Yawn Contagion
Rohan Kapitany,
University of Queensland, Australia
[email protected]
Abbreviated Introduction and summary of goals:
The three goals of this investigation are: Exactly how contagious are yawns?
For how long is it appropriate to consider a spontaneous yawn a ‘trigger’? And how
often do we miscategorise a spontaneous yawn (which necessarily must occur at a
base frequency) as a contagious yawn? I first present an Agent-Based Model in
depth, then provide data the model produces based on informed (but hypothetical)
input. Second, I present a behavioural study closely matched to the model’s design
in order to ground both the input and output. Finally, I present results, by matching
real yawning data to the agent-based model of yawning in order to address (and
resolve) empirical and theoretical questions associated with the topic of contagious
yawning.
THIS IS NOT THE FINAL FORM OF WRITTEN INFORMATION. IT IS PROVIDED HERE TO
CLEARLY DESCRIBE THE METHODS AND PROCEDURES.
IT SHOULD BE CONSIDERED A DRAFT COPY. WHILE THE FORM AND EXPRESSION MAY
CHANGE BETWEEN REGISTRATION AND PUBLICATION, THE INTEGRITY OF THE
METHODS AND PROCEDURES WILL NOT BE ALTERED.
Methods: The Model
We used NetLogo to create our computer model (Wilensky, 1999). NetLogo is
an Agent-Based Modelling (ABM) language which allows authors to program virtual
environments populated by agents who follow specific rules. The following overview
is in accordance with the ‘Standard Protocol’ for report ABMs, as described by
Grimm et al. (2006)
Overview of the Model
Purpose of the Model
The purpose of the model was to create a simulation in which agents yawn,
where those yawns could cause other agents to yawn. Importantly, this model
specifies a base-rate likelihood that an agent will yawn, allowing us to quantify what
proportion of ‘contagious yawns’ were actually spontaneous yawns that only appear
to be ‘contagious’. We term these ‘incidental yawns’ – yawns which are actually
spontaneous, but by virtue of the timing to ‘trigger yawns’ are miscategorised.
Additionally, this model allows us to quantify the ‘contagion factor’ of a yawn (i.e.,
how many more times an agent is to yawn in response to observing a trigger).
Variables of the Model
A global variable is a variable applicable or accessible to all agents in the
model. This model has 4 global variables, each described in Table Z.
In addition to global variables, each agent ‘owns’ one idiosyncratic variable
called ‘latent-yawn’. This variable is based on the ‘Yawn-Latency’ variable. After an
agent yawns it sets it’s ‘latent-yawn’ variable to the value of the ‘yawn-latency’. Each
tick of the model this value decreased until it reach 0, at which point it ceased to
influence other agents who could ‘see’ it.
Table Z. A description of global variables used in the agent-based model of
contagious yawns.
Variable
Description
Base-Rate
This value specified how likely (%) an agent was to yawn on
any given tick (or ‘minute’) of the model (e.g., 1.66% chance
to yawn per tick = 1 yawn per hour).*
Contagion-Factor
This value determines to what extent the Base-Rate is
modified when an agent ‘observes’ a yawn (e.g., a
contagion-factor 2.00 would increase the base-rate to
3.32).*
Sees-how-many
This value determines how many other [immediately
adjacent] agents each agent can ‘see’. This value can be set
from 0 to 28. Our behavioural experiment will focus on 8
immediately-neighbouring others.
Yawn-latency
This value determines how long a ‘spontaneous’ yawn can
influence those who observed it. It corresponds
theoretically to ‘minutes’.*
*In section 3 we will attempt to fit these value based on our behavioural data.
Process overview and Scheduling
This is an exceptionally simple Agent-Based Model. The environment itself
does not influence the agents, and as such, there are no variables to describe. Time
proceeds in ‘ticks’ (analogous to minutes). The order of operations in which the
agents act within the world is also simple:
1. Any agent who yawned on the previous tick (red agents) turns blue. The
patch on which this agent sits turns black.
2. Agents reduce their ‘latent-yawn’ value by 1.
3. Agents calculate whether they yawn on this tick, and if appropriate, they
yawn (and set their ‘latent-yawn’ value correspondingly).
Design Concepts
The model described thus far calculates the likelihood of a given agent
yawning in any given ‘tick’. The model asks each of the 1024 agents in random order
to calculate this value in turn. To calculate a yawn each agent generates a randomfloating number between 0 and 99.999… (e.g., 65.2432 or .0396). If this randomly
generated number is less than the base-rate (e.g., 1.666), then the agent
spontaneously yawns. Agents who had neighbours yawn before their opportunity to
calculate their own yawn have their base-rate probability influenced by the
contagion-factor. If the contagion-factor is set to 1.75 (for example), then the baserate for that agent becomes 2.9155. If the random number is less than this
[increased] value the agent yawns, but the yawn is only considered a ‘contagious
yawn’ if the value is greater than the base-rate and less than the base-rate
multiplied by the contagion-factor (e.g., 2.0132). Contagious yawns are represented
by a yellow patch. If the random number is less than the base-rate (i.e., falling below
the threshold of the unaltered base-rate; e.g., 1.4448) the agent yawns, but it is
considered an ‘incidental yawn’ (and is represented by a white patch). Incidental
yawns are yawns that would have occurred independent of the observation of a
neighbours’ yawn. A latency of x ticks means agents are influenced by the
observation of a yawning neighbour for a duration of ticks equal to this value. See
Figure X.
Figure X. Example of visual output of the Yawn Contagion Model.
Note: Blue agents are agents who are not yawning. Red agents are yawning (1).
Agents surrounded by yellow are agents who are yawning contagiously as a result of
a yawn on this tick (2) or on a previous tick (4). Agents with a white background are
agents who are spontaneously yawning in proximity to a preceding yawn on this tick
(2) or a previous tick (3) - these would ordinarily be counted as contagious yawns,
when they are, in fact, ‘incidental’. This world wraps horizontally and vertically, such
that agents on the left threshold are functionally adjacent to agents on the right, and
the agents on the top are adjacent to agents on the bottom.
Model Data
In order to demonstrate how the model works, the data it produces, and the
potential extent of the operational problems in the yawning literature, we generated
some data based on hypothetical input values.
We assumed all agents were able to ‘see’ 8 other agents (as per the proposed
behavioural experiment; see ‘Methods: The Behavioral Experiment’). The simulation
was run assuming yawns were contagious, and again assuming they were not. As
noted by Demuru and Palagi (2012) and Norscia & Palagi (2011), it is standard to
count any yawn as ‘contagious’ if it occurs within 5-minutes of a ‘trigger’ yawn. We
ran our model using 3 yawn-latencies (1-minute, 3-minutes, and 5-minutes) over a
total of 120 simulated minutes. Latency corresponds both to the amount of time an
agent will actually be susceptible to a contagious yawn, and the time in which it may
be classified as contagious/incidental. The simulation was run 25 times for each set
of values. Results are described in Table Xa and Xb. Naturalistic research has shown
that, on average, people yawn up to once an hour (Zilli et al., 2007) but experimental
rates vary considerably (Baenninger & Greco, 1991; Provine, 1986)
Table Xa. Hypothetical Data: 120-simulated minutes, with no contagion, and a Base-Rate =
1.666% (simulated 25 times).
Mean
Incidental
Mean
Contagious
Mean
Spontaneous
Mean Total
Percent
Incidental
Percent
Spontaneous
Latency 1
2.4 (1.47)
0
14.48 (3.54)
16.88 (3.87)
14.22%
85.78%
Latency 3
4.96 (2.72)
0
12.84 (3.52)
17.8 (4.48)
27.87%
72.13%
Latency 5
8.88 (2.12)
0
9.44 (4.78)
18.32 (4.51)
48.47%
51.53%
Table Xb. Hypothetical Data: 120-simulated minutes, with contagion of 1.5x, and a Base-Rate =
1.666% (simulated 25 times).
Mean
Incidental
Mean
Contagious
Mean
Spontaneous
Mean Total
Percent
Percent
Incidental Spontaneous
Latency 1
3.52 (1.81)
9.6 (2.68)
15.32 (2.44)
28.44 (4.43)
12.38%
53.87%
33.76%
Latency 3
7.16 (2.36)
9.48 (2.62)
9.36 (3.01)
26.00 (5.35)
27.54%
36.00%
36.46%
Latency 5
11.96 (3.83)
9.36 (2.55)
6.4 (2.16)
27.72 (5.10)
43.15%
23.01%
33.77%
If we assume the extreme null hypothesis (that yawns are not contagious,
and the appearance of contagion is an illusion), and assuming a 1 minute latency,
then 14.22% of all yawns would be incidental, but would otherwise be incorrectly
Percent
Contagious
categorized as a contagious. This value explodes to 48.87% if we assume yawns are
contagious for 5 minutes.
If yawns are contagious (as the evidence suggests) then the true rate of
contagious yawns (at 1 minute latency) is 33.76%, with 12.38% of all yawns being
incidental (and miscategorised as contagious). With a 5-minute latency we are wrong
more often than we are right, with 43.15% of yawns being incidental and only
33.77% being contagious. Put another way, with a 1-minute latency, there is a 3 in 4
chance that a contagious yawn is actually a contagious yawn. With a 5-minute
latency, there is a less than 1 in 2 chance that it is actually contagious.
Critically, if yawns are actually contagious for 1 minute, but we incorrectly
assume that yawns are contagious for longer, then these values explode. If we
assume yawns are contagious for longer than they really are, then the true rate of
contagion stays constant at around 1/3rd (as per Table Xb), but the proportion of
incidental yawns increases from a 12.38% (for 1-minute latency) to 33.52% (for a 3minute latency) and to 40.31% (for a 5-minute latency). Note that this values are
very similar to the values in Table Xb, but are theoretically different – the data in
Table Y (below) do not allow [true] contagious yawns to cause other contagious
yawns (whereas, in Table Xb, this relationship is possible).
Table Y. Hypothetical Data: 120-simulated minutes, with contagion of 1.5x and a Base-Rate =
1.666% (simulated 25 times), but with a true contagious latency of 1.
Incident
Latency 3
Incident
Latency 5
Mean
Incidental
Mean
Contagious
Mean
Spontaneous
Mean Total
Percent
Incidental
Percent
Spontaneous
Percent
Contagious
9.44 (3.15)
9.16 (3.50)
9.56 (2.27)
28.16 (5.05)
33.52%
33.95%
32.53%
11.56 (3.07)
9.56 (2.71)
7.56 (3.11)
28.68 (5.12)
40.31%
26.36%
33.33%
Discussion
The rate at which spontaneous yawns are miscategorised as contagious,
when they are actually incidental, is non-trivial. This is true assuming the extreme
null hypothesis (that yawns are not contagious), assuming yawns truly are
contagious, as well as assuming they are contagious and our understanding of their
latency is incorrect. Due to the fact that no study has empirically quantified the
degree to which a yawn makes another susceptible to yawning, the modest value of
1.5 was selected – observing a yawn makes one only 50% more likely to yawn than
they would ordinarily. However, the rate of 1-yawn an hour broadly corresponds to
work by Zilli et al. (2007) who asked participants to record the frequency of their
yawning over many days. The model does not assume relatedness or social
connectedness between agents, even though this has been shown to influence
susceptibility. Our corresponding behavioural experiment is setup such that it also
bears these assumptions. Conservatively, our model assumes all agents are
susceptible to contagious yawns, even though research shows that only about 50%
of people actually are. If our model were to correctly assume this fact, it follows that
the Type 1 rate would skyrocket. Finally, our model treats instances in which an
agent observes more than 1 yawn before they calculate whether they yawn as nonadditive – an assumption unaddressed by the empirical literature.
Methods: The Behavioral Experiment
Participants
Participants were undergraduate students at a large Australian University
who received course credit in exchange for their time.
Experimental Design and Protocol
This was a within-subjects design with two levels of one factor. Participants
sat listening to an audio programme on their personal devices while wearing a blindfold, or not wearing a blind-fold.
Participants were tested in groups of 9 [provide mean/mode here]. Upon
arrival they were asked to sit in chairs arranged in a circle facing inwards (thus, each
participant could see 8 others when not wearing a blind-fold). They were told they
would be asked to remain seated for the duration of the experiment. Then they
were asked to download/stream [classical music] and listen to it using earphones on
their personal devices (iPhones, Androids, etc.). This was done for two reasons, 1) in
the event that someone yawned during the blind-fold condition, the earphones
would prevent them from hearing it (which has been shown to elicit a contagious
yawn; REF), 2) it would ensure that participants interest levels were comparable. The
order of blind and non-blind conditions was counterbalanced over testing sessions.
After X-MINUTES [APPROXIMATELY HALF THE STUDY TIME] of the first session
participants were asked to pause their podcasts and either remove/don their blindfolds (according to condition). At the completion of the testing session
(approximately 1 hour) participants filled out a brief survey assessing to what extent
they found the programme interesting, how sleepy they were, their recollection of
whether they yawned in the preceding hour, and self-report measures of their own
tendency to yawn contagiously.
While evidence shows that being observed moderates the rate of yawning,
participants were subtly informed (in the written brief sheet) that they were being
filmed. Two cameras were placed inconspicuously in the testing room. The
experimenter sat at a remove from the group, facing a wall. At no point prior to the
survey was any cue given that the experiment was about yawning (i.e., the
experimenter did not yawn, and the recruiting and briefing process gave no
indication as to the research topic).
Behavioral Coding
The lead author coded the behaviour of all participants based on the video
data (available in full online at …). % OF VIDEOS WERE CODED BY INDEPENDENT
CODERS. ALPHA = XX. Each time a participant yawned it was counted, and it’s timing
recorded. Data was split by condition.
Results
Behavioral Results

t-test on ‘interest’ of the programme. Should be null between session 1 and
session 2.

t-test on non-blind yawns that occurred in part 1 vs part 2

t-test on blind-fold yawns that occurred in part 1 vs part 2
o If blind-yawns are sig higher in part 2 vs part 1 it could be that a)
yawns remain contagious for a very long time (or at least, yawns
prime further yawns even if they don’t lead to ‘contagion’), or b)
participants are just more tired/bored in part 2 than part 1, if so…

t-test on all yawns in part 1 vs part 2. If part 2 is greater than
part 1, then it’s not that yawns are long-term contagious, it’s
that sitting down for that long increases yawning behaviour.

Base-rate frequency of average spontaneous yawns (yawns/per min)
reported (blind condition).

Frequency of yawns (yawns/per min) in the non-blind condition.

T-test comparing (base-rate) blind yawns to non-blind yawns
o Predict a sig differences such that more yawns occurred in the nonblind condition.

Survey data: how long do people feel a trigger yawn is likely to elicit a
contagious response
Fitting the Data to the Model
NOTE: WELL, THERE SHOULDN’T BE MANY DF’S ON HOW TO DO THIS. THE PROGRAM I
WROTE IS PRE-REGISTERED, AS IS THE METHODS FOR DATA COLLECTION. FITTING THESE
TWO TOGETHER SHOULD SIMPLY BE A TRANSPARENT MATTER OF PRESENTING DECISIONS
AND DATA.
References
Anderson, J., & Meno, P. (2010). Psychological Influences on Yawning in Children.
Current Psychology Letters, 26(1), 1 - 9.
Baenninger, R., & Greco, M. (1991). Some antecedents and consequences of
yawning. The Psychological Record, 41(1980), 453 - 460.
Bartholomew, A. J., & Cirulli, E. T. (2014). Individual variation in contagious
yawning susceptibility is highly stable and largely unexplained by
empathy or other known factors. PLoS One, 9(3), e91773.
Conway, J. (1970). "The game of life". Scientific American, 223(4), 4.
Daquin, G., Micallef, J., & Blin, O. (2001). Yawning. Sleep Medicine Reviews, 5(4),
299 - 312.
Demuru, E., & Palagi, E. (2012). In bonobos yawn contagion is higher among kin
and friends. PLoS One, 7(11), e49613.
Gallup, A. C. (2011). Why do we yawn? Primitive versus derived features.
Neurosci Biobehav Rev, 35(3), 765-769.
Gallup, A. C., & Gallup, J., G. G. (2008). Yawning and thermoregulation. Physiology
and Behavior, 95, 10 - 16.
Grimm, V., Berger, U., Bastiansen, F., Eliassen, S., Ginot, V., Giske, J., et al. (2006). A
standard protocol for describing individual-based and agent-based
models. Ecological Modelling, 198(1-2), 115-126.
Guggisberg, A. G., Mathis, J., Schnider, A., & Hess, C. W. (2010). Why do we yawn?
Neurosci Biobehav Rev, 34(8), 1267-1276.
Guggisberg, A. G., Mathis, J., Schnider, A., & Hess, C. W. (2011). Why do we yawn?
The importance of evidence for specific yawn-induced effects. Neurosci
Biobehav Rev, 35(5), 1302-1304.
Marewski, J., & Olsson, H. (2009). Beyond the Null Ritual. Zeitschrift für
Psychologie / Journal of Psychology, 217(1), 49 - 60.
Norscia, I., & Palagi, E. (2011). Yawn contagion and empathy in Homo sapiens.
PLoS One, 6(12), e28472.
Platek, S. M., Critton, S. R., Myers, T. E., & Gallup, G. G. (2003). Contagious
yawning: the role of self-awareness and mental state attribution.
Cognitive Brain Research, 17(2), 223-227.
Platek, S. M., Mohamed, F. B., & Gallup, G. G., Jr. (2005). Contagious yawning and
the brain. Brain Res Cogn Brain Res, 23(2-3), 448-452.
Provine, R. R. (1986). Yawning as a Stereotyped Action Pattern and Releasing
Stimulus. Ethology, 72(2).
Provine, R. R. (1989). Faces as releasers of contagious yawning- An approach to
face detection using normal human subjects. Bulletin of the Psychonomic
Society, 27(3).
Provine, R. R., & Hamernik, H. (1986). Yawning: Effects of stimulus interest.
Bulletin of the Psychonomic Society, 24(6), 437 - 438.
Rundle, B. K., Vaughn, V. R., & Stanford, M. S. (2015). Contagious yawning and
psychopathy. Personality and Individual Differences, 86, 33-37.
Senju, A., Maeda, M., Kikuchi, Y., Hasegawa, T., Tojo, Y., & Osanai, H. (2007).
Absence of contagious yawning in children with autism spectrum
disorder. Biology Letters, 3, 706 - 708.
Shoup-Knox, M., Gallup, A. C., Gallup, J., & McNay, E. (2010). Yawning and
stretching predict brain temperature chagnes in rats: support for the
thermoregulatory hypothesis. Frontiers in Evolutionary Neuroscience, 2, 1
- 5.
Walusinkski, O. (2009). Yawning in diseases. European Neurology, 62, 180 - 187.
Wilensky, U. (1999). NetLogo. Centre for Connected Learning and ComputerBased Modeling: Northwestern University. Evanston, IL.
Yoon, J. M. D., & Tennie, C. (2010). Contagious yawning: a reflection of empathy,
mimicry, or contagion? Animal Behaviour, 79(5), e1-e3.
Zilli, I., Giganti, F., & Salzarulo, P. (2007). Yawning in morning and evening types.
Physiol Behav, 91(2-3), 218-222.