Behavioural drivers of survey bias: interactive effects of personality, the perceived risk and device properties

Detecting small mammal species for wildlife research and management typically depends on animals deciding to engage with a device, for instance, by entering a trap. While some animals engage and are detected, others do not, and we often lack a mechanistic understanding of what drives these decisions. As trappability can be influenced by traits of personality, personality has high potential to similarly influence detection success for non-capture devices (chew-track cards, tracking tunnels, etc.). We present a conceptual model of the detection process where animal behaviours which are detected by different devices are grouped into tiers based on the degree of intimacy with a device (e.g., approach, interact, enter). Each tier is associated with an increase in the perceived danger of engaging with a device, and an increase in the potential for personality bias. To test this model, we first surveyed 36 populations of free-living black rats (Rattus rattus), a global pest species, to uniquely mark individuals (n = 128) and quantify personality traits. We then filmed rat behaviour at novel tracking tunnels with different risk-reward treatments. As predicted, detection biases were driven by personality, the bias increased with each tier and differed between the risk treatments. Our findings suggest that personality biases are not limited to live-capture traps but are widespread across devices which detect specific animal behaviours. In showing that biases can be predictable, we also show biases can be managed. We recommend that studies involving small mammal sampling report on steps taken to manage a personality-driven bias.

Automated summary

This study surveyed free-living black rat populations, filmed rat behavior in novel tracking tunnels with different risk-reward treatments, and found that detection biases were driven by personality, increased with each tier, and differed between risk treatments. This suggests that biases driven by personality are predictable and can be managed across devices detecting specific animal behaviors.