When aggressiveness could be too risky: linking personality traits and predator response in superb fairy-wrens

Introduction

In animals, individuals with distinct personality traits can engage in different risk-taking strategies (Jones & Godin, 2010; Quinn & Cresswell, 2005; Van Oers et al., 2004). Personality, also known as temperament or coping strategy, refers to between-individual differences in behavioural traits that are consistent over time and across contexts (Bókony et al., 2012; Hall et al., 2015; Réale et al., 2007). Personality traits have been demonstrated in a number of taxa (Bell, Hankison & Laskowski, 2009), may be partly heritable (Dingemanse et al., 2002), and can remain consistent for a significant portion of an individual’s lifespan (Hall et al., 2015; Wuerz & Krüger, 2015). Personality traits can be measured along five broad axes, which may correlate with each other to form a behavioural syndrome along the ‘proactive-reactive’ axis (Sih, Bell & Johnson, 2004): (a) the shyness-boldness axis (how an individual reacts during a risky situation, such as encounters with predators and humans); (b) the exploration-avoidance axis (how an individual reacts to a new situation, such as a novel environment, food, or object); (c) the low-high activity axis (the general level of activity of an individual in a non-risky and non-novel environment); (d) the low-high aggressiveness axis (an individual’s agonistic reaction towards conspecifics); and (e) the low-high sociability axis (how an individual reacts to the presence or absence of conspecifics when excluding aggressive behaviour, Réale et al., 2007; Table 1; Fig. 1). Most studies examining animal personality traits in different contexts have focused on the shyness-boldness and exploration-avoidance axes (e.g., Amy et al., 2010; Jacobs et al., 2014; Snijders et al., 2015; van Asten, Hall & Mulder, 2016) or how one personality trait may relate to other behaviours or personality traits (Duckworth, 2006; Hall et al., 2017; Hollander et al., 2008). Although individual differences in personality traits have been shown to correlate with survival and fitness (Dingemanse et al., 2004; Dingemanse & Réale, 2005; Smith & Blumstein, 2008), studies explicitly linking animal personality measured in captivity with behaviours in the wild are sparse, and much less is known about how personality traits in one context relate to the same traits in other contexts (but see Cain et al., 2011; Huntingford, 1976; Witsenburg et al., 2010).

In risky situations, such as encounters with predators, different personality types may have distinct adaptive benefits (Aplin et al., 2014; Coleman & Wilson, 1998; Hedrick & Riechert, 1993; Sih, Kats & Maurer, 2003; Verbeek, Drent & Wiepkema, 1994). Fast-exploring individuals often at the proactive end of the ‘proactive-reactive’ axis, where a ‘proactive’ bird is not just fast-exploring, but also bold, aggressive and active, (Sih, Bell & Johnson, 2004; Table 1; Fig. 1) tend to engage in riskier behaviours, such as investigating novel objects/environments (Huntingford, 1976; Kluen et al., 2012; Koolhaas et al., 2001; Verbeek, Drent & Wiepkema, 1994), acting conspicuously in the presence of a predator (Cole & Quinn, 2014; Greig, Spendel & Brandley, 2010; Langmore & Mulder, 1992; Sih, Kats & Maurer, 2003), and returning to typical behaviour more quickly after a threat has passed (Hedrick & Riechert, 1993; Verbeek, Drent & Wiepkema, 1994). In contrast, slow-exploring individuals (at the reactive end of the axis) tend to prioritise their own survival and are, therefore, more risk-averse (Coleman & Wilson, 1998; Hall et al., 2015; Quinn & Cresswell, 2005). Individuals’ ability to avoid potential predators may thus drive differential selection patterns amongst personality types (Jones & Godin, 2010; Quinn & Cresswell, 2005), especially if individuals have to trade off between predation risk and other important behaviours such as foraging (Jones & Godin, 2010; Sih, Bell & Johnson, 2004; Verbeek, Drent & Wiepkema, 1994). For instance, mole salamanders (Ambystoma barbouri) that were highly active foragers were also active in the presence of a perceived predator, making them more vulnerable to predation (Sih, Kats & Maurer, 2003). For this reason, risk-taking behaviours that enhance activity may be maladaptive when predation risk is high.

An individual’s sex, regardless of its personality, may also affect how it responds to potential predators (McQueen et al., 2017; Samia et al., 2015; Tanis, Bott & Gaston, 2018; Zelano, Tarvin & Pruett-Jones, 2001). Many bird species are sexually dimorphic, with males displaying more conspicuous plumage than females, which can lead to sex-specific differences in both predation risk and anti-predator behaviours (McQueen et al., 2017; Powolny et al., 2014; Ruiz-Rodríguez et al., 2013; Stuart-Fox et al., 2003, but see Cain et al., 2011). However, the relationship between conspicuousness or sex and anti-predator behaviours is not straightforward, simply because individuals also differ in other factors that influence risk-taking behaviours. For example, when early parental investment is higher for female birds (through the costs of egg laying and incubation), mothers may tolerate higher levels of risk when defending their investment and nest (Trivers, 1972). During the breeding season, females may also engage in riskier behaviours because they have higher caloric needs than males (Powolny et al., 2014). The interaction between personality, conspicuousness and sex is relatively unexplored, but, in general, conspicuous individuals tend to be more vigilant and risk-averse in the presence of predators than their less colourful conspecifics (Hart & Freed, 2005; Ibáñez, López & Martín, 2014; Journey et al., 2013; McQueen et al., 2017; Ortega, López & Martín, 2014).

In this study, we used wild superb fairy-wrens (Malurus cyaneus), a common Australian songbird with a cooperative breeding system, to assess whether sex and three personality traits (boldness, exploration, and aggressiveness), measured during short-term captivity, correlated with boldness in the wild (Table 1; Fig. 1). We assessed boldness in wild fairy-wrens by using vocal playback to simulate the presence of a local predator of nestlings and adults, the grey currawong (Strepera versicolor) (Colombelli-Négrel, Robertson & Kleindorfer, 2009a; Colombelli-Négrel, Robertson & Kleindorfer, 2010a; Greig, Spendel & Brandley, 2010). A previous study found that fast-exploring superb fairy-wrens (tested in a novel environment test) were less likely to be present in the wild one year after initial sampling (Hall et al., 2015). Because males and breeding females are highly philopatric in this species, the authors suggested that these disappearances were likely explained by increased mortality (Hall et al., 2015; see also Cockburn et al., 2008), due to fast-exploring individuals being less cautious and taking more risks (i.e., approaching faster and closer or resuming foraging sooner once the threat has supposedly passed) when faced with potential threats (Dingemanse & Réale, 2005; Hall et al., 2015; Jones & Godin, 2010). Here, we tested this hypothesis by broadcasting predator vocalisations close to the nests of fairy-wrens with known personalities and assessing each group member’s behavioural response. Based on Hall et al. (2015), we predicted that fast-exploring individuals would be more likely to show risky behaviours in the presence of a predator, thereby increasing their mortality risk. These behaviours may include responding more slowly to the predator playback or approaching closer to the speaker and staying longer in its proximity (see also Jones & Godin, 2010; Quinn & Cresswell, 2005) than slow-exploring individuals. We also predicted that males would be more risk-averse in response to the predator playback compared to females, due to the males’ more conspicuous colouration (Hart & Freed, 2005; Ibáñez, López & Martín, 2014; Journey et al., 2013; McQueen et al., 2017; Ortega, López & Martín, 2014).

Methods

Personality assays

To assess individual personality traits, adult superb fairy-wrens were captured using mist-nests and taken to a nearby banding station. We first assessed their boldness phenotype (personality trait 1) by measuring the birds’ response to human handling, following procedures modified from Brommer & Kluen (2012). Individual responses to human handling have been identified as repeatable behavioural traits in other species, including in adult and nestling birds (Hall et al., 2015; Kluen, Siitari & Brommer, 2014). During handling, we recorded two behavioural variables. First, we tilted the bird onto its back in the bander’s grip and counted the number of individual struggles over a 30 s period (‘back-test response’, also called ‘docility’) (Brommer & Kluen, 2012; Hall et al., 2015; Hessing et al., 1993) (Fig. 2A). Second, we noted whether the subject struggled while recording five morphological measurements, and then assigned it a discrete ordinal score from 0 (did not struggle during any measurement procedure) to 5 (struggled during all five measurement procedures). The morphological measurements were: (1) tarsus length (to the nearest 0.1 mm, using a sliding calliper), (2) head-bill length (nearest 0.1 mm, sliding calliper), (3) tail length (nearest one mm, ruler), (4) wing length (nearest one mm, butt-ended ruler), and (5) mass (nearest 0.01 g, electronic scale). We named this behavioural variable ‘processing response’ (also called ‘handling aggression’; see Brommer & Kluen, 2012). If not already banded, the bird was fitted with a numbered aluminium leg-band issued by the Australian Bird and Bat Banding Scheme and a unique combination of three coloured bands to allow visual identification in the field.

All captured adults were then transferred in a cotton bag to an onsite building, where we quantified their exploration phenotype (personality trait 2) during a novel environment test (Fig. 2B) and their aggressiveness phenotype (personality trait 3) using a mirror stimulation test. Each subject was placed alone in a small wooden release box (170 × 115 × 90 mm in size) and allowed 5 min to acclimate. The door to the release box was then opened, allowing the bird to enter the novel environment: a metal flight cage (700 × 450 × 450 mm) with three wooden perches (Fig. 2B). The cage was covered with an opaque fabric on all but one side, visually isolating the birds from its surroundings. On the remaining side, we had already placed a camera (GoPro Hero7 Black, GoPro, Inc., San Mateo, CA, USA) to observe the subject’s movements. The novel environment was divided into 13 distinct sectors that the subject could visit: three perches, four floor quadrats, four cage walls, the cage ceiling, and the release box. For each subject, we recorded (1) ‘activity’ (the total number of sectors that the bird visited in the 5 min following emergence, including repeat visits) and (2) ‘sector visitation’ (the number of unique sectors visited over the same period, also referred to as ‘exploration’; see Hall et al., 2015). After 5 min, we remotely raised a curtain at one end of the cage to reveal a mirror (300 × 400 mm), which commenced the mirror stimulation test (Fig. 2B). Over the next 3 min, we then recorded two measures of aggressiveness: (1) ‘mirror attacks’ (the number of times the bird contacted the mirror) and (2) ‘time close to mirror’ [the total time (in seconds) the bird spent close to the mirror (i.e., in the three nearest sectors excluding time spent attacking the mirror)]. Video recordings of these assays were scored using the software Solomon Coder v. beta 19.08.02. To avoid any observer bias in the testing procedure, the novel environment and mirror stimulation tests were conducted and scored by a single experimenter (ACK). Some birds were tested multiple times; however, for this study, we only analysed performance in their first trial, when the novel environment and mirror presentation were still truly novel. In our study population, responses to the novel environment were significantly repeatable (activity: adjusted R = 0.303, P = 0.005; sector visitation: adjusted R = 0.311, P = 0.005, comparable to the adjusted repeatability of 0.37 previously found in superb fairy-wrens by Hall et al. (2015)) and the repeatability of aggressiveness (likelihood of attacking the mirror) during the mirror-stimulation test was marginally non-significant (adjusted R = 0.152, P = 0.062), based on N = 226 trials by 150 birds (Katsis, Colombelli-Négrel & Kleindorfer, pers. obs. 2019–2021).

Results

We analysed the predator playback responses of 40 birds (29 males and 11 females) with known personality across 15 territories. During the pre-playback period, there was no difference in the number of terrestrial alarm calls produced between the control and predator playback trials (Mann–Whitney: U = 780.50, N = 40, P = 0.57). As we predicted, individuals responded faster (U = 452.50, P < 0.0001) and were more likely to produce terrestrial alarm calls (U = 460.00, P <  0.0001) during the predator playback compared to the control playback. They also spent more time within 5 m of the speaker (U = 628.50, P = 0.04) in response to the predator playback, compared to the control. The minimum distance of approach did not differ between the two playback treatments (U = 607.00, P = 0.06). Among our personality traits measured in captivity, there was no significant correlation between boldness (PC_Handling) and exploration (PC_Exploration) (r² = 0.02, P = 0.88), nor between boldness (PC_Handling) and aggressiveness (‘time close to mirror’: r² = 0.06, P = 0.71; ‘mirror attacks’: U = 149, P = 0.19). There was a correlation between aggressiveness and exploration (PC_Exploration), with birds spending more time close to the mirror being more exploratory (‘time close to mirror’: r² = 0.41, P = 0.01; Fig. 3A) and those that attacked the mirror tending to be more exploratory (U = 132, P = 0.07; Fig. 3B). Our two measures of aggressiveness were highly correlated, with birds that attacked the mirror many times also spending more time close to the mirror (U = 40.50, P < 0.0001).

When considering only birds’ response to the predator playback, individuals that attacked the mirror during the mirror stimulation test showed a stronger response to the playback: they approached faster (mean approach time: 88 s versus 215 s), closer (mean minimum distance: 6.4 m versus 10.1 m), spent more time within 5 m of the speaker (mean time spent: 20 s versus 16 s), and were more likely to alarm call than those that did not attack the mirror (53% versus 50%) (Table 3; Fig. 4A). Birds that spent an intermediate amount of time near the mirror during the mirror stimulation test also responded more strongly to playback—they approached faster, closer, spent more time within 5 m of the speaker, and were more likely to alarm call—compared to birds at both extremes (Table 3; Fig. 4B). Boldness (PC_Handling), exploration (PC_Exploration), sex, playback order and the number of responders did not predict PC_Playback (Table 3). Territory ID did not significantly predict PC_Playback (Wald Z = 1.37; P = 0.17).

Discussion

Previous research has shown that individuals with distinct personality traits differ in their risk-taking strategies (Jones & Godin, 2010; Quinn & Cresswell, 2005; Van Oers et al., 2004). Here, too, we found that an individual’s aggressiveness, measured as its response towards an apparent conspecific during short-term captivity, was associated with boldness and risk-taking in the wild. Specifically, superb fairy-wrens that attacked their mirror image and spent an intermediate period near the mirror during the mirror stimulation test responded most vigorously to predator playback. Conversely, exploration during a novel environment test and boldness (measured as response to human handling) did not predict an individual’s response to predator playback. These results suggest that specific personality traits, rather than an overall behavioural syndrome (where personality traits correlate with each other; Sih, Bell & Johnson, 2004), predict behavioural responses to a perceived predator in superb fairy-wrens.

Superb fairy-wrens that attacked their mirror image during the mirror stimulation test, and those that spent an intermediate duration near the mirror, responded most vigorously to the predator playback, approaching the speaker faster and closer, spending more time near the speaker, and being more likely to alarm call. This suggests that conspecific aggressiveness is associated with predator-response behaviours in this species, albeit in a non-linear fashion. Responding to a predator quickly and closely could reduce the risk of nest predation, by alerting other group members to the threat and facilitating the performance of distraction displays to draw the predators’ attention (Kleindorfer, Fessl & Hoi, 2005; LaBarge et al., 2021; Rowley, 1962; Zuberbühler, 2001). However, such a strong response would presumably also place the responders at greater risk of being predated themselves, making the net fitness benefits unclear. Aggressive phenotypes may be maintained in populations not because their predator responses are beneficial but because aggressiveness is associated with positive outcomes in other contexts: for example, during contests with conspecifics (Godin & Dugatkin, 1996; Hedrick & Riechert, 1993; Sih, Bell & Johnson, 2004) or during foraging, perhaps by helping individuals compete for shared food resources (Aplin et al., 2014; Coleman & Wilson, 1998; Hedrick & Riechert, 1993; Sih, Kats & Maurer, 2003). In the case of superb fairy-wrens, neighbouring groups probably compete with each other for the high-resource territories, and higher aggressiveness may be advantageous for these disputes (Cain & Langmore, 2016). The fitness benefits of aggressiveness are still to be fully investigated in superb fairy-wrens, although Hall et al. (2015) reported that fast-exploring birds (which also tended to be more aggressive) were less likely to be present in the wild one year after initial sampling, which suggests lower survivorship (although this remains to be fully tested). If aggressiveness is beneficial for competition with conspecifics, it may give individuals an adaptive benefit where predation risk is low (Hedrick & Riechert, 1993; Sih, Kats & Maurer, 2003). Conversely, where the risk of predation is high, more aggressive phenotypes that respond riskily to potential predators may be selected against. Unexpectedly, birds that spent the longest duration close to the mirror did not exhibit the strongest responses to predator playback in the wild. This could indicate that, beyond a certain threshold, very high scores in the mirror stimulation test no longer quantify a bird’s aggressiveness but perhaps also its intelligence (the ability to eventually recognise its mirror-image) or its perseveration (the tendency to perform recurring or redundant responses without appropriate stimulus; Vickery & Mason, 2005).

There was no significant relationship between the number of responders to the predator playback and the strength of individuals’ response. This contrasts with previous studies in fairy-wren species, where social context influenced individuals’ responses to a simulated predator or other potential threat (Teunissen et al., 2021; van Asten, Hall & Mulder, 2016). van Asten, Hall & Mulder (2016) found that fast-exploring superb fairy-wrens responded more strongly to a novel object than did slow-explorers, but only when multiple group members were present. It is possible that our study achieved different results due to the type of stimulus used to simulate a predator. As predatory birds will typically hunt in silence, it is possible that the perceived threat from our vocal playback trials was lower compared to what a physical model might produce (Abbey-Lee et al., 2016), which could in turn have influenced the relationship between the number of responders and playback response.

Fairy-wrens responded more strongly to the predator (grey currawong) playback than to the control (willie wagtail) playback. This indicates, consistent with other studies, that subjects recognised the playback track as a risk to the nest or to themselves without any accompanying visual stimuli (Abbey-Lee et al., 2016; Kleindorfer et al., 2013a; Kleindorfer, Fessl & Hoi, 2005; McQueen et al., 2017; Williamson & Fagan, 2017; Zuberbühler, 2001). Many previous studies assessing avian predator response used physical models to simulate a predator (Colombelli-Négrel, Robertson & Kleindorfer, 2010a; Krams et al., 2014; Magrath, Pitcher & Gardner, 2009; Mutzel et al., 2013). As described previously, we cannot discount that a visual predator stimulus might have evoked a different, and perhaps stronger, defensive response from our subjects, and potentially displayed a clearer relationship between proactive personalities and playback response. Because most avian predators do not call while hunting, our predator playback may have been interpreted to mean that a predator was present in the area but did not necessarily pose an immediate threat (see also Abbey-Lee et al., 2016). Even so, the use of audio playback does have potential benefits. Model presentations, while effective for eliciting alarm responses and defensive behaviours, also require the bird to have a clear line of sight to the model, which is often impractical in the dense vegetation in which many fairy-wrens build their nests (Magrath, Pitcher & Gardner, 2007; Magrath, Pitcher & Gardner, 2009).

Contrary to our predictions, sex did not predict an individual’s playback response, despite most males being in their nuptial plumage at the time of testing. Highly conspicuous individuals tend to be risk-averse in the presence of predators, choosing areas of higher cover when calling for danger (Katz et al., 2015; Ximenes & Gawryszewski, 2020). Our results may be explained by the dense vegetation of our study area, where even highly conspicuous males were able to respond to the predator playback without significantly increasing their risk compared to their counterparts. A future investigation may account for this by considering the vegetation cover of the area surrounding the simulated predator. We also cannot exclude that males perceived the lower risk posed by the playback (compared to a visually present predator) and used the opportunity to display their potential as mates, as found by Langmore & Mulder (1992) in superb fairy-wrens and by Greig & Pruett-Jones (2008) in splendid fairy-wrens (Malurus splendens). Future experiments with model presentations may be able to demonstrate this.

Contrary again to our predictions, an individual’s response to predator playback did not correlate with its exploration behaviour in a novel environment (measured in captivity) nor its response to human handling (boldness measured in captivity). This contrasts with previous work in great tits (Parus major) showing associations between a bird’s exploration behaviour in captivity and its response to risky situations in the wild. Specifically, more exploratory great tits were more responsive to playback simulating a conspecific intruder (Amy et al., 2010), were less neophobic towards a novel object at their nest (Cole & Quinn, 2014), and called more in response to a nest intruder (Hollander et al., 2008). In superb fairy-wrens, it may be the case that exploration in a controlled captive environment simply does not correlate with behaviour in the wild, as has been reported in zebra finches (Taeniopygia guttata) (McCowan et al., 2015). However, the negative relationship between fairy-wren exploration behaviour and apparent survival (see Hall et al., 2015) implies that personality traits measured in captivity correlate at least with some aspect of their wild behaviour. Clearly, further work is needed to determine how fast- and slow-exploring individuals differ in their behaviour in the field and the fitness consequences associated with each behavioural phenotype. Similarly, an individual’s boldness in captivity (struggle response to human handling) did not predict its behaviour in the field. Our handling assays were intended to simulate the experience of being seized by a predator and having to escape. Because of this, the fitness benefits associated with a strong or weak handling response may only be apparent during close physical altercations with a predator, and not necessarily during the earlier stages of a predator encounter. In a study of blue tits (Cyanistes caeruleus), Fresneau, Kluen & Brommer (2014) found a sex-specific relationship between handling response and nest defence, with less responsive females defending their nestlings more vigorously. Notably, nest defence was measured while experimenters removed a nestling from the nest, simulating a nest predator more directly and intensely than the vocal playback used in our study. Therefore, the potential fitness implications of a fairy-wren’s response to handling still need to be tested in scenarios in which predators are perceived to pose a direct and immediate threat.

Conclusion

This study adds to the expanding literature of how individual personality may correlate with behavioural responses to predators (Dingemanse & Réale, 2005; Hollander et al., 2008; Krams et al., 2014). Of the three personality traits we measured in captivity, only one (aggressiveness, measured in a mirror stimulation test) significantly predicted that bird’s behavioural response to a simulated predator in the wild. Our study suggests that overarching behavioural syndromes may not be as static as previously suspected (Hall et al., 2015; Wuerz & Krüger, 2015), with specific personality traits predicting situational behavioural responses. With only one of three traits predicting wild behaviours, it is likely that researchers’ choice of personality trait helps determine whether they find significant relationships between personality and other wild behaviours. Hence, in future investigations of this kind, we suggest it may be useful, or even essential, to measure multiple personality traits. Future studies should investigate this relationship between multiple populations and examine how risk-taking behaviours correlate with actual mortality and breeding/foraging success.

Supplemental Information