Auditory contagious yawning in domestic dogs (Canis familiaris): first evidence for social modulation
- First Online:
- Cite this article as:
- Silva, K., Bessa, J. & de Sousa, L. Anim Cogn (2012) 15: 721. doi:10.1007/s10071-012-0473-2
Dogs’ capacity to ‘catch’ human yawns has recently attracted the attention of researchers in the field of animal cognition. Following recent studies suggesting that contagion yawning in humans, and some other primates, is empathy-related, some authors have considered the possibility that the same mechanism may underlie contagious yawning in dogs. To date, however, no positive evidence has been found, and more parsimonious hypotheses have been put forward. The present study explored the ‘contagion-only’ hypothesis by testing whether the mere sound of a human yawn can be sufficient to elicit yawning in dogs, in a way that is unaffected by social–emotional factors. Unexpectedly, results showed an interesting interplay between contagion and social effects. Not only were dogs found to catch human yawns, but they were also found to yawn more at familiar than unfamiliar yawns. Although not allowing for conclusive inferences about the mechanisms underlying contagious yawning in dogs, this study provides first data that renders plausible empathy-based, emotionally connected, contagious yawning in these animals.
KeywordsCanis domesticusContagionEmpathySocial modulationYawning
In recent years, a renewed interest in the study of contagious yawning has been growing among researchers from various fields of investigation (see Campbell and de Waal 2010), mostly because of a theoretical link with empathy that has been supported empirically. Platek et al. (2003) showed that humans who performed better at self-recognition and theory-of-mind, two abilities that contribute to complex empathy, performed more contagious yawning. More recently, Palagi et al. (2009) found that yawn contagion in gelada baboons is best predicted by emotional closeness to the yawner—which is consistent with the observation that empathic tendencies are strongest or most likely to arise, as a function of familiarity (Preston and de Waal 2002; de Waal 2008)—and Campbell and de Waal (2011) demonstrated an ingroup–outgroup bias in contagious yawning in chimpanzees.
Apart from primate species [(humans (Provine 1986, 1989; Anderson and Meno 2003; Senju et al. 2007), chimpanzees (Anderson et al. 2004), stump-tail macaques (Paukner and Anderson 2006), and gelada baboons (Palagi et al. 2009)], contagious yawning has, to date, only been demonstrated in dogs. In 2008, Joly-Mascheroni et al. published a study in which twenty-one out of twenty-nine dogs yawned after seeing a human experimenter acting full yawning movements with vocalizations while none yawned after seeing that same experimenter displaying non-yawning mouth opening actions without vocalizations. Even though, in more recent studies, Harr et al. (2009), and also O’Hara and Reeve (2010), found very limited evidence for contagious yawning in dogs, it has been recognized that at least some individuals do yawn contagiously at humans.
Interestingly, a first attempt to test the empathy-based, emotionally connected, contagious yawning in dogs found no evidence in support of the prediction that dogs should be more likely to yawn contagiously at familiar than unfamiliar humans (see O’Hara and Reeve 2010). Instead, it was suggested that if dogs ‘catch’ human yawns, then contagion might rely on a less cognitively stringent mechanism than empathy, such as some behavioral fixed action pattern that is hard-wired and simply needs a releasing stimulus (Yoon and Tennie 2010). The present study aimed at further exploring this so-called ‘contagion-only’ hypothesis by testing whether an isolated presentation of the sound of a human yawn may be sufficient to induce yawning in domestic dogs, in an ‘encapsulated way that is unmodulated by social contextual factors’ [as proposed by Yoon and Tennie (2010)]. Interestingly, results were obtained that add more conflicting data to the debate on the possible mechanism underlying the phenomenon of dogs yawning contagiously at humans.
Twenty-nine domestic dogs of various ages and breeds participated in the present study. All dogs had resided with their owner for at least 6 months, were current on standard immunization, and were comfortable around strangers. Informed consent for participation in this study was obtained from the owners.
Four types of experimental stimuli were used in this study: (1) audio sequences of familiar yawns, (2) audio sequences of unfamiliar yawns, (3) audio sequences of familiar control sounds, and (4) audio sequences of unfamiliar control sounds. Familiar yawns and familiar control sounds were recorded from the owners of the dogs. Unfamiliar yawns and unfamiliar control sounds were recorded from one female researcher. To help induce natural (i.e. not faked) yawning during recordings, an audio loop of pre-recorded yawns was made available over headphones. All natural yawns were recorded at a 48 kHz sampling rate (16 bit quantization) and had a mean duration of 5 s. Control sounds were selected on the basis of a previous study on auditory contagious yawning in humans (Arnott et al. 2009) and consisted of reversed yawns. They were created from each recorded natural yawn by temporally reversing the segment with Audacity® 1.3.12-beta software. Thus, for any given yawn (familiar and unfamiliar), there existed a corresponding control that had virtually the same amplitude, overall temporal information, and frequency content.
Sequences used as stimuli were assembled in GarageBand™ software so that they included five sounds (e.g., five familiar yawns), each followed by 5 s of silence. Sequences used for each dog were amplitude normalized with Audacity® 1.3.12-beta software. All sequences were played at a constant volume of 70 dB.
The experimental procedure followed the general lines of that described in O’Hara and Reeve (2010). It is known that stress can induce ‘tension yawns’ in dogs, which can be distinguished from ‘natural’ yawns as they are more intense and often associated with other signs of anxiety (see Harr et al. 2009). Therefore, particular attention was paid to take all the measures against dog fatigue and stress that have been recommended in previous studies (e.g., O’Hara and Reeve 2010).
Each dog was exposed, in a random order, to the four types of experimental stimuli described above, on two experimental sessions separated by a time interval of 7 days. Two sequences of sounds (e.g., one sequence of familiar yawns and one sequence of unfamiliar control sounds), separated by a 2-min period of resting for the dogs, were presented per session. All dogs were exposed to the same sequences of unfamiliar sounds. Testing was conducted at the dogs’ homes, in rooms in which they were familiar and always in the presence of a familiar person (other than the owner from whom yawns were recorded). After a 10-min period of habituation to the researchers, the test dog was put on a leash and positioned facing two speakers plugged into a laptop. The familiar person stood quietly next to the dog, ensuring that it did not move away during testing (note that this person was given headphones with white loud noise so as not to know which stimulus was being played). A researcher sat behind the laptop and recorded, in real time, the number of yawns performed by the dog during the presentation of each audio sequence. All experimental sessions were recorded so that the number of yawns could be verified as scored accurately, by an independent observer naive to the purpose of the study (100% agreement; no ‘tension’ yawns were coded throughout the entire study). Following Campbell and de Waal (2011), the amount of attention that each dog paid to each stimulus was also coded (by both the researcher and the independent observer) in terms of time head cocking and time looking toward the speakers.
All statistical analyses were performed using STATISTICA 7.0 (StatSoft, Tulsa, Oklahoma, USA), and a significance level of 0.05 was used.
Results of the McNemar chi-square test on the number of dogs that yawned/did not yawn in response to yawning stimuli and controls
No. of dogs that yawned
No. of dogs that did not yawn
No of dogs that yawned
No of dogs that did not yawn
To test whether dogs were more likely to yawn contagiously at familiar than unfamiliar stimuli, analyses were conducted across stimuli that differed only in familiarity/unfamiliarity [as in O’Hara and Reeve (2010)]. Obtained results showed that the sounds of familiar yawns elicited significantly more yawns than the sounds of unfamiliar yawns (mean number of yawns elicited by familiar yawns: 0.45; mean number of yawns elicited by unfamiliar yawns: 0.10; Wilcoxon matched pairs test: N = 29, Z = 2.34, P = 0.02). Contrastingly, no significant differences in the number of yawn responses were observed across controls differing in familiarity/unfamiliarity.
To test whether attention effects could have accounted for the obtained differences in the number of yawns elicited by familiar and unfamiliar yawns, time head cocking and time looking at the speakers were compared between the two conditions. Results of Wilcoxon matched pairs tests showed no significant differences (mean time head coking ± S.D. recorded during the presentation of familiar yawns: 4.52 ± 5.01 s; mean time head cocking ± S.D. recorded during the presentation of unfamiliar yawns: 5.07 ± 4.61 s, N = 29, Z = 0.36, P = 0.72; mean time looking toward the speakers ± S.D. recorded during the presentation of familiar yawns: 19.07 ± 24.80 s; mean time ± S.D. recorded during the presentation of unfamiliar yawns: 15.72 ± 6.27 s; N = 29, Z = 0.41, P = 0.68). Also, there were no correlations between time head coking and number of yawns performed during the presentation of familiar and unfamiliar yawns (familiar yawns: Person’s r = 0.08, P = 0.68; unfamiliar yawns: Person’s r = 0.08, P = 0.68). Nor were correlations found between time looking at the speakers and number of yawns (familiar yawns: Person’s r = 0.15, P = 0.43; unfamiliar yawns: Person’s r = 0.15, P = 0.44).
Besides showing that the mere sound of a human yawn can elicit yawning in dogs, obtained results also highlight the effect of a social variable (familiarity/unfamiliarity) on the level of contagion, with dogs yawning more at familiar yawns than at unfamiliar ones. Interestingly, obtained data on attention seems to rule out the possibility that attention differences per se may have accounted for the observed pattern of yawning responses. It could have been that the voice of the owners attracted more attention orienting and/or engagement, which prompted dogs to process familiar yawns more deeply or effectively. Results, however, show that dogs seemed to pay similar attention to both the familiar and the unfamiliar yawn stimuli [for a similar observation in chimpanzees, see Campbell and de Waal (2011)].
What does the observed social bias suggest about the mechanism underlying the phenomenon of dogs yawning contagiously at humans? Social modulation of contagious yawning, as observed in primate species [gelada baboons (Palagi et al. 2009) and chimpanzees (Campbell and de Waal (2011)], has been interpreted as supporting the suggesting that catching another’s yawn may be an empathic response. The idea that dogs, as a domestic species living in close contact with humans, may be capable of some level of empathy toward people is not new. According to Joly-Mascheroni et al. (2008), it is conceivable that these animals have developed the capacity to represent humans’ actions and to modulate their own behavioral and autonomic responses accordingly, and that this underlies contagious yawning [see also Topál et al. (2009) and Silva and de Sousa (2011) for an extensive theoretical discussion on dogs’ potential to empathize with humans].
If empathy is the mechanism underlying contagious yawning also in dogs, then, how can we explain the conflicting results presented by O’Hara and Reeve (2010), which showed no evidence for empathy-based, emotionally connected yawning contagion in dogs? According to Campbell et al. (2009), the first challenge in discussing results across studies on contagious yawning relates to the methodological details that differ between studies and to the fact that these details have not been studied for their effect on contagious yawning. To test for social modulation of contagious yawning, O’Hara and Reeve (2010) exposed dogs to familiar and unfamiliar yawns, presented by a live model or in an LCD screen, and compared the number of elicited yawns between conditions that differed only in familiarity/unfamiliarity. It is important to note, as it contrasts with the procedure follow in the present study, that, during experimental trials, dogs tested by these authors were encouraged to focus on these visual stimuli, meaning that there was some human–dog interaction that may have diverted the animals’ attention.
But what if there is no unitary underlying releasing mechanism across all species capable of yawning contagiously and the phenomenon of dogs yawning contagiously at humans is indeed not empathy-related [as suggested by the results pointed out in O’Hara and Reeve’s study (2010)]? It is known that dogs can actively generate an internal representation of the owner’s face when they hear the owner’s voice (Adachi et al. 2007). Therefore, if one considers that it may be the perceptual image of yawn that triggers yawning (see Sarnecki 2008), then the explanation for the social bias observed here could be differences in the capacity to form mental representations from familiar and unfamiliar auditory input. That is, the perceptual image generated from hearing unfamiliar yawns would then be more obscure and less defined than that generated from hearing familiar stimuli, thus eliciting less yawning.
Clearly, there is a need for future research to focus on the key variables in the presentation of yawns, as this will facilitate comparisons across studies (see Campbell and de Waal 2010) and will allow firm conclusions to be drawn as to whether or not contagious yawning in dogs is empathy-related. If it emerges that it is, then giving the growing range of roles being played by dogs in human society (as service and therapy dogs, for example), it could turn out a useful complementary tool for selecting the most appropriate dogs (in terms of empathic processing) for specific tasks. Finally, it would also be important that future studies explore the potential effects of a vast range of individual traits (from basic life history to temperament and degree of attachment to humans) on dogs’ susceptibility to contagion. Following Sarnecki’s view (2004), a theory of yawning contagion should explain not merely why human yawns are contagious to dogs, but also why, for some individuals, they are not.
The authors are grateful to all the owners of the dogs who volunteered to participate in the present study. Also the authors want to thank Claudio Tennie for his valuable suggestions as well as three anonymous referees for their valuable comments on a previous version of this manuscript. Fundação para a Ciência e a Tecnologia funded the participation of Karine Silva (FCT-SFRH/BPD/37017/2007).