Genetic Variation and Causes of Genotype-Environment Interaction in the Body Size of Blue Tit (Parus caeruleus)

Corresponding author: Juha Merilä, Section of Population Biology, Department of Zoology, Uppsala University, Villavägen 9, S-752 36 Uppsala, Sweden, juha.merila{at}zoologi.uu.se (E-mail).

Communicating editor: L. PARTRIDGE

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In several studies of natural populations of birds, the heritability of body size estimated by parent-offspring regression has been lower when offspring have developed in poor feeding regimens than when they developed in good feeding regimens. This has led to the suggestion that adaptation under poor regimens may be constrained by lack of genetic variation. We examined the influence of environmental conditions on expression of genetic variation in body size of nestling blue tits (Parus caeruleus) by raising full sibs in artificially reduced and enlarged broods, corresponding to good and poor feeding regimens, respectively. Individuals grown in the poor regimen attained smaller body size than their sibs grown in the good regimen. However, there was among-family variation in response to the treatments—i.e., genotype-environment interactions (GEIs). Partitioning the GEI variance into contributions attributable to (1) differences in the among-family genetic variance between the treatments and (2) imperfect correlation of genotypic values across treatments identified the latter as the main cause of the GEI. Parent-offspring regressions were not significantly different when offspring were reared in the good environment (h² = 0.75) vs. when they were reared in the poor environment (h² = 0.63). Thus, there was little evidence that genetic variance in body size was lower under the poor conditions than under the good conditions. These results do not support the view that the genetic potential for adaptation to poor feeding conditions is less than that for adaptation to good conditions, but they do suggest that different genotypes may be favored under the different conditions.

GENOTYPE-environment interaction (GEI) exists when different genotypes have different phenotypic responses to environmental variation (e.g., STEARNS 1992 ; VIA 1993 , VIA 1994 ). From the evolutionary point of view, GEI is important for two reasons. First, it can permit populations to evolve ultimately to have the optimum phenotypic mean in different environments, thereby promoting adaptation to heterogeneous environments (VIA and LANDE 1985 , VIA and LANDE 1987 ). However, in certain situations, GEI can also slow down the rate of adaptation of a population to a variable environment in the short run (FRY 1996 ). Second, GEI can lead to maintenance of genetic variation (GILLESPIE and TURELLI 1989 ), and thereby at least partly help to explain the high genetic variance in many fitness-related traits. Consequently, the amount of genetic variance in the slope of reaction norms (i.e., GEI) may be a strong determinant of the fitness of the population and ultimately of its survival. A lack of variation may lead to a failure to respond adaptively to environmental changes and ultimately to extinction. It is therefore of fundamental importance to investigate the amount and proximate causes of genetic variation in reaction norms in natural populations. Apart from the pioneering studies in amphibians (e.g., BERVEN 1987 ; NEWMAN 1988 ), and a few studies of birds (GEBHARDT-HENRICH and VAN NOORDWIJK 1991 ; GEBHARDT-HENRICH and VAN NOORDWIJK 1994 ; PRICE 1991 ; SMITH and WETTERMARK 1995 ; MERILA 1996 , MERILA 1997 ), there have been almost no attempts to evaluate the significance of GEIs in natural vertebrate populations. The lesson from these studies is mixed: some have found evidence for the presence of GEI (GEBHARDT-HENRICH and VAN NOORDWIJK 1991 ; PRICE 1991 ; MERILA 1997 ), while others have not (SMITH and WETTERMARK 1995 ; MERILA 1996 ). Additional evidence for GEI in natural bird populations comes from studies showing lower heritability estimates of morphological traits when offspring are reared in poor environments (low food supply) than when they are reared in good environments (see MERILA 1997 for a review; LARSSON et al. 1997 ). This had led to the suggestion that the potential for adaptation to poor environments is lower than the potential for adaptation to good environments (LARSSON 1993 ; MERILA 1997 ). However, interpretation of heritability estimates obtained by parent-offspring regressions when offspring are reared in different environments must be made with caution (LANDE 1987 ; RISKA et al. 1989 ). If offspring of the same or nearly the same set of parents are studied in both good and poor environments, then the denominator of the midoffspring-midparent regression, which is the variance of the midparental values, is the same regardless of offspring environment (MERILA 1997 ). Any difference in the regressions between good and poor environments must therefore involve the numerator of the regression, which is the covariance of midparent and midoffspring values. When parental and offspring environments are the same, the numerator estimates half of the additive genetic variance (FALCONER and MACKAY 1996 ). If parental and offspring environments differ, as they must in at least one case if two offspring environments are involved, then the regression numerator estimates half the additive genetic covariance of the trait between parental and offspring environments. This genetic covariance is the covariance of breeding values for the trait as expressed in the different environments. It may be written as _poApAo, where _po is the additive genetic correlation across parental and offspring environments, _Ap is the square root of the additive genetic variance in the parental environment, and _Ao is the square root of the additive genetic variance in the offspring environment. If there are two offspring environments, denoted by o1 and o2, any difference in the covariances implies that either _{po1 po2} or _{Ao1 Ao2}. Hence, a lower regression in one offspring environment than in the other could reflect a difference in the genetic correlations, not a difference in the additive variances. Either of these inequalities implies that GEI in the formal sense exists, but only in the case of a difference in the additive variances can one conclude that the population has lower potential to adapt to the poor environment than to the good environment. The main aims of this study were (1) to investigate how different environmental conditions (brood size) affect the causal components of phenotypic variance in blue tit body size, (2) to investigate the importance of the component of variance attributable to GEI in determining total phenotypic variance in body size of this species, and (3) to distinguish between the different possible causes of GEI—namely, cross-environment genetic correlations of less than one and differences between genetic variances among environments. In addition, by using cross-fostered nestlings, we asked to what degree the resemblance between offspring and parents might be of environmental origin.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The study species and population:
The blue tit (Parus caeruleus) is a small, cavity-nesting passerine that inhabits a wide range of habitats, from mixed deciduous forests to urban parks, with a clear preference for rich old deciduous forests (CRAMP and PERRINS 1993 ). It is a partial migrant in continental Europe; most juveniles disperse far from their natal areas, while adult birds are territorial and remain on breeding grounds all year round. Blue tits readily accept nest-boxes as breeding sites and, in our population, lay only one clutch annually. Although polygamy is common in some parts of the breeding range (DHONDT 1987 ), in our study area 99% of the males were monogamous (J. MERILÄ, unpublished data). The mean clutch size in first clutches is about 11 eggs (range 4–17 eggs), and incubation is by the female alone. The male feeds the female during incubation, and when the young hatch, both parents feed the nestlings. The nestling period is 16–20 days, and the young reach independence about 2 weeks after fledging.

We studied a breeding population of blue tits on the southern part of the island of Gotland (about 57°10' N, 18°20' E), off the east coast of Sweden, during April–June 1993–1995. Our study area consisted of 16 wood lots, which were mostly rich deciduous forest dominated by oak (Quercus robur) and ash (Fraxinus excelsior), sometimes with a dense understory of hazel (Corylus avellana). Some parts of the area were mixed coniferous forest, and a few nest-boxes were placed in suburban gardens located close to rich deciduous woodlands.

The data:
The data were collected during regular nest-box inspections beginning in late April until all young had fledged by mid-June. For each brood, the date of clutch initiation, clutch size, and the date of hatching were recorded. Except for 1995, when all birds were measured by another person, J.M. measured the tarsus length of all 14-day-old young (to an accuracy of 0.1 mm) using digital calipers as detailed in MERILA 1997 . By this time, the nestling tarsus length has usually attained its final size (DHONDT 1982 ), as confirmed by repeated measurement of individuals first measured as nestlings in 1992 or 1993, and their remeasurement as adults in 1993 or 1994. There was no systematic change in tarsus length after fledging (mean change ± SE = 0.002 ± 0.051 mm, paired t-test: t₅₀ = 0.04, P = 0.97), and the repeatability (i.e., intra-class correlation; SOKAL and ROHLF 1981 ; see also BAILEY and BYRNES 1990 ) calculated using nestling and adult measurements was high (R = 0.83, F_25,26 = 10.48, P < 0.001). However, in 1994, when the growing conditions for nesting tits were suboptimal, nestlings in most broods failed to develop normally. Hence, in 1994, most 14-day-old nestlings were underdeveloped both in terms of plumage (J. MERILÄ, personal observation) and tarsal growth. The latter was confirmed by repeated measurements of nestlings from 22 families, first measured at day 14 post-hatch, and then remeasured 2–4 days after this. In all different experimental groups, nestlings continued their growth [average increment ±SE: 0.32 ± 0.11 mm, 0.25 ± 0.03 mm, and 0.56 ± 0.15 mm for reduced (n = 6), control (n = 7), and enlarged (n = 9) broods, respectively]. Hence, though nestlings were still growing, there was no evidence that average growth increments would have differed among experimental groups (one-way ANOVA: F_2,19 = 1.80, P = 0.19). Comparison of repeated measurements of females measured twice during the breeding season in 1993 (~3 weeks apart) and of adults measured both in 1993 and 1994 (~1 year apart) further confirmed the high repeatability of measurements (1993: R = 0.88, F_34,35 = 16.07, P < 0.001; 1993–1994: R = 0.91, F_15,16 = 22.39, P < 0.001). Likewise, the repeatability of measurements in 1995, calculated using the same birds captured twice, was high (R = 0.89, F_13,14 = 17.58, P < 0.001). However, comparison of birds that J.M. measured in 1993 or 1994 with the measurements made in 1995 revealed that although the measurements were highly correlated (r = 0.84, P < 0.001), the measurements of tarsus length from 1995 were on average 0.39 mm longer than those made in previous years (paired t-test: t₃₈ = 8.12, P < 0.0001). Hence, a correction was made for all measurements made in 1995 by subtracting the average among-year difference (0.39 mm) from all 1995 measurements.

Mothers were captured for the first time when they had incubated eggs for about 9 days (1993–1994), and for the second time with the putative father when feeding the 2-wk-old young. Hence, the family relationships are based on the assumptions that (1) the adults feeding the young in a particular nest-box were their true parents and (2) all nestlings originally in the same nest-box were full-sibs. However, as extra-pair paternity (EPP) is known to occur in this species (KEMPENAERS et al. 1992 ), some of the presumed full-sibs might have been half-sibs not related to their putative father. However, one way to evaluate the possible importance of EPP is provided by comparison of father-offspring and mother-offspring regressions (cf. ALATALO et al. 1989 ) because EPP should result in lower father-offspring than mother-offspring regressions.

Cross-fostering and brood size manipulation experiments:
To separate genetic and environmental causes of resemblance, we performed reciprocal cross-fostering experiments, creating broods that consisted of approximately equal numbers of nestlings from two different families (Figure 1). As the aim was also to evaluate how different growth conditions might affect heritability estimates and causal components of variance, brood size was simultaneously manipulated by reducing or increasing the original clutch size by about 1/3. This was accomplished by moving about 2/3 of the young from a "reduced" brood to an "enlarged" brood and switching back 1/3 of the young in the recipient (enlarged) nest to the donor (reduced) nest (Figure 1). Hence, both reduced and enlarged broods consisted of foreign and own young in approximately equal numbers. The experimental pairs of nests were created by matching two nests according to their hatching dates and clutch sizes. However, as it was not always possible to match nests by these criteria, some exchanges were made between nests that differed in clutch size by (at most) two eggs, but never between nests that differed in their hatching date. Most exchanges took place among nests situated 200–1500 m apart, but some exchanges took place among nests situated up to 5 km apart. All young were banded with aluminum rings at the age of 6 days; until this age, foreign young were made identifiable by painting their claws or clipping some of the downy feathers on their head. In 1993, 1994, and 1995, 32, 23, and 29 pairs of broods, respectively, were manipulated; hence, in total, 168 families were subjected to experimentation. However, to reduce variation caused by differences in the genetic constitution of the two experimental groups, only those nests where at least one own and one foreign young survived in both nests of the pair (= dyad) were included in the analyses. Therefore, after excluding all dyads where one nest was lost because of predation, death of the whole brood, death of all young from one family, 29, 19, and 25 pairs of experimental broods were available for analyses in 1993, 1994, and 1995, respectively. To check that the transfer itself did not affect the size test nestlings attained, we also performed swaps where young were exchanged without changing the original brood size (Figure 1). Thus, these nests served as controls to evaluate possible effects of moving. Using the same criteria as for experimental broods, 13 and 6 pairs of control broods were available for analysis in 1993 and 1994, respectively. No control broods were created in 1995. There was no difference in initial clutch size between reduced and enlarged broods (Table 1), although clutch size in control broods in 1993, but not in 1994, was significantly smaller than in experimental broods (Table 1). However, reduced broods contained significantly fewer young both immediately after manipulation (day 2 post-hatch) and at day 14 post-hatch, than control broods, while the enlarged broods contained significantly more young than reduced and control broods—except in 1994, when enlarged broods experienced heavy mortality (Table 1). Despite the heavy mortality in 1994, and significantly lower fledging success in 1995 as compared with other years, the results in Table 1 suggest that our manipulations were successful in creating differential growth environments, as attempted.

View larger version (44K): In this window In a new window Download PPT slide   Figure 1. —Schematic representation of the cross-fostering protocol in control (no brood size manipulation) and experimental (brood reduction or enlargement) broods. Each box represents a nest-box, and the arrows indicate the origin (upper row) of the young in the resultant broods (lower row).

Full-sib analyses:
We used three different types of full-sib analyses to investigate the relative importance of genetic and environmental factors in determining nestling size. First, following MERILA 1996 , MERILA 1997 we performed mixed-model analyses of variance on the data from both offspring environments. In the full model, the main effects were Year and Dyad within Year (both random), Experiment (reduced or enlarged broods, a fixed effect), and Box of origin within Dyad and Year (random). We also analyzed reduced models for each of the different years to reveal possible among-year heterogeneity in different effects. Second, we performed nested, random-effect ANOVAs for each offspring environment separately, using Box of rearing and Box of origin (nested within Box of rearing) as factors. In this analysis, the box-of-origin effect estimates 1/2 of the additive genetic variance (V_A), plus 1/4 of the dominance variance and maternal effects if present. Third, we used the control broods to perform a two-way nested (random effects) analysis of variance (ANOVA) following ATCHLEY and RUTLEDGE 1980 to investigate GEI in response to differences between nest-boxes within dyads. In this analysis, the terms Box of rearing, Box of origin and their interaction were nested within Dyads. ATCHLEY and RUTLEDGE 1980 give a fuller account of the biological interpretation of the effects in this model.

All full-sib analyses were performed using type III sums-of-squares as obtained from PROC GLM in SAS (SAS INSTITUTE, INC. 1989 ). The variance components for all effects were calculated by equating the observed mean squares with expected mean squares; and, if they occurred, negative variance components were set to zero. The "variance" explained by the experimental environment was also calculated, although the environments were considered a fixed effect as a result of their nonrandom sampling. The estimates based on type III sums-of-squares were usually in good agreement with estimates derived by restricted maximum-likelihood methods.

Cross-environment genetic correlations:
The genetic correlation across environments quantifies the degree to which expression of a trait in one environment shares a heritable genetic basis with the expression of the same trait in a different environment (VIA and LANDE 1985 ; FALCONER and MACKAY 1996 ). There are two ways of estimating the cross-environment genetic correlation from the experimental design used in this study. First, cross-environment genetic correlations in each year can be estimated as

(1)

FRY 1992

ROBERTSON 1959

(2a)

Here, the numerator includes the average of the genetic variances in the two environments [(V_BOXO,RED + V_BOXO,ENL)/2], and the GEI variance (V_BOXOxEXP) estimated by the Experiment x Box of origin interaction (Table 4 or Table 5). The denominator is the same as in the first method. This equation can be also written as (ROBERTSON 1959 )

Parent-offspring regressions:
Parent-offspring regressions (using both midparental and single-parent values) were used to compare heritability estimates for tarsus length in the reduced and enlarged offspring environments (FALCONER and MACKAY 1996 ). Because the same set of parental values was used for offspring in the two environments, we used multivariate regression (SAS INSTITUTE INC. 1989) to take into account the correlation between the offspring of the same parents in the two environments. The multivariate regression procedure was used to test two null hypotheses (1) that the two slopes are equal and (2) that both of the slopes equal zero. We did not perform parent-offspring regressions for control broods, as the sample sizes were too small for any reasonable inference. To combine data from different study years, we first transformed midoffspring and midparent values to zero mean within each of the study years, leaving the variances unchanged. As the single-parent-midoffspring estimates of heritability can be biased by assortative mating or unequal variances between the two sexes, these possible sources of bias were checked. There was no assortative mating with respect to tarsus length in any of the study years when analyzed separately (1993: r = -0.01, n = 55, P > 0.10; 1994; r = -0.14, n = 30, P > 0.10; 1995: r = 0.10, n = 45, P > 0.10) or together (r = 0.03, n = 130, P > 0.10), and the variances of male and female tarsus length were homogeneous (F_129,129 = 1.103, P = 0.29).

All analyses were performed with the SAS statistical package (SAS INSTITUTE, INC. 1989 ).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Effects of cross-fostering:
To check that the exchange of young among nests did not affect the size they attained, we first compared the mean tarsus length of the fostered and nonfostered full sibs that were raised in control broods. There was no evidence that fostered and nonfostered young differed in tarsus length in either of the years for which data on control broods were available (Table 2). Hence the moving itself did not affect the size young attained.

Full-sib analyses of control broods:
Two-way nested ANOVAs performed for control broods revealed that under "normal" conditions, environmental (Box of rearing, Dyad) and GEI effects (Box of rearing x Box of origin) were not significant and always accounted for less than 5% of the phenotypic variance in body size (Table 3). Indeed, the only significant effect in these analyses was that attributable to Box of origin (Table 3), which is assumed to estimate 1/2 of the additive genetic variance (ATCHLEY and RUTLEDGE 1980 ).

Full-sib analyses of experimental broods:
We first investigated factors affecting offspring size by performing a four-factorial, mixed-model ANOVA, which revealed that only the main effect of Box of origin was significant, estimating 1/2 of the additive genetic covariance of tarsus length across the two offspring environments, whereas year, experiment, and dyad effects did not account for any significant proportion of variance (Table 4). All of the interactions in the model were significant (Table 4). The significant Year x Experiment interaction shows that although nestlings raised in enlarged broods attained shorter tarsi than their full-sibs raised in reduced broods in each of the 3 study years (Figure 2), the magnitude of this effect differed between the years. The significant Dyad(Year) x Experiment effect can be interpreted as a common-environment effect of Box of rearing, as each experimental treatment (reduced or enlarged) is represented by a single rearing box per dyad. Hence, there was between-dyad heterogeneity in the success of the experiment in affecting tarsus length. The significant Experiment x Box of origin interaction suggests that there were differences among families in how the young responded to different experimental conditions—i.e., genotype-environment interactions, or equivalently, genetic variation in slopes of reaction norms (Table 4). However, this effect accounted for less than 5% of all phenotypic variance in tarsus length (Table 4). To investigate these effects further, we repeated the analysis for each of the study years separately. These analyses revealed that in each of the study years, the brood-size manipulations (= experiment) had a significant effect on offspring tarsus length (Table 5). Likewise, genetic influences were strong in all study years, as revealed by the significant effect of box of origin (Table 5). There was statistical evidence for significant GEI (Box of origin x Experiment interaction) only in 1993, although a similar tendency (P < 0.10) was observed in 1995 (Table 5). In all cases, however, the variance components associated with the Experiment x Box of origin interactions were small compared to the Box of origin main effects (Table 4 and Table 5). There were also significant Experiment x Dyad interactions in two of the three years. The effect of Dyad was never significant, as could be expected since the families were paired at random with respect to their phenotypic size.

View larger version (40K): In this window In a new window Download PPT slide   Figure 2. —Mean tarsus length (± SE) of blue tit nestlings in two experimental environments (reduced and enlarged broods) in 3 different study years. Within each year, the two means are based on mean values of two full-sib groups raised either in reduced or enlarged broods. In each of the years, the experiments' main effect was significant (Table 5), but there was also a significant Year x Experiment interaction in the combined analysis (Table 4), showing that the effect of the experiment differed in different years.

Full-sib heritabilities:
To estimate environment-specific heritabilities and variance components from the full-sib data, we performed separate ANOVAs for each of the experimental environments (Table 6). Box of origin variance components, reflecting 1/2 the additive genetic variance plus dominance and maternal-effect contributions (if present), were significant and fairly similar between the two treatments (Table 6). However, residual (error) variance components were always much higher in the enlarged than in the reduced environment. Box of rearing variance components, reflecting common-environment effects, varied considerably but tended to be larger in the enlarged environment (Table 6). Consequently, estimated heritabilities were always higher in the reduced environment than in the enlarged environment (Table 6).

Cross-environment correlations:
Cross-environment genetic correlations (r_g) estimated using Equation 1 (see MATERIALS AND METHODS) were 0.65, 0.82, and 0.91 for 1993, 1994, and 1995, respectively. These estimates are in good agreement with those obtained with Equation 2a, which gave 0.64, 1.0, and 0.89 for 1993, 1994, and 1995, respectively. Notably, the lowest estimates occurred in 1993, when there was evidence for significant GEI from the mixed-model ANOVA (Table 5). Furthermore, combining the data from all years gives estimates of cross-environment r_g of 0.80 and 0.82 with Equation 1 and Equation 2a, respectively. Solving the two terms in Equation 2b and expressing them as a percentage of the total GEI variance (Table 4 and Table 5), the proportional contributions of cross-environmental correlation terms to GEI variance were 82%, 0%, 83%, and 87% in 1993, 1994, 1995, and combined data, respectively. Hence, these estimates indicate that the cause of significant GEI from mixed-model ANOVA in 1993 (nearly significant in 1995 and in combined data) was lack of genetic correlation between the environments, and not differential expression of genetic variance. If so, we would expect that genetic variances for the two offspring environments estimated by parent-offspring regression should be similar (see introduction).

Parent-offspring regressions:
Before examining whether parent-offspring regressions would reveal the same difference in heritability estimates between the two experimental environments as the full-sib analyses did, or whether offspring-parent covariances differed for offspring grown in different environments, we first investigated whether there was any suggestion that a shared environment between parents and offspring might have increased their resemblance. By regressing the midoffspring tarsus length of cross-fostered offspring against the tarsus length of their biological parents and foster parents, we found no similarity between the offspring and foster-parent tarsus lengths, while in both experimental environments, resemblance between offspring and their biological parents was high and significant (Table 7). Likewise, there was no suggestion that EPP would have been frequent (cf. ALATALO et al. 1989 ), as the contributions of both sexes of parents to offspring size were about equal (Table 8). Therefore, in our subsequent tests, we have combined the cross-fostered and home-grown young into the same analyses to increase the power of the tests.

Midparent-midoffspring regressions performed using the same set of biological parents for the two groups of full sibs raised in different environments revealed that the heritability estimates were lower in the "poor" offspring environment than in the "good" offspring environment, although not significantly so (Table 9). There was no evidence for nonlinearity in parent-offspring resemblance (cf. GIMELFARB and WILLIS 1994 ), as the quadratic components in each of the regressions (cf. Table 9) were far from significant (F 0.94, P > 0.34, in all cases). However, even when the data from different years were combined, there was no significant difference in the regression slopes between the two treatment groups (Table 9).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Genotype-by-environment interaction:
In this study, we found evidence for significant GEI in body size of nestling blue tits in 1 of the 3 study years and in the combined data. Hence, although the overall variance due to GEI always accounted for less than 10% of all phenotypic variance, there appeared to be some genetic variation in the sensitivity to environmental effects among different blue tit families. Dissection of the GEI variance into its causal components (ROBERTSON 1959 ) revealed that the interaction was apparent due primarily to the genetic correlation between the reduced and enlarged environments being less than one, and not due to lowered additive genetic variance in the enlarged environment. Hence, given the equal genetic variances under poor and good environments, the results reported here do not support the view that the genetic potential for adaptation under poor feeding conditions in natural bird populations is less than the genetic potential for adaptation under good feeding conditions (cf. MERILA 1997 ). Still, however, it is worth emphasizing that the lower mean fitness in poor environments might tend to make selection in poor environments less important in determining how the population evolves (KAWECKI et al. 1997 ).

The presence of GEI may allow adjustment of development toward the trajectory that maximizes fitness in a particular environment (e.g., STEARNS 1992 ). For example, under poor feeding conditions, it may be advantageous to cease growth at an early stage and allocate the available energy to maintenance instead of to further growth. This is what has been called an "adaptive flexible strategy," but a similar phenotypic response may also take place as a simple consequence of stress, without any fitness benefits (KEARSEY and POONI 1996 ). Likewise, an "inflexible" or highly canalized genotype that produces the same phenotype irrespective of environmental conditions may be the most fit in certain situations, but it can be difficult to demonstrate that its fitness is higher than that of a more flexible genotype. Although we cannot address the possible adaptive significance of the observed GEI in this study, the bottom line is that genotypes differing in the shape and/or orientation of their reaction norms have the potential to influence the course and pace of evolution in heterogeneous environments (VIA and LANDE 1987 ; GILLESPIE and TURELLI 1989 ). Hence, if different blue tit genotypes are favored under different environmental conditions, there should be no severe genetic constraints for adaptation to each of the environments.

Heritability of body size:
The heritability estimates obtained from offspring-parent regressions were very similar to those obtained by DHONDT 1982 in a study of Belgian blue tits (h ² = 0.70 ± 0.27), and to that reported earlier from this same population using a different data set (h ² = 0.81 ± 0.17); MERILA and WIGGINS 1995 ). Although these h ² values for tarsus length are in the upper range reported for birds in any population (see BOAG and VAN NOORDWIJK 1987 ; MERILA 1997 for reviews), we found no evidence that the offspring-parent resemblance was due to genotype-environment correlation, as there was no resemblance between offspring and their foster parents. Such an environmentally induced similarity among parents and offspring could arise, for example, if the largest parents occupied the best territories and thereby enhanced the growth performance of their offspring. However, apart from a few notable exceptions ( JAMES 1983 ; LARSSON and FORSLUND 1992 ), there is little evidence for such effects from natural bird populations, as none of the cross-fostering studies performed so far have found any indications of genotype-environment correlation in body-size traits (SMITH and DHONDT 1980 ; ALATALO and LUNDBERG 1986 ; WIGGINS 1989 ; GUSTAFSSON and MERILA 1994 ; SMITH 1994 ).

One of the striking findings of this study was the conspicuous disagreement of heritability estimates from full-sib and parent-offspring analyses. While the estimates from full-sib analyses and parent-offspring regressions were more similar when based on offspring grown in reduced broods, full-sib estimates obtained using offspring grown in enlarged broods were much lower than equivalent estimates from parent-offspring regressions. This contrasts with the observation that full-sib estimates of heritability usually exceed parent-offspring estimates (e.g., LARSSON 1993 ; MERILA and GUSTAFSSON 1993 ). However, the low full-sib estimates of heritability in enlarged broods resulted from much higher environmental—and, hence, total phenotypic variances (and possibly genetic variances)—in enlarged broods compared with reduced broods and the parents. This suggests that either much selection occurs between fledging and first breeding, and/or that growth of smaller nestlings continues after fledging. It is quite likely that both processes will account for part of the higher phenotypic variance in enlarged broods, as some growth was observed in the poor weather year (1994) after the nestlings were measured, and selection against small-sized nestlings has been repeatedly observed in this (NUR 1984 ) as well as in other phylogenetically and ecologically closely related species (ALATALO et al. 1990 ; LINDEN et al. 1992 ).

It is worth noting that the estimated midparent- midoffspring covariances tended always to be higher for offspring in the reduced environment than for offspring in the enlarged environment (Table 9). This suggests either that (1) the additive genetic variance is higher in the reduced environment and/or that (2) the genetic correlation between parental and offspring environment is higher when offspring are in the reduced environment than when they are in the enlarged environment (see Introduction). For both environments, the estimated covariances were always smaller than the corresponding Box of origin variance components from the full-sib analyses (Table 5). This difference could have a variety of explanations. If additive genetic variances are the same in each environment, and if maternal and dominance effects are not present, then the parent-offspring covariances should be smaller than the full-sib variance components by a fraction equal to the genetic correlation between parent and offspring environment (see Introduction). Alternatively, the Box of origin variance components could be inflated by dominance or maternal effects. It is also possible that additive genetic variances are lower among parents than among offspring because selection has reduced the variance among parents.

Confounding factors:
There is ever-increasing evidence that EPP is common in many socially monogamous birds, including the blue tit. KEMPENAERS et al. 1992 found that about 31% of the blue tit nests (11% of all young) contained nestlings sired by extra-pair males. In theory, this should lead to lowered father-offspring resemblance as compared with mother-offspring resemblance (ALATALO et al. 1989 ; but see GEBHARDT-HENRICH and NAGER 1991 ). We found no indications that EPP was common in our population, as the contributions of both putative parents to offspring-parent resemblance were similar. Hence, unless we assume strong maternal effects that increased the resemblance between mothers and their offspring, there is no reason to suspect that EPP has biased our estimates of heritability. Indeed, it is possible that EPP is much more frequent among the Belgian blue tits, which also show much higher frequency ( 11%; DHONDT 1987 ) of polygamy than that observed in our population (<1% of pairs; J. MERILÄ, unpublished data). In any case, as the treatments in this study were allocated randomly relative to EPP, it should not affect our inferences about relative magnitudes of genetic variance in these environments.

Maternal and common-environment effects in the form of genotype-environment correlations are a potential source of bias in all quantitative genetic studies. As discussed earlier, we found no evidence for genotype-environment correlations between parents and their offspring, but we cannot exclude the possibility that our full-sib estimates of genetic variances could include a component of maternal or early common-environment effects. As the young were swapped when they were 2 days old, effects of early caretaking and prehatching maternal effects, if present, have been included in our estimates of genetic variances and correlations from the full-sib data. However, such effects are, according to our best knowledge, likely to be small. Likewise, full-sib estimates also include a component of dominance variance, if such was present. Again, empirical data to date indicates that if such effects are present, they are typically small (CLAYTON et al. 1957 ; MOUSSEAU and ROFF 1987 ).

Offspring size and number:
Our results also bear relevance to discussions of one of the major trade-offs in life history evolution: namely, that between offspring size and number (STEARNS 1992 ). In organisms with parental care, this trade-off is assumed to arise because of the inability of parents to provide adequate food for the young in large broods (STEARNS 1992 ). Our results show that parents caring for large broods were less successful in providing food for the offspring than those caring for smaller broods. If we assume that the size attained by the offspring at the end of the nestling period predicts their future survival prospects (LINDEN et al. 1992 ), then the observed among-family variation in the response to the two environments means also that the optimal brood size for different parental genotypes may differ depending on the environmental conditions. This might at least partly help to explain the large individual variation in blue tit clutch size (4–17 eggs, e.g., MERILA and WIGGINS 1995 ).

We further found that the experiment had different effects on offspring size in different years, the effect being most pronounced in the year when the weather conditions were particularly bad for breeding (1994). The 1994 breeding season was cold and rainy (mean temperature and total rainfall during the first two weeks of June: 11.0° and 21.1 mm), whereas 1993 and 1995 were warmer—and except for 1995, less rainy (1993: 13.5° and 0.2 mm; 1995: 13.4° and 62.2 mm). As the weather will affect the timing and abundance of caterpillar larvae, which are the main food source of tits (CRAMP and PERRINS 1993 ), we believe that the experiment-year interaction in offspring size was caused by between-year differences in food availability, so that the effect of the experiment became amplified when the feeding conditions were particularly bad.

In conclusion, our results demonstrate a weak but statistically significant GEI in body size of nestling blue tits. This interaction, representing genetic variation in sensitivity to environmental effects, was explained mainly by the genetic correlation between the two environments being less than unity. Hence, our results give no support for the view that the amount of genetic variance expressed in poor environments is necessarily less than that expressed in good environments—and, consequently, that the genetic potential for adaptive evolution differs between poor and good environments. Further, our results, together with circumstantial evidence from several other studies of natural bird populations (reviewed in MERILA 1997 ), suggest that GEIs may be common in natural vertebrate populations—and, consequently, that the relative fitness of different genotypes may vary according to prevailing environmental conditions.

	FOOTNOTES

¹ Present address: Utah State University, Department of Biology, Logan, UT 84322-5305.

	ACKNOWLEDGMENTS

We thank MARIE BJÖRKLUND, JESSICA LINDSTRÖM, JANNE KUMPULAINEN, MIKA KOLEHMAINEN, ROBERT PRZYBYLO, and DAVE WIGGINS for their help with the fieldwork as well as LINDA PARTRIDGE and two anonymous reviewers for useful comments on this paper. Financial support from Stiftelsen Zoologisk Forskning (Uppsala, Sweden), the Royal Swedish Academy of Sciences, and the Kone Foundation (Finland) is gratefully acknowledged.

Manuscript received July 28, 1997; Accepted for publication December 1, 1997.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ALATALO, R. V. and A. LUNDBERG, 1986  Heritability and selection on tarsus length in the pied flycatcher Ficedula hypoleuca. Evolution 40:574-583.

ALATALO, R. V., L. GUSTAFSSON, and A. LUNDBERG, 1989  Extra-pair paternity and heritability estimates of tarsus length in pied and collared flycatchers. Oikos 56:54-58.

ALATALO, R. V., L. GUSTAFSSON, and A. LUNDBERG, 1990  Phenotypic selection on heritable size traits: environmental variance and genetic response. Am. Nat. 135:464-471.

ATCHLEY, W. R. and J. J. RUTLEDGE, 1980  Genetic components of size and shape. I. Dynamics of components of phenotypic variability and covariability during ontogeny in the laboratory rat. Evolution 34:1161-1173.

BAILEY, R. C. and J. BYRNES, 1990  A new, old method for assessing measurement error in both univariate and multivariate morphometric studies. Syst. Biol. 39:124-130.

BERVEN, K. A., 1987  The heritable basis of variation in larval development patterns within populations of the wood frog (Rama sylvatica). Evolution 41:1088-1097.

BOAG, P. T., and A. J. VAN NOORDWIJK, 1987 Quantitative genetics, pp. 45–78 in Avian Genetics, edited by F. COOKE and P. A. BUCKLEY. Academic Press, New York.

CLAYTON, G. A., J. A. MORRIS, and A. ROBERTSON, 1957  An experimental check on quantitative genetic theory. I. Short term responses to selection. J. Genet. 55:131-151.

CRAMP, S., and C. M. PERRINS, 1993 The Birds of the Western Palearctic, Vol. VII. Oxford University Press, Oxford.

DHONDT, A. A., 1982  Heritability of blue tit tarsus length from normal and cross-fostered broods. Evolution 36:418-419.

DHONDT, A. A., 1987  Reproduction and survival of polygynous and monogamous blue tits Parus caeruleus. Ibis 129:327-334.

FALCONER, D. S., and T. MACKAY, 1996 Introduction to Quantitative Genetics. Longman, New York.

FRY, J. D., 1992  The mixed-model analysis of variance applied to quantitative genetics: biological meaning of the parameters. Evolution 46:540-550.

FRY, J. D., 1996  The evolution of host specialization: are trade-offs overrated? Am. Nat. 148:S84-S107.

GEBHARDT-HENRICH, S. G. and R. NAGER, 1991  Can extra-pair fertilizations be detected by differences in maternal and paternal heritability estimates? Ibis 133:93-94.

GEBHARDT-HENRICH, S. G. and A. J. VAN NOORDWIJK, 1991  The genetical ecology of nestling growth in the great tit. I. Heritability estimates under different environmental conditions. J. Evol. Biol. 3:341-362.

GEBHARDT-HENRICH, S. G. and A. J. VAN NOORDWIJK, 1994  The genetical ecology of nestling growth in the great tit. Environmental influences on the expression of genetic variances during growth. Funct. Ecol. 8:469-476.

GILLESPIE, J. H. and M. TURELLI, 1989  Genotype-environment interactions and the maintenance of polygenic variation. Genetics 114:333-343[Abstract/Free Full Text].

GIMELFARB, A. and J. H. WILLIS, 1994  Linearity versus nonlinearity of offspring-parent regressions: an experimental study of Drosophila melanogaster. Genetics 138:343-352[Abstract].

GUSTAFSSON, L. and J. MERILÄ, 1994  Foster parent experiment reveals no genotype-environment correlation in external morphology of the collared flycatcher, Ficedula albicollis. Heredity 73:124-129.

JAMES, F. C., 1983  Environmental component of morphological differentiation in birds. Science 221:184-186[Abstract/Free Full Text].

KAWECKI, T. J., N. H. BARTON, and J. D. FRY, 1997  Mutational collapse of fitness in marginal habitats and the evolution of ecological specialisation. J. Evol. Biol. 10:407-429.

KEARSEY, M. J., and H. S. POONI, 1996 Genetical Analysis of Quantitative Traits. Chapman & Hall, London.

KEMPENAERS, B., G. R. VERHEYEN, M. VAN DEN BROECK, T. BURKE, and C. VAN BROECKHOVEN et al., 1992  Extra-pair paternity results from female preference for high-quality males in the blue tit. Nature 357:494-496.

LANDE, R., 1987  Appendix to COYNE and BEECHAM 1987. Genetics 117:737-738.

LARSSON, K., 1993  Inheritance of body size in the barnacle goose under different environmental conditions. J. Evol. Biol. 6:195-208.

LARSSON, K. and P. FORSLUND, 1992  Genetic and social inheritance of body and egg size in the barnacle goose (Branta leucopsis). Evolution 46:235-244.

LARSSON, K., K. RATISTE, and V. LILLELEHT, 1997  Heritability of head size in the common gull Larus canus in relation to environmental conditions during offspring growth. Heredity 79:201-207.

LINDÉN, M., L. GUSTAFSSON, and T. PÄRT, 1992  Selection of fledging mass in the collared flycatcher and the great tit. Ecology 73:336-343.

MERILÄ, J., 1996  Genetic variation in offspring condition—an experiment. Funct. Ecol. 10:465-474.

MERILÄ, J., 1997  Expression of genetic variation in body size of collared flycatcher under different environmental conditions. Evolution 51:526-536.

MERILÄ, J. and L. GUSTAFSSON, 1993  Inheritance of size and shape in natural population of collared flycatchers, Ficedula albicollis. J. Evol. Biol. 6:375-398.

MERILÄ, J. and D. A. WIGGINS, 1995  Offspring number and quality in the blue tit (Parus caeruleus)—a quantitative genetic approach. J. Zool. 237:615-623.

MOUSSEAU, T. A. and D. A. ROFF, 1987  Natural selection and the heritability of fitness components. Heredity 59:181-197.

NEWMAN, R. A., 1988  Adaptive plasticity in development of Scaphiopus couchii tadpoles in desert ponds. Evolution 42:774-783.

NUR, N., 1984  The consequences of brood size for breeding blue tits. II. Nestling weight, offspring survival and optimal brood size. J. Anim. Ecol. 53:497-517.

PRICE, T. D., 1991  Environmental and genotype-by-environment influences on chick size in the Yellow-browed leaf warbler Phylloscopus inornatus.. Oecologia 86:535-541.

RISKA, B., T. PROUT, and M. TURELLI, 1989  Laboratory estimates of heritabilities and genetic correlations in nature. Genetics 123:865-871[Abstract/Free Full Text].

ROBERTSON, A., 1959  The sampling variance of the genetic correlation coefficient. Biometrics 15:469-485.

SAS INSTITUTE, INC., 1989 SAS/STAT User's Guide, Version 6. SAS Institute Inc., Cary, NC.

SMITH, H. G., 1994  Heritability of tarsus length in cross-fostered broods of the European starling (Sturnus vulgaris). Heredity 71:318-322.

SMITH, H. G. and K.-J. WETTERMARK, 1995  Heritability of nestling growth in cross-fostered European starlings Sturnus vulgaris. Genetics 141:657-665[Abstract].

SMITH, J. N. M. and A. A. DHONDT, 1980  Experimental confirmation of heritable morphological variation in a natural population of song sparrows. Evolution 34:1155-1158.

SOKAL, R. R., and F. J. ROHLF, 1981 Biometry. W. H. Freeman, New York.

STEARNS, S. J., 1992 The Evolution of Life Histories. Oxford Unviersity Press, Oxford.

VIA, S., 1993  Adaptive pheontypic plasticity: target or by-product of selection in a variable environment. Am. Nat. 142:352-365.

VIA, S., 1994 The evoultion of phenotypic plasticity: what do we really know?, pp. 35–57 in Ecological Genetics, edited by L. A. REAL. Princeton University Press, Princeton, NJ.

VIA, S. and R. LANDE, 1985  Genotype-environment interaction and the evolution of phenotypic plasticity. Evolution 39:505-522.

VIA, S. and R. LANDE, 1987  Evolution of genetic variability in a spatially heterogeneous environment: effects of genotype-environment interaction. Genet. Res. 49:147-156[Medline].

WIGGINS, D. A., 1989  Heritability of body size in cross-fostered tree swallow broods. Evolution 43:1808-1811.

This article has been cited by other articles:

				L. C. Emebiri, V. Matassa, and D. B. Moody GENSTAT Programs for Performing Muir's Alternative Partitioning of Genotype-by-Environment Interaction J. Hered., January 1, 2005; 96(1): 78 - 79. [Abstract] [Full Text] [PDF]

				N. Saino, C. Suffritti, R. Martinelli, D. Rubolini, and A. P. Moller Immune response covaries with corticosterone plasma levels under experimentally stressful conditions in nestling barn swallows (Hirundo rustica) Behav. Ecol., May 1, 2003; 14(3): 318 - 325. [Abstract] [Full Text] [PDF]

				J. Blondel, P. C. Dias, P. Perret, M. Maistre, and M. M. Lambrechts Selection-Based Biodiversity at a Small Spatial Scale in a Low-Dispersing Insular Bird Science, August 27, 1999; 285(5432): 1399 - 1402. [Abstract] [Full Text]