Genetic and environmental components of growth in nestling blue tits (Parus caeruleus)


Camilla Kunz Evolutionary Biology Center, Department of Population Biology, Uppsala University, Norbyvägen 18D, S – 752 36 Uppsala, Sweden. E-mail:


We investigated the effect of brood-size mediated food availability on the genetic and environmental components of nestling growth in the blue tit (Parus caeruleus), using a cross-fostering technique.

We found genetic variation for body size at most nestling ages, and for duration of mass increase, but not of tarsus growth. Hence, nestling growth in our study population seems to have the potential to evolve further. Furthermore, significant genotype–environment interactions indicated heritable variation in reaction norms of growth rates and growth periods, i.e. that our study population had a heritable plasticity in the growth response to environmental conditions.

The decreasing phenotypic variance with nestling age indicated compensatory growth in all body traits. Furthermore, the period of weight increase was longer for nestlings growing up in enlarged broods, while there was no difference to reduced broods in the period of tarsus growth. At fledging, birds in enlarged broods had shorter tarsi and lower weights than birds in reduced broods, but there was no difference in wing length or body condition between the two experimental groups. The observed flexibility in nestling growth suggests that growing nestlings are able to respond adaptively to food constraint by protecting the growth of ecologically important traits.


In birds, postnatal growth rates vary not only among species, but also among populations of the same species ( O’Connor, 1978; Ricklefs, 1983). Differences in growth rates can reflect genetic adaptations to different ecological environments. For example, high rates of nest predation and competition among nestmates select for rapid growth, whereas unpredictable food availability is believed to favour slow growth (reviewed by Ricklefs, 1983, 1984b).

For growth patterns to evolve, genetic variation for growth parameters is required ( Endler, 1986; Falconer, 1989). While studies on poultry have shown that growth rates may be heritable (e.g. Singh et al., 1991 ), most studies of natural populations have failed to find significant genetic variation for growth parameters, except for asymptotic size (e.g. Gebhardt-Henrich & van Noordwijk, 1994; Smith & Wettermark, 1995). Instead, large effects of rearing environment on growth curves have been observed. This may suggest that growth rates in most natural bird populations do not have the potential to evolve further.

If populations regularly encounter variable environments, it could be favourable for individuals to be phenotypically plastic in order to be able to develop in a range of environments, instead of being genetically adapted to only a single environment ( Stearns, 1989; Via et al., 1995 ). Indeed, many studies on birds have demonstrated that nestling growth rates vary with food availability ( Gebhardt-Henrich & van Noordwijk, 1994), food quality ( Johnston, 1993) or parasite load ( Dufva & Allander, 1996).

Still, if the ‘genetically determined’ final size is an important fitness component, nestlings should aim at reaching this size and compensate for poor early growth by increased later growth rates or a prolonged growth period. Compensatory growth has been found in some studies ( Berthold, 1976; Wiggins, 1990), but was incomplete or absent in others ( Larsson & Forslund, 1991; Konarzewski et al., 1996 ).

The ability to respond to variable environments with plastic growth may differ between individuals as well as between populations. Few studies have addressed the question of whether plasticity of nestling growth has a heritable component, and thus a potential to evolve ( Smith & Wettermark, 1995; Gebhardt-Henrich & van Noordwijk, 1994).

Blue tits are altricial birds frequently faced with variable or unpredictable food abundance during the nestling period (e.g. Perrins, 1991). Food availability has been found to affect nestling growth and adult size in this species ( Dias & Blondel, 1996; Merilä & Fry, 1998). To reduce the impact of poor feeding conditions on fitness, developing nestlings may have evolved flexible morphological or physiological responses to variation in food supply.

In the present study, we investigated genetic and environmental components of nestling growth and size under different environmental conditions. We created ‘good’ and ‘poor’ rearing conditions by reducing and by enlarging broods, respectively. Cross-fostering of nestlings allowed us to study: (i) to what extent variation in nestling growth was due to genetic and environmental effects, respectively; (ii) whether reaction norms of growth showed any heritable variation, as manifested in genotype–environment interactions; (iii) whether poorly grown nestlings showed any compensatory growth; and (iv) whether brood size-mediated food availability had any permanent effect on fledging size. We measured four different morphological traits to see whether nestlings growing up under different food regimes allocated resources differently to different body structures.


Study site

The study was carried out in Sweden in the vicinity of Uppsala (59°51′, 17°38′E) in several mixed-deciduous forest patches provided with nest-boxes. The patches were rich habitats, dominated by old oaks (Quercus robur), birches (Betula pendula, B. pubescens) and aspen (Populus tremula), with some occurrence of Scots pine (Pinus sylvestris) and Norway spruce (Picea abies). Some patches included pastures grazed by cattle or sheep. The study area (≈120 ha) was provided with a total of ≈650 nest-boxes.

Cross-fostering design

All nest boxes were controlled regularly to record egg-laying and hatching dates (date when the first egg in each brood was hatched). Pairs of broods with the same hatching date were matched to form a ‘duplicate’, regardless brood size. For females laying the number of eggs which maximizes the number of surviving young ( Pettifor, 1993), it is the degree of change in, rather than absolute brood size which should be important. At the age of 2 days (hatching day = day 0), parts of the broods were swapped among matched nests, thereby enlarging and reducing the broods by 1/4, respectively ( Fig. 1). This brood size manipulation will henceforth be abbreviated as BSM. Nestlings were randomly assigned to the different treatments.

Figure 1.

 Schematic presentation of cross-fostering protocol. Brood sizes were manipulated by swapping parts of broods among matched nests.

The rationale behind this design was to induce brood size dependent feeding regimes among nestlings ( Nur, 1984a, b), and to be able to partition the total phenotypic variance of nestlings in its genetic and environmental components. Heritabilities can be inflated by maternal effects or common environment, which both can increase similarity among siblings in a nongenetic way ( Falconer, 1989). Cross-fostering can control for possible maternal effects occurring after the transfer to foster parents, and eliminates the problem of shared rearing environment by placing part of the offspring in an environment different from that of their siblings.

Before the BSM, no significant difference in brood size was present between experimental groups (‘enlarged’ 9.6 ± SE 2.0, ‘reduced’ 9.6 ± SE 2.4, N = 32, t = 0.10, P = 0.92). After the swapping of nestlings, enlarged broods contained significantly more nestlings than reduced broods (enlarged 11.8 ± SE 2.9, reduced 7.7 ± SE 2.1, N = 32, t = 4.69, P = 0.0001). Differences in brood size remained significant throughout the nest- ling period (day 15 post-hatch: enlarged 9.2 ± SE 4.0, reduced 5.6 ± SE 2.0, N = 23, t = 2.70, P = 0.016). Hence, our manipulation was not offset by differential nestling mortality.

Data on body size

On day 2 post-hatch, we individually marked all young with nail varnish on their claws and weighed them with a spring balance to the nearest 0.1 g. At the age of 4, 6, 8, 10, 12, 14 and 15 days, we measured wing length with a ruler to the nearest 0.5 mm, tarsus length with callipers to the nearest 0.1 mm, and body mass of all young in a brood with a Pesola spring balance. At the age of 10 days, we individually marked all young with aluminium rings. When the offspring were 15 days of age, we captured the parents with mist nets at the nest, or with traps in the nest-boxes, and measured and ringed them in the same way as their offspring. Since nestlings may fledge from day 16 onwards, we refer to body size on day 15 as ‘fledging size’.

A condition index was calculated as the residual from a regression of log weight on log tarsus length. This index was chosen as it has been found to be a good predictor of lipid reserves in blue tits ( Merilä & Svensson, 1997) and other birds ( Johnson et al., 1985 ; Blem, 1990), as well as a predictor of individual birds’ survival probability (e.g. Lindén et al., 1992 ).

Statistical analysis

Although data for individual nestlings were not always perfectly normally distributed due to some nestlings lagging behind in growth, the Wilks’ test statistic W was always found to be between 0.9 and 1.0. Therefore, untransformed data were used in statistical analyses. The data based on brood means did not differ significantly from a normal distribution. Certainly, size at different ages is partly correlated, as are the four different characters. However, we have chosen to present the analysis of nestling growth age by age.

Genetic and environmental components of nestling size

Genetic and environmental effects on nestling size were tested using SAS General Linear Model (GLM) procedures ( SAS Institute Inc., 1990). We used the RANDOM statement to estimate random-model expected mean squares and error estimates for F-tests. Variance components and their standard errors were obtained from VARCOMP procedure in SAS using Restricted Maximum Likelihood (REML) ("bbr rid="b73">SAS Institute Inc., 1990).

To partition phenotypic variation into genetic and environmental effects in each environment separately, we performed a nested analysis with random factors ‘box of rearing’ and ‘box of origin’. Effects of ‘box of origin’ were nested within ‘box of rearing’, and estimate half of the additive genetic variance, but include also one quarter of dominance variance and all maternal effects up to the age of 2 days. ‘Nest of rearing’ accounts for the variance due to parental care after day 2, plus the variance due to common nest-box, including the effects of BSM. Heritabilities for size at different ages, and for the length of the growth period were obtained by dividing the variance due to ‘nest of origin’ by the total variance of the sample, and multiplying this ratio by two ( Falconer, 1989). Standard errors of heritability estimates were obtained with a jackknife procedure ( Sokal & Rohlf, 1995).

Genetic variation in reaction norms

Genetic variation in reaction norms, manifested as genotype–environment interactions, was tested for in a mixed-model nested analysis. ‘BSM’ was used as a fixed effect and measured the effect of brood size manipulations on nestling size, i.e. whether size was plastic. The term ‘nest of origin’ (random effect) tested for possible genetic effects on nestling size, and was nested within ‘duplicates’. The term ‘duplicate’ (random effect) controlled whether there was any variation in body size among duplicates, which could arise, e.g. from changes in food availability during the season. The interaction term ‘duplicate*BSM’ investigated whether different duplicates were affected differently by BSM, while ‘Origin*BSM’, with ‘Origin’ nested within ‘duplicate’, tested whether different genotypes responded differently to BSM, indicating genetic variation for plasticity. Non-significant interaction terms were deleted from the model. Because the last factor was of main interest, only statistics of the ‘Origin*BSM’ interaction term are shown in the results.

Compensatory growth

Nestlings growing up under poor conditions may compensate for retarded growth by prolonging the period of growth. The growth period of weight and tarsus was determined as the time up to the age when there was no longer any increase in the trait. Tarsus length could not be measured with the same precision as weight. Therefore, it was arbitrarily decided that growth of individual birds continued as long as tarsus length increased with at least 0.2 mm. The growth period for wing could not be determined, since wing growth is not completed at fledging. Because consecutive measurements of size around the critical ages were not available for all nestlings, sample sizes for paired t-tests were considerably reduced (N = 11 and 12 for weight and tarsus, respectively). Therefore, we used unpaired t-tests for comparing the growth period of nestlings reared in the two experimental environments (N = 34 and 39). The results obtained from the two tests were similar, although P-values were lower in unpaired tests.


Estimates of heritabilities from full-sib analysis rest on the assumption that there is (i) no extra-pair paternity, (ii) no assortative mating, (iii) no maternal effects, (iv) no dominance variation, and (v) no competition among young present.

(i) Extra-pair paternity leads to reduced similarity among nest-mates and, therefore, heritabilities are underestimated. Extra-pair paternity rates are low (6–11%, Gullberg et al., 1992 ; Kempenaers et al., 1992 ), and possibly absent in some blue tit populations ( Merilä & Fry, 1998). Furthermore, if extra-pair young occurred in our study population, we assumed that they were distributed randomly over both treatment groups, and did not affect comparisons of heritabilities between the groups. Brood parasitism in form of egg dumping by foreign females has so far not been found in this species ( Kempenaers et al., 1995 ).

(ii) Assortative mating of parental birds could bias heritability estimates upwards ( Falconer, 1989). We tested for assortative mating by correlating male and female body size within breeding pairs. Correlations were nonsignificant for all body traits (tarsus length: r = 0.18, N = 26, wing length: r = –0.05, N = 25, weight: r = 0.10, N = 26, condition: r = –0.12, N = 26).

(iii) Non-genetic maternal effects can increase similarity among siblings and thereby inflate heritabilities obtained from full-sib analyses ( Falconer, 1989). To test for maternal effects we correlated offspring body size with their mother’s body condition to see whether her nutritional state, or another trait closely correlated to it, could have influenced offspring growth. Such a correlation might be expected, since the female’s nutritional state has been found to affect egg size ( Slagsvold & Lifjeld, 1989) which in turn is known to affect nestling size (reviewed by Williams, 1994). Female body condition may also affect brooding or feeding efficiency. However, correlation of female body condition with mean nestling weight on day 2 was nonsignificant (r = 0.04, P = 0.88, N = 20). Furthermore, we failed to find any maternal effects on mean tarsus length on day 15 (r = 0.024, P = 0.93, N = 16), regardless of brood size (enlarged: r = 0.09, P = 0.74, N = 15; reduced: r = –0.16, P = 0.57, N = 14), or among nestlings raised by true or foster parents (true: r = 0.08, P = 0.74, N = 21; foster: r = –0.04, P = 0.89, N = 14). This suggests that direct effects of maternal nutrition must have been small. Although we cannot rule out the possibility that nestling size was affected by maternal effects unrelated to maternal condition, a common finding has been that maternal effects on offspring size in birds are short-lived (e.g. Price & Grant, 1985; Bolton et al., 1992 ; Schwabl, 1996). We therefore believe that maternal effects did not seriously confound our results.

(iv) A potential problem with heritability estimates obtained from full-sib analysis is that additive genetic variances (VA) may be confounded by dominance variances (VD), with heritabilities becoming biased upwards ( Falconer, 1989). Theoretical and empirical investigations suggest that dominance contribution is high mainly in fitness traits under strong selection ( Merilä & Sheldon, 1999). While VD may indeed be high for life-history traits, VD has been found to be very low for morphological traits in wild outbred populations of animals and plants ( Crnokrak & Roff, 1995). Accordingly, Mousseau & Roff (1987) found differences between heritabilities derived from parent–offspring regression (PO) and full sib analysis (FS) to be small (<10%) in wild animals, including populations of birds. Because we only analysed morphological traits, we expect our FS heritability estimates to be close to narrow-sense heritabilities. This assumption turned out to be correct for tarsus length, the only trait comparable in parents and 15-day-old skeletally fully grown offspring. (FS enlarged: 0.60 ± SE 0.29, reduced: 0.34 ± SE 0.34; PO enlarged: 0.62 ± SE 0.30, reduced: 0.58 ± SE 0.29.)

(v) Any size differences between unrelated sib-groups within a nest, e.g. due to slight age differences, may affect competition and inflate sibling similarity. In contrast, an increase in variation among nestlings regardless of relatedness, such as that due to hatching asynchrony, would make it more difficult to detect similarity among related nestlings. Correlations of nestling weight on day 2 with tarsus length on day 15 were on average not different from zero (mean r = 0.28, N = 19; 95% confidence limits –0.004, 0.566), suggesting that any size hierarchies that would indicate competition were not maintained. Correlations tended to be higher in enlarged than reduced broods (Wilcoxon Rank Test, N enlarged = 10, N reduced = 9, z = 1.84, P = 0.066, N total = 19). Hence, there is a possibility that we underestimated genetic effects on nestling size and growth in enlarged broods. However, because P > 0.05, this bias is likely to be small.


Size differences before swapping

There was no significant difference in mass day 2 post-hatch between chicks raised in enlarged and reduced broods (paired t-test, N = 22, t = 1.43, P = 0.17). Therefore, any effects of initial size are probably negligible.

Genetic and environmental components of nestling growth

(a) Nestling size

‘Good’ environments (reduced brood size) did not significantly affect nestling tarsus length, wing length or weight, except for weight on days 12 and 14 (Fig. 22). Instead, large genetic components were present for all these traits. However, genetic effects tended to decrease towards fledging. Correspondingly, heritability estimates tended to be higher at early than later nestling ages, significantly so for weight (day 4 vs. 12, day 6 vs. 12) and wing (day 6 vs. 14) ( Fig. 3). The opposite pattern emerges for body condition. ‘Good’ environments significantly affected early body condition, with genetic effects being absent. Towards fledging age, environmental effects on body condition decreased, paralleled by an increase of genetic effects. This matches with increasing heritabilities for body condition with age, with significantly higher values on day 14 than on days 4, 6 and 8.

Figure 2.

 Total phenotypic variance and contribution by environment (E) and genetic (G) factor for (a) tarsus length, (b) weight, (c) wing length, (d) body condition. Significance levels of environmental (E) and genetic (G) variance components and degrees of freedom (d.f.) at different nestling ages were obtained from a nested two-factorial analysis, and are given below the figures (°P < 0.1, *P < 0.05, **P < 0.01, ***P < 0.001). Bar shading: white = G, light grey = E, dark grey = residual error variance.

Figure 3.

Fig. 2

Figure 3.

 Heritabilities and standard errors of nestling size at different ages. (a) Tarsus length, (b) weight, (c) wing length, (d) body condition. Filled circles  = reduced broods, open circles  =  enlarged broods. Significance levels of heritabilities in enlarged (enl) and reduced (red) broods: °P < 0.1, *P < 0.05, **P < 0.01, ***P < 0.001.

‘Poor’ growth conditions (enlarged brood size) had significant effects on all traits (Fig. 22). These effects tended to be strongest at intermediate ages, when genetic effects were weak or absent. Towards fledging, environmental components became smaller, while genetic effects increased and became significant again. Accordingly, heritabilities for nestling size tended to be lowest at intermediate ages ( Fig. 3). However, this pattern was less pronounced for wing growth, where genetic effects were significant at almost all ages. Hence, for wing heritabilities, no obvious trend of change with age could be seen. Due to large error variation, heritabilities were never significantly different from zero in any trait, except on day 14 post-hatch.

Generally, heritabilities tended to be higher under good than under poor conditions, significantly so for tarsus on days 8 and 10, weight on day 6, and wing on days 6 and 8.

Since significant genetic effects on nestling size were found more often than expected by chance at the level of P = 0.05 (in 33 out of 56 tests), there seems to be genetic variation for body size in our study population.

(b) Growth periods

When testing for genetic and environmental effects on the length of the growth period for each treatment group separately, the period of weight increase was not affected by growth conditions, but instead showed significant genetic variation under ‘good’ conditions (Table 1). Thus, under ‘poor’ conditions, heritability estimates did not differ from zero (Table 1). The period of tarsus growth was not affected by growth conditions or by ‘nest of origin’ in either environment, although weak genetic effects may have been present under ‘poor’ conditions (significant at the level of 0.1). Accordingly, growth periods for tarsus were not heritable in any environment (Table 1).

Table 1.   Results from a nested two-factorial analysis to partition genetic and environmental effects on the growth period, and heritability estimates with standard errors (h2 ± SE). Thumbnail image of

Genetic variation in reaction norms

When testing for genotype–environment interactions at all nestling ages, significant effects of the term ‘BSM*Origin’ were found for all traits (Table 2). The number of significant interactions (6 out of 28) exceeded the number expected by chance at the level of P = 0.05. Thus, the degree of plastic responses to environmental variation seems to be heritable in all traits. We also found significant genotype–environment interactions for the growth period of weight, but not of tarsus length (Table 3). For tarsus length, significant ‘BSM*duplicate’ interaction showed that ‘BSM’ affected nestlings from different duplicates to a different extent (Table 3).

Table 2.   Levels of significance of genetic variation in plasticity during growth, tested for by genotype–environment interaction (‘BSM*Origin’) in a nested three-factorial analysis. Thumbnail image of
Table 3.   Results from a nested three-factorial analysis to partition genetic and environmental components of the growth period (nonsignificant interaction terms were deleted from the model (–)). Thumbnail image of

Compensatory growth

Nestlings in enlarged broods increased in weight on average 0.7 days longer than nestlings in reduced broods, while it was not possible to show prolonged tarsus growth (Table 4). Furthermore, phenotypic variances, which tended to be largest at intermediate ages, tended to decrease towards fledging age (Fig. 22), indicating compensatory growth in all traits (e.g. Atchley, 1984).

Table 4.   Length of the growth period (in days). Thumbnail image of

Effects of BSM on fledging size

Nestlings in enlarged broods had significantly shorter tarsi and lower weights at fledging than nestlings in reduced broods, but there were no significant differences in wing length or body condition between the two groups (Fig. 44, Table 5).

Figure 4.

 Mean nestling size and standard errors at different ages. (a) Tarsus length, (b) weight, (c) wing length, (d) body condition. Filled circles  = enlarged broods, open circles  =  reduced broods.

Figure 6.

Fig. 4 (Contd.)

Table 5.   Phenotypic variation of nestlings on day 15. Thumbnail image of


We have shown that nestling blue tits responded to brood size manipulations with plastic growth patterns, and that growth and plasticity had heritable components. Growing nestlings gave priority to wing development and a good body condition over weight and tarsus length.

Genetic variation in nestling size and growth periods

We found significant genetic variation in nestling size under both experimental conditions, and in the period of weight increase under ‘good’ feeding conditions. We could not detect any nongenetic maternal effects, assorta- tive mating, effects of sibling competition, or indications of large dominance component to variation, all which may bias heritability estimates. Hence, our estimates of additive genetic effects seem to be reliable in this study, and both growth rates and growth periods have the potential to evolve further. Heritabilities varied with nestling age, meaning that responses to selection on nestling size may vary also, depending on when selection takes place during growth ( Falconer, 1989). Because changes in genetic variation during ontogeny may alter strength or direction of correlations between traits (e.g. Atchley & Rutledge, 1980), correlated responses to selection are likely to vary with nestling age as well.

Although adult size has been found heritable in many studies (review in Boag & van Noordwijk, 1987; Wiggins, 1989; Larsson, 1993; Merilä & Gustafsson, 1993; Smith, 1993; Potti & Merino, 1994), nestling size during growth and other growth parameters rarely have shown any genetic variation. In an experimental study on Great tits (Parus major) ( Gebhardt-Henrich & van Noordwijk, 1994), genetic effects on nestling size were significant only on days 5 and 14, with early effects being interpreted as maternal effects. However, genetic effects approached the significance level of 0.1 at many ages during growth, suggesting the presence of heritable variation for nestling growth in that population. In a cross-fostering experiment with starlings (Sturnus vulgaris) ( Smith & Wettermark, 1995), genetic effects were significant on days 1 and 14, but not in between. In both studies, home-reared and swapped nestlings were reared in similar environments (unmanipulated brood sizes). In our experiment, genetic effects during growth were mainly found under ‘good’ rearing conditions, which may not have been present in the studies named above.

Other studies on the genetics of growth have fitted data to growth curves. Rhymer (1992) found that growth rates in Mallard ducks (Anas platyrhynchos) varied within populations at the family level, but in other species, no genetic variation in growth parameters could be detected ( Ricklefs & Peters, 1981; Ricklefs, 1984a; Gebhardt-Henrich & van Noordwijk, 1994; Smith & Wettermark, 1995), except for asymptotic size ( Henrich, 1989; Smith & Wettermark, 1995). Curve fitting requires many precise data points to obtain good parameter estimates; small changes in single measurements can profoundly affect the estimates ( Zach, 1988). Insufficient sampling frequency and measurement error may therefore make it difficult to detect any genetic variation, especially for the shape parameter m of the Richard’s growth curve ( McCallum & Dixon, 1990). Alternatively, genetic variation for growth rates may be diminished by natural selection ( Falconer, 1989). For example, sibling competition and high predation risk are both assumed to select for rapid growth ( Werschkul & Jackson, 1979; Ricklefs, 1984b). As a consequence, remaining variation in growth rates would be largely nongenetic.

Little is known about the genetics of growth periods. Gebhardt-Henrich & Marks (1993) found genetic variation for the growth period of tarsus length in Japanese quail (Coturnix coturnix japonica) under poor experimental conditions, but not for weight or wing length. However, results from this laboratory study can not be applied to a natural situation. In a wild population of great tits, the growth period of weight did not show any heritable variation ( Gebhardt-Henrich & van Noordwijk, 1994). With the few data available to date, it is not possible to state whether growth periods commonly could respond to selection pressures, such as predation risk, food availability or competition.

Genetic variation in plasticity of nestling growth

Genetic variation in plasticity of nestling size was indicated by significant genotype–environment interactions in a three-factorial analysis. Furthermore, this term was significant for the growth period of weight. This means that reaction norms of growth have the potential to evolve in our study population.

Information on genotype–environment interactions for nestling growth in the wild are largely lacking. Significant interaction terms for growth and developmental time have been found for weight in Mallard ducks ( Rhymer, 1992), but not in Great tits or European starlings ( Gebhardt-Henrich & van Noordwijk, 1991; Smith & Wettermark, 1995). For adult size, genotype–environment interactions seem to be present in some bird populations ( van Noordwijk et al., 1988 ; Gebhardt-Henrich & van Noordwijk, 1991, 1994; Price, 1991; Rhymer, 1992; Larsson, 1993; Larsson et al., 1997 ; Merilä, 1997), but not in others ( Boag, 1983; Merilä & Gustafsson, 1996; Merilä, 1996a; Merilä & Fry, 1998).

Phenotypic plasticity in growth patterns

Compensatory growth

We observed compensatory growth in enlarged broods, where nestlings extended the period of weight increase as compared to nestlings in reduced broods. Compensatory growth under poor rearing conditions is also indicated by a decrease of environmental effects on size with age (e.g. Atchley, 1984; Riska et al., 1984 ), a trend we observed in all traits. Still, compensation for retarded growth was not complete for tarsus length and weight. Convergence of growth curves may be observed if rates or cessation of growth depend on body size rather than age (e.g. Riska et al., 1984 , and references therein).

Plasticity in growth patterns in response to variation in rearing conditions has been found in a number of avian species. In some cases, effects of rearing were permanent (zebra finch: Boag, 1987; great tit: Allander, 1995), but in other cases, complete compensatory growth occurred (blackbird: Berthold, 1977; tree-swallow: Wiggins, 1990; bee-eater: Emlen et al., 1991 ). Whether compensatory growth is possible may depend partly on the severity of growth conditions. For example, in contrast to our results, great tits growing up when feeding conditions were poor ceased weight increase earlier than individuals growing up under good feeding conditions ( Gebhardt-Henrich & van Noordwijk, 1994). Furthermore, in populations where food shortage does not occur frequently enough, flexible growth patterns, in the sense of slowed growth followed by compensatory growth, may simply not evolve ( Konarzewski et al., 1996 ).

However, lack of compensation could also be due to internal constraints. For instance, patterns of skeletal development seem to show little variation in birds in general, possibly due to strong dependencies among different stages of development ( Starck, 1993). In accord- ance with our results, no compensatory tarsus growth under poor rearing conditions has been observed in nestling crows (Corvus corone corone)( Richner, 1989) or tits (P. major: van Noordwijk et al., 1988 , P. montanus: Thessing & Ekman, 1994). In contrast, levels of fat reserves can be altered long after growth is completed ( Haftorn, 1992).

Resource allocation

Selection on adult size may be stronger for some traits than others, and in that way favours greater developmental stability, or target-seeking growth in these traits. Our findings that fledging wing length and body condition were given priority before weight and tarsus length may indicate the greater importance of the former traits for fledged young. Indeed, body weight and condition have been shown to be good predictors of fledgling survival in the blue tit ( Nur, 1984b) as well as in other species ( Hochaka & Smith, 1991; Magrath, 1991; Lindén et al., 1992 ). Unfortunately, in many studies, no distinction is made between body weight and body condition, but observations on collared flycatchers (Ficedula albicollis) and willow tits suggest that fat reserves are more important for juvenile survival than structural size ( Alatalo et al., 1990 ; Thessing & Ekman, 1994). Notably, wing length has been shown to affect take-off speed in starlings ( Swaddle et al., 1996 ). The ability to fly well directly after fledging is likely crucial to avoid predation by, e.g. sparrowhawks (Accipiter nisus), which prey heavily upon newly fledged blue tits ( Perrins & Geer, 1980; Geer, 1982). In contrast, no clear-cut pattern of selection on tarsus length has been found (selection for larger size: Garnett, 1981; Alatalo & Lundberg, 1986; Richner, 1989; Hõrak, 1994; selection for smaller size: Price et al., 1994 ; Meriläet al., 1997 ; stabilizing selection: Alatalo & Lundberg, 1986).

Protection of wing development at the expense of weight or skeletal growth has been found in several other studies on nestling growth ( Boag, 1987; Donazar & Ceballos, 1989; Gebhardt-Henrich & Marks, 1993; Dufva & Allander, 1996; Oyan & Anker-Nilssen, 1996; De Kogel, 1997; but see Husby, 1991). However, in contrast to our results, body condition has been found usually among the traits most affected by poor rearing conditions ( Richner et al., 1993 ; Merilä & Wiggins, 1995; Dias & Blondel, 1996; Merilä, 1996).


We found genetic variation for nestling size and for the period of weight increase in a wild population of blue tits. Furthermore, genotype–environment interactions suggested that reaction norms of growth patterns were heritable and, hence, that both growth and reaction norms had the potential to evolve in our study population. To this end, our study adds to the few other studies on wild avian populations, which have investigated whether the time window of juvenile growth or its plasticity has a heritable component. The degree of developmental stability, or the ability to compensate for retarded growth can vary among different body traits. We found that growing blue tits protected the development of wings and body reserves at the expense of other body traits, which might be a result of internal constraints, but could also be an adaptive strategy.


The authors thank N. Uhlin and G. Carlsson for assistance in the field, and A. Forsman, J. Merilä, K. Larsson and two anonymous reviewers for valuable comments on earlier drafts of this manuscript.