Professional Documents
Culture Documents
Daily Diary, Attachemnt, Marital Interaction and Rel Satisfaction
Daily Diary, Attachemnt, Marital Interaction and Rel Satisfaction
Daily Diary, Attachemnt, Marital Interaction and Rel Satisfaction
JUDITH A. FEENEY
University of Queensland
Abstract
The relations among measures of attachment, spouse behavior, and marital satisfaction were assessed in a broad sample
of 193 married couples, using both questionnaire and diary methods. Insecure attachment was associated with less
favorable reports of spouse behavior, as assessed by diary checklists. Marital satisfaction was predicted by attachment
measures and reports of spouse behavior. The relation between attachment security and marital satisfaction was
moderated, but not mediated, by reported spouse behavior. Specifically, insecure individuals’ evaluations of their
relationships were more reactive to recent spouse behavior, an effect that was especially marked for fearful participants
and for those in longer-term marriages. Some gender differences in patterns of prediction were obtained. The results are
discussed in terms of the working models associated with attachment styles, and the processes by which relationship
satisfaction may be eroded over time.
A large body of research supports the pro- More specifically, research into the pre-
position that stable and satisfying couple dictors of marital quality has assessed the
relationships promote health and well-being importance of both positive and negative
(Baumeister & Leary, 1995). Hence, it is not spouse behaviors. This research has produced
surprising that theoretical and empirical stu- some conflicting findings. For example, there
dies have investigated the factors leading to is evidence that confrontation of marital
satisfying relationships. Although the predic- problems, including the expression of anger,
tors of relationship satisfaction are complex, is beneficial to long-term satisfaction, despite
dyadic interaction patterns play a central role. the fact that such confrontation may be quite
In their comprehensive model of marriage, for negative in tone (Gottman & Krokoff, 1989;
example, Lewis and Spanier (1979; Spanier & Karney & Bradbury, 1997). In contrast, other
Lewis, 1980) conceptualized ‘‘rewards from studies have associated negative communica-
spousal interaction’’ as one of three major sets tion behaviors, such as coercion, with declines
of variables predictive of marital quality. The in satisfaction (Julien, Markman, & Lindahl,
importance of spousal interaction patterns is 1989; Noller, Feeney, Bonnell, & Callan,
also highlighted by practical considerations: 1994). These conflicting results may be
These variables tend to be more amenable to largely attributable to differences in the way
intervention than the other sets of predictors that constructs are measured (Gill, Christen-
discussed by these researchers (e.g., the sen, & Fincham, 1999); in particular, opera-
personal and social resources that partners tional definitions of negativity vary widely.
bring to the relationship). Nevertheless, there is substantial support for
the broad assertion that marital satisfaction is
related directly to positive behaviors such as
acceptance, recognition, and compromise, and
inversely to negative behaviors such as blame,
Correspondence should be addressed to the author at
School of Psychology, University of Queensland, Queens- criticism, and coercion (Gill et al., 1999;
land 4072 Australia; e-mail: J. Feeney@psy.uq.edu.au. Smith, Vivian, & O’Leary, 1990).
39
40 J. A. Feeney
style differences should be most pronounced Third, secure individuals may perceive
under such conditions. Recent research sup- specific partner behaviors more positively, even
ports this proposition. In a study of long-term if partners do not behave in particularly positive
dating couples, Feeney (1998) found that links ways. According to attachment theory, indivi-
between attachment dimensions and observer- duals gradually internalize working models of
rated affect and behavior were confined relationships, based on repeated interactions
to interactions in which one partner’s behavior with close partners. These working models
threatened the development of intimacy. Simi- embody expectations of self and others, and are
larly, Rholes, Simpson, and Stevens (1998) thought to have direct effects on cognitive
reported that attachment style differences processing and emotional appraisal of relation-
in conflict resolution behaviors were more ship events (Collins & Read, 1994). These
pronounced when couples were asked to propositions are supported by the finding that,
discuss issues of major importance to their when asked to explain hypothetical relationship
relationships. Finally, in a diary study in events, secure individuals are more likely than
which undergraduate students reported their insecure individuals to provide explanations
social interactions (across all relationship that reflect expectations of partners’ availability
types) for one week, attachment differences and responsiveness (Collins, 1996).
in intimacy and disclosure were accentuated in
high-conflict interactions (Pietromonaco &
Romantic attachment, spouse behavior, and
Barrett, 1997).
satisfaction: Exploring the links
Importantly, theory and research suggest
that secure individuals not only behave more An important research need in this area is to
constructively in their romantic relationships, integrate the diverse findings linking romantic
but also describe their partners’ behaviors attachment style with both global relationship
more positively. These attachment style differ- quality and specific relationship behaviors.
ences in reports of partner behaviors are likely Two possible integrative models are suggested
to reflect three interrelated effects. by the existing findings, but require further
First, most studies of intact couples reveal investigation.
partner matching in attachment characteris- First, it is possible that the association
tics, whereby secure individuals tend to be between secure attachment and relationship
paired with secure partners (e.g., Senchak & quality is mediated by perceptions of partner
Leonard, 1992). Hence, secure individuals’ behavior. That is, secure individuals may
reports of their partners are likely to be reports report greater satisfaction with their relation-
of securely attached persons who, as just ships because they perceive their partners as
noted, seem to behave more constructively behaving in constructive and well-intentioned
toward their partners, especially under condi- ways. As noted earlier, these favorable
tions of stress or conflict. perceptions may reflect actual partner beha-
Second, it is possible that the constructive vior, together with a tendency to interpret
attitudes and behaviors of secure individuals relationship events more positively. Several
tend to bring out the best in partners (even previous studies have tried to identify vari-
when those partners are not, themselves, ables that mediate the link between attachment
secure). Some evidence of this effect is security and relationship satisfaction. To date,
provided by the work of Cohn, Silver, Cowan, there is evidence that this link may be
Cowan, and Pearson (1992), who found that mediated in part by mutual negotiation of
couples consisting of one secure and one conflict (Feeney, 1994), emotional expressive-
insecure partner were rated by observers as ness (Feeney et al., 1998), and appropriate
behaving similarly to those with two secure patterns of disclosure (Keelan, Dion, & Dion,
partners, in terms of levels of conflict and 1998). However, no previous research has
positive behaviors. In other words, a secure assessed overall perceptions of positive and
individual may buffer the negative effects of negative spouse behavior as potential media-
insecure attachment. tors of the attachment-satisfaction relation.
42 J. A. Feeney
Second, it is possible that attachment records. The use of the diary method to assess
security and spousal behavior have interactive spouse behavior was considered particularly
effects on global satisfaction. The literature important, for although diary reports may be
reviewed above shows that attachment groups influenced to some extent by the perceptions
differ in their responses to hypothetical rela- of the reporter, the requirement to record
tionship events and to events in their own specific behaviors soon after their occurrence
relationships, particularly when there is an minimizes the effects of global response biases
element of conflict. These group differences (Reis, 1994). Finally, the study was designed
are thought to reflect underlying working to test the possible mediating and moderating
models of attachment. Specifically, the work- roles of spouse behavior in the relation
ing models associated with insecure attachment between attachment and marital satisfaction.
may guide patterns of inference that attribute Identification of the variables that mediate and
negative partner behavior to internal, stable, moderate the effects of attachment is of prime
and global factors. In this way, negative partner importance, because such work clarifies the
behavior is likely to confirm existing doubts mechanisms by which secure individuals
about the partner’s love and commitment and maintain strong and satisfying bonds.
about the viability of the relationship. Hence,
insecure individuals may be more reactive to Hypotheses. The study was designed to test
such relational events, in terms of their overall three hypotheses derived from the literature
evaluation of the relationship. The link be- reviewed above. A general research question
tween insecurity and reactions to positive was also addressed, for which existing
behaviors may be more complex. Individuals data were insufficient to generate specific
who are uncomfortable with closeness may predictions.
tend to discount positive partner behavior, as Hypothesis 1 stated that attachment mea-
due to external, unstable, and specific factors. sures would be related to reports of spouse
In contrast, those who are anxious about their behavior. Specifically, secure attachment (high
relationships strive for reciprocation and ‘‘un- comfort with closeness, low anxiety over
ion,’’ and hence may be highly reactive to relationships) was expected to relate directly
positive partner behaviors. No previous studies to reports of positive spouse behavior, and
have explored these possibilities. inversely to reports of negative spouse beha-
vior. However, it is important to note that these
relations should not be particularly strong,
The present study
given that spouse behavior was assessed in a
The present study explored the links among relatively objective way and in an everyday
measures of attachment, spouse behavior, and (rather than high-conflict) situation.
marital satisfaction. The study had four Hypothesis 2 concerned the correlates of
important features. First, the sample of couples marital satisfaction. Satisfaction was expected
varied widely in terms of education, occupa- to be related to attachment dimensions (posi-
tion, and age (in contrast to many studies of tively to comfort with closeness and inversely
romantic attachment that have relied on to anxiety over relationships; Hypothesis 2a).
samples of college students). Second, the Further, satisfaction was expected to be related
sample included roughly equal numbers of to diary reports of spouse behavior (positively
shorter-term and longer-term marriages. This to positive behavior and inversely to negative
feature enabled an assessment of the generality behavior; Hypothesis 2b).
of findings across length of marriage; previous Hypothesis 3 concerned the links among
research has indicated that some effects of attachment, spouse behavior, and marital
attachment (e.g., interactive effects of own and satisfaction, and tested two contrasting predic-
partners’ attachment dimensions) may be tions. Hypothesis 3a stated that the association
restricted to particular stages of marriage between attachment and marital satisfaction
(Feeney, 1994). Third, participants completed would be mediated by perceptions of spouse
both self-report questionnaires and diary behavior. That is, secure spouses would report
Attachment, marital interaction, and relationship satisfaction 43
higher marital satisfaction because they per- and 44% of wives held managerial or profes-
ceive their partners’ behavior more favorably. sional positions. The remainder were not
In contrast, Hypothesis 3b stated that attach- engaged in paid work (being full-time students,
ment and spouse behavior would have inter- unemployed, or involved in home duties).
active effects on marital satisfaction. That is,
insecure spouses would be more reactive to
Measures
recent partner behavior, as evidenced by a
stronger association between marital satisfac- Diary. Each diary booklet contained two
tion and reports of spouse behavior (especially sets of records, to be filled out on two
negative behavior). consecutive days, one during the week and
The general research question concerned one on the weekend (this approach was taken
whether the observed associations (attachment to ensure an adequate sampling of the various
with spouse behavior; attachment and spouse spouse behaviors). Each record required the
behavior with marital satisfaction) would vary participant to note the number of hours spent in
across stage of marriage. the presence of the spouse that day, to read
through a checklist of 95 spouse behaviors and
check all those that had occurred during the
Method
day, and finally, to rate the day on a Likert scale
(1 to 9) reflecting overall subjective satisfac-
Participants
tion with the relationship. The behaviors listed
Participants were 193 married couples re- were designed to be widely applicable to
cruited by third-year psychology students as married couples, and to include both positive
part of a class project. The students worked in (pleasant) and negative (unpleasant) behaviors.
pairs, with each pair required to recruit one The behaviors were drawn from a larger
couple from each of two stages of marriage checklist of marital behaviors (Weiss & Perry,
(married between 1 and 10 years, n = 92; and 1983), and were selected so as to retain an
married more than 10 years, n = 101). Length adequate sampling of the domains of behavior.
of marriage ranged from 1 to 35 years (M = The current measure contained 49 positive
10.85; SD = 7.58), and age of participants behaviors (e.g., greeted me affectionately;
ranged from 20 to 61 years (M = 39.26, SD = expressed understanding or support of my
9.03). Students were encouraged to recruit feelings or mood), and 46 negative behaviors
couples from varying socioeconomic back- (e.g., called me derogatory names; got angry
grounds (rather than relying on close friends and wouldn’t tell me why). The numbers of
and family members), but were not asked to behaviors checked were summed across the
report their recruitment sources. Fifty-six of two days’ records, so that the highest possible
the couples were childless. For the remaining number of positive behaviors was 98, and the
couples, the number of children ranged from 1 highest possible number of negative behaviors
to 9, with a modal value of 2. was 92.1
Of the total sample, 13% of husbands and
10% of wives had not completed secondary
school; 17% of husbands and 24% of wives
had completed secondary school but received 1. On the diary checklists, participants were asked to
indicate whether any of the spouse behaviors were ‘‘not
no further training; 24% of husbands and 22%
applicable’’ on the particular day. For example, the item
of wives had some postsecondary training; and ‘‘spouse apologized to me’’ might be considered
46% of husbands and 44% of wives held inapplicable if no cause for apology had arisen during
university degrees. Similarly, the sample was the day. All analyses reported in this paper were also
reasonably diverse in terms of occupational conducted on conditionalized scores, obtained by
expressing the number of positive and negative spouse
categories. For current occupation, 18% of
behaviors as proportions of the applicable positive and
husbands and 6% of wives reported holding negative behaviors, respectively. The results of these
manual positions; 17% of husbands and 24% of analyses were similar to those based on the absolute
wives held clerical positions; 56% of husbands frequencies of behaviors.
44 J. A. Feeney
the links among attachment, spouse behavior, siduals showed that this association reflected
and satisfaction) was addressed using correla- two specific effects: an overrepresentation of
tional and hierarchical regression analyses. the secure-secure pairing (69 couples) and an
under-representation of dismissing wives
paired with secure husbands (9 couples). In
Attachment characteristics of the sample
other words, of the 101 secure husbands in
The numbers of husbands and wives endorsing the sample, 68% were paired with secure
each description of the forced-choice measure wives, and only 8.9% were paired with
of attachment style were as follows: secure dismissing wives. Further, three of the four
(101 and 109, respectively), preoccupied (27 correlations between partners’ attachment
and 26), dismissing (37 and 30), and fearful dimensions were significant: husbands’ anxi-
(28 and 28). The scales assessing the two ety with wives’ anxiety (r = .17, p < .01),
attachment dimensions (comfort with close- husbands’ comfort with wives’ comfort
ness and anxiety over relationships) were (r = .24, p < .001), and husbands’ anxiety
approximately normally distributed (see the with wives’ comfort (r = .22, p < .001). All
Method section for mean scores and standard these ‘‘matching’’ effects were apparent in both
deviations). These two scales were negatively shorter-term and longer-term marriages.
correlated, r = .38 for husbands, r = .30 Because most of the focal analyses
for wives; p <.001 in each case. reported in the following sections involve
As would be expected from the partial comparing the two groups defined by length
overlap in item content, responses to the of marriage, it was considered important to
forced-choice measure were strongly related assess whether attachment characteristics
to scores on the attachment dimensions, differed for these groups. Chi square ana-
multivariate F (6, 372) = 14.08 for husbands lyses showed that the groups did not differ
and 17.43 for wives, MSE = 15.29 and 14.67 in terms of either husbands’ responses (2 (3)
respectively, p <.001 in each case. Consistent = 2.68) or wives’ responses (2 (3) = 0.34)
with previous research, post hoc tests showed to the forced-choice measure. Further, a
that secure and preoccupied participants re- MANOVA comparing group scores on the
ported higher levels of comfort with closeness two attachment dimensions was also non-
than did dismissing and fearful participants; significant, multivariate F (2, 180) = 1.77;
further, preoccupied and fearful participants MSE = 24.69, ns.
reported higher levels of anxiety over relation-
ships than did secure and dismissing partici-
Association between attachment and reports of
pants. (Because similar results have been
spouse behavior
reported in previous studies, these results are
not detailed further.) As noted earlier, the hypothesized link between
There was no reliable association between attachment and diary-based reports of spouse
gender and responses to the forced-choice behavior was assessed using MANOVAs and
measure, 2 (3) = 1.03. Further, a MANOVA APIM analyses, depending on the measure of
in which gender was varied within-dyad attachment being analyzed.
revealed no gender differences on the two
attachment dimensions, multivariate F (2,181) Forced-choice attachment and reports of
= 2.29, MSE = 15.80, ns. In short, husbands spouse behavior. The association between
and wives responded in similar ways to the responses to the forced-choice measure and
various attachment measures. reports of spouse behaviors was assessed by
Both the forced-choice measure and the conducting separate MANOVAs for each
attachment dimensions provided evidence of gender. In these analyses, forced-choice
partner matching. Cross-tabulation of hus- attachment style and length of marriage were
bands’ and wives’ responses to the forced- varied between subjects, and the dependent
choice measure showed a reliable association, variables were the number of positive and
2 (9) = 24.63, p < .005. Standardized re- negative spouse behaviors.
46 J. A. Feeney
The analysis for husbands revealed no scores on the outcome variable). The method
significant effects. For wives, the only involves conducting two regression equations:
significant effect was an interaction between a between-dyads regression, in which the
attachment style and length of marriage mean of the two partners’ scores on each
for negative behaviors, F(3,184) = 3.10, predictor variable is used to predict the mean
MSE = 23.12, p < .05. Simple effects tests outcome score for the dyad, and a within-
indicated no effect of attachment style for dyads regression, in which the difference
wives in the shorter-term marriages. In the between the two partners’ scores on each
longer-term marriages, the effect of attach- predictor variable is used to predict the
ment style was quite marked, reflecting the difference between partners’ scores on the
uniquely high scores of fearful wives (see outcome measure. The unstandardized regres-
Figure 1). sion coefficients and standard errors from
these analyses are then used to calculate the
Attachment dimensions and reports of spouse actor and partner effects (for details, see
behavior. The association between attach- Kashy & Kenny, 2000).
ment dimensions and the diary reports was In the present study, this method was used
assessed using the Actor-Partner Inter- to assess the association between partners’
dependence model as described by Kashy and attachment dimensions and reports of positive
Kenny (2000). As the name implies, this and negative spouse behaviors, separately for
method of data analysis is designed to deal shorter-term and longer-term marriages. (A
with nonindependent data. It enables assess- preliminary MANOVA revealed that the
ment of both actor effects (e.g., the effect groups defined by length of marriage did not
of individuals’ comfort with closeness on differ in their perceptions of spouse behavior.)
their own scores on an outcome variable), The results of these APIM analyses are
and partner effects (e.g., the effect of indi- summarized in the top sections of Table 1.
viduals’ comfort with closeness on partners’ As the table shows, significant relations
Figure 1. Interaction of attachment style and length of marriage for wives’ reports of negative
spouse behaviors.
Attachment, marital interaction, and relationship satisfaction 47
between the attachment dimensions and the between attachment measures and reports of
diary reports were confined to the reports of negative spouse behaviors, though only for
negative spouse behavior in longer-term longer-term marriages. In these longer-term
marriages. Specifically, in these marriages, marriages, reports of negative spouse behav-
own comfort with closeness was associated iors were associated inversely with own
with reports of fewer negative spouse beha- comfort, and positively with own and part-
viors, and own and partner’s anxiety over ner’s anxiety. Moreover, fearful wives in longer-
relationships were associated with reports of term marriages (i.e., wives low in comfort and
more negative spouse behaviors. high in anxiety) reported particularly high
It should be noted that these analyses also levels of negative spouse behaviors. The
included tests for main effects of gender, and latter finding, although gender-specific, is
interactive effects of gender and attachment consistent with those from the APIM.
dimensions. The only main effect occurred in Together, these results support the first
longer-term marriages, with wives reporting hypothesis.
more negative spouse behaviors than did
Association between attachment and
husbands, t (89) = 2.57, p < .05. Impor-
marital satisfaction
tantly, no interactive effects of gender attained
significance, indicating that the actor and A similar method to that reported in the
partner effects tested in the analyses were previous section was used to test for the effects
not dependent on gender. To simplify pre- of attachment dimensions (actor and partner)
sentation, the tests for main and interactive on marital satisfaction. Before running these
effects of gender have not been included in analyses separately for shorter- and longer-
Table 1. term marriages, it was considered important to
ensure that evaluations of marital quality
Attachment and reports of spouse behavior: showed reasonable variability, even in the
Summary. Overall, the analyses reported in shorter-term marriages. Inspection of the QMI
this section provide evidence of an association data for the two groups revealed a small
Table 1. Actor-partner effects for the prediction of spouse behavior and marital satisfaction
from attachment dimensions
Effects being tested (predictor; criterion) Standard error t test (actor) t test (partner)
Comfort; positive spouse behavior .111 .225 1.234
.156 .545 .282
Anxiety; positive spouse behavior .167 .425 .796
.197 .579 .995
Comfort; negative spouse behavior .072 1.139 .181
.101 2.089* .168
Anxiety; negative spouse behavior .109 .615 .725
.128 3.406*** 2.008*
Comfort; marital satisfaction .098 1.408 1.724
.170 .959 1.265
Anxiety; marital satisfaction .122 2.189* .648
.178 2.640** 2.129*
Note. In each cell, the top entry refers to shorter-term marriages, and the bottom entry to longer-term marriages. Degrees
of freedom for actor and partner effects vary slightly across analyses, but are approximately 120 for shorter-term
marriages and 160 for longer-term marriages.
* p <.05, **p <.01, ***p <.001.
48 J. A. Feeney
Husbands Wives
Predictor r Beta r Beta
Positive spouse behavior .27* .20* .21* .20*
.31** .18* .54*** .44***
Negative spouse behavior .36*** .35*** .38*** .37***
.58*** .51*** .39*** .30***
Time together .25* .17+ .27* .18+
.34** .19* .36*** .24**
Note. In each cell, the top entry refers to shorter-term marriages, and the bottom entry to longer-term marriages.
+
p < .10, *p < .05, **p < .01, ***p < .001.
Attachment, marital interaction, and relationship satisfaction 49
pattern of findings, the results of the hierarch- coefficients provide considerable support for
ical analyses simply replicate the APIM and the proposition that insecure persons are more
standard regression analyses already reported, reactive to spousal behavior. The most striking
and hence have not been tabulated. findings were the high reactivity coefficients
for positive behavior for fearful husbands and
Spouse behavior as a moderator: Attachment wives, and for negative behavior for preoccu-
and reactivity to spouse behavior. Hypo- pied husbands, dismissing husbands and
thesis 3b, concerning the link between wives, and fearful wives. All six of these
attachment and reactivity to spouse behavior, coefficients were significantly higher than the
was tested in two ways to enable the conver- corresponding coefficients for secure spouses.
gent validity of the results to be assessed. First, The second method for assessing the link
reactivity coefficients for positive and negative between attachment and reactivity to spouse
spouse behavior were calculated using the behavior involved defining marital satisfaction
diary reports alone, and following the method in terms of QMI scores, rather than ratings of
described by Jacobson, Follette, and McDonald daily satisfaction. Specifically, hierarchical
(1982). Specifically, a reactivity coefficient for (moderated) multiple regression analysis was
positive behavior was obtained by correlating used to test for interactive effects of attach-
reports of positive spouse behavior with daily ment dimensions and reports of positive and
satisfaction ratings, across the two days’ negative spouse behavior. To minimize the
records (i.e., the correlation between positive complexity of these analyses, only the effects
behavior and daily satisfaction [r-to-z trans- of own attachment dimensions were investi-
formed] was averaged across the two days). gated. At Step 1, own attachment dimensions
Similarly, a reactivity coefficient for negative and reports of positive and negative spouse
behavior was obtained by correlating reports behaviors were entered. At Step 2, the four
of negative spouse behavior with daily satis- product terms (comfort by positive behaviors,
faction ratings. comfort by negative behaviors, anxiety by
The reactivity coefficients were calculated positive behaviors, anxiety by negative beha-
separately for each attachment group, as viors) were added to the equation. The product
defined by the forced-choice measure; to terms were based on standardized scores; this
maintain adequate cell sizes, the coefficients procedure avoids multicollinearity, and also
were based on the full sample (regardless of prevents variables with greater variability
length of marriage). The resulting coefficients contributing unduly to the product term.
are shown in Table 3; the final entry in this At Step 1, both husbands’ and wives’
table indicates, for example, that fearful wives’ marital satisfaction were related positively to
ratings of daily satisfaction were negatively reports of positive spouse behavior, and
correlated (r = .65) with their reports of inversely to anxiety and reports of negative
negative spouse behavior. These reactivity spouse behavior. (These results are consistent
Note. In each column, coefficients with the subscript a differ significantly from the corresponding coefficient for the
secure group.
50 J. A. Feeney
with the findings already discussed, and are insecure husbands (see Table 4; first entry in
not detailed further.) Addition of the product each cell).
terms added reliably to the variance explained For comparison purposes, similar analyses
for husbands; R2 change = .089, F (inc) = were conducted for wives (see Table 4, entries
4.90, p < .001. All four product terms reliably in parentheses). For insecure wives, marital
predicted husbands’ satisfaction, with beta satisfaction was consistently associated with
weights ranging from .17 to .27. For reports of recent spouse behavior (as was the
wives, the combined set of product terms did case for insecure husbands). Unlike the
not add reliably to the variance explained, findings for husbands, however, secure wives
although the beta weight for the interaction also showed a link between marital satisfac-
between comfort and positive behaviors was tion and reports of spouse behavior, but only
significant ( .16, p < .02). when security was defined in terms of levels of
To further examine the significant interac- relationship anxiety. In short, the only reliable
tions, husbands were classified as either secure gender differences among these correlations
or insecure, defined as one standard above and were for low-anxious spouses: Low-anxious
below the mean on the relevant attachment wives were more reactive to recent spouse
dimension. That is, in the case of an interaction behavior than were low-anxious husbands.
involving the comfort dimension, secure hus- Finally, to assess whether the observed
bands were those scoring at least one standard interactions between attachment and spouse
deviation above the mean, and insecure hus- behavior held across stage of marriage, the
bands were those scoring at least one standard moderated regression analyses (own attach-
deviation below the mean. Conversely, in the ment and spouse behavior at Step 1; product
case of an interaction involving anxiety, secure terms at Step 2) were repeated separately for
husbands scored at least one standard deviation the groups defined by length of marriage. For
below the mean and insecure husbands scored shorter-term marriages, there were no reliable
at least one standard deviation above the mean. product terms for either gender. By contrast, in
Bivariate correlations between marital satisfac- longer-term marriages, there were three sig-
tion and the relevant measure of spouse nificant effects: for husbands, comfort by
behavior were then calculated, separately for positive behaviors and comfort by negative
secure and insecure husbands. These analyses behaviors; and for wives, comfort by positive
revealed a very consistent picture of the behaviors. (Tests of the beta weights confirmed
interaction effects, which supported Hypoth- that these effects were significantly stronger in
esis 3b. In all four cases, the association longer-term than in short-term marriages.) The
between marital satisfaction and reports of follow-up tests of the three significant interac-
recent spouse behavior was significant only for tions paralleled those for the full sample. In
each case, the association between marital
Table 4. Follow-up tests of interactions between attachment security and spouse behavior in
predicting marital satisfaction
Note. Main cell entries are for husbands; those in parentheses are for wives.
*p < .05, **p < .01.
Attachment, marital interaction, and relationship satisfaction 51
satisfaction and reports of spouse behavior was about their relationships across the marital
significant only for insecure spouses (with life-cycle, and more attuned to potential signs
correlations ranging from .41 to .72). of negativity and rejection. However, long-
itudinal research is clearly needed to test this
possibility.
Discussion
As expected, however, the associations
Responses to the questionnaire measures between attachment and spouse behavior were
supported the representativeness of the sample generally not very strong. As noted earlier,
recruited for this study. Although scores on the diary-based checklists provide a relatively
global measure of marital satisfaction were objective picture of spouse behavior; unlike
negatively skewed, the sample included a global self-report measures, they do not
substantial number of couples who were quite require participants to make highly subjective
dissatisfied with their relationships. Further, evaluations of behavior or to aggregate ratings
responses to the attachment measures indicated over a large number of past episodes. For these
a good representation of the various attachment reasons, diary reports are unlikely to be
styles. In terms of the forced-choice measure, influenced strongly by the general patterns of
just over 50% of husbands and wives described expectation and inference that probably con-
themselves as secure, with the remainder tribute to the strong attachment style differ-
spread fairly evenly over the three remaining ences reported in some studies. Further, much
categories. Similarly, although mean scores on of the previous research into romantic attach-
the two attachment dimensions suggested a ment style has focused specifically on conflict
relatively secure sample, scores spanned close situations, in which attachment differences in
to the full possible range. behavior are expected to be more pronounced
In summary, unlike many samples of married (Feeney, 1998; Rholes et al., 1998). It is
couples, the present sample did not contain a important to note, however, that the relations
strong overrepresentation of secure individuals between attachment and spouse behavior may
or of highly satisfied couples. Previous studies have been stronger if participants themselves
have also found that the recruitment method had been allowed to rate whether specific
used in this study yields representative samples spouse behaviors were positive or negative.
and highly reliable data (e.g., Feeney, 1994;
Noller, Law, & Comrey, 1987).
Predicting marital satisfaction: Main effects
The present results support previous findings
Attachment and reports of spouse behavior
linking attachment security with greater rela-
The results of this study supported the tionship satisfaction (Shaver & Hazan, 1993).
proposed association between attachment However, anxiety over relationships consis-
measures and reports of spouse behavior. tently emerged as the more important dimen-
Within longer-term marriages, reports of sion of attachment in this regard. Although
negative spouse behaviors were related in- comfort with closeness is undoubtedly a
versely to own comfort with closeness, and crucial variable when it comes to establishing
positively to both own and partner’s anxiety an intimate relationship, it may be somewhat
over relationships; the strongest effect was less important in terms of relationship main-
for own anxiety. Moreover, within longer- tenance (Feeney, 1994). Moreover, the im-
term marriages, fearful wives reported parti- portance of the anxiety dimension to marital
cularly high levels of negative spouse relationships can be explained by longitudinal
behaviors. It is not clear why these relations data, which point to a link between anxiety
were confined to the longer-term marriages, and increasingly negative patterns of couple
given that neither the attachment measures communication (Feeney et al., 1994).
nor the reports of spouse behavior varied Not surprisingly, reported spouse behavior
with length of marriage. It is possible that was also an important predictor of marital
insecure spouses become less optimistic satisfaction. For both husbands and wives, and
52 J. A. Feeney
expectations of their partners which are tied should be the case. One possibility is that
less closely to specific daily events. levels of relationship satisfaction are generally
Insecure individuals’ greater reactivity to high in shorter-term marriages, making it
partner behavior is likely to have deleterious difficult to detect such interactions. However,
effects on their relationships in the long term. In the small difference in satisfaction across the
particular, as already noted, the tendency to two stages of marriage lends little support to
notice small transgressions and to ruminate over this explanation. Another possibility is that
them seems bound to result in a gradual erosion insecurity (especially in more extreme in-
of relationship satisfaction. High reactivity to stances) is associated with a gradual increase
positive behaviors may also have negative in perceptions of spousal negativity; these
effects, however, given that all relationship negative perceptions are likely to reflect
partners show at least occasional lapses of escalating relationship conflict, together with
attention, consideration, and support (Miller, negative expectations. (These propositions are
1997). The results of the present study suggest supported, in part, by the present results for
that the relationships of avoidant individuals, wives’ reports of negative spouse behavior;
and particularly those from the fearful group, see Figure 1). There may be a certain critical
may be most prone to the negative effects of point in this process at which insecure
reactivity to partner behavior. spouses’ evaluations of their relationships
In terms of the regression analyses, the become more reactive to daily events.
interactive effects of attachment and spouse
behavior added to the prediction of marital
General discussion and conclusions
satisfaction for husbands only. These interac-
tions indicated that links between marital The present study has three limitations that
satisfaction and recent spouse behavior were should be noted. First, all the measures
restricted to insecure men; in other words, obtained were based on self-reports, although
husbands who were anxious about their the diary method is considered to be less prone
relationships or low in comfort with closeness to global response sets. Second, the diary
were more reactive to their spouse’s behavior. reports of spouse behavior may have con-
For wives, links between marital satisfaction taminated the questionnaire measures,
and recent spouse behavior applied more although this possibility was minimized by
generally: The only wives who were not the time interval between the two assessments.
reactive to spouse behavior were those high in Third, use of a cross-sectional design makes it
comfort with closeness. The reason for this difficult to establish developmental trends in
gender difference is not clear. Overall, women marriage. There may be cohort effects that
may be somewhat more sensitive to the contribute to differences in findings across the
nuances of their partners’ relationship beha- groups defined by length of marriage. In
vior (a difference that may stem from sex-role particular, there is likely to be selective loss
socialization). Such sensitivity may, at times, of the most unhappy relationships, due to
be a two-edged sword; for example, it may be separation and divorce. However, in the
associated with greater relationship awareness, present study, group comparisons were not
but also with the tendency to monitor and confounded by attachment security, and the
evaluate daily transactions. Because comfort group difference in marital satisfaction was
with closeness is associated with perceptions small.
of relationship partners as dependable and It is interesting to note that the effects of
trustworthy (Feeney, 1999), it may play an insecure attachment reported in this study show
important role in reducing reactivity to recent parallels with those from studies comparing
relationship events. happy and unhappy couples. Specifically,
It is interesting to note that the interactive previous studies have shown that couples who
effects of attachment and spouse behavior are unhappy with their relationships are more
were also restricted to longer-term marriages. likely to monitor (or ‘‘track’’) the partner’s
At this stage, it is not entirely clear why this behavior, to expect strict reciprocation of
54 J. A. Feeney
favors, and to be reactive to immediate such direction would involve assessing the
relational events, particularly negative ones stability of spouses’ expectations and evalua-
(e.g., Jacobson et al., 1982). The present results tions of each other over relatively short
suggest that attachment insecurity may often periods of time. In this way, the proposition
underlie these destructive behaviors. This that secure individuals hold more stable
finding is important because it not only perceptions and expectations of their partners
provides a theory-based explanation of these could be tested more directly. In addition, the
patterns but also suggests appropriate strategies suggestion that avoidant individuals are parti-
for intervention. Sperling and Lyons (1994), for cularly reactive to recent relational events
example, have discussed therapeutic ap- requires further investigation. Longer-term
proaches particularly suited to modifying longitudinal research is also needed, to
internal representations of attachment. These investigate more fully the interactive effects
approaches focus specifically on addressing the of attachment security and spouse behavior.
network of expectations, affects, and memories Such research should help to clarify the
associated with clients’ close relationships. mechanisms by which insecurities about
The results of this study suggest some attachment issues are either played out, or
important directions for future research. One resolved, in marital relationships.
References
Bartholomew, K. (1990). Avoidance of intimacy: An Attachment theory and close relationships (pp. 189–
attachment perspective. Journal of Social and Perso- 218). New York: Guilford.
nal Relationships, 7, 147–178. Feeney, J. A. (1999). Adult romantic attachment and couple
Bartholomew, K., & Horowitz, L. M. (1991). Attachment relationships. In J. Cassidy & P. R. Shaver (Eds.), The
styles among young adults: A test of a four-category handbook of attachment: Theory, research, and clinical
model. Journal of Personality and Social Psychology, applications (pp. 355–377). New York: Guilford.
61, 226–244. Feeney, J. A., & Noller, P. (1991). Attachment style and
Baumeister, R. F., & Leary, M. R. (1995). The need to verbal descriptions of romantic partners. Journal of
belong: Desire for interpersonal attachments as a Social and Personal Relationships, 8, 187–215.
fundamental human motivation. Psychological Bulle- Feeney, J. A., Noller, P., & Callan, V. J. (1994).
tin, 117, 497–529. Attachment style, communication and satisfaction in
Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self- the early years of marriage. In K. Bartholomew & D.
report measurement of adult attachment: An integrative Perlman (Eds.), Advances in personal relationships
overview. In J. A. Simpson & W. S. Rholes, (Eds.), Vol.5: Attachment processes in adulthood (pp. 269–
Attachment theory and close relationships (pp. 46–76). 308) London: Jessica Kingsley.
New York: Guilford. Feeney, J. A., Noller, P., & Roberts, N. (1998). Emotion,
Cohn, D. A., Silver, D. H., Cowan, C. P., Cowan, P. A., & attachment, and satisfaction in close relationships. In P.
Pearson, J. (1992). Working models of childhood A. Andersen & L. K. Guerrero (Eds.), Handbook of
attachment and couple relationships. Journal of Family communication and emotion: Research, theory, appli-
Issues, 13, 432–449. cations, and contexts (pp. 473–505). San Diego, CA:
Collins, N. L. (1996). Working models of attachment: Academic Press.
Implications for explanation, emotion, and behavior. Gill, D. S., Christensen, A., & Fincham, F. D. (1999).
Journal of Personality and Social Psychology, 71, 810– Predicting marital satisfaction from behavior: Do all
832. roads really lead to Rome? Personal Relationships, 6,
Collins, N. L., & Read, S. J. (1990). Adult attachment, 369–387.
working models, and relationship quality in dating Gottman, J. M., & Krokoff, L. J. (1989). Marital interaction
couples. Journal of Personality and Social Psychology, and satisfaction: A longitudinal view. Journal of
58, 644–663. Consulting and Clinical Psychology, 57, 47–52.
Collins, N. L., & Read, S. J. (1994). Cognitive rep- Hazan, C., & Shaver, P. R. (1987). Romantic love con-
resentations of attachment: The structure and function ceptualized as an attachment process. Journal of
of working models. In K. Bartholomew & D. Perlman Personality and Social Psychology, 52, 511–524.
(Eds.), Advances in Personal Relationships (Vol. 5, Jacobson, N. S., Follette, W. C., & McDonald, D. W.
pp. 53–90). London: Jessica Kingsley. (1982). Reactivity to positive and negative behavior in
Feeney, J. A. (1994). Attachment style, communication distressed and nondistressed married couples. Journal
patterns, and satisfaction across the life cycle of of Consulting and Clinical Psychology, 50, 706–714.
marriage. Personal Relationships, 1, 333–348. Julien, D., Markman, H. J., & Lindahl, K. M. (1989). A
Feeney, J. A. (1998). Adult attachment and relationship- comparison of a global and a microanalytic coding
centered anxiety: Responses to physical and emotional system: Implications for future trends in studying
distancing. In J. A. Simpson & W. S. Rholes, (Eds.), interactions. Behavioral Assessment, 11, 81–100.
Attachment, marital interaction, and relationship satisfaction 55
Karney, B., & Bradbury, T. (1997). Neuroticism, marital Pistole, M. C. (1989). Attachment in adult romantic
interaction, and the trajectory of marital satisfaction. relationships: Style of conflict resolution and relation-
Journal of Personality and Social Psychology, 72, ship satisfaction. Journal of Social and Personal
1075–1092. Relationships, 6, 505–510.
Kashy, D. A., & Kenny, D. A. (2000). The analysis of data Reis, H. T. (1994). Domains of experience: Investigating
from dyads and groups. In H. T. Reis & C. M. Judd relationship processes from three perspectives. In R.
(Eds.), Handbook of research methods in social and Erber & R. Gilmour (Eds.), Theoretical frameworks for
personality psychology (pp. 451–477). Cambridge, personal relationships (pp. 87–110). Hillsdale, NJ:
UK: Cambridge University Press. Erlbaum.
Keelan, J. P. R., Dion, K. K., & Dion, K. L. (1998). Rholes, W. S., Simpson, J. A., & Stevens, J. G. (1998).
Attachment style and relationship satisfaction: Test of a Attachment orientations, social support, and conflict
self-disclosure explanation. Canadian Journal of resolution in close relationships. In J. A. Simpson &
Behavioural Science, 30, 24–35. W. S. Rholes (Eds.), Attachment theory and close
Kirkpatrick, L. E., & Davis, K. E. (1994). Attachment relationships (pp. 166–188). New York: Guilford.
style, gender, and relationship stability: A longitudinal Senchak, M., & Leonard, K. E. (1992). Attachment styles
analysis. Journal of Personality and Social Psychology, and marital adjustment among newlywed couples.
66, 502–512. Journal of Social and Personal Relationships, 9,
Levy, M. B., & Davis, K. E. (1988). Lovestyles and 51–64.
attachment styles compared: Their relations to each Shaver, P. R., & Brennan, K. A. (1992). Attachment styles
other and to various relationship characteristics. Journal and the ‘‘big five’’ personality traits: Their connections
of Social and Personal Relationships, 5, 439–471. with each other and with romantic relationship
Lewis, R. A., & Spanier, G. B. (1979). Theorizing about outcomes. Personality and Social Psychology Bulletin,
the quality and stability of marriage. In W. R. Burr, R. 18, 536–545.
Hill, F. I. Nye, & I. L. Reiss (Eds.), Contemporary Shaver, P. R., & Hazan, C. (1993). Adult romantic
theories about the family (Vol. 1, pp. 268–294). New attachment: Theory and evidence. In D. Perlman & W.
York: Free Press. Jones (Eds.), Advances in personal relationships
Mikulincer, M., & Nachshon (1991). Attachment styles (Vol. 4, pp. 29–70). London: Jessica Kingsley.
and patterns of self-disclosure. Journal of Personality Smith, D. A., Vivian, D., & O’Leary, K. D. (1990).
and Social Psychology, 61, 321–331. Longitudinal prediction of marital discord from
Miller, R. S. (1997). We always hurt the ones we love: premarital expressions of affect. Journal of Consulting
Aversive interactions in close relationships. In R. M. and Clinical Psychology, 58, 790–798.
Kowalski (Ed.), Aversive interpersonal behaviors (pp. Smith, G. T., Snyder, D. K., Trull, T. J., & Monsma, B. R.
11–29). New York: Plenum. (1988). Predicting relationship satisfaction from cou-
Noller, P., Feeney, J. A., Bonnell, D., & Callan, V. J. ples’ use of leisure time. American Journal of Family
(1994). A longitudinal study of conflict in early Therapy, 16, 3–13.
marriage. Journal of Social and Personal Relation- Spanier, G. B., & Lewis, R. A. (1980). Marital quality: A
ships, 11, 233–252. review of the seventies. Journal of Marriage and the
Noller, P., Law, H., & Comrey, A. L. (1987). Cattell, Family, 42, 825–839.
Comrey, and Eysenck personality factors compared: Sperling, M. B., & Lyons, L. S. (1994). Representations of
More evidence for the five robust factors? Journal of attachment and psychotherapeutic change. In M. B.
Personality and Social Psychology, 53, 775–782. Sperling & W. H. Berman (Eds.), Attachment in adults:
Norton, R. (1983). Measuring marital quality: A critical Theory, assessment, and treatment (pp. 331–347). New
look at the dependent variable. Journal of Marriage York: Guilford.
and the Family, 45, 141–151. Weiss, R. L., & Perry, B. A. (1983). The Spouse
Pietromonaco, P. R., & Barrett, L. F. (1997). Working Observation Checklist: Developments and clinical
models of attachment and daily social interactions. applications. In E. E. Filsinger (Ed.), Marriage and
Journal of Personality and Social Psychology, 73, family assessment: A sourcebook for family therapy
1409–1423. (pp. 65–84). Beverly Hills, CA: Sage.