Researchers often use artificial models of animals to elicit and study behavior. Until recently, these models were typically handcrafted; however, 3D-printing technology has been adopted by researchers looking to create accurate and consistent animal models from scans of living animals, taxidermies, or existing models. While 3D-printing techniques create models with accurate and repeatable shape and size, applying coloration to these models is still typically achieved with traditional methods, such as painting by hand. These approaches can be time-consuming and require high levels of artistic skill, creating a barrier to producing realistic models, especially when more than one model or standardized coloration is required. Here, we present a simple workflow to avoid these issues by creating a photograph-accurate paper "skin" that can be glued onto 3D-printed animal models to provide surface coloration. We have used this methodology to create avian models for several experiments, and found that it can create highly detailed and standardized models with minimal training and is independent of artistic skill. Additionally, this method allows the files needed to accurately recreate models to be shared digitally with other researchers, further enhancing repeatability in the field.

Effective responses to predation or competition are expected to increase survival and the overall reproductive success and thus fitness of individuals. Therefore, most animals are expected to exhibit some form of defence, whether physical and/or behavioural, against predators and competitors. In cooperative species, these defences are often adapted to groupliving species; for example, alarm calling to warn nearby conspecifics, or mobbing as part of a group to chase off a predator. The Noisy Miner (Manorina melanocephala) is an extreme example of these behaviours; they are a colonial species, with a range of alarm calls to alert conspecifics to potential threats, and aggressive mobbing behaviour used to exclude a large variety of predators, competitors, and other species from their colony space.

In this thesis, I describe the research that I have done on the anti-threat behaviours of Noisy Miners. More specifically, I first describe a novel method for creating high-detail artificial “model” animals to elicit threat-response behaviour. I then test how variation in four traits of model conspecifics affects observed territorial responses from wild Noisy Miners, in order to better understand how model construction and fidelity impacts behavioural responses in a free-living system. Next, I examine the mobbing response of Noisy Miners to models of different predator and competitor species, to determine if mobbing responses are moderated according to threat level and target species. Finally, I take a closer look at miner mobbing recruitment calls to investigate what information these may encode in relation to the stimulus presented, and how this relates to the function of the call and the successful exclusion of many avian species from habitat occupied by Noisy Miners.

We found that artificial models elicit a similar response from Noisy Miners compared to a taxidermized mount. Additionally, we found that the colour-accuracy, detail-accuracy, and posture of Noisy Miners models affected response from wild birds, with low colour-accuracy being associated with more time spent in the vicinity of the model, high detail-acccuracy being associated with fewer social “Q4” calls, and aggressive posture leading to an overall increase in calling. However, all models elicited a ‘territorial’ response appropriate to a conspecific. This implies that models do not need be ‘perfect’ to elicit the intended behaviour in experimental contexts, but that less accurate models might elicit weaker or stronger responses than natural stimuli, a potential experimental confound that should be considered when comparing between behaviour elicited by different models in behavioural experiments.

When mobbing heterospecifics, I found that Noisy Miners mobbed models of high-risk predators for longer than other potential predators or competitors, and were slower to approach the model of a predator that was most likely to prey on Noisy Miners during the period of the model presentation. I therefore conclude that Noisy Miner mobbing behaviour is threat-sensitive, and that Noisy Miners do distinguish between (at least some of) the many species that they mob. However, I also found that mobbing behaviour was not directed most strongly at the competitors that had the highest dietary overlap with miners (and therefore should be the strongest threat), nor at the smallest competitor, despite smaller competitors being more likely to be excluded from Noisy Miner colonies in situ, highlighting that our overall understanding of how Noisy Miners exclude competitors from their colonies remains incomplete.

Finally, examining the structure of Noisy Miner mobbing recruitment calls found that they call in a lower frequency towards high-threat predators. This I speculated to be adaptive – low frequencies are associated with improved transmission through closed habitat, and can be an honest signal of body size, suggesting two possible explanations for our finding. Overall, this shows that Noisy Miner mobbing calls, like mobbing itself, is modulated according to the relative threat of a stimuli. Interestingly however, the difference in frequency did not appear to represent two separate call subtypes, the existence of which in this call has been described previously in the literature. Noisy Miner threat recognition and response behaviours are quite complex, and further research that uncovers more of the drivers behind variation in their chur calls and mobbing would provide a further-improved understanding of anti-threat behaviour and the role of Noisy Miners in the ecosystem. The new method herein for model construction will be very useful in future study of anti-threat behaviour.