You are on page 1of 16

Personal Relationships,1 (1994), 333-348. Printed in the United States of America.

Copyright Q 1994 Cambridge University Press. 1350-4126/94 $5.00 + .OO

Attachment style, communication


patterns, and satisfaction across the
life cycle of marriage

JUDITH A. FEENEY
Department of Psychology, University of Queensland, Australia

Abstract
Qucstionnaire mcasures of attachment, communication patterns, and rclationship satisfaction were
administered to 361 married couples, sampled across the life cycle of marriagc. Individuals who were securc in
attachment (defined in terms of comfort with closeness and low anxiety over relationships) tended to be paired
with secure spouses. Security of attachment was associated with one’s own relationship satisfaction, although
husbands’ satisfaction was related primarily to the anxiety dimension. The most consistent effect of partners’
attachment was an inverse relationship between wives’ anxiety and husbands’ satisfaction. Thc negative effect
of wives’ anxiety on perccivcd relationship satisfaction (for both partners) was evident primarily for couples in
which husbands were uncomfortable with closeness. The association between attachment dimensions and
relationship satisfaction was largely mediated by communication patterns for wives, but only partially
mediated by communication patterns for husbands; for both husbands and wives, a measure of mutually
constructive communication emerged as the strongest correlate of satisfaction. These findings were generally
consistent across thc life cycle of marriage, and they arc important in clarifying the nature of the link between
attachment and satisfaction in a broad samplc.

Hazan and Shaver’s (1987) ground-break- item. In particular, a number of researchers


ing studies applied the principles of attach- have reported that measures of attachment
ment theory to adult romantic relationships style reflect two key dimensions: comfort
using a simple forced-choice measure of at- with closeness (a bipolar dimension that
tachment style designed to classify individu- contrasts secure and avoidant attachment)
als into the three major attachment groups and anxiety over relationships (a dimension
(secure, avoidant, and anxious-ambivalent). corresponding closely to anxious-ambiva-
These studies provided preliminary evi- lent attachment) (see Feeney, Noller, &
dence that adult attachment groups can be Callan, 1994; Simpson, Rholes, & Nelligan,
differentiated in terms of remembered rela- 1992; Strahan, 1991).
tionships with parents, internal working Despite the variety of measures em-
models (expectations of relationships), and ployed, research has consistently supported
relationship quality. the attachment perspective on dating rela-
Subsequent research has employed a tionships (Feeney & Noller, 1990,1991;Levy
range of measures of adult attachment & Davis, 1988; Simpson, 1990; Strahan,
(both categorical and continuous) based on 1991). The implications of attachment di-
Hazan and Shaver’s original forced-choice mensions,for relationship quality may, how-
ever, depend on gender. Simpson (1990)
Rcqucsts for reprints should be addressed to Judith A. found that relationship satisfaction was
Feeney, Department of Psychology, University of
positively associated with secure attach-
Queensland, Queensland, Australia 4072.
Electronic mail may be sent to: judy@psych.psy. ment and negatively associated with
uq.oz.au. avoidant attachment for both men and

333
334 J.A. Feeney

women in dating relationships, but anxious- bands’ security was also positively related
ambivalent attachment was linked with low to their use of support and validation.
satisfaction for women only. Using empiri- In a questionnaire-based study of newly
cally derived attachment scales of anxiety, married couples, Senchak and Leonard
close, and depend (together, the latter two (1992) investigated the implications of at-
scales correspond to the bipolar comfort tachment pairing for intimacy, perceptions
scale described above), Collins and Read of partner functioning, and conflict resolu-
(1990) similarly found that women’s anxiety tion. Spouses who described themselves as
was inversely related to own and partner’s secure in attachment (using the forced-
evaluations of their dating relationships, choice measure) tended to be paired with
whereas men’s closeness predicted positive secure partners. Couples consisting of two
relationship evaluations by self and partner. secure partners showed better marital ad-
Researchers have also investigated at- justment than did other couples, reporting
tachment pairing in dating relationships. greater intimacy, more favorable evalu-
Collins and Read (1990) found that men ations of the spouse, and less aggression and
who scored high on the close scale tended to withdrawal in response to conflict.
be paired with women who rated high on the Feeney, Noller, and Callan (1994) con-
close and depend scales and low on the anxi- ducted a longitudinal study examining the
ety scale. A longitudinal study of dating cou- relations among attachment, communica-
ples (Kirkpatrick & Davis, 1994) found that tion, and relationship satisfaction during the
avoidant individuals tended to be paired first 2 years of marriage. Communication
was assessed in terms of self-reported inter-
with anxious-ambivalent partners, and that
action patterns during conflict and accuracy
the relationships of anxious-ambivalent
at decoding nonverbal messages from the
women paired with either secure or avoid-
spouse. These researchers also found that
ant men were relatively enduring.
the implications of attachment dimensions
The present study examines attachment
were dependent on gender. Comfort with
style and communication patterns as pre-
closeness was related to reports of relation-
dictors of relationship satisfaction across
ship satisfaction and constructive communi-
the life cycle of marriage, consistent with
cation processes for husbands and to accu-
the focus of recent studies extending the racy at decoding nonverbal messages for
attachment perspective to marital bonds. wives. By contrast, anxiety was inversely re-
Kobak and Hazan (1991) examined the link lated to accuracy at decoding nonverbal
between attachment security (defined by messages for husbands and was associated
scales measuring the reliance on partner with low satisfaction and destructive forms
and psychological availability) and marital of communication for both husbands and
quality (i.e., marital satisfaction, and ob- wives. Wives’ anxiety also predicted hus-
servers’ ratings of spouses’ communication bands’ reports of negative couple communi-
during conflict). Marital satisfaction was as- cation.
sociated with reports of both one’s own re- In presenting their research, Feeney et
liance on her or his partner and the part- al. (1994) proposed a mediational model
ner’s reliance on one’s self; in addition, whereby the influence of attachment style
husbands’ reports of wives’ psychological on relationship satisfaction may be partially
availability were associated with both part- accounted for by communication variables.
ners’ marital satisfaction, whereas wives’ Contrary to the mediational model, how-
reports of husbands’ psychological avail- ever, their results suggested that attach-
ability were associated only with wives’ sat- ment style exerts a direct influence on
isfaction. As expected, husbands’ and relationship satisfaction, which is largely in-
wives’ security was negatively related to dependent of couple communication. Spe-
their use of rejection during conflict; hus- cifically, while husbands’ satisfaction was
Satisfaction in marriage across the life cycle 335

related to earlier (as well as to concurrent) tive than about secure partners), whereas
attachment measures, this relationship was avoidant subjects report feeling more
not influenced by communication variables; positive about relationships with anxious-
wives’ satisfaction was related only to con- ambivalent partners than with avoidant
current measures of attachment and com- partners (Pietromonaco & Carnelley, in
munication. press).
Within the attachment perspective on in- The implications of attachment pairings
timate relationships, at least two issues of for marital quality have been assessed only
theoretical and practical importance re- for newlyweds and only in terms of the rela-
main to be addressed. First, further consid- tively crude division of couples into “se-
eration needs to be given to the contribu- cure,” “insecure,” and “mixed” (i.e., one
tion of partner attachment to marital secure and one insecure partner) (see Sen-
quality. Partners’ attachment style influ- chak & Leonard, 1992). Additional infor-
ences couple interaction and individuals’ mation relevant to this issue may be pro-
perceptions of relationships, although this vided by studying the key dimensions
influence again appears to depend on gen- underlying adult attachment. It is possible,
der. Women’s anxious-ambivalence (or for example, that the influence on marital
anxiety over relationships) is positively re- quality of one spouse’s anxiety depends on
lated to partners’ perceptions of conflict the partner’s degree of comfort with close-
and negatively related to partners’ satisfac- ness: A partner who is uncomfortable with
tion and commitment; the influence of intimacy is less likely to be available and
men’s attachment style on partners’ rela- responsive to the anxious spouse’s particu-
tionship evaluation is less clear, although lar needs and thus less able to provide the
there is evidence that men’s comfort with comfort and reassurance that fulfill the
closeness predicts partners’ ratings of trust “safe haven” function of attachment (cf.
and communication quality (Collins & Bowlby, 1969, 1973,1980). This issue takes
Read, 1990; Feeney et al., 1994). the attachment perspective beyond a sim-
Whereas the research outlined above in- ple consideration of individual differences
dicates that attachment dimensions of both to a focus on the dynamics occurring within
one’s self and one’s partner are important in couples; as such, it is clearly relevant to pro-
understanding dyadic communication and grams of counselling and therapy aimed at
satisfaction, the possibility that attachment ameliorating relationship difficulties.
styles of subjects and partners may inferact Second, it is important to understand
to predict relationship outcomes has not more fully the nature of the relations
been fully tested. As noted earlier, the link among measures of attachment, communi-
between pairings on attachment style and cation, and relationship satisfaction and to
the stability of dating relationships has been recognize that the pattern of relations may
investigated (Kirkpatrick & Davis, 1994), alter over the course of marriage. The rela-
with longitudinal data suggesting that the tionship satisfaction of newly married cou-
relationships of anxious-ambivalent fe- ples (as studied by Senchak & Leonard,
males who are paired with either secure or 1992, and Feeney et al., 1994) may be di-
avoidant males tend to be relatively endur- rectly influenced by security of attachment,
ing. In addition, empirical evidence shows as partners confront the tasks of living to-
that attachment styles of subjects and hypo- gether in marriage and establishing the
thetical dating partners interact in their course of their relationship. Later in mar-
effects on some measures of relationship riage, the influence of attachment variables
evaluation: For example, secure subjects may weaken or may be mediated by com-
report feeling equally positive about hypo- munication processes, as suggested by
thetical relationships with avoidant and Feeney et al. (1994). Again, this issue has
anxious-ambivalent partners (but less posi- important implications for approaches to
336 J.A. Feeney

relationship counselling, and it can be ad- H4: It was expected that, in marriages estab-
dressed only by studying broader samples lished beyond the “honeymoon”stage, the
of married couples. influence of attachment dimensions on re-
The present study focused on the asso- lationship satisfaction would be mediated,
ciation between attachment style and rela- at least in part, by reported communication
tionship satisfaction across the life cycle of patterns.
marriage. In line with the above review, two
key issues were addressed: (1) the influence
Method
of partner attachment (both alone and in
interaction with subjects’ attachment) in
Subjects and procedure
predicting satisfaction, and (2) the role of
communication variables as possible me- Subjects were 361 married couples re-
diators of the association between attach- cruited by third-year psychology students
ment and relationship satisfaction. Several as part of a class project. Students worked
hypotheses were proposed. in pairs, with each pair required to recruit
six couples: one working-class and one
H1: Given evidence of partner matching on middle-class couple from each of three
attachment security (Collins & Read, stages of marriage (married between 1 and
1990; Senchak & Leonard, 1992), it was 10 years, n = 122; married between 11 and
expected that subjects’ and partners’ 20 years, n = 120; and married more than 20
scores on each attachment dimension years, n = 119). The students were encour-
(comfort with closeness; anxiety over re- aged to recruit couples from a range of
lationships) would be positively corre- sources (family, relatives, friends, col-
lated, and that subjects’ scores on comfort leagues, etc.), and because the student
with closeness would be negatively cor- group was predominantly middle class,
related with partners’ scores on anxiety many couples in the working-class catego-
over relationships. ries were recruited from outside students’
H2: It was hypothesized that secure attach- immediate social networks. Students were
ment (defined by high comfort with close- not required to report their methods of re-
ness and low anxiety over relationships) cruitment or to state whether couples who
would be associated with marital satisfac- were approached had declined to partici-
tion. I n particular, it was predicted that pate. Sampling across length of marriage
wives’ anxiety over relationships would be and social class was adopted to provide a
negatively related to both their own and broadly based subject sample-the social
their husbands’ satisfaction. Given the class variable was not focal, however, and
conflicting findings reported to date, no was not incorporated in data analyses.
specific predictions were made concerning (Newlyweds were excluded from the sam-
the relative importance of husbands’ com- ple because the focus was on established
fort and anxiety f o r one’s own orpartners’ marriages; note also that the size of subsam-
satisfaction. ples varies for specific analyses because of
H3: In line with the argument presented missing data.)
earlier, it was expected that subjects’ and Of the total sample, 56% of husbands
partners’ attachnzent dimensions would and 55% of wives had received some terti-
interact in their effect on relationship satis- ary education;23% of husbands and 26% of
faction, with the combination of a spouse wives had completed only high school; 17%
high in anxiety over relationships and a of husbands and 15% of wives had some
spouse low in comfort with closeness being high school education; and 4% of both hus-
associated with low satisfaction. No pre- bands and wives had attended primary
dictions were made concerning other pos- school only. In terms of current occupation,
sible interaction effects. 46% of husbands and 35% of wives were
Satisfaction in marriage across the life cycle 337

employed in managerial, professional, or partner won’t want to stay with me”; “I find
semiprofessional positions; 45% of hus- that others are reluctant to get as close as I
bands and 31% of wives were working in would like”; “I often worry that my partner
clerical or manual positions; the remainder doesn’t really love me”; “Sometimes people
were either unemployed or engaged in are scared away by my wanting to be too
home duties. close to them”; and “I don’t often worry
Students were given copies of the ques- about being abandoned” (reverse scored).
tionnaire package to distribute to the mar- Coefficient alphas were .81 (comfort)
ried couples. The order of the various instru- and .SO (anxiety); as in the study by Feeney
ments in the package was counterbalanced. et al. (1994), these scales were based on 8
A cover sheet explained the purpose of the and 5 items, respectively, because internal
project and emphasized the confidential na- consistency was increased by dropping the
ture of the material and the importance of item ‘‘I feel comfortable having other peo-
both spouses completing the measures in- ple depend on me” from the comfort scale
dependently. Subjects could either return and the item “I want to merge completely
the completed questionnaires to the student with another person” from the anxiety
who recruited them or mail them directly scale. Scores on comfort (possible range 8
back to the researcher. A telephone contact to 40) varied from 9 to 40, with a mean of
was also provided to enable the researcher 28.2 and a median of 29.0. Scores on anxiety
to deal with any queries concerning the ma- (possible range 5 to 25) varied from 5 to 23,
terials. with a mean of 9.9 and a median of 9.0.

Measures
Communication patterns. Communication
Attachment. Attachment was measured us- was assessed using the Communication Pat-
ing the 15 separate statements from the terns Questionnaire (Christensen & Sul-
original forced-choice measure (Hazan & laway, 1984), which measures the extent to
Shaver, 1987), with items rated on a Likert which couples (both one’s self and partner)
scale ranging from (1)not at all like me to (5) use various interaction strategies during
very much like me. As reported by Feeney et conflict. The inventory taps four factors
al. (1994), subjects were given scores on the (Noller & White, 1990): mutuality (9 items
two factors of comfort with closeness (re- measuring mutual expression and under-
ferred to as comfort) and anxiety over rela- standing); coercion (9 items assessing
tionships (referred to as anxiety) obtained threat, blame, and aggression); destructive
empirically with a separate sample of 352 process (7 items assessing patterns of de-
undergraduates (Feeney, 1990). mand-withdraw, pressure-resist, etc.); and
Items forming the comfort scale were: “I postconflict distress (4 items measuring
feel comfortable depending on other peo- guilt, hurt, and withdrawal). The four fac-
ple”; “I find it relatively easy to get close to tors discriminate groups differing in marital
others”; “I find it difficult to depend on adjustment (Noller & White, 1990). Coeffi-
others” (reverse scored); “Love partners cient alphas for the present sample were .88
often want me to be more intimate than I (mutuality), .86 (coercion), .84 (destructive
feel comfortable being” (reverse scored); “I process), and .75 (postconflict distress). To
am nervous when anyone gets too close” facilitate comparison of scores across the
(reverse scored); “I am somewhat uncom- four scales, scale scores were divided by the
fortable being close to others” (reverse number of items forming the scale. Scores
scored); “I find it easy to trust others”; and for each scale spanned the full possible
“I don’t often worry about someone getting range from 1 to 9, with the following mean
too close to me.” The anxiety scale con- and median scores respectively: 6.4 and 6.7
sisted of the items: “I often worry that my (mutuality); 2.3 and 2.0 (coercion); 3.7 and
338 J.A. Feeney

3.7 (destructive process); and 4.3 and 4.5 couple, and gender was varied within-cou-
(postconflict distress). ple (this procedure takes account of the fact
that data from husbands and wives are not
Relationship satisfaction. Relationship sat- statistically independent).
isfaction was assessed using the Quality The first MANOVA employed the two
Marriage Index (QMI) (Norton, 1983). The attachment scales as dependent variables.
QMI is a highly reliable, unidimensional There was a main effect of length of mar-
scale, consisting of 6 items that evaluate the riage, multivariate F(4,656) = 4 . 5 4 , ~< .002,
marriage relationship as a whole (e.g., “Our which applied only to the anxiety scale. Post
marriage is strong”; coefficient alpha for hoc tests using the Scheffe technique indi-
the scale = .96). Scores on the QMI cated that couples married for more than 20
spanned the full possible range from 6 to 42, years reported less anxiety ( M = 9.13) than
with a mean of 34.48 and a median of 36. those married for 1 to 10 years ( M = 10.29).
There was also a main effect of gender, mul-
tivariate F(2, 327) = 8.82, p < .001, which
Results
applied only to the comfort scale. Husbands
reported less comfort with closeness ( M =
Overview of analyses
27.53) than did wives ( M = 28.96).
Four sets of analyses were conducted. First, A main effect of gender was also ob-
a series of MANOVAs assessed the effects tained for relationship satisfaction, F(1,
of gender and length of marriage on the 357) = 6 . 5 0 , ~< .02, with husbands report-
measures of attachment, communication ing greater relationship satisfaction ( M =
patterns, and relationship satisfaction. 34.83) than did wives ( M = 34.03). By
These MANOVAs were conducted so as to contrast, a MANOVA performed on the
assist interpretation of possible differences, four scales of the Communication Patterns
in subsequent focal analyses, between the Questionnaire revealed no significant ef-
results of groups defined by gender and fects.
length of marriage. Second, partner match-
ing on attachment dimensions was assessed
using correlational and cross-tabulation Partner matching on attachment dimensions
techniques. Third, a series of hierarchical
multiple regression analyses (conducted Partner matching on attachment dimen-
separately by gender and length of mar- sions was assessed using correlations and
riage) examined subjects’ and partners’ at- contingency tables. Spouses’ scores on the
tachment dimensions as predictors of rela- two attachment dimensions were first corre-
tionship satisfaction. Finally, another series lated separately for each group defined by
of hierarchical multiple regressions as- length of marriage. Wives’ anxiety was posi-
sessed communication variables as possible tively correlated with husbands’ anxiety and
mediators of the relation between attach- negatively correlated with husbands’ com-
ment and relationship satisfaction. fort at all stages of marriage. (Correlations
with husbands’ anxiety for marriages of 1 to
10 years, 11to 20 years, and over 20 years, re-
Effects of gender and length of marriage spectively, were r(120) = .26,p < .01; r(118)
‘onattachment, communication, and = .19,p < .05; and r(117) = .32,p < .01. Cor-
satisfaction relations with husbands’ comfort were
As outlined above, the effects of gender r(120) = -.19,p <.05;r(118) = -.30,p < .01;
and length of marriage on attachment,com- and r(117) = -.24, p < .01, respectively.)
munication patterns, and relationship satis- Wives’ comfort was inversely related to hus-
faction were investigated using MANOVA. bands’ anxiety for couples married more
Length of marriage was varied between- than 10 years, r(118) = -.37,p < .001, and
Satisfaction in marriage across the life cycle 339

r(117) = -.24,p < .Ol,for 11 to 20 years and binations of attachment scores occurred
over 20 years, respectively; it was not reli- more frequently than others, regardless of
ably related to husbands’ comfort at any length of marriage. Specifically, there were
stage of marriage. These findings generally relatively high numbers of husbands low in
support the hypothesised link between hus- anxiety who were paired with wives high in
bands’ and wives’ attachment security, al- comfort (see c in Table 1 ) or with wives low
though the nonsignificant association be- in anxiety (see d in Table 1);note that these
tween spouses’ scores on comfort was patterns reflect high levels of self-reported
unexpected. attachment security in both partners. By
Additional information concerning part- contrast, there were relatively few couples
ner matching is provided by the frequencies consisting of wives low in comfort who
of couples in each combination of attach- were paired with secure husbands (i.e., hus-
ment scores (e.g., husbands high or low in bands high in comfort [a in Table 11 or hus-
comfort by wives high or low in comfort; bands low in anxiety [c in Table 11).
see Table 1). High and low attachment Second, some combinations of attach-
scores were defined as those falling at least ment scores occurred more frequently at
one standard deviation above and below certain stages of marriage. For example,
the mean, respectively. (Each quadrant of couples in which husbands reported low
Table 1 represents the full subject sample, comfort with closeness were generally
omitting those scoring within one standard more frequent in marriages of 1to 10 years’
deviation of the mean on the relevant at- duration than in longer term marriages. In
tachment dimensions.) particular, couples consisting of a husband
Two main points can be noted with re- low in comfort and a wife high in anxiety
gard to these frequencies. First, some com- were over-represented in this group of early

Table 1. Frequencies of couples in each combination of attachment scores

Years Years
(a) H Comfort by (b) H Comfort by
W Comfort 1-10 11-20 Over20 W Anxiety 1-10 11-20 Over20

H low comfort/ H low comfort/


W low comfort 11 5 8 W low anxiety 14 6 9
H low comfort/ H low comfort/
W high comfort 13 6 9 W high anxiety 21 9 12
H high comfort/ H high comfort/
W low comfort 5 3 6 W low anxiety 12 14 14
H high comfort/ H high comfort/
W high comfort 12 18 13 W high anxiety 7 8 5

(c) H Anxiety by (d) H Anxiety by


W Comfort 1-10 11-20 Over20 W Anxiety 1-10 11-20 Over20

H low anxiety/ H low anxiety/


W low comfort 7 1 5 W low anxiety 19 16 27
H low anxiety/ H low anxiety/
W high comfort 15 21 21 W high anxiety 9 11 5
H high anxiety/ H high anxiety/
W low comfort 8 7 5 W low anxiety 4 5 9
H high anxiety/ H high anxiety/
W high comfort 8 5 6 W high anxiety 19 14 12

Note: H = husband: W = wife.


340 J.A. Feeney

marriages: In the 1-to-10 years group, 21 wives’ anxiety and husbands’ relationship
couples consisted of this combination, com- satisfaction, as hypothesized.
pared with the 9 couples in the 11-to-20 The strongest evidence of an interaction
years group and the 12 couples in the over- between subjects’ and partners’ attachment
20 years group; x* (2) = 5 . 9 7 , ~< .05. dimensions (Step 2) was found for mar-
riages of 10 years’ duration or less, F(inc) =
3.51 for husbands and 3.37 for wives,p < .02
in each case. In the remaining four analyses,
Subjects ’ and partners’ attachment
the product terms provided a nonsignificant
dimensions as predictors of satisfaction
or marginally significant increment in ex-
Six hierarchical multiple regression analyses plained variance. For couples in these more
were conducted, one for each group defined recent marriages, there was a highly signifi-
by gender and length of marriage. The de- cant interaction between husbands’ comfort
pendent variable was relationship satisfac- and wives’ anxiety in predicting the relation-
tion, and subjects’ and partners’ scores on ship satisfaction of both husbands and wives.
the two attachment scales were entered at To explore this interaction further, addi-
Step 1. At Step 2, the four multiplicative tional analyses were conducted. First, cou-
terms between subjects’ and partners’ at- ples married for 10 years or less were di-
tachment scores were entered: Husband’s vided into two groups, according to whether
Comfort X Wife’s Comfort; Husband’s the husband reported high versus low com-
Comfort X Wife’s Anxiety;Husband’s Anxi- fort with closeness (defined as one standard
ety X Wife’s Comf0rt;Husband’s Anxiety X deviation above and below the mean, re-
Wife’s Anxiety. These product terms as- spectively). Bivariate correlations between
sessed the interactions between attachment wives’ anxiety and both husbands’ and
dimensions of subjects and their spouses. wives’ relationship satisfaction were calcu-
The interaction terms were based on stand- lated separately for these two groups of cou-
ardized scores to prevent any one variable ples. For couples in which the husband re-
from contributing an undue amount of vari- ported high comfort with closeness, there
ance to the product term. These analyses are was a nonsignificant relationship between
summarized in Table 2. wives’ anxiety and both husbands’ and
At Step 1,subjects’ and partners’ scores wives’ satisfaction. For couples in which the
on the two attachment scales provided reli- husband reported low comfort, however,
able prediction of relationship satisfaction wives’ anxiety was negatively correlated
for all six groups (p < .001 in each case). The with both husbands’ and wives’ satisfaction,
regression coefficients indicate that anxiety r (35) = - .57 and r = - .63, respectively,p <
was associated with own reports of low re- .01 in each case. This pattern of results, al-
lationship satisfaction for both husbands though gender specific, was consistent with
and wives across the life cycle of marriage. the third hypothesis, namely, that the combi-
Comfort was associated with own reports of nation of a spouse high in anxiety and a
satisfaction for wives only; it should be spouse low in comfort would be associated
noted, however, that the bivariate correla- with low satisfaction.
tion between husbands’ comfort and their As noted earlier, 21 couples in the 1-to-
own satisfaction (i.e., not controlling for 10 years group consisted of this combina-
anxiety) was highly reliable at all stages of tion of a husband low in comfort with close-
marriage. These results support the hy- ness and a wife high in anxiety over
pothesized link between security of attach- relationships, and this combination was less
ment and relationship satisfaction. The common in longer term marriages (see b in
most consistent finding for partner attach- Table 1). Furthermore, the reverse pattern
ment was an inverse relationship between of a wife low in comfort and a husband high
Satkfaction in marriage across the life cycle 341

Table 2. Hierarchical regression analyses of relationship satisfaction on own and partners’


attachment
Husbands, 1-10 years Husbands, 11-20 years Husbands, over 20 years

(a) Variable r Beta r Beta r Beta

Step 1
H comfort .27** .ll .31** .10 .26** .03
H anxiety -.39*** -.25** -.36*** -.25* -.39*** - .26*
W comfort .24* .09 .29** .09 .37*** .20*
W anxiety -.42*** -.30*** - .30** -.19t -.42*** - .23*
Full model
H comfort .27** .05 .31** .13 .26** .04
H anxiety -.39*** - .24* -.36*** -.25* -.39*** - .27*
W comfort .24* .09 .29** .08 .37*** .23*
W anxiety - 42*** -.27** -.30** -.18 -.42*** -.25*
H C X WC .12 .16 .12 - .O1 - .05 - .05
H C X WA .27** .31** - .08 - .06 .lo - .14
H A X WC .10 .13 .02 .10 .10 .03
H A X WA - .04 .16 .08 .09 -.19* - .26*

Wives, 1-10 years Wives, 11-20 years Wives, over 20 years

(b) Variable r Beta r Beta r Beta

Step 1
H comfort ,177 .06 .38*** .21* .12 -.13
H anxiety -.26** -.14 - .33*** - .17 -.35*** -.32**
W comfort .30** .19* .29** .07 .38*** .21*
W anxiety -.36*** -.25** -.32*** -.19$ -.39*** - .22*
Full model
H comfort .17t .O1 .38*** .23t .12 -.14
H anxiety - .26** -.14 -.33*** -.18 -.35*** -.35**
W comfort .30** .19* .29** .09 .38*** .21*
W anxiety -.36*** - .24* -.32*** -.17 -.39*** - .22*
H C X WC .04 .ll .08 - .08 - .06 - .12
H C X WA .24* .31** .04 .06 .ll - .08
H A X WC .ll .11 .01 .06 .02 -.11
H A X WA - .06 .13 .10 .16 -.13 - .23*

Note: HC X WC = husbands’ comfort by wives’ comfort; HC X WA = husbands’ comfort by wives’ anxiety;


HA X WC = husbands’ anxiety by wives’ comfort; HA X WA = husbands’ anxiety by wives’ anxiety.
t p < .06, * p < .05, **p < .01, ***p < ,001.

in anxiety occurred infrequently at all ious wives) is also pertinent to the interpre-
stages of marriage (with a total of 20 cou- tation of this interaction. Mean QMI scores
ples in all). The low frequencies of these were relatively low for both husband and
combinations may explain both the gender- wife (28.93 and 29.53, respectively), in com-
specific nature of the interaction between parison with those for couples consisting of
comfort and anxiety and its restriction to husbands high in comfort and anxious
the 1-to-10 years group. wives (37.00 and 35.50 for husbands and
An examination of mean levels of satis- wives, respectively). However, although
faction reported by these particular couples comparisons must be made with caution
(husbands low in comfort paired with anx- because of the small cell sizes, the satisfac-
342 J.A. Feeney

tion ratings of husband low in comfort/anx- portance of the attachment scales should de-
ious wife couples were comparable to those crease from Step 1to Step 2, as reflected in a
of anxious husbandlwife low in comfort reduction in their standardized regression
couples (29.89 and 27.67 for husband and weights (cf. Baron & Kenny, 1986). Intercor-
wife, respectively) and those of anxious relations among the predictor variables are
husband/wife high in comfort couples shown in Table 3, and the regression analy-
(30.50 and 30.25, respectively). Thus, it ap- ses are summarized in Table 4.
pears that husbands’ anxiety is associated For all six groups, subjects’ scores on the
with own and partners’ dissatisfaction, irre- two attachment scales provided reliable
spective of wives’ comfort with closeness, prediction of relationship satisfaction at
whereas wives’ anxiety is associated with Step 1 (p < .001 in each case). Husbands’
dissatisfaction only for couples in which the anxiety was negatively associated with rela-
husband is uncomfortable with intimacy. tionship satisfaction, irrespective of length
of marriage. For wives, both attachment
scales provided reliable prediction of rela-
Attachment and communication as tionship satisfaction, with satisfaction being
predictors of relationship satisfaction positively related to comfort and negatively
Six additional hierarchical multiple regres- related to anxiety.
sion analyses were conducted, one for each A t Step 2, addition of the communica-
combination of gender and length of mar- tion variables provided a significant in-
riage. The dependent variable was relation- crease in explained variance for all six
ship satisfaction, and subjects’ scores on the groups (again,p < .001 in each case). In the
two attachment scales were entered at Step full regression model, mutuality was the
1.A t Step 2 , the four scores derived from the most important predictor of husbands’ sat-
Communication Patterns Questionnaire isfaction, with anxiety maintaining a signifi-
were added. These analyses tested the pro- cant or near-significant relationship with
posed mediational model: If the relationship the dependent variable. These results sug-
between attachment and satisfaction is me- gest that, for husbands, the negative asso-
diated by communication patterns, the im- ciation between anxiety and satisfaction is

Table 3. Correlations among measures of relationship satisfaction, attachment, and


communication

Satisfaction Comfort Anxiety Mutuality Coercion Destructive Distress

Satisfaction 1.00 .27** -.41** .65** -.43** - .48** - .09


1.00 .32** -.36** .64** -.so** -.57** -.19**
Comfort 1.00 -.41** .34** -.28** -.29** - .20**
1.00 - .39** .31** -.32** -.29** -.23**
Anxiety 1.oo -.40** .32** .39** .30**
1.00 -.38** .41** .35** .28**
Mutuality 1.oo -.51** -.62** -.16**
1.00 -.54** -.56** -.23**
Coercion 1.00 .66** .38**
1.00 .64** .42**
Destructive 1.oo .38**
1.oo .39**
Distress 1.00
1.00

Note: Correlations are shown for husbands (top rows) and wives (bottom rows).
**p < .01.
Satisfaction in marriage across the life cycle 343

Table 4. Hierarchical regression analyses of relationship satisfaction on own attachment


and communication
Husbands, 1-10 years Husbands, 11-20 years Husbands, over 20 years

(a) Variable r Beta r Beta r Beta


Step 1
Comfort .27** .15 .33*** .17 .27** .03
Anxiety -.38*** -.33*** -.41*** -.33*** -.43*** -.41***
Full model
Comfort .27** .03 .33*** .O1 .27** - .07
Anxiety -.38*** -.15* -.41*** -.17t -.43*** -.19t
Mutuality .69*** .54*** .66*** .60*** .68*** .59***
Coercion -.48*** - .04 -.39*** - .04 -.42*** -.15
Destructive -.59*** -.19 -.45*** .04 -.41*** .07
Distress - .12 .13 -.15 .02 .02 .12

Wives, 1-10 years Wives, 11-20 years Wives, over 20 years

(b) Variable r Beta r Beta r Beta

Step 1
Comfort .27** .18* .33*** .21* .38*** .24*
Anxiety -.35*** -.29** -.34*** - .24* -.39*** -. .27**
Full model
Comfort .27** .09 .33*** .02 .38*** .08
Anxiety -.35*** - .04 -.34*** - .O1 -.39*** - .06
Mutuality .69*** .58*** .69*** .67*** .69*** .52***
Coercion -.51*** - .07 - .56*** - .08 -.47*** - .05
Destructive -.53*** -.15 -.61*** -.13 -.56*** -.25**
Distress -.26** .12 -.23* .08 - .09 .07

jp< .06, * p < .05, **p < .01, ***p < ,001

only partially mediated by communication predictor variables was reversed (i.e., com-
patterns adopted during conflict. Mutuality munication measures were entered at Step
was also the most important predictor of 1, and attachment scales added a t Step 2).
wives’ satisfaction in the full model, with In each of these analyses, the communica-
both attachment scales becoming nonsigni- tion measures accounted for a significant
ficant predictors. These results, which were proportion of the variance in relationship
consistent across the three stages of mar- satisfaction, R2 at Step 1 ranged from .44to
riage, suggest that, for wives, the relation- .59, p < .001 in each case. A t Step 2, the
ship between security of attachment and attachment scales did not provide a signifi-
satisfaction is largely mediated by commu- cant increase in explained variance for any
nication patterns. This finding supports the group; for husbands, however, a trend
fourth hypothesis. emerged for anxiety to be negatively re-
To test further whether the variance in lated to satisfaction, and this trend was con-
relationship satisfaction explained by at- sistent over length of marriage, p ranged
tachment dimensions was attributable from -.15 to -.17, p < .06 in each case.
largely to communication patterns, a final These results provide further evidence that
series of regression analyses were con- for wives, the variance in relationship satis-
ducted (separately by gender and length of faction explained by attachment dimen-
marriage) in which the order of entry of the sions is attributable largely to communica-
344 J.A. Feeney

tion patterns, and that for husbands, anxiety direct interest, but rather were assessed for
over relationships may have deleterious ef- their possible relevance to analyses of the
fects over and above those attributable to correlates of satisfaction in each group. For
communication patterns. this reason, further exploration of the links
among length of marriage, family stage, and
satisfaction was not undertaken.
Discussion
Husbands reported less comfort with
The present findings regarding the effects closeness than did wives, although the mean
of gender and length of marriage were gen- difference was again small. This finding is
erally consistent with previous research. consistent with that of Feeney et al. (1994),
The groups defined by length of marriage who reported a small gender difference in
did not differ in terms of reported commu- comfort in their sample of respondents after
nication patterns. Because the research re- 21 months of marriage, and may reflect so-
ported here was based on a cross-sectional cialization patterns that emphasize women’s
design, it cannot directly assess develop- expressive and nurturant roles.
mental changes in communication patterns. In the present sample, couples married
Nevertheless, this result is consistent with for more than 20 years reported lower anxi-
longitudinal studies demonstrating the sta- ety than did those married for up to 10
bility of marital communication patterns years. Again, given the cross-sectional na-
(Markman, 1984; Noller, Feeney, Bonnell, ture of the study, various interpretations of
& Callan, 1994). It should also be noted that this result can be offered. A developmental
no gender difference appeared in reported process may be involved, whereby the secu-
communication patterns. Because the Com- rity afforded by marriage allows spouses
munication Patterns Questionnaire asks who are initially anxious about their rela-
each respondent to report on the communi- tionships to revise their negative internal
cation of both husband and wife, gender working models (Hazan & Shaver, 1987;
effects would simply indicate a sex differ- Ricks, 1985) and, hence, to become less anx-
ence in reporting, rather than a difference ious. Alternatively, the marriages of couples
in interaction patterns. with at least one highly anxious partner
Consistent with the findings of several may be less likely to endure, given that
previous researchers (e.g., Campbell, Con- anxiety is associated with reports of marital
verse, & Rodgers, 1976; Rhyne, 1981), hus- conflict and with destructive responses to
bands in the present study reported greater this conflict (Feeney et al., 1994). A third
relationship satisfaction than did wives, al- possibility is that a cohort effect is involved,
though the size of the mean difference was with younger couples, having grown up in
small. Length of marriage was unrelated to an era of increasing marital disruption, ex-
satisfaction, despite previous reports of a periencing greater anxiety about the future
U-shaped curve for satisfaction over the life of their own marital relationships.
cycle of marriage (e.g., Anderson, Russell, Correlations between spouses’ attach-
& Schumm, 1983; Argyle, 1986). Failure to ment dimensions suggested matching in
find such an effect may be due to methodo- terms of attachment security. For both hus-
logical factors: The present study defined bands and wives, one’s own anxiety was
length of marriage using a very simple cate- positively related to the spouses’ anxiety
gorization that did not take account of fam- and negatively related to the spouses’ com-
ily stages (preparental, childbearing, etc.). fort (although the link between husbands’
The group married for up to 10 years, for anxiety and wives’ comfort was not evident
example, contained a mix of preparental in short-term marriages). Given the general
and childbearing couples. In the present re- tendency for secure individuals to be paired
search, however, differences in satisfaction with secure spouses, the lack of association
related to length of marriage were not of between husbands’ and wives’ levels of
Satisfaction in marriage across the life cycle 345

comfort is surprising. Because this study fo- research has shown high relationship stabil-
cused on established relationships, the role ity for dating couples consisting of an anx-
of partner choice in the observed patterns ious woman and an avoidant man, despite
of attachment matching is unclear. That is, the low satisfaction associated with this pair-
the link between husbands’ and wives’ ing (Kirkpatrick & Davis, 1994). Consistent
anxiety and spouses’ insecurity may reflect with the argument made earlier, these re-
a process of partner selection; alternatively, searchers suggest that anxious women may
it may stem from dynamics occurring later strive hard to elicit attention and affection
within the relationship, whereby one part- from their partners without this behavior
ner’s basic fears about loss and abandon- being seen as violating sex-role stereotypes;
ment come to undermine the other part- furthermore, anxious women paired with
ner’s attachment security. Low comfort avoidant men (those uncomfortable with
with closeness appears to be less salient closeness) may compensate to some extent
than high anxiety in terms of partner selec- for their partners’ emotional distance by
tion and/or less likely subsequently to un- persistently attending to relationship prob-
dermine the partner’s security; this differ- lems and attempting to solve them.
ence may arise because low comfort is less The present results generally replicate
strongly linked with overt signs of jealousy, previous work in pointing to the link be-
coercion, and conflict (Feeney et al., 1994). tween secure attachment and relationship
The partner matching data also pointed satisfaction. For husbands and wives, direct
to relatively large numbers of husbands low associations existed between satisfaction
in anxiety who were paired with secure and both attachment dimensions; satisfac-
spouses (that is, wivcs high in comfort and tion was associated with high comfort and
wives low in anxiety) and to relatively small low anxiety. When the independent contri-
numbers of wives low in comfort who were butions of attachment dimensions to rela-
paired with secure spouses. In addition, hus- tionship satisfaction were assessed, most of
bands reporting low levels of comfort were these associations remained significant: sat-
less likely to be found in mature marriages, isfaction was negatively correlated with
and the specific pairing of husbands low in anxiety for both husbands and wives and
comfort and wives high in anxiety was very positively correlated with comfort for wives.
infrequent in mature marriages. In addition, wives’ anxiety was negatively
Again, given the cross-sectional design of associated with husbands’ satisfaction.
this study, interpretations of these latter One unexpected feature of these results
findings must be made with caution. It is is the strong negative relationship between
possible that some husbands who are un- husbands’ anxiety and their satisfaction, a
comfortable with closeness work through relationship not apparent in previous work.
their relationship difficulties (perhaps with Previous research, however, has focused
the assistance of counselling) and, hence, re- primarily on dating and newly married cou-
vise their negative working models of rela- ples, and, as noted by Kirkpatrick and Davis
tionships. Alternatively (or in addition),hus- (1994), establishing an intimate relation-
bands who are uncomfortable with ship may be quite a different issue from
closeness may be likely to experience mari- successfully maintaining one. Specifically,
tal disruption within the first decade. (Both males’ comfort with closeness may be an
these explanations may apply especially to important factor in establishing and devel-
those paired with anxious wives.) Although oping an intimate relationship, but a rela-
the latter alternative may appear more plau- tively unimportant factor in terms of rela-
sible, it is important, in considering the im- tionship maintenance. Moreover, in marital
plications of attachment dimensions, to dis- relationships, anxiety is associated with in-
tinguish between relationship quality and creasingly negative patterns of communica-
stability. As noted previously, longitudinal tion for both males and females (Feeney et
346 J.A. Feeney

al., 1994) and, hence, may engender dissat- may appear contrary to the results of Collins
isfaction. and Read (1990). These researchers found
The possibility that subjects’ and part- that evaluations of dating relationships were
ners’ attachment dimensions interact to positively related to men’s comfort with
predict marital quality received some sup- closeness and negatively related to women’s
port. In particular, the results point to the anxiety; they argued that females’ anxiety
problematic combination of husbands low may be especially detrimental because it is
in comfort and wives high in anxiety. The manifested in jealousy and is seen by men as
fact that this pattern was gender specific- a threat to their independence and freedom.
with the interaction between husbands’ Collins and Read (1990), however, did not
anxiety and wives’ comfort failing to pre- test for interactions between subjects’ and
dict either partner’s satisfaction-appears partners’ attachment styles; in this respect,
to be an artifact of the small number of the main effect of wives’ anxiety found in the
couples showing this reverse pattern of at- present study is consistent with their results,
tachment pairings. Mean satisfaction scores although it is qualified by the interaction
suggest that anxiety in husbands is associ- with husbands’ comfort.
ated with low satisfaction for both partners, The proposed mediational model re-
regardless of wives’ level of comfort; wives’ ceived considerable support. For wives, the
anxiety, on the other hand, is associated relationship between attachhent dimen-
with husbands’ and wives’ dissatisfaction sions and satisfaction was largely accounted
only for couples in which the husband is for by reported communication patterns
uncomfortable with intimacy. during conflict. In particular, attachment di-
Hence, as other researchers have sug- mensions were eclipsed as correlates of rela-
gested (Collins & Read, 1990;Kirkpatrick & I
tionship satisfaction by mutuality, a scale
Davis, 1994), there seems to be an interplay that measures open and constructive com-
between gender roles and attachment di- munication and lack of avoidance and with-
mensions in predicting relationship quality. drawal. This support for the mediational
The clinging, demanding style associated model contrasts with the results of Feeney et
with anxiety (and anxious-ambivalence) is al. (1994). As suggested earlier, direct effects
similar to the gender-role stereotype of fe- of attachment on satisfaction may be pre-
male relationship behavior-thus, anxiety sent mainly in the earliest stages of marital
may be less detrimental to relationship qual- relationships, when couples are most ac-
ity when evidenced by females because it is tively involved in defining the nature of
seen as stereotype-confirming. Conversely, their relational life and struggling with
husbands’ anxiety may be linked with poor issues concerning interdependence and au-
relationship quality (regardless of wives’ tonomy (Fitzpatrick, 1988). The mediated
comfort with closeness) because it violates relationship found in the present study sug-
traditional gender-role stereotypes. In the gests that, in the case of established mar-
case of anxious wives, the presence of a hus- riages, insecure wives who report dissatis-
band who is uncomfortable with intimacy faction may be helped by receiving training
appears to negate the protection otherwise in more adaptive forms of communication.
afforded by the stereotype-confirming na- For husbands, the relationship between
ture of the wife’s behavior, presumably be- anxiety and marital satisfaction appeared
cause the husband who seeks to maintain to be only partially mediated by communi-
distance is unlikely to provide the reassur- cation patterns, although mutuality again
ance and support craved by the anxious emerged as the strongest correlate of satis-
spouse. faction. It is unclear at this stage why this
The finding that females’ anxiety is detri- gender difference in relations with satisfac-
mental to relationship quality only in the tion was obtained. As noted earlier, how-
presence of an emotionally distant partner ever, the demanding relationship style asso-
Satiifaction in marriage across the life cycle 347

ciated with anxiety violates the traditional of an anxious wife and a husband low in
gender-role stereotype for men; for this rea- comfort with closeness was associated with
son, it may be less acceptable for men to dissatisfaction. The effects of attachment on
show their anxiety openly in terms of jeal- relationship satisfaction were largely medi-
ous and coercive behavior. Thus husbands’ ated by communication patterns for wives,
anxiety may manifest itself in more subtle but only partially mediated by communica-
ways, such as in negative attributions for tion patterns for husbands.
partner behavior and/or negative expecta- These issues could usefully be further
tions concerning the relationship. An im- examined with long-term longitudinal de-
portant direction for further research is to signs. In the present study, we were clearly
clarify the ways in which husbands’ anxiety able to assess only concurrent relations
affects relationship satisfaction. among variables, and available data suggest
In summary, the present study provides that the correlates of concurrent and later
important clarification of the link between satisfaction may differ (Feeney et al., 1994;
attachment and relationship satisfaction in Gottman & Krokoff, 1989). Longitudinal
established marriages. Wives’ satisfaction research is also needed to clarify the nature
was associated with both attachment di- of the relationshirj between length of mar-
mensions, whereas husbands’ satisfaction riage and the attachment ratings of both
was associated primarily with low anxiety. individuals and couples. Nevertheless, this
Wives’ anxiety was also inversely related to research has been useful in clarifying the
their husbands’ satisfaction. In marriages of link between attachment and satisfaction in
relatively short duration, the combination a broadly based sample.

References
Anderson, S. A,, Russell, C. S., & Schumm, W. R. Feeney, J. A., & Nollcr, P. (1990). Attachmcnt style as
(1983). Pcrccivcd marital quality and family life- a predictor of adult romantic relationships. Journal
cycle categories: A further analysis. Journal of of Personality and Social Psychology, 58,281-291.
Marriage and the Family, 46, 127-139. Feeney, J. A., & Nollcr, P. (1991). Attachmcnt style
Argylc, M. (1986). Social behavior problems in adoles- and verbal descriptions of romantic partners. Jour-
cence. In R. K. Silbcrciscn (Ed.), Development as ac- nal of Social and Personal Relationships, 8,187-215.
tion in context (pp. 55-86). Berlin: Springcr-Vcrlag. Fccney, J. A., Nollcr, P., & Callan, V. J. (1994). Attach-
Baron, R. M., & Kcnny, D. A. (1986). The moderator- ment style, communication and satisfaction in the
mediator variable distinction in social psychologi- early ycars of marriage. In K. Bartholomcw & D.
cal rcsearch: Conceptual, strategic, and statistical Perlman (Eds.), Advances in personal relationships:
considerations. Journal of Personality and Social Vol. 5. Adult attachment relationships (pp. 269-
Psychology, 51,1173-1182. 308). London: Jessica Kingsley Publishers.
Bowlby, J. (1969). Attachment and loss: Vol. 1. Attach- Fitzpatrick, M. A. (1988). Between husbands and wives.
ment. New York: Basic Books. Ncwbury Park, CA: Sage.
Bowlby, J. (1973). Attachment and loss: Vol. 2. Separa- Gottman, J. M., & Krokoff, L. J. (1989). Marital inter-
tion: Anxiety and anger. New York: Basic Books. action and satisfaction: A longitudinal view. Jour-
Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss. nal of Consulting and Clinical Psychology, 57,
New York: Basic Books. 47-52.
Campbcll, A., Converse, P. E., & Rodgers, W. L. Hazan, C., & Shaver, P. (1987). Romantic love concep-
(1976). The quality of American life: Evaluations tualizcd as an attachment proccss. Journal of Per-
and satisfactions. New York: Russell Sage. sonality and Social Psychology, 52,511-524.
Christensen, A,, & Sullaway, M. (1984). Commnnica- Kirkpatrick, L. A,, & Davis, K. E. (1994). Attachment
tion Patterns Qucstionnaire. Unpublishcd manu- stylc, gcndcr, and relationship stability: A longitu-
script, University of California, Los Angclcs. dinal analysis. Journal of Personality and Social
Collins, N. L., & Rcad, S. J. (1990). Adult attachment, Psychology, 66,502-512.
working modcls, and relationship quality in dating Kobak, R. R., & Hazan, C. (1991). Attachment in
couples. Journal of Personality and Social Psychol- marriage: Effects of security and accuracy of work-
ogy, 58,644-663. ing models. Joitrnal of Personality and Social Psy-
Fecney, J. A. (1990). The attachment perspective on chology, 60,861-869.
adult romantic relationships. Unpublished doc- Levy, M. B., & Davis, K. E. (1988). Lovestyles and
toral dissertation, University of Quecnsland, Bris- attachment styles compared: Their relations to each
banc. other and to various relationship characteristics.
Copyright of Personal Relationships is the property of Wiley-Blackwell and its content may
not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for
individual use.

You might also like