Metabolic rates of aggressive and submissive phenotypes are colour blind in the polymorphic Gouldian finch

Some of the earliest evidence that social status of free-living vertebrates covaried with resting metabolic rate (RMR) came from studies of winter flocks of small passerines. The social ranks of three species correlated with their daytime RMR, with the highest-ranked individuals consistently having the highest RMR (Roskaft et al., 1986; Hogstad, 1987). Similar associations between dominance/aggression and energy metabolism have been found in mammals (Turbill et al., 2013), lizards (Friessen et al., 2017), fish (Metcalfe et al., 1995; McCarthy, 2001) and crustaceans (Brown et al., 2003). A recent meta-analysis evaluated the extent of covariance between maintenance metabolic rates and a variety of behavioural traits from over 70 studies (Mathot et al., 2019). It concluded that traits associated with net energy gain (dominance, foraging) were among those having the strongest positive association with metabolic rate. As these were correlative studies, robust conclusions about the functional significance to these outcomes are unwarranted, as are assumptions that particular behavioural traits associated with dominance promotes higher metabolism or vice versa (Biro and Stamps, 2010).

Studies of inter-individual differences in metabolic rate and behavioural traits of endotherms are more often based on measurements of resting (RMR) than basal metabolic rates (BMR). While both pertain to inactive animals, they can differ markedly owing to the much broader range of conditions permitted for determining RMR. BMR measurements stipulate use of adult (non-growing) animals that are: non-breeding; exposed to thermoneutral temperatures (thus not requiring regulatory thermogenesis or evaporative cooling); post-absorptive; and asleep during the inactive phase of their circadian cycle (e.g. at night for diurnal species; Benedict, 1938; Aschoff and Pohl, 1970). These conditions specify an endotherm's minimal cost of self-maintenance and provide a standardised physiological status for comparisons of metabolic rate between individuals, populations or species (Hulbert and Else, 2004). By contrast, RMR measurements only require that animals be measured while inactive without regard to factors such as time of day, reproductive status, state of digestion, temperature, etc., all of which can elevate metabolic rate above BMR and thus confound comparative studies of metabolic rates.

Irrespective of which form of metabolic measurement is used to determine dominance costs, of greater concern is that RMR and BMR are almost by necessity measured on isolated individuals. Thus, such measurements may not accurately indicate metabolic rates in normal social conditions. Experiments on dominant fish housed with a subordinate found that the metabolic rates of subordinates increased more than that of the dominants, and the magnitude of increases in the metabolic rates of subordinate individuals was directly related to the level of aggression expressed by the dominant (Sloman et al., 2000). Similarly, Millidine et al. (2009) found that the rate of opercular movement (a proxy for metabolic rate) in individual fish within groups was size related, with opercular movement decreasing in large fish surrounded by smaller fish but increasing in small fish surrounded by larger individuals (Millidine et al., 2009). These examples demonstrate the need to consider the metabolic consequences of dominant–subordinate interactions before assigning dominance costs based on metabolic measurements of animals in isolation.

Accordingly, there remain many unanswered questions regarding the extent of difference in metabolic rates between aggressive and submissive phenotypes and how these might be influenced by social history as well by social context. A potentially fruitful experimental approach is to examine species with naturally evolved and genetically distinct phenotypes in appearance and behaviour. For example, in sympatric sibling cichlid species there was no metabolic difference between males of the dominant red-coloured species and blue-coloured males of the submissive species when measured in isolation (Dijkstra et al., 2011). By contrast, the red species had unexpectedly lower energy expenditure per unit effort than the blue species during aggressive responses while viewing another fish (Dijkstra et al., 2011). Gouldian finches [Erythrura gouldiae (Gould 1844)], a highly social estrildid finch native to tropical savannah woodlands in northern Australia, provide another excellent opportunity to examine the influences of social history and social context on the metabolic rate of behaviourally and morphologically distinct individuals. This species is unusual in that the two predominant colour morphs, red-headed and black-headed, occur sympatrically and have ratios within free-living populations that are spatially and temporally consistent (Gilby et al., 2009). The colour polymorphism, which occurs in both sexes, has provoked much interest in characterising traits that could account for its persistence (e.g. Kokko et al., 2014). Recent work has identified an extremely limited (approximately 72 kbp) level of genomic divergence between red- and black-headed morphs (Toomey et al., 2018; Kim et al., 2019). It is believed that this ‘red locus’ acts as a regulatory region that controls a pleiotropic gene, affecting a range of morphological, physiological and behavioural phenotypes, and is maintained by balancing density-dependent selection (Kim et al., 2019). For example, red-headed males respond to nutritional and social stresses with greater corticosterone secretion and reduced immune responsiveness than black-headed males (Pryke et al., 2007, 2012) and also have much higher testosterone levels when placed in competitive environments (Pryke et al., 2007). The two morphs also differ in levels of oxidative stress biomarkers following exposure to high temperatures (Fragueira et al., 2019). Multiple studies in both captive and field conditions show red-headed morphs of both sexes to be significantly more aggressive than black-headed morphs when contesting limited resources (Pryke and Griffith, 2006, 2009; Pryke, 2007; Williams et al., 2012; Brazill-Boast et al., 2013).

Here, we characterised three indices of aerobic metabolism of red- and black-headed first-year virgin males: daytime resting metabolic rate (RMR), nocturnal basal metabolic rate (BMR) and exercise-induced maximum metabolic rate (MMR). We also examine the influence of social history by repeated measures of RMR of birds never exposed to opposite colour morphs (intermorph naïve) and after opposite morphs shared a cage (intermorph exposed), both when isolated and while viewing unfamiliar birds of each colour morph. Quantifying both BMR and MMR allowed determination of absolute aerobic scope (AAS), which is believed to represent the aerobic power in excess of maintenance costs that is available for other activities (Brett, 1972). In some species, AAS better predicts dominance rank than proxies of either minimal or maximal oxygen consumption rates (Killen et al., 2014).

Given the putative linkages and the implied functional association between dominance and elevated metabolic rate (Biro and Stamps, 2010; Mathot et al., 2019), and the findings of Killen et al. (2014), we predict that BMR, RMR and likely AAS will be higher in red-headed than in black-headed morphs.

We used 24 black-headed and 24 red-headed captive-reared wild-type male Gouldian finches. Birds had been raised by parents of the same colour and, upon fledging, were housed alone in complete social isolation (Pryke et al., 2007). While in seclusion and prior to experiments, juveniles transitioned from the drab grey/olivaceous immature plumage common to both morphs to the brightly coloured adult plumage of their male parent. Two cohorts of first-year virgin males were studied (one in 2007 and one in 2009); both involved 12 intermorph naïve birds of each colour morph. Upon arrival at the University of Wollongong, visual isolation of the two colour morphs was maintained by partitions between cages. Birds of the same colour morph were placed two per cage (34×45×45 cm; Hoei Cage Co., Japan). Cages were fitted with multiple perches and one dispenser each for commercial finch seed (Golden Cob, Mars Birdcare Australia), mineralised grit and water, all available ad libitum. Individuals of both cohorts remained with their initial cage partners for 2 weeks before metabolic measurements commenced.

All aspects of animal care and their use in experimental procedures were approved by the University of Wollongong Animal Ethics Committee.

This involved two sets of metabolic measurements that took place on consecutive days (tests 1 and 2), with test 1 involving a pair of birds of the same or opposite colour morphs; for test 2, each bird was exposed to a morph of the opposite colour of that viewed in test 1. Experimental pairs were randomly selected from all birds other than either bird's cage partner. The sequence in which each bird viewed the same or opposite colour morph was varied among birds to balance morph colour combinations for tests 1 and 2.

Upon completion of phase 1, birds were transferred to cages containing an individual of the opposite colour morph for 5 days. Subsequently, two sets of metabolic measurements (tests 3 and 4) were made on consecutive days (we define these individuals as ‘intermorph exposed’). As in phase 1, the sequence in which each bird viewed the same or opposite colour morph to itself on consecutive days was varied among all birds to balance morph colour combinations for tests 3 and 4.

Birds from the 2009 cohort were placed as mixed morph groups in outdoor flight cages (2.5×3×4 m, 12 birds per cage) following completion of experiment 1. Cages had multiple feeding sites, extensive perching locations and shelter from weather. After at least 4 days in aviaries, individuals were collected about mid-morning through midday to measure exercise-induced MMR. Upon completing MMR measurements, birds were placed in individual holding cages with free access to food and water. The following day, food was removed three hours before initiation of nocturnal measurements of BMR.

We measured RMR during the day using an open-circuit respirometry system that allowed us to continuously monitor two metabolic chambers simultaneously. For a given metabolic series (tests 1 through 4), birds were removed from holding cages, had body mass (M_b) measured on a digital scale (±0.01 g; model OHAUS AV413C), and were then placed individually in 1.5 l chambers fashioned from sealable polycarbonate plastic food containers. Chambers were optically transparent, rectangular and fitted with a perch at one-third height, as well as inlet and outlet tubes for airflow. The chamber containing the first bird was positioned in the constant-temperature cabinet and then screened with an opaque partition before positioning the second bird's chamber within 2 cm of one another. The perches were oriented perpendicularly to the partition, resulting in mutual viewing irrespective of perching location. Overhead fluorescent lighting within the temperature-controlled cabinet illuminated the birds evenly and a webcam was used to monitor behaviour.

Cabinet temperature was regulated at 30°C, which is within the thermoneutral zone for Gouldian finches (Burton and Weathers, 2003). Mass-flow controllers (Mykrolis, model FC-2902V-T) provided a constant air supply of 500 ml min^–1 STP of dry air into each metabolic chamber. The outflows of each chamber and sequential sampling of inlet air were subsampled (approximately 100 ml min^–1). Subsampled air was passed through Drierite and soda lime to remove water and CO₂, respectively, en route to a two-channel O₂ analyser (Oxilla II; Sable Systems International, Henderson, NV, USA). We used LabHelper software (warthog.ucr.edu) to control the multiplexer outputs and read chamber O₂ concentration at 1 s intervals. We used LabAnalyst (warthog.ucr.edu) to correct oxygen readings for drift between consecutive baselines and to calculate O₂ consumption rates (V̇_O2, ml min⁻¹) according to eqn 2 of Hill (1972).

Oxygen consumption was recorded continuously after closure of the cabinet door. Video monitoring showed that birds usually settled within a few minutes, and their metabolic rates declined steadily, reaching a plateau approximately 30 to 60 min after door closure. Approximately 10 min after both birds exhibited a stable RMR, the partition was lifted so birds could see one another, and V̇_O2 was recorded for a further 60 min. RMR was identified as the lowest 3-min running mean V̇_O2 recorded over the entire measurement period. This duration was selected as birds tended to be restive under full illumination and periods of sustained rest levels of metabolic rate rarely exceeded 5 min. Peak MR (MR_peak) was designated as the highest 3-min running mean V̇_O2 recorded after the partition was lifted. These measurements were used to evaluate the maximum factorial increase in MR associated with birds viewing one another, which we calculated as MR_peak/RMR. The average V̇_O2 over the entire 60-min period that birds viewed one another was designated as MR_mean.

Procedures used for measuring BMR and exercise-induced MMR are described in detail elsewhere (Buttemer et al., 2019). In brief, for MMR tests, birds were collected from the flight cages and placed individually in a hop-flutter wheel that was rotated to elicit maximal oxygen consumption during exercise. Rotation speed was dynamically adjusted to each bird's pattern of movement to achieve maximal activity until they exhibited exhaustion. Data were adjusted with ‘instantaneous’ conversion procedures to account for gas mixing characteristics of the wheel and accurately resolve short-term changes in V̇_O2 (Chappell et al., 1999; Buttemer et al., 2019). The peak 30-s instantaneous rate of V̇_O2 during exercise was designated as MMR. Following MMR tests, birds were placed in holding cages (two birds per cage) with free access to food and water. BMR tests were performed on these birds the following evening, with food removed from their holding cages 3 h prior to measurements. Birds were placed in 2-l metal metabolic chambers for overnight measurements of V̇_O2 under thermoneutral conditions (30°C) in total darkness. BMR was calculated as the minimum 5-min running average V̇_O2 over the entire night; BMR obtained using this averaging interval was significantly repeatable in another study of similar-sized passerine birds (Careau et al., 2014a).

Analyses were conducted using linear mixed models in ASReml-R version 4 (Butler et al., 2018). All continuous variables (M_b, RMR, MR_mean and factorial increase in MR) were log₁₀ transformed to improve normality of the residuals and subsequently standardised to a mean of 0 and a variance of 1 to facilitate model convergence and allow direct comparison of effect sizes (in units variance) across variables. All models included a fixed effect of metabolic test sequence as a continuous variable (1 to 4). This controlled for potential temporal trends and habituation across the four sequential metabolic tests of phases 1 and 2 of this experiment, but included different sets of fixed effects. In these paired metabolic measurements, each bird represented the other bird's opponent. For the factorial increase in MR and MR_mean, we included all possible interactions between: (1) M_b, (2) colour morph of the focal individual, (3) intermorph social status and (4) colour morph of the opponent. We initially included the body mass of the opponent (M_b-opponent), but it was never significant and was left out of the final model. Significance of fixed effects was tested with a conditional Wald F-statistic and the denominator degrees of freedom (d.f.) were determined following Kenward and Roger (1997). The two colour morph variables (focal and opponent) were coded as centred continuous variables, with black=−1 and red=1. Similarly, the intermorph social state of focal individuals were coded as centred continuous variables, with naïve=−1 and exposed=1. Therefore, all main effects are estimated for the average bird and are biologically interpretable even in the presence of interactions in the model (Schielzeth, 2010; Cleasby and Nagagawa, 2011).

To account for the non-independence in the data (each bird had four sequential metabolic evaluations in phases 1 and 2 of this experiment, except one that had three), we included the identity of the focal individual as a random effect and therefore partitioned the phenotypic variance into the among-individual variance (V_ID) and residual variance (V_e). We also included the identity of the opponent individual as a random effect to capture any variance (V_ID-opponent) that might be caused through indirect effects of particular opponents. A substantial V_ID-opponent component would imply that some individuals consistently triggered larger or smaller increases in the MR of the focal individual. We also included a correlation between the direct (V_ID) and indirect (V_ID-opponent) effects (r_{focal–opponent}). If positive, this would imply that individuals who have higher factorial increase in MR (direct effect) also trigger a greater increase in others (indirect effect). If negative, then this would mean that individuals that have a higher factorial increase in MR trigger a smaller increase in opponents. Although the model may seem complex, it easily converged in only six iterations and there were no signs of overfitting (all but one of the 48 individuals were measured four times, for a total of 191 observations, with varying focal–opponent combinations, such that all parameters were estimable). Finally, we calculated repeatability as the ratio of V_ID over total phenotypic variance (conditioned on the fixed effects). For M_b and RMR, we used the equation R=V_ID/(V_ID+V_e), whereas for factorial increase in MR we used R=V_ID/(V_ID+V_ID-opponent+V_e). Note that repeatability estimates should be interpreted as being conditioned on the variables and interactions included as fixed effects in the model (Wilson, 2018). The approximate standard error of R estimates was calculated using the vpredict() function (Butler et al., 2018). Unless stated otherwise, all values are presented as means±1 s.e.m.

The influence of colour morph on BMR, MMR, factorial aerobic scope (FAS) and AAS was determined using linear mixed models that included body mass, morph colour and their interactions as independent variables.

RMR increased throughout experiment 1, as indicated by a significant and positive effect of metabolic test sequence (Table 1 and Fig. 1D). The two-way interaction between M_b and intermorph social state was statistically significant (Table 1) such that RMR increased with M_b in both colour morphs when intermorph naïve (Fig. 1A), but there was no relationship in either group when intermorph exposed (Fig. 1B). No other main effect or interaction was significant (Table 1). Black-headed and red-headed morphs did not differ from each other in RMR at either intermorph social stage (Fig. 1C), with black-headed and red-headed birds averaging 1.03±0.03 and 1.04±0.04 ml O₂ min⁻¹, respectively, when intermorph naïve, and both averaging 1.08±0.03 ml O₂ min⁻¹ when intermorph exposed. Among- and within-individual variance estimates in RMR were very similar, resulting in a moderate repeatability estimate of R=0.474±0.086 (estimate±s.e.).

Body mass at the time of RMR determinations did not differ between colour morphs (F_1,42.7=1.833, P=0.18), but did increase significantly for both groups after transfer to cages with a different colour morph (F_1,136.8=13.06, P=0.0004). Body mass averaged 15.68±0.12 g for both morphs during phase 1 RMR measurements and 16.37±0.13 g during phase 2.

The 3-min peak factorial increase in MR after the partition was lifted (MR_peak/RMR) declined throughout this experiment, as indicated by a significant and negative effect of metabolic test sequence (Table 2 and Fig. S1d). As for RMR, the two-way interaction between M_b and intermorph social state was highly significant (Table 2), but this time it had a negative estimate, which means that the factorial increase in MR was not related to M_b in naïve individuals (Fig. S1a), but positively related to M_b in intermorph exposed individuals (Fig. S1b). The main effect for the variable ‘opponent’ was significant and positive, revealing that red-headed morphs triggered a higher increase in MR than black-headed morphs (Table 2). The ‘intermorph social state×opponent’ interaction was also significant with a negative estimate (Table 2), indicating a weaker effect of seeing a red-headed morph in intermorph exposed than naïve birds. This was clearly evident in both morphs, with metabolic reaction of naïve birds being much higher when viewing red-headed than black-headed opponents (Fig. 2). Using data on naïve birds only, the difference between viewing a red-headed versus black-headed was similar in red-headed morphs (estimate=0.29±0.13) and black-headed morphs (estimate=0.26±0.13) (note: otherwise the ‘focal×intermorph social state×opponent’ interaction would be significant). By contrast, the metabolic response to red-headed opponents was significantly reduced in both morphs when intermorph exposed and did not differ from their reaction to viewing black-headed morphs, which was unaffected by intermorph social history (Fig. 2). There was significant among-individual variation in MR_peak, resulting in a repeatability estimate of R=0.301±0.082. Interestingly, there was also significant variance associated with the identity of the opponent (V_ID-opponent), and such indirect effects were positively correlated with direct effects (r_{focal–opponent}; Table 2).

Unlike the limited periods of MR_peak when birds viewed one another, MR_mean over the entire 60-min viewing period was unaffected by morph colour, intermorph social history or metabolic test sequence (Table 3). The only significant interaction was between body mass and intermorph social state, but the estimate was negative (opposite effect from MR_peak). Similarly, the estimates for some of the interactions that approached significance were of opposite sign (Table 3). Repeatability of MR_mean was R=0.528±0.80, and the V_ID-opponent values were low and not significantly greater than 0 (Table 3).

Colour morphology was unrelated to BMR, exercise-induced MMR, as well as AAS and FAS (MMR–BMR and MMR/BMR, respectively; Table S1). Body mass had a significant effect on MMR (Table S1), but did not differ between colour morphs (t₂₂=1.63, P=0.117). The mean variable values were: M_b=15.3±0.2 g, BMR=0.74±0.03 ml O₂ min⁻¹, MMR=4.48±0.12 ml O₂ min⁻¹, AAS=3.74±0.12 ml O₂ min⁻¹ and FAS=6.17±0.25.

All birds used in this study were bred from the same captive population used for studies that revealed consistent differences between black-headed and red-headed morphs in aggression, dominance and a range of physiological traits (Pryke and Griffith, 2006; Pryke, 2007; Pryke et al., 2007, 2012). Given that consistent individual differences in behaviour (e.g. Dingemanse et al., 2002; Petelle et al., 2015; Edwards et al., 2017) and metabolic rate (e.g. Sadowska et al., 2005; Ronning et al., 2007; Nilsson et al., 2009) are known to be heritable and that dominant individuals are typically associated with higher metabolic rates (Biro and Stamps, 2010; Mathot et al., 2019), the lack of difference between the two morphs in all our metabolic measures was unexpected.

Our initial measurements of RMR involved both morphs while intermorph naïve, and these revealed RMR to be indistinguishable between them. This alone suggests that elevated RMR among individual Gouldian finches was not an antecedent of their subsequent dominance status, unlike the ontogeny of dominance documented in some salmonid fish (Metcalfe et al., 1995; Cutts et al., 1999). Based on the well-characterised differences in dominance between the two morphs, it was surprising that intermorph social history did not differentially affect RMR. This contrasts with the responses of juvenile brown trout (Salmo trutta) sharing a tank with another fish, with metabolic rates of dominant fish decreasing and those of subordinates increasing after being paired for 24 h (Sloman et al., 2000). Similarly, Cristol (1995) found that transfer of individual dark-eyed juncos (Junco hyemalis) between captive flocks changed the RMR of dominant and submissive birds differently, even in controls that were returned to their original group. In our study, the main effect of intermorph social history on RMR was a change in the relationship between body mass and RMR. However, this response was similar in both morphs, with each morph showing an increase in RMR with respect to body mass when intermorph naïve (Fig. 1A), but not when intermorph exposed (Fig. 1B). For both morphs, thermoneutral RMR was 43% higher than BMR, which is consistent with night-time thermoneutral metabolic rate differences between fasted birds compared with daytime measures after feeding and photostimulation (Buttemer et al., 1986).

Our procedures for measuring BMR eliminated metabolic variance resulting from differences in digestive status, and were conducted during the nocturnal rest phase of the circadian cycle. Hence, the BMR data indicate ‘true’ minimum metabolic rate. BMR has previously been determined for Gouldian finches (Marschall and Prinzinger, 1991; Burton and Weathers, 2003), but those studies did not distinguish between colour morphs. Our BMR measurements were 6 and 11% higher than those made by the latter two studies, respectively, and were statistically indistinguishable between the two morphs when intermorph exposed. This reinforces our conclusion that differences in dominance between the two morphs are unrelated to metabolic rate.

As previously discussed, metabolic tests on isolated individuals cannot account for potential effects of social interaction. Therefore, we examined the metabolic consequences of birds viewing an unfamiliar red- or black-headed morph to identify potential costs of behavioural differences in a more natural context. In intermorph naïve birds, both morphs reacted more to the presence of a red-headed than a black-headed morph. This is consistent with experimental evidence that physiological and behavioural responses to red are genetic traits in Gouldian finches. In dyadic feeding contests between intermorph naïve juvenile males that were unmarked and those whose head colour was artificially coloured either red, black or blue, red colouring elicited far greater submissive behaviour and higher corticosterone secretion in unmanipulated individuals, regardless of either bird's genetic makeup (Pryke, 2009). This suggests an innate sympathetic nervous response to viewing a red-headed individual, which might be expected to elicit a rise in metabolic rate. However, we found no effect of opponent head colour on MR_peak in intermorph exposed birds. Importantly, the extent and duration of these metabolic rate rises during these 60-min viewing periods were small. This is reflected by the lack of difference in MR_mean between morphs when viewing another bird, regardless of intermorph history or opponent head colour.

Finally, we examined the possibility that differences in dominance between morphs might be correlated with maximal metabolism in forced exercise (MMR) or with absolute or factorial aerobic scope (MMR–BMR or MMR/BMR, respectively). Although the highest aerobic metabolic rates of volant species are apparently achieved during flight (McKechnie and Swanson, 2010), we assume that MMR determined using a hop-flutter wheel is a useful index of aerobic performance for several reasons. First, MMR measurements made with hop-flutter wheels are highly repeatable over time, implying that birds tested with this method reach consistent aerobic limits (Chappell et al., 1996; 2011; Careau et al., 2014a). Second, hop-flutter-based MMR values of free-living house sparrows progressively declined after they were transferred to small cages restricting their locomotor activities (Buttemer et al., 2008), as was found for maximal V̇_O2 of humans after reducing aerobic exercise training (Neufer, 1989). Finally, hop-flutter MMR values are substantially higher than maximum theromogenic V̇_O2 in birds (McKechnie and Swanson, 2010) and sometimes approach values for flight metabolism (Chappell et. al. 2011). In a tropical damselfish, aerobic scope is a predictor of dominance rank (Killen et al., 2014). However, in our Gouldian finches, exercise-induced MMR as well as factorial and absolute aerobic scopes were indistinguishable between the two colour morphs, despite these measurements being made after birds had interacted with opposite colour morphs for 10–14 days.

Based on the range of metabolic measures made under the same conditions for all birds, we conclude that aggressive versus submissive personality differences between the two Gouldian finch colour morphs (sensuRoche et al., 2016) are not aligned with differences in aerobic metabolic rate. We believe that our findings are pertinent to natural conditions in that the first antagonistic interactions free-living Gouldian finches have with unrelated adult birds occur after juveniles moult from their drab monomorphic plumage to their colour morph. The lack of correspondence between aggressive traits and elevated metabolic rate in these birds contrasts with the findings for many other species (Biro and Stamps, 2010; Mathot et al., 2019), and may be a consequence of the long-term sympatric co-existence between these genetically and chromatically distinct morphs.

Rohwer and Ewald (1981) describe ways that individuals of species with morphological dominance markers can derive mutual benefit from such ‘badging’. In the shepherd's hypothesis, subordinates are superior food finders (‘sheep’) and dominants (‘shepherds’) exploit this trait, but also protect subordinates from being overwhelmed by too many dominants. The behavioural morphs of Gouldian finches broadly fit this model, with black-headed birds being more exploratory (see below) and red-headed birds being more aggressive towards other red-headed morphs than to black-headed morphs (Pryke, 2007; Williams et al., 2012; King et al., 2015). In this context, it is noteworthy that free-living populations of Gouldian finches always contain a mix of both morphs, with red-headed birds outnumbering black-headed birds by approximately 2:1 (Franklin and Dostine, 2000; Gilby et al., 2009). In dyadic contests for food, black-headed birds typically yield passively to red-headed birds by displaying submissive or avoidance behaviours, whereas contests between red-headed birds invariably involve aggressive physical confrontations (Pryke and Griffith, 2006; Pryke, 2007).

Although submissive to red-headed morphs, black-headed Gouldian finches are much less neophobic and risk-averse than red-headed birds (Mettke-Hofmann, 2012; Williams et al., 2012) and act more like leaders in unfamiliar settings in both captive and free-living situations (King et al., 2015; O'Reilly et al., 2019). This contrasts with many other species, where risk taking and dominance positively covary (Briffa et al., 2015). This may be a consequence of red colouration being more conspicuous to predators than black in open landscapes, thus favouring more tentative exploration by the more dominant morph (Mettke-Hofmann, 2012).

Thus it is possible that the behavioural differences between these two morphs are mutually beneficial and represent evolutionary stable strategies favouring their continued co-existence (Maynard Smith, 1974; Kokko et al., 2014). The absence of metabolic rate differences between the two morphs may reflect this long-term stability, but our study did not evaluate the potential for metabolic rates to differ when resources become limited. In such circumstances, the frequency and intensity of aggressive activities will likely increase in red-headed birds, which could result in them having higher daily energy requirements than their black-headed counterparts. This could lead to divergence in flight and digestive organ sizes between the two morphs and, consequently, differences in maintenance energy metabolism (Daan et al., 1990; Chappell et al., 1999; Bai et al., 2016).

We thank Sarah Pryke for facilitating this study and Bethany Hoye and Simeon Lisovski for discussions on the statistical analyses. The constructive feedback from Neil Metcalfe and an anonymous reviewer on an earlier version of the manuscript is gratefully acknowledged.