montgomerie & Weatherhead 1988). For instance, males, who are larger in many passerine species (Ranta, Laurila & Elmberg 1994; Mills 2008), have been suggested to engage more in risky defense against predators than females, either because they are more effective and/or have lower risk of injury (Andersson & Norberg 1981). Thus, sex differences,

and variation in environmental conditions, can alter the costs and benefits of anti-predator behavior, and shape the trade-offs between investment in current and future reproduction. A potentially important component of predator defense strategies is that individuals may respond differently to predators depending on the social context (Clutton‐Brock 1991). For instance, jointly confronting a predator might increase defense success or reduce in-jury risk (Weatherhead 1989). Therefore individuals that encounter predators alone could choose not to engage a predator, but to use more risk-averse tactics, such as alarm calls. The benefits of joint predator defense have been considered as one of the major benefits of group-living in many social bird species (Krause & Ruxton 2002). Moreover, the presence or additional investment by other individuals might also lead to changes in anti-predator be-havior. For instance, the presence of subordinates in cooperatively breeding species could lead to lower investment for the breeders (i.e. load lightening; Johnstone 2011). However, similar and increased levels of investment by dominants in the presence of subordinates have also been shown (Hatchwell 1999; Valencia et al. 2006). The effects of the social en-vironment on the expression of individual anti-predator behaviors are therefore complex and not well understood.

In the facultative cooperatively breeding Seychelles warbler (Acrocephalus sechellensis) egg predation is the primary cause of nest failure and an important aspect of fitness (Komdeur & Kats 1999). The main nest predator is the Seychelles fody (Foudia sechellarum; hereafter ‘fody’), an endemic weaver that has been observed taking eggs from unattended warbler nests (Komdeur & Kats 1999). Seychelles warblers on Cousin Island typically lay single egg clutches (91% of clutches; Komdeur 1996b; Bebbington et al. 2017), which means that a pre-dation event in most cases renders the entire breeding attempt unsuccessful (Komdeur 1996b). Thus, nest predation is both common and costly for Seychelles warblers, and in response the species has evolved direct (attacks and alarms) and indirect (nest guarding) anti-predator behaviors (Komdeur 1991; Veen et al. 2000). Seychelles warblers are entirely insectivorous and insect availability is variable across the island (Komdeur & Daan 2005), consequently, local food availability may play an important role in modulating the expres-sion of anti-predator behaviors.

Several social components of the Seychelles warbler system might be central in driving an-ti-predation behaviors. First, nest defense tactics are sex-specific: males are often observed nest guarding (showing vigilance behavior close to the nest; Slack 1976), which reduces the likelihood of nest predation (Komdeur & Kats 1999). Females rarely nest guard, but incuba-tion (a female-only behavior) also prevents egg predaincuba-tion as it prevents fodies from gaining access to the egg (Komdeur 1991; Komdeur & Kats 1999). As such, it is the combined effort of males and females that determines the extent to which the nest is protected against

pred-ators. However, since nest guarding has evolved specifically to counter nest predation, this behavior should be much more flexible than modifications of incubation, which is also de-termined by the thermal requirements of the eggs. Second, males mostly guard the nest in the absence of incubation by the female, but there is some overlap between nest guarding and incubation, particularly at the beginning and end of female incubation bouts. Preda-tors can therefore be confronted by either the male or female alone, or cooperatively, and variation in anti-predator responses can indicate different costs of nest defense due to the social environment. Lastly, dominants can be accompanied by 0-4 subordinates (Komdeur 1991; Kingma et al. 2016a), which can help with incubation (females only) and provisioning (Komdeur 1991, 1994a). However, much less is known about the role of subordinate Sey-chelles warblers in mitigating predation risk.

Here we used an experimental approach to increase the perceived risk of egg predation in Seychelles warblers by using a mounted fody model in combination with fody audio play-back to simulate an imminent threat at the nest. A similar method has been used in the past to successfully test for innate nest defense behaviors in this species (Veen et al. 2000). We then assessed the direct anti-predator responses of individuals to these mounted mod-els (attacks and alarm calls), as well as the subsequent changes in indirect anti-predator behavior (incubation and nest guarding behavior), and compared these to a presentation with a non-predator control model. We ask three main questions: (i) Do parents increase anti-predator behaviors (nest guarding or incubation) in response to experimentally in-creased nest predation risk and are these responses dependent on the availability of food or parental sex? (ii) Do individuals respond differently to a direct predator threat depending on whether they confront a predator alone or together? (iii) Do helpers contribute towards nest defense, and how does helper presence affect the dominant birds’ anti-predator be-havior? Our results shed light on how group members engage in different types of nest defense behaviors and how the trade-off between these behaviors are affected by social and/ or environmental contexts.

MATERIALS AND METHODS

The Seychelles warbler is a small cooperatively breeding passerine endemic to several is-lands in the Seychelles archipelago. The main study island of Cousin (ca 29 ha; 4°19’53.6”S 55°39’43.3”E) is saturated with Seychelles warbler territories, and the population is relative-ly stable around 320 adult birds (Brouwer et al. 2009). The long-term monitoring effort on this population means that since 1997 nearly all birds (97%) on the island are individually

identifiable by a unique combination of color rings and a metal ring (Hammers et al. 2015; Komdeur et al. 2016). The sex of all ringed individuals was confirmed by molecular sexing (Richardson et al. 2001). To find nests, dominant breeding females in each territory were fol-lowed for at least 15 minutes every 3-4 days. To determine the date of egg laying, we checked the nest in each territory at least every fourth day before nest completion and every other day after that.

Seychelles warblers are strictly insectivorous and take most of their prey from the under-side of leaves (Komdeur 1991). Therefore territory quality can be accurately estimated in terms of the availability of their arthropod prey, according to the methods described in Komdeur (1992) and Brouwer et al. (2006). Briefly, we counted the number of arthropods on the underside of 50 leaves of all main tree species at 13 different locations, representative of each part of the island. We then estimated the cover of each of these tree species at dif-ferent strata of the canopy for each territory. Arthropod counts per tree species were then multiplied by the cover of each tree species for each territory, and the resulting measure of territory quality (i.e. arthropod density) was log transformed and mean centered.

Nest predator presentation

Predator presentation experiments were performed on Cousin Island between the 19th _of July and the 2nd _{of September 2015, between 10-12 am or 2-5 pm. The Seychelles fody is} cur-rently listed as “near-threatened” on the IUCN red list (BirdLife International 2013), so we were unable to obtain a taxidermic model of this species, and instead used a mounted fe-male house sparrow (Passer domesticus), which is very similar to the Seychelles fody in size and appearance. An earlier investigation into predator recognition in the Seychelles war-bler showed no differences in anti-predator responses between a caged Seychelles fody and a caged mounted female house sparrow (Veen et al. 2000). Two different mounted house sparrows were used to increase generalizability (Johnson & Freeberg 2016). Similar to Veen et al. (2000), we used a mounted model of a barred ground dove (Geopelia striata), which occurs naturally on Cousin, as a non-predator control. Using an eight meter long fiberglass telescopic pole, we presented either a mounted house sparrow (N = 19) or a mounted barred ground dove (N = 11) approximately one meter from the Seychelles warbler nest during in-cubation. Practical constraints in the field meant that experiments were performed during different stages of nest incubation (mean number of days after onset of incubation = 8.9, range = 3–15 days). All but one nest treatment event were performed on different nests in different territories, but in one territory we used two different predator models on sub-sequent nests. Simultaneously with the presentation, we played calls of the species mod-el used (obtained from the Xeno-canto bird sound database (www.xeno-canto.org)) on a portable mp3 player between 5-10 m from the nest. Audio playbacks were standardized by

removing background noise and repeating two call bouts every thirty seconds for the full length of the presentation using Adobe Audition CC (see supplementary information). We used different recordings for each of the two mounted house sparrow models. We recorded the number of attacks – pecking and dive bombing (i.e. rapidly flying overhead and pecking at the model in flight) – and alarm calls by the dominant male, dominant female and sub-ordinates of either sex (when present) using a GoPro (Hero 3+) mounted on the telescopic pole, one meter from the model. We used a voice recorder, in addition to the video record-ings, to record the identity and behavior of birds during the experiment. These recordings were later processed and the number of alarm calls and attacks was quantified using the software BORIS (Friard & Gamba 2016). The presentation ended 5 minutes after the arrival of the first territory group member at the nest area, visible to the observer.

Nest guarding was defined as individuals perching < 2.5 meters from the nest while no fe-male was on the nest (Komdeur & Kats 1999). To assess whether individuals showed more nest guarding (males) or incubation behavior (females) after an encounter with the simu-lated nest predator, we recorded the behaviors of the group individuals for one hour both before and after the presentation of the mounted bird. During the second observation – which started five minutes after the end of the predator or control presentation – we used the same playback to simulate the continued presence of the predator or control bird in the territory. In all but two cases, observations before and after the presentation experiment were recorded by one of three different observers.

Statistical analyses Attacks and alarms

Attacks towards the non-predator model (dove) were rare: in all 11 non-predator presen-tations, only three individuals (in three different territories) attacked the model. There-fore in the analysis of attacks, we focused on responses towards the predator model only, while alarm call analysis also included the non-predator model. We fitted either alarms, or attacks, as the response variable in separate generalized linear mixed models assuming a Poisson error. We fitted group member status (i.e. dominant male, dominant female, or subordinate) and the presentation type (fody versus dove; for alarms only) as predictors. To determine whether rates of alarm calls or attacks differed when individuals were alone or together, we also included the presence of other defenders as a binary variable. Individuals were counted as ‘arriving alone’ when they arrived at least 10 seconds before another group member. Individuals that ‘arrived together’ were either those that joined a partner that was already present, or that arrived with another individual within 10 seconds of each other. To account for the time spent either alone or together we included this variable as an offset in both models. We also analyzed whether there were differences between the two different

predator models and we included territory ID as a random effect to account for the non-inde-pendence of observations within territories. We included incubation stage (i.e. the number of days after the onset of incubation) to account for potential differences in aggression as a result of motivational state affected by, for example, renesting potential. Following Veen et al. (2000), we only analyzed the behaviors during the first two minutes of observations for each individual after arrival.

Incubation and nest guarding responses

To investigate whether individuals increased nest guarding (dominant males) or incuba-tion (dominant females) after being confronted with a nest predator, we analyzed these behaviors separately using linear mixed models with varying intercepts for each territo-ry. We included the interaction between presentation type (predator or non-predator) and watch type (before or after the presentation) to account for different responses between the nest predator and the non-predator control. We also analyzed the effect of arthropod density, and whether that territory had an incubating subordinate (since nest guarding by subordinates was uncommon; see results), on the incubation and nest guarding responses of dominants. We included the interactions watch type x arthropod density, and watch type x incubating subordinate present to test whether responses varied under these environmen-tal and variables. We allowed for random intercepts between observers to account for be-tween-observer variation.

We used an information theoretic model selection approach based on the Akaike Infor-mation Criterion (Akaike 1973) with small sample size correction (AICc; Hurvich & Tsai 1989)). We fitted full models as described above, and dropped variables if doing so led to a reduction in out-of-sample deviance (i.e. AICc) sensu Burnham and Anderson (2002), and Burnham et al. (2010), starting with higher level interactions. Variables that were of par-ticular interest (e.g. presentation type) for inference were not removed. Variables that were removed, including interactions, were re-entered for estimation of their effects using likeli-hood ratio tests (LRT) on nested models assuming a χ2_{- distribution. All models were fitted} using the package lme4 (Bates et al. 2014) and model selection and predictions performed with AICcmodavg (Mazerolle 2013). We used package multcomp (Hothorn et al. 2008) to test whether slope estimates contained in higher level interactions differed significantly from zero. All effect sizes given in the results section are means ± standard errors.

RESULTS

All dominant males (N = 28; one male died just prior to the experiment) and all dominant females (N = 29) showed nest guarding and incubation behavior respectively during the observation periods, and 27 dominant males and 26 dominant females were present during the actual model presentation. Of the eight subordinates seen during either stage of the ex-periment, three displayed both attacks/alarms and incubation behavior (all females), two showed alarms or attacks but no incubation or nest guarding, two showed nest guarding or incubation but no nest defense, and one individual did not contribute in any way (this latter individual was harassed by the dominants when he tried to participate in nest de-fense). Interestingly, the five subordinates that participated in direct defense (i.e. attacks or alarms) always arrived after the dominant breeding female or dominant breeding male: therefore differences in attack or alarm rates depending on whether these subordinates confronted the model presentation together or alone could not be estimated. Of the total time spent in either incubation or nest guarding, dominant females spent most of their time incubating (97.0%), while dominant males almost exclusively nest-guarded (99.2%) and subordinates (four females) showed a mixed investment (82.4% incubation and 17.6% nest guarding). One subordinate male nest guarded for 84 seconds, but showed no direct anti-predator behaviors.

Alarm calls and attacks in response to model presentation

Dominant females alarm called more than breeding males (mean ± SE = -1.01 ± 0.09, z = -10.51, DF = 1, p < 0.001, Fig. 6.1A) and subordinates (-1.04 ± 0.17, z = -6.07, DF = 1, p < 0.001, Fig. 6.1A), but there was no difference in alarm calling rate between dominant males and subor-dinates (-0.03 ± 0.18, z = -0.16, DF = 1, p = 0.99, Fig. 6.1A). Individuals alarm called less when they confronted the mounted model together than when they were alone (-0.36 ± 0.13, χ2 ₌ 7.97, DF = 1, p = 0.005, Fig. 6.1A), and this effect was stronger for dominant females than for dominant males (interaction effect: 0.82 ± 0.31, χ2 _{= 7.18, DF = 1, p = 0.005, Fig. 6.1A).} Individ-uals tended to alarm more during the predator presentation than during the non-predator presentation, but this effect was not significant (0.84 ± 0.49, χ2 _{= -3.02, p = 0.08, DF = 1, Fig.} 6.1A), and there were no differences between the two predator models used (-0.07 ± 0.56, χ2 = 0.12, DF = 1, p = 0.90). The number of alarms was independent of incubation stage (-0.08 ± 0.05, χ2 _{= 2.16, DF = 1, p = 0.14).}

There was no difference in the number of attacks by the dominant females, dominant males or subordinates (χ2 _{= 0.45, DF = 2, p = 0.80). Individuals attacked the predator model more} often when they were together than when they were alone (0.78 ± 0.27, χ2 _{= 9.33, DF = 1, p <}

0.01, Fig. 6.1B), but this did not differ between dominant females and dominant males (0.42 ± 0.83, χ2 _{= 0.51, DF = 1, p = 0.61). Individuals tended to attack one of the predator models} more than the other, but this effect was not significant (1.59 ± 1.04, χ2 _{= 2.83, DF = 1, p = 0.09).} Individuals attacked the predator model less if the eggs had been incubated for more days (-0.29 ± 0.10, χ2 _{= 8.23, DF = 1, p < 0.01).}

Nest guarding and incubation responses to increased nest predation risk Nest guarding

Dominant males increased their nest guarding duration significantly after the nest pred-ator presentation (0.21 ± 0.04, z = 5.47, p < 0.001), while males nest guarded less after the non-predator presentation (-0.09 ± 0.04, z = -2.03, p = 0.04). Consequently, dominant males increased their time spent nest guarding (i.e. the slope between before and after the model presentation) more after a nest predator presentation than after a non-predator presenta-tion (0.29 ± 0.06, χ2 _{= 19.79, DF = 1, p < 0.001; Fig. 6.2). Although the number of territories} with incubating subordinates was low (n = 4), dominant males had a significantly smaller nest-guarding response when there was an incubating subordinate present in the territory (-0.22 ± 0.09, χ2 _{= 5.97, DF = 1, p = 0.02; Fig. 6.3A). Consistent with previous results (Komdeur} & Kats 1999), nest guarding by dominant males before the predator presentation increased with arthropod density, but was independent of arthropod density after the presentation (-0.07 ± 0.03, χ2 _{= 5.21, DF = 1, p = 0.02; Fig. 6.3B).} DF DM SUB 0.0 0.5 1.0 1.5 2.0

Number of attacks (mean ± SE)

A

B

Number of alarms (mean ± SE)

FIGURE 6.1 The mean model predicted (± SE) number of alarm calls (A), and the number of attacks (B) per minute for Seychelles warblers when they were alone (open circles) or together (filled circles) during an experimental presentation of a nest predator (N = 19) or non-predator (N = 11). Individuals arrived alone when they were present at least 10 seconds before the arrival of another bird. DF = dominant female, DM = dominant male, SUB = subordinate.

Before After Before After

Predator Non predator

0 10 20 30 40 50 60

Nestguarding/incubating (mean ± SE min hour

-1)

Incubation by breeding female Nest guarding by breeding male

***

ns

FIGURE 6.2 Time spent nest guarding or incubating by Seychelles warbler dominant males (blue circles) and dominant females (red triangles), respectively, before and after a nest predator (N = 19) or non-predator (N = 11) presentation. Thick lines represent mean predicted responses with standard errors, while thin lines show individual responses. Significance indicators (*** = P < 0.001, NS = not significant) relate to the hypothesis that slopes differ from each other.

without subordinate with subordinate Before After 0 5 10 15 20 25

Time spent nest guarding (mean ± SE min hour

-1)

A

p = 0.02

−1.5 −0.5 0.0 0.5 1.0 1.5

Arthropod density (centered)

0 10 20 30

40 _{Before nest predator} After nest predator

B

Time spent nest guarding (mean ± SE min hour

-1)

−1.0

p = 0.02

FIGURE 6.3 Changes in time spent nest guarding by dominant male Seychelles warblers as a result of an experimental presen-tation with a nest predator in relation to (A) having an incubating subordinate present in the territory (N = 4) or not (N = 15), and (B) territory food availability (i.e. centered log arthropod density). P-values relate to the hypothesis that slopes differ from each other.

Incubation

In contrast to male nest guarding behavior, there was no difference in female incubation duration after presentation of a nest predator or non-predator control (-0.02 ± 0.06, χ2 ₌ 0.14, DF = 1, p = 0.71; Fig. 6.2), and dominant females did not increase their incubation

dura-tion after either the predator (0.02 ± 0.03, z = 0.58, DF = 1, p = 0.56; Fig. 6.2), or the non-preda-tor presentation (0.04 ± 0.05, z = 0.89, DF = 1, p = 0.37; Fig. 6.2). There was some evidence that dominant females showed less incubation behavior overall when there was an incubating subordinate present in the territory (-0.11 ± 0.06, χ2 _{= 2.96, DF = 1, p = 0.09), but incubation} responses to the model presentation were not dependent on whether there was an incu-bating subordinate present (interaction effect: 0.06 ± 0.10, χ2 _{= 0.51, DF = 1, p = 0.48). Female} incubation responses were slightly higher in territories with higher arthropod density, but this was not significant (0.06 ± 0.03, χ2 _{= 3.24, DF = 1, p = 0.07). There was a small negative} relationship between arthropod density and incubation duration before the model presen-tations (-0.11 ± 0.06, z = -2.00, DF = 1, p = 0.05), but there was no relationship between arthro-pod density and the duration of incubation after the model presentations (0.10 ± 0.06, z = 1.60, DF = 1, p = 0.11).

DISCUSSION

Being able to accurately assess and respond to nest predation risk is important when nest predation risk varies and anti-predator response are costly. Our results show that males in-crease the time spent guarding the nest – which is an effective way of reducing egg preda-tion in natural condipreda-tions (Komdeur & Kats 1999) – after we experimentally increased the perceived risk of nest predation (Fig. 6.2). In contrast to males, females showed no change in incubation duration after being presented with a nest predator. Male nest guarding re-sponses were lower in territories with high food availability – because males in such terri-tories already showed high levels of nest guarding – and with subordinates present (Fig. 6.3A, B), but female incubation duration did not depend on these factors. The number of alarm calls towards the nest predator and non-predator model did not differ significantly, but physical attacks towards the non-predator model were rare, in contrast to attacks to-wards the predator model. This suggests that our predator model was perceived as a bigger threat than our non-predator control model (Fig. 6.1A, B) as shown before (Veen et al. 2000). Breeders showed more attacks when they confronted the nest predator together than when either of them did so alone, which suggests benefits of joint defense for Seychelles warbler parents. We discuss our results further below.

Direct anti-predator responses: alarm calls and attacks

Individuals attacked more when presented with the nest predator model than the non-pred-ator control (Fig. 6.1). No strong effect was found for alarm calls, but alarm calls were more common and occurred at higher rates than attacks. This could indicate that attacks are more costly in terms of energy expenditure, or potential injury risk, although injuries have

never been recorded as a result of nest defense against Seychelles fodies (F.G., J.K., S.A.K. D.S.R., personal observations). In contrast to Veen et al. (2000), who found that males had higher attack rates than females, we found no such differences, but we did find that females alarm called more than males overall. Subordinates did not always participate in nest de-fense, neither by direct defense (alarms and attacks) towards a predator, nor by incubation or nest guarding. It is possible that subordinate nest defense strategies may be conditional (e.g. perhaps based on relatedness or body condition as observed for provisioning in this species; Richardson et al. 2003b; van de Crommenacker, Komdeur & Richardson 2011) as they are less consistent than that of the breeders. This is further illustrated by the fact that even when subordinates participated in nest predator defense, they always arrived after the dominant female or dominant male. However, when they did participate in defense, they alarm called as often as dominant males, and attack rates were similar to dominant males and females (Fig. 6.1B).

In species where more than one individual provides parental care, individuals might alter their anti-predator responses depending on the social context (Chase 1980; Clutton‐Brock 1991). Interestingly, Seychelles warblers attack a nest predator model more often when they are together than when they are alone (Fig. 6.1B). Similar patterns have been found in other species, e.g. great tits Parus major (Regelmann and Curio 1986). It is likely that (i) individ-uals are more likely to attack together because of the benefits of additional vigilance by others. This argument is supported by our own observations, where group members would take turns in attacking the nest predator, and one individual would remain at a distance and usually alarm call (FG, personal observations). Alternatively, (ii) individuals might be signaling a willingness to invest in the current brood in the hope that their partner will also increase investment (Johnstone & Hinde 2006; Johnstone et al. 2013). Dominant females alarm called more when they were alone than when they were together, but no such effect was present for dominant males, who also showed fewer alarm calls than females overall. That individuals alarm call more when they are alone is consistent with at least two func-tions that have been ascribed to alarm calls in other species: (i) when alarm calls function to signal to the predator that it has been seen, but attacking alone is too risky, and (ii) to signal the presence of a threat to other group members (Caro 2005). The additional func-tion of trying to attract other group members might thus explain the increased alarm rate of dominant females when they confronted the model alone. Interestingly, dominant males showed no differences in alarm rates depending on whether they confronted the model pre-sentation together or alone. Dominant males generally alarm called less than females and did not compensate for this by showing more attacks than dominant females. Our results therefore suggest that Seychelles warbler breeders (females) show more risk-averse an-ti-predator behaviors when they are alone, switching to more direct aggression when they

confront a nest predator when they have social support. Surprisingly, we found that attacks towards the nest predator model decreased when eggs had been incubated for more days, which is counter to the general hypothesis that nest defense should increase with increased reproductive value of the clutch or reduced nesting potential (Montgomerie & Weather-head 1988). However, results for this hypothesis have been mixed, with some species show-ing no change in nest defense behavior as the brood ages, and others showshow-ing decreased investment, similar to our results (reviewed in Caro 2005). One possible explanation for the decline of anti-predator responses with incubation stage, in our study and elsewhere, is a decrease in parental body condition as the brood ages due to high investment in incubation and nest guarding.

Indirect anti-predator responses: incubation and nest guarding

Dominant males increased their nest guarding behavior in response to our nest predator presentations and playback, but no such response was found in terms of incubation behav-ior in dominant females. The lack of response in females is likely the result of the higher overall female investment and the trade-off between incubation and time spent foraging (Reid et al. 2002; Tinbergen & Williams 2002). On average females incubate ca 50% of their time, while males do not incubate and only spent 17% of their time on nest guarding (only pre-experiment nest watches), leaving males with more opportunity to respond to the in-creased threat, compared to females. Additionally, females are often on strict incubation schedules to create the optimal conditions for proper embryonic development, which should further limit females’ ability to respond to increased nest predation risk (Deeming & Reynolds 2015). Although we only conducted our experiment in four territories with in-cubating subordinates, our results suggest that males can also benefit from the presence of incubating subordinates: males without incubating subordinates showed a much stronger response after the simulated nest predation threat, while such an effect was smaller and not significant for females (Fig. 6.3A). Load-lightening is a common benefit of subordinate help, and observed in many cooperative breeders (e.g. Hatchwell 1999), including the Sey-chelles warbler (Komdeur 1994a), Our results suggest that load lightening can be exacerbat-ed under increasexacerbat-ed nest prexacerbat-edation risk. Although we are currently unaware of any survival benefits due to load-lightening for Seychelles warbler males, it is possible that such effects only become apparent when the magnitude of nest predation risk is taken into account (Brouwer et al. 2006). Additionally, the reduced time investment by males could allow them to pursue extra-pair matings, as is the case in the superb fairy-wren Malurus cyaneus (Green et al. 1995); this is plausible in the Seychelles warbler, where extra-pair mating is common (Richardson et al. 2001).

time spent nest guarding in high quality territories was already high and did not change much as a result of our model presentations, while dominant males in low quality territo-ries showed a significant increase in nest guarding (Fig. 6.3B). This result is in line with a previous study in the Seychelles warbler that showed a similar correlation between male nest guarding investment and territory quality (Komdeur & Kats 1999). Our results thus in-dicate that male nest guarding behavior can be temporally increased when the risk of nest predation is high, but that the trade-off between nest guarding and other activities prohib-its dominant males from keeping up such high levels of close nest guarding over a longer period of time (Komdeur & Kats 1999). Interestingly, where the latter study found no rela-tionship between territory quality and female incubation, we found that dominant females tended to show a decrease in incubation duration with increasing territory quality, which was similar in strength to the increase in nest guarding behavior for dominant males. This suggests that at least part of female incubation behavior is compensatory and functions to reduce nest predation risk, but male removal experiments would be necessary to show this conclusively. The main difference between Komdeur and Kats (1999) and this study is that their measure of territory quality included territory size (i.e. is a measure of total arthropod abundance), while our study did not (i.e. measures arthropod density). The latter could be a better reflection of female foraging efficiency during incubation off bouts and, therefore, of the trade-off between incubation and territory quality.

Conclusion

Our results show differential responses to short-term increased nest predation risk between different group members in the cooperatively breeding Seychelles warbler. In our study, we have addressed both direct responses to predators and changes in breeding and vigilance behavior, before and after a simulated nest predation threat. This study differs from previ-ous investigations, that have looked primarily at alterations to incubation and feeding be-haviors, by investigating nest guarding, a form of vigilance that has evolved specifically to reduce the risk of nest predation (Komdeur & Kats 1999). We show that direct responses by dominant males and females to nest predators in the Seychelles warbler differ depending on whether the threat is confronted together or not. Furthermore, only dominant males respond to simulated nest predation risk by increasing vigilance, while females show no such response in time spent incubating. We highlight the fact that male vigilance is likely much more flexible perhaps because, (i) it is unconstrained by thermal requirements to the egg(s) and (ii) the initial investment is much lower, leaving more opportunity for males to respond to increased risk. This is further illustrated by the finding that male nest guarding behavior is conditional on arthropod density and on subordinate help, suggesting that this behavior is costly, and that these costs can be alleviated under favorable food and social conditions. Together, these results show that anti-predator behavior can differ substantially

according to individual, social, and environmental conditions. Acknowledgements

We thank Nature Seychelles, the Seychelles Bureau of Standards and the Department of En-vironment for making our fieldwork on Cousin Island possible, and we particularly thank Megan Pendred for help during fieldwork. We thank Alex Jansen, Moisès Sánchez-Fortún and Julia Schroeder for providing the taxidermic mounts. This research was supported by Netherlands Organisation for Scientific Research (NWO) TOP grant (854.11.003) and ALW grant (823.01.014) in the name of JK. SAK was supported by an NWO Veni fellowship (863.13.017) and KB by a Natural Environment Research Council (NERC) Ph.D. studentship. DSR was supported by a NERC grant (NE/K005502/1).