Brood parasites usually reduce their host's breeding success, resulting in strong selection for the evolution of host defences. Intriguingly, some host individuals/populations show no defence against parasitism, which has been explained within the frame of three different evolutionary hypotheses. One of these hypotheses posits that intermediate levels of defence at the population level may result from nonrandom distribution of parasitism among host individuals (i.e. structured parasitism). Empirical evidence for structured brood parasitism is, however, lacking for hosts of European cuckoos due to the absence of long-term studies.
Here, we seek to identify the patterns of structured parasitism by studying great spotted cuckoo parasitism on individual magpie hosts over five breeding seasons. We also aim to identify whether individual characteristics of female magpies and/or their territories were related to the status of repeated parasitism.
We found that 28·3% of the females in our population consistently escaped from cuckoo parasitism. Only 11·3% of females were always parasitized, and the remaining 60·4% changed their parasitism status. The percentage of females that maintained their status of parasitism (i.e. either parasitized or nonparasitized) between consecutive years varied over the study. Females that never suffered cuckoo parasitism built bigger nests than parasitized females at the beginning of the breeding season and smaller nests than those of parasitized females later in the season. Nonparasitized females also moved little from year to year and preferred areas with different characteristics over the course of the breeding season than parasitized females. Overall, females escaping from cuckoo parasitism reared twice as many chicks per year than those that were parasitized.
In conclusion, our study reveals for first time the existence of a structured pattern of cuckoo parasitism based on phenotypic characteristics of individual hosts and of their territories.
Parasites exert major selective pressures on their hosts. Therefore, natural selection is expected to favour individual hosts that effectively counteract the effect of parasitism. Understanding the spatial and temporal dynamic of parasite–host interactions and identifying the factors affecting the evolution of host defences and parasite virulence are major challenges in current evolutionary ecology (Thompson 2006). Theoretical models aiming to investigate co-evolution of parasites and their hosts have largely assumed that all host individuals in a population have the same chance of being infected (e.g. van Baalen 1998; Gandon, Agnew & Michalakis 2002; Best, White & Boots 2009). Most natural populations, however, experience some degree of social and/or spatial structure (Thompson 2006), and therefore, it is expected that one host will be more likely infected if its close neighbours or individuals within its social group are infected. Only recently, space and population viscosity were incorporated into theoretical models to approach the question of how host within-population spatial and social structure may affect the evolution of host resistance (e.g. Best et al. 2011; Débarre et al. 2012). Yet empirical evidence supporting structured parasitism and its effect on parasite–host co-evolution in natural populations is scant (see, however, Kerr et al. 2006; Boots & Mealor 2007; Martínez-Padilla et al. 2012).
Interspecific avian brood parasitism is a particular form of parasitism in which a species, the parasite, lays its eggs in the nest of another species, the host, which carries out all the parental care, from incubating parasite eggs to feeding parasite chicks. Brood parasites usually reduce their host's breeding success (Davies 2000; Payne 2005). For instance, parasite eggs can hatch considerably earlier than hosts ones, because of their shorter incubation period, and parasite hatchlings may either remove all host eggs and nestlings from the nest (Davies 2000), or outcompete host nest siblings in their competition for food (Soler & Soler 2000). There is a large body of evidence showing that hosts can evolve behavioural mechanisms to respond to these selective pressures (Rothstein 1990; Davies 2000). These are nest defence against adult parasites (e.g. Røskaft et al. 2002; Davies et al. 2003; Welbergen & Davies 2009) and/or discrimination and removal of parasitic eggs (e.g. Brooke & Davies 1988; Soler & Møller 1990; Avilés et al. 2010; Spottiswoode & Stevens 2010) and/or nestlings (e.g. Langmore, Hunt & Kilner 2003; Grim 2007; Sato et al. 2010). On the other hand, the evolution of host defences can select for further counter-defences in the parasite, such as highly mimetic parasitic eggs to evade host detection (Brooke & Davies 1988), causing a co-evolutionary arms race (Davies 2000).
Although brood parasites exert strong selection for the evolution of host defences, many hosts display a striking lack of antiparasite defences. For instance, British dunnocks (Prunella modularis) do not discriminate common cuckoo (Cuculus canorus) eggs, despite their very different appearance and that dunnocks were affected by cuckoo parasitism for the last 600 years (Davies & Brooke 1989). Many other hosts of different cuckoos and cowbird species exhibit a noticeable absence of defensive behaviours (Davies 2000; Payne 2005). Understanding why some hosts accept avian brood parasitism despite its costs remains a challenge in this co-evolutionary scenario (Rothstein & Robinson 1998; Payne 2005).
Three main evolutionary hypotheses can explain the apparent paradox of absence of antiparasite defences in the face of costly parasitism. The evolutionary lag hypothesis attributes the absence of defence to an evolutionary lag in the development of the defensive mechanisms by hosts either due to an absence of the genetic variants needed to evolve the defence, or because there has been not enough time for the defence to spread out (Rothstein 1975). The evolutionary equilibrium hypothesis suggests that, given cognitive and physiological constraints on defence, the advantage of bearing it would be context dependent. Accordingly, hosts would accept parasitism when the costs of avoiding it will exceed the costs of accepting it (Rohwer & Spaw 1988; Lotem, Nakamura & Zahavi 1992; Avilés, Rutila & Møller 2005; Krüger 2011). Finally, limited transmission of genetic variants due to spatially structured parasitism between and within populations may also lead to apparent maladaptive absence of defences at the population level (Soler et al. 1999; Røskaft et al. 2002; Hauber, Yeh & Roberts 2004; Hoover, Yasukawa & Hauber 2006). Indeed, theoretical models predicted that accepter and rejecter phenotypes may coexist within a population as a consequence of nonrandom distribution of parasitism among host individuals (‘repeated parasitism’ sensu Hauber, Yeh & Roberts 2004). Although empirical evidence in support of the evolutionary lag and the equilibrium hypotheses has been reported for a wide variety of brood parasitic systems (reviewed in Winfree 1999; Krüger 2007), empirical evidence for limited transmission of parasitism only exists for hosts of the brown-headed cowbird (Molothrus ater). These studies showed a higher than expected probability of parasitism for previously parasitized individuals between first and second breeding attempts within the same year and in consecutive breeding years (Hauber, Yeh & Roberts 2004; Hoover, Yasukawa & Hauber 2006; Hoover & Hauber 2007). Brown-headed cowbird parasitism inflicts comparatively lower costs on its hosts than the parasitic cuckoo species in Europe (Payne 2005). Therefore, more studies with different brood parasite–host systems in which parasites exert strong selection on their hosts are needed before general trends about the occurrence of repeated parasitism and its role on the evolution of host defences can be disclosed.
Here, we first seek to identify long-term patterns of repeated parasitism by great spotted cuckoos (Clamator glandarius) on female magpies (Pica pica) from 2007 to 2011 in a scenario of increased parasitism pressure. Hitherto, repeated parasitism was identified by studying individual host and/or territory exposure to cowbird brown-headed parasitism between pairs of consecutive years (Hauber, Yeh & Roberts 2004; Hoover, Yasukawa & Hauber 2006). Identifying individual hosts' exposure to parasitism over long-term data frames will allow ascertainment of the extent to which parasitism status of individuals and the emerging patterns of repeated parasitism in the population are affected by parasitism levels.
Secondly, we aim to identify what individual characteristics of female magpies and/or of the territories they hold across their life relate to the probability of suffering repeated parasitism. There is overwhelming empirical evidence that a variety of ecological and host phenotypic traits influence the exposure to brood parasites over the course of a breeding season (reviewed in Cherry, Bennett & Moskat 2007; Parejo & Avilés 2007). However, previous work has failed to identify phenotypic differences between individuals suffering repeated parasitism and those escaping from brown-headed cowbirds (Hoover, Yasukawa & Hauber 2006; Hoover & Hauber 2007). These studies targeted morphological characteristics and age of hosts but did not study the differences in explicit aspects known to reveal host quality (e.g. Parejo & Avilés 2007). Here, we will focus on studying repeated parasitism in relation to phenotypic and reproductive magpie traits for which empirical and experimental evidence supports a link with individual quality (Birkhead 1991). In addition, we also examine the differences in nest size in magpie hosts in relation to repeated parasitism, as experimental evidence supports the observation that nest size is a post-mating sexually selected signal revealing parental abilities in magpies (de Neve et al. 2004) and may be used as a cue to choose profitable hosts by cuckoos (Soler et al. 1995a).
Thirdly, we examine the relationship between repeated parasitism and level of defence in magpie hosts. Theoretical and empirical work suggests that limited transmission of parasitism may greatly impact the spatial dynamics of parasite–host interactions (Lenormand 2002), and in a scenario of brood parasitism, theoretical models have shown that structured parasitism may affect the dynamic of cuckoo–host interaction at the population level by retarding the evolution of host defences (Grim 2002; Hauber, Yeh & Roberts 2004; Røskaft et al. 2006; Avilés & Parejo 2011). Limited horizontal transmission of brood parasitism is expected to diminish the fitness benefits of egg rejection and to shape the ontogeny of hosts' recognition systems (Hauber, Yeh & Roberts 2004; Hoover, Yasukawa & Hauber 2006). An empirical premise remaining to be tested in this theoretical scenario is that host phenotypes differing in their exposure to parasitism also differed in their baseline levels of defences against brood parasites. Here, we provide a first empirical test of this assumption by studying the relationship between long-term exposure to great spotted cuckoo parasitism and level of defences in individual magpies. We will specifically target two known key components of magpie defence against cuckoo parasitism, namely egg discrimination (Soler & Møller 1990; Soler et al. 1999) and intraclutch variation in egg appearance (Soler, Soler & Møller 2000; Avilés et al. 2004).
Finally, we will examine for the first time the effect of repeated parasitism on fitness and dispersal of magpie hosts. So far, empirical studies have demonstrated deleterious effects of cuckoo parasites on host reproduction (reviewed in Davies 2000; Payne 2005), and found some support for a link between host dispersal and brood parasitism (e.g. Hoover 2003; Sedgwick 2004; Molina-Morales et al. 2012), over the course of a breeding season. However, previous studies disregarded analysing individual long-term exposure to cuckoo parasitism in relation to host productivity and dispersal, which may hamper a realistic assessment of the strength of cuckoo selection on host defences.
Materials and methods
Study Area and System
The study was conducted in La Calahorra (37°10′N, 3°03′W, Hoya de Guadix, Southern Spain) during the years 2007–2011. This is a patchy area where groves of almond trees (Prunus dulcis), in which magpies preferentially build their nests, are very common. Magpies are territorial, sedentary and socially monogamous long-lived passerines (Birkhead 1991).
In our study area, magpies lay one clutch during April–May and are the main host of the great spotted cuckoo. Cuckoo parasitism severely reduces magpie reproductive success through early hatching and effective competition for parental food delivery to cuckoo nestlings (e.g. Soler, Martínez & Soler 1996; Soler, Soler & Martínez 1997), and has selected for host recognition and rejection of cuckoo eggs, which in turn has selected for punitive cuckoo behaviours promoting parasitism acceptance (Soler et al. 1995b). Thus, the interaction between magpies and great spotted cuckoos can be regarded as an example of co-evolution (Soler & Soler 2000).
Temporal Variation in Prevalence of Great Spotted Cuckoo Parasitism
The percentage of parasitized nests in our population (i.e. parasitism rate) varied between years (15·9% in 2007, 25·4% in 2008, 65·6% in 2009, 50·7% in 2010, 55·8% in 2011) and increased through the breeding season affecting in average almost 70% of magpie nest in the last 10 days of the laying season (Spearman's correlation, rp = 0·88, P =0·01, N = 6, Fig. 1).
Individual Marking and Monitoring
Some of the adult magpies were captured and colour ringed and thus monitored by observation. Monitoring of nonringed females was based in parentage analyses (details of captures, molecular methods and parentage analyses can be found in the study by Molina-Morales et al. 2012). Briefly, we assigned particular breeding attempts in different years to the same female when the nestlings in those broods were found to be all full siblings to each other. Also, we could assign breeding attempts to marked females using paternity analyses, so that females could be matched to their nest in a given year even if they had not been marked in that year but later on, and thus observations were not available (Molina-Morales et al. 2012).
Nest Monitoring and Individual Characteristics of Female Magpie Hosts and of the Territories
Magpie nests were monitored from 1 March to the beginning of July each breeding season. Nests were found by careful inspection of all trees in the area, and GPS positioned. Each nest was observed with a telescope from a hide around 100 m away during nest building in order to assign marked birds to each nesting attempt. Nests were visited at 5-day intervals, although during egg laying and hatching the nests were visited every 2–3 days to check whether the nest was parasitized by great spotted cuckoos and to record all required data. Nests were categorized as parasitized if at least one cuckoo egg was detected in the nest. Replacement clutches were not included in this study in order to obtain unbiased estimates for individual characteristics of female magpies known to decline with season (e.g. clutch size, nest productivity, nest size). For each breeding event, we recorded the following data:
Laying date, estimated as the number of days from the first of April, clutch size and number of fledglings.
Average egg volume, as the mean value of the volume of all magpie eggs in a clutch. The volume of each individual egg was estimated as 4/3(Π × a × b2)/1000 (in litres), where a is the largest radius of the ellipsoid egg and b is half of the egg width measured with a calliper (precision 1 mm).
Nest volume. The size of nests was estimated using a measuring tape (precision 1 cm); we measured height and width and calculated nest volume using the same ellipsoid formula used to estimate egg volume.
Intraclutch variation in egg appearance. We took a picture of each clutch using a CANON 350D digital camera (Canon Inc., Tokyo, Japan). All photographs were taken under standardized light conditions on a Kodak grey card. Intraclutch variation was estimated on these photographs following the scale of Øien, Moksnes & Røskaft (1995). Briefly, this method attributes increasing levels of variation in egg appearance within a clutch on an ordinal scale (1, no variation, to 5, all the eggs were different from one another) based on human perception. Seven experienced observers scored intraclutch variation for all clutches (103 clutches). The different assessments of a clutch were moderately consistent (repeatability = 0·428, F103,624 = 6·23, P <0·001), thus justifying the use of the mean values attained for the seven observers as an estimate of the degree of intraclutch variation in all further analyses.
Response to mimetic model eggs. We tested magpie responses to mimetic model eggs to classify females as acceptors or rejecters (e.g. Soler & Møller 1990; Soler et al. 1999). We introduced one mimetic model egg during magpie egg laying and revisited the nest after 6–7 days. Previous work in our magpie population has shown that 75% of all rejection of artificial models occurs in the first 24 h after parasitism and that after 72 h all eggs have been rejected (Avilés et al. 2004). The response was regarded as a rejection if the model egg disappeared from the nest or acceptance if the model egg was incubated with the host's clutch.
Host density. For each nest, we measured the distance to the two nearer conspecific nests (nearest neighbour distance) and used the average of both distances as an estimate of magpie density in the area surrounding each nest.
Breeding dispersal distance. Following Molina-Morales et al. (2012), this was calculated as the shortest distance in metres between two nests occupied by the same bird in consecutive years.
Spatial information. We used GIS software ArcGIS 9.3 version (ESRI 2008) to obtain environmental data based on aerial photographs and Vegetation Cover and Land Use Databases for the Province of Granada that were freely available from Junta de Andalucía (Junta de Andalucía 2003). We recorded the following information regarding the spatial situation of each nest: (i) distance in metres to the closest pine forests as an estimate of distance to great spotted cuckoo's feeding site as great spotted cuckoos feed almost exclusively on Pine Processionary, Thaumetopoea pityocampa (Soler 2003), (ii) distance to the closest track, (iii) distance to dry riverbeds, open habitats, (iv) distance to the motorway, (v) distance to the nearest village, (vi) percentage of wooded surface within 100 m around the nest, (vii) percentage of herbaceous crop within 100 m around the nest and (viii) percentage of surface occupied by almond trees within 100 m around the nest.
Analyses were performed using Statistica 7.0 (Statsoft Inc., Tulsa, OK, USA). We characterized each magpie female regarding its parasitism status along all her known breeding attempts as never parasitized, sometimes parasitized or always parasitized. Firstly, we tested for differences in number of years that a magpie female was sampled in relation to status of parasitism as a categorical predictor using a Poisson generalized linear model which allowed us to know whether the probability of classifying a given female in relation to parasitism status depended on the number of years it was monitored.
In addition, we ran a Monte Carlo simulation based on the null model that, within each year, parasitism would occur randomly among the individuals. We wrote a program in r to perform the following algorithm:
Set a counter (‘count’) to 0.
Within each year, randomly permute the parasitic status (nonparasitized = 0, parasitized = 1) among the magpies.
Once step 2 is carried out for each year, recalculate the random frequencies of birds that are ‘always’, ‘sometimes’ or ‘never’ parasitized.
Compare the randomly calculated frequencies of ‘always’ and ‘never’-parasitized females with the observed ones. If both random frequencies are identical to or smaller than, the observed frequencies, add one unit to the counter: counti + 1 = counti + 1.
Repeat 1000 times steps 2 through 4.
Calculate the final P-value as P = count (1/1000). If P < 0·05, the observed frequencies will be significantly different from what would be expected by chance alone.
Because the parasitic status among females is reshuffled within each year, the above algorithm is neither affected by the number of years that a bird was observed nor by the interannual variability in parasitic frequencies.
Secondly, for each identified female, we determined whether it maintained its parasitism status (i.e. if it was either parasitized or nonparasitized) or if its status was reversed (if it passed either from parasitized to nonparasitized or from nonparasitized to parasitized) in two consecutive years. This yielded a dichotomous variable revealing parasitism status (changed/maintained). We then determined whether the frequency of females repeating parasitism status in consecutive years differed from that the other two consecutive years by using contingency analyses (Zar 1996). Given that the parasitism rate steadily increased over the course of the study, this approach allowed us to detect how fluctuations in prevalence of parasitism may affect female status of parasitism by comparing the frequency of changes in status between consecutive years with similarly low parasitism (i.e. 2007–2008), with that in consecutive years with similarly high parasitism (i.e. 2009–2010), or with that in consecutive years after a sudden increase in parasitism (i.e. 2008–2009).
Aiming to relate long-term parasitism status with the phenotypic and spatial variables previously described, and because we have several breeding attempts (2–5) for each female, we calculated the average value of each phenotypic and spatial variable for each female. This is justified because repeatability analysis revealed that all these variables were repeatable (r > 0·25, F53,74 > 2·16, P >0·0001), except intraclutch variation in egg appearance (F1,53 = 1·48, P =0·09). When calculating mean clutch size and number of fledglings of females sometimes parasitized, we excluded those years when females were parasitized, because it is well-established that parasitized magpie nests have smaller clutch and brood sizes than unparasitized ones (Soler, Martínez & Soler 1996). In this way, we can test for the first time whether the breeding outcome of never-parasitized females differed from that of females that are parasitized but eventually escaped parasitism. However, because we were also interested in the long-term fitness consequences of repeated parasitism in a second analysis, we also averaged the number of fledglings a female produced during her life including years in which it was parasitized for females that sometimes were parasitized. In this way, we can contrast long-term female productivity in relation to repeated parasitism. Finally, because we tested each female several times for model egg rejection, females were classified as rejecters if they rejected the model egg at least one time, or acceptors if they always accepted the model egg.
Mean laying date, clutch size, number of fledglings and volume of eggs of each female followed a normal distribution (Shapiro–Wilk tests: W >0·95, P >0·06). Body condition, tarsus length and bill length were approximately normally distributed (Shapiro–Wilk tests: W >0·94, P >0·29). Mean nest volume and mean breeding dispersal distance were log-transformed to fit a normal distribution (Shapiro–Wilk tests: W >0·96, P >0·13). Regarding spatial information, we first calculated the mean value of each variable for the different nests of each female. After that, we used principal components analysis (PCA) in order to transform several correlated variables into a few orthogonal variables (the principal components). We obtained three PCA factors with eigenvalues > 1. The first principal component (PC1) was negatively related to percentage of herbaceous crop and positively to percentage of woody surface and percentage of surface occupied by almond trees, and thus represents a gradation in the area surrounding the nest from clear to wooded spaces (Table 1). PC2 was positively related to distance to tracks and negatively to distance to villages and to dry riverbeds. PC3 was positively related to distance to parasites feeding places (Table 1).
Table 1. Results of the principal components analysis on spatial variables
Factor loadings for the first three axes of a principal component analysis of spatial variables. Loadings in boldface indicate the most important factors (score > |0·60|). Percentage of variance explained by each axis is also shown.
Distance to the closest pine forests
Distance to closest track
Distance to motorway
Distance to nearest village
Distance to dry riverbed
Percentage of herb growing
Percentage of tree cover
% of variance
All the females were divided into two groups. The first group included nonparasitized females, and the second group was compounded by always and sometimes parasitized females. We did this because of the low number of cases for always parasitized females (only six cases), and because our aim was to identify the factors that may explain why some females systematically escaped from cuckoo parasitism and others did not. General linear models were used to look for differences among individuals parasitized or not in phenotypic and spatial variables. Previous work has shown that laying date is a correlate of individual quality in magpies (Soler et al. 1995a). Therefore, aiming to account for individual variation in quality in our analyses, we entered laying date and its interaction with parasitism status in all the models except in the analysis of breeding dispersal distances, because dispersal movements occur after the breeding season. Finally, we used a logistic regression model to test whether rejection of parasite eggs in magpies (i.e. rejection vs. acceptance) was explained by parasitism status while accounting for laying date.
Parasitism Status of Females in Different Breeding Attempts
We assessed parasitism status for 53 females that were monitored on average for 2·41 years (range 2–5 years; standard deviation: 0·718). The number of years that a female was monitored did not significantly differ between females always parasitized, sometimes parasitized and never parasitized ( = 0·193, P =0·908), suggesting that differences in parasitism status were not due to the number of years a given female was monitored.
Interestingly, 15 of 53 females of our population (28·3%) were never parasitized. Of the remaining 38 females, only six (i.e. 11·3%) were always parasitized and 32 (i.e. 60·4%) switched their parasitism status. The Monte Carlo simulation analysis showed that, in 90% of cases, the observed frequency of always, never and sometimes parasitized females was different from that expected by chance (P =0·10).
The percentage of females that maintained their status of parasitism (i.e. either parasitized or nonparasitized) between consecutive years varied over the course of the study. Up to 77·8% of females maintained their parasitism status from 2007 to 2008, when parasitism rate was low in the population. This percentage doubled that of females maintaining parasitism status between 2008 and 2009 (37·5%; = 5·67, P =0·017), when cuckoo parasitism markedly increased in the population. However, the proportion of magpie females maintaining their parasitism status between pairs of years with similarly low (i.e. 77·8%, 2007–2008) and high (i.e. 56·3%, 2009–2010) parasitism did not significantly differ ( = 1·79, P =0·18). Similarly, nonsignificant differences existed in the proportion of females maintaining parasitism status between 2008 and 2009 and between 2009 and 2010 ( = 1·13, P =0·28).
Parasitism Status and Phenotypic and Reproductive Traits of Magpie Hosts
Among the analysed traits, only the relationship between nest size and laying date was different for parasitized and never-parasitized females (Table 2, Appendix S1, Supporting Information). The size of the nests modestly decreased across the season, although nonsignificantly, for females escaping from cuckoo parasitism (R =0·345, F13 = 1·754, P =0·207), whereas among the parasitized females those breeding later in the season had larger nests than those breeding at the beginning of the season (R =0·357, F36 = 5·259, P =0·028, Fig. 2).
Table 2. Relationships between long-term parasitism status (i.e. either parasitized sometimes or always vs. never parasitized) and reproductive and defensive traits of magpie hosts
Analyses are general linear models and logistic regressions with reproductive and defensive variables of magpies as dependent variables. Laying date (1 = 1 April) was introduced as a covariate (see methods).
Laying date × parasitism
Number of fledglings
Laying date × parasitism
Laying date × parasitism
Laying date × parasitism
Laying date × parasitism
Laying date × parasitism
Laying date × parasitism
Parasitism(never, sometimes, always)
Number of breeding attempts
PC1 habitat score
Laying date × parasitism
Parasitism Status and Host Defensive Traits
Parasitized and nonparasitized females did not differ in intraclutch variation in egg appearance (Table 2) nor in their capacity to reject model eggs (Table 2).
Parasitism Status and Host Dispersal and Density
Never-parasitized and parasitized females differed in their average breeding dispersal distances (Table 2). Magpies escaping from cuckoo parasitism moved less than those suffering cuckoo parasitism (never parasitized: mean = 151·60, SD = 143·70, n = 15; parasitized: mean = 327·60, SD = 521·96, n = 38).
Long-term conspecific density experienced by females did not differ between parasitized and never-parasitized magpies (Table 2).
Parasitism Status and Long-Term Host Breeding Productivity
Long-term host productivity was affected by repeated parasitism (Table 2). Indeed, never-parasitized females produced on average 4·49 (SD = 1·57) fledglings per breeding attempt across their life, whereas females that were occasionally or always parasitized produced 2·45 (SD = 1·69) and 0·83 (SD = 1·16) fledglings, respectively.
Parasitism Status and Long-Term Host Habitat Preference
We found that long-term habitat preference of magpie hosts was related to parasitism status in interaction with laying date (Table 2). Magpies escaping from cuckoo parasitism bred in areas with a smaller percentage of wooded surface (i.e. high positive PC1 scores) at the beginning of the breeding season, whereas those breeding late in the season preferred to breed in areas with a larger percentage of wooded surface and smaller percentage of herbaceous crops (Fig. 3).
Theoretical models predict that spatially and temporally structured patterns of parasitism may influence co-evolutionary dynamics of parasite–host interactions (Røskaft et al. 2002; Hoover, Yasukawa & Hauber 2006; Best et al. 2011; Débarre et al. 2012). An obvious prerequisite of these models is that most natural host populations are somehow socially and/or spatially structured and therefore that not all host phenotypes have the same chance of being infected across their life. So far, studies of avian brood parasitism aiming to detect structured parasitism have considered a short temporal scale, usually of two consecutive years, and targeted on hosts of the brown-headed cowbird (e.g. Hauber, Yeh & Roberts 2004; Hoover, Yasukawa & Hauber 2006). However, hosts of brood parasites often breed in more than 2 years, and parasitism level may vary greatly from one year to another at the population scale (see methods). Thus, it is only by performing long-term studies that we will be able to ascertain the true occurrence of repeated parasitism (sensu Hauber, Yeh & Roberts 2004) and to establish the causes and consequences of a long-term status of parasitism in individual females. Here, we monitored female magpies during several years, detecting repeated parasitism over several breeding seasons tested for consistency of the emerging patterns over a scenario of variable parasitism pressure. We also looked for differences in individual characteristics of females and of the territories they hold across their life and in long-term nestling productivity, as a correlate of fitness, between females suffering parasitism and those that have never been parasitized.
We found that almost 30% of the sampled females systematically escaped from cuckoo parasitism, even when parasitism level steadily increased in the population, and neighbour conspecific nests were parasitized to some extent over the five study years. In addition, we report that 11·3% of sampled females were always parasitized and that the probability of characterizing the status of parasitism of individual magpies as never parasitized or sometimes/always parasitized was not due to the differences in sampling effort. Furthermore, our simulation analysis revealed that random expectation had a low chance (10%) of producing the observed pattern of repeated parasitism. Altogether, this can be interpreted as an evidence of moderate limited horizontal transmission of parasitism at the within-population scale (sensu Hoover, Yasukawa & Hauber 2006), suggesting that selection for the evolution of host defences is not uniform within our magpie population, and therefore, setting the scenario for the study of phenotypic and reproductive differences between females that are able to avoid parasitism and those that are parasitized. Probability of repeating parasitism status (either parasitized or not parasitized) in consecutive years varied across the study. Almost 80% of females repeated their parasitism status between 2007 and 2008, when great spotted cuckoo parasitism affected about 25% of the host population. Previous studies with cowbird hosts have found evidence of a nonrandom pattern of parasitism across breeding seasons irrespective of parasitism rates (Hauber, Yeh & Roberts 2004; Hoover, Yasukawa & Hauber 2006). Indeed, these studies show parasitism rates of cowbirds that are around 40% in eastern phoebes (Hauber, Yeh & Roberts 2004) and around 60% in prothonotary warblers (Hoover, Yasukawa & Hauber 2006). In the same vein, here we reported that the proportion of females maintaining their parasitism status between 2 years with low level of parasitism (i.e. 2007–2008) did not differ from that between years with high level of parasitism (i.e. 2009–2010), which suggests that cuckoo parasitism may be temporally structured within host populations provided parasitism levels, either high or low, were homogeneous over time.
In a scenario of changing parasitism pressure, however, our results show a significant lower consistency in the pattern of repeated status of parasitism between consecutive years at the level of individual. Indeed, the proportion of females that repeated their parasitism status suffered a drastic decrease between 2008 and 2009 after a marked increase in parasitism level in 2009. Therefore, our findings constitute the first empirical support for the contention that patterns of repeated status of parasitism within host populations are not consistent over time and can be influenced by parasitism-level oscillations at the population scale.
Our results show that there are no differences between magpies consistently escaping parasitism and those parasitized (either sometime or always) in terms of most phenotypic (i.e. body condition, body size, bill length or egg volume) or reproductive traits (i.e. clutch size, laying date and number of fledglings). Interestingly, however, we found that among the earlier breeding magpies, those building bigger nests had a lower chance of being parasitized. The trend was the opposite for the late breeders, as those having in average larger nests across their life had the highest chance of suffering parasitism (Fig. 2). This pattern emerging from long-term monitoring of status of parasitism and of phenotypic and reproductive magpie host traits contrasts with previous findings showing that great spotted cuckoos selected larger magpie nests to lay their eggs (Soler et al. 1995a). On the contrary, our results show that some females breeding early in the season escape from cuckoo parasitism, even those with large nests. This nest-size-mediated change in status of parasitism over the course of the season may be explained in terms of seasonal changes in nest conspicuity. At the beginning of the season, the number of magpie nests available (i.e. laying eggs) for cuckoos is small (see Fig. 1), and because trees are yet leafless (M. Molina-Morales, pers. obs.), all nests may be easily located by cuckoos. At this time, cuckoos may simply parasitize smaller nests because they are more accessible, because they have fewer sticks in the roof and the cup may have several entrances. However, later in the season, the amount of nests available increases (Fig. 1) as trees grow leaves and, therefore, nests become less conspicuous. The highest availability of host's nests corresponds with the higher level of parasitism, and then cuckoos may parasitize more conspicuous (large) nests. Alternatively, assuming active selection by nest size reveals parental quality (sensu Soler et al. 1995a), there may be not enough differences in quality between pairs with different nest size at the beginning of the season to select for ‘choosing’ cuckoos, whereas later on larger nests correspond to better magpie pairs. Indeed, variation in nest size among earlier breeding magpies was lower than among late breeding magpies (see Fig. 2). In any case, irrespective of the host selection mechanism behind, our results do point to nest size as one of the traits explaining differences in long-term status of parasitism of magpie hosts.
Our results provide support for the existence of a pattern of structured parasitism related to habitat characteristics. Indeed, we have found that never-parasitized females chose clearer areas at the beginning of the season, whereas later they bred in wooded places. There are no differences between parasitized and unparasitized females in other characteristics of the habitat, such as distance to cuckoo feeding areas, to the village or roads. The spatial habitat structure hypothesis (Røskaft et al. 2002) predicts that different habitat characteristics such as the presence of perches (trees or other vantage points) will influence the probability of nests of being parasitized. Our results support this hypothesis, because individual hosts escaping from cuckoo parasitism across their life breed in areas with different characteristics along the season than those suffering some degree of cuckoo parasitism. Why more clearer areas should be better than wooded areas at the beginning of the season to escape parasitism (and the opposite) remains to be explained, but may be related to changes in detectability and availability of host nests across the season (see above).
Limited horizontal transmission of brood parasitism has been suggested to slow down the evolution of host resistance (Hauber, Yeh & Roberts 2004). The theoretical model proposed in Hauber, Yeh & Roberts (2004) predicts that both the selective consequences and the time-frame for the evolution of host resistance strategies depend on the costs of brood parasitism and the magnitude of repeated parasitism. Because the costs of parasitism in cowbird are not too high and there are consistent patterns of repeated parasitism at the individual level, the model predicted a low chance for the evolution of host resistance that was empirically confirmed with field data in two cowbird hosts (Hoover, Yasukawa & Hauber 2006). In the great spotted cuckoo–magpie system, the costs of parasitism are higher than in cowbirds (Davies 2000; Payne 2005) and the magnitude of repeated parasitism is low or moderate, as more than half the females studied changed their parasitism status over their lifetime. Thus, the benefits of rejection must be large enough for this trait to spread in populations, as previous work shows (Soler & Møller 1990; Soler & Soler 2000). Accordingly, we have not found evidence that long-term parasitism status was related to host defences at the analysed spatial scale. Indeed, rejection behaviour and intraclutch variation in egg appearance, two traits related to the defensive capacity of magpie hosts (Soler et al. 1999; Avilés et al. 2004), did not differ between magpies never parasitized and those parasitized. This may suggest that the detected degree of structured parasitism is not high enough to select for a structured pattern of host defence at the within-population level. This is not surprising given the close spatial proximity, and thus probably high gene flow, between parasitized and nonparasitized host phenotypes within the population, which would impede genetic or phenotypic differentiation (Slatkin 1985; Lenormand 2002). Alternatively, it cannot be discarded that structured parasitism was only a recent phenomenon in the population and, therefore, that the absence of a structured level of defence was due to an evolutionary lag sensu Rothstein (1975).
In conclusion, females systematically escaping from cuckoo parasitism in our population presented a breeding strategy that allow them to evade parasitism, characterized by building a nest of different size and settling in areas of different characteristics over the breeding season in comparison with parasitized females. In addition, we report that unparasitized females move little between years as compared to females experiencing parasitism. Although our sample size is small and must thus be regarded with caution, this might suggest that a combination of nest size, laying date and characteristics of the habitat makes unparasitized magpies successful at avoiding parasitism, and this might explain why they disperse less than parasitized females. That they are very successful is clear since these females were able to rear double the number of chicks than those that were parasitized once or more.
A final theoretical inference emerging from our study is that more long-term individual-based studies are needed to establish the variation in spatial and temporal structure of brood parasitism in different systems, and thus to understand the long-term dynamics of avian brood–host parasite interactions. Critically, long-term individual-based studies will provide a more accurate picture of individual host success, allowing the possibility of adding up offspring production over the whole life span (Brooker & Brooker 1996; Krüger & Lindström 2001) and so of estimating the real costs of brood parasitism, which is the relevant parameter needed to understand the evolution of defensive mechanisms.
This work was financed by Junta de Andalucía (project P06-RNM-01862). We thank Tomás Pérez-Contreras, Alfredo Sánchez, Marta Precioso, Vicente Serrano and Juan Salvador for their help during field work. Marker development, genotyping and sex typing were performed at the NERC Biomolecular Analysis Facility–Sheffield supported by the Natural Environment Research Council, UK. Finally, we thank two anonymous referees for their constructive comments. We thank Jordi Moya-Laraño for his invaluable help with the simulation analysis.