Contagious yawning has been suggested to be
a potential signal of empathy in non-human
animals. However, few studies have been able to
robustly test this claim. Here, we ran a
Bayesian multilevel reanalysis of six studies of
contagious yawning in dogs. This provided robust
support for claims that contagious yawning is
present in dogs, but found no evidence that dogs
display either a familiarity or gender bias in
contagious yawning, two predictions made by the
contagious yawning&endash;empathy hypothesis.
Furthermore, in an experiment testing the
prosociality bias, a novel prediction of the
contagious yawning&endash;empathy hypothesis,
dogs did not yawn more in response to a
prosocial demonstrator than to an antisocial
demonstrator. As such, these strands of evidence
suggest that contagious yawning, although
present in dogs, is not mediated by empathetic
mechanisms. This calls into question claims that
contagious yawning is a signal of empathy in
mammals.
Résumé
Le bâillement contagieux a
été suggéré comme un
signe potentiel d'empathie chez les animaux non
humains. Cependant, peu d'études ont
été en mesure de tester de
manière robuste cette affirmation. Les
auteurs ont effectué une réanalyse
bayésienne à plusieurs niveaux de
six études sur le bâillement
contagieux chez le chien. Cela a fourni un
solide argument aux affirmations selon
lesquelles le bâillement contagieux est
présent chez les chiens, mais n'a
trouvé aucune preuve que les chiens
affichent une familiarité ou un parti
préférence liée au sexe
concernant l'hypothèse du
bâillement contagieux et de l'empathie. De
plus, dans une expérience testant le
biais de prosocialité, une nouvelle
prédiction de l'hypothèse
contagieuse du bâillement et de
l'empathie, les chiens n'ont pas
bâillé plus en réponse
à un manifestant d'un
congénère familier qu'à un
étranger. Ceci suggère que le
bâillement contagieux, bien que
présent chez les chiens, n'est pas
médié par des mécanismes
empathiques. Cela remet en cause les
affirmations selon lesquelles le
bâillement contagieux est un signe
d'empathie chez les mammifères.
1. Introduction
Empathy, broadly defined as emotional and
mental sensitivity to another's state
[1], appears to play a key role in
humans' prosocial and cooperative behaviour
[2&endash;4]. While there are many
aspects of empathy, all empathetic responses are
ultimately predicated upon an observer having
some form of access to another individual's
emotional state [1].
Perception&endash;action mechanism accounts
argue that this access is granted by
state-matching: when an observer perceives a
particular state in another individual, neural
representations of that state are activated in
the observer, resulting in the experience of a
similar state [1,5,6]. It has been
suggested that this perception&endash;action
mechanism evolved early in mammals
[1,5,6] and underpins the helping
behaviour seen in many non-human mammals
[7&endash;14]. However, it is difficult
to rule out simpler explanations for helping
behaviour that do not require empathy, such as
associative learning [12,15] or
alternative motivations such as a desire to seek
social contact [16&endash;18] or explore
novel objects [19].
Contagious yawning has emerged as a
potentially powerful tool to help resolve this
impasse. Contagious yawning occurs when
observing another individual yawn induces
yawning in the observer [20]. The
contagious yawning&endash;empathy hypothesis
posits that both contagious yawning and empathy
are mediated by the same
perception&endash;action mechanism
[1,6,21&endash;23]. This hypothesis
makes several testable predictions about the
propensity of individuals to engage in
contagious yawning. First, individuals who
report lower levels of empathy should be less
likely to engage in contagious yawning. Second,
as people show greater empathy for kin and
friends, they should be more likely to yawn when
exposed to yawning from familiar yawners
compared to unfamiliar yawners [21,24].
Finally, it has been hypothesized that, due to
more direct involvement in offspring care,
female mammals may experience a greater level of
empathy than males [25&endash;28] and
thus it has been predicted that female mammals
should be more likely to engage in contagious
yawning [29].
Evidence of each of these predictions has
been found in humans. Firstly, subpopulations
reporting lower levels of empathy, such as
children with autism spectrum disorder (ASD)
[30] or adults who score highly in
psychopathic traits [31], are less
likely to engage in contagious yawning.
Secondly, both empathy and contagious yawning
appear to share a familiarity bias: people
experience more empathy for friends and family
than strangers and are more likely to
contagiously yawn when familiar people yawn
[21,24] (but see [32]). Finally,
although both men and women are equally likely
to engage in contagious yawning, women who do
engage in contagious yawning yawn more
frequently than men who engage in contagious
yawning [29]. While this provides some
evidence of a gender bias in human contagious
yawning, there is currently much debate
surrounding this finding [33,34].
Based on this evidence from humans, it has
been suggested that the widespread presence of
contagious yawning across mammals demonstrates
that the perception&endash;action mechanism
underpinning empathy is phylogenetically ancient
and thus that the helping behaviours seen in
mammals are driven by at least a rudimentary
form of empathy [1,35,36]. This
conclusion rests on two key premises. First, it
assumes that contagious yawning is indeed
widespread across mammals. Second, it assumes
that contagious yawning in animals is
underpinned by the same perception&endash;action
mechanism mediating empathy. However, there is
currently a lack of robust evidence for either
of these assumptions [37].
Firstly, there is a lack of conclusive
evidence that contagious yawning is widespread
across mammals. A comprehensive review of the
contagious yawning literature by Massen &
Gallup [37] demonstrates that the
majority of studies on contagious yawning are
focused on non-human primates, particularly
chimpanzees, with few studies looking at
contagious yawning in non-primates. Furthermore,
research into contagious yawning in other
species has mostly been restricted to a single
study. Chimpanzees are the only non-human
species of mammal to have consistently shown
contagious yawning across multiple studies
[38&endash;45]. There is no evidence of
contagious yawning in gorillas [44,46]
and for both bonobos [24,36,44,47,48]
and dogs [35,49&endash;54], the evidence
is mixed: some studies have shown evidence for
contagious yawning while other studies have
found no effect. As well as making it difficult
to draw conclusions regarding the presence of
contagious yawning in these species, this lack
of consistency in replication brings into
question claims for contagious yawning in other
species that are based on a single study
[24,35,36,44,47,49&endash;64]. Given the
small sample sizes used in many of these studies
[65], their findings may simply reflect
either false positives or negatives rather than
the true presence or absence of contagious
yawning. Such issues are particularly compounded
by the variation in study design across
different species [66]. In particular,
relying on observational rather than
experimental methodologies can be problematic
due to the influence of synchronized circadian
rhythms making it more likely that animals in a
group may spontaneously yawn at the same time
[37]. As such, it is difficult to say
with certainty that there is evidence for
contagious yawning in non-human mammals outside
of chimpanzees, let alone discern the
phylogenetic pattern of contagious yawning
across all mammals.
Secondly, contagious yawning in non-human
animals may not be mediated by empathetic
mechanisms.
Contagious yawning in animals may be the
result of stress [54,57], an affiliation
strategy [67], a means of communication
[61], or a mechanism to improve
collective vigilance within groups
[37,68,69] rather than being related to
empathy via a perception&endash;action
mechanism. A powerful way to test between these
hypotheses is to examine if the patterns of
behaviours predicted by the contagious
yawning&endash;empathy hypothesis, such as the
familiarity bias and the gender bias, are also
seen in animals. Such biases can be thought of
as cognitive signatures; a particular suite of
behaviours that should be seen if contagious
yawning is mediated by a
perception&endash;action mechanism but not if it
is mediated by another cognitive mechanism
[70]. Evidence for the familiarity bias
has been found in studies on chimpanzees
[40,41], dogs [50,64], bonobos
[36], gelada baboons [58], and
wolves [64], but other studies on
chimpanzees [42,43] and dogs
[51,53] have not found evidence for this
bias, and one study on rats has even found the
opposite pattern [61]. In terms of the
gender bias, there is currently no consistent
support for the prediction that females are more
likely to contagiously yawn than males across
mammals [37]. Instead of females being
more likely to engage in contagious yawning,
there appears to be some evidence of an
interaction between the gender of the observed
yawner and the subject, but this pattern is not
consistent. Bonobos yawn more when an individual
of the opposite gender yawns and yawn more in
general when the observed yawner is female
[36]. Female geladas are also more
likely to engage in contagious yawning than
males but only when the observed yawner is
female [58]. By contrast, in
chimpanzees, male yawning is more contagious for
other males than female yawning is for other
females [42], while there is no evidence
of any gender bias in dogs [35]. As
such, there is currently no conclusive evidence
for either signature across mammals, which
brings into question whether contagious yawning
is mediated by a perception&endash;action
mechanism shared with empathetic processes.
Here, we tested the contagious
yawning&endash;empathy hypothesis in dogs by
reanalysing past data and also employing a novel
experimental paradigm. First, we established
that there is robust evidence for contagious
yawning outside of chimpanzees and humans by
conducting a Bayesian multilevel reanalysis of a
combined dataset from six studies of contagious
yawning in dogs. We then examined our combined
dataset for evidence of the familiarity bias and
the gender bias. Finally, we ran a study to
search for a novel signature predicted by the
contagious yawning&endash;empathy hypothesis:
the prosociality bias. In humans, empathy is
modulated by social experience: people
experience greater empathy for people who
interact with them in a fair or prosocial manner
[71,72]. Similarly, dogs show a
preference for those who interact with them in a
positive manner rather than a negative manner
[73]. Therefore, we carried out an
experiment to test whether dogs show a
prosociality bias (i.e. yawn more contagiously
when in the presence of a human who has been
nice to them rather than one who has ignored
them). We predicted that, if contagious yawning
is mediated by empathic processes in dogs, we
would find evidence for the familiarity, gender,
and prosociality biases in our study.
4. Discussion
By combining the data from six different
studies, the resulting dataset is the largest
used to date to examine the presence of
contagious yawning in a non-human mammal. This
allowed us to draw conclusions about the
presence and absence of contagious yawning and
the signatures predicted by the contagious
yawning&endash;empathy hypothesis with a greater
level of certainty than by relying on individual
studies alone. Our reanalysis shows that dogs do
exhibit contagious yawning, showing higher
probabilities and rates of yawning for yawning
demonstrators compared to control demonstrators.
This provides robust support for the claims that
contagious yawning is present in dogs
[35,49&endash;51]. In order to test
whether this contagious yawning is related to
mechanisms underpinning empathy, we examined
this dataset for evidence of the familiarity
bias and gender bias. However, dogs in our
reanalysis showed no evidence of either of these
biases. Similarly, when we ran a novel
experiment to look for a prosociality bias, we
found that the dogs in our experiment were no
more likely to yawn for prosocial demonstrators
than antisocial demonstrators. Dogs, therefore,
show no evidence for any of the familiarity,
gender, or prosociality biases predicted by the
contagious yawning&endash;empathy hypothesis.
This suggests that contagious yawning in dogs is
not mediated by an empathy-related
perception&endash;action mechanism
[52&endash;54]. The presence of
contagious yawning in non-human animals,
therefore, cannot be assumed to be evidence for
a perception&endash;action mechanism shared
between humans and other mammals, as has been
previously proposed [1,35,41,58]. That
is not to say that some non-human animals do not
necessarily experience some form of empathy but
that contagious yawning cannot be taken as a
diagnostic signal for the presence of these
empathetic processes. Furthermore, these
results, alongside the arguments put forward by
Massen & Gallup in their recent review
[37], bring into question the validity
of the contagious yawning&endash;empathy
hypothesis more broadly.
It is important to acknowledge several
caveats to our conclusions. Firstly, in both our
reanalysis and experiment, the subjects were
primarily responding to interspecific yawns from
human demonstrators. While it is possible that
dogs would respond differently to conspecific
and interspecific yawning, there are several
reasons to believe that this is not the case.
Research in other species such as chimpanzees
suggests that they respond similarly to
conspecific and interspecific yawns
[41], and, in our reanalysis,
controlling for demonstrator type did not
improve model fit. Nevertheless, more rigorous
comparisons between how dogs respond to
conspecific and interspecific yawning would be a
useful future line of research. Secondly, it is
important to note that the familiarity, gender,
and prosociality biases are indirect measures of
empathy [37]. As such, care needs to be
taken in interpreting these biases and there
remains substantial debate over how to do so.
For example, it has been argued that both the
tendency for children with ASD to be less prone
to contagious yawning [83] and the
familiarity bias [37,84,85] can be
explained in terms of differences in attending
to yawners rather than differences in empathetic
response. Similarly, the gender bias reported in
humans [29] is not straightforward to
interpret and there is debate over whether it
simply reflects a false positive in the
literature [33,34]. By contrast,
proponents of the contagious
yawning&endash;empathy hypothesis argue that the
familiarity bias continues to be found even when
controlling for differences in subjects'
attention [40,41] and that the negative
results for the gender bias in previous studies
reflects methodological issues with prior
experiments [34]. Furthermore, although
alternative hypotheses such as the attentional
hypothesis could explain the presence of a
single bias such as the familiarity bias, only
the contagious yawning&endash;empathy hypothesis
predicts the presence of all three biases. As
such, testing for all three biases represents a
powerful test of the contagious
yawning&endash;empathy hypothesis. Finally,
searching for a novel signature, the
prosociality bias, required a novel experimental
methodology where dogs were exposed to a
prosocial experimenter that interacted with them
and an antisocial experimenter that ignored
them. Previous work which used a similar
methodology demonstrated that dogs do show a
preference for the prosocial demonstrator
[73], and so if the contagious
yawning&endash;empathy hypothesis is correct,
dogs should have reacted with increased yawning
to the prosocial demonstrator. However, further
work would be useful in confirming the presence
or absence of the prosociality bias in dogs and
other species such as humans.
Research into contagious yawning has been
dominated by the contagious
yawning&endash;empathy debate [37].
However, contagious yawning is an interesting
phenomenon in its own right as its evolutionary
roots and ultimate function remain a mystery
[20]. Contagious yawning in animals may
be the result of stress [54,57], an
affiliation strategy [67], a means of
communication [61], or a mechanism to
improve collective vigilance within groups
[37,68,69] rather than being related to
empathy via a perception&endash;action
mechanism. Future research into contagious
yawning should include a greater focus on
testing between these and other hypotheses. For
example, the affiliation hypothesis might
predict that contagious yawning should be seen
more frequently during reconciliation periods
after conflict while the collective vigilance
hypothesis posits that contagious yawning should
increase in response to external disturbances
[37,86]. However, it is important to
note that these theories are not necessarily
mutually exclusive [87] and that factors
such as stress appear to influence yawning
propensity in complex ways [88,89].
Additionally, an important next step is to
consider evidence of contagious yawning outside
of mammals. While there has been some work
looking at contagious yawning in budgerigars
[86,90] and tortoises [91],
research has otherwise been sparse outside of
the mammalian class.
Future research would benefit from
systematically testing contagious yawning across
multiple species. One barrier to such projects
is that studying a range of different species
often requires different experimental set-ups to
make such testing feasible. There is a concern
that such a range of methodological approaches
may make cross-species and cross-study
comparisons difficult, if not impossible
[35,66]. However, our finding that the
effect of treatment on yawning probabilities and
rates remains stable when controlling for
various aspects of study design suggests that
the presence of contagious yawning is relatively
robust to differences in experimental design. As
such, while it is important to use broadly
similar designs (for instance, comparing
animals' yawning rates when exposed to either a
yawning demonstrator or control demonstrator),
there could be considerable flexibility in other
aspects of study design. For example, our
results suggest that animals' yawning
probabilities and rates to either live
demonstrators or recorded demonstrators are
comparable. Therefore, our findings suggest that
more ambitious cross-species work can be carried
out with confidence in the validity of the
subsequent comparisons.
To conclude, our results provide
robust support for the hypothesis that
contagious yawning is found in dogs, the first
non-human species of mammal where it has been
clearly shown outside of chimpanzees. However,
we found no evidence that dogs yawn more in
response to either familiar human yawners
compared to unfamiliar human yawners, or to
prosocial human yawners compared to antisocial
human yawners. Additionally, we found no
evidence that female dogs were more likely to
yawn in response to a yawning demonstrator than
male dogs. As such, these findings cast doubt on
the widespread assertion that contagious yawning
is mediated by the same perception&endash;action
mechanism as empathy [1,6,35,41,58].
Instead, they support recent claims that there
is no link between contagious yawning and
empathetic processes [37,67] and
underline the importance of developing more
direct measures of empathy in non-human animals
[37,92]. However, while our results
suggest that researchers cannot rely on
contagious yawning as a diagnostic signal of
empathy, our additional findings that the effect
of contagious yawning appears to be robust to
variations in experimental methods suggest that
cross-species comparisons may be a powerful way
to disentangle the evolutionary roots of this
behaviour.
