Adjusting behavioural responses to that of important social partners in novel or ambiguous situations can be advantageous for young/inexperienced individuals. The process by which individuals rely on emotional displays of social partners in evaluating and responding to a novel, ambiguous stimulus or situation is called social referencing (Walden 1993). Obtaining information via observation and behaving accordingly may be considered as a first step of social learning (Heyes 1994) in such contexts. Social referencing may include different behavioural components, such as referential looking, which is defined as looking at an informant immediately preceded and/or followed by looking at a novel stimulus (Russell et al. 1997), and specific behavioural regulation, which is described as the subject’s behaviour consistently influenced by the (emotional) cues provided by the social partner upon encountering a new stimulus (Mumme et al. 1996; Russell et al. 1997). In many situations, however, simpler forms of social learning, such as stimulus enhancement, or behavioural synchronisation—exhibiting similar behaviour at the same time (Duranton and Gaunet 2016)—may also evoke similarly efficient behavioural regulation on the part of the naive subject. For example, young chimpanzees (Pan troglodytes) touched and interacted more with a remote controlled toy after observing their mother interacting with it, compared to when they faced the novel object alone (Tomonaga et al. 2004).
Social referencing has been extensively investigated in infant-parent dyads, the parent being a source of information to which the infant is paying attention. Most paradigms involved infants being presented with an ambiguous object, person or situation in the presence of the caregiver who expressed either happy or fearful facial expressions. In studies using novel or ambiguous objects, when the mother displayed a fearful facial expression, infants avoided an object (Sorce et al. 1985) or played less with ambiguous toys (Gunnar and Stone 1984). Infants were friendlier to a stranger when their mothers spoke positively to them about the stranger (Feinman and Lewis 1983), whereas infants, who observed anxious interactions between their mothers and the stranger, behaved anxiously towards the stranger (de Rosnay et al. 2006).
Merola and colleagues reported that dogs (Canis familiaris) alternated their gaze between their owner and a strange object (fan), but did not find behaviour regulation corresponding to that of their owners’ (Merola et al. 2012a, 2013). However, in another study dogs also adjusted their behaviours in a way that corresponded to their owners’ signals (Merola et al. 2012b). Further, the same study reported that social referencing may also occur when unknown people provide cues for the dogs. These observations were interpreted as dogs showing social referencing in ambiguous situations. Investigating the development of social referencing Fugazza et al. (2018) reported that 8-week-old dog puppies were more likely to approach a novel stimulus and interact with it in the presence of a human showing positive vocal and facial cues, compared to a human showing neutral vocal and facial cues. An hour later, when being alone, puppies regulated their behaviour according to the humans’ previous behaviour indicating that they had learnt about the stimulus in the presence of humans’ signalling.
However, there could be several different social mechanisms contributing to the behavioural change observed in a naïve individual.
In an ambiguous situation, naïve dogs may stay close to their owners even in the absence of any relevant emotional signal, simply because they apply a sort of freezing strategy (Walker et al. 1997) and/or due to the safe haven effect, that is, the owner provides security for the dog (Gácsi et al. 2013a; Cimarelli et al. 2016). These explanations have been supported by results showing that the increase in family dogs’ heart rate was lower when they had to face a threatening stranger in the presence of their owners compared to the condition when they were alone (Gácsi et al. 2013a).
In the case of emotionally loaded human vocalisations, dogs might react to their owner's emotional signal by approaching the signaller irrespective of the novel/ambiguous stimulus in the environment (e.g., dogs tend to look at or approach a crying person: Custance and Mayer 2012). Yong and Ruffman (2015) demonstrated that in such situations dogs may respond to the emotional content of the person’s communication without later connecting it to the novel/ambiguous stimulus. Thus social referencing does not necessarily result in social learning. Although dogs behaved depending on whether an unfamiliar experimenter displayed fear or happiness during an encounter with a small robot, when the dogs were left alone, there was no difference in dogs’ proximity to the robot depending on the emotion the human displayed previously.
Behavioural synchronisation could also play a role in the behaviour of the naïve individual. When movement cues are provided, i.e., stepping forward or backward, dogs may show behaviour synchronisation following the movements of their owner (Duranton et al. 2017a), which might be erroneously interpreted as a reaction to the novel stimulus. Recently Duranton et al. (2016) tested dogs’ reactions towards a stranger in the presence of their owners. In the positive condition, the owners took three steps forward (approach condition), whereas in the negative condition they took three steps back (retreat condition). The owners were not permitted to show any facial expressions, emotions or speak, while the stranger looked at the owners. Most dogs alternated gaze between the owner and the stranger, and they approached the stranger later in the retreat condition than in the approach condition. However, due to the experimental setup, dogs might alternate gaze between the stranger and the owner simply because the owner showed intense behaviours. This could have elicited looking behaviour even in the lack of the stranger and dogs’ moving together with their owners could likely be due to behavioural synchronization rather than social referencing.
The evolutionary function of all these social processes is to guide the young, inexperienced individual in novel/ambiguous situations, and make fast reactions and efficient social learning possible. In lifelike situations, for example, when the owner shows a complex response (vocalisation, movement, facial expression) to a frightening stimulus, these processes may work in combination.
Here, we aimed to extend the social referencing paradigm for dog-owner dyads modelling real life situations, when the owner provides complex responses. Our intention was to examine the overall reactions of dogs in such situations and not to assess the emergence of individual components of social referencing in more controlled and therefore unnatural scenarios. Since well-socialised companion dogs meet strangers in their everyday life in multiple places, we applied a stimulus that could be ambiguous enough to evoke social referencing. The ‘threatening approach’ paradigm (Vas et al. 2005) has been widely used to investigate family dogs’ responses to an ambiguous social stimulus. The name of the procedure is misleading; it refers to its differences from the ‘friendly approach’. Actually, the stranger does not display any direct threat, such as raising her hands or shouting, but silently and slowly approaches the dog while staring at it. Based on several studies, dogs show variable responses towards the stranger, ranging from displaying friendly behaviours (e.g., play bow: Győri et al. 2010), through neutral, and submissive behaviours to fear or aggression related behaviours (Vas et al. 2005; Gácsi et al. 2013b; Kis et al. 2014; Klausz et al. 2014). Thus the novel experimental conditions applied in our experiments were the stranger’s ‘threatening approach’ type of behaviour and the provision of lifelike complex behavioural responses on the part of the owner. So far, dogs’ reactions to strangers based on their owners’ cues have only been examined in situations when the owners provided only movement cues to the dogs (Duranton et al. 2016, 2017b).
To assess family dogs’ responses to their owners’ suspicious and/or reassuring reactions we carried out two experiments, which allowed for applying both between- and within-subjects designs and also to model two slightly different lifelike situations. In Experiment 1, the dogs faced the stranger with their owners in a context modelling an office situation where the owners were sitting at a fixed location during the encounter. Applying a between-subjects design, the owners displayed reassuring behaviours in the reassuring owner (RO) group and suspicious behaviours in the suspicious owner (SO) group with their actions limited (by the situation) to voice intonations and body posture changes when talking to the threatening stranger.
In Experiment 2, to avoid the effects of potential uncontrolled factors, dogs (different from those involved in Experiment 1) were tested in a within-subjects design so that the same dogs participated in both the RO and SO conditions. Dogs faced the stranger with their owners in a large hall modelling a street or public place (e.g., shopping malls), where the larger distances (compared to Experiment 1) allowed the dog more time to perceive and respond to the behaviour of both the stranger and the owner. Depending on the condition, owners made one small step towards or away from the threatening stranger and talked to her with a happy or worrying intonation, respectively.
We predicted that in such social situations, dogs would respond to the ambiguous stimulus according to their owners’ behaviour. Specifically, when the owner displayed worrying behaviours, dogs would show more avoidance towards the stranger and/or more proximity seeking with the owner, and in contrast, when receiving reassuring cues, dogs would approach the stranger.