The bold and the biased: the influence of behaviour on capture probability in woylies (Bettongia penicillata ogilbyi)
Natasha D. Harrison
A B *
A
B
Abstract
Consistent differences in behaviour among individual animals are commonly observed. These behavioural types have important implications for monitoring populations as they can have a profound impact on detection probability. Behaviour-driven sampling bias has potential to greatly influence the conservation and management of woylies (Bettongia penicillata ogilbyi), that are primarily monitored using live trapping. This study explicitly tests whether there is any correlation between agitation behaviour (a repeatable trait) and cage trap capture probability in wild woylies, finding no evidence for an effect. This suggests that studies of woylies employing cage trapping are unlikely to be confounded by behaviour-driven sampling biases arising from this trait.
Keywords: animal personality, applied behaviour, bettong, capture probability, non-random sampling, population monitoring, sampling bias, trapping bias.
Introduction
Consistent differences in behaviour among individual animals are commonly observed and often correlated across contexts (Gosling and John 1999; Réale et al. 2007). These behavioural types have profound impacts on many aspects of life history and ecology, shaping how an individual responds to predators, competitors and changes in its environment (Dingemanse and Réale 2005; Réale et al. 2007). Behavioural types can have important implications for studies of wild populations, as they can influence the detectability of individuals (Biro and Dingemanse 2009; Augustine et al. 2014; Hammond et al. 2021). For example, exploratory flycatchers (Ficedula albicollis) were more likely to enter a trap than risk-averse ones (Garamszegi et al. 2009), and bolder agama lizards (Agama planiceps; Carter et al. 2012) and stoats (Mustela erminea; Johnstone et al. 2024) had higher capture probability than shy individuals. This can result in behaviour-driven sampling biases (non-random sampling of a population) (Webster and Rutz 2020), particularly for sampling methods that require individuals to respond to novelty (Stuber et al. 2013), such as live trapping (Wilson et al. 2011; Michelangeli et al. 2016; Jolly et al. 2019; Johnstone et al. 2021; Šmejkal et al. 2022).
Live trapping is the primary method of monitoring wild woylies (Smith et al. 2020; Harrison et al. 2024b), and behaviour-driven sampling bias has the potential to greatly influence the conservation and management of this species. Erroneous estimates of population parameters such as density and survival could confound our understanding of how populations respond to threats and management interventions. For example, quantifying woylie anti-predator behaviours and their effect on fitness has been a focus of recent work (Page et al. 2019; Harrison et al. 2023, 2024a, 2025a, 2025b; Tay et al. 2023). In most cases, behavioural traits are evaluated in individuals during capture (e.g. Page et al. 2019; Harrison et al. 2024a) or on release (e.g. Tay et al. 2023), and survival is determined by the recapture of individuals (e.g. Harrison et al. 2025a). Studies of this nature rest on the assumption that the population has been evenly sampled (so as to adequately represent the range and variation of behavioural phenotypes), and that each behavioural phenotype is equally likely to be detected. Conclusions drawn about associations between behaviour and survival could be fundamentally flawed if there is an underlying correlation between behaviour and capture probability that is not properly accounted for. Flawed population estimates could also arise as non-homogenous detection probability violates a key assumption of capture–recapture modelling frameworks (Jolly et al. 2019), which are commonly used to estimate demographic parameters in woylie populations. Furthermore, woylies are the species most translocated for conservation in Australia (Harrison et al. 2024b), and behaviour-driven capture bias risks reducing phenotypic and genetic diversity in founding populations if only a subset of the population is susceptible to trapping.
Here, I explicitly test whether there is any correlation between agitation behaviour (a trait consistent within individuals; Harrison et al. 2022) and cage trap capture probability in wild woylies at five sites across their extant range.
Materials and methods
Data collection
Woylies (Fig. 1) were captured using wire cage traps (approximately 22 cm × 22 cm × 55 cm; Sheffield Animal Traps, Western Australia) baited with universal bait (peanut butter and rolled oats) as part of a larger study exploring anti-predator behaviours (Harrison 2024). Trapping was conducted across three sites within their remnant indigenous range (Dryandra −32.7819, 116.9238, October 2022; Boyicup −34.2884, 116.5816, March 2022; and Moopinup −34.1455, 116.6623, March 2022), as well as within two fenced havens free from invasive predators (Perup Sanctuary −34.1762, 116.5709, March 2022; and Dryandra Numbat Woylie Sanctuary −32.7712, 117.1119, October 2022). Between 50 and 100 traps were deployed (and checked the following day) for a single session at each site comprising three or four consecutive nights (for further detail see Harrison et al. 2024a). Once captured, woylies were marked for individual identification with uniquely numbered titanium ear tags (Monel 1005-1 ear tags, National Tag and Brand Co., Newport, KY, USA) or a microchip (Microchips Australia), and their sex, age and agitation behaviour were recorded. Woylie agitation behaviour (none, low, medium or high) was quantified at five different time points throughout the handling process: when the handler approaches the trap, places the bag over the trap door, ushers the animal into the bag, when the animal is in the bag before handling and when the animal is in the bag during handling (specific definitions for each score can be found in Supplementary File S1). This particular assay was designed specifically for quantifying agitation behaviour during the capture and processing of woylies (Harrison et al. 2022), and was shown to be repeatable within individuals among numerous woylie populations (Harrison et al. 2025a) meaning that individuals show consistent responses across multiple observations. Scores were combined to give a cumulative agitation score per capture and averaged over all recorded observations for each individual. There were a total of 581 captures of 344 individuals. For each marked individual captured at least once during the session, I recorded which individual trapping nights within the session that they were captured (1) or not (0), resulting in 901 observations. This assumes that the population is closed (no mortality, recruitment, immigration or emigration) during each session. Capture probability was calculated as the proportion of nights each animal was caught within a trapping session (noting that this only includes individuals who were captured at least once, so the minimum capture probability will be 0.25).
Statistical analyses
All statistical analyses were conducted in the R environment (R Core Team 2024). To explore whether agitation behaviour was consistent within individuals, within-individual repeatability of agitation scores was calculated using a mixed model repeatability estimate fitted with restricted maximum likelihood and Poisson distribution using the package ‘rptR’ (Nakagawa and Schielzeth 2010). Error estimates were calculated with 1000 parametric bootstraps. Potential effects of intrinsic and extrinsic factors on woylie capture probability were explored using a binomial mixed effects model from the package ‘glmmTMB’ (Brooks et al. 2017). This model tested for the effects of sex (male or female), age (adult or subadult; based on the presence of a fully developed pouch or testes), the cumulative number of times an individual had been captured previously, mean agitation score and a measure of trap availability. Trap availability was calculated by dividing the number of traps by the area of the trap array (including a 500 m buffer either side of the trap array to account for the size of an average woylie home range; Yeatman and Wayne 2015) multiplied by woylie density at the site (for further details of density calculations see Harrison et al. 2024a). Individual identity nested within population was incorporated as a random effect as there were multiple observations from each individual. Model assumptions were evaluated using the ‘DHARMa’ package (Hartig 2022), and the significance of each parameter was evaluated by performing likelihood ratio tests of the full model with and without each variable.
Ethics
All trapping was conducted with approval from the University of Western Australia Animal Ethics Committee (2021_ET000428), Department of Biodiversity, Conservation and Attractions (DBCA) Animal Ethics Committee (2018/22F) and under DBCA Licence to Disturb Threatened Species (TFA_2223-0127; TFA 2223-0058).
Results
Mean agitation scores ranged from 0 to 15 (the minimum and maximum scores possible using this assay), with a median mean score of 6 (±2.84 s.d.). Agitation scores were repeatable within individuals across multiple captures (R = 0.355; 0.240–0.429 95% CI; P = <0.001). Mean capture probability across all individuals was 0.5 (±0.28 s.d.). There was no evidence for an influence of agitation behaviour on woylie capture probability. Capture probability was higher in males compared to females, but no effect of age, number of previous captures, trap availability or mean agitation score could be detected (Table 1). The variance attributed to population was 2.182 (±1.477 s.d.), indicating substantial variation in capture probability across populations. In contrast, the variance for individual within population was negligible (1.720 × 10−9 (±4.148 × 10−5 s.d.), suggesting minimal individual-level variation beyond that explained by population.
| Parameter | Estimate | s.e. | df | P-value | |
|---|---|---|---|---|---|
| Intercept | 2.348 | 1.648 | – | – | |
| Number previous captures | 0.001 | 0.002 | 1 | 0.745 | |
| Sex_Male | 0.315 | 0.151 | 1 | 0.037 | |
| Age_Subadult | 0.049 | 0.379 | 1 | 0.898 | |
| Trap availability | −7.016 | 8.399 | 1 | 0.371 | |
| Mean agitation score | −0.034 | 0.029 | 1 | 0.247 |
The level of each variable is listed after the ‘_’, e.g. Sex_Male. The estimate, corresponding standard error (s.e.) and the degrees of freedom (df) are presented, as well as P-values resulting from likelihood ratio tests of the full model with and without each respective variable.
Discussion
Animal behaviour is an important driver of how individuals interact with their environment (Réale et al. 2007) and has the potential to introduce sampling bias (Biro and Dingemanse 2009). Despite woylies exhibiting consistent, repeatable agitation responses across multiple captures, agitation behaviour had no detectable effect on capture probability, suggesting that studies of woylies using live trapping are unlikely to be confounded by sampling biases resulting from this trait. This study underscores the importance of incorporating animal behaviour in ecology and conservation (Blumstein and Fernández-Juricic 2004; Réale et al. 2007), and further supports existing work suggesting that behaviour-driven sampling bias is not inevitable (Michelangeli et al. 2016; Jolly et al. 2019).
While there was no evidence for a correlation between agitation behaviour and capture probability, this analysis only considers a single behavioural trait. The agitation assay evaluated here represents boldness in woylies (Harrison et al. 2022), a behaviour that is commonly found to influence capture probability across taxa (Wilson et al. 2011; Carter et al. 2012; Johnstone et al. 2021, 2024). Nonetheless, the lack of correlation found here does not rule out the possibility that other behavioural traits could influence the detectability of this species, and as such, it is recommended that behaviour still be a consideration in ecological studies of woylies. For example, exploring how the detection of individuals may vary across age, locations and seasons, which are known to influence mammal trappability (Ham et al. 2019), and incorporating learned responses to detectors in capture-mark recapture studies (Augustine et al. 2014).
A limitation of this study, (and more generally of studies involving animals that lack unique visually identifiable features, such as spot or stripe patterns), is that individuals must be captured at least once to be marked or identified (Biro 2013). It is possible that this dependence on initial capture could introduce sampling bias (Biro 2013), whereby some individuals persistently avoid entering traps at all, and that such individuals could plausibly have a different behavioural phenotype. This is unlikely to be the case here, given the generally high capture probability, vast range of behavioural phenotypes captured, and near-zero contribution of individual identity to variation in capture probability. Repeating this study using a method of identification and behavioural assay that is not conditional on initial capture would be informative. Where there is doubt about the potential for a confounding bias between a behavioural type and sampling method, consideration should be given to expanding the sampling regime to attract a wider variety of individuals (Webster and Rutz 2020), for example, by varying the trap type, sampling location and timing, or bait type.
Data availability
Data relevant to this study are available at: https://github.com/natasha-harrison/Woylie.
Declaration of funding
This project was supported by the Hermon Slade Foundation (HSF21054), the Holsworth Wildlife Research Endowment, the Royal Society of Western Australia John Glover Research Support Grant and the Wettenhall Environmental Trust.
Acknowledgements
I acknowledge the Kaniyang, Minang, Pibelman, Whadjuk and Wilman Aboriginal people as the Traditional Owners of the lands on which this study took place. Thank you to the numerous collaborators, DBCA staff, and volunteers involved with the collection of these data and with the broader project, especially Nicola Mitchell, Ben Phillips and Adrian Wayne. Thanks also to two anonymous reviewers for their constructive feedback.
References
Augustine, B. C., Tredick, C. A., and Bonner, S. J. (2014). Accounting for behavioural response to capture when estimating population size from hair snare studies with missing data. Methods in Ecology and Evolution 5(11), 1154-1161.
| Crossref | Google Scholar |
Biro, P. A. (2013). Are most samples of animals systematically biased? Consistent individual trait differences bias samples despite random sampling. Oecologia 171(2), 339-345.
| Crossref | Google Scholar | PubMed |
Biro, P. A., and Dingemanse, N. J. (2009). Sampling bias resulting from animal personality. Trends in Ecology & Evolution 24(2), 66-67.
| Crossref | Google Scholar | PubMed |
Blumstein, D. T., and Fernández-Juricic, E. (2004 Available at http://www.jstor.org/stable/3588984). The Emergence of Conservation Behavior. Conservation Biology 18(5), 1175-1177.
| Google Scholar |
Brooks, M., Kristensen, K., van Benthem, K., Magnusson, A., Berg, C., Nielsen, A., Skaug, H., Mächler, M., and Bolker, B. (2017). glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. R Journal 9, 378-400.
| Crossref | Google Scholar |
Carter, A. J., Heinsohn, R., Goldizen, A. W., and Biro, P. A. (2012). Boldness, trappability and sampling bias in wild lizards. Animal Behaviour 83(4), 1051-1058.
| Crossref | Google Scholar |
Dingemanse, N. J., and Réale, D. (2005 Available at http://www.jstor.org/stable/4536295). Natural Selection and Animal Personality. Behaviour 142(9/10), 1159-1184.
| Google Scholar |
Garamszegi, L. Z., Eens, M., and Török, J. (2009). Behavioural syndromes and trappability in free-living collared flycatchers, Ficedula albicollis. Animal Behaviour 77(4), 803-812.
| Crossref | Google Scholar |
Gosling, S. D., and John, O. P. (1999). Personality Dimensions in Nonhuman Animals. Current Directions in Psychological Science 8(3), 69-75.
| Crossref | Google Scholar |
Ham, C., Donnelly, C. A., Astley, K. L., Jackson, S. Y. B., and Woodroffe, R. (2019). Effect of culling on individual badger Meles meles behaviour: Potential implications for bovine tuberculosis transmission. Journal of Applied Ecology 56(11), 2390-2399.
| Crossref | Google Scholar | PubMed |
Hammond, T. T., Curtis, M. J., Jacobs, L. E., Tobler, M. W., Swaisgood, R. R., and Shier, D. M. (2021). Behavior and detection method influence detection probability of a translocated, endangered amphibian. Animal Conservation 24(3), 401-411.
| Crossref | Google Scholar |
Harrison, N. D. (2024). Loss of anti-predator traits from havened Australian mammal populations: drivers, consequences, and mitigation tools. Ph.D. thesis, University of Western Australia, Western Australia, Australia 10.26182/hgyx-6j84
Harrison, N. D., Frick, C. H., and Wayne, A. F. (2022). Repeatable measure of cage trap behaviour to quantify boldness and agitation in a macropod. Australian Mammalogy 45(2), 237-240.
| Crossref | Google Scholar |
Harrison, N. D., Phillips, B. L., Mitchell, N. J., Wayne, J. C., Maxwell, M. A., Ward, C. G., and Wayne, A. F. (2023). Perverse outcomes from fencing fauna: Loss of antipredator traits in a havened mammal population. Biological Conservation 281, 110000.
| Crossref | Google Scholar |
Harrison, N. D., Phillips, B. L., Wayne, A. F., and Mitchell, N. J. (2024a). Sustained predation pressure may prevent the loss of anti-predator traits from havened populations. Ecology and Evolution 14(7), e11668.
| Crossref | Google Scholar | PubMed |
Harrison, N. D., Thorn, S. M., Maxwell, M. A., Ward, C. G., Wayne, J. C., and Wayne, A. F. (2024b). Insuring woylies (Bettongia penicillata ogilbyi) against extinction: establishment of Perup Sanctuary. Wildlife Research 51(7), WR23056.
| Crossref | Google Scholar |
Harrison, N. D., Frick, C. H., Wayne, A. F., Mitchell, N. J., Valentine, L., Wayne, J. C., Sandow, D., Brewster, R., and Phillips, B. L. (2025a). The quick and the dead: behavioural plasticity of anti-predator responses in an Endangered mammal. Conservation Science and Practice 7(4), e70028.
| Crossref | Google Scholar |
Harrison, N. D., Phillips, B. L., Mitchell, N. J., Macmahon, B., Rasmussen, D., Lacey, P., Fanson, K., and Wayne, A. F. (2025b). Dancing with the devil: Could native predators inside ‘predator‐free’ havens be good for the conservation of threatened native prey species? Journal of Applied Ecology 62(3), 444-457.
| Crossref | Google Scholar |
Hartig, F. (2022). ‘DHARMa: Residual Diagnostics for Hierarchical (Multi-Level/Mixed) Regression Models.’ R package version 0.4.6. Available at https://github.com/florianhartig/dharma
Johnstone, K. C., McArthur, C., and Banks, P. B. (2021). Catch me if you can: personality drives technique-specific biases during live-capture trapping. Wildlife Research 48(8), 713-721.
| Crossref | Google Scholar |
Johnstone, K. C., Price, C., and Garvey, P. M. (2024). Personality, sex and capture biases: Challenges for predator monitoring and management. Journal of Applied Ecology 61(9), 2207-2218.
| Crossref | Google Scholar |
Jolly, C. J., Webb, J. K., Gillespie, G. R., Hughes, N. K., and Phillips, B. L. (2019). Bias averted: personality may not influence trappability. Behavioral Ecology and Sociobiology 73(9), 129.
| Crossref | Google Scholar |
Michelangeli, M., Wong, B. B. M., and Chapple, D. G. (2016). It’s a trap: sampling bias due to animal personality is not always inevitable. Behavioral Ecology 27(1), 62-67.
| Crossref | Google Scholar |
Nakagawa, S., and Schielzeth, H. (2010). Repeatability for Gaussian and non‐Gaussian data: a practical guide for biologists. Biological Reviews 85(4), 935-956.
| Crossref | Google Scholar | PubMed |
Page, K. D., Ruykys, L., Miller, D. W., Adams, P. J., Bateman, P. W., and Fleming, P. A. (2019). Influences of behaviour and physiology on body mass gain in the woylie (Bettongia penicillata ogilbyi) post-translocation. Wildlife Research 46(5), 429-443.
| Crossref | Google Scholar |
R Core Team. (2024). ‘R: A Language and Environment for Statistical Computing.’ (R Foundation for Statistical Computing: Vienna, Austria). Available at https://www.R-project.org/.
Réale, D., Reader, S. M., Sol, D., McDougall, P. T., and Dingemanse, N. J. (2007). Integrating animal temperament within ecology and evolution. Biological Reviews 82(2), 291-318.
| Crossref | Google Scholar | PubMed |
Šmejkal, M., Bartoň, D., Brabec, M., Sajdlová, Z., Souza, A. T., Moraes, K. R., Blabolil, P., Vejřík, L., and Kubečka, J. (2022). Behaviour affects capture probability by active sampling gear in a cyprinid fish. Fisheries Research 249, 106267.
| Crossref | Google Scholar |
Smith, M., Volck, G., Palmer, N., Jackson, C., Moir, C., Parker, R., Palmer, B., and Thomasz, A. (2020). Conserving the endangered woylie (Bettongia penicillata ogilbyi): Establishing a semi‐arid population within a fenced safe haven. Ecological Management & Restoration 21(2), 108-114.
| Crossref | Google Scholar |
Stuber, E. F., Araya-Ajoy, Y. G., Mathot, K. J., Mutzel, A., Nicolaus, M., Wijmenga, J. J., Mueller, J. C., and Dingemanse, N. J. (2013). Slow explorers take less risk: a problem of sampling bias in ecological studies. Behavioral Ecology 24(5), 1092-1098.
| Crossref | Google Scholar |
Tay, N. E., Warburton, N. M., Moseby, K. E., and Fleming, P. A. (2023). Predator escape behaviour in threatened marsupials. Animal Conservation 26(4), 587-601.
| Crossref | Google Scholar |
Webster, M. M., and Rutz, C. (2020). How STRANGE are your study animals? Nature 582(7812), 337-340.
| Crossref | Google Scholar | PubMed |
Wilson, A. D. M., Binder, T. R., McGrath, K. P., Cooke, S. J., and Godin, J.-G. J. (2011). Capture technique and fish personality: angling targets timid bluegill sunfish, Lepomis macrochirus. Canadian Journal of Fisheries and Aquatic Sciences 68(5), 749-757.
| Crossref | Google Scholar |
Yeatman, G. J., and Wayne, A. F. (2015). Seasonal home range and habitat use of a critically endangered marsupial (Bettongia penicillata ogilbyi) inside and outside a predator-proof sanctuary. Australian Mammalogy 37(2), 157-163.
| Crossref | Google Scholar |
