Parasitism, host immune defence and dispersal


Anders Pape Møller, Laboratoire de Parasitologie Evolutive, CNRS UMR 7103, Université Pierre et Marie Curie, 7 quai St Bernard, Case 237, F-75252 Paris Cedex 05, France.
Tel.: (+33) 1 44 27 25 94; fax: (+33) 1 44 27 35 16;


Host-parasite interactions have been hypothesized to affect the evolution of dispersal by providing a possibility for hosts to escape debilitating parasites, and by influencing the level of local adaptation. We used a comparative approach to investigate the relationship between a component of host immune function (which reflects the evolutionary history of parasite-induced natural selection) and dispersal in birds. We used a sample of 46 species of birds for which we had obtained field estimates of T-cell response for nestlings, mainly from our own field studies in Denmark and Spain. Bird species with longer natal, but not with longer breeding dispersal distances had a stronger mean T-cell-mediated immune response in nestlings than species with short dispersal distances. That was also the case when controlling for the potentially confounding effect of migration from breeding to wintering area, which is known from previous studies to be positively associated with dispersal distance. These relationships held even when controlling for similarity among species because of common ancestry. Avian hosts with a larger number of different breeding habitats had weaker mean T-cell-mediated immune responses than habitat specialists. This relationship held even when controlling for similarity among species because of common ancestry. Therefore, T-cell-mediated immunity is an important predictor of evolutionary changes in dispersal ability among common European birds.


Dispersal is a life history trait with important consequences for ecology and evolution of populations (Clobert et al., 2001). Dispersal will obviously affect the way in which individuals are distributed in space, which in turn will affect gene flow and genetic population structure. Ecological plasticity has been suggested to be associated with dispersal and the ability to locate new environments (Rosenzweig, 1975; Owens et al., 1999), while others have suggested that ecological plasticity may restrict dispersal (Murren et al., 2001).

Dispersal abilities will not only affect the number and identity of conspecifics with which an individual interacts, but also the intensity of such interactions, if individuals are not randomly distributed in space. Multiple causes for the evolution of dispersal have been proposed, including kin selection and kin competition (Gandon & Michalakis, 2001). Dispersal will also affect interactions with other species such as predators and parasites, and such interactions may have profound effects on the evolution of dispersal (reviews Boulinier et al., 2001; Van Baalen & Hochberg, 2001). Here we investigate the prediction that the evolution of long dispersal distances are associated with large impact of parasites on their hosts and hence on the evolution of strong immune responses.

Ecological studies of the highly colonial cliff swallow Petrochelidon pyrrhonota have shown that local levels of ectoparasitism in breeding colonies affect natal dispersal distances, with fledglings dispersing further when intensities of local parasitism are high (Brown & Brown, 1992). Similarly, colonially breeding kittiwakes Rissa tridactyla infested with ticks of the species Ixodes uriae have higher breeding dispersal, but not natal dispersal, when the prevalence of the breeding cliffs is high (Boulinier et al., 2001). In less social hosts, Sorci et al. (1994) found that offspring dispersal of the common lizard Lacerta vivipara depended on tick infestation of their mothers. Studies of great tit Parus major nests experimentally infested with hen fleas Ceratophyllus gallinae showed that male offspring were more likely to recruit locally than females, when raised in infested nests (Heeb et al., 1999). A tendency for greater dispersal distances of yellow-bellied marmots Marmota flaviventris infested with ectoparasites also suggested that parasites affect dispersal (Van Vuren, 1996). In contrast, Drosophila infected with nematodes was not related to dispersal under field conditions, despite the fact that parasites reduced host survival (Jaenike et al., 1995).

Long dispersal distances may have evolved in situations where local parasite impact on reproductive success is severe (Møller & Erritzøe, 2001). At the interspecific level we might predict a positive association between dispersal distance and immune response, as a direct link between parasite-induced mortality and host level of immune defence exists (Martin et al., 2001). Several potential examples exist of hosts moving out of the range of parasites (Freeland, 1980; Folstad et al., 1991), resulting in a decrease of parasite impact on host fitness.

The level of local adaptation in host–parasite systems has been hypothesized to depend on the relative rate of migration of hosts and parasites (Gandon et al., 1996; Kaltz & Shykoff, 1998). Based on the meta-population model of Gandon et al. (1996), host populations are expected to be locally adapted when parasites migrate more than hosts, while local mal-adaptation should occur when hosts migrate more than parasites. These predictions are supported by some empirical studies (Kaltz & Shykoff, 1998). However, migration rates of hosts may change in response to the presence of parasites, making the interacting parties change their relative benefits. For example, a study of interactions between the brood parasitic great spotted cuckoo Clamator glandarius and its magpie Pica pica host revealed little difference in migration rate between host and parasite in sympatry, while allopatric host populations had considerably smaller migration rates than sympatric host populations (Martínez et al., 1999). This suggests that host migration rate may change in response to the presence of parasites. Thus, long genetic migration distances by hosts (which may equate with long dispersal distances) are expected to provide hosts with a selective advantage in their coevolutionary interactions with parasites. Hence, we might predict that long dispersal distances should be associated with stronger immune responses of individual hosts.

The degree of ecological plasticity may affect host–parasite interactions because specialization by parasites on particular host species is more likely when the host is restricted to one or a few habitats (Tripet et al., 2002). Superficially, this might occur because only a few parasite species are encountered. However, that is not the case as many colonially breeding avian hosts in fact encounter more species of parasites than solitary species because mixed colonies of different host species with different specialist parasites commonly breed in the same colony (Tripet et al., 2002). As habitat specialization is associated with parasites specializing on particular hosts, a direct link between ecological plasticity and immune response can be predicted at the interspecific level. The sign of this hypothetical relationship would depend on the relative selection pressures exerted by parasites on hosts varying in ecological plasticity. A small range of habitats will increase the intensity of interactions between hosts and parasites, and may therefore result in specialization and speciation by the parasite counteracting host defences. Specific hosts should increase their defences against parasites because increases in host defence will select for increases in offence by the parasite. Therefore, a negative relationship between ecological plasticity and immune response is predicted at the interspecific level. A habitat generalist host is likely to encounter many different kinds of potential generalist parasites, and the rate of horizontal transmission of parasites among different species of hosts may therefore be larger in generalist than in habitat specialist hosts. Generalist parasites may be restricted to host species with relatively low levels of immune defence, because they are unable to cope with the strong defences of specialists. The coevolutionary process between parasites and hosts may be relatively slow when several hosts and parasites are involved (diffuse coevolution) in comparison to cases when only a couple of species are involved (strict coevolution) (Thompson, 1994). As it is likely that selection pressures on hosts exerted by specialist parasites are stronger than those exerted by generalist parasites, and habitat specialist hosts are likely to suffer more from specialist parasites than habitat generalist hosts, the relationship between ecological plasticity as reflected by the number of habitats and immune response should be negative.

The aims of this study were to assess the relationships between dispersal distance and immune function, using birds as a model system. Specifically, we investigated the interspecific relationship between dispersal distance and a component of immune response, after taking a number of potentially confounding variables into account. Information on dispersal distances for 75 bird species was reported by Paradis et al. (1998). As a measure of immune response, we used a standard measure of T-cell-mediated immune response to subcutaneous challenge of nestlings at a standardized age with a novel antigen (phytohaemagglutinin). A previous comparative study has shown that bird species that suffer from intense parasite-induced nestling mortality indeed have evolved stronger cell-mediated immune responses than species with little or no mortality (Martin et al., 2001). We included a measure of ecological plasticity estimated as the number of breeding habitats used by each species, according to a standard handbook on Western Palaearctic birds. Furthermore, previous studies have shown that migratory bird species have longer dispersal distances than resident birds (Paradis et al., 1998; Belliure et al., 2000). Hence, a measure of migration distance was included in the analyses. Finally, hole nesting and colonially breeding birds encounter parasites that induce higher degrees of nestling mortality than open nesting and solitarily breeding birds (Møller & Erritzøe, 1996; Møller et al., 2001). As a consequence, the former categories of species have evolved stronger immune responses than the latter species (Møller & Erritzøe, 1996; Møller et al., 2001). As hole nesting and coloniality may restrict the ecological plasticity of a species, we controlled for these variables when analysing the relationship between immune function and ecological plasticity.

Material and methods

Dispersal distance

We used dispersal distances of the bird species breeding in Britain, as reported by Paradis et al. (1998). This data set contains information on mean dispersal distances to the nearest kilometre, as recorded by the British Trust for Ornithology during an extended period 1909–1994. Natal dispersal was defined as the distance moved by individuals ringed in their year of hatching and recovered at breeding age, while breeding dispersal was defined as the distance moved by individuals ringed at breeding age. Detailed information about data compilation and calculations is provided by Paradis et al. (1998). We used the geometric mean dispersal distances in the present analyses.

T-cell-mediated immunity

We included all species for which information on T-cell-mediated immune response had been published by 31 July 2001 (three species C. glandarius, Luscinia svecica, Serinus serinus ; Soler et al., 1999; Johnsen et al., 2000; Hoi-Leitner et al., 2001), or for which we or our collaborators had recorded responses to the phytohaemagglutinin test (the remaining species). Only the tests on S. serinus (0.5 mg phytohaemagglutinin in 0.10 mL physiological water) and L. svecica (0.1 mg in 0.04 mL physiological water) were not performed in the same way as the test for the remaining species, but exclusion of these species did not change the findings for number of breeding habitats (there was no information on dispersal distance for these species). During the breeding seasons 2000–2001 we spent large parts of April to June searching for nests of birds, in which nestlings could be tested for cell-mediated immune response. This was done in Southern Spain around Granada and Sierra Nevada in 2000 and in Northern Denmark in 2000–2001. We requested information on nests from a number of amateur ornithologists with a good knowledge of birds (see Acknowledgments), and the nests recorded in this way were monitored until the test was being performed. As a measure of immune response we used the T-cell-mediated immune response to a challenge with phytohaemagglutinin. This is a standard estimate from the poultry literature of the ability to produce a T-cell-mediated immune response (Goto et al., 1978; McCorkle et al., 1980; Parmentier et al., 1993; Dietert et al., 1996). Injection with phytohaemagglutinin results in local activation and proliferation of T-cells, followed by local recruitment of inflammatory cells and increased expression of major histocompatibility complex molecules (Goto et al., 1978; Abbas et al., 1994; Parmentier et al., 1998). Before injection we removed the feathers from a small spot of skin on the wing web (patagium) of the right and the left wings and marked the sites of injection with a permanent, water-resistant colour marker. We then measured the thickness of the skin to the nearest 0.01 mm with a pressure-sensitive calliper (Teclock SM112, Japan). For each wing web we made three measurements to quantify measurement error. As in previous studies we found highly repeatable measurements, with repeatabilities above 0.95. This is similar to our previous studies of T-cell-mediated immunity (e. g. Saino et al., 1997; Soler et al., 1999; Martin et al., 2001; Merino et al., 2001). Subsequently, we injected 0.02 mg phytohaemagglutinin dissolved in 0.04 mL physiological water in one wing web, and 0.04 mL physiological water in the other wing web. Approximately 24 h later we remeasured the thickness of the skin at the two sites of injection, as described above. The index of cell-mediated immune response was simply calculated as the difference in thickness of the wing web injected with phytohaemagglutinin 24 h after and just before injection, minus the difference in thickness of the wing web injected with physiological water. Thus, the measure of response is expressed in millimetre. We calculated mean responses for each brood and then calculated an overall mean based on these brood mean values. The data set is presented in the Appendix.

Table Appendix.  Information on nestling T-cell response (mm), sample size, number of breeding habitats defined as the number of habitats listed in Cramp & Simmons (1977–1994), migration distance from breeding area to wintering area (in multiple of 500 km), nest site (0, open; 1, hole), coloniality (0, solitary; 1, max. 10 pairs in colony; 2, max. 100 pairs in colony; 3, max. 1000 pairs in colony), and body mass (g) for bird species included in the present study.
SpeciesT-cell response (mm)No. nestlingsNo. breeding habitatsMigrationNest siteColonialityBody mass (g)Natal dispersal (km)Breeding dispersal (km)
Acrocephalus scirpaceus0.395415120012.35.222.94
Alauda arvensis0.37991030042.70.300.17
Apus apus1.90245121237.611.472.63
Carduelis chloris0.82914900027.80.950.73
Clamator glandarius1.193277400167.0
Columba livia2.80884012393.0
Columba palumbus2.26168100490.02.281.81
Corvus corone2.14116110005703.240.93
Corvus frugilegus2.68574103488.01.960.65
Corvus monedula2.053102012246.02.130.72
Cuculus canorus1.1301151200106.0
Delichon urbica2.186614120314.53.190.69
Emberiza citrinella0.62841210026.52.030.16
Erithacus rubecula1.2383311018.20.570.36
Falco tinnunculus2.3481614200186.014.011.69
Fringilla coelebs0.7277820021.90.790.30
Hirundo daurica0.94023230019.0
Hirundo rustica1.26510734120216.23.190.56
Lanius excubitor1.2973720062.1
Luscinia svecica0.430717100017.0
Motacilla alba1.04791151021.05.160.89
Motacilla cinerea1.48811061018.015.80
Muscicapa striata1.398913120014.63.380.80
Parus caeruleus0.54313801013.30.800.23
Parus major0.62525801019.00.800.25
Passer domesticus1.328121001128.00.210.15
Passer montanus2.8109411122.00.720.28
Phoenicurus ochruros0.93216331016.4
Phoenicurus phoenicurus1.13288121014.73.87
Phylloscopus bonelli0.341838007.1
Pica pica1.556837000189.01.030.25
Pyrrhocorax graculus1.65992001245.0
Riparia riparia0.990787131314.66.651.22
Saxicola rubetra0.206610120016.9
Saxicola torquata0.63515920014.4
Serinus serinus0.61984410011.2
Sitta europaea1.3845601022.01.800.25
Streptopelia decaocto2.45555000152.02.500.57
Sturnus vulgaris1.23031021084.71.100.27
Sylvia atricapilla0.93361250015.517.548.03
Sylvia cantillans0.5976109008.5
Sylvia communis0.512813120014.52.821.15
Sylvia melanocephala0.7358610011.5
Sylvia undata0.348350009.5
Troglodytes troglodytes0.6255143008.91.170.53
Turdus merula1.3532711000113.00.260.22

We have found highly significant, consistent differences in cell-mediated immune response in nestlings among species when testing these differences with a one-way anova (calculations based on residual T-cell response from a regression of log-transformed T-cell response on log-transformed body mass: F65,604 = 6.375, SS (species) = 35.780, SS (residual) = 65.192, P < 0.001, data from Denmark 2001). This implies that there is considerably variation among species in relative immune response. Furthermore, we have found for 18 species with mean estimates available from both Spain and Denmark that there is significantly more variation among than within species in mean cell-mediated immune response (one-way anova based on residual T-cell response from a regression of log-transformed T-cell response on log-transformed body mass: F17,18 = 7.675, SS (species) = 1.623, SS (residual) = 0.224, P < 0.001). Nestlings were injected at a standard relative age during their ontogeny (when they were two thirds through their normal nestling period) rather than at a similar absolute age. This procedure ensured that nestlings were tested at a similar developmental stage, approximately equalling break of wing feathers through the feather pins, which reduces the effects of recorded responses being dependent on developmental age. Preliminary studies of age-dependent cell-mediated immunity in barn swallow Hirundo rustica nestlings have shown little variation in intensity of response during the period 10–16 days (in a species with a 20 days nestling period) (A. P. Møller unpublished data).

Although a thorough characterization of immunocompetence (the ability to raise an efficient response to a parasite attack) requires that T- and B-cell mediated and innate immunity are quantified (National Research Council, 1992), we suggest that a single measure recorded in a standardized across a range of species is superior to no measure at all. Furthermore, we note that at the interspecific level T-cell-mediated immune response is positively correlated with antibody production to a challenge with sheep red blood cells across a sample of 13 species of hirundines (Møller et al., 2001). Thus, two components of immunocompetence covary positively across species of birds.

Ecological plasticity

We recorded the number of breeding habitats by using the habitat preferences that Cramp & Simmons (1977–1994) reported for all species. These authors define the principal habitat categories in a glossary, and only use these for characterization of the habitat use of the species. We enumerated these habitat categories in the breeding habitat descriptions of each species. The index used was simply the total number of habitat categories reported. As our study was made independently of these habitat preferences published by Cramp and Simmons, we can assume that we have not introduced bias in the classification.

Ecological variables

Migratory bird species are known to disperse further than resident bird species (Paradis et al., 1998; Belliure et al., 2000). Hence, we quantified migration distance as the distance between the centre of the breeding distribution and the centre of the winter distribution. The centre was simply calculated as the mean value of the northernmost and the southernmost latitude of the breeding and the wintering distributions, respectively. This measure of distribution has previously been shown to provide a biologically meaningful measure (Gaston & Blackburn, 1996). The difference between these two mean latitudes was considered to represent the migration distance. These distances were recorded to the nearest 500 km on a map. Information on breeding and wintering ranges was obtained from Cramp & Simmons (1977–1994).

Bird species with hole nests suffer more from the effects of parasites than open nesting species, and they have stronger immune responses as shown by their larger relative size of bursa of Fabricius and spleen (Møller & Erritzøe, 1996). Likewise, hole nesters have more specific habitat requirements than open nesters, which can place their nests in many more different sites than hole nesters (Lack, 1968), and this may affect their ecological plasticity. Nest sites were classified as hole nests, when the nest is placed in a cavity, crevice or a similar structure, while all other species were classified as open nesters. Information was obtained from Cramp & Simmons (1977–1994).

Colonially breeding bird species suffer more from the effects of parasites than solitarily nesting species (Møller et al., 2001), and they have stronger immune responses as shown by their larger relative size of bursa of Fabricius and spleen and stronger T- and B-cell responses (Møller & Erritzøe, 1996; Møller et al., 2001). Species were classified as colonially nesting if more than a single pair regularly breed aggregated in space, while all other species were classified as solitarily nesting. likewise, colonial species breed in sites with specific requirements with respect to safety from predators and proximity to suitable foraging sites (Lack, 1968), and this may affect their ecological flexibility. Colony size was recorded on a logarithmic scale with maximum colony sizes of 10, 100, 1000 10 000 and 100 000 being scored as 1, 2, 3, 4 and 5. Information was obtained from Cramp & Simmons (1977–1994).

We used information on body mass from Dunning (1993). All data are reported in the Appendix.

Statistical methods

Information on natal and breeding dispersal was not available for all species, and sample sizes therefore differ among analyses. Mean geometric dispersal distances, cell-mediated immune response and adult body mass were log10-transformed before analysis to achieve distributions that did not differ from normal distributions. As we have found a consistent intraspecific difference in immune response between Denmark and Spain, we included a dummy variable in the preliminary analyses, by assigning a value of 0.0 to a mean immune response from Spain, a value of 0.5 to a mean immune response from Spain and Denmark, and a value of 1.0 to a mean immune response from Denmark. Inclusion of this variable did not change any of the conclusions presented in the results. Hence, our findings were robust with respect to the geographical origin of the individual data points.

Allometry effects of cell-mediated immune responses were controlled by using log-transformed body mass as a covariate in the analyses. Across species we found a significant positive relationship between log10-transformed immune response and log10-transformed body mass [F1,64 = 20.31, r2 = 0.241, P < 0.001, slope (SE) = 0.228 (0.051)]. However, a similar regression between log10-transformed immune response and log10-transformed skin thickness before injection was not significant [F1,64 = 2.25, r2 = 0.051, P = 0.08, slope (SE) = 0.202 (0.134)]. A multiple linear regression with log10-transformed immune response as the dependent variable and log10-transformed body mass and log10-transformed skin thickness before injection as independent variables only revealed a significant partial regression coefficient for body mass, but not for skin thickness [F2,63 = 11.14, r2 = 0.352, P < 0.0001, slope (SE) for body mass = 0.347 (0.079), t = 4.37, P < 0.0001; slope (SE) for skin thickness = 0.043 (0.118), t = 0.37, P = 0.716]. Use of log10-transformed body mass thus corrects efficiently for interspecific differences in body size without causing any bias due to initial thickness of skin among species.

Phenotypic mean values for species cannot be considered statistically independent observations because cases of convergent evolution are mixed with cases of similarity because of common ancestry. We calculated statistically independent linear contrasts for each variable according to the method developed by Felsenstein (1985). We used a composite phylogeny based on the phylogenies in Sibley & Ahlquist (1990), Sheldon & Winkler (1993), Blondel et al. (1996), Martin & Clobert (1996) and Leisler et al. (1997). The phylogeny is shown in Fig. 1. We adopted the software CAIC to make the calculations of contrasts (Purvis & Rambaut, 1995). All branches were assigned the same length, although a second set of analyses based on uneven branch lengths, assuming a gradual evolution model as implemented in the software by Purvis & Rambaut (1995), produced qualitatively similar results. We tested for violations of statistical assumptions by regressing standardized contrasts against their standard deviations (Garland et al., 1992). None of these tests revealed any significant deviations, after Bonferroni adjustment for multiple tests. Contrasts were analysed by forcing regressions through the origin (Purvis & Rambaut, 1995). Multiple regression models and forward stepwise multiple regression models were used to control for the effects of confounding variables.

Figure 1.

Composite phylogeny used for the comparative analyses. See Material and methods for sources.

Sample sizes differ among tests because information on dispersal distance was only available for a sub-sample of the species.


Natal and breeding dispersal distance were not significantly related to cell-mediated immune response in a multiple regression with log10-transformed body mass as an additional independent variable, when using species as independent observations [Fig. 2a, b; multiple linear regression: natal dispersal: F2,28 = 1.41, r2 = 0.091, P = 0.261, slope (SE) = 0.752 (0.454), t = 1.657, P = 0.109; breeding dispersal: F2,26 = 0.28, r2 = 0.021, P = 0.759, slope (SE) = 0.295 (0.407), t = 0.726, P = 0.474]. That was also the case when basing the calculations on contrasts [multiple linear regression: natal dispersal: F2,27 = 2.90, r2 = 0.172, P = 0.072, slope (SE) = 0.180 (0.082), t = 2.198, P = 0.036; breeding dispersal: F2,25 = 0.72, r2 = 0.052, P = 0.498, slope (SE) = 0.074 (0.122), t = 0.605, P = 0.550].

Figure 2.

Dispersal distance in relation to T-cell-mediated immune response of nestling birds. (a) Natal dispersal distance (km). (b) Breeding dispersal distance (km). (c) Residual natal dispersal distance after controlling for number of breeding habitats and log10-transformed body mass in a multiple regression analysis and residual T-cell response after controlling for the effect of log10-transformed body mass.

As migratory birds disperse longer than resident birds (Paradis et al., 1998; Belliure et al., 2000), we used migration distance as a third independent variable. Dispersal and migration distance was not significantly related to cell-mediated immunity (Table 1). A similar conclusion was reached by a forward stepwise linear regression analysis.

Table 1.  Multiple regression analyses with T-cell-mediated immune response as the dependent variable and natal dispersal distance, breeding dispersal distance, migration distance and body mass as independent variables.
SourceSlope (SE)tP
Species as observations
Model: F = 4.62, d.f. = 4,24, r2 = 0.435, P = 0.0066
 Natal dispersal0.172 (0.156)1.1070.279
 Breeding dispersal−0.145 (0.180)0.8050.429
 Migration0.039 (0.065)0.6030.552
 Body mass0.317 (0.086)3.6900.0011
Contrasts as observations
Model: F = 1.05, d.f. = 4,24, r2 = 0.149, P = 0.403
 Natal dispersal0.211 (0.128)1.6480.112
 Breeding dispersal−0.072 (0.150)0.4800.635
 Migration−0.026 (0.102)0.2560.801
 Body mass0.192 (0.174)1.1060.280

Cell-mediated immunity was smaller is bird species with a larger number of breeding habitats. When controlling for similarity because of common ancestry by using contrasts as independent observations we found a highly significant negative relationship between cell-mediated immunity and number of breeding habitats (Table 2). That was also the case when controlling statistically for the potentially confounding effect of hole nesting and coloniality on number of breeding habitats (Table 2). However, in the complete data set including species without information on dispersal distance there was no such relationship [Fig. 3a; multiple linear regression including body mass as an additional independent variable: F2,43 = 20.02, r2 = 0.482, P < 0.001, partial regression coefficient for number of breeding habitats: slope (SE) = −0.015 (0.008), t = 1.801, P = 0.079; analysis based on contrasts (F2,41 = 1.68, r2 = 0.076, P = 0.199, partial regression coefficient for number of breeding habitats: slope (SE) = −4.577 (3.312), t = 1.382, P = 0.175]. Thus, it was only when entering dispersal distance as a covariate that the relationship between number of breeding habitats and cell-mediated immunity became significant.

Table 2.  Multiple regression analyses with T-cell-mediated immune response as the dependent variable and number of breeding habitats, natal dispersal distance, migration distance, hole nesting, coloniality and body mass as independent variables.
SourceSlope (SE)tP
Species as observations
Model: F = 7.32, d.f. = 6,24, r2 = 0.647, P = 0.0002
 No. breeding habitats−0.020 (0.012)1.7430.094
 Natal dispersal0.103 (0.078)2.0850.048
 Migration−0.027 (0.055)0.4920.628
 Hole nesting0.026 (0.073)0.3630.719
 Coloniality0.136 (0.094)1.4440.162
 Body mass0.236 (0.077)3.0640.0053
Contrasts as observations
Model: F = 6.06, d.f. = 6,24, r2 = 0.603, P = 0.0006
 No. breeding habitats−0.029 (0.008)3.4180.0023
 Natal dispersal0.235 (0.077)3.0690.0053
 Migration−0.021 (0.062)0.3310.744
 Hole nesting−0.055 (0.067)0.8250.418
 Coloniality0.096 (0.084)1.1510.261
 Body mass0.214 (0.122)1.7600.091
Figure 3.

T-cell-mediated immune response of nestlings (mm) in relation to number of breeding habitats in birds. (a) Number of breeding habitats. (b) Residual number of breeding habitats after controlling for the effect of natal dispersal distance and log10-transformed body mass in a multiple regression analysis and residual T-cell response after controlling for the effect of log10-transformed body mass.

When cell-mediated immunity of the different species was entered as the dependent variable and both natal dispersal distance and number of breeding habitats as independent variables, immune response was significantly positively related to natal dispersal distance (Fig. 2c) and significantly negatively related to the number of breeding habitats (Fig. 3b; Table 2). That was even the case when using contrasts as independent observations (Table 2). Breeding dispersal distance was not significantly related to cell-mediated immunity when added as an independent variable to the multiple regression in Table 2 [F7,21 = 4.34, r2 = 0.592, P = 0.199, partial regression coefficient: slope (SE) = 0.010 (0.117), t = 0.084, P = 0.934], while the partial regression coefficients for natal dispersal and number of habitats remained statistically significant.


The main findings of the present study were that interspecific variation in a measure of T-cell-mediated immune response was positively correlated with variation in natal dispersal distance, but not with breeding dispersal distance. Furthermore, cell-mediated immune response was negatively correlated with the number of breeding habitats. Further analyses showed that they were independent of any relationship between migration and dispersal, and any relationship between ecological specialization and hole nesting or colonial breeding. The magnitude of the effect size measured in terms of Pearson's correlation coefficient [where r2 = t2/(t2 + d.f.) (Rosenthal, 1994)] was 28.2% of the interspecific variance in T-cell response explained by natal dispersal and 32.7% of the variance explained by ecological plasticity (data from Table 2). Effects are considered by Cohen (1984) to be strong, when the amount of variance explained is 25%. As all these analyses were correlational in nature, we do not invoke any causation in the relationships discussed here. Obviously, our measure of T-cell-mediated immune response is only one component of immunocompetence (National Research Council, 1992), but our approach should be feasible even for other components, although these will be more difficult to obtain for a large number of species under field conditions.

Natal dispersal distance is positively related to breeding dispersal distance across species (r = 0.781; Paradis et al., 1998; Belliure et al., 2000), although breeding dispersal is much shorter than natal dispersal. We found a positive relationship between natal dispersal distance and cell-mediated immunity (Fig. 2c), while a similar relationship was absent for breeding dispersal (Fig. 2b). Birds may develop adaptive immune responses to the parasite fauna of their breeding sites (and migration and wintering sites), and such phenotypic adaptation may be particularly prominent among juvenile and prebreeding life stages (Møller & Erritzøe, 2001). Once established in a breeding site, with well-developed immune responses to local genotypes of parasites, any further long-distance movement may be detrimental because of age-specific ‘vaccination’ and antibody production towards local parasite strains (Møller & Erritzøe, 2001). The positive relationship between cell-mediated immunity and natal dispersal distance, but not between cell-mediated immunity and breeding dispersal distance, is therefore expected.

It may seem surprising that we find so clear results, given that dispersal distances were recorded in Britain, breeding habitat preferences throughout the distributional range of the species, and cell-mediated immune responses mainly were recorded in Denmark and Spain. However, Belliure et al. (2000) have previously shown that natal dispersal distance is a good predictor of relative subspecies richness in birds, implying that dispersal distances of birds from Britain are predictors of phenomena well beyond the British Isles. Furthermore, we have data showing strong geographical differences in phytohaemagglutinin response of the same species in Denmark and Spain, although analyses of variance clearly indicate consistency in response among populations of the same species (see Materials and methods). Thus, we have shown for dispersal distance and cell-mediated immune response that estimates for different species are reliable predictors of ecological and evolutionary relationships beyond the populations investigated.

Ecological generalists were found to have weaker T-cell-mediated immunity than habitat specialists (Fig. 3). This relationship held even when taking nest site and breeding aggregation into account as potentially confounding variables. We predicted this negative relationship because a larger number of habitats would allow escape from specialist parasites. Ecological flexibility has been proposed to reflect colonizing ability and the probability that a taxon will diversify (Rosenzweig, 1975; Owens et al., 1999). As short dispersal distances are associated with a high degree of subspeciation and hence of diversification (Belliure et al., 2000), and as bird species with short dispersal distances have weak cell-mediated immune responses (Fig. 2), we can infer that a large number of breeding habitats indeed seems to reflect the probability of a taxon diversifying.

Belliure et al. (2000) found no significant relationship between the number of breeding habitats and dispersal distance in birds, suggesting that long distance dispersers do not have more opportunities to colonize a novel habitat. However, as short dispersal distances are associated with a higher degree of sub-speciation and hence of diversification (Belliure et al., 2000), and as the impact of parasitism measured as the level of cell-mediated immune response predicts both dispersal distance and number of breeding habitats (see Results), we can hypothesize that parasites have a dampening effect on diversification within host species and therefore possibly on speciation among hosts. Although the mere existence of subspecies as a biological entity has been disputed (Mayr, 1942, 1963), morphological differentiation can be considered to represent initial steps in divergence among populations (Houde & Endler, 1990). Thus, parasitism through its effects on the evolution of dispersal distances may reduce the rate of speciation in hosts.

In conclusion, we have provided evidence for relationships between natal dispersal distance, breeding habitat specialization and a component of immune function in birds. These findings suggest that parasites have either directly or indirectly played a significant role in the evolution of dispersal distances in common species of European birds.


W. C. Aarestrup, Ph. Christe, E. Flensted-Jensen, M. Houmøller, K. Klarborg, J. M. Marín, P. Pap and T. Pérez kindly helped to search for nests and/or make phytohaemagglutinin tests. Their efforts were indispensable for the success of our study. J. Clobert and T. Martin provided constructive criticism. M. M. V. was supported by a grant from Fundación Caja de Madrid.