Children Expect Leaders to Oust Intruders, Refrain From Unprovoked Aggression, but Not to Be Generally Prosocial

Thomas, Ashley J.; Navarro Hernandez, Silvia; Sumner, Emily; Sarnecka, Barbara

doi:10.1525/collabra.33956

Humans in every society encounter social hierarchies. Previous research has shown that children use various cues to say who is, “in charge”. In the present studies we ask how children think those who are in charge treat others. In Studies 1 and 2, children heard stories and saw drawings of social groups where one character wore a crown and sat on a throne (Study 1) or just wore a crown (Study 2) and three other characters stood alongside with no special clothing or markings. The children were told about an action, and then had to guess who had done it. When asked who pushed someone down, most 6- to 8-year-old children guessed that one of the uncrowned characters had done the pushing, while 3- to 5-year-old children seemed to consider the crowned and uncrowned characters equally likely to push someone. When asked who kicked out a hostile intruder, children of all ages chose the crowned character more often than an uncrowned character. Study 3 asked whether children expect crowned characters to put themselves in harm’s way to protect uncrowned characters and whether they expect crowned characters to be actively prosocial. Here, neither the older nor the younger children expected the crowned characters to put themselves in harm’s way to protect the others, and they thought the crowned character was less likely than others to perform prosocial actions such as helping someone up or sharing a cookie. These data suggest that children expect leaders, at least in this context, to provide specific benefits such as expelling hostile intruders, but not to be more prosocial than other people overall. These data are consistent with the hypothesis that children use inferences about social roles to predict people’s behavior.

Humans across societies and in many different social settings find themselves in social hierarchies (Boehm, 1999; Fiske, 1992). In order to navigate hierarchies, humans must be able to recognize relative social rank and use information about rank to predict people’s behavior (Kaufmann & Clément, 2014). In practice, social rank can be grounded in different systems, broadly referred to as dominance hierarchies or prestige-based hierarchies (Fiske, 1992; Henrich & Gil-White, 2001). Hierarchies are referred to as dominance hierarchies when high-ranking people are deferred to because they are perceived to be more dangerous than others (e.g., stronger or more aggressive; Pratto et al., 1994; van Vugt & Tybur, 2014). For example, a stronger sibling might maintain control over the television because their weaker sibling does not want to risk getting hurt in a fight for the remote. This form of hierarchy occurs not only in humans but also in many other species where rank is a function of an individual’s potential to inflict harm on others (Boehm, 1999; Cummins, 2005; Huntingford, 2013; Smith & Price, 1973; van Vugt & Tybur, 2014). Dominant individuals in both human hierarchies and hierarchies in other species often maintain their rank through threat of violence. For example, in some non-human primate species, high-ranking individuals act like school-yard bullies, committing unprovoked acts of aggression which helps them maintain their social rank (Silk, 2003).

Hierarchies are referred to as prestige-based hierarchies when people are deferred to because others perceive them as able and willing to provide benefits such as knowledge, material resources or protection (e.g., Cheng & Tracy, 2014). People who hold high rank through prestige are often held to a higher standard of behavior than low-ranking people. For example, in societies that rely on hunting for food, the best hunters often enjoy more decision-making power. However, to maintain this status, the best hunters must also provide benefits by sharing most or all their meat with lower-status people (von Rueden et al., 2014). Of course, dominance and prestige cannot always be clearly distinguished. Take, for example, the relationship between an employer and an employee. While the employee may perceive their boss as someone who provides benefits in the form of guidance and resources, the employer can inflict harm on the employee by taking away income or opportunities. Because each person’s rank is defined in relation to other people, the ultimate determinants of ‘dominance’ and ‘prestige’ are the perceptions of the lower-ranking people in the relationship and their expectations about how the higher-ranking people will treat them.

Both dominant and prestigious individuals are often seen as stronger than others, but prestigious people are expected to use their strength for their group’s good, such as to protect subordinates or enforce rules (Fiske, 1992; Fiske & Rai, 2014). In contrast, dominant individuals are expected to use their strength to get what they want. The actions of dominant and prestigious individuals’ also influence perceptions of them: A dominant individual who is especially aggressive may instill fear in subordinates, whereas a prestigious individual who is notably non-aggressive may instill trust in subordinates (Fiske, 1992; Fiske & Rai, 2014).

In the current study, we ask whether children (visitors to a science museum in Southern California), have expectations about how high-ranking individuals (hereafter, ‘leaders’¹) will treat other individuals. While much prior work in developmental psychology has focused on how children understand others in terms of their traits (e.g., Heyman & Gelman, 1998; Kiley Hamlin & Steckler, 2015) or group membership (e.g., Dunham, 2018; Gelman & Taylor, 2000), less work has focused on children’s understanding of relationships (see, for example, Afshordi & Liberman, 2020; Olson & Spelke, 2008), including whether they use existing relationships to predict behavior. Thus, the first question the current studies set out to test is whether children use an existing hierarchical relationship to predict behavior. Second, these studies investigate whether children expect leaders to act in ways that agree more with dominance or prestige-based hierarchies. That is, without information about whether a high-ranking person is high-ranking because of dominance or prestige, would children nonetheless expect them to act more in line with one of these models? Thus, the actions we use in the current study were chosen with the distinction between dominance and prestige in mind, however as noted above, these types of social rank cannot always be cleanly distinguished. In line with prestige-based hierarchies we specifically ask, do children expect leaders to provide certain benefits? If so, what types of benefits? In line with dominance-based hierarchies, we ask, do they expect leaders to be responsible for unprovoked aggression? These studies can shed light on whether children are biased to see leaders, at least in this context, in line with one of these two models.

These questions build on previous work asking whether infants and children identify leaders given commonly recognized cues of social rank. Infants are sensitive to cues that relate to dominance, such as physical size and coalition size (Pun et al., 2016; Thomsen et al., 2011; Thomsen & Carey, 2013). These dominance cues also influence infants’ social evaluations: Infants avoid those who win zero-sum conflicts, suggesting that they may see ‘winners’ as dominant and thus threatening (Thomas & Sarnecka, 2019). Older children are sensitive to dominance cues as well. They use these cues to decide who is in charge. For example, when children ages 5 and 6 years are shown pictures of two people—one in a dominant posture with hands on hips and chest puffed out, the other in a submissive posture with shoulders rolled forward and hands in front of them—most children say the person with the dominant posture as being ‘in charge’ (Brey & Shutts, 2015). Related to the current question are studies suggesting that children may associate those who are in charge with antisocial actions. In these studies, children were told about an interaction that included a cue of social rank (e.g., giving permission). Children more consistently identified the high-ranking character when the interaction was bad for the subordinate (e.g., the subordinate asked if they could play, and the high-ranking character said no) than when it benefitted the subordinate (e.g., the subordinate asked if they could play, and the high-ranking character said yes; Gülgöz & Gelman, 2016). Similarly, three to five year-olds also say that a person is ‘the boss’ if they are stronger, have more resources, or win a conflict, all of which may benefit ‘the boss’ (Charafeddine et al., 2014).

These studies show that children make inferences about who is in charge. They also suggest that children may think leaders are those who benefit from interactions as opposed to those who provide benefits. In other work, children distinguish dominance from prestige (or at least distinguish aggression from dominance) from a young age. For example, toddlers ages 21 to 31 months prefer those who are deferred to in zero-sum conflicts, but not those who use force to win (Thomas et al., 2018). Toddlers of the same age expect subordinates to follow the directions of absent leaders who were deferred to, but not those who used aggression to impose their will (Margoni et al., 2018). Moreover, infants expect leaders to intervene when one subordinate does something antisocial to another (Stavans & Baillargeon, 2019). Interestingly, older children do not seem to distinguish dominance from prestige in their preferences and resource allocations (Enright et al., 2020). Here we ask whether children expect high-ranking individuals to either provide benefits to others (consistent with prestige-based ranking) or to commit aggression toward others (consistent with dominance-based ranking).

One reason to think that children may expect leaders will provide benefits is that children do expect high-ranking people to provide at least one benefit–information. When a high-ranking person and a low-ranking person disagree about the name of a novel object, children endorse the high-ranking person’s opinion (Bernard et al., 2016). However, it is not yet known whether children have general expectations about how high-ranking individuals will treat subordinates, including whether they expect them to be more or less likely to commit unprovoked acts of aggression, whether they expect leaders to provide benefits, and if so, what those benefits might be.

We start by asking the following questions: Do children expect leaders to commit unprovoked acts of aggression? Do children expect leaders to provide protection? Broadly, these studies address the question of whether children expect leaders to use their social status for the benefit of others or to their detriment. All studies were approved by the UC Irvine IRB; protocol number #2013-9945; Titled Concepts of Social Relationships. All participants were recruited from a museum in Orange County in Southern California, USA.

Study 1

Method

Participants

In Study 1, we tested 195 children between the ages of 3 and 9 years old who were visitors to a children’s museum in Orange County, Southern California. Of these, we excluded data from 12 children who did not answer one or more of the main test questions (e.g., “Who is in charge”; “Who pushed someone down”; and “Who kicked someone out”). Data from 7 other children were excluded because of interference from another child or parent, or because they had seen another child do the experiment before doing it themselves. These criteria were decided ahead of data collection.

This left 176 children in the sample (93 boys, 83 girls, by parental report). This sample included 19 three-year-olds, 32 four-year-olds, 44 five-year-olds, 34 six-year-olds, 29 seven-year-olds, 15 eight-year-olds, and 3 nine-year-olds. Our recruitment goal, decided before testing began, was to test at least 30 children each at ages four, five, six, and seven. Because we recruited participants at a museum, some three-, eight- and nine-year-olds also joined the study, but the sample sizes are smaller for those groups.

Table 1. Parent’s response when asked about child’s racial background.

Racial Background

Frequency

Percent

Valid Percent

Asian

38

21.591

38.384

Asian, Black/African American

1

0.568

1.010

Asian, white

9

5.114

9.091

Black, Native American

1

0.568

1.010

Black, Native American, white

1

0.568

1.010

Black/African American

5

2.841

5.051

Black/African American, white

3

1.705

3.030

Native Hawaiian/Pacific Islander

2

1.136

2.020

white

38

21.591

38.384

white, Native Hawaiian/Other Pacific Islander

1

0.568

1.010

No response

77

43.750

Total

176

100.000

Table 2. Parent’s response when asked about child’s ethnic background.

Ethnic Background

Frequency

Percent

Valid Percent

Another Hispanic, Latino, or Spanish Origin

2

1.136

2.222

Asian

1

0.568

1.111

Cambodian-Chinese

1

0.568

1.111

Central American

1

0.568

1.111

Colombian

1

0.568

1.111

Mexican/ Mexican-American/ Chicano

27

15.341

30.000

Mexican/ Mexican-American/ Chicano; Another Hispanic, Latino, or Spanish Origin

2

1.136

2.222

Mexican/ Mexican-American/ Chicano; Peruvian

2

1.136

1.111

Mexican/ Mexican-American/ Chicano; Puerto Rican

2

1.136

2.222

Mexican/Mexican-American/Chicano; Guatemalan

1

0.568

1.111

Not Hispanic/Latino

45

25.568

50.000

Pakistan Origin

1

0.568

1.111

Peruvian

1

0.568

1.111

Spanish

3

1.705

3.333

No response

86

48.864

Total

176

100.000

Materials

Children saw a series of illustrations depicting human characters wearing colored t-shirts. The characters were designed to prevent inferences about race or gender. Each group was introduced with a novel name, e.g., ‘the Wugs’. In the illustration we used to introduce each group there was one leader – a character sitting on a throne wearing a crown, and three non-crowned characters. We illustrated the leader this way because sitting higher than others or wearing something to make oneself appear larger or brighter is a common cue of high rank across cultures (see Fiske, 1992 for review). During pilot-testing, children consistently identified the crowned, throned character as the “leader.”

Illustrations accompanied the spoken stories. For example, the story about pushing featured an illustration of two arms in a t-shirt, pushing a character down. All the illustrations with the critical action were designed to avoid giving hints about the identity of the character doing the action. (See Figure 1 for an example and see https://osf.io/4vezr/ for all illustrations used in the study.)

View large Download slide

Figure 1. Sets of illustrations used in Stories 1a and 1b in Study 1. [Left Panel] Illustrations used in Story A [Right Panel] Illustrations used in Story B.

Procedure

Participants were recruited from among the visitors to a children’s museum during regular business hours. Experimenters approached parents on the museum floor and invited them to participate if their child was in the target age range. Parents filled out a consent form while the experimenter interacted with the child before leading them into the testing room. Before entering the testing room, parents were briefed about the procedure and asked to sit outside the testing room. The testing room had clear glass doors so that parents could watch the procedure without distracting the children or affecting their answers.

The participating child sat across from the experimenter. Before the experiment began, the experimenter introduced herself and said, “I’m going to tell you some stories and ask you some questions, ok?” Then the experimenter presented the stories in an order that was randomized by shuffling the packs of story cards. (The text of all stories is available at https://osf.io/4vezr/)

Each story began by describing a pair of novel groups (e.g., “These are the Zazzos and these are the Rookas”). All the stories included two groups because one story talked about the leader ousting an intruder from another group (see below). We introduced two groups in all conditions for consistency. In the illustrations accompanying the stories, each group wore shirts of a single color (See Figure 1). One character in each group was drawn wearing a crown and sitting on a throne. We did not explicitly label this character for the children, but for ease of reference we will hereafter refer to this character as ‘the leader.’ In Study 1, children heard four stories. Two stories described someone pushing another character down—in Story 1a, an in-group character pushed another in-group character down (“Uh oh, one of the Wugs pushed another Wug down!”) We used in-group aggression because in dominance-based hierarchies, dominant individuals often use unprovoked aggression toward in-group members to maintain rank. In Story 1b, an out-group character pushed an in-group character down, and was ousted from the group’s territory (“Uh oh, A Wug came and pushed one of the Flurps down! Look! The Wug is leaving, he got kicked out.”) Here we used out-group aggression as a relatively straightforward case of protection. The other two stories involved a character stealing a cookie: In Story 1c, an in-group character stole a cookie from another in-group character (“Uh oh, one of the Koopas stole another Koopa’s cookie!) (See Figure 3). We used an in-group transgression because high-ranking individuals often control resources in dominance-based hierarchies. In Story 1d, an out-group character stole the cookie of an in-group character and was ousted from the group’s territory (”Uh oh, a Koopa came and stole a cookie from one of the Zazzos. Look! The Koopa is leaving, he got kicked out.”) (See Figure 4). At the end of each story, we asked the children to point at the character who they thought did the action. Children’s answers were coded as ‘leader’ if they chose the character wearing the crown, and ‘subordinate’ if they chose one of the other two characters. Children’s choices were recorded with pen and paper by the experimenter administering the study, as well as by a second experimenter who was observing and recording the session on video. They compared their recorded answers at the end of each session. If their answers disagreed, a third experimenter decided the answer based on the video recording.

After the test trials came several control questions. We showed children an illustration of a new, novel group with four characters and asked (1) “Who is in charge?” (2)”Who do you think is the nicest?“* and (3)”Who would you most want to be friends with?“* The first question served as a manipulation check, to confirm that children saw the crowned character as a leader. The second and third questions checked whether children had an overall bias for or against choosing the leader.

Analysis Approach

We analyzed the data using the ‘brms’ and ‘BayesFactor’ packages in R (Bürkner, 2017; Morey et al., 2014; R Core Team, 2017). When considering whether children chose the leader more or less often than the other characters, we used the function ‘ProportionBF’ in the Bayes Factor package to compare the likelihood of the data given the null hypothesis, which was that children chose each character (Leader, Subordinate #1 and Subordinate #2) at the same rate, against the alternative hypothesis that children chose the leader more or less often than the other characters. For the questions that came after the stories about a target action (e.g., ‘pushing’), chance was ⅓ because children chose among three characters (one leader and two subordinates). For the control questions, chance was ¼ because children were choosing among four characters (one leader and three subordinates). We used default priors and these choices were decided before analysis was carried out. For ease of comparison with other published studies, we also use frequentist tests, using binom.test in R to do frequentist binomial tests.

We used brms to test whether a child’s age and gender predicted their answers to the questions that came at the end of each story. Based on pilot data we decided to split children into two age groups. For alternative analyses treating age as a continuous variable, see SM. For each of the main test questions, we asked whether age (in two categories: 3- to 5- year-olds and 6- to 8-year-olds) and gender (as reported by the child’s parent), predicted children’s answers to the dependent measure questions. To do this, we used the R package brms (Bürkner, 2017) in R (R Core Team, 2017) to fit Bernoulli Bayesian regression models with age group and gender as factors. We did not include interactions in the model and did not have random factors. We used default priors and used 6 sampling chains for 5000 iterations with a warm-up period of 1000 iterations. In cases where we found evidence that children in the two age groups answered the questions differently, we report the answers from children in the two age groups separately. See supplemental materials for results broken down by age group and gender for all dependent variables. This analysis was pre-registered (https://osf.io/p8q7g), but only after we conducted a different analysis: During the review process, we realized our original analyses were flawed, and the authors had since taken up the practice of pre-registering all their analyses and studies.

Following recommendations by Kass & Raftery, 1995, we consider Bayes Factors of over 10 as strong evidence, over 30 as very strong evidence, and over 100 as decisive evidence, Bayes Factors between 3 and 10 are considered are considered moderate, and Bayes Factors between 1/3 and 3 are considered weak, inconclusive, or anecdotal. All data and R scripts can be found on the open science framework. In these analyses, we include all children who answered the questions in the analysis but see SM for analyses that only include children that answered, “Who is in charge?” correctly.

Results and Discussion

Story 1a. “Who pushed someone down?”

When asked which in-group character pushed another in-group character down, children chose the subordinates more often than the leader. Of the 176 children tested in Study 1, only 38 (21.59%) guessed that the leader (i.e., the crowned figure) pushed someone down, as opposed to the other two characters (BF=58.26 in favor of the alt. hypothesis that the children chose leader less than 1/3 of the time; two-sided binomial test, where the null value was ⅓, p<.001, see Figure 2).

Children chose among three characters, so chance in this case was 1/3. * = BF10 >10, ** = BF10>50, *** = BF10>1000

View large Download slide

Figure 2. Percentage of children who chose the leader.

Children chose among three characters, so chance in this case was 1/3. * = BF₁₀ >10, ** = BF₁₀>50, *** = BF₁₀>1000

View large Download slide

Figure 3. Illustration used in Story 1c where one of the group members steals from another group member.

View large Download slide

Figure 4. Illustration used in the story about one of the groups kicking someone out for stealing a cookie.

We found inconclusive evidence as to whether gender influenced children’s answers (BF=1.36 in favor of a model that does not include gender). We found weak evidence that age group influenced children’s answers (BF=3.11 in favor of a model that does include age group, see SM for analysis treating age as continuous).

Looking at the two age groups separately, we found inconclusive evidence as to whether younger children chose the leader less frequently or as frequently as the other two characters: 25 out of 95 (26.3%) of the three- to five-year-old children we tested chose the leader (BF=1.33 in favor of the null; p=.1037.) In contrast, we found strong evidence that older children chose the leader less frequently: Only 13/81 (15.11%) of the 6 to 9-year-olds chose the leader (BF=82.26 in favor of the alt. hypothesis that they chose the leader less than ⅓ of the time; p<.001). Thus, older children seemed to think that the leader was less likely to push someone down. It is unclear from these data whether younger children shared this belief or not (see Figure 2).

Story 1b. “Who kicked someone out for pushing?”

When asked who ejected an out-group character for pushing an in-group character down, most children guessed it was the leader: 120 out of 176 (68.68%) children chose the leader (BF>1000 in favor of the alt. hypothesis; p<.001). We found inconclusive evidence as to whether children’s gender influenced their answers (BF=1.03 favor of a model that includes gender). We did find strong evidence in favor of a model that included age group (BF=340.34 in favor of a model that included age group). However, both age groups chose the leader more often than the other two characters: 54 out of 95 (56.84%) of 3 to 5-year-olds chose the leader (BF>1000 in favor of the alt. hypothesis, p<.001) and 66 out of 81 (81.48%) of the 6 to 8-year-olds chose the leader (BF>1000 in favor of the alt. hypothesis, p<.001). Thus, it seems that both younger and older children expect leaders to kick someone out for behaving in an antisocial way, suggesting that children see leaders as protecting in-group members from outside aggressors (see Figure 2).

Story 1c. “Who stole someone’s cookie?”

In Story 1c (see Figure 3), when an in-group member stole another member’s cookie, we found inconclusive evidence about children’s expectations: 50 out of 176 (28.44%) of the children chose the leader (BF=1.92 in favor of the null; p=.174). We did find strong evidence against the idea that children chose the leader more often than the other characters (BF=29.14 in favor of the hypothesis that children chose the leader either as often or less often than the other two characters). Thus, children do not think leaders are more likely than subordinates to steal a cookie, but they may think that leaders are equally likely to do so. We found inconclusive evidence as to whether gender (BF=1.25 in favor of a model that includes gender) or age group affected children’s answers (BF=1.04 in favor of a model that does not include age group). We found inconclusive evidence as to whether older children chose the leader or subordinate more frequently (22/81, 27%; BF=1.78 in favor of the null, p=0.288) and weak evidence that younger children chose the leader as frequently as a subordinate (28/95, 27%; BF=2.76 in favor of the null; p=0.445).

Story 1d. “Who kicked someone out for stealing a cookie?”

In Story 1d (see Figure 4), we asked children who kicked an intruder out for stealing a cookie. Here, as in Story 1b, children were much more likely to choose the leader: 116 out of 176 (65.9%) chose the leader (BF>1000 in favor of the alternative, which is considered decisive evidence; p<.001). There was inconclusive evidence as to whether gender influenced the children’s answers (BF=1.24 in favor of a model that did include gender). There was strong evidence that age group affected their answers to this question (BF=10.61 in favor of a model that included age group as a factor). Both age groups were much more likely to choose the leader than a subordinate when answering who kicked someone out for stealing a cookie: 56 out of 95 (57.75%) 3-to 5-year-olds chose the leader (BF>1000 in favor of the alternative; p<.001) and 60 out of 81 (74.00%) 6-to-9-year-olds chose the leader (BF>1000 in favor of the alternative; p<.001).

Manipulation Check: Who is in charge?

After the children had answered the questions following the four stories, we checked whether they viewed the character who was wearing a crown and sitting on a throne as the leader. To do this, we showed an illustration of a new group of characters and asked, “Who is in charge?” (For this question and the following two questions, the children chose among four characters, so the chance level of responding was 25%.) Overall, children were much more likely to choose the leader than the other three characters: 155 out of 176 (88.00%) children chose the crowned character (BF>1000 in favor of the alt. hypothesis; p<.001).

There was inconclusive evidence as to whether gender influenced children’s answers (BF= 1.22 in favor of a model that does include gender), but there was strong evidence that age group influenced children’s answers (BF=57.28 in favor of a model that does include age group). However, both younger and older children were more likely to choose the leader than the other three characters: 78 out of 95 (82.10%) 3- to 5-year-olds chose the leader (BF>1000 in favor of the alternative; p<.001) and 77 out of 81 (95.01%) 6- to 9-year-olds chose the leader (BF>1000 in favor of the alternative; p<.001).

Control Question 1: Who is nicest?

Using the same picture, we also asked, children, “Who do you think is the nicest?” Here, there was inconclusive evidence that children chose each character equally often. This suggests that children only chose the person wearing a crown and sitting on the throne in response to specific questions. Of the 170 children who answered this question, 49 chose the leader (28.8%) (BF=2.63 in favor of the null that they chose the leader ¼ of the time). We found moderate evidence that age influenced children’s answers (BF=4.38 in favor of a model that does include age group) and inconclusive evidence as to whether gender did (BF=1.23 in favor of a model that does include gender). We found inconclusive evidence as to whether older children chose the leader more often, 28/80 (35%) of them chose the leader (BF=1.68 in favor of the alt. hypothesis that children chose the leader more than ¼ of the time), and weak evidence that younger children chose the leader as often as the other characters (28/80; 23%; BF=3.18 in favor of the null).

Control Question 2: Who would you rather be friends with?

Finally, we asked the children, “If you could be friends with anyone, who would you pick?” Here, children chose the leader somewhat more frequently than the other characters. Of the 171 children who answered the question, 56 chose the leader (34.5%; BF=2.32 in favor of the alternative hypothesis; p=.02). We found inconclusive evidence as to whether gender influenced children’s answers (BF=1.57 in favor of a model that does include gender). We found strong evidence that age group affected children’s answer to this question (BF=11.58 in favor of a model that does include age group). Older children chose the leader more often than subordinates, 32/78 (41%) chose the leader (BF=21.43 in favor of the alternative, p<.0015) while younger children chose the leader as often as the subordinates (24/93 (24%); BF=3.48 in favor of the null, p=.881).

Study 2

In Study 2 we checked whether children might be using cues in the illustration to answer the questions. Specifically, children might have thought that the leader was less likely to push another character because the leader was sitting on a throne. Or they might think that the leader was more likely to eject the intruder because the leader was physically closer to the intruder than the other characters. In Study 2, children heard Stories 1a and 1b from Study 1, but we modified the pictures slightly so that the leader was on the right instead of the left side of the picture and stood next to the other characters instead of sitting on a throne (see Figure 5).

View large Download slide

Figure 5. Example illustration in Study 2 The leader was moved to the right side of the illustration and is standing on the same level as the others.