The contribution of an avian top predator to selection in prey species

Authors


Summary

  1. Natural selection can vary in magnitude, form and direction, yet the causes of selection, and of variation in selection, are poorly understood.

  2. We quantified the effect of a key predator (Eurasian sparrowhawks) on selection on fledging body mass in two bird species (blue tits and great tits). By partitioning selection into within- and between-brood components, we were able to separate individual from brood-level effects of fledging mass on predation probability and recruitment.

  3. In blue tits, selection on fledging mass by sparrowhawk predation was nonsignificant and could not explain selection to recruitment. In contrast, in great tits, sparrowhawk predation selected for increased fledging mass at the individual level and could explain 73% of individual-level selection on fledging mass to recruitment.

  4. Moreover, in great tits, individual-level selection on fledging mass was significantly stronger in years in which sparrowhawks were present compared to years when sparrowhawks were absent. Selection at the brood level was independent of sparrowhawk presence.

  5. These results provide compelling evidence that sparrowhawk predation acts as an important causal agent of selection on fledging mass in great tits but not in blue tits. Variation in predation pressure can therefore account for variation in selection, but specific patterns may not be easily generalized across species.

Introduction

Natural selection can vary in magnitude, form and direction, with far-reaching consequences for the genetic architecture and evolutionary trajectories of natural populations (Kingsolver et al. 2001, 2012; Siepielski, DiBattista & Carlson 2009; Bell 2010). However, the causes of natural selection, and the reasons for its variation, are currently poorly understood (MacColl 2011). Since it is the environment that acts as the agent of selection by causing differences in fitness between phenotypes (Wade & Kalisz 1990; MacColl 2011), identifying environmental factors that are responsible for observed selection patterns will greatly increase our understanding of variation in selection (e.g. Calsbeek & Cox 2010).

Among the range of biotic and abiotic factors that can act as agents of selection (e.g. climate, food availability, habitat, inter- and intraspecific competition), predators can have profound impact, as they are a direct cause of mortality and often select their prey nonrandomly (Losos, Schoener & Spiller 2004; Carlson, Rich & Quinn 2009; Genovart et al. 2010). However, the impact of predators can be overestimated if they predominantly target individuals that would have low fitness for other reasons (e.g. injury, disease, starvation; Temple 1987). Prey body mass is a trait recognized as sensitive to selection imposed by predators, because heavy individuals may lack the agility and speed to escape predation (Witter & Cuthill 1993), or, alternatively, because light individuals may be more sensitive to starvation and take more risks, or be less able to escape predation due to a lower body condition (Lima 1998). Selection on body mass, by predators, can therefore be hypothesized to take almost any shape.

In birds, body mass just prior to fledging is a trait routinely measured in population studies, and several studies have found it to be positively correlated with the probability of recruitment to the natal population (reviewed by Gebhardt-Henrich & Richner 1998). Studies that quantified selection on fledging mass found it to vary between species, populations and years, from no or stabilizing selection to positive directional selection (e.g. Garnett 1981; Nur 1984; Tinbergen & Boerlijst 1990; Lindén, Gustafsson & Pärt 1992), and at present, our understanding of the cause of this variation is lacking. The presence/absence of predators (e.g. Eurasian sparrowhawks Accipiter nisus) has been suggested as a potential cause of variation in selection, as predators may select against high fledging mass and cause selection to be stabilizing, or to level off at high fledging mass (Tinbergen & Boerlijst 1990; Adriaensen et al. 1998). Yet, these studies have not confirmed the predators as agents of selection by testing whether they indeed take the heavier fledglings. Studies that investigated individual sparrowhawk prey found that, in certain years, sparrowhawks more often took the lighter fledglings (Geer 1982; Götmark 2002).

Most studies on fledging mass and survival (but see Tinbergen & Boerlijst 1990) do not distinguish between soft selection (selection operating only within groups; Wallace 1968) and selection where individual fitness is dependent on group properties (Goodnight, Schwartz & Stevens 1992). This distinction is, however, especially important in relation to predation, because predation will cause different selection patterns within and between broods, if, for example, predators take the lightest fledglings in a brood, but are randomly attracted to broods (soft selection), or, alternatively, if they are attracted to greater expression of conspicuous behaviour (e.g. loud begging) in broods that are light on average, but randomly take fledglings from those broods (a form of group selection). Both mechanisms will manifest as population-level selection against low fledging mass, but may have different consequences for evolutionary adaptation (Goodnight, Schwartz & Stevens 1992). Moreover, if the occurrence of soft selection and group selection can be excluded, it is predicted that selection on fledging mass per se at the individual level would lead to similar selection patterns within and between broods (Tinbergen & Boerlijst 1990).

Here we study prey selection among fledglings great tits (Parus major) and blue tits (Cyanistes caeruleus) by Eurasian sparrowhawks in Wytham Woods (Oxfordshire, UK). Great tits breeding in this forest have been monitored with a constant research protocol since 1960, while blue tits have been monitored in shorter periods for specific projects. In this population, great tit fledging mass is heritable and consistently under positive directional selection (Perrins 1965; Garnett 1981; Garant et al. 2004), while selection on fledging mass in blue tits was either absent or positive but levelling off at high mass (Nur 1984). In this study, we quantify selection on fledging mass imposed by sparrowhawks in both species and compare selection by sparrowhawks to overall selection on fledging mass to recruitment. For great tits, we additionally test whether patterns of selection on fledging mass differ between periods when sparrowhawks were absent from, and re-established in, Wytham Woods. In all analyses, we partition selection into within- and between-brood components (following van de Pol & Wright 2009) to separate individual- and brood-level effects of fledging mass on predation probability and recruitment.

Materials and methods

Study Species and General Data Collection

Blue and great tits are small (~11 and 18 g, respectively) passerine birds that readily accept nestboxes for breeding (Cramp & Perrins 1993). The data we present here come from a study population breeding in nestboxes in the c. 385 ha mixed deciduous woodland of Wytham Woods (Oxfordshire, UK, 51°46′ N1 °20′W). Routine data collection in the breeding season (April–June) consists of nestbox checks to record occupation, laying date, clutch size, hatching date and breeding success. Nestlings are weighed (to the nearest 0·1 g) and ringed (with uniquely numbered metal leg rings) when the oldest chick in the nest is 15 days of age, that is, when mass has approached an asymptote (van Balen 1973). Brood size is defined as the number of nestlings alive at this visit. After fledging, young tits roam around in family flocks for up to several weeks, while still dependent on parental care (Verhulst & Hut 1996). Parents are caught in the nestbox while feeding nestlings, and identified by their ring, or newly ringed if immigrant. Recruits to the natal population are identified as locally hatched birds that were caught as parents in subsequent years.

In the breeding season of 2011, all fledglings were additionally fitted with passive integrative transponder (PIT) tags integrated in plastic rings (IB Technology, Aylesbury, UK), fitted around the unringed leg, to allow automatic identification with RFID antennae (Dorset ID, Aalten, the Netherlands). For the analyses of the 2011 recruitment probability, we defined recruits as fledglings from 2011 that were identified as parents feeding nestlings in Wytham Woods in 2012, either by standard capture or through automated PIT-tag detection.

Sparrowhawk Predation

Eurasian sparrowhawks are forest-living raptors that mainly hunt on small passerines (Newton 1986). They are the principal predator of tits and considered the top predator in the well-studied oak-caterpillar–tit–sparrowhawk food chain (e.g. Both et al. 2009). They breed annually and roughly time offspring hatching around the peak in tit fledging (Newton 1986).

To identify tit fledglings that were caught by sparrowhawks, we aimed to locate all sparrowhawk nests in Wytham Woods in 2011 by intensive searches at known previous nesting locations and other suitable habitat. Nesting areas located were thoroughly searched for regurgitated sparrowhawk pellets and other prey remains on the ground. Because sparrowhawks specialize on live prey (only occasionally taking very large items as carrion, Newton 1986), metal rings and PIT tags obtained from these remains were used to infer which tits were caught by sparrowhawks. Nesting locations were visited at least once a week, after fledging of the majority of tit nestlings, and visits continued until no new rings or PIT tags were found. In practice, the first rings and PIT tags were found at the end of May and the last ones at the end of July. However, because we cannot be certain that rings or PIT tags were not missed at earlier visits, we cannot estimate the timing of the actual predation event. Since tits are typical prey species for male sparrowhawks (females generally take larger prey) and the female does not hunt in the incubation and early nestling phase (Newton 1986), the majority of selection on fledglings tits that we report here will have been exerted by male sparrowhawks. Note that, despite considerable effort, we may not have located all sparrowhawk nests, and the tit prey identified near the nests may only represent a small, but likely random, fraction of the total number of tits caught.

Sample Sizes

In 2011, a total of 3424 ringed blue tits and 1684 ringed great tits fledged in Wytham Woods and surrounding area (Fig. S1, Supporting information). We located five sparrowhawk nests in the area (Fig. S1) at which we found rings and/or PIT tags that originated from 108 tits (79 blue tits and 29 great tits), of which 83 (76·9%) were fledglings from 2011 (59 blue tits and 24 great tits). From 59 tit broods, one fledglings was identified as caught by a sparrowhawk; from 9 tit broods, two fledglings were identified as caught by sparrowhawks; and from two tit broods, three fledglings were identified as caught by sparrowhawks. In four of the cases where at least two fledglings of the same brood were caught, they were caught by different sparrowhawks.

Because we did not locate any sparrowhawk nest in the east of Wytham Woods (Fig. S1), we excluded tit fledglings from this area (i.e. Pasticks, Marley, Marley Plantation and Higgins, see Fig. S1) in our analysis of the probability to be caught by a sparrowhawk, as well as our analysis of the probability of recruiting as a breeder in the following year. In this way, we avoid potential covariance between fledging characteristics and sparrowhawk predation that might be caused by variation in area quality; note that this is also avoided by specifically looking at the predation probability within broods (see ‘'Statistical Analyses'’). This data selection results in 2525 blue tit fledglings from 313 broods of which 56 (2·2%) recruited in 2012, and 1208 great tit fledglings from 165 broods of which 86 (7·1%) recruited in 2012, for the analyses of predation probability and recruitment probability, but sample sizes for specific analyses can slightly differ because of missing information on certain covariates for a small fraction of broods (see tables for final sample sizes).

To analyse whether selection to recruitment on fledging mass differed in years with and without sparrowhawks present, we used a data set of all great tits fledged from nestboxes in Wytham Woods from 1960 to 2010 (= 51 years, 9822 broods, 73 121 fledglings, 6926 recruits). Sparrowhawks were absent from Wytham Woods in the early years (1960–1972), but re-established from 1973 onwards (Gosler, Greenwood & Perrins 1995).

Statistical Analyses

We used generalized linear mixed models (GLMM) with a binomial error distribution and a logit link function, with brood identity as a random effect, to test for effects of fledging mass on predation probability. Final models were derived through backwards elimination of least significant terms, starting with highest-order interaction terms. Only significant terms (< 0·05) were included in final models. Nonsignificance of excluded terms was confirmed by re-entry in the final model.

To separate effects of fledging mass operating at the within- and between-brood levels, we included both the average fledging mass of a brood and the deviation from that average (referred to as ‘within-brood fledging mass’), for each individual, as linear and quadratic explanatory variables (following van de Pol & Wright 2009). We included brood size and hatching date as covariates to correct for the possibility that associations between fledging mass and predation probability are caused by brood size and/or seasonal differences in predation pressure, instead of fledging mass per se. To analyse whether effects of fledging mass differed significantly between the within- and between-brood levels, we tested the effect of average brood fledging mass in models together with individual fledging mass (following van de Pol & Wright 2009).

To analyse whether effects differed between prey species, we included species (blue tit or great tit) as a categorical variable and its interactions with all other terms. All variables were standardized, by calculating z-scores, to account for species differences in the average and variance of these terms. For the average brood fledging mass, this was done using the average and standard deviation of all brood averages, while for within-brood fledging mass, we used zero as the average and the average of all within-brood standard deviations.

To compare selection by sparrowhawks with overall selection to recruitment, we calculated standardized linear selection differentials (S) following Arnold & Wade (1984a,b). To this end, we used standardized average and within-brood fledging mass, as described above, in a regression model with relative fitness as the dependent variable, and a normal error distribution. Relative fitness for selection by predation (0 = predated, 1 = not predated) and selection to recruitment (0 = not recruited, 1 = recruited) was calculated by dividing the individual value by the species mean.

To test whether effects of fledging mass on great tit recruitment differed between years with and without sparrowhawks, we tested for interaction effects between sparrowhawk presence (absent: 1960–1972, present: 1973–2010; Gosler, Greenwood & Perrins 1995) and (quadratic) fledging mass at the between-year, between-brood and the within-brood level. This was done by including the average fledging mass of a year, the difference between year average and brood average, and the difference between brood average and individual fledging mass (following van de Pol & Wright 2009), and their quadratic terms and interactions with sparrowhawk presence, as explanatory variables. This model used a binomial error distribution with a logit link function, with year and brood identity as hierarchically structured random effects, and the probability to recruit as a breeder as dependent variable.

All models were run in mlwin 2.0 (Rasbash et al. 2004), and significance (two-tailed) of all explanatory variables was determined using the Wald statistic, which approximates the χ2-distribution.

Results

Prey Selection by Sparrowhawks

In great tits, fledglings from on average heavier broods were less likely to be caught by a sparrowhawk (Fig. 1a, Table 1a). Similarly, within broods, the heaviest fledglings were least likely to be caught (Fig. 1b, Table 1a). There was, however, a significant quadratic effect of fledging mass within broods (Table 1a), suggesting that the effect of fledging mass levelled off at low mass (Fig. 1b). The linear effects of average and within-brood fledging mass on predation probability did not differ significantly (coefficient ± SE = 0·02 ± 0·22, χ2 = 0·01 Δdf = 1, P = 0·920), indicating that both effects may be caused by the same mechanism; that is, sparrowhawks taking individual great tit fledglings with low body mass.

Table 1. Summary of logistic regression models testing for effects on the probability to be caught by a sparrowhawk, for (a) great tit fledglings only, (b) blue tit fledglings only and (c) fledglings of both species combined. Statistics of nonsignificant terms are presented as when added to the final model. Explanatory variables are standardized in the analyses of both species combined (see ‘'Materials and methods'’)
Final modelCoefficient (SE) χ 2 Δdf P
(a) Dependent variable: caught by sparrowhawk (only great tit fledglings)
Intercept1·84 (2·49)   
Average brood fledging mass−0·32 (0·14)5·0810·024
Within-brood fledging mass−1·57 (0·61)6·5710·010
Within-brood fledging mass2−1·07 (0·52)4·2110·040
Nonsignificant
Average brood fledging mass20·02 (0·06)0·1110·740
Hatching date−0·13 (0·08)2·8210·093
Brood size (n = 1182 fledglings, 162 broods)0·03 (0·12)0·0510·823
(b) Dependent variable: caught by sparrowhawk (only blue tit fledglings)
Intercept−3·74 (0·14)   
Nonsignificant
Average brood fledging mass0·45 (0·23)3·8310·050
Average brood fledging mass2−0·33 (0·30)1·1610·281
Within-brood fledging mass0·08 (0·24)0·1010·752
Within-brood fledging mass2−0·45 (0·39)1·3310·249
Hatching date0·02 (0·04)0·1810·671
Brood size (= 2525 fledglings, 313 broods)0·08 (0·08)1·0110·315
(c) Dependent variable: caught by sparrowhawk (great tit and blue tit fledglings)
Intercept−3·79 (0·15)   
Species (BT = ref)−0·32 (0·30)1·1910·275
Average brood fledging mass0·37 (0·18)4·1710·041
Species × Average brood fledging mass−0·93 (0·28)10·8410·001
Hatching date0·13 (0·16)0·6710·413
Species × Hatching date−0·81 (0·41)3·9710·046
Nonsignificant
Average brood fledging mass2−0·09 (0·10)0·7710·380
Species × Average brood fledging mass20·19 (0·21)0·7610·383
Within-brood fledging mass−0·09 (0·11)0·6210·431
Within-brood fledging mass2−0·19 (0·12)2·5710·109
Species × Within-brood fledging mass−0·33 (0·20)2·8210·093
Species × Within-brood fledging mass2−0·79 (0·49)2·5610·110
Brood size0·12 (0·15)0·6410·424
Species × Brood size (n = 3707 fledglings, 475 broods)−0·35 (0·33)1·1910·275
Figure 1.

Proportion of fledglings caught by sparrowhawks (a) plotted against the average fledging mass of the brood, for blue tits and great tits. Data points are averages for x-axis categories ≤9·5, >9·5–10·5, >10·5–11·5, >11·5 in blue tits and ≤16·5, >16·5–17·5, >17·5–18·5, >18·5–19·5, >19·5 in great tits. Sample sizes per category are 119, 282, 1528, 587 and 142, 235, 388, 280, 147 fledglings, respectively. Trend lines represent the linear effects of average brood fledging mass (see Table 1). (b) Plotted against individual fledging mass relative to the brood average (within-brood fledging mass), for blue and great tits. Data points are averages for x-axis categories ≤−1, >−1–0·5, >−0·5–0·0, >0·0–0·5, >0·5–1·0, >1·0 in both blue and great tits. Sample sizes per category are 95, 295, 826, 831, 383, 51 fledglings in blue tits and 122, 182, 282, 272, 185, 139 fledglings in great tits, respectively. Trend lines represent the linear and quadratic effects of within-brood fledging mass (see Table 1). In both graphs, data points are located at the average fledging mass in the category. Fledging mass is categorized for visual purpose only, and all statistical analyses are on fledging mass as continuous variable.

In contrast, in blue tits, there was no significant effect of fledging mass on predation probability, but there was a trend for sparrowhawks to take more fledglings from heavier broods (Fig. 1, Table 1b). However, the linear effects of average and within-brood fledging mass on predation probability were not significantly different (coefficient ± SE = 0·39 ± 0·34, χ2 = 1·29, Δdf = 1, = 0·256), suggesting that sparrowhawks do not directly select blue tit fledglings at the brood level on the basis of fledging mass.

The effect of average brood fledging mass on sparrowhawk predation probability in great tits was highly significantly different from its effect on predation probability in blue tits (Table 1c). The effect of hatching date on predation probability was also significantly different between the two species (Table 1c), but this should be treated with caution as these contrasting effects were not significantly different from zero within the two species (Table 1a,b). The nonsignificance of the interaction between within-brood fledging mass and species (Table 1c) is likely due to a lack of statistical power, because when we increased the power to detect within-brood effects, by only including broods with at least one fledglings taken by a sparrowhawk, this interaction effect was significant (coefficient ± SE = 1·01 ± 0·38, χ2 = 7·07, Δdf = 1, = 0·008).

Selection by Sparrowhawks vs. Overall Selection to Recruitment

In great tits, the standardized linear selection differential (S) for fledging mass, caused by sparrowhawk predation, was 0·010 (SE = 0·005) between broods and 0·008 (SE = 0·004) within broods. In blue tits, S caused by sparrowhawks was -0·007 (SE = 0·003) between broods and −0·001 (SE = 0·003) within broods. These estimates of S caused by sparrowhawk predation were in most cases considerably lower than the estimates for selection to recruitment (S ± SE: GTbetween broods = 0·472 ± 0·117, GTwithin broods = 0·011 ± 0·105, BTbetween broods = 0·377 ± 0·143, BTwithin broods = 0·240 ± 0·134, Fig. 2a). However, for selection at the within-brood level in great tits, selection by sparrowhawk predation explained 73% of the overall selection to recruitment on fledging mass (Fig. 2b).

Figure 2.

(a) Standardized selection differentials, and standard errors, of overall selection to recruitment, plotted against standardized selection differentials, and standard errors, caused by sparrowhawk predation, for blue tit and great tit between-brood and within-brood selection on fledging mass. (b) Relative importance of between-brood and within-brood selection on fledging mass caused by sparrowhawk predation (Spredation/Soverall) for blue tits (open bars) and great tits (filled bars).

Selection on Fledging Mass with and Without Sparrowhawks

The linear within-brood effect of fledging mass on recruitment probability in great tits varied significantly with sparrowhawk presence (Table 2). In years in which sparrowhawks were present, the heavier fledglings from a brood were more likely to recruit, while there was no such effect in years in which sparrowhawks were absent (Fig. 3). This pattern could not alternatively be explained by a gradual change in the form of selection on fledging mass over time, unrelated to sparrowhawk presence, as the interaction between year and within-brood fledging mass was not significant (coefficient ± SE = 0·00 ± 0·00, χ2 = 0·39, Δdf = 1, = 0·532) when added to the final model (Table 2). Independent of sparrowhawk presence, recruitment probability was higher for great tits fledged in years with greater average fledging mass, and for fledglings from on average heavier broods (Table 2). However, the positive effect of fledging mass levelled off at highest mass, both between and within broods (Table 2).

Table 2. Summary of a logistic regression model testing for effects on the probability to recruit as a breeder, for all great tit fledglings hatched in Wytham Woods from 1960 to 2010. Statistics of nonsignificant terms are presented as when added to the final model
Final modelCoefficient (SE) χ 2 Δdf P
Dependent variable: recruited as a breeder (only great tit fledglings)
Intercept−13·72 (3·07)   
Average year fledging mass0·61 (0·16)13·931< 0·001
Average brood fledging mass0·18 (0·01)150·361< 0·001
Average brood fledging mass2−0·06 (0·01)58·911< 0·001
Within-brood fledging mass−0·10 (0·04)5·6310·018
Within-brood fledging mass2−0·07 (0·01)34·141<0 ·001
Sparrowhawks present (no = ref)0·30 (0·14)4·2710·039
Within-brood fledging mass × Sparrowhawks present0·19 (0·05)18·161<0 ·001
Nonsignificant
Average year fledging mass20·00 (0·00)0·0011·000
Average year fledging mass × Sparrowhawks present0·18 (0·34)0·2710·603
Average year fledging mass2 × Sparrowhawks present0·00 (0·00)0·0011·000
Average brood fledging mass × Sparrowhawks present0·03 (0·04)0·6310·427
Average brood fledging mass2 × Sparrowhawks present0·02 (0·03)0·7510·386

Within-brood fledging mass2 × sparrowhawks present

(= 73 121 fledglings, 9822 broods, 51 years)

0·02 (0·03)0·2710·603
Figure 3.

Proportion of great tit fledglings recruited as a breeder plotted against individual fledging mass relative to the brood average (within-brood fledging mass) in years with and without sparrowhawks. Data points are averages for x-axis categories ≤−2, >−2–1·5, >−1·5–1·0 > −1·0–0·5, >0·5–0·0, >0·0–0·5, >0·5–1·0, >1·0–1·5, >1·5–2·0, >2·0. Sample sizes per category are 303, 268, 634, 1435, 2008, 2241, 1586, 837, 310, 175 fledglings in years without sparrowhawks and 1138, 1509, 4270, 9515, 14 089, 14 976, 10 704, 4976, 1678, 559 fledglings in years with sparrowhawks, respectively. Trend lines represent the final model in Table 2, for average year and brood fledging mass. Data points are located at the average fledging mass in the category. Fledging mass is categorized for visual purpose only, and all statistical analyses are on fledging mass as continuous variable.

Discussion

We found that sparrowhawks selectively took light great tit fledglings, while blue tit fledglings were taken at random with respect to their body mass. Selection by a predator against low mass as observed in great tits is not in agreement with results from other taxa, in which predation-induced selection against large body size is often reported (reviewed by Blanckenhorn 2000). This discrepancy may be explained by the fact that fledging mass is a composite trait, consisting of structural size and body reserves or condition (Green 2001). Since it is unlikely that sparrowhawks specifically target great tit fledglings with small structural size, the dependence on body condition to avoid predation may override all advantages a small structural size can entail in escaping predators. More specifically, louder begging of hungry fledglings, allowing easier detection by sparrowhawks, or poorer flight-feather or muscle development, diminishing the chances of escape once detected, are potential mechanistic causes of selection against low fledging mass by sparrowhawks in great tits. Selection against individuals in poor condition is also found in other species (Temple 1987). Perhaps the overall smaller size of blue tits, and therefore higher mass-specific metabolism (West, Brown & Enquist 1997), may force them to take more risks irrespective of individual body condition, causing their predation probability to be less dependent on body mass.

An alternative explanation for our result would be that sparrowhawks do not return the heaviest great tit fledglings to the nesting area, such that the prey we identified is biased towards lighter individuals. Although very large prey items may indeed be consumed at the site of capture, great tits are by no means a large prey item for male sparrowhawks (they can take prey up to 120 g, Newton 1986) and this potential bias does not appear to occur in the mass range of fledglings tits (Newton & Marquiss 1982). Moreover, the rings and PIT tags we found mainly come from regurgitated pellets of adults (i.e. found under a favourite roosting tree). Hence, even if prey is consumed elsewhere, the remains found near the nest are likely to represent a random sample of prey taken.

For both prey species, the linear within-nest covariance between fledging mass and predation probability did not differ from the linear between-nest covariance between fledging mass and predation probability. Hence, we found no suggestion that predator-induced selection on fledging mass acts exclusively within groups of siblings from the same brood (soft selection) or exclusively between broods (a form of group selection). Our results are therefore most parsimoniously explained by selection acting at the individual level, such that the between-brood effect in great tits is only an indirect consequence of a higher predation probability of lighter individuals that are, by definition, over-represented in on average lighter broods.

While selection on fledging mass by sparrowhawk predation is similar at the within- and between-brood level, selection on great tit fledging mass to recruitment is on average 2·6 times greater between than within broods, in Wytham Woods' great tits (S. Bouwhuis et al., unpublished results). Considerably greater selection to recruitment on fledging mass between than within broods was also found in this study (Fig. 2). Consequently, predation by sparrowhawks could only explain within-brood or individual-level selection to recruitment on fledging mass. However, if there are no agents of selection that directly target group properties (and we argue that we can exclude the specific group-level selection that predators may impose as outlined in the introduction to this study), then the between-brood covariance between fledging mass and recruitment is unlikely to represent selection on fledging mass per se (Tinbergen & Boerlijst 1990). Instead, the covariance between recruitment and fledging mass at the brood level is more likely to result from independent effects of the early brood environment on both fledging mass and recruitment probability (S. Bouwhuis et al., unpublished results). Sparrowhawk predation may thus be the agent of selection responsible for the majority of selection on fledging mass per se, in great tits. This is supported by other lines of evidence; selection on great tit fledging mass predominantly takes place in the first few weeks after fledging (S. Bouwhuis & B. C. Sheldon, unpublished results), and the main source of mortality in this period is predation (Naef-Daenzer, Widmer & Nuber 2001), and sparrowhawks are estimated to take 18–34% of all great tit fledglings in this period (Perrins & Geer 1980). However, since within-brood selection to recruitment on fledging mass was greater for blue tits, which could not be explained by sparrowhawk predation, other agents of selection on avian fledging mass also exist, and their relative importance may vary between species.

Inferring causality for agents of selection can be difficult. Covariance between body condition and predation probability does not necessarily imply that predators are causal agents of selection, because predators might simply take individuals that due to their poor condition would eventually have died of other causes in the absence of predators (Temple 1987). The causal effects of predators on selection are therefore best established with randomized trials with and without predators in a natural environment (Brönmark & Miner 1992; Losos, Schoener & Spiller 2004; Reznick & Ghalambor 2005; Calsbeek & Cox 2010). However, such experiments may be difficult, if not impossible, when subjects cannot be confined to a limited environment (e.g. birds), as well as raise considerable ethical problems. In this study, we instead made use of natural variation in predator presence and found that directional selection to recruitment on great tit fledging mass, at the individual level, was absent in years without sparrowhawks (Fig. 3), following the pattern predicted based on individual sparrowhawk prey. Hence, we showed that (i) sparrowhawks take individual great tit fledglings based on their mass, indicating that sparrowhawk presence is not merely correlated with selection on fledging mass, and (ii) without sparrowhawks present, there was no selection on fledging mass at the individual level, indicating that sparrowhawks do not merely take individuals that would have died of other causes anyway. As such this study may represent the best possible evidence in birds, for predation to act as a causal agent of selection on a key predictor of individual fitness.

Acknowledgements

This work was supported by two ‘Rubicon’ fellowships of the Netherlands Organisation for Scientific Research (NWO) to OV and SB, and recent data collection by an ERC advanced grant (AdG 251064) to BCS. We thank the many people who helped to collect field data analysed here, and, in particular, Ross Crates for locating one of the sparrowhawk nests and John Quinn for providing detailed information on previous sparrowhawk nesting locations. Ross MacLeod and two anonymous reviewers provided helpful comments on the manuscript.

Ancillary