mise à jour du
19 décembre 2010
Anim Behav
 A test of the yawning contagion and emotional connectedness hypothesis in dogs, Canis familiaris
Sean J. O'Hara, Amy V. Reeve
School of Environment and Life Sciences, University of Salford


The idea that yawning contagion in humans is predicated on an empathetic response was recently extended to dogs following the report of cross-species contagious yawning in this species (human-dogs). We attempted to replicate contagious yawning in dogs and to investigate what conditions promote yawning contagion. We tested the emotional connectedness hypothesis, which predicts that contagious yawning is precipitated by exposure to (1) familiar model (primary caregiver) and/or (2) similar model (conspecific) stimuli, if contagious yawning is empathy related. In our sample of family-owned and rescue centre dogs we further predicted that rescue centre dogs would yawn contagiously less than owned dogs on the basis that they are more lacking in emotional closeness and have more weakly bonded relationships with their primary care provider than owned dogs. We also controlled for stress since stress can induce yawning. While we found that five of 19 dogs did yawn in response to an unfamiliar human yawning, we were unable to confirm contagion. Our results provide no support for empathy-based, emotionally connected yawning contagion in dogs and casts doubt on the recently documented phenomenon of crossspecies contagious yawning. We interpret our findings as showing that if dogs are seen yawning contagiously then the contagion must be explained on less cognitively stringent grounds than empathy.
Tous les articles sur la contagion du bâillement
All articles about contagious yawning

A contagion effect of yawning is well established in humans. Even mention of yawning can be sufficient to induce yawning in some people (Provine 1986). Recent attempts to understand what underlies the contagion have focused on the suggestion that catching' another's yawn may be an empathetic response (e.g. JolyMascheroni et al. 2008; Arnott et al. 2009; Palagi et al. 2009) which, in turn, potentially requires self-awareness (Platek et al. 2003). According to this view, although yawning is widespread in vertebrates, the capacity to yawn contagiously is expected to be limited to those species capable of inferring the mental states of others and possessing a concept of self.
The first empirical evidence for contagious yawning in nonhuman subjects came from a study of chimpanzees, Pan troglodytes (Anderson et al. 2004; see also Anderson & Matsuzawa 2006) where the contagion effect was only marginally lower than those reported for humans (Provine 1986; Platek et al. 2003); while a more recent attempt to test for contagious yawning in stumptailed macaques, Macaca arctoides, proved equivocal, as although the macaques yawned significantly more when exposed to conspecific yawning stimuli than to controls, the authors themselves acknowledged that they could not exclude the possibility that some yawns were tension induced (i.e. anxiety yawns), and not real' yawns (Paukner & Anderson 2006). These results are consistent with current thinking that higher mental state attribution (e.g. theory of mind) may be limited, among primates, to humans and perhaps great apes (Call & Tomasello 2008; Kaminski et al. 2008), species known also to demonstrate empathy (de Waal 1996, 2008).
It was thus somewhat surprising then that Joly-Mascheroni et al. (2008) recently claimed that domestic dogs yawned contagiously and did so at rates 12-32% higher than had been reported for humans. The authors speculated that dogs might possess rudimentary forms of empathy. This view arises from suggestions that dogs may have the capacity to empathize with humans since they adjust their behaviour to coincide with human behaviour (Topal et al. 1998), and, through the process of domestication, dogs have acquired exceptional abilities to decode human signals through positive selection for this trait (Hare & Tomasello 2005; Hare 2007). Dog sensitivity to human directives could thus potentially make them capable of interpreting human emotions through behavioural cueing. A recent similarly framed duplication of the dog study by Harr et al. (2009) failed to replicate Joly-Mascheroni et al's (2008) original findings. One problem that dogged the macaque study was the presence of tension yawns in the sample. The same problem could not be ruled out from Joly-Mascheroni et al's. (2008) dog study. Harr et al. (2009) attended to this concern in their study design by visually monitoring the study subjects for behavioural indicators of anxiety so that they could exclude 'tension' yawns. They differed also by presenting their stimuli visually on an LCD screen rather than in person and they added a conspecific test alongside the human yawn stimulus. Harr et al. (2009) found very limited evidence for contagious yawning in dogs. With such conflicting results where do these findings leave us?
Empathy research in humans asserts that empathic tendencies are strongest or most likely to arise as a function of familiarity (Kahlil 2002; Preston & de Waal 2002; de Waal 2008), and Bates et al.'s (2008) analyses of the emotions of wild elephants, Loxodonta africana, suggest the same might be true for them. In a recent study of contagious yawning in gelada baboons, Theropithecus gelada, Palagi et al. (2009) found that contagion was best predicted by emotional closeness to the yawner. However, although widely assumed, direct tests of the familiarity hypothesis are rare.
Yoon & Tennie (2010) are also open to the possibility that contagious yawning can occur in vertebrates lacking a theory of mind. They recently proposed that nonconscious mimicry, and 'an even lower-level mechanism', contagion, could also provide potential explanations for contagious yawning (Yoon & Tennie 2010, page e2) and outlined ways in which to distinguish them. For example, mimicry would occur in social contexts with preferred or desirable partners and would be likely to involve copying a wider range of behaviours than are normally attributed to contagion. Contagion under these criteria would apply if a full response occurs in an agent even when only a part of the full action was observed in the target.
In our study we used domestic dogs. Through cohabitation, dogs have acquired a putative shared emotional connectedness to humans as well as social and environmental proximities. Archaeological dog burial evidence suggests sharing our environment, dependence and companionship have been features of the human-dog relationship for ca. 8500 years (Morey & Wiant 1992). During this time there has been selection for levels of communication and interaction with humans that are unparalleled in any other nonhuman animal.
Here, we tested for a yawning contagion in dogs. We built into our study design trials using familiar and similar yawner stimuli (owners and conspecifics, respectively) with the prediction that dogs should be more likely to yawn contagiously at (1) familiar humans (i.e. owners) than unfamiliar humans (i.e. a female researcher) and (2) to members of their own species (conspecifics) than heterospecifics (especially unfamiliar humans), if dogs are showing familiarity/similarity-based empathy-induced yawning contagion. We further hypothesized that dogs in human-bonded relationships (i.e. owned pets) should be more likely to yawn contagiously than dogs living in rescue shelters, which are potentially more contact deprived and/or have more weakly bonded relationships with their primary care provider(s) (i.e. more lacking in emotional closeness), if contagious yawning is empathy related.
Our results show that if dogs yawn contagiously, then they do so only at a low level. Although we detected some contagion effect in rescue centre dogs (less likely to yawn at controls than owned dogs), we did not find any evidence for contagion at the individual level or in our overall sample. Our tests of familiarity, similarity and bonding to humans all failed to predict yawning and thus we found little support for an empathy-related yawning contagion in dogs. These results, together with those of Harr et al. (2009), cast further doubt on the recently documented phenomenon of cross-species contagious yawning reported by Joly-Mascheroni et al. (2008).
Importantly, while our trial conditions were shown not to affect subjects' heart rates adversely, they do hint at the possibility that first exposure to a test condition elevates heart rate. This suggests one-off testing procedures might be susceptible to stressing dogs if other systems are not put in place to mitigate against this risk. Because stressing has the potential to induce 'tension' yawns we took several steps to ensure against this: (owned) dogs were tested in familiar surroundings: we used a single researcher who was a female: the dog's companion was present throughout: and each trial condition component was limited to a fixed time. From our methods it is not possible to distinguish whether experimental order or first testing was responsible for the higher heart rate in week 1: however, we suggest the latter as experimental order would predict the reverse effect: a higher heart rate in week 2 when dogs were exposed to the unfamiliar researcher rather than in week I trials where dogs were tested with a familiar person. Knowing baseline heart rates for each subject could have given us clearer indications of deviations from 'normal' levels; however, our main interest here was simply to evaluate how stressful or otherwise the trial condition procedure might be on subjects and it was not practical for us to gather baseline data.
Twenty-one of the 29 pet dogs (72%) in Joly-Mascheroni et al's (2008) study yawned at the unfamiliar human and none yawned in response to the control. By this criterion the equivalent percentage of yawning dogs here is 26% (five dogs). Six of our subjects yawned in mouth movement control trials and three of these did not yawn when presented with yawn stimuli. This to us suggests the lack of a contagion effect in our sample. Furthermore, when dogs did yawn it was unclear exactly to what they were responding. Each trial condition prompted a yawning response in two or more of our subjects and the audio-only stimuli elicited more yawn responses than any other condition. Familiarity also did not induce more yawning than unfamiliarity as was predicted. Indeed, the trend was in the opposite direction: rescue centre dogs yawned more at the unfamiliar researcher. It has been argued that familiar companions (Preston & de Waal 2002) and, in particular, those with whom one has a close emotional connection (Palagi et al. 2009) are the most likely agents to promote contagious yawning in a target. Dogs appear not to follow this rule, which could suggest that dogs lack the putative emotional connectedness impetus. Humans form close and strong emotional bonds to their dogs (Steward 1983) and dog owners might often claim a mutual, emotionally bonded relationship with their pets. However, the notion that dogs show an empathic reciprocity towards their owners has not been proven and the emotional connectivity bond is likely to be only unidirectional, from owner to dog. Thus, if contagious yawning is underscored by empathy, a lack of emotional connection on the part of the dogs could account for our result. Testing subjects in reverse, however, might yield interesting results: that is might owners yawn contagiously when exposed to their dog's yawns?
Similarity also failed to induce yawning in our dogs. Until recently, yawning contagion studies tested subjects only against conspecifics and reported contagion, but in the present study just two dogs responded only to the conspecific stimulus (10.5%) and not the control stimulus. There are limitations that we acknowledge. First, we used an unfamiliar conspecific model (as have other studies). Second, we used a video recording played back on a laptop screen, which may not be optimal for dogs. While this method has been successfully pioneered in human and nonhuman primate studies (and LCD playback has been successfully applied in other dog studies e.g. Pongrácz et al. 2003), this form of playback is potentially less ecologically relevant for dogs. Furthermore, our dogs might have been perceptually disadvantaged as, unfortunately, we did not compensate for dogs' faster eye rate flicker. Coile et al. (1989) have suggested an equivalent playback to what humans see on a laptop screen requires the flicker rate of the screen to be adjusted upwards of 70-80 Hz. One way future trials could address both similarity and familiarity issues is by using subjects from homes with more than one dog.
If yawning in dogs is not empathy related but a low level of yawning contagion is observed, what might account for it? Yoon & Tennie (2010) posited a further alternative for catching another's yawn: nonconscious mimicry, also referred to as priming by response facilitation (Byrne 1994, 2002; Hoppitt et al. 2007). The ability to imitate another's behaviour, motivated by the desire to promote affiliative interaction in social contexts (Chartrand & Bargh 1999), might explain some cases of contagious yawning (Yoon & Tennie 2010). Such social behaviour copying is well known in humans when attempting to establish a rapport with another and could be adaptive (e.g. van Baaren et al. 2004). In our study it was the rescue centre dogs that yawned most at human models, and mostly at the unfamiliar one. Nonconscious mimicry could explain this observation: when attention is directed towards rescue dogs by a newly arrived human stranger, contact-deprived dogs may recognize the potential for new social affiliation and thus respond by returning nonconscious mimicking actions. Overall, however, domesticated dogs may have a reduced response to such priming as they have become less adult-like in form, psychology and behaviour, at least relative to wild canids such as wolves, Canis lupus. An absence of contagious yawning in young children (Anderson & Meno 2003) potentially reflects the fact that they, as followers', have less of a requirement to be sensitive to synchronizing and coordinating needs within their social unit (Palagi et al. 2009). The same may be true for dogs. We suggest one way to test this hypothesis is to investigate yawning contagion in wild social-living canids with fully adult personalities, comparing adult and young animal responses to peer yawning. Furthermore, the communicative nature of the social contagion hypothesis can be tested by observing the regulating effect of yawning on synchronized group behaviour.
Contagion studies are providing productive gateways to helping illuminate social cognition and contagion is an emerging field of research (Yoon & Tennie 2010). In our contagion study we found little evidence of cross-species contagious yawning in dogs and our results left the emotional connectedness hypothesis unsupported. Furthermore, the variation we observed in responsiveness among the different groups of dogs corresponds with the findings of UdelI et al. (2010) where shelter dog performance differed from owned dog performance and lends weight to their caution against regarding the dog population as a homogeneous grouping and extrapolating inferences to all dogs after testing only owned dogs. We conclude that if dogs yawn contagiously then the releasing mechanism appears unrelated to empathy. We think it seems most unlikely that yawning contagion is limited to only those species capable of inferring the mental states of others and possessing a concept of self; and there is no compelling case to suggest a unitary underlying releasing mechanism across all vertebrates. For dogs, we have outlined alternative approaches to shedding light on this problem and we encourage researchers to shift focus on to social-living canids and other carnivores.