Abstract
Current research on acoustic encoding of emotional content suggests that there are universal cues, allowing for decoding within and across taxa. This is particularly important for human-animal relationships, wherein domestic animals are supposed to be particularly efficient in decoding human emotions. Here we investigated whether the decoding of the emotional content in human voices shared universal acoustic properties, or whether it could be influenced by experience. Emotional human voices were presented to two populations of horses, in which behavioral, cardiac, and brain responses were measured. The two populations differed in their living and working conditions: one population lived in naturalistic conditions (stable social groups in pastures) and were ridden occasionally for outdoor trail riding with one to a few different riders, while the other was kept in more restricted conditions (individual stalls) and participated in riding lessons involving many different riders. Assessment of the horses' welfare state (animal-based measures) and their relationships with humans, performed independently of the playback experiments, revealed that the populations differed in both aspects. Whereas both populations appeared to react to the angry human voice, the population with the best welfare state and relationship with humans showed little differentiation between the different emotional voices and exhibited low behavioral reactions. On the contrary, the other population showed high behavioral and cardiac reactions to all negative voices. Brain responses also differed, with the first population showing higher responses (increased gamma, i.e., excitation) for the happy voice and the second for fear and anger (increased theta, i.e., alarm). Thus, animals' affective state and past experiences appear very influential for their perception of (cross-taxa) acoustic emotional cues.