THE INTERPRETATION OF EMOTIONS IN SMALL RUMINANTS

Ensuring that animals have a life worth living means not only reducing their exposure to stressful and/or negative situations, but above all ensuring that they can experience predominantly positive affective states and emotions. The aim of this thesis was to improve human knowledge of how small ruminants (goats, Capra hircus, and sheep, Ovis aries) express emotions, with a particular focus on the acoustic features of vocalisations that can be used as animal-based indicators of the affective states of these animals. To pursue this goal, a multidisciplinary approach was adopted in this thesis, ranging from the study of the direct perception and interpretation of vocalisations by humans to the use of traditional acoustic analysis to decipher the emotional content of vocalisations, through the extraction of acoustic features and the application of statistical models, and their practical application for the development of up-to-date Precision Livestock Farming (PLF) technologies. The general introduction of the thesis provided an overview of the state of the art of research on small ruminant emotions and included a review paper on the topic of human-goat relationship. In fact, in farmed animals like sheep and goats, the quality of the human-animal relationship (HAR) depends on the emotional connotations of the interactions that occur between the two parties. The review focused on how humans and goats communicate, which factors influence this relationship, which methods are available to measure it, and how it can be improved. The review showed that certain interspecies communication channels, particularly the acoustic ones, have received little attention so far, and highlighted the importance of early, frequent and positive interactions with goats to improve HAR, and the role of some socio-demographic variables, such as gender, in influencing behaviour and attitude towards these animals. The review also highlighted the poor feasibility of the validated measures available for farmers to use for self-assessment of HAR, and suggested that farmers’ sense of commitment to goats could be improved by increasing their awareness of goats’ ability to express emotional states. The first research chapter presented the development of a wireless acoustic sensor network for the continuous and non-invasive monitoring of goat vocalisations. The design, architecture and implementation of the system were presented, together with the hierarchical classification of goat vocalisations adopted for the manual annotation of these acoustic signals. In this context, the study also presented the development of a web-based annotation tool of goat vocalisations, to allow the association of labels, namely the context of emissions, to goat vocalisations. Finally, the study presented the development of a smartphone application to alert farmers and inform them of the emotional state of their goats. 5 The second research chapter is an investigation of human perception of goat emotions based on their vocalisations. This human ability was investigated taking into account the respondents’ previous experience with goats, their level of empathy towards these animals, and other individual characteristics that may influence the recognition of animal emotions, and by considering the dimensional framework of animal emotions, which considers them as characterised by the dimensions of valence (positive vs negative) and arousal (low vs high). It was found that humans are able to recognise the emotional dimension of valence in goat vocalisations, as suggested by the consistency of the qualitative description of these vocalisations with their context of emission. The study also revealed a bias in the human ability to classify goat vocalisations emitted in negative contexts, presumably due to the fact that signals emitted in negative emergency situations carry more important and urgent information than those emitted in positive situations. The human ability to classify goat vocalisations was found to be influenced by the level of experience with these animals, with people with frequent direct contact with goats showing higher correct classification rates of the context of emission. Other individual characteristics, such as the level of empathy towards goats, seemed to improve the association of bleats to the correct emission contexts, although further studies are needed to clarify the role of empathy, attitudes and their underlying factors. Having children, gender and university education were not found to influence the human ability to interpret goat vocalisations. In the third research chapter, acoustic, behavioural and physiological markers of negative arousal in lambs were investigated in a two-phases isolation test (partial vs full) designed to induce negative emotions, but characterised by different intensities of emotional arousal. Results showed that lambs produced more vocalizations per time unit and with higher frequencies (energy quartiles, formants, fundamental frequency) and Wiener entropy when experiencing negative arousal. These acoustic features were also found to increase with the bodily activation of the individual (namely the level of activity) as the arousal increased. Although the use of thermography as a method of detecting emotional states in animals seems promising, the results using temperature as a physiological indicator were inconsistent. Finally, the individual level of sociability did not affect the behavioural and physiological response of lambs to isolation, while an effect of the size of the lamb was found; both factors deserve further investigation. Overall, the results of these studies advance the understanding of the expression of emotions in small ruminants, with practical implications for improving the quality of the human-animal relationship and the level of animal welfare, and highlight an important and potential role for PLF technologies in the automated collection and interpretation of emotions in sheep and goats.