Oecologia (1999) 118:265 ± 276 Ó Springer-Verlag 1999 Juan Jose Soler á Anders Pape Mùller á Manuel Soler A comparative study of host selection in the European cuckoo Cuculus canorus Received: 16 July 1998 / Accepted: 27 October 1998 Abstract Certain kinds of hosts are commonly regarded as being more suitable than other for rearing European cuckoos (Cuculus canorus) ± insectivores that lay small eggs and have open, shallow nests ± although empirical tests of cuckoo host selection are lacking. We analysed host use by the European cuckoo in 72 British passerines that are potential hosts and for which there was information available on life-history variables and variables related to cuckoo-host coevolution, such as rate of parasitism, rejection rate of non-mimetic model eggs and degree of cuckoo-egg mimicry of host eggs. The relative population size of the host species a€ected parasitism rate most strongly, followed by relatively short duration of the nestling period, and the kind of nest, with cuckoos selecting open-nesting hosts. However, the e€ect of the nestling period could be related to host body size and the kind of nest used, because hole-nesting species normally have longer nestling periods than open-nesters. We re-analysed the data excluding hole nesters and corvid species (species with larger body mass), but the results remained identical. The European cuckoo may bene®t from selecting hosts with short nestling periods because such hosts provide food for their nestlings at a very high rate. When only those species known as cuckoo hosts were analysed, the variable that best accounted for the parasitism rate was duration of the breeding season. Therefore, availability of potential hosts in both time and space is important for cuckoos in selecting hosts. J.J. Soler (&) á M. Soler Departamento de Biologõ a Animal y Ecologõ a, Facultad de Ciencias, Universidad de Granada, E-18071 Granada, Spain e-mail: jsolerc@goliat.ugr.es, Fax: +34-58-243238 A.P. Mùller Laboratoire d'Ecologie, CNRS URA 258, Universite Pierre et Marie Curie, BaÃt.A, 7eÁme eÂtage, 7 quai St. Bernard, case 237, F-75252 Paris Cedex 5, France Key words Brood parasitism á Cuckoo Host abundance á Host characteristics á Host-parasite coevolution Introduction Studies of specialisation by parasites on particular hosts have expanded our understanding of host-parasite coevolution (Price 1980). Haldane (1949) pointed out that the abundance of hosts should be important in host selection by a parasite, and when an exploited host genotype evolved a high degree of immunity against the parasite, this should result in a change in the preference of the parasite to more abundant and less immune host genotypes. Although Haldane (1949) speci®cally considered particular genotypes of hosts in his arguments, these could readily be applied to selection by parasites of particular host species. Work by parasitologists has identi®ed some of the factors responsible for variation in host speci®city. One such rule is that the more specialised the host group, the more specialised are its parasites (Eichler 1948). The degree of specialisation by parasites may provide information on the relative phylogenetic age of hosts (Noble and Noble 1976). Therefore, to gain a good understanding of host selection, it is important to know the phylogenetically independent values of the characters that make hosts suitable for exploitation by parasites. Here we study such characteristics of the hosts of a generalist brood parasite, the European cuckoo (Cuculus canorus). Interspeci®c brood parasitism is a reproductive strategy which consists of laying eggs in the nests of another species, known as the host, which usually provides care for the eggs and chicks of the parasite. This reproductive strategy is used by approximately 1% of all birds species (Payne 1977), and many bird species su€er from being hosts of brood parasites. For example, the shiny cowbird (Molothrus bonariensis), an American obligate brood parasite, has been known to parasitise more than 180 host species (Friedmann et al. 1977). The 266 European cuckoo is also known to parasitise a large number of host species, with a few main hosts (Glue and Morgan 1972; Wyllie 1981). Di€erent strains of cuckoos (gentes) have evolved eggs that mimic those of their hosts in appearance, and there is thus clear evidence of specialisation on particular host species (Brooke and Davies 1988; Mason 1986; Moksnes and Rùskaft 1995). Species which reject cuckoo eggs have been regarded as unsuitable hosts on the basis of this single criterion (Rothstein 1982). Egg mimicry could have evolved a posteriori as a counter-defence against the parasite, and if this were so, rejecter hosts would then be the agents selecting for mimetic cuckoo eggs, through natural selection (Brooke and Davies 1988; éien et al. 1995; but see Brooker and Brooker 1996). The main hosts may have particular characteristics, other than similar eggs, that make them suitable as foster parents for the European cuckoo nestling. The function of host preferences remains unknown, but it is likely that cuckoos prefer hosts that provide suitable food for parasite o€spring, or that build nests from which the cuckoo nestling is able to eject host o€spring, or that do not physically attack the female cuckoo when she is laying. Some potential host characteristics that are traditionally accepted as important for cuckoos were summarised by Payne (1977) as follows: 1. The food regime of the host species should facilitate proper development of the parasite. However, the growth rate of some brood-parasite nestlings is similar when fed by di€erent host species (Payne 1977). 2. Parental behaviour of the parasite and the foster parents should be compatible, e.g. there should be some similarity between the begging behaviour of host and parasite nestlings. 3. Host egg size may be decisive, because eggs are incubated by contact with the hosts' brood patch, and contact may be poor if parasite eggs are smaller than those of the host; hence the cuckoo has evolved a small egg relative to its body size. 4. Hosts or their nests must be readily available. 5. Host size must be compatible with that of the parasite. Since the review by Payne (1977) was published, more characteristics have been added to this list, such as the type of host nest and duration of sympatry with the parasite. However, no study has yet analysed host characteristics, other than those related to host-parasite coevolution such as host defences, in relation to parasitism by any brood parasite. More detailed studies are needed on parasitism of di€erent hosts, because the use of parasitism rate as an index of host suitability may cause a major error ± that is, apparently unparasitised nests may earlier have held a cuckoo egg that has already been ejected by the host. Thus, it can be dicult to estimate the real parasitism rate of each species with available information. Therefore, although parasitism and rejection rates are clearly interrelated (they are not really independent variables), the only accurate way to estimate the parasitism rate is to compensate for the ejection rate of the species as determined from ®eld experiments simulating parasitism (Rothstein 1982). On the other hand, there are some problems with using such parasitism rates corrected for experimental rejection rates because: 1. The rate of rejection is determined by the degree of mimicry of the cuckoo egg and, therefore, we should correct for rejection rates measured in the same populations as the rates of parasitism. However, such data are not available in the literature, which more often reports information only on experimental rejection rates of non-mimetic model eggs. 2. Parasitism rates are likely to be underestimates for common hosts (because mimetic eggs will be missed by observers), while parasitism rates for uncommon hosts are likely to be overestimates. These arguments suggest that it is more convenient simply to use data on parasitism rate as a measure of host suitability. We attempted to tackle the problem of using the parasitism rate in two ways. Firstly, we corrected the parasitism rate for the rejection rate of the host species based on rejection of experimentally introduced non-mimetic cuckoo eggs (see Appendix 1) by forcing the entry of this variable into the ®nal multiple regression model. However, most of the species for which this information is available are hosts of the European cuckoo, and, therefore, it can only be used to analyse di€erences in parasitism rates of known host species. Secondly, we used the parasitism rate directly in the comparative analyses. In this paper: 1. We analyse biological features of potential hosts in order to pinpoint those that may be important in the process of host selection by the European cuckoo. 2. We also analyse this question by excluding hole nesters as potential hosts of the European cuckoo, because some features of the host could be related to the type of nest used by potential hosts, and also by excluding corvids, which, due to their great body mass, are outside the range of potential hosts. 3. Finally, we reanalyse the data using only species that have been found parasitised by the European cuckoo, both forcing and not forcing the entry of the rejection rate into the ®nal multiple-regression model, in order to determine which variables are able to explain variance in parasitism rate of European cuckoo hosts. Materials and methods Potential host species used in the analyses To analyse possible features of potential cuckoo hosts, we used as potential hosts of the European cuckoo all British passerine species for which there was information available in the literature (n = 72); 16 of these potential hosts were hole nesters. We found data in the literature on rejection rate of experimental non-mimetic 267 eggs for 19 host species with a known duration of breeding season in Britain (Appendix 1). Variables analysed in the model We assembled information on the following variables for each potential cuckoo host: 1. Rate of parasitism in Great Britain collected from di€erent sources (Davies and Brooke 1989b; Glue and Murray 1984; Lack 1963; Wyllie 1981). 2. For a measure of host abundance, we used the number of breeding pairs during 1988±1991 in Britain, reported by Gibbons et al. (1993). When a minimum and a maximum number of breeding pairs were reported, we used mean values. 3. For a measure of host geographic range size we used number of 10 km ´ 10 km squares where the species was found breeding between 1988 and 1991 in Britain (Gibbons et al. 1993). This estimate also provides a relative measure of population size, since overall abundance and distribution of birds are generally positively correlated (Brown 1984; Maurer 1994; Blackburn et al. 1997). However, for the study of cuckoos selecting hosts, it could be of interest to distinguish between these two variables, because, while host density is a local feature, host geographic range size would be an index of availability of hosts at di€erent locations. Local abundance and geographic range size have been shown to be positively related (Blackburn et al. 1997). 4. Body mass (g): the mean value of those reported for male and female by Perrins (1987). 5. Clutch size, as the mean value of the maximum and the minimum reported by Perrins (1987). 6. Duration of incubation period, as a mean value of the maximum and the minimum reported by Perrins (1987). 7. Duration of nestling period, as a mean value of the maximum and the minimum reported by Perrins (1987). 8. Number of broods raised per breeding season (data from Perrins 1987). 9. Duration of breeding season, as the season for the occurrence of eggs without the margins for early eggs and late broods reported in annual cycle diagrams by Cramp (1985±1992) and Cramp and Perrins (1993±1994). 10. Hatching asynchrony: whether the duration of the hatching period exceeded 1 day (data from Clark and Wilson 1981; Cramp 1985±1992; Cramp and Perrins 1993±1994). 11. Type of nest (open, semi-open or hole nest) from Cramp (1985± 1992) and Cramp and Perrins (1993±1994). 12. Sociality (solitary, semi-colonial or colonial) from Cramp (1985±1992) and Cramp and Perrins (1993±1994). 13. Type of food that the potential host brings to the nestlings, from Cramp (1985±1992) and Cramp and Perrins (1993±1994). 14. Rejection rate, as the mean value of those reported from various sources (Davies and Brooke 1989a; von Haartman 1981; JaÈrvinen 1984; Moksnes et al. 1990) not only from studies in the British Isles, but also from other European countries. We used the mean value for the rejection rate because Soler and Mùller (1996) demonstrated a high repeatability between values from di€erent studies (repeatability = 0.73; SE = 0.13; F = 7.12; df = 13,16; P = 0.0002; Soler and Mùller 1996). 15. Degree of mimicry of European cuckoo eggs parasitising different host species as the percentage of cuckoo egg found in museum collections (in England, Sweden, Germany, Denmark, Switzerland, Finland, United States, The Netherlands, The Czech Republic, Hungary, Austria, Serbia (Vojvodina) and Norway) that mimics those of each host species reported by Moksnes and Rùskaft (1995). Statistical procedures Because we were interested in the characteristics that make passerines suitable as hosts of the European cuckoo (features inde- pendent of common phylogenetic ancestry), we used available comparative methods. In a previous article we calculated the repeatability of parasitism and rejection rate variables for species for which there were data available in the literature from at least two di€erent countries, and both repeatability values were statistically signi®cant (Soler and Mùller 1996), thereby being candidates for comparative analyses. No comparative method allows analysis of potential relationships of more than a single discrete variable with continuous variables (Harvey and Pagel 1991). On the other hand, in analyses of the possible relationship between discrete variables and parasitism rate (one by one), because the phylogeny of the hosts is poorly known, only a few contrasts were calculated (hatching asynchrony: 9; type of nest:5; sociality: 5; type of food that the potential host brings to the nestlings: 3). Moreover, the contrasts were estimated for di€erent nodes for each variable (depending on whether the character changed), and thus the contrast values for each variable cannot be combined and used in multivariate analyses. To solve problems with discrete variables in the analyses, we carried out a canonical correspondence analysis for the discrete variables (ter Braak 1987), which produced a continuous canonical axis representing the values of the discrete variables. We introduced four discrete variables into the correspondence analysis: (1) hatching asynchrony (yes or no), (2) type of nest (hole-, semi-open, or open nests), (3) sociality (colonial, semi-colonial, or solitary), and (4) kind of food brought to the nestlings (insects, mixture of insects and seeds, or seeds). The analysis provided us with three axes that explained 100% of the variance (total eigenvalue = 0.12, percentage of variance explained: ®rst axis = 55.1%, second axis = 26.0%, third axis = 19.0%), and one coordinate for each species on each axis. To identify the relationships between each axis and the di€erent variables, i.e. to interpret the canonical axes, we carried out Spearman rank correlation analyses between the categorised data and the coordinates for all three axes (see Table 1 for the values of the Spearman coecients for each axis). We performed one correspondence analysis for all species, another that excluded hole nesters and corvids, and a third one for European cuckoo host species for which we found data on experimental rejection rate and duration of breeding. In this way, we were able to calculate contrasts for each of the canonical axes. Because we did not ®nd information on duration of the breeding season for all 72 species, we analysed the data including and excluding this variable. For all canonical analyses performed, the resulting axes explained 100% of the variance. Although Sibley and Ahlquist (1991) published a bird phylogeny based on DNA sequences, many of the passerine species for which we found data on the previously described variables are not in this phylogeny, and thus it was not possible to use that phylogenetic information in this study. Therefore, we used the passerine classi®cation given in Howard and Moore (1991) as a phylogeny. Although the use of phylogenies based on morphology (traditional taxonomic classi®cation) could cause some problems, it is preferable to use the most available complete information rather than making no analyses at all, although analyses should be revised when the phylogeny becomes better known (Garland et al. 1991). Moreover, several recent studies have suggested that phylogenies based on molecular changes may also include inaccuracies such as branching patterns (Harvey et al. 1992; Nee et al. 1993) and there are many examples in the literature where traditional morphological classi®cation has been used in comparative studies (e.g. Hartley and Davies 1994; Keller and Genoud 1997; Owens and Bennett 1994). In the analysis, we assumed polytomies among di€erent species within the same genus, and among di€erent genera within the same family; i.e. we assumed that all species in the same genus (or all genera in the same family) evolved simultaneously from a common ancestor (multiway speciation events; see Purvis and Garland (1993) for problems with polytomies, their implications, and possible solutions). Hence, we have set branch lengths of all species to the same value (=1) (Garland et al. 1993; Purvis and Garland 1993). We also used two methods to solve polytomies and assign branch lengths, one method developed by Grafen (1989) that gives 268 Table 1 Relationships between the discrete variables and the value of each canonical axis from the correspondence analysis for A all species in the analysis, B all species excluding hole nesters and corvids, and C only species known to be cuckoo hosts and with information on experimental rejection rate. Values are Spearman rank order correlation coecients. Open nesters were assigned a value of 1, semi-open nesters a value of 2, and hole nesters a value of 3. Insectivorous nestlings were assigned a value of 1, semi-insectivorous nestlings a value of 2, and seed eating nestlings a value of 3. Synchronous species were assigned a value of 1, and asynchronous species a value of 2. Solitary species were assigned a value of 1, semi-colonial species a value of 2, and colonial species a value of 3 Axis 1 A All species in the analysis (n=72) Hatching asynchrony 0.067 Type of nest )0.860 Sociality 0.262 Type of food 0.662 Axis 2 Axis 3 0.593 )0.260 )0.516 )0.310 0.200 )0.200 0.456 )0.271 B Non-hole nesters and non-corvids (n=51) Hatching asynchrony )0.725 )0.396 Sociality 0.491 0.275 Type of food 0.709 )0.390 C Cuckoo host species (n=22) Hatching asynchrony )0.098 Type of nest )0.891 Sociality 0.408 Type of food 0.625 0.653 0.051 0.472 )0.488 0.576 )0.196 )0.381 )0.121 greater weight to those species or nodes whose data points are not well explained by the phylogeny or by the other predictor variables under consideration, and another developed by Pagel and Harvey (1989) which relies on the assumption that the X (or Y) variable can provide useful information about the hidden phylogenetic structure in the multiple node. These methods can be applied to imperfectly resolved phylogenies, such as might be the case if a taxonomy were used instead of a phylogeny, as in the present work. To control for the possible e€ects of common phylogenetic descent we used the independent-comparison method of Felsenstein (1985) as implemented in a computer program written by Garland et al. (1993). This method ®nds a set of independent pairwise di€erences or contrasts, assuming that changes along the branches of the phylogeny can be modelled by a Brownian motion process (successive changes are independent of one another), and that the expected total change added together over many independent changes is zero (Harvey and Pagel 1991). Therefore, pairwise di€erences in the phylogenetic tree are independent of each other (Harvey and Pagel 1991). The advantage of the independent-comparison approaches is that, by partitioning the variation appropriately, all contrasts can be used to assess a hypothetical comparative relationship (Harvey and Pagel 1991). We calculated Felsenstein's independent contrasts for all the potential confounding variables, including those non-continuous variables. For the calculation of contrasts, we used normalised values of continuous variables and the coordinates on the axes from the correspondence analysis for non-continuous variables. Thus, we obtained a set of contrasts for each variable estimated from the same node or species pair. To study the possible relationships between the level of parasitism that each host is su€ering by the European cuckoo (parasitism rate) and di€erent life-history variables of the potential hosts, we used the standardised independent contrasts (from the independent-contrast method of Felsenstein 1985) in a forward stepwise multiple regression analysis (F value to enter in the model additional independent variables = 3.00) forced through the origin, where parasitism rate was the dependent variable and all the others were independent variables. Some variables in the analysis could be interrelated, and, in order to solve this problem, we carried out a principal component analysis (PCA) with the values of contrasts of all dependent variables derived from a phylogeny with all branch lengths equalling one. However, the variables shared little variance, the eigenvalue of the second factor being less than 2 (eigenvalue factor 2 = 1.68) and the three ®rst axes explaining only 58.7% of the variance. Therefore, for a better understanding of the results, we used the contrast value for each variable instead of the principal-component coordinate for each factor. It is known that the kind of nest used by potential hosts is one of the most important physical constraints on parasitism by the European cuckoo, because the opening of a nest hole is usually too small to allow the young cuckoo to ¯edge (éien et al. 1995). Furthermore, hole-nesting birds have a reduced risk of predation that has resulted in evolution of a long nestling period (Bosque and Bosque 1995; Lack 1968) and large clutches (Lack 1968). Corvids are not potential hosts of the European cuckoo (although Cyanopica cyana is a cuckoo host in Japan: Nakamura 1990). Therefore, we analysed the data including and excluding hole nesters and corvids from the analysis. To identify variables explaining di€erences in parasitism rate su€ered by species known as a hosts, we used those for which information was available for all variables (n = 19), and in order to control for problems related to host ejection of cuckoo eggs before nests were checked, we forced, the rejection rate in the multiple-regression analyses as an independent variable. All variables introduced in the analyses were normalised (Kolmogorov-Smirnov tests, n.s.): parasitism rate was transformed to log(n + 0.001), population density of the host species was fourth-root transformed. Host geographic range size and duration of breeding season were already normally distributed, body mass, clutch size and incubation period were log(n) transformed, the duration of the nestling period was log10(n), rejection rates and degree of mimicry were transformed to arcsin(n), axes 1 and 3 were already normally distributed, and axis 2 and number of broods raised per breeding season were binomial. All tests were two-tailed. Results Variables explaining cuckoo host selection The analysis of the relationship between the standardised contrasts of the di€erent life history variables of the potential hosts and the standardised contrasts of the parasitism rate showed that only three variables explained a signi®cant proportion of the variance: host geographic range size (positively related), the duration of the nestling period (negatively related), and the kind of nest selected (®rst canonical axis) (selecting open nesters). This was the case regardless of the method used to estimate branch length (Table 2). These three variables explained more than 28.1% of the variance (multiple R > 0.52, F > 8.7, df = 3,68, P < 0.00006; Table 2). When duration of the breeding season was used as an additional independent variable (thereby reducing sample size, see Methods), the result was similar and only geographic range size of the hosts and the duration of the nestling period explained a signi®cant proportion of the variance independently of the method used to estimate branch length (R2 > 0.39, F > 14.5, df = 2,45, P < 0.000011; Table 2; Fig. 1). Because the duration of the nestling period could be related to the kind of nest used, and corvids fell outside the body-size range of potential hosts, we re- 269 analysed the data for non-hole nesters and non-corvids only. The results were similar, showing that, regardless of the method used to estimated branch length, the most important predictor of parasitism rate was host density (partial correlation coecients, R > 0.41, t > 3.19, df = 49, P < 0.0024; Table 2) and the duration of the nestling period (partial correlation coef®cients, R < )0.327, t>2.4, df = 49, P < 0.02; Table 2) (only these two variables were statistically signi®cant). Both variables explained more than 24% of the variance in the parasitism rate (multiple R > 0.49, F > 7.69, df = 2,48, P < 0.0013; Table 2). When the duration of the breeding season was used as an additional independent variable, geographic range size of the hosts, but not host density, and duration of the nestling period were the variables that best accounted for the parasitism rate (geographic range size, partial correlation coecient, R > 0.56, t > 4.1, df = 38, P < 0.00017; duration of the nestling period, R < )0.45, t > 3.12, df = 38, P < 0.0034; Table 2). Therefore, host population size (geographic range size or density) and the duration of the nestling period explain why some passerine species are more suitable hosts of the European cuckoo than others. Variables explaining di€erences in parasitism rate of known European cuckoo hosts In this analysis of 19 species of hosts of the European cuckoo, we introduced the level of mimicry of cuckoo eggs and the rejection rate as additional variables. The results of a stepwise multiple-regression analyses of the phylogenetically independent contrasts showed that the European cuckoo preferentially selects hosts with a longer breeding season, and those with eggs that the cuckoo can mimic. No other variables were statistically signi®cant (Table 3). However, when rejection rate was forced to enter into the ®nal model (see Methods), thereby controlling the ®nal model for this variable, duration of breeding season was the only signi®cant variable (Table 3). Discussion Although the European cuckoo is able to parasitise a large number of passerine species, it has evolved clear preferences for particular species. For example, the number of hosts reported in Britain exceeds 50, but only three species, dunnock (Prunella modularis), reed warbler (Acrocephalus scirpaceus) and meadow pipit (Anthus pratensis) accounted for 77% of all cases of parasitism (Glue and Morgan 1972). These host preferences must be related to di€erences in ®tness obtained by parasitism of di€erent hosts, and particular host characteristics may be responsible for such di€erential parasite ®tness. One of the most obvious host characteristics, directly related to parasite ®tness, is whether the host is able to recognise and reject cuckoo eggs. Although it has been shown that cuckoos have decreased their use of some rejector species, because of the great costs for the brood parasite of host rejection of foreign eggs (Rothstein 1990), brood parasites are still using particular host species rather than switching to acceptor species. A fundamental question regarding host use by cuckoos is whether host use is genetically inherited or arises by imprinting. However, an analysis of genetic di€erentiation of host race using mitochondrial and microsatellite DNA variation found no signi®cant di€erences among cuckoos from di€erent hosts (Gibbs et al. 1996). Recognition of foreign eggs by common cuckoo hosts is a non-intrinsic host characteristic resulting from hostparasite interactions possibly depending on the duration of coevolution between parasite and host populations (Davies and Brooke 1989a,b; éien et al. 1995; Soler and Mùller 1996; but see Brooker and Brooker 1996). Moreover, di€erent populations of the same species can be rejectors or acceptors depending on the duration of sympatry between parasite and host, such as meadow pipit and pied/white wagtail (Motacilla alba) (Davies and Brooke 1989a) or magpies (Soler and Mùller 1990). Therefore, the fact that the parasite does not dramatically change host species, despite the high costs of egg rejection, may be because of bene®cial host characteristics that are important for the proper development of the o€spring of the parasite, but perhaps also because cuckoos have a genetically based host choice. Some host characteristics that did not explain a signi®cant proportion of the variance in parasitism could still be important in the process of host selection by the cuckoo. The low variance of these variables among hosts, and the relationships between di€erent life-history variables of hosts, may mask their importance in the comparative analyses. For example, all hole nesters in the analyses feed their chicks with insects, and all seedeaters are non-hole nesters. Since the European cuckoo rarely parasitises hole nesters (see Appendix 1), a preference of the brood parasite for non-hole-nesting hosts would be con®rmed by the comparative analyses. Due to the relationship between hole-nesting and the kind of food brought to nestlings, the analysis could even detect an apparent relationship between parasitism and food provided for nestlings. Other important interrelated variables are (1) the body mass of the potential host species and the duration of the incubation and the nestling periods, and clutch size (Lack 1968), (2) the duration of the nestling period, clutch size and incubation period (SiikamaÈki 1995), and (3) relative population size and degree of sociality. Our results showed that host population size (mostly geographic range size) and the duration of the nestling period of the host were related (positively and negatively, respectively) to parasitism rate su€ered by different potential host species. The relatively large importance of the population size of potential hosts as a predictor of parasitism by the European cuckoo makes intuitive sense. This result is consistent with the predic- 270 Table 2 Results of the forward stepwise multiple regression analyses forced through the origin between standardised contrasts of parasitism rate (dependent variable), standardised contrasts of life-history variables [population density, population geographic range, body mass, clutch size, duration of incubation period, duration of nestling period, number of broods per season, hatching asynchrony, type of nest (open, semi-open or hole nests), sociality (solitary, semi-colonial or colonial), and type of food brought to the nestlings] and standardised contrasts of host variables related to coevolution (mimicry of cuckoo eggs to those of the host and host rejection-rate of non-mimetic egg) of potential host species calculated using three di€erent methods to assign branch lengths. R^ is the multiple regression coecient of the variables included in the model All branch lengths equal to 1 method Analysis of all potential host species Dependent variable: Parasitism rate Independent variables: Geographic range size Duration of the nestling period Preference for open nesters Including duration of the breeding season in the analyses Dependent variable: Parasitism rate Independent variables: Geographic range size Duration of the nestling period Analysis of potential non-hole nester and non-corvid host species Dependent variable: Parasitism rate Independent variables: Population density Duration of the nestling period Including duration of the breeding season in the analyses Dependent variable: Parasitism rate Independent variables: Geographic range size Duration of the nestling period Using the method of Pagel and Harvey (1989) Using the method of Grafen (1989) R Test P R Test P R Test P R^ = 0.534 F(3,68) = 9.02 0.00004 R^ = 0.531 F(3,68) = 8.92 0.000046 R^ = 0.528 F(3,68) = 8.76 0.000055 R = 0.427 R = )0.299 R = 0.235 t(69) = 3.90 t(69) = 2.59 t(69) = 2.00 0.0002 0.012 0.0498 R = 0.429 R = )0.271 R = 0.241 t(69) = 3.91 t(69) = 2.33 t(69) = 2.05 0.0002 0.023 0.044 R = 0.427 R = )0.276 R = 0.240 t(69) = 3.89 t(69) = 2.37 t(69) = 2.04 0.00023 0.021 0.045 R^ = 0.630 F(2,45) = 14.8 0.000011 R^ = 0.632 F(2,45) = 15.0 0.00001 R^ = 0.631 F(2,45) = 14.9 0.000011 R = 0.569 R = )0.509 t(46) = 4.64 t(46) = 3.98 0.00003 0.00025 R = 0.579 R = )0.492 t(46) = 4.77 t(46) = 3.79 0.00002 0.00044 R = 0.577 R = )0.491 t(46) = 4.75 t(46) = 3.78 0.000021 0.00045 R^ = 0.493 F(2,48) = 7.69 0.0013 R^ = 0.507 F(2,48) = 8.28 0.0008 R^ = 0.507 F(2,48) = 8.32 0.0008 R = 0.422 R = )0.327 t(49) = 3.22 t(49) = 2.40 0.0023 0.02 R = 0.424 R = )0.342 t(49) = 3.24 t(49) = 2.52 0.0022 0.015 R = 0.419 R = )0.347 t(49) = 3.20 t(49) = 2.57 0.0024 0.013 R^ = 0.602 F(2,37) = 10.5 0.00025 R^ = 0.613 F(2,37) = 11.1 0.00016 R^ = 0.613 F(2,37) = 11.1 0.00016 R = 0.565 R = )0.456 t(38) = 4.17 t(38) = 3.12 0.00017 0.003 R = 0.576 R = )0.457 t(38) = 4.28 t(38) = 3.12 0.00012 0.0034 R = 0.575 R = )0.459 t(38) = 4.26 t(46) = 3.15 0.00013 0.0032 0.009 0.08 0.13 t(16) = 2.98 t(16) = 1.87 t(16) = 1.60 R = 0.610 R = 0.434 R = )0.383 0.009 0.08 0.12 t(16) = 2.99 t(16) = 1.87 t(16) = 1.63 R = 0.611 R = 0.435 R = )0.388 t(16) = 2.79 t(16) = 2.04 t(16) = 1.65 R = 0.584 R = 0.466 R = )0.392 0.013 0.06 0.12 0.008 F(3,15) = 5.68 R^ = 0.729 0.008 F(3,15) = 5.71 R^ = 0.730 0.011 F(3,15) = 5.26 model R^ = 0.716 Forcing entry of rejection rate into the Dependent variable: Parasitism rate Independent variables: Duration of the breeding season Mimicry Rejection rate 0.009 0.047 t(17) = 2.93 t(17) = 2.14 R = 0.590 R = 0.472 0.009 0.048 t(17) = 2.94 t(17) = 2.13 R = 0.592 R = 0.470 t(17) = 2.91 t(17) = 2.1 R = 0.588 R = 0.464 0.01 0.05 0.008 F(2,15) = 6.58 R^ = 0.672 0.008 F(2,16) = 6.56 R^ = 0.671 0.012 F(2,16) = 5.90 R^ = 0.652 Final model Dependent variable: Parasitism rate Independent variables: Duration of the breeding season Mimicry P Test R P Test Using the method of Grafen (1989) Using the method of Pagel and Harvey (1989) R P Test R tion of Haldane (1949) that parasites should specialise on the commonest host genotype. Host availability must be an important factor for the cuckoo choosing hosts because one European cuckoo female lays more than 12 eggs per season (Payne 1977), presumably to compensate for rejection and predation rates, and, if the host is not suciently widespread and/or abundant, it could be dicult for a female cuckoo to ®nd nests in which to lay all eggs. The scarcity of individual potential host species has been proposed as an important factor a€ecting the use of secondary host species (Riddiford 1986). When species known to be hosts of the European cuckoo were analysed, variation in parasitism rate was explained by the duration of the breeding season. Species with a long breeding season were frequently parasitised. These two results suggest that availability of potential hosts both in time and space is important for cuckoos selecting hosts. The second variable explaining cuckoo host selection is the duration of the nestling period. That was the case when analysing all potential host species, but not when only known host species were used in the analyses. One of the possible factors explaining why the European cuckoo selects hosts with relatively short nestling peri- All branch lengths equal to 1 method Fig. 1 Relationships between residuals of parasitism rates and A duration of the nestling period and B geographic range size of species for which we found information on duration of the breeding season. Standardised contrasts (st C), estimated with all branch lengths equal to 1 and politomies (see Methods), were used in the analyses Table 3 Result of stepwise regression analyses between parasitism rate (dependent variable) and life history, and other variables related to coevolution (mimicry by cuckoo eggs of those of the host and host rejection-rate of non-mimetic eggs) of 19 host species, using standardised contrasts calculated according to di€erent methods to resolve polytomies. We also controlled for the possibility of hosts ejecting the cuckoo egg before checking the nest by forcing standardised contrasts of rejection rate to enter in the ®nal model. R^ is the multiple regression coecient of the variables included in the model 271 272 ods is the low risk of nest predation compared to species with longer nestling periods (Bosque and Bosque 1995; Lack 1968). However, nestlings of the European cuckoo spend around 17 days in the nest, independently of the host species (Wyllie 1981), and therefore, the predation risk of a nest with a cuckoo nestling would be independent of the host species. However, species with a relatively short nestling period also have high growth rates (Lack 1968), and, therefore, the food requirements by the chicks of such species are higher than for species with a low growth rate. Hence, the growth rate may re¯ect a trade-o€ between selection for rapid growth to escape predation and selection for slow growth to reduce food requirements (Lack 1968). An example of this e€ect is the low growth rate of hole nesters (Gill 1990). The family Cuculidae, which contains both brood parasites and non-parasitic species, has a short nestling period for their body mass (Lack 1968). This means that their chicks need a large amount of food per day to support their fast growth rate. Therefore, broodparasitic cuckoos should select hosts based on their growth rate and, because they need foster parents that are able to provide chicks suciently for a high growth rate, they should select hosts with relatively short nestling periods. The duration of the nestling period is also related to clutch size, and larger clutches are commoner in species with low predation rates (hole nesters) and long nestling periods (Gill 1990; Lack 1968). However, clutch size did not enter the ®nal models explaining variation in parasitism rates, but it was related to other life-history variables such as duration of the breeding season (independently of the method used to resolve polytomies, R < )0.29, n = 47, P < 0.05). Therefore, clutch size could also be an important factor per se due to the ejection behaviour of the cuckoo chick (Wyllie 1981), and it would be bene®cial for the brood parasite to select hosts with a small clutch size, thereby reducing the cost of ejecting host eggs and/or chicks for the cuckoo nestling. This activity is very costly (the eggs or the chicks of the host are very heavy, normally more than half the mass of the cuckoo chick), and sometimes the cuckoo chick is exhausted after ejecting one host egg or chick (Palomino, personal communication). For example, it has been proposed that their deep nests explain the low parasitism rate su€ered by thrushes (Turdus spp., Moksnes et al. 1990), because of the diculty for the cuckoo chick of ejecting the eggs and chicks of the host. Another possible reason why the cuckoo prefers host species with small, fast-growing broods is that such species are more likely to be able to renest the same season and, therefore, are better able to cope with the costs of parasitism (Brooker and Brooker 1996) and o€er new opportunities for the parasite. Although the number of broods raised during the breeding season did not signi®cantly improve the model explaining parasitism rate, the number of broods is signi®cantly related to the duration of the breeding season, given that species with more than one breeding attempt have longer breeding seasons than those with only one attempt. Thus, both the number of breeding attempts and the duration of the breeding season are related to temporal availability of hosts. Recently, Blackburn et al. (1996) showed that the abundance of British birds correlated with life-history variables corresponding to rate of o€spring development: more abundant species are those that develop faster. This suggests an interrelationship between the two principal factors in¯uencing host choice in cuckoos. However, because both host availability (geographic range size or density) and duration of the nestling period are included in the same multiple regression model, the partial correlation coecients are by de®nition controlled for each other's e€ects. Therefore, although interrelationships among these two variables were not investigated, the detected e€ects of geographic range size (or density) of potential hosts and the duration of the nestling period on parasitism rate are independent of the relationship between those variables. With respect to the discrete variables such as hatching asynchrony, type of nest, sociality, and type of food that the potential host brings to the nestlings, we found no signi®cant relationship with parasitism rates. However, we do not exclude the possibility that they are important for cuckoo host selection because (1) the phylogeny of potential hosts is poorly known and, therefore, the number of independent contrasts is very small, and (2) the relationships between the discrete variables and other life-history variables are sometimes strong. In conclusion, we provide comparative evidence suggesting that host availability (in time and space) and the duration of the nestling period of potential hosts explain host selection by the European cuckoo. Acknowledgements We are most grateful to M.G. Brooker, N.B. Davies, M. Martõ n-Vivaldi, E. Rùskaft, S.I. Rothstein and Carmen Zamora for valuable comments on the manuscript. Funds were provided by the European communities Net work (ERBCHRXCT940649) and a European Communities post-doctoral grant (ERBCHBCT930307) to J.J.S. Appendix 1 Information on parasitism rate (%) (sources: a Davies and Brooke 1989b, b Lack 1963, c Glue and Murray 1984, and d Wyllie 1981), rejection rate (%) (sources: e Davies and Brooke 1989a, f Moksnes et al. 1990, g JaÈrvinen 1984, h von Haartman 1981, i Cramp 1985; when more than one reference was available for a species we calculated the mean value), geographic range size, population density (number of pairs breeding in Britain), body mass (g), clutch size, incubation period (days), nestling period (days), hatching asynchrony, type of nest, sociality, and type of food that the potential hosts bring to the nest to feed their chicks Species Parasitism Rejection rate rate Geographic range Population Body density mass Clutch Incubation Nestling No. of Hatch Nest size period period broods asyntype chrony Sociality Food type Alaudidae Alauda arvensis Lullula arborea 0.04(c) 0.01(b) 2571 162 2000000 350 38.6 27.2 4 4 12 14 19 12 2 2.5 Sync Sync Open Open Solitary Solitary Hirundinidae Riparia riparia Hirundo rustica Delichon urbica 0.00(c) 0.01(d) 0.00(c) 991 2457 2227 250000 570000 375000 13.4 18.6 17.0 5 5 5 14 15 14 19 21 21 2 2.5 2.5 Sync Sync Sync Hole Colonial Insects 2.25 Semi-open Semi-col Insects 3 Hole Colonial Insects Motacillidae Anthus trivialis A. pratensis Motacilla ¯ava M. cinerea M. alba 0.74(c) 2.66(a) 0.13(c) 0.03(c) 0.42(a) 1215 2261 759 1657 2467 120000 1900000 50000 34000 300000 21.6 18.2 16.8 18.4 22.6 5 5 6 5 6 14 14 13 14 14 13 14 13 12 15 1.5 2 2 2 2 Sync Sync Sync Sync Sync Open Open Open Semi-open Semio-pen Solitary Solitary Solitary Solitary Solitary Insects Insects Insects Insects Insects Laniidae Lanuius collurio 0.01(b) 2 2 28.2 6 15 14 1 Async Open Solitary Insects 2 Cinclidae Cinclus cinclus 0.00(c) 1097 14000 63.1 5 16 23 2 Sync Semi-open Solitary Insects 4 0.0(ef) 26.8(ef) 80.0(f) 73.2(ef) Duration of the breeding season Egg mimicry Insects 3 Insects 4 0.0 74.9 64.0 54.2 Troglodytidae Troglodytes troglodytes 0.04(a) 0.0(e) 2650 7100000 9.4 6 15 17 1.5 Async Semi-open Solitary Insects 3 0.0 Prunellidae Prunella modularis 1.93(a) 3.1(ef) 1317 2000000 20.8 5 12 12 2.5 Sync Open Solitary Insects 4 0.4 2536 4200000 18.1 5 14 14 2.5 Sync Open Semi-col Insects 2.5 0.01(b) 303 5500 20.7 5 14 13 1 Async Open Solitary Insects 1 0.00(c) 57 100 16.4 5 13 17 2 Sync Hole Solitary Insects 1.75 1019 1062 850 1341 90000 21000 15000 55000 14.3 16.9 14.4 25.9 6 6 6 6 14 13 15 14 15 13 15 15 1.5 1.5 2.5 1.5 Sync Sync Sync Sync Hole Open Open Hole Solitary Solitary Solitary Solitary Insects Insects Insects Insects 1.75 1.75 3 2.75 401 2583 18 1491 46 2153 8250 4400000 25 990000 60 230000 112.0 106.0 98.5 73.4 67.2 125.0 5 4 5 4 6 4 14 14 14 13 13 14 14 14 14 14 13 15 2 2.5 1.5 2.5 2 2 Sync Async Async Sync Sync Sync Open Open Open Open Open Open Solitary Solitary Colonial Solitary Solitary Solitary Insects Insects Insects Insects Insects Insects 2.25 4 2.5 4.25 1.75 2.5 0.29(a) 0.03(a) 0.01(b) 0.12(c) 0.00(a) 0.11(c) 0.01(a) 0.00(c) 0.01(b) 0.00(c) 0.00(c) 20.0(e) 31.5(efgh) 5.9(ef) 63.9(ef) 9.1(f) 62.7(ef) 34.9(f) 4.4 0.0 0.0 0.0 0.0 273 Turdidae Erithacus rubecula Luscinia megarhynchos Phoenicurus ochruros P. phoenicurus Saxicola rubetra S. torquata Oenanthe oenanthe Turdus torquatus T. merula T. pilaris T. philomelos T. iliacus T. viscivorus Species Sylviidae Cettia cetti Acrocephalus palustris A. scirpaceus Phylloscopus sibilatrix P. collybita P. trochilus Sylvia atricapilla S. borin S. communis S. curruca S. undata Regulus ignicapillus R. regulus 274 Appendix 1 (contd.) Parasitism Rejection rate rate 0.00(c) 1.88(c) 86.8(i) Geographic range Population Body density mass Clutch Incubation Nestling No. of Hatch Nest size period period broods asyntype chrony 57 8 450 12 14.2 12.0 5 5 16 12 15a 12 1 1 Sync Async Open Open Sociality Food type Solitary Solitary Duration of the breeding season Insects Insects 2 19.1 5.54(a) 0.09(c) 61.8(e) 638 859 60000 17200 12.0 8.9 4 6 12 13 12 12 2 1 Sync Sync Open Solitary Semi-open Solitary Insects 2.75 Insects 1.5 47.4 0.00(c) 0.06(c) 0.17(c) 0.32(c) 0.07(c) 0.01(b) 0.00(c) 0.00(c) 100(f) 90.0(f) 76.9(f) 66.7(f) 1662 2446 1757 1477 1934 978 40 48 640000 2300000 580000 200000 660000 80000 950 165 7.5 8.8 17.0 19.4 14.7 11.9 9.5 5.5 6 5 5 5 5 5 4 8 13 13 11 12 12 11 13 15 14 14 12 10 12 11 14 20 1.5 1 1.5 1.5 1.5 1.5 2 2 Sync Sync Sync Sync Sync Sync Sync Async Semi-open Semi-open Open Open Open Open Open Semi-open Insects Insects Insects Insects Insects Insects Insects Insects 0.0 0.0 65.0 86.9 1930 560000 5.6 8 16 19 2 Async Semi-open Solitary Insects 1.75 2097 547 120000 37500 15.0 12.1 5 7 13 13 14 16 2 1 Sync Sync Semi-open Solitary Hole Solitary Insects 2.25 Insects 0.01(b) Muscicapidae Muscicapa striata 0.12(c) Ficedula 0.00(a) hypoleuca 72.2(ef) 0.0(ef) Solitary Solitary Solitary Solitary Solitary Solitary Solitary Solitary 2.25 2 2 2.25 2.5 2.5 2.75 1.75 Panuridae Panurus biarmicus 0.00(c) 52 400 15.5 6 13 12 2 Sync Open Aegithalidae Aegithalos caudatus 0.00(c) 1868 210000 9.0 10 14 16 1 Sync Semi-open Solitary Insects 2.25 858 789 2317 1382 60000 25000 1600000 3300000 10.5 10.3 18.8 10.8 7 8 8 9 13 13 13 14 17 18 20 19 1 1 1.5 1 Async Async Async Async Hole Hole Hole Hole Solitary Solitary Solitary Solitary Insects Insects Insects Insects Paridae Parus palustris P. montanus P. major P. caeruleus 0.00(c) 0.00(c) 0.00(c) 0.00(a) Sitiidae Sitta europaea 0.0(f) 16.7(e) 0.0(ef) Semi-col Insects 2.75 0.00(c) 1063 130000 22.7 7 15 24 1 Sync Hole Solitary Insects Certhiidae Certhia familiaris 0.00(c) 1675 200000 8.7 6 15 15 1.5 Async Hole Solitary Insects Ploceidae Passer domesticus 0.00(c) 1431 3600000 28.2 5 13 15 2.5 Sync Hole 0.00(c) 1040 110000 22.4 5 13 13 2.5 Sync Hole Colonial Some seed Colonial Some seed P. montanus Egg mimicry 33.0 Emberizidae Emberiza citrinella E. schoeniclus 0.01(b) 0.15(a) 100(f) 95.0(f) 1962 1200000 26.9 4 13 13 2.5 Sync Open Solitary 1846 220000 19.4 5 14 13 2 Sync Open Solitary 18 229 22.8 4 12 12 2.5 Sync Open Solitary Some seed Some seed Some seed 2.25 4.1 1.25 4.2 E. cirlus 0.00(c) Fringillidae Fringilla coelebs 0.01(b) 61.3(ef) 2503 5400000 22.7 5 13 14 1 Sync Open Solitary Carduelis chloris 0.05(a) 24.1(ef) 2056 530000 28.6 5 14 15 2.5 Sync Open Semi-col C. spinus C. carduelis 0.00(c) 0.01(b) 728 1888 300000 220000 13.8 5 16.4 6 12 13 15 14 2 2.5 Sync Async Open Open Semi-col Semi-col Acanthis ¯ammea A. ¯avirostris Loxia curvirostra Pyrrhula pyrrhula 0.00(c) 0.14(c) 0.00(c) 0.04(c) 1184 420 406 1769 160000 65000 1500 190000 12.0 18.0 42.4 24.5 5 6 4 5 12 13 15 13 12 15 21 14 1.5 2 1 2 Async Async Sync Async Open Open Open Open Semi-col Semi-col Semi-col Semi-col Coccothraustes coccothraustes 0.01(b) 170 4725 52.9 5 13 13 1 Sync Open Semi-col 2498 1100000 80.0 5 13 21 1.5 Async Hole Colonial Insects 1.25 0.0(e) Some seed Some seed Seeds Some seed Seeds Seeds Seeds Some seed Some seed 3 1.25 11.8 1.25 8.2 2.25 2 1.75 2 Sturnidae Sturnus vulgaris 0.00(a) Oriolidae Oriolus oriolus 0.00(c) 14 32 76.7 4 15 15 1 Async Open Solitary Insects Corvidae Garrulus glandarius0.00(c) Pica pica 0.00(c) 1267 1775 160000 590000 162.0 6 220.0 6 16 18 20 25 1 1 Synca Async Open Open Solitary Solitary Pyrrhocorax pyrrhocorax Corvus monedula C. corone 0.00(c) 64 315 338.0 4 18 38 1 Async Open Solitary Seeds 1.5 Some seed Insects 0.00(c) 0.00(c) 2149 2124 390000 790000 234.0 5 530.0 5 19 19 33 31 1 1 Async Async Hole Open C. corax 0.00(c) 785 7000 1193.0 5 21 38 1 Sync Open Colonial Insects 0.75 Colonial Some 0.75 seed Solitary Insects 23.8(ef) a From Cramp (1985±1992) and Cramp and Perrins (1993±1994) 275 276 References Blackburn TM, Lawton JH, Gregory RD (1996)Relationships between abundances and life histories of British birds. J Anim Ecol 65:52±62 Blackburn TM, Gaston KJ, Quinn RM, Arnold H, Gregory RD (1997) Of mice and wren ± the relation between abundance and geographic range size in British mammals and birds. Philos Trans R Soc Lond B 352:419±427 Bosque C, Bosque MT (1995) Nest predation as a selective factor in the evolution of developmental rates in altricial birds. Am Nat 145:234±260 Braak CJF ter (1987) CANOCO ± a FORTRAN program for canonical community ordination by [partial] [detrended] [canonical] correspondence analysis, principal components analysis and redundancy analysis (version 2.1). TNO Institute of Applied Computer Science, Statistical Department, Wageningen Brooke M de L, Davies NB (1988) Egg mimicry by cuckoos Cuculus canorus in relation to discrimination by hosts. Nature 335:630±632 Brooker M, Brooker L (1996) Acceptance by the splendid fairywren of parasitism by Hors®eld's bronze cuckoo: further evidence for evolutionary equilibrium in brood parasitism. Behav Ecol 7:395±407 Brown JH (1984) On the relationship between abundance and distribution of species. Am Nat 124:225±279 Clark AB, Willson DS (1981) Avian breeding adaptations: hatching asynchrony, brood reduction, and nest failure. Q Rev Biol 56:253±277 Cramp S (1985±1992) The birds of the Western Palearctic, vols 5±7. Oxford University Press, Oxford Cramp S, Perrins CM (1993±1994) The birds of the Western Palearctic, vols 8±9. Oxford University Press, Oxford Davies NB, Brooke M de L (1989a) An experimental study of coevolution between the cuckoo, Cuculus canorus and its hosts. I. Host egg discrimination. J Anim Ecol 58:207±224 Davies NB, Brooke M de L (1989b) An experimental study of coevolution between the cuckoo, Cuculus canorus and its hosts. II. Host egg marking, chicks discrimination and general discussion. J Anim Ecol 58:225±236 Eichler W (1948) Some rules in ectoparasitism. Annu Mag Nat Hist Ser 12 1:588±598 Felsenstein J (1985) Phylogenies and the comparative method. Am Nat 125:1±15 Friedmann H, Ki€ LF, Rothstein SI (1977) A further contribution to the knowledge of the host relations of the parasitic cowbirds. Smithsonian Contrib Zool 235:1±75 Garland T Jr, Huey RB, Bennett AF (1991) Phylogeny and coadaptation of thermal physiology in lizards: a reanalysis. Evolution 45:1969±1975 Garland T Jr, Dickerman AW, Janis CM, Jones JA (1993) Phylogenetic analysis of covariance by computer simulation. Syst Biol 42:265±292 Gibbons DW, Reid JB, Chapman RA (1993) The new atlas of breeding birds in Britain and Ireland: 1988±1991. Poyser, London Gibbs HL, Brooke M de L, Davies NB (1996). Analysis of genetic di€erentiation of host races of the common cuckoo Cuculus canorus using mitochondrial and microsatellite DNA variation. Proc R Soc Lond B 263:89±96 Gill FB (1990) Ornithology. Freeman, New York Glue D, Morgan R (1972) Cuckoo hosts in British habitats. Bird Study 19:187±192 Glue D, Murray E (1984). Cuckoo hosts in Britain. Br Trust Ornithol News 134:5 Grafen A (1989) The phylogenetic regression. Philos Trans R Soc Lond B 205:581±598 Haartman L von (1981) Coevolution of the cuckoo Cuculus canorus and a regular cuckoo host. Ornis Fenn 58:1±10 Haldane JBS (1949) Disease and evolution. Ric Sci Suppl 19:68±76 Hartley IR, Davies NB (1994) Limits to cooperative polyandry in birds. Proc R Soc Lond B 257:67±73 Harvey PH, Pagel M (1991) The comparative method in evolutionary biology. Oxford University Press, Oxford Harvey PH, Nee S, Mooers Aé, Partridge L (1992) These hierarchical views of life: phylogenies and metapopulations. In: Berry RJ, Crawford TJ, Hewitt GM (eds) Genes in ecology. Blackwell, Oxford, pp 123±137 Howard R, Moore A (1991) A complete checklist of the birds of the world. Academic Press, London JaÈrvinen A (1984) Relationship between the ``common cuckoo'' Cuculus canorus and its host, the redstart Phoenicurus phoenicurus. Ornis Fenn 61:84±88 Keller L, Genoud M (1997) Extraordinary lifespans in ants: a test of evolutionary theories of ageing. Nature 389:958±960 Lack D (1963) Cuckoo hosts in England. Bird Study 10:185±201 Lack D (1968) Ecological adaptation for breeding in birds. Methuen, London Mason P (1986) Brood parasitism in a host generalist, the shiny cowbird. I. The quality of di€erent species as hosts. Auk 103:52±60 Maurer BA (1994) Geographical population analysis. Blackwell, Oxford Moksnes A, Rùskaft E (1995) Egg-morphs and host preference in the common cuckoo (Cuculus canorus): an analysis of cuckoo and host eggs from European museum collections. J Zool Lond 236:625±648 Moksnes A, Rùskaft E, Braa AT, Korsnes L, Lampe HM, Pedersen HC (1990) Behavioural responses of potential hosts towards arti®cial cuckoo eggs and dummies. Behaviour 116:64±89 Nakamura H (1990) Brood parasitism by the cuckoo Cuculus canorus in Japan and the start of new parasitism on the azurewinged magpie Cyanopica Cyana. Jpn J Ornithol 39:1±18 Nee S, Mooers Aé, Harvey PH (1993) The tempo and mode of evolution revealed from molecular phylogenies. Proc Natl Acad Sci USA 89:8322±8326 Noble ER, Noble GA (1976) Parasitology, 4th edn. Lea and Febiger, Philadelphia éien IJ, Moksnes A, Rùskaft E (1995) Evolution of variation in egg colour and pattern in European passerines: adaptations in a coevolutionary arms race with the cuckoo, Cuculus canorus. Behav Ecol 6:166±174 Owens IPF, Bennet PM (1994) Mortality costs of parental care and sexual dimorphism in birds. Proc R Soc Lond B 257:1±8 Pagel MD, Harvey PH (1989) Comparative methods for examining adaptation depend on evolutionary models. Folia Primatol 53:203±220 Payne R (1977) The ecology of brood parasitism in birds. Annu Rev Ecol Syst 8:1±28 Perrins C (1987) Aves de EspanÄa y de Europa (Nueva generacioÂn de guõ as). Omega, Barcelona Price PW (1980) Evolutionary biology of parasites. Princeton University Press, New Jersey Purvis A, Garland T Jr (1993) Polytomies in comparative analyses of continuous characters. Syst Biol 42:569±575 Riddiford N (1986) Why do cuckoos Cuculus canorus use so many species of host? Bird Study 33:1±5 Rothstein SI (1982) Successes and failures in avian and nestling recognition with comments on the utility of optimality reasoning. Am Zool 21:547±560 Rothstein SI (1990) A model system for coevolution: avian brood parasitism. Annu Rev Ecol Syst 21:481±508 Sibley CG, Ahlquist JE (1991) Phylogeny and classi®cation of birds: a study of molecular evolution. Yale University Press, New Haven SiikamaÈki P (1995) Are large clutches costly to incubate ± the case of the pied ¯ycatcher. J Avian Biol 26:6±80 Soler M, Mùller AP (1990) Duration of sympatry and coevolution between great spotted cuckoo and its magpie host. Nature 343:748±750 Soler JJ, Mùller AP (1996) A comparative analysis of the evolution of variation in appearance of eggs of European passerines in relation to brood parasitism. Behav Ecol 7:89±94 Wyllie I (1981) The cuckoo. Batsford, London