by Hugo Mercier
Passion within Reason?
From the evolutionary perspective I have adopted in this book,
emotional contagion is implausible. If emotions were truly con-
tagious, if they forced irrepressible mimicry, they would be too
easily abused. Cheaters could laugh until those they have cheated
laughed with them. Mortal enemies could get their opponents
to empathize with and care for them. If our emotions were so
easily manipulated, we would be much better off not paying any
attention to emotional signals.
Con temporary emotion researchers, such as Hatfield and her
colleagues, are quick to point out the limits of the basic view of
contagion, in which like begets like, as scholar Alfred Espinas
suggested in the nineteenth century: “The repre sen ta tion of an
emotional state gives rise to this same state in those who witness
it.”19 Such contagion makes no sense whatsoever for some emo-
tions. Take anger. We express anger to impress on others that
we have been wronged, and that this better not happen again.20
w h at t o f e e l? 99
If the only effect of expressing anger were to make others angry
in turn, expressing anger would be counterproductive.
We could thus imagine a version of contagion that would
be much less strict than “like begets like.” An emotion would
simply have to provoke a reaction, even if that reaction were
diff er ent from the initial emotion. But that wouldn’t solve the
prob lem raised by contagion. If displays of anger consistently led
the onlookers to yield to the angry individual, the weakest
among us could show their anger and make any opponent submit,
regardless of their relative strengths.
There has to be something that keeps emotional signals
broadly reliable, that is, beneficial on average for those who re-
ceive them. Charles Darwin, who devoted a book to the expres-
sion of emotions, was well aware of the prob lem. In the case of
blushing he cites one of his colleagues, Thomas Burgess, who
suggested that blushing serves the function of putting our shame-
ful wrongdoings in full view. For Burgess, the creator provided
the soul with the “sovereign power of displaying in the cheek, that
part of the human body which is uncovered by all nations, the
vari ous internal emotions of the moral feelings whenever they
are infringed upon either by accident or design.”21 Blushing
would be an honest signal because of a direct line between the
soul and our cheeks, one that could not be tampered with by
conscious volition. Surprisingly enough, the common answer to
why emotional signals remain reliable hasn’t changed much since
Burgess attributed this reliability to the benign hand of the lord.
Economist Robert Frank wrote the most thought- provoking
book about the function of emotions since Darwin: Passions
within Reason.22 In this work, he argues that displaying emotions
can be rational. Consider the prob lem raised by making credi-
ble threats of retaliation. To prevent others from wronging us,
we want them to believe that if they wrong us, we will retaliate.
100 ch ap t er 7
However, as I explained in the preceding chapter, once we have
been wronged, retaliation is often not the most sensible option.
Imagine you get scammed when buying a cheap product online,
and the seller is adamant that they won’t issue a refund. You could
sue them, but that might cost a lot of money, and it is sure to take
up a lot of time. It is often more sensible to just drop it. Know-
ing this, scammers can take advantage of you. But if you could
convince scammers that you would retaliate, regardless of the
costs, then you wouldn’t get scammed— and so you wouldn’t
even have to retaliate.23
For Frank, emotions and their expression evolved by solving
these types of prob lems. Anger would have evolved to express
one’s commitment to retaliate if wronged, regardless of the costs
of retaliation. The obvious question, then, is what makes the ex-
pression of anger credible? For Frank— and others before and
after him— the answer is that the expression of emotions, like
our reaction to these expressions, is automatic, outside of con-
scious control: “If all the facial muscles were perfectly subject to
conscious control, facial expressions would be robbed of their
capacity to convey emotional information.”24 Other cues are also
described as being intrinsically honest because they are outside
of conscious control: pupil dilation as a cue to arousal, blushing
as a cue to guilt.
The cir cuit of emotional contagion would thus be complete:
people can afford to react automatically to emotional signals
because emotional signals are sent automatically and thus are
impossible to fake. By this means, emotions could spread until
they affect a whole crowd, in an avalanche of automatic and com-
pel ing signals.
From an evolutionary point of view this reasoning does not
hold water, as it is irrelevant whether or not a signal is sent con-
sciously. Take Thomson’s gazelles trying to communicate to
w h at t o f e e l? 101
wild dogs that they are too fit to be caught. Instead of stotting,
which requires some energy, why not emit some kind of call?
Whether or not the call was under the gazelles’ conscious con-
trol would not affect its reliability: unfit gazelles would soon
evolve to give the same call, voluntarily or not, and dogs would
pay the call no heed.
Similarly, if some be hav iors or emotional displays were able
to reliably elicit a reaction in their audience, individuals would
evolve to send these signals whenever it was in their interest to
do so, even if that meant sending unreliable signals. The gazelles’
stotting is a reliable signal because it is simply impossible for unfit
gazelles to evolve the capacity to stot convincingly. By contrast,
there is no such obstacle for emotional signals, however auto-
matic they are. So why not display anger even when we would
never retaliate? Why not blush even if we wouldn’t hesitate for
one minute to repeat our shameful actions?
If automaticity is no guarantee of reliability, why do we pay
any attention to emotional signals? What keeps them honest?
Emotional Vigilance
The answer lies in clearing up the confusion between two
closely related concepts: automatic and mandatory.25 A cogni-
tive mechanism is automatic if it functions outside of conscious
control. Fortunately, this is true of the vast majority of cogni-
tion, most of the time: we couldn’t consciously focus on all
the steps necessary to make sense of an utterance or interpret
a visual scene. A cognitive mechanism is mandatory if it cannot
help but run its course once the right stimulus is pre sent.
Mandatory mechanisms would be like reflexes: if the doctor’s
hammer hits you on the right spot under the knee, you will al-
ways raise your foot.
102 ch ap t er 7
It is tem
pting to think that if a cognitive mechanism is auto-
matic, then it has to also be mandatory. That’s only because we
focus too much on conscious control; in fact, if most cognitive
mechanisms are automatic, very few, if any, are mandatory.26
Seeing a slice of scrumptious choco late cake makes most
people hunger for it. This reaction is hard to repress—it is
automatic— even when we’re on a diet (especially when we’re
on a diet). However, the same slice of choco late cake might elicit
only disgust after a heavy meal capped by two portions of cheese-
cake. Again, this reaction would be wholly automatic. Yet,
because the same stimulus can yield opposite reactions in dif-
fer ent contexts, neither reaction is mandatory.
If our reactions to emotional signals aren’t mandatory, then
there is room for what Guillaume Dezecache, Thom Scott-
Phil ips, and I have called emotional vigilance— mechanisms of
open vigilance dedicated to emotional signals.27 Even if they do
so unconsciously, people should be able to adjust their reactions
to emotional signals so as to stop responses that are not in their
best interest. The application of this emotional vigilance would
then provide incentives for senders to avoid sending unreliable
emotional signals.
How should emotional vigilance function? There is likely not
a one- size- fits- all recipe. Emotional vigilance should be attuned
to the properties of diff er ent emotions. For instance, disgust
might offer fewer opportunities for manipulation than anger:
think of how useful many (all?) would find it to make anyone
they want submit to them. By contrast, making people disgusted
appears less useful— except maybe to get all of that scrumptious
choco late cake for yourself. Stil , when reacting to emotional sig-
nals, the following three factors should be relevant across all
emotions: what our prior beliefs and plans are, in what context
the signals are produced, and whether the sender is trustworthy.
w h at t o f e e l? 103
Even babies and toddlers— not the creatures one would grant
with the most sophisticated emotional control— can take these
factors into account when reacting to emotional signals.
Not many parents will be surprised to hear that children are
masters of selective ignorance. They pay attention to what their
parents say only when it suits them. An experiment conducted
by psychologist Catherine Tamis- LeMonda and her colleagues
offers a nice demonstration with eighteen- month- olds.28 The ba-
bies had to choose whether or not to walk down a slope. Their
mothers were either encouraging them to walk or tel ing them
not to, with a mix of emotional signals. The babies could not
avoid the signals because their mothers were right in front of
them, gesticulating and making faces. Indeed, they understood
their mothers perfectly. For slopes that were neither too steep
nor too flat, the babies paid attention to the mothers: only a quar-
ter ventured down the slope when the mother was sending
negative signals, while three- quarters went down if she was send-
ing positive signals. Other wise, the babies completely ignored
their mothers. If the slope was perfectly safe— only a few degrees
of inclination— the babies went for it even if their mother was
urging them not to. If the slope was clearly dangerous— a fifty-
degree angle— the babies stopped, irrespective of what the
mother was signaling (an experimenter was there to catch them
had they de cided other wise; no babies were hurt in the making
of this experiment). The babies were engaging in a simple form
of plausibility checking.
Young toddlers also understand when an emotional display
is justified. In their article “Cry Babies and Pollyannas,” psycholo-
gists Sabrina Chiarel a and Diane Poulin- Dubois describe an
experiment, also conducted with eighteen- month- olds.29 The
babies were shown videos of an actress expressing either a justi-
fied emotion (happiness after being handed a nice toy) or an
104 ch ap t er 7
unjustified one (sadness in the same conditions). The babies
were more curious about the unjustified display of sadness, look-
ing back and forth between the object and the actor, trying to
figure out what was happening. They also expressed less concern,
and were less likely to call for help, when the emotion seemed
unjustified.
A similar experiment with toddlers— three- year- olds—
revealed that they not only adjusted their reactions as a function
of whether or not the emotional displays were justified but also
held senders of unreliable signals accountable. Psychologist Rob-
ert Hepach and his colleagues had the toddlers interact with adults
who expressed consistently justified or unjustified emotions.30
The adult would express distress if her sleeve (rather than her
hand) got caught under the heavy lid of a box, she would cry after
a drawing of hers was slightly dented (rather than torn apart), and
she would complain at having received a fair (rather than unfair)
share of marbles in a game. Later on, the adult would start crying
from behind a screen. When that adult had consistently sent un-
justified signals, only a third of the toddlers checked on her,
whereas more than 80 percent did so when her complaints had
been well founded. In a later task, the toddlers were less likely to
help the adult who had sent unreliable emotional signals.
Here lies the key to the stability of emotional signals: many
are rejected outright, and those who abuse the signals end up
paying the price. Abusers might not be punished in the typical
sense of being physically punished, but their reputation suffers,
as does the reputation of those who break more explicit commit-
ments. Senders of unreliable emotional signals are trusted less
when they send emotional signals, and possibly when they em-
ploy other forms of communication as wel .
What about adults, then? Don’t the experiments reported
earlier in the chapter show that adults inevitably mimic the
w h at t o f e e l? 105
emotions they perceive in others? Could humans lose the abil-
ity to discriminate between reliable and unreliable emotional
signals as they grow up?
No. Adults also adjust their reactions to emotional signals
as a function of their source, and of the context in which the
signals are emitted. Lanzetta and Englis had shown that partici-
pants automatically mimic the smile or frown expressed by a
confederate, but only when the participants expected to cooper-
ate with the confederate later on. When the participants expected
to compete with him instead, they tended to show opposite
reactions, smiling when the confederate received a shock and
frowning when he was rewarded— what Lanzetta and Englis
called counterempathy.31
Source effects have been reported in many experiments. Tears
are taken to indicate sadness more reliably if they are shed by an
adult rather than a toddler.32 Women do not mimic the expres-
sions of those who behave unfairly toward them.33 Men express
positive emotions when others show fear, and negative emotions
when others show joy—if the others are fans of a rival sports
team.34 Even catching yawns, a seemingly perfect example of ir-
resistible contagion, is not as reflexive as it seems: people are
more likely to start yawning when they see people they know,
rather than strangers, yawn.35 And, like toddlers, adults increas-
ingly mistrust those who mispresent their emotions— for in-
stance, people who feign anger to obtain strategic advantages in
negotiations.36
Contagion Is a Catchy but Misleading Analogy
Our reactions to emotional signals might be automatic—we do
not consciously control our emotional reactions— but they are
far from being mandatory. Instead, they adapt according to a
106 ch ap t er 7
number of factors, including our preexisting plans or beliefs, the
context, and the source’s credibility. This calls into question the
contagion analogy.37 We evolved neither to send nor to receive
pathogens— indeed, a good chunk of our evolution is devoted
to avoiding the effects of pathogens. By contrast, we did evolve
to send and receive emotional signals.38 It therefore makes little
sense to talk about contagion when talking about people’s reac-
tions to emotions.
Describing the transmission of emotions as resulting from
contagion is merely giving a new name to the phenomenon of
transmission, without any explanatory purchase: no facet of the
phenomenon is better understood thanks to the analogy (on the
contrary!).39 Because contagion from actual pathogens is rela-
tively well understood, this sleight of hand provides an illusion
of understanding, but in fact pathogen contagion and emotional
communication have more differences than commonalities.40
What about the costs, though? Can’t contagion, whether by
pathogens or by emotional signals, be costly for those affected?
How can we reconcile the idea that our reactions to emotional
signals are adaptive, fine- tuned to protect us from unreliable
senders, with the epidemics of weird be hav iors, or the way
crowds are supposed to turn individuals into bloodthirsty hoo-
ligans or panicked sheep? In fact, the current adaptive perspec-
tive is not incompatible with the supposed cases of emotional
