Death Studies, 28: 5977620, 2004

Copyright # Taylor & Francis Inc.

ISSN: 0748-1187 print / 1091-7683 online
DOI: 10.1080/07481180490476425

????????????????????????????????????????????????????

CONTINUING BONDS IN COPING WITH

THE DEATH OF A HUSBAND
????????????????????????????????????????????????????

NIGEL P. FIELD
Pacific Graduate School of Psychology, Palo Alto, California, USA
MICHAEL FRIEDRICHS
Deer Oaks Mental Health Associates,Victoria,Texas, USA

This study examined the continuing bond (CB) to the deceased in coping with the death of
a husband. Fifteen early-bereaved widows whose husband had died 4 months previously
and 15 later-bereaved widows whose husband had died more than 2 years ago were electronically signaled every 3 hours to complete a set of measures that included the PANAS positive and negative mood scales and CB coping. Participants completed these measures 4
times each day for 14 successive days. Following from an attachment theory perspective
on the role of CB in providing felt security, it was hypothesized that CB would be effective
as a way ofcoping in mood regulation, but that its effectiveness would be moderated by time
since the death. Consistent with predictions, a positive within-person relationship was
found between CB coping and positive mood for the later-bereaved group, but not for
early-bereaved widows. CB coping was also positively related to negative mood for both
early, and later-bereaved widow groups, however. Finally, in a lagged analysis, greater
use of CB was predictive of a shift toward more negative mood among early-bereaved
widows, but not for later-bereaved widows.The results were discussed in the context of
previous literature on the function of CB in adaptation to bereavement.

There has been increasing attention in the bereavement literature

devoted to the function of the continuing bond (CB) with the deceased
in coping with the death of a loved one (Field, Nichols, Holen, &
Horowitz, 1999; Klass, Silverman, & Nickman, 1996; Klass & Walter,
Received 14 February 2004; accepted 26 November 2003.
Address correspondence to Nigel P. Field, Pacific Graduate School of Psychology, 935 East
Meadow Drive, Palo Alto, CA 94303. E-mail: n.field@pgsp.edu

597

598

N. P. Field and M. Friedrichs

2001; Stroebe, Gergen, Gergen, & Stroebe, 1992). For example,

bereaved individuals often report a sense of the deceaseds presence as
a source of comfort in dealing with the stress over the loss. Reports of
such experiences well on after the death have been cited as evidence that
the attachment bond to the deceased loved one does not end with the
death but continues on and is an integral part of successful adjustment
to bereavement (Klass & Walter, 2001; Stroebe et al., 1992).
Despite much anecdotal evidence for the adaptiveness of CB, recent
empirical studies on the relationship between measures of CB and
bereavement-related distress have called into question the more blanket
claims for its adaptiveness. Such factors as the type of CB expression, as
to whether it involves use of the deceaseds possessions in an attempt to
gain comfort as opposed to a focus of fond memories (Field et al., 1999)
or the prominence of various CB expressions at a later point after the
death (Field, Gal-Oz, & Bonanno, 2003), may have important bearing
on whether CB is found to be adaptive. The present study builds on this
previous work in more explicitly examining CB expressions as ways of
coping through the use of a within-person process-oriented design that
examines individuals intensively over time. As elaborated below, such
methodologies are increasingly recognized in the coping literature as
necessary in being able to draw conclusions that can be applied to the
individual (Tennen, Affleck, Armeli, & Carney, 2000). To our knowledge, this study represents the first attempt to use this type of design in
a bereavement context.
A major limitation of existing studies on the function of CB is their
chief reliance on cross-sectional designs focusing exclusively on
between-person associations. In a few studies that have used longitudinal designs to examine CB in adjustment, the findings suggest that
excessive use of some CB expressions may actually interfere with grief
resolution (Field et al., 1999). However, such designs do not provide a
means for determining the effectiveness of CB as a way of coping at a
microanalytic process level. Furthermore, these designs typically
require having participants retrospectively summarize the extent to
which they have used various coping strategies over a long interval and
consequently are more likely to produce inaccurate results due to memory bias than measures focusing on much shorter time intervals (Stone
et al., 1998).
Recent work in the coping literature has emphasized the value of
microanalytic process designs involving repeated-measure indicators of

Continuing Bonds inWidowhood

599

coping and adjustment assessed at daily or within-day intervals in order

to identify temporally unfolding relationships between these variables
within the individual and their generalizability across individuals. This
work has demonstrated that such within-person comparisons are essential in being able to draw conclusions about the individual and that
within-person questions cannot be answered with exclusively betweenperson designs. In effect, exclusive reliance on between-person designs
may produce misleading results when applied to the individual; withinand between-person associations may differ not only in magnitude but
even in direction (Tennen et al., 2000). Inspired by this literature, the
present study used a within-person process design to examine the role
of CB as a way of coping in mood regulation.
Attachment theory provides a framework in explicating the function
of CB in coping with the loss of a loved one. According to attachment
theory, two important criteria that define an attachment bond include
the use of an attachment figure as a target for proximity maintenance
and in serving a safe haven function (Fraley & Davis, 1997; Hazan &
Shaver, 1994). As a target for proximity maintenance, individuals are
motivated to maintain physical proximity to an attachment figure and
to resist separations. Consistent with this, research findings have shown
that significant periods of separation from a spouse, such as during military leave, is a major source of distress (Vormbrock, 1993). Safe haven
refers to a tendency to seek out the attachment figure for contact, reassurance, and safety when under stress or threat. Proximity maintenance
and safe haven functions are said to provide the individual with a sense
of felt security (Kobak, 1999).
Although maintenance of physical proximity to an attachment figure
such as a spouse is no longer possible following his or her death, it nevertheless may be possible for the bereaved to experience felt security
through evoking a mental representation of the attachment figure.
Consistent with this, Bowlby (1980) cited evidence from a number of
bereavement studies (e.g., Glick,Weiss, & Parkes,1974; Rees,1971) showing that many well-adjusted widows and widowers reported an ongoing
sense of presence of their deceased spouse as a constant companion well
on after the death; this suggests that proximity maintenance at the
representational level is an integral part of long-term adjustment to
bereavement. Such sense of presence phenomena can provide the
bereaved with a sense of felt security despite the physical separation.
Similarly, support for the use of the deceased for safe haven is shown in

600

N. P. Field and M. Friedrichs

anecdotal accounts of bereaved individuals who report calling up the

image of the deceased under conditions of threat, such as when having
major surgery (Stroebe et al., 1992). Therefore, to the degree that the
bereaved is able to experience felt security through proximity to the
representation of the deceased, CB should serve as an effective way of
coping in mood regulation.

Present Study
Inspired by this recent attention to within-person designs in the coping
literature (Tennen et al., 2000), the present study used a processoriented approach to examine the impact of CB expressions as way of
coping on mood. Each bereaved participant completed a set of coping
and mood measures at various points during the day over the course of
two weeks. This provided a means for assessing the effectiveness of CB
expressions as ways of coping in mood regulation within the individual
in a way not possible in relying exclusively on between-person designs.
In addition to using a within-person process design, the present study
included a group of recently bereaved widows who had lost their spouse
on average within the previous four months and a group of laterbereaved widows whose husband had died on average about two years
ago. The reason for including these two groups is that time since the
death might be an important factor in the effectiveness of CB in mood
regulation. It is widely recognized in the bereavement literature that
reminders of the deceased earlier on after the death are considerably
more emotionally painful than later on (Attig, 2001; Parkes, 1998). This
is reflected in characteristic decreases in stress-specific symptoms and
depression over time (Field & Horowitz,1998). From an attachment theory perspective, because the permanence of the separation is not fully
registered at the attachment system level in the initial period following
the death (Archer, 1999; Shaver & Tancredy, 2001), the attachment goal
of regaining physical proximity to the deceased is less likely to have been
relinquished at an earlier point after the death. Consequently, attempts
to make use of CB as a way of coping may be less effective in mood regulation earlier on after the death because their use will remind the
bereaved of the deceaseds physical absence, which is discrepant with
the goal of regaining physical proximity. Later on after the death, in
having had the time to more fully accommodate to the reality of

Continuing Bonds inWidowhood

601

the new life situation without the spouse and to let go of the attachment
goal toward regaining physical proximity, the bereaved should be able
to make more effective use of CB in coping. In other words, at a point
well on after the death, the bereaved may be more capable of experiencing felt security exclusively through psychological proximity to the
representation of the deceased. To the extent that the bereaved makes
use of CB in this capacity, one can say that the CB endures after the
death and is adaptive.
Beyond considering time after the death as a possible moderator of
the effectiveness of CB coping expressions in mood regulation, this study
focused on the impact of CB on positive mood in addition to negative
mood. Coping researchers have more recently extended their work
beyond focusing exclusively on negative affect during a stressful period
in giving greater attention to coping processes that generate positive
affect (Folkman & Moskowitz, 2000). This interest in positive affect
stems from empirical findings showing that individuals encountering
stressful life events report experiencing positive affect in addition to
distress, and that both positive and negative affect can occur concurrently during a stressful period (Folkman, 1997; Folkman & Moskowitz,
2000). Furthermore, empirical evidence has been mounting for the
function of positive affect in offsetting the debilitating physiological
and psychological effects of stress (Folkman & Moskowitz, 2000;
Moskowitz, Folkman, Colette, & Vittinghoff, 1996). CB expressions as
ways of coping may be expected to generate positive affect for a number
of reasons. For example, to the degree that CB is an outcome of a rewarding past relationship with the deceased, the use of CB expressions may
serve as a reminder of this, thereby enabling the bereaved to experience
a sense of fulfillment through the enduring legacy of having had a happy
relationship.This may be especially important for older bereaved individuals who, in having had a good past relationship with the deceased,
can experience the satisfaction of knowing this and thus are less likely
to be plagued with regrets over lost opportunities. Furthermore, effective
use of CB in coping implies having internalized the relationship, such
that the bereaved can feel emotionally sustained by the internal presence
of the deceased in a similar fashion to the function that the deceased
might have served in contributing to the well-being of the bereaved in
the past when he or she was alive.
The following hypothesis is therefore offered regarding the use of CB
as a way of coping in mood regulation as derived from an attachment

602

N. P. Field and M. Friedrichs

theory based perspective on the function of CB in providing a secure

base and safe haven. CB should prove to be an effective way of coping
in mood regulation shown in a within-person relationship between the
extent of CB usage and decreased negative and increased positive mood;
however, this relationship is expected to be moderated by time since the
death such that bereaved widows who have lost their husband a longer
time ago (i.e., more than two years ago) will be able to make more effective use of CB in mood regulation relative to bereaved widows who have
recently lost their husband (i.e., approximately four months ago).
Method
Participants
Thirty widows over the age of 55 were recruited via notices posted in
community centers and other public places as well as through advertisements in local newspapers. Inclusion criteria specified that the death
must have occurred either within the previous six months or more than
one year ago, that participants must be over the age of 50, and have lived
with the deceased for at least five years prior to the loss.
The sample consisted of 15 early-bereaved widows, who had lost their
husbands on average within the previous 4.1 months (SD 1.9), and a
later bereaved group, whose husbands had died on average 25.6 months
ago (SD 8.1). All of the women were Caucasian. The two groups did
not differ on any of the demographic variables, summarized inTable 1.
Procedures
An initial face-to-face meeting was arranged by Michael Friedrichs
with each participant in which she was administered a packet of questionnaires.The packet of questionnaires contained a set of demographics
items, symptom inventories that included the Beck Depression Inventory (Beck & Steer, 1987) and the Impact of Event Scale (Horowitz,
Wilner, & Alvarez, 1979), as well as other measures that are not part of
the present report. After completing the questionnaire packet, each participant was given a three-ring binder containing daily process measures
upon which was affixed a light-weight electronic timer originally
designed to ensure medication compliance (i.e., the Medi-Minder
Medicine Timer). This alarm was set to remind the participant when to

603

Continuing Bonds inWidowhood

TABLE 1 Demographic Characteristics
Characteristic
Age
Years cohabitation
Education
High school
Some college
Bachelors degree
Masters degree
Income
10 K730 K
30 K750 K
50 K770 K
Over 70 K
Employment
Unemployed
Part-time
Full-time
Retired
Mode of death
Stroke
Heart disease
Alzheimers
Cancer
Other

Early bereaved (n 15)

Later bereaved (n 15)

M 65.5 SD 10.8
M 33.6 SD 20.2

M 66.3 SD 8.9
M 35.6 SD 15.5

3
5
2
5

5
5
2
3

3
6
3
3

3
3
3
6

0
4
6
5

2
3
3
7

2
4
2
6
0

1
4
1
4
4

Note. M Mean; SD Standard Deviation; K $1000.

fill out the questionnaire.The binder contained 56 identical forms. Each

of these forms was clearly marked with the appropriate day and time at
which it was to be filled out. Each participant was informed that she
would be completing these measures four times each day over the course
of 14 successive days. The electronic pill timer was set to remind each
participant to complete a form at designated times three hours apart
each day.
Participants completed the within-daily measures over a two-week
period. At the one-week point, the investigator met each participant
face-to-face to ensure that she was not experiencing undue emotional
strain from participating in the study and to collect the first weeks questionnaires. The investigator met with each participant a final time upon
completing the second week of data collection. Participants were then
debriefed and paid for their involvement in the study.

604

N. P. Field and M. Friedrichs

In light of the intensive nature of the repeated-measures data collection, participants were alerted to the possibility of annoyance or unpleasant side effects that could accompany the frequent completion of these
measures. The investigator also provided a phone number where he
could be contacted at any time during the data collection. In addition,
participants were provided with telephone numbers where free crisis
counseling could be obtained, as well as free and sliding-fee clinical
and community psychological resources, should they deem it necessary.
Measures
Impact of Event Scale (IES)
The (IES; Horowitz et al., 1979) is a widely used measure of subjective stress, answered in response to a specified stressful life event. In the
context of the present study, the IES was answered with respect to the
death of the husband. The IES is composed of an Intrusion and Avoidance subscale. Intrusion addresses the involuntary entry into awareness
of ideas, memories, and emotions associated with the event. Avoidance is
the conscious attempt to divert attention from thoughts and feelings
related to the distressing life event. Respondents are asked to estimate
the degree to which each item describes their experience over the past
week on a 4-point scale ranging from not at all to often. Support for
the reliability and validity of the IES has been obtained with diverse
populations (see Horowitz, Field, & Classen, 1993). In the context of
bereavement, the IES has been capable of distinguishing those seeking
psychotherapy for complicated grief from a non-clinical bereaved group
who experienced similar interpersonal losses (Horowitz et al., 1984).
The IES has also been found to be sensitive of changes in stress symptom
severity over the course of bereavement (Field & Horowitz, 1998).
Beck Depression Inventory (BDI)
The BDI (Beck & Steer, 1987) is a 21-item scale that assesses the presence and severity of various affective, cognitive, motivational, vegetative, and psychomotor symptoms of depression over the past week. It is
widely used and psychometrically sound.
Repeated Measures Questionnaire
This measure was developed for the present study to examine withinperson relationships between the use of CB as a way of coping and mood.

Continuing Bonds inWidowhood

605

It included the Positive and Negative Affect Schedule (PANAS; Watson,

Clark, & Tellegen, 1988) as a measure of positive and negative mood.
A CB coping measure was specially developed for this study. The
repeated-measure questionnaire also included other items that are not
part of the present study and therefore are not discussed here. Participants answered each of the mood and CB coping items with respect to
their experience during the past three hours.
PANAS
Watson et al. (1988) designed the PANAS to be a short, reliable, and
easy-to-administer scale for assessing positive and negative mood. The
10-item Positive Affect (PA) scale reflects the extent to which a person
feels alert, enthusiastic, and active. The 10-item Negative Affect (NA)
scale describes a variety of aversive mood states such as anger, contempt,
disgust, guilt, fear, and nervousness (Watson et al., 1988). The PANAS
is flexible in that the time interval for which it is answered can be
tailored to the specific purposes of the research endeavor. Watson et al.
confirmed that the two subscales remained highly internally consistent
(Cronbachs coefficient alpha > .80) when answered for different time
intervals. In the present study context, in which participants were
instructed to answer to the PANAS with respect to thepast three hours,
similarly high internal consistency reliabilities were obtained (.94 for
PA and .88 for NA over a total of 1605 samplings). The PANAS has also
been shown to possess good convergent and discriminant validity
(Watson et al., 1988). The discriminant validity of the PANAS was confirmed in the present study in which the PA and NA scales were shown
to be only weakly correlated with each other (
.20 over a total of 1605
samplings).
CB Coping
This 6-item measure captures a range of CB expressions identified in
the bereavement literature as commonly reported ways of maintaining
a connection to the deceased following the death of a significant other
(e.g., Field et al., 2003; Parkes, 1998; Shuchter & Zisook, 1988). These
consisted of the following CB expressions: (a) had an imaginary (or if
you believe, real) conversation with my husband; (b) tried to do things
that that would have made my husband happy; (c) tried to think of good
memories of my husband; (d) imagined reuniting with him in heaven;
(e) used photos or his belongings to feel closer to him; (f ) imagined or

606

N. P. Field and M. Friedrichs

felt him guiding me, watching over me, or comforting me. Participants
were instructed to rate how often during the past three hours they used
each of these CB expressions as ways of coping to help them deal with
their feelings related to the loss of their husband, on a 5-point scale from
0 not at all to 4 constantly. Because this measure was shown to be internally
consistent (Cronbachs alpha coefficient of .85 over a total of 1605
samplings), a composite score was calculated by averaging the scores
across the six items. This served to cut down the number of analyses,
and thereby decrease the likelihood for obtainingType 1 errors.
Compliance
Of the total number of diary forms administered to all participants
over the course of the study, consisting of a total of 1,680 (i.e., 30656),
1,605 were fully completed for a 96% completion rate. Forms with
missing data were deleted from the analysis. There were no noteworthy
differences between participants in the number of completed diary forms
and no relationship shownbetweenthe number of completed diary forms
and the symptom inventory scores or time since the death. Therefore,
the data for all 30 participants was included in the analysis. The HLM5
(Raudenbush, Bryk, Cheong, & Congdon, 2000) statistical software
used in the analyses was capable of accommodating such missing data.
Results
Time Since Death and Symptomatic Distress
Independent t tests were conducted to compare the early- and laterbereaved widows in the severity of stress and depressive symptoms. As
shown in Table 2, the early bereaved reported significantly higher
intrusion on the IES relative to the later-bereaved group. The results
for intrusion are similar to longitudinal bereavement studies using the
IES in showing a decrease in intrusion scores from early to later points
after the death (e.g., Field & Horowitz, 1998). The two groups do not
differ in avoidance or in depression, however.
Descriptive Analyses of Repeated Measures
To compare the average ratings of the early- and later-bereaved groups
on the repeated-measures data, we obtained aggregated scores for each

607

Continuing Bonds inWidowhood

TABLE 2 Comparing Early and Later Bereaved on Symptom Inventory Mean

Scores

Measure
IES-Intrusion
IES-Avoidance
BDI

Early bereaved

Late bereaved

SD

SD

t value

21.40
16.73
10.93

3.04
5.27
5.87

15.73
15.33
9.47

3.96
5.63
6.46

4.40***
0.70
0.65

Note. Values in parentheses represent standard deviations. IES Impact of Event Scale;
BDI Beck Depression Inventory.
*p < .05. **p < .01. ***p < .001.

participant on each repeated measure and then we performed independent t tests on these aggregated scores. The results of this analysis for
the PANAS negative and positive mood and continuing bonds (CB)
repeated-measures variables are presented inTable 3.
Noteworthy here is the much higher average score for positive mood
relative to negative mood in both the early- and later-bereaved groups.
The early and later bereaved do not differ in their average levels of negative and positive mood. A marginally significant relationship is shown
for CB, however, indicating a tendency for the early bereaved to make
greater use of CB expressions than later bereaved widows. Independent
t tests were also performed comparing the early- and later-bereaved
groups intraindividual variance scores for each of the repeated
measures, indicating no such difference as a function of group.

TABLE 3 Comparing Early and Later Bereaved on Aggregate Repeated Measures

Scores

Measure
PANAS Positive Affect
PANAS Negative Affect
CB Composite

Early bereaved

Late bereaved

SD

SD

t value

27.59
12.79
0.81

7.99
3.38
0.75

27.75
11.58
0.42

3.96
1.79
0.11

0.03
1.22
1.76{

Note. Values in parentheses represent standard deviations. CB Continuing Bonds Scale;

PANAS Positive and Negative Affect Schedule.
{
p < .10.

608

N. P. Field and M. Friedrichs

Symptoms and Aggregate Repeated Measures Intercorrelations

Intercorrelations among the symptom and aggregate repeated-measures scores are presented in Table 4. A significant correlation is shown
for the aggregate CB composite measure with IES-Intrusion (r .45,
p < .01) but not for IES-Avoidance (r .02, ns) or for the BDI
(r .04, ns). The significant result for CB with Intrusion is consistent
with Field et al. (2003) who found a significant positive relationship
between CB and grief-specific symptoms. No relationship exists
between the aggregate CB composite measure and the PANAS negative
mood (r .24, ns) or positive mood (r .14, ns) measures, however.
Hierarchical Linear Modeling Analyses for the Repeated Measures
Because of possible reactivity effects of intensive repeated measurement
of CB coping over the two week assessment period, a latent growth
curve analysis was performed prior to the main set of analyses to identify
any systematic changes in responding to the CB items over the course
of the study. This involved estimating a slope score for each participant
based on her CB composite scores for the 56 repeated assessments. The
slope represents a measure of change over time in CB ratings. If indeed
reactivity effects exist, a significant linear change in the extent of CB
coping should be shown, as reflected in a significant slope value pooled
across participants. This analysis was undertaken using the hierarchical
linear modeling (HLM) analysis using the HLM5 statistical software
(Raudenbush et al., 2000). A non-significant slope value of 0.00

TABLE 4 Symptoms and Aggregate Repeated Measures Intercorrelations

Measure
1. IES Intrusion
2. IES Avoidance
3. BDI
4. PANAS Negative
5. PANAS Positive
6. CB

0.24
0.36*
0.32

0.21
0.45*

0.41*
0.44*

0.34
0.02

0.36*

0.47**
0.04

0.28
0.24

0.14

Note. IES Impact of Event Scale; BDI Beck Depression Inventory; PANAS Positive and
Negative Affect Schedule; CB Continuing Bonds Composite.
*p < .05. **p < .01, two-tailed.

Continuing Bonds inWidowhood

609

(t
0.42, ns) indicated no evidence for reactivity effects of repeated

assessments on CB ratings. Similar analyses were conducted on the
repeated mood measures. A significant slope value of
0.01 (t
2.65)
for negative mood indicated a tendency toward a slight decrease in the
extent of negative affectivity over the course of the study, whereas the
slope value of
.02 for positive mood was statistically non-significant
(t
0.90, ns).
The main analysis involved an HLM analysis to examine the relationship between CB coping and mood in the repeated measures data.
This analysis permits a simultaneous analysis of Level 1 within-person
and Level 2 between-person variation. At Level 1, a unique regression
equation is estimated for each individual. Specifically, the within-person
Level 1 equation specifies the predicted value of the criterion variable
Yij for a particular participant i on a given occasion j, as a linear
function of her level of Y at the point b0j on the regression line where
the predictor variable Xij is equal to the participants mean score, plus
the effect of the predictor variable bljXij on that occasion, and a random
residual component rij.1 The specification of the Level 1 within-person
model is shown in Equation 1:
Yij b0j b1jXij rij

The Level 2 between-person model is specified in Equations 2 and 3

below. Here the Level 1 coefficients are regressed on the between-person
variable of interest,Wj. The Equation 2 g00 represents the mean Y score
of the entire sample, g01Wj represents the contribution of the betweenperson variable of interest as a predictor of the Level1 b0j, and u0j represents a random residual component. The Equation 3 slope term g10
represents the mean slope score for the entire sample, g11Wj represents
the contribution of the between-person variable of interest as a predictor
of the Level 1 slope term b1j, and u1j represents a random residual
component.
b0j g00 g01Wj u0j

b1j g10 g11Wj u1j

1
In the main set of HLM analyses, the predictor variables in the Level 1 equation were centered prior to the analysis by transforming each participants scores to deviations from her mean
score for this variable.

610

N. P. Field and M. Friedrichs

In the present study context, the Level 1 within-person analysis

entailed regressing each participants repeated-measures mood ratings
on her composite CB rating scores.The Level 2 between-person analysis
involved treating each participants estimated coefficient scores from the
Level 1 analysis as criterion variables and introducing between-person
predictors in the regression analysis in an attempt to explain interindividual variation in the Level 1 estimated intercept and slope parameters.
In the present study, time since the death (i.e., early vs. later bereaved
groups) was the between-person variable of interest.
Because the Level 1 residuals were not independent (i.e., were significantly correlated), a first-order auto-regressive model correcting for
autocorrelations among the Level 1 residuals was used in all of the
HLM analyses (Raudenbush et al., 2000). The first HLM analysis
involved examining the relationship between the CB composite measure
and PANAS negative mood. The results of this analysis are shown in
Table 5. The baseline intercept value of 12.20 represents the predicted
average negative mood score for the entire sample when the CB composite measure score is equal to the average score for the sample as a whole.
The group intercept value of
1.19 indicates that the later bereaved
group has a lower average negative mood score than the early bereaved
group but this difference is not statistically significant. The significant
baseline slope value of .45 indicates that a one-unit increase in the CB
composite score is associated with a .45 increment in negative mood for
the bereaved sample as a whole. The non-significant group slope value
of
.13 indicates that the significant relationship between CB and negative mood is not moderated by time since the death. In other words,

TABLE 5 Hierarchical Linear Modeling Analysis Predicting PANAS Negative

Mood

Intercept
Mean CB
Group
Slope
Mean CB
Group

Coefficient

SE

t ratio

12.20

1.19

0.47
0.95

25.77***

1.26

0.45

0.13

0.13
0.27

3.41**

0.50

Note. CB Continuing Bonds Composite.

**p < .01. ***p < .001, two-tailed.

Continuing Bonds inWidowhood

611

greater CB usage is predictive of more negative mood, independent of

time since the death.
A second HLM analysis involved examining the relationship
between the CB composite measure and PANAS positive mood. The
results are shown inTable 6. The baseline intercept value of 27.74 represents the predicted average negative mood score for the sample as a
whole when the CB composite measure score is equal to the mean score
for the entire sample. The non-significant group intercept value of 0.10
indicates that the early-and later-bereaved groups do not differ in average positive mood scores. The significant baseline slope value of 0.96
indicates a tendency for higher CB scores to be associated with more
positive mood ratings for the sample as a whole. However, a significant
group slope value of 2.49 indicates that time since the death moderates
the relationship between CB and positive mood. In light of this significant interaction, a separate HLM analysis was conducted on the earlyand later-bereaved groups. A significant slope value of 2.22,
t(804) 6.20, p < .001, was found for the later-bereaved group whereas
the slope value of
0.26, t(797)
.67, for the early-bereaved group
was non-significant. In effect, these results suggest that the use of CB as
a way of coping is effective in facilitating positive mood but only at a later
point after the death.
Although CB and mood were shown to be related in the above series
of HLM analyses, it was not possible to determine whether, in fact, CB
had an effect on changes in mood based on these synchronous analyses
in which all scores that were entered into the Level 1 within-person
regression equation were obtained at the same point in time. Therefore,
an additional set of HLM lagged analyses was conducted to address
TABLE 6 Hierarchical Linear Modeling Analysis Predicting PANAS Positive Mood

Intercept
Mean CB
Group
Slope
Mean CB
Group

Coefficient

SE

t ratio

27.74
0.10

1.35
2.71

20.50***
0.04

0.96
2.49

0.27
0.54

3.59**
4.64***

Note. CB Continuing Bonds Composite.

**p < .01. ***p < .001, two-tailed.

612

N. P. Field and M. Friedrichs

more directly the impact of CB in mood regulation. If indeed CB is

assumed to have a causal impact on changes in mood, this should be
evident in analyses that contain a criterion variable representing change
in mood.
A lagged analysis would provide a means for assessing the impact of
the use of CB at an earlier point in the day on changes in mood from this
earlier point to a later point in the day. If CB is shown to influence such
changes in mood, this would provide more direct support for its role in
mood regulation. This involved conducting an HLM analysis in which
mood assessed at an earlier point in the day (Time 1) was entered in the
Level 1 regression model in addition to the CB composite variable in
the prediction of the same mood variable obtained at the next administration of data collection (Time 2). In having included Time 1 mood as
a predictor in the regression model, the Time 2 mood criterion variable
serves as a residualized change score from the previous wave to the next
wave of sampling. The informal specification of the Level 1 within-person model for this analysis is as follows:
Time 2 mood b0 b1Time 1 mood b2Time 1 CB error
Specifically, this analysis involved entering predictors obtained in the
first sampling each day in predicting mood assessed in the second
sampling of that day, predictors from the second sampling in predicting
mood in the third sampling that day, and predictors from the third
sampling in predicting mood in the final sampling that day. A separate
analysis was conducted for negative and positive mood.
The resultsofthe analysis for negativemoodare summarizedinTable 7.
The marginally significant baseline slope value of 0.24 for CB indicates
a tendency for higher CB scores to be predictive of shift toward more
negative mood for the sample as a whole. However, a significant group
slope value of
.62 indicates that time since death moderates the relationship between CB and negative mood. Given this significant interaction effect, a separate follow-up HLM lagged analysis was conducted
on the early- and later-bereaved groups. A significant slope value of
0.54, t(577) 2.46, p < .05, found for the early bereaved group verified
that greater within-person use of CB was predictive of an increase in
negative mood at an earlier point after the death. On the other hand, a
non-significant slope value of
0.07, t(587)
0.36, for the laterbereaved group indicated no effect of CB on change in negative mood

613

Continuing Bonds inWidowhood

TABLE 7 Hierarchical Linear Modeling Lagged Analysis Predicting PANAS

Negative Mood

Intercept
Mean CB
Group
PANAS Negative Slope
Mean PANAS Negative
Group
CB Slope
Mean CB
Group

Coefficient

SE

t ratio

12.11

1.08

0.48
0.95

25.47***

1.14

0.25
0.04

0.03
0.06

8.76***
0.77

0.24

0.62

0.14
0.29

1.66{

2.16*

Note. CB Continuing Bonds Composite.

{
p < .10. *p < .05. ***p < .001, two-tailed.

at a later point after the death. In effect, these results show that the
impact of CB on subsequent increase in negative mood holds true only
for the early-bereaved group.
In the analysis involving positive mood, the results indicated a nonsignificant baseline slope value of 0.16 for CB, t(1164) 0.58, and a
non-significant value of 0.50 for the effect of group, t(1164) 0.87.
Therefore, no support was found for the impact of CB on changes in
positive mood in either the early- or later-bereaved groups.
Finally, for comparative purposes, a second pair of lagged analyses
were performed in which CB was treated as the criterion variable. Here,
the Time 1 CB and mood measures were entered into the regression
equation in predicting Time 2 CB. The results for negative mood as a
predictor are summarized inTable 8.The significant baseline slope value
of 0.01 for the impact of negative mood on changes in use of CB indicate
that greater within-person negative mood is predictive of a subsequent
increase in the use of CB for the bereaved sample as a whole. The nonsignificant group slope value of .00 indicates that this significant
relationship between CB and negative mood is not moderated by time
since the death. In other words, greater negative mood is predictive of
a shift toward greater use of CB, and this relationship holds true for both
early- and later-bereaved widows.
In the HLM lagged analysis involving positive mood as a predictor, a
non-significant baseline slope value of 0.00 was shown for positive
mood, t(1164) 0.91, and a non-significant value of 0.01 for the effect

614

N. P. Field and M. Friedrichs

TABLE 8 Hierarchical Linear Modeling Lagged Analysis Predicting Changes in CB

Intercept
Mean PANAS Negative
Group
CB Slope
Mean CB
Group
PANAS Negative Slope
Mean PANAS Negative
Group

Coefficient

SE

t ratio

0.59

0.41

0.11
0.21

5.53***

1.94{

0.33

0.16

0.03
0.05

12.20***

2.86**

0.01

0.00

0.00
0.01

2.43*
0.48

Note. CB Continuing Bonds Composite.

{
p < .10. *p < .05. **p < .01. ***p < .001, two-tailed.

of group, t(1164) 1.52.Therefore, no support was found for the impact

of positive mood on changes in CB in either the early- or later-bereaved
groups.
Discussion
This study examined the role of CB expressions as ways of coping in
mood regulation following the death of a husband. The results highlighted the complexity of the within-person relationships between CB
and mood. The results also suggested that the effectiveness of CB in
mood regulation may be partly contingent on time since the death.
The results for the early-bereaved widows in both the synchronous
and lagged HLM analyses indicated that, if anything, CB was not effective in mood regulation. In particular, the positive slope value for CB
in predicting negative mood in the lagged analysis suggested that greater
use of CB in coping precipitated a shift toward increased negative mood.
The results for the effect of CB on mood were more ambiguous in the
later-bereaved group. Although a significant positive slope for the effect
of CB in predicting positive mood in the synchronous analysis is suggestive that CB is effective in mood regulation among later bereaved
widows, the non-significant results for CB on positive mood in the
lagged analysis argues against drawing any firm conclusions regarding
the causal role of CB in mood regulation. Moreover, the significant association between CB and negative mood in the synchronous analysis

Continuing Bonds inWidowhood

615

weakens further any strong claim as to the effectiveness of CB in mood

regulation within the later bereaved group.
The significant moderating effect of time since the death on the relationship between CB and mood in both the synchronous and lagged
analyses is consistent with our hypothesis that CB may be less effective
in mood regulation earlier on than later on after the death. If anything,
however, the results are more suggestive of the detrimental effect of CB
in mood regulation at an earlier point after the death than they are of
the beneficial effects of CB later on after the death. In this respect, at
best, these results provide limited support for the adaptiveness of CB as
a way of coping in bereavement.
Another noteworthy finding was the bidirectional relationship
between CB and mood in the lagged analyses such that negative mood
was also predictive of a subsequent increase in the use of CB. This may
not be all that surprising if considering that negative affect should be
expected to incite greater use of CB as a way of coping in an attempt to
regulate negative mood.
Although it might be argued that the lagged analyses are more informative than the synchronous analyses in identifying causal relationships
between variables, any strong conclusions regarding directionality is
not warranted in a correlational study. Moreover, the temporal intervals
that were selected in conducting lagged analyses in the present study,
and for that matter the temporal interval chosen in the synchronous
analyses, may have had direct bearing on the results yielded. In fact, it
is not known whether 3-hour sampling intervals would be the most
suitable length of time to use in determining the impact of CB in mood
regulation. Many fluctuations in both mood and the use of CB could
have occurred within this time frame. For example, it is possible within
a 3-hour interval for a distressing reminder of the spouses death initially
to incite negative affect that, in turn, boosts greater use of CB in order
to cope with this negative affect that, consequently, induces positive
affect. This might explain the positive relationship found between CB
and both negative and positive mood in the later-bereaved widows in
the synchronous analyses. This explanation is also consistent with an
attachment perspective outlined earlier wherein greater distress (i.e.,
increased negative mood) would be expected to activate the attachment
system, expressed in increased use of CB coping, that leads to improved
mood. It is therefore compelling that the lagged analyses yielded significant results despite these limitations. In future research, it might be

616

N. P. Field and M. Friedrichs

worthwhile to consider using shorter interval samplings in investigating

causal relationships.
The results of the synchronous analyses for the later bereaved group
also clarify the independence of positive and negative mood in showing
that greater use of CB expressions among widows can be associated with
increases in both positive and negative mood within a relatively narrow
time frame. They are in keeping with those of Folkman & Moskovitz,
(2000) in showing that it is possible to experience both positive and
negative affect within a stressful encounter and the importance of
including measures of positive affect in addition to negative affect in
assessing coping in adjustment to stressful life events. This significant
finding for positive mood in the later-bereaved group is also in keeping
with anecdotal accounts in the bereavement literature of how over time
the bereaved is capable of experiencing more sustained emotions of joy
and nostalgia when thinking about the deceased relative to earlier on
after the death (Attig, 2001).
The within-person positive relationship found between CB and negative mood together with the between-person positive correlation shown
between the aggregate CB measure and the IES-Intrusion stress symptom measure are in keeping with the results of previous studies indicating that CB is associated with greater bereavement-related distress
(Field et al., 1999; Field et al., 2003). On the other hand, the significant
within-person positive relationship between CB and positive mood in
the later-bereaved widow group in the synchronous analysis suggests its
possible adaptiveness. If we had relied exclusively on the between-person
relationship involving the aggregate CB measure and IES-Intrusion,
we might have simply concluded that CB was maladaptive. Our results
therefore offer further credence to the recent emphasis in the coping
literature on the importance of using within-person daily process
designs in order to uncover the dynamic relationships among stress and
coping measures (Tennen et al., 2000).
There are a number of limitations in this study that may have bearing
on the conclusions that can be drawn from the results. Firstly, as
previously mentioned, one cannot draw strong conclusions regarding
the direction of causality in the relationship between variables in a
correlational study.
A second limitation is the failure to include a stressor measure such as
a daily hassles measure in the repeated-measures questionnaire.
According to attachment theory, proximity seeking should become

Continuing Bonds inWidowhood

617

more prominent when a person is under stress or threat. If evoking a

representation of the deceased serves an attachment function of providing felt security when under stress, use of CB as a way of coping should
be more prominent on exposure to a stressor and this, in turn, should
be effective in reducing stress. Therefore, a more conclusive test of an
attachment theory based perspective on CB would involve incorporating an assessment of stressors in the repeated measure to examine
whether the use of CB coping moderated the impact of stressor severity
on mood. Presumably, if CB were effective in mood regulation, it would
be shown to serve as an important personal resource in buffering the
effect of a stressor.
Thirdly, because participants answered the repeated measures on
standard questionnaire forms, it is not possible to know whether they
completed these questionnaires at the designated times. In future
research it would be advantageous to administer the repeated measures
using hand-held computer technology to address this limitation. The
use of such technology, in which participants complete the measures
directly on the hand-held computers, can remedy this potential problem
(Barrett & Barrett, 2001).
Fourthly, the repeated measures may have caused reactivity effects
given how frequently they had been administered. For example, in having to answer the CB items repeatedly, this may have led participants to
make greater use of these as ways of coping than would have occurred
naturally if they had not been repeatedly exposed to these items. However, because there was no evidence of change over time in the ratings
of the CB items in the growth curve analysis, this may not have been a
serious problem.
Fifthly, despite the fact that the present study incorporated an attachment theoretical framework in generating the hypothesis and in
explaining the results, no individual difference attachment measures
were included in the study. Because individual differences in attachment
security are throught to be an important risk factor in adjustment to
bereavement (Field & Sundin, 2001; Neimeyer, Prigerson, & Davies,
2002), and considering that it may be so partly through its effect on
CB, it would be important to address this in future research.
Sixthly, the adaptiveness of different CB expressions was not
addressed. Because there is evidence that the type of CB expression has
implications for whether or not it is adaptive (Field et al.,1999), it would
be worthwhile in future research to examine the relative effectiveness of

618

N. P. Field and M. Friedrichs

different CB expressions in mood regulation.This would require including a more comprehensive set of CB expressions than in the present
study.
Seventhly, the homogeneous sample that consisted exclusively of older
Caucasian widows places restrictions on the generalizability of the
results. It would be important in future work to replicate the findings
on both men and women of different ages and ethnic backgrounds.
Finally, it could be argued that the research benefits of a daily
measures study are outweighed by the demands of this methodology on
participants, especially so for those who are recently bereaved. Although
this is indeed an important ethical concern, the growth curve analysis
for the mood measures did not indicate that widows levels of distress
increased over the course of the study. Furthermore, during the debriefing, the majority of participants mentioned that repeatedly answering
the questionnaires was helpful in working through the loss. The
availability of the investigators throughout the two-week data collection
period and face-to-face contact at weekly intervals may have played an
important role in minimizing potential detrimental effects of participating in this kind of study. Certainly, considerable sensitivity and care must
be taken in conducting this kind of research.
In summary, the present study contributes to the existing literature on
the function of CB in adaptation to bereavement by directly examining
the use of CB as a way of coping in mood regulation. The results highlight further the increasing recognition of the complex relationship
between CB and adjustment in the bereavement literature, and the value
of within-person process designs in extending our understanding of
these relationships.
References
Archer, J. (1999).The nature of grief:The evolution and psychology of reactions to loss. London:
Routledge.
Attig,T. (2001). Relearning the world: Making and finding meanings. In R. Neimeyer
(Ed.), Meaning construction and the experience of loss (pp. 33753). Washington, DC:
American Psychological Association.
Barrett, L. F. & Barrett, D. J. (2001). An introduction to computerized experience sampling in psychology. Social Science Computer Review, 19, 1757185.
Beck, A.T. & Steer, R. A. (1987). Beck Depression Inventory manual. San Antonio,TX:The
Psychological Corporation.
Bowlby, J. (1980). Attachment and loss.Vol. 3 Loss: Sadness and depression. London: Hogarth.

Continuing Bonds inWidowhood

619

Field, N. P. & Horowitz, M. J. (1998). Applying an empty-chair monologue paradigm

to examine unresolved grief. Psychiatry, 61, 2797287.
Field, N. P., Gal-Oz, E., & Bonanno, G. A. (2003). Continuing bonds and adjustment
at 5 years after the death of a spouse. Journal of Consulting and Clinical Psychology, 71,
1107117.
Field, N. P., Nichols, C., Holen, A., & Horowitz, M. J. (1999). The relation of continuing attachment to adjustment in conjugal bereavement. Journal of Consulting and Clinical Psychology, 67, 2127218.
Field, N. P. & Sundin, E. (2001).The role of adult attachment in adjustment to conjugal
bereavement. Journal of Social and Personal Relationships, 18, 3477361.
Folkman, S. (1997). Positive psychological states and coping with severe stress. Social
Science and Medicine, 45, 120771221.
Folkman, S. & Moskowitz, J. T. (2000). Positive affect and the other side of coping.
American Psychologist, 55, 6477654.
Fraley, R. C. & Davis, K. E. (1997). Attachment formation and transfer in young
adults close friendships and romantic relationships. Personal Relationships, 4,
1317144.
Glick, I. O., Weiss, R. S., & Parkes, C. M. (1974).The first year of bereavement, NewYork:
JohnWiley.
Hazan, C. & Shaver, P. R. (1994). Attachment as an organizational framework for
research on close relationships. Psychological Inquiry, 5, 1722.
Horowitz, M. J., Field, N. P., & Classen, C. C. (1993). Stress response syndromes and
their treatment. In L. Goldberger & S. Breznitz (Eds.), Handbook of stress:Theoretical
and clinical aspects (2nd ed., pp. 7577773). NewYork: Free Press.
Horowitz, M. J.,Weiss, D. S., Kaltreider, N., Krupnick, J., Marmar, C.,Wilner, N., &
DeWitt, K. (1984). Reactions to the death of a parent: Results from patients and
field subjects.TheJournal of Nervous and Mental Disease, 172, 3837392.
Horowitz, M. J.,Wilner, N., & Alvarez,W. (1979). Impact of Event Scale: A measure of
subjective stress. Psychosomatic Medicine, 41, 2097218.
Klass, D., Silverman, P., & Nickman, S. L. (Eds.) (1996). Continuing bonds: New understandings of grief.Washington, DC: Taylor & Francis.
Klass, C. & Walter, T. (2001). Processes of grieving: How bonds are continued. In
M. S. Stroebe, R. O. Hansson, W. Stroebe, & H. Schut (Eds.), Handbook of bereavement research: Consequence, coping, and care (pp. 4317448). Washington, DC: American
Psychological Association.
Kobak, R. (1999). The emotional dynamics of disruptions in attachment relationships.
In J. Cassidy & P. Shaver (Eds.), Handbook of attachment: theory, research, and clinical
applications (pp. 21743). NewYork: Guilford Press.
Moskowitz, J.T., Folkman, S., Collette, L., & Vittinghoff, E. (1996). Coping and mood
during AIDS-related caregiving and bereavement. Annals of Behavioral Medicine,
18, 49757.
Neimeyer, R. A., Prigerson, H. G., & Davies, B. (2002). Mourning and meaning. American Behavioral Scientist, 46, 2357251.
Parkes, C. M. (1998). Recovery from bereavement (3rd ed.). Madison, CT: International
Universities Press.

620

N. P. Field and M. Friedrichs

Raudenbush, S., Bryk, A., Cheong, Y. F., & Congdon, R. (2000). HLM5: Hierarchical
linear and nonlinear modeling. Lincolnwood, IL: Scientific Software International.
Rees,W. D. (1971). The hallucinations of widowhood. British MedicalJournal, 4, 37741.
Shaver, P. R. & Tancredy, C. M. (2001). Emotion, attachment, and bereavement: A
conceptual commentary. In M. S. Stroebe, R. O. Hansson,W. Stroebe, & H. Schut
(Eds.), Handbook of bereavement research: Consequence, coping, and care (pp. 63788).
Washington, DC: American Psychological Association.
Shuchter, S. R. & Zisook, S. (1988).Widowhood: The continuing relationship with the
dead spouse. Bulletin of the Menninger Clinic, 52, 2697279.
Stone, A. A., Schwartz, J. E., Neale, J. M., Shiffman, S., Marco, C. A., Hickcox, M.,
Paty, J., Porter, L. S., & Cruise, L. J. (1998). A comparison of coping assessed by
ecological momentary assessment and retrospective recall. Journal of Personality and
Social Psychology, 74, 167071680.
Stroebe, M., Gergen, M. M., Gergen, K. J., & Streobe, W. (1992). Broken hearts or
broken bonds: Love and death in historical perspective. American Psychologist, 47,
120571212.
Tennen, H., Affleck, G., Armeli, S., & Carney, M. A. (2000). A daily process approach
to coping. American Psychologist, 55, 6267636.
Vormbrock, J. (1993). Attachment theory as applied to war-time and marital separation. Psychological Bulletin, 114, 1227144.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief
measures of positive and negative affect: The PANAS scales. Journal of Personality
and Social Psychology, 54, 106371070.