You are on page 1of 9

Journal of Family Psychology

2005, Vol. 19, No. 2, 180 188

Copyright 2005 by the American Psychological Association


0893-3200/05/$12.00
DOI: 10.1037/0893-3200.19.2.180

Attributions and Anger in Early Marriage: Wives Are


Event-Dependent and Husbands Are Schematic
Keith Sanford
Baylor University
Two types of attribution believed to predict anger in married couples were investigated.
Wives anger was expected to be predicted by event-dependent attributions, appraisals based
on the unique aspects of ones current situation. Husbands anger was expected to be
predicted by schematic attributions, appraisals based on ones global sentiment in the
relationship. Seventy-seven recently married couples attended 2 assessment sessions, and
each couple identified 4 incidents pertaining to unresolved relationship issues. Participants
rated their event-dependent attributions and their anger prior to a discussion for each incident.
They also completed questionnaires regarding schematic attributions and relationship sentiment. Hierarchical linear modeling was used to distinguish between the 2 types of attribution.
Strong support was found for the expected gender differences. Results suggest that wives are
particularly attentive to the details of interpersonal interaction.
Keywords: attributions, marriage, couples, emotion, gender differences

There is a robust relationship between satisfaction in


marriage and the types of attribution couples make for
negative relationship events (Bradbury & Fincham, 1990).
When there is conflict in a marriage, distressed couples tend
to view each other as the cause of the problem and to ascribe
blame to each other. When couples make negative attributions, they are more likely to engage in destructive forms of
communication when discussing areas of conflict (Bradbury, Beach, Fincham, & Nelson, 1996; Bradbury & Fincham, 1992; Miller & Bradbury, 1995), which in turn is
predictive of relationship dissatisfaction and instability
(Karney & Bradbury, 1995). Along this line, Karney and
Bradbury (2000) demonstrated that attributions in marriage
often change over time and that changes in attributions
predict longitudinal changes in relationship satisfaction.
Although it is clear that attributions play an important
role in relationship process, one issue that has received
relatively little attention is how attributions are related to
emotion in marriage. Presumably, emotions are a product of
human evolutionary history and serve important adaptive
functions (Smith & Lazarus, 1990). To the extent that
attachment relationships are essential for human survival,
emotions are likely to play an important role in the maintenance and dissolution of close interpersonal relationships.
Emotions prepare and motivate a person to act, shape the

way a situation is perceived, and, when expressed, influence


the responses of others. Thus, it is likely that emotion plays
an important role in marital process. It is possible that one
reason attributions are strongly predictive of marital outcomes is that attributions are closely linked with emotion.
Along this line, Weiner (1985) suggested that one important function of attributions is that they lead to emotion.
Specifically, attributions are expected to produce anger
when a negative event is attributed to the partners controllable behavior. Similarly, Lazaruss (2001) appraisal theory
of emotion suggests that a person is likely to feel anger if he
or she blames another for a salient negative event. Accordingly, research indicates that attributions or appraisals regarding blame, locus, and controllability are correlated with
anger (Smith, Haynes, Lazarus, & Pope, 1993). In research
using laboratory and field simulations, participants report
feeling more anger when led to believe that a confederates
negative behavior is internal and controllable rather than
external and uncontrollable (Weiner, Amirkhan, Folkes, &
Verette, 1987). In addition, when primed to make blaming
attributions, participants are more likely to experience anger
in response to a confederates rebuff (Neumann, 2000). In
research specifically involving married couples, negative
attributions have been correlated with self-report ratings of
trait hostility (Senchak & Leonard, 1993) and negative
affect (Fincham, Beach, & Nelson, 1987). In addition, Fincham and Bradbury (1992) found that attributions correlate
with both self-report ratings of anger and behavioral ratings
of anger made during problem-solving discussions.
Taken together, it appears that negative attributions are
indeed correlated with anger in marriage. However, to fully
understand the link between attributions and anger, it is
necessary to consider an important distinction made by
Bugental and Johnston (2000) between schematic and

Valuable input and support was provided by Kristen Sanford,


who unexpectedly passed away on November 16, 2002. This study
was supported in part by a grant from the Baylor University
Research Committee.
Correspondence concerning this article should be addressed to
Keith Sanford, Department of Psychology and Neuroscience, Baylor University, One Bear Place, No. 97334, Waco, TX 76798
7334. E-mail: keith_sanford@baylor.edu

180

ATTRIBUTIONS AND ANGER IN MARRIAGE

event-dependent cognition. Schematic cognition includes


attributions that are made not in response to a particular
context but on the basis of ones schemata or memories
regarding how things typically are in ones relationship.
According to Smith and Kirbys (2001) appraisal theory of
emotion, this type of cognition can become activated automatically and largely outside of conscious awareness
through priming or spreading activation. When schemata or
memories are activated, the associated appraisals also become activated, and this is believed to produce emotion. To
the extent that schemata tend to be stable over time, this
process would produce a link between attribution and emotion that remains relatively consistent across contexts. For
example, a negative relationship event could automatically
activate a persons schema of spouse blame or marital
dissatisfaction, which in turn could produce anger, and this
could occur regardless of the conscious attributions the
person might make for his or her spouses behavior in that
particular situation. At the schematic level, then, the link
between attributions and emotion is similar to Weisss
(1980) concept of sentiment override, whereby couples respond to a situation on the basis of their overall sentiment
toward the relationship as a whole rather than on the basis
of the unique features of the situation.
In contrast, event-dependent cognition includes attributions that are made when a person consciously attends to the
unique characteristics of a particular situation and engages
in a deliberate appraisal. In their appraisal theory of emotion, Smith and Kirby (2001) describe this as cognition
based on reasoning. Importantly, event-dependent attributions will change or fluctuate from one situation to the next,
whereas schematic attributions do not. This fluctuation
could be produced by changes in the situation, changes in
the relationship, changes in a persons mood or goals, or
changes in a persons knowledge. The defining feature of an
event-dependent effect is that when the attributions change,
there is a corresponding change in emotion. For example, if
a wife consciously appraises a negative relationship event
and concludes that her husband is to blame, she may feel
angry for that particular event, but she may not feel angry
for a negative event occurring a few days later where she
excuses him from blame.
Importantly, event-dependent and schematic effects are
not mutually exclusive. It is likely that any given attribution
will be part schematic and part event dependent. This raises
a question, however, of whether both parts are active ingredients in predicting anger. At the event-dependent level, do
within-person changes in attribution predict corresponding
changes in anger? At the schematic level, do schematic
attributions or measures of relationship sentiment predict
anger over and above what can be explained by eventdependent attributions?
Recent research on gender differences in marriage suggests that the answers to these questions might be different
for husbands and wives. An interesting and consistent finding in attribution research is that correlations between attributions and behavior are stronger for wives than for husbands (Bradbury et al., 1996; Bradbury & Fincham, 1992;
Miller & Bradbury, 1995). This research finds that wives

181

attributions predict behavior during problem-solving conversations and that husbands attributions are sometimes
unrelated to behavior. Other research on social support in
marriage has also found that wives are more responsive to
the immediate context than are husbands (Carels & Baucom, 1999; Cutrona & Suhr, 1994). In previous research
specifically distinguishing between event-dependent and
schematic cognition, Sanford (2003a) found that only
wives event-dependent (or proximal) expectancies predict
behavior. One explanation for this gender difference is that
wives engage in more bottom-up processing, whereas husbands engage in more top-down processing. That is, wives
may be more attentive to the subtle details of interpersonal
interaction (Acitelli, 1992), whereas husbands may be less
sensitive to changes in the relationship and respond on the
basis of their overall sentiment. If this is true, then the link
between event-dependent attributions and anger should be
stronger for wives than for husbands. In contrast, the link
between schematic attributions and anger should be stronger
for husbands than for wives. In addition, husbands anger
should be strongly related to their overall relationship sentiment, whereas this may not be true for wives.
Following an approach outlined by Sanford (2003a,
2003b), this study used hierarchical linear modeling (HLM)
to empirically distinguish between event-dependent and
schematic effects. To do this, it was necessary to assess both
event-dependent attributions and anger in response to real
relationship events on multiple occasions. This made it
possible to determine the extent to which attributions show
within-person fluctuations over time, and whether these
fluctuations correspond to within-person changes in anger.
In addition, this study investigated two schematic variables:
schematic attributions and relationship sentiment. Schematic attributions were assessed by asking participants to
rate the attributions they would make in response to a set of
hypothetical negative events. Relationship sentiment was
assessed via a self-report rating of marital quality. This
made it possible to determine the extent to which individual
differences in schematic attributions and in relationship
sentiment predict emotion after controlling for the eventdependent effects. Taken together, this study investigated
the following hypotheses:
1.

Within-person changes in event-dependent attributions are expected to predict corresponding


changes in anger.

2.

Schematic attributions are expected to explain


unique variance in anger after controlling for
event-dependent attributions.

3.

The effect for event-dependent attributions is expected to be stronger for wives than for husbands.

4.

The effect for schematic attributions is expected to


be stronger for husbands than for wives.

5.

For husbands, overall relationship sentiment is expected to predict unique variance in anger after

182

SANFORD

controlling for both schematic and eventdependent attributions.

Method
Participants
Participants were 77 recently married couples. The majority of
participants were recruited through letters sent to all couples under
the age of 40 who had applied for marriage licenses in the previous
2 years in a mediumsmall Texas city. These letters described the
research study, invited couples to schedule two assessment sessions in a communication laboratory, and stated that couples would
receive $60.00 for their participation. The first 60 couples who
scheduled and kept both assessment appointments were included
in the sample. Previous research has found that couples recruited
in this way are often slightly better adjusted than couples recruited
through other methods, such as newspaper advertisements (Karney
et al., 1995). Thus, to reduce the overall risk of sampling bias in
the present study, an additional 17 couples were recruited through
a newspaper advertisement that read Focus on your marriage.
Contribute to our research program and earn $60.00. The advertisement also stated that couples had to be recently married to
participate. In the present study, there were no significant differences between couples recruited via the two different strategies on
any of the salient variables considered. All participating couples
had been married 3 years or less at the time of participation. The
participants average age was 26.48 years (SD 6.72), and they
had an average income of $40,387 (SD $21,606). Forty-nine
percent had no children, 23% had been previously married, and
58% lived together prior to marriage. The sample was 76% Caucasian, 16% Hispanic, 3% African American, and 5% other races.

Procedure
All participants took part in two assessment sessions that were
held at least 2 weeks apart. Each session lasted about 1 to 1.5 hr,
and both involved engaging in several videotaped communication
exercises and completing a number of questionnaires. During the
first assessment session, spouses were taken to separate rooms and
given a Specific Incident Form printed on carbon paper. This
form instructed participants to identify a specific incident that
illustrates an important unresolved issue in your relationship and
to answer the following questions: (a) When did the incident take
place? (b) Where did the incident take place? (c) What happened?
and (d) What is the issue? After both spouses had completed this
form, the experimenter set the husbands form aside for later and
separated the carbon copy from the wifes original form. While
couples remained seated in separate rooms, one copy of the wifes
incident form was given to the husband, and the other copy was
left in front of the wife. Participants were told that they would soon
have a conversation with their spouse about the wifes chosen
incident, and the husband was given time to read the wifes
completed incident form. Each spouse was then given a preconversation questionnaire regarding his or her thoughts and feelings
about the particular incident described on the wifes form. The
copies of the wifes incident form were left in front of both spouses
while they completed this questionnaire to ensure that there was no
confusion regarding the specific incident under consideration. After completing the questionnaire, couples were reunited and asked
to engage in a 10-min conversation about the wifes incident.
Couples were then taken back to separate rooms, and this time
each spouse was given a copy of the husbands specific incident
form. Couples then completed a second preconversation question-

naire, this time in regard to the husbands incident. Subsequently,


couples were reunited to have a 10-min conversation about the
husbands incident.
The second assessment session was scheduled at least 2 weeks
after the first. The procedure was identical to the first session with
two exceptions. First, at the beginning of the second assessment
session, couples completed a questionnaire packet that included a
measure of overall relationship satisfaction and a measure of
schematic attributions. Second, the order of conversations was
reversed so that the husbands issue was considered and discussed
before the wifes issue. This order was the same for all couples
(not counterbalanced), in part to keep the procedure from becoming unnecessarily complex and in part because previous research
using this particular protocol has found the ordering of incidents to
be inconsequential (Sanford, 2003b). Altogether, couples considered a total of four different topics (two on the first visit and two
on the second), and they completed four separate preconversation
questionnaires.

Measures
Relationship sentiment. The Quality Marital Index (QMI;
Norton, 1983) was used to assess participants relationship sentiment, or satisfaction. The items on this questionnaire ask participants to evaluate the overall goodness of their relationship. For
example, participants rate the extent to which they agree with
statements such as Our marriage is strong. This questionnaire
was specifically designed to assess relationship quality without
asking respondents to describe particular events or behaviors that
might overlap with other aspects of relationship functioning, such
as agreement or communication. As such, the QMI provides a
good index of a persons global, subjective evaluation of his or her
relationship. The QMI contains six items, and in the present study
it produced alphas of .94 for wives and .93 for husbands. The
participants in this study were relatively nondistressed, with mean
QMI scores of 34.23 (SD 6.82) and 34.19 (SD 6.15) for wives
and husbands, respectively.
Schematic attributions. The Responsibility/Blame subscale of
Fincham and Bradburys (1992) Relationship Attribution Measure
(RAM) was used to measure a persons current, general tendency
to make blaming attributions for his or her partners behavior. This
questionnaire was chosen as a measure of schematic attributions
because it was specifically designed to eliminate situation-specific
responses by having participants rate hypothetical events and not
real events. The questionnaire includes four hypothetical events,
each representing a common form of negative partner behavior
(e.g., Your partner criticizes something you say). For each
hypothetical event, participants rate on a 6-point scale the extent to
which they agree with three different attributions. In the present
study, this 12-item scale produced alphas of .87 for wives and .88
for husbands.
Preconversation questionnaire. A separate preconversation
questionnaire was completed for each of the four incidents spouses
identified on the specific incident forms. Spouses were instructed
to look at the specific incident form under consideration and
answer questions in response to the particular incident identified
on that form. To assess event-dependent attributions, spouses were
instructed to look at the specific incident form and think about
what your partner did during the specific incident described.
Participants then indicated their agreement with six different attributions drawn from Fincham and Bradburys (1992) RAM: My
partner did this on purpose rather than unintentionally, My
partners behavior was motivated by selfish rather than unselfish
concerns, My partner deserves to be blamed for this, My
partners behavior is caused by something about him/her, The

ATTRIBUTIONS AND ANGER IN MARRIAGE


cause of my partners behavior is something that is not likely to
change, and The cause of my partners behavior is something
that affects other areas of our marriage. Each item was rated on
a 6-point scale ranging from 1 (disagree strongly) to 6 (agree
strongly). Reliability estimates across the four separate assessments ranged between .76 and .85 for wives and between .77 and
.79 for husbands.
Anger was assessed using the Hard Emotion subscale from a
questionnaire developed by Sanford and Rowatt (2004) for assessing emotional responses to conflict in close interpersonal relationships. Participants were instructed to look at the specific incident
form and take a moment to consider how you feel about the issue
described. They then rated their agreement with the following
four items: I feel angry, I feel irritated, I feel annoyed, and
I feel aggravated. Each item was rated on a 6-point scale ranging
from 1 (disagree strongly) to 6 (agree strongly). The four emotion
items are intended to include generic and socially acceptable ways
of describing anger. Previous research indicates that these items
form a single factor that is distinct from other negative emotions
and that this scale is related to measures of attachment and relationship satisfaction in theoretically expected ways (Sanford &
Rowatt, 2004). Reliability estimates across the four separate assessments ranged between .87 and .92 for wives and between .89
and .91 for husbands.

Results
Correlations were computed between each of the eight
event-dependent variables (four assessments of attribution
and four assessments of anger). In addition, correlations
were computed between the schematic variables (schematic
attributions and relationship quality) and the eventdependent variables. These correlations are reported in Table 1 for wives and Table 2 for husbands. As would be
expected, event-dependent attributions were moderately
correlated with schematic attributions, although the correlations were not so high as to suggest that these variables are
redundant. For wives, each event-dependent attribution correlated with the corresponding event-dependent anger rating, and most of the correlations between event-dependent
attributions and relationship quality were significant. Husbands event-dependent attributions produced fewer signif-

183

icant correlations with other variables. However, husbands


schematic attributions, as well as husbands relationship
quality, correlated with anger at all four assessments,
whereas most of these correlations were nonsignificant for
wives.
In general, these results are consistent with the hypothesized gender differences. Event-dependent attributions produced strong results for wives, whereas schematic attributions produced strong results for husbands. However, these
correlations are somewhat ambiguous because they cannot
fully isolate the effects of event-dependent attributions from
the effects of schematic attributions. This is because there is
a degree of overlap between the measures of the two types
of attribution. Participants event-dependent appraisals
could be influenced by their schematic attributions, and
conversely, ratings of schematic attributions could merely
reflect a cumulative sum of event-dependent appraisals
rather than a stable, schematic-level process. To obtain an
accurate estimate of event-dependent effects, it is necessary
to control for individual differences in schematic attributions, which can be done by limiting the analysis to explaining changes occurring within the person. At the eventdependent level, changes in attributions should predict
corresponding changes in anger. Likewise, to obtain an
accurate estimate of schematic effects, it is necessary to
control for the effects of event-dependent attributions. As is
demonstrated shortly, accurate estimates of effects for
event-dependent and schematic attributions can be obtained
using HLM.
As a preliminary step in the hierarchical analysis, intraclass correlations were computed for event-dependent attributions and anger. These correlations were useful because
they indicate (a) the proportion of variance due to eventdependent changes within persons and (b) the proportion of
variance due to stable, individual differences. To compute
intraclass correlations, the variance between people was
divided by the total variance (variance between people plus
variance within people). The intraclass correlations for attributions were .53 for wives and .48 for husbands. This

Table 1
Correlations Between Event-Dependent and Schematic Variables for Wives
Event-dependent attributions
Variable

Time 1

Event-dependent attributions
Time 1
Time 2
Time 3
Time 4
Event-dependent anger
Time 1
Time 2
Time 3
Time 4
Schematic variables
Relationship sentiment
Attributions
* p .05.

** p .01.

.57***
.49***
.52***

Time 2

Time 3

.35**
.29**

.44***

.23*
.14
.02
.12

.17
.38***
.07
.00

.00
.11
.35**
.18

.28*
.58***

.13
.42***

.27*
.40***

*** p .001.

Event-dependent anger
Time 4

Time 1

Time 2

Time 3

Time 4

.25*
.33**
.11
.50***
.37***
.50***

.45***
.15
.36***
.10
.21

.25*
.21
.14
.25*

.41***
.21
.17

.30**
.20

184

SANFORD

Table 2
Correlations Between Event-Dependent and Schematic Variables for Husbands
Event-dependent attributions
Variable

Time 1

Event-dependent attributions
Time 1
Time 2
Time 3
Time 4
Event-dependent anger
Time 1
Time 2
Time 3
Time 4
Schematic variables
Relationship sentiment
Attributions
* p .05.

** p .01.

Time 2

Time 3

Event-dependent anger
Time 4

.48***
.40***
.31**

.30**
.22

.40***

.30**
.21
.17
.15

.01
.19
.10
.05

.25*
.28*
.40***
.25*

.12
.05
.21
.17

.29**
.35**

.00
.33**

.11
.64***

Time 1

.62***
.45***
.33**

.11
.48***

.52***
.39***

Time 2

.35**
.25*
.41***
.38***

Time 3

Time 4

.50***
.37***
.36***

.36***
.31**

*** p .001.

suggests that approximately half of the variance in attributions reflected changes occurring within persons and half of
the variance reflected individual differences between people. The intraclass correlations for anger were .29 for wives
and .41 for husbands. This suggests that there was somewhat less consistency in anger across assessments, with
more than half of the variance representing within-person
change.
HLM (Bryk & Raudenbush, 1992) was used to examine
the relationship between attributions and anger across two
levels of analysis: the event-dependent level and the schematic level. This data analysis procedure is best described as
a process that begins by creating a separate regression
equation for each individual couple in the sample. Following the multivariate hierarchical model described by Raudenbush, Brennan, and Barnett (1995), this study used a
single equation to estimate a separate intercept and slope for
the wife and husband of each couple. In the present study,
this equation took the following form:
y ct b w0 wife bh0 husband bw1 wife attributionct
bh1 husband attributionct ect .
For each couple, anger is assessed at 8 times (4 times for
each spouse), and yct is the anger score for couple c at
assessment t (t 1, . . . , 8). For the wife, the variable
labeled wife is a dummy variable equal to 1, and the
variable labeled wife attribution is the wifes attribution z
score at assessment point t. For the husband, both of these
variables are set to equal zero, and they drop out of the
equation when predicting outcomes for the wife. This means
that bw0 is the intercept for the wife, and bw1 is the slope for
the wife. Similarly for the husband, the variable labeled
husband is a dummy variable equal to 1 (and set to zero
for the wife), and the variable labeled husband attribution
is the husbands attribution z score at assessment point t (set
to equal zero for the wife). Thus, bh0 is the intercept for the
husband, and bh1 is the slope for the husband. Taken together, it is possible to estimate four parameters for each

couple in the sample: the wifes intercept and slope and the
husbands intercept and slope. The slope indicates the extent
to which within-person fluctuations in event-dependent attributions predict corresponding within-person changes in
emotion. The intercept gives a persons expected emotion,
holding event-dependent attributions constant. Notably,
event-dependent attributions scores were turned into z
scores in this study so that the intercept gives a persons
expected emotion when his or her event-dependent attribution is equal to the sample average attribution.
After the four parameter estimates were computed for
each couple in the sample (two intercepts and two slopes),
these parameters were then used as outcome variables at the
second level of analysis. The following second-level equations were used to estimate each of the first-level parameters:
b w0c w00 w01 wife attributionc uw0c
b h0c h00 h01 husband attributionc uh0c
b w1c w10
b h1c h10,
where w00 is the average expected anger for wives when all
attributions are held constant; h00 is the average expected
anger for husbands when all attributions are held constant;
w01 is the slope indicating the extent to which wives
schematic attributions predict their expected anger scores
after controlling for event-dependent attributions; h01 is the
slope indicating the extent to which husbands schematic
attributions predict their expected anger scores after controlling for event-dependent attributions; w10 is the average
within-person slope for event-dependent attributions for all
wives; and h10 is the average within-person slope for
event-dependent attributions for all husbands.
Taken together, these equations provide parameters indicating (a) the expected anger for wives and husbands after

ATTRIBUTIONS AND ANGER IN MARRIAGE

holding all attributions constant, (b) the effect of schematic


attributions for wives and husbands after holding eventdependent attributions constant, and (c) the within-person
effect for event-dependent attributions for wives and husbands. The HLM 5.05 software program (Raudenbush,
Bryk, & Congdon, 2001) was used to estimate the parameters for the equations listed above, and t tests were used to
test the significance of each coefficient or, specifically, to
test the null hypothesis that the coefficient is zero. In addition, chi-square tests were used to test for significant differences between wives and husbands. Each chi-square test
contrasted a selected coefficient for wives with the corresponding coefficient for husbands and tested the null hypothesis that the two coefficients were equal. To aid in the
interpretation of the results, all variables were standardized
prior to analysis.
Results from the HLM analysis are reported in Table 3.
Although no specific hypotheses were made regarding the
intercepts, the results are worth noting. After holding attributions constant, the expected z score for wives anger is
significantly higher than zero, and the expected z score for
husbands anger is significantly lower than zero. Consistent
with these findings, the chi-square test indicates that, holding attributions constant, wives report significantly more
anger than do husbands. This gender difference is congruent
with previous research using the same emotion questionnaire, which also found that wives report more emotion than
husbands (Sanford & Rowatt, 2004).
The slopes reported in Table 3 directly address the specific hypotheses for this study. The results for schematic
attributions were significant for husbands but not for wives,
and the chi-square test indicates that the effect is significantly stronger for husbands than for wives. In contrast, the
event-dependent attributions produced a large effect for
wives and a small, albeit significant, effect for husbands.
The difference between wives and husbands in eventdependent attributions was also significant. Taken together,
the results are consistent with the hypotheses that the eventdependent effects will be strongest for wives and the schematic effects will be strongest for husbands.
At the schematic level, relationship sentiment was also
expected to be predictive of husbands anger, but not nec-

185

essarily wives anger. To test this, relationship sentiment, as


assessed by the QMI, was added to the HLM analysis. The
second-level equations were simply expanded so that both
schematic attributions and relationship sentiment were used
to predict the expected emotion score for each person. The
results from this analysis are reported in the lower half of
Table 3. The results for relationship sentiment were strong
and significant for husbands, yet they were close to zero for
wives. Moreover, the difference between wives and husbands was significant. The results for husbands schematic
attributions were attenuated after controlling for relationship sentiment. Although husbands schematic attributions
remained significant, the difference between wives and
husbands schematic attributions became nonsignificant after controlling for relationship sentiment. Moreover, the
coefficient for husbands relationship sentiment was significantly larger than the coefficient for husbands schematic
attributions, 2(1, N 77) 31.57, p .01.

Discussion
This study builds on previous research indicating that
attributions are correlated with anger in marriage (Fincham
et al., 1987; Fincham & Bradbury, 1992; Senchak &
Leonard, 1993), and it distinguished between two types of
attribution: event dependent and schematic (Bugental &
Johnston, 2000). The results indicate that the processes that
link attributions and anger are not the same for husbands
and wives. Event-dependent attributions were the best predictor of wives anger, whereas schematic attributions were
the best predictor of husbands anger. This means that for
wives, changes and fluctuations in attributions from one
context to the next are strongly predictive of corresponding
changes in emotion. Moreover, variables that are distal to
the current context, such as relationship sentiment and schematic attributions, appear to have no additional influence on
wives emotion after controlling for their event-dependent
appraisals. In contrast, husbands emotion was closely related to schematic-level variables and especially to overall
sentiment in the relationship. For husbands, changes or
fluctuations in attributions were only weakly related to
changes in emotion. Thus, wives anger appears to be most

Table 3
Hierarchical Linear Modeling (HLM) Results Using Event-Dependent and Schematic
Variables to Predict Anger
HLM coefficients
Equation

Wives

First equation
Expected anger Z score (intercept)
Schematic attributions (slope)
Event-dependent attributions (slope)
Second equation
Expected anger Z score (intercept)
Schematic attributions (slope)
Relationship sentiment (slope)
Event-dependent attributions (slope)
* p .05.

** p .01.

*** p .001.

.20**
.04
.43***
.20**
.04
.01
.43***

Husbands

2(1, N 77) for


gender difference

.18*
.25**
.19**

15.45***
3.75*
8.00**

.19**
.17*
.33***
.17**

19.19***
1.44
11.19**
9.66**

186

SANFORD

closely connected to event-dependent appraisals of the current context, whereas husbands anger appears to be most
closely connected with schematic appraisals of the overall
relationship.
The distinction between event-dependent and schematic
cognition is similar to distinctions made by other researchers from varying lines of inquiry using somewhat different
terminology. For example, Sanford (2003a, 2003b) distinguishes between proximal and distal cognition, Solomon
(2001) distinguishes between phasic and tonic attributions,
and Smith and Kirby (2001) distinguish between cognition
based on reasoning and that based on associative processing. Although researchers have used differing terminology
to describe the two types of cognition, there are three
common assumptions suggested by many researchers in this
area.
The first common assumption is that there is an important
distinction between attributions made in response to a specific event and attributions based on memories or on schemata for how things typically are in the relationship. The
correlation results in the present study indicate that these
two types of attribution are moderately correlated. This
makes assessment of attributions difficult, because the two
types of attribution can be easily confounded. That is, any
measure of event-dependent attributions is likely to be influenced, to a degree, by schematic attributions. Conversely,
a measure of schematic attributions may reflect, to an extent, a simple cumulative sum of event-dependent attributions. Consequently, if attributions are assessed for only a
single occasion, it is impossible to determine what percentage of the variance actually represents an event-dependent
appraisal and what percentage represents a schematic appraisal. However, the present study demonstrates a method
for distinguishing between these two types of attribution.
This requires assessing both types of attribution and using
HLM to identify the unique effects associated with each
type. The results based on this procedure confirm the validity of both event-dependent and schematic attributions.
Specifically, the results indicate that the function and correlates of event-dependent attributions are not the same as
the function and correlates of schematic attributions.
A second common assumption suggested by researchers
is that event-dependent cognition will change from one
context to the next, whereas schematic cognition is a relatively stable characteristic of the person. In other words,
event-dependent attributions pertain to within-person
change, whereas schematic attributions pertain to individual
differences. Results from the present study suggest that
attributions include a degree of both stability and withinperson change. For both wives and husbands, approximately
50% of the variance in event-dependent attributions represented individual differences between people, and approximately 50% of the variance represented changes occurring
within people. Given the strong reliability estimates for the
attribution questionnaire, most of this variance appears to be
true variance. This means that the weak effect for husbands
event-dependent attributions was not the result of a lack of
reliable event-dependent variance. Husbands did change
their attributions across situations, but these changes were

only weakly related to husbands emotion. Taken together,


the results of this study fit with the assumption that eventdependent attributions show substantial within-person
change from one context to the next. Although it would
have been useful to also investigate stability or change in
schematic attributions, the present study was not able to do
this because it was conducted over a relatively brief time
span. In addition, couples made attributions in response to
four different incidents that were not necessarily ordered in
a particular linear sequence. Thus, the focus of this study
was more on understanding short-term variability in attributions than on linear change in attributions over time.
Previous research, however, has found that measures of
schematic attributions demonstrate high testretest correlations (Fincham & Bradbury, 1987) yet can also change over
the period of a year or longer (Karney & Bradbury, 2000).
Thus, schematic attributions may remain relatively stable,
especially over short periods of time, yet also show gradual
change over long periods of time.
A third common assumption among theorists is that
event-dependent cognition is a deliberate, conscious process, whereas schematic cognition is automatic and often
occurs outside of conscious awareness. This issue, however,
was not directly addressed in the present study. Both eventdependent and schematic attributions were assessed via
self-report questionnaires, and participants needed to be
consciously aware of their attributions to report them. Thus,
it is not clear whether schematic attributions are typically
based on subconscious priming and spreading activation (as
suggested by Smith & Kirby, 2001) and, if so, whether these
attributions are changed when they are made conscious for
the purpose of assessment. It is also possible that schematic
attributions are sometimes conscious and deliberate. For
example, a spouse may effectively cope with a relationship
conflict by deliberately thinking, We have a good relationship, and therefore there must be a good reason for my
partners behavior. The findings of the present study suggest that future work in this area would be promising.
It is important to note that this study focused on understanding attributions and anger in relatively nondistressed,
recently married couples. Thus, it is not clear whether the
results of this study will generalize to older or more distressed relationships. Previous research using a variety of
sample types has found that attributions are correlated with
both affect and communication behavior. This includes
studies of recently married couples (Miller & Bradbury,
1995; Senchak & Leonard, 1993), studies of couples drawn
from the community (Fincham & Bradbury, 1992), and
studies including both distressed, clinical couples and nondistressed couples (Bradbury et al., 1996; Bradbury & Fincham, 1992; Fincham et al., 1987). Thus, it is common for
findings from attributional research to generalize across
different types of couples. Nevertheless, it remains possible
that specific findings regarding event-dependent and schematic attributions may be moderated by relationship characteristics. For example, it is possible that, over time, attributions become increasingly schematic in a relationship.
This study investigated several hypotheses regarding differences between wives and husbands. In general, event-

ATTRIBUTIONS AND ANGER IN MARRIAGE

dependent effects were expected to be strongest for wives,


whereas schematic effects were expected to be strongest for
husbands. These hypotheses were supported with striking
consistency. As such, the results are congruent with previous research indicating that husbands cognition tends to
function at a global level, whereas wives cognition tends to
function at a context-specific level. For example, husbands
marital attributions tend to spill over into interactions with
their children, whereas wives marital attributions are more
limited to the marital context (Brody, Arias, & Fincham,
1996). In predicting behaviors during specific marital
problem-solving discussions, wives attributions tend to
produce stronger effects than husbands attributions (Bradbury et al., 1996; Bradbury & Fincham, 1992; Miller &
Bradbury, 1995). Consistent with the results of the present
study, husbands behavior is better predicted by their global
sentiment than by their attributions (Bradbury & Fincham,
1992). In other research specifically using HLM to differentiate between event-dependent and schematic effects,
Sanford (2003a) found that only wives event-dependent (or
proximal) cognitive expectancies predicted behavior during
a problem-solving conversation.
Taken together with these previous findings, the results of
the present study are consistent with the possibility that, in
comparison with husbands, wives are more cognizant of
relationship processes (Acitelli, 1992) and engage in more
meaningful appraisals of relationship events. This means
that wives may be most influenced by the specific context of
relationship events, whereas husbands may be most influenced by their overall sentiment in the relationship. Thus,
wives cognition may be better than husbands cognition in
predicting event-dependent variables, such as emotion for a
particular event or behavior in a particular conversation.
The results of the present study can also provide clues for
interpreting other gender differences reported in marital
research. For example, researchers have observed a
demandwithdrawal pattern in distressed couples where one
partner (usually the wife) is frequently critical and angry,
and the other partner (usually the husband) feels overwhelmed by this criticism and frequently withdrawals from
conflict (Heavey, Layne, & Christensen, 1993). Gottman
(1994) suggested that this is because husbands are easily
overwhelmed, or flooded, by negative emotion and that
this leads to withdrawal from conflict. The results of the
present study suggest that if a husband develops an extremely negative relationship schema, he may respond to all
conflicts with intense negative emotion. In this case, the
husband may not respond to aspects of the particular context
that might help him regulate negative emotions. Instead,
even trivial incidents may be appraised as being extremely
negative, produce flooding, and lead to withdrawal. This, in
turn, could set in motion and/or maintain a demand
withdrawal pattern of interaction.
The results of this study also have implications for current approaches to couples therapy. For example, it is possible that interventions targeting variables that function at
the event-dependent level, such as wives attributions, will
produce the most immediate change in a relationship. This
raises the question of how to change event-dependent attri-

187

butions. The close connection between event-dependent attributions and emotion for wives could indicate that wives
attributions reflect accurate appraisals of the current context
(Bradbury et al., 1996). If it turns out that wives attributions are indeed generally accurate, then the best way to
change an event-dependent attribution would be to change
the context. This suggests that a particularly effective intervention might be to teach husbands to behave differently
and teach wives to make positive attributions for these
changes. The results of the present study indicate that a
change in a wifes attributions would likely correspond to a
change in her emotion. These implications for couples therapy remain relatively speculative at this point, in part because this study used a relatively nondistressed sample and
the findings may not generalize to a clinical setting. The
results of the present study are important in that they demonstrate a method for distinguishing between eventdependent and schematic attributions, show how these types
of attributions function differently in wives and husbands,
and provide direction for future research regarding clinical
applications.

References
Acitelli, L. K. (1992). Gender differences in relationship awareness and marital satisfaction among young married couples.
Personality and Social Psychology Bulletin, 18, 102110.
Bradbury, T. N., Beach, S. R. H., Fincham, F. D., & Nelson, G. M.
(1996). Attributions and behavior in functional and dysfunctional
marriages. Journal of Consulting and Clinical Psychology, 64,
569 576.
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, 613 628.
Brody, G. H., Arias, I., & Fincham, F. D. (1996). Linking marital
and child attributions to family processes and parent child relationships. Journal of Family Psychology, 10, 408 421.
Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear
models: Applications and data analysis methods. Thousand
Oaks, CA: Sage.
Bugental, D. B., & Johnston, C. (2000). Parental and child cognitions in the context of the family. Annual Review of Psychology,
51, 315344.
Carels, R. A., & Baucom, D. H. (1999). Support in marriage:
Factors associated with on-line perceptions of support helpfulness. Journal of Family Psychology, 13, 131144.
Cutrona, C. E., & Suhr, J. A. (1994). Social support communication in the context of marriage: An analysis of couples supportive interactions. In B. R. Burleson (Ed.), Communication of
social support: Messages, interactions, relationships, and community (pp. 113135). Thousand Oaks, CA: Sage.
Fincham, F. D., Beach, S. R., & Nelson, G. (1987). Attribution
processes in distressed and nondistressed couples: III. Causal and
responsibility attributions for spouse behavior. Cognitive Therapy and Research, 11, 71 86.
Fincham, F. D., & Bradbury, T. N. (1987). The impact of attributions in marriage: A longitudinal analysis. Journal of Personality
and Social Psychology, 53, 510 517.
Fincham, F. D., & Bradbury, T. N. (1992). Assessing attributions
in marriage: The Relationship Attribution Measure. Journal of
Personality and Social Psychology, 62, 457 468.

188

SANFORD

Gottman, J. M. (1994). What predicts divorce? The relationship


between marital processes and marital outcomes. Hillsdale, NJ:
Erlbaum.
Heavey, C. L., Layne, C., & Christensen, A. (1993). Gender and
conflict structure in marital interaction: A replication and extension. Journal of Consulting and Clinical Psychology, 61, 16 27.
Karney, B. R., & Bradbury, T. N. (1995). The longitudinal course
of marital quality and stability: A review of theory, methods, and
research. Psychological Bulletin, 118, 334.
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.
Karney, B. R., Davila, J., Cohan, C. L., Sullivan, K. T., Johnson,
M. D., & Bradbury, T. N. (1995). An empirical investigation of
sampling strategies in marital research. Journal of Marriage and
the Family, 57, 909 920.
Lazarus, R. S. (2001). Relational meaning and discrete emotions.
In K. R. Scherer, A. Schorr, & T. Johnstone (Eds.), Appraisal
processes in emotion: Theory, methods, research (pp. 37 67).
London: Oxford University Press.
Miller, G. E., & Bradbury, T. N. (1995). Refining the association
between attributions and behavior in marital interaction. Journal
of Family Psychology, 9, 196 208.
Neumann, R. (2000). The causal influences of attributions on
emotions: A procedural priming approach. Psychological Science, 11, 179 182.
Norton, R. (1983). Measuring marital quality: A critical look at the
dependent variable. Journal of Marriage and the Family, 45,
141151.
Raudenbush, S. W., Brennan, R. T., & Barnett, R. C. (1995). A
multivariate hierarchical model for studying psychological
change within married couples. Journal of Family Psychology, 9,
161174.
Raudenbush, S. W., Bryk, A. S., & Congdon, R. (2001). Hierarchical linear and nonlinear modeling (Version 5.05). Lincolnwood, IL: Scientific Software International.
Sanford, K. (2003a). Expectancies and communication behavior in
marriage: Distinguishing proximal-level effects from distal-level
effects. Journal of Social and Personal Relationships, 20, 391
402.

Sanford, K. (2003b). Problem-solving conversations in marriage:


Does it matter what topics couples discuss? Personal Relationships, 10, 97112.
Sanford, K., & Rowatt, W. C. (2004). When is negative emotion
positive for relationships? An investigation of married couples
and roommates. Personal Relationships, 11, 329 354.
Senchak, M., & Leonard, K. E. (1993). The role of spouses
depression and anger in the attributionmarital satisfaction relation. Cognitive Therapy and Research, 17, 397 409.
Smith, C. A., Haynes, K. N., Lazarus, R. S., & Pope, L. K. (1993).
In search of the hot cognitions: Attributions, appraisals, and
their relation to emotion. Journal of Personality and Social
Psychology, 65, 916 929.
Smith, C. A., & Kirby, L. D. (2001). Toward delivering on the
promise of appraisal theory. In K. R. Scherer, A. Schorr, & T.
Johnstone (Eds.), Appraisal processes in emotion: Theory, methods, research (pp. 121138). London: Oxford University Press.
Smith, C. A., & Lazarus, R. S. (1990). Emotion and adaptation. In
L. A. Pervin (Ed.), Handbook of personality: Theory and research (pp. 609 637). New York: Guilford Press.
Solomon, D. H. (2001). Affect, attribution, and communication:
Uniting interaction episodes and global relationship judgments.
In J. H. Harvey (Ed.), Attribution, communication behavior, and
close relationships (pp. 79 90). Cambridge, England: Cambridge University Press.
Weiner, B. (1985). An attributional theory of achievement motivation and emotion. Psychological Review, 92, 548 573.
Weiner, B., Amirkhan, J., Folkes, V. S., & Verette, J. A. (1987).
An attributional analysis of excuse giving: Studies of a naive
theory of emotion. Journal of Personality and Social Psychology, 52, 316 324.
Weiss, R. L. (1980). Strategic behavioral marital therapy: Toward
a model for assessment and intervention. In J. P. Vincent (Ed.),
Advances in family intervention, assessment and theory (Vol. 1,
pp. 229 271). Greenwich, CT: JAI Press.

Received August 8, 2003


Revision received January 20, 2004
Accepted March 26, 2004

You might also like