Documente Academic
Documente Profesional
Documente Cultură
Society for
Research in
Child Development
Child Development,
Accessed: 2 2 / 1 2 / 2 0 1 2 06:14
Your use of the JSTOR archive indicates your acceptance of the Tenns & Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/tenns.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use infonnation technology and tools to increase productivity and facilitate new fonns
of scholarship. For more infonnation about JSTOR, please contact support@jstor.org.
Wiley and Society for Research in Child Development are collaborating with JSTOR to digitize, preserve and
extend access to Child Development.
STOR
http://www.jstor.org
Pennsylvania
State
University
D U N N , J U D Y ; B R O W N , JANE; S L O M K O W S K I , C H E R Y L ; T E S L A , C A R O L I N E ; a n d Y O U N G B L A D E , L I S E . Young
Children's
Understanding of Other People's Feelings and Beliefs:
Individual Differences
and
Their Antecedents. CHILD DEVELOPMENT, 1991, 62, 1 3 5 2 - 1 3 6 6 . Individual d i f f e r e n c e s in y o u n g
children's understanding of others' feelings and in their ability to explain human action in terms
of b e l i e f s , and the earlier correlates of these d i f f e r e n c e s , w e r e studied with 50 children o b s e r v e d
at h o m e with mother and sibling at 33 months, then tested at 40 months on affective-labeling,
perspective-taking, and false-belief tasks. Individual d i f f e r e n c e s in social understanding w e r e
marked; a third of the children o f f e r e d explanations of actions in terms of false b e l i e f , though
f e w p r e d i c t e d actions on the basis of beliefs. T h e s e differences w e r e associated with participation
in family discourse about feelings and causality 7 months earlier, verbal fluency of mother and
c h i l d , and c o o p e r a t i v e interaction with the sibling. D i f f e r e n c e s in understanding f e e l i n g s w e r e
also associated with the discourse measures, the quality of mother-sibling interaction, SES, and
g e n d e r , with girls m o r e successful than boys. T h e results support the v i e w that d i s c o u r s e about
the social w o r l d may in part mediate the key conceptual advances reflected in the social c o g n i t i o n
tasks; interaction b e t w e e n child and sibling and the relationships b e t w e e n other family m e m b e r s
are also implicated in the growth of social understanding.
The very early stages of children's understanding of other minds and others' feelings have in the last few years received rapidly growing attention (Astington, Harris, &
Olson, 1988; Frye & Moore, 1990; Whiten,
1990); particular interest has been shown in
the question of when young children are
aware of mental states and of the psychological causes of behavior (Miller & Aloise,
1989). Research strategies have included the
documentation of the recognition and differentiation of emotions by infants (Harris,
1989); the development of children's use
of, first, feeling state terms and, later, mental terms (Bretherton, McNew, & BeeghlySmith, 1981; Ridgeway, Waters, & Kuczaj,
1985; Shatz, Wellman, & Silber, 1983); the
distinctions very young children make between the mental and physical worlds involving knowledge states (Wellman, 1985);
and their ability to reason about and explain
human action in terms of beliefs and desires.
A focus of special interest has been the development of children's ability to predict
and explain actions that are premised on
The subjects were 50 second-born children, participating with their older siblings
and their mothers in a longitudinal study
in central Pennsylvania. There were 23
boys (13 with older brothers, 10 with older
sisters) and 27 girls (10 with older sisters
and 17 with older brothers). The mean age
gap between the siblings was 43 months
(range 1673). The families, recruited from
sequential birth announcements in the local
newspaper, included a wide range of backgrounds; the occupational status of the fathers was assessed with the National Opinion Research Corporation prestige ratings
based on the 1970 census (Hauser & Featherman, 1977): the mean paternal prestige rating was 51.5 (SD = 16.3), the range from 15
to 88. For the U.S. white labor force, the
mean NORC rating is 41.7 (SD = 13.9); thus
although the Pennsylvania sample has a
higher occupational prestige rating than the
working U.S. population, the variance is
similar. Mean values for parental education
were 14.7 (SD = 2.4) years and 15.4 (SD =
2.9) years for mothers and fathers, respectively.
PROCEDURES
& Wellman, 1989), and the affective perspective-taking tasks (Denham, 1986), were
conducted. On the first visit, the false belief
tasks were carried out with the children.
A questionnaire was also administered to
mothers to decide on appropriate content for
the affective perspective-taking task for each
child (see below). On the second visit, the
affective perspective-taking task was carried
out; information on the parents' occupations
and education was obtained from an interview with the mother. Six observers were
involved in the data collection.
MEASURES
Conversational Measures
The transcripts provided a data base of
on average 316 child speaker turns, 356
turns of mother and sibling to the child
(Table 1). A speaker turn was defined as all
of one speaker's utterances bounded by the
utterances of another speaker. The correlations between the measures on the first
and second visits at time 1 gave an estimate
of test-retest reliability for the different
speaker turns; these were as follows: child
to mother r(50) = .69, mother to child r(50)
= .74, child to sibling r(50) = .68, sibling to
child r(50) = .68, all significant at p < .05.
Mean Length of Utterance
The mean length and the upperbound
mean length of each child's utterances were
coded following standard procedures (Shatz
& Gelman, 1973) from the 100 consecutive
child utterances that followed the child's
first 10 conversational turns. The number of
words in the 10 longest of these utterances
was used to determine the upperbound
MLU. The mothers' MLU and upperbound
MLU were similarly calculated. The MLU
and upperbound MLU were positively correlated for both mothers and children, r(50)
= .86 and .85, for mothers and for children,
respectively, both p < .001, and the pattern
of correlations with these measures and the
other time 1 measures to be reported was
very similar. In the analyses that follow we
therefore report just the results for the upperbound MLU.
Feeling State Talk
A coding system was designed for the
analysis of conversations in which family
members referred to feeling states (see
Dunn etal., 1991). References included conversational turns in which the speaker used
a feeling state term (e.g., "sad" or "happy"),
those in which the speaker used a phrase
that connoted a feeling state (e.g., "made a
TABLE
STATE
TURNS
TOTAL
SPEAKERS
Child to mother
and sib
Mother and sib to
child
Mother and child
conversation
CAUSAL
(%)
TURNS
TURNS
(%)
Mean
SD
Mean
SD
Mean
SD
316.4
162.5
4.1
2.7
2.7
2.3
355.9
166.4
7.8
3.3
7.4
3.1
441.5
286.3
6.4
3.0
5.9
3.1
UPPERBOUND
MLU
Mean
Child
Mother to child
MLU
SD
Mean
SD
3.67
.82
7.17
1.53
5.38
.78
11.93
2.20
ing state turns) as an index of the child's participation in family talk about feelings; total
feeling state turns to child (i.e., mother to
child 4- sibling to child feeling state turns)
as an index of the talk directly addressed
to the target child concerning feelings;
mother-child total feeling state turns (i.e.,
mother to child + child to mother feeling
state turns) as an index of the conversations
between mother and child about feelings;
this was included because the previous
study had found mother-child conversational turns about feeling states to be the
most powerful predictor of later affective
perspective taking. Frequencies of these
measures, and of the total turns in the observations, are shown in Table 1.
A conversational turn referring to feeling states was defined as all of one speaker's
utterances bounded by the utterances of another speaker in which an explicit reference
to a feeling state was made. If an individual's
utterances within one conversational turn
referred to more than one emotional theme,
or to more than one individual's feelings,
each reference was coded separately. Each
conversational turn that referred to a feeling
state was coded in terms of a number of variables (speaker, addressee, theme, dispute,
causal reference, pragmatic context). In the
present article, we consider the categories of
speaker and addressee, and employ measures of total feeling state turns. In our initial
examination of the data we employed three
measures: total child feeling state turns
(i.e., child to mother + child to sibling feel-
TABLE
MEANS,
STANDARD DEVIATIONS,
2
INTER-RATER AGREEMENT,
Mean
SD
Interrater
Agreement
Test-Retest
Correlation
...
...
...
...
3.87
3.28
3.13
3.57
.88
.92
.77
1.02
.82*
.80*
.80*
.83*
.74*
.73*
.19
.80*
...
...
...
...
3.48
2.86
2.32
3.03
.89
.80
.87
.94
.82*
.80*
.80*
.83*
.76*
.58*
.42*
.59*
...
...
...
...
...
2.52
2.81
2.11
1.98
2.86
.77
.86
.90
.82
1.11
.84*
.79*
.76*
.95*
.85*
.34*
.58*
.52*
.58*
.66*
...
...
...
...
...
2.73
2.88
3.22
2.41
2.89
.83
.94
1.02
.99
1.15
.84*
.79*
.76*
.95*
.85*
.39*
.55*
.60*
.57*
.63*
Rating Scales
Mother to child:
Responsiveness
Attention
Control
Affection
Mother to sibling:
Responsiveness
Attention
Control
Affection
Child to sibling:
Conflict
Cooperation
Control
Competition
Affection
Sibling to child:
Conflict
Cooperation
Control
Competition
Affection
* p < .05.
Correlations were next conducted between the two social cognition measures
from time 2 and (a) total number of talk turns
per hour of observation and the measures of
feeling state talk and causal talk during the
time 1 observations; (b) mother's education,
father's occupation, gender of the child, and
the upperbound MLU of child and mother;
and (c) the rating scales of family members'
interaction with each other.
There was no significant relation between total talk and the two social cognition
measures. Table 3 shows the correlations for
the discourse measures, the MLU measures,
gender, the education and SES measures,
and the significant correlations with the rating scales. For the affective aggregate score,
significant correlations were found with the
child's talk about feelings and about causality in the observations 7 months earlier, and
with mother-child conversations about feelings and the total talk to the child about feel-
ings. The score was also related to the children's and mothers' upperbound MLU and
the SES level, as reflected in the fathers' occupational prestige scores, and to gender.
Girls scored higher on the affective understanding measure.
Offering explanations in the false belief
tasks was related to the child's talk about
feeling states and to mother-child conversations about feeling states and about causality
at the earlier time point. Mothers' upperbound MLU was correlated with explanations in the false belief tasks, but the SES,
mothers' education, child MLU, and gender
associations did not reach significance.
The relations between the social cognition scores and the rating scales of family
members' interaction were examined next.
Mother-to-Child Ratings
The four maternal rating scales (responsiveness, attention, control/intrusiveness, affection) were each correlated with the two
social cognition measures from time 2. Of
the eight correlations, none was significant.
Mother-to-Sibling Ratings
Three of the eight correlations between
mother-to-sibling rating scales and the social
cognition measures were significant at p <
.05 (see Table 3). The attention and responsiveness that the mothers showed toward the
sibling was correlated with the children's
affective aggregate score at the later time
point, and the mothers' control/intrusiveness
toward the sibling was related to the children offering explanations in the false belief
tasks.
Child-to-Sibling Ratings
Three of the 10 correlations between
the five child-to-sibling scales and the two
social cognition measures were significant
(see Table 3): The social cognition measures
were related to the cooperation shown toward the sibling 7 months earlier, and affection to the sibling was also related to the
affective tasks.
Sibling-to-Child Ratings
Of the 10 correlations between the five
scales and the two social cognition scores,
two were significant; sibling affection and
cooperation were related to the child's later
score on the affective aggregate measure.
MULTIVARIATE ANALYSES OF THE
CORRELATES OF THE SOCIAL
COGNITION MEASURES
TABLE 3
CORRELATIONS BETWEEN TIME 1 MEASURES AND TIME 2 SOCIAL
UNDERSTANDING
MEASUHES
Affective
Aggregate
F e e l i n g state talk:
C h i l d total
Total to child
Mother-child
Causal talk:
C h i l d total
Total to child
Mother-child
Child upperbound M L U
Mother u p p e r b o u n d M L U
Mother education
Father occupational prestige .
G e n d e r (girls = 1; b o y s = 2)
Rating scales: 3
Mother to sibling:
Responsiveness
Attention
Control
Affection
C h i l d to sibling:
Conflict
Cooperation
Control
Competition
Affection
Sibling to child:
Conflict
Cooperation
Control
Competition
Affection
.45*
.29*
.36*
.42*
False B e l i e f
Explanations
.31*
.14
.28*
.19
.06
.22
.51*
.35*
.26
.45*
-.36*
.20
.33*
.14
.24
.18
31*
28*
26
25
.09
-.06
.34*
.08
-.15
51*
04
06
33*
-.16
.38*
-.10
.04
.18
- .06
27*
26
- .03
35*
-.04
.15
.15
-.03
.20
.20
.42*
* Note that correlations for rating scales of mother to child are not included in the
table; all were nonsignificant.
* p < .05.
four sets. On the basis of the preceding correlational analyses, the variables selected to
enter in the multiple regression for this variable were the rating of child-to-sibling cooperation, father's occupational prestige, the
upperbound MLU of child and mother entered as one step, and the discourse variables child talk about causality and child talk
about feelings (minus those turns in which
cause was discussed) entered as one step.
Table 4 shows that together these variables
accounted for 56% of the variance in the affective aggregate measure. By next conducting separate hierarchical multiple regression
analyses systematically excluding each of
the four sets of variables in turn, the unique
variance contributed by each variable was
determined. Table 4 shows that for the
affective aggregate variable, the discourse
variables independently contributed 10% of
the variance, the child-sibling cooperation
T A B L E
Social U n d e r s t a n d i n g
Measure
Discourse
Measures
Change
in R ?
RATINGS OF INTERACTION,
R2
df
A f f e c t i v e aggregate
.56*
10.99
5.44
Child feeling
state" +
child
causal
.10*
Child cooperate
with sib
.07*
F a l s e b e l i e f task: e x p l a nations
.39*
7.08
4.45
Mother-child
causal +
child
feeling
state"
.04 +
Child cooperate
with sib
M o t h e r control
sib
.11*
Ratings
Change
in R ?
.07*
SES
Father
prestige
S E S , AND UPPERBOUND M L U
Change
in R 2
.04 +
Upperbound
MLU
Change
in R
C h i l d and
mother
.05 +
Mother
.01
Stewart, R. B., & Marvin, R. S. (1984). Sibling relations: T h e role of conceptual perspectivetaking in the ontogeny of sibling caregiving.
Child Development, 55, 1322-1332.
Stacker, C., Dunn, }., & Plomin, R. (1989). Sibling
relationships: Links with child temperament,
maternal behavior, and family structure.
Child Development, 60, 7 1 5 - 7 2 7 .
Tinsley, H. E. A., & Weiss, D. J. (1975). Inter-rater
reliability and agreement of subjective j u d g e ments. Journal
of Counselling
Psychology,
22, 3 5 8 - 3 7 6 .
Wellman, H. M. (1985). T h e origins of metacognition. In D. Forrest-Pressley, C. McKinnon, &
T. Waller (Eds.), Metacognition, cognition
and human performance. N e w York: Acad e m i c Press.
Wellman, H. M. (1988). First steps in the child's
theorizing about the mind. In J. W. Astington,
P. L. Harris, & D. R. Olson (Eds.), Developing
theories of mind (pp. 6 4 - 9 2 ) . Cambridge:
Cambridge University Press.
Whiten, A. (Ed.). (1990). Natural theories of mind.
Oxford: Basil Blackwell.
Zahn-Waxler, C., Radke-Yarrow, M., & King, R. A.
(1979). Child rearing and children's prosocial
initiations towards victims of distress. Child
Development, 48, 3 1 9 - 3 3 0 .