Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

Journal of Evolutionary Psychology, 5(2007)1–4, 213–234

DOI: 10.1556/JEP.2007.1022

RELATIONSHIPS AND EMOTIONS IN CHILDREN’S


UNDERSTANDING OF SOCIAL CONTRACT
VIOLATIONS
H. CLARK BARRETT1,2*, MONIKA KELLER2, MASANORI TAKEZAWA3,2,
SZYMON WICHARY4,2
1
University of California, Los Angeles
2
Max Planck Institute for Human Development, Berlin
3
Tilburg University, Tilburg
4
Warsaw School of Social Psychology, Warsaw

Abstract. Previous research on children’s understanding of social contracts has shown that
children are able to identify violations of social contracts from an early age, that they attribute
negative feelings including anger to victims of contract violations, and that attributions of
negative (moral) feelings to violators increase with age. This study examines two questions that
have not been systematically investigated in prior work. First, do children attribute specific types
of negative emotion, guilt and anger, to violators and victims of social contracts, and do these
attributions change with age? Second, do children’s attributions of guilt and anger depend on the
social roles of the parties in the contract, including parents and offspring, siblings, and friends?
Using a bilateral social contract scenario with first and fourth graders, we found that children in
both age groups were able to identify contract violations, and attributed guilt to violators and
anger to victims of social contract violations. Although anger and guilt attributions increased with
age, they were high across both ages, and relatively unaffected by type of role relationship. These
results suggest that children understand the moral emotions associated with violations of social
contracts.

Keywords: evolutionary psychology, cognitive development, social contract reasoning,


perspective taking, social emotions

*
Corresponding author: H. CLARK BARRETT, FPR-UCLA Center for Culture, Brain and
Development, UCLA Center for Behavior, Evolution, and Culture, UCLA Department of
Anthropology, 341 Haines Hall, Box 951553, Los Angeles, CA, 90095-1553. E-mail:
barrett@anthro.ucla.edu.

1789–2082 © Akadémiai Kiadó, Budapest


214 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

RELATIONSHIPS AND EMOTIONS IN CHILDREN’S


UNDERSTANDING OF SOCIAL CONTRACT VIOLATIONS

The ability to conditionally exchange benefits is a skill that is rare in the animal
kingdom. Humans are exceptions to this rule. The capacity to engage in cooperative
exchange agreements, or social contracts, is arguably one of the reasons for our
success as a social species. In recent years, a substantial body of research has
examined social contract reasoning in children, to determine when and how the
capacity for contingent cooperation develops. This research has shown that
children, like adults, are remarkably skilled at reasoning in the domain of social
exchange.
In developmental research, social contracts are typically framed using
vignettes where a child character agrees to do something that is not desirable as a
condition for getting something. Such contracts come in two forms, unilateral and
bilateral. In unilateral contracts, a rule is imposed which the child must follow in
order to obtain a benefit, such as wearing a coat to play outside (HARRIS and
NÚÑEZ 1996). In such scenarios, only one character can violate the contract.
Bilateral contracts involve mutual exchange of benefits, such as tidying a room in
order to get a bicycle (KELLER et al. 2004). In such scenarios, either party could
violate the contract by failing to provide the agreed-upon benefit. Previous research
has shown that adults are particularly good at detecting violations of social contract
agreements, even though they are not good at detecting violations of rules that are
not framed as social contracts, suggesting a special competence in the domain of
social exchange (COSMIDES 1989; GIGERENZER and HUG 1992). A variety of
studies have shown that this is a competence that develops early in childhood, as
young as age 3 (CUMMINS 1996; HARRIS 2000; HARRIS and NÚÑEZ 1996; NÚÑEZ
and HARRIS 1998).
Social contract reasoning in adults is studied using the four-card Wason
selection task. This task asks subjects to consider cards that represent one part of the
outcome of a social contract scenario. In the tidying room / bicycle scenario
described above, for example, cards would show a tidy room, a not tidy room, the
child getting a bicycle, and the child not getting a bicycle. Subjects are then asked
to select cards depicting potential violations of the contract. Simplified versions of
this task have been developed for use with children. Girotto and colleagues
(GIROTTO et al. 1989; LIGHT et al. 1990) showed that children aged 7 to 10 were
able to select cards representing violations of authority-imposed deontic rules (e.g.,
“all buzzing bees must stay in the hive”; GIROTTO et al. 1989; LIGHT et al. 1990).
CUMMINS (1996) developed a very simple version of the task in which children
were given the rule “all squeaky mice must stay in the house,” and shown toy mice
that had to be squeezed to see whether they were squeaky or not. Children as young
as age 3 knew that only the mice outside had to be squeezed, suggesting
development of the capacity to detect violations of social contracts by age 3.
HARRIS and NÚÑEZ (1996) showed that children could identify violations in

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 215

seemingly arbitrary rule scenarios, such as having to wear a helmet in order to


paint. A follow-up study replicated this result across cultures, with Colombian
children, and showed that children at this age are able to distinguish between
accidental and intentional violations of rules (HARRIS 2000; NÚÑEZ and HARRIS
1998).
In the adult literature, there is debate over the particular contexts that facilitate
reasoning. For example, COSMIDES (1989) argues that adults show a particular
competence in detecting violations of social contract rules, but CHENG and
HOLYOAK (1985) argue that the competence is broader, extending to deontic rules
(rules involving permission and obligation more generally). To some degree, this
issue is glossed in the child literature, because the rules involved could be construed
as social contracts or deontic rules depending on the framing. For example,
authority-imposed rules such as “if you play outside, you must wear a coat” could
be interpreted simply as unilateral deontic rules (the child is required by an
authority to perform a certain action), or as a bilateral social contract (the parent
offers the child the benefit of being able to play in exchange for the benefit of
knowing that child will not get sick). Typically, social contracts are agreed upon by
two parties rather than being imposed unilaterally by an authority figure. To
examine whether children could reason about social contracts between peers,
HARRIS et al. (1999) constructed scenarios in which two children agreed to
exchange toys. Children age 3 and 4 succesfully identified violations on this task,
and the result was replicated cross-culturally, in Nepal.
In addition to examining peer–peer contracts, HARRIS et al. (1999) examined
another important feature of social contract reasoning that has been demonstrated in
adults: perspective switching. GIGERENZER and HUG (1992) showed that when
given social contracts involving bilateral cheating options, adults switched their
choice of cards indicating a violation depending on the perspective of the party to
which they were cued. HARRIS and NÚÑEZ (1997) and HARRIS et al. (1999) showed
that 3 and 4 year olds are also competent at reasoning about bilateral contracts.
Children were able to correctly identify who was “naughty” depending on which
outcome they were shown (party A versus party B cheating). Using the Wason four
card task, JANKE (1999) showed that older children, aged 10 to 15, were able to
successfully detect violations of bilateral contracts.
These studies suggest that there is a moral component to children’s reasoning
about social contracts. The fact that young children are able to correctly identify the
“naughty” party in social contract violations suggests that children view violations
of social contracts as morally wrong. Even young children appear to construe social
contracts as agreements that there is a moral obligation to honour, and parties that
break them are morally wrong (HARRIS and NÚÑEZ 1997; HARRIS et al. 1999). An
entailment of the moral nature of contracts is that people might feel specific moral
emotions when they break them. An important part of competence in social
exchange interactions is the ability to anticipate what others might feel if the
contract is broken. This competence helps children to avoid moral breaches and to

JEP 5(2007)1–4
216 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

make up for them if they have occurred intentionally or unintentionally (KELLER


1984).
There is a substantial literature on the development of moral reasoning and
feelings, exploring the link between children’s reasoning about morality and about
emotions such as happiness, sadness, anger, and guilt (ARSENIO and LEMERISE
2001; KELLER et al. 2003; NUNNER-WINKLER and SODIAN 1988; KOHLBERG 1976;
NÚÑEZ 1999; SMETANA 1995; SY et al. 2003; YOUNISS 1980). To date, relatively
little work has been done explicitly linking social contract reasoning with reasoning
about morality and emotions. Exceptions include the work discussed above by
HARRIS et al. (1999) and NÚÑEZ (1999), which explicitly asked children to identify
“naughty” parties and began to explore the emotions associated with social contract
violations, including anger.
Previous research has examined positive and negative emotions associated
with the violation of moral rules. NUNNER-WINKLER and SODIAN (1988) showed
that most 4-year-olds attributed positive emotions to violators of moral rules, but
most 8-year-olds attributed negative emotions to violators, suggesting a deve-
lopmental change in the evaluation of moral wrongdoing. In particular, attribution
of negative emotions to victims appears to develop early, whereas attributions of
negative emotions to violators develops later. SY et al. (2003) also showed that
young children often attribute “happy” feelings to violators of prosocial obligations,
though attribution of negative feelings to victims develops early. NÚÑEZ (1999)
showed that three-year-olds attribute anger to victims of social contract violations.
KELLER et al. (2003) examined children’s attributions of feelings to both
violators and victims of broken promises and supported the finding that whereas
attribution of negative feelings to a victim of a violation develops early, attribution
of negative feelings to violators occurs relatively later in development. More
recently, KELLER et al. (2004) conducted a study in which children ranging from 3
to 10 years evaluated bilateral social contracts and were asked to judge the emotions
of the parties involved, as well as to identify contract violations. In one scenario the
social contract was between a parent and a child and in another it was between two
peers. Keller et al. found that children were able to identify contract violations
correctly for both mother–child and peer–peer contracts. They also found,
consistent with earlier studies, that children of all ages attributed negative feelings
to the victim of contract violations, but the attribution of negative feelings to the
violator of the contract was more frequent in older children.
Interestingly, KELLER et al. (2004) also found an effect of type of relationship
on emotion attribution to the violator. Both younger and older children attributed
positive feelings to the child as contract violator more frequently in the context of a
mother–child contract than in the context of a peer–peer contract. Keller et al.
suggest that this might be because mother–child relationships differ from
relationships between unrelated peers in that they are certain to persist despite
contract violations, whereas peer–peer relationships are not. Additionally, KELLER
et al. (2003) found that in all age groups, negative feelings were attributed more

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 217

frequently to the self than to the hypothetical transgressor, consistent with research
on self–other attributions in social psychology. For example, EPLEY and DUNNING
(2000) show that people rate themselves less selfish and more generous than others,
even when they do not differ in their behavior.
A feature of existing work on children’s reasoning about emotions and social
contracts is that, with few exceptions (e.g. NÚÑEZ’S (1999) work on anger), it has
not examined children’s judgments about specific emotions such as happiness,
guilt, sadness or anger, instead focusing only on the distinction between “positive”
emotions (“feeling good”) and “negative” emotions (“feeling bad”). This matters
because emotions within the positive and negative categories can have very distinct
functions. For example, sadness, anger, and guilt are all regarded as “negative”
emotions, yet they have very different functions in the regulation of social behavior.
A mature understanding of social contract relationships likely entails the ability to
anticipate the particular emotions, as well as possible behaviors, of both social
contract violators and victims. In particular, the understanding that victims of social
contract violations may feel angry, and that violators may feel guilty, might be
important in regulating one’s own behavior in the context of social contract
interactions (FESSLER and HALEY 2003). In addition, the specific reactions that
individuals will have to the violation of a contract may depend on the features of
those individuals, including their relationship vis-à-vis each other. For example,
there might be differences in reactions by the violator and the victim depending on
whether the parties differ in degree of relatedness (e.g. siblings, friends) or authority
role (e.g. parent and offspring). These differences have not been systematically
examined in prior research. With these considerations in mind we designed the
study reported here.

The Present Study

Our objectives were twofold. First, with the exception of Núñez’ work on anger,
prior studies of children’s reasoning about emotions in social contract violations
have looked only at “positive” and “negative” emotions. The negative emotions that
could be associated with contract violations include anger, guilt, and sadness. While
these are all “negative” emotions, they have very different functions in regulating
moral behavior. Guilt is an emotion normally attributed to moral norm violators, not
victims, and anger is just the opposite (GIBBARD 1990). If children did not
distinguish between these negative emotions, and were just as likely to attribute
either one to either party after the violation of a contract, this would exhibit poor
understanding of the relationship between moralized acts and specific emotions.
Therefore, in this study, we examined children’s attributions specifically of guilt
and anger.
Second, KELLER et al. (2004) showed that the pattern of attribution of
emotions differed depending on the nature of the relationship between the parties in
the social contract. Keller et al. examined two scenarios, one involving a mother–

JEP 5(2007)1–4
218 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

child relationship, and another involving a peer–peer relationship. In the mother–


child scenario, children attributed fewer negative feelings and more positive
feelings to the child as violator than they did in the peer–peer scenario. When the
mother was the violator, they attributed more bad feelings to her than they did to the
violator in the peer–peer scenario. Thus, children judge that kids do not feel as bad
when violating a deal with their mother as they do when the deal is with a friend,
and they judge that mothers feel particularly bad when violating a deal with their
own child. However, there are two possible factors that could account for this
asymmetry, which were confounded in the prior study: kinship and power. In kin
relationships, giving is often non-reciprocal. For evolutionary reasons, parents often
invest in offspring without expecting reciprocal giving in exchange, because the
benefit comes in the form of helping one’s own genes, or “inclusive fitness”
(HAMILTON 1964). Parents might be less inclined to punish contract violations on
the part of their own offspring, and they almost never discontinue a relationship
because of lack of compliance with an agreement on the part of their child. This
may explain why children do not expect guilt on the part of a child who takes a
benefit from a parent without reciprocating. However, the logic of inclusive fitness
also holds for other kin relationships, such as those between siblings. Even though
siblings may be “peers,” their relationships may differ from those of non-kin peers,
such as unrelated friends. For example, like parent–offspring relationships, sibling
relationships are not as likely to end following a broken agreement as are friend-
ships. In turn, while sibling–sibling and parent–offspring relationships are both kin
relationships, they are not the same in terms of power asymmetries. Frequently
parents transfer resources to children because parents are more competent at
obtaining resources, such as food, clothes, or toys, and these resources are valued
differently by children and adults. For siblings of similar age, on the other hand,
resources such as toys are more equally valued. Similar power and similar
preferences lead to a greater potential for conflict and need for fair outcomes.
These two dimensions, which we will call power (adults versus peers) and
kinship (kin versus unrelated individuals), have not been disentangled in previous
research. KELLER et al. (2004) compared a parent–child contract scenario to a peer–
peer contract scenario. In the parent–child scenario, the parent is both related to and
more powerful than the child. In the peer–peer scenario, the children are both
unrelated to each other and equal in power. Furthermore, Keller et al. presented
different contracts in the mother–child and peer contexts. Therefore, we designed a
study with a 2×2 design of power and kinship, holding type of contract constant, to
determine how the specific nature of social role relationships impacts children’s
emotion attributions.
Our study used the same basic design as KELLER et al. (2004). Children were
presented with a vignette about a bilateral contract, in this case, tidying the garden
in exchange for being able to play a computer game. They were then shown pictures
of different possible outcomes (garden tidy / not tidy, computer game / no computer
game). This allowed us to assess violation detection, and also to ask questions about

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 219

the emotional reactions of different parties in the scenario. In addition, the nature of
the relationship between the parties in the social contract was systematically varied
in two dimensions, power and kinship, in order to examine effects of relationship
type. We investigated two age groups, 1st graders and 4th graders. The age range of
these children, approximately 6 years and older, made us confident that they would
possess the basic social contract reasoning skills that have been shown in other
studies to develop by age 3. Because previous research indicated that changes in
reasoning about emotions and moral wrongdoing occur in this age range (KELLER
1984; KELLER et al. 2004; NUNNER-WINKLER and SODIAN 1988), we expected that
we might observe developmental differences between the two age groups.
Our predictions were as follows. Based on previous work (CUMMINS 1996;
HARRIS and NÚÑEZ 1996; KELLER et al. 2003), we expected children across age
groups and conditions to be successful at judging contract violations. We also
expected them to correctly distinguish between different negative emotions
associated with contract violation. Specifically, we expected children to attribute
guilt to contract violators and anger to victims. However, based on previous studies
(ARSENIO and LOVER 1995; KELLER et al. 2003; NUNNER-WINKLER and SODIAN
1988), we expected that older children would attribute guilt more frequently, and
that attribution of anger to the victim would be more frequent than attribution of
guilt to the violator. Additionally, we anticipated possible perspective effects, in
particular, on guilt. Our experiment included a perspective manipulation in which
children were cued to the perspective of a protagonist. In some scenarios, it was this
character who violated the rule. In other scenarios, the other character violated the
rule, in which case the self character would be a victim. Previous studies suggested
that the perspective to which children were cued might affect their attribution of
emotions, for example, with more frequent attribution of guilt when the self
character was the violator (KELLER et al. 2003). Finally, whereas KELLER et al.
(2004) showed effects of relationship type (parent / offspring versus friends) on
children’s judgments, they could not determine whether this was due to kinship or
power differences between the relationships. We manipulated these factors
independently to determine which, if either, is responsible for the effect.

METHOD

Participants

Two groups of children were recruited for the study from a public school in Berlin,
Germany. The two groups consisted of 40 first graders (age M = 6.6 years, range 6–
8 years) and 40 fourth graders (age M = 9.9 years, range 9–11 years), about equally
distributed according to sex. Participants came from middle- to upper-middle-class
families. The ethnicity of participants was approximately 90% German and 10%
Turkish.

JEP 5(2007)1–4
220 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

Materials and Procedure

Participants were interviewed individually, by three native German-speaking


research assistants. Children were told that they would see pictures describing a
story about either a little boy (Simon) or a little girl (Simone). The sex of the story
protagonist was matched to the sex of the participant. The experimental design
included four different relationships. Power and relatedness were varied in a 2 × 2
design (high and low power and related / not related). Because of the number of
conditions and sample size considerations, we manipulated relatedness as a
between-subjects variable, and power as a within-subjects variable. Participants
were randomly assigned to one of two between-subjects conditions: the “related”
condition, in which the other character in the story was the protagonist’s same-sex
sibling or mother, respectively; or the “unrelated” condition, in which the other
character in the story was the protagonist’s same-sex friend or a neighbor,
respectively. Within these conditions, each participant saw two power conditions,
counterbalanced: one about the protagonist interacting with another child, and one
with an adult. Each child was thus presented two stories, representing different
situations.
The full text of the social contract scenario is given in Appendix 1. In the
story, the female or male protagonist makes a deal with the other character (we will
use the name of the male protagonist, Simon, and the term “other person” to refer to
the other character in the story). The other person (mother, sibling, adult neighbor,
or friend) wants to have the garden tidied and offers to allow Simon to use a new
video game in exchange for this work. After an introductory passage framing the
story, the other person proposes this deal to Simon as follows (translated from
German):
“It would be nice if you would clean the garden for me. Then I have a new video game that
you may play. But, if you want to play the video game, you must first clean the garden.”

The contract was thus bilateral, with the potential for cheating on both sides:
Simon could cheat if he used the game without cleaning the garden, and the other
person could cheat if Simon cleaned the garden but was not given access to the
game (participants were not explicitly told this). After presentation of the scenario,
participants were shown examples of the four logically possible combinations of
outcomes (garden clean/not clean, video game used/not used), presented on picture
cards (see Appendix 2) as a warm-up. We showed participants all four cards so that
they would understand that all possible outcomes of the scenario could occur, and
that characters’ behavior was not constrained by the agreement.
After this warm-up, children were asked for their judgments about all possible
outcomes, presented in random order. For each card, we asked whether or not the
contract was broken in that situation, and then we asked specific questions about the
emotional reactions of the characters. The target questions were: violation

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 221

judgments (“was the deal broken here?”), attributions of anger to victims (“does
Simon/the other person feel angry?”), and attributions of guilt to violators (“does
Simon/the other person have a bad conscience?”). Note that in the two unilateral
contract violations depicted in Appendix 1, Simon was the violator and the other
person was the victim (garden not clean/video game used). In the other unilateral
violation scenario, the converse was true (garden clean/video game not used).

RESULTS

Judgments of Contract Violation

Figure 1 shows the frequency of children’s contract violation judgments for each
type of relationship pooled across all other factors. In the figure, four outcomes are
represented in the following way: “P” is tidying the garden and “Q” is playing the
video game. The four outcomes are therefore mutual compliance (P and Q), self as
violator (not P and Q), other as violator (P and not Q), and mutual non-compliance
(not P and not Q).
100
Judgment of contract violations (%)

80 Friend
Sibling
60 Neighbor
Mother

40

20

0
Mutual Self’s Violation The Other’s Mutual Non-
Compliance (P & D) (not P & Q) Violation (P & not Q) compliance
(not P & not Q)
Figure 1. Percentages of children who judged that a contract was violated

In general, children were able to judge violations for all three kinds of
violations correctly. Average proportions of violation judgment were 96.8% for self
(protagonist) violating the rule, 81.9% for the other person violating the rule, and
82.6% for both parties violating the rule. In contrast, when both parties complied
with the rule, few children rendered violation judgments (mean percent of violation
judgments: 6.9%).

JEP 5(2007)1–4
222 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

Hierarchical log-linear analyses were performed to look for effects of grade,


sex, and relationship scenario on children’s judgments. First, hierarchical log-linear
analyses were conducted to find the best model, using p > 0.05 as a significance
criterion. If the best model produced a significant effect, further log-linear analyses
were conducted to obtain parameter estimates for individual effects.
We examined the following model: Judgment of contract violation (2) × Sex
(2) × Grade (2) × Type of relationship (4). The same model was examined for each
of the three outcomes. Results of the model were as follows. For mutual compliance
(P and Q), the final model was Judgment × Sex, L2 = 14.7, df = 28, n.s. Log-linear
analysis further revealed that the estimated parameter of this effect was marginally
significant (z = 1.87, p < 0.06). Female participants were more likely to misjudge
that a contract was broken when it was not (10.8% of girls, n = 83, versus 2.6% of
boys, n = 78).
We also looked for possible perspective effects on children’s judgments. The
two target unilateral violations were included in the hierarchical log-linear analyses:
self violates (not P and Q) and other violates (P and not Q). The initial model
included Violation Judgment (2) × Type of Violator (2) × Sex (2) × Grade (2) ×
Relationship (4). The final model was: Type of Violator × Violation Judgment, L2 =
35.2, df = 60, n.s. Log-linear analysis further showed that the estimated parameter
of this effect was statistically significant (z = 3.9, p < 0.01), indicating that children
are more likely to judge that the contract was broken when self is violator (96.9% of
children judged the contract was broken, n = 161) than when other is violator
(82.0% violation judgments, n = 161).

Attribution of Emotions

In scenarios in which Simon (self) was a violator of the social contract (Figure 2a),
a majority of children attributed guilt to violators of social contracts (first graders:
64.0%, n = 75; fourth graders: 82.6%, n = 86; pooled: 73.9%). An even greater
majority of children attributed anger to victims of social contract violations (first
graders: 84.0%, n = 75; fourth graders: 96.5%, n = 86; pooled: 90.7%).
In scenarios in which the other party was a violator of the social contract
(Figure 2b), the pattern of attributions was slightly different. A majority of children
attributed guilt to violators, though fewer than when self was the violator (first
graders: 54.1%, n = 74; fourth graders: 72.9%, n = 85; pooled: 64.2%). A majority
of children attributed anger to victims of social contract violations, but again, fewer
than when self was the violator (first graders: 78.7%, n = 75; fourth graders: 91.9%,
n = 86; pooled: 85.7%).
Thus, a majority of children in all age groups attribute guilt to violators and
anger to victims of social contract violations. However, there appear to be effects of
both age (grade) and perspective, with more guilt and anger attributions being made
by fourth graders than first graders, and with a higher frequency of attributions

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 223

when self is the contract breaker than when the other person violates the contract.
To examine these effects, we conducted log-linear analyses for each emotion, anger
and guilt, separately. The hierarchical model for anger was Anger (2) × Age (2) ×
Sex (2) × Type of violator (2; self/other) × Relatedness (2; kin/non-kin) × Power (2;
adult/child). The final best-fit model was: Anger × Power, Anger × Age, L2 = 33.9,
df = 58, n.s. Parameter estimates from the final model were Anger × Power, z = 2.0,
p < 0.05; Anger × Age, z = 3.4, p < 0.01.
Self As Violator
(a)

100
1st Grade
Attribution of emotions (%)

80 4th Grade

60

40

20

0
Guilt to Violator Anger to Victim
The Other as Violator
(b)

100
Attribution of emotions (%)

1st Grade
80 4th Grade

60

40

20

0
Guilt to Violator Anger to Victim

Figure 2. Percentages of children who attributed guilt to a violator and anger to a victim either (a)
when self violated the contract or (b) when other violated the contract

JEP 5(2007)1–4
224 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

These results (Table 1) indicate that children were more likely to attribute
anger to the victim when the other person was a child (91.9%, n = 163) than when
the other person was an adult (84.6%; n = 162). Furthermore, older children were
more likely to attribute anger to the victim (94.2%, n = 172) than younger children
(81.3%, n = 153).
In the hierarchical log-linear analysis of children’s attributions of guilt, the
final best-fit model was complex: Guilt × Sex × Age × Relatedness, Guilt × Type of
violator × Sex × Relatedness, Guilt × Type of violator × Sex × Age, Guilt × Type of
violator × Sex × Power; L2 = 11.0, df = 28, n.s. Because of the complexity of this
model, and because each interaction involved sex, males and females were analyzed
separately, with the following variables entered into each analysis: Guilt × Type of
violator × Relatedness × Power.

Table 1. Percentages of attributions of anger

Type of relationship
Grade
Child Adult All
1 85.7 77.6 81.3
(n = 77) (n = 76)

4 97.7 90.7 94.2


(n = 86) (n = 86)
Total 91.9 84.6 84.9

For boys, the best-fit model was: Guilt × Type of violator, Guilt × Age; L2 =
15.9, df = 26, n.s. Parameter estimates for the final model were Guilt × Type of
violator, z = 2.1, p < 0.05; Guilt × Age, z = 3.7, p < 0.01. Boys were more likely to
attribute guilt when self violated the contract (78.2%, n = 78) than when the other
person violated the contract (62.3%, n = 77). Older boys were more likely to
attribute guilt (82.8%, n = 87) than younger boys (54.4%, n = 68; see also Table 2).

Table 2. Percentages of attributions of guilt by boys

Grade
Type of violator
1 4 Total
70.6 84.0 78.2
Self
(n = 34) (n = 44)
38.2 81.4 62.3
Other
(n = 34) (n = 43)
Total 54.4 82.8

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 225

(a)
(b)
100 1st Grade 1st Grade
100
Attribution of Guilt (%)

Attribution of Guilt (%)


4th Grade
80 4th Grade
80
60
60
40
40
20
20
0
0
Kin Non-kin
Self Other person
Relatedness
Type of violator

(c)
100
1st Grade
Attribution of Guilt (%)

80 4th Grade

60

40

20

0
Self & Kin Self & Non-kin Other person & Other person &
Kin Non-kin
Type of violator x Relatedness

Figure 3. Percentages of attributions of guilt by girls. (a) Relatedness by grade; (b) Type of
violator by grade; (c) Relatedness by type of violator by grade

For girls, the best fit model was: Guilt × Type of violator × Age, Guilt ×
Relatedness × Age, Guilt × Type of violator × Relatedness; L2 = 8.8, df = 18, n.s.
Parameter estimates for the final model were Guilt × Relatedness; z = 2.7, p < 0.05;
Guilt × Relatedness × Age; z = 2.1, p < 0.05; Guilt × Type of violator × Age; z =
2.2, p < 0.05; Guilt × Type of violator × Relatedness; z = 2.2, p < 0.05. The nature
of these effects was as follows. Guilt × Relatedness: Girls were more likely to
attribute guilt when self violated the contract with non-kin (76.1%, n = 88) than
with kin (58.4%, n = 77). However, the Guilt × Relatedness × Age interaction
reveals (Figure 3a) that this effect obtains only in the fourth-grade girls who
attributed guilt more frequently when the other person was non-kin (86.4%, n = 44)

JEP 5(2007)1–4
226 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

than kin (57.5%, n = 40; χ2 (df = 1) = 8.9, p < 0.01). Guilt × Type of violator × Age
(Figure 3b): Fourth-grade girls were also more likely to attribute guilt when self
violated a contract (81.0%, n = 42) than when the other person violated a contract
(64.3%, n = 42; χ2 (df = 1) = 2.9, p < 0.07), but first-grade girls were not. Guilt ×
Type of violator × Relatedness (Figure 3c): For girls, when the violator is self,
kinship matters, with more attributions of guilt when self violates a contract with
non-kin (84.1%, n = 44) than with kin (53.8%, n = 39; χ2 (df = 1) = 9.0, p < 0.01).
This effect was not found for attributions of emotion to the other person.

Comparison of attribution of anger and guilt

We examined if attribution of anger is more frequent than attribution of guilt. First,


we conducted a hierarchical log-linear analysis on Attribution (2; yes/no) × Type of
emotion (2; anger/guilt) × Sex (2) × Age (2). The saturated model, which includes
all main and interaction effects, was selected as the best fit model. As all effects
were included, the indicator of the goodness of fit (L2) becomes zero. Log-linear
analysis revealed two interaction effects regarding the difference between anger and
guilt attributions. The interaction effect of Attribution × Emotion (z = 5.80, p <
0.01) showed that, across age and sex of children, attribution of anger to the victim
was more frequent than attribution of guilt to the violator (see Table 3). The effect
of Attribution × Emotion × Sex × Age (z = 2.51, p < 0.05) indicates that younger
boys were the least likely to attribute guilt (Table 3).

Table 3. Percentages of anger and guilt attributions

Grade
1 4
Anger 74.4 95.2
Girls (n = 82) (n = 84)
Guilt 63.0 72.6
(n = 81) (n = 84)
Anger 89.7 93.2
Boys (n = 68) (n = 88)
Guilt 54.4 82.8
(n = 68) (n = 87)

The Relationship between Attributions of Emotion and


Judgments of Contract Violation

We computed chi-square and phi coefficients to assess whether there was a


relationship between judging that a particular outcome constituted a contract

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 227

violation and the attribution of emotions to violators and victims in that outcome.
One might expect there to be such a relationship if attribution of the appropriate
emotions is contingent upon understanding that a contract has been broken.
Chi-square and phi coefficients were calculated on Judgment of contract
violation × Judgment of anger (or guilt) for each of the unilateral contract violation
situations: not P and Q (self violates) and P and not Q (other violates). Phi
coefficients reach either 1 or –1 if all cases fall in diagonal cells in 2 × 2 tables.
Thus, we can use phi coefficients as a measure of correlation, that is, an index of
how strongly judgment of contract violation coincides with attribution of anger (or
guilt).
For both self and other as violator, correlations between violation judgments
and attribution of the appropriate emotions were significant. In scenarios involving
contract violations by self (not P and Q), judgments that a violation had occurred
were significantly correlated with attributions of guilt to self, that is, the Simon
character (χ2 (df = 1) = 14.6, p < 0.001; phi = 0.30, p < 0.001), as well as
attributions of anger to other (χ2 (df = 1) = 30.5, p < 0.001; phi = 0.44, p < 0.001).
Similarly, in scenarios involving contract violations by the other person (P and not
Q), judgments that a violation had occurred were significantly correlated with
attributions of anger to self (χ 2 (df = 1) = 40.5, p < 0.001; phi = 0.50, p < 0.001), as
well as guilt to the other person (χ2 (df = 1) = 22.7, p < 0.001; phi = 0.38, p <
0.001).

Discussion

The main results of this study were twofold. First, the study demonstrated that
children judge that violators of social contracts feel guilt, and that victims feel
anger. While previous research had shown that children associate negative
emotions, including anger, with contract violation (KELLER et al. 2004; NÚÑEZ
1999; NUNNER-WINKLER and SODIAN 1988), this is the first study to systematically
show that distinct negative emotions are attributed to the victim and violator of
social contract agreements. The attribution of these emotions depends on children’s
judgments of particular outcomes as violations of contracts, suggesting that there is
a moral element to attributions of guilt and anger. In addition, a variety of factors
were found that modulate the attribution of these emotions, including sex, age, and
the perspective to which children are cued.
The second major result is that kinship and power have small but distinct
effects on how children judge social contract violations. Children’s judgments
differed relatively little across different types of social relationship, suggesting that
the social contract itself is what creates the primary moral obligation between two
parties involved in such a contract. However, kinship and power relationships did
have effects on children’s judgments, and there were sex differences in these
effects. Here, we briefly review the results.

JEP 5(2007)1–4
228 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

As expected, children in this age range were all skilled in their ability to detect
contract violations, consistent with earlier literature (e.g. CUMMINS 1996; HARRIS
2000; NÚÑEZ and HARRIS 1998; HARRIS et al. 2001). Interestingly, children judged
mutual non-compliance to be a violation, suggesting that they construe each party
as having an obligation to comply independently of what the other party does (this
would not necessarily have to be the case; children could have judged that if one
party does nothing, the other party has no obligation to follow through). There was
also a small effect of perspective, with a larger proportion of violation judgments
when the rule was broken by the “Simon” character than when the rule was broken
by the other character in the story, in keeping with the literature on perspective
effects in children’s attribution of moral emotions (KELLER et al. 2003; EPLEY and
DUNNING 2000).
As predicted, the majority of children attributed anger to victims of social
contract violations, and guilt to violators, and did so increasingly more frequently
with age. Previous studies have found that moral emotion attributions change with
age, with more frequent attributions of negative emotions to violators beginning at
6 years and up (NUNNER-WINKLER and SODIAN 1988; ARSENIO and LOVER 1995;
KELLER et al. 2003). However, previous studies have only looked at negative
emotions or “bad feelings” in general. This is the first study to examine attributions
of guilt to violators of social contracts. Guilt is of particular importance because of
its moral valence. As opposed to merely feeling bad because, for example, one
regrets being foolish enough to have gotten caught, guilt is specifically associated
with the knowledge that one has violated a moral norm (GIBBARD 1990). Thus, it is
significant that the majority of first graders attributed a “bad conscience” to
violators of a social contract, and not just bad feelings in general. These attributions
increased with age from first to fourth grade, suggesting that development in this
domain continues into the second decade of life. This was true of anger attributions
to victims as well. While anger attributions have previously been demonstrated in
children as young as three (NÚÑEZ 1999), previous studies have not demonstrated a
continued increase in these attributions in older children, as documented here.
It is important to note, however, a caveat to these results. These were not free
response questions; children were asked to make yes / no judgments about specific
terms, “bad conscience” and “angry” respectively. Therefore, children’s judgments
could have been different if the choices available to them had been different. For
example, the frequency with which children responded “angry” might have been
lower if other options were available, or if the question were framed as a free
response. Future studies should attempt to remedy this problem either by offering
more response choices, or by not restricting responses at all. They should also
attempt to examine whether these emotions are conceptualized in the same way by
children and adults (for example, does “bad conscience” mean the same as “guilt”
does to an adult?)
A second purpose of this study was to examine factors influencing emotions
attributions including, especially, the social role relationship between the parties

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 229

involved in the contract. The previous study by KELLER et al. (2003, 2004) had
found that children attributed positive feelings to children violating agreements with
their mothers more frequently than to children violating agreements with peers.
Children may feel more entitled to break contracts with parents because parent–
offspring interactions are usually less strictly reciprocal than peer–peer interactions.
Because of the inclusive fitness benefits of offspring care, parents will usually
continue to invest in children regardless of their compliance, whereas this is not
necessarily true for peer–peer relationships. However, relationships between
siblings do involve kinship, despite relatively equal power, and like parent-
offspring relationships, are likely to persist despite transgressions. We therefore
asked whether power asymmetries or kinship played a greater role in attribution of
moral emotions to contract violators and victims.
For attributions of anger to victims, only the power dimension was significant.
When the contract was between two children, anger attributions to victims were
greater than when the contract was between a child and an adult. This is consistent
with the hypothesis that fairness is a more important factor when the parties are
equal in power. Power is balanced, so moral emotions like anger may play a more
important role in stabilizing cooperation.
Guilt attributions increased with age, but there were no other simple effects,
including of relationship type. Additionally, results differed for boys and girls. For
boys, guilt attributions to self as violator were more common than to the other
person as violator. This replicated the perspective effect found by KELLER et al.
(2003), and suggests that emotion attributions may be greater, for boys at least,
when making judgments about a character whose perspective one adopts. However,
we did not replicate KELLER et al. (2004) finding of an effect of relationship type on
attributions of guilt (see below for discussion).
For girls, factors mediating guilt attributions were even more complex. Older
but not younger girls attributed guilt more frequently to violators that were non kin
than to kin. Perhaps this reflects a growing sense in older girls of moral duty
towards non-kin with whom one has made an agreement than towards non-kin.
Why this appears only in girls is unclear, but perhaps reciprocally balanced non-kin
social relationships become more important in the ten-year age range for girls than
they do for boys. In addition, older girls tended to attribute guilt more frequently to
self as violator than to the other person as violator.
These findings diverge from those of earlier findings in which both boys’ and
girls’ attributions of bad feelings to children as violators were lower than to adults
as violators in parent–child contracts (KELLER et al. 2004). In this study, however,
role relationships in social contracts had relatively little effect either on violation
judgments or on emotion attributions. Although further research will be needed to
determine whether these findings are robust, it seems safe to conclude that role
differences in social contracts are relatively unimportant compared to the overall
logic of social contract violations and the emotions involved. In other words, for all
relationship types, children in the age range of 6 to 10 years find social contract

JEP 5(2007)1–4
230 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

agreements to be binding, and the moral emotions of anger and guilt are attributed
to victim and violator when such an agreement is broken, regardless of who the
victim or violator is. There are small effects of relationship and self / other
perspective, but these produce only small variations in overall attributions, which
are high.
One possible explanation for differences in findings between this study and
earlier studies, including differences in the effect that relationship type had on
emotion attributions, could be that the specific contracts differed between the
studies. In the KELLER et al. (2004) study, the agreement between parent and child
was that the child would tidy his room in exchange for a bike. In that scenario, it
would be reasonable for children to assume that it was not in the protagonist’s
power to get a bike unless the parent wanted him to have it. Therefore, one could
interpret the “violation” scenario, in which the child gets the bike even though he
has not tidied his room, as implicit agreement on the part of the parent to give the
bike anyway. Under this interpretation, fewer attributions of bad feelings on the part
of the child would be expected. In the peer–peer contract in KELLER et al. (2004),
however, the protagonist could be interpreted as simply taking the bike from his
friend, without permission, which could be expected to lead to greater attributions
of guilt to the protagonist. In our study, it is ambiguous just how access to the video
game is controlled, but it is relatively easy to imagine the child having free access
to the game, making it more similar to the peer–peer situation in KELLER et al.
(2004). If subjects assumed the resource (the video game) could be used freely by
the child in all situations, this might have led to lower differences in emotion
attributions across relationship types, because there would not be the appearance of
tacit approval by parents who bestow a benefit on a child even when the child has
not complied, as in KELLER et al. (2004). The effects of different contract types on
judgments of moral emotions in social relationships is an important area for future
studies to explore.

CONCLUSION

While previous studies have shown that children are able to detect social contract
violations, and that they associate negative emotions including anger with contract
violation, no study has yet examined in detail the attribution of the specific negative
emotions of guilt and anger in social contract situations. This matters because
different negative emotions have different functions in regulating moral behavior.
In this study, we showed that children do, in fact, attribute guilt (a “bad
conscience”) to violators of social contracts. Guilt attributions are very high overall,
though both guilt and anger attributions increased with age from first to fourth
grade, suggesting ongoing development of reasoning about emotions in this
domain. Interestingly, mediating effects on these emotion attributions were
relatively small, suggesting that by first grade, children view social contracts as
morally binding for both parties. The finding that children distinguish between guilt

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 231

and anger in the context of social exchange agreements adds to the literature
demonstrating the remarkable competence of children in this complex, and perhaps
uniquely human, form of social interaction.

REFERENCES

ARSENIO, W. and LEMERISE, E. (2001): Varieties of childhood bullying: Values, emotion


processes, and social competence. Social Development, 10, 59–73.
ARSENIO, W. F. and LOVER, A. (1995): Children’s conceptions of sociomoral affect: Happy
victimizers, mixed emotions and other expectancies. In Killen, M. and Hart, D. (eds.):
Morality in Everyday Life: Developmental Perspectives. Cambridge: Cambridge University
Press, pp. 87–128.
CHENG, P. W. and HOLYOAK, K. J. (1985): Pragmatic reasoning schemas. Cognitive Psychology,
17, 391–416.
COSMIDES, L. (1989): Has natural selection shaped how humans reason? Studies with the Wason
selection task. Cognition, 31, 187–276.
CUMMINS, D. D. (1996): Evidence of deontic reasoning in 3- and 4-year-old children. Memory &
Cognition, 24, 823–829.
EPLEY, N. and DUNNING, D. (2000): Feeling “holier than thou”: Are self-serving assessments
produced by errors in self or social prediction? Journal of Personality and Social
Psychology, 79, 861–875.
FESSLER, D. M. T. and HALEY, K J. (2003): The strategy of affect: Emotions in human
cooperation. In Hammerstein, P. (ed.): Genetic and Cultural Evolution of Cooperation.
Cambridge, MA: MIT Press, pp. 7–36.
GIBBARD, A. (1990): Wise Choices, Apt Feelings: A Theory of Normative Judgment. Cambridge,
MA: Harvard University Press.
GIGERENZER, G. and HUG, K. (1992): Domain-specific reasoning: Social contracts, cheating and
perspective change. Cognition, 42, 127–171.
GIROTTO, V., BLAYE, A. and FARIOLI, F. (1989): A reason to reason: Pragmatic bases of
children’s search for counter-examples. European Bulletin of Cognitive Psychology, 9, 297–
321.
HAMILTON, W. D. (1964): The genetical evolution of social behavior I, II. Journal of Theoretical
Biology, 7, 1–52.
HARRIS, P. L. (2000): The Work of the Imagination. Oxford: Blackwell.
HARRIS, P. L. and NÚÑÉZ, M. (1996): Understanding of permission rules by pre-school children.
Child Development, 67, 1572–1591.
HARRIS, P.L. and NÚÑEZ, M. (1997): Children’s understanding of permission and obligation. In
Smith, L., Dockrell, J. and Tomlinson, P. (eds.): Piaget, Vygotsky and beyond. London:
Routledge, pp. 211–223.
HARRIS, P.L., NÚÑEZ, M. and BRETT, C. (2001): Let’s swap: Early understanding of social
exchange by British and Nepali children. Memory & Cognition, 29(5), 757–764.
JANKE, B. (1999): Cheating and Conditional Reasoning in Children and Adults: Context Effects
in the Solution of the Wason Selection Task. Poster presented at the biennial meeting of the
Society for Research in Child Development, Albuquerque, NM.
KELLER, M. (1984): Resolving conflicts in friendship: The development of moral understanding
in everyday life. In Kurtines, W.M. and Gewirtz, J.L. (eds): Morality, Moral Behaviour and
Moral Development. New York: Wiley, pp. 140–158.

JEP 5(2007)1–4
232 H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

KELLER, M., GUMMERUM, M., WANG, X. T. and LINDSEY, S. (2004): Understanding perspectives
and emotions in contract violation: Development of deontic and moral reasoning. Child
Development, 75, 614–635.
KELLER, M., LOURENÇO, O., MALTI, T. and SAALBACH, H. (2003): The multifaceted phenomenon
of ‘happy victimizers’: a cross-cultural comparison of moral emotions. British Journal of
Developmental Psychology, 21, 1–18.
KOHLBERG, L. (1976): Moral stages and moralization: The cognitive-developmental approach. In
Lickona, T. (ed.): Moral Development and Behavior: Theory, Research, and Social Issues.
New York: Holt, Rinehart, and Winston, pp. 31–53.
LIGHT, P. H., GIROTTO, V. and LEGRENZI, P. (1990): Children’s reasoning on conditional
promises and permissions. Cognitive Development, 5, 369–383.
NÚÑEZ, M. (1999). Young Psychologists and Dealers: Early conditional reasoning on reciprocal
exchanges and the emotions involved. Biennial meeting of the Society for Child
Development, Albuquerque, NM, April.
NÚÑEZ, M. and HARRIS, P. L. (1997): Finding the naughty child: 5-and-6-year-old reasoning in
the Selection Task. VIII European Conference on Developmental Psychology, Rennes,
September.
NÚÑEZ, M. and HARRIS, P. (1998): Young children’s reasoning about prescriptive rules: Spotting
transgressions through the Selection Task. Xvth Biennial ISSBD Meetings, Berne,
Switzerland, July.
NUNNER-WINKLER, G. and SODIAN, B. (1988): Children’s understanding of moral emotions. Child
Development, 59, 1323–1338.
SMETANA, J. G. (1995): Context, conflict, and constraint in adolescent–parent authority
relationships. In Killen, M. and Hart, D. (eds.): Morality in Everyday Life: Developmental
Perspectives. Cambridge: Cambridge University Press, pp. 225–255.
SY, S., DEMEIS, D.K. and SCHEINFIELD, R. (2003): Pre-school children’s understanding of the
emotional consequences for failures to act prosocially. British Journal of Developmental
Psychology, 21, 259–272.
YOUNISS, J. (1980): Parents and Peers in Social Development: A Sullivan-Piaget Perspective.
Chicago: University of Chicago Press.

JEP 5(2007)1–4
RELATIONSHIPS AND EMOTIONS 233

Acknowledgement

The order of the authors is alphabetical. All authors contributed equally. We are
grateful for support from the Center for Adaptive Behavior and Cognition at the
Max Planck Institute for Human Development. We are also grateful to the research
students for conducting the experiments and processing data, with special thanks to
Michaela Gummerum. We also thank the school headmaster, teachers and students
for their participation in the research and Anita Todd for editing the paper.

Appendix 1: Text of social contract story

I would like to tell you a story about a boy/girl called Simon/Simone.


Here is a picture of Simon/Simone (show picture of Simon/Simone at the
table).
Simon/Simone likes to play video games.
Here he/she is playing a video game (show the picture with video).

Condition 1: Mother
(No text)
Condition 2: Brother/Sister
The story deals with Simon/Simone and his/her brother/sister.
Condition 3: Neighbor
Simon/Simone knows Mrs. Müller, who lives in the neighborhood. She
sometimes asks a child to help her with the garden and then she gives the child
something.
Condition 4: Peer
Simon/Simone has a good friend Peter/Petra, who lives in the neighborhood.
They often play together and sometimes they help each other.

(Each story was identical but the other character changed.)


One day the mother said to Simon/Simone: “Look, the garden is very messy”
(show picture of messy garden).

“It would be nice if you would clean the garden for me (show picture of
clean garden). Then I have a new video game that you may play. But, if you
want to play the video game, you must first clean the garden.”

Simon/Simone and the mother made an agreement on this. After that the
mother went shopping.
Now I will show you some pictures about how this story may have ended.
Here is one ending of the story:
The left side of the picture shows how the garden looks and the right side of
the picture shows what Simon/Simone is doing.

JEP 5(2007)1–4
Appendix 2: Materials
234

(a) (b)

JEP 5(2007)1–4
(c) (d)

Four cards of picture used in the experiment: (a) P and Q (mutual compliance: the child cleaned the garden and was allowed to play a
video game); (b) P and not Q (the other violates: the child cleaned the garden but was not allowed to play a video game); (c) not P and Q
(self violates: the child did not clean the garden but was allowed to play a video game); (d) not P and not Q (mutual violation: the child did
not clean the garden and was not allowed to play a video game)
H. CLARK BARRETT, MONIKA KELLER, MASANORI TAKEZAWA, SZYMON WICHARY

You might also like