The Role of Learning in Making Precision Livestock Technologies Welfare-Friendly: A Virtual Fencing Case Study

Abstract

<p>Virtual fencing is a new technology which requires sheep to learn to respond to an audio cue in order to avoid receiving an aversive (unpleasant) electrical pulse stimulus. The welfare impacts of this new technology have not yet been investigated in sheep. Through the papers presented in this thesis, this project aimed to investigate the welfare impacts of virtual fencing on sheep using commonly used measures of stress and behaviour to make inferences about the affective state of the animals. From their initial encounter with the stimuli, through the learning process if their situation becomes predictable and controllable, and how controllability impacts on their perception of the stimuli. Naïve sheep were exposed to virtual fence stimuli and their stress and behavioural responses were compared with that of sheep exposed to commonly encountered stimuli, and control sheep with no exposure to stimuli. It was found that sheep exposed to the virtual fence audio cue perceived it to be largely benign, and less aversive than the sound of a dog barking. Sheep exposed to the electrical pulse stimulus perceived it to be mildly aversive, but less aversive than a commonly used inverted restraint procedure. When sheep learn to respond to the audio cue to avoid receiving an electrical pulse (predictable controllability), their stress responses upon hearing the cue are not different to that of control animals which did not receive stimulus. Animals that learned to associate the audio cue with receiving an electrical pulse that was not altered by behaviour (predictable uncontrollability) showed increased stress responses to the audio cue, more so when uncontrollability was confirmed by receiving the electrical pulse. Finally, the project investigated the role of maternal learning in the ability of lambs to later learn a virtual fence. The role of social influences on learning was investigated and it was found that lambs which had observed naïve maternal demonstrators interacting with the virtual fence later learned the system more rapidly than their peers which had not had a demonstrator, and peers which had observed trained maternal demonstrators. This suggests that the behaviour modelled by the demonstrator has an impact on the ability of virtual fence training to be transmitted to the lambs. Overall, this project has demonstrated that how the animal learns about the associations between the audio cue and the electrical pulse strongly affects the welfare implications of this technology. Rapid learning of the virtual fence system reduces the time spent in the unpredictable and uncontrollable phase of learning, leading to better welfare as the animals are able to predict and control their experience of the virtual fence. </p>