Consequences of ‘load-lightening’ for future indirect fitness gains by helpers in a cooperatively breeding bird

Authors


Correspondence author. E-mail: j.meade@sheffield.ac.uk

Summary

1. Helpers that invest energy in provisioning the offspring of related individuals stand to gain indirect fitness benefits from doing so. First, if the helper’s effort is additional to that of the parents (additive) the productivity of the current breeding attempt can be increased. Secondly, if the parents reduce their workload (compensation) this can result in future indirect fitness gains to the helper via increased breeder survival; termed ‘load-lightening’.

2. Long-tailed tits (Aegithalos caudatus) have a cooperative breeding system in which helpers assist kin and parents exhibit both additive and compensatory reactions in the presence of helpers. Offspring from helped nests are heavier and more likely to recruit into the breeding population, thus helpers gain indirect fitness benefits from increasing the productivity of the current breeding attempt. Despite breeders’ reduction of feeding effort in the presence of helpers, previous investigations found no subsequent increase in breeder survival.

3. The aim of this study was to test the hypothesis that load-lightening resulted in indirect fitness benefits for helpers. We used data from a 14-year study to investigate the provisioning rate, survival and future fecundity of male and female long-tailed tits that did and did not receive help at the nest.

4. We found an asymmetrical response to the presence of helpers at large brood sizes. Males reduced their feeding rate more than females, and this differential response was reflected in a significant increase in male survival when provisioning large broods assisted by helpers. We found no evidence of any increase in future fecundity for helped breeders.

5. The finding that males reduce their provisioning rate in the presence of helpers (at large brood sizes) to a greater degree than females, and that this is reflected in an increase in survival rate for males only, implies that the survival increase is caused by the reduction in work-rate rather than a non-specific benefit of a larger group size.

6. The marginal benefits of help for breeder survival are likely to be more difficult to identify than the increased productivity at helped nests, but should not be overlooked when investigating the potential indirect fitness gains that supernumeraries can accrue by helping.

Introduction

In cooperatively breeding birds, some individuals, usually referred to as helpers, defer their own breeding attempts and help to raise offspring that are not their own. In some species these helpers gain direct benefits from helping, such as access to paternity or group augmentation (for reviews, see Cockburn 1998; Dickinson & Hatchwell 2004). If helping behaviour is directed towards kin, as is the case in most cooperative systems (Brown 1987; Emlen 1997; Hatchwell 2009), helpers can also gain indirect fitness benefits (Hamilton 1964) either by increasing the productivity of the current breeding attempt, or by reducing the reproductive costs of the breeders thus increasing their chances of survival or future fecundity, referred to as ‘load-lightening’ (Brown 1978; Crick 1992; Heinsohn 2004).

The cooperative behaviour of helpers can take diverse forms, but the most common and easily quantified kind of helping is nestling provisioning. The rate at which helpers provision nestlings varies both within and across species, so the overall contribution of helpers to the food demands of a brood is extremely variable (Heinsohn 2004). Furthermore, the response of breeders to the presence of helpers at their nest determines whether the helpers’ work results in the production of more or higher quality offspring in the current breeding attempt, or whether it allows breeders to use the opportunity created by these additional resources to reduce their own work-rate and enhance their future survival or fecundity (Hatchwell 1999). At one extreme, breeders can continue to feed at the same rate regardless of the number of helpers so that the helpers’ care is additional to parental care and the brood’s total provisioning rate increases, e.g. white-fronted bee eaters Merops bullockoides (Emlen & Wrege 1991). The recruitment of fledglings into the breeding population is positively related to nestling mass in many bird species (e.g. Garnett 1981; Magrath 1991) and the increased provisioning rate because of the presence of helpers can increase the number of offspring recruited (e.g. Brown et al. 1982; Komdeur 1994; Hatchwell et al. 2004). Thus, breeders benefit from increased direct fitness, and, provided that helpers assist kin, they can gain indirect fitness through increased productivity of the current brood.

However, parental care has also been shown to be costly in many species, reducing body condition, the likelihood of survival and future fecundity (Dijkstra et al. 1990; Clutton-Brock 1991; Ketterson & Nolan 1994). Therefore, at the other extreme of potential reactions to the care provided by helpers, breeders may reduce their work-rate to completely compensate for the presence of helpers, leaving the overall provisioning rate unchanged, as found in grey-crowned babblers Pomatostomus temporalis (Brown et al. 1978). This load-lightening effect could be converted into direct fitness benefits for breeders through enhanced survival or future fecundity, and future indirect fitness benefits for helpers, again assuming that they are helping relatives (Brown 1978; Crick 1992).

Between these extremes, breeders in many cooperative species show a range of responses both within and between species, reducing their work-rate in the presence of helpers but not compensating fully for their presence, thereby reducing their own feeding rate but still increasing the total feeding rate. In a comparative analysis of the response of breeders to the presence of helpers, Hatchwell (1999) found that additive care was associated with a high rate of nestling starvation, while compensation (a reduction in work-rate by breeders) occurred when nestling starvation was relatively rare.

Several studies have demonstrated that load-lightening can reduce breeders’ inter-clutch interval within years, and thus increase the fecundity of a breeding pair by facilitating a greater number of breeding attempts per year (Rowley 1965; Brown & Brown 1981; Woxvold & Magrath 2005; Blackmore & Heinsohn 2007).

The presence of helpers has also been shown to have a beneficial effect on breeder survival (e.g. Reyer 1984; Woolfenden & Fitzpatrick 1984; Russell & Rowley 1988; Rabenold 1990), although others (e.g. Ligon & Ligon 1978; Rabenold & Christensen 1979; Eden 1987; Koford, Bowen & Vehrencamp 1990) have found no effect of helper presence on survival. However, these were correlative studies, and as Koenig & Mumme (1987) and Cockburn (1998) note, it is difficult to disentangle any helper effect from potentially confounding effects of territory quality, group size and individual quality on survival. A study of acorn woodpeckers Melanerpes formicivorus by Koenig & Mumme (1987) did take into account territory quality, breeder experience and group composition and found that male but not female survival increased with an increasing number of helpers. By contrast, in superb fairy-wrens Malurus cyaneus, Cockburn et al. (2008) found that female but not male survival was increased when supernumeraries were present, concluding that this effect was likely to be causative because the increase in survival was observed in the dispersive rather than the philopatric sex. Cockburn (1998) also noted that the long life-spans of some cooperative breeders can complicate survival predictions based on load-lightening.

The cooperative breeding system of the long-tailed tit Aegithalos caudatus (L.) differs from many other cooperative breeders in several ways that make the investigation of reproductive behaviour less sensitive to confounding variables. Birds do not defend exclusive territories so habitat or territory quality should not bias the data in a systematic fashion. In addition, there is no evidence that provisioning or survival rates vary in relation to habitat within our study site (MacColl & Hatchwell 2003). Helpers are not offspring that have delayed dispersal; instead, each year all birds attempt to breed in pairs initially, but a seasonal decline in reproductive success acts as a constraint on independent reproduction so that some birds whose own nests have been depredated become helpers and redirect their care towards relatives (MacColl & Hatchwell 2002; Hatchwell et al. 2004). The stochastic nature of predation (Hatchwell et al. 1999a) means that there is unlikely to be any correlation between the quality of the area that a nest occupies and the likelihood of a breeding attempt succeeding or the breeders at the nest receiving help (Hatchwell et al. 2004). Helpers are most likely to be a sibling of one of the breeding pair (Russell & Hatchwell 2001) so there is no strong link between reproductive success in one year and the likelihood of being helped in the next. Long-tailed tits over-winter in flocks of 6–20 birds and for adults there is no evidence that flock size and composition is influenced by breeding status in the preceding season (McGowan et al. 2007). Long-tailed tits also have relatively short life-spans compared with many cooperative breeding species and fecundity, survival and helping are unrelated to age (McGowan, Woodburn & Hatchwell 2003; Hatchwell et al. 2004; Hatchwell & Sharp 2006).

Long-tailed tit helpers gain indirect fitness benefits by increasing productivity in the current breeding attempt. The presence of helpers does not increase the number of offspring fledged, but fledglings from helped nests are significantly heavier than nestlings from broods without helpers, and are significantly more likely to recruit into the breeding population (Hatchwell et al. 2004). However, the possibility that helpers may gain delayed or ‘future’ indirect fitness benefits by increasing the survival or fecundity of helped breeders remains unresolved. It has been shown experimentally that long-tailed tit parents significantly increase their provisioning rate if a helper is removed (Hatchwell & Russell 1996), and observational data showed that parental provisioning decreases significantly (by c. 20%) in the presence of a single helper, and after this the care of additional helpers is additive (MacColl & Hatchwell 2003). Thus, the presence of helpers at the nests of long-tailed tits allows breeders to decrease their workload and this load-lightening response of parents would be predicted to result in increased residual fitness. However, a previous study investigating the relative survival rates of failed breeders, helpers and successful breeders with, and without helpers did not detect a survival difference between successful breeders that received help and those that did not (McGowan et al. 2003). However, the nature of the analysis neither allowed individual covariates to be examined, nor were abiotic factors considered. Given the strong effect of provisioning rate on offspring fitness it seems paradoxical that long-tailed tits would choose to reduce their work-rate in the presence of helpers if it did not confer some sort of future reproductive advantage whether through increased survival or increased fecundity.

In this study, first we re-examine the relationship between the presence of helpers-at-the-nest and provisioning rate. Then, we test the hypothesis that helpers lighten the reproductive load; for successful long-tailed tit breeders with and without helpers in year n, we investigate survival to year + 1, timing of reproduction in year + 1, and clutch size in year + 1, controlling for individual covariates and environmental conditions.

Materials and methods

Species and study site

For this investigation we used data collected between 1995 and 2009 from a study population of 25–72 pairs of long-tailed tits in the Rivelin Valley, Sheffield, UK (53°23′ N, 1°34′ W). All long-tailed tit pairs within the study site are monitored and their nests located by observation. Any unringed individuals are caught once nest sites have been located. Individuals are then uniquely colour-ringed, weighed and blood samples are taken from the brachial vein (under UK Home Office Licence). Nests are monitored approximately every second day. If nest failure occurs the study site is searched intensively for subsequent nesting attempts. Timing of laying, clutch size, hatch date, brood size and fledge date are recorded, and accessible nests are observed every second day, typically for 1 h per observation period, from hatching (day 0) to fledging (days 16–17) so the identity of all carers and their provisioning rates can be recorded (for more details on provisioning observations, see MacColl & Hatchwell 2003). Nestlings in accessible nests are weighed, colour-ringed and blood samples are taken on day 11. Blood samples are subsequently used to sex adults and nestlings.

Breeders were classed as successful in a given year if their brood fledged (long-tailed tits are always single-brooded). A helper was any bird observed feeding the nestlings of another pair. Because males are the philopatric sex, birds were classified as having raised recruits only if any of their male offspring were found to be attempting to breed within the study site in subsequent years. Birds ringed as nestlings could be accurately aged using the database, and unringed immigrants were assumed to be 1-year olds at the time of ringing. For more details about the study site and species, see Hatchwell et al. (2004) and Hatchwell & Sharp (2006). All statistical analyses were conducted in the r environment, ver. 2.8.1.

Statistical analysis

Provisioning rate

We first investigated the provisioning rate of long-tailed tit parents in relation to the presence or absence of helpers-at-the-nest. This was performed using a linear mixed model with a normal error structure and an identity link function fitted using the function lmer in the r package lme4 (Bates, Maechler & Dai 2008). Bird identity and nest identity (subsequently referred to as ‘nest’) were included as random factors to control for non-independence of repeated observations of feeding rates by the same bird, and repeated observations of feeding rates at the same nest. We used the square-root of the provisioning rate (no. feeds h−1) as our response variable (this transformation facilitated normal and heteroscedastic errors). The fixed effects were: year, date, nestling age, brood size, sex and whether helpers provisioned the nest. ‘Year’ was included as a categorical variable to control for non-specific factors that may affect food availability between years. ‘Date’ (the number of days after 1 March) was included to control for factors affecting food availability within years. ‘Nestling age’ was measured in days, taking hatching day as day 0 (long-tailed tit broods hatch synchronously). Previous research (MacColl & Hatchwell 2003) suggests that the relationship between provisioning rate and nestling age is curvilinear (quadratic) so the quadratic term for this explanatory variable was also added to the model. ‘Brood size’ was the number of chicks present in the nest on day 11. This is a reliable indicator of brood size because nestling starvation in long-tailed tits is rare (Hatchwell 1999). ‘Sex’ was the gender of the breeder. In our data set, the total number of helpers at a nest ranged from 1 to 4, but most had just one (53%) or two (31%) helpers. Previous analyses found that parents significantly reduce their provisioning rate in the presence of one helper with subsequent help being additive (MacColl & Hatchwell 2003), so we used a binary variable set to 1 if a bird other than the parents fed the brood during each period of observation, otherwise set to 0 – this explanatory variable will subsequently be referred to as ‘helped?’.

It is important to identify potential sex differences in a ‘load-lightening’ response because if load-lightening is restricted to one gender and subsequent improvement in survival or future fecundity is restricted to the same sex, there is better evidence for direct causation within a correlative analysis (Cockburn 1998). We therefore tested the hypothesis that the provisioning rates of males and females differed and that the response to the presence of helpers also depended on brood size via a three-way interaction between ‘sex’, ‘brood size’ and ‘helped?’. Previous research has shown that birds with larger broods work harder (MacColl & Hatchwell 2003), therefore the presence of supernumerary birds may be more beneficial when brood sizes are large.

We also tested specifically the hypothesis that the two sexes adjust their provisioning rate differentially during the breeding season by specifying a sex × date interaction. The overall feeding rate at nests falls during the breeding season (MacColl & Hatchwell 2003), and late fledging long-tailed tit broods are less valuable because of lower overwinter survival (MacColl & Hatchwell 2002). However, in a previous study based on only 8 years of data, ‘date’ did not significantly affect provisioning rate (MacColl & Hatchwell 2003). Here, we use data on provisioning rates from 2043 observation periods of 267 individual birds at 182 nests over 14 years.

Survival

We modelled the probability that a bird that bred successfully in year n would survive to year + 1. We fitted generalized linear models with a binomial error and a logit link function with a binary response variable of ‘survived?’ which took a value of 1 if the bird was seen as part of the breeding population in subsequent years and a value of 0 otherwise. From 1995 to 2009 full information was available for 256 colour-ringed individuals; 128 males and 128 females. We fitted separate models for male and female breeders. The explanatory variables were: age, brood size, fledge date, winter rainfall, whether the birds produced recruits, and whether or not a helper was present at the nest. ‘Age’ was the age in years of the breeder in year n. Age is known precisely for birds ringed as chicks, immigrant birds are assumed to be 1-year olds at the time of ringing. ‘Fledge date’ was the date of fledging of the successful breeding attempt in year n relative to the overall median fledge date from 1995 to 2008 (22 May); a negative number indicates an early fledge date. ‘Winter rainfall’ was the mean daily rainfall in mm, from 1 October in year n to 1 March in year + 1, as measured by the Weston Park Meteorological Station, 5 km from the centre of the Rivelin Valley study site. We used winter rainfall as an environmental predictor of survival because preliminary analyses suggested that rainfall rather than temperature is the principal correlate of survival for our study site (J. Meade, unpublished). Whether or not the birds produced recruits was a binary variable set to 1 if a male fledgling from the successful breeding attempt in year n was seen in the breeding population in subsequent years, otherwise set to 0. We included male offspring only because males are the philopatric sex so the probability of seeing female offspring as recruits is low. We subsequently refer to this variable as ‘recruits?’ We used a binary variable set to 1 if a bird other than the parents fed the brood in year n, otherwise set to 0 – referred to subsequently as ‘helped?’.

We also tested two hypotheses concerning the interaction of variables. First, research on white-browed scrubwrens Sericornis frontalis by Magrath (2001) showed that young females accrue greater benefits from receiving help than older females. Therefore, we examined whether younger individuals benefit more from help than older individuals by specifying an age × helped interaction. Secondly, provisioning larger brood sizes can have a negative effect on survival (see Dijkstra et al. 1990), thus it is likely that help would be more beneficial to birds provisioning larger broods. Therefore, we specified a helped × brood size interaction to determine whether the effect of receiving help varies with brood size.

Timing of breeding

We modelled the timing of clutch initiation in year + 1, for birds that bred successfully in year n and survived to year + 1 using a linear model. The response variable was the lay date (in days from 1 March) of the first egg in year + 1. Only first nesting attempts were included. Full information was available for 75 individuals (39 males and 36 females). We fitted models separately for male and female breeders. The explanatory variables were: helped?, age, brood size, recruits?, fledge date, winter rainfall and the mean minimum daily temperature in degrees centigrade from 15 February to 15 March in year + 1 [shown to have a significant effect on lay date by MacColl & Hatchwell (2002)], recorded at Weston Park Meteorological Station. We did not include any interactions in this analysis because of the small sample size.

Clutch size

We modelled the clutch size in year + 1, for females that bred successfully in year n and survived to year + 1 using a generalized linear model with Poisson error and a logarithm link function, where the response variable was clutch size in year + 1. Only the first clutches were included in our analysis. Full information was available for the successful breeding attempts of 38 females. The explanatory variables were: helped?, age, brood size, recruits?, fledge date, spring temperature and winter rainfall. We did not include any interactions in this analysis because of the small sample size.

Results

Provisioning rate

First, we found that the effect of sex on provisioning rate depended on brood size and that this dependence was mediated by the presence or absence of helpers (χ= 15·4, d.f. = 1, < 0·001). Fig. 1 illustrates, for each sex, how provisioning rate varies with brood size and the presence or absence of helpers. Both sexes feed at a higher rate when brood sizes are larger, and both sexes reduce their provisioning rate in the presence of helpers. This reduction varies differentially with brood size between the two sexes. For males there is a greater reduction in work-rate in the presence of helpers at larger brood sizes. For females the greatest reduction in provisioning rate in the presence of helpers occurs when brood sizes are small.

Figure 1.

 The relationship between feeding rate and brood size for parents with and without helpers-at-the-nest. Panels a and c show the predicted values from the model (χ2sex × brood size × helped = 15·4, d.f. = 1, < 0·001) for females (a) and males (c). Dotted lines (inline image) indicate helped birds and solid lines (inline image) indicate un-helped birds. All other explanatory factors are set to their medians. The vertical dashed lines indicate the median brood size. Panels b and d show the mean of the square-root of the feeding rate ±SE at each brood size for females (b) and males (d). inline image = helped birds and inline image = un-helped birds.

We also found that the differences in provisioning between males and females depended on date (estimate = 0·02, SE = 0·003, χ= 27·8, d.f. = 1, < 0·001). Figure 2 shows that male breeders reduced their provisioning rate as the breeding season progressed, whereas provisioning by female breeders was largely unaffected by date. The model also showed a significant curvilinear relationship between provisioning rate and nestling age, the visit rate tending to level off as chicks became older (estimate = −0·01, SE = 0·0007, χ2 = 46·3, d.f. = 1, < 0·001). Finally, provisioning rates varied by year (χ= 28·0, d.f. = 13, = 0·009) with birds feeding chicks at a significantly lower rate in 2000, 2001, 2004, 2006 and 2007.

Figure 2.

 The relationship between provisioning rate and date (in days after 1 March) for helped males (inline image) and helped females (inline image). Lines are the predicted values from the model (χ2sex × date = 27·8, d.f. = 1, < 0·001). All other explanatory variables are set to their medians.

Survival

We modelled the survival of successful breeders using generalized linear models to determine whether the presence of helpers-at-the-nest affected survival rates. For female breeders, survival was significantly negatively affected by winter rainfall (estimate = −1·07, SE = 0·34, χ= 10·8, d.f. = 1, < 0·001), but by no other factors, although increasing brood size had a marginally significant negative effect on survival (estimate = −0·182, SE = 0·14, χ= 3·5, d.f. = 1, = 0·062). The survival of male breeders was strongly affected by fledge date: males were significantly more likely to survive to year + 1 if they fledged broods later in year n (estimate = 0·08, SE = 0·03, χ= 11·7, d.f. = 1, <0·001). There was also a marginally significant negative effect of both winter rainfall (estimate = −0·64, SE = 0·36, χ2 = 3·4, d.f. = 1, = 0·067), and whether males successfully recruited male offspring from year n into the breeding population in year + 1 (estimate = −0·76, SE = 0·43, χ2 = 3·3, d.f. = 1, = 0·070).

We tested two hypotheses regarding interactions. First, we found no evidence for the hypothesized effect of breeder age on the consequences of help for survival, for either sex (females: estimate = −0·33, SE = 0·50, χ= 0·4, d.f. = 1, = 0·508; males: estimate = 0·32, SE = 0·40, χ= 0·5, d.f. = 1, = 0·420). However, in the second analysis, we found that the effect of the presence of helpers on male survival varied significantly as brood size changed (estimate = 0·41, SE = 0·18, χ= 5·5, d.f. = 1, = 0·019). Brood size had a negative effect on the survival of male breeders without helpers, but this relationship was reversed for males receiving help. This effect was not found for female breeders (estimate = 0·08, SE = 0·18, χ= 0·2, d.f. = 1, = 0·650). To illustrate these relationships, we controlled the other explanatory variables (age, brood size, fledge date, winter rainfall were set to their medians, recruits was set to ‘no’) and plotted survival in relation to brood size for birds with and without helpers, for females and males separately (Fig. 3).

Figure 3.

 The relationship between brood size and survival for helped (inline image) and un-helped (inline image) birds, for females (a) and males (b). The lines show predicted values from the models (females: = ns; males: = 0·019) with other explanatory variables controlled (age, brood size, fledge date, winter rainfall are set to their medians, recruits? is set to ‘no’). Symbols show the raw data ○ = helped, • = un-helped.

Timing of breeding

Spring temperature was the only variable that affected the timing of laying for female (< 0·001) or male breeders in year + 1 (= 0·002; Table 1), with clutches initiated earlier when the month preceding laying was warmer. This is the period during which nest-building usually starts (McGowan, Sharp & Hatchwell 2004) and this result replicates that of MacColl & Hatchwell (2002). There was no significant difference in the timing of laying of successful breeders that had and had not received help in year n (Tables 1 and 2).

Table 1.   Model estimates from linear model analyses (normal error, identity link) of the date of initiation of the first clutch in year + 1
Model termFemalesMales
d.f.EstimateSEScaled deviancePd.f.EstimateSEScaled devianceP
  1. The timing of clutch initiation in year + 1 for birds that bred successfully in year n and survived to year + 1 was modelled separately for male and female breeders. Scaled deviance was obtained by fitting terms last. Significant P values are shown in bold.

Age10·6311·7560·170·68310·5681·2950·240·624
Brood size10·0540·5740·010·91610·2020·7060·100·749
Helped?1−0·4562·4910·040·8361−4·4803·0062·700·100
Fledge date10·0790·1260·500·47810·2510·1902·140·144
Recruits?1−2·8602·0972·310·1281−4·2972·7712·910·088
Spring temperature1−3·9340·87219·65<0·0011−2·7080·9389·290·002
Winter rainfall10·6152·1980·100·75111·2512·6660·280·599
Table 2.   The timing of laying (both sexes) and clutch size (females only) in year + 1 for long-tailed tits that did or did not receive help in year n
 Lay date in year + 1Clutch size in year + 1
 FemalesMalesFemales
 Mean ± SE(n)Mean ± SE(n)Mean ± SE(n)
Helped41·37 ± 1·60(19)39·18 ± 1·88(22)9·95 ± 0·13(21)
Un-helped39·41 ± 2·05(17)41·41 ± 2·27(17)9·76 ± 0·18(17)

Clutch size

There were no effects of any of the factors included in the linear model investigating clutch size in year + 1. Successful female breeders that received help in year n did not produce a larger clutch size in year + 1 than successful female breeders that had not received help (Tables 2 and 3). In general, there was very little variation in clutch sizes in our data set; all but one of the 38 first clutches in the analysis contained 9–11 eggs.

Table 3.   Model estimates from a linear model analysis (Poisson error, logarithm link) of clutch size in year + 1 for females that bred successfully in year n
Model termEstimateSEd.f.χ2-ValueP-value
  1. Chi-square values were obtained by fitting terms last.

Age0·00010·08410·0000·999
Brood size−0·00020·02410·0000·993
Helped?0·02240·11010·0420·838
Fledge date0·00030·00710·0030·960
Recruits?−0·02020·11110·0330·855
Lay date (year + 1)−0·00110·00810·0210·885
Spring temperature−0·01280·03810·1140·735
Winter rainfall0·03300·10910·0920·762

Discussion

In this paper, our objective was to investigate whether helpers-at-the-nest gain future indirect fitness benefits by increasing the survival or future fecundity of the breeders they help. We have shown that although both male and female breeders reduce their provisioning rate in the presence of helpers-at-the-nest, this reduction in work-rate is affected by the size of the brood that the parents are provisioning. Males reduce their provisioning rate to the greatest degree when feeding large broods, whereas the greatest reduction in female provisioning rate in the presence of helpers is at small brood sizes (Fig. 1). As provisioning rate increases with an increasing brood size (MacColl & Hatchwell 2003), males make their largest helper-mediated reduction in work-rate when their provisioning rate is high, whereas females take advantage of the presence of helpers at small brood sizes when their work-rate is already low. This differential reduction in work-rate in the presence of helpers is reflected in an increased likelihood of survival for males with larger brood sizes while there is no such increase in the survival probability of female breeders. It should be noted that although the increase in male survival is evident only for brood sizes of eight or more, broods of this size are typical in our study population, the median for the data reported here is eight, and in our entire data set of known brood sizes (n = 220) the median brood size is nine, with 71% of broods containing eight or more nestlings. Thus, we have shown that as well as gaining substantial indirect fitness benefits by increasing the productivity of related breeders’ current breeding attempts (Hatchwell et al. 2004; MacColl & Hatchwell 2004), long-tailed tit helpers may also gain future indirect fitness benefits by enhancing the survival of closely related male breeders.

We found links between provisioning rate and survival probability throughout our analyses. Provisioning rates increased with increasing brood size, and there was a corresponding decrease in survival with increasing brood size for females and un-helped males. Similarly, we found that, unlike females, males decreased their feeding rates as the breeding season progressed, and that male survival to year + 1 was significantly correlated with the date of fledging in year n, whereas female survival was not. Males fledging broods later in the season were more likely to survive than males fledging early broods, presumably because they put less effort into provisioning late broods.

Longevity has been shown to be a significant component of lifetime fitness estimates in birds and mammals (Clutton-Brock 1988) and even small increases in yearly survival can strongly affect individual fitness (Woolfenden & Fitzpatrick 1984; Mumme, Koenig & Ratnieks 1989). Small increases in annual survival are much harder to detect than a helper effect on productivity and also harder to dismiss. This is especially true of long-lived species, as many cooperative species are (Arnold & Owens 1998), where the marginal effect of helpers on longevity may be small but still have significant consequences (Emlen 1991). Across our data set, the median clutch size is nine and the mode is 10. At these brood sizes a male breeder’s probability of survival increases by 9% and 19%, respectively, when a helper is present. Individual fitness of both sexes of long-tailed tit is significantly positively correlated to lifespan (MacColl & Hatchwell 2004); therefore, this increase in survival probability is likely to have a substantial effect on an individual’s lifetime reproductive success. It is also worth noting that the individual fitness of females is positively correlated with their mean provisioning rate, but there is no such relationship for males (MacColl & Hatchwell 2004). It may be that the larger reduction in provisioning rate made by males in the presence of helpers (at large brood sizes) causes any positive relationship between individual fitness and provisioning rate to break down.

Unlike many other species in which the effects of load-lightening have been investigated (e.g. Reyer 1984; Woolfenden & Fitzpatrick 1984; Russell & Rowley 1988; Rabenold 1990), we think that the long-tailed tit system is very unlikely to be confounded by group-size or habitat effects. During the breeding season, long-tailed tits breed in pairs without exclusive territories, and they over-winter in flocks composed of more than one family, plus immigrants, with extensive overlapping home ranges (Hatchwell et al. 2001). The formation of these flocks is not fully understood, but membership is not influenced by status in the preceding breeding season (McGowan et al. 2007). Helpers are adults that have failed in their own breeding attempt and then typically redirect their care towards siblings, whose nest is often some distance away (Russell & Hatchwell 2001). Therefore, the presence of helpers is unpredictable because of the stochastic nature of nest predation and because it is not closely related to the outcome of the previous breeding season. Divorce is also common, occurring in c. 63% of surviving pairs (Hatchwell et al. 2000). For these reasons, group size and composition between years is extremely changeable and therefore unlikely to have any confounding effect on our results. Similarly, particular individuals or families are not closely associated with exclusive habitats of variable quality. In addition, the sexually asymmetrical response to the presence of helpers (at large brood sizes) strengthens the argument that the increase in male survival is caused by load-lightening in the presence of helpers rather than some non-specific group-size effect (Cockburn 1998).

Our finding that load-lightening and subsequent survival is male-biased is intriguing. Asymmetrical parental responses to the presence of helpers-at-the-nest has been found in several other species (for review, see Hatchwell 1999); for example, male Seychelles warblers Acrocephalus sechellensis reduce their feeding rate in response to an increased number of helpers-at-the-nest whilst female breeders continue to feed at the same rate (Komdeur 1994). In long-tailed tits, helpers are usually males (c. 85%; Hatchwell et al. 2004) who are related to the male breeder (c. 75%; K.-B. Nam and B.J. Hatchwell, unpublished) and as helpers stand to gain deferred indirect fitness benefits only by increasing the survival of breeders that they are closely related to, helpers would rarely gain a fitness benefit by increasing female survival. Therefore, a male bias in the effect of load-lightening is best for helpers. It is difficult to imagine what mechanism allows this asymmetry in survival benefit to occur, although there are several possibilities.

First, we have no evidence that either breeders or helpers are able to enforce provisioning by other nest attendants, as is thought to occur in superb fairy-wrens (Mulder & Langmore 1993). Likewise, we have no evidence of false-feeding, as might be expected to occur under an enforcement regime (Boland, Heinsohn & Cockburn 1997). The most obvious sanction in species with stable groups may be aggression or eviction, but in the fluid social system of long-tailed tits, eviction from a ‘group’ is likely to incur little cost. Nevertheless, it would be interesting to explore the reaction of male and female breeders to the extra care provided by helpers to examine whether parental response does differ according to whether a helper is related to the male or the female. Unfortunately, our sample size of helpers related only to females is too small to permit this analysis.

Secondly, sex differences in the response of breeders to helper care could be related to patterns of parental investment (Cockburn et al. 2008). Nest-building is by both parents, but females invest more than their own body weight in producing a clutch of eggs, and take sole responsibility for incubation (Hatchwell et al. 1999b) and for brooding chicks up to day 6. In contrast the average provisioning rate of males during the nestling period is greater than that of females (MacColl & Hatchwell 2003). However, although these sex differences suggest that the costs of provisioning may differ between sexes because of their prior investment, they still do not explain why females with large broods don’t reduce their effort when helped.

The final possibility is that responses are driven by sexual conflict over investment within the breeding pair (Houston & Davies 1985), females taking advantage of helper provisioning to reduce her effort at small brood sizes, and males doing so when broods are large. The behavioural dynamics of sexual conflict over investment are currently being explored and may provide insights into this asymmetrical response to helpers.

Of the other factors affecting survival, winter rainfall had a significant negative effect on survival of females and was close to significance for males. This negative effect of rainfall on survival has been reported for other small European passerines (e.g. Grosbois et al. 2006) and is likely to be as a result of increased energetic costs associated with being wet, and with lower and more unpredictable invertebrate food availability in wetter winters (Speight & Wainhouse 1989).

We found no indication that long-tailed tits of different ages respond differently to the presence of helpers. Moreover, in previous studies, we have found no effect of age on productivity with or without helpers (Hatchwell et al. 2004; MacColl & Hatchwell 2004). This contrasts with a study of scrubwrens, where yearling females had low productivity and benefited significantly from the presence of helpers, while older females did not; an effect that is fairly widespread among cooperative species (Magrath 2001). Long-tailed tits are very short-lived relative to most other cooperative breeders, including scrubwrens. Of 285 female long-tailed tits of known life-history, only 20% lived longer than 2 years. Therefore, the absence of age-related productivity or helping effects is not surprising.

In single-brooded species, early breeding is often advantageous (Crick, Gibbons & Magrath 1993), and in long-tailed tits brood size and offspring survival probability decline through the season (MacColl & Hatchwell 2002), so a reduction in reproductive costs was predicted to increase measures of breeder fecundity. However, we found no evidence that receiving help in year n allowed breeders to attempt to breed earlier in year + 1. Similarly, we found no evidence that clutch size in year + 1 was affected by the presence of helpers in year n, there was little variation in the sizes of first clutches with all but one varying from 9 to 11 (the anomaly being eight). Both timing of breeding and clutch size are likely to be strongly affected by several biotic and abiotic factors, including the condition of an individual’s partner, the time of pairing, as well as microclimatic effects of temperature. Therefore, any variation in a breeder’s condition as a result of reduced reproductive costs in the previous year might be masked by other factors, especially given the relatively small sample sizes involved in this analysis. Any effect of load-lightening on future fecundity is likely to be greater (or easier to detect) in species which have multiple broods in a single season, as in the grey-crowned babbler (Blackmore & Heinsohn 2007).

In conclusion, although empirical data suggests that future indirect fitness benefits are potentially important in cooperative breeding systems, analyses may be confounded by breeder and territory quality (Cockburn 1998). Several recent studies show clearly that load-lightening occurs when group size is controlled for (Woxvold & Magrath 2005; Blackmore & Heinsohn 2007; Russell et al. 2008), but these studies did not investigate whether load-lightening had an effect on breeder survival. Effects of long-tailed tit helpers on productivity are measurable after just a few years of study (Glen & Perrins 1988; Hatchwell et al. 2004), but the effect of helpers on breeder survival that we report here was more elusive (McGowan et al. 2003), even though behavioural responses to the care of helpers make sense only if such effects do occur (Hatchwell & Russell 1996; Hatchwell 1999). Therefore, given the difficulty of measuring marginal differences in adult survival, the potential for future indirect gains from helping kin may be far more widespread than is currently recognized.

Acknowledgements

We thank Andy Bamford, Martin Fowlie, Nicky Green, Jin-Won Lee, Andrew MacColl, Andy McGowan, Dan Richardson, Douglas Ross, Andy Russell, Stuart Sharp and Michelle Simeoni for invaluable assistance with data collection, and Sheffield City Council, Yorkshire Water and the Hallamshire Golf Club for permission to work on their land. We thank D. Gillespie for helpful statistical advice and three anonymous reviewers for their helpful comments. This work was funded by grants from the Natural Environment Research Council, Nuffield Foundation, Association for the Study of Animal Behaviour, and the University of Sheffield, to whom we are most grateful.

Ancillary