You are on page 1of 29

JOURNAL OF RESEARCH ON ADOLESCENCE, 14(3), 257284

Copyright r 2004, Society for Research on Adolescence

He Did It on Purpose! Family


Correlates of Negative Attributions
About an Adolescent Sibling
Lisa S. Matthews
Texas A&M University

Rand D. Conger
University of California, Davis

Scholars have suggested that family life may influence childrens attributions
about close relationships. Using a sample of 369 two-parent families with 2
children (a target adolescent in the 8th grade and a sibling aged 10 to 18), we
investigated whether the siblings negative attributions regarding the target
adolescent were associated with mothers and fathers negative attributions
regarding the target adolescent, or the gender constellation and birth order of
the sibling pair. In addition, we used the self-reported negative attributions
of both siblings and adolescents to investigate whether attributions predicted
ones own behavior, or whether the behavior of the partner predicted the
perceivers negative attributions. Structural equation models showed that
mothers and fathers negative attributions regarding the target adolescent
predicted concurrent negative attributions for girls, but not boys. In addition,
siblings negative attributions regarding each other predicted their negative
and positive behavior toward the partner 2 years later. The positive behavior
of the attribution target also predicted the perceivers negative attributions,
but its effect was weaker. Girl sibling pairs were less negative in their attributions than cross-sex sibling pairs and boy sibling pairs. Finally, although
younger siblings were less negative in their attributions than older siblings,
this difference decreased over time. This study advances the knowledge of
negative attributions in close relationships by identifying family correlates of
an understudied, but important, close relationship.
Requests for reprints should be sent to Lisa Matthews, Department of Sociology, Texas
A&M University, College Station, TX 77843-4351. E-mail: matthews@neo.tamu.edu

258

MATTHEWS AND CONGER

Evidence regarding the important role that attributions play in close relationships continues to accumulate. For example, studies of married
couples show that individuals who interpret a spouses aversive behavior
as intentional or selfishly motivated, or as caused by dispositional characteristics are likely to experience a more conflictual and lower quality
marriage (Bradbury & Fincham, 1990; Fincham, 2001; Fincham, Bradbury,
& Scott, 1990; Fletcher & Fincham, 1991). Attributions also predict changes
in marital quality (Fincham & Bradbury, 1987, 1991; Karney & Bradbury,
2000). Although rarely studied, it seems likely that attributions would also
affect the quality of other important family relationships, including sibling
relationships. However, although there are clear differences between
marriages and sibling relationships, it is not unreasonable to assume that
negative attributions (e.g., believing that another persons aversive behavior is intentional) would have a similarly negative effect on the sibling
relationship.
Two sources of child attributions about parents and peers proposed in
the literatureparentss attributions and the behavior of the attribution
target (Bugental & Johnston, 2000; Grusec & Mammone, 1995)would be
likely to influence childrens negative attributions regarding siblings. In
addition, family characteristics, such as the gender constellation of siblings and birth order of children, have been found to influence family
interactional dynamics and could affect negative attributions. However, as
yet there has been little empirical research examining associations between child attributions and these components of family life, and no research to our knowledge examining these factors within sibling
relationships. In the present report we used information from a longitudinal study of two parents and their two children (a target adolescent and
his or her near-age sibling) to investigate whether the siblings negative
attributions regarding the target adolescent are associated with mothers
and fathers negative attributions regarding the target adolescent, and
with the gender constellation and birth order of the sibling pair. In addition, we used the negative attributions of both siblings and adolescents
to investigate the direction of influence between attributions and behavior
(i.e., do negative attributions predict the perceivers own behavior, or does
the behavior of the attribution target predict the perceivers negative
attributions?).
Parents Negative Attributions, AdolescentSibling Interaction, and
Sibling Constellation Factors
We defined attributions in global terms as the individuals assessment that
the attribution target was not generally a good-natured person and that his

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

259

or her aversive behavior was intentional or nonaccidental. Thus, our study


focused on negative attributions. Several family scholars have hypothesized that children may learn to make particular types of attributions
regarding other family members (e.g., negative attributions) through observing their parents (Bugental & Johnston, 2000; Bugental, Johnston,
New, & Silvester, 1998; Fincham, Beach, Arias, & Brody, 1998; Grusec,
Hastings, & Mammone, 1994; Grusec & Mammone, 1995). Social learning
theory suggests that children learn what kinds of attributions to make in
particular situations by hearing others in their social environment articulate attributions (Bandura, 1971; Miller & Dollard, 1941). For example, a
child who hears her mother accuse others of selfishness when late
for dinner may learn that people who are late are acting out of a selfish
motivation.
In addition, family systems theory suggests that family members establish a shared definition of family events, including evaluations of the
behaviors of individual family members, to create and maintain a smoothly functioning family system (Munton & Antaki, 1988). This theory suggests that parents and children would exert a reciprocal effect on each
other in terms of attribution making. That is, daily conversation affords all
family members the opportunity to voice their interpretations of family
interpersonal events, raising the possibility that parents and children may
influence each other in their attributions. Although family researchers are
becoming more cognizant of the possibility of bidirectional socialization
effects between parents and children (Bugental & Goodnow, 1998; Hogan,
Etz, & Tudge, 1999), we believe that parents, as the more powerful individuals, will have more of an influence on children than vice-versa. Parents may also have stronger motives for influencing childrens attribution
than the reverse. Because negative attributions (i.e., deciding that a family
member is being irritating on purpose) often lead to conflict, parents may
actively encourage their children to make benign attributions for negative
sibling behavior. In particular, it seems likely that parenting goals (e.g., of
maintaining family peace) would motivate parental attempts to influence
childrens attributions (Hastings & Grusec, 1998).
To our knowledge, one only study has examined the association among
parents and childrens attributions regarding others in close relationships. Fincham et al. (1998) found a significant correlation between the
level of negativity of fathers and mothers spousal attributions (i.e.,
whether parents attributed aversive spousal behavior to negative intent,
selfish motivation, etc.) and the level of negativity of childrens attributions for parents. Correlations were significant only within same-sex parentchild dyads (e.g., girls attributions regarding her mother correlated
significantly with her mothers attributions regarding her husband).

260

MATTHEWS AND CONGER

Fincham et al. concluded that children model attributions that same-sex


parents articulate in interaction with the child or other family members.
Although the socialization approach just discussed suggests that one
source of attributions lies in the learning history of the individual, other
scholars have found an association between attributions and behavior.
Classic attribution theory predicts that the behavior of the attribution target influences the perceivers attributions. For example, the theory of correspondent inferences (Jones & Davis, 1965) states that people are likely to
make correspondent inferences (i.e., to assume that an individuals behavior reflects an underlying dispositional trait) when that persons behavior seems freely chosen and intentional. Children may be especially
likely to make correspondent inferences regarding family members, including siblings, because the emotional closeness that characterizes family
relationships tends to make their behavior especially salient (Dix & Grusec, 1985). Furthermore, living in close proximity to the sibling means that
the siblings behaviors may have high hedonic relevance (i.e., they are
supportive of or opposed to the adolescent perceivers interests) and are
more likely to be personalized (i.e., the child perceiver believes that the
behavior is directed specifically at himself or herself). Both of these factors
increase the likelihood of correspondent inferences and suggest that the
behavior of the attribution target would predict attributions.
Other studies have indicated that a perceivers attributions predict his
or her behavior toward the attribution target. For example, studies have
shown that children and adolescents who attribute hostile intent to peers
are more likely to behave in an aggressive manner toward the attribution
target (Dodge, 1980; Dodge & Frame, 1982; Lochman, 1987; Nasby, Hayden, & DePaulo, 1980; Quiggle, Garber, Panak, & Dodge, 1994; Waas,
1988). Although siblings behavior may influence the recipients attributions about that sibling, we believe that the relation between siblings
attributions and their own future behavior will be stronger. Cognitivebehavioral theories of close relationships (Baucom & Epstein, 1990; Johnson, Karney, Rogge, & Bradbury, 2001) suggest that attitudes and cognitions are more stable than behavior, and empirical studies have lent some
support to this contention (Achenbach, McConaughy, & Howell, 1987;
Schwarz, Barton-Henry, & Pruzinsky, 1985). For example, Gottman (1993)
hypothesized that in romantic couples, perceptions of partner hostility do
not change until a certain threshold of behavioral negativity is reached.
Similarly, perceptions of spouses hostility and warmth appear to change
more slowly than spousal behavior (Matthews, Conger, & Wickrama,
2000). Finally, in this study the attribution-to-behavior linkage is within
person, whereas the behavior-to-attribution linkage is between person
(i.e., partners behavior predicting the perceivers negative attributions).

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

261

Thus, it makes sense that the attribution-to-behavior path would be


stronger. Taken together, the evidence suggests that the relation between a
siblings initial attributions and his or her future behavior will be stronger
than the relation between the partners behavior and the siblings future
attributions.
Finally, childrens and adolescents negative attributions are probably
associated with sibling constellation variables, especially birth order and
gender constellation. Studies show that birth order has a strong influence
on sibling relationships (Hoffman, 1991). Older siblings are generally described as more dominant than younger siblings, with the younger children looking up to their older brothers and sisters as admired role models
and as providers of guidance, instruction, and caregiving (Buhrmester,
1992; Hoffman, 1991; Minnett, Vandell, & Santrock, 1983; Pepler, Corter, &
Abramovitch, 1982; Updegraff, McHale, & Crouter, 2000). Younger siblings generally feel closer to older siblings than the reverse, and older
siblings report more conflict than younger siblings (Buhrmester, 1992).
Studies also show that sibling relationships usually become more egalitarian and less emotionally intense with increasing age (Buhrmester &
Furman, 1990; Vandell, Minnett, & Santrock, 1987). The gender constellation of the two siblings also exerts an important influence on the relationship, with same-sex pairs demonstrating higher levels of warmth and
intimacy than cross-sex pairs, especially for girls (Buhrmester, 1992;
McHale, Updegraff, Tucker, & Crouter, 2000; Minnett et al., 1983).
Hypotheses
Based on the theoretical and empirical considerations outlined here, we
developed five hypotheses. First, we hypothesized that fathers and
mothers negative attributions regarding the target adolescent (i.e., their
assessment that he or she was not good-natured or deliberately engaged in
aversive behavior to irritate others) would predict sibling negative attributions about the adolescent, after controlling for earlier sibling negative
attributions. Our second hypothesis was that the siblings and the adolescents negative attributions about each other would predict their negative and positive behavior toward each other. Theory (e.g., Jones & Davis,
1965) suggests that attributions are based on the behavior of the attribution
target. However, because cognitions tend to be more stable than behavior
(Achenbach et al., 1987; Schwarz et al., 1985), we predicted that siblings
and adolescents negative attributions about each other would predict
their own behavior toward their partner, whereas the effect of the behavior
of the attribution target on attributions would be weak or nonexistent
(Davis, 1985).

262

MATTHEWS AND CONGER

Our remaining three hypotheses concern the effects of birth order and
sibling gender constellation on negative attributions. Prior studies, discussed earlier, showed that same-sex siblings experience greater warmth
and intimacy in their relationship, compared with opposite-sex siblings,
and that this is especially the case for girls. Thus, our third hypothesis was
that same-sex siblings would show lower mean levels of negativity in their
attributions regarding the target adolescent than opposite-sex siblings and
that girlgirl sibling pairs would be less negative in their attributions than
boyboy pairs. Prior research regarding birth order suggests that younger
siblings view their brother or sister in a more positive light than older
siblings but that the sibling relationship becomes less emotionally intense
over time. Consistent with these earlier studies, our fourth hypothesis was
that younger siblings, compared with older siblings, would show lower
mean levels of negativity in their attributions regarding the target adolescent. However, we also expected to see this difference grow smaller
over time, as the younger siblings matured and became less dependent on
their older brother or sister. Thus, our fifth hypothesis was that the difference in mean attributional negativity between younger and older siblings would be significantly smaller at Time 2 than at Time 1.
METHOD
Sample
The sample for this study was derived from a larger sample of 451 initial
families participating in a 4-year longitudinal study. The Iowa Youth and
Families Project was designed to examine the individual characteristics
and family interactional patterns that contribute to successful adaptation
to rural economic stress. Families were recruited for the study in Year 1
(1989) from 34 public and private schools in north central Iowa. The names
and addresses of all seventh-grade students were obtained from all
schools in communities with populations of 6,500 or less in the identified
counties. Parents were sent a letter explaining the project and then were
contacted by phone. To participate in the study, families had to include
two parents, their seventh-grade child (the target adolescent), and a sibling within 4 years of age of the seventh grader. Few minority families live
in the rural area where the research was conducted; thus, they were not
selected to participate. Of the qualifying families originally contacted, 79%
agreed to be part of the study. In the present analysis, listwise deletion
resulted in a final sample of 369 parents and siblings, and 324 sibling
target adolescent pairs. In our final sample, husbands and wives averaged
13.5 years of education each. Gross per capita income ranged from 0

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

263

(a negative income reported) to $38,000, with a median of $8,466.00. The


target adolescents were 13 to 15 years old during the 1st year of the study,
whereas siblings ranged in age from 10 to 18 years of age. Bias analyses
using all study variables showed that families who were excluded from
the sample because data were missing or because they dropped out of the
study were similar to families who were retained. For example, excluded
fathers and mothers had approximately the same level of education and
per capita income as retained fathers and mothers. The only difference
was that excluded adolescents were slightly more negative in their
behavior toward the sibling in 1990 than were retained adolescents.
Procedure
In the 4 years of the study (from 1989 to 1992), the families were visited
twice each year by trained field interviewers. During the first home visit,
family members filled out questionnaires asking them about a variety of
topics, including recent life events, their physical and mental health, and
their relationships with family members and friends. Each family member
filled out his or her questionnaire independently. A second visit occurred
within 2 weeks of the first. During this visit, family members first filled out
additional questionnaires, including the questionnaire containing the
attributional measures and then participated in four videotaped interactional tasks. For each task, the interviewer turned on and tested the video
equipment, explained the task to the family, and helped them complete a
sample question. The interviewer then retired to a part of the house where
he or she could not hear the family until the task was over.
Each task consisted of family members sitting around a table and taking
turns reading questions from cards. The family members then would answer and discuss the questions together. Data used in these analyses came
from Task 3, which involved only the target adolescent and his or her
sibling, and was 15 min long. The children were asked to talk about their
relationships with each other and with their parents and how they were
doing in school. The videotaped tasks were evaluated by trained video
coders. Coders received 2 months of training and were required to pass
several written and viewing tests before being allowed to code tapes.
Separate, independent coders were used for each task.
Measures
The study involved measurement in two conceptual areas: negative
attributions and adolescent and sibling behavior. In addition, parents

264

MATTHEWS AND CONGER

educational level, per capita income, and the age of the sibling were used
as control variables. Negative attribution measures were available only in
the second (1990) and fourth (1992) waves of the study; thus, all measures
were taken from these two assessments.
Negative attributions. We measured negative attributions using
three questions. Fathers, mothers, and siblings were asked to report
whether the following three statements characterized the target
adolescent: He/she sometimes does mean things just to irritate
people, When he/she does something that bothers other people, it is
usually just an accident, and He/she is good-natured and always tries
to be helpful and considerate toward others. The adolescent was asked
the same questions regarding the sibling. Responses were made on a
5-point scale ranging from 1 (strongly agree) to 5 (strongly disagree). The
items were coded so that a high score indicated the presence of negative
attributions (i.e., as indicating negative intent). Reliabilities for the
summed scale ranged from .68 to .75 for mothers and fathers.
Reliabilities for the children were .62 and .75 for siblings, and .63 and
.80 for adolescents, for 1990 and 1992, respectively.
Negativity and positivity in observed behavior. All observer
measures were coded according to the Iowa Family Interaction Rating
Scales (IFIRS; Melby et al., 1989). The IFIRS is designed to uncover patterns
of behavior in the actor.1 That is, rather than assessing discrete behavioral
acts (i.e., counting the number of times an actor says something positive)
or the content of what is actually said, the IFIRS focuses on behavioral
subtext, that is, on gestures, facial expressions, tones of voice, and so on
that convey an impression. Thus, the IFIRS assesses behaviors that are
characteristic of the actor, or are habitual. Approximately 25% of the
interactions were coded by pairs of independent coders to obtain
intraclass correlations. Within-coder alpha coefficients for negative and
positive behavior were obtained from the three scales constituting each
measure.
Observer measures were assessed on a scale ranging from 1 (the behavior
is not at all characteristic of the individual) to 9 (the behavior is very characteristic
of the individual). Negative behavior was measured with three scales. The
hostility scale measures the degree of hostility and anger the focal person
displays toward the recipient and the degree to which he or she is critical,
rejecting, or disapproving of the recipient. The antisocial scale measures
1

See Melby et al. (1995) for validation of the negative behavior scales.

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

265

the degree of resistance or defiance the focal person displays toward the
recipient through inconsiderate, noncompliant, insensitive, or obnoxious
behavior. The angry-coercive scale measures the degree to which the focal
person tries to influence the other person through the use of anger or
hostility (e.g., by using power plays, making demands or hostile commands, or being stubborn or resistant; Melby et al., 1989). These measures
had excellent reliability. Cronbachs alphas for the hostility, angrycoercive, and anti-social scales were .91 and .91 for sibling and adolescent,
respectively, at Time 1, and .94 and .94 for sibling and adolescent, respectively, at Time 2. Intraclass correlations for the negative behavior scales in
1990 and 1992 ranged from .70 to .84 for adolescent and from .69 to .86 for
sibling.
Positive behavior was also measured with three scales. The warmth
scale measures the degree to which the focal person reacts favorably toward, takes an interest in, or enjoys being with the recipient of the behavior. The prosocial scale measures the focal persons ability to relate
competently with other family members and includes cooperativeness,
sensitivity, helpfulness, a willingness to change for the sake of others, or a
willingness to comply with the requests of others. The listener-responsiveness scale measures the degree to which the focal person listens to and
shows interest in the recipient by acknowledging and validating what the
other person has to say (e.g., by using nonverbal backchannels and verbal
assents; Melby et al., 1989). These measures also had excellent reliability.
Cronbachs alphas for the warmth, prosocial, and listener-responsiveness
scales were .85 and .84 for sibling and adolescent, respectively, at Time 1,
and .84 and .80 for sibling and adolescent, respectively, at Time 2. Intraclass correlations for the positive behavior scales measured in 1990 and
1992 ranged from .48 to .70 for adolescent and from .53 to .75 for sibling.
Control variables. Finally, we used self-report of sibling age, the
mean of fathers and mothers self-reported years of education, and
mothers report of per capita income as control variables.
Plan for Data Analysis
We tested our first and second hypotheses (that parents negative attributions would predict siblings negative attributions and that siblings
and adolescents negative attributions would predict their own behavior)
using structural equation modeling as our primary analytic procedure. All
structural equation models were estimated with the maximum-likelihood
algorithm from the LISREL 8.30 software. In addition, using the modification

266

MATTHEWS AND CONGER

indexes provided by the LISREL 8.30 software as a guide, we correlated


those errors of same-reporter indicators that significantly reduced the chisquare statistic of each model. This is a procedure that improves the
overall fit of the model to the data (Bollen, 1989), and we performed it on
all structural equation models estimated in this study. In none of the
models did correlating errors alter the magnitude or significance of factor
loadings or path coefficients. Overall fit of models to the data was evaluated with several fit indexes. Models are considered to fit the data well if
they have a chi-square that is less than twice the degrees of freedom, have
goodness-of-fit indexes (GFI) and comparative fit indexes (CFI) of .90 or
greater, or have a root mean square of approximation (RMSEA) of .08 or
less (Bentler, 1990; Bollen, 1989; Browne & Cudeck, 1993). Finally, for all
models, individual control variables were dropped if preliminary analyses showed they had no effect on the model. This was done to simplify
analysis and presentation.
As a first step in the structural equation modeling procedure, we specified and tested measurement models. Latent constructs were formed using individual measures as separate indicators. For example, the
measurement model used to test the first hypothesis consisted of six latent constructs: one each for father, mother, and sibling negative attributions at two time points (1990 and 1992). Each latent construct was formed
by using the three questionnaire items that made up the respective negative attribution scale as separate indicators for each respondent and year
(e.g., fathers 1990 responses to the three negative attribution questionnaire items were used to form the 1990 father negative attribution latent
construct). The measurement models used in the test of the second
hypothesis were formed in a similar fashion.
We had two waves of data for both the negative attribution measures
and the behavioral measures. Having two waves of data allowed us to do
two things. First, it allowed us to test our hypotheses regarding the direction of influence between parent and child negative attributions (that
parent negative attributions would predict child negative attributions)
and between child negative attributions and behavior (that negative attributions would predict behavior). Second, it allowed us to predict
change in the rank ordering of the dependent variable from 1990 to 1992 by
controlling for earlier levels of the dependent variable (Kessler & Greenberg, 1981). Predicting change, compared with simply predicting the dependent variable at a later time point (i.e., without controlling for earlier
levels of the dependent variable), provides a stronger test of our hypotheses. Finally, additional analyses using the multisample procedure provided by the LISREL 8.30 software showed that none of the models was
moderated by either birth order (younger vs. older sibling) or gender

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

267

constellation (boy pairs vs. girl pairs, or same-sex versus cross-sex


pairs).2
Hypotheses 3 through 5 were tested with a mixed-model procedure
available in the SPSS 11.0 software. Details of this procedure are outlined
in the Results section.
RESULTS
Do Father and Mother Negative Attributions Predict Change in Sibling
Negative Attributions?
As a first step in examining whether father and mother negative attributions predict sibling negative attributions, we examined intercorrelations
among scale scores for fathers, mothers, and siblings. Correlations, presented in Table 1, were estimated separately for boys and girls, and show a
significant relationship between parents and siblings negative attributions
about the target adolescent. The correlations between parent and sibling
negative attributions are all statistically significant, and range from .21 to
.43 for girls (below diagonal) and from .24 to .41 for boys (above diagonal).
On the basis of this initial support for our first hypothesis (that parent
negative attributions would predict change in sibling negative attributions), we proceeded with structural modeling analyses. Standardized
factor loadings for the latent constructs, obtained from the measurement
model, are presented in the top three sections of Table 2. All factor loadings
are statistically significant.3 The measurement model fit the data well
(w2 5 159.13 with 110 df, GFI 5 .96, CFI 5 .98, RMSEA 5 .035).

2
The multisample procedure available in Lisrel 8.30 simultaneously estimates a pair of
identical models conducted with separate subsamples (e.g., a pair might consist of one model
with boys and one model with girls), obtaining a single chi-square statistic for both models.
To see whether two subsamples differ significantly on a particular path, the researcher first
estimates one pair of models, constraining parallel paths to be equal to one another across the
two subsamples (e.g., the predictive path from father attributions to boy attributions is constrained to be equal to the path from father attributions to girl attributions). The researcher then
estimates a pair of models without the equality constraint, obtaining a second chi-square. The
difference between two chi-square statistics is itself distributed as a chi-square. Thus, if the
difference between the two chi-square statistics is 3.84 or greater (the .05 significance level with 1
df ), the researcher can conclude that the subsamples differ significantly from each other on that
path.
3
Factor loadings of .5, in which the latent construct accounts for 25% of the variance of the
measure (.52), are considered good. Factor loadings of .70 and higher, indicating that the latent
construct accounts for about 50% (.72 5 .49) of the variance of the measure, are considered
excellent.

268

MATTHEWS AND CONGER

TABLE 1
Intercorrelations, Means, and Standard Deviations Among Attribution Scale Scores and Control
Variables
1
1. Father 1990
2. Father 1992
3. Mother 1990
4. Mother 1992
5. Sibling 1990
6. Sibling 1992
7. Parents ed.
8. PCI
9. Siblings age
M
SD

56n
45n
36n
34n
28n
09
08
03
6.97
1.77

63n

59n
36n

44n
39n
64n

43n
47n
21n
34n
23n
03
02
6.51
1.72

63n
42n
33n
07
03
02
6.80
2.03

34n
43n
04
04
11
6.27
1.90

6
31n
24n
41n
25n

53n
17n
13
13
7.52
2.48

31n
26n
36n
27n
49n
22n
06
02

7
14
12
04
05
10
11
16
10

04
08
02
04
04
02
07

04 7.27
02 6.51
06 7.03
12 6.20
18n 8.21
03 7.36
03 13.44
11 11.36
12.96

06

6.91 13.56 11.32


2.50 1.60
.79

SD
1.91
1.84
1.95
1.89
2.24
2.30
1.63
.14
2.45

12.7
2.5

Note. Boys are above diagonal (n 5 182); girls are below diagonal (n 5 187). PCI 5 per capita
income.
n
po05, two-tailed.

The next step involved estimating the structural equation models.4


Preliminary models confirmed our hypothesis that parents would have a
stronger effect on siblings than the reverse. Models with paths predicting
father and mother negative attributions from sibling negative attributions
showed that neither path reached significance, whereas models with paths
predicting sibling negative attributions from father and mother negative
attributions were both significant. We therefore proceeded with models
containing unidirectional (parent to sibling) effects only.
Boys and girls were analyzed separately. Preliminary models examining
effects within sibling gender demonstrated that fathers and mothers attributional activity had similar effects on their childrens negative attributions.5

4
One concern in conducting these analyses was the low reliabilities for the attribution
measures in siblings in 1990 (Cronbachs alpha 5 .62). Additional analyses (not reported)
showed that excluding siblings under age 12 in 1990 (the group with the lowest reliability) did
not substantially alter results. Thus, to maximize the sample size, we conducted analyses with
the entire sample of boys and girls.
5
We determined that there was no difference between mothers and fathers by constraining
the mother and father predictive paths to be equal to one another and comparing the resulting
chi-square with a chi-square from a model without the equality constraints. A difference equal
to or greater than 3.84 (the .05 significance level with 1 df) indicates a significant difference in the
paths. The difference was .61 for the boy model and .25 for the girl model.

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

269

TABLE 2
Standardized Maximum Likelihood Estimates of Factor Loadings for Latent Constructs

Father attributions
Mean
Accident
Good-natured
Mother attributions
Mean
Accident
Good-natured
Sibling attributions
Mean
Accident
Good-natured
Adolescent attributions
Mean
Accident
Good-natured
Sibling observed negative behavior
Hostility
Antisocial behavior
Angry coercive behavior
Adolescent observed negative behavior
Hostility
Antisocial behavior
Angry coercive behavior
Sibling observed positive behavior
Warmth and supportiveness
Prosocial behavior
Listener responsiveness
Adolescent observed positive behavior
Warmth and supportiveness
Prosocial behavior
Listener responsiveness

1990

1992

.66
.67
.61

.71
.62
.64

.69
.57
.67

.72
.71
.70

.63
.50
.64

.60
.75
.80

.61
.51
.70

.68
.80
.77

.95
.87
.84

.97
.90
.87

.96
.89
.82

.98
.91
.87

.81
.84
.79

.74
.91
.75

.79
.82
.79

.75
.91
.66

Note. See text for description of measures; n 5 370 for father, mother, and sibling attributions;
n 5 324 for adolescent attributions and behavior measures. All factor loadings po.05, two-tailed.

Therefore, we estimated all models with the father and mother paths
constrained equal to one another. Results for both models are presented in
Figure 1 (results for girls are in parentheses). All stability paths were high
(standardized path coefficients were .73 or higher for fathers and mothers,
and were .68 for boys and .62 for girls). The high stabilities for siblings

270

MATTHEWS AND CONGER

.77*

Father Attributions
1990

(.74*)

Father Attributions
1992
.12*
(.24*)

.76*
(.70*)

.49*
(.46*)

Mother Attributions
1990

.73*
(.85*)

Mother Attributions
1992
.08
.08

.60*
(.60*)

(.12*)

(.12*)

Sibling Attributions
1990

.68*
(.62*)

Sibling Attributions
1992

FIGURE 1 Standardized maximum likelihood estimates for model examining associations


among parent and sibling attributions (results for boys are outside parentheses; results for
girls are inside parentheses). Ns 5 182 (boys model) and 187 (girls model). Indicators for
latent constructs are not depicted. Paths from father and mother attributions to sibling attributions are constrained to be equal to one another. Sibling age is used as a control variable
(not depicted). npo.05, two-tailed.

suggested that a 2-year time lag may be too short for an effect of parent
negative attributions on sibling negative attributions to show up. Instead,
analyses showed that Time 1 parent negative attributions predicted sibling
negative attributions indirectly through Time 2 parent negative attributions. Parents Time 2 negative attributions were significantly associated
with girls, but not boys, negative attributions (standardized beta coefficients were .08, ns, for both father and mother paths in the boy model,
and .12 for both father and mother paths in the girl model, both pso.05,
two-tailed). The indirect effect of parentss Time 1 negative attributions
on girls Time 2 negative attributions was significant (standardized
effects 5 .08 and .09 for father and mother, respectively, both pso.05,

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

271

two-tailed). The boys model fit the data well (w2 5 157.84 with 148 df,
GFI 5 .92, CFI 5 .99, RMSEA 5 .019). The girls model also fit the data well
(w2 5 217.14 with 148 df, GFI 5 .90, CFI 5 .94, RMSEA 5 .050). On the basis
of these analyses, we accepted the first hypothesis and concluded that
parents negative attributions about an adolescent child predicted girls
(but not boys) negative attributions about the same adolescent child.

Do Sibling and Adolescent Negative Attributions Predict Own


Behavior?
Our second hypothesis was that the siblings and the adolescents negative
attributions about each other would predict changes in their respective
negative and positive behavior toward each other, whereas the behavior of
the attribution target would have little or no effect on the perceivers
negative attributions. Intercorrelations among sibling and adolescent attribution scale scores and negative and positive behavior scale scores are
presented in Table 3. Correlations showed that negative attributions were
significantly associated with both negative and positive behavior. Correlations for siblings ranged from .12 to .29 for negative behavior, and from
.22 to .40 for positive behavior. Correlations for adolescents were of
similar magnitude and were statistically significant (see Table 3).
On the basis of these correlations, we proceeded with the structural
equation models. We estimated two measurement models. The first had
eight latent constructs (siblings and adolescents negative behavior, and
sibling and adolescent negative attributions at Time 1 and Time 2). The
second measurement model was similar to the first but substituted positive behavior for negative behavior. Standardized factor loadings for the
latent constructs are presented in the bottom section of Table 2. The negative behavior measurement model fit the data well (w2 5 316.56 with 235
df, GFI 5 .93, CFI 5 .98, RMSEA 5 .033). The positive behavior measurement model also fit the data well (w2 5 315.79 with 232 df, GFI 5 .93,
CFI 5 .98, RMSEA 5 .033).
Next, we estimated the structural models. Preliminary analyses using
the multisample analysis procedure in the LISREL 8.30 software indicated
that there were no significant differences between boys and girls in any of
the models; therefore, they were combined to maximize the sample size
and simplify presentation.
First, we evaluated our hypothesis that negative attributions would
predict negative behavior, and not vice-versa. To maintain the five cases
per parameter estimated that is recommended for reliable results (Carmines & McIver, 1981), we constrained parallel siblingadolescent paths to

M
SD

50n
31n
31n

29n
38n

49n

25n
12n
20n
11n

21n
29n
14n
28n

40n
22n
36n
25n
19n
15n
10

25n
36n
26n
40n
00
18n
05

7.91
2.46

7.22
2.48

16n
19n
18n
20n

23n
29n
25n
30n

46n
74n
40n

45n
82n

47n

22n
22n
24n
22n
20n
09
03

31n
40n
27n
36n
18n
19
04

31n
28n
33n
22n
06
19n
11n

14n
44n
17n
39n
09
18n
01

12.43
7.23

13.95
6.84

7.98
2.29

7.19
2.43

Note. See text for descriptions of measures; N 5 324.


po.05, two-tailed.

10

34n
30n
35n
24n
22n
15n
01

14n
41n
22n
46n
19n
23n
04

47n
78n
35n
09
16n
06

40n
71n
17n
24n
00

13.01
7.26

14.22
7.10

13.64
4.81

11.58
4.6

11

45n
08
17n
04
13.77
4.84

12

14n
25n
04
11.55
4.57

13

14

07
05

03

12.40
2.36

13.44
1.59

15

11.33
.60

MATTHEWS AND CONGER

Attributions
1. Sibling 1990
2. Sibling 1992
3. Adol 1990
4. Adol 1992
Neg behavior
5. Sibling 1990
6. Sibling 1992
7. Adol 1990
8. Adol 1992
Pos behavior
9. Sibling 1990
10. Sibling 1992
11. Adol 1990
12. Adol 1992
13. Sibling age
14. Parents ed.
15. PCI

272

TABLE 3
Intercorrelations, Means, and Standard Deviations Among Attribution Scale Scores, Observer-Rated Behavior Scale Scores, and Control Variables

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

273

be equal to one another.6 Results are presented in Figure 2 (results for the
model examining negative behavior are outside the parentheses). Models
showed that, as expected, siblings negative attributions predicted their
negative behavior toward the partner, whereas negative behavior by the
partner did not predict the other siblings negative attributions. Time 1
sibling and adolescent negative attributions predicted own Time 2 negative behavior (standardized betas 5 .10 and .09 for siblings and adolescents, respectively, po.05, two-tailed). Additional analyses (not shown)
indicated that adolescents Time 1 negative behavior did not significantly
predict the siblings Time 2 negative attributions (standardized beta 5 .03,
p4.05), nor did the siblings Time 1 negative behavior predict Time 2
adolescent negative attributions (standardized beta 5 .02, p4.05). The
negative behavior structural model fit the data well (w2 5 466.75 with 291
df, GFI 5 .90, CFI 5 .97, RMSEA 5 .042).
Results for the positive behavior model are also presented in Figure 2
(coefficients are in parentheses). Multicollinearity in the positive behavior
models (specifically, high correlations between sibling and adolescent
warm behavior) created difficulty in estimating the models. To estimate
the models, we assumed that positive behavior is no more likely to be
exchanged between siblings than negative behavior, and we set the reciprocal paths between Time 2 positive behavior equal to those of Time 2 negative behavior. Comparison of the chi-square statistics of the models with
and without set paths showed that setting the positive paths equal to the
negative paths did not significantly affect model fit (difference in w2 5 2.02).
Results presented in Figure 2 (in parenthees) show that, as with the
negative behavior model, Time 1 sibling and adolescent negative attributions predicted own Time 2 positive behavior (standardized beta 5 .10
for both siblings and adolescents, po.05, two-tailed). Next, we used the
same procedure used earlier in the negative behavior models to test
whether the positive behavior of the attribution target predicted negative
attributions. For example, we added a predictive path from Time 1
adolescent behavior to Time 2 sibling negative attributions to a model
already containing the attribution-to-behavior paths (i.e., to the model
6
We felt confident that for most of the pathsspecifically, the stability paths and the path
predicting change in behavior from attributionssiblings and adolescents could reasonably be
considered interchangeable, and thus equality constraints would be appropriate. However,
studies cited earlier in the literature review indicated that older siblings tend to influence
younger siblings more strongly than the reverse (Updegraff et al., 2000), raising a question
regarding the appropriateness of assuming that adolescents and siblings could be considered
interchangeable. However, tests comparing the chi-square statistics of models with equality
constraints with those without equality constraints showed no differences between parallel
sibling and adolescent paths. Thus, we imposed the equality constraints on parallel paths in all
models.

274

MATTHEWS AND CONGER

Sibling Attributions
about Adolescent
1990

.52*
(.54*)

.28*
(-.48*)

.20*
(-.45*)

.78*
(.77*)

.15*
(.13*)

.10*
(-.10*)

Sibling ObserverRated Behavior


to Adolescent
1990

.31*
(.31*)

Sibling ObserverRated Behavior


to Adolescent
1992

.33*
(.33+)

.80*
(.81*)

Adolescent
Observer-Rated
Behavior to
Sibling 1990

.23*
(-.37*)

Sibling Attributions
about Adolescent
1992

.28*
(-.38*)

.31*
(.30*)

.35*
(.35+)

.15
(.07)

Adolescent
Observer-Rated
Behavior to
Sibling 1992

.09*
(-.10*)

Adolescent
Attributions about
Sibling 1990

.69*
(.69*)

Adolescent
Attributions about
Sibling 1992

FIGURE 2 Standardized maximum likelihood estimates for models examining sibling and
adolescent attributions and observer-rated sibling and adolescent behavior (results for negative behavior are outside parentheses; results for positive behavior are inside parentheses).
N 5 324 for both models. Indicators for latent constructs are not depicted. Only hypothesized
paths are included (see the Results section for details). Parallel paths are constrained to be
equal across sibling and adolescent, including: sibling and adolescent attribution stability
paths, sibling and adolescent behavior stability paths, sibling and adolescent behavior reciprocal paths, and attribution-to-behavior paths. Positive behavior reciprocal paths are set
equal to those of negative behavior reciprocal paths (see the Results section for further
information). Sibling age and parents education are used as control variables (not depicted).
n
po.05, two-tailed; 1path coefficients set (see text for details).

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

275

presented in Figure 2). The new paths were marginally significant (standardized betas from behavior to attributions were both
.10, p 5 .05,
t 5 1.96). The paths from negative attributions to behavior were not
reduced in magnitude, and remained significant (t 5 2.31). The magnitude and significance of the standardized betas suggest that the two sets
of paths are equivalent. However, the addition of the behavior-to-attribution paths was just under significance (change in chi-square was 3.64). In
contrast, additional analyses adding attribution-to-behavior paths to a
model already containing behavior-to-attribution paths did improve
model fit (change in chi-square was 4.70). These results suggest that although both sets of paths have an effect, the paths predicting behavior
from negative attributions are somewhat stronger. The final positive behavior structural model (shown in Figure 2) fit the data well (w2 5 501.07
with 285 df, GFI 5 .90, CFI 5 .95, RMSEA 5 .047).
On the basis of these analyses, we concluded that siblings and adolescents negative attributions predicted their own negative and positive behaviors toward one another. In contrast, the path from sibling positive
behavior to partners negative attributions was only marginally significant.

Does Gender Constellation or Birth Order Influence Negative


Attributions?
To test the three hypotheses concerning the effects of gender constellation
and birth order on negative attributions (Hypotheses 3 through 5), we
performed a 4 (gender constellation: boyboy, girlgirl, boygirl, girl
boy)  2 (birth order: older sibling, younger sibling)  2 (time)  3 (family
member: father, mother, and sibling) mixed-model analysis of variance
(ANOVA) using the mixed model procedure available in the SPSS 11.0
computer program. The negative attribution scale score (of siblings, adolescents, mothers, and fathers) was used as the dependent variable, and
the repeated measures were family member and time. Our third hypothesis was that same-sex sibling pairs would be less negative in their attributions than opposite-sex siblings. Contrary to expectations, a planned
contrast showed that there was no significant difference between same-sex
and opposite-sex gender pairs (t 5 .426, p4.05). However, consistent
with expectations, planned contrasts showed that girl pairs were significantly less negative than boy pairs in their attributions (t 5 3.59,
po.05). Mean negative attribution scale scores are presented in Table 4.
Our fourth hypothesis stated that older siblings would be less positive
in their attributions about the adolescent than younger siblings, and our
fifth hypothesis stated that this difference would lessen over time. The

276

MATTHEWS AND CONGER

TABLE 4
Comparison of Mean Attribution Scale Scores by Gender Constellation of Sibling Pair
Gender Constellation of Sibling Pair

Time 1 (1990)
(SD)
Time 2 (1992)
(SD)

GirlGirl
(n 5 100)

BoyBoy
(n 5 89)

BoyGirl
(n 5 87)

GirlBoy
(n 5 93)

7.11
(2.48)
6.39
(2.57)

8.53
(2.17)
7.58
(2.08)

8.00
(2.40)
7.52
(2.29)

7.91
(2.28)
7.15
(2.49)

Note. A high score indicates a more negative attribution. Planned contrasts show no difference between same-sex and opposite-sex sibling pairs. Girlgirl pairs are significantly less
negative than boyboy pairs.

mixed-model ANOVA showed that the birth order main effect was not
significant, F(1, 2272) 5 2.709, p4.05. However, birth order had significant
interactions with two other variables: family member, F(2, 2272) 5 3.196,
po.05, and time, F(1, 2272) 5 4.134, po.05. Independent samples t tests
showed significant differences in mean negative attribution score by birth
order for siblings, but not for fathers or mothers, and only in 1990
(t 5 2.76, po.05). This significant interaction (which confirmed Hypothesis 5) is presented in Figure 3, which shows a significant difference
between younger and older siblings in 1990, with older siblings being
more negative in their attributions than younger siblings. By 1992, both
younger and older siblings had become less negative, with older siblings
showing the greatest decrease in negativity, coming to equal younger siblings in their evaluations of the adolescent.
In sum, in contrast to expectations that same-sex sibling pairs would be
less negative in their attributions than opposite-sex sibling pairs, our results showed that there was no difference. Thus, we rejected our third
hypothesis. These analyses also showed that younger siblings made less
negative attributions than older siblings but that this difference declined
over time. Thus, we accepted our fourth and fifth hypotheses.

DISCUSSION
Although the sibling bond is recognized as an important close relationship, it is relatively understudied (Jenkins, 1992; Leder, 1991). Furthermore, despite the demonstrated importance of attributions in other close
relationships, such as marriage (Bradbury & Fincham, 1990; Fincham,

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

277

FIGURE 3 Mean attribution scale score for younger and older siblings in 1990 and 1992.
Ns 5 173 younger siblings and 191 older siblings. Differences between older and younger
siblings are statistically significant in 1990 (po.05).

2001; Fincham et al., 1990; Fletcher & Fincham, 1991), no study to our
knowledge has specifically examined attributions within the sibling relationship. In this study, we addressed this lacuna in the literature by
examining two possible sources of child or adolescent negative attributions about other family members: parent attributions, which have been
proposed as a correlate of child or adolescent attributions about other
family members (Bugental & Johnston, 2000; Grusec & Mammone, 1995),
and birth order and gender constellation of the sibling pair, which has
been shown to have an important influence on sibling relationship dynamics (Buhrmester, 1992; Hoffman, 1991). In addition, we investigated
associations among negative attributions and behavior, both the possibility

278

MATTHEWS AND CONGER

that the behavior of the attribution target predicts negative attributions,


which is supported by classical attribution theory ( Jones & Davis, 1965),
and the possibility that negative attributions predict own behavior, which
is suggested by empirical studies of marriage and of child interactions
with peers outside the family (Dodge, 1980; Dodge & Frame, 1982; Gottman,
1993; Jones & Davis, 1965; Lochman, 1987), and the possibility that negative
attributions predict own behavior, which is suggested by empirical studies
of marriage (Gottman, 1993). The results from this study add to our
knowledge of the correlates of negative attributions within the family and
of the role that negative attributions play in sibling relationships.
Our findings indicated that fathers and mothers negative attributions
predicted daughters concurrent negative attributions, but not sons. As
expected, we found that the effect was unidirectional, with parents negative attributions predicting siblings attributions, but not the reverse.
Although family scholars have suggested the possibility that relationship
attributions are transmitted intergenerationally (e.g., Grusec et al., 1994),
to our knowledge this is only the second study to document a direct link
between parents and childrens attributions regarding a family member.
Fincham et al. (1998) found correlations between parents marital attributions and childrens attributions about their parents for both boys and
girls, but only within same-sex pairs (i.e., motherdaughter and father
son). Our findings highlight the importance of fathers, as well as mothers,
in the cognitive lives of daughters, thus giving us a fuller picture of the role
of parents in transmitting relationship attributions to their children.
What could account for our lack of findings regarding boys, especially
in light of Fincham et al.s (1998) finding that boys attributions were
correlated with those of their fathers? Social scientists propose that women
are socialized to be more focused on the relational aspects of interpersonal
interactions than are men (Chodorow, 1978; Gilligan, 1982). It is possible
that girls are more interested in the underlying motives and intent of
interactional partners than are boys and for this reason would be more
attentive to attributions articulated in their presence. In addition, the siblings in our sample, aged 10 to 18, were for the most part older than the
children sampled by Fincham et al. (10- to 12-year-olds). Thus, one possible explanation for our conflicting results is that boys are more easily
influenced by parents when younger (and are relatively impervious to
influence as teenagers), whereas girls are influenced by parents throughout childhood and adolescence.
Our analyses also indicated that birth order and the gender constellation of the sibling pair structured siblings negative attributions. Prior
studies have shown that younger siblings generally view their brother or
sister in more positive terms than do older siblings but that the sibling

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

279

relationship also tends to become more egalitarian and emotionally


distant over time. Our findings were consistent with previous research:
Younger siblings tended to evaluate their brother or sister as possessing
more positive motives and intent than older siblings. However, the
present study also showed that this statistically significant difference had
disappeared 2 years later, with older siblings coming to adopt a more
benign view of their brother or sister. It may be that as the younger target
adolescent matured, he or she became more independent, and less of a
bother or hassle to the older sibling, who then formed more positive
attributions about the adolescent. Our results also indicated that contrary
to our expectations, same-sex sibling pairs were just as negative in their
attributions as cross-sex sibling pairs. However, consistent with our hypothesis and with the prior literature (e.g., McHale et al., 2000), girl samesex pairs were less negative than boy same-sex pairs. This finding could
also be a reflection of girls being more attuned to the relational aspects of
interpersonal bonds than boys, as discussed earlier. Indeed, this possibility seems more likely given that the two cross-sex sibling pairs (boysgirls
and girlsboys) fall in between the boyboy and girlgirl pairs in their
level of attributional negativity. That is, the more girls there are in the
sibling pair, the less negative are the attributions.
Finally, consistent with our hypotheses, study results indicated that for
both the sibling and the adolescent, negative attributions predicted their
own level of negativity in behavior 2 years later, whereas the negative
behavior of the attribution target did not predict the perceivers negative
attributions. Other studies have demonstrated that children who make
hostile attributions about peers outside of the family behave negatively
toward the attribution target (e.g., Dodge, 1980; Dodge & Frame, 1982;
Lochman, 1987). Our results extend these findings to peer relationships
within the family as well (i.e., siblings). Results from the positive behavior
models were more ambiguous, with the positive behavior-to-attribution
path being statistically significant and only slightly weaker than the attribution-to-behavior path. How can we explain this disparity between
negative and positive behavior? One possibility is that positive behavior
may more unusual than negative behavior and thus more likely to elicit
attributional activity than negative behavior.
The greater stability of cognitions relative to negative behavior (and
thus the tendency for attributions to drive negative behavior rather than
the reverse) has important implications for relationship functioning.
Studies from the marital literature show that husbands and wives tend to
interpret spousal behavior in a way that confirms their overall sentiments
regarding the marriage. For example, distressed individuals may interpret
negative partner behavior as due to stable, dispositional causes, and thus

280

MATTHEWS AND CONGER

as characteristic of the spouse. At the same time, they may dismiss positive
spousal behavior as due to temporary, situational causes, concluding that
it is not reflective of the spouses true character or feelings (Bradbury &
Fincham, 1992; Gottman, 1994). These studies suggest that the importance
of attributions for relationships may lie in their relative invulnerability to
disconfirming evidence. Although we did not have the data to test formally whether participants were more likely to perceive attribution-congruent (vs. attribution-incongruent) behavior in their interaction partner,
our finding that siblings negative attributions predict their own behavior
more consistently than the behavior of the partner predicts negative attributions is consistent with this process. Further studies should examine
whether this process, found in married couples, extends to other types of
close relationships, such as siblings.
An important limitation of this study concerns our measurement of
negative attributions. Our measure was short and not derived from prior
theorytwo factors that probably contributed to its low reliability. Although much progress was made in the 1990s with regard to the measurement of relationship attributions (Fincham et al., 1998; Fincham &
Bradbury, 1992), we were limited by the attribution measures that were
available to us. Future research should take advantage of the new attribution measures that are now available, such as the childrens relationship
attribution measure (Fincham et al., 1998).
Another factor that should be kept in mind when reviewing these results concerns the fact that behavior was measured with an observer rating
of a 15-min interaction task, whereas attributions were measured with a
self-report. Although using different measurement modalities to assess
study concepts solves one set of problemsthose due to method variancethey introduce other potential problems. For example, the self-reported attributions drew on a longer time frame and encompassed more
social situations than the observational task. Both of these factors could
have increased the stability of attributions relative to behavior, which in
turn could have influenced our finding that attributions predict behavior
rather than the reverse. In addition, conclusions regarding causation can
never be drawn from correlational analyses, such as structural equation
modeling. Thus, future research should examine alternative explanations
for our findings, such as the possibility that associations between parent and
daughter attributions are due to shared genetics or social contextual factors.
Finally, the generalizability of our results is limited because our sample is
exclusively White and rural. Replication using non-White and urban samples will extend our understanding of attribution processes in families.
This study can be extended in several ways. Scholars have suggested
that early dyadic experiences can establish conceptual models and

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

281

expectations regarding close relationship functioning that influence future


relationships, resulting in continuity in close relationship functioning
across the life course (Dodge, 1985; Dunn, 1993; Hartup & Stevens, 1997;
Howes, Hamilton, & Philipsen, 1998; Ladd, Kochenderfer, & Coleman,
1996; Vandell, 2000). Thus, an important next step would be to investigate
whether attributions learned in the family of origin are applied to relationships outside of the family, such as friends or romantic partners. It may
be, for example, that relationship attributions are a factor in the oft-cited
finding that children of divorce are more likely to get divorced themselves
(McClanahan & Bumpass, 1988; White, 1990).
The present report underscores the importance of negative attributions in
close relationships by addressing a previously unstudied relationship:
the sibling bond. Although many studies have demonstrated the role that
attributions play in marriage, this is the first study to our knowledge that
shows that negative attributions play a similarly important role in the sibling
relationship by predicting both positive and negative behavior. This study
has also advanced our knowledge by showing that both fathers and mothers
have a direct influence on childrens negative attributions, at least for
daughters, and adds to our knowledge regarding intergenerational transmission processes in general. Finally, this study adds to our knowledge of
the effects of birth order and gender constellation, showing that they structure not only behavior, as found in previous studies, but also cognitions.

ACKNOWLEDGMENTS
During the past several years support for this research has come
from multiple sources including the National Institute of Mental
Health (MH00567, MH19734, MH43270, MH48165, MH51361), the National Institute on Drug Abuse (DA05347), the Bureau of Maternal and
Child Health (MCJ-109572), the MacArthur Foundation Research Network on Successful Adolescent Development among Youth in High-Risk
Settings, and the Iowa Agriculture and Home Economics Experiment
Station (Project 3320). The authors would like to thank Thomas Bradbury,
James Burk, Roger Clark, Howard Kaplan, and Alex McIntosh for their
comments on an earlier version of this paper.
REFERENCES
Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral
and emotional problems: Implications of cross-informant correlations for situational
specificity. Psychological Bulletin, 101, 213232.

282

MATTHEWS AND CONGER

Bandura, A. (1971). Psychological modeling: Conflicting theories. Chicago: Aldine-Atherton.


Baucom, D. H., & Epstein, N. (1990). Cognitive-behavioral marital therapy. New York: Brunner/
Mazel.
Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107,
238246.
Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.
Bradbury, T. N., & Fincham, F. D. (1990). Attributions in marriage: Review and critique.
Psychological Bulletin, 107, 333.
Bradbury, T. N., & Fincham, F. D. (1992). Attributions and behavior in marital interaction.
Journal of Personality and Social Psychology, 63, 613628.
Brown, M. W., & Cudeck (1993). Alternative ways of assessing model fit. In K. A. Bollen, &
J. S. Lang (Eds.), Testing Structural equation models (pp. 136162). Newbury Park, CA: Sage.
Bugental, D. B., & Goodnow, J. J. (1998). Socialization processes. In N. Eisenberg (Ed.),
Handbook of child psychology: Vol. 3. Social, emotional, and personality development (pp.
389462). New York: Wiley.
Bugental, D. B., & Johnston, C. (2000). Parental and child cognitions in the context of the
family. Annual Review of Psychology, 51, 315344.
Bugental, D. B., Johnston, C., New, M., & Silvester, J. (1998). Measuring parental attributions:
Conceptual and methodological issues. Journal of Family Psychology, 12, 459480.
Bugental, D. B., & Martorell, G. (1999). Competition between friends: The joint influence of
the perceived power of self, friends, and parents. Journal of Family Psychology, 13, 260273.
Buhrmester, D. (1992). The developmental courses of sibling and peer relationships. In
F. Boer, & J. Dunn (Eds.), Childrens sibling relationships: Developmental and clinical issues
(pp. 1940). Hillsdale, NJ: Erlbaum.
Buhrmester, D., & Furman, W. (1990). Perceptions of sibling relationships during middle
childhood and adolescence. Child Development, 61, 13871398.
Carmines, E. G., & McIver, J. P. (1981). Analyzing models with unobserved variables: Analysis
of covariance structures. In G. W. Bohrnstedt, & E. F. Borgatta (Eds.), Social measurement:
Current issues (pp. 65115). Beverly Hills, CA: Sage.
Chodorow, N. (1978). The reproduction of mothering. Berkeley: University of California Press.
Davis, J. A. (1985). The logic of causal order. Beverly Hills, CA: Sage.
Dix, T. H., & Grusec, J. E. (1985). Parent attribution processes in the socialization of children.
In I. E. Siegal (Ed.), Parental belief systems: The psychological consequences for children (pp.
201233). Hillsdale, NJ: Erlbaum.
Dodge, K. A. (1980). Social cognition and childrens aggressive behavior. Child Development,
51, 162170.
Dodge, K. A. (1985). Attributional bias in aggressive children. Advances in Cognitive-Behavioral
Research and Therapy, 4, 73110.
Dodge, K. A., Bates, J. E., & Pettit, G. S. (1990). Mechanisms in the cycle of violence. Science,
250, 16781683.
Dodge, K. A., & Frame, C. L. (1982). Social cognitive biases and deficits in aggressive boys.
Child Development, 53, 620635.
Dodge, K. A., Pettit, G. S., Bates, J. E., & Valente, E. (1995). Social information-processing
patterns partially mediate the effect of early physical abuse on later conduct problems.
Journal of Abnormal Psychology, 104, 632643.
Dunn, J. (1993). Young childrens close relationships: Beyond attachment. Newbury Park, CA:
Sage.
Fincham, F. D. (2001). Attributions in close relationships: From Balkanization to integration.
In G. J. O. Fletcher, & M. S. Clark (Eds.), Blackwell handbook of social psychology (pp. 331).
Malden, MA: Blackwell.

FAMILY CORRELATES OF NEGATIVE ATTRIBUTIONS

283

Fincham, F. D., Beach, S. R. H., Arias, I., & Brody, G. H. (1998). Childrens attributions in the
family: The childrens relationship attribution measure. Journal of Family Psychology, 12,
481493.
Fincham, F. D., & Bradbury, T. N. (1987). The impact of attributions in marriage: A longitudinal analysis. Journal of Personality and Social Psychology, 53, 481489.
Fincham, F. D., & Bradbury, T. N. (1991). Cognition in marriage: A program of research on
attributions. Advances in Personal Relationships, 2, 159203.
Fincham, F. D., & Bradbury, T. N. (1992). Assessing attributions in marriage: The relationship
attribution measure. Journal of Personality and Social Psychology, 62, 457468.
Fincham, F. D., Bradbury, T. N., & Scott, C. K. (1990). Cognition in marriage. In F. D. Fincham, &
T. N. Bradbury (Eds.), The psychology of marriage (pp. 118149). New York: Guilford Press.
Fletcher, G. J. O., & Fincham, F. D. (1991). Attribution processes in close relationships. In
G. J. O. Fletcher, & F. D. Fincham (Eds.), Cognition in close relationships (pp. 735). Hillsdale,
NJ: Erlbaum.
Gilligan, C. (1982). In a different voice: Psychological theory and womens development. Cambridge,
MA: Harvard University Press.
Gottman, J. M. (1993). A theory of marital dissolution and stability. Journal of Family Psychology, 1, 5775.
Gottman, J. M. (1994). What predicts divorce? The relationship between marital processes and marital
outcomes. Hillsdale, NJ: Erlbaum.
Grusec, J. E., Hastings, P., & Mammone, N. (1994). Parenting cognitions and relationship
schemas. In J. G. Smetana (Ed.), Beliefs about parenting: Origins and developmental implications (pp. 519). San Francisco: Jossey-Bass.
Grusec, J. E., & Mammone, N. (1995). Features and sources of parents attributions about
themselves and their children. Review of Personality and Social Psychology, 15, 4973.
Hartup, W. W., & Stevens, N. (1997). Friendships and adaptation in the life course. Psychological Bulletin, 121, 355370.
Hastings, P. D., & Grusec, J. E. (1998). Parenting goals as organizers of responses to parent
child disagreement. Developmental Psychology, 34, 465479.
Hoffman, L. W. (1991). The influence of the family environment on personality: Accounting
for sibling differences. Psychological Bulletin, 110, 187203.
Hogan, D. M., Etz, K. E., & Tudge, J. R. H. (1999). Reconsidering the role of children in family
research: Conceptual and methodological issues. In C. L. Shehan (Ed.), Contemporary
perspectives on family research: Vol. 1. Through the eyes of the child: Re-visioning children as
active agents in family life (pp. 329351). Stanford, CT: JAI Press.
Howes, C., Hamilton, C. E., & Philipsen, L. C. (1998). Stability and continuity of child caregiver and childpeer relationships. Child Development, 69, 418426.
Jenkins, J. (1992). Sibling relationships in disharmonious homes: Potential difficulties and
protective effects. In F. Boer, & J. Dunn (Eds.), Childrens sibling relationships (pp. 125138).
Hillsdale, NJ: Erlbaum.
Johnson, M. D., Karney, B. R., Rogge, R., & Bradbury, T. N. (2001). The role of marital behavior
in the longitudinal association between attributions and marital quality. In V. Manusov, &
J. H. Harvey (Eds.), Attribution, communication behavior, and close relationships (pp. 173192).
New York: Cambridge University Press.
Jones, E. E., & Davis, K. E. (1965). From acts to dispositions. In L. Berkowitz (Ed.), Advances in
experimental social psychology (Vol. 2, pp. 220266). New York: Academic Press.
Karney, B. R., & Bradbury, T. N. (2000). Attributions in marriage: State or trait? A growth
curve analysis. Journal of Personality and Social Psychology, 78, 295309.
Kessler, R. C., & Greenberg, D. F. (1981). Linear panel analysis: Models of quantitative change.
New York: Academic Press.

284

MATTHEWS AND CONGER

Ladd, G. W., Kochenderfer, B. J., & Coleman, C. C. (1996). Friendship quality as a predictor of
young childrens early school adjustment. Child Development, 67, 11031118.
Leder, J. M. (1991). Brothers and sisters: How they shape our lives. New York: St Martins Press.
Lochman, J. E. (1987). Self- and peer perceptions and attributional biases of aggressive and
nonaggressive boys in dyadic interactions. Journal of Consulting and Clinical Psychology, 55,
404410.
Matthews, L. S., Conger, R. D., & Wickrama, K. A. S. (2000). Marital commitment and marital
interaction: Trajectories of growth and decline. Manuscript submitted for publication.
McClanahan, S., & Bumpass, L. (1988). Intergenerational consequences of family disruption.
American Journal of Sociology, 94, 130152.
McHale, S. M., Updegraff, K. A., Tucker, C. J., & Crouter, A. C. (2000). Step in or stay out?
Parents roles in adolescent siblings relationships. Journal of Marriage and the Family, 62,
746760.
Melby, J., Conger, R. D., Book, R., Rueter, M., Lucy, L., Repinski, D., et al. (1989). The Iowa family
interaction coding manual (1st ed.). Unpublished manuscript, Iowa State University at Ames.
Melby, J., Conger, R. D., Ge, X., & Warner, T. D. (1995). The use of structural equation modeling
in assessing the quality of marital observations. Journal of Family Psychology, 9, 280293.
Miller, N. E., & Dollard, J. (1941). Social learning and imitation. New Haven, CT: Yale University
Press.
Minnett, A. M., Vandell, D. L., & Santrock, J. W. (1983). The effects of sibling status on sibling
interaction: Influence of birth order, age spacing, sex of child, and sex of sibling. Child
Development, 54, 10641072.
Munton, A. G., & Antaki, C. (1988). Causal beliefs amongst families in therapy: Attributions at
the group level. British Journal of Clinical Psychology, 27, 9197.
Nasby, W., Hayden, B., & DePaulo, B. M. (1980). Attributional bias among aggressive boys to
interpret unambiguous social stimuli as displays of hostility. Journal of Abnormal Psychology, 89, 459468.
Pepler, D., Corter, C., & Abramovitch, R. (1982). Social relations among children: Siblings
and peers. In K. Rubin, & H. Ross (Eds.), Peer relationships and social skills in childhood
(pp. 20922). New York: Springer-Verlag.
Quiggle, N. L., Garber, J., Panak, W. F., & Dodge, K. A. (1994). Social information processing in
aggressive and depressed children. Child Development, 63, 13051320.
Schwartz, J. C., Barton-Henry, M. L., & Pruzinsky, T. (1985). Assessing child-rearing behaviors: A comparison of ratings made by mother, father, child, and sibling on the CRPBI.
Child Development, 56, 462479.
Updegraff, K. A., McHale, S. M., & Crouter, A. C. (2000). Adolescents sex-typed friendship
experiences: Does having a sister versus a brother matter? Child Development, 71,
15971610.
Vandell, D. L. (2000). Parents, peer groups and other socializing influences. Developmental
Psychology, 36, 699710.
Vandell, D. L., Minnett, A. M., & Santrock, J. W. (1987). Age differences in sibling relationships
during middle childhood. Journal of Applied Developmental Psychology, 8, 247257.
Waas, G. A. (1988). Social attributional biases of peer-rejected and aggressive children. Child
Development, 59, 969992.
Weiss, B., Dodge, K. A., Bates, J. E., & Pettit, G. S. (1992). Some consequences of early harsh
discipline: Child aggression and a maladaptive social information processing style. Child
Development, 63, 13211335.
White, L. K. (1990). Determinants of divorce: A review of research in the eighties. Journal of
Marriage and the Family, 52, 904912.

You might also like