Size and immune function as predictors of predation risk in nestling and newly fledged jackdaws

Prey choice by predators may be based on the potential prey's condition, for example resulting in sub- standard individuals running a higher risk of being predated. Over 5 years, we studied young jackdaws, Corvus monedula , to determine whether size and innate baseline immune function may predict predation risk by goshawks, Accipiter gentilis , during the nestling and early ﬂ edging phases. We measured body mass, wing length, tarsus length and four innate immune indices two to four times when nestlings were 12 e 29 days old. To determine which individuals had been predated during the nestling phase and shortly after ﬂ edging, we searched for metal rings of the jackdaws in the only goshawk territory close to the jackdaw colony. Nestling mortality before 12 days of age was entirely due to starvation, whereas between 12 days of age and ﬂ edging, mortality was mainly due to predation. Nestlings with smaller size (mass, wing, tarsus) and low lysis titre and haptoglobin concentrations were at a higher risk of being predated before ﬂ edging. Directly after ﬂ edging, individuals with short wings were preferentially pre- dated, with no effects of body mass, tarsus length or any of the four immune indices measured at day 29 (i.e. shortly before ﬂ edging). That lower immune function and smaller size predict predation risk in nestlings may re ﬂ ect that these individuals are of poor quality and/or lag behind in development. We hypothesize that hunger makes these nestlings sit closest to the entrance hole and hence become the ﬁ rst to be predated. For ﬂ edglings, our results suggest that poor ﬂ ight ability makes individuals with short wings

Predation is one of the major reasons for mortality and hence is a major factor influencing fitness in wild animal populations (Newton, 1998). Predation events can be seen either from the perspective of a predator that needs to select individual prey from a pool of potential prey or from the perspective of a potential prey individual seeking to avoid predation. Predators often actively select prey and selecting the right type of prey can be the difference between feeding and starving. For example, if insects are abundant, insectivorous predators (e.g. European starlings, Sturnus vulgaris, and common shrews, Sorex araneus) tend to prefer larger insects (Barnard & Brown, 1981;Tinbergen, 1981). A preference for larger prey has also been observed in Geoffroy's cats, Oncifelis geoffroyi, which predate larger prey species; the likely reason for this is to maximize energy intake and reduce daily foraging costs (Canepuccia et al., 2007). Studies analysing prey size in sparrowhawks, Accipiter nisus, short-toed eagles, Circaetus gallicus, and sharp-shinned hawks, Accipiter striatus, indicated that these three raptors prefer prey species of intermediate size over smaller and larger ones (Gil & Pleguezuelos, 2001;Roth et al., 2006;Selås, 1993). Apart from the size of individual prey items, there might be other characteristics that make a certain prey more favoured than other potential prey. Based on the few studies on prey selection in top predators available at the time (Newton, 1979), Temple (1987) proposed that vertebrate predators would more often choose substandard (e.g. young, sick, disabled, low-quality/ condition) prey, because such individuals are easier to catch and handle. More recent studies have indicated that individuals with increased parasite load (Murray et al., 1997;Møller & Nielsen, 2007;Navarro et al., 2004), small spleens (Møller & Erritzøe, 2000) or poor condition (Genovart et al., 2010;Penteriani et al., 2008;Tucker et al., 2016) are at a much higher risk of being predated. However, the physiological mechanisms behind these patterns are largely unknown. Furthermore, most studies ignore predation of nestlings and many combine juveniles and adults, limiting our knowledge of what individual characteristics at different life stages (nestling, juvenile, adult) drive predation patterns.
Viewed from the perspective of the prey, predation should be avoided. Various plastic responses and evolved traits, such as behaviour (e.g. escaping, Perkins et al., 2014; change in foraging strategy, Eccard et al., 2020), morphology (e.g. phenotypic changes, Br€ onmark & Miner, 1992) and physiological capacity can accomplish this. Physiological capacity (or 'condition') includes the capacity of the immune system. The immune system is a major physiological component of self-maintenance and promotes survival by reducing the probability of disease-related mortality (Roitt et al., 1998) but it also incurs costs and is therefore a powerful driver in many evolutionary and ecological processes (Hasselquist & Nilsson, 2012). As such, several studies have reported predictive capacities of immune parameters on subsequent survival in free-living birds (Hegemann et al., 2013, Roast et al., 2020Wilcoxen et al., 2010). This may be the result of direct links between immune function and the ability to withstand disease outbreaks (Wilcoxen et al., 2010); however, the immune system may also have direct consequences for predation probability. In particular, individuals mounting an immune response to an infection will often be less physically active (Hart, 1988). While this could mean they are less exposed to predators for some species, many species will not have access to predator-safe refuges or still need to be active to some extent during an immune challenge even though they modify their behaviour (Hegemann et al., 2018). In the latter cases, potential prey may be less vigilant and/or pose less resistance during a predation event, which may reduce the effort the predator will have to invest in capturing and handling the prey (Alzaga et al., 2008;Penteriani et al., 2008;Temple, 1987). Yet, few studies have quantified whether animals showing indications of ongoing immune responses are indeed exposed to increased predation risk. Alternatively, and not mutually exclusive, individuals with a poor baseline (constitutive) immune system might have high levels of parasites (Whiteman et al., 2006), resulting in poor body condition (Alzaga et al., 2008). Such a link is suggested as small passerine birds with low levels of baseline immune function often show low annual survival (Hegemann et al., 2013(Hegemann et al., , 2015. However, current studies have only linked levels of baseline immune function to survival rates, without being able to quantify the ultimate cause of death, that is, whether mortality is directly caused by the pathogen or whether disease-related reductions in condition increase the risk of predation. In this study, we investigated prey choice by studying which type of individuals are more likely to be predated. We identified predated individuals by ring recoveries. Compared to most predation studies, which rely on postpredation observations, we followed a free-living population of jackdaws, Corvus monedula, from egg to fledgling and we measured body mass, structural size and innate immune parameters before predation took place. Consequently, we were able to compare characteristics of predated individuals in relation to the entire population of young jackdaws. In our study population, the main predator is the goshawk, Accipiter gentilis, which predates jackdaws both before and directly after fledging, and we searched for jackdaw rings in the goshawk territory to identify those jackdaw individuals that were predated. To determine the characteristics of individual prey preferentially selected by the goshawk, we measured structural body size (wing length and tarsus length) and body mass in young jackdaws of various ages over the course of five breeding seasons (2014e2018) and innate immune function (haptoglobin concentration, bacterialkilling capacity, haemagglutination and haemolysis titre) over three breeding seasons (2014,2015,2018). The unique number of the aluminium ring each individual received when ca. 12 days old made it possible to identify predated individuals before and just after fledging. Since the probability of predation by the goshawk may differ between nestlings and fledglings, we analysed these life stages separately by dividing the data into two sets: predation occurring at the nestling and at the early fledging stage. First, we examined when nestling mortality due to starvation or predation occurred and at what stages starvation and predation were the main causes of mortality in nestlings. Second, we compared structural body size and four parameters of innate baseline immune function between predated and nonpredated individuals. We did this comparison for three age classes (days of age: 12, 17 and 22) during the nestling phase as the characteristics predicting predation may differ with nestling development, particularly since nestling mobility within the nestbox changes from immobile to highly mobile during this period. Finally, we did the comparison for postfledging predation using measurements at day 29 which is just before nestling jackdaws fledge. Following the predictions and terminology of Temple (1987), we predicted that the goshawks may favour substandard individuals (Hart & Hamrin, 1990;Murray, 2002) and/or prey with signs of an immune response, which is usually connected to lower condition and mobility (Hart, 1988). Here, we specifically defined substandard individuals as individuals with smaller body size, lower body mass and lower immune function compared to the rest of the population.

Study Site
We studied jackdaws breeding in a nestbox colony at Revingehed (55 43 0 1.938 00 N, 13 26 0 22.477 00 E) in southern Sweden (see Aastrup & Hegemann, 2021) in five consecutive breeding seasons in 2014e2018. The colony is located in a small woodlot surrounded by open grasslands extensively grazed by cows and used for military exercises. The colony consisted of 57 boxes in 2014 and 2015, and 63 boxes in 2016e2018. Additionally, one nest in a natural cavity was included in 2015e2018. The inside of the nestboxes was 30.5 Â 29 cm and 32 cm (front) to 38 cm (back) high. The entrance hole was 9 cm in diameter and located 24 cm above the bottom plate of the nestbox. The extent of nest material inside each box varied greatly among boxes/pairs and ranged from a thin layer (3e5 cm) to a thick layer as high as (or even higher than) the entrance hole.

Nest Monitoring
In each year, all nestboxes were checked at 3e4-day intervals starting in mid-April to determine laying of the first egg and clutch size. Jackdaws lay three to six eggs with a laying interval of 24 h (Glutz von Blotzheim, 1985). For details on clutch size in our own colony see the Results and Table 1. A clutch was considered complete when no additional eggs had been laid in 3 or more consecutive days. Seventeen days after the first egg was laid, nestboxes were checked daily to determine the day of hatching, allowing us to determine the exact age of the nestlings. In 2016, the hatching date of nestlings from some of the nestboxes and in 2017 the hatching date for all nestboxes were estimated based on the average incubation length in our colony in 2014 and 2015.
Nestlings were ringed with an aluminium ring from the Swedish Ring Centre usually when the oldest nestlings in a brood were about 12e15 days old ( . Structural measurements (see below) and body mass were always taken during ringing, and in 2014, 2015 and 2018 also when nestlings were 29 days old. Jackdaws fledge when they are 30e35 days old (Glutz von Blotzheim, 1985; A. Hegemann, personal observations) and day 29 is the last day of the nestling phase to reliably collect data from the entire population without risking premature fledging. Hence, we took the measurements on day 29 as an indicator of postfledging condition. We know from the great number of plucking remains found within the colony that many of the predation events occur within the colony right after fledging. We have also observed that adults move away from the colony with their fledglings within a couple of days; hence we assume that most of the postfledging predation happens within the first days of fledging.
In 2018, to quantify nestling mortality (from the first hatching day) in more detail, the nestlings were counted every third day, with the first hatching set as day 0. The 3-day interval between nest visits was maintained until day 12. Afterwards, nest checks were done also on days 17, 22 and 29, when nestlings were also measured (see below). We defined nestling age as the same for all nestlings within a nestbox starting from when the first nestling hatched. Even though jackdaws have asynchronous hatching for up to 3 days, later hatched nestlings usually die within the first few days of life. In 24 nestboxes we individually marked nestlings on day 3 by painting their claws with different colour combinations using nail polish and measured survival and body mass on days 3, 6 and 9. In 75% of the nestboxes (N ¼ 18), it was always the lightest (smallest) nestlings that disappeared before day 12. Assuming that the smallest (¼youngest) nestlings were also those that disappeared in the other boxes during the first few days, age variation between siblings was greatly reduced at 12 days after the first nestling hatched. Of the 60 nestboxes with successful hatching in 2018, 38.3% (N ¼ 23) had at day 12 only nestlings with no age differences, a further 58.3% (N ¼ 35) had nestlings with only 1 day of age variation and only 3.3% (N ¼ 2) had nestlings with age variation of up to 2 days. Thus, in the limited number of broods where age variation existed at older nestling stages, the smallest nestling within a nestbox might have been the youngest. However, even if a small proportion of the individuals that were smallest in our data set were small because they were younger, this does not affect our analyses because we were interested in which characteristics make an individual more prone to predation compared to other individuals within the population (for further details on body mass variation in 12-day-old jackdaws, see Ring Recovery).

Structural Body Size and Mass
We measured maximum wing chord (henceforth wing length) to the nearest 1 mm with a wing ruler and tarsus length (except during 2016) with a calliper to the nearest 0.1 mm. Tarsus length is here defined as the length between the bended intertarsal joint and the bended toes. This measure is greater than the true tarsus (Svensson, 2005) but is subject to a lower measurement error when measured on living birds (Alatalo & Lundberg, 1986). Body mass was measured on an electronic scale to the nearest 0.1 g. From 2014 to 2017 all measurements were performed by A.H. and in 2018 by C.A. after extensive training to reduce the variation in measurements between A.H. and C.A.

Immune Assays
To measure innate immune function, nestlings were bled in 2018 on the day they were ringed, and again on day 29. In 2014 and 2015, nestlings were bled on day 29. The nestlings were bled from the brachial vein, and we collected 100e300 ml of blood into heparinized capillary tubes. The blood samples were stored on ice for up to 10 h before centrifugation with a relative centrifugal force of 1073 g for 10 min in the laboratory using a Mechanika Precyzyjna MPW 211 (Mechanika Precyzyjna, Warsaw, Poland). Plasma and red blood cells were stored separated at À50 C until analysis. To quantify each bird's baseline immune function, which is the important first line of defence (Janeway et al., 2005), we measured four parameters of innate immune function: haptoglobin concentration, bacterial-killing capacity, haemagglutination and haemolysis (see Aastrup & Hegemann, 2021). To quantify the haptoglobin concentration, a commercially available colorimetric assay was used (Tp801; Tridelta Development Ltd., Maynooth, Ireland). Haptoglobin is a protein of the acute phase response that binds free haemoglobin to prevent it from providing nutrients to pathogens. Normally this protein circulates at low concentrations in blood, but it has been shown to increase rapidly in response to infection, inflammation or trauma in avian species (Millet et al., 2007;van de Crommenacker et al., 2010). We followed the instructions of the manufacturer with minor modifications in accordance with Matson et al. (2012). For one plate of the 2018 samples, there was a mishap in the microplate reader making the 5 min measurement unusable, and a second reading was therefore performed at 7:50 min. For the three following plates a measurement was done after 5 min (according to the manufacturer's instructions) and again at 7:50 min. Values increased slightly from the reading after 5 min to the reading after 7:50 min but were highly correlated (Pearson r: 0.77; 95% confidence interval: 0.71e0.82; t 230 ¼ 18.3, P < 0.001). Hence, we used the average difference between the values from these three plates to correct the values from the plate for which we only had one measurement after 7:50 min. For the between-year analyses, corrected values were used. For the analyses including only the 2018 data set, we used the values after 7:50 min as these values did not require correction for any individual.
To test the individuals' ability to remove pathogens, an ex vivo bacterial-killing capacity assay (BKA) was performed (French & Neuman-Lee, 2012) using Escherichia coli as the pathogen and with modifications as described in Eikenaar and Hegemann (2016). The latter modifications include reducing all liquids by one-third and changing the wavelength from 300 nm to 600 nm. We determined the optimal plasma volume and E. coli concentration beforehand which showed that 3 ml of plasma and 4 ml of an E. coli concentration of 10 5 was optimal for jackdaw samples (Aastrup & Hegemann, 2021). Haptoglobin and BKA plates were analysed in a FLUOstar Omega microplate reader (BMG Labtech, Aylesbury, U.K.).
The complement and nonspecific natural antibodies (mainly IgM) were quantified using a haemagglutinationehaemolysis assay . Rabbit blood used in this assay was purchased from Envigo (Indianapolis, IN, U.S.A.; rabbit blood in Alsever's solution, S.BC-0009). All plate scans were cut into individual strips, randomized and scored blindly to age, predation status and individual (2014e2015 samples: A.H.; 2018 samples: C.A.). Based on a chicken standard run on all plates, the coefficient of variation for the among-plate variation was estimated: agglutination titre 0.18 (2014 batch), 0.17 (2015 batch) and 0.15 (2018 batch) and lysis titre 0.16 (2014 batch), 0.13 (2015 batch) and 0.18 (2018 batch). Plasma was refrozen between the different immune assays as these parameters are robust to repeated freezeethaw cycles (Hegemann et al., 2017).

Ring Recovery
To quantify which jackdaws had been predated, the ground in the centre of the local goshawk territory, including under and in their nests, and the ground under three red kite, Milvus milvus, nests, as well as parts of the jackdaw colony, were systematically searched with a metal detector (Garrett ACE 250 Metal Detector, Garrett, Garland, TX, U.S.A.) to recover aluminium rings from jackdaws that were predated during the pre-or postfledging periods. To prevent disturbance, searches with the metal detector in the goshawk and red kite territories were conducted after the end of their breeding season. The jackdaw colony was searched after all nestlings had fledged and left the colony. The goshawk nest was 3 km from the jackdaw colony and was the only known goshawk territory in the surroundings. The male goshawk (average tarsometatarsus: 81.4 mm, Reynolds et al., 1994) predated on nestling jackdaws in the nestbox by reaching through the entrance hole of the box. This was confirmed by visual observations and by the recovery of 33 rings belonging to nestlings that went missing before fledging from 2014 to 2018 and found under or close to the goshawk nest. The three red kite nests were close (100e1000 m) to the jackdaw colony. We recovered rings from six young jackdaws below red kite nests (two rings of prefledging and four rings of postfledging individuals). We do not know whether these recoveries resulted from direct predation by the red kites or by kleptoparasitism or carcasses originating from goshawk predation.
Of all predated jackdaws during the pre-and postfledging periods (see Results for details), a total of 104 rings from 2014e2018 (2014: N ¼ 21; 2015: N ¼ 25; 2016: N ¼ 21; 2017: N ¼ 17; 2018: N ¼ 20) were recovered from the sites we searched (14 in the jackdaw colony, 84 under/in or close to the goshawk nest and six under/in or close to red kite nests). A further 77 nestlings were assumed to have been predated because they weighed !100 g when they went missing from the nestbox. We assumed that parents are unable to remove dead nestlings of this size from a nestbox, and of the 35 nestlings we found dead in the nestbox after day 12, 29 had a body mass >100 g at their previous measurement (range 101.2e243.4 g). At the age of 12 days, jackdaws weigh on average 153.9 g (minimum ¼ 37.8, maximum ¼ 210.5, N ¼ 345). Those nestlings found dead in the nestbox were usually the (by far) smallest nestling of the brood during the previous measurement, making starvation the most likely cause of death. Nestlings that were confirmed predated and nestlings that we assumed to be predated did not differ in any of the structural measurements, body mass or immune parameters (all t < 0.90 and P > 0.38).
During the early nestling phase when all nestlings weighed below 100 g, nestlings often disappeared, most likely because the parents discarded dead nestlings from the box. This was confirmed by observations of small dead nestlings under the nestbox or by dead nestlings that were in the nestbox during one check but had disappeared during the next check. At such an early age, we consider predation highly unlikely because the female jackdaw is still brooding and hence a predator would also catch or harm the female, but we never found any indications of this. Furthermore, when nestlings are small, we would expect predation of the entire brood in a single event, which never occurred in the 5 years of our study.
For postfledging predation, we assumed that the jackdaw individuals for which we did not recover any ring had not been predated.

Ethical Note
This study was performed under permit 2014/M56-14 from the Malm€ o/Lund ethical committee and was in accordance with ASAB/ ABS Guidelines for the Treatment of Animals in Behavioural Research and Teaching. Our presence in the colony, during both nest checks and sampling, was reduced to a minimum and did not have any detrimental effects on the jackdaws' breeding attempts or efforts. All nestlings in a brood were removed from the nestbox for sampling at the same time and kept in a cloth bag during this time. Nestlings were kept in shade to prevent hyperthermia. Predation events were natural and not influenced or manipulated by us.

Data Analysis and Statistics
We analysed the data using R version 3.6.3 (R Core Team, 2020) applying generalized linear mixed models (GLMM; function glme from package lme4; Bates et al., 2015). For each data set (see below), we always first tested whether sex, number of nestlings or Julian date explained variation in any of our data sets by running a GLMM with binomial error structure and predation status (predated or nonpredated, excluding all starved individuals) as the dependent variable and nestbox nested in year as a random effect. None of these variables could explain any significant part of the variation in predation risk for any of the models in any of the data sets (likelihood ratio test, LRT < 2.89, P > 0.09) and hence we did not consider these variables any further.
We used all data from 2014e2018 and applied a GLMM with mortality status as the dependent factor to test whether nestlings that died due to predation during the nestling period, that is, between day 12 and day 29, differed from nonpredated/surviving conspecifics in body mass or wing length (tarsus length was not available for 2016 and hence excluded from the analyses). Since ringing was not standardized to the age of 12 days for all individuals and years (see above), for this analysis we included all birds ringed at 10e16 days (N ¼ 708). We corrected body mass and wing length for age, by running two linear models, one for age against body mass and one for age against wing length. Time of day and year were used as covariates in both models. We then extracted the residuals from the latter models by subtracting the predicted values from the observed values. We then applied the method of withinand between-subject centring (van de Pol & Wright, 2009) to these residuals to distinguish between-brood from within-brood effects. We used them as independent variables in two separate GLMM models with binomial error structure and mortality status (nonpredated coded as 1, predated coded as 0) as the dependent factor and nestbox nested in year as a random effect. Owing to a high variance inflation factor (VIF: 4.69), we were unable to include residual body mass and residual wing length in the same model. We also ran this analysis including starved individuals as a third category and these results are presented in the Appendix.
Using the detailed data from 2018, we first tested whether immune function or structural body size had an effect on predation risk during the nestling phase (i.e. between day 12 and day 29). To do so, we ran a GLMM model with binomial error structure and predation status (nonpredated coded as 1, predated coded as 2) as the dependent factor and haptoglobin concentration, bacterialkilling capacity, haemagglutination and haemolysis titre and body mass index (body mass/tarsus length) as independent variables and nestbox as a random effect. Body mass index was used as it can reflect a measure of condition, by relating mass to structural size rather than using mass independently of size (Labocha & Hayes, 2012;Labocha et al., 2014). We could not enter body mass, wing length and tarsus length into the same model simultaneously, because this resulted in a high VIF (range 4.12e7.09). Therefore, we instead ran separate models for each of these three parameters. We did not use a principal component analysis because we specifically wanted to investigate each of the three factors (body mass, wing length and tarsus length) separately, as these factors could have separate effects on predation risk.
To further zoom into the nestling period to account for the fast morphological development and hence also the change from being immobile to being highly mobile within the nestbox, we compared nestlings predated between day 12 and day 17 with those that survived this period by running a GLMM with predation status (nonpredated, predated) as the dependent variable and the immune parameters and body mass index as independent variables, with nestbox as the random effect. Then, we tested whether structural body size had an effect on predation risk between day 17 and day 22, using the structural data from day 17. Again, we ran separate models for body mass, wing length, tarsus length and body mass index. We did the same analyses also for the period between day 22 and 29; however, we only found four predated nestlings during this period, limiting the statistical power. For each time period that we analysed, the first (and last) days of the time period are always included from (and until) we visited the nestbox. We obviously cannot rule out any predation events on the same day but shortly after (or before) this visit.
To test whether predated fledglings differed from nonpredated ones, we used the data collected on day 29 in 2014, 2015 and 2018 and compared groups using a GLMM model with binomial error structure and predation status as the dependent factor and haptoglobin concentration, bacterial-killing capacity, haemagglutination and haemolysis titre, body mass, wing length and tarsus length as independent variables. In this model, the VIF was low (range 1.89e1.99), allowing inclusion of body mass, wing length and tarsus length simultaneously. Because body mass and tarsus length are highly correlated with body mass index, it resulted in high VIF values (range 116.91e1064.34); therefore, we ran a separate model for body mass index. In all models, nestbox nested in year was used as a random effect.
We always started with the full model and removed nonsignificant terms (P > 0.05) based on backwards elimination and the drop1 function in R. Odds ratios were extracted from models using the package broom.mixed (Bolker & Robinson, 2022). All significance tests are chi-square tests.

Hatching and Mortality Data
For 3 years (2014,2015,2018), we have detailed data on hatching success (Table 1). In these years, the clutch size averaged 4.8 eggs of which 89.5% hatched. Further details can be found in Table 1.
In 2018, we recorded early nestling mortality (0e12 days of age) in detail. The highest mortality percentage occurred between day 0 and day 3 followed by the period between day 9 and day 12 ( Table 2). All other age classes had an even mortality rate. Around half (52.9%) of all hatchlings fledged; of those that fledged 52.4% were males (N ¼ 77) and 47.6% were females (N ¼ 70). Of the nestlings in 2018 that died before fledging, 72.5% (N ¼ 95) died from starvation and 27.5% (N ¼ 36) were predated (days 12e17: N ¼ 19; days 17e22: N ¼ 7; days 22e29: N ¼ 4; six nestlings were predated but they were not measured: days 12e17: N ¼ 3; days 17e22: N ¼ 1; days 22e29: N ¼ 1; for one it is unclear when exactly it was predated). Of the 2018 nestlings that died from starvation, 91.5% (N ¼ 87) did so before day 12. In the 5 study years, we found no signs of predation before day 12.
In all years, 181 jackdaws were predated, of which 60.8% (N ¼ 110) were predated before fledging and 39.2% (N ¼ 71) postfledging, and 696 jackdaws fledged. For further details for predation and fledging in each year see Table 3.

Predation During the Nestling Phase: All Years
Using body mass and the structural data collected in 2014e2018 (excluding starved individuals; see the Appendix for the analysis including starved nestlings), we found that between ringing (age range 10e16 days) and day 29, both residual body mass (LRT: c 2 1 ¼ 9.2, P ¼ 0.002, odds ratio, OR 1.02, 95% confidence interval, CI [1.01, 1.04]) and residual wing length were negatively related to predation risk within broods (LRT: c 2 1 ¼ 8.65, P ¼ 0.003, OR 1.05, 95% CI [1.02, 1.09]), with smaller individuals more likely to be predated (Fig. 1). There was no effect of residual body mass (LRT:  N total is the total number of nestlings in each age class. N dead is the total number of nestlings that died between two time points, e.g. N dead on day 3 is the number of nestlings that died or went missing between day 3 and day 6. Day 29 is not represented as all mortality for this age class occurred postfledging.
Between day 17 and day 22, we found that both measures of structural body size as well as body mass to be highly negatively correlated with predation risk (wing length: predated: mean ± -SE ¼ 94 ± 4.88 mm; nonpredated: mean ± SE ¼ 111 ± 0.66 mm; c 2 1 ¼ 1.09, P ¼ 0.296, estimate ± SE ¼ 0.50 ± 0.48) did not affect the risk of being predated between day 22 and day 29.

Postfledging Predation
In 2014, 2015 and 2018, a total of 43 newly fledged jackdaws (Table 3) were predated after leaving the nestbox. In terms of this Table 3  Predation data of ringed jackdaws for all years separately   2014  2015  2016  2017  2018   Total predated  31  41  44  23  42  Predated prefledging  18  23  26  13  30  Predated postfledging  13  18  18  10  12  Survived to fledging  134  102  170  144  146 In 2018, an additional six individuals were predated prefledging, but they were not ringed or measured and therefore not included in the table (see Results). postfledging predation, there was no significant predictive effect of body mass, tarsus length or body mass index (Table 4) measured at the age of 29 days. However, wing length was negatively related to predation risk (Table 4), and it was the fledglings with shorter wings that were at a higher risk of being predated (Fig. 3). None of the immune parameters measured at day 29 and thus shortly before fledging were correlated with predation risk at the early fledging stage (all P > 0.197; Table 4, Fig. 3).

DISCUSSION
The aim of this study was to investigate prey selection in a top predator (goshawk) by following a free-living population of prey (jackdaws). The results suggest that relatively small nestlings with reduced complement activity and low haptoglobin concentrations have the highest risk of being predated from the nestbox. In the early postfledging period, wing length was the only significant predictor of predation risk, where fledglings with shorter wings experienced a higher risk of predation.
In the first 12 days of life, nestling mortality is most likely solely due to starvation. However, after this first nestling period, predation becomes the main cause of mortality. We had no single event of complete simultaneous nest predation (i.e. all nestlings predated in a single event) in the 5 years of our study as would have been predicted if pine marten, Martes martes, had been the predator. Asynchronous hatching likely explains why starvation mortality is highest during the first 12 days of life. Within a brood there is hatching asynchrony of up to 3 days and hence the first hatchlings within a brood have a head start during which they can grow rapidly (Arnold & Griffiths, 2003;Gibbons, 1987). This is likely to affect the runts early in life, explaining why we found starvation mortality to be highest between hatching and day 3.
Compared to other time periods during the nestling stage, the highest predation rate was between day 12 and day 17. During this  period, jackdaw nestlings are still fairly immobile and hence easy to handle and remove from the nestbox. As the nestlings grow, they will not only grow longer wings which makes them more difficult to remove from the nestbox through the entrance hole, but will also begin to move around more within the box and may thus try to escape from a possible attack which will likely make it harder for the goshawk to capture them within the box. Goshawks are opportunistic predators that predate on a variety of different prey based on availability (Opdam et al., 1977;Tornberg, 1997;Tornberg et al., 2006). Over the course of the breeding season, the proportion of different prey items in the goshawk diet changes (Opdam et al., 1977). That we did not see any predation before day 12 may be related not only to the size of the jackdaw nestlings (and hence the profitability as prey for the goshawk), but also to the brooding behaviour of jackdaw females. Female jackdaws brood for roughly 11e12 days (Glutz von Blotzheim, 1985) and after that participate in foraging trips. Once both parents begin feeding the nestlings, the nest will be left unattended for longer periods and the begging calls of the unattended nestlings are likely to attract the goshawk to predate the nestlings (see also McDonald et al., 2009).
Throughout the nestling phase, smaller nestlings were at a higher risk of predation, possibly because they were hungrier and therefore sat closest to the nest entrance, which would increase their chance of obtaining food from their parents. Thus, they might be more exposed to nest predation than their larger siblings which might wait further away from the nestbox entrance. This is supported by our analyses showing that size predicted predation risk at the within-brood level but not at the between-brood level, meaning that it is the smallest nestlings within a brood that run the highest risk of predation independent of how their size relates to that of nestlings in other broods. Hence, predation during the nestling phase might not be active selection by the goshawk, but rather a result of within-brood competition forcing smaller, weaker nestlings to take more risks to increase their chances of obtaining food from their parents. Whether the nestlings are smaller because they are the youngest or because of poorer development remains unknown. Yet the outcome of the predation patterns is still important from a fitness perspective.
The few studies that have investigated predation risk in nestlings have found that nestlings mount an immune response when the threat of predation becomes severe (Roncalli et al., 2018;Tilgar et al., 2010). These studies, however, investigated immune function and the immune response during a simulated predation attempt, while we looked at the baseline immune function before any predation event had occurred. We found that predated nestlings had lower lysis and haptoglobin concentrations prior to the predation event. Nestlings' innate immune function develops substantially with age and is not yet fully developed at the time of fledging (Aastrup & Hegemann, 2021;Dowling & Levy, 2014;Mauck et al., 2005;Palacios et al., 2009). As predated nestlings had a lower body mass index and were structurally smaller, they might have been less developed than their nonpredated siblings. Whether this is caused by small (1-day) differences in age (i.e. resulting from hatching asynchrony) or other factors (e.g. food shortage, lower genetic quality) remains to be investigated, but in either case it signals that less developed nestlings are more likely to be predated.
The fact that predated nestlings had lower haptoglobin concentration than nonpredated nestlings was contrary to our prediction. We expected predated nestlings to have higher haptoglobin concentration than nonpredated nestlings, indicating either an ongoing inflammation or infection (Quaye, 2008). This suggests that acute sickness does not play a role in nestling predation although we acknowledge that a positive relationship might be difficult to detect due to the rareness of sampling sick individuals at the peak of infection in free-living animals (but see Hegemann et al., 2018). Instead, our results lend further support to the hypothesis that the predated nestlings were less developed as haptoglobin levels usually develop early in life (Aastrup & Hegemann, 2021).
Fledgling body mass is often an important predictor of postfledging survival with heavier fledglings more likely to survive  (Møller et al., 2010;Ronget et al., 2018;Suedkamp Wells et al., 2007;Tinbergen & Boerlijst, 1990). However, we did not find any indication that body mass influenced postfledging survival in jackdaws which is in accordance with a study by Boonekamp et al. (2014) on the same species (but see Verhulst & Salomons, 2004 for a study on jackdaws reporting a positive association between survival and body mass). Most previous studies finding a positive relationship (Both et al., 1999;Losdat et al., 2013;Monr os et al., 2002;Naef-Daenzer et al., 2001;Perrins & McCleery, 2001) included smaller passerines where a higher body mass can protect against starvation and conspecific competition (Vermeulen et al., 2016). However, in larger species body mass at fledging often does not predict subsequent survival (e.g. Kersten & Brenninkmeijer, 1995;Stienen & Brenninkmeijer, 2002;Arizaga et al., 2015;Nebel et al., 2021;this study). Instead, predation may account for most of the early postfledging mortality in our jackdaws.
We did, however, find a negative correlation between wing length and predation risk in newly fledged jackdaws. Whether this difference in wing length is based on suboptimal development or simply rooted in the fact that birds with shorter wings are those that hatched 1 day later remains unknown. However, even if the reason for the differences remains unclear, the outcome in terms of higher predation on less developed nestlings will have an important impact on fitness patterns. The individual variation in wing length might affect the flight capacity of fledglings (Martin et al., 2018) and hence their ability to escape predators. While we did find a similar association between wing length and predation risk in the nest, postfledging wing length was the only significant predictor of predation risk. This indicates that the ability to fly might be a more important indicator for predation risk in fledglings than their overall development. For an aerial hunter like the goshawk, predating fledglings with relatively poor flight skills might make them easy prey to capture compared to fledglings with longer wings. Our results therefore suggest that the goshawk does not randomly select prey, but actively choose prey that appear easier to catch (see also Kenward, 1978;see table 3 in Temple, 1987). Thus, the predator tries to maximize the net energetic gain from an attack, and in the present case preferentially predates fledglings with shorter wings.
During the postfledging period, we found that there were no differences in mean values for any of the immune indices between predated and nonpredated fledglings. Thus, innate baseline immune function did not appear to be an important determinant of predation risk in the first few days after fledging. An experiment on collared doves, Streptopelia decaocto, indicated that immunechallenged nestlings had reduced survival and a higher risk of becoming predated during the postfledging period (Eraud et al., 2009). We had predicted that especially haptoglobin, an indicator of inflammation (Quaye, 2008), might be positively related to predation risk. However, we did not find such a relationship, suggesting that acute sickness does not play a role in predation in nestling or fledgling jackdaws in our correlational study; we acknowledge, however, that a positive correlation between haptoglobin values during peak infection and predation risk might be hard to find in free-living animals (see above).

Conclusions and Perspectives
A common method for determining the immune function of prey has been to use secondary indicators (e.g. spleen size in birds, Møller & Erritzøe, 2000; white tuft on the tail of rabbits, Oryctolagus cuniculus, Penteriani et al., 2008). In these studies, substandard individuals were more likely to be predated, suggesting that predated individuals also had lower immune function.
However, using these indirect traits might only provide a general view of the overall quality of the prey over a long period rather than immune function or a physiological mechanism per se. Morphometric traits also integrate quality over longer timeframes while measuring immune function directly (as in this study) reflects shorter time intervals that are more closely connected to the predation events (Hegemann et al., 2012). Here, we have provided the first evidence that direct measures of innate immune function (i.e. lysis titre and haptoglobin concentration) are also related to predation in nestling jackdaws. However, given that predated individuals were also the smallest, we cannot rule out the possibility that these nestlings were of generally lower quality or of a slightly younger age, rather than indicating that innate immune function is a key physiological mechanism increasing predation risk via, for example, behavioural changes that modulate vulnerability to predators. Furthermore, we found no evidence that direct measures of immune function predicted predation risk in fledglings during the first few days after leaving the nestbox. It remains to be investigated though whether immune challenges with their associated behavioural effects (in particular lethargy) can alter the predation risk.
Previously, prey choice studies often did not include nestlings, or used a combined data set of juveniles and adults, where juveniles were often considered substandard individuals in comparison to adults (Genovart et al., 2010;Temple, 1987). Here, we have shown that even within juveniles there are differences in predation risk, and that substandard individuals of the juvenile population are at a higher risk of predation. Repeating the study with nestlings, juveniles and adults, as well as experimental immune challenges, would provide a better understanding of which factors govern predation risk at different life stages, and would help tease apart the effects of generally low body condition from those of other physiological mechanisms.

Data Availability
Data are available upon request.

Declaration of Interest
None.