An experimental test of the causes of small-scale phenotypic differentiation in a population of great tits


D. Garant, Department of Zoology, Edward Grey Institute, University of Oxford, Oxford, OX1 3PS, UK.
Tel.: +44 1865 281999; fax: +44 18765 271168;


Phenotypic differentiation between populations is thought to occur mainly at spatial scales where gene-flow is restricted and selection regimes differ. However, if gene flow is nonrandom, dispersal may reinforce, rather than counteract, evolutionary differentiation, meaning that differences occurring over small scales might have a genetic basis. The purpose of this study was to determine the cause of differences in mean phenotype between two parts of a population of great tits Parus major, separated by <3 km. We conducted a partial cross-fostering experiment between two contrasting parts of this population to separate genetic and environmental sources of variation, and to test for gene–environment interaction. We found strong environmental effects on nestling size, mass and condition index, with nestlings reared in a low density part of the population being larger, heavier and in better condition, than those in a high density part, irrespective of their origin. In addition, we found smaller, but significant, differences in nestling condition and shape associated with the areas that birds originated from, suggesting the presence of genetic differences between parts of this population. There was no evidence of gene–environment interaction for any character. This experiment is thus consistent with previous analyses suggesting that differences between parts of this population had evolved recently, apparently due to phenotype-dependent dispersal, and indicates that population differentiation can be maintained over small spatial scales despite extensive dispersal.


Adaptive population differentiation is thought to be rare over small scales because of the homogenizing effect of gene-flow between populations (Endler, 1977; Slatkin, 1987). However, recent theoretical and laboratory work has suggested that speciation could occur in the presence of gene-flow, even between sympatric populations (the so-called ‘divergence-with-gene-flow’ model; Rice & Hostert, 1993). The model predicts that strong divergent selection pressures across different environments should counter the homogenizing effect of gene-flow and allow population differentiation to be maintained (see Smith et al., 1997 for an example). Recent theory has also supported the possibility for differentiation to occur along an environmental gradient with different phenotypes varying in fitness in different environments (Doebeli & Dieckmann, 2003).

Evidence of divergence despite gene flow at small spatial scales is limited, but there are some examples, potentially indicating that such processes may occur in natural populations. For instance, in two populations of blue tits (Parus caeruleus) separated by only 25 km, genetic differentiation is maintained by behavioural restrictions on gene-flow and strong differential selection between the two habitats (Blondel et al., 1999). Similarly, two populations of parasitoid waSPS (Diaeretiella rapae), separated by only 1 km, show extensive genetic differentiation between populations, mainly due to their differential use of two available aphid hosts (analogous to habitats) (Vaughn & Antolin, 1998). Lastly, Postma & van Noordwijk (2005) have shown that differences in clutch size between island sub-population of the great tit (Parus major) reflect the balance between local adaptation and immigration of nonadapted birds from the mainland.

More recently, a different role for gene flow in differentiation was shown in a population of great tits inhabiting a single continuous woodland (Garant et al., 2005). Using data from a long-term study of great tits P. major in Wytham wood, Oxfordshire, UK, Garant et al. (2005) showed that gene-flow could be nonrandom and could thus generate genetic differentiation despite extensive gene-flow. Specifically, birds in one part of this population increased in relative fledging mass over time, whereas birds in another part decreased in relative fledging mass. This differentiation was suggested to have been driven by differences in habitat quality, with larger, potentially socially-dominant birds settling in a preferred habitat (e.g. see Verhulst et al., 1997; Braillet et al., 2002). Gene-flow thus generated an association between certain phenotypes and certain habitats, with larger birds settling in a habitat where density was constrained to be lower by lower artificial nest site availability, which is presumably better quality habitat because of reduced inter and intraspecific competition (Minot & Perrins, 1986; Garant et al., 2005).

However, this previous evidence of phenotypic and genotypic differentiation because of nonrandom dispersal was based on an analysis of nonexperimental data (Garant et al., 2005), with limited phenotypic measures available, and the analysis did not exclude the possibility of gene-by-environment interactions in which the best genotype would be matched to better environments. Gene-by-environment interaction might also offer an alternative explanation for the mechanism by which differences between areas arise. Rather than the environment sorting individuals by genotype, the effect might be driven by a form of individual optimization, in which individuals settle in areas in which their genotypes will perform best. The goal of the current study was to disentangle and quantify genetic and environmental effects on morphometric traits for great tits breeding in the two most contrasting habitats of the previous study. We achieved this by performing a reciprocal partial cross-fostering experiment between the two habitats.

Materials and methods

Study area and experimental design

The study was conducted in April–June 2004 in Wytham Wood, Oxfordshire, UK (51°47′N, 1°20′W), where a great tit population breeding in nest-boxes has been studied since 1947, with nest-box numbers and locations kept constant since 1965. All breeding attempts in this population were monitored from the date of egg-laying until all nestlings had fledged by regular visit of nest boxes to determine the initiation of egg-laying, clutch size, hatching date and fledging date. The two contrasting parts of the population are henceforth referred to as the East (E) and the North (N). The East is the part of the population that has shown a consistent decline in relative fledging mass, whereas the North is the part that has shown the most consistent increase, when an overall negative effect of population density is controlled for (see Garant et al., 2004, 2005).

A total of 323 nestlings from 38 nests (19 in N; 19 in E) were subjected to partial cross-fostering (76 transferred from E to N; 76 from N to E; 75 left in E; 96 left in N). Using a technique similar to traditional cross-fostering experiments (e.g. Gebhardt-Henrich & Van Noordwijk, 1994; Smith & Wettermark, 1995), reciprocal transfers were carried out between broods in N and E with matching hatch dates and clutch size (±2 chicks) (Fig. 1). While aiming to maximize the number of cross-fostered young, we aimed not to transfer more than half a brood into a foreign nest to minimize possible effects of competition between resident and foster nestlings (van Noordwijk, 1988). All transfers were conducted on day 3 after hatching (hatch date = day 0). The nestlings were individually marked by clipping a unique combination of claw-tips. In order to avoid skewing the masses of transferred nestlings in any directions, the nestlings to be transferred were weighed, ranked for weight, and then sequentially assigned to be either cross-fostered or remain in the source nest. Furthermore, to minimize variation from extremely early and late (or second) broods, which often suffer from unfavourable environmental conditions (Gebhardt-Henrich & Van Noordwijk, 1991), transfers were all performed from nests that hatched during the ‘peak’ of hatching (in this case from 7 May to 17 May).

Figure 1.

Map of each nest box territory estimated from Theissen polygons and mean (with SE) territory area (m2) of breeding pairs of great tits in North and East areas involved in the cross-fostering. This measure is inversely related to the breeding density experienced by breeding birds.

Data collection

Fledging mass was measured on day 3 and day 14, and tarsus and wing measurements were taken on day 14 after hatching, at which time Great tit nestlings are expected to have achieved their final fledging body size (Perrins, 1979). All mass measurements were performed by the same person using a Pesola balance accurate to 0.1 g. Wing measurements, defined as the distance between the carpus and the tip of the longest primary feather, were taken to the nearest 0.5 mm with a standard 150 mm wing rule and minimum tarsus measurements were measured with digital calipers to the nearest 0.1 mm. Repeatability for the fledging mass measurements was 0.998 (P < 0.001), based on repeated measures and balanced one-way anova comparing within-individual and between-individual variation (Lessells & Boag, 1987). Repeatability was not estimated for tarsus and wing length, but these traits typically show repeatabilities in our study populations of >0.90 (B.C. Sheldon, unpublished data). Time of day was always recorded when collecting data. A single feather was taken from each nestling (under Home Office licence) for subsequent sex identification. All nestlings having survived to day 14 had successfully fledged by day 18–21.

Sex determination

Male great tits are on average larger than females (Björklund & Linden, 1993), and this size difference is apparent at fledging. In order to control for this difference, the 276 nestlings surviving at day 14 had DNA extracted from collected feathers using 5% Chelex 100 chelating resin (Walsh et al., 1991). Among these, 272 individuals were successfully sexed by polymerase chain reaction amplification of the sex linked genes CHD1-W and CHD1-Z (Griffiths et al., 1998).

Density estimates

One candidate cause for the different temporal trends in fledging mass demonstrated in Garant et al. (2005) is breeding density. To verify whether birds did indeed experience variable density among areas, Theissen polygons (Rhynsburger, 1973) were formed around each occupied nest box in Wytham in 2004 and their areas calculated (m2). Nest boxes were considered occupied if a confirmed breeding attempt was made by a pair of great tits. In this case, Theissen polygons contain all points closer to a given nest box than to any other nest box, such that polygon boundaries are equidistant between occupied nest boxes. The area of each polygon is thus inversely related to the intraspecific breeding density experienced by the breeding pair, and while these polygons are likely to over-estimate territory size in the case of birds breeding at very low densities, they nevertheless provide a reasonable index of population density.

Statistical analysis

All statistical analyses were performed using GenStat Release 7.2 (VSN Intl., 2003). We were specifically interested in the determinants of day 14 mass, but we also examined the condition index that was calculated by including tarsus length as a covariate in all analyses involving day 14 mass. We also performed Principal Component Analysis (PCA) from a correlation matrix (scaled) of the combination of day 14 mass, wing length and tarsus length. Such analysis finds the linear combinations of the traits that successively maximize the variation contained within them, collapsing original variability to a smaller number of dimensions thus allowing us to determine if there were differences among areas in the main principal components of the characters measured.

To identify factors contributing significantly to explaining variation in the above variables we used general linear mixed models (GLMM) with restricted maximum likelihood estimation procedures. In all cases, nest of origin was nested within nest of rearing and both were included as random effects in the model to control for nonindependence of chicks within nests. Habitat of origin (N or E) and habitat or rearing (N or E) and their interaction were included as fixed effects to separate the genetic (origin) and environmental (rearing) effects and to test for a potential gene-by-environment interaction. Sex was also included as a fixed effect. Tarsus length was included as a covariate in all analyses of condition index. Mass at day 3 was also included as a covariate, in order to control for any size asynchrony introduced by cross-fostering nestlings of slightly different ages, or of different sizes due to maternal effects via egg size. Sex-specific interactions were also tested a posteriori in each model. The significance of each factor as a predictor of each response variable was assessed from its Wald statistic, which is distributed as inline image (VSN Intl., 2003).

A GLMM with binomial error was used to assess variation in mortality between areas. The model included survival (0 = dead, 1 = survived) as the dependent variable, habitat of rearing and habitat of origin as fixed effects, and nest of origin and rearing as random effects to control for nonindependence of nest-mates. Again, the significance of the fixed effects was determined from their Wald statistic.


Starting conditions

There was no significant difference in nestling mass between different habitats of origin at the time of cross-fostering (inline image = 0.29, P = 0.59) and, as nestlings were exchanged after being ranked with respect to their day 3 mass, there was no difference in mass between the two rearing habitats at day 3 (inline image = 1.07, P = 0.30). The mean Theissen polygon area of the nest boxes where nestlings were raised in the North part of the wood was almost three times larger than those in the East (mean occupied nest box Theissen polygon area ± SE: east = 7604 ± 689 m2, north = 21456 ± 3143 m2, t19.7 = 4.30, P < 0.001; see Fig. 1), indicating (as expected) a much lower local breeding density in the North compared to the East.

Differential mortality among areas

A total of 47 nestlings died over the study period. Results from the GLMM analysis suggested that mortality was higher in the East than in the North (inline image = 2.87, P = 0.09), supporting the idea that this area is presumably a poorer environment, although the difference was not statistically significant. Any mortality differences that were present were only due to the environmental influence (habitat of rearing) as there was no difference in nestling mortality among habitat of origin (inline image = 0.00, P = 0.99).

Effects of origin and rearing environments

The GLMM analysis of Day 14 mass revealed a strong and significant effect of the habitat of rearing (Table 1; see Fig. 2a), with nestlings reared in the North being heavier than East-reared individuals (difference = 0.92 g ± 0.31 SE). There was no significant effect of the habitat of origin on fledging mass at day 14 (Table 1), even if North-born nestlings were larger than those born in the East (see Fig. 2a). Sex and day 3 mass also had a significant effect on day 14 mass (Table 1). Further analyses revealed that day 3 mass showed a significant interaction with sex on day 14 mass (the day 3 mass-day 14 mass relationship was less pronounced for males than for females: inline image = 4.41, P < 0.05).

Table 1.  Significance of the fixed effects and relative proportions of variance (% Var) explained by the random effects of the habitat of rearing and the habitat of origin (nested within rearing) included in general linear mixed models for (a) nestling mass at day 14, (b) condition index at day 14, (c) PC1-size at day 14 and (d) PC2-shape at day 14, for great tit nestlings cross-fostered between habitats. Two hundred and seventy-two individuals from 38 nests are included in the analyses.
 Habitat of rearing  8.861N > E0.003
 Habitat of origin  0.881N > E0.347
 Day 3 mass 68.901+<0.001
 Sex 44.441M > F<0.001
 % Var   
 Rearing 52.3   
 Origin  3.8   
 Habitat of rearing 10.401N > E0.001
 Habitat of origin  4.351N > E0.037
 Tarsus length 24.351+<0.001
 Day 3 mass 64.941+<0.001
 Sex 12.871M > F<0.001
 % Var   
 Rearing 47.4   
 Origin  4.3   
 Habitat of rearing  5.501N > E0.019
 Habitat of origin  0.361E > N0.547
 Day 3 mass138.051+<0.001
 Sex 80.761M > F<0.001
 % Var   
 Rearing 49.8   
 Origin  8.0   
 Habitat of rearing  1.771E > N0.183
 Habitat of origin 17.201E > N<0.001
 Day 3 mass 70.511<0.001
 Sex 37.511M > F<0.001
 % Var   
 Rearing 18.5   
 Origin  7.0   
Figure 2.

The effects of habitat of rearing and habitat of origin on the different traits measured in the cross-fostering experiment. (a) Mass at day 14 (g), (b) condition index (defined as mass at day 14 with tarsus length as a covariate in the model), (c) PC1 (size), (d) PC2 (shape). Values are least square means with their standard errors. Black squares: East-born offspring; White circles: North-born offspring.

The GLMM analysis of condition index also showed a significant effect of the habitat of rearing (Table 1). This effect was in the same direction as for fledging mass with offspring being raised in the North having a larger condition index than those in the East (difference between N and E = 0.89 g ± 0.28 SE; Fig. 2b). The habitat of origin effect was significant for this trait, indicating that offspring born in the North were relatively heavier than those originating from the East (difference between N and E = 0.23 g ± 0.11 SE; Table 1; Fig. 2b). Further analyses also revealed interactions between habitat of origin (inline image = 4.96, P < 0.05) and day 3 mass (inline image = 6.02, P < 0.05) with sex on condition index of nestlings. The interaction between habitat of origin and sex is largely due to the fact that females showed a significant difference in condition index between habitats, whereas males were similar (see Fig. 3a).

Figure 3.

The interaction between the habitat of origin and sex on nestlings from the cross-fostering experiment for (a) condition index (g) (defined as the mass at day 14 with tarsus length as a covariate in the model) and (b) PC2 (shape). Values are least square means with their standard errors. Black squares: Males; White circles: Females.

The principal component analyses returned a PC1 that was equally representative of all three body size traits (all with positive loadings: tarsus = 0.520, wing = 0.557 and day 14 mass = 0.648) and contributed to 60.4% of the variation (henceforth called PC1-size). The PC2 component was mainly representative of tarsus length (positive loading = 0.759) and wing length (negative loading = −0.649) and contributed to 25.9% of the variation (henceforth called PC2-shape).

The mixed model analysis of PC1-size showed that the habitat of origin effect was not significant, but that the habitat of rearing was significant (Table 1). In this case, birds reared in the North were larger than those reared in the East (difference between E and N = 0.71 ± 0.30; see Fig. 2c). There was also a marginally nonsignificant sex-specific interaction of PC1-size with day 3 mass (inline image = 3.75, P = 0.05). The analysis of PC2-shape, on the other hand, revealed that only the habitat of origin was significant (Table 1) with individuals from the North having significantly more negative values than individuals from the East (in other words having relatively shorter tarsi relative to their wing lengths; see Fig. 2d). There was significant sex-specific interaction of PC2-shape with habitat of origin (inline image = 3.86, P = 0.05) where the difference among females was again more important than between males (Fig. 3b). The interaction between habitat of rearing and habitat of origin was not significant in any of the previous models (all P ≥ 0.40; Fig. 2a–d).


Our results partially confirm earlier evidence of phenotypic and genotypic differentiation between North and East areas of Wytham wood (Garant et al., 2005). Importantly, we found no significant gene-by-environment interaction, suggesting that apparent genetic differences among areas are not likely to be produced by gene–environment covariance where ‘good’ genotypes are simply matched to ‘good’ habitats (see Kruuk et al., 2003). Although demonstrating the absence of gene-by-environment interaction ultimately requires showing that the genetic correlation for a given trait across habitats is not different from one, we believe that the fact that we found no significant origin-by-environment effects suggests that such interactions were limited in this study.

Our results showed a strong environmental effect on fledging mass, with nestlings reared in the North (an area of low population density) achieving larger mass at day 14 and a greater condition index. PC1 (size) was also influenced by the environment, suggesting that the environment shapes other aspects of the morphology than fledging mass. These significant environmentally induced differences show that environmental quality can vary greatly over very short distances (in this case only a few kilometres; see Fig. 1). This effect is perhaps most likely to be due to a density-related effect, as larger territory sizes were occupied in the North area (see Fig. 1), although other causes cannot be excluded without experimental manipulation. In fact, territory size had a significant positive effect on fledging mass (inline image = 4.76, P = 0.03), when tested in a GLMM controlling for nest of origin and nest of rearing as random effects and sex and mass at day 3 as fixed effects.

In addition to effects of rearing environment, however, we also found significant habitat of origin effects for some traits, suggesting that the differences between nestlings reared in these two areas is not purely environmental. The difference we observed for fledging mass at day 14, although not being significant, was in the range of the previously documented divergence. In the current study, the lower 95% confidence interval (CI) for nestling mass in the East was 17.95 g and the upper 95% CI in North was 18.70 g, which represent a potential difference of 0.58 phenotypic standard deviations (SD), not inconsistent with the 0.39 phenotypic SD difference among areas documented by Garant et al. (2005). Other lines of evidence also suggest the presence of a genetic differentiation in morphology among areas. First, chicks born in the North had a significantly better condition index in both rearing environments. Previous analyses, reported in Garant et al. (2005), were based on day 14 mass of nestlings as it was the only character available with sufficient sample size for analysis over the study period. Hence, it is not clear from that study whether mass was indeed the most important variable changing among areas. For example, the change could potentially have involved relative fledging condition which might be detected as a change in mass, due to the strong phenotypic correlation between these characters.

The technique employed here, that of partial cross-fostering, cannot exclude the possibility of maternal effects, or environmental effects occurring before the experimental transfer occurs. For example, females breeding in lower density areas might lay larger, better-nourished, eggs, an effect that might persist throughout development. Consequently, the most that we can safely conclude from the cross-fostering employed here is that the results are consistent with genetic differentiation among areas, even if this differentiation is not formally established. However, the agreement obtained by different quantitative genetic methods (in the present case: cross-fostering; in Garant et al., 2005 by an animal model analysis) suggests that this conclusion may be reasonable.

Analyses of variation in PC2 (shape) revealed a striking habitat of origin component, due to nestlings originating in the East having longer tarsi but shorter wings at 14 days of age compared with those born in the North. Sexual dimorphism in shape also differed between the two habitats, being more pronounced in birds originating from the North (low density population: Fig. 3b). The functional significance of these differences is not clear. One possibility is that within higher density areas, where food may be more limited, selection has favoured relatively faster development of tarsi relative to wings, if the possession of longer tarsi yields an advantage in competition for food within families.

While we aimed to match nests from the two parts of the wood for clutch size, clutch size of nests in the East were smaller than those in the North (East: mean = 7.95 ± 0.35; North: mean = 9.05 ± 0.27; t36 = 2.48, P < 0.05), as generally expected under high density conditions (Perrins & McCleery, 1989; Both, 1998). Larger clutches of origin also gave rise to larger fledglings (effect: 0.145 ± 0.052, inline image = 7.94, P < 0.01; when including clutch size of origin, while holding other terms constant, in the linear mixed model of day 14 mass; see Materials and methods). The existence of this positive relationship between clutch size and fledging size is surprising for two reasons. First, in this population, there is generally a negative effect of clutch size on mean fledging mass (when analysing all clutches from 1965 to 2000: r = −0.0923 ± 0.0264, P < 0.001; D. Garant, unpublished data). Secondly, these two traits are also negatively genetically correlated (−0.256 ± 0.120, D. Garant, unpublished data), so that smaller nestlings are expected to derive from larger clutches wherever these nestlings are grown. The positive relationship we found could thus indicate some degree of variation in maternal quality, potentially due to the presence of maternal effects in our sample, where larger females were able to lay larger clutches (see also Haywood & Perrins, 1992) that would ultimately produce larger chicks. The presence of such an effect makes it more difficult to clearly separate the genetic effect from the maternal component, particularly as maternal effects might vary depending on density (see Mazuc et al., 2003 for example).

Although many studies describing patterns of phenotypic differentiation among populations of birds have been published, few have been able to exclude the possibility that the observed among-population differences are attributable to phenotypic plasticity, rather than to genetically determined differences (see Merilä & Sheldon, 2001 for a review). The few studies that performed reciprocal transplants to specifically separate environmental and genetic effects had mixed outcome with some showing only environmentally-based differences (Komdeur, 1996) and others finding mixed results depending on the population studied (James, 1983). However, a reasonably strong genetic basis was found among two phenotypically divergent populations of Coal Tits (Parus ater), in Sweden, separated by 300 km and a marine barrier (Alatalo & Gustafsson, 1988). In this case, complete clutches were transferred between the two locations and tarsus length differences were best explained by the population of origin effect.

Our results were also supportive of previous suggestions that there is a real difference in habitat quality among areas. Indeed, we found that survival tended to be lower in the East area than in the North. More work is required to determine the exact causes behind this differential mortality, and whether this might ultimately lead to different selection pressures depending on the area. It is already known that the East is a much more dense habitat (Garant et al., 2005; this study), which probably results in greater competition for food (Minot & Perrins, 1986), but other factors might also directly affect the amount of food or the time available to gather food, such as density of habitat where prey species are located and predation pressures.

A closer look at the higher mortality in the East might also help to explain the sex-dependence of the genetic effect (habitat of origin) on fledging mass. Males often have higher nutritional requirements than females, and might suffer disproportionately in a habitat where food is scarce (Clutton-Brock et al., 1985; Teather & Weatherhead, 1989; see also Sheldon et al., 1998). Although further study is required to confirm any sex-biased mortality, our results suggest that such phenomenon is possible, as we found a significantly female-biased sex ratio in offspring from the East compared with the North (East: 51 males, 75 females; North: 73 males 73 females; inline image = 7.47, P < 0.01). The high density habitat is thus clearly lacking males, but because we did not sample chicks that died before day 14, we are not able to separate a potential sex-biased mortality from other alternative explanations, such as possible sex-biased maternal investment (West & Sheldon, 2002). The interaction between condition index and sex, however, potentially suggests that mortality targeted smaller males in the high density habitat (see Fig. 3a). As such, these smaller males would have been removed from the analysis, acting to reduce the potential magnitude of the genetic effect between areas.

In conclusion, we have demonstrated that nestlings from different areas within a single population may differ for both genetic and environmental reasons. Our findings are important for the understanding of local adaptation and ultimately speciation, and underline the importance of integrating fine scale environmental heterogeneity into such studies. While the environmental causes of these differences remains to be tested experimentally, the large differences in population density between the two sub-populations suggests that density, or differences in environmental quality causing different density populations to develop are ultimately the cause of these differences.


The authors would like to thank two anonymous referees for comments on the manuscript. During this work DG was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Postdoctoral Research Fellowship and by a Biotechnology and Biological Sciences Research Council (BBSRC) grant to BCS and Loeske E. B. Kruuk, and BCS by a Royal Society University Fellowship. BJS and TAW were respectively supported by scholarships from NSERC and BBSRC.