The quick and the dead: Behavioral plasticity of anti‐predator responses in an Endangered mammal

Abstract

Populations isolated from predation inside predator‐free havens often exhibit a reduction in anti‐predator traits. The loss of such traits has a critical bearing on strategic conservation management, and so it is important to understand the basis of trait shift and how anti‐predator traits may be retained or restored. We explored plasticity in anti‐predator behaviors in an Endangered mammal, the woylie (Bettongia penicillata ogilbyi) at both the individual and population level. We quantified anti‐predator responses in individual woylies sourced from either a predator‐free haven (havened) or from an indigenous wild population (non‐havened) before and after translocation to a site with low densities of introduced predators, providing the first experimental test of anti‐predator responses and corresponding survival consequences in this species. Initially, the havened woylies had weak anti‐predator responses (lower agitation) compared with non‐havened animals. After exposure to predators, apparent survival was lower in the havened cohort compared to the non‐havened cohort. Those havened individuals that did survive, however, had stronger anti‐predator responses at the end of the study, approaching the level of response shown by their non‐havened counterparts. This within‐individual shift in behavior provides evidence for behavioral plasticity in this particular trait, suggesting that some aspects of anti‐predator behavior may be regained following exposure to predators and that this behavioral flexibility can be advantageous. At the same time, evidence of lower survival in previously havened individuals indicates that fixed differences in other traits remain and likely also contribute to survival. We discuss the implications of these findings for conservation management.

INTRODUCTION

In many ecological systems, animal populations are becoming isolated from predators. This can occur directly, for example, when species inhabit predator-free havens (Legge et al., 2018) or are raised in captivity (Williams & Hoffman, 2009), as well as indirectly, when predators are removed from a land- or seascape (Berger et al., 2001; Bond et al., 2019). Without predation pressure, relaxed selection can result in populations that are naïve to predators that they once feared, with weakening of behavioral and morphological anti-predator traits (Bannister et al., 2018; Blumstein, 2002; Blumstein, 2006; Blumstein et al., 2004; Blumstein et al., 2006; Blumstein & Daniel, 2005; Harrison, Phillips, et al., 2023; Jolly et al., 2021; Jolly & Phillips, 2021; Moseby et al., 2018; Muralidhar et al., 2019). While we know these traits can be rapidly lost (Bannister et al., 2018; Blumstein et al., 2004; Harrison, Phillips, et al., 2023; Jolly et al., 2018; Moseby et al., 2018), we have a poor understanding of how quickly they may be regained (but see West et al., 2018 and Ross et al., 2019). As the loss of anti-predator responses has a critical bearing on strategic conservation management (Blumstein, 2000), it is important to understand the basis of trait shift, and how anti-predator traits may be retained or restored (Harrison, Wayne, et al., 2023). Knowing these attributes would greatly improve our ability to manipulate the management of populations and individuals to improve conservation outcomes (Donelson et al., 2023).

Shifts in anti-predator response may represent phenotypic plasticity or an evolved, genetic shift. For example, common garden experiments have revealed that offspring of predator-naïve northern quolls (Dasyurus hallucatus) continue to be naïve, and offspring of predator-exposed quolls demonstrate predator aversion, even when raised in captivity (Jolly et al., 2018), suggesting a genetic basis to predator recognition. If trait loss represents an evolved, genetically based shift, however, it may be difficult to reinstate anti-predator traits once they have been lost. But some anti-predator responses are experience-dependent and the product of behavioral plasticity (Blumstein, 2002). Generally, if trait loss represents a plastic shift in response to the changed predatory environment, then anti-predator traits will likely return in the presence of predators (Blumstein, 2002).

Critical weight range mammals (35–5500 g) are a major focus of this field in the Australian context as they are especially susceptible to predation from invasive eutherian predators (Burbidge and McKenzie, 1989), and as a result, are often conserved within predator-free havens (Legge et al., 2018). Exposing predator-naïve havened populations to low and controlled levels of predation pressure has been proposed as a strategy to promote desired anti-predator responses in Australian mammals (Moseby et al. 2015). This approach aims to provide prey species with predator experience to facilitate learning, as well as imposing direct selection pressures through predation of individuals with weak anti-predator traits, thus working to strengthen anti-predator traits at the population level. Rapid shifts in response to controlled predator exposure have been observed in numerous anti-predator traits, including increased body size (Blumstein et al. 2019) and escape behavior (Tay et al. 2021) in burrowing bettongs (Bettongia lesueur), improved vigilance behavior in greater bilbies (Macrotis lagotis; Ross et al., 2019) and shark bay bandicoots (Perameles bougainville; Waaleboer et al. 2024), and increased wariness in spinifex hopping mice (Notomys alexis; Stepkovitch et al. 2024). Yet, despite these widespread observations, the mechanisms driving such changes following predator exposure remain poorly understood.

In this study, we experimentally test for adaptive plastic anti-predator behaviors using woylies (Bettongia penicillata ogilbyi) as a model system. Woylies are macropod marsupials, also known as brush-tailed bettongs, and are listed as Endangered under Australia's Environment Protection and Biodiversity Conservation Act (1999) due to widespread population declines. We reintroduced woylies to an area of their former range where invasive predators (feral cats Felis catus and foxes Vulpes vulpes) occur at low densities through ongoing management. Woylies were sourced from two locations: (1) from a predator-free haven, and (2) where they currently persist in the presence of invasive predators that are being actively managed. The havened woylies were sampled from a population isolated from predators for over 10 years (up to 20 generations; Harrison, Wayne, et al., 2023), whose anti-predator traits have weakened over time whereby successive individuals have become smaller (reduction in weight of ~250 g and in pes length of ~3 mm after 10 years) and less reactive (Harrison, Phillips, et al., 2023). By quantifying proxy anti-predator responses (measures of agitation) in individuals from each of these populations before and after their reintroduction, we reveal the survival consequences of weakened anti-predator responses and examine the ability of havened woylies to adapt their anti-predator behaviors once faced with predators. We expect that havened woylies will have weakened anti-predator responses (as a consequence of both evolved and plastic shifts) compared to non-havened woylies, and that they will have reduced survival once exposed to predators. If agitation scores are plastic and respond to the presence/absence of predators, then we would expect these scores to rapidly shift within individual woylies upon moving from a predator-free to a predator-exposed environment.

METHODS

Study system and site

Our study took place in collaboration with the Marna Banggara program (), aimed at reintroducing locally extinct indigenous species to areas of their former range in Dhilba Guuranda-Innes National Park (DGI) on Yorke Peninsula in South Australia (−35.269, 136.865). While the park is heavily predator-controlled, foxes and feral cats persist in the landscape at low densities (mean activity of 1.9% and 2.7% for foxes and cats, respectively, measured in detections per track-facing camera per night between June 2022 and November 2023). While it is unlikely that every woylie individual would have experienced a direct encounter with a predator in DGI National Park, we assume that all woylies encountered signs of predators (e.g., feces or scent marks) that could trigger a plastic behavioral response (a meta-analysis of anti-predator responses found that Australian mammals consistently responded to visual and olfactory cues of predators; Banks et al., 2018).

In 2021 and 2022, 84 woylies (yalgiri in the local Narrungga language) were reintroduced to DGI National Park from a havened population on Wedge Island in South Australia (hereafter SA woylies) (Frick et al., 2023). An additional 36 woylies (sourced from a haven n = 16, F:10, M:6; or outside a haven n = 20, F:7, M:13) were translocated to DGI National Park in June 2022 from Perup Sanctuary and Tone-Perup Nature Reserve in Western Australia (hereafter WA woylies) (Figure 1a). Perup Sanctuary was initially founded in 2010 from the surrounding indigenous woylie populations (including Tone-Perup Nature Reserve; Harrison, Thorn, et al., 2024), so individuals from both WA populations shared, until very recently, the same evolutionary history. When WA woylies were captured for translocation to DGI National Park, we recorded their sex, weight, and quantified their anti-predator behavior using an assay that measures agitation in the presence of a large unfamiliar ‘predator’, the human handler (described by Harrison et al., 2022). This assay quantifies agitation at five points: when the handler (1) approaches the trap, (2) places the capture bag over the trap, (3) opens the trap door, (4) has the animal in the capture bag before handling, and (5) handles the animal for processing. Each of these scores was summed to give a cumulative agitation score (as per Harrison, Phillips, et al., 2024).

[IMAGE OMITTED. SEE PDF]

The WA woylies were transported by road and aircraft to the release site (located approximately 2500 km away) within 24 h. Twenty-two woylies sourced from WA were fitted with VHF transmitters (radio collars initially that were removed and replaced with tail-mounted transmitters at trappings 2–4 weeks post release) programmed with a mortality signal and tracked continuously using autonomous towers (Frick et al., 2023) to monitor survival for up to 92 days (3 months post-release). The area monitored by the towers had a radius of approximately 6 km, and there were no recorded instances where an individual permanently dispersed outside of this range (Frick, unpublished data). Between 3 and 7 months post-release, cage-trapping was opportunistically carried out across the release sites to remove tail transmitters and to perform welfare checks (approximately 100 trap nights).

After 7 months, an extensive trapping effort (approximately 900 trap nights) took place over multiple nights. The sampling array for this trapping was designed to encompass the area monitored by the autonomous towers (within which all tagged animals were known to reside at 3 months post release), covering all known locations of individuals recorded from the radiotracking, plus an ~800 m buffer, to maximize the chances of capturing all surviving animals. Using the date that an individual was last seen based on VHF tracking and cage trapping, we calculated the maximum number of days that each individual was known to be alive for. During the 7-month trapping, we recorded each individual's identity (every individual was microchipped) and reassessed anti-predator behavior of the WA woylies using the same behavioral assay. We also assessed the anti-predator behavior of the SA woylies (across multiple captures) to increase our sample size when calculating the within-individual repeatability of agitation behavior. The behavioral component of this experiment, aimed at investigating the plasticity of anti-predator behaviors in WA woylies, follows a BACI (before-after, control-intervention) design, where the non-havened cohort, already familiar with predators, acts as the control treatment (accounting for additional environmental variation). This allowed us to make population-level comparisons between havened and non-havened cohorts, and to examine within-individual changes in behavior following translocation and likely exposure to predators (Table 1).

TABLE 1 Research questions explored in this study, along with the specific data used (WA and/or SA populations) and level (population or individual) of analysis.

	Question	Populations included in analysis	Level	Location in results
Survival	Is there a difference in survival probability between havened and non-havened woylies following translocation?	Havened (WA) and non-havened (WA) woylies	Population	Figure 1b, section (a) text
Is there a difference in the number of days known to be alive between havened and non-havened woylies following translocation?	Havened (WA) and non-havened (WA) woylies	Population	Table 4
Behavior	Is agitation behavior repeatable within individual woylies/yalgiri across multiple captures?	Havened (WA) and non-havened (WA) woylies, havened yalgiri (SA)	Individual	Section (b) text
Is there a difference in agitation behavior between havened and non-havened woylies before translocation?	Havened (WA) and non-havened (WA) woylies	Population	Figure 2a, Table 3a
Is there a difference in agitation behavior between havened and non-havened woylies after translocation?	Havened (WA) and non-havened (WA) woylies	Population	Figure 2a, Table 3a
Do individual woylies shift their agitation behavior following translocation?	Havened (WA) and non-havened (WA) woylies	Individual	Figure 2b, Table 3b

Statistical analysis

All statistical analyses were conducted in the R environment (R Core Team, 2020).

Survival

To compare survival between the two WA cohorts (while accounting for any differences in capture probability), we built a Cormack-Jolly-Seber (CJS) model using the package ‘openCR’ (Efford, 2021). This model is conditional on a first capture, which we treated as the release of individuals, during which all translocated woylies were sighted (session 0). We then calculated apparent survival (ϕ) and detection probability (p) for each cohort in the single trapping session 7 months after their release (session 1). This analysis relied on capture data only, and the model is hierarchical, modeling survival while accounting for capture probability. We first determined the best fitting parameters for the capture probability model by comparing AIC values of models where p could vary by session (this was coded into all models as session 0 has perfect detection), as well as cohort, sex, and the interaction between them, while ϕ was held constant. We then fixed the best performing model for p and determined the best full model by comparing AIC values from models where ϕ could vary by cohort, sex, and the interaction between them. We used likelihood ratio tests between the full model with and without each respective parameter to evaluate their significance. This analysis was based on recaptures: those that were not recaptured may have dispersed away from the site, died, or simply not have been detected. Imperfect detection is accounted for in the CJS model, so to make inferences on survival probabilities, we make the assumption that dispersal is equal between cohorts. To test this, we compared the distance between the respective release site and the furthest point where each individual was captured among cohorts and sexes using a one-way ANOVA.

Behavior

To explore whether agitation behavior was consistent within individuals, with repeat observations from the same individuals, we calculated within-individual repeatability of the agitation scores using a linear mixed model repeatability estimate fitted with restricted maximum likelihood, controlling for handler (random effect) using the package ‘rptR’ (Nakagawa & Schielzeth, 2010). Because we expected agitation behavior to change following translocation to a new environment, we split the data by whether the observation occurred before or after translocation. As most individuals were only captured a single time before translocation, we were unable to calculate the within-individual repeatability of agitation behaviors in WA woylies before translocation (although the assay has proved to be repeatable in this population previously; Harrison et al., 2022). To calculate within-individual repeatability of agitation behavior after translocation, we included observations of both WA woylies and SA yalgiri to increase the sample size.

To compare agitation scores between havened and non-havened WA woylie populations, we built a linear mixed effects model, testing for the effect of the interaction between time (before or after translocation) and population cohort (haven or non-havened), controlling for sex (fixed effect), as well as handler and individual identity (random effects) as this model contained repeat measures from multiple individuals. To compare the change in agitation scores within WA individuals only, we used linear regression to test for the effect of population type (haven or non-haven) on change in agitation scores within individuals (mean score after translocation minus score before translocation), controlling for sex. Respective model fits were tested using the ‘DHARMa’ package (Hartig, 2022). We used likelihood ratio tests between the full model with and without each respective parameter to calculate significance. Significant interaction effects were explored by conducting Tukey post-hoc pairwise comparisons using the package ‘emmeans’ (Lenth, 2024).

Effect of behavior on survival

To explore the influence of agitation behaviors on survival, we built a linear regression model testing for the effect of agitation scores before translocation, sex, and population cohort on the maximum number of days each individual was known to be alive (based on radio tracking and capture data). Respective model fits were tested using the ‘DHARMa’ package (Hartig, 2022). We used likelihood ratio tests between the full model with and without each respective parameter to calculate significance.

RESULTS

Survival

Individuals that carried radio collars or tail-tag transmitters were tracked for a maximum of 92 days, during which time no predator-related mortalities were detected (one havened female died by car strike and has been excluded from all analyses here). Seven months after the translocation of the WA woylies, 47% of havened individuals (n = 7; 3F, 4 M) and 63% of non-havened individuals (n = 12; 4F, 8 M) were recaptured and so known to be alive. There was no significant difference in the distance between the release site and the furthest capture point among cohorts (F(1, 14) = 0.22, p = 0.648) or sexes (F(1, 14) = 0.11, p = 0.749). The best CJS model allowed p to vary by session, cohort, and sex (Table S1a). While including sex as a predictor of detection probability improved the model, there was no significant difference in detection probability between males and females (p = 0.19). Detection probability was higher in individuals from the havened cohort (p = 0.003; Table 2). The top four models (each within ΔAIC <1; see Table S1b) included various combinations of cohort and sex on ϕ. Apparent survival estimates were lower in the cohort sourced from the haven compared to those sourced from outside the haven (p = 0.07) and lower in females compared to males (p = 0.09; Figure 1).

TABLE 2 Estimates of detection probability and apparent survival in translocated woylies from Western Australia.

Cohort	Sex	Apparent survival (ϕ)	Detection probability (p)
Havened	Male	0.55 (0.23–0.84)	0.95 (0.73–0.99)
Female	0.30 (0.10–0.61)	0.92 (0.55–0.99)
Non-havened	Male	0.80 (0.42–0.96)	0.58 (0.33–0.79)
Female	0.56 (0.21–0.87)	0.39 (0.16–0.68)

Behavior

Based on 144 captures of 83 individuals (1–4 captures per individual, mean = 2) in South Australia during the intensive trapping period conducted 7 months after releasing WA woylies, agitation scores were repeatable within individuals (R = 0.361, s.e. 0.108, p < 0.001).

Before translocation, at the population level, the havened WA woylies (36 observations from 33 individuals) had lower agitation scores than the non-havened WA woylie population (67 observations from 65 individuals) (t(7.97) = −3.94, p = 0.018; Figure 2 and Table 3a). After translocation, agitation scores in the previously havened population (18 observations from 6 individuals) increased (t(25.72) = −4.09, p = 0.002) to a similar level as in the non-havened population (24 observations from 10 individuals) (t(96.04) = 0.748, p = 0.877), while agitation levels in the non-havened population were unchanged (t(15.17) = −0.358, p = 0.984) (Figure 2 and Table 3a). Within-individual changes in behavior among WA individuals were higher in havened woylies (n = 5) compared to non-havened woylies (n = 10) (Table 3b).

[IMAGE OMITTED. SEE PDF]

TABLE 3 Effects of source population on agitation behavior before and after translocation in woylies from Western Australia at the population (a) and individual (b) levels.

Fixed effect	Population level behavior	Within-individual changes in behavior
Estimate	S.E	df	p-value	Estimate	S.E	df	p-value
Intercept	3.975	0.793	–	–	6.400	2.381	–	–
Sex_Male	0.470	0.797	1	0.521	−2.000	2.291	1	0.151
Population_Non-haven	2.301	1.099	1	0.786	−4.600	2.381	1	0.077
Translocation_After	3.945	0.933	1	0.011	–	–	–	–
Population_Non-haven: Translocation_After	−3.092	1.283	1	0.020	–	–	–	–

Effect of behavior on survival

There was no effect of agitation score on survival, and males were known to be alive longer than females (Table 4). No effect of cohort on the number of days known to be alive was detected.

TABLE 4 Effects of source population and behavior on the number of days known to be alive in translocated woylies from Western Australia.

Fixed effect	Estimate	S.E	df	p-value
Intercept	73.382	40.206	–	–
Cohort_Non-haven	65.222	49.120	1	0.200
Sex_Male	103.804	47.562	1	0.042
InitialAgitationScore	0.768	7.732	1	0.922

DISCUSSION

Our experiment allows a rare glimpse into behavioral plasticity in anti-predator traits in an Endangered marsupial. We found that woylies display clear behavioral plasticity in agitation behavior (a proxy for anti-predator behavior): predator-naïve woylies increased their agitation behavior after translocation to an environment with predators, indicative of improved anti-predator responses. While apparent survival was lower in the havened cohort, agitation responses before translocation did not predict survival, but those havened individuals known to survive increased their agitation to the same level as their predator-exposed counterparts. This suggests that behavioral flexibility to adapt to changed environmental pressures is present and could be advantageous. These findings also demonstrate that predator-naïve individuals sourced from predator-free havens can have reduced survival compared to non-havened individuals when released into areas where predators persist.

Plastic behavioral shifts in response to predation pressure, such as that seen here in agitation behavior, could well be common, having likely evolved in response to historical spatiotemporal heterogeneity in predation pressure. While we might expect some plastic traits to persist for longer in havens if they come with low maintenance costs (Lahti et al., 2009; Murren et al., 2015), it is encouraging that this plasticity has been retained in a population isolated from predators for up to 20 generations (Harrison, Wayne, et al., 2023). But of course, our findings here do not rule out the possibility that genetic adaptations to the predator-free environment have also occurred. Notably, survival of havened woylies was estimated to be substantially lower (0.4 vs. 0.7 for non-havened animals) on release into an environment in which predators were present, despite exhibiting a rapid (<7 months) plastic shift in agitation scores. It is likely that other traits (such as size) are not as plastic as behavior. In this regard, it is noteworthy that even after 7 months, havened animals had demonstrably higher detection (i.e., rates of entering traps) than did non-havened animals. The less responsive nature of these other traits to shift leaves the havened cohort more susceptible to predation despite the clear plasticity they have displayed in agitation scores. Consequently, future studies should aim to delineate the relative effects of plastic versus fixed (non-plastic) traits in determining fitness in the presence of predators. Experimental testing, for example, measuring a larger suite of traits (plastic and non-plastic) and examining their correlation with individual survival in havened and non-havened populations, would shed light on this question.

Our findings are consistent with similar studies conducted at a population level. Burrowing bettongs (Bettongia lesueur; a marsupial closely related to woylies) displayed evidence of behavioral adaptation whereby docility scores increased in a population after 12 months exposure to feral cats (West et al., 2018). Where we found an increase in agitation behavior, the burrowing bettong study found a behavioral shift in the opposite direction, where docility increased. This likely reflects variation in optimum predator escape strategies between species. Prey species may either immediately escape from an approaching predator or attempt to remain undetected, and the opportunity cost of these strategies depends on the availability of energy resources (Broom & Ruxton, 2005) among other biological and ecological factors. We suggest that the increase in agitation observed here in woylies may represent an increased affinity to escape in the presence of predators, as woylies tend to inhabit resource-rich areas where the cost of fleeing may be small. Species such as burrowing bettongs may adopt the opposite strategy—increasing docility in response to predators to remain undetected (West et al., 2018)—as they inhabit arid and often resource-limited environments. Overall, our work here suggests that such population-level shifts in behavior reflect, at least in part, plastic shifts in individual behavior.

Our study joins a growing body of literature that has experimentally tested for fitness costs of havening (Bannister et al., 2021; Jolly et al., 2018; Ross et al., 2019) and found evidence to suggest it can have negative consequences. Similarly, a population of bilbies (Macrotis lagotis) exposed to predators had higher survival when translocated into an environment with predators compared to havened bilbies who were naïve (Ross et al., 2019). Our interpretation here, whereby estimated survival in havened woylies is almost half that of the non-havened woylies (0.4 vs. 0.7) rests on the assumption that dispersal is equal between cohorts (i.e., those that are not captured have dispersed away from the site or died: we are assuming equal dispersal, yielding unequal survival). Given that the trapping array that informed this model encompassed the entire area within which all collared individuals were known to persist at 3 months post release, we expect that few, if any, individuals would have dispersed outside of this area. Moreover, of the individuals captured at 7 months, there was no difference in the distance between release and capture sites among cohorts. While we did not detect an effect of cohort on the number of days an animal was known to be alive, the model we used could not take into account the substantial difference in detection probability between cohorts, which likely led to an underestimation of survival in the non-havened cohort. Experimental designs capable of following the specific fate of individuals (i.e., obtaining confirmed mortalities) are best suited to exploring questions of survivorship in future studies of this nature.

Similar behaviors to those measured here have been linked to survival in other small mammal species; for example, shyness and fearfulness have been associated with improved survival probability in swift foxes (Vulpes velox; Bremner-Harrison et al., 2004) and brushtail possums (Trichosurus vulpecula; May et al., 2016), respectively. Here, we were unable to detect a correlation between agitation behavior before translocation and post-release survival. While surviving woylies displayed an increased level of agitation, without assessing the behavior of woylies who died, we cannot suggest that agitation behavior is the main driver of survival. There is likely a suite of traits contributing to how an individual detects and evades predators that were not measured. For example, there is considerable size variation between these cohorts (Harrison, Phillips, et al., 2024), which facilitates predator escape (Tay et al., 2023), and size is an important predictor of survival in the face of invasive predators for other bettong species (Evans et al., 2021; Moseby et al., 2018, 2023). It is also possible that the degree of plasticity is under selection here, whereby individuals that could rapidly adjust their behavior survived, and those that could not perished. Further exploration of how traits associated with anti-predator responses correlate with post-release survival in woylies may help to improve their translocation success by facilitating the selection of individuals with higher survival probability.

Our findings are important for the conservation of Australia's threatened mammals, where several species survive only in havens (Legge et al., 2018), and where havening has led to some species recovering to the point of a potential conservation delisting (Woinarski et al., 2023). We provide evidence to suggest that havened individuals have reduced survival probability when released into the wild where invasive predators persist. If predator-naïve populations can quickly improve some aspects of their anti-predator responses following experience with predators, they may be better suited for release into areas with low predator densities, although likely with a transient low survival probability as behaviors adjust and with a fixed lower survival reflecting the non-plastic traits they may now carry. In reintroduction programs, the loss of even a small proportion of individuals can be detrimental to a newly establishing population. To improve survival outcomes, in-situ predator exposure before translocation could prevent the loss or facilitate the adaptation of appropriate behaviors, as has been seen in closely related species (Harrison, Phillips, et al., 2024; Moseby et al., 2023; Ross et al., 2019; West et al., 2018). Maintaining low predator densities at release sites when species are reintroduced will likely also be crucial to allow individuals to re-establish the appropriate behavioral responses. It is important to note that in this study we only assessed a single plastic behavioral trait, yet appropriate anti-predator responses are likely a compound of multiple fixed and plastic traits, some of which may have driven the reduced survival we found in havened woylies. The ability of populations to maintain a full suite of anti-predator responses in havens will ultimately depend on the lability of such traits and the strength of selective pressures operating in havens. These are factors about which we currently know very little.

ACKNOWLEDGMENTS

We acknowledge the traditional custodians of the lands and animals on which this work was conducted, the Noongar and Narrungga people. Thanks to the many staff and volunteers involved in the Marna Banggara project. All animal procedures were approved by the University of Adelaide Ethics Committee (34990), The Department for Environment and Water Wildlife Ethics Committee (18-2021) and Scientific Research Permit (M27091-3), and the University of Western Australia Animal Ethics Committee (2021_ET000428).

FUNDING INFORMATION

Funding for this study was provided by WWF Australia, the Hermon Slade Foundation, Holsworth Wildlife Research Endowment, Royal Society of Western Australia, Northern & Yorke Landscape Board, and the South Australian Department of Environment and Water.

References

Banks, P. B., Carthey, A. J. R., & Bytheway, J. P. (2018). Australian native mammals recognize and respond to alien predators: A meta‐analysis. Proceedings B, 285(1885), [eLocator: 20180857]. [DOI: https://dx.doi.org/10.1098/rspb.2018.0857]