The identification of facial expressions of emotions in dogs (Canis familiaris) using a Facial Action Coding System

Objectively assessing animal emotions is challenging and requires the development of valid and reliable indicators of emotions. Emotional states are accompanied by behavioural expressions. If specific expressions reliably accompany specific emotional states across contexts, then those expressions have potential to serve as indicators of the emotional states. In human emotion research, particularly facial expressions have been studied extensively for this purpose and therefore may help infer animal emotions as well. This project was aimed at studying facial expressions of dogs exposed to situations that are likely to elicit positive anticipation and frustration. Both emotions can be triggered in situations related to the expectation of a reward: while positive anticipation can occur between signalling and delivery of a reward, it may turn into frustration when the reward remains inaccessible. In a series of studies, using the contingencies described, two contextual features were systematically varied – the expected reward type (food/toys) and the social context (non-social/social; i.e. whether the reinforcement was associated with a human or not). The main goal was to identify facial expressions that are consistently associated with either positive anticipation or frustration across contexts, as these expressions may serve as indicators of the respective emotional states. To measure dogs’ facial expressions, the Dog Facial Action Coding System (DogFACS) was used, which is the dog�specific adaptation of FACS, the gold standard for measuring facial expressions in human emotion research. Ears adductor was the only variable that was more common during positive anticipation, and it was consistently associated with this state across (reward and social) contexts. The antagonistic movement to the Ears adductor, Ears downward, as well as Ears flattener and Nose lick were more common during frustration across reward and social contexts. Despite the consistent association of those facial expressions with positive anticipation or frustration, none of them would have allowed consistent, correct designations of the associated emotional state when used as individual indicators. Diagnostic accuracy assessments showed that validity estimates of the Ears adductor varied greatly across contexts: whereas sensitivity was low and specificity high in a non-social context, it was the other way round in a social context. Similar to the Ears adductor, validity estimates of Ears downward showed an inverse pattern between contexts: while in the non-social context Ears downward was more sensitive than specific, this was the other way round in the social context. Accuracy estimates of Ears flattener were more consistent, despite some variation. Nose lick was the variable with the most stable accuracy measures across contexts. On their own, these facial expressions would not serve as highly reliable, robust, and valid indicators of positive anticipation or frustration in dogs. However, they may be potential candidates for the future development of indicators of these states, e.g. when combined with other facial or body expressions. Additional facial expressions that were associated with frustration were Blink, Tongue show, Lip corner puller, Jaw drop, and Lips part. However, although they were not affected by the expected reward type, they only accompanied frustration in a non-social context. Therefore, these facial expressions are less likely robust candidates for the development of indicators of frustration in dogs. The Upper lip raiser also accompanied frustration in non-social contexts, but it was influenced by the expected reward type and may therefore be more related to the associated motivational state. The last study of this thesis focussed exclusively on the Inner brow raiser, which is a facial expression that has received considerable attention related to its role in dog-human communication. When its production was compared between a social and a non-social context, it occurred more frequently in the non-social context, challenging the previous hypothesis of a communicative function. We also found the Inner brow raiser to be strongly associated with eye movements, which suggests a proximate mechanism behind this facial expression. In this project, facial expressions, an infrequently studied modality in animal (emotion) research, were identified that consistently accompanied either positive anticipation or frustration across different (reward and social) contexts in dogs. Although the Ears adductor was consistently associated with positive anticipation and Ears downward, Ears flattener, and Nose lick with frustration, they do not seem to constitute reliable, robust, and valid indicators of the respective emotions in their own. Nonetheless, they are potential candidates that provide a starting point for the future development of emotion indicators by systematically examining them in combination with other expressive behaviours. The introduction of diagnostic accuracy assessments is a pioneering approach to animal emotion research, providing novel methods to advance the evaluation of the validity of putative indicators of animal emotions