Studies of genetic variation in metabolic traits have so far not focused on birds. In our study population of captive zebra finches we found evidence for a significant heritable genetic component in basal metabolic rate (BMR). Heritability of all morphological traits investigated (body mass, head length, tars length and wing length) was significantly larger than zero. All traits were positively phenotypically correlated. Eight of 10 genetic correlations presented in this study differed significantly from zero, all being positive, suggesting the possibility of correlated responses to any selection acting on the traits. When conditioned on the genetic variance in body mass, the heritability of BMR was reduced from 25% to 4%. Hence, our results indicate that genetic changes in BMR through directional selection are possible, but the potential for adaptation independent of body mass may be limited.
In spite of being one of the most measured physiological traits in animals, surprisingly little is known about the genetics underlying basal metabolic rate (BMR), especially in birds. BMR is the minimal metabolic rate measured for an endothermic animal (IUPS Thermal Commission, 2001). Hence, BMR represents an animal's maintenance cost, and is useful as a physiological standard for animal performance.
Environmental factors like temperature and precipitation have been suggested to be important determinants of BMR (Lovegrove, 2003). These environmental factors are likely to change drastically in future because of the ongoing climatic changes. Hence, obtaining knowledge about the genetics underlying individual variation in BMR is an important task, because this might help us understand both previous selection pressure as well as the ability for species to energetically adapt to environmental changes expected in the future. There is large variation in BMR between different bird species, and this variation seems to be related to their habitat and their mode of living (Bennett & Harvey, 1987). There are also differences in metabolism between populations of the same bird species living at different latitudes (Wikelski et al., 2003; Broggi et al., 2005). Some of this variation is believed to stem from genetic differences between populations, and indicates that natural selection has acted on metabolism at least in the past (Furness, 2003).
The genetics of morphological traits in avian populations have been studied extensively, and most traits have been estimated to be moderately to highly heritable (see Merilä & Sheldon, 2001). On the other hand, there are few studies investigating the heritability of physiological traits, such as metabolism. As metabolic traits show phenotypic flexibility, it may be more difficult for selection to work upon them compared with more fixed morphological traits (Hayes & O'Connor, 1999). In spite of this phenotypic flexibility, several studies report significant repeatabilities of BMR in birds (Bech et al., 1999; Hõrak et al., 2002; Rønning et al., 2005; Vézina & Williams, 2005), showing that between individual variation exist, upon which selection could potentially work. The results from heritability studies of metabolic traits are, however, equivocal. There is evidence for a significant heritability of maximal oxygen consumption (Vo2max) in rodents (Dohm et al., 2001; Konarzewski et al., 2005; Nespolo et al., 2005; Sadowska et al., 2005). In humans both Vo2max and the Vo2max response to exercise training is found to be moderately heritable (Bouchard et al., 1998, 1999). On the other hand, most studies examining BMR in mammals have found the heritability to be absent or low (Lacy & Lynch, 1979; Dohm et al., 2001; Nespolo et al., 2003; Bacigalupe et al., 2004; Nespolo et al., 2005). However, two recent studies on rodents report a moderately high heritability of BMR (h2 ≈ 0.4, Konarzewski et al., 2005; Sadowska et al., 2005). Artificial selection on both BMR and aerobic capacity has proven effective in mice (Swallow et al., 1998; Henderson et al., 2002; Książek et al., 2004; Wisløff et al., 2005), supporting the presumption of the presence of additive genetic variance in metabolic traits. Heritability studies examining metabolism have mainly been conducted on mammals. We are not aware of any studies on heritability of metabolic traits in birds other than in domestic chicken (Gallus gallus), where Damme et al. (1986) found evidence for a genetic component in fasting metabolic rate, a trait which resembles BMR.
For a trait to be heritable some additive genetic variance has to be present (Falconer & Mackay, 1996). The presence of additive genetic variation in BMR would indicate that intraspecific BMR variation between animals living in different environments not only reflects phenotypic adjustments but may include some genetic differentiation as well. Because selection rarely operates on only one trait at a time, it is important to identify the size and sign of genetic correlations between traits. When there is a genetic correlation between traits, selection on one trait will influence traits with which it is correlated (Lande & Arnold, 1983). Hence, traits genetically correlated with BMR could influence the evolution of BMR through indirect selection.
The aim of the present study was to investigate the genetics underlying BMR in a bird species. We estimated heritability of, as well as the phenotypic and genetic correlations between, BMR and morphological traits in a laboratory population of zebra finches (Taeniopygia guttata).
Materials and methods
Maintenance and breeding
The study was conducted using captive zebra finches Taeniopygia guttata Vieillot, which is a small finch native to Australasia. Between breeding periods the birds were held in six large (10 m3) sex-specific walk-in aviaries. Ambient temperature in the rooms used both for breeding and non-breeding birds was 24 °C and the relative humidity was kept at 40%. There was a 12 : 12 h light–dark regime with light on at 07:00 local time. When not breeding, all birds were provided with a mixed seed diet (Life Care, Total Pet Care, Aalestrup, Denmark) and drinking water ad libitum. During the breeding periods the birds were provided a commercial protein supplement (Eggfood, Witte Molen, Moleneind, the Netherlands) in addition to the mixed seed diet.
The pedigree used in our study consists of 349 birds from three generations where the first generation birds (n = 63) came from three different breeders and was purchased through a local pet shop at the end of 2000 and early 2001. These birds were held in sex-specific aviaries until breeding. In the first generation birds a significant deviation from Hardy–Weinberg expectation was found in four out of six loci (Fisher exact test; P < 0.05, GENEPOP 1.2, Raymond & Rousset, 1995), indicating genetic differences between the birds from the different breeders. We have no information regarding potential kinship between the birds supplied by the individual breeders, but the mean observed heterozygosity (H) and mean number of alleles at the six typed microsattelite loci showed that the genetic variation presented was relatively large (mean observed H = 0.605; mean no. alleles = 16.7; n = 63). To increase the amount of genetic variation in each population and reduce the probability of inbreeding, the breeding aviaries were supplied with a random mix of birds from all three different breeders. The birds in the first generation bred in two periods, spring 2001 and autumn 2001/spring 2002. The second generation birds (n = 174) bred in autumn 2002 and spring 2003. During breeding periods males and females were placed together and provided with nest boxes inside the aviaries. All birds were breeding in large aviaries, each containing 10–12 breeding pairs, except for 21 pairs from the second generation that bred in separate breeding cages. Chicks were removed from their parents when they were fully independent at an age of 6 weeks and placed in separate sex-specific aviaries.
Some chicks (n = 106) in the second generation, hatched in autumn 2001/spring 2002, were raised under different feeding regimes. In the first 6 weeks after hatching these chicks were provided either a low quality diet (seeds only) or a high quality diet (seeds and commercial protein supplement, supplied with boiled eggs). The purpose of this diet manipulation was to examine the effect of food quality during early development (see Bech et al., 2004). Because we used the diet manipulated birds in the present study as well, we controlled for the effect of different diet quality in our analyses.
Basal metabolic rate and morphological traits
Basal metabolic rate was measured as O2-consumption rates using an open flow system. The measurements were conducted at night during the birds’ normal resting phase. Oxygen concentration in effluent air was measured using a Servomex Xentra, type 4100, two channel oxygen analyser (Servomex Controls Ltd., Crowborough, England). Four birds were measured simultaneously using four metabolic chambers (1.5 L), and a flow rate of 400 mL min−1. The measuring protocol is described in detail elsewhere (Rønning et al., 2005). The metabolic measurements were conducted in the period from March 2001 to March 2004. Most birds (n = 253) used in our analyses were measured only once. However, 32 birds were measured six times over a 2.5 years period. Because of the proportionally small sample size of birds with multiple measurements we were unable to obtain any estimates of the permanent environmental effect on BMR. As the repeatability of BMR was relatively high (R = 0.51 ± 0.09, P < 0.001, n = 32, calculated following Lessells & Boag, 1987) over the 2.5 years period (see also Rønning et al., 2005), we simply used the mean BMR values from these birds in our analyses. To prevent potentially metabolic adjustments during breeding from obscuring the metabolic measurements, no bird were measured earlier than 4 months after breeding.
Body mass was measured with an accuracy of 0.01 g using a Sartorius BP 610 digital weight (DWS, Elk Grove, IL, USA). Tarsus length and head length (including the bill) were measured using a slide calliper (to the nearest 0.1 mm), and wing length was measured to the nearest 1.0 mm using a standard ruler. All measures are of adult birds.
Genetic determination of parenthood
Because the majority of pairs used in this study were breeding together in large aviaries containing 10–12 pairs the parenthood of most birds was not certain and based on observations of social parents only. Blood samples (c. 25 μL) were collected from individual zebra finches and stored in 96% ethanol. Birds that died before the blood sampling were kept in a freezer (−25 °C). From these birds a small liver sample was collected and stored frozen. A Chelex (Sigma-Aldrich, St. Louis, MO, USA) based extraction procedure was used to extract DNA from the blood and liver samples. The subsequent polymerase chain reaction (PCR) amplification of polymorphic microsatellite loci was carried out on a GeneAmp PCR system 9700 (Applied Biosystems, Foster City, CA, USA). Each 10 μL PCR mix included 0.5 units of Taq DNA polymerase (AH Diagnostics, Aarhus, Denmark), 20 mm (NH4)2SO4, 75 mm Tris-HCl pH 8.8, 0.15 mg mL−1 DNAse free BSA, 10 mmβ-mercaptoethanol, 2.5 mm MgCl2, 0.14 mm dNTPs (AH Diagnostics), 0.7 μm of each of the forward and reverse primers, and approximately 20 ng of genomic DNA.
Individuals were genotyped on six microsatellite loci: Ase18 originally isolated from the Seychelles warbler, Acrocephalus sechellensis, genome (Richardson et al., 2000), BF03 originally isolated from the Bengalese finch, Lonchura striata var. domestica, genome (Yodogawa et al., 2003), FhU2 originally isolated from the pied flycatcher, Ficedula hypoleuca, genome (Primmer et al., 1996), INDIGO 29 and INDIGO 41 originally isolated from the village indigobird, Vidua chalybeata, genome (Sefc et al., 2001) and LS2 originally isolated from the loggerhead shrike, Lanius ludovicianus, genome (Mundy & Woodruff, 1996).
The products were separated by capillary electrophoresis on an ABI Prism 3100 Genetic Analyzer (Applied Biosystems). To resolve alleles, reverse primers were fluorescence-labelled with either 6-FAM (FhU2 and INDIGO 29) (Invitrogen, Carlsbad, CA, USA), VIC (BF03 and INDIGO 41) or NED (Ase18 and LS2) (Applied Biosystems). The software Genotyper v.2.1 (Applied Biosystems) was used to score alleles.
The parenthood analysis software Cervus 3.0 (Kalinowski et al., 2007) was used to determine the genetic parents of birds that hatched within each of the aviaries (see description of breeding set-up above). All the adult birds present in the aviary at the time of hatching of a given offspring were classified as a potential parent. Cervus used this information, combined with information on the estimated proportion of adult birds that was marked (i.e. 100% in this study), information on allele frequencies at the microsatellite loci, and the individual genotypes of offspring and their potential parents. The software then calculated a LOD score (log-likelihood ratio) for the likelihood that a pair of potential parents were the true genetic parents of the offspring. The difference in LOD scores of the two most likely potential parent pairs were compared with a criterion (Δ LOD) generated in the simulation module of Cervus. If the observed difference in LOD scores was larger than the criterion generated by Cervus, the assigned maternity or paternity was correct in at least 95% of cases (Marshall et al., 1998; Kalinowski et al., 2007). If a parent pair could not be determined with 95% confidence Cervus instead determined either maternity or paternity with 95% confidence, if possible. In addition to the aviary populations, where parenthood was determined genetically as outlined above, 40 offspring were born in cages where only one adult male and female were present. These birds were also included in the analyses. Hence, using these procedures we identified one parent of 14 and both parents of 229 of the 349 birds included in the pedigree. The birds in the first generation (n = 63) did not have their parents included in the population. Hence, one or both genetic parents were identified for 85.0% (i.e. 243 out of 286) of the descendants from birds in the first generation. The rate of parent assignment was 16% lower than the predictions from Cervus, possibly due to the existence of some null alleles and related parents.
Estimation of genetic parameters
The heritability of different traits and the genetic correlations between them were estimated using restricted maximum likelihood estimation procedures (VCE4 software; Neumaier & Groeneveld, 1998). We used an animal model, where the column vector of phenotypic values of n individuals (y) were expressed in terms of its additive genetic value and other random and fixed effects
where β’ and u were vectors of fixed (e.g. sex and diet quality) and random effects (e.g. additive genetic values), respectively, e was a vector of residual values, and X and Z were the corresponding incidence matrices (Lynch & Walsh, 1998). Because the animal model utilizes all relationships between individuals in a pedigree it is more powerful than conventional methods in estimating quantitative genetic parameters (Lynch & Walsh, 1998; Kruuk, 2004). Moreover, it can include fixed and random effects which otherwise may bias estimates of heritability, and allows use of unbalanced data sets with missing observations or missing pedigree links in the estimation procedure (Lynch & Walsh, 1998; Kruuk, 2004; Garant & Kruuk, 2005). Because the animal model corrects for the pattern of flow of genetic information across generations it also provides estimates that are less biased by the occurrence of selection, inbreeding or nonrandom mating (Sorensen & Kennedy, 1984, 1986).
In our models, the total phenotypic variance in a trait (VP) was partitioned into the sum of two components (VP = VA + VR), where VA was the additive genetic variance and VR was the residual variance. However, we accounted for differences between sexes (Table 1) and diet regimes (Bech et al., 2004) in phenotypic means by including sex and diet as fixed effects in the models where additive genetic variances and covariances (i.e. the G-matrix) were estimated. The common environment experienced by offspring growing up in the same brood or having the same social mother can be a substantial source of phenotypic variation, and may bias estimates of additive genetic variances and covariances (Kruuk, 2004). One way to test for this is to include either brood identity or identity of social mother as random factors in the animal model. The 286 descendants from the first generation birds in our pedigree originated from 170 different broods, where only one or two offspring were measured from most (i.e. 82%) broods. Consequently, preliminary analyses showed that the proportion of variance explained by the common brood environment was not significant for any of the traits. On the other hand, our model did not converge when maternal identity was included as a random factor. Hence, the variance because of maternal identity could not be properly quantified. Any variation because of maternal effects and common brood environment was therefore disregarded in the analyses presented.
Table 1. Mean and standard deviation (SD) of basal metabolic rate (BMR) and morphological traits in male and female zebra finches.
|BMR (mL O2 h−1)||136||38.122||3.695||149||41.102||4.166||40.510||1|| < 0.001 |
|Head length (mm)||152||23.357||0.662||176||23.197||0.677||4.655||1|| 0.032 |
|Tarsus length (mm)||152||14.153||0.627||178||14.171||0.636||0.063||1||0.802|
|Wing length (mm)||152||57.125||1.788||178||56.916||1.627||1.238||1||0.267|
|Body mass (g)||158||14.236||1.574||190||14.661||1.602||6.158||1|| 0.014 |
The heritability (h2) of each trait was calculated as the ratio of the additive genetic variance to the total phenotypic variance (Falconer & Mackay, 1996). In order to test whether estimates of heritability or genetic correlations were significantly different from zero, we calculated z-scores
where x was the estimate of either heritability or genetic correlation, and σ was the respective standard error. The resulting z-scores were then tested against a large sample standard normal distribution. All significance tests were two-tailed. Please note that calculating significance based on z-scores assumes a normal distribution of the data, hence the significance levels reported should be interpreted with some caution.
For comparison of additive genetic and residual variance across different traits we calculated coefficients of additive genetic variance (CVA) and residual variance (CVR) following Houle (1992). To estimate the amount of additive genetic variance in BMR that was independent of the genetic variance in body mass, we calculated conditional genetic variance (VA(y|x)) in which the additive genetic variance in BMR (y) conditional on the genetic values of body mass (x) was
where VA(y) and VA(x) were the additive genetic variance in BMR and body mass, respectively, and CovA(xy) was the additive genetic covariance between the traits (Hansen et al., 2003; Jensen et al., 2003). The heritability of BMR conditional on body mass (i.e. conditional heritability) was then calculated, substituting VA for VA(y|x).
Basal metabolic rate (mL O2 h−1) and all morphological traits analysed were found to have an h2 significantly different from zero (Table 2). In a model including only sex-specific male and female BMR, the heritability of BMR did not differ between the sexes (z = 1.55, P = 0.12), and the correlation between male-BMR and female-BMR was not different from one (z = −1.33, P = 0.18, data not shown). Because BMR was highly genetically correlated with body mass (Table 3), we calculated the conditional heritability of BMR from the VA in BMR that was independent of the additive genetic variance in body mass. Accordingly, the conditional heritability of BMR was 0.041. Thus, BMR retained 16.4% of its additive genetic variation when conditional on body mass, suggesting a rather limited genetic variation in this trait that is independent of body mass.
Table 2. Estimates of additive genetic variance, residual variance, heritability (h2), coefficient of additive genetic variance (CVA) and coefficient of residual variation (CVR) of basal metabolic rate (BMR) and morphological traits in zebra finches.
|BMR (mL O2 h−1)||285||39.68 (4.21)||3.794||11.364||0.250 (0.043)***||4.91||8.50|
|Head length||328||23.27 (0.67)||0.156||0.253||0.382 (0.050)***||1.70||2.16|
|Tarsus length||330||14.16 (0.63)||0.120||0.252||0.322 (0.046)***||2.45||3.55|
|Wing length||330||57.01 (1.70)||1.358||1.487||0.477 (0.046)***||2.04||2.14|
|Body mass||348||14.47 (1.60)||0.742||1.465||0.336 (0.059)***||5.95||8.36|
Table 3. Phenotypic correlations (above) and additive genetic correlations (below) between basal metabolic rate (BMR)and morphological traits in zebra finches.
|Head length||0.271***|| || || |
|0.751 (0.117)***|| || || |
|Tarsus length||0.215***||0.211***|| || |
|0.429 (0.129)***||0.188 (0.102)|| || |
|Wing length||0.251***||0.314***||0.173**|| |
|0.512 (0.122)***||0.435 (0.096)***||0.289 (0.081)***|| |
|0.914 (0.081)***||0.755 (0.109)***||0.180 (0.135)||0.429 (0.106)***|
Correlations between traits
There were positive phenotypic correlations between all traits examined in our study (Table 3). Furthermore, among the 10 additive genetic correlations estimated in this study, eight were significantly larger than zero. There were for example positive and highly significant genetic correlations between BMR and all morphological traits. Furthermore, when comparing genetic and phenotypic correlations there seemed to be some differences regarding the magnitude of the correlations, but none of the correlations were of different directions, all being positive (Table 3). For example, the genetic correlations between BMR and morphological traits were stronger than the corresponding phenotypic correlations, suggesting that the environment disrupts the rather strong underlying genetic link between BMR and general morphological size.
To our knowledge, this is the first study to examine the genetic basis of BMR in birds. We found a significant heritable component of individual variation in BMR in a captive zebra finch population (Table 2). In contrast to physiological traits, there is a large number of studies investigating heritability of morphological traits in avian populations, with tarsus length, wing length and body mass being some of the most common traits studied (Merilä & Sheldon, 2001). Although we found highly significant heritabilities in all morphological traits examined in our study, the estimated h2 values for the captive zebra finches were somewhat lower than the values typically reported in wild populations, which often are between 0.4 and 0.6 (Merilä & Sheldon, 2001). This could be caused by a true lower relative amount of additive genetic variance, or it could be due to a somewhat higher measurement error in these traits in the zebra finches compared with other bird species. The latter is highly unlikely, because all morphological measurements were conducted by the same person to ensure minimum measurement error. Another aspect is that we were not able to estimate and account for any maternal effects in our models. Preliminary analyses suggested, however, that environmental effects because of common brood were negligible (see Materials and methods). Moreover, if any maternal effects were present and not controlled for, this would lead to an overestimation of the heritabilities reported. We also have to emphasize that our study was conducted under laboratory conditions and as a consequence may not be directly comparable with studies on wild populations (Simons & Roff, 1994; Weigensberg & Roff, 1996; Blanckenhorn, 2002).
Phenotypic and genetic correlations
Basal metabolic rate scales with body mass in animals, and BMR was as expected positively phenotypically and genetically correlated with body mass and size-related traits in zebra finches (Table 3). The strong genetic correlations indicate pleiotropic effects of genes coding for e.g. BMR and body mass, or that genes coding for these traits are in linkage disequilibrium (Lynch & Walsh, 1998). All significant phenotypic and genetic correlations between morphological traits were positive. Thus, BMR and the morphological traits studied, if selected in the same direction, do not constrain each other either phenotypically or genetically. Another consequence of these positive correlations is that direct selection on one of the traits would result in an indirect selection on other traits (Lande & Arnold, 1983; Price & Langen, 1992).
Selection on basal metabolic rate
Data on genetics of avian energetics is almost absent in the literature. However, in domesticated chicken, fasting metabolic rate was found to be moderately heritable (Damme et al., 1986). Studies on mammals have given ambiguous results regarding heritability of BMR. Some studies have estimated the heritability of BMR to be low and not significantly different from zero, and it has been suggested that this is because of either previous strong natural selection (Lacy & Lynch, 1979) or very high environmental variance in the trait (Nespolo et al., 2003, 2005; Bacigalupe et al., 2004). In our study BMR showed the highest CVR (Table 2), indicating that environmental variance in this trait could be high compared with less flexible morphological traits. However, another aspect that could have inflated the residual variance in BMR is that measurement error is likely to be greater in metabolic measurements relative to morphological measurements. Two recent studies of rodents, however, report a moderately high h2 (c. 0.4) for BMR (Konarzewski et al., 2005; Sadowska et al., 2005). Furthermore, artificial selection experiments on rodents have proven effective for BMR (Selman et al., 2001; Książek et al., 2004), giving further evidence for a heritable genetic component of BMR because a response to selection depends on a genetic basis for variation in the trait (Falconer & Mackay, 1996). Results in the present study (Table 2) add to the number of studies indicating that individual metabolic intensity in general is transferred across generations, and suggests that this is true for both mammals and birds.
The whole body BMR is essentially determined by the metabolic intensity and mass of all different tissues and organs of the body. Consequently a strong correlation between BMR and body mass can be expected. BMR showed a highly significant heritability in our study, and CVA was high compared with the size-related traits (Table 2). However, because of the strong genetic correlation between body mass and BMR, selection on this trait would be strongly influenced by selection on body mass as well. Although environmental changes might result in directional selection in some traits, it is believed that most traits are likely to be under stabilizing selection (e.g. Hansen, 1997). The potential strength of directional selection in a trait is thus largely determined by the proportion of its genetic variation that is not constrained by stabilizing selection on genetically correlated characters (Hansen et al., 2003). Thus, it is interesting to look at the heritable genetic variation in BMR that is not shared with body mass. We therefore calculated conditional heritability of BMR controlled for the additive genetic variance shared with body mass. The heritability of BMR was substantially reduced when conditional on body mass (i.e. accounting for the strong genetic link between these two traits). Nevertheless, some heritable genetic variation in basal metabolism remained. A possibility for an independent adaptation in this metabolic trait seems to be present, but may be a slow process. These results suggest that observed differences in BMR between populations of similar sized animals could to some extent be a result of genetic differentiation because of different selection regimes.
Basal metabolic rate in birds is known to show physiological adaptations to ecological conditions and different life cycle events (e.g. Langseth et al., 2000; Lindström & Rosen, 2002; Lindström & Klaassen, 2003). A difference in resting metabolic rate among four different stonechat (Saxicola torquata) populations living at different latitudes persisted when nestlings were raised under identical conditions (Wikelski et al., 2003). This common garden study provides evidence for a genetically determined difference in metabolic rate, and indicates that different environments induce different selection regimes causing variation in BMR among bird populations (Furness, 2003). Although some studies indicate that genetic differentiation is one reason for inter-population variation in BMR it is not fully established if this is a general result, or if metabolic differences between populations more often reflect acclimatization to local environments. In any case, our study shows that a potential for BMR to change genetically is present. However, because BMR is likely to be a complex polygenetic trait it may be difficult to fully predict how this trait will evolve under different selection regimes.
In conclusion, our study shows that basal metabolism is positively correlated both phenotypically and genetically to several morphological traits in the zebra finch. All significant correlations were positive, implying that BMR and the morphological traits studied do not constrain each other either phenotypically or genetically, if selected in the same direction. Furthermore, in this study we found evidence for a heritable genetic component in BMR. Although we have not examined how selection acts on BMR, our results show that there is additive genetic variation in this metabolic trait, giving it an opportunity to change genetically across generations. However, only a small part of the additive genetic variance in BMR was independent of the genetic variance in body mass. Thus, a potential for adaptation in BMR independent of body mass may be present, but it seems to be limited.
We thank O.A. Indset, B. Simensen and voluntary students for providing the housing condition for the zebra finches. Thanks also to two anonymous reviewers for providing useful comments on an earlier version of this manuscript. The study was supported by grants from the Norwegian Science Research Council (Grant nos 138698/410 and 159584/V40).