Developmental Psychology

© 2018 American Psychological Association 2018, Vol. 54, No. 9, 1723–1734

0012-1649/18/$12.00 http://dx.doi.org/10.1037/dev0000548

Sensitivity to the Evaluation of Others Emerges by 24 Months

Sara Valencia Botto and Philippe Rochat
Emory University

Although the human proclivity to engage in impression management and care for reputation is ubiqui-
tous, the question of its developmental outset remains open. In 4 studies, we demonstrate that the
sensitivity to the evaluation of others (i.e., evaluative audience perception) is manifest by 24 months. In
a first study, 14- to 24-month-old children (N ⫽ 49) were tested in situations in which the attention of
an audience was systematically manipulated. Results showed that when the experimenter was inattentive,
as opposed to attentive, children were more likely to explore an attractive toy. A second study (N ⫽ 31)
explored whether same-aged children would consider not only the attention of the experimenter but also
the values the experimenter expressed for two different outcomes when exploring a toy. We found that
children reproduced outcomes that were positively valued by the experimenter significantly more when
the experimenter was attentive but were more likely to reproduce negatively valued outcomes when the
experimenter was inattentive. A third control study (N ⫽ 30) showed that the significant effect of Study
2 disappeared in the absence of different values. Lastly, Study 4 (N ⫽ 34) replicated and extended the
phenomenon by showing toddler’s propensity to modify their behavior in the presence of 2 different
experimenters, depending on both the experimenter’s evaluation of an outcome and their attention.
Overall, these data provide the first convergent demonstration of evaluative audience perception in young
children that precedes the full-fledged normative, mentalizing, and strong conformity psychology
documented in 4- to 5-year-old children.

Keywords: evaluative audience perception, impression management, self-consciousness, social cog-

nition, social development

Adults report the fear of public speaking more often than any These well-documented phenomena are inseparable from what
other fear, including death (Dwyer & Davidson, 2012). This in- we refer to here as evaluative audience perception. Specifically,
triguing finding highlights a defining human characteristic: We are evaluative audience perception captures the basic human propen-
concerned with how others might perceive and evaluate us. Indeed, sity to (a) assume that one’s own behavior or appearance could be,
numerous studies show adults’ inclination to promote, enhance, or will be, evaluated by others either positively or negatively; and
and manage their own public image in the presence of others (b) have a general preference toward eliciting positive as opposed
(Leary & Allen, 2011; Mead, Baumeister, Stillman, Rawn, & to negative evaluations from others (Cooley, 1902; Goffman,
Vohs, 2011; Tennie, Frith, & Frith, 2010). For instance, adult 1959). Although much evidence documents evaluative audience
participants tend to display enhanced generosity (Dana, Weber, & perception in adults, it is unclear as to when this central psycho-
Kuang, 2007) as well as enhanced performance in trivial tasks logical trait emerges in development.
when others are watching as opposed to being alone (Cottrell, A few recent studies suggest that this phenomenon is clearly
Wack, Sekerak, & Rittle, 1968; Triplett, 1898). Strong conformity evident by 4 to 5 years of age (Banerjee, Bennett, & Luke, 2012;
to a majority judgment or opinion is also prevalent, even when Engelmann, Herrmann, & Tomasello, 2012; Leimgruber, Shaw,
such judgment is blatantly wrong or untrue (Asch, 1956; Chartrand Santos, & Olson, 2012; Piazza, Bering, & Ingram, 2011). In a
& Bargh, 1999; Cialdini & Trost, 1998). replication of Asch’s (1956) classic study, Haun and Tomasello
(2011) showed that 4-year-olds, like adults, tended to conform to
a majority opinion in public but not in private. By the age of 5,
children are more generous (Engelmann et al., 2012; Piazza et al.,
Sara Valencia Botto and Philippe Rochat, Department of Psychology, 2011) and cheat less in the presence of an observer (Engelmann et
Emory University. al., 2012). Furthermore, children also display explicit reputational
We thank all of the research assistants in the Infant and Child Lab for concerns by 5 years (Engelmann, Over, Herrmann, & Tomasello,
their help in coding. We also thank Theresa Nettles and Natalie Eldred, 2013; Fu, Heyman, Qian, Guo, & Lee, 2016; Zhao, Heyman,
Emory Infant and Child Lab Managers, for their support. We are also Chen, & Lee, 2018). Fu and colleagues (2016) showed that when
grateful to Stella Lourenco, Freddy Kamps, and Daniel Dilks for reading
primed with a reputational cue, such as telling the child that their
and providing their insights on previous versions of the manuscript. Lastly,
we are grateful to all the parents and infants who participated in this
peers think positively about them, 5-year-olds were less likely to
research as part of the Emory Child Study Center. cheat compared with when not primed. Because 4- to 5-year-olds
Correspondence concerning this article should be addressed to Sara begin to modulate their behavior depending on the presence or
Valencia Botto, Department of Psychology, Emory University, 36 Eagle absence of an audience, evaluative audience perception is typically
Row, Atlanta, GA 30322. E-mail: sara.botto@emory.edu thought to emerge around the preschool years.
1723
1724 BOTTO AND ROCHAT

However, research on the development of social emotions, like perspective, as recent evidence suggests that already by 3 years,
embarrassment, point to the possibility of an earlier emergence. children begin to show sensitivity to reputational cues (Zhao et al.,
Embarrassment expressed in coy smiles, gaze aversion, and blush- 2018). Because it is necessary to understand that others can eval-
ing is the quintessential emotional expression of the human sen- uate your behavior in order to understand aspects of reputation, it
sitivity to the evaluation of others (Keltner, 1996; Leary & Mead- would be necessary for children to develop evaluative audience
ows, 1991; Sattler, 1966). When describing blushing in his 3-year- perception prior to the age of 3.
old son, Charles Darwin (1872/1965) remarked, “It is not the Given that the cognitive “ingredients” necessary for evaluative
simple act of reflecting on our own appearance, but the thinking of audience perception are in place by 2 years, we designed four
what others think of us, which excites a blush” (p. 325). Indeed, studies investigating whether signs of this phenomenon are already
research documents embarrassment as a social emotion expressed evident by the end of infancy. Specifically, we explored whether
in situations that can potentially bring unwanted evaluations from 14- to 24-month-olds, like older children and adults, would also
others, indexing evaluative audience perception (Lewis & Ramsay, modify their behavior depending on whether or not they were
2002; Sabini, Siepmann, Stein, & Meyerowitz, 2000). being observed. In Study 1, we allowed children to play with a
Interestingly, like older children and adults, toddlers will also novel toy while systematically manipulating the attention of an
reliably express embarrassment in contexts that might bring about adult observer. We expected children to have differential behavior
potential evaluations—such as looking at mirror reflections and (i.e., show more restrain or inhibition) toward an attractive novel
discovering a mark on their face (Amsterdam, 1972; Amsterdam & toy in the attentive but not in the inattentive audience condition.
Levitt, 1980), being complimented or asked to perform in public Study 2 further probed whether children, in their exploration of the
(Lewis, Sullivan, Stanger, & Weiss, 1989), or failing to meet a toy, are not just sensitive to the attention of the observer but also
standard or goal— by 18 to 21 months (Kagan, 1981; Stipek, to how the observer evaluated possible outcomes of the toy during
Recchia, McClintic, & Lewis, 1992). Despite these observations, an initial toy demonstration. We predicted that if children engage
embarrassment in toddlers is not considered to index evaluative in evaluative audience perception by 24 months, then they should
audience perception. For example, Lewis (2011) stated that be- modulate their behavior as a function of not only audience atten-
cause toddlers lack the cognitive capacities to understand norms, tion but also audience evaluation of different outcomes, reproduc-
embarrassment prior to the third year would be a response to being ing the positively valued outcome more frequently in the attentive
an object of attention and, therefore, not yet evaluative (Lewis, condition but the negatively valued outcome more frequently in
2011; Lewis et al., 1989). Although this might be the case, few the inattentive condition. A third control experiment (Study 3)
studies have directly assessed the possibility that signs of embar- reproduced Study 2 but in a condition in which the experimenter
rassment in the second year might already have an evaluative- did not differentially evaluate the toy outcomes. Lastly, Study 4
audience-perception underpinning. extended the findings of the first three studies by probing whether
In fact, several relevant social and cognitive abilities already toddlers would factor both the differential feedback (positive vs.
established by the second year support the possibility that evalu- negative) of two experimenters toward a toy’s action as well as the
ative audience perception could emerge by 24 months. By 7 to 9 attention of each experimenter when the child proceeded to inter-
months, infants seem to acknowledge others as evaluators of the act with the toy. We reasoned that if toddlers were indeed sensitive
external world, incorporating others’ affective reactions (i.e., eval- to the evaluation of others, they should be significantly more
uations) toward ambiguous objects and circumstances into their inclined to reproduce the toy’s action when the experimenter who
own behavior (e.g., social referencing; Campos & Steinberg, 1981; assigned a positive value (as opposed to the negative value) was
Striano & Rochat, 2000). By 10 months, infants distinguish be- attentive. Together, these studies probed whether evaluative audi-
tween an attentive versus inattentive audience, engaging more with ence perception—the basic foundation of impression management
individuals who direct their gaze toward them (Jones, Collins, & and reputational concern—is already evident by the child’s second
Hong, 1991; Striano & Rochat, 2000; Tomasello, 1995). Further- birthday, pointing to a much earlier developmental origin of what
more, Repacholi and Meltzoff (2007) showed that by 18 months, is a particularly enhanced, if not a unique trait of our species
infants are less likely to imitate an action if an adult previously (Rochat, 2009).
responded with anger toward another adult who performed that
action, suggesting sensitivity to indirect emotional signals from
Study 1
others. Finally, by 18 to 21 months, infants develop an objectified
sense of self, passing the mirror mark test with, seemingly, others We tested 49 14- to 24-month-old children in a novel paradigm
in mind (Rochat, 2009; Rochat, Broesch, & Jayne, 2012). It is also (robot task). The experimenter demonstrated how to activate a novel
at this developmental juncture that infants’ first sensitivity to toy robot by pressing a button on a remote, not giving any instructions
norms and standards are reported (Kagan, 1981; Stipek et al., to the child. After the demonstration, participants were then free to
1992). In all, these capacities could, in theory, allow toddlers to play with the toy robot for 30 s while the experimenter either observed
recognize others not only as evaluators of the external environment the child (attentive condition) or turned 45° sideways while pretend-
(i.e., social referencing) but also as evaluators of their own behav- ing to read a magazine (inattentive condition; see Figure 1a). With no
ior (i.e., evaluative audience perception). In other words, by 24 clear instructions given to the child, and as index of evaluative
months, toddlers could recognize that others can have positive or audience perception, we expected children to behave differently be-
negative reactions in relation to their behavior as well, and thus tween audience conditions (within-subject comparison), showing
might now be inclined to behave in a way that would yield a more inhibition in the attentive than inattentive condition. We ex-
positive as opposed to negative response when they are being pected more inhibition in the attentive condition (i.e., refraining from
observed. This proposition would complement a developmental activating the robot) because there were no explicit cues given to the
 EVALUATIVE AUDIENCE PERCEPTION 1725

Figure 1. (a) Illustration of the three successive robot task conditions, from left to right: (A) demonstration, (B)
inattentive condition, and (C) attentive condition. The authors received signed consent from each individual for
their likeness to be published in this article. (b) Bird’s-eye-view diagram of the experimental setup. See the
online article for the color version of this figure.

child as to whether activating the toy would be positively or nega- (n ⫽ 6), or parental interference (n ⫽ 3). An a priori GⴱPower
tively evaluated by the experimenter. To measure differential behav- 3.1 analysis was run to determine the appropriate sample size,
ior between conditions, we compared the amount of button pressing which showed that a sample of 42 participants was sufficient to
between audience conditions. achieve 80% power and a medium effect size (Faul, Erdfelder,
To examine whether an early display of embarrassment by Lang, & Buchner, 2007). In a single testing session, all children
the second year is indicative of a sensitivity to other’s evalua- completed both conditions (attentive and inattentive) of the
tion, we measured the amount of embarrassment in the attentive robot task (within-subject design, order of conditions counter-
versus the inattentive condition. To provide contrast, we also balanced across participants). All studies were approved by the
coded for negative (fear), positive (happy), and control (neutral) Emory University Internal Review Board under the project
emotions. name Origins of the Intentional Stance (Study Number
IRB00041083).
Method Materials. A red circular remote-controlled toy robot with a
Participants. Participants were 49 14- to 24-month-old 13 ⫻ 13 cm diameter was used in both conditions. The robot was
(M ⫽ 18.27, SD ⫽ 2.97; 30 females) healthy children from placed in a 32 ⫻ 32-cm blue box that was covered by a yellow
predominantly White, middle-class families living in a large cloth. The 13.5h ⫻ 9w ⫻ 2d-cm remote control, which contained
urban area, recruited from a large Child Study Center database. one button on each extremity, was placed in a red slip so that only
Eleven additional children were tested but were excluded from one button was exposed and was placed to the right of the blue
the analysis because of experimental error (n ⫽ 2), fussiness box. When the button was pressed, the toy robot lit up and rotated
1726 BOTTO AND ROCHAT

in a circular motion. Both the robot and the remote were placed on display. Coders were trained on identifying emotion using the
a 92 ⫻ 76-cm moon-shaped white table (see Figure 1a and 1b). following criteria:
All sessions were video recorded for coding purposes by two • Embarrassment: combination of gaze aversion, attempts to
small video cameras. One was placed behind the experimenter to hide face, blushing, and/or coy smile.
capture a frontal view of the child and the other was placed behind • Happiness: smiling or giggling.
the child to capture both the back view of the child and the frontal • Fear: leaning back and widening eyes.
view of the experimenter. Both camera views were used to record • Neutral: no distinct emotion identified.
and code participants’ behavior, including their emotional expres-
In addition, coders recorded the level of engagement in each
sions across conditions (see Figure 1b).
task using the same Likert scale (0 – 4, with 4 being the highest).
Procedure. To get the child acclimated, the experimenter Interrater reliability for all emotion ratings in the robot task had an
interacted with the child for 5 min, exploring toys on the floor in intraclass correlation coefficient of 0.8 or higher.
the waiting area. After this brief warm-up with the experimenter,
the child and caregiver were brought to the testing room. Before
Results
the start of the study, the experimenter asked the caregiver to
remain neutral and quiet during the duration of the experiment. Robot task. A preliminary mixed factorial analysis of cova-
The child then sat on the caregiver’s lap facing a table with the riance (ANCOVA), factoring gender and order of condition as
experimental props (see Figure 1a and 1b). The experimenter sat between-subjects variables, button pressing between conditions as
78 cm directly across from the child and began the demonstration the within-subject variable, and age as a covariate, showed no
phase by lifting the yellow cloth and uncovering the toy robot on main effect of age, F(1, 46) ⫽ 1.40, p ⫽ .245, gender, F(1, 46) ⫽
the table. The experimenter then demonstrated how to activate the 3.05, p ⫽ .089, or order of condition, F(1, 46) ⫽ 2.75, p ⫽ .106.
toy robot by pressing the button on the remote control three times, There were also no significant interactions between any of these
with a 3-s lapse between each press. Contingent with the robot factors and button pressing (all ps ⬎ .05). Thus, all were collapsed
lighting up and rotating, and after each button press, the experi- in subsequent analyses. A repeated measures analysis of variance
menter smiled and said, “Isn’t that great?” We included this (ANOVA) comparing button presses between attentive and inat-
feedback after button presses because the toy robot was novel to tentive conditions showed that children were significantly more
most children and we wanted to prevent children from being afraid likely to activate the remote in the inattentive condition (M ⫽ 7.0,
of the robot. This verbal feedback remained consistent throughout SE ⫽ 0.89) compared with the attentive condition (M ⫽ 5.7, SE ⫽
the task, contingent on each button press from either the experi- 0.85), F(1, 48) ⫽ 4.86, p ⫽ .03, ␩2p ⫽ 0.09 (see Figure 2).
menter during demonstration or the child during test. After the Emotion. To examine whether self-conscious emotions would
demonstration phase, the experimenter placed the remote control change as a function of audience (attentive vs. inattentive), we
on the table within reach of the child. Immediately after placing the performed a 2 (condition: attentive vs. inattentive) ⫻ 4 (emotion:
remote down on the table, the experimenter either observed the happy, embarrassed, fear, neutral) repeated measures ANOVA.
child with a neutral expression (attentive condition) or turned 45° Results yielded a main effect of emotion, F(3, 46) ⫽ 13.82, p ⬍
sideways and pretended to read a magazine (inattentive condition). .001, ␩2p ⫽ 0.23, and audience condition, F(1, 46) ⫽ 5.23, p ⫽
Note that during the attentive condition, the experimenter looked at .027, ␩2p ⫽ 0.1. Importantly, there was a significant interaction
the child, but not in an insistent way, deviating her gaze from the between emotion and condition, F(3, 46) ⫽ 3.99, p ⫽ .009, ␩2p ⫽
child briefly throughout the condition. Both conditions lasted 30 s 0.08. Follow-up pairwise comparisons after Bonferroni corrections
each (see Figure 1a). revealed that children displayed significantly more embarrassment
Coding. All coding was done using prerecorded videos of
each task by research assistants who were blind to both conditions
and hypotheses. Modifying behavior as a function of audience in
the robot task was measured via button presses. Research assistants
recorded the number of presses in each condition using an event-
recorder software (Datavyu Team, 2014). Only those presses that
successfully activated the robot counted. A second coder coded
20% of participants (the intraclass correlation coefficient was
0.92).
Three additional research assistants coded for emotion using
preset criteria based on existing infant and toddler emotion de-
scriptions (Lewis et al., 1989; Reddy, 2000). Although we were
particularly interested in the emotion of embarrassment, we also
coded for a negative emotion (fear), a positive emotion (happy),
and a control emotion (neutral) for contrasts. For each condition,
coders used a Likert scale from 0 to 4 (with 4 being the highest)
Figure 2. Modifying behavior in the attentive versus inattentive condi-
and rated the presence of embarrassment, happiness, fear, or neu- tion as indexed by button pressing in Study 1. Bars represent standard
tral expressions in the attentive versus inattentive audience condi- error. A repeated measures ANOVA indicated that children were more
tions of the robot task. A “0” rating indicated the absence of that likely to press the remote in the inattentive versus attentive condition, F(1,
emotion within the condition, whereas “4” indicated a strong 48) ⫽ 4.86, p ⫽ .03, ␩2p ⫽ 0.09. ⴱ p ⬍ .05.
 EVALUATIVE AUDIENCE PERCEPTION 1727

in the attentive condition (M ⫽ 0.92, SE ⫽ 0.19) compared with tion across attentive versus inattentive conditions in the robot task
the inattentive condition (M ⫽ 0.31, SE ⫽ 0.12), F(1, 46) ⫽ 10.41, is an important question to explore.
p ⫽ .002, ␩2p ⫽ 0.18 (see Figure 3). In contrast, children did not A second possible explanation is that children were fearful of
display differential amounts of happy, fear, or neutral emotions publicly transgressing a rule, not pressing the button while the
between conditions (p ⫽ .195, .177, and .103, respectively). experimenter was attentive because they did not know whether
they were “allowed” to do so, thus avoiding potential punishment
Discussion and not just a negative evaluation (i.e., affective reaction). Though
related, these two motivations to inhibit behavior in the first study
By systematically manipulating the attention of the experi- are quite different. On the one hand, children could have inhibited
menter toward participants (attentive vs. inattentive), this first their behavior because they might associate touching new things
study documents 14- to 24-month-olds’ propensity to modify their without permission with a negative consequence. In this instance,
behavior when another is watching. Specifically, children showed children would simply manifest in their inhibition a generalized
more inhibition in activating an attractive toy robot in the presence association instead of an evaluative audience perception as defined
of an attentive as opposed to an inattentive audience (i.e., exper- here. A crucial characteristic of modifying behavior as a function
imenter). This pattern of behavior occurred in the absence of a of audience in the context of evaluative audience perception is that
clear instruction or standard that might have prevented or discour- behavior is changed according to what others value (i.e., approve
aged the child from activating the toy robot. Children also tended or disapprove of). For example, both adults and children are more
to display significantly more embarrassment when the experi- generous when someone is watching because generosity is gener-
menter was attentive as opposed to inattentive while interacting ally positively valued (Engelmann et al., 2012). Because there
with the remote. These findings are in line with previous works were no explicit values given by the experimenter in Study 1, it is
showing that young children tend to display self-conscious emo- difficult to discern whether children were displaying inhibition in
tions when they are the object of social attention (Lewis et al., the attentive condition as a true expression of evaluative audience
1989; Rochat, 2013; Stipek et al., 1992). perception.
Although this first study clearly shows that prior to the second Finally, although our goal for Study 1 was to create an ambig-
birthday, children already tend to modify their behavior as a uous task with no explicit norms or instruction to see what chil-
function of an attentive versus inattentive audience, important dren’s spontaneous proclivity would be, the fact that no clear
questions remain. In particular, what drove children to inhibit their instructions were given could have simply refrained children from
behavior (i.e., less button presses) when the experimenter was activating the remote in the attentive condition because they were
attentive? There are three possible explanations. One is tempera- expecting the experimenter to keep playing with the remote. After
ment, with children possibly being too shy to interact with the the experimenter turned sideways in the inattentive condition,
remote while a stranger was watching them. Behavioral inhibition children could have been more inclined to activate the remote
in the presence of others is indeed a characteristic of a shy because it was “their turn.” Again, this explanation does not
temperament (Rothbart & Mauro, 1990). Although it is not likely demonstrate young children’s sensitivity to how others might
that most children in our sample had a predominantly shy temper- evaluate their behavior. Study 2 was thus designed to have better
ament, the extent that temperament could predict relative inhibi- control over factors that might still explain the results of Study 1,
including variable temperament, fear of reprimand, lack of instruc-
tions, and absence of explicit values expressed by the experi-
menter. Replication of these results while controlling for these
variables would serve as better test of evaluative audience percep-
tion prior to the second birthday.

Study 2
Study 2 incorporated two remotes: one associated with a posi-
tive value and one associated with a negative value. During an
initial demonstration, the experimenter pressed one of the two
remotes paired with the feedback “Wow! Isn’t that great?” and
smiling (i.e., positive outcome), whereas the other remote was
paired with the negative feedback “Uh oh! Oops oh no!” and
frowning (negative outcome; see Method section for details). We
predicted that if children were indeed sensitive to another’s eval-
uation, they would selectively reproduce the positive outcome
Figure 3. Mean rating of four different emotional expressions as a when the experimenter was looking but be more inclined to re-
function of attentive versus inattentive conditions in Study 1. Error bars
produce and explore the negative outcome when the experimenter
represent standard error. Mixed factorial analyses revealed an interaction
between emotion and audience condition, F(3, 46) ⫽ 3.40, p ⫽ .009, ␩2p ⫽
was not looking.
0.19. Follow-up Bonferroni-corrected pairwise comparisons showed that To control for temperament, we asked a parent of each child to
children displayed significantly more embarrassment in the attentive con- fill out Putnam, Gartstein, and Rothbart’s (2006) short tempera-
dition, F(1, 46) ⫽ 10.41, p ⫽ .002, ␩2p ⫽ 0.18. All other emotions were not ment assessment form of the Early Childhood Questionnaire. In
significant (p ⬎ .05). ⴱ p ⬍ .05. our analysis, we were particularly interested in the dimensions of
1728 BOTTO AND ROCHAT

Fear, Inhibitory Control, Shyness, and Sociability to see whether Reliability testing for both button presses and emotion ratings
temperament would have any predictive power on change of yielded high intraclass correlation coefficients (all rs ⬎ .80).
behavior in the robot task. Finally, to reduce the ambiguity of the
first study linked to the lack of instructions, following the dem- Results
onstration phase, the experimenter pushed the remote control to-
ward the child, explicitly stating, “Your turn!” With this proce- Robot task. A preliminary mixed factor ANOVA, with gen-
dure, the child was explicitly permitted and encouraged to interact der and condition order as between-subject variables, and remote
with the remotes. and audience condition as within-subject variables, yielded no
We also coded for emotions in the attentive versus inattentive significant main effect or interactions for gender or condition order
conditions but also in relation to the positive versus negative (all were p ⬎ .05). Age was also not a significant covariate (p ⫽
remotes. We predicted that if embarrassment by the second year is .669). These factors were thus not included in subsequent analyses.
just a by-product of being the object of attention, then toddlers To see whether children chose to press the negative versus positive
should display embarrassment significantly more in the attentive remote as a function of audience, we ran a 2 (condition: attentive
condition, as they did in Study 1. However, if toddlers were to vs. inattentive) ⫻ 2 (remote: positive vs. negative) repeated mea-
reproduce positive outcomes significantly more in the attentive sures ANOVA. Results yielded a significant interaction between
condition, then we should observe reduced embarrassment, be- remote and condition, F(1, 30) ⫽ 5.02, p ⫽ .03, ␩2p ⫽ 0.14 (see
cause toddlers would be acting in a way that was in line with what Figure 4). Pairwise comparisons after Bonferroni corrections re-
the experimenter valued (i.e., no potential for a negative evalua- vealed that children pressed the negative remote significantly more
tion). Both toddlers’ strategic use of remotes and reduced embar- in the inattentive condition (i.e., when the experimenter was not
rassment would show that by 24 months, children take into con- looking; M ⫽ 4.87, SE ⫽ 1.17) than in the attentive condition
sideration both the values and the attentiveness of an observer. (M ⫽ 2.13, SE ⫽ 0.65), F(1, 30) ⫽ 4.74, p ⫽ .03, ␩2p ⫽ 0.14.
Inversely, children chose the positive remote significantly more in
the attentive condition (M ⫽ 5.26, SE ⫽ 1.46) than in the inatten-
Method tive condition (M ⫽ 2.45, SE ⫽ 0.50), F(1, 30) ⫽ 7.67, p ⫽ .01,
Participants. A total of 31 participants were included in ␩2p ⫽ 0.1. Children were thus strategic in their button pressing,
Study 2 (14 –24 months; M ⫽ 20.65, SD ⫽ 2.83; 21 females). Four considering both the values placed on each remote and the atten-
additional children were tested but were excluded from the anal- tion of the experimenter.
yses because they were fussy. An a priori GⴱPower 3.1 analysis Temperament. A preliminary correlation matrix exploring
was run to determine the appropriate sample size, which showed the relation between total button pressing and the distinct dimen-
that a sample of 30 participants was sufficient to achieve 80% sions of the temperament questionnaire (Sociability, Inhibitory
power and a medium effect size (Faul et al., 2007). Control, Fear, and Shyness) indicated that there was a significant
Procedure. Two remotes were placed side by side in a clear relation between shyness and total button presses across conditions
tray on the table, with two buttons triggering different robot (r ⫽ ⫺0.467, p ⫽ .029). However, follow-up analyses showed that
motions (one spun in circles and the other moved forward, leading shyness was not a significant covariate for positive or negative
the robot to step out of the surrounding display box). We incor-
porated these different robot motions to entice the child to explore
both remotes. To clearly differentiate the remotes, we covered one
remote with an orange sleeve and the other remote with a blue
sleeve. For the demonstration phase, the experimenter picked up
one remote, pushed its button, and gave positive feedback (“Wow!
Isn’t that great?”) and smiled, or negative feedback (“Uh oh! Oops,
oh no!”) and frowned, after setting the robot in motion. After
pushing the first remote twice, the experimenter then picked up
and demonstrated the effect of pressing the button of the other
remote, with accompanying positive or negative feedback. The
experimenter gave this feedback only during the demonstration
phase. The color of remote, movement of the robot, type of
feedback, as well as left–right location of the remote on the tray
was counterbalanced across participants. After the demonstration
phase, the experimenter smiled and pushed the tray with the two
remotes toward the child and said, “Your turn!” The experimenter
then either turned 45° and pretended to read a magazine for 30 s Figure 4. Mean button pressing as a function of condition (attentive vs.
inattentive), and both types of remote (positive vs. negative in Study 2;
(inattentive condition) or looked at the child with a neutral face for
blue vs. orange in control Study 3). Error bars represent standard error.
30 s (attentive condition), order counterbalanced.
Overall, children in Study 2 pressed the positive button significantly more
Coding. The same dependent measures (button presses and in the attentive condition, F(1, 59) ⫽ 6.17, p ⫽ .013, ␩2p ⫽ 0.1, but pressed
emotions) as in Study 1 were coded by a research assistant, adding the negative button significantly more in the inattentive condition, F(1,
in the analysis of the different remotes that were activated between 59) ⫽ 7.77, p ⫽ .007, ␩2p ⫽ 0.12. In contrast, children in Study 3 (control)
conditions (positive vs. negative). A second coder coded button did not differentially choose one remote over the other between conditions
pressing and emotion for 20% of randomly selected participants. (p ⬎ .05). ⴱ p ⬍ .05.
 EVALUATIVE AUDIENCE PERCEPTION 1729

button presses between audience conditions of the robot task, F(1, that the effects observed were because of the values that were
29) ⫽ 0.46, p ⫽ .503. given by the experimenter. We thus tested an additional control
Emotion. The analysis for emotion (embarrassment) did not group of 30 children for which no differential feedback was
yield a significant main effect or interaction (all ps ⬎ .05; see given for either of the two remotes. If toddlers were indeed
Discussion section). being strategic in their choices depending on both the values
and the attention of an audience in the experimental group
(Study 2), we would expect no differential button pressing in
Discussion
this control group because the experimenter gave no distinct
In the second study, we explored whether children would be values to either remote during the demonstration.
strategic in their button pressing as a function of audience attention
when the experimenter placed a positive versus a negative value on Study 3
each remote. We predicted that if 14- to 24-month-old children
were truly modifying their behavior in Study 1 as an expression of
evaluative audience perception, they should consider both the Participants
differential values of the remotes (positive vs. negative) as well as Thirty 14- to 24-month-old children were included in Study 3
the attention of the experimenter. Supporting our hypothesis, re- (M ⫽ 19.17, SD ⫽ 3.23; 17 males). There were no significant
sults indicated that children pressed the positive remote signifi- differences in age, gender, or temperament between the partici-
cantly more in the attentive condition but, importantly, pressed the pants of Study 2 (experimental) and Study 3 (control; p ⬎ .05).
negative remote significantly more in the inattentive condition. In
particular, children in Study 2, in contrast to Study 1, did not
simply inhibit their button-pressing behavior in the attentive con- Method
dition but were rather strategic. For example, children were not The methodology in Study 3 was identical to Study 2, except
inhibited in pressing the negative remote in both conditions but that the experimenter demonstrated each remote with no distin-
instead chose to activate the negative remote more in the inatten- guished verbal comments or values (same feedback). Following
tive compared with the attentive condition, and vice versa. Unlike the action of either remote, the experimenter said, “Oh wow!” The
the results of Study 1, these results cannot be explained by either color of remote, order of remote demonstration, as well as left–
general inhibition related to the mere presence of a stranger, task right location of the remote on the tray were counterbalanced
ambiguity, or a generalized negative association with either re- across participants. Interrater reliability for button presses and
mote. emotion ratings was high (intraclass correlation coefficient greater
The expression of evaluative audience perception in Study 2 than 0.8).
rests on the fact that children not only pressed the positive remote
significantly more in the attentive condition but, importantly,
Results
pressed the negative remote significantly more in the inattentive
condition. This pattern of behavior resembles that of older children Robot task. To explore button-pressing behavior of each re-
and adults, who tend to reproduce behavior that is positively mote between conditions, we ran a 2 (condition: attentive vs.
valued by others when others are attentive but are more likely to inattentive) ⫻ 2 (remote: blue vs. orange) repeated measures
behave in a way that might be negatively evaluated (like cheat or ANOVA. There was no main effect of remote, F(1, 29) ⫽ 0.45,
steal more) when they are not being observed. Lastly, interindi- p ⫽ .510, or audience condition, F(1, 29) ⫽ 0.21, p ⫽ .653, as well
vidual differences in temperament are not a significant factor, as as no significant interaction between remote and condition, F(1,
we found no evidence of a significant relation between tempera- 29) ⫽ 0.08, p ⫽ .782. There was also no significant effect of
ment and overall button pressing in either condition. gender, condition order, or age (all ps ⬎ .05).
As predicted, children in Study 2 showed no significant To determine whether button-pressing behavior in Study 2
emotional differences between conditions of the robot task. We differed from that of Study 3, we ran a 2 (group: Study 2 vs.
hypothesized that if children in Study 2 were to be strategic in Study 3) ⫻ 2 (condition: attentive vs. inattentive) ⫻ 2 (remote:
their button pressing, then they would show no embarrassment positive [blue for Study 3] vs. negative [orange for Study 3])
because they were behaving in a way that aligned with the mixed factorial ANOVA. Results yielded a significant three-
values set forth by the experimenter. However, if embarrass- way interaction of group, condition, and remote, F(1, 59) ⫽
ment was simply a by-product of attention, as currently pro- 5.35, p ⫽ .024, ␩2p ⫽ 0.09 (see Figure 4). As expected,
posed in the literature, then we should have observed embar- follow-up Bonferroni-corrected contrasts revealed that, unlike
rassment in the attentive condition. Because children in Study 2 Study 2, children in Study 3 had no differential button pressing
showed strategy in their button pressing (increase of positive between the two remotes in either the attentive, F(1, 59) ⫽ 0.27,
button pressing in the attentive condition), no embarrassment p ⫽ .630, or the inattentive, F(1, 59) ⫽ 0.05, p ⫽ .816,
was displayed in this context. This finding challenges the view condition. In contrast, children in Study 2 were strategic in their
that toddlers’ display of embarrassment prior to the second year button pressing, pressing the positive remote more often but the
can be reduced to the mere emotional reaction to being an negative remote less often in the attentive condition, F(1, 59) ⫽
object of attention and not an index of being sensitive to others’ 6.17, p ⫽ .013, ␩2p ⫽ 0.1, and vice versa in the inattentive
evaluation. condition, F(1, 59) ⫽ 7.77, p ⫽ .007), ␩2p ⫽ 0.12 (see Figure 4).
Although Study 2 provides strong evidence for evaluative In all, these results indicate that when the experimenter gave no
audience perception in the second year, we wanted to assure differential values to either remote (control Study 3), the selec-
1730 BOTTO AND ROCHAT

tivity demonstrated in Study 2 disappeared, providing further granting or withholding permission (i.e., without “yes” or “no”
evidence of evaluative audience perception by children younger words).
than 24 months. After each experimenter’s paired demonstrations, both experi-
Emotion. Measures of all four emotions, including embarrass- menters placed one hand on the remote and pushed the remote
ment, did not yield any significant main effect or interaction (all within reach of the child while simultaneously saying, “Your
ps ⬎ .05). turn!” E1 then proceeded to turn their back while E2 faced the
child for 30 s (negative-attentive condition). After the 30 s, the
experimenters switched conditions: E1 would turn to face the child
Discussion for 30 s while E2 turned their back (positive-attentive condition;
Results of the third control study confirm that when differential see Figure 5a). The experimenters’ side (left or right in relation to
feedback from the experimenter was removed, children did not the child), the order in which either the positive and negative value
systematically vary in their button pressing between conditions. was demonstrated first, and in which experimenter turned their
Therefore, it was indeed the positive and negative values demon- back or faced the child first, were counterbalanced across partic-
strated by the experimenter that drove the differential behavior in ipants.
Study 2. In a fourth and last experiment, we further assessed All coding was done using prerecorded videos of each task by
toddlers’ expressions of evaluative audience perception by probing research assistants who were blind to both conditions and hypoth-
whether they would differentially modify their behavior when eses. Modifying behavior as a function of audience in the robot
facing two different experimenters who expressed either positive task was measured via button presses. Research assistants recorded
or negative values toward the same action (i.e., button pressing to the number of presses in each condition using Boris, an event-
activate the robot). recorder software (Friard & Gamba, 2016). Only those presses that
Our rationale was that if, indeed, toddlers demonstrate evalua- successfully activated the robot counted. For reliability, a second
tive audience perception, they should differentially act on the coder coded 20% of randomly selected participants, yielding an
remote depending on who is watching and, specifically, whether intraclass correlation coefficient of 0.88.
this audience expressed positive or negative values toward the
remote. Unlike the previous three studies, the child would now Results
have an attentive audience throughout the task; therefore, children A mixed factorial ANOVA, with gender, two equally distributed
modifying their behavior between the two conditions could not just age groups (14 –18 months and 19 –24 months),1 and order of
depend on whether they were being observed or not but rather on condition as between-subjects variables, and condition (positive-
whether a negative or positive evaluator was observing them. attentive vs. negative-attentive condition) as the within-subject
variables, yielded no significant main effect or any significant
Study 4 interaction of age or condition order (all ps ⬎ .05). There was a
main effect of gender, F(1, 25) ⫽ 4.8, p ⫽ .03, ␩2p ⫽ 0.16, whereby
male participants tended to produce more button presses overall.
Participants More importantly, the analysis yielded a significant main effect of
condition (positive-attentive vs. negative-attentive condition), F(1,
Thirty-four 14- to 24-month-old children were included in Study
22) ⫽ 6.7, p ⫽ .016, ␩2p ⫽ 0.21. As shown in Figure 6, children
4 (M ⫽ 19.57, SD ⫽ 3.31; 19 males). Seven additional children
were significantly more likely to activate the remote in the
were tested but were excluded from the analysis because they were
positive-attentive (M ⫽ 7.7, SE ⫽ 1.04) compared with the
fussy (n ⫽ 2) or because they failed to interact with the remote
negative-attentive (M ⫽ 5.7, SE ⫽ 0.92) condition.
across both conditions (n ⫽ 5). There were no significant differ-
ences in age and gender between the participants of Studies 1 and
Discussion
2 (p ⬎ .10).
Using a different, more complex paradigm, results of Study 4
confirm the phenomenon we observed in the first three studies,
Method reinforcing the interpretation that by 24 months, toddlers do man-
The setup and materials for Study 4 were similar to that of Study ifest evaluative audience perception. Specifically, when engaging
1, except that two experimenters (E1 and E2) were involved, with a novel toy (remote) that was either positively or negatively
sitting side by side directly in front of the child (see Figure 5a, 5b). valued by two different experimenters, toddlers not only factored
To control for appearance, both experimenters were brunette fe- this value in their subsequent behavior but also considered whether
males wearing either blue or orange scrubs (color counterbalanced the positive or negative experimenter was attentive toward them.
across participants). During the demonstration phase, E1 (positive
feedback) would pick up the remote, direct the child’s attention to 1
Unlike the preceding three studies, there was an equal number of
the remote by saying “Look!” and press the button. After the robot children in both of these age groups, allowing us to use age as a between-
would spin once, E1 would smile, look at the child, and say, “Yay! subjects variable to better explore an age effect. We used 14 to 18 years and
The toy moved!” This was repeated twice. After E1’s demonstra- 19 to 24 years because this captured an equal 5-month range. Note that it
tion, E2 (negative feedback) repeated the same sequence as E1, but is also by the age of 18 to 21 months, and not earlier, that toddlers are
typically reported to show clear signs self-conscious emotions such as
would say “Yuck! The toy moved!” while looking at the child and embarrassment (Kagan, 1981; Lewis et al., 1989; Stipek et al., 1992). Age
frowning. We chose the expletives “yay” and “yuck” to match for entered separately as a covariate did not yield a significant interaction or
syllable length as well as to use words that gave values without main effect (p ⬎ .6).
 EVALUATIVE AUDIENCE PERCEPTION 1731

Figure 5. (a) Illustration of the successive phases of Study 4 from the child’s perspective. Top row depicts E1
and E2 demonstration with either positive (E1) versus negative (E2) feedback after pressing the button. Bottom
row depicts the two test conditions following demonstration with either attentive E1 or attentive E2. The remote
is within reach of the child. The authors received signed consent from both individuals for their likeness to be
published in this article. (b) Bird’s-eye-view diagram of the experimental setup for Study 4. E1 ⫽ Experimenter
1; E2 ⫽ Experimenter 2. See the online article for the color version of this figure.

This is evident in the fact that children were significantly more button pressing. This corroborates the idea that already by 2 years,
inclined to activate the remote when the positive, as opposed to the children are sensitive to how the adult might evaluate them,
negative, experimenter was attentive. In contrast, when the nega- presumably anticipating a positive or negative response to their
tive experimenter was attentive, children tended to inhibit their behavior, and thus adjusting their behavior accordingly.
1732 BOTTO AND ROCHAT

evaluative audience perception is evident before a child’s second

birthday.
Our findings add to a growing body of literature on the emer-
gence of impression management and self-conscious emotions in
early development. They indicate that evaluative audience percep-
tion, which we consider the foundation for the development of
self-presentation and reputation management documented at a later
age (3–5 years; Engelmann et al., 2012; Fu et al., 2016), emerges
already by the end of the second year. Our data suggest that from
this age on, children (a) understand that their behavior can be
positively or negatively evaluated by others (i.e., others can have
positive or negative reactions to their behavior), and (b) have a
general preference toward eliciting positive as opposed to negative
evaluations from others. This is an important developmental mile-
stone, as it suggests that by 24 months, children understand that
their own behavior can elicit affective reactions from others based
Figure 6. Mean button presses as a function of condition (positive atten-
on prior expressed values, and accordingly modify their behavior
tive vs. negative attentive) in Study 4. Bars represent standard error. A
to elicit positive as opposed to negative reactions from attentive
repeated measures ANOVA showed that children were more likely to
activate the robot when the positive experimenter was attentive as opposed others. We propose that these two early developing propensities
to when the negative experimenter was attentive, F(1, 22) ⫽ 6.7, p ⫽ .016, form the building block of the human concern for reputation and
␩2p ⫽ 0.21. ⴱ p ⬍ .05. self-presentation.
Evidence for an early emerging evaluative audience perception
has consequences at both a methodological and theoretical level.
If, by the end of the second year, children do manifest evaluative
General Discussion
audience perception, which potentially influences their behavior,
Self-image, reputation, and impression management prevails then the presence of onlookers might be a significant factor in what
from early childhood (Banerjee et al., 2012; Fu et al., 2016) and is is measured in children. Therefore, more consideration should be
a major trademark in adulthood (Schlenker & Leary, 1982; Wa- given as to how the presence of others may affect the young child’s
tling & Banerjee, 2007). Although the human proclivity to care for behavior in a task. From a theoretical standpoint, our studies
reputation and impression management is ubiquitous, the question challenge the interpretation that by the second year, the kind of
of its origins remains an open question. The present four studies embarrassment expressed by the young child is not yet self-
explored signs of an evaluative audience perception (i.e., sensitiv- conscious or evaluative proper. Accordingly, such early expres-
ity to others’ evaluations) in 14- to 24-month-olds, an age at which sions of embarrassment would correspond to the mere direct
social emotions like embarrassment emerge (Kagan, 1981; Lewis avoidant response to being an object of attention—an unfamiliar,
et al., 1989; Stipek et al., 1992). We asked whether the age at hence, uncanny, experience—and therefore not yet evaluative.
which children first manifest such emotions would also correspond With this interpretation, it would not be until later in development,
to first signs of systematic change in behavior as a function of an with first signs of guilt and shame, that a child would perceive
attentive versus an inattentive audience—an index of evaluative others as evaluators (Lewis & Ramsay, 2002). Our studies offer
audience perception. strong support for the alternative account that social emotions
In Study 1, we found that toddlers were indeed sensitive to the emerging by the second year are not simply stress responses to
presence of an audience; when an experimenter was attentive, they insistent gazing by others but can already be considered as the
tended to inhibit their proclivity to explore a novel remote- expression of a sensitivity to how others might respond to the
controlled toy. We also found that children displayed more em- child’s behavior.
barrassment when the experimenter was watching as opposed to Although our studies provide clear evidence of evaluative au-
not watching. Study 2 further supports that what might drive such dience perception in children younger than 24 months, many
differential behavior is an evaluative audience perception. We questions remain. For instance, could evaluative audience percep-
found that children strategically chose to activate a button that was tion emerge even earlier than 14 months? Because we found no
associated with a positive as opposed to a negative value from the significant age effect in our studies, we have not established the
experimenter in the attentive condition. The reverse was true in the point at which this emerges in development. Furthermore, what
inattentive condition. As a crucial control, when the experimenter might be the cognitive prerequisites of evaluative audience per-
did not assign differential values to the remotes (Study 3), we ception? And how does evaluative audience perception develop to
found that children’s choice of remote did not vary systematically become more explicit and metacognitive? Although our studies
as a function of attentive versus inattentive audience conditions. show that toddlers are sensitive to how others might respond to
Lastly, Study 4 further corroborated these findings by showing that their behavior, and engage in leveraging positive as opposed to
children not only factored whether they were being observed but negative responses, the degree to which this represents a theory of
also who was observing them. Specifically, when reproducing mind or metacognitive ability is an open question. In particular,
button presses, children factored both the experimenters’ attention more research is needed to probe how and when children begin to
and the value they had previously expressed toward the remote. understand that others’ evaluation entails not only positive or
Together, these four studies provide the first demonstration that negative reactions but also mental representations, like opinions or
 EVALUATIVE AUDIENCE PERCEPTION 1733

judgments, about the self. As it stands, our data do not provide Engelmann, J. M., Over, H., Herrmann, E., & Tomasello, M. (2013).
sufficient evidence to elucidate this critical developmental ques- Young children care more about their reputation with ingroup members
tion. Further investigation should examine how evaluative audi- and potential reciprocators. Developmental Science, 16, 952–958.
ence perception becomes more explicit in early childhood. Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). GⴱPower 3: A
flexible statistical power analysis program for the social, behavioral, and
In summary, evaluative audience perception should be consid-
biomedical sciences. Behavior Research Methods, 39, 175–191. http://
ered as an important index for the study of human social cognition,
dx.doi.org/10.3758/BF03193146
particularly in regard to self-consciousness and the concern for Friard, O., & Gamba, M. (2016). BORIS: A free, versatile open-source
reputation (Rochat, 2009). Our findings suggest that the basis for event-logging software for video/audio coding and live observations.
these features is present early in human development, and emerges Methods in Ecology and Evolution, 7, 1325–1330. http://dx.doi.org/10
prior to children passing language-based theory of mind tests or .1111/2041-210X.12584
manifesting strong conformity. As an important index of social– Fu, G., Heyman, G. D., Qian, M., Guo, T., & Lee, K. (2016). Young
cognitive development, future studies should investigate predictors children with a positive reputation to maintain are less likely to cheat.
of this early outset, how it might vary across individuals, and how, Developmental Science, 19, 275–283. http://dx.doi.org/10.1111/desc
as a function of age, evaluative audience perception develops to .12304
Goffman, E. (1959). The presentation of self in everyday life. Garden City,
become increasingly normative and metacognitive.
NY: Author.
Haun, D., & Tomasello, M. (2011). Conformity to peer pressure in pre-
References school children. Child Development, 82, 1759 –1767. http://dx.doi.org/
10.1111/j.1467-8624.2011.01666
Amsterdam, B. (1972). Mirror self-image reactions before age two. De- Jones, S. S., Collins, K., & Hong, H. W. (1991). Audience effect on smile
velopmental Psychobiology, 5, 297–305. http://dx.doi.org/10.1002/dev production in 10-month-old infants. Psychological Science, 2, 45– 49.
.420050403 http://dx.doi.org/10.1111/j.1467-9280.1991.tb00095.x
Amsterdam, B. K., & Levitt, M. (1980). Consciousness of self and painful Kagan, J. (1981). The second year: The emergence of self-awareness.
self-consciousness. The Psychoanalytic Study of the Child, 35, 67– 83. Cambridge, MA: Harvard University Press. http://dx.doi.org/10.4159/
http://dx.doi.org/10.1080/00797308.1980.11823105 harvard.9780674181410
Asch, S. E. (1956). Studies of independence and conformities. Psycholog- Keltner, D. (1996). Evidence for the distinctness of embarrassment, shame,
ical Monographs: General and Applied, 70, 1–70. http://dx.doi.org/10 and guilt: A study of recalled antecedents and facial expressions of
.1037/h0093718 emotion. Cognition and Emotion, 10, 155–172. http://dx.doi.org/10
Banerjee, R., Bennett, M., & Luke, N. (2012). Children’s reasoning about .1080/026999396380312
self-presentation following rule violations: The role of self-focused Leary, M. R., & Allen, A. B. (2011). Belonging motivation: Establishing,
attention. Child Development, 83, 1805–1821. http://dx.doi.org/10.1111/ maintaining, and repairing relational value. In D. Dunning (Ed.), Fron-
j.1467-8624.2012.01813.x tiers of social psychology. Social motivation (pp. 37–55). New York,
Campos, J. J., & Steinberg, C. (1981). Perception, appraisal, and emotion: NY: Psychology Press.
The onset of social referencing. In M. E. Lamb & L. R. Sherrod (Eds.), Leary, M. R., & Meadows, S. (1991). Predictors, elicitors, and concomi-
Infant social cognition (pp. 237–314). Hillsdale, NJ: Erlbaum. tants of social blushing. Journal of Personality and Social Psychology,
Chartrand, T. L., & Bargh, J. A. (1999). The chameleon effect: The 60, 254 –262. http://dx.doi.org/10.1037/0022-3514.60.2.254
perception-behavior link and social interaction. Journal of Personality Leimgruber, K. L., Shaw, A., Santos, L. R., & Olson, K. R. (2012). Young
and Social Psychology, 76, 893–910. http://dx.doi.org/10.1037/0022- children are more generous when others are aware of their actions. PLoS
3514.76.6.893 ONE, 7(10), e48292. http://dx.doi.org/10.1371/journal.pone.0048292
Cialdini, R. B., & Trost, M. R. (1998). Social influence: Social norms, Lewis, M. (Ed.). (2011). The self-conscious emotions. In Encyclopedia on
conformity and compliance. In D. T. Gilbert, S. T. Fiske, & G. Lindzey Early Development (pp. 11–14). Montreal, Quebec, Canada: Center of
(Eds.), The handbook of social psychology (pp. 151–192). New York Excellence for Early Childhood Development.
NY: McGraw-Hill. Lewis, M., & Ramsay, D. (2002). Cortisol response to embarrassment and
Cooley, C. H. (1902). Human nature and the social order. New York, NY: shame. Child Development, 73, 1034 –1045. http://dx.doi.org/10.1111/
Scribner. 1467-8624.00455
Cottrell, N. B., Wack, D. L., Sekerak, G. J., & Rittle, R. H. (1968). Social Lewis, M., Sullivan, M. W., Stanger, C., & Weiss, M. (1989). Self
facilitation of dominant responses by the presence of an audience and the development and self-conscious emotions. Child Development, 60, 146 –
mere presence of others. Journal of Personality and Social Psychology, 156. http://dx.doi.org/10.2307/1131080
9, 245–250. http://dx.doi.org/10.1037/h0025902 Mead, N. L., Baumeister, R. F., Stillman, T. F., Rawn, C. D., & Vohs,
Dana, J., Weber, R., & Kuang, J. (2007). Exploiting moral wiggle room: K. D. (2011). Social exclusion causes people to spend and consume
Experiments demonstrating an illusory preference for fairness. Eco- strategically in the service of affiliation. Journal of Consumer Research,
nomic Theory, 33, 67– 80. http://dx.doi.org/10.1007/s00199-006-0153-z 37, 902–919. http://dx.doi.org/10.1086/656667
Darwin, C. (1965). The expression of the emotions in man and animals. Piazza, J., Bering, J. M., & Ingram, G. (2011). “Princess Alice is watching
Chicago, IL: Chicago University Press. (Original work published 1872) you”: Children’s belief in an invisible person inhibits cheating. Journal
Datavyu Team. (2014). Datavyu: A video coding tool. Databrary Project. of Experimental Child Psychology, 109, 311–320. http://dx.doi.org/10
New York: New York University. .1016/j.jecp.2011.02.003
Dwyer, K. K., & Davidson, M. M. (2012). Is public speaking really more Putnam, S. P., Gartstein, M. A., & Rothbart, M. K. (2006). Measurement
feared than death? Communication Research Reports, 29, 99 –107. of fine-grained aspects of toddler temperament: The early childhood
http://dx.doi.org/10.1080/08824096.2012.667772 behavior questionnaire. Infant Behavior & Development, 29, 386 – 401.
Engelmann, J. M., Herrmann, E., & Tomasello, M. (2012). Five-year olds, http://dx.doi.org/10.1016/j.infbeh.2006.01.004
but not chimpanzees, attempt to manage their reputations. PLoS ONE, Reddy, V. (2000). Coyness in early infancy. Developmental Science, 3,
7(10), e48433. http://dx.doi.org/10.1371/journal.pone.0048433 186 –192. http://dx.doi.org/10.1111/1467-7687.00112
1734 BOTTO AND ROCHAT

Repacholi, B. M., & Meltzoff, A. N. (2007). Emotional eavesdropping: Stipek, D., Recchia, S., McClintic, S., & Lewis, M. (1992). Self-evaluation
Infants selectively respond to indirect emotional signals. Child Devel- in young children [Monograph]. Monographs of the Society for Research
opment, 78, 503–521. http://dx.doi.org/10.1111/j.1467-8624.2007 in Child Development, 57, 1–98. http://dx.doi.org/10.2307/1166190
.01012.x Striano, T., & Rochat, P. (2000). Emergence of selective social referencing
Rochat, P. (2009). Others in mind—Social origins of self-consciousness. in infancy. Infancy, 1, 253–264. http://dx.doi.org/10.1207/S153270
New York, NY: Cambridge University Press. http://dx.doi.org/10.1017/ 78IN0102_7
CBO9780511812484 Tennie, C., Frith, U., & Frith, C. D. (2010). Reputation management in the
Rochat, P. (2013). The gaze of others. In M. Banaji & S. Gelman (Eds.), age of the world-wide web. Trends in Cognitive Sciences, 14, 482– 488.
Navigating the social world: What infants, children, and other species http://dx.doi.org/10.1016/j.tics.2010.07.003
can teach us (pp. 205–211). New York, NY: Oxford University Press. Tomasello, M. (1995). Joint attention as social cognition. In C. Moore &
http://dx.doi.org/10.1093/acprof:oso/9780199890712.003.0037 P. J. Dunham (Eds.), Joint attention: Its origins and role in development
Rochat, P., Broesch, T., & Jayne, K. (2012). Social awareness and early (pp. 103–130). Hillsdale, NJ: Lawrence Erlbaum Associates Inc.
self-recognition. Consciousness and Cognition: An International Jour- Triplett, N. (1898). The dynamogenic factors in pacemaking and compe-
nal, 21, 1491–1497. http://dx.doi.org/10.1016/j.concog.2012.04.007 tition. The American Journal of Psychology, 9, 507–533. http://dx.doi
Rothbart, M. K., & Mauro, J. A. (1990). Temperament, behavioral inhibi- .org/10.2307/1412188
tion, and shyness in childhood. In H. Leitenberg (Ed.), Handbook of Watling, D., & Banerjee, R. (2007). Children’s understanding of modesty
social and evaluation anxiety (pp. 139 –160). Boston, MA: Springer. in front of peer and Adult audience. Infant and Child Development, 16,
http://dx.doi.org/10.1007/978-1-4899-2504-6_5 227–236. http://dx.doi.org/10.1002/icd.450
Sabini, J., Siepmann, M., Stein, J., & Meyerowitz, M. (2000). Who is Zhao, L., Heyman, G. D., Chen, L., & Lee, K. (2018). Telling young
embarrassed by what? Cognition and Emotion, 14, 213–240. http://dx children they have a reputation for being smart promotes cheating.
.doi.org/10.1080/026999300378941 Developmental Science. Developmental Science, 21, e12585. http://dx
Sattler, J. M. (1966). Embarrassment and blushing: A theoretical review. .doi.org/10.1111/desc.12585
The Journal of Social Psychology, 69, 117–133. http://dx.doi.org/10
.1080/00224545.1966.9919711
Schlenker, B. R., & Leary, M. R. (1982). Social anxiety and self- Received October 30, 2017
presentation: A conceptualization and model. Psychological Bulletin, Revision received February 28, 2018
92, 641– 669. http://dx.doi.org/10.1037/0033-2909.92.3.641 Accepted March 21, 2018 䡲