Document type: scientific article published in Animal
Authors: I. Reimert, G.A. Franchi, M. Bagaria, L. Canario, J. Elizabeth Bolhuis, S. Düpjan, E. Verbeek, E. Fàbrega, L.J. Pedersen, H. Telkänranta, A. Eggert, L.R. Moscovice
Preview: Evaluating emotions in livestock is challenging due to their subjective nature and the lack of robust indicators, especially for positive emotions. We developed experiments across four facilities, all designed to induce positive and negative emotional reactions in n = 124 female domestic pigs tested in a within-subject design. Our goals were to test for robust, non-invasive indicators of short-term emotional responses across pigs representing different ages and breeds and kept in different husbandry conditions. We developed a common ethogram for behavioural recording and standard operating procedures for collecting saliva samples, which were analysed in the same lab to measure short-term changes in cortisol, as an indicator of arousal, and oxytocin, as a putative indicator of valence. We used linear and generalised linear mixed models (LMM/GLMM) to test for effects of predicted differences in valence and arousal across different experiments on behavioural and physiological responses. We also ran a multilevel principal component analysis (PCA) to assess whether several indicators could be consolidated into a smaller set of principal components representing different dimensions of emotional responses. More established indicators varied as predicted across facilities, with increases in locomotor play in positive contexts (GLMM, P ≤ 0.001) and increases in escape attempts in negative contexts (GLMM, P ≤ 0.001). Pigs more often had both ears pointed forward in positive contexts (GLMM, P ≤ 0.001) and both ears pointed backwards in negative contexts (GLMM, P ≤ 0.001). Pigs had greater increases in cortisol in response to negatively than to positively valenced contexts (LMM, P ≤ 0.001), and regardless of valence, pigs exposed to more novel social or physical contexts during experiments also had greater increases in cortisol (LMM, P ≤ 0.001). Oxytocin did not vary consistently across differently valenced contexts. Pigs’ PCA scores clustered according to the differently valenced contexts that they experienced and not to their different facilities. The two principal components with the highest eigenvalues had moderate loadings for a combination of more established and novel indicators related to fearful vs playful responses. In conclusion, this cross-institute approach helped to validate the use of locomotor play and escape behaviour as robust indicators of short-term valenced responses across pig populations. When combined with other measures, short-term changes in ear posture also have promise as a valence indicator. Changes in cortisol were related to both valence and arousal. Cross-institute approaches are useful for increasing the validity of non-invasive indicators of emotion in farm animals.
