Citation
Trajectories of self-report well-being before and after a broken engagement

Material Information

Title:
Trajectories of self-report well-being before and after a broken engagement
Creator:
Kenny, Jessica A. ( author )
Language:
English
Physical Description:
1 electronic file (67 pages). : ;

Subjects

Subjects / Keywords:
Betrothal ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Abstract:
Engagement is often thought of as the happiest time in a couple’s life, yet the concept of “cold feet” is also well known. In fact, it has been estimated that 15-20% of engagements are broken every year (Safier, 2003). Despite the prevalence of this important life event, little empirical research has focused on engagement distress and breakup. This study was the first to look at potential changes in individuals’ psychological well-being from an engaged status (Time 1/pre-breakup timepoint) to a broken engagement status (Time 2/post-breakup timepoint). Data from 74 individuals were collected through mailed surveys completed before and after their breakup. Participants were between the ages of 18 and 39, 69% were female, and 72.9% were White. Results of a paired t-test indicate that the overall trajectory of well-being decreased from pre-broken engagement to post-broken engagement (66.2% of sample). However, using a Growth Mixture Model analysis, five overall trajectory groups emerged, including a stable low well-being group (10.8%), an increasing well-being group (23%), and three distinct declining groups with different slopes and intercepts. Of those who experienced a decline in well-being from pre to post broken engagement, only 16.3% experienced a clinically significant change. Overall, 21.6% of the total sample was in a range of possible clinical distress at post-breakup, and half of those participants (10.8%) were already experiencing possible clinical distress at pre-breakup. Characteristics of each trajectory group, limitations of the study, and implications for both clinical work and further research are discussed.
Thesis:
Thesis (M.A.)--University of Colorado Denver.
Bibliography:
Includes bibliographic references
System Details:
System requirements: Adobe Reader.
General Note:
Department of Psychology
Statement of Responsibility:
by Jessica A. Kenny.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
951985235 ( OCLC )
ocn951985235

Downloads

This item has the following downloads:


Full Text
1
TRAJECTORIES OF SELF-REPORT WELL-BEING
BEFORE AND AFTER A BROKEN ENGAGEMENT
by
JESSICA KENNY
B.A., Pepperdine University, 2010
M.A., University of Denver, 2013
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Master of Arts
Clinical Health Psychology Program
2015


This thesis for the Master of Arts degree by
Jessica Kenny
has been approved for the
Clinical Health Psychology program
by
Elizabeth Allen, Chair
Edward Dill
Galena Rhoades
Joan Bihun
Date: December 21, 2015


Kenny, Jessica, M.A.
Trajectories of Self-Report Well-Being Before and After a Broken Engagement
Thesis directed by Associate Professor Elizabeth Allen
m
ABSTRACT
Engagement is often thought of as the happiest time in a couples life, yet the concept
of cold feet is also well known. In fact, it has been estimated that 15-20% of engagements
are broken every year (Safier, 2003). Despite the prevalence of this important life event, little
empirical research has focused on engagement distress and breakup. This study was the first
to look at potential changes in individuals psychological well-being from an engaged status
(Time 1/pre-breakup timepoint) to a broken engagement status (Time 2/post-breakup
timepoint). Data from 74 individuals were collected through mailed surveys completed
before and after their breakup. Participants were between the ages of 18 and 39, 69% were
female, and 72.9% were White. Results of a paired t-test indicate that the overall trajectory of
well-being decreased from pre-broken engagement to post-broken engagement (66.2% of
sample). However, using a Growth Mixture Model analysis, five overall trajectory groups
emerged, including a stable low well-being group (10.8%), an increasing well-being group
(23%), and three distinct declining groups with different slopes and intercepts. Of those who
experienced a decline in well-being from pre to post broken engagement, only 16.3%
experienced a clinically significant change. Overall, 21.6% of the total sample was in a range
of possible clinical distress at post-breakup, and half of those participants (10.8%) were
already experiencing possible clinical distress at pre-breakup. Characteristics of each
trajectory group, limitations of the study, and implications for both clinical work and further
research are discussed.


IV
The form and content of this abstract are approved. I recommend its publication.
Approved: Elizabeth Allen


V
TABLE OF CONTENTS
CHAPTER
I. INTRODUCTION.......................................................1
Hypothesis 1.......................................................7
Hypothesis 2.......................................................7
Hypothesis 3......................................................11
II. METHODS...........................................................15
Procedure.........................................................15
Sample............................................................17
Instruments.......................................................17
Data Analysis.....................................................25
III. RESULTS...........................................................27
Hypothesis 1......................................................29
Hypothesis 2......................................................29
Hypothesis 3......................................................35
Post-Hoc Analyses.................................................44
IV. DISCUSSION........................................................46
Limitations.......................................................49
Implications......................................................50
Future Directions.................................................53
REFERENCES..............................................................56
APPENDIX
59


1
CHAPTERI
INTRODUCTION
Conventional wisdom would say that engagement is supposed to be the happiest time
in a couples life. This may be due to some degree of over-idealization of the partner during
this time in the relationship, as Bonds-Raacke, Bearden, Carriere, Anderson and Nicks
(2001) found that individuals who were engaged reported significantly higher idealistic
distortion scores about their partners than did either married individuals or those who were in
long-term dating relationships. These authors reported that engaged individuals tend to agree
with their partner in an overly positive and enthusiastic manner and reinterpret negative
characteristics in a positive light.
Although there is, on average, a higher level of this type of positivity and idealistic
distortion among engaged couples, there is also variability in the experiences of individuals
in the engagement period, including the well-known concept of cold feet and
accompanying conflict and distress. In a non-scientific survey of engaged women, only 12%
of brides said they felt really happy during their engagement, and 23% were sadder, 24%
were more afraid, and 41% were more anxious than they expected (Moir-Smith, 2006).
Moreover, Lavner et al. (2012) retrospectively asked newly married spouses about their
premarital uncertainties, and responding to a yes/no item, 47% of husbands (n = 108) and
38% of wives (n = 87) reported being uncertain about getting married. More specific patterns
of difference and conflict can be identified through premarital questionnaires that many
engaged couples take in preparation for marriage. Fowers and Olsen (1992) identified four
types of engaged couples by comparing scores on a premarital inventory, which included
subscales across a range of couple issues. Vitalized couples (28%) scored high in


2
compatibility on all dimensions assessed, Present-Oriented Couples (27%) had moderately
positive relationship quality and compatibility but were somewhat unrealistic in their views
of marriage and didnt have consensus on plans for their future, Future-Oriented couples
(23%) had moderately low compatibility scores but scored higher on scales assessing
planning for marriage and future plans and goals, and Conflicted Couples (22%) were
characterized by pervasively low compatibility scores on all scales assessed. Thus, conflict,
distress, and uncertainty do often occur for a notable number of couples in the engagement
period. In fact, it has been estimated that 15-20% of engagements are broken every year
(Safier, 2003).
Whereas we know that a broken engagement is both a major life change and fairly
common, we lack knowledge regarding how a broken engagement might relate to individual
psychological adjustment, and what features of the relationship predict such adjustment.
Thus, I examined both trajectories and predictors of change in well-being for individuals who
have experienced a broken engagement, drawing from the larger Relationship Development
Study (RDS; Rhoades, Stanley, Markman, 2010). RDS is a longitudinal survey study of
1,295 individuals in various stages of relationship formation and commitment. In an
exploratory (unpublished) study of individuals within this sample with a broken engagement,
Kenny, Knopp, Rhoades, Stanley, and Markman (2013) found that having experienced a
broken engagement at some point in the past is associated with higher current psychological
distress and lower life satisfaction than those who have not experienced a broken engagement,
controlling for age. However, these analyses cannot reveal whether this distress and low life
satisfaction actually represent a decrease following the broken engagement, or whether this
distress preceded the broken engagement. There are reasons to believe that both trajectories


3
of stable low well-being or declining low well-being from pre to post-broken engagement
could be true.
Also using the RDS sample, Rhoades, Kamp Dush, Atkins, Stanley, and Markman
(2011) assessed the impact that breakups in general have on mental health and life
satisfaction, and found declines in well-being from pre-breakup to post-breakup. However,
they did not examine this specifically for those who broke up from an engaged relationship.
Rather, they included any non marital breakup, including dating relationships. Rhoades et al.
(2011) modeled living together and having had plans for marriage as moderators of these
declines and found that these variables were significantly associated with larger declines in
well-being following a breakup compared to couples who werent living together and didnt
have plans for marriage. Within this sample, they defined participants who had plans for
marriage as those who answered affirmative to have the two of you together made a
specific commitment to marry? Four hundred and one individuals who experienced a
breakup also reported plans for marriage prior to the breakup. However, only 60 individuals
stated that they were actually engaged prior to breakup. Therefore, there is clearly a
difference between stating that one has plans for marriage and stating that ones relationship
status is engaged, perhaps in terms of increased commitment that comes with being formally
engaged versus having general plans for marriage. This current study isolated and analyzed
only the people whose reported relationship status was engaged when they experienced a
breakup. The findings of Rhoades et al. (2011) suggest that an overall (average) declining
trajectory will be found with this sample as well. That is, this finding supports the idea that,
for many individuals, the overall low well-being found by Kenny et al. (2013) for those who


4
have experienced a break up from an engagement may be the result of a decline in well-being
following a breakup, rather than a static low well-being level prior to break up.
On the other hand, there is research that supports the possibility that, for other
individuals, self-reported low well-being after a breakup may reflect ongoing distress that
was present even before the breakup. Given the lack of empirical research on outcomes of
broken engagements, literature on well-being after both divorce and dating breakups may
help guide hypotheses about engagement breakups. For example, Waite et al. (2003) found
that unhappily married adults who then divorced were no personally happier at follow up
than those who have stayed married. This suggests that, if there was already distress related
to individual or relationship problems prior to breakup, ending that relationship may not
automatically relieve this distress. For some individuals, there may be relatively stable
individual distress problems (e.g., depression), or ending a relationship may trade one set of
problems (e.g., conflict in the relationship) for another set of problems (e.g., isolation or
financial strain). In these scenarios, personal distress would be evident both before and after
breakup. This type of trajectory could characterize a number of individuals who experience a
broken engagement. Thus, although the dominant hypothesized trajectory is a decreasing
level of well-being following a broken engagement, there may be a number of people who
have a stable low trajectory, with low well-being present at both pre-breakup and post-
breakup timepoints.
Of course, for others there may be entirely different courses over time. Whereas I
have discussed potential trajectories of declining well-being and stable low well-being, some
individuals may not show either of these trajectories. Crisis theory proposes that major life
events, such as role transitions and separations from significant others, create obstacles for


5
meeting basic needs and therefore increases the probability of either interpersonal
disturbances or new adaptations and increased functional capacity (Selig, 1976). Thus, crisis
events in a family create the potential for post crisis deteriorating, status quo, or enhanced
functioning. By extension, in addition to the two trajectories already discussed (i.e., a
declining course and a stable low course), there could also be an improving course.
Illustrating these ideas, Veevers (1991) suggests that, whereas the definition of
divorce as a traumatic crisis has led to a focus on the deleterious effects, a review of the
literature indicates that it may also be seen as a strengthening experience. In fact, Bourassa,
Sbarra, and Whisman (2015) found that women in the lowest quality marriages increase in
life satisfaction following a divorce. The notion of divorce as a positive outcome has also
been noted by Markman, Rhoades, Stanley, and Peterson (2013), who analyzed divorce
outcomes of couples who had participated in a premarital relationship education program and
found that the likelihood of divorce increased for couples who had physical aggression in
their relationship. The authors suggest that there may be some situations, such as when there
is aggression in the relationship, in which divorce could be considered a positive outcome.
Unmarried breakups may operate similarly. For example, whereas Rhoades et al. (2011)
found an average decline in life satisfaction after breakup, 16.1% of the sample experienced
an increase in life satisfaction after breakup. Similarly, 19.6% of their participants
experienced a decrease in psychological distress after breakup. Therefore, it is reasonable to
believe that some individuals in my sample may report similar positive outcomes,
particularly when getting out of a potentially bad relationship.
Lastly, there is reason to believe that a stable moderate/high trajectory of well-being
could be possible. This could be explained, in part, by the possibility of an adaptation to a


6
traumatic event as proposed by crisis theory (Selig, 1976). That is, there may be a group of
people within the engaged sample who have moderate/high well-being at both pre-breakup
and post-breakup. While this may seem counterintuitive, Rhoades et al. (2011) found that
higher relationship quality pre-breakup was associated with smaller declines in life
satisfaction following a breakup in an unmarried breakup sample. The authors discussed this
as consistent with findings that happy people endorse greater adjustment throughout
various life challenges. Therefore, there may be a subset of individuals in this broken
engagement sample who show a similar stable moderate or high well-being trajectory over
time.
Overall, multiple individual trajectories of well-being are possible and may be
distinguishable from pre to post-breakup. In fact, Rhoades et al. (2011) provided percentages
of individuals (of the 401 analyzed) who fell into three distinct groups before and after an
unmarried breakup: decreasing psychological distress (19.6%), increasing psychological
distress (30.7%), and stable psychological distress (49.7%). Because the percentages
provided by Rhoades et al. (2011) were not isolated for engaged (or planning for marriage)
participants, a separate analysis would be needed to identify these types of rates for broken
engagement. A separate analysis of only engaged individuals is warranted, considering this is
a critical period in a persons life, the sparse empirical data available on the subject of broken
engagements, and the apparent distinction between engagement and planning for marriage.
Therefore, the first aim of this study is to determine if the average overall trajectory
of well-being for pre-breakup to post-breakup increases or decreases from pre broken
engagement to post broken engagement. The second aim of this study is to evaluate whether,
and how many, individuals fall into distinct trajectories of well-being before and after broken


engagement. The third aim is to assess what characterizes individuals in the different
trajectories on a range of relationship factors.
7
Aim 1: To determine if the average overall trajectory of well-being decreases from pre to
post breakup for individuals who were engaged pre-breakup.
Hypothesis 1
As suggested by Rhoades et al. (2011), there will be a significant decline in level of
well-being from pre-breakup to post-breakup for people who have experienced a broken
engagement.
Aim 2: To evaluate whether, and how many, individuals fall into distinct trajectories within
this sample.
Hypothesis 2
As suggested by crisis theory, in addition to the average declining trajectory of well-
being (i.e., that may characterize the modal slope) posited in the first hypothesis, three
additional distinguishable trajectories of well-being from pre-breakup to post-breakup
(steady moderate/high well-being, steady low well-being, improving well-being) will emerge.
See Figure 1.


8
Level of
well-being
Blue = Declining well-being
Red = Steady moderate/high well-being
Green = Steady low well-being
Purple = Improving well-being
Figure 1. Hypothesized trajectories of well-being from pre to post broken engagement
Aim 3: To assess what characterizes individuals in the different trajectories of self-
report well-being before and after a broken engagement.
Although empirical research on distinguishing characteristics of different well-being
trajectories following a broken engagement is sparse, there are several reasons to believe that
certain characteristics about the relationship will have an influence on the trajectory of ones
well-being following a broken engagement. As noted above, Rhoades et al. (2011) conducted
a similar study with a broader unmarried relationship sample and found that cohabitating and
having had plans for marriage were associated with larger declines in life satisfaction from
pre-breakup to post-breakup whereas having higher relationship quality at pre-breakup was
associated with smaller declines in life satisfaction following a breakup. In addition, having
begun to date someone new was associated with smaller declines from pre-breakup to post-
breakup. Whereas Rhoades et al. (2011) analyzed several important variables related to a


broad, unmarried sample, I focused on variables that may be particularly relevant to an
engaged sample.
9
A potential important moderating factor for pre to post broken engagement well-
being may be approval of friends and family. Elison (2011) found that relationship
progression to marriage is best predicted by several factors, including approval of
relationship by friends and family. If friends and family approval of the relationship is
typically a marker of engagement making it to marriage, in the instance of a broken
engagement, it may be more of a violation of ones expectancies and contribute to the
breakup being more traumatic or distressing. Thus, it may be that individuals who have high
friends and family approval at pre-breakup would experience a greater decrease in well-being
following the breakup. On the other hand, if the approval is low at pre-breakup, this may be
associated with a less distressing breakup and higher level of well-being following the
breakup.
Another possible moderating factor may be the presence of humiliating relationship
events, defined as events that devalue the individual. Cano and OLearys (2000) research on
humiliating marital events found that, controlling for marital discord, women who
experienced a humiliating marital event (husband infidelity, husband initiation of separation
or divorce, separation or divorce due to husband infidelity or violence) were six times more
likely to be diagnosed with a major depressive episode than those who had not experienced a
humiliating marital event. Thus, if certain humiliating events occur within a relationship,
they may intensify a drop in well-being even if the breakup is a positive experience for them
in the long term. Thus, it may be that infidelity, physical aggression, or partner desired
breakup may be more characteristic of the declining or stable-low trajectory groups, whereas


10
groups with steady moderate/high levels or improving trajectories of well-being may be less
likely to report aggression, infidelity, or that the partner desired the breakup. Although
Rhoades et al. (2011) found that a desire to breakup (from the participants point of view)
was not significantly related to changes in life satisfaction from pre to post breakup, this
study will specifically look at the participants partners desire to breakup in relation to lower
or higher well-being of the participant. If individuals who experience physical aggression,
partner infidelity, or a breakup initiated by the partner also experience a decrease in well-
being following the breakup, this may mirror the work of Cano and OLeary (2000).
Thus, there are multiple possible moderators of change from pre to post breakup. As
noted above, a third aim is to assess what characterizes individuals in the different
trajectories of self-report well-being that emerge before and after a broken engagement. The
exact execution of this aim was conditional on the results of the second aim. That is,
assuming I found adequate population for different trajectories of well-being, then the third
aim is to analyze how the groups representing the distinct trajectories differ on the following
variables: pre-breakup levels of friends and family approval of the relationship (social
pressure), desire to marry their partner, felt constraint, dedication, relationship satisfaction,
perceived likelihood of marriage, presence of a set marriage date, reports of physical and
psychological aggression from the partner (towards the participant), reports of infidelity of
the partner, and post-breakup levels of perception of partner relative (to self) desire for the
breakup, and reports of infidelity by the participants partner. Whereas Rhoades et al. (2011)
previously analyzed two similar moderators that I propose to look at (relationship quality,
partner initiated breakup), this study will attempt to understand if those moderators operate in
a similar way in an engaged sample as opposed to a broad unmarried sample. The remainder


11
of my moderators are distinct from those that Rhoades et al. (2011) used, and will provide
further understanding of the unique relationship stage of a broken engagement.
Hypothesis 3
The four groups representing individuals in the four distinguished trajectories of well-
being following a broken engagement will show significant differences in certain variables.
Not all measured variables will be relevant for all groups. Using the relevant variables, the
variables that may differentiate the four hypothesized trajectory groups are described below.
See Table 1.
1. (Red line) Stable moderate/high well-being: This is a group of individuals who report
moderate to high well-being at both pre-break and post-breakup. It is hypothesized that
they may be conceptually characterized in the following way: Although there were no
obvious markers for distress in their relationship, they may also have not been entirely
invested in the relationship or in the upcoming marriage. Therefore, after breakup, they
continue to experience moderate to high well-being. Specifically, before breakup, they
are hypothesized to be relatively (compared to the other groups): (1) low in felt constraint,
(2) low in dedication, (3) high in relationship satisfaction, (4) low in perceived likelihood
of marriage, and (5) low in psychological aggression from their partner. Moreover, they
are posited to be more likely to be unsure of their desire to marry their partner, less likely
to have a marriage date set before breakup, less likely to report that their partner desired
the breakup, less likely to report physical aggression towards them in the relationship,
and less likely to report partner infidelity before or after breakup. As mentioned above,
high relationship satisfaction may be counter-intuitive, but Rhoades et al. (2011) found
that higher relationship quality pre-breakup was associated with smaller declines in life


12
satisfaction following a breakup. They discussed this as consistent with findings that
happy people endorse greater adjustment throughout various life challenges.
2. (Blue line) Moderate to high well-being before, low well-being after: This is a group of
individuals who report moderate to high well-being at pre-breakup but decline after
breakup. It is hypothesized that they may be conceptually characterized in the following
way: These individuals were relatively happy in their relationship, and they were also
entirely invested in their relationship and upcoming marriage. Therefore, after breakup,
their well-being decreased after an unexpected breakup or accompanying events.
Specifically, before breakup, they are hypothesized to be relatively (compared to the
other groups): (1) high in friends and family approval of relationship/partner (social
pressure), (2) low in felt constraint, (3) high in dedication, (4) high in relationship
satisfaction, (5) high in perceived likelihood of marriage, and (6) low in psychological
aggression from their partner. Moreover, they are posited to be more likely to be sure of
their desire to marry their partner, more likely to have a marriage date set before breakup,
more likely to report that their partner desired the breakup, less likely to report physical
aggression towards them in the relationship before breakup, less likely to report partner
infidelity before breakup, and more likely to report partner infidelity at the post-breakup
timepoint.
3. (Purple line) Low well-being before, moderate to high well-being after: This is a group of
individuals who report low well-being pre-breakup and higher well-being post breakup. It
is hypothesized that they may be conceptually characterized in the following way: There
were markers for distress in their relationship and they werent entirely invested in the
relationship or the upcoming marriage. Therefore, after breakup, they experienced an


13
increase in well-being as they got out of a presumably bad situation. However, they may
not have experienced humiliating or de-valuing (see Cano & OLeary, 2000) that would
contribute to a potentially chronic low well-being. Specifically, before breakup, they are
hypothesized to be relatively (compared to the other groups): (1) low in friends and
family approval of relationship/partner (social pressure), (2) high in felt constraint, (3)
low in dedication, (4) low in relationship satisfaction, and (5) low in psychological
aggression from their partner. Moreover, they are posited to be more likely to be unsure
of their desire to marry their partner, more likely to have a marriage date set before
breakup, less likely to report that their partner desired the breakup, less likely to report
physical aggression towards them in the relationship before breakup, and less likely to
report partner infidelity in the relationship before or after breakup.
4. (Green line) Stable low well-being: This is a group of individuals who report low well-
being at both pre- and post breakup. It is hypothesized that they may be conceptually
characterized in the following way: There were markers for distress in their relationship
and they also werent entirely invested or committed to their upcoming marriage.
However, although they avoided a presumably bad situation, they didnt experience an
increase in well-being, possibly because they still have other distressing factors in their
life that are contributing to their low well-being, and/or because they experienced
humiliating or devaluing events (see Cano & OLeary, 2000) that contributed to a more
chronic low well-being. Specifically, before breakup, they are hypothesized to be
relatively (compared to the other groups): (1) high in felt constraint, (2) low in dedication,
(3) low in relationship satisfaction, and (4) high in psychological aggression from their
partner. Moreover, they are posited to be more likely to have no marriage date set before


14
breakup, more likely to report that their partner desired the breakup, more likely to report
physical aggression towards them in the relationship before breakup, and more likely to
report partner infidelity before and after breakup.
Table 1. Hypothesized distinguishing varia jles of well-being trajectories
Measures #1 Red: Stable moderate/ high well-being #2 Blue: High to low well- being #3 Purple: Low to high well- being #4 Green: Stable low well- being
Friends/family approval (social pressure) atpre breakup N/A High Low N/A
Desire to marry partner at pre breakup Unsure Sure Unsure N/A
Felt constraint at pre breakup Low Low High High
Dedication at pre breakup Low High Low Low
Relationship satisfaction at pre breakup High High Low Low
Perceived likelihood of marriage at pre breakup Low High N/A N/A
Marriage date set at pre breakup No Yes Yes No
Partner desired the breakup at post breakup No Yes No Yes
Physical aggression (from partner) at pre breakup No No No Yes
Psychological aggression (from partner) at pre breakup Low Low Low High
Partner infidelity at pre breakup Low Low Low High
Partner infidelity at post breakup Low High Low High
*Note: For each row, the hypothesis refers to group comparisons. For example, for the friends and family
approval variable: group #2 is hypothesized to have higher friends and family approval than people in group
#3. For the desire to marry partner variable: a greater proportion of people in group #2 are hypothesized to be
100% sure than of people in either group #1 or group #3. For the marriage date set: a greater proportion of
people in group #2 and group #3 are hypothesized to have set a marriage date than people in either group #1
or group #4.


15
Thus, the overall goal of all three aims is to: (a) evaluate if there is a statistically
significant change between the overall pre breakup and post breakup levels of well-being (b)
see if there are distinct well-being trajectory groups that emerge, and (c) understand how
these groups differ on certain pre-breakup and/or post-breakup variables.


16
CHAPTER II
METHODS
Procedure
Participants in the current study (N =74) were participants who are part of a larger
longitudinal study (N = 1,295) that completed 12 time points of surveys over the course of
four to five years (Rhoades, Stanley, Markman, 2010). To recruit participants for the larger
parent study on relationship development, a calling center used a targeted listed telephone
sampling strategy to call households within the contiguous United States. After a brief
introduction to the study, respondents were screened for participation. To qualify,
respondents needed to be between 18 and 34 years of age and be in an unmarried relationship
with a member of the opposite sex that had lasted 2 months or longer. The study was
designed to examine ways that premarital experiences impacted marriage. When the study
started, same-sex marriage wasnt widely available and therefore, only those in opposite-sex
relationships were included. Those who qualified, agreed to participate, and provided
complete mailing addresses (N = 2,213) were mailed forms within 2 weeks of their phone
screening. Of those who were mailed forms, 1,447 individuals returned them (65.4%
response rate); however, 152 of these survey respondents indicated on their forms that they
did not meet requirements for participation, either because of age or relationship status,
leaving a sample of 1295 for the first wave (Tl) of data collection. These 1295 individuals
were mailed the second wave (T2) of the survey about four months after returning their Tl
surveys. Participants received a survey every four months for the 12 waves.
At each timepoint, participants were asked if they have ever been engaged to
someone who they did not end up marrying. Of the overall sample, 22% of individuals


17
indicated that this was true for them (N = 291). Of the sample of 291, participants for whom
there was a survey while they were engaged and before their breakup, and another survey
post-breakup were included (N = 75). The exact pre and post timepoints are different for each
participant.
Of the 75 people who had experienced a broken engagement over the course of the
study, there were 5 people who had two or more broken engagements over the course of the
study (sometimes with the same partner). For these participants with multiple broken
engagements, I used the pre and post timepoints of the first broken engagement that they
experienced during the study. Each participant had a pre and post timepoint that was
sequential (i.e., no skipped timepoints in between), except for one person. This person was
engaged at T9, didnt complete a T10 packet, and was broken up at T11.1 initially sought to
include this person in analyses. However, after running initial Growth Mixture Model
analyses within Mplus, I was consistently getting an outlier in the form of one person in their
own trajectory class. After matching the ID number to the outlier case, I discovered that the
outlier was this participant. Mplus results indicated that this participant was statistically
different from the rest of the sample and thus this outlier affected the trajectory classification.
When I removed this person from analyses and ran the analyses with 74 people, the
trajectories were much more evenly distributed. Thus, this person was excluded as an outlier.
Therefore, the final sample for this study was 74 people who were engaged at one timepoint
and broken up with that same partner at the subsequent timepoint.
Sample
Participants were aged 18-39, with an average age of 27.51 (SD = 4.66). Sixty-nine
percent were female; 72.9% were White, 16.2% Black or African American, 6.8% Hispanic


18
or Latino, 1.4% Asian, and 2.7% indicated more than one race. Of the 74 participants, 48.0%
had completed at least one year of college. Approximately 75% of this group was employed.
The median yearly income, for those who were employed, was about $30,000. 22.9% of the
sample reported that they were dating someone new at their post-breakup timepoint, and
4.1% of the sample reported that they were still dating the person who they had just broken
their engagement with.
Instruments
Measures included in original hypotheses:
Well-being: To measure well-being, I used 12 items from the longer Mood and
Anxiety Symptom Questionnaire (MASQ; Clark & Watson, 1991), which yields a
psychological distress score. I then reverse scored this measure to reflect less distress,
thereby becoming a marker of psychological adjustment or well-being (i.e., higher scores
indicate higher well-being). This was assessed at each participants pre-breakup and post-
breakup timepoints. Example items include, During the last week, I felt dissatisfied with
everything, During the last week, I felt tense or high strung and During the last week I
felt discouraged. Items are rated on a 5-point scale, ranging from 1 (not at all) to 5
(extremely). Scores are the average of the 12 items. The 12 items were chosen to be used in
the larger parent study from the overall MASQ based on factor analyses that indicate that
they capture a more general psychological distress rather than symptoms specific to anxiety
or depressive disorders (Keogh & Reidy, 2000; Rhoades, Stanley, & Markman, 2011). The
12-item measure has been shown to be reliable in measuring psychological distress in similar
samples (a = .94), demonstrating excellent internal consistency reliability (Wortel & Rogge,


19
2010). This measure has also shown substantial convergent validity with self-report symptom
measures of anxious and depressive symptomatology (Clark & Watson, 1991).
Friends andfamily approval of the relationship (social pressure): To assess friends
and family approval of the relationship, or social pressure from friends and family for the
relationship to endure, the 6-item social pressure subscale from the Commitment Inventory
was used (Stanley & Markman, 1992). This was assessed at each participants pre-breakup
timepoint. Example items include, My friends would not mind it if my partner and I broke
up (or divorced) (reverse scored), My family would not care either way if this relationship
ended (reverse scored), It would be difficult for my friends to accept it if I ended the
relationship with my partner, and My family really wants this relationship to work. Each
item is rated on a 7-point scale, ranging from 1 (strongly disagree) to 7 (strongly agree). This
measure has established excellent internal consistency reliability by demonstrating alphas
of .92 and .88 in similar samples (Stanley & Markman, 1992). This measure has
demonstrated validity by convergence with theoretically related variables, such as a
correlation of .63 (p < .01) with a one-item commitment measure and a correlation of .39 (p
< .01) with a one-item relationship satisfaction measure in similar samples (Stanley &
Markman, 1992). This measure is scored by reverse scoring the items that reflect family and
friends not caring about or approving of the relationship, and then averaging the items.
Higher scores indicate both higher approval and higher social pressure from friends and
family for the relationship to endure.
Desire to marry ones partner: A categorical item, Do you want to marry your
current partner? was used to assess desire to marry ones partner at each participants pre-
breakup timepoint. Participants indicated their response on a 4-point scale from 1 (Yes, I am


20
sure I want to marry my partner), 2 (Not sure), 3 (No, I do not want to marry my partner),
and 4 (I havent thought about it). This scale will be scored as participants being sure they
want to marry their partner (1), or unsure they want to marry their partner (2, 3, or 4).
Felt constraint: Felt constraint was measured using three items: I feel trapped in this
relationship but I stay because I have too much to lose if I leave, I would leave my partner
if it was not so difficult to do and I feel stuck in this relationship at each participants pre-
breakup timepoint. Each item is rated on a response scale from 1 (strongly disagree) to 7
(strongly agree). Rhoades et al. (2010) found this measure to have good internal consistency
reliability in the parent study (a = .82). This measure has demonstrated validity through
divergence with theoretically non-related measures, such as dedication, and convergence
with theoretically related measures, such as relationship adjustment and stability (Rhoades et
al., 2010). Higher dedication is associated with lower felt constraint (Rhoades et al., 2010).
This measure is scored by averaging the items. Higher scores indicate a greater felt constraint.
Dedication: Dedication was measured using the 4-item dedication subscale from the
Commitment Inventory (Stanley & Markman, 1992). This was assessed at each participants
pre-breakup timepoint. Example items include, I want this relationship to stay strong no
matter what rough times we encounter, I like to think of my partner and me more in terms
of us and we than me and him/her, It makes me feel good to sacrifice for my
partner and My relationship with my partner is clearly part of my future life plans. Each
item was rated on a 1 (strongly disagree) to 7 (strongly agree) scale. Rhoades et al. (2010)
found the measure to be internally consistent in the parent study (a = .88) and Rhoades et al.
(2012) found similar internal consistencies for men (a = .86) and women (a = .87). The
dedication scale has also demonstrated validity through convergence with theoretically


21
related characteristics, such as commitment and relationship satisfaction (Stanley &
Markman, 1992; Owen et al., 2011). This measure is scored by averaging the items. Higher
scores are indicative of more dedication in the relationship.
Relationship satisfaction: The 4-item version of the Dyadic Adjustment Scale was
used to measure relationship satisfaction (Sabourin, Valois, & Lussier, 2005). This was
assessed at each participants pre-breakup timepoint. This scale was shortened from the 32-
item Dyadic Adjustment Scale (Spanier, 1976). This measure includes items about thoughts
about dissolution, a general item about how well the relationship is going, frequency of
confiding in one another, and the degree of happiness of the relationship. Response options
vary across the four items. Rhoades et al. (2010; 2011) found an alpha of .81 using this same
version in the parent study, demonstrating internal consistency reliability. The DAS has been
shown to correlate well with other marital relationship tests such as a .88 correlation with the
Locke-Wallace Marital Adjustment Scale (Graham, Liu, & Jeziorski, 2006). Further, the
DAS shows predictive and concurrent validity by predicting likelihood of divorce and
consistently distinguishing distressed couples from nondistressed couples (Graham et al.,
2006). As suggested by the authors, the scale is scored by summing the items (Sabourin et al.
2005). Higher scores indicate higher relationship quality.
Perceived likelihood of marriage: Perceived likelihood of marriage was assessed
using a one item measure, How likely is it that you and your partner will get married? at
each participants pre-breakup timepoint. Participants are asked to indicate their responses on
a 5-point Likert scale from Very unlikely to Very likely. This item is based off an item
used in the National Survey of Families and Households. Test retest reliability has been
shown to be high in research with the parent sample (Rhoades et al., 2010; 2012). Higher


22
scores indicate a greater perceived likelihood of marriage.
Marriage date set: To assess whether the participants have set a marriage date or not,
one item was given, Have you and your partner set a date for getting married? and was
assessed at each participants pre-breakup timepoint. The options are Yes, and it is____
(mm/dd/yy) or No.
Partner desired breakup: To assess whether the partner initiated the breakup, one
item was given, Who wanted to end the relationship more? at each persons post-breakup
timepoint. Participants are asked to indicate their responses on a 7-point scale from 1 (Me) to
7 (My partner) with 4 being labeled as Equal. This item was scored dichotomously, such
that anyone who answers 5 or above was coded as 1 and anyone who answers below 5 was
coded as 0. Therefore, a code of 1 means that the partner wanted to breakup more than the
participant, and a code of 0 means that the partner did not want to breakup more than the
participant (i.e., the participant wanted to breakup more than or equal to their partner).
Physical aggression from partner: To measure physical aggression in a relationship
towards the participant (from the partner), two subscales from the Revised Conflict Tactics
Scale were used; the minor injuries received subscale and the minor physical aggression by
partner subscale (Straus, Hamby, Boney-McCoy, & Sugarman, 1996). This was assessed at
each participants pre-breakup timepoint. The minor injuries received subscale (e.g., I had a
sprain, bruise, or small cut because of a fight with my partner) and the minor physical
aggression by partner (e.g., My partner pushed or shoved me were used to create a scale of
20 items for participants to indicate if this never happened (0) or to indicate (on a scale of 1-
7) the frequency of the action, ranging from not in the past year, but it did happen before to
more than 20 times in the past year. Straus et al. (1996) found the minor injuries scale to


23
have an alpha of .95 and the minor physical aggression by partner scale to have an alpha of
.86 in a similar sample, indicating good internal consistency reliability. The scales also
demonstrated good validity by convergence with theoretically non-related constructs, such as
negotiation. The higher the level of negotiation in a couple, the less they report injuries or
aggression (Straus et al., 1996). Participants who indicated that they had never sustained
injuries due to a fight with their partner and had never been the recipient of physical
aggression from their partner were coded as having no history of physical aggression towards
them in the relationship. These participants were coded as 0 for this scale. Participants who
reported any of these behaviors were coded as having a history of physical aggression
towards them at some point in their relationship. These participants were coded as 1 for
this scale.
Psychological aggression from partner: To measure psychological aggression
towards the participant (from the partner) in a relationship, the minor psychological
aggression received subscale from the Revised Conflict Tactics Scale was used (Straus,
Hamby, Boney-McCoy, & Sugarman, 1996). This was assessed at each participants pre-
breakup timepoint. The minor psychological aggression received subscale includes four
items: My partner insulted or swore at me, My partner shouted or yelled at me, My
partner stomped out of the room or house or yard during a disagreement, and My partner
said something to spite me. Participants are asked to indicate if this never happened (0) or to
indicate (on a scale of 1-7) the frequency of the action, ranging from not in the past year, but
it did happen before to more than 20 times in the past year. Straus et al. (1996) found the
minor psychological aggression scale to have an alpha of .79 in a similar sample, indicating
sufficient internal consistency reliability. In another study using the parent sample, the as for


24
the psychological aggression subscale were .77 and .82 for men and women, respectively
(Rhoades, Stanley, Markman, 2009). This measure is scored by averaging the items. Higher
scores indicate a greater prevalence of psychological aggression received.
Partner infidelity: Infidelity by the participants partner was assessed using one item
at each participants pre-breakup and post-breakup timepoints. The pre-breakup item is, Has
your partner had sexual relations with someone other than you since you began seriously
dating? The post-breakup item is, Did your ex-partner have sexual relations with someone
other than you while you were dating? The response options are No, Probably not,
Probably so and Yes, I know for sure. Higher scores indicate a greater certainty that the
participants partner was unfaithful at either pre-breakup or post-breakup timepoints.
Additional measures used for descriptive purposes:
Length of engagement: The time in days of the length of the engagement is a
calculated variable. This variable was created by subtracting the reported breakup date from
the date of the engagement. There was only enough data to calculate this for 57 (77%) of the
participants. When this variable is used with only participants who experienced a drop in
well-being (see post hoc section, below), this percentage decreases slightly (76%). The initial
range was 25 3737 days, with a mean of 591.80 days (SE = 85.55) and a median of 363
days. Due to the high level of skew, two extreme values (greater than or equal to 2,553 days)
were re-coded into the next highest value (1,460 days), to represent an outer value of 1,460
days or greater. This recoding dropped the skew to an acceptable level. The median for the
recoded value remained at 363 days and the mean became 536.74 (SE = 56.07).


25
Time since breakup: The time in days since the broken engagement is a calculated
variable. This variable was created by subtracting the reported breakup date from the date of
the post survey. Due to missing dates on some survey packets, this variable could only be
calculated for 93% of the sample. When used with only participants who experienced a drop
in well-being (see post hoc section, below), this percentage decreases slightly (92%).
Reasons for breakup: To ascertain a participants reason(s) for the breakup,
respondents were asked Why did the relationship end? (Fill in all that apply). The options
that participants could choose to endorse were, (1) substance abuse issues, (2) religious
differences, (3) domestic violence, (4) economic hardship, (5) infidelity, (6) too much
arguing, (7) lack of commitment, and (8) sexual incompatibility. This was given at each
participants post-breakup timepoint. Percentages of each reason endorsed will be given for
each trajectory of well-being as a way to further identify these groups. The results of this
variable are displayed in Table 5.
Data Analysis
The three aims of my thesis were to: (a) determine the average, overall trajectory of
self-report well-being from pre-broken engagement to post-broken engagement, (b) see if
there are distinct well-being trajectory groups that emerge, and (c) understand how these
groups differ on certain pre-breakup and/or post-breakup timepoint variables.
I tested my first hypothesis with a simple paired t-test. I addressed my second aim
(and provided further evidence regarding the first aim) with Mplus using Growth Mixture
Modeling (GMM). Given that I was able to proceed to Aim 3,1 addressed this aim using
either an ANOVA or Chi-Square analysis for each hypothesized distinguishing variable (see
Table 2, below).


26
Table 2. Hypothesized analyses for distinguishing variables of well-being trajectories
Measures #1 Red/ Stable high well- being #2 Blue / High to low well- being #3 Purple / Low to high well- being #4 Green / Stable low well-being Analyses for each hypothesis:
Friends/famil y approval (social pressure) at pre breakup N/A High Low N/A ANOVA: group 2 vs. group 3
Desire to marry partner at pre breakup Unsure Sure Unsure N/A Chi Square: 2 options (sure, unsure) x 3 groups
Felt constraint at pre breakup Low Low High High ANOVA: post hoc groups 3 and 4 higher than groups 1 and 2
Dedication at pre breakup Low High Low Low ANOVA: post hoc group 2 higher than groups 1,3, and 4
Relationship satisfaction at pre breakup High High Low Low ANOVA: post hoc groups 1 and 2 higher than groups 3 and 4
Perceived likelihood of marriage at pre breakup Low High N/A N/A ANOVA: group 2 higher than group 1
Marriage date set at pre breakup No Yes Yes No Chi Square: 2 options (Yes, No) x 4 groups
Partner desired the breakup at post breakup No Yes No Yes Chi Square: 2 options (Yes, No) x 4 groups
Physical aggression (from partner) at pre breakup No No No Yes Chi Square: 2 options (Yes, No) x 4 groups


27
Measures #1 Red/ Stable high well- being #2 Blue / High to low well- being #3 Purple / Low to high well- being #4 Green / Stable low well-being Analyses for each hypothesis:
Psychological aggression (from partner) atpre breakup Low Low Low High ANOVA: post hoc group 4 higher than groups 1, 2, and 3
Partner infidelity at pre breakup Low Low Low High ANOVA: post hoc group 4 higher than groups 1, 2, and 3
Partner infidelity at post breakup Low High Low High ANOVA: post hoc groups 2 and 4 higher than groups 1 and 3


28
CHAPTER III
RESULTS
Prior to hypothesis testing, I analyzed basic descriptive and reliability statistics for all
possible distinguishing variables (see Table 3). Results indicate that all internal consistencies
were acceptable, when relevant.
Table 3, Initial overall results for possible distinguishing variables
Mean (SD) Percentages Reliability (a)
Well-being atpre breakup (N=74) 3.69 (.95) N/A .92
Well-being at post breakup (N=74) 3.19(1.16) N/A .95
Friends/family approval (social pressure) at pre breakup (N=73) 4.26(1.43) N/A .82
Desire to marry partner at pre breakup (N=74) N/A 1.4% No I do not want to marry my partner 79.7% Yes I am sure I want to marry my partner 18.9% -1 am not sure if I want to marry my partner N/A
Felt constraint at pre breakup (N=74) 2.20(1.45) N/A .88
Dedication at pre breakup (N=74) 5.60 (1.33) N/A .86
Relationship satisfaction at pre breakup (N=71) 15.39 (3.93) N/A .80
Perceived likelihood of marriage at pre breakup (N=73) N/A 1.4% Very unlikely 13.7% Neutral 20.5% Likely 64.4% Very likely N/A
Marriage date set at pre breakup (N=74) N/A 60.8% Marriage date set 39.2% No marriage date set N/A


29
Mean (SD) Percentages Reliability (a)
Partner desired breakup at post breakup (N=71) N/A 67.6% Participant wanted the breakup more 32.4% Partner wanted the breakup more N/A
Physical aggression (from partner) at pre breakup (N=73) N/A 49.3% No history of physical aggression 50.7% History of some physical aggression .84
Psychological aggression (from partner) at pre breakup (N=72) 2.57(1.89) N/A .79
Partner infidelity at pre breakup (N=74) N/A 55.4%-No 13.5% Probably not 9.5% Probably so 21.6% Yes I know for sure N/A
Partner infidelity at post breakup (N=70) N/A 32.9%-No 11.4% Probably not 22.9% Probably so 32.0% Yes I know for sure N/A
Time since breakup in days (N=69) 72.26 (57.55) Median: 61.00 Range: 1-230 N/A N/A
Length of engagement in days (recoded) (N= 57) 536.74 (423.32) Median: 363.00 Range: 25 1460 N/A N/A
Next, I ran correlations among key variables (see Table 4, below). Variables appear to
converge as expected (e.g., dedication is highly correlated with relationship satisfaction),
with a few exceptions (e.g., partner infidelity is unrelated to relationship satisfaction at pre
breakup).
Table 4. Correlation matrix with all relevant variables
1. 2. 4. 5. 6. 7. 8.
1. Partner infidelity pre
2. Partner infidelity post .58**
3. Relationship satisfaction pre -.08 .23


30
1. 2. 3. 4. 5. 6. 7. 8.
4. Felt constraint pre -.05 -.25* _ 54**
5. Likelihood of marriage pre -.22 .08 .66** . 50**
6. Dedication pre .02 .20 7Q** -.75** .51**
7. Psychological aggression (from partner) pre .28* .00 _ 38** .28* -.27* -.30*
8. Friends/family approval (social pressure) pre -.21 .06 49** 24** .61** .45** 24**
9. Length of engagement (recoded) .08 -.04 -.25 .31* _ 41 ** -.06 -.12 -.18
Hypothesis 1
My first hypothesis was that the average overall trajectory of well-being, as measured
by psychological adjustment, for pre-breakup and post-breakup, within people who have
experienced a broken engagement, will show a decrease from before the broken engagement
to after the broken engagement. I tested this hypothesis with a paired t-test. Results of the
paired t-test confirmed that there is a significant drop between pre-broken engagement levels
of well-being and post-broken engagement levels of well-being (t 1,73 = 4.85, p < .001).
Hypothesis 2
My second hypothesis was that in addition to the average declining trajectory of well-
being, three additional distinguishable trajectories of well-being from pre-breakup to post-
breakup (steady moderate/high well-being, steady low well-being, improving well-being)
will emerge within the sample.


31
GMM is a method for identifying both the overall trajectory and unobserved
subgroups. Although I hypothesized four distinct trajectories, I chose GMM due to this
models ability to describe possible latent trajectories via longitudinal change in a post-hoc
manner. That is, GMM allows latent trajectories to emerge as opposed to identifying the
trajectories myself. Thus, I used exploratory GMM models with the amount of classes set at
3, 4, 5, and 6 to ensure I chose the number of trajectories that best fit the data. I used Mplus
version 7.3. Between 200 and 1000 random starts were used, the best log-likelihood value
was replicated, and the model estimation terminated normally for the models run with
number of classes set at 3, 4 or 5. For analyses with the number of classes set at 6, the model
did not terminate normally and the best log-likelihood value was not replicated (see below).
The model fit results of each model are as follows:
3 classes: entropy = .81, log-likelihood value = -181.96, AIC = 385.91, BIC = 411.26,
sample-size adjusted BIC = 376.59.
4 classes: entropy = .82, log-likelihood value = -177.45, AIC = 382.90, BIC = 415.16,
sample-size adjusted BIC = 371.04.
5 classes: entropy = .84, log-likelihood value = -173.35, AIC = 380.71, BIC = 419.88,
sample-size adjusted BIC = 366.30.
6 classes: This model did not terminate normally due to misfit between parameters and
degrees of freedom. The best log-likelihood value was not replicated. Thus, this model is not
plausible in the data.
When a Chi-Square difference test is not relevant to decide on best model fit, as is the
case with GMM, using a combination of statistical information such as Akaikes Information


32
Criteria (AIC; Akaike, 1987) and Bayesian Information Criteria (BIC; Schwartz, 1978), as
well as agreement with substantive theory, is the best approach (Nylund, Asparouhow &
Muthen, 2007). The BIC and AIC have meaning only in comparison to other BIC and AIC
values; the smaller the numbers, the better the model fit.
Entropy is another indication of model fit. Given an estimated model, estimated
posterior probabilities for each individual and each class are produced. Individuals are
classified into the class with the highest probability. The classification quality is summarized
into the class with the highest probability, and this classification quality is summarized in an
entropy value ranging from 0-1 with 1 corresponding to the situation where all individuals
have 100% probability for one class and 0% probability for the other classes (Muthen, Brown,
Leuchter & Hunter, 2010). In other words, entropy is a measure of confidence in class
assignment (Grimm et al., 2010). Thus, the closer to 1 the entropy value is, the better.
Given that there is no real a priori theory for trajectories of well-being following a
broken engagement, this criterion became less relevant for me. According to the AIC values,
sample-size adjusted BIC values, and entropy values, the results show that 5 classes fit the
data better than 3, 4 or 6 classes.
Thus, my hypotheses regarding the four specific trajectories from pre-breakup to
post-breakup (declining well-being, steady moderate/high well-being, steady low well-being,
improving well-being) was not supported exactly as predicted. However, different
trajectories did emerge. The 5 trajectories, or classes, are summarized below. By specifying
the estimator function as MLR within MPlus, a maximum likelihood estimator with robust
standard errors, using a numerical integration algorithm, was used for missing data. That is,
MPlus used all data that were available to estimate the model using full information


33
maximum likelihood. Thus, each parameter is estimated directly without first filling in
missing data values for each individual.
The following class information is what this model yields (recall thatMASQ was
reverse scored to reflect well-being; p value associated with slope):
Group (class) 1: Initial moderate well-being, significant drop = 8 people. Average pre-
breakup well-being score = 4.32 (SD = .51), Average post-breakup well-being score = 2.13
(SD = .63), Intercept = 4.32; Slope = -2.19, p < .001.
Group (class) 2: Initial moderate well-being, significant increase =17 people. Average pre-
breakup well-being score = 3.87 (SD = .52), Average post-breakup well-being score = 4.35
(SD = .41), Intercept = 3.87; Slope = .48, p < .001.
Group (class) 3: Initial moderate well-being, significant drop = 16 people. Average pre-
breakup well-being score = 3.19 (SD = .68), Average post-breakup well-being score = 2.63
(SD = .41), Intercept = 3.19; Slope = -.56, p < .001.
Group (class) 4: Initial moderate well-being, significant drop = 25 people. Average pre-
breakup well-being score = 4.40 (SD = .64), Average post-breakup well-being score = 3.70
(SD = .72), Intercept = 4.40; Slope = -.70, p < .001.
Group (class) 5: Stable low well-being = 8 people. Average pre-breakup well-being score =
1.69 (SD = .50), Average post-breakup well-being score = 1.59 (SD = .48), Intercept = 1.69;
Slope = -.10, p = .55.


34
Figure 2. Trajectory group results of GMM
The results of the GMM analyses provide further support for Hypothesis 1, in that
most people fell into a group characterized by a higher level of well-being at pre-breakup and
a lower level of well-being at post-breakup (N= 49, 66.2% of the sample, blue, purple, and
green lines in Figure 2 above, groups 1, 3, 4). There was mixed support for Hypothesis 2. As
noted above, GMM did not reveal the four hypothesized trajectories specifically; however,
the groups that emerged did represent three of the proposed trajectories. As hypothesized,
there was a stable low well-being group and an increasing well-being group. There were
three groups that represented the hypothesized decreasing well-being from pre to post
breakup, each with different intercepts and slopes. The only hypothesized trajectory that did
not emerge in this sample was a stable high well-being group.


35
Interpretation of Well-Being Scores
Our understanding of these groups can be improved by considering how to interpret
the levels of the MASQ and degree of change of the MASQ. Few have attempted to provide
clinical utility or cut-off scores for the MASQ. Findings by Buckby et al. (2007) would
suggest that a score below 2.55 on the MASQ (when reverse scored to reflect well-being)
would represent possible depression (see Figure 3). Whereas Bucky et al. (2007) used the
larger MASQ scale, the 12 MASQ items used in this study were chosen to be included in the
larger parent study based on factor analyses that indicate that they capture a more general
psychological distress rather than symptoms specific to anxiety or depressive disorders
(Keogh & Reidy, 2000; Rhoades, Stanley, & Markman, 2011). Nonetheless, the cut-off score
suggested by Buckby and colleagues may provide a potential indicator of the level of distress
(or lack thereof) that these participants may be experiencing at their pre and post broken
engagement timepoints. Group 5 falls in this range at both pre (1.69) and post breakup (1.59),
and group 1 falls in this range at post-breakup (2.13). Thus, 10.8% of this sample is in a
range of possible clinical or diagnosable distress at pre-breakup, and 21.6% of this sample is
in a range of possible clinical or diagnosable distress at post-breakup (see Figure 3). Using
this potential cut-off, the majority of people who experienced a broken engagement in this
study were not experiencing a clinical or diagnosable level of distress at either pre-breakup or
post-breakup.
Jacobson, Follette, and Revenstorf (1984) initially defined clinically significant
change as the extent to which therapy moves someone outside the range of the dysfunctional
population or within the range of the functional population. Jacobson and Truax (1991)
operationalize this definition and provide both the Reliable Change Index (RCI), and criteria


36
for clinical significance. The four groups that the GMM analysis identified as having
statistically significant change between pre and post would also be considered reliable
change via the RCI, meaning that a change of that magnitude would not be expected due to
the unreliability of the measure. However, following the criteria given by Jacobson and
Truax (1991), the only group who experienced clinically significant change was group 1.
Figure 3. Possible clinical cut-off for levels of well-being
*Note: Levels of well-being in the shaded area, at or below 2.55, may be indicative of clinical or
diagnosable distress.
Hypothesis 3
My third hypothesis was contingent upon the second hypothesis, in that if the
hypothesized trajectories emerged, they would show differences in the variables delineated in
Table 2: pre-breakup levels of variables such as level of friends and family approval of the
relationship (social pressure), desire to marry the partner, felt constraint, dedication,
relationship satisfaction, perceived likelihood of marriage, presence of a set marriage date set,


37
reports of psychological aggression from partner, reports of physical aggression from partner,
and reports of infidelity by the partner, and post-breakup levels of perception of partner
relative (to self) desire for the breakup and reports of infidelity by the partner. Not all
variables were hypothesized to be relevant for all groups.
I proposed to use a series of ANOVAs for the continuous variables (i.e., friends and
family approval (social pressure), felt constraint, dedication, relationship satisfaction,
perceived likelihood of marriage, partner infidelity at pre and post, and psychological
aggression), and a series of Chi Square tests for the categorical variables (i.e., desire to marry
partner, physical aggression, marriage date set, partner desired breakup) to understand how
the variables significantly vary between the groups. See the analyses for each hypothesis
column in Table 2, above.
When I proposed my thesis, there were a few important contributing factors that were
still in flux, including my final sample size, the number/type of trajectories that would
emerge, and the number of people that would populate each trajectory that emerged. Based
on the GMM results, the specific trajectories that emerged were not a perfect match to the
ones I hypothesized. Further, the Ns for the groups were fairly small (two groups with 8
people) and thus power for ANOVAs and Chi-Square analyses across all five groups was
unacceptably low. There was also considerable variability in the size of the groups (range of
8 to 25 people). Analytically, I could not run the proposed analyses with the five trajectories
found, because the basic test assumptions would have been violated (i.e., several cells in Chi-
Square tests did not have expected cell count >5, lack of homogeneity of variance and
unequal number in each group for ANOVA tests). Thus, comparing all five groups
statistically in the manner proposed for Hypothesis 3 was inappropriate. Further, collapsing


38
the three declining groups into one group was not appropriate given the GMM findings that
these were three statistically distinct trajectories.
However, regardless of the planned statistical tests, I had still painted a hypothesized
conceptual picture of each group based on the measured variables. Thus, despite my lack of
ability to compare the groups statistically, I chose to proceed in a few different ways in an
attempt to still characterize and understand the people who fell into the different well-being
trajectory groups. I first attempted a rough narrative, based on visual examination of the
means or percentages, regarding the convergence or lack of convergence with my hypothesis
(see Appendix A). However, this approach became too problematic, given that results
indicated more groups than I initially hypothesized and the difficulty in objectively
quantifying what high or low meant in relation to the obtained group means or how
much would represent a meaningful level of relatively higher or lower among the groups.
Therefore, I instead chose to highlight the patterns, unique characteristics, and what
stood out to me among each trajectory group or relevant clusters of groups. This is based on
my subjective review of the results in Tables 5 and 6.1 will then present some post hoc
analyses to statistically predict the degree of drop in well-being for the three trajectory
groups who all experienced a drop in well-being from pre-broken engagement to post-broken
engagement.
Trajectory groups 1,3,4: Drop in well-being from pre to post broken engagement
When I was originally conceptualizing possible trajectory groups, I thought that there
would be a group who was doing relatively well at pre-breakup, and then would experience a
drop in well-being at post-breakup. Trajectories 1, 3 and 4 are all groups of people who
experienced a drop in well-being from pre-broken engagement to post-broken engagement.


39
Although three trajectories emerged in place of the one that I hypothesized, there were
certain themes present among all three groups that suggest that, compared to groups 2 and 5,
these three groups were doing relatively well at pre-breakup and thus, their drop in well-
being may be a sign that they were relatively blind-sided by the breakup. Evaluating the
results in Table 6, it appears that these three groups, compared to groups 2 and 5, had
relatively higher overall pre-breakup desire to marry their partner, likelihood of marriage,
and friends and family approval (social pressure) for their relationship to work. They also
seemed to have the lowest rates of pre-breakup infidelity and they completed their post-
breakup surveys at the shortest amount of time since their breakup, perhaps representing less
recovery time after breakup.
It appears that group 1 looks most like my hypothesized declining group. Group 1
experienced the greatest drop in well-being from pre to post-breakup, their average post-
breakup level of distress falls in the possible range of clinical or diagnosable distress, and
their drop in well-being was clinically significant. Their relatively higher level of well-being
at pre-breakup may be supported by the following: they seemed to have the highest level of
pre-breakup relationship satisfaction, were the only group who had every single participant
indicate a full desire to marry their partner, and seemed to experience the lowest pre-breakup
amount of psychological and physical aggression from their partner. Considering their
seemingly positive relationships, they may then have been surprised or blind-sided by the
breakup or certain events that may have contributed to their drop in well-being. For example,
they appeared to endorse the highest percentage (along with group 5, stable low) of people
reporting, yes I know for sure that their partner was unfaithful at breakup.


40
Results from the reasons for breakup variable (see Table 5) provide further potential
evidence that possible negative circumstances surrounding their breakup may not have been
present throughout their relationship (and thus the breakup or events such as infidelity may
have come as a shock to the participants). For example, this was the only group where 0%
endorsed substance abuse, domestic violence, economic reasons, and sexual incompatibility,
and they also had the lowest percentage of people endorsing arguing and lack of
communication (along with group 5, stable low) as reasons for breakup. These are all factors
which, had they been present in the relationship, may have given the participants some
indication that their relationship wasnt doing well at the pre assessment. Furthermore, they
had the highest percentage of people endorsing partner didnt want to marry on the reasons
for breakup variable, which may be another possible indication that this breakup was
unexpected and/or particularly hurtful or distressing. However, it should be noted that they
also had the highest ratings on participant (self) didnt want to marry on the reasons for
breakup variablegiven the pre levels of relationship positives, this high level may have
emerged from upsetting and unforeseen events such as infidelity.
Trajectory group 2: Increase in well-being from pre to post broken engagement
Group 2 is the only group of people who experienced an increase in well-being from
pre to post breakup. I had initially hypothesized that this group may have experienced their
broken engagement as a relief because they may have experienced markers for distress in
their relationship at pre-breakup. Results (see Table 6) indicate there were indeed markers for
distress in their relationship at pre-breakup, such as appearing to have the lowest percentage
of people who had a marriage date set and being the only group where some participants
indicated that they did not want to marry their partner (5.9%) at pre-breakup. Further, they


41
reported the second lowest level of likelihood of marriage, dedication, and social pressure
from friends and family. Thus, they may have been somewhat aware of various negative
aspects of their relationship, and therefore their broken engagement may have been
somewhat of a relief.
Results from the reason for breakup variable (see Table 5) may provide further
evidence that this group may have felt relief from getting out of a potentially bad relationship
and were not experiencing factors that may have prolonged their low well-being even after
the breakup. For example, this group had the highest percentage of people endorsing
economic reasons and lack of communication as reasons for breakup. These reasons may
have been long-standing and were likely not a one time event. Thus, the participants may
have felt relief when the breakup happened, but the dynamics in their relationship werent so
extreme (i.e., substance abuse, infidelity, physical or psychological abuse) that their impact
lasted well beyond the breakup, as was perhaps the case with the stable low group (group 5).
This group, on average, completed their post-breakup surveys in the longest amount
of time since breakup, which could suggest that this increase in well-being may not be an
initial increase (e.g., relief) but may in fact be the result of greater time for recovery. They
were also engaged, on average, for the shortest amount of time.
Trajectory group 5: Stable low well-being from pre to post broken engagement
Group 5 is the only group of people who experienced a low level of well-being at
both pre and post broken engagement. Both of their pre-breakup and post-breakup levels of
distress were in the possible range of clinical or diagnosable distress. I had initially
hypothesized that this was a group of people who have had had markers for distress in their
relationship, but they may not experience an increase in well-being following avoiding a


42
presumably bad situation because they may still have had other distressing factors in their life
that were contributing to their well-being. Results (see Table 6) indicate that this groups
somewhat chronic distress may be evidenced by having the highest percentage of participants
who reported at pre-breakup that, yes I know for sure that their partner was unfaithful,
were unsure if they wanted to marry their partner, and who reported psychological aggression
from their partner. They also had the lowest perceived likelihood of marriage and friends and
family approval (social pressure) for their relationship to work, and the second lowest level
of relationship satisfaction. However, although they got out of a seemingly bad relationship
or environment, they still may have had factors that got in the way of their well-being
increasing after breakup, such as having the highest percentage of people who reported yes
I know for sure that their partner was unfaithful at post-breakup (along with group 1) and
whose partner desired the breakup.
Results from the reason for breakup variable (see Table 5) provides further possible
evidence for their on-going difficult situation that may contribute to their low well-being at
both pre and post breakup: they had the highest percentage of people endorsing substance
abuse, domestic violence, and infidelity as reasons for breakup, all events which may
intensify a drop in well-being due to the devaluing aspect of the event. Specifically, the work
done by Cano and OLeary (2000) suggest that the humiliating events of infidelity, violence,
or partner initiated breakup may especially prolong a breakup, even if getting out of the
relationship is a positive thing in the long-run. This group, on average, completed their post-
breakup surveys in the second longest amount of time since breakup and were engaged for
the longest amount of time. This could add further evidence to the idea that this low well-
being is somewhat chronic, and although they may be getting out of a potentially bad


43
relationship, other distressing life factors may still be contributing to their persistent low
well-being.
Table 5, Percentages of people endorsing different reasons for breakup within the 5 groups
Reasons for Breakup // Trajectories 1. (Initial moderate well-being, significant drop) 8 people 2. (Initial moderate well-being, significant increase) 17 people 3. (Initial moderate well-being, significant drop) 16 people 4. (Initial moderate well-being, significant drop) 25 people 5. (Stable low well- being) 8 people
Substance abuse 0% 11.8% 6.3% 20% 62.5%
Religious reasons 12.5% 5.9% 6.3% 16% 12.5%
Domestic violence 0% 5.9% 6.3% 4% 12.5%
Economic reasons 0% 29.4% 25% 16% 0%
Infidelity 37.5% 35.3% 25% 36% 62.5%
Arguing 37.5% 47.1% 68.8% 68% 62.5%
Lack of communication 25% 64.7% 31.3% 32% 25%
Sexual incompatibility 0% 17.6% 25% 16% 37.5%
Partner didnt want to marry 25% 23.5% 0% 12% 12.5%
Participant (self) didnt want to marry 12.5% 11.8% 6.3% 12% 12.5%


44
Table 6. Results of proposed distinguishing variables within the 5 groups
1. (Initial moderate well-being, significant drop) 8 people 2. (Initial moderate well-being, significant increase) 17 people 3. (Initial moderate well-being, significant drop) 16 people 4. (Initial moderate well-being, significant drop) 25 people 5. (Stable low well- being) 8 people
Partner infidelity atpre breakup No = 75% For sure = 12.5% No = 47.1% For sure = 29.4% No = 56.3% For sure = 12.5% No = 64% For sure = 16% No = 25% For sure = 50%
Partner infidelity at post breakup No = 12.5% Yes I know for sure = 37.5% No = 29.4% Yes I know for sure = 35.3% No = 50% Yes I know for sure = 18.8% No = 32% Yes I know for sure = 32% No = 12.5% Yes I know for sure = 37.5%
Partner desired breakup at post breakup 50% partner desired breakup 29.4% partner desired breakup 18.8% partner desired breakup 24% partner desired breakup 62.5% partner desired breakup
Marriage date set at pre breakup 37.5% date set 35.5% date set 37.5% date set 44% date set 37.5% date set
Relationship satisfaction at pre breakup Range 9-21 Mean=16.86 Range 5-21 Mean= 15.05 Range 3-19 Mean=13.93 Range 8-21 Mean = 16.24 Range 9-19 Mean = 14.75
Felt constraint at pre breakup Range 1- 3.67 Mean =1.67 Range 1-6 Mean=2.43 Range 1-6.33 Mean=2.96 Range 1-3.67 Mean =1.63 Range 1-5 Mean= 2.54
Desire to marry partner at pre breakup 100% have full desire to marry 70.6% have full desire to marry, 23.5% are unsure, 5.9% do not want to marry partner 75% have full desire to marry, 25% are unsure 88% have full desire to marry, 12% are unsure 62.5% full desire to marry, 37.5% are unsure
Perceived likelihood of marriage at pre breakup 62.5% very likely 47.1% very likely 81.3% very likely 68% very likely 37.5% very likely
Dedication at pre breakup Range: 3.14- 7 Mean = 5.70 Range: 1.75-7 Mean=5.37 Range 1.12- 6.75 Mean=5.09 Range 4-7 Mean = 6.03 Range 3.13-7 Mean = 5.67


45
1. (Initial moderate well-being, significant drop) 8 people 2. (Initial moderate well-being, significant increase) 17 people 3. (Initial moderate well-being, significant drop) 16 people 4. (Initial moderate well-being, significant drop) 25 people 5. (Stable low well- being) 8 people
Psychological aggression (from partner) at pre breakup Range: 0 -4 Mean = 2.09 Range: 0-6.25 Mean=2.44 Range: .5- 5.75 Mean = 3.02 Range 0-6 Mean = 2.23 Range 0-7 Mean = 3.57
Friends/family approval (social pressure) at pre breakup Range: 2.25- 6.75 Mean = 4.06 Range: 1-7 Mean = 3.98 Range: 3- 6.75 Mean = 4.71 Range 2-7 Mean = 4.48 Range 1.75-5.5 Mean = 3.53
Physical aggression (from partner) at pre breakup 37.5% had some aggression 58.8% had some aggression 62.5% had some aggression 40.0% had some aggression 50.0% some aggression
Average time since breakup in days 39.63 days Range 1- 116 Median = 16.50 101.59 days Range 16-230 Median = 79.00 80.00 days Range 6-152 Median = 93.00 50.00 days Range 1-152 Median = 36.00 100.29 days Range 1- 142 Median = 124.00
Average length of engagement in days (recoded) 722.17 days Range 145 1288 Median = 715.00 411.00 days Range 25- 1460 Median = 292.00 515.40 days Range 76- 1289 Median = 350.00 530.24 days Range 94- 1460 Median = 497.00 995.00 days Range 599- 1406 Median = 980.00
Post Hoc Analyses (predicting level of drop in well-being among those who experienced a
drop in well-being):
In order to determine if certain variables are related to the amount of drop in well-
being from pre to post-breakup, I isolated the members of the three groups who all
experienced a drop in well-being (N = 49) and created a variable which reflected the amount
of drop, with higher scores indicating a greater drop in well-being from pre-breakup to post-
breakup. I then ran analyses with this well-being drop variable as the dependent variable
and each individual hypothesized variable as the predictor variable in separate regressions.


46
Using G*power for a linear regression with one predictor, in order to determine the
sample size needed to detect medium effects (.15, Cohens f2), the required sample size is 89.
Therefore, my sample of 49 people who all experienced a drop in well-being was not large
enough to detect small or medium effects. The required sample size needed to detect large
effects (.35, Cohens f2) is 40. Therefore, if large effects exist in my data, I have adequate
power to detect them.
Results indicated that none of the variables examined significantly predicted the
amount of drop in well-being. Specifically, drop in well-being was not significantly predicted
by pre breakup levels of partner infidelity, relationship satisfaction, felt constraint, perceived
likelihood of marriage, psychological aggression from partner, dedication, or friends/family
approval (social pressure). Further, the two calculated variables, time since breakup and
length of engagement, did not significantly predict a drop in well-being. However, these two
variables should be interpreted with caution due to the amount of missing data for each.
Finally, partner infidelity at post also did not significantly predict a drop in well-being.
Following Cohens effect size guidelines for correlation coefficients (.5= large, .3 =
medium, .1= small), it is clear from the correlation matrix below (Table 7) that there are a
few variables that had a small effect on the level of drop in well-being (see the first column
in Table 7). These small effects did not result in significant results due to low power.
Variables with small effects include: partner infidelity at post breakup (r = .16), pre breakup
levels of relationship satisfaction (r = .12), felt constraint (r = -.13), perceived likelihood of
marriage (r = .16), and friends/family approval (social pressure) (r = .19), as well as time
since breakup (r = -.24).


47
Table 7. Correlations between amount of drop in well-being and other variables for people
who had a drop in well-being from pre to post broken engagement__________________________
Amount of w drop in ~ -.09 Partner infidelity at pre breakup
.16 Partner infidelity at post breakup
.12 Relationship satisfaction at pre breakup
-.13 Felt constraint at pre breakup
.16 Perceived likelihood of marriage at pre breakup
well-being ^ (higher scores indicate greater
-.01 Psychological aggression (from partner) at pre breakup
-.01 Dedication at pre breakup
.19 Friends/family approval (social pressure) at pre breakup
-.24 Time since breakup (days)
.08 Length of engagement (days)
*None of these correlations were significant.


48
CHAPTER IV
DISCUSSION
Engagement is often thought of as an idealistic phase in a relationship, and yet it has
been estimated that 15-20% of engagements are broken every year (Safier, 2003). Although
cold feet before marriage is a common lay idea, research into the outcomes and associated
factors of broken engagements is far less common. This study was the first to look at
potential changes in individuals well-being from an engaged status (Time 1/pre-breakup
timepoint) to a broken engagement status (Time 2/post-breakup timepoint). Based on
findings on unmarried breakups from Rhoades et al. (2011), I first predicted that there would
be a significant decline in level of well-being from pre-breakup to post-breakup for people
who have experienced a broken engagement, and results of a simple paired t-test supported
this prediction. Results revealed that 66.2% of the sample experienced a significant decline in
well-being from pre to post broken engagement. None of the variables examined in this study
significantly predicted the degree of drop in well-being, which may be an issue of low power
(see Limitations).
My second hypothesis was that, in addition to the average declining trajectory of
well-being, three additional distinguishable trajectories of well-being from pre-breakup to
post-breakup (steady moderate/high well-being, steady low well-being, improving well-
being) would emerge. Results of a Growth Mixture Model analysis revealed five distinct
trajectories, representing three of the four hypothesized trajectories, including a stable low
well-being group (10.8%), an increasing well-being group (23%), and three distinct declining
well-being trajectories, with different slopes and intercepts. The only hypothesized trajectory
that did not emerge was the steady moderate/high well-being. Although 66.2% of the sample


49
experienced a drop in well-being, only one of the declining trajectory groups (10.8%, group
1) experienced clinically significant change in their pre to post broken engagement well-
being. In other words, of those who experienced a decline in well-being from pre to post
breakup, 16.3% experienced a clinically significant drop.
Using a clinical cut-off score suggested by Buckby et al. (2007), 10.8% of the entire
sample was in the range of possible clinical or diagnosable distress at pre-breakup, and
21.6% of this sample was in the range of possible clinical or diagnosable distress at post-
breakup. Thus, the majority of people who experienced a broken engagement were not in a
clinical or diagnosable range of distress at either pre or post-breakup. Further, it appears that
half of the people who experienced clinical distress after a broken engagement were already
experiencing clinical distress before the breakup. In addition, humiliating relationship events
involved with the breakup, as detailed by Cano et al., (2000), appear to influence ones level
of well-being after a broken engagement as well. The 21.6% of the sample who experienced
clinical distress at post-breakup were also those who, on average, did not desire the breakup
and who reported the highest level of partner infidelity at post-breakup.
My third hypothesis was to assess what characterizes individuals in the different
trajectories of self-report well-being before and after a broken engagement. Although
statistical analyses were not appropriate due to the number and unequal composition of the
trajectories, and the small number of people in some trajectories, there were various
descriptive ways in which the conceptual story of the different trajectories emerged.
Declining well-being trajectory
This declining well-being trajectory can be conceptualized as a group of people who
were doing relatively well at pre-breakup, and then experienced a drop in well-being at post-


50
breakup. Examples of doing well at pre-breakup include relatively high levels of pre-
breakup relationship satisfaction, desire to marry their partner, and dedication. Given their
seemingly high well-being at pre-breakup, the breakup may have represented a relatively
abrupt and traumatic loss or rupture. For example, at post-breakup, the group with the
greatest declines also had the highest percentage (along with the stable low group) of people
endorsing that their partner was unfaithful. However, it is also true that this group completed
their post-breakup survey in the shortest amount of time since their breakup, perhaps
representing less recovery time from the breakup compared to other groups who had higher
levels of well-being at post-broken engagement.
Increasing well-being trajectory
This increasing well-being group may have experienced the broken engagement as a
relief due to some distress or ambivalence at pre-breakup. (However, the markers for distress
may not have been humiliating or devaluing which may have contributed to more longer,
chronic low well-being (see Cano & OLeary, 2000).) For example, this group had the lowest
percentage of people with a set wedding date and was the only group of people wherein
participants indicated they did not want to marry their partner. They were also engaged, on
average, for the shortest amount of time, which could suggest that they had less time to invest
or plan for their upcoming wedding/marriage, and therefore may have experienced less
overall loss as it relates to both the relationship and upcoming marriage. Moreover, this
increasing well-being group, on average, completed their post-breakup surveys in the longest
amount of time since breakup, which could suggest that this increase in well-being may not
only be a reflection of relief, but could also be the result of greater time for recovery.


51
Stable low well-being trajectory
The stable low well-being trajectory could represent a group of people who were
unhappy in their relationship at pre-breakup, but who may not have experienced an increase
in well-being or relief from the breakup because they may still have had sources of distress
and/or may have experienced certain humiliating events (see Cano & OLeary, 2000) that left
them feeling devalued and contributed to their stable low well-being. For example, this group
had the highest percentage of partner infidelity and psychological aggression, as well as
reported substance abuse and domestic violence as reasons for breakup. This group was also
engaged for the longest amount of time, which may mean that the negative aspects of their
relationship affected their well-being for a considerable amount of time.
Limitations
Before discussing potential implications of the results, it is important to highlight
several limitations of the study. All measures were asked from the participants point of view.
The participants understanding of the relationship may differ from his or her partners
understanding of the relationship. Due to the inclusion criteria for the study, the 74
participants represent only those persons with a broken engagement who continued to
participate after the engagement ended. It may be that individuals with a broken engagement
who discontinued participation would show higher rates of distress and destabilization. That
is, the method may have selected for individuals with greater average levels of adjustment
post break up. The opposite is also possible; that is, persons with higher post breakup distress
being more motivated to complete the measure and share their experiences. In either case, the
method introduces a selection factor that may undermine the generalizability of the results.
Missing data are also a limitation of the study (particularly in the case of length of


52
engagement). Thus, it is important to interpret any results with that variable with more
caution than usual.
The study design has obvious flaws in terms of isolating the time line of events. It is
possible that there were other events that happened within the pre-breakup to post-breakup
assessment points that caused changes in their post-breakup level of well-being. That is, it
could be that a participant filled out their pre-breakup survey, experienced a decrease in well-
being, and then broke up and filled out their post-breakup survey. Thus, extraneous history or
maturation effects may have influenced the results. Moreover, the extent to which the
observed outcomes for the groups will continue or change is not known due to the limited
time-frame of this study.
Due to the relatively small sample size and the five trajectories that emerged, weak
power prohibited me from being able to statistically analyze differences between groups, and
was the reason that the majority of the Aim 3 results were speculative. Using such a
subjective process represents additional risk of focusing only on confirming evidence or
glossing over disconfirming evidence.
Implications
These findings could be relevant for currently engaged individuals experiencing,
approaching, or contemplating a breakup, as well as clinicians working with such individuals.
Although it is unknown what longer term outcomes would be, and several other areas of
inquiry are needed to further understand the clinical utility of these findings, certain general
themes can be gleaned from this study. Overall, about half of the people who experience
clinical distress after a broken engagement were already experiencing clinical distress before
the broken engagement. For individuals who were individually distressed prior to breakup


53
(10.8%), breaking an engagement may not change their level of low well-being. The small
subset of individuals (also 10.8% of this sample) who dropped into the clinically distressed
range may be blind-sided by their breakup or events such as infidelity, but the time-frame for
this distress is not known given the relatively short duration of follow up for this group.
In contrast, a subset of individuals (23% of the current study) actually improved in
their well-being pre to post breakup. This group was also characterized by the shortest
duration of engagement prior to breakup. This relatively short duration could represent less
investment in the engagement, and then less loss/disruption upon breakup. It is also possible
that breaking an engagement sooner (perhaps before set wedding plans are made) is better for
individual outcomes.
In general, for the majority of people who are distressed or ambivalent in their
engaged relationship (e.g., have relatively low levels of dedication), but who are also not
already experiencing clinically significant distress or certain de-valuing events or long-
standing issues (i.e., substance abuse, infidelity, physical or psychological abuse), there is a
good chance that their initial level of well-being following a broken engagement may not
change in a clinically significant way either up or down. This appears to describe the
majority of people who experience a broken engagement (78.4% of this sample).
In fact, given longer time frames of follow up, an even smaller proportion of the
sample may have evidenced clinical distress at post breakup. Specifically, the group who
experienced a clinically significant level of decreased well-being answered their post-
breakup survey in the shortest amount of time since breakup, and the group who experienced
an increase in well-being answered their post-breakup survey in the longest amount of time


54
since breakup. Both of these findings suggest that more recovery time from the event may be
associated with greater adjustment.
This could be useful information to a person interested in how one will fare after a
broken engagement. Overall, if a person was doing relatively well prior to the broken
engagement or didnt experience humiliating or devaluing events associated with their
breakup, it may be that time will help in healing from this event and their well-being may not
drop to a level of clinical distress. On the other hand, the degree to which ones breakup is
associated with existing distress and/or humiliating or devaluing events (i.e., partner desired
breakup, infidelity, physical or psychological abuse, substance abuse), it may be beneficial to
seek help and extra support during this time, as their level of well-being may drop to a level
of clinical distress immediately following the breakup. As Cano et al. (2000) suggest,
effective clinical interventions aimed at coping with feelings of betrayal, humiliation, and
shame may be particularly helpful for individuals who experience humiliating or devaluing
events associated with their breakup. Likewise, clinical interventions aimed at alleviating
chronic distress (likely compounded by but not fully attributable to the breakup) may be
helpful for individuals who were already experiencing clinical distress before their breakup.
At times, it may also be useful to reframe a broken engagement as a positive outcome in the
long run, even if there is current distress over the breakup. That is, it may be adaptive to
move on from relationships characterized by infidelity, abuse, drug or alcohol problems, or
uncertain commitment (e.g., Bourassa et al., 2015; Lavner et al., 2012; Markman et al., 2013).
In fact, a Huffington Post author discussing her decision to call off a wedding stated that one
of the five pieces of guidance she wished she had available when she was weighing the


55
options and questioning her decision was you will survive (Sturges, 9/28/15). The current
study helps inform this particular aspect of broken engagements.
Future Directions
The current study provides useful information on trajectories of well-being before and
after engagement. Yet, it does not address certain questions that individuals may have as they
face significant decisions regarding engagement and marriage. For example, one may wonder
if some level of doubt about the relationship or certain relationship conflicts are a predictor
of relationship problems in the future, or whether they are typical of couples who go on to
have happy and stable relationships. Although people generally use cold feet and doubt
synonymously when discussing premarital doubts, there is not a clear definition of cold feet.
Even Websters dictionary: apprehension or doubt strong enough to prevent a planned
course of action, and Dictionary.com: a loss or lack of courage or confidence; an onset of
uncertainty or fear provide somewhat opposing definitions of cold feet. What is clear
from each definition is that cold feet during an engagement may lead one to question their
upcoming marriage. What is unclear is whether there is a certain level of cold feet that is
predictive of marital distress that may be wise to pay attention to (i.e., a broken engagement),
versus lower levels of cold feet that are not necessarily indicative of future distress (i.e.,
just cold feet). In the same Huffington Post article noted above, the author writes: I know
there are people out there who are worried about the small voice inside begging them to wait,
and I wish there were more readily available resources to help figure out what that voice
means (Sturges, 9/28/15).
This absence of clarity could be addressed if a prospective, longitudinal study of
relationship development were conducted in which specific questions about levels and types


56
of pre-marital doubts were asked, along with additional variables. The sample could then be
split into 3 groups: (1) broken engagements (those who got engaged and then broke off
their engagement), (2) successful marriages (those who got engaged, married, and remain
relatively satisfied), and (3) distressed marriages (those who got engaged, married, and
then divorced or remain married with high levels of distress). With these data, we could then
determine what the pre-broken engagement levels of certain variables are for the three groups.
Although we already know that six months into marriage, endorsing the retrospective yes/no
question, were you ever uncertain or hesitant about getting married? predicts poorer
marital outcomes after four years of marriage (Lavner et al., 2012), this is a very broad and
generic marker of doubt. A comparison study following the three groups noted above may
help us determine levels and types of doubt that may be predictive of marital distress. That is,
we may be able to better define cold feet before marriage and provide certain cutoff levels
of cold feet that may or may not be predictive of future marital distress.
A comparison study like this may allow both clinicians and engaged individuals to
consider this information through a cost-benefit analysis lens, by considering the likelihood
of and degree of drop in well-being over this specific time in life (as suggested by the current
results) against the potential for a broken engagement or divorce later on. Thus, if ones
engagement levels are closer to the successful marriage group levels, they may feel more
confident in pursuing marriage despite their premarital doubts, whereas if their levels are
more similar to the distressed marriages group they may want to consider avoidance of a
potential divorce later on down the road, including avoiding the possible psychological
consequences that a divorced couple and their family may experience (Waite et al., 2002).
That is, such data may help individuals put their pre-marital distress or doubt into context and


57
allow them to decide whether it would be worth it to break their engagement (even after
they bought their dress/tux, sent out invitations, etc.), given the likelihood of certain well-
being trajectories.
There are several other important future directions for this study. Although this
current study did not collapse different trajectory groups together or look at isolated
comparisons among just a few of the trajectory groups, rationale could be developed for
either of those data analysis methods. Further, it may be helpful to better understand what a
self-report relationship status of engaged means to different people (i.e., for those who go
back and forth between engaged and dating, or for those who equate making an oral
commitment to get married to their partner someday with being engaged), and whether a
broken engagement looks different depending on ones definition of engaged.
In conclusion, results of this study are the first to demonstrate that there are different
trajectories of well-being from pre broken engagement to post broken engagement. Although
this adds to the scientific understanding of this unique time in a relationship, there is still
much to learn about the characteristics of the different well-being groups in order to advise,
in an empirically sound way, individuals considering or experiencing a broken engagement,
or clinicians working with these individuals.


58
REFERENCES
Bonds-Raacke, J. M., Bearden, E. S., Carriere, N. J., Anderson, E. M., & Nicks, S. D.
(2001). Engaging distortions: Are we idealizing marriage? Journal Of Psychology:
Interdisciplinary And Applied, 135(2), 179-184. doi: 10.1080/00223980109603689
Bourassa, K.J., Sbarra, D. A. & Whisman, M. (2015). Women in Very Low Quality
Marriages Gain Life Satisfaction Following Divorce. Journal of Family Psychology,
29(3): 490-499. doi: 10.1037/fam0000075
Buckby, J. A., Yung, A. R., Cosgrave, E. M., & Killackey, E. J. (2007). Clinical utility of the
Mood and Anxiety Symptom Questionnaire (MASQ) in a sample of young help-
seekers. BioMed Central Psychiatry, 7:50.
Cano, A., & OLeary, D. K. (2000) Infidelity and Separations Precipitate Major Depressive
Episodes and Symptoms of Nonspecific Depression and Anxiety. Journal of
Consulting and Clinical Psychology, 68(5) : 774-781.
Clark, L.A., & Watson D. (1991). Tripartite model of anxiety and depression: Psychometric
evidence and taxonomic implications. Journal of Abnormal Psychology, 100:316-336.
Elison, C. C. (2011). Disharmony: Premarital relationship dissolution. Dissertation
Abstracts International Section A, 72
Fowers, B. J., & Olson, D. H. (1992). Four types of premarital couples: An empirical
typology based on PREPARE. Journal Of Family Psychology, 6(1), 10-21.
doi: 10.1037/0893-3200.6.1.10
Graham, J., Liu, Y., & Jeziorski, J. (2006). The Dyadic Adjustment Scale: A Reliability
Generalization Meta-Analysis. Journal of Marriage and Family, 68(3): 701-717.
Grimm, K. J., Ram, N., and Estabrook, R. (2010). Nonlinear Structured Growth Mixture
Models in Mplus. Multivariate Behavioral Research, 45:6. doi:
10.1080/00273171.2010.531230.
Jacobson, N. S., Foilette, W C., & Revenstorf, D. (1984). Psychotherapy outcome research:
Methods for reporting variability and evaluating clinical significance. Behavior
Therapy 15, 336-352.
Jacobson, N. S. & Truax, P. (1991) Clinical Significance: A Statistical Approach to Defining
Meaningful Change in Psychotherapy Research. Journal of Consulting and Clinical
Psychology, 59(1): 12-19.
Kenny, J. J, Kayla, K., Rhoades, G.K., Stanley, S.M., & Markman, H.J. (2013). I (do)nt:
Causes and Effects of Broken Engagements. Poster presented at the Association for
Behavior and Cognitive Therapies conference; Nashville, TN.


59
Keogh, E, & Reidy J. (2000). Exploring the factor structure of the Mood and Anxiety
Symptom Questionnaire (MASQ). Journal of Personality Assessment, 74:106-125.
Lavner, J. A., Karney, B. R., Bradbury, T. N. (2012). Do Cold Feet Warn of Trouble Ahead?
Premarital Uncertainty and Four-Year Marital Outcomes. Journal of Family
Psychology, 26(6): 1012-1017.
Markman, H.J., Rhoades, G.K., Stanley, S.M., & Peterson, K.M. (2013). A Randomized
Clinical Trial of the Effectiveness of Premarital Intervention: Moderators of Divorce
Outcomes. Journal of Family Psychology, 27(1): 165-172.
Moir-Smith, A. (2006). Emotionally engaged: A bride s guide to surviving the
happiest time of her life. New York, New York: Penguin Group.
Muthen, B., Brown, H. C., Hunter, A. M., Cook, I. A., & Leuchter, A. F. (2010) General
Approaches to Analysis of Course: Applying Growth Mixture Modeling to
Randomized Trials of Depression Medication. In Shrout, P. E., Keyes, K. M.,
Ornstein, K. (Eds.) Causality and Psychopathology, Finding the Determinants of
Disorders and Their Cures (pp. 159-179). Washington, DC: American Psychiatric
Publishing.
Nylund, K. L., Asarouhov, T. & Muthen, B. O. (2007). Deciding on the Number of Classes
in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation
Study. Structural Equation Modeling, 14(4): 535-569.
Owen, J., Rhoades, G.K., Stanley, S.M., & Markman, H.J. (2011). The Revised Commitment
Inventory: Psychometrics and use with unmarried couples. Journal of Family Issues,
32(6), 820-841.
Rhoades, G. K., Stanley, S. M. & Markman, H. J. (2009). Couples Reasons for
Cohabitation: Associations with Individual Well-being and Relationship Quality.
Journal of Family Issues: 30(2): 233-258.
Rhoades, G. K., Stanley, S. M., & Markman, H. J. (2010). Should I stay or should I go?
Predicting dating relationship stability from four aspects of commitment. Journal of
Family Psychology, 24, 543-550. doi: 10.1037/a0021008
Rhoades, G. K., Kamp Dush, C. M., Atkins, D. C., Stanley, S. M., & Markman, H. J.
(2011). Breaking up is hard to do: The impact of unmarried relationship dissolution
on mental health and life satisfaction. Journal Of Family Psychology, 25(3), 366-374.
doi:10.1037/a0023627
Rhoades G.K., Stanley, S.S., & Markman, H.J. (2012). A Longitudinal Investigation of
Commitment Dynamics in Cohabiting Relationships. Journal of Family Issues, 33(3),
369-390.


60
Sabourin, S.P., Valois, P., & Lussier, (2005). Development and validation of a brief version
of the Dyadic Adjustment Scale with a nonparametric item analysis model.
Psychological Assessment, 17(1): 15-17.
Safier, R. (2003). There goes the bride: Making up your mind, calling it off and
moving on. San Francisco, CA: Jossey-Bass.
Sturges, E. (2015, September 28). The Advice I Wish I Was Given: How to Cancel Your
Wedding. Huffington Post. Retrieved from_http://www.huffingtonpost.com/eve-
sturges/thi s-i s-the-advice-i-wi sh_b_8191242. html.
Scott, S. B., Rhoades, G. K., Stanley, S. M. Allen, E. S., & Markman, H. J. (2013). Reasons
for Divorce and Recollections of Premarital Intervention: Implications for Improving
Relationship Education. Couple and Family Psychology: Research and Practice,
2(2): 131-145.
Selig, A. L. (1976). Crisis Theory and Family Growth. The Family Coordinator, 25 (3),
pp. 291-295.
Spanier, G.B. (1976). Measuring dyadic adjustment: New scales for assessing the quality of
marriage and similar dyads. Journal of Marriage and Family, 38(1): 15-28
Stanley, S.M, & Markman, H.J. (1992). Assessing commitment in personal
relationships. .Journal of Marriage and Family, 54:595-608.
Straus, M. A., Hamby, S. L., Boney-McCoy, S., & Sugarman, D. B. (1996). The Revised
Conflict Tactics Scales (CTS2): Development and Preliminary Psychometric Data.
Journal of Family Issues, 17(3): 283-316.
Veevers, J. (1991). Traumas versus strens: A paradigm of positive versus negative
divorce outcomes. Journal of Divorce and Remarriage, 15(1-2), pp. 99-126.
Waite, L. J., Browning, D., Doherty, W. J., Gallagher M., Luo, Y., & Stanley S. M.
(2002). "Does Divorce Make People Happy? Findings from a Study of Unhappy
Marriages". Institute for American Values, 2002.
Wortel, S.N., & Rogge, R.D. (2010). Can you feel the love tonight? Links between sexual
behavior and both individual and relationship well-being over time. Poster presented
at the Association for Behavior and Cognitive Therapies conference; San Francisco,
CA.


61
APPENDIX A
Results of proposed distinguishing variables with original hypotheses in parentheses and
1 (High, decreasing well-being) 8 people Group #2, Blue Hypothesis 2 (Increasing well-being) 17 people Group #3, Purple Hypothesis 3 (Mid, minimal decreasing well-being) 16 people Group #2, Blue Hypothesis 4 (High, minimal decreasing well-being) 25 people Group #2, Blue Hypothesis 5 (Stable low well-being) 8 people Group #4, Green Hypothesis
Partner infidelity at pre breakup No = 75% For sure = 12.5% (low) No = 47.1% For sure = 29.4% (low) No = 56.3% For sure = 12.5% (low) No = 64% For sure = 16% (low) No = 25% For sure = 50% (high)
Partner infidelity at post breakup No = 12.5% Yes I know for sure = 37.5% (high) No = 29.4% Yes I know for sure = 35.3% (low) No = 50% Yes I know for sure = 18.8% (high) No = 32% Yes I know for sure = 32% (high) No = 12.5/a Yes I know for sure = 37.5% (high)
Partner desired the breakup at post breakup 50% partner desired breakup (yes) 29.4% partner desired breakup (no) 18.8/J partner desired breakup ryes] 24% partner desiredl breakup ryes) 62.5% partner desired breakup (yes)
Marriage date set at pre breakup 87.5% date 85.5% datej &et(yes| 57.5% date 44% date sell 37.5% date set (no)
pet ("ySl sel| ryesj fyesi
Relationship satisfaction at pre breakup Range 9- 21 Mean=16.8 6 (high) Range 5-21 Mean=15.0 5 (low) feange 3 Mean= rhi&h; -il 3.93 Range 8-21 Mean = 16.24 (high) Range 9-19 Mean =l 14.75 (low)
Felt constraint at pre breakup Range 1- 3.67 Mean = 1.67 (low) Range 1-6 Mean=2.43 (high) ibnge 1 Mean=2 [low) -6.33J -.9^ Range 1-3.67 Mean =1.63 (low) Range 1-5 Mean= 2.54 (high)
Desire to marry partner at pre breakup 100% have full desire to marry (sure) 70.6% have full desire to marry, 23.5% are 75% have full desire to marry, 25% are unsure 88% have full desire to marry, 12% are unsure 62.5% full desire to marry, 37.5% are unsure


62
unsure, 5.9% do not want to marry partner (unsure) (sure) (sure)
Perceived likelihood of marriage at pre breakup 62.5% very likely (high) 47.1% very likely 81.3% very likely (high) 68% very likely (high) 37.5% very likely
Dedication at pre breakup Range: 3.14-7 Mean = 5.70 (high) Range: 1.75-7 Mean=5.37 (low) Range 1.12- 6.75 Mean=5.09 (high) Range 4-7 Mean = 6.03 (high) Range 3.13-7 Mean = 5.67 (low)
Psychological aggression (from partner) at pre breakup Range: 0 -4 Mean = 2.09 (low) Range: 0- 6.25 Mean=2.44 (low) Ram 5.75 Mea £e: .5- n = 3.02 n Range 0-6 Mean = 2.23 (low) Range 0-7 Mean = 3.57 (high)
Friends/famil y approval (social pressure) at pre breakup Range: 2.25-6.75 Mean = 4.06 (high) Range: 1-7 Mean = 3.98 (low) Range: 3- 6.75 Mean = 4.71 (high) Range 2-7 Mean = 4.48 (high) Range 1.75- 5.5 Mean = 3.53
Physical aggression (from partner) at pre breakup 37.5% had some aggression (no) 58.8% had some aggression (no) 62.5 SOITK Iggr [no Vo had a ession 40.0% had some aggression (no) 50.0% some aggression (yes)
Time since breakup in days 39.63 days Range 1- 116 Median = 16.50 101.59 days Range 16- 230 Median = 79.00 80.00 days Range 6-152 Median = 93.00 50.00 days Range 1-152 Median = 36.00 100.29 days Range 1-142 Median = 124.00
Length of engagement in days (recoded) 722.17 days Range 145 1288 Median = 715.00 411.00 days Range 25- 1460 Median = 292.00 515.40 days Range 76- 1289 Median = 350.00 530.24 days Range 94- 1460 Median = 497.00 995.00 days Range 599- 1406 Median = 980.00
Note: Variables with no colors were not initially hypothesized but may help to characterize the different
trajectories. Words in quotes and parenthesis are what was initially hypothesized, see also Table 1. Green
means the results map on relatively well to the hypothesis for how that variable would function within the
groups. Yellow means that it is unclear or I am unable to determine if the results map onto the hypothesis for
that variable. Red means that the results do not map on well to the hypothesis for that variable.


Full Text

PAGE 1

i TRAJECTORIES OF SELF REPORT WELL BEING BEFORE AND AFTER A BROKEN ENGAGEMENT b y JESSICA KENNY B.A., Pepperdine University, 2010 M.A., University of Denver, 2013 A thesis submitted to the Faculty of the Graduate School of the Un iversity of Colorado in partial fulfillment of the requirements for the degree of Master of Arts Clinical Health Psychology Program 2015

PAGE 2

ii This thesis for the Master of Arts degree by Jessica Kenny h as been approved for the Clinical Health Psychology program by Elizabeth Allen, Chair Edward Dill Galena Rhoades Joan Bihun Date: December 21 2015

PAGE 3

iii Kenny, Jessica, M.A. Trajectories of Self Report Well Being Before and After a Broken Engagement Thesis directe d by Associate Professor Elizabeth Allen ABSTRACT Engagement is often thought of as the happiest time in a couple's life, yet the concept of "cold feet" is also well known In fact, it has been estimated that 15 20% of engagements are broken every year (Safier, 2003). Despite the prevalence of this important life event little empirical research has focused on engagement distress and breakup. This study was the first to look at potential changes in individuals' psychological well being from an engaged st atus (Time 1/pre breakup timepoint) to a broken engagement status (Time 2/post breakup timepoint) Data from 74 individuals w ere collected through mailed surveys completed before and after their breakup Participants were between the ages of 18 and 39, 69% were fe male, and 72.9% were White. Results of a paired t test indicate that the overall trajectory of well being decrease d from pre broken engagement to post broken engagement (66. 2 % of sample). However, using a Growth Mixture Model analysis, five overall trajectory groups emerged, including a stable low well being group (10.8%) an increasing well being group ( 23 %) and three distinct declining groups with different slopes and intercepts. Of those who experienced a decline in well being from pre to post b roken engagement, only 1 6.3 % experienced a clinically significant change Overall, 21.6% of the total sample was in a range of possible clinical distress at post breakup, and half of those participants ( 10.8% ) were already experiencing possible clinical di stress at pre breakup. Characteristics of each trajectory group l imitations of the study and implications for both clinical work and further research are discussed

PAGE 4

iv The form and content of this abstract are approved. I recommend its publication. App roved: Elizabeth Allen

PAGE 5

v TABLE OF CONTENTS CHAPTER I. INTRODUCTION... 1 Hypothesis 1.....7 Hypothesis 2.....7 Hypothesis 3.. .11 II. METHODS....15 Procedure.......15 Sample....17 Instruments.....17 Data Analysis.....25 III. RESULTS.. ....27 Hypothesis 1.......29 Hypothesis 2.......29 Hypothesis 3.......35 Post Hoc Analyses.............44 IV. DISCUSSION..... ...46 Limitations.........49 Implications....50 Future Directions..... ..53 REFERENCES.... .. ..56 APPENDIX.... .. 59

PAGE 6

1 CHAPTER I INTRODUCTION Conventional wisdom would say that engagement is supposed to be the happiest time in a couples' life. This may be due to some degree of over idealization of the partner during this time in the relationship, as Bonds Raacke, B earden, Carriere, Anderson and Nicks (2001) found that individuals who were engaged reported significantly higher idealistic distortion scores about their partners than did either married individuals or those who were in long term dating relationships. The se authors reported that engaged individuals tend to agree with their partner in an overly positive and enthusiastic manner and reinterpret negative characteristics in a positive light. Although there is, on average, a higher level of this type of positiv ity and idealistic distortion am ong engaged couples, there is also variability in the experiences of individuals in the engagement period, including the well known concept of "cold feet" and accompanying conflict and distr ess. In a non scientific survey of engaged women, only 12% of brides said they felt "really happy" during their engagement, and 23% were sadder, 24% were more afraid, and 41% were more anxious than they expected (Moir Smith, 2006). Moreover, Lavner et al. (2012) retrospectively asked newly married spouses about their premarital uncertainties, and r esponding to a "yes/no" item, 47% of husbands (n = 108) and 38% of wives (n = 87) reported being uncertain about getting married. More specific patterns of difference and conflict can be identifie d through premarital questionnaires that many engaged couples take in preparation for marriage. Fowers and Olsen (1992) identified four types of engaged couples by comparing scores on a premarital inventory, which included subscales across a range of coupl e issues. Vitalized couples (28%) scored high in

PAGE 7

2 compatibility on all dimensions assessed, Present Oriented Couples (27%) had moderately positive relationship quality and compatibility but were somewhat unrealistic in their views of marriage and didn't hav e consensus on plans for th eir future, Future Oriented couples (23%) had moderately low compatibility scores but scored higher on scales assessing planning for marriage and future plans and goals, and Conflicted Couples (22%) were characterized by pervasiv ely low compatibility scores on all scales assessed. Thus, conflict, distress and uncertainty do often occur for a notable number of couples in the engagement period In fact, it has been estimated that 15 20% of engage ments are broken every year (Safier, 2003). Whereas we know that a broken engagement is both a major life change and fairly common, we lack knowledge regarding how a broken engagement might relate to individual psychological adjustment, and what features of the relationship predict such adj ustment Thus, I examined both trajectories and predictors of change in well being for individuals who have experienced a broken engagement, drawing from the larger Relationship Development Study (RDS; Rhoades, Stanley, Markman, 2010). RDS is a longitudina l survey study of 1 295 individuals in various stages of relationship formation and commitment. In an exploratory (unpublished) study of individuals within this sample with a broken engagement, Kenny, Knopp, Rhoades, Stanley, and Markman (2013) found that having experienced a broken engagement at some point in the past is associated with higher current psychological distress and lower life satisfaction than those who have not experienced a broken engagement controlling for age. However, these analyses can not reveal whether this distress and low life satisfaction actually represent a decrease following the broken engagement, or whether this distress preceded the broken engagement. There are reasons to believe that both trajectories

PAGE 8

3 of stable low well being or declining low well being from pre to post broken engagement could be true. Also using the RDS sample, Rhoades, Kamp Dush, Atkins, Stanley, and Markman (2011) assessed the impact that breakups in general have on mental health and life satisfaction, and found declines in well being from pre breakup to post breakup. However, they did not examine this specifically for those who broke up from an engaged relationship. Rather, they included any non marital breakup, including dating relationships. Rhoades et al (2011) modeled living together and having had plans for marriage as moderators of these declines and found that these variables were significantly associated with larger declines in well being following a breakup compared to couples who weren't living to gether and didn't have plans for marriage. Within this sample, they defined participants who had "plans for marriage" as those who answered affirmative to "have the two of you together made a specific commitment to marry?" Four hundred and one individuals who experienced a breakup also reported plans for marriage prior to the breakup. However, only 60 individuals stated that they were actually engaged prior to breakup. Therefore, there is clearly a difference between stating that one has plans for marriage and stating that one's relationship status is engaged, perhaps in terms of increased commitment that comes with being formally engaged versus having general plans for marriage. This current study isolated and analyzed only the people whose reported relati onship status was engaged when they experienced a breakup. The findings of Rhoades et al. (2011) suggest that an overall (average) declining trajectory will be found with this sample as well. That is, this finding supports the idea that, for many individu als, the overall low well being found by Kenny et al. (2013) for those who

PAGE 9

4 have experienced a break up from an engagement may be the result of a decline in well being following a breakup rather than a static low well being level prior to break up On the other hand, there is research that supports the possibility that, for other individuals, self reported low well being after a breakup may reflect ongoing distress that was present even before the breakup. Given the lack of empirical research on outcomes o f broken engagements, literature on well being after both divorce and dating breakup s may help guide hypotheses about engagement breakup s For example, Waite et al. (2003) found that unhappily married adults who then divorced were no personally happier at follow up than those who have stayed married. This suggests that, if there was already distress related to individual or relationship problems prior to breakup, ending that relationship may not automatically relieve this distress. For some individuals, the re may be relatively stable individual distress problems (e.g., depression) or ending a relationship may trade one set of problems (e.g., conflict in the relationship) for another set of problems (e.g., isolation or financial strain). In these scenarios, personal distress would be evident both before and after breakup. This type of trajectory could characterize a number of individuals who experience a broken engagement. Thus, although the dominant hypothesized trajectory is a decreasing level of well bein g following a broken engagement, there may be a number of people who have a stable low trajectory, with low well being present at both pre breakup and post breakup timepoints. Of course, for others there may be entirely different courses over time. Whereas I have discussed potential trajectories of declining well being and stable low well being, some individuals may not show either of these trajectories. Crisis theory proposes that major life events, such as role transitions and separations from significant others, create obstacles for

PAGE 10

5 meeting basic needs and therefore increases the probability of either interpersonal disturbances or new adaptations and increased functional capacity (Selig, 1976). Thus, crisis events in a family create the potential for post crisis deteriorating, status quo, or enhanced functioning. By extension i n addition to the two trajectories already discussed (i.e., a declining course and a stable low course), there could also be an improving course. Illustrating these ideas Veevers (1991) suggests that whereas the definition of divorce as a traumatic crisis has led to a focus on the deleterious effects, a review of the literature indicates that it may also be seen as a strengthening experience. In fact, Bourassa, Sbarra, and Whisman (2015) found that women in the lowest quality marriages increase in life satisfaction following a divorce. The notion of divorce as a positive outcome has also been noted by Markman, Rhoades, Stanley, and Peterson (2013), who analyzed divorce outcomes of couples who had participated in a premarital relationship education program and found that the likelihood of divorce increased for couples who had physical aggression in their relationship. The authors suggest that there may be some situations, such as wh en there is aggression in the relationship, in which divorce could be considered a positive outcome. Unmarried breakups may operate similarly. For example, whereas Rhoades et al. (2011) found an average decline in life satisfaction after breakup, 16.1% of the sample experienced an increase in life satisfaction after breakup. Similarly, 19.6% of their participants experienced a decrease in psychological distress after breakup. Therefore, it is reasonable to believe that some individuals in my sample may repo rt similar positive outcomes, particularly when getting out of a potentially bad relationship Lastly, there is reason to believe that a stable moderate/high trajectory of well being could be possible This could be explained, in part, by the possibility of an adaptation to a

PAGE 11

6 traumatic event as proposed by crisis theory (Selig, 1976). That is, there may be a group of people within the engaged sample who have moderate/high well being at both pre breakup and post breakup. While this may seem counterintuitive Rhoades et al. (2011) found that higher relationship quality pre breakup was associated with smaller declines in life satisfaction following a breakup in an unmarried breakup sample. The authors discussed this as consistent with findings that "happy" peo ple endorse greater adjustment throughout various life challenges. Therefore, there may be a subset of individuals in this broken engagement sample who show a similar stable moderate or high well being trajectory over time. Overall, multiple individual tr ajectories of well being are possible and may be distinguishable from pre to post breakup. In fact, Rhoades et al. (2011) provided percentages of individuals (of the 401 analyzed) who fell into three distinct groups before and after an unmarried breakup: d ecreasing psychological distress (19.6%) increasing psychological distress (30.7%) and stable psychological distress (49.7%) Because the percentages provided by Rhoades et al. (2011) were not isolated for engaged (or "planning for marriage") participant s, a separate analysis would be needed to identify these types of rates for broken engagement. A separate analysis of only engaged individuals is warranted, considering this is a critical period in a person's life, the sparse empirical data available on th e subject of broken engagements, and the apparent distinction between engagement and "planning for marriage." Therefore, the first aim of this study is to determine if the average overall trajectory of well being for pre breakup to post breakup increases or decreases from pre broken engagement to post broken engagement. The second aim of this study is to evaluate whether, and how many, individuals fall into distinct trajectories of well being before and after broken

PAGE 12

7 engagement The third aim is to assess what characterizes individuals in the different trajectories on a range of relationship factors Aim 1: To determine if the average overall trajectory of well being decreases from pre to post breakup for individuals who were engaged pre breakup. Hypoth esis 1 As suggested by Rhoades et al. (2011), there will be a significant decline in level of well being from pre breakup to post breaku p for people who have experienced a broken engagement. Aim 2: To evaluate whether, and how many, individuals fall into distinct trajectories within this sample. Hypothesis 2 As suggested by crisis theory, i n addition to the average declining trajectory of well being (i.e., that may characterize the modal slope) posited in the first hypothesis, three additional distinguisha ble trajectories of well being from pre breakup to post breakup (steady moderate/high well being, steady low well being, improving well being) will emerge. See Figure 1.

PAGE 13

8 Figure 1. Hypothesized trajectories of well being from p re to post broke n engagement Aim 3: To assess what characterizes individuals in the different trajectories of self report well being before and after a broken engagement. Although empirical research on distinguishing characteristics of different well being trajectori es following a broken engagement is sparse, there are several reasons to believe that certain characteristics about the relationship will have an influence on the trajectory of one's well being following a broken engagement. As noted above Rhoades et al. (2011) conducted a similar study with a broader unmarried relationship sample and found that cohabitating and having had plans for marriage were associated with larger declines i n life satisfaction from pre breakup to post breakup whereas having higher rel ationship quality a t pre breakup was associated with smaller declines in life satisfaction following a breakup In addition, having begun to date someone new was associated with smaller declines from pre breakup to post breakup. Whereas Rhoades et al. (201 1) analyzed several important variables related to a Pre Breakup Post Breakup Blue = Declining well being Red = Steady moderate/high well being Green = Steady low well being Purple = Improving well being Level of well being

PAGE 14

9 broad, unmarried sample, I focused on variables that may be particularly relevant to an engaged sample. A potential important moderating factor for pre to post broken engagement well being may be approv al of friends and family. Elison (2011) found that relationship progression to marriage is best predicted by several factors, including approval of relationship by friends and family. If friends and family approval of the relationship is typically a marker of engagement making it to marriage, in the instance of a broken engagement, it may be more of a violation of one's expectancies and contribute to the breakup being more traumatic or distressing. Thus, it may be that individuals who have high friends and family approval at pre breakup would experience a greater decrease in well being following the breakup On the other hand, if the approval is low at pre breakup, this may be associated with a less distressing breakup and higher level of well being followin g the breakup. Another possible moderating factor may be the presence of "humiliating" relationship events, defined as events that devalue the individual. Cano and O'Leary's (2000) research on humiliating marital events found that, controlling for marital discord, women who experienced a humiliating marital event (husband infidelity, husband initiation of separation or divorce, separation or divorce due to husband infidelity or violence) were six times more likely to be diagnosed with a major depressive ep isode than those who had not experienced a humiliating marital event. Thus, if certain humiliating events occur within a relationship, they may intensify a drop in well being even if the breakup is a positive experience for them in the long term. Thus, i t may be that infidelity, physical aggression, or partner desired breakup may be more characteristic of the declining or stable low trajectory groups, whereas

PAGE 15

10 groups with steady moderate/high levels or improving trajectories of well being may be less likely to report aggression, infidelity, or that the partner desired the breakup. Although Rhoades et al. (2011) found that a desire to breakup (from the participant's point of view) was not significantly related to changes in life satisfaction from pre to post b reakup, this study will specifically look at the participant's partner's desire to breakup in relation to lower or higher well being of the participant. If individuals who experience physical aggression, partner infidelity, or a breakup initiated by the pa rtner also experience a decrease in well being following the breakup, this may mirror the work of Cano and O'Leary (2000). Thus, there are multiple possible moderators of change from pre to post breakup. A s noted above, a third aim is to assess what chara cterizes individuals in the different trajectories of self report well being that emerge before and after a broken engagement. The exact execution of this aim was conditional on the results of the second aim That is, assuming I f ound adequate population f or different trajectories of well being, then the third aim is to analyze how the groups representing the distinct trajectories differ on the following variables: pre breakup level s of friends and family approval of the relationship (social pressure) desi re to marry their partner, felt constraint, dedication, relationship satisfaction, perceived likelihood of marriage, presence of a set marriage date, reports of physical and psychological aggression from the partner ( towards the participant ) reports of in fidelity of the partner, and post breakup levels of perception of partner relative (to self) desire for the breakup and reports of infidelity by the participant's partner. Whereas Rhoades et al. (2011) previously analyzed two similar moderators that I pro pose to look at (relationship quality, partner initiated breakup ), this study will attempt to understand if those moderators operate in a similar way in an engaged sample as opposed to a broad unmarried sample. The remainder

PAGE 16

11 of my moderators are distinct f rom those that Rhoades et al. (2011) used, and will provide further understand ing of the unique relationship stage of a broken engagement. Hypothesis 3 The four groups representing individuals in the four distinguished trajectories of well being following a broken engagement will show significant differences in certain variables. Not all measured variables will be relevant for all groups. Using the relevant variables, the variables that may differentiate the four hypothesized trajectory groups are describe d below. See Table 1. 1. (Red line) Stable m oderate / high well being: This is a group of individuals who report moderate to high well being at both pre break and post breakup. It is hypothesized that they may be conceptually characterized in the following way: Alt hough there were no obvious markers for distress in their relationship, they may also have not been entirely invested in the relationship or in the upcoming marriage. Therefore, after breakup, they continue to experience moderate to high well being. Sp ecifically, before breakup, they are hypothesized to be relatively (compared to the other groups): ( 1 ) low in felt constraint, ( 2 ) low in dedication, ( 3 ) high in relationship satisfaction, ( 4 ) low in perceived likelihood of marriage and ( 5 ) low in psychol ogical aggression from their partner. Moreover, they are posited to be more likely to be un sure of their desire to marry their partner, less likely to have a marriage date set before breakup, less likely to report that their partner desired the breakup, le ss likely to report physical aggression towards them in the relationship, and less likely to report partner infidelity before or after breakup. As mentioned above, high relationship satisfaction may be counter intuitive, but Rhoades et al. (2011) found th at higher relationship quality pre breakup was associated with smaller declines in life

PAGE 17

12 satisfaction following a breakup. They discussed this as consistent with findings that "happy" people endorse greater adjustment throughout various life challenges. 2. (B lue line) Moderate to high well being before, low well being after: This is a group of individuals who report moderate to high well being at pre breakup but decline after breakup. It is hypothesized that they may be conceptually characterized in the follow ing way: These individuals were relatively happy in their relationship and they were also entirely invested in their relationship and upcoming marriage. Therefore, after breakup, their well being decreased after an unexpected breakup or accompanying event s Specifically, before breakup, they are hypothesized to be relatively (compared to the other groups): (1) high in friends and family approval of relationship/partner (social pressure) ( 2 ) low in felt constraint, ( 3 ) high in dedication, ( 4 ) high in relat ionship satisfaction, ( 5 ) high in perceived likelihood of marriage and ( 6 ) low in psychological aggression from their partner. Moreover, they are posited to be more likely to be sure of their desire to marry their partner, more likely to have a marriage d ate set before breakup, more likely to report that their partner desired the breakup, less likely to report physical aggression towards them in the relationship before breakup, less likely to report partner infidelity before breakup and more likely to rep ort partner infidelity at the post breakup timepoint. 3. ( Purple line) Low well being before, moderate to high well being after: This is a group of individuals who report low well being pre breakup and higher well being post breakup. It is hypothesized tha t they may be conceptually characterized in the following way: There were markers for distress in their relationship and they weren't entirely invested in the relationship or the upcoming marriage. Therefore, after breakup, they experienced an

PAGE 18

13 increase in well being as they got out of a presumably bad situation. However, they may not have experienced humiliating or de valuing (see Cano & O'Leary, 2000) that would contribute to a potentially chronic low well being. Specifically, before breakup, they are hypo thesized to be relatively (compared to the other groups): (1) low in friends and family approval of relationship/partner (social pressure), (2) high in felt constraint, (3) low in dedication, (4) low in relationship satisfaction, and (5) low in psychologic al aggression from their partner. Moreover, they are posited to be more likely to be unsure of their desire to marry their partner, more likely to have a marriage date set before breakup, less likely to report that their partner desired the breakup, less l ikely to report physical aggression towards them in the relationship before breakup, and less likely to report partner infidelity in the relat ionship before or after breakup. 4. (Green line) Stable l ow well being: This is a group of individuals who report low well being at both pre and post breakup. It is hypothesized that they may be conceptually characterized in the following way: There were markers for distress in their relationship and they also weren't entirely invested or committed to their upcoming mar riage. However, although they avoided a presumably bad situation, they didn't experience an increase in well being, possibly because they still have other distressing factors in their life that are contributing to their low well being and/or because they experienced humiliating or devaluing events (see Cano & O'Leary, 2000) that contributed to a more chronic low well being Specifically, before breakup they are hypothesized to be relatively (compared to the other groups): (1) high in felt cons traint, (2) low in dedication, (3) low in relationship satisfaction and (4) high in psychological aggression from their partner. Moreover, they are posited to be more likely to have no marriage date set before

PAGE 19

14 breakup, more likely to report that their partner desired the breakup, more likely to report physical aggression towards them in the relationship before breakup and more likely to report partner infidelity before and after breakup. Table 1. Hypothesized distinguishing variables of well being trajectories Measu res #1 Red: Stable moderate/ high well being #2 Blue: High to low well being #3 Purple: Low to high well being #4 Green: Stable low well being Friends/family approval (social pressure) at pre breakup N/A High Low N/A Desire to marry partner at pre break up Unsure Sure Unsure N/A Felt constraint at pre breakup Low Low High High Dedication at pre breakup Low High Low Low Relationship satisfaction at pre breakup High High Low Low Perceived likelihood of marriage at pre breakup Low High N/A N/A Marriag e date set at pre breakup No Yes Yes No Partner desired the breakup at post breakup No Yes No Yes Physical aggression ( from partner ) at pre breakup No No No Yes Psychological aggression ( from par t ner ) at pre breakup Low Low Low High Partner infidelity at pre breakup Low Low Low High Partner infidelity at post breakup Low High Low High *Note: For each row, the hypothesis refers to group comparisons. For example, for the friends and family approval variable: group #2 is hypothesized to have higher frien ds and family approval than people in group #3. For the desire to marry partner variabl e: a greater proportion of people in group #2 are hypo thesized to be 100% sure than of people in either group #1 or group #3. For the marriage date set: a greater propor tion of people in group #2 and group #3 are hypothesized to have set a marriage date than people in either group #1 or group #4.

PAGE 20

15 Thus, the overall goal of all three aims is to: (a) evaluate if there is a statistically significant change between the overal l pre breakup and post breakup levels of well being (b) see if there are distinct well being trajectory groups that emerge, and (c) understand how these groups differ on certain pre breakup and/or post breakup variables.

PAGE 21

16 CHAPTER II METHO DS Procedure Participants in the current study (N =74) were participants who are part of a larger longitudinal study (N = 1,295) that completed 12 time points of surveys over the course of four to five years (Rhoades, Stanley, Markman, 2010). To recruit pa rticipants for the larger parent study on relationship development, a calling center used a targeted listed telephone sampling strategy to call households within the contiguous United States. After a brief introduction to the study, respon dents were screen ed for participation. To qualify, respondents needed to be between 18 and 34 years of age and be in an unmarried relationship with a member of the opposite sex that had lasted 2 months or longer. The study was designed to examine ways that premarital exper iences impacted marriage. When the study started, same sex marriage wasn't widely available and therefore, only those in opposite sex relationships were includ ed. Those who qualified, agreed to participate, and provided complete mailing addresses (N = 2,21 3) were mailed forms within 2 weeks of their phone screening. Of those who were mailed forms, 1,447 individuals returned them (65.4% response rate); however, 152 of these survey respondents indicated on their forms that they did not meet requirements for p articipation, either because of age or relationship status, leaving a sample of 1295 for the first wave (T1) of data collection. These 1295 individuals were ma iled the second wave (T2) of the survey about four months after returning their T1 surveys Parti cipants received a survey e very four months for the 12 waves. At each timepoint, participants were asked if they have ever been engaged to someone who they did not end up marrying. Of the overall sample, 22% of individuals

PAGE 22

17 indicated that this was true for them (N = 291). Of the sampl e of 291, participants for whom there was a survey while they were engaged and before their breakup and another survey post break up were included (N = 75). The exact pre and post timepoints are different for each participant. Of the 75 people who had experienced a broken engagement over the course of the study, there were 5 people who had two or more broken engagements over the course of the study (sometimes with the same partner) F or these participants with multiple broken engagements, I used the pre and post timepoints of the first broken engagement that they experienced during the study. Each participant had a pre and post timepoint that was sequential (i.e., no skipped timepoints in between) except for one person. This p erson was engaged at T9, didn't complete a T10 packet, and was broken up at T11. I initially sought to include this person in analyses. However, after running initial Growth Mixtu re Model analyses within Mp lus, I was consistently getting an outlier in the form of one person in their own trajectory class. After matching the ID number to the outlier case, I discovered that the outlier was this p articipant Mplus results indicated that this participant was statistically different fr om the rest of the sample an d thus this outlier affecte d the trajectory classification. When I removed this person from analyses and ran the analyses with 74 people, the trajectories were much more evenly distributed. Thus, this person was excluded as an outlier. Therefore, the final sample for this study was 74 people who were engaged at one timepoint and broken up with that same partner at the subsequent timepoint. Sample Participants were aged 18 39 with an average age of 27.51 (SD = 4.66 ) Sixty nine percent were female; 72.9 % w ere White, 16.2 % Black or African American, 6.8% Hispanic

PAGE 23

18 or Latino, 1.4 % Asian, and 2.7% indicated more than one race. Of the 74 participants, 48.0 % ha d completed at least one year of college Approximately 75 % of this group was employed. The median yearl y income, for those who were employed, was about $30,000 22.9 % of the sample reported that they were datin g someone new at their post breakup timepoint, and 4.1% of the sample reported that they were still dating the person who they had just broken their engagement with. Instruments Measures included in original hypotheses: Well being : To measure well being I used 12 items from the longer Mood and Anxiety Symptom Questionnaire (MASQ; Clark & Watson, 1991), which yields a psyc hological distress score. I t hen reverse score d this measure to reflect less distress, thereby becoming a marker of psychological adjustment or well being (i.e., higher scores indicate higher well being) This was assessed at each participant's pre breakup and post breakup timepoints. Example items include, "During the last week, I felt dissatisfied with everything," "During the last week, I felt tense or "high strung and "During the last week I felt discouraged." Items are rated on a 5 point scale, ranging from 1 (not at all) to 5 (e xtremely). Scores are the average of the 12 items. The 12 items were chosen to be used in the larger parent study from the overall MASQ based on factor analyses that indicate that they capture a more general psychological distress rather than symptoms spe cific to anxiety or depressive disorders (Keogh & Reidy, 2000; Rhoades, Stanley, & Markman, 2011). The 12 item measure has been shown to be reliable in measuring psychological distress in similar samples ( = .94) demonstrating excellent internal consistency reliability (Wortel & Rogge,

PAGE 24

19 2010). This measure has also shown substantial convergent validity with self report symptom measures of anxious and depressive symptomatology ( Clark & Watson 1991). Frien ds and family approval of the relationship (social pressure) : To assess friends and family approval of the relationship or social pressure from friends and family for the relationship to endure the 6 item social pressure subscale from the Commitment Inve ntory was used (Stanley & Markman, 1992). This was assessed at each participant's pre breakup timepoint. Example items include, "My friends would not mind it if my partner and I broke up (or divorced)" (reverse scored), "My family would not care either way if this relationship ended" (reverse scored), "It would be difficult for my friends to accept it if I ended the relationship with my partner," and "My family really wants this relationship to work." Each item is rated on a 7 point scale, ranging from 1 (s trongly disagree) to 7 (strongly agree). This measure has established excellent internal consistency reliability by demonstrating alphas of .92 and .88 in similar samples (Stanley & Markman, 1992). This measure has demonstrated validity by convergence with theoretically related variables, such as a correlation of .63 (p < .01) with a one item commitment measure and a correlation of .39 (p < .01) with a one item relationship satisfaction measure in similar samples (Stanley & Markman, 1992 ). This measure is s cored by reverse scoring the items that reflect family and friends not caring about or approving of the relationship, and then averaging the items. Higher scores indicate both higher approval and higher social pressure from friends and family for the relat ionship to endure. Desire to marry one's partner: A c ategorical item, "Do you want to marry your current partner?" was used to assess desire to marry one's partner at each participant's pre breakup timepoint. Participants indicated their response on a 4 po int scale from 1 (Yes, I am

PAGE 25

20 sure I want to marry my partner), 2 (Not sure), 3 (No, I do not want to marry my partner), and 4 (I haven't thought about it). This scale will be scored as participants being sure they want to marry their partner (1), or unsure they want to marry their partner (2, 3, or 4). Felt constraint: Felt constraint was measured using three items: "I feel trapped in this relationship but I stay because I have too much to lose if I leave," "I would leave my partner if it was not so difficul t to do" and "I feel stuck in this relationship" at each participant's pre breakup timepoint. Each item is rated on a response scale from 1 (strongly disagree) to 7 (strongly agree). Rhoades et al. (2010) found this measure to have good internal consistenc y reliability in the parent study ( = .82). This measure has demonstrated validity through divergence with theoretically non related measures, such as dedication, and convergence with theoretically related measures, such as relationship adjustment and sta bility (Rhoades et al., 2010). Higher dedication is associated with lower felt constraint (Rhoades et al., 2010). This measure is scored by averaging the items. Higher scores indicate a greater felt constraint. Dedication: Dedication was measured using th e 4 item dedication subscale from the Commitment Inventory (Stanley & Markman, 1992). This was assessed at each participant's pre breakup timepoint. Example items include, "I want this relationship to stay strong no matter what rough times we encounter," "I like to think of my partner and me more in terms of us' and we' than me' and him/her,'" "It makes me feel good to sacrifice for my partner" and "My relationship with my partner is clearly part of my future life plans." Each item was rated on a 1 ( st rongly disagree ) to 7 ( strongly agree ) scale. Rhoades et al. (2010) found the measure to be internally consistent in the parent study ( = .88) and Rhoades et al. (2012) found similar internal consistencies for men ( = .86) and women ( = .87). The dedication scale has also demonstrated validity through convergence with theoretically

PAGE 26

21 related characteristic s such as commitment and relati onship satisfaction (Stanley & Markman, 1992; Owen et al., 2011). This measure is scored by averaging the items. Higher scores are indicative of more dedication in the relationship. Relationship satisfaction: The 4 item version of the Dyadic Adjustment Sca le was used to measure relationship satisfaction (Sabourin, Valois, & Lussier, 2005). This was assessed at each participant's pre breakup timepoint. This scale was shortened from the 32 item Dyadic Adjustment Scale (Spanier, 1976). This measure includes it ems about thoughts about dissolution, a general item about how well the relationship is going, frequency of confiding in one another, and the degree of happiness of the relationship. Response options vary across the four items. Rhoades et al. (2010; 2011) found an alpha of .81 using this same version in the parent study demonstrating internal consistency reliability. The DAS has been shown to correlate well with other marital relationship tests such as a .88 correlation with the Locke Wallace Marital Adjus tment Scale (Graham, Liu, & Jeziorski, 2006). Further, the DAS shows predictive and concurrent validity by predicting likelihood of divorce and consistently distinguishing distressed couples from nondistressed couples (Graham et al., 2006). As suggested by the authors, the scale is scored by summing the items (Sabourin et al. 2005). Higher scores indicate higher relationship quality. Perceived likelihood of marriage: Perceived likelihood of marriage was assessed using a one item measure, "How likely is it that you and your partner will get married?" at each participant's pre breakup timepoint. Participants are asked to indicate their responses on a 5 point Likert scale from "Very unlikely" to "Very likely." This item is based off an item used in the Nationa l Survey of F amilies and Households. Test retest reliability has been shown to be high in research with the parent sample (Rhoades et al., 2010; 2012). Higher

PAGE 27

22 scores indicate a greater perceived likelihood of marriage. Marriage date set: To assess whethe r the participants have set a marriage date or not, one item was given, "Have you and your partner set a date for getting married?" and was assessed at each participant's pre breakup timepoint. The options are "Yes, and it is _____ (mm/dd/yy)" or "No." Pa rtner desired breakup: To assess whether the partner initiated the breakup, one item was given, "Who wanted to end the relationship more?" at each person's post breakup timepoint. Participants are asked to indicate their responses on a 7 point scale from 1 (Me) to 7 (My partner) with 4 being labeled as "Equal." This item was scored dichotomously, such that anyone who answers 5 or above was coded as "1" and anyone who answers below 5 was coded as "0." Therefore, a code of 1 means that the partner wanted to b reakup more than the participant, and a code of 0 means that the partner did not want to breakup more than the participant (i.e., the participant wanted to breakup more than or equal to their partner) Physical aggression from partner : To measure physical aggression in a relationship towards the participant ( from the partner ) two subscales from the Revised Conflict Tactics Scale were used; the minor injuries received subscale and the minor physical aggression by partner subscale (Straus, Hamby, Boney McCo y, & Sugarman, 1996). This was assessed at each participant's pre breakup timepoint. The minor injuries received subscale (e.g., "I had a sprain, bruise, or small cut because of a fight with my partner") and the minor physical aggression by partner (e.g., My partner pushed or shoved me were used to create a scale of 20 items for participants to indicate if this never happened (0) or to indicate ( on a scale of 1 7 ) the frequency of the action ranging from "not in the past year, but it did happen before" to "more than 20 times in the past year. Straus et al. (1996) found the minor injuries scale to

PAGE 28

23 have an alpha of .95 and the minor physical aggression by partner scale to have an alpha of .86 in a similar sample, indicating good internal consistency relia bility. The scales also demonstrated good validity by convergence with theoretically non related constructs, such as negotiation. The higher the level of negotiation in a couple, the less they report injuries or aggression (Straus et al., 1996). Participan ts who indicated that they had never sustained injuries due to a fight with their partner and had never been the recipient of physical aggression from their partner were coded as having no history of physical aggression towards them in the relationship Th ese participants were coded as "0" for this scale. Participants who reported any of these behaviors were coded as having a history of physical aggression towards them at some point in their relationship. These participants were coded as "1" for this scale. Psychological aggression from partner : To measure psychological aggression towards the participant ( from the partner ) in a relationship, the minor psychological aggression received subscale from the Revised Conflict Tactics Scale was used (Straus, Hamby, Boney McC oy, & Sugarman, 1996). This was assessed at each participant's pre breakup timepoint. The minor psychological aggression received subscale includes four items: "My partner insulted or swore at me," "My partner shouted or yelled at me," "My partne r stomped out of the room or house or yard during a disagreement," and "My partner said something to spite me." Participants are asked to indicate if this never happened (0) or to indicate ( on a scale of 1 7 ) the frequency of the action ranging from "not in the past year, but it did happen before" to "more than 20 times in the past year Straus et al. (1996) found the minor psychological aggression scale to have an alpha of .79 in a similar sample, indicating sufficient internal consistency reliability. I n another study using the parent sample the !s for

PAGE 29

24 the psychological aggression subscale were .77 and .82 for men and women, respectively (Rhoades, Stanley, Markman, 2009). This measure is scored by averaging the items. Higher scores indicate a greater prevalence of psychological aggression re ceived. Partner i nfidelity: Infidelity by the participant's partner was assessed using one item at each participant's pre breakup and post breakup timepoints. The pre breakup item is, "Has your partner had sexual relations with someone other than you since you began seriously dating?" The post breakup item is, "Did your ex partner have sexual relations with someone other than you while you were dating?" The response options are "No," "Probably not," "Probably so" and "Yes, I know for sure." Higher scores in dicate a greater certainty that the participant's partner was unfaithful at either pre breakup or post breakup timepoints Additional measures used for descriptive purposes: Length of engagement: The time in days of the length of the engagement is a cal cul ated variable. This variable was created by subtracting the reported breakup date from the date of the engagement. There was only enough data to calculate this for 5 7 (7 7 %) of the participants. When this variable is used with only participants who experien ced a drop in well being (see post hoc section, below), this per centage decreases slightly (76 %). The initial range was 25 3737 days, with a mean of 591 .80 days (SE = 85 55 ) and a median of 363 days. Due to the high level of skew, two extreme values ( gre ater than or equal to 2,553 days ) were re cod ed into the next highest value (1,460 days) to represent an outer value of 1,460 days or greater. This recoding dropped the skew to an acceptable level. T he median for the recoded value remained at 363 days and the mean became 53 6 74 (SE = 5 6 07 ).

PAGE 30

25 Time since breakup: The time in days since the broken engagement is a calculated variable. This variable was created by subtracting the reported breakup date from the date of the post survey. Due to missing dates on some survey packets, this variable could only be calculated for 93 % of the sample. When used with only participants who experienced a drop in well being (see post hoc section, below), this percentage decreases slightly (92%). Reasons for breakup : To ascer tain a participant's reason(s) for the breakup respondents were asked "Why did the relationship end? (Fill in all that apply)." The options that participants could choose to endorse were, "(1) substance abuse issues, (2) religious differences, (3) dom esti c violence, (4) economic hardship, (5) infidelity, (6) too much arguing, (7) lack of commitment, and (8) sexual incompatibility." This was given at each participant's post breakup timepoint. Percentages of each reason endorsed will be given for each trajec tory of well being as a way to further identify these groups. The results of this variable are displayed in Table 5. Data Analysis The three aims of my thesis were to: (a) determine the average, overall trajectory of self report well being from pre broken engagement to post broken engagement, (b) see if there are distinct well being trajectory groups that emerge, and (c) understand how these groups differ on certain pre breakup and/or post breakup timepoint variables. I tested my first hypothesis with a s imple paired t test. I addressed my second aim (and provide d further evidence regarding the first aim) with Mplus using Growth Mixture Modeling (GMM). Given that I was able to proceed to Aim 3 I addressed this aim using either an ANOVA or Chi Square analy sis for each hypothesized distinguishing variable (see Table 2, below).

PAGE 31

26 Table 2. Hypothesized analyses for distinguishing vari ables of well being trajectories Measures #1 Red/ Stable high well being #2 Blue / High to low well being #3 Purple / Low to high well being #4 Green / Stable low well being Analyses for each hypothesis: Friends/famil y approval (social pressure) at pre breakup N/A High Low N/A ANOVA: group 2 vs. group 3 Desire to marry partner at pre breakup Unsure Sure Unsure N/A Chi Square: 2 options ( sure, unsure) x 3 groups Felt constraint at pre breakup Low Low High High ANOVA: post hoc grou ps 3 and 4 higher than group s 1 and 2 Dedication at pre breakup Low High Low Low ANOVA: post hoc group 2 higher than group s 1, 3, and 4 Relationship satisfaction at pre breakup High High Low Low ANOVA: post hoc group s 1 and 2 higher than group s 3 and 4 Perceived likelihood of marriage at pre breakup Low High N/A N/A ANOVA: group 2 higher than group 1 Marriage date set at pre breakup N o Yes Yes No Chi Square: 2 options (Yes, No) x 4 groups Partner desired the breakup at post breakup No Yes No Yes Chi Square: 2 options (Yes, No) x 4 groups Physical aggression ( from partner ) at pre breakup No No No Yes Chi Square: 2 options (Yes, No) x 4 groups

PAGE 32

27 Measures #1 Red/ Stable high well being #2 Blue / High to low well being #3 Purple / Low to high well being #4 Green / Stable low well being Analyses for each hypothesis: Psychological aggression ( from partner ) at pre breakup Low Low Low High ANOVA: post hoc group 4 higher than groups 1 2, and 3 Partner infidelity at pre breakup Low Low Low High ANOVA: post hoc group 4 higher than group s 1, 2, and 3 Partner infidelity at post breakup Low High Low High ANOVA: post hoc groups 2 and 4 higher than group s 1 and 3

PAGE 33

28 CHAPTER III RESULTS Prior to hypothesis testing I analyzed basic descriptive and reliability statistics for all possible distinguishing variables (see Table 3 ). Results indicate that all internal cons istencies were acceptable, when relevant. Table 3 Initial overall results for possible distinguishing variables Mean (SD) Percentages Reliability ( ) Well b eing at p re breakup (N=74) 3.69 (.95) N/A .92 Well b eing at p ost breakup (N=74) 3.19 (1.16) N/A .95 Friends/family approval (social pressure) at pre breakup (N=73) 4.26 (1.43) N/A .82 Desire to marry partner at pre breakup (N=74) N/A 1.4% No I do not want to marry my partner 79.7% Yes I am sure I want to marry my partner 18.9% I am not sure if I want to marry my partner N/A Felt constraint at pre breakup (N=74) 2.20 (1.45) N/A .88 Dedication at pre breakup (N=74) 5.60 (1.33) N/A .86 Relationship satisfaction at pre breakup (N=71) 15.39 (3. 93) N/A .80 Perceived likelihood of marriage at pre breakup (N=73) N/A 1.4% Very unlikely 13.7% Neutral 20.5% Likely 64.4% Very likely N/A Marriage date set at pre breakup (N=74) N/A 60. 8% Marriage date set 39.2% No marriage date set N/A

PAGE 34

29 Mean (SD) Percentages Reliability ( ) Partner desired breakup at post breakup (N=71) N/A 67.6% Participant wanted the breakup more 32.4% Partner wanted the breakup more N/A Physical a ggression (from partner) at pre breakup (N=73) N /A 49.3% No history of physical aggression 50.7% History of some physical aggression .84 Psychological aggression (from partner) at pre breakup (N=72) 2.57 (1.89) N/A .79 Partner infidelity at pre breakup (N=74) N/A 55.4% No 13.5% Probably not 9. 5% Probably so 21.6% Yes I know for sure N/A Partner infidelity at post breakup (N=70) N/A 32.9% No 11.4% Probably not 22.9% Probably so 32.0% Yes I know for sure N/A T ime s ince breakup in days (N=69) 72.26 (57.55) Median: 61.00 Range: 1 230 N/A N/A L ength of engagement in days (recoded) (N= 57) 536.74 (423.32) Median: 363.00 Range: 25 1460 N/A N/A Next, I ran correlations among key variables (see Table 4, below). Variables appear to converge as expected (e.g., dedication is highly corr elated with relationship satisfaction) with a few exceptions (e.g., partner infidelity is unrelated to relationship satisfaction at pre breakup ). Table 4 Correl ation matrix with all relevant variables 1. 2. 3. 4. 5. 6. 7. 8. 1. Partner infidelity pre 2. Partner infidelity post .58** 3. Relationship satisfaction pre .08 .23

PAGE 35

30 1. 2. 3. 4. 5. 6. 7. 8. 4. Felt constraint pre .05 .25* .64** 5. Likelihood of marriage pre .22 .08 .66** .50** 6. Dedication pre .02 .20 .70** .75** .51** 7. Psychological aggression (from partner) pre .28* .00 .38** .28* .27* .30* 8. Friends/family approval (social pressure) pre .21 .06 .49** .34** .61** .45** .34** 9. Length of engagement (recoded) .08 .04 .25 .31* .41** .06 .12 .18 Hypoth esis 1 My first hypothesis was that the average overall trajectory of well being, as measured by psychological adjustment, for pre breakup and post breakup, within people who have experienced a broken engagement, will show a decrease from before the broken engage ment to after the broken engagement. I tested th is hypothes i s with a paired t test. Results of the paired t test confirmed that there is a significant drop between pre broken engagement levels of well being and post broken engagement levels of well being (t 1,73 = 4.85, p < .00 1 ). Hypothesis 2 My second hypothesis was that in addition to the average declining trajectory of well being, three additional distinguishable trajectories of well being from pre breakup to post breakup (steady moderate/high well being, steady low well being, improving well being) will emerge within the sample.

PAGE 36

31 GMM is a method for identifying both the overall trajectory and unobserved subgroups. Although I hypothesized four distinct trajectories, I chose GMM due to this model 's ability to describe possible latent trajectories via longitudinal change in a post hoc manner. That is, GMM allows latent trajectories to emerge as opposed to identifying the trajectories myself. Thus, I used exploratory GMM models with the amount of cl asses set at 3, 4, 5, and 6 to ensure I chose the number of trajectories that best fit the data. I used Mplus version 7.3. B etween 200 and 1000 random starts were used, the best log likelihood value was replicated, and the model estimat ion terminated norma lly for the models run with number of classes set at 3, 4 or 5. For analyses with the number of classes set at 6, the model did not terminate normally and the best log likelihood value was not replicated (see below). The model fit results of each model a re as follows: 3 classes: e ntropy = .8 1 log likelihood value = 181.9 6 AIC = 385.91 BIC = 411.2 6 s ample size adjusted BIC = 376.59 4 classes: e ntropy = .82 log likelihood value = 177.45 AIC = 382.90 BIC = 415.1 6 sample size a djusted BIC = 371.0 4 5 classes: e ntropy = .8 4 log likelihood value = 173.35 AIC = 380.7 1 BIC = 419.8 8 sample size a djusted BIC = 366.30 6 classes: This model did not terminate normally due to misfit between parameters and degrees of freedom. The best log likelihood v alue was not replicated. Thus this model is not plausible in the data. When a Chi Square difference test is not relevant to decide on best model fit, as is the case with GMM, using a combination of statistical information such as Akaike's Information

PAGE 37

32 Cri teria (AIC ; Akaike, 1987) and Bayesian Information Criteria (BIC ; Schwartz, 1978), as well as agreement with substantiv e theory, is the best approach (Nylund, Asparouhow & Muth ÂŽ n, 2007). The BIC and AIC have meaning only in comparison to other BIC and A IC values; the smaller the numbers, the better the model fit Entropy is another indication of model fit. Given an estimated model, estimated posterior probabilities for each individual and each class are produced. Individuals are classified into the class with the highest probability. The classification quality is summarized into the class with the highest probability, and this classification quality is summarized in an entropy value ranging from 0 1 with 1 corresponding to the situation where all individua ls have 100% probability for one class and 0% p robability for the other classes ( MuthÂŽn Brown, Leuchter & Hunter, 2010). In other words, entropy is a measure of confidence in class assignment (Grimm et al., 2010). Thus, the closer to 1 the entropy value is, the better. Given that there is no real a priori theory for trajectories of well being following a broken engagement, this criteri on became less relevant for me. According to the AIC values, sample size adjusted BIC values, and entropy values, th e res ults show that 5 classes fit the data better tha n 3, 4 or 6 classes. Thus, my hypotheses regarding the four specific trajectories from pre breakup to post breakup (declining well being, steady moderate/high well being, steady low well being, improving wel l being) was not supported exactly as predicted. However, different trajectories did emerge. The 5 trajectories, or classes are summarized below By specifying the estimator function as MLR within MPlus, a maximum likelihood estimator with robust standard errors, using a numerical integration algorithm, was used for missing data. That is, MPlus used all data that w ere available to estimate the model using full information

PAGE 38

33 maximum likelihood. Thus, each parameter is estimated directly without first filling in missing data values for each individual. The following class information is what this model yields (r ecall that MASQ was reverse scored to reflect well being ; p value associated with slope) : Group (c lass ) 1 : Initial moderate well being, significant dr op = 8 people Average pre breakup well being score = 4. 32 ( SD = .51 ), Average post breakup well being score = 2. 13 ( SD = .63 ), Intercept = 4 32 ; S lope = 2. 19 p < .00 1 Group (c lass ) 2 : Initial moderate well being, significant increase = 17 people Aver age pre breakup well being score = 3.8 7 ( SD = .52 ), Average post breakup well being score = 4.3 5 ( SD = .41 ), Intercept = 3 .8 7 ; Slope = .48, p < .00 1 Group (c lass ) 3 : Initial moderate well being, significant drop = 16 people Average pre breakup well bein g score = 3.1 9 ( SD = .68 ), Average post breakup well being score = 2.6 3 ( SD = .41 ), Intercept = 3 .1 9 ; S lope = .56, p < .00 1 Group (class) 4 : Initial moderate well being, significant drop = 25 people Average pre breakup well being score = 4.40 ( SD = .64 ), Average post breakup well being score = 3. 70 ( SD = .72 ), Intercept = 4 .40; S lope = .70, p < .00 1 Group (class) 5 : Stable low well being = 8 people Average pre breakup well being score = 1.69 ( SD = .50 ), Average post breakup well being score = 1.59 ( SD = .48 ), Intercept = 1 .69; S lope = .10, p = .55

PAGE 39

34 Figure 2. Trajectory group results of GMM The results of the GMM analyses provide further support for Hypothesis 1, in that most people fell into a group characterized by a higher level of well bei ng at pre breakup and a lower level of well being at post breakup (N= 49, 66.2% of the sample, blue, purple, and green lines in Figure 2 above, groups 1, 3, 4). There was mixed support for Hypothesis 2. As noted above, GMM did not reveal the four hypothesi zed trajectories specifically; however, the groups that emerged did represent three of the proposed trajectories. As hypothesized, there was a stable low well being group and an increasing well being group. There were three groups that represented the hypo thesized decreasing well being from pre to post breakup, each with different intercepts and slopes. The only hypothesized trajectory that did not emerge in this sample was a stable high well being group. !"# $ $"# % %"# & &"# # Pre Breakup Well Being Post Breakup Well Being N = 8, Group 1 N = 17, Group 2 N = 16, Group 3 N = 25, Group 4 N = 8, Group 5

PAGE 40

35 Interpretation of Well Being Scores Our understa nding of these groups can be improved by considering how to interpret the levels of the MASQ and degree of change of the MASQ Few have attempted to provide clinical utility or cut off scores for the MASQ. Findings by Buckby et al. (2007) would suggest th at a score below 2.55 on the MASQ ( when reverse scored to reflect well being ) would represent possible depression (see Figure 3) Whereas Bucky et al. (2007) used the larger MASQ scale, the 12 MASQ items used in this study were chosen to be inc luded in the larger parent study based on fa ctor analyses that indicate that they capture a more general psychological distress rather than symptoms specific to anxiety or depressive disorders (Keogh & Reidy, 2000; Rhoades, Stanley, & Markman, 2011). Nonetheless, the cut off score suggested by Buckby and colleagues may provide a potential indicator of the level of distress (or lack thereof) that these participants may be experiencing at their pre and post bro ken engagement timepoints. Group 5 falls in this range at bot h pre (1.69) and post breakup (1.59) and group 1 falls in this range at post breakup (2.13) Thus, 10.8% of this sample is in a range of possible clinical or diagnosable distress at pre breakup, and 21.6% of this sample is in a range of possible clinical or diagnosable distress at post breakup (see Figure 3) Using this potential cut off, the majority of people who experienced a broken engagement in this study were not experiencing a clinical or diagnosable level of distress at either pre breakup or post b reakup. Jacobson, Follette, and Revenstorf (1984) initially defined clinically significant change as the extent to which therapy moves someone outside the range of the dysfunctional population or within the range of the functional population. Jacobson and Truax (1991) operationalize this definition and provide both the Reliable Change Index (RCI), and criteria

PAGE 41

36 for clinical significance. The four groups that the GMM analysis identified as having statistically significant change between pre and post would al so be considered reliable change via the RCI, meaning that a change of that magnitude would not be expected due to the unreliability of the measure. However, following the criteria given by Jacobson and Truax (1991), the only group who experienced clinical ly significant change was group 1. Figure 3. Possible clinical cut off for levels of well being Note: Levels of well being in the shaded area, at or below 2.55, may be indicative of clinical or diagnosable distress. Hypothesis 3 My third hypothe sis was contingent upon the second hypothesis, in that if the hypothesized trajectories emerged, they would show differences in the variables delineated in Table 2 : pre breakup levels of variables such as level of friends' and family approval of the relati onship (social pressure) desire to marry the partner, felt constraint, dedication, relationship satisfaction, perceived likelihood of marriage, presence of a set marriage date set, !"# $ $"# % %"# & &"# # Pre Breakup Well Being Post Breakup Well Being N = 8, Group 1 N = 17, Group 2 N = 16, Group 3 N = 25, Group 4 N = 8, Group 5

PAGE 42

37 repor ts of psychological aggression from partner reports of physical aggr ession from partner and reports of infidelity by the partner, and post breakup levels of perception of partner relative (to self) desire for the breakup and reports of infidelity by the partner. Not all variables were hypothesized to be relevant for all g roups. I proposed to use a series of ANOVAs for the continuous variables (i.e., friends and family approval (social pressure) felt constraint, dedication, relationship satisfaction, perceived likelihood of marriage partner infidelity at pre and post an d psychological aggression ) and a series of Chi Square tests for the categorical variables (i.e., desire to marry partner, physical aggression, marriage date set, partner desired breakup) to understand how the variables significantly vary between the grou ps. See the "analyses for ea ch hypothesis" column in Table 2 above. When I proposed my thesis there were a few important contributing factors that were still in flux, including my final sample size, the number /type of trajectories that would emerge, an d the number of people that would populate each trajectory that emerged. Based on the GMM results, the specific trajectories that emerged were not a perfect match to the ones I hypothesized. Further, the Ns for the groups were fairly small (two groups with 8 people) and thus power for ANOVAs and Chi Square analyses across all five groups was unacceptably low. There was also considerable variability in the size of the groups (range of 8 to 25 people). Analytically I could not run the proposed analyses with the five trajectories found, because the basic test assumptions would have been violated (i.e., several cells in Chi Square tests did not have expected cell count >5, lack of homogeneity of variance and unequal number in each group for ANOVA tests) Thus, comparing all five groups statistically in the manner proposed for Hypothesis 3 was inappropriate. Further, collapsing

PAGE 43

38 the three "declining" groups into one group was not appropriate given the GMM findings that these were three statistically distinct traje ctories. However, regardless of the planned statistical tests, I had still painted a hypothesized conceptual picture of each group based on the measured variables. Thus, d espite my lack of ability to compare the groups statistically, I chose to proceed in a few different ways in an attempt to still characterize and understand the people who fell into the different well being trajectory groups. I first attempted a rough narrative, based on visual examination of the means or percentages regarding the conver gence or lack of convergence with my hypothesis (see Appendix A ). However, this approach became too problematic, given that results indicated more groups than I initially hypothesized and the difficulty in objectively quantif ying what "high" or "low" mean t in relation to the obtained group means or how much would represent a meaningful level of relatively "higher" or "lower" among the groups Therefore, I instead chose to highlight the patterns unique characteristics and what stood out to me among each trajectory group or relevant clusters of groups. This is based on my subjective review of the results in Table s 5 and 6 I will then present some post hoc analyses to statistically predict the degree of drop in well being for the three trajectory groups w ho all experienced a drop in well being from pre broken engagement to post broken engagement. Trajector y groups 1,3,4: Drop in well being from pre to post broken engagement When I was originally conceptualizing possible trajectory groups, I thought that t here would be a group who was doing relatively well at pre breakup, and then would experience a drop in well being at post breakup. Trajectories 1, 3 and 4 are all groups of people who experienced a drop in well being from pre broken engagement to post bro ken engagement

PAGE 44

39 Although three trajectories emerged in place of the one that I hypothesized, there were certain themes present among all three groups that suggest that, compared to groups 2 and 5, these three groups were doing relatively well at pre breaku p and thus, their drop in well being may be a sign that they were relatively blind sided by the breakup Evaluating the results in Table 6, it appears that these three groups c ompared to groups 2 and 5, had relatively higher overall pre breakup desire to marry their partner, likelihood of marriage, and friends and family approval ( social pressure ) for their relationship to work They also seemed to have the lowest rates of pre breakup infidelity and they completed their post breakup surveys at the shortest amount of time since their breakup perhaps representing less "recovery time" after breakup It appears that group 1 looks most like my hypothesized declining group. Group 1 experienced the greatest drop in well being from pre to post breakup their aver age post breakup level of distress falls in the possible range of clinical or diagnosable distress and their drop in well being was clinically significant T heir relatively higher level of well being at pre breakup may be supported by the following: they seemed to have the highest level of pre breakup relationship satisfaction were the only group who had every single participant indicate a full desire to marry their partner and seemed to experience the lowest pre breakup amount of psychological and physi cal aggression from their partner Considering their seemingly positive relationships, t hey may then have been surprised or blind sided by the breakup or certain events that may have contribut ed to their drop in well being. For example, they appeared to en dorse the highest percentage (along with group 5, stable low) of people reporting, "yes I know for sure" that their partner was unfaithful at breakup

PAGE 45

40 Results from the reasons for breakup variable (see Table 5) provide further potential evidence that poss ible negative circumstances surrounding their breakup may not have been present throughout their relationship ( and thus the breakup or events such as infidelity may have come as a shock to the participants ) For example, this was the only group where 0% en dorse d substance abuse, domestic violence, economic reasons, and sexual incompatibility and they also had the lowest percentage of people endorsing arguing and lack of communication (along with group 5, stable low) as reasons for breakup. These are all fa ctors which, had they been present in the re lationship, may have given the participants some indication that their relationship wasn't doing well at the pre assessment Furthermore th ey had the highest percentage of people endorsing "partner didn't want t o marry" on the reason s for breakup variable which may be a nother possible indication that this breakup was unexpected and/or particularly hurtful or distressing However, it should be noted that they also had the highest ratings on "participant (self) di dn't want to marry" on the reasons for breakup variable given the pre levels of relationship positives, this high level may have emerged from upsetting and unforeseen events such as infidelity. Trajectory group 2 : Increase in well being from pre to post b roken engagement Group 2 is the only group of people who experienced an increase in well being from pre to post breakup. I had initially hypothesized that this group may have experienced their broken engagement as a re lief because they may have experience d markers for distress in their relationship at pre breakup. Results (see Table 6) indicate there were indeed markers for distress in their relationship at pre breakup such as appearing to have the lowest percentage of people who had a marriage date set a nd being the only group where some participants indicated that they did not wa nt to marry their partner (5.9%) at pre breakup Further, they

PAGE 46

41 reported the second lowest level of likelihood of marriage, dedication, and social pressure from friends and family Thus, they may have been somewhat aware of various negative aspects of their relationship, and therefore their broken engagement may have been somewhat of a relief Results from the reason for breakup variable (see Table 5) may provide further evidence t hat this group may have felt relief from getting out of a potentially bad relationship and were not experiencing factors that may have prolonged their low well being even after the breakup For example, this group had the highest percentage of people endor sing economic reasons and lack of communication as reasons for breakup. These reasons may have been long standing and were likely not a one time event. T hus the participants may have felt relief when the breakup happened, but the dynamics in their relatio nship weren't so extreme (i.e., substance abuse, infidelity, physical or psychological abuse) that their impact lasted well beyond the breakup, as was perhaps the case with the stable low group (group 5). This group, on average, completed their post break up surveys in the longest amount of time since breakup which could suggest that this increase in well being may not be an initial increase (e.g., relief) but may in fact be the result of greater time for recovery They were also engaged, on average, for t he shortest amount of time Trajectory group 5: Stable low well being from pre to post broken engagement Group 5 is the only group of people who experienced a low level of well being at both pre and post broken engagement. Both of their pre breakup and p ost break up levels of distress we re in the possible range of clinical or diagnosable distress. I had initially hypothesized that this was a group of people who have had had markers for distress in their relationship, but they may not experience an increase in well being following avoiding a

PAGE 47

42 presumably bad situation because t hey may still have had other distressin g factors in their life that were contributing to their well being. Results (see Table 6) indicate that this group's somewhat chronic distress may be evidenced by having the highest percentage of participants who reported at pre breakup that, "yes I know for sure" that their partner was unfaithful were unsure if they wanted to marry their partner and who reported psychological aggression from their partner. They also had the lowest perceived likelihood of marriage and friends and family approval (social pressure) for their relationship to work and the second lowest level of relationship satisfaction. However although they got out of a seemingly ba d relationship or environment, they still may have had factors that got in the way of their well being increasing after breakup, such as having the highest percentage of people who reported "yes I know for sure" that their partner was unfaithful at post b reakup (along with group 1) and whose partner desired the breakup. Results from the reason for breakup variable (see Table 5) provides further possible evidence for their on going difficult situation that may contribute to their low well bein g at both pre and post breakup: they had the highest percentage of people endorsing substance abuse, domestic violence, and infidelity as reasons for breakup all events which may intensify a drop in well being due to the devaluing aspect of the event. Specifically, th e work done by Cano and O'Leary (2000) suggest that the humiliating events of infidelity, violence, or partner initiated breakup may especially prolong a breakup, even if getting out of the relationship is a positive thing in the long run. This group, on a verage, completed th eir post breakup surveys in the second longes t amount of time since breakup and were engaged for the longest amount of time This could add further evidence to the idea that this low well being is somewhat chronic and although they may be getting out of a potentially bad

PAGE 48

43 relationship, other distressing life factors may still be contributing to their persistent low well being. Table 5 Percentages of people endorsing different reasons for breakup within the 5 groups Reasons for Breakup // Trajectories 1 ( Initial moderate well being, significant drop ) 8 p eople 2 ( Initial moderate well being, significant increase ) 17 p eople 3 ( Initial moderate well being, significant drop ) 16 p eople 4 ( Initial moderate well being, significant dr op ) 25 p eople 5 (Stable low well being) 8 p eople Substance a buse 0% 11.8% 6.3% 20% 62.5% Religious r easons 12.5% 5.9% 6.3% 16% 12.5% Domestic v iolence 0% 5.9% 6.3% 4% 12.5% Economic reasons 0% 29.4% 25% 16% 0% Infidelity 37.5% 35.3% 25% 36% 62.5 % Arguing 37.5% 47.1% 68.8% 68% 62.5% Lack of c ommunication 25% 64.7% 31.3% 32% 25% Sexual i ncompatibility 0% 17.6% 25% 16% 37.5% Partner didn't want to marry 25% 23.5% 0% 12% 12.5% Participant (self) didn't want to marry 12.5% 11.8% 6.3% 12% 12.5%

PAGE 49

44 Table 6 Results of proposed distinguishing variables within the 5 groups 1. (Initial moderate well being, significant drop) 8 p eople 2. (Initial moderate well being, significant increase) 17 p eople 3. (Initial moderate well being, si gnificant drop) 16 p eople 4. (Initial moderate well being, significant drop) 25 p eople 5. (Stable low well being) 8 p eople Partner infidelity at pre breakup No = 75% For sure = 12.5% No = 47.1% For sure = 29.4% No = 56.3% For sure = 12.5% No = 64% For sure = 16% No = 25% For sure = 50% Partner infidelity at post breakup No = 12.5% Yes I know for sure = 37.5% No = 29.4% Yes I know for sure = 35.3% No = 50% Yes I know for sure = 18.8% No = 32% Yes I know for sure = 32% No = 12.5% Yes I know for sure = 37.5% Partner desired breakup at post breakup 50% partner desired breakup 29.4% partner desired breakup 18.8% partner desired breakup 24% partner desired breakup 62.5% partner desired breakup Marriage date set at pre breakup 37.5% date set 35.5% date set 37.5% date set 44% date set 37.5% date set Relationship satisfaction at pre breakup Range 9 21 Mean=16.86 Range 5 21 Mean=15.05 Range 3 19 Mean=13.93 Range 8 21 Mean = 16.24 Range 9 19 Mean = 14.75 Felt constraint at pre breakup Range 1 3.67 Mean = 1.67 Range 1 6 Mean=2.43 Range 1 6.33 Mean=2.96 Range 1 3.67 Mean = 1.63 Range 1 5 Mean= 2.54 Desire to marry partner at pre breakup 100% have full desire to marry 70.6% have full desire to marry, 23.5% are unsure, 5.9% do not want to marry partner 75% h ave full desire to marry, 25% are unsure 88% have full desire to marry, 12% are unsure 62.5% full desire to marry, 37.5% are unsure Perceived l i kelihood of marriage at pre breakup 62.5% very likely 47.1% very likely 81.3% very likely 68% very likely 37.5% very likely Dedication at pre bre akup Range: 3.14 7 Mean = 5.70 Range: 1.75 7 Mean=5.37 Range 1.12 6.75 Mean=5.09 Range 4 7 Mean = 6.03 Range 3.13 7 Mean = 5.67

PAGE 50

45 1. (Initial moderate well being, significant drop) 8 p eople 2. (Initial moderate well being, significant increase) 17 p eople 3. (Initial moderate well being, si gnificant drop) 16 p eople 4. (Initial moderate well being, significant drop) 25 p eople 5. (Stable low well being) 8 p eople Psychological aggression (from partner) at pre breakup Range: 0 4 Mean = 2.09 Range: 0 6. 25 Mean=2.44 Range: .5 5.75 Mean = 3.02 Range 0 6 Mean = 2.23 Range 0 7 Mean = 3.57 Friends/family approval (social pressure) at pre breakup Range: 2.25 6.75 Mean = 4.06 Range: 1 7 Mean = 3.98 Range: 3 6.75 Mean = 4.71 Range 2 7 Mean = 4.48 Range 1.75 5.5 Mean = 3.53 Physical aggression (from partner) at pre brea kup 37.5% had some aggression 58.8% had some aggression 62.5% had some aggression 40.0% had some aggression 50.0% some aggression Average time since breakup in days 39.63 days Range 1 116 Median = 16.50 101.59 days Range 16 230 Median = 79.00 80.00 days Range 6 152 Median = 93.00 50.00 days Range 1 152 Median = 36.00 100.29 days Range 1 142 Median = 124.00 Average length of engagement in days (recoded) 722.17 days Range 145 1288 Median = 715.00 411.00 days Range 25 1460 Median = 292.00 515.40 days Range 76 1289 Median = 350.00 530.24 days Range 94 1460 Median = 497.00 995.00 days Range 599 1406 Median = 980.00 Post Hoc Analyses (predicting level of drop in well being amo n g those who experienced a drop in well being) : In order to determine if certain variables are related to the amount of drop in well being from pre to post breakup I isolated the members of the three groups who all experienced a drop in well being (N = 49) and create d a variable which reflected the amount of drop with higher scores indicating a greater drop in well being from pre breakup to post breakup. I then ran analyses with this "well being drop" variable as the dependent variable and each individual hypothesize d variable as the predictor variable in separate regressions.

PAGE 51

46 Using G*power for a linear regression with one predictor, in order to determine the sample size needed to detect medium effects (.15, Cohen's f 2 ), the required sample size is 89. Therefore, my sample of 49 people who all experienced a drop in well being was not large enough to detect small or medium effects. The required sample size needed to detect large effects (.35, Cohen's f 2 ) is 40. Therefore, if large effects exist in my data, I have adequ ate power to detect them. Results indicated that none of the variables examined significantly predicted the amount of drop in well being Specifically, drop in well being was not significantly predicted by pre breaku p l evels of partner infidelity, relatio nship satisfaction, felt constraint, perceived likelihood of marriage, psychological aggression from partner, dedication, or friends/family approval ( social pressure ) Further, the two calculated variables, time since breakup and length of engagement, did not significantly predict a drop in well being. However, these two variables should be interpreted with caution due to the amount of missing data for each. Finally, p artner infidelity at post also did not significantly predict a drop in well being. Followi ng Cohen's effect size guidelines for correlation coefficients (.5= large, .3 = medium, .1= small), it is clear from the correlation matrix below (Table 7) that there are a few variables that had a small effect on the level of dro p in well being (see the f irst column in Table 7). These small effects did not result in significant results due to low power. Variables with small effects include: partner infidelity at post breakup ( r = .1 6 ), pre breakup levels of relationship satisfaction ( r = .12), felt constra int ( r = .13), perceived likelihood of marriage ( r = .16), and friends / family approval ( social pressure ) ( r = .19) as well as time since breakup (r = .2 4 ).

PAGE 52

47 Table 7. Correlations between amount of drop in well being and other variables for people who had a drop in well being from pre to post broken engagement A mount of drop in well being (higher scores indicat e greater drop) .09 Partner i nfidelity at p re breakup .1 6 Partner i nfidelity at p ost breakup .12 Relationship s atisfaction at pre breakup .13 Felt c onstraint at pre breakup .16 Perceived l ikelihood of m arriage at pre breakup .01 Psychological a ggression ( from p artner ) at pre breakup .01 Dedication at pre breakup .19 Friends/family approval (social pressure) at pr e breakup 2 4 T ime s ince b reakup (days) 08 L ength of e ngagement (days) *None of these correlations were significant.

PAGE 53

48 CHAPTER IV DISCUSSION Engagement is often thought of as an idealistic phase in a relationship, and yet it has been estimated that 15 20% of engagements are broken every year (Safier, 2003). Although "cold feet" before marriage is a common lay idea, research into the outcomes and associated factors of broken engagements is f ar less common. This study was the first to l ook at potential changes in individuals' well being from an engaged status (Time 1/pre breakup timepoint) to a broken engagement status (Time 2/post breakup timepoint). Based on findings on unmarried breakup s from Rhoades et al. (2011), I first predicted t hat there would be a significant decline in level of well being from pre breakup to post breakup for people who have experienced a broken engagement, and results of a simple paired t test supported this prediction. Results revealed that 66.2% of the sample experienced a significant decline in well being from pre to post broken engagement. None of the variables examined in this study significantly predicted the degree of drop in well being, which may be an issue of low power (see Limitations). My second hyp othesis was that, in addition to the average declining trajectory of well being, three additional distinguishable trajectories of well being from pre breakup to post breakup (steady moderate/high well being, steady low well being, improving well being) wou ld emerge. Results of a Growth Mixture Model analysis revealed five distinct trajectories, representing three of the four hypothesized trajectories including a stable low well being group (10.8%), an increasing well being group ( 23 %), and three distinct d eclining well being trajectories, with different slopes and intercepts The only hypothesized trajectory that did not emerge was the steady moderate/high well being. Although 66.2% of the sample

PAGE 54

49 experienced a drop in well being, only one of the declining t rajectory groups (10.8%, group 1) experienced clinically significant change in their pre to post broken engagement well being. In other words of those who experienced a decline in well being from pre to post breakup, 16.3% experienced a clinically signifi cant drop. Using a clinical cut off sc or e suggested by Buckby et al. (2007), 10.8% of the entire sample was in the range of possible clinical or diagnosable distress at pre breakup, and 21.6% of this sample was in the range of possible clinical or diagnos able distress at post breakup. Thus, the majority of people who experienced a broken engagement were not in a clinical or diagnosable range of distress at either pre or post breakup. Further, it appears that half of the people who experience d clinical dist ress after a broken engagement were already experiencing clinical distress before the breakup. In addition, h umiliating relationship events involved with the breakup, as detailed by Cano et al., (2000), appear to influence one's level of well being after a broken engagement as well. The 21.6% of the sample who experienced clinical distress at post breakup were also those who, on average, did not desire the breakup and who reported the highest level of partner infidelity at post breakup. My third hypothesi s was to assess what characterizes individuals in the different trajectories of self report well being before and after a broken engagement. Although statistical analyses were not appropriate due to the number and unequal composition of the trajectories, a nd the small number of people in some trajectories, there were various descriptive ways in which the conceptual story of the different trajectories emerged. Declining well being trajectory This declining well being trajectory can be conceptualized as a gr oup of people who were doing relatively well at pre breakup, and then experienced a drop in well being at post

PAGE 55

50 breakup. Examples of "doing well" at pre breakup include relatively high levels of pre breakup relationship satisfaction, desire to marry their p artner and dedication. Given their seemingly high well being at pre breakup, the breakup may have represented a relatively abrupt and traumatic loss or rupture. For example, at post breakup, the group with the greatest declines also had the highest percen tage (along with the stable low group) of people endorsing that their partner was unfaithful. However, it is also true that this group completed their post breakup survey in the shortest amount of time since their breakup, perhaps representing less "recove ry time" from the breakup compared to other groups who had higher levels of well being at post broken engagement. Increasing well being trajectory This increasing well being group may have experienced the broken engagement as a relief due to some distress or ambivalence at pre breakup. (However, the markers for distress may not have been humiliating or devaluing whic h may have contributed to more longer, chronic low well being (see Cano & O'Leary, 2000).) For example, this group had the lowest percentage of people with a set wedding date and was the only group of people wherein participants indicated they did not want to marry their partner. They were also engaged, on average, for the shortest amount of time, which could suggest that they had less time to in vest or plan for their upcoming wedding/marriage, and therefore may have experienced less overall "loss" as it relates to both the relationship and upcoming marriage. Moreover, this increasing well being group, on average, completed their post breakup surv eys in the longest amount of time since breakup, which could suggest that this increase in well being may not onl y be a reflection of "relief", but could also be the result of greater time for recovery.

PAGE 56

51 Stable low well being trajectory The stable low w ell being trajectory could represent a group of people who were unhappy in their relationship at pre breakup, but who may not have experienced an increase in well being or relief from the breakup because they may still have had sources of distress and/or m ay have experienced certain humiliating events (see Cano & O'Leary, 2000) that left them feeling devalued and contributed to their stable low well being. For example, this group had the highest percentage of partner infidelity and psychological aggression as well as reported substance abuse and domestic violence as reasons for breakup. This group was also engaged for the longest amount of time, which may mean that the negative aspects of their relationship affected their well being for a considerable amoun t of time. Limitations Before discussing potential implications of the results, it is important to highlight several limitations of the study. All measures were asked from the participant's point of view. T he participant's understanding of the relationshi p may differ from his or her partner's understanding of the relationship. Due to the inclusion criteria for the study, the 74 participants represent only those persons with a broken engagement who continued to participate after the engagement ended. It may be that individuals with a broken engagement who discontinued participation would show higher rates of distress and destabilization. That is, the method may have selected for individuals with greater average levels of adjustment post break up. The opposit e is also possible; that is, persons with higher post breakup distress being more motivated to complete the measure and share their experiences. In either case, the method introduces a selection factor that may undermine the generalizability of the results Missing data are also a limitation of the study (particularly in the case of length of

PAGE 57

52 engagement ) Thus, it is important to interpret any results with that variable with more caution than usual. The study design has obvious flaws in terms of isolating t he time line of events. It is possible that there were other events that happened within the pre breakup to post breakup assessment points that caused changes in their post breakup level of well being. That is, it could be that a participant filled out the ir pre breakup survey, experienced a decrease in well being, and then broke up and filled out their post breakup survey. Thus extraneous history or maturation effects may have influenced the results Moreover, the extent to which the observed outcomes for the groups will continue or change is not known due to the limited time frame of this study. Due to the relatively small sample size and the five trajectories that emerged, weak power prohibited me from being able to statistically analyze differences be tween groups and was the reason that the majority of the Aim 3 results were speculative. Using such a subjective process represents additional risk of focusing only on confirming evidence or glossing over disconfirming evidence. Implications T hese findin gs could be relevant for currently engaged individuals experiencing, approaching, or contemplating a break up, as well as clinicians working with such individuals. Although it is unknown what longer term outcomes would be and s everal other areas of inquir y are needed to further understand the clinical utility of these findings, certain general themes can be gleaned from this study. Overall, about half of the people who experience clinical distress after a broken engagement were already experiencing clinica l distress before the broken engagement. For individuals who were individually distressed prior to breakup

PAGE 58

53 (10.8%) breaking an engagement may not change their level of low well being. The small subset of individuals ( also 10.8% of this sample) who dropped into the clinically distressed range may be blind sided by their breakup or events such as infidelity, but the time frame for this distress is not known given the relatively short duration of follow up for this group In contrast, a subset of individuals ( 23 % of the current study ) actually improved in their well being pre to post breakup. This group was also characterized by the shortest duration of engagement prior to breakup. This relatively short duration could represent less investment in the engageme nt, and then less loss/disruption upon breakup. It is also possible that breaking an engagement sooner (perhaps before set wedding plans are made) is better for individual outcomes. In general, for the majority of people who are distressed or ambivalent i n their engaged relationship ( e.g., have relatively low levels of dedication), but who are also not already experiencing clinically significant distress or certain de valuing events or long standing issues (i.e., substance abuse, infidelity, physical or ps ychological abuse), there is a good chance that their initial level of well being following a broken engagement may not change in a clinically significant way either up or down. This appears to describe the majority of people who experience a broken engage ment (78.4% of this sample). In fact, given longer time frames of follow up, an even smaller proportion of the sample may have evidenced clinical distress at post breakup. Specifically, t he group who experienced a clinically significant level of decreased well being answered their post breakup survey in the s hortest amount of time since breakup and the group who experienced an increase in well being answered their post breakup survey in the longest amount of time

PAGE 59

54 since breakup Both of these findings sugg est that mo re recovery time from the event may be associated with greater ad justment This could be useful information to a person interested in how one will fare after a broken engagement. Overall, if a person was doing relatively well prior to the broke n engagement or didn't experience humiliating or devaluing events associated with their breakup, it may be that time will help in healing from this event and their well being may not drop to a level of clinical distress On the other hand, the degree to wh ich one's breakup is associated with existing distress and/or humiliating or devaluing events (i.e., partner desired breakup, infidelity, physical or psychological abuse, substance abuse) it may be beneficial to seek help and extra support during this tim e as their level of well being may drop to a level of clinical distress immediately following the breakup As Cano et al. (2000) suggest, effective clinical interventions aimed at coping with feelings of betrayal, humiliation, and shame may be particularl y helpful for individuals who experience humiliating or devaluing events associated with their breakup. Likewise, clinical interventions aimed at alleviating chronic distress (likely compounded by but not fully attributable to the breakup) may be helpful f or individuals who were already experiencing clinical distress before their breakup At times, it may also be useful to reframe a broken engagement as a positive outcome in the long run, even if there is current distress over the breakup That is, it may b e adaptive to move on from relationships characterized by infidelity, abuse, drug or alcohol problems, or uncertain commitment (e.g., Bourassa et al., 2015; Lavner et al., 2012; Markman et al., 2013). In fact, a Huffington Post author discussing her decisi on to call off a wedding stated that one of the five pieces of guidance she wished she had available when she was "weighing the

PAGE 60

55 options and questioning" her decision was y ou will survive" (Sturges, 9/ 28 /15). The current study helps inform this particular aspect of broken engagements. Future Directions The current study provides useful information on trajectories of well being before and after engagement. Yet, it does not address certain ques tions that individuals may have as they face significant decisio ns regarding engagement and marriage. For example, one may wonder if some level of doubt about the relationship or certain relationship conflicts are a predictor of relationship problems in the future, or whether they are typical of couples who go on to ha ve happy and stable relationships. Although people generally use "cold feet" and "doubt" synonymously when discussing premarital doubts, there is not a clear definition of "cold feet." Even Webster's dictionary: "apprehension or doubt strong enough to prev ent a planned course of action," and Dictionary.com: "a loss or lack of courage or confidence; an onset of uncertainty o r fear" provide somewhat opposing definitions of "cold feet What is clear from each definition is that "cold feet" during an engagemen t may lead one to question their upcoming marriage. What is unclear is whet her there is a certain level of "cold feet" that is predictive of marital distress that may be wise to pay attention to (i.e., a broken engagement) versus lower levels of "cold fee t" that are not necessarily indicative of future distress (i.e., "just" cold feet ). In the same Huffington Post article noted above, the author writes: I know there are people out there who are worried about the small voice inside begging them to wait, an d I wish there were more readily available resources to help figure out what that voice means (Sturges, 9/ 28 /15). This absence of clarity could be addressed i f a prospective, longitudinal study of relationship development w ere conducted in which specifi c q uestions about levels and types

PAGE 61

56 of pre marital doubts were asked, along w ith additional variables The sample could then be split into 3 groups: (1) "broken engagements" (those who got engaged and then broke off their engagement), (2) "successful marri ages" (those who got engaged, married, and remain relatively satisfied), and (3) "distressed marriages" (those who got engaged, married, and then divorced or remain married with high levels of distress). With th ese data, we could then determine what the pr e broken engagement levels of certain variables are for the three groups Although we already know that six months into marriage, endorsing the retrospective yes/no question, "were you ever uncertain or hesitant about getting married ?" predicts poorer mari tal outcomes after four years of marriage (Lavner et al., 2012), this is a very broad and generic marker of doubt. A comparison study following the three groups noted above may help us determine levels and types of doubt that may be predictive of marital d istress. That is, we may be able to better define "cold feet" before marriage and provide certain cutoff levels of "cold feet" that may or may not be predictive of future marital distress. A comparison study like this may allow both clinicians and engaged individuals to consider this information through a cost benefit analysis lens, by considering the likelihood of and degree of drop in well being over this specific time in life (as suggested by the current results) against the potential for a broken engag ement or divorce later on. Thus, if one's engagement levels are closer to the "successful marriage" group levels, they may feel more confident in pursuing marriage despite their premarital doubts, whereas if their levels are more similar to the "distressed marriages" group they may want to consider avoidance of a potential divorce later on down the road, including avoiding the possible psychological consequences that a divorced couple and their family may experience (Waite et al., 2002). That is, such data may help individuals put their pre marital distress or doubt into context and

PAGE 62

57 allow them to decide whether it would be "worth it" to break their engagement (even after they bought their dress/tux, sent out invitations, etc.), given the likelihood of certa in well being trajectories. There are several other important future directions for this study Although this current study did not collapse different trajectory groups together or look at isolated comparisons among just a few of the trajectory groups, ra tionale could be developed for either of those data analysis methods. Further, it may be helpful to better understand what a self report relationship status of "engaged" means to different people (i.e., for those who go back and forth between "engaged" and "dating," or for those who equate making an oral commitment to get married to their partner someday with being engaged) and whether a broken engagement looks different depending on one's definition of "engaged." In conclusion, results of this study are the first to demonstrate that there are different trajectories of well being from pre broken engagement to post broken engagement. Alt hough this adds to the scientific understanding of this unique time in a relationship there is still much to learn about the characteristics of the different well being groups in order to advise in an empirically sound way, individuals considering or experiencing a broken engagement, or clinicians working with these individuals.

PAGE 63

58 REFERENCES Bonds Raacke, J. M., Bear den, E. S., Carriere, N. J., Anderson, E. M., & Nicks, S. D. (2001). Engaging distortions: Are we idealizing marriage? Journal Of Psychology: Interdisciplinary And Applied, 135(2), 179 184. doi:10.1080/00223980109603689 Bourassa, K.J., Sbarra, D. A. & Whi sman, M. (2015). Women in Very Low Quality Marriages Gain Life Satisfaction Following Divorce. Journal of Family Psychology, 29 (3): 490 499. doi: 10.1037/fam0000075 Buckby, J. A., Yung, A. R., Cosgrave, E. M., & Killackey, E. J. (2007). Clinical utility of the Mood and Anxiety Symptom Questionnaire (MASQ) in a sample of young help seekers. BioMed Central Psyc hiatry, 7:50. Cano, A., & O'Leary, D. K. (2000) Infidelity and Separations Precipitate Major Depressive Episodes and Symptoms of Nonspecific Depr ession and Anxiety. Journal of Consulting and Clinical Psychology, 68 (5): 774 781. Clark, L.A., & Watson D. (1991). Tripartite model of anxiety and depression: Psychometric evidence and taxonomic implications. Journal of Abnormal Psychology ,100:316 336. Elison, C. C. (2011). Disharmony: Premarital relationship dissolution. Dissertation Abstracts International Section A, 72 Fowers, B. J., & Olson, D. H. (1992). Four types of premarital couples: An empirical typology based on PREPARE. Journal Of Family P sychology, 6(1), 10 21. doi:10.1037/0893 3200.6.1.10 Graham, J., Liu, Y., & Jeziorski, J. (2006). The Dyadic Adjustment Scale: A Reliability Generalization Meta Analysis. Journal of Marriage and Family, 68(3): 701 717. Grimm, K. J., Ram, N., and Estab rook, R. (2010). Nonlinear Structured Growth Mixture Models in Mplus. Multivariate Behavioral Research 45:6. doi : 10.1080/00273171.2010.531230. Jacobson, N. S., Foilette, W C., & Revenstorf, D. (1984). Psychotherapy outcome research: Methods for repor ting variability and evaluating clinical significance. Behavior Therapy 15, 336 352. Jacobson, N. S. & Truax, P. (1991) Clinical Significance: A Statistical Approach to Defining Meaningful Change in Psychotherapy Research. Journal of Consulting and Clini cal Psychology, 59 (1): 12 19. Kenny, J. J, Kayla, K., Rhoades, G.K., Stanley, S.M., & Markman, H.J. (2013). I (do)n't: Causes and Effects of Broken Engagements. Poster presented at the Association for Behavior and Cognitive Therapies conference; Nashvil le, TN.

PAGE 64

59 Keogh, E, & Reidy J. (2000). Exploring the factor structure of the Mood and Anxiety Symptom Questionnaire (MASQ). Journal of Personality Assessment 74:106 125. Lavner, J. A., Karney, B. R., Bradbury, T. N. (2012). Do Cold Feet Warn of Trouble Ahead? Premarital Uncertainty and Four Year Marital Outcomes. Journal of Family Psychology, 26 (6): 1012 1017. Markman, H.J., Rhoades, G.K., Stanley, S.M., & Peterson, K.M. (2013). A Randomized Clinical Trial of the Effectiveness of Premarital Intervent ion: Moderators of Divorce Outcomes. Journal of Family Psychology, 27 (1): 165 172. Moir Smith, A. (2006). Emotionally engaged: A bride's guide to surviving the "happiest" time of her life. New York, New York: Penguin Group. MuthÂŽn, B., Brown, H. C., Hunt er, A. M., Cook, I. A., & Leuchter, A. F. (2010) General Approaches to Analysis of Course: Applying Growth Mixture Modeling to Randomized Trials of Depression Medication. In Shrout, P. E., Keyes, K. M., Ornstein, K. (Eds.) Causality and Psychopathology, F inding the Determinants of Disorders and Their Cures (pp. 159 179). Washington, DC: American Psychiatric Publishing. Nylund, K. L., Asarouhov, T. & MuthÂŽn, B. O. (2007). Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Model ing: A Monte Carlo Simulation Study. Structural Equation Modeling, 14 (4): 535 569. Owen, J., Rhoades, G.K., Stanley, S.M., & Markman, H.J. (2011). The Revised Commitment Inventory: Psychometrics and use with unmarried couples. Journal of Family Issues, 32 (6), 820 841. Rhoades, G. K., Stanley, S. M. & Markman, H. J. (2009). Couples' Reasons for Cohabitation: Associations with Individual Well being and Relationship Quality. Journal of Family Issues: 30 (2): 233 258. Rhoades, G. K., Stanley, S. M., & Mar kman, H. J. (2010). Should I stay or should I go? Predicting dating relationship stability from four aspects of commitment. Journal of Family Psychology, 24 543 550. doi: 10.1037/a0021008 Rhoades, G. K., Kamp Dush, C. M., Atkins, D. C., Stanley, S. M., & Markman, H. J. (2011). Breaking up is hard to do: The impact of unmarried relationship dissolution on mental health and life satisfaction. Journal Of Family Psychology 25(3), 366 374. doi:10.1037/a0023627 Rhoades G.K., Stanley, S.S., & Markman, H.J. (2 012). A Longitudinal Investigation of Commitment Dynamics in Cohabiting Relationships. Journal of Family Issues, 33 (3), 369 390.

PAGE 65

60 Sabourin, S.P., Valois, P., & Lussier, (2005). Development and validation of a brief version of the Dyadic Adjustment Scale with a nonparametric item analysis model. Psychological Assessment, 17(1): 15 17. Safier, R. (2003). There goes the bride: Making up your mind, calling it off and moving on. San Francisco, CA: Jossey Bass. Sturges, E. (2015, September 28). "The Advice I Wish I Was Given: How to Cancel Your Wedding." Huffington Pos t. Retrieved from http://www.huffingtonpost.com/eve sturges/this is the advice i wish_b_8191242.html. Scott, S. B., Rhoades, G. K., Stanley, S. M. Allen, E. S., & Markman, H. J. (2013). Reaso ns for Divorce and Recollections of Premarital Intervention: Implications for Improving Relationship Education. Couple and Family Psychology: Research and Practice, 2 (2): 131 145. Selig, A. L. (1976). Crisis Theory and Family Growth The Family Coordina tor 25 (3), pp. 291 295. Spanier, G.B. (1976). Measuring dyadic adjustment: New scales for assessing the quality of marriage and similar dyads. Journal of Marriage and Family 38(1):15 28 Stanley, S.M, & Markman, H.J. (1992). Assessing commitment in p ersonal relationships. Journal of Marriage and Family 54:595 608. Straus, M. A., Hamby, S. L., Boney McCoy, S., & Sugarman, D. B. (1996). The Revised Conflict Tactics Scales (CTS2): Development and Preliminary Psychometric Data. Journal of Family Issues, 17 (3): 283 316. Veevers, J. (1991). Traumas versus strens: A paradigm of positive versus negative divorce outcomes. Journal of Divorce and Remarriage, 15(1 2), pp. 99 126. Waite, L. J., Browning, D., Doherty, W. J., Gallagher M., Luo, Y., & Stanley S. M. (2002). "Does Divorce Make People Happy? Findings from a Study of Unhappy Marriages" Institute for American Values, 2002. Wortel, S .N., & Rogge, R.D. (2010). Can you feel the love tonight? Links between sexual behavior and both individual and relationship well being over time. Poster presented at the Associat ion for Behavior and Cognitive Therapies conference; San Francisco, CA

PAGE 66

61 APPENDIX A Results of proposed distinguish ing variables with original hypotheses in parentheses and color coding to show convergence (or lack thereof) with initial hypothesis 1 (High, decreasing well being) 8 p eople Group #2 Blue Hyp othesis 2 (Increasing well being) 17 p eople Group #3 Pu rple Hyp othesis 3 (Mid, m inimal decreasing well being) 16 p eople Group #2 Blue Hyp othesis 4 (High, m inimal decreasing well being) 25 p eople Group #2 Blue Hyp othesis 5 (Stable low well being) 8 p eople Group #4 Green Hyp othesis Partner infidel ity at pre breakup No = 75% For sure = 12.5% ("low") No = 47.1% For sure = 29.4% ("low") No = 56.3% For sure = 12.5% ("low") No = 64% For sure = 16% ("low") No = 25% For sure = 50% ("high") Partner infidelity at post breakup No = 12.5% Yes I know for sure = 37.5% ("high") No = 29.4% Yes I know for sure = 35.3% ("low") No = 50% Yes I know for sure = 18.8% ("high") No = 32% Yes I know for sure = 32% ("high") No = 12.5% Yes I know for sure = 37.5% ("high") Partner desired the breakup at post breakup 50% par tner desired breakup ("yes") 29.4% partner desired breakup ("no") 18.8% partner desired breakup ("yes") 24% partner desired breakup ("yes") 62.5% partner desired breakup ("yes") Marriage date set at pre breakup 37.5% date set ("yes") 35.5% date set ("yes" ) 37.5% date set ("yes") 44% date set ("yes") 37.5% date set ("no") Relationship satisfaction at pre breakup Range 9 21 Mean=16.8 6 ("high") Range 5 21 Mean=15.0 5 ("low") Range 3 19 Mean=13.9 3 ("high") Range 8 21 Mean = 16.24 ("high") Range 9 19 Mean = 14 .75 ("low") Felt constraint at pre breakup Range 1 3.67 Mean = 1.6 7 ("low") Range 1 6 Mean=2.43 ("high") Range 1 6.33 Mean=2.9 6 ("low") Range 1 3.67 Mean = 1.63 ("low") Range 1 5 Mean= 2.54 ("high") Desire to marry partner at pre breakup 100% have full d esire to marry (" sure ") 70.6% have full desire to marry, 23.5% are 75% have full desire to marry, 25% are unsure 88% have full desire to marry, 12% are unsure 62.5% full desire to marry 37.5% are unsure

PAGE 67

62 unsure, 5.9% do not want to marry partner (" unsure ") (" sure ") (" sure ") Perceived likelihood of marriage at pre breakup 62.5% very likely ("high") 47.1% very likely 81.3% very likely ("high") 68% very likely ("high") 37.5% very likely Dedication at pre breakup Range: 3.14 7 Mean = 5.70 ("high") Range: 1.75 7 Mean=5.37 ("low") Range 1.12 6.75 Mean=5.0 9 ("high") Range 4 7 Mean = 6.03 ("high") Range 3.13 7 Mean = 5.67 ("low") Psychological aggression (from partner) at pre breakup Range: 0 4 Mean = 2.09 ("low") Range: 0 6.25 Mean=2.44 (" low ") Range: .5 5.75 M ean = 3.0 2 ("low") Range 0 6 Mean = 2.23 ("low") Range 0 7 Mean = 3.57 ("high") Friends/famil y approval (social pressure) at pre breakup Range: 2.25 6.75 Mean = 4.06 ("high") Range: 1 7 Mean = 3.9 8 ("low") Range: 3 6.75 Mean = 4.71 ("high") Range 2 7 Mean = 4.48 ("high") Range 1.75 5.5 Mean = 3.53 Physical aggression (from partner) at pre breakup 37.5% had some aggression ("no") 58.8% had some aggression ("no") 62.5% had some aggression ("no") 40.0% had some aggression ("no") 50.0% some aggression ("yes") Time since breakup in days 39.63 days Range 1 116 Median = 16.50 101.59 days Range 16 230 Median = 79.00 80.00 days Range 6 152 Median = 93.00 50.00 days Range 1 152 Median = 36.00 100.29 days Range 1 142 Median = 124.00 Length of engagement i n days (recoded) 722.17 days Range 145 1288 Median = 715.00 411.00 days Range 25 1460 Median = 292.00 515.40 days Range 76 1289 Median = 350.00 530.24 days Range 94 1460 Median = 497.00 995.00 days Range 599 1406 Median = 980.00 Note: Variables with no colors were not initially hypothesized but may help to characterize the different trajectories. Words in quotes and parenthesis are what was initially hypothesized, see also Table 1. Green means the results map on relatively well to the hypothesis for how that variable would function within the groups. Yellow means that it is unclear or I am unable to determine if the results map onto the hypothesis for that variable. Red means that the results do not map on well to the hypothesis for that variable.