There is robust experimental evidence for
contagious yawning, yet observational studies of
naturalistic behavior have been fewer. Without
data from real-world behavior, researchers have
questioned the existence of contagious yawning
and made assumptions about some parameters
(e.g., the duration of the effect). The authors
observed contagious yawning in chimpanzees to
confirm/disconfirm its existence in the
behavioral repertoire of this species, and if
present, provide some of the missing
descriptives. They recorded yawns on an
all-occurrence basis from 18 captive-reared
chimpanzees at the Los Angeles Zoo. They
recorded identity, time, and individuals who
could have been affected. They calculated a
threshold for contagion by taking the mean and
adding 1.96 standard deviations, constructing a
response curve. Across multiple measures they
see a consistent pattern in which there is a
strong effect of contagion for 1.5 minutes, a
less strong but still significant effect lasting
up to 3.5 minutes in some measures, and no
evidence of contagion beyond 3.5 minutes. From
the time stamp on each yawn they were able to
rule out temporal synchrony as an alternative
hypothesis. Thus, contagious yawning appears to
be a natural phenomenon in chimpanzees lending
support to the myriad experimental and
observational studies to date.
Il existe de solides preuves
expérimentales que les bâillements
peuvent être contagieux, mais des
études observationnelles de ce
comportement sont peu nombreuses. Sans
données sur les comportements du monde
réel, certains chercheurs ont mis en
doute l'existence de bâillements
contagieux et ont formulé des
hypothèses sur certains paramètres
(par exemple, la durée de l'effet). Les
auteurs ont observé des bâillements
contagieux chez les chimpanzés pour
confirmer / infirmer son existence dans le
répertoire comportemental de cette
espèce et, s'il existe, fournir certaines
des descriptions manquantes. Ils ont
enregistré tous les bâillements de
18 chimpanzés élevés en
captivité au zoo de Los Angeles.
Ils ont enregistré l'identité,
l'heure et les congénères qui
auraient pu être touchés. Ils ont
calculé un seuil pour la contagion en
prenant la moyenne et en ajoutant 1,96
écart-type, en construisant une courbe de
réponse.
À travers plusieurs mesures, ils
observent, de façon constante, une
réponse à un bâillement dans
la minute et demi après
l'émetteur, un effet moins puissant mais
toujours significatif pouvant durer
jusqu'à 3,5 minutes dans certaines
mesures, et aucune preuve de contagion
au-delà de 3,5 minutes.
La mesure de ce délai, leur a permis
d'exclure la synchronisation temporelle comme
hypothèse alternative. Ainsi, le
bâillement contagieux apparaît comme
un phénomène naturel en accord
avec les études expérimentales et
observationnelles antérieures.
Contagious yawning is when a yawn by one
individual induces a yawn in another individual.
An assumption of this definition is that the
second individual (termed here: the observer)
would not have yawned except for the
stimulation, visual or auditory, by the first
individual (termed here: the trigger). Evidence
for the existence of contagious yawning has
typically come from experimental approaches: in
the most robust designs subjects are exposed to
yawns and some other stimulus for control, and
rates of yawning in each condition are
compared.
This approach has yielded consistent,
replicated evidence for contagious yawning in
humans (Homo sapiens) and chimpanzees (Pan
troglodytes). There is conflicting experimental
evidence of contagious yawning in bonobos (Pan
paniscus), with one study finding an effect
while another did not, but the larger sample
size of Tan et al. may explain why they found a
significant population-level effect while Amici
et al. did not. It is notewor- thy that Amici et
al. did find a significant effect in
chimpanzees, which they had many more of than
bonobos. Budgerigars (Melopsittacus undulates)
have shown experimental and observational
(discussed below) evidence of contagion.
Stumptail macaques (Macaca arctoides) showed
statistically significant contagion in response
to video, but it is not clear if the yawning
resulted from increased arousal as opposed to
copying the expression or affective state of the
yawner. Some studies on dogs (Canis lupus
familiaris) have found evidence of contagious
yawning while others have not, which may stem
from the different methods used. Lastly, two
species of lemur (Lemur catta and Varecia
variegata), lowland gorillas (Gorilla gorilla),
orangutans (Pongo abelii), and red-footed
tortoises (Geochelone carbonaria) did not show
evidence of contagion in experimental settings,
although the sample size issues mentioned for
the bonobo studies also pertain to the
orangutans and tortoises.
One issue is that experimental paradigms
present subjects with a supernormal stimulus:
they encounter more yawns in a shorter amount of
time than they would in real life. Anderson
raised the issue of whether contagious yawning
is an artifact of experimental designs and is
not present in the normal social behavior of the
species tested. Taking chimpanzees as an
example, despite decades of observational
studies in both the field and captivity, no one
has reported contagious yawning in their data.
Most likely this is because no one has looked
for it. As Vick & Paukner pointed out,
previous ethograms of chimpanzee facial
expressions did not include yawning. Thus, the
lack of contagious yawning in natural behavior
that concerned Anderson could stem from the
issue of supernormal stimuli that he raised, but
it could just as well stem from researchers not
measuring the behavior in the first place, as
there are no published data, positive or
negative. Whatever the cause, the criticism by
Anderson has thus far not been directly and
adequately addressed.
There is some evidence that contagious
yawning exists in the normal behavioral
repertoires of at least some species.
Observational methods have allowed researchers
to record yawns of individuals without resorting
to playback. These studies found that gelada
baboons (Theropithecus gelada), humans, bonobos,
and wolves (Canis lupus) were more likely to
catch yawns from individuals they had stronger
bonds with. The primary goal of these studies
was to examine the social relationships of
contagion as a means of studying empathy. Three
of the four studies also performed some kind of
yawn vs. control statistical comparison, which
is essential to establishing the existence of
contagion. Gelada baboons were more likely to
yawn within 5 minutes of perceiving a yawn than
other expressions, bonobos who could perceive a
yawn were more likely to yawn within 3 minutes
than bonobos who could not perceive the yawn,
and wolves yawned more within 3 minutes of
perceiving a yawn than during a 3-minute matched
control. One limitation is that the time window
of 3 minutes or 5 minutes is an arbitrary one.
There is no empirical evidence that we can find
that demonstrates that these are the exact
windows of contagion in these, or any, species.
Thus, the windows used seem to be judgement
calls rather than determined by data.
Using observational methods but a different
analysis, budgerigars showed a nonrandom
distribution of yawns such that short inter-yawn
latencies were more common than would be
expected by chance. The benefit of this method
is that it lets the data determine the rate of
yawning at different time spans post yawn. The
authors found a strong effect of yawning within
40 seconds of another yawn, after which the
response tapered off. However, it is unclear
exactly what the statistical baseline rate of
yawning was and when the yawning became
statistically indistinguishable from it.
Together, these 5 studies using observational
methods provide varying support that contagious
yawning is not merely an artifact of
experimental designs.
The two approaches have resulted in an
empirical jump from experimental evidence of the
existence of con- tagious yawning to
observational applications of contagious yawning
as a means of studying empathy. What is missing
is a rigorous assessment of the observational
evidence of contagious yawning and contagious
yawning alone, comparing the rate of contagious
yawning to a baseline, and ruling out
alternative explanations, like yawns being
synchronized over the time of day by similar
circadian patterns of activity. An exception is
the observational study of budgerigars, which
identified a window of contagion (40 s) and
ruled out circadian synchrony from their data.
The difference in window of contagion between
budgerigars (40 s) and those used for mammals (3
or 5 min) is quite large, and it could result
from the different methodologies or species
differences. Whatever the reason, the budgerigar
window of 40 s does not appear to relate well to
the mammals that have been studied. Perhaps it
is the lack of comparative data that has led
some to question the existence of contagious
yawning as a natural phenomenon in some species
or at all.
Establishing whether contagious yawning
exists in behavioral repertoires is important
because before we can apply contagious yawning
as a method to examine social relationships, we
need to know whether it is present in real-world
data. If contagious yawning is not present in
the daily lives in these species, whether
captive or wild, then the applications of
contagious yawning as a measure of empathy rests
on faulty assumptions. The social relationships
in the pattern of contagious yawning may reflect
other, unidentified variables, not empathy-based
contagion.
In this study we sought to identify whether
there is statistical evidence of contagious
yawning in observational data of a
well-established group of captive-living
chimpanzees. Our hypothesis was that if yawns
are contagious, then there should be a
significantly greater chance of a yawn occurring
after observing a yawn than the baseline rate of
yawning. Rather than selecting an arbitrary time
window for contagion, we analyzed all latencies
between a yawn and when the individual last saw
a yawn and let the data build a response curve.
Rates of yawning significantly above the
baseline rate of yawning would be evidence of
contagion.
In addition to the presence or absence of
contagious yawning, our data allow us to
describe the parameters of contagion, if it
exists. One aspect of contagious yawning that
appears different from contagion of other
expressions is the extended latency between
observing and performing a yawn. With copying
other facial expressions (e.g., smiles, frowns,
expressions of fear), the window of contagion is
0.5&endash;1.0 seconds 41. Yawning appears
different, however, with an observer yawning
some seconds, possibly even minutes, later. This
difference between copying yawns and copying
other expressions has been enough for some to
question the existence of contagious yawning.
The window of contagion for yawns is not known
for any species other than the budgerigars.
Within the mammals, we do not know how quickly
one yawn can affect another or for how long as
the observational studies used durations that
had not been empirically established. Our data
allow us to measure how soon yawns may affect
observer chimpanzees within our temporal
resolution. Our data also show how long yawns
may be contagious for, but this is only in
relation to the population-level rate. Our
method is not sensitive to and therefore cannot
supply the maximum duration possible for an
individual to yawn in response to another.
We also measured some other variables that
could affect the rate of contagion, including
the number of yawns seen by the observer, the
physical proximity to the yawning individual,
and the sex of both the observer and the
trigger. Since each yawn was time stamped, we
examined yawning by time within our
09:30&endash;12:30 observation window. This
allowed us to evaluate temporal synchrony (due
to internal circadian rhythms or external cues)
within our observation times as an alternative
hypothesis to contagion if yawns appear to be
clustered. In addition, we used the timestamps
to examine whether rates of contagion change
with time (other studies will need to fill in
times for the rest of the day). The group of
chimpanzees we studied included 5 infants.
Experimental studies have shown an absence
of contagious yawning in chimpanzees aged
1&endash;4 years old with emergence beginning at
5 years old. The limitation is that these were
experimental studies, and the artificial setting
could potentially result in evidence of
contagious yawning at an earlier or later age
than what occurs naturally. Observational data
can reveal when individuals begin to show
contagious yawning in their normal social
interactions, which could have implications for
their social-emotional development. With these
data we hope to find out whether statistical
evidence of contagious yawning can be identified
from observational data, and if so, what its
parameters are like.
Discussion
The results show evidence that yawns are
contagious in the normal daily life of one
captive group of chimpanzees. Our observations
allow us to dispel a couple of concerns about
contagious yawning. First of all, Anderson was
rightly concerned that experimental studies
presented supernormal stimuli to subjects. He
raised the question of whether contagion was an
artifact of experimental designs, or if it
functioned in the normal lives of the animals
studied, whether captive or wild. Our results
show that contagious yawning does exist in the
normal lives of this captive group of
chimpanzees, that the effect is strong (6
standard deviations above baseline during the
first minute after viewing a yawn), and that
experiments on contagious yawning have probed a
natural phenomenon, not an artifact.
Further support that yawns are contagious
comes from the chains of transmission analysis.
There would be no reason to predict that
contagion should stop at only two individuals
(one trigger and one observer), and it should be
possible for yawns to continue to be transmitted
to additional individuals. With our conservative
criteria we observed 4 instances of triadic
contagion in which yawning proceed from a
trigger to observer 1 to observer 2. With a
slightly more relaxed definition of contagion,
the longest chain moved from a trigger to 4
successive observers. We believe that
documenting chains of transmission also supports
the existence of contagious yawning as a natural
phenomenon.
A second concern is that previous
observational studies of contagious yawning did
not test for the existence of contagion
statistically, as we have done. Three studies
chose a window of contagion and found more
yawning during that window than baseline,
whereas another did not run an analysis of this
sort 33. The limitation here is that the window
of 3 minutes or 5 minutes was informed by
experimental studies, rather than being
determined by their own data or other
observational studies. Thus, there is some
arbitrariness in the choice of times. Only
Miller et al. used their data to examine a
window of contagion. The problems here are that
it is not clear when yawning statistically
deviates or returns to baseline from their
analysis, the use of one-tailed statistics, and
the window identified of only about 40 seconds
for budgerigars does not appear to be applicable
to the mammals studied.
Because of the issues with observational
data, and a lack of it, Anderson accepted the
existence of contagious yawning under
experimental conditions and questioned its
existence in natural behavior (at least for
nonhu- mans). However, Kapitány &
Nielsen went so far as to question whether
contagious yawning exists at all, for any
species, including humans. With our results we
can dispel the concern of Kapitány &
Nielsen as well. We provide strong statistical
support for the existence of contagious yawning
in chimpanzees through observational data, which
augments the multiple experiments also finding
strong statistical support for contagious
yawning in this species. In addition, our
results serve as support for the observational
studies that did not test for contagion
directly, as we have. The window we identified
is similar to ones that have been used
previously, and having identified contagious
yawning in the behavioral repertoire of two
species (chimpanzees in the current study and
budgerigars), we expect that similar response
curves would be found in the data of others,
though the exact numbers may vary by
species.
It would also be a mistake to question the
findings of the studies that used 5 minutes or 3
minutes as a window of contagious yawning when
our evidence extends to 3.5 minutes. For one
thing, our results describe only this group
(housed at the Los Angeles Zoo and Botanical
Gardens) of this species (Pan troglodytes). Our
exact numbers may not be indicative of the
species as a whole. The group we studied is
large by captive standards (N = 18) and diverse
(8 adult females, 5 adult males, 5 infants),
especially for zoos, but 18 is not enough to
establish a species-wide rate of response. In
small samples, one or two high functioning or
low functioning individuals can skew the precise
numbers. The overall pattern should be
consistent (yawns are contagious for some time
after viewing a yawn), but the exact time window
measured might vary between groups. Future
testing will be needed to show how
representative our parameters of contagion are
for chimpanzees as a whole, as well as if they
vary at other times of the day.
In addition, there is no reason to expect
the parameters we measured to be the same for
all species. Different species could have longer
or shorter windows of contagion based on their
evolution. The other observational studies were
on geladas, humans, bonobos, and wolves, and
each species may have its own parameters of
contagion. We would have to measure them to find
out. Lastly, and perhaps most importantly from a
statistical perspective, these studies were
looking at social patterns of contagion. The
expectation was that contagious yawning, if
based on empathy mechanisms, should be biased
toward closer social affiliates than distant
ones.
Thus, contagious yawning should show a
social pattern, but noncontagious yawning should
not. Noncontagious yawning should be randomly
distributed in the group. Using a window of
contagion that is too long would be adding
noncontagious yawns (i.e., noise) to the
experimental condition, and using one that is
too short would be adding some contagious yawns
(i.e., signal) to the control condition. If the
experimenters still find a significant
difference, that implies that the signal is so
strong that it can be identified statistically
even with extra noise or reduced signal in the
input. The implication is that if 5 minutes is
too long or 3 minutes is too short and a signal
is found, the signal should be even stronger for
the optimum window. The problem in including too
long a window of contagion is the possibility
that the noise (noncontagious yawns) masks the
signal (contagious yawns) leading to a null
result. With too short of a window there would
be a large amount of signal (contagious yawns)
in both the contagion condition and the
baseline, leading to no significant difference
between them. Thus, the problem of using too
long or too short of a window of contagion is
the risk of a type 2 error (false negative), not
a type 1 error (false positive). In our view,
our results support the previous observational
studies by showing that contagious yawning is
present in observational data and do not
undermine them by identifying a slightly
different and more precise window of
contagion.
Unlike previous studies, our results allow
us to describe the natural parameters of
contagious yawning. We applied numerous
different approaches to operationally defining
contagious yawning, and the results are remark-
ably consistent. Using a conservative subset of
our data, we found that yawning was contagious
for 3.5 minutes after observing a yawn. Applying
different criteria changed the outcome only
subtly, and a larger scale pat- tern emerged
with a very strong effect of contagion up until
1.5 min, a less strong but still above threshold
effect of contagion after 1.5 min and lasting as
long as 3.5 min, and then yawning falling below
threshold and staying there. Different measures
disagree exactly where the contagion effect
falls below threshold (e.g., 1.5 min, 3 min, or
3.5 min), but none extend the window of
contagion beyond 3.5 min after viewing a yawn.
Since our primary measure of contagion, which we
established with a conservative subset of the
total yawns, yielded a window of contagion up
until 3.5 min, and since this number was
supported by other measures, we feel confident
in accepting 3.5 min as a duration for which
yawns are measurably contagious in this group of
chimpanzees. Nonetheless, other researchers may
disagree and prefer a more stringent window. The
window from using only yawns from
10:00&endash;12:00 to avoid feeding times is
tighter, with strong evidence only for the 1.5
minutes post yawn exposure. We have provided our
data across multiple criteria for contagious
yawning so that readers can come to their own
conclusions, based on the evidence we gathered,
as to what constitutes the most appropriate
window of measuring yawn contagion.
One note, it would be the wrong
interpretation to conclude that all yawns within
3.5 minutes of viewing a yawn are contagious,
and that no yawns after 3.5 minutes are
contagious. We cannot know for certain,
currently, whether any given yawn is contagious
or non-contagious. There are no behavioral cues
that have been identified to distinguish them,
and the only way to determine this conclusively
would be through differing neural signatures
between yawns originating internally
(noncontagious) or through visual or auditory
stimulation (contagious). These neural
signatures have not yet been identified, and, of
course, they would not be helpful outside of
laboratory studies. Instead, the window shows
the duration of time for which reliable
assumptions of pooled data can be made. As done
already, the window of the time is the one in
which to look for social biases in contagion.
The window says that a high number of these
yawns will be contagious (signal), and that
after 3.5 minutes many/ most of the yawns will
be noncontagious (noise). Nothing says that
yawns within 3.5 minutes of view another must be
contagious. Only that if we are going to look
for social patterns, this window provides us
with the greatest signal-to-noise contrast with
which to do so. In addition, our data do not say
that it is impossible for yawns to be contagious
beyond 3.5 minutes in chimpanzees. That is also
the wrong interpretation. Currently, there is no
known difference in the motor pattern between
contagious and noncontagious yawns, and one may
not exist. Without an observable difference,
only a neural method could conceivably label
each yawn as contagious or non-contagious, so
only that method could establish the absolute
maximum duration for which yawns are contagious
in any individual or species.
An alternative hypothesis to yawns being
contagious is that they are temporally
synchronized, perhaps due to circadian rhythms
or regular, daily events. Yawning varies over
the time of day. Theoretically, a tight enough
peak could produce a pattern that looks
statistically like contagion but arises from
yawns being temporally synchronized, rather than
contagious. However, yawning was not tightly
grouped enough in our observations to account
for the greater chance of observing a yawn after
another yawn. The window we measured was only
3.5 minutes long, at its longest. That would
require the temporal pattern of yawning to have
a similar peak lasting no more than a few
minutes. Yawning does appear to occur at higher
rates at the very beginning and end of our
observations, which differs from Vick &
Paukner. The increased rates we observed
correspond with feedings (just after morning
feed and just before lunch), so it is possible
that they are displacement behaviors related to
tension or anticipation of meal times. Increased
yawning around feeding times was also observed
in lions (Panthera leo) and mandrills
(Mandrillus). However, the times close to the
chimpanzees' feedings (9:30&endash;10:00 and
12:00&endash;12:30) do not correspond
exclusively with the time stamps of the
contagious yawns we observed. Yawns meeting our
criteria for contagion appear throughout the
observation times, even with a rise toward the
end. Furthermore, if we remove these two
half-hour ranges and only analyze yawns
occurring between 10:00&endash;12:00, we find
similar evidence of contagion with a strong
effect for 1.5 minutes after viewing a yawn,
responses bordering threshold until 3.5 minutes,
and then dropping well below threshold
thereafter. Thus, patterns of temporal synchrony
cannot explain our data.
The only social data we have in this
analysis are the sex and early developmental age
of the yawner. Males yawned significantly more
than females, which was not recorded in
chimpanzees previously. These differences may
reflect how a few individuals can influence
population-level statistics in the group sizes
available in captivity. Larger sample sizes,
perhaps from wild groups or sanctuaries, may
clarify sex differences in yawning in
chimpanzees. Using the higher baseline rate of
yawning by males as the expected values, males
did not deviate from this rate in their
propensity for contagion. In an observational
study, female humans showed higher
susceptibility to yawn contagion, and the
authors linked this difference to females being
more empathetic than males in general. We did
not make this same observation in chimpanzees as
both sexes showed rates of contagion similar to
their baseline rate of yawning. Looking at the
triggering individual, male chimpanzees and
female bonobos were more likely to serve as
triggers than the opposite sex. However, we did
not observe this result, either, as neither sex
was more likely to serve as a trigger than
expected based on the baseline rate of yawning.
Thus, our data do not support either of the
previously identified sex differences in
contagious yawning. Two of the studies were on
different species, so it is possible a female
bias in contagion 49 and triggering 34 exists in
those species, but not in chimpanzees. As for
Massen et al., that study was experimental, and
ours was observational. One possibility is that
the supernormal exposure of video playback 30
draws out a difference that does not exist in
natural behavior. Conversely, the difference may
have been too subtle for us to measure with our
sample size. The latter seems unlikely because
we did not even observe a trend in the direction
of males being triggers more often than females.
Our observed values match our expected values
very closely. Another possibility is that there
are differences between the groups of
chimpanzees; each study may accurately describe
an effect for the group observed, which is not
unreasonable when thinking about the diversity
of chimpanzee behavior and the relatively small
sample sizes of captive groups. Lastly, our
study could have a false negative, or Massen et
al. could have a false positive. Only
replication will solve the discrepancy in these
observed effects.
Regarding development, we have too few cases
to analyze them statistically, so we need to be
cautious with our discussion. We can say that we
did record potential contagious yawns from
individuals slightly younger than has been seen
in experimental studies. Anderson et al. used
video playback with infants (3 years old)
accompanying their parents, and the infants did
not yawn at all. The authors did not provide
data on attention, so we do not know to what
extent the infants perceived the yawns. It is
possible the infants were distracted by the
presence of their mother, or perhaps the
artificiality of video playback does not
stimulate contagious yawning at this young age.
Madsen et al. used live yawn demonstrations from
humans (familiar and unfamiliar) and did not see
evidence of contagious yawning under the age of
5 years old. As with Anderson et al., perhaps
the artificiality of experimentation fails to
elicit behaviors that are developmentally in
transition. It is also possible that
empathy-based contagion expresses itself sooner
for members of the social group, which we can
assume would be socially closer to the
chimpanzees than members of another species. One
last possibility that occurs to us, Madsen et
al. studied orphaned chimpanzees, and
differences in emotional development have been
found between mother-reared and orphaned
bonobos. It is interesting that of the potential
contagious yawns by infants that we recorded,
none of the triggers were the infant's mother.
Because our numbers are so small, we cannot
conclude that contagious yawning is in fact
present at younger ages than has been recorded
previously, but our data do suggest that
observational methods may reveal different
developmental patterns than
experimentation.
In conclusion, we found strong evidence that
contagious yawning is a natural phenomenon in a
captive group of chimpanzees. We can dispel with
the concerns about the very existence of
contagious yawning and accept previous
experimental and observational studies as
measuring aspects of behaviors that exist in the
normal rep- ertoires of the species studied. Our
methods can be applied widely, and we hope
researchers will build response curves for
different species. It would be very interesting
to know whether the parameters of contagion
differ between groups of the same species and
across species. Combining the comparative method
and observational studies of contagious yawning
could tell us about the evolution of contagious
expressions, empathy as a possible mechanism,
and the socioecological factors that may have
shaped this behavior.
