Content uploaded by Yang Wu
Author content
All content in this area was uploaded by Yang Wu on Sep 13, 2018
Content may be subject to copyright.
Toddlers Connect Emotional Responses to Epistemic States
Yang Wu (yangwu@mit.edu), Laura E. Schulz (lschulz@mit.edu),
Rebecca Saxe (saxe@mit.edu)
Department of Brain and Cognitive Sciences, MIT
77 Massachusetts Avenue, Cambridge, MA 02139 USA
Abstract
Emotional expressions are typically transient; while we may
react emotionally to a new event, we are unlikely to respond
with the same emotion once the event becomes familiar. Here
we look at whether toddlers understand the relationship
between people’s epistemic states and their emotional
responses. Younger (12-17-month) and older (18-24-month)
toddlers were familiarized with a movie in which an observer
was knowledgeable or ignorant about a recurring event. On the
test trial, the observer saw the event and either remained neutral
or changed to a valenced emotional reaction (positive or
negative). We predicted that the change from a neutral to a
valenced expression would be more surprising if the event was
familiar to the observer than if the event was novel. We found
an interaction between epistemic state and emotion for older
but not younger toddlers. These results suggest that before age
two, children begin to understand the transient nature of
emotional reactions and their dependence on people’s
epistemic states.
Keywords: emotion understanding; epistemic state; ignorance;
causal reasoning; toddlers
Introduction
The ability to understand others’ emotions is a critical
component of theory of mind. Although considerable
research has looked at how we might recognize others’
emotions from overt emotional displays (including facial
expressions, vocalizations, body posture, and gait; e.g.,
Ekman & Friesen, 1971; Bachorowski & Owren, 2003;
Meeren, van Heijnsbergen, & de Gelder, 2005; Dael,
Mortillaro & Scherer, 2012), an increasing body of work
suggests that adults have a rich, abstract intuitive
understanding of the conditions that tend to elicit different
emotions (Skerry & Saxe, 2015; Ong, Zaki & Goodman,
2015; Wu, Baker, Tenenbaum & Schulz, 2018; Fontaine,
Scherer, Roesch, & Ellsworth, 2007; Houlihan & Saxe 2017).
This intuitive theory allows us to connect others’ probable
emotional response to an event to their appraisal that the
event was goal congruent, expected, familiar, fair,
controllable, etc. (e.g., Fontaine, Scherer, Roesch, &
Ellsworth, 2007; Scherer & Meuleman, 2013; Ortony, Clore,
& Collins, 1990).
How do children learn this rich intuitive theory of others'
emotions? Prior research suggests that even toddlers
1 We use the term “goal congruence” in a broad sense here. “Goal”
can refer to not only individuals’ goals or desires such as losing
weight, but also more abstract goals such as being alive, healthy and
happy. Thus, although some events investigated in the literature
recognize the importance of goal congruence for predicting
others' emotions in response to events. It is less clear whether
toddlers can incorporate inferences about others' appraisals of
event expectedness, fairness, or controllability into emotion
predictions. As a first step, here we ask whether and when
toddlers incorporate inferences about the subjective novelty
or expectedness of an event into their expectations of others'
emotional reactions.
Developmental research suggests that infants and toddlers
understand the relationship between the goal congruence1 of
an event and the valence of someone’s emotional responses
(Chiarella & Poulin-Dubois, 2013, 2018; DeLoache &
LoBue, 2009; Repacholi & Gopnik, 1997; Reschke, Walle,
Flom, & Guenther, 2017; Skerry & Spelke, 2014; Wu,
Muentener, & Schulz, 2017). For example, ten-month-olds
look longer if an agent expresses a negative (versus positive)
reaction to achieving her objective (Skerry & Spelke, 2014),
and 18-month-olds can use an agent’s positive and negative
emotional responses to identify which of two foods she wants
(Repacholi & Gopnik, 1997). A recent study (Wu, Muentener,
& Schulz, 2017) suggests a more fine-grained understanding
of this relationship: twelve to seventeen-month-olds can
differentiate not only cross-valence, but also within-valence
emotional expressions, connecting diverse positive emotional
vocalizations to their probable external events (e.g., linking
funny events to a laughing response, light-up toys to
excitement, adorable babies to “Aww…”, etc.).
For adult perceivers however, goal congruence is only one
feature of events that is used when predicting others'
emotional reactions. A second key dimension is whether the
event is subjectively familiar and/or expected, versus novel
and/or unexpected. Toddlers are certainly able to track which
events are subjectively novel to other people (even if the
event is not novel for the toddler herself). For instance, in one
classic paradigm, a child plays with two objects and an agent,
then the agent leaves the room and a third object is introduced.
If the agent returns, looks at the three objects and says either
“A modi!” or “Wow! Cool!" followed by "Can you give it to
me?” 18- and even 12-month-olds selectively assign the
object label or hand off the object new to the observer
(Tomasello & Haberl, 2003; Tomasello & Akhtar, 1995; see
also, O’Neill, 1996; Akhtar, Carpenter, & Tomasello, 1996;
(e.g., seeing something funny or scary) were not explicitly framed
as goal congruent or incongruent, the valence of such events was
defined implicitly by their congruence with some conventional goals.
2711
Liszkowski, Carpenter, Tomasello, 2008; Luo & Baillargeon,
2007).
Thus, toddlers can tell which objects or events are
subjectively novel or unexpected to other people. But do
toddlers incorporate this epistemic state into their predictions
of other's emotional reactions to events? Evidence to date is
unclear. On one hand, a recent study found that by twenty
months, toddlers expect that someone with a false belief will
express surprise (rather than satisfaction) on observing an
unexpected outcome (Scott, 2017). On the other hand, like
many claims about early false belief understanding, this result
is in tension with earlier studies finding that children fail to
predict that someone will be surprised by the unexpected
contents of a container until they are five or six years old
(MacLaren & Olson, 1993; see also Hadwin & Perner, 1991;
Ruffman & Keenan, 1996; Wellman & Banerjee, 1991;
Wellman & Bartsch, 1988).
In the current study, we test directly whether toddlers can
incorporate another person's Knowledge or Ignorance of an
event to predict her emotional reaction. Specifically, adults
expect that both Positive and Negative emotional reactions
are likely to be more intense to novel or unexpected events,
whereas familiar or expected events are more likely to evoke
muted or even Neutral reactions (e.g., Fontaine, Scherer,
Roesch, & Ellsworth, 2007; Scherer & Meuleman, 2013;
Ortony, Clore, & Collins, 1990). Thus, we test toddlers’
ability to predict emotional reactions based on a person's
epistemic state, independent of any prior information about
the goal congruence of the event, or the valence of the
emotional reaction. Do toddlers expect that any vigorous
emotional reaction to an event is more likely if the person was
previously ignorant than if the agent was previously
knowledgeable, regardless of whether the emotional reaction
is positive or negative?
To measure these expectations, we use a looking time task.
Within participants, all toddlers see one sequence in which an
agent remains Ignorant about an event over four
familiarization trials and another sequence in which an agent
is Knowledgeable about the event (order counterbalanced,
with different agents and events between conditions).
Between participants, we manipulate whether the agent’s
final emotional expression to the event is Positive, Negative,
or remains Neutral.
We predict that both Positive and Negative emotional
reactions will be more surprising given an agent who already
knows about the events than given an agent seeing them for
the first time. Thus the primary prediction is that given a
valenced reaction (either Positive or Negative) toddlers will
look longer in the Knowledgeable condition (where the agent
has seen the events and responded neutrally four times before)
than in the Ignorant condition (where the agent first notices
the event on the test trial). Conversely, the Neutral reaction
should be more surprising given an agent first seeing the
events than an agent who has seen the events all along. Thus
the pattern should reverse: given a Neutral reaction, toddlers
should look longer in the Ignorant condition than the
Knowledgeable condition. However, since the events
themselves are not especially emotive, a neutral reaction may
be relatively unsurprising even as a first response, and the
effect in the Neutral condition may well be attenuated. The
Neutral condition serves primarily as a control to ensure that
any change in the predicted direction in infants’ looking is
driven by the valenced emotional reaction, not merely the
shift from ignorance to knowledge. Consistent with previous
work suggesting young toddlers’ ability to distinguish
ignorant and knowledgeable agents (Tomasello & Haberl,
2003; Tomasello & Akhtar, 1995; O’Neill, 1996), we start by
testing these predictions with 12-17-month-olds (Experiment
1; pre-registered on the Open Science Framework:
https://osf.io/xae5f/?view_only=7b655cac3f744bd3a299591
a856301f6). We then run an exploratory experiment
(Experiment 2) testing older toddlers: 18-23-month-olds.
Experiment 1
Method
Participants
A pre-registered sample of N = 48 12-17-month-olds (mean:
15.2 months, range: 12.0-17.7 months; 48% girls) were
recruited from an urban children’s museum. Toddlers were
assigned to three emotion conditions (order counterbalanced):
positive (n = 16, mean: 15.5 months, range: 12.0-17.7
months), negative (n = 16, mean: 15.2 months, range: 13.1-
17.5 months), and neutral (n = 16, mean: 14.9 months, range:
12.1-17.4 months). Twenty-two infants were replaced due to:
fussiness (n = 18), family interference (n = 1), and eyes not
visible from videotapes (n = 3). While most of the children
were white and middle class, a range of ethnicities and
socioeconomic backgrounds reflecting the diversity of the
local population (47% European American, 24% African
American, 9% Asian, 17% Latino, 4% two or more races) and
the museum population (29% of museum attendees receive
free or discounted admission) were represented throughout.
Stimuli & Procedures
We created two comparable sets of movie stimuli. A female
actor and a pink box with a monkey puppet were in one set,
and another female actor and a blue box with a tiger puppet
were in the other (see Figure 1). All movie stimuli can be
downloaded here:
https://osf.io/va8xg/?view_only=b36670f1ace54908affecee
5b1d11f90.
Toddlers were tested in a dimly-lit room at a children’s
museum. The child’s parent sat in a chair, approximately 63
cm in front of the screen (93 cm wide, 56 cm high), holding
the child. A laptop was used to execute the stimulus
presentation and was concealed behind the screen. After the
parent and child were seated, the experimenter went behind
the screen and started the presentation. The experimenter was
blind to the epistemic state condition (i.e., Knowledgeable or
Ignorant) throughout but could see the child and code her
looking time via a webcam mounted on the screen. A multi-
colored pinwheel was used as an attention getter and was
presented first. Once the child looked at it, the experimenter
pressed a button to initiate a trial.
2712
Familiarization Phase An actor appeared on the screen and
said: “Hi, baby!” Then the screen turned black and a chime
sound was played to get the child’s attention. Then the actor
reappeared. She sat on a chair holding a book. On her left,
there was a box with a lid closed. The actor looked at the child
for 2 seconds. Then she turned left, looking at the box for 5
seconds. Then the actor looked back at the child for 2 seconds.
Then she turned right, reading the book for 5 seconds. Her
expression remained neutral throughout. She repeated this
procedure four times. In the meantime, a puppet popped up
from inside the box but the timing differed by condition. In
the Ignorant condition, the puppet always popped up when
the actor was reading the book. Thus, the actor never saw the
puppet. In the Knowledgeable condition, the puppet always
popped up when the actor was looking at the box. Thus, the
actor was familiar with the puppet. See Figure 1. The
familiarization phase takes about 70 seconds in total.
Test Phase After familiarization, the screen turned black and
a chime sound was played again to get the child’s attention.
Then the actor reappeared. She looked at the child for 2
seconds. She then turned left, looking at the box. The puppet
popped up from inside the box in both the Ignorant and
Knowledgeable conditions. In the positive condition, the
actor expressed a happy facial expression accompanied by a
positive vocalization. In the negative condition, the actor
leaned backward and showed a negative face accompanied
by a negative vocalization. In the Neutral condition, the actor
showed a neutral response, maintaining the same she had
during the familiarization phase. See Figure 1. Then the
screen froze and the experimenter started to code the child’s
looking. Once the child looked away for two consecutive
seconds, the program automatically moved on to the next trial.
Each child participated in an Ignorant condition using one set
of stimuli (randomly selected) and a Knowledgeable
condition using the other set of stimuli. The order of the
conditions was randomized within participants, and the order
of the agent’s responses (Positive, Negative or Neutral) was
counterbalanced across participants.
Coding
A coder, blind to the epistemic state condition (i.e., Ignorant
or Knowledgeable), coded infants’ looking offline from
videotapes. This coding corroborated the experimenter’s
decision to end a trial in all but three cases; these three
children were dropped and replaced due to premature
termination of a trial. Only offline coding was used for further
analyses. Another coder coded whether each child looked at
the agent and the puppet at least once during the
Familiarization phase of each trial. All infants passed this
criterion. A different coder, blind to the epistemic state
condition, coded how attentive children were during the
familiarization phase; children were equally attentive to the
ignorant (looking time: M = 42.89s, SD = 11.50) and
knowledgeable (looking time: M = 43.52s, SD = 10.91; t(46)
= -.30, p = .769) familiarization events.
Results and discussion
We pre-registered a mixed-effects model to look at the effects
of Epistemic State (Ignorant or Knowledgeable) and Emotion
(Positive, Negative or Neutral). The fixed factors were
Epistemic State, Emotion and their interaction, and the
random factor was Subject. No slope was defined. We used
the lme function in the R package nlme, version 3.1-131. For
brevity, throughout the paper we only report the main effects
of the fixed factors; detailed summary (e.g., estimated effects,
standard errors, ts, and ps) of our models can be found here:
https://osf.io/s2kzx/?view_only=0db2376533974c28b5d48c
1886e1d47f. Our mixed-effects model suggests that the main
effect of Epistemic State was significant (F(1, 45) = 9.71, p
= .003). There was a non-significant trend towards a main
effect of Emotion (F(2, 45) = 2.90, p = .065). Contrary to our
predictions however, the interaction between Epistemic State
and Emotion was not significant (F(2, 45) = 1.27, p = .289).
As pre-registered, we looked at the effect of Epistemic
State for each emotion separately. However, there was no
significant difference between the Ignorant and
Knowledgeable conditions for either the Negative (t(15)
= .96, p = 1.000, 95% CI [-2.50, 6.57]; paired-sample t test;
p values were corrected with the Bonferroni method
throughout) or Neutral test condition (t(15) = 1.35, p = .587,
95% CI [-1.97, 8.84]). Also contrary to our predictions, in the
Positive condition, infants looked longer in the Ignorant than
the Knowledgeable condition (t(15) = 3.06, p = .024, 95% CI
[2.17, 12.12]).
X 4
X 4
TEST
FAMILIARIZATION
Ignorant Condition NeutralPositive Negative
Within-subjects
Between-subjects
Knowledgeable Condition
Figure 1 Experimental design
2713
We then ran an exploratory analysis, using the same mixed-
effects model but collapsing across the two valenced
conditions (since we predicted no difference between the
Positive and Negative conditions). Additionally, because
each child participated in both the Ignorant and Knowledge
conditions, and some children looked longer overall than
others, we analyzed both children’s raw looking time and
their proportional looking time in each of the two conditions
(calculated by dividing their looking time in each condition
by their total looking time in both conditions). Again
however, for both the raw looking time and the proportional
looking time, only the main effect of Epistemic State was
significant (raw: F(1, 46) = 9.42, p = .004; proportional: F(1,
46) = 7.47, p = .009). The interaction between Epistemic
State and Emotion was not significant (raw: F(1, 46) = .16, p
= .693; proportional: F(1, 46) = .00, p = .982). Nor was the
main effect of Emotion (raw: F(1, 46) = 2.26, p = .140;
proportional: F(1, 46) = .00, p = 1.000). See Figure 2 a.
Figure 2. Toddlers’ looking time in Experiments 1 and 2.
Error bars indicate bootstrapped 95% confidence intervals.
Thus, 12-17-month-olds showed no evidence of
understanding that a suddenly expressed emotional
expression would be more likely given a previously ignorant
agent than a knowledgeable one. Instead, regardless of the
agent’s emotional response, these toddlers looked longer
when a previously ignorant agent noticed something new
than when a knowledgeable agent saw something familiar.
These results are consistent with previous work (Vaish &
Woodward, 2010) showing that 14-month-olds can use
attention but not emotional cues to predict others’ actions:
they looked longer when an agent reached for an unattended
object than an attended one, regardless of whether she
previously expressed a positive or negative response to the
attended object. As noted however, Scott (2017) suggests that
by 20 months, toddlers are sensitive to the relationship
between surprise and false belief; so in Experiment 2, we test
18-23-month-olds, to see whether older toddlers connect
others’ epistemic states to their emotional reactions.
Additionally, we made a minor change on our stimuli. In
our original stimuli, the screen turned black after the
Familiarization phase (accompanied by a chime sound) and
then the agent reappeared for the Test phase. The short black
screen may have made it more difficult to construe the events
as a continuous sequence and may also have increased the
memory demands of our task. To react to the Test phase as
predicted, children would have to hold their representations
of the familiarization events in mind during this interval. In
our new stimuli, we removed the black screen to reduce some
task demands. We retained the chime sound to ensure we had
children’s attention at the beginning of the Test phase.
Experiment 2
Method
Participants
Forty-eight 18-23-month-olds (mean: 20.7 months, range:
18.2-23.8 months; 46% girls) were recruited from the same
children’s museum. As Experiment 1, they were assigned to
three emotion conditions (order counterbalanced): positive (n
= 16, mean: 20.5 months, range: 18.2-23.7 months), negative
(n = 16, mean: 20.6 months, range: 18.2-23.8 months) and
neutral (n = 16, mean: 21.1 months, range: 18.6-23.1 months).
Sixteen children were replaced due to: fussiness (n = 11),
family interference (n = 3), distraction (n = 1) and not looking
at the test events (n = 1).
Stimuli & Procedures
The stimuli and procedures were the same as Experiment 1
except for one minor change. We removed the black screen
at the beginning of the Test phase to increase continuity
between the Familiarization and Test phases. The modified
stimuli can be downloaded here:
https://osf.io/va8xg/?view_only=b36670f1ace54908affecee
5b1d11f90.
Coding
As Experiment 1, a coder, blind to the epistemic state
condition, coded infants’ looking offline from videotapes.
This coding corroborated the experimenter’s decision to end
a trial in all cases. Only offline coding was used for further
analyses. Another coder coded whether each child looked at
the agent and the puppet at least once during the
familiarization phase of each trial. All infants passed this
criterion. A different coder, blind to the epistemic state
condition, coded how attentive children were during the
familiarization phase; children were equally attentive to the
ignorant (looking time: M = 52.89s, SD = 12.49) and
knowledgeable (looking time: M = 56.46s, SD = 13.32; t(46)
= -1.47, p = .149) familiarization events.
Results and discussion
We used the same mixed-effects model as Experiment 1 to
look at the effects of Epistemic State (Ignorant or
Knowledgeable) and Emotion (Positive, Negative, and
0
5
10
15
20
25
Valenced Neutral
Emotional Response
Looking Time (s)
Raw Looking Time
0.00
0.25
0.50
0.75
1.00
Valenced Neutral
Emotional Response
Proportional Looking Time
Proportional Looking Time
0
10
20
30
40
Valenced Neutral
Emotional Response
Looking Time (s)
Epistemic Status Ignorant Knowledgeable
Raw Looking Time
0.00
0.25
0.50
0.75
1.00
Valenced Neutral
Emotional Response
Proportional Looking Time
Proportional Looking Time
Epistemic Status Ignorant Knowledgeable
(a) Experiment 1 (12-17-month-olds)
(b) Experiment 2 (18-23-month-olds)
2714
Neutral). There was no main effect of Epistemic State (F(1,
45) = 1.01, p = .320) or Emotion (F(2, 45) = .21, p = .812);
nor was there an interaction (F(2, 45) = 2.33, p = .109).
However, because we did not have different predictions for
the positive and negative conditions, we collapsed data across
these two valenced conditions. Specifically, we looked at
whether there was a significant interaction between
Epistemic State (ignorant or knowledge) and Emotion
(valenced or neutral). For the Valenced conditions, we
predicted that children would look longer in the
Knowledgeable than the Ignorant condition but for the
Neutral condition, we predicted that the effect would be, if
anything, reversed. As in Experiment 1, we analyzed both
children’s raw looking time and their proportional looking
time in each of the two conditions.
The older toddlers showed the predicted interaction
between Epistemic State and Emotion both considering their
raw looking time (F(1, 46) = 4.05, p = .050) and the
proportional looking time (F(1, 46) = 6.38, p = .015). In the
Valenced conditions, there was a marginally significant
effect in the predicted direction for the raw looking time (t(31)
= 2.38, p = .076, 95% CI [-12.33, -.39]; paired-sample t tests;
p values were corrected with the Bonferroni method
throughout), and a significant effect for the proportional
looking time (t(31) = 2.74, p = .020, 95% CI [-.30, -.04]). For
the Neutral condition, the Knowledgeable and Ignorant
conditions did not differ either by raw (t(15) = .92, p = .748,
95% CI [-6.37, 15.97]) or proportional looking time (t(15)
= .39, p = 1.000, 95% CI [-.21, .30]).
In sum, collapsing across valence, the results suggest that
18-23-month-old toddlers connect agents’ emotions to their
epistemic states: toddlers are more likely to expect a new
emotional response from a previously ignorant agent than
from a knowledgeable one. However, the results are not
robust to considering each valenced response separately.
Further research must replicate the design and analyses to
establish the strength of the effect.
General Discussion
In two experiments, we found that toddlers between 12 and
17 months old were sensitive to an agent’s epistemic state:
looking longer when an ignorant agent saw something new
than when a knowledgeable agent saw something familiar.
However, they appeared insensitive to the agent’s emotional
response to those events (Experiment 1). By contrast, 18-23-
month-old toddlers looked longer when a knowledgeable
agent had an emotional response to events she had previously
viewed neutrally than when a previously ignorant agent
changed her expression from neutral to valenced on first
observing the event (Experiment 2). Note that this cannot be
due to ancillary features of contrast between the Ignorant and
Knowledgeable conditions because the pattern did not
emerge when the agent’s expression remained neutral. These
results suggest that at least by the second half of the second
year, toddlers have some understanding of the emotional
consequences of an agent’s epistemic state. They understand
that emotional reactions are more likely to novel than familiar
events even when they cannot specifically predict anything
about the content of the emotion.
This study differs from earlier research showing that
infants as young as 12 months expect someone to attend to
and be excited by novel rather than familiar objects
(Tomasello & Haberl, 2003). In that study, both attentional
and emotional cues were available, and infants may have
primarily used attentional, not emotional, cues to decide that
“Wow! Cool!” referred to the object that was new to the
speaker. By contrast, in our study, we separated the effects of
attention and emotion by looking at the interaction between
the agent’s epistemic state and her emotional response.
Although the younger toddlers were sensitive only to an
agent’s epistemic state, there was an interaction between
epistemic state and emotion in the older toddlers, suggesting
that the emotional cues made a distinct contribution. These
results are consistent with other findings suggesting that
infants and children have an early understanding of some
antecedent causes of others’ emotions (e.g., Chiarella &
Poulin-Dubois, 2013, 2018; DeLoache & LoBue, 2009;
Repacholi & Gopnik, 1997; Reschke et al., 2017; Scott, 2017;
Skerry & Spelke, 2014; Wu et al., 2017).
Note that in the positive and negative conditions, the agent
expressed both an emotional facial expression and an
emotional vocalization to the test event; in the neutral
condition however, the agent maintained her neutral face
without any vocalization. Thus, the valenced conditions
differed from the neutral one both in a change of facial
expression and the presence of an emotional vocalization. We
are agnostic of whether the facial, vocal or both types of cues
drove the effects of our study. However, our data suggest that
the older toddlers did not simply responded to the presence
or absence of novel emotional cues. That is, they did not look
longer overall in the valenced conditions than the neutral
condition (see Figure 2 b). Instead, their looking time was
influenced jointly by whether the agent expressed novel
emotional responses, and whether the agent was ignorant of
or knowledgeable about the event, suggesting that the
toddlers incorporated the agent’s epistemic state in expecting
her emotional reactions to events.
Additionally, the interaction between Epistemic State and
Emotion became significant in Experiment 2 only when we
collapsed across the two valenced conditions (positive and
negative conditions). We did not have enough statistical
power to analyze the effects in the two valence domains
separately. Future work could take a closer look at this as well
as replicating our findings. Our current data, however,
provides initial evidence that at least by the second half of the
second year, toddlers begin to understand that people’s
emotional reactions depend on what they do or do not know
about events. Much work remains to be done to understand
how these early abilities develop into the rich, intuitive theory
of emotion found in human adults (Skerry & Saxe, 2015;
Scherer & Meuleman, 2013; Fontaine et al., 2007; Ortony et
al., 1990; Ong et al., 2015; Wu et al., 2018). However, our
study begins to shed some light on the early emerging
2715
abilities that may be fundamental to our later-developing
sophisticated understanding of emotions.
Acknowledgments
This study is supported by the Center for Brains, Minds and
Machines (CBMM), funded by NSF STC award CCF-
1231216. Warm thanks to the Boston Children’s Museum
and participating parents and children. Thanks to Elizabeth
Spelke and Josh Tenenbaum for helpful feedback, and to
Elizabeth Rizzoni, Sydney Kuo, Miranda Sachi Fry,
Catherine Wu, Jennah A. Haque and Anya Keomurjian for
help with data collection and coding.
References
Akhtar, N., Carpenter, M., & Tomasello, M. (1996). The role
of discourse novelty in early word learning. Child
Development, 67(2), 635-645.
Bachorowski, J. A., & Owren, M. J. (2003). Sounds of
emotion. Annals of the New York Academy of Sciences,
1000(1), 244-265.
Chiarella, S. S., & Poulin-Dubois, D. (2013). Cry babies and
pollyannas: Infants can detect unjustified emotional
reactions. Infancy, 18(s1), E81-E96.
Chiarella, S. S., & Poulin-Dubois, D. (2018). “Are You
Really Sad?” Infants Show Selectivity in Their Behaviors
Toward an Unconventional Emoter. Infancy, 23(3), 453-
470.
Dael, N., Mortillaro, M., & Scherer, K. R. (2012). Emotion
expression in body action and posture. Emotion, 12(5),
1085-1101.
DeLoache, J. S., & LoBue, V. (2009). The narrow fellow in
the grass: Human infants associate snakes and fear.
Developmental Science, 12(1), 201-207.
Ekman, P., & Friesen, W. V. (1971). Constants across
cultures in the face and emotion. Journal of Personality
and Social Psychology, 17(2), 124-129.
Fontaine, J. R., Scherer, K. R., Roesch, E. B., & Ellsworth,
P. C. (2007). The world of emotions is not two-
dimensional. Psychological Science, 18(12), 1050-1057.
Hadwin, J., & Perner, J. (1991). Pleased and surprised:
Children's cognitive theory of emotion. British Journal of
Developmental Psychology, 9(2), 215-234.
Liszkowski, U., Carpenter, M., & Tomasello, M. (2008).
Twelve-month-olds communicate helpfully and
appropriately for knowledgeable and ignorant partners.
Cognition, 108(3), 732-739.
Luo, Y., & Baillargeon, R. (2007). Do 12.5-month-old
infants consider what objects others can see when
interpreting their actions?. Cognition, 105(3), 489-512.
MacLaren, R., & Olson, D. (1993). Trick or treat: Children's
understanding of surprise. Cognitive development, 8(1),
27-46.
Meeren, H. K., van Heijnsbergen, C. C., & de Gelder, B.
(2005). Rapid perceptual integration of facial expression
and emotional body language. Proceedings of the National
Academy of Sciences of the United States of America,
102(45), 16518-16523.
O’Neill, D. K. (1996). Two-year-old children’s sensitivity to
a parent’s knowledge state when making requests. Child
Development, 67, 659-677.
Ong, D. C., Zaki, J., & Goodman, N. D. (2015). Affective
cognition: Exploring lay theories of emotion. Cognition,
143, 141-162.
Ortony, A., Clore, G.L., & Collins, A. (1990). The cognitive
structure of emotions. Cambridge university press.
Repacholi, B. M., & Gopnik, A. (1997). Early reasoning
about desires: evidence from 14-and 18-month-olds.
Developmental Psychology, 33(1), 12-21.
Reschke, P. J., Walle, E. A., Flom, R., & Guenther, D. (2017).
Twelve-Month-Old Infants’ Sensitivity to Others’
Emotions Following Positive and Negative Events. Infancy,
22(6), 874-881.
Ruffman, T., & Keenan, T. R. (1996). The belief-based
emotion of suprise: The case for a lag in understanding
relative to false belief. Developmental Psychology, 32(1),
40.
Saxe, R., & Houlihan, S. D. (2017). Formalizing emotion
concepts within a Bayesian model of theory of mind.
Current opinion in Psychology, 17, 15-21.
Scherer, K. R., & Meuleman, B. (2013). Human emotion
experiences can be predicted on theoretical grounds:
evidence from verbal labeling. PloS One, 8(3), e58166.
Scott, R. M. (2017). Surprise! 20-month-old infants
understand the emotional consequences of false beliefs.
Cognition, 159, 33-47.
Skerry, A. E., & Saxe, R. (2015). Neural representations of
emotion are organized around abstract event features.
Current Biology, 25(15), 1945-1954.
Skerry, A. E., & Spelke, E. S. (2014). Preverbal infants
identify emotional reactions that are incongruent with goal
outcomes. Cognition, 130(2), 204-216.
Tomasello, M., & Akhtar, N. (1995). Two-year-olds use
pragmatic cues to differentiate reference to objects and
actions. Cognitive Development, 10(2), 201-224.
Tomasello, M., & Haberl, K. (2003). Understanding
attention: 12-and 18-month-olds know what is new for
other persons. Developmental Psychology, 39(5), 906.
Vaish, A., & Woodward, A. (2010). Infants use attention but
not emotions to predict others’ actions. Infant Behavior
and Development, 33(1), 79-87.
Wellman, H. M., & Banerjee, M. (1991). Mind and emotion:
Children's understanding of the emotional consequences of
beliefs and desires. British Journal of Developmental
Psychology, 9(2), 191-214.
Wellman, H. M., & Bartsch, K. (1988). Young children's
reasoning about beliefs. Cognition, 30(3), 239-277.
Wu, Y., Baker, C. L., Tenenbaum, J. B., & Schulz, L. E.
(2018). Rational inference of beliefs and desires from
emotional expressions. Cognitive Science, 42(3), 850-884.
Wu, Y., Muentener, P., & Schulz, L. E. (2017). One- to four-
year-olds connect diverse positive emotional vocalizations
to their probable causes. Proceedings of the National
Academy of Sciences, 114(45), 11896-11901.
2716
