The evolution of cooperation is a persistent problem for evolutionary biologists. In particular, understanding of the factors that promote the expression of helping behaviour in cooperatively breeding species remains weak, presumably because of the diverse nature of ecological and demographic drivers that promote sociality.
In this study, we use data from a long-term study of a facultative cooperative breeder, the long-tailed tit Aegithalos caudatus, to investigate the factors influencing annual variation in helping behaviour. Long-tailed tits exhibit redirected helping in which failed breeders may become helpers, usually at a relative's nest; thus, helping is hypothesised to be associated with causes of nest failure and opportunities to renest or help.
We tested predictions regarding the relationship between annual measures of cooperative behaviour and four explanatory variables: nest predation rate, length of the breeding season, population-level relatedness and population density.
We found that the degree of helping was determined principally by two factors that constrain successful independent reproduction. First, as predicted, cooperative behaviour peaked at intermediate levels of nest predation, when there are both failed breeders (i.e. potential helpers) and active nests (i.e. potential recipients) available. Second, there were more helpers in shorter breeding seasons when opportunities for renesting by failed breeders are more limited.
These are novel drivers of helping behaviour in avian cooperative breeding systems, and this study illustrates the difficulty of identifying common ecological or demographic factors underlying the evolution of such systems.
The ecological factors that promote the evolution or expression of cooperative breeding behaviour in birds have been a focus of research in social evolution for nearly 50 years (Cockburn 1998; Hatchwell & Komdeur 2000; Ekman et al. 2004). Theoretical work, originating with Hamilton's (1964) introduction of inclusive fitness theory, predicts that cooperative behaviour is most likely to evolve in kin-structured populations (Lehmann & Keller 2006; West, Griffin & Gardner 2007), and this prediction is supported by the observation that cooperative breeding behaviour typically occurs within family groups (Stacey & Koenig 1990; Hughes et al. 2008; Hatchwell 2009). Thus, a key question that arises in attempts to solve the long-standing evolutionary problem of cooperative behaviour is what factors generate kin-structured populations where kin selection might operate?
Hamilton (1964) proposed that population viscosity, that is, limited dispersal, was likely to be a critical factor in kin-selected systems, and most theories regarding the evolution of cooperative breeding behaviour have drawn directly from this idea, emphasising the importance either of constraints on dispersal (e.g. Selander 1964; Emlen 1982; Koenig et al. 1992) or the benefits of philopatry (e.g. Stacey & Ligon 1991; Ekman et al. 2004; Covas & Griesser 2007) in causing offspring to delay dispersal and become helpers. However, among birds, a substantial proportion of species (c. 9%; Cockburn 2006) exhibit some form of cooperative breeding, and the diversity of social organisation within such systems means that the specific ecological and/or demographic factors that limit dispersal are also likely to be diverse. For example, many studies have attributed philopatry to a shortage of vacant breeding territories or of some critical resource essential for successful reproduction (Koenig et al. 1992; Ekman et al. 2004). Such constraints may be particularly prevalent in unpredictable environments where the probability of successful independent reproduction is low or uncertain (Jetz & Rubenstein 2011). These ecological constraints may be exacerbated by demographic factors, such as a low adult mortality rate that reduces the turnover of breeding opportunities, as envisaged by the life-history hypothesis (Arnold & Owens 1998). Furthermore, although there have been numerous evolutionary transitions to and from cooperative breeding in the avian phylogeny, cooperative breeding is also patchily distributed among taxa (Ligon & Burt 2004; Cockburn 2006). Thus, a suite of phylogenetic, ecological and demographic factors is likely to promote the evolution of cooperative breeding.
In the vast majority of cooperatively breeding birds, helping is facultative (Brown 1987), allowing the various factors that covary with the degree of cooperation within a species to be investigated. Such studies have proved extremely successful in identifying several correlates of cooperative behaviour. One approach has been to experimentally relax specific constraints on independent reproduction to examine whether there is a corresponding reduction in helping behaviour. For example, removal experiments showed that a shortage of territories and mates constrained independent reproduction in superb fairy wrens Malurus cyaneus (Pruett-Jones & Lewis 1990). Using the same method, Komdeur (1992) found that the relative quality of territories, as well as their availability, was important in dispersal decisions of Seychelles warblers Acrocephalus sechellensis. Similarly, food supplementation for communally living sociable weavers Philetairus socius was used to reduce constraints on reproduction, resulting in a lower incidence of cooperation (Covas, Doutrelant & DuPlessis 2004); provision of a critical resource (nest cavities) had a similar effect in red-cockaded woodpeckers Picoides borealis (Walters, Copeyon & Carter 1992) and green woodhoopoes Phoenicurus purpureus (DuPlessis 1992).
An alternative approach is to use correlational studies that compare the prevalence of cooperation in two or more populations of the same or very closely related species that differ ecologically or demographically. There have been few such studies, notable exceptions including studies of four congeneric Galapagos mockingbirds Nesomimus spp. (Curry 1989), acorn woodpeckers Melanerpes formicivorus (Koenig & Stacey 1990), pied kingfishers Ceryle rudis (Reyer 1990), pukeko Porphyrio porphyrio (Jamieson 1997), carrion crow Corvus corone (Baglione, Marcos & Canestrari 2002) and long-tailed tits Aegithalos caudatus (Sharp et al. 2011). However, one difficulty in such interpopulation correlative studies is that any variation in cooperative behaviour among populations may not be caused by the factor investigated, but instead by some other ecological or genetic difference among populations.
This problem is largely avoided in correlational analyses using long-term data from the same population in which variation in the prevalence of cooperation across years can be investigated with respect to specific ecological and/or demographic factors. However, there have been remarkably few such analyses (e.g. Emlen 1984; Koenig & Mumme 1987; Pasinelli & Walters 2002; Koenig, Walters & Haydock 2011), perhaps because of the problems of maintaining field studies for sufficient time. Such between-year comparisons are likely to work particularly well in species where helping is a function of factors operating in the same year (i.e. species where help is ‘redirected’ care from failed breeders), rather than a function of breeding success in previous years (i.e. species where help is from grown offspring that have delayed dispersal). Here, we take this approach to investigate the ecological and social correlates of helping behaviour using long-term data from a single population of long-tailed tits A. caudatus (Linnaeus), a facultative cooperative breeder with redirected care by failed breeders.
Long-tailed tits experience a high rate of nest failure; if nest failure occurs early in the breeding season, pairs attempt to breed again, but later in the season, failed breeders may become helpers at the nests of another pair (Glen & Perrins 1988; MacColl & Hatchwell 2002; Hatchwell et al. 2004). Helpers prefer to help close relatives (Russell & Hatchwell 2001) and they work harder when they are more closely related to the recipients (Nam et al. 2010). Helpers increase the rate at which broods are provisioned, so nestlings fledge in better condition and have a higher probability of recruitment into the breeding population (McGowan, Hatchwell & Woodburn 2003). Helpers also lighten the reproductive load of parents, increasing the over-winter survival of male breeders (Meade et al. 2010). Thus, long-tailed tits gain substantial indirect fitness benefits from their cooperative behaviour (MacColl & Hatchwell 2004). By contrast, no direct fitness benefits of helping have been detected (Meade & Hatchwell 2010). In the light of these previous studies, here we use long-term data to investigate whether interannual variation in the degree of cooperation in long-tailed tits is correlated with four features of their ecology and demography: nest predation rate, length of the breeding season, availability of kin and population size.
Specifically, we test the following four hypotheses regarding the factors influencing helping. (i) Most helpers care for close kin (Russell & Hatchwell 2001; Nam et al. 2010), so the availability of relatives should influence the probability of a failed breeder becoming a helper. Elsewhere, we have considered the demographic processes that generate kin structure (Sharp, Simeoni & Hatchwell 2008; Sharp et al. 2008, 2011; Beckerman, Sharp & Hatchwell 2011), but here we focus solely on the outcome of these processes, that is, the availability of close kin. (ii) Nest predation creates potential helpers and determines the number of nests available to be helped (Hatchwell et al. 1999, 2004). We predict that cooperative behaviour will peak at intermediate levels of predation because at low rates of predation, there will be few failed breeders, hence few potential helpers, and at high levels of predation, there will be few nests available to be helped, hence little helping opportunity. (iii) Cooperative behaviour should also be influenced by the time available for personal reproduction. The decision to switch from breeding to helping occurs when the expected pay-off from breeding independently drops below the expected pay-off from helping close kin (MacColl & Hatchwell 2002). Therefore, the length of the breeding season will determine the time available to raise a brood to fledging before breeders abandon independent reproduction for that year, and hence influence the probability of becoming a helper or being helped. (iv) The ecological constraints hypothesis (Emlen 1982) proposes that helping will be more frequent at higher densities because breeding vacancies are scarcer. This may not apply to the non-territorial long-tailed tit, but there may be some generalised competitive effect that reduces the probability of successful breeding at higher population densities. Therefore, we also included population size as a potential predictor of cooperative behaviour.
Materials and methods
Study species and site
A population of 25–72 pairs of long-tailed tits living in the Rivelin Valley, Sheffield, UK (53°23′N, 1°34′W) have been studied since 1994. The data used here were collected from 1995 to 2011, during which time the size of the study area has not changed. The study site comprises c. 3 km2 of farmland, oak and birch woodland and birch and hawthorn scrub. At the start of each breeding season (March), all birds form socially monogamous pairs and attempt to breed independently. Both sexes build nests, and we locate nest sites at this stage of the breeding cycle. Any unringed breeders are caught in mist-nets, uniquely colour-ringed and blood samples are taken from the brachial vein (under UK home office licence); >95% of the adult breeding population is ringed each year. Nests are closely monitored and lay date, clutch size (typically 9–11), hatch date, brood size and fledge date are recorded. Hatching within broods is synchronous and nests are observed from hides (c. 10 m from nests) or from a distance of 20–30 m using binoculars or telescope, on alternate days from day 2 (hatching date = day 0) until fledging (usually day 16 or 17) or nest failure. Observation periods usually last for 1 h, during which time all carers at a nest are identified and recorded. Nestlings are individually marked with unique colour-ring combinations at day 11 of the nestling period, when a blood sample is also taken. Adults and nestlings are sexed using molecular techniques (Griffiths et al. 1998).
Cooperative behaviour and predictor variables
In this study, we test first, how the prevalence of helping behaviour and, second, how the intensity of helping behaviour are influenced by four ecological and demographic variables: predation, relatedness, length of the breeding season and population size.
The prevalence of helping in each year was defined as the proportion of the colour-ringed population that became helpers. Prevalence is, thus, a binomial response variable including the number of ringed birds that became helpers in year n and the number of ringed birds in the population in year n that did not become helpers. The intensity of helping behaviour was defined as the mean number of helpers per helped nest in each year. The number of helpers observed per helped nest ranged from 1 to 8.
We define helpers as any adults other than the breeding pair that were observed to feed nestlings between hatching and fledging. Helpers may arrive at a nest at any stage between hatching and fledging, depending on the time at which their own nest fails, but once a bird starts helping at a nest, it usually continued to do so until the nest failed or the brood fledged. (Hatchwell et al. 2004). Therefore, the contribution of helpers varies considerably according to the stage of the nestling period when they start helping; furthermore, the rate at which individual helpers provision nestlings varies considerably (MacColl & Hatchwell 2002; Nam et al. 2010). However, for the purposes of this study, we simply recorded whether broods were helped or not and how many helpers they had.
‘Predation’ – Long-tailed tit nests have a high probability of failure. A small proportion of failures result from nest abandonment, usually weather-induced and occurring before egg-laying, and in a few cases, failure coincides with the disappearance and presumed death of a breeder. The great majority of failures, however, occurs through depredation of eggs and/or nestlings by a variety of predators, including magpies Pica pica, jays Garrulus glandarius, carrion crows C. corone, weasels Mustela nivalis, stoats Mustela erminea and grey squirrels Sciurus carolinensis (Hatchwell et al. 1999). Any event in which eggs or nestlings disappeared from a nest or the remains of either were found within a nest was recorded as a predation event. In general, predators cause considerable damage to nests, so predation could be unequivocally determined as the cause of failure. If nests were depredated within 2 days of normal fledging age (i.e. day 14 onwards), surrounding areas were searched thoroughly for fledglings, and if any were found, we recorded the nest as having been partially depredated. Thus, ‘predation’ provides a measure of the proportion of pairs that were failed breeders in a given year. We expected a nonlinear relationship between the level of helping in the population and predation rate because at 0% predation, there would be no birds available to help, and at 100% predation, there would be no birds available to be helped. Therefore, we included the proportion of nests depredated as a quadratic term.
‘Relatedness’ – The dyadic relatedness of members of the adult population was determined from social pedigrees using ringing data, assuming that social relatedness provides a close approximation to genetic relatedness. This is a valid assumption because the rate of extra-pair paternity in the study population is low (<7% of nestlings) and intraspecific brood parasitism is negligible (Hatchwell et al. 2002; Simeoni 2011). A substantial majority (73%) of helpers assist at a nest belonging to a first-order relative, typically a sibling (Nam et al. 2010). Relatedness of the population was, therefore, measured as the proportion of the marked population that had at least one-first-order relative alive in the breeding population. There are several alternative ways of measuring population-level relatedness, but we think this is the most relevant metric in the context of the long-tailed tit's cooperative breeding system (Beckerman, Sharp & Hatchwell 2011; Sharp et al. 2011); this issue is considered further in the discussion.
‘Length of the breeding season’ – This was defined as the range of the mid-80% of first egg lay dates for all clutches recorded in each season. This measure effectively reduces the influence of outlying early and late nests. At our study site, the start of the breeding season varies among years by up to 25 days and is influenced by early spring temperature, while the termination of breeding is less variable (MacColl & Hatchwell 2002), although it has advanced over the course of this study (P. Gullett, unpublished data).
‘Population size’ – The size of the adult population was simply the number of birds of breeding age present in the population in a given year, including any unringed birds and birds that only appeared as helpers.
To investigate the effects of ecological and demographic factors on the prevalence of helping, we fitted a generalised linear model with binomial error and a logit link. We modelled the intensity of helping behaviour using a linear model with Gaussian error and an identity link function. Heteroscedasticity and normality of errors were assessed following Crawley (2007). All analyses were conducted within r Version 2.10.1 (2009). Variance inflation factors of the explanatory variables were <2 indicating a lack of collinearity between explanatory variables. For both analyses, our global, maximal model took the form: Response variable ~ predation + predation2 + relatedness + length of breeding season + population size. No other interactions were included as preliminary analysis suggested that none were important. The input variables were standardised following Gelman (2008). We calculated all permutations of the model resulting in 25 candidate models (including the null model) and ranked them in order of their support by the data using the Akaike Information Criterion corrected for small sample sizes (AICc) and the AICc weights. We used 95% confidence to generate our set of top models (Burnham & Anderson 2002). These are the models where the sum of the AICc weight adds to 0·95. We then computed model-averaged predictions across the 95% confidence set of top models using the r package MuMin (Bartón 2011).
Annual variation in response and explanatory variables
The degree of cooperation varied considerably across years, whether measured as the prevalence of helping behaviour (range: 5·4–27·0% of the breeding population; mean = 15·2% ± 5·2 SD, N =17 years; Table 1) or as the intensity of helping (range: 1·2–2·5 helpers per helped nest; mean = 1·8 helpers ± 0·36 SD, N =17 years; Table 1). Thus, there was substantial variation in our measures of helping behaviour, as required for the following analyses.
Table 1. Annual measures of cooperative behaviour and ecological and demographic variables in the study population from 1995 to 2011
Helper prevalence (% of population that help)
Helping intensity (mean number of helpers per helped nest)
Relatedness (% of population with first-order relatives)
Length of breeding season (range (in days) of mid-80% of first egg lay dates)
Predation (% of nests depredated)
The ecological and demographic factors that we considered as potential predictors of helping behaviour also varied across years. The rate of nest predation was consistently high (mean = 71·9% ± 9·5 SD of nests depredated; range: 51·9–84·9%, N =17), so that a minority of pairs bred successfully in each year of the study (Table 1). The proportion of the adult population that was known to have first-order relatives in the adult population ranged from 22·0 to 48·0% (mean = 31·9% ± 7·73 SD, N =17), presumably reflecting variation in demographic rates and productivity in the previous year (Beckerman, Sharp & Hatchwell 2011). The length of the breeding season was also variable (range of mid-80% of first egg lay dates: 13–33 days, N =17), such that in some years, the window of opportunity for independent breeding was more than twice as long as in other years. Finally, population size varied almost threefold across years (Table 1).
Prevalence of cooperative behaviour
Prevalence of helping behaviour was influenced most strongly by predation level. As predicted, our modelling revealed a nonlinear effect of predation on the prevalence of cooperation, with the peak likelihood of cooperation occurring at intermediate levels of predation (Fig. 1). The final averaged model was based on the 95% confidence set of 13 models (see Table 2). Table 3 shows model-averaged results for the 95% confidence set. Predation and predation2 have the greatest relative importance. The confidence intervals of the remaining three variables (‘relatedness’, ‘length of breeding season’ and ‘population size’) include zero, so there is little evidence that they have any effect on the prevalence of helping behaviour in this system.
Table 2. The 95% confidence set of models for (a) the prevalence (proportion of the population helping) and (b) the intensity (number of helpers at helped nests) of helping. For both dependent variables, the full model included all fixed effects: predation (Pred), length of breeding season (LBS), proportion of the population with relatives (Rel), population size (Pop) and predation2. The minimum adequate model that minimised the AICc value was selected using the r package MuMIn
AICc, Akaike Information Criterion corrected.
(a) Prevalence of helping
Pred + Pred2
Pred + Pred2 + Rel
Pop + Pred + Pred2
LBS + Pred + Pred2
Pred + Rel
LBS + Pred
Pred + Pred2 + LBS + Rel
Pred + Pred2 + LBS + Pop
(b) Intensity of helping
LBS + Pop
LBS + Rel
LBS + Pop + Rel
LBS + pred
LBS + Pop + Pred
LBS + Pred + Rel
Pred + Rel
Rel + Pred + Pred2
Rel + Pred
Pop + Rel
Table 3. Summary results after model averaging over the 95% confidence set for (a) the prevalence and (b) the intensity of helping behaviour
The parameter of the greatest relative importance is shown for each model in bold.
Pred, predation; Rel, proportion of the population with living relatives; Pop, population size; LBS, length of breeding season.
Effect sizes have been standardised on two SD following Gelman (2008).
(−1·85 to −1·33)
(−0·43 to 0·36)
(−0·31 to 0·38)
(−0·78 to −0·01)
(−1·52 to −0·15)
(−0·54 to 0·33)
(1·60 to 1·94)
(−0·78 to −0·07)
(−0·61 to 0·06)
(−0·51 to 0·34)
(−1·30 to 0·06)
(−0·64 to 0·11)
Intensity of cooperative behaviour
The intensity of helping behaviour was influenced most strongly by length of the breeding season. The mean number of helpers at helped nests increased with decreasing season length (Fig. 2). The final averaged model was based on the 95% confidence set of 13 models (see Table 2) and those containing length of the breeding season, and combinations of relatedness and population size, received the most support. Confidence intervals of population size, relatedness and predation all include zero implying these factors had little effect on the intensity of helping behaviour.
Using observational data from 17 years of continuous study of cooperative breeding in a population of long-tailed tits, we identified predation rate and breeding season length as the key factors driving annual variation in the level of cooperation. First, as predicted, the prevalence of helping was significantly related to the rate of nest predation, peaking at intermediate predation rates. In contrast, the number of helpers per helped nest was unrelated to predation, but was significantly negatively correlated with the length of the breeding season. Population size and population-level relatedness did not predict prevalence of helping, but were related to the intensity of helping, although to a lesser degree than length of the breeding season.
The absence of an effect of population size on either measure of cooperation in long-tailed tits was unsurprising given the absence of territorial behaviour in long-tailed tits. In typical cooperative breeders, helpers are philopatric offspring who help until eventually dispersing to become breeders when the opportunity arises (Brown 1987). In such species, constraints are more severe at higher population densities, reducing the probability of independent reproduction and thereby increasing the incidence of helping behaviour; this is the essence of the ‘habitat saturation’ hypothesis (Selander 1964; Emlen 1982; Koenig et al. 1992). There is some empirical support for this prediction (e.g. Curry 1989; Komdeur 1992), although not as much as might be expected. However, long-tailed tits do not defend territories or critical, limited resources (e.g. nest sites) for breeding. Furthermore, the adult sex ratio is usually 1 : 1 (Nam, Meade & Hatchwell 2011), and we have little evidence of competition for reproductive opportunities in either sex, although death of a partner may result in mate-switching caused by widowed males usurping a paired male (Simeoni 2011). Thus, most or all adults are probably able to attempt breeding from their first year onwards, regardless of breeding density.
Instead, our results indicate that cooperative breeding in long-tailed tits is promoted by two other constraints on successful reproduction: predation and a short breeding season. Predation risk has often been invoked as an ecological constraint on dispersal (Emlen 1982; Koenig et al. 1992; Kokko & Ekman 2002) and has been demonstrated experimentally in cooperatively breeding cichlids (Heg et al. 2004), where predation effectively sets the scene for helping by causing family formation. In the case of long-tailed tits, the context is different because predation determines the identity of potential helpers and recipients, and the relationship between cooperation and predation was nonlinear, with helping prevalence peaking at intermediate predation rates. Although some other cooperative species exhibit redirected helping as a principal source of helpers (Dickinson & Hatchwell 2004; Hatchwell 2009), we are not aware of any previous study showing a peak in cooperation at intermediate nest failure rates, even though it would be predicted for such species. There is an equivalent process in white-fronted bee-eaters Merops bullockoides, where the pool of potential helpers that may redirect their care to help kin is supplemented by erstwhile breeders who are coerced into helping through harassment and interference (Emlen & Wrege 1992). This has substantial fitness benefits for the coercers and marginal costs for the coerced helpers. We have no direct evidence that long-tailed tit breeders adopt a similar coercive strategy to recruit helpers, but it would be interesting to investigate the relative fitness returns of coercion in this species to determine whether such behaviour would be adaptive.
In contrast to the effect of predation, we found no significant effect of the length of the breeding season on the prevalence of helping. We had predicted that helping would be more frequent in short seasons when there was less opportunity to breed successfully. It is possible that the hypothesised relationship is masked by the effect of predation, which is the ultimate determinant of whether an individual becomes a potential helper or recipient of help. On the other hand, we did find that our other measure of cooperation, the intensity of helping, increased in shorter breeding seasons, as predicted. This finding is consistent with the hypothesis that a temporal constraint on the probability of successful independent reproduction is a key driver of cooperation in this species (MacColl & Hatchwell 2002).
The final predictor of cooperation that we investigated was population-level relatedness. Observations and experiments show that helpers bias their care towards close kin (Russell & Hatchwell 2001; Sharp et al. 2005; Nam et al. 2010), increasing offspring recruitment (Hatchwell et al. 2004) and reducing breeders’ reproductive costs (Meade et al. 2010); hence, helpers derive substantial indirect fitness benefits (MacColl & Hatchwell 2004). By contrast, we have no evidence that helpers gain direct fitness benefits from their cooperative behaviour (Meade & Hatchwell 2010). Therefore, helping in long-tailed tits is a product of kin selection so helping was predicted to correlate positively with relatedness. However, we found a minor effect of population-level relatedness only on the intensity of helping. There are several potential reasons for this. First, we used pedigrees to measure population relatedness, but the pedigree is inevitably incomplete due to immigration. Furthermore, immigrants often disperse in sibling groups, and immigrants may occasionally help each other following their coordinated dispersal (Sharp, Simeoni & Hatchwell 2008), albeit at a low frequency. Secondly, we included both sexes when estimating relatedness, while helpers are predominantly male (85%). The sex ratio of helpers could simply reflect the sex bias in dispersal (Sharp et al. 2008), but there is some evidence that females are less likely to help than males are (Sharp et al. 2011). Therefore, it could be argued that relatedness should be measured only in males, but this would disregard the fact that female helpers are regularly observed. Third, in the Rivelin Valley, most helpers move a relatively short distance from their last failed attempt to the nest where they help (mean = 290 m; Hatchwell et al. 2004). Our measure of relatedness included relatives who were present in the study area but may have dispersed too far from their relatives to be available to help or be helped (Sharp et al. 2011). Finally, although a large majority of helpers are first-order kin of at least one of the helped breeders, some helpers are more distantly related or even unrelated (Nam et al. 2010); these other relationships were not included when estimating relatedness. Nevertheless, among the various ways in which population relatedness can be measured, the proportion of the population with close kin matches the observed pattern of helping best (Beckerman, Sharp & Hatchwell 2011). It should also be noted that the contrast between the effect of kinship on helping at the individual and population level in long-tailed tits emphasises the importance of identifying the appropriate spatial scale when investigating the effect of kinship on the expression of cooperative behaviour (Sharp et al. 2011).
The other important point to make regarding relatedness is that while the level of cooperation was not positively correlated with relatedness in the manner predicted, the presence of kin in the population is obviously critical for the evolution of kin-directed helping. Long-tailed tits are no more philopatric than many ecologically similar non-cooperative species living in similar habitats (Russell 1999; Hatchwell 2009), and Beckerman, Sharp & Hatchwell (2011) show that kin-structured populations can arise instead from the pattern of offspring mortality. Empirical tests of the life history and demographic factors generating kin structure are beyond the scope of this study, but in addition to the direct consequences of predation for the availability of potential helpers and recipients shown here, there may potentially be longer-term effects of nest predation rate in one season on the occurrence of helping in the following season, mediated via population relatedness.
Finally, the analyses presented here emphasise the difficulty facing researchers attempting to seek general explanations for the evolution of cooperative breeding. Many comparative studies have sought to identify more or less specific correlates of cooperative breeding across the avian phylogeny with equivocal results, for example, Ford et al. (1988), DuPlessis, Siegfried & Armstrong (1995), Cockburn (1996), Arnold & Owens (1998, 1999), Rubenstein & Lovette (2007) and Jetz & Rubenstein (2011). It could be argued that the factors influencing the degree of cooperation in long-tailed tits are unusual because they are atypical cooperative breeders, but many species of cooperative breeder have helping that occurs within ‘kin neighbourhoods’, sometimes involving redirected helping (Dickinson & Hatchwell 2004; Hatchwell 2009), in which nest predation, coupled with temporal constraints on breeding, is a likely source of potential helpers. This study also illustrates the fact that even a simple factor such as predation rate (and hence the probability of successful independent reproduction – a cornerstone of the ecological constraints hypothesis) may not be related to cooperation in a linear fashion. Accommodating such diverse drivers of cooperation within comparative analyses remains a significant challenge.
We thank A.F. Russell, D.J. Ross, M.K. Fowlie, A.D.C. MacColl, A. McGowan, K.-B. Nam, M. Simeoni, N. Green, A. Bamford, D. Richardson, J.W. Lee, P. Gullett and C. Napper for their invaluable assistance with data collection, and Sheffield City Council, Yorkshire Water and Hallamshire Golf Club for permission to work on their land. We also thank A.F. Russell and B.K. Woodward for discussion of ideas presented here. T. Coulson and two anonymous reviewers provided helpful comments on the manuscript. Long-term data were collected under grants awarded to BJH by Natural Environment Research Council, Nuffield Foundation, Association for the Study of Animal Behaviour and the University of Sheffield for which we are most grateful. Birds were ringed under BTO permit C3770, and blood samples were taken under Home Office project licence 4003214. This study was initiated during award of a Leverhulme Trust Research Fellowship to BJH.