`If I Had It to Do Over Again .. .": Midlife Review, Midcourse

Journal of Personality and Social Psychology
1999, Vol. 76, No. 2, 270-283
Copyright 1999 by the American Psychological Association, Inc.
0022-3514/99/53.00
'If I Had It to Do Over Again .. .": Midlife Review, Midcourse
Corrections, and Women's Weil-Being in Midlife
Abigail J. Stewart
Elizabeth A. Vandewater
University of Michigan
The University of Texas at Austin
Regrets about early adult life choices, expressed in midlife, are examined as a source of motivation for
life changes in later midlife in 2 samples of women. Replicated findings with longitudinal data indicate
that regret motivates goal setting but is not associated with actually making desired life changes. In both
samples, women who had regrets about early adult life choices but did not make relevant life changes
were lower in later well-being than both women with regrets who did make such changes and women
without regrets. Compared with women who transformed regrets into life changes, women who did not
were lower in effective instrumentality and higher in rumination, though they did not appear to face more
barriers to change. Analyses with longitudinal data indicated that both rumination and effective instrumentality mediated the relationship between regret and well-being for women who did not translate regret
into life changes.
Like psychologists, cultural commentators have been divided in
their views on the value of frankly acknowledging regret. Edith
Piaf (n.d.) seemed proud when she sang, "Je ne regrette rien" ("I
regret nothing") and Katherine Mansfield apparently felt "Regret
is an appalling waste of energy . . . it's only good for wallowing
in" (as cited in Landman, 1993, p. 9). On the other hand, Thoreau
(1906/1949) advised as follows: "Make the most of your regrets;
never smother your sorrow, but tend and cherish it till it comes to
have a separate and integral interest. To regret deeply is to live
afresh" (p. 95). In this article we explore the possibility that frank
acknowledgement of regret—at least in the context of a midlife
review—may indeed permit some women to make changes and
thereby-"live afresh." We also explore why for others the process
of midlife review and regret about early adult life choices does not
result in a "midcourse correction" of the life trajectory.
Abigail J. Stewart, Department of Psychology, Program in Women's
Studies, and Institute for Research on Women and Gender, University of
Michigan; Elizabeth A. Vandewater, Department of Human Ecology, The
University of Texas at Austin.
The research reported in this article was conducted with support from
Boston University Graduate School, National Science Foundation Visiting
Professorships for Women, the Society for the Psychological Study of
Social Issues, the MacArthur Foundation Network for Research on Successful Midlife Development, Radcliffe Research Support and Midlife
Program Grants from the Henry A. Murray Research Center, the University
of Michigan Horace H. Rackham Graduate School, National Institute of
Mental Health subgrants under prime grants 1-RO1-MH43948 and 1-RO1MH47408, and National Institute of Aging Training Grant T32-AG0017.
Computer-accessible data and copies of some of the raw data for several
waves of both studies have been archived at the Henry A. Murray Research
Center, Radcliffe College.
We are grateful to the participants in the studies for their contributions
of time and personal reflections over many years; to Sandra Tangri for her
generous collaboration on the Michigan sample follow up; to Janet Landman and Janet Malley for their colleagueship and conceptual contributions;
to Olga Favreau, Susan Nolen-Hoeksema, Joan Ostrove, Isis Settles, and
Amy Young for their helpful comments on previous versions of this article;
to Gail Agronick, Thomas Popoff, and David Winter for their cheerful and
untiring assistance; and to the Radcliffe Longitudinal Study group and the
Women's Life Paths Study group for feedback, support, and help at many
points in the process.
Correspondence concerning this article should be sent to Abigail J.
Stewart, Department of Psychology, University of Michigan, 525 East
University Street, Ann Arbor, Michigan 48109-1125. Electronic mail may
be sent to [email protected].
The Notion of a Midlife Review
Psychologists are fairly well-agreed that there may be a "life
review" in old age (Butler, 1974; Coleman, 1974; Erikson, Erikson, & Kivnick, 1986; Meacham, 1977) and that it may go well (in
which case it helps an individual come to terms with aging and
death) or badly (in which case it can result in despair). We propose
that a life review conducted in middle age may be one element of
the much-vaunted "midlife crisis" and that it may have similar
potential costs and benefits (see Chiriboga, 1989, for a related
argument). This sort of review or "stocktaking" is clearly a component of many accounts of midlife crises. Levinson (1978) argued
that middle-aged adults experience an intense period of selfevaluation when "Every aspect of their lives comes into question,
and they are horrified by much that is revealed" (p. 199). Similarly,
if a bit less drastically, Jung (1934/1954) pointed out that at around
age 40 "one begins to take stock, to see how one's life has
developed up to this p o i n t . . . . The critical survey of himself and
his fate enables a man to recognize his peculiarities" (p. 193).
Although some theorists like these have stressed the inwardturning aspect of this process of life-evaluation, we note that there
is sometimes a much more active element to it, reflected in visible
change. For example, Levinson pointed out that some people
"cannot go on as before, but need time to choose a new path or
modify an old one" (p. 199).
In his later research with women, Levinson (1996) claimed that
270
MIDLIFE REVIEW AND WELL-BEING
"Most homemakers went through a 'rock bottom' experience of
marriage and life in the Mid-Life Transition" (p. 173). He pointed
out that they "attempted to make new choices" (p. 174) at this
stage of life. The career women in his study had less extreme
experiences, but they also often "felt the need for more limited
changes" (p. 372). Levinson concluded that there was a qualitative
difference between the career women he interviewed and the
homemakers. The homemakers, who centered their early adult
lives on marriage and motherhood, generally "recognized that their
marriage enterprise had been a partial or massive failure and that,
whatever its previous value, they wanted a different kind of marriage, family, and life structure in the next season" (p. 408).
According to his account, the career women more often had
included elements of work, marriage, and motherhood in their
lives throughout early adulthood and sought more modest adjustments in their lives around age 40. These findings echo those of
Helson and Picano (1990) that the traditional role may be "bad," at
least for many contemporary women. These accounts suggest that
for some women the process of midlife review may be intensified
by the absence of a role in the labor force.
There is, then, reason to believe that a process of midlife review
may take place in early middle age, among women. This review
may result in regret about past sacrifices and a desire for drastic
life changes under some circumstances (perhaps particularly when
early adulthood has involved the sacrifice of some central roles)
and more modest changes under others.
The Role of Life Regrets
Although many psychologists are interested in the cognitive
features of regret (as a form of counterfactual thought), others have
recognized that acknowledged regrets can motivate changes in
behavior. Lecci, Okun, and Karoly (1994) defined regrets as
"unfulfilled or unattainable intentions or goals" (p. 731), thereby
placing regrets clearly in the domain of motivation. There is no
doubt that sometimes regret is associated with shame and rumination about personal inadequacies and, therefore, with inaction (see,
e.g., Landman, 1993; Niedenthal, Tangney, & Gavanski, 1994). It
can also, though, lead to corrective actions. For example, Landman, Vandewater, Stewart, and Malley (1995) suggested that
acknowledging "missed opportunities" might sometimes reflect
simple acknowledgment of (past) facts with implications for the
present, but at other times it might permit development of a plan
for the future: "Like pain, counterfactual thought about a regrettable past may serve instructional and motivational purposes—
telling us that something is wrong and moving us to do something
about it" (Landman et al., 1995, p. 89).
This perspective is consistent with evidence that regrets may provide information about a standard to be exceeded in future performance (Karoly, 1993) and that apparently negative thinking can
facilitate positive responding (Boninger, Gleicher, & Strathman,
1994; Cantor, 1990; Roese, 1994). Further, Lecci et al. (1994) provided important information about when regret is likely to serve such
a function: when the person is young or middle-aged (vs. older) and
when the focus of thought about regret includes a sense of potential
personal efficacy. The importance of this kind of efficacy or instrumentality, and of factors that impede it like rumination, is supported
by a fairly large body of research (e.g., Bandura, 1977,1986; Bandura
& Schunk, 1981; Lyubomirsky & Nolen-Hoeksema, 1995; Nolen-
271
Hoeksema, 1991; Nolen-Hoeksema, Parker, & Larson, 1994). Situational and dispositional factors (like age, efficacy, and ruminative
tendencies) may, then, increase or decrease the likelihood that regrets
serve as positive motivations for actions rather than stimuli for paralysis. Because we know that age may be one important situational
factor, we propose that when a midlife review (in contrast to a life
review in old age) includes acknowledged regrets, it may also more
easily motivate life changes. Even among middle-aged individuals,
though, this is more likely if the regret is accompanied by positive
expectancies of success about making those changes and not by
ruminative doubt.
The Role of Regret in Women's Midlife Well-Being
Research on women from several generations has suggested that
women's life regrets often focus on missed opportunities for
education and career development. For example, Sears (1979)
found that the Terman gifted women (born around 1910) regretted
not having pursued careers throughout their adult lives. Similarly,
Landman and Manis (1992) found that two thirds of their sample
of adult women (surveyed in the early 1980s) had educational and
career regrets. Among women born in this century, then, there is
evidence that educational and career regrets are particularly common (see Landman, 1993, for a review).
We may also suspect, though, that educational and career regrets
may have increased motivational importance for the current generation of middle-aged women. It is this generation that first
benefited from the educational and work opportunities associated
with the women's movement of the late 1960s and early 1970s.
However* not all women personally benefited from those opportunities; those women who feel they did not may be particularly
prone to midlife regret. Levinson's (1996) research suggests that
contemporary middle-aged women (women of the "baby boom"
generation) have more serious regrets at midlife if they pursued
marriage and motherhood roles, but not careers, during early
adulthood. Landman et al. (1995), reporting on a sample of
college-educated women of the same cohort, found that midlife
acknowledgment of "missed opportunities" for career and education was associated with higher levels of concurrent depression
and anxiety, even controlling for dispositional depression and
anxiety. Presumably the women most likely to have regrets about
such missed opportunities were precisely those who did not in fact
take advantage, as young adults, of the new opportunities for
women.
Regret, Life Changes, and Midlife Well-Being
We have seen that midlife is often a period of life review,
including identification of regrets about the past. In addition, we
have seen that regrets can motivate goal setting and active efforts
to actualize goals. Finally, we have seen that contemporary midlife
women who have significant regrets about sacrifices of their own
educations and careers in early adulthood may be especially prone
to emotional difficulty, and the formulation of goals for life
changes, in middle age.
In this study, we tested several hypotheses based on these ideas.
We drew on two longitudinal studies of midlife women who are
college educated. These two samples of women were part of the
first generation of women to come of age as new opportunities for
272
STEWART AND VANDEWATER
education and careers were opening up for women (see Hulbert &
Schuster, 1993). For that reason, they may be expected to be
particularly vulnerable to regrets about choices made under one set
of social pressures, when—later in life—they were subject to a
different set. In addition, these well-educated women possessed
good resources for making life changes they desired.
Among these midlife women, we predicted that regrets about
"traditional" role choices (that sacrificed women's educations or
careers) would serve as motivation for making changes in these
areas. It is, of course, possible that even in these relatively privileged samples, external barriers to change stand in the way of those
wishing to act on their regrets; we examined this possibility.
Second, we proposed that acting on regrets would enhance wellbeing in later midlife, whereas failure to act on them would lower
midlife well-being. Third, we anticipated that the ability to act on
regrets would be related to individual differences in personality
resources related to expectancies for success at making life
changes. More specifically, we expected to find that the personality resource of effective instrumentality enhances women's ability to make desired life changes, whereas the personality characteristic of rumination interferes with their ability to do this. In fact,
we proposed that these personality characteristics would mediate
between unsuccessful midcourse corrections and lower midlife
well-being; we tested this hypothesis with the sample that has
longitudinal data permitting such a test.
Study 1
Method
Participants
The first study includes participants in Tangri's longitudinal study of
women from the University of Michigan Class of 1967 (Tangri, 1969,
1972; Tangri & Jenkins, 1986, 1993). The initial sample of 200 was
randomly stratified to produce three equal groups differentiated by their
career aspirations as college seniors (nontraditional, moderately traditional,
or very traditional; see Tangri, 1969, 1972). They were first studied in 1967
and have been followed up three times (1970, 1981, 1992) since then. In
each wave the women were asked basic demographic questions, as well as
a number of closed- and open-ended questions about their life choices and
life events. In addition, personality characteristics and well-being were
assessed in the 1992 wave.
This study used data collected in 1981 (n = 117) and 1992 (« = 109),
when the women were, respectively, ages 36 and 47. Analyses for this
study focus on the 83 women with complete data in these two waves (i.e.,
at both age 36 and 47). As is common in longitudinal studies, the entire
sample did not participate in every wave of data collection. Analyses aimed
at assessing potential bias in this sample conducted on all available variables (including demographic information, personality, and well-being) did
not reveal systematic differences between participants selected for this
study and those omitted because of missing data. Similarly, Cole and
Stewart (1996) reported no detectable bias in the 1992 wave, on the same
variables. In addition, Jenkins (1989, 1994) reported that sample attrition
from 1967 to 1981 did not systematically bias the follow-up sample in
terms of demographic and background variables available to her.
about age 29, and they had an average of two children. In terms of highest
educational degree earned, by age 47, 28% had a bachelor's, 47% had a
master's, and 25% had a doctoral level degree. At age 36, 69% were
engaged in the paid labor force; at age 47, 90% of them were. Those
women in the paid labor force at age 47 were distributed fairly evenly
across a work status scale (Hollingshead & Redlich, 1956): 27% were
major professionals (e.g., physicians, college professors), 37% were minor
professionals (e.g., certified public accountants, social workers), and 36%
were in administrative or clerical jobs. By age 47, their median personal
income was between $30,000 and $40,000 a year.
Measures
In order to examine our hypotheses, and to capitalize on the longitudinal
nature of our data, we used measures generated from a variety of sources.
These included coding of open-ended material, use of standardized scales,
and creation of psychometrically sound measures where needed. This
multimethod approach using an unusually rich array of instruments (i.e.,
standardized quantitative scales, Q sorts, and projective measures, described in detail below) is an important strength of this research.
Regrets about pursuing traditional roles for women. On the basis of
literature suggesting these would be especially prominent, our interest was
in regrets about having pursued traditionally feminine roles (as opposed to
other kinds of regrets—e.g., regretting having chosen a law career rather
than a career in medicine). In order to assess this, we coded women's
responses at age 36 (in 1981) to the following question: "In retrospect, are
there any things you would have done differently? If yes, please tell us
what." Similar questions, in open- and closed-ended formats, have been
asked in Gallup polls (in 1965; see Erskine, 1973), other studies of
educated women (Baruch, Barnett, & Rivers, 1983; Ginzberg & Associates, 1966; Yohalem, 1993), and recent research with cross-sectional
samples (DeGenova, 1992; Metha, Kinnier, & McWhirter, 1989).
Statements coded as traditional role regret included (a) indications that
the woman would have planned or thought more seriously about a career,
would have chosen career over home, would have gone to graduate school
or would have gone earlier, or would have chosen a career in a high-status,
traditionally male-dominated field, rather than a traditionally femaledominated occupation; (b) indications that the woman would have married
after establishing a career (rather than before), would have married someone who shared equally in childcare and household responsibilities, or
would not have put her husband's needs above her own; or (c) indications
that the woman would have gone back to work sooner (or at all) after
having children or would have established a career before having children.
Examples of traditional role regrets were as follows: "I would not have
let my husband take sole responsibility for determining the course of our
lives. His. career has always been the only deciding factor in our lives[,]
which has not been fair to me or the children." "I would have chosen a
profession less traditionally female and would have gone into something
with more long-range challenge." "When I was in college I considered a
career in medicine. Due to lack of support from family members and future
husband, I didn't pursue it. I would have at least tried it if I were 20 today."
Examples of regrets that were not about traditional roles were "choose a
career without night shifts" and "would have married earlier."
On the basis of this content coding, participants were categorized into
one of two groups: (a) traditional role regret (women who specifically
regretted having pursued traditional roles, n = 51) and (b) no traditional
role regret (women who either had no regrets, n = 23, or who had regrets
that were not about pursuing traditional roles for women, n = 9, total N =
32).' The coding was completed by two coders, and interrater reliability
Description of the Sample
At age 47, 86% percent of the women were married or living with a
partner, 8% were currently divorced, and 6% had never been married.
Eighty-nine percent of the sample had become mothers, on average at
1
A few women (n = 5) expressed more than one regret; they were
categorized as having traditional role regrets if at least one regret was
coded as such.
MIDLIFE REVIEW AND WELL-BEING
coefficients calculated at various times during the coding were quite high,
ranging from .91 to .94.
Consistent with research on previous cohorts of women (see Erskine,
1973; Landman & Manis, 1992; Landman et al., 1995), traditional role
regrets for this cohort mainly centered on education (44%) and career
(38%), with smaller proportions in the areas of marriage (15%) and
parenting (3%). Of the women without traditional role regrets, nearly three
fourths had no regrets, whereas all of those with regrets lamented that their
family role involvements were not more traditional (such as getting married and starting a family earlier or staying home with children).2
Because traditional role regret at age 36 was measured with a single
item, we were concerned about reliability. However, in 1992 (at age 47) the
women responded to five different life review questions used in various
studies of women's lives, including the question asked at age 36. Thus,
using the age 47 data, we were able to address this issue by examining the
reliability of a multi-item measure of traditional role regret that included
the single item asked at age 36 (in 1981).
All five life review items from the 1992 (age 47) wave were coded for
traditional role regret (by the same expert coders with interrater reliability
coefficients of .90 and above), and women were categorized as either
having or not having traditional role regrets for each item, using the
systems previously described. The five items were asked in the following
order; response distributions associated with them follow: (a) "Were there
any attractive opportunities for career or other long-range activities which
you did not pursue? If yes, what are they? Why did you not pursue them?
Do you have any regrets about not pursuing them?" (yes, n = 21; no, n =
57); (b) "Are there any ways in which your life has been different from
what you thought it would be when you left college? If yes, in what way?"
(yes, n = 17; no, n = 59); (c) "If you had to do it over again, would you
choose the same lifestyle pattern—with respect to your home versus career
decisions? Why/why not? What would you change?" (yes, n = 18; no, n =
58); (d) "In retrospect, are there any things you would have done differently? If yes, please tell us what." (yes, n = 26; no, n = 50); and (e) "Any
regrets?" (yes, n = 14; no, n = 56).3 Because the first question asked
specifically about career-related opportunities missed, responses to subsequent questions may well have been biased toward content in that domain.
Analyses of these five items suggest that they comprise a highly internally consistent measure (a = .76) reflecting regrets about traditional roles.
Moreover, the item used to assess traditional role regrets in 1981 (Item 1
in 1992) was highly correlated (r = .71) with a scale composed of the other
four 1992 items (Items 2 through 5; a = .64). These analyses foster
confidence in the reliability of the age 36 measure of traditional role regret.
Career-related goals at age 36. Career-related goals were coded from
open-ended responses to the question "As you think of your future life,
what is your picture of the way you'd like life to work out for you?" The
coding focused solely on expressions of goals related to career, education,
or work opportunities, which were coded as none (no career goal mentioned or continue on the same track, n = 39), incremental change desired
(continue on same track with advancement or promotion, n = 15), or major
change desired (new job trajectory wanted, n = 27). Examples of responses coded none were "continuing as a psychologist," and "more of the
same." Examples of incremental change desired were "become general
counsel of a major corporation," and "get tenure." Examples of major
change desired were "return to graduate school" and "start my own small
business." The coding was completed by two coders, and interrater reliability coefficients calculated at various times during the coding ranged
from .89 to .93.
Desired career-relevant changes made from age 36 to age 47. To
assess whether women's career-relevant changes following regret were
desired, or chosen, by themselves versus imposed on them by situational
exigencies or another person (such as a spouse or partner), we coded their
year-by-year activities from age 36 to age 47, which they related in
response to the following open-ended question:
273
In the spaces on the next two pages, please indicate as fully as possible
the places you have lived, and the activities that have occupied your
time, each year, for the last twelve years. Some examples might be
"got divorced and went to work for the March of Dimes working on
legislation for funding," "Took a course in modern dance while living
in California—also decided to apply for teaching jobs," and so on. We
know that in some cases you will have difficulty recalling your
activities; however, we would appreciate as much detail as possible.
One way of recalling might be to locate a pivotal event that you do
recall, and then work backward and forward in time from it.
The coding focused on career and educational changes (referred to as
career-relevant changes) such as entering or leaving graduate school or
career training programs, entering or leaving jobs, and receiving job
promotions or demotions. Such changes were coded as either desired or not
desired by two coders with interrater reliability coefficients taken at
various times during the coding ranging from .87 to .91.
A career-relevant change was coded as desired if (a) the woman indicated that the change was something she had been striving for, working
toward, or simply desired (i.e., there was clear indication that the change
was desired) or (b) there was no indication that the change wasn't desired,
and it was generally the kind of change someone would choose (e.g., "got
promoted"). Examples of desired career-relevant changes included "finished doctorate and began teaching at a university," "finally realized that I
had always wanted to be a doctor and went to medical school," and "left
present company to try a U.S. multinational."
A career-relevant change was coded as not desired if (a) there was clear
indication that the change was not desired by the woman herself, was the
result of other people's decisions or influences on her life (such as having
to give up a rewarding job as a consequence of a spouse's career move), or
both or (b) the woman did not explicitly state that the change was not
desired, but it was generally not the kind of change someone would choose
(e.g., "got demoted," was fired, was laid off, etc.). Examples of careerrelevant "changes that were not desired were as follows: "My husband left
me so I started a bed and breakfast because I had to do something to keep
from starving." "I took the church secretary job because I hate being 'on
the dole' from my husband. 'We' have lots of money, T don't!"
Over the 11-year period from 1981 to 1992 (i.e., between ages 36 and
47), participants could obviously have had both kinds of career-relevant
changes (desired or not). For the purpose of examining our hypotheses,
participants were categorized into one of two groups: (a) women with any
desired career-relevant changes (n = 57) or (b) women who had no desired
career-relevant changes (n = 26).
Contextual barriers to change. We recognized that some women's
lives included external obstacles to the changes they might desire. Some of
these might be impossible to assess with our data, such as a partner's
psychological need for an at-home spouse. However, others could be
examined. Using demographic information that reflected the women's
family responsibilities at age 36, we coded the following contextual barriers to change as present or absent (1 = present, 0 = absent): (a)
responsibility for children under the age of five (yes, n = 39; no, n = 45),
(b) single parenthood (yes, n = 3; no, n = 80), (c) unemployed spouse
contributing no income (yes, n = 1; no, n = 82), and (d) spouse with a
2
Thus, among women who indicated that they would do things differently (n = 60), most (85%) expressed regrets about the traditionality of
their role involvements (mainly in terms of education and career). A much
smaller proportion of women with regrets (15%) lamented their lack of
traditional role involvements, and these were uniformly centered around
family (rather than any aspect of work or education).
3
Question 1 is identical to the question used to measure traditional role
regret at age 36 in Study 1 (Michigan sample); Question 2 is identical to the
question used to measure traditional role regret at age 37 in Study 2
(Radcliffe sample).
274
STEWART AND VANDEWATER
high-level and demanding executive or professional position (e.g., president or vice president, diplomat, law firm partner, etc.; yes, n = 26; no, n =
57). Because individual barriers were often low in frequency (presence
ranged from 1 to 39), the four barriers were summed to create total
contextual barrier scores (M = 0.84, SD = 0.72, range = 0 to 3). Where
frequency allowed, we also examined the presence of individual barriers
separately.
Well-Being at Age 47
Because scholars in this area have advocated a multidimensional approach to well-being (e.g., Andrews & Robinson, 1991; Ryff, 1989), we
examined four different types of indicators generally viewed as important
components of overall well-being: symptoms of psychological distress
(i.e., depression and anxiety), physical well-being, overall life satisfaction,
and adult adjustment.
Symptoms of psychological distress. At age 47, the women completed
the Zung self-rating depression (1965) and anxiety scales (1971b). The
scales ask participants to rate emotional and physical symptoms associated
with anxiety (feeling panicky, having shaking hands) or depression (feeling
blue, loss of appetite), respectively, on a 4-point scale ranging from never
(1) to nearly all the time (4). Scores are derived by dividing the sum of the
raw scores by the maximum score of 80 (most anxious or most depressed,
respectively) and are thus expressed as a proportion ranging from .25
to 1.00 (anxiety scale a = .80, M = .45, SD = .07; depression scale a =
.85, M = .41, SD = .07).
Physical well-being. At age 47, the women were asked to rate their
general state of health on a 5-point scale ranging from poor (1) to excellent
(5), as well as their energy level on a 5-point scale ranging from low (1) to
high (5). Measures of overall health and energy are generally highly
reliable and strongly correlated with "objective" measures such as physicians' assessments (Bird & Fremont, 1991; Okun & George, 1984). The
items were standardized and combined to create a two-item scale assessing
physical well-being (scale a = .78, M = .11, SD = .87).
Life satisfaction. At ages 36 and 47 participants responded to a widely
used and highly reliable measure of overall life satisfaction (Robinson,
Shaver, & Wrightsman, 1991) that asks, "Overall, are you satisfied with the
way your life has turned out thus far?" on a 3-point scale (1 = not very
satisfied, 2 = fairly satisfied, 3 = very satisfied; age 36, M = 2.50,
SD = 0.55; age 47, M = 2.42, SD = 0.64). Robinson et al. advocated the
use of single-item measures of life satisfaction as they are highly correlated
with multiple-item measures, are just as reliable, and have the added
advantage of being less taxing for respondents.
Index of adult adjustment. Active engagement in different areas of life
(work, intimate relationships, family, etc.) has long been conceptualized as
a measure of effective social functioning (see, e.g., Levinson, 1978; Vaillant, 1977). To assess this, we used Picano's (1986) Life History Index of
Adult Adjustment (IAA), developed specifically for adult women using
Vaillant's criteria for men.
The IAA is scored to reflect positive adjustment, with each item scored
dichotomously; respondents receive 1 point for responses reflecting positive adjustment on each item. Items are summed to yield an overall score
reflecting adjustment in four areas of life (career, social-interpersonal,
psychological health, and physical health). We used a short form of the
total index (20 items), using the 6 items (from all four areas) that accounted
for 80% of the variance in the total adjustment score as reported by Picano
(1986). These 6 items were (a) no disruption or curtailment of daily
activities by illness, (b) enjoyment of work, (c) absence of drug use, (d) 10
years or more of stable marriage, (e) satisfaction with remuneration at
work, and (f) absence of a period of significant psychological distress
(M = 3.84, SD = 1.29). The IAA has good reliability and external validity
(reported in detail in Picano, 1986), and other researchers have found that
this approach yields a measure of adult adjustment that is related to
well-being in adulthood (e.g., Franz, McClelland, & Weinberger, 1991).4
Relationships among well-being indices. The four indices of wellbeing are intercorrelated to some degree. The anxiety and depression scales
are in fact highly correlated (.73 in this sample; it is in this range in most;
see, e.g., Zung, 1971a). However, the average intercorrelation of each
well-being measure with all other measures of well-being is quite modest
(anxiety mean r = .30; depression mean r = .36; physical well-being mean
r = .38; life satisfaction mean r = .22; adult adjustment mean r = .30).
Given these significant but modest average intercorrelations and their
conceptual distinctness, it seemed wise to retain them as separate
indicators.
Personality Resources
Because we were interested in the extent to which personality resources
were related to the women's ability to act on their regrets and make desired
changes, we examined two personality characteristics at age 47: effective
instrumentality and rumination (see, e.g., Bandura, 1986; Nolen-Hoeksema
et al., 1994).
Individual differences in effective instrumentality and rumination were
measured using data from the California Adult Q-set (CAQ; Block 1961/
1978). The 1992 questionnaire data were studied by three to five raters,
who then formulated personality descriptions of each woman by means of
Q-sorting the 100-item CAQ. The comprehensive set of questionnaires
consisting of open- and closed-ended measures provided a rich and differentiated impression of each woman's life to that point. This was true even
though all material was removed from rating files that provided the basis
for measures of regret, future goals, changes, and well-being. These exclusions were, of course, essential to ensuring that no two variables were
based on the same or overlapping data. Block and Haan (1971) and Helson
(1992) have shown that Q sorts based on questionnaire files such as these
provide extremely useful summaries of observers' impressions of personality. Though no attempt is made to induce consistency of judgments,
because"'the goal is to aggregate somewhat different views of a person,
interjudge reliabilities for the CAQ composites in this sample ranged from
.58 to .91, with a mean of .77.
Effective instrumentality. We conceptualized effective instrumentality
as a broad sense of personal efficacy coupled with a repertoire of coping
skills and investment in life projects, which substantially overlaps with the
views of many others (e.g., efficacy, Bandura, 1977; locus of control,
Rotter, 1966; personal causation, de Charms, 1976; perception of control,
Langer, 1983; and helplessness, Seligman, 1975, among others), and we
explicitly relied on such theories to choose Q-sort items indicative of
effective instrumentality. A measure of effective instrumentality was created by averaging the women's scores on seven Q-sort items: assertive,
4
It is important to note that this index is most appropriately thought of
as a measure of conventional adult adjustment. That is, in addition to signs
of illness and stress, respondents acquire "adjustment points" for such
things as being legally married and remaining in that marriage for at
least 10 years (thus, they must be in a legally sanctioned heterosexual
relationship) and enjoying their work and feeling happy with their pay
(thus, they must be in the paid labor force). On the one hand, these kinds
of engagements are certainly used by many (both by self and by others) to
judge the success or achievement of individuals in our society. On the other
hand, it is important to acknowledge that these kinds of behaviors can also
occur in the presence of significant distress (e.g., individuals who remain
in poor or abusive marriages, the lesbian or gay person who legally marries
because of fear of the very real social consequences of being "discovered").
In addition, the lack of engagement in these kinds of conventional social
accomplishments can occur in the presence of significant well-being (e.g.,
women who choose not to marry or remarry after divorce, etc.). Thus, the
particular meaning of this measure should be kept in mind when interpreting the results.
275
MIDLIFE REVIEW AND WELL-BEING
productive, high aspirations, acts decisively, gives up easily (reverse
scored), delays action (reverse scored), and self-defeating (reversed scored;
scale a = .86, M = 6.81, SD = 1.13).
Rumination. On the basis of the work of Nolen-Hoeksema and her
colleagues (e.g., Nolen-Hoeksema, 1991; Nolen-Hoeksema et al., 1994),
we conceptualized rumination as a tendency to focus on negative life
events or life circumstances. This tendency amounts to a self-defeating
(and possibly depressive) personal style. Although Nolen-Hoeksema and
her colleagues (e.g., Nolen-Hoeksema et al., 1994) were careful to distinguish between rumination about a depressed mood (i.e., depressive rumination) and the tendency to focus on one's failures, our measure makes no
such distinction, as it is based on observer ratings of a set of comprehensive
questions. Thus, we labeled it simply rumination and created scale scores
by averaging responses to six Q-sort items: tends to ruminate, self-pitying,
interprets clear-cut situations in complicated ways, negativistic, moody,
and cheerful (reverse scored; scale a = .71, M = 3.45, SD = .82).
Analysis Plan
To address our hypothesis that traditional role regrets would serve as
motivation for change, we examined the relationships between concurrent
traditional role regrets and career-related goals and between earlier traditional role regrets and subsequent desired career-relevant changes using
nonparametric statistics.
We also focused on a particularly important group: those with regrets
who did not make changes in subsequent years. Our hope is that we can
identify how this particular group differs from other groups—specifically,
those with regrets who did make changes and sometimes also those without
regrets.
In the first instance, we asked whether women with regrets who did not
make changes faced more external barriers to change in their lives than
those who had regrets and did make changes. In parallel fashion, we
hypothesized that acting on regrets would enhance midlife well-being,
whereas failing to act on them would lower midlife well-being. Finally, we
predicted that the personality resource of effective instrumentality enhances women's capacity to act on their regrets, whereas rumination
interferes with this capacity. These hypotheses involve comparisons (of
barriers, well-being, and personality, respectively) between two specific
pairs of groups: (a) women with traditional role regrets who did not make
desired career-relevant changes and those who did (i.e., regret-no change
vs. regret with change) and (b) women with traditional role regrets who did
not make desired career-relevant changes and all other groups combined
(i.e., regret-no change vs. regret with change, no regret-no change, and no
regret with change). Because we expected that women with traditional role
regrets who did not make desired career-relevant changes would have
lower well-being, lower levels of effective instrumentality, and higher
levels of rumination than any of the three comparison groups, we performed planned comparisons to examine these hypotheses.
Results
Results of analyses examining relationships between concurrent
traditional role regrets and career-related goals and between earlier
traditional role regrets and subsequent desired career-relevant
changes are presented in Table 1. Having traditional role regrets
was significantly related to having concurrent career-related goals,
^ ( 2 , N = 81) = 15.66, p < .001, but was not related to later
making desired career-relevant changes, ^ ( 1 , N = 83) = 0.22, ns.
Thus, it seems that although traditional role regrets may be related
to an inner desire or motivation for change, such regret by itself is
not necessarily related to actually making desired career-relevant
changes.
We also considered whether contextual barriers to change were
more prominent in the lives of women who had regrets, but did not
make changes, when compared with women with regrets who did.
Because these two groups did not significantly differ in whether or
not they experienced any single contextual barrier, we present the
average number of barriers experienced by each group in Table 1.
As can be seen in this table, the difference between these groups
on the average number of contextual barriers was also not significant, ^49) = 0.60. Thus, it does not appear that women in this
cohort with traditional role regrets who did not make subsequent
desired career-relevant changes experienced any more contextual
barriers than those with regrets who did make such changes—at
least in terms of barriers reflecting family responsibilities.
To address our second set of hypotheses, we examined differences in midlife well-being and personality resources of groups
Table 1
Michigan Sample: Relationships Among Traditional Role Regret, Career-Related Goals, Desired Career-Relevant Changes,
and Contextual Barriers to Change
Career-related goals (age 36)
Variable (age 36)
Absence of
traditional role
regret
Presence of
traditional role
regret
Mean contextual
barriers to
change
Desired career-relevant changes
(age 36 to age 47)
None
Incremental
change desired
Major change
desired
No desired
changes made
Any desired
changes made
74% (23)
16% (5)
10% (3)
34% (11)
66% (21)
32% (16)
20% (10)
48% (24)
29% (15)
71% (36)
Traditional rolf : regret (age 36)
by desired career-relevant changes
(age 36 Ito age 47)
Regret-no change
(n = 15)
Regret with change
(n = 36)
1.00(0.55)
0.86 (0.79)
Note. For career-related goals and desired career-relevant changes, row percentages add up to 100; numbers in parentheses indicate the number of
participants associated with each percentage. For contextual barriers to change, numbers in parentheses indicate the standard deviation associated with each
276
STEWART AND VANDEWATER
defined according to earlier traditional role regret and subsequent
desired career-relevant change. The means and standard deviations
of all groups on all well-being and personality variables are presented in Table 2.
The results of the planned comparisons between the two groups
of women with traditional role regrets at age 36 indicate that those
who made no desired career-relevant changes from age 36 to
age 47 were significantly more depressed, ?(79) = -3.30, p <
.001, were significantly more anxious, r(79) = -3.00,p < .01, and
felt significantly less physically healthy, r(78) = 2.90, p < .01, at
age 47 than those who had made desired changes. However, these
two groups did not differ in terms of life satisfaction or adult
adjustment at midlife.
This overall pattern of results was confirmed when the midlife
well-being of women with earlier traditional role regrets who made
no subsequent desired career-relevant changes was compared with
the other three groups as a whole. Again, women who did not
make changes were more depressed, ;(79) = 3.42, p < .001, were
more anxious, f(79) = 3.18, p < .01, and felt less physically
healthy, r(78) = 2.21, p < .05, than all others. Also as before, these
two comparison groups did not differ on overall life satisfaction or
adult adjustment.
In order to rule out the possibility that prior well-being might be
a more important determinant of later well-being than regret and
desired change, we performed the same analyses, using life satisfaction at age 36 (the only prior well-being variable available in
this study) as a covariate for all well-being indicators at age 47.
Using earlier life satisfaction as a covariate did not change the
results of the well-being analyses reported above.
Finally, the results of the planned comparisons examining personality differences (see Table 2 for group means and standard
deviations) indicate that women with traditional role regrets at
age 36 who did not make desired career-relevant changes from
age 36 to age 47 scored significantly lower on effective instrumentality at age 47, *(78) = 2.39, p < .05, and significantly higher
on rumination, ?(78) = —2.02, p < .05, than women with traditional role regrets who did make desired career-relevant changes.
When women with traditional role regrets who did not make
desired career-relevant changes were compared with the other
three groups together, they scored significantly lower on effective
instrumentality, f(78) = 2.35, p < .05, but were not significantly
different on rumination.
Discussion
The results of our analyses provide some support for all of our
hypotheses. First, there is evidence of regret-motivated goal setting
for this sample of women, but it did not seem to motivate successful change. Making the kinds of changes we examined was not
rare in this sample; many of the women with and without regrets
made changes. However, it is clear that some of the women with
traditional role regrets did not make life changes relevant to those
regrets, despite having expressed the desire to make those changes.
Perhaps actually making changes requires additional resources.
Second, those with regrets who did make changes were better
off in terms of well-being than those who did not. They did not
differ in the presence of contextual barriers to change, at least
insofar as we were able to measure them. However, they did differ
in terms of the personality characteristics that might account for
their capacity to make life changes successfully.
Though these results are encouraging, we note that the sample is
relatively small and we used new measures of effective instrumentality and rumination. In addition, the measures of well-being,
change, and personality resources were all obtained at the same
time. These limitations in the data underscored the usual motive to
replicate intriguing new results.
Study 2
Method
Participants
The women in this sample are participants in Stewart's longitudinal
study of the graduates of the Radcliffe College Class of 1964 (Stewart,
1974, 1978, 1980; Stewart & Salt, 1981; Stewart & Vandewater, 1993).
The current study used data collected from the women in 1979 (n = 133),
Table 2
Michigan Sample: Planned Comparisons of Traditional Role Regret (at Age 36) by Desired Career-Relevant Change Groups
(at Age 36 to 47) on Later Well-Being and Personality in Midlife
Comparison 1
Regret-no change
Variable
Well-being (age 47)
Depression
Anxiety
Physical well-being
Life satisfaction
Adult adjustment
Personality (age 47)
Effective instrumentality
Rumination
Comparison 2
Regret with change
Regret-no change
All other groups
M
SD
M
SD
M
SD
M
SD
.48,
.50,
-.41,
2.33
3.33
.07
.09
1.12
0.48
1.11
.40,
•44,
•35,
2.43
3.86
.08
.06
.76
0.60
1.37
•48,
•50,
-•41.
2.33
3.33
.07
.09
1.12
0.48
1.11
.40,
•43,
2.48
3.96
.07
.07
.78
0.62
1.30
6.15,
3.80,
1.52
1.22
6.96,
3.29,
0.85
0.61
6.15,
3.80
1.52
1.22
6.99,
3.42
0.99
0.70
•23,
Note. Means with different subscripts differ significantly at p < .05. Regret-no change, n = 15; regret with change, n = 36; no regret-no change,
n = 21; no regret with change, n = 11.
MIDLEFE REVIEW AND WELL-BEING
1986 (n = 103), and 1991 (n = 149), when they were, respectively,
ages 37,43, and 48. As with the Michigan sample, in each wave the women
were asked basic demographic questions and a number of open-ended
questions about their lives and (after the first wave) completed standardized measures of well-being. In addition, personality characteristics were
assessed by observers at age 43.
Analyses for this study focus on the 76 women with complete data who
participated in all three waves examined here. Analyses aimed at assessing
potential bias in this sample were conducted on all available variables
(including demographic information, well-being variables, and personality
variables). These analyses did not reveal systematic differences between
the participants selected for this study and those omitted because of
missing data. Vandewater, Ostrove, and Stewart (1997) reported no detectable bias in this sample on similar variables.
Description of the Sample
At age 48, 74% of the sample were married, 21% were currently
divorced, and 5% had never been married. Eighty-seven percent of the
women had become mothers (on average, at about age 29), and they had an
average of two children. By age 48, 29% of these women had earned a
doctoral-level degree, and 42% had earned a master's-level degree. At
age 37, 83% of the sample were engaged in the paid labor force; at age 48,
86% of them were. Those women who were employed in the paid labor
force at age 48 tended to be concentrated in the upper end of a work status
scale (Hollingshead & Redlich, 1956); 49% of them were major professionals (e.g., physicians, college professors), 22% were minor professionals (e.g., certified public accountants, social workers), and 28% were in
administrative or clerical jobs. At age 48, the women's median income was
between $30,000 and $49,000 a year.
Measures
Most measures obtained for this sample were identical with those used
in Study 1 and are described in detail in the Study 1 section. These include
desired career changes from age 37 to age 43 (asked with an identical
question and coded by the same expert coders used in the Michigan
sample; any desired career-relevant changes, n = 45; no desired careerrelevant changes, n = 31); contextual barriers to change at age 37—
namely, responsibility for children under the age of five (yes, n = 35; no,
n = 41), single parenthood (yes, n = 10; no, n = 66), unemployed spouse
contributing no income (yes, n = 4; no, n = 72), and spouse with a
high-level and demanding executive or professional position (yes, n = 38;
no, n = 38; sum across barriers M = 1.14, SD = 0.78, range = 0 to 3);
symptoms of psychological distress at ages 33 and 43 (anxiety at age 33
M = .38, SD = .06; anxiety at age 43 M = .39, SD = .07; depression at
age 33 M = .38, SD = .07; depression at age 43 M = .40, SD = .09);
physical well-being at age 47 (M = - . 0 1 , SD = .89, a = .78); life
satisfaction at ages 37 and 48 (age 37 M = 2.48, SD = 0.55; age 48
M = 2.49, SD = 0.60); adult adjustment at age 43 (M = 4.69, SD = 1.31);
and Q-sort measures of effective instrumentality and rumination at age 43
(interjudge reliabilities for this sample were equivalent to those for Michigan: range = .60 to .91, M = .77; effective instrumentality M = 6.81,
SD = 1.35, scale a = .89; rumination M = 3.48, SD = 1.06, scale
a = .78).
In addition, although the coding systems for traditional role regrets and
career-related goals were identical to those used in the Michigan sample
(and the coding was completed by the same expert coders), the questions
themselves were slightly different. The measurement of these variables in
the Radcliffe sample are discussed in detail below.
Regrets about pursuing traditional roles for women. The item used to
assess traditional role regrets in this sample at age 37 (in 1979) was "If you
had it to do over again, would you choose the same lifestyle pattern—with
respect to your home versus career decision(s)?" (yes, n = 26; no, n = 50).
277
Examples of traditional role regrets for this cohort were "I would have
planned a career other than teaching, because now I know that you can do
other things and be a mother—if you marry a helpful mate," "I would
change my devotional attitude towards my husband—which I assumed
from my mother," and "I regret not having a career." Examples of regrets
that were not about traditional roles were "I would have gotten married and
had children," "I might just have picked a career without night call" (from
a physician), and "I wouldn't change anything—no one in their right mind
would stay home and change diapers if they could be a scientist instead"
(an example of no regret).
As in the Michigan sample, the majority of these women expressed
regret in one specific area (85%) whereas a minority (15%) expressed two
regrets. For the Radcliffe cohort, traditional role regrets overwhelmingly
focused on career (65%), with far fewer in the area of education (19%; in
contrast to the Michigan cohort, in which traditional role regrets were more
evenly distributed between these areas). Similar to the Michigan sample,
smaller proportions of the Radcliffe women's traditional role regrets focused on marriage (8%) and parenting (27%). Among women without
regrets about traditional roles, 81% had no regrets at all. Of the remaining
19%, roughly one third indicated that they would change careers, one third
indicated that they would make parenting changes, and one third indicated
that they did not have enough leisure time.
Because both of the slightly different questions used in the two samples
were collected in the 1992 wave of the Michigan study, we were able to
address the issues of measure similarity and reliability to some extent.
When the reliability of the regret items identical to those in the Michigan
(at age 36) and Radcliffe (at age 37) samples was analyzed as a two-item
scale in the later Michigan data, they had an alpha coefficient of .72. Not
only is this a highly reliable coefficient for a two-item measure, it is not
appreciably lower than the internal consistency of the five-item measure of
traditional role regret reported earlier (a = .76). In addition, the item used
in the Radcliffe sample was fairly highly correlated (r = .48) with a
reliable {et = .74) scale composed of the other four life review items asked
of the Michigan sample in 1992. These findings foster confidence in the
regret measure used in the Radcliffe sample and also support the notion
that although questions were worded slightly differently for the two samples, they captured fairly comparable responses.
Career-related goals at age 37. In 1979, the women were asked, "If
you could do anything you wished in the next ten years, what would you
do?" For this sample, the coding of career-related goals produced the
following results: none, n = 34; incremental change desired, n = 20;
major change desired, n = 22.
Comparison of the Samples
In order to aid interpretation of results, in the two studies, we
conducted analyses comparing the samples on key variables (i.e., traditional role regret, career-related goals, career-relevant changes, barriers to change, well-being, and personality; see Table 3). As can be
seen in Table 3, nearly twice as many women in the Michigan sample
had traditional role regrets in their late thirties compared with the
Radcliffe sample. However, the samples did not differ in terms of
having career-related goals at that age period nor in terms of having
made career-relevant changes by their mid-forties. On the other hand,
the proportions of women with varying combinations of traditional role
regret and career changes were significantly different in the two cohorts. In the Michigan sample, the largest percentage of women who
made career-relevant changes were those with traditional role regrets; in
the Radcliffe sample, the largest percentage of women who made
career-relevant changes had no traditional role regrets.
In addition, women in the Radcliffe sample experienced significantly
more contextual barriers to change than did women in the Michigan
sample. In terms of the presence of individual barriers to change in their
mid-thirties, significantly more Radcliffe women were single parents—
278
STEWART AND VANDEWATER
Table 3
Comparison of Michigan and Radcliffe Samples on Key Variables
Michigan
(n = 83)
Variable
Radcliffe
(n = 76)
Significance test
a
Regrets, goals, and change
Traditional role regret
Career-related goalsb
Desired career-relevant changes
Traditional role regret by career change
No regret-no desired change
No regret-desired change
Regret-no desired change
Regret-desired change
Contextual barriers to change
Well-being0
Depression
Anxiety
Physical well-being
Life satisfaction
Adult adjustment
Personality11
Effective instrumentality
Rumination
Late thirties life outcomes0
Marital status
Single, never married
Married
Divorced or separated
Became a mother*
Number of children
Work status8
Not working for pay
White collar job
Minor professional
Major professional
Education
Bachelor's level
Master's level
Doctoral level
Personal income
61% (51)
52% (42)
69% (57)
36% (28)
55% (42)
59% (45)
13% (11)
25% (21)
18% (15)
44% (36)
.84 (.72)
26%
37%
13%
24%
1.14
.41 (.07)
.45 (.07)
.10 (.87)
2.46 (0.67)
3.84 (1.29)
.40 (.09)
.39 (.07)
-.01 (.89)
2.49 (0.55)
4.69(1.31)
6.81 (1.13)
3.45 (0.82)
6.81 (1.35)
3.48 (1.06)
10% (8)
83% (69)
7% (6)
76% (63)
1.55(1.10)
. 5% (4)
82% (62)
13% (10)
80% (61)
1.81 (1.37)
24%
35%
22%
19%
18% (14)
25% (19)
16% (12)
41% (31)
(22)
(28)
(9)
(17)
(.78)
^ ( l , N = 159) = 11.78***
ns
ns
^ ( 1 , N = 159) = 10.32**
r(157) = -2.53*
ns
?(157) = -4.35***
ns
ns
f(157) = 4.14***
ns
ns
ns
(20)
(29)
(18)
(16)
ns
ns
^ ( 3 , N = 159) = 8.83*
ns
48% (40)
29% (24)
23% (19)
$14,668.75
(14,299.23)
37% (28)
37% (28)
26% (19)
$15,539.47
(14,102.43)
ns
Note. For percentages, numbers in parentheses indicate the number of participants associated with each
percentage; for means, numbers in parentheses indicate the standard deviation associated with each mean.
a
Michigan sample regrets and goals at age 36; desired career-relevant changes from age 36 to age 47. Radcliffe
sample regrets and goals at age 37; desired career-relevant changes from age 37 to age 43. Percentages and
sample sizes refer to number of participants who have traditional role regrets, or who made desired careerrelevant changes, and so forth.
b
Women with either incremental or major career goals were combined in this analysis as indicative of having
career-related goals.
c
Michigan well-being measured at age 47. Radcliffe depression, anxiety, and adult adjustment measured at age
43; physical well-being and life satisfaction measured at age 48.
d
Personality measured at age 47 in Michigan and age 43 in Radcliffe.
e
All life outcome variables measured at age 36 in Michigan and age 37 in Radcliffe.
f
Includes biological and adopted children.
s
From HoIIingshead and Redlich (1956). Examples of white collar are administrative or clerical jobs, examples
of minor professionals are certified public accountants or social workers, and examples of major professionals
are physicians or college professors.
*/>< .05. * * / > < . 0 1 . * * * / ? < . 0 0 1 .
13% versus 4% in the Michigan sample, r(157) = 2.17, p < .05—and had
spouses with extremely demanding careers—76% versus 35% in the Michigan sample, f(157) = 4.84, p < .001. However, they did not differ on the
other two barriers we examined. Thus, it appears that the larger number of
total barriers among the Radcliffe women in early midlife is largely due to
sample differences in these two particular contextual barriers.
Compared with the Radcliffe cohort, the Michigan cohort was more
anxious and scored lower on adult adjustment. However, the samples did
not differ on well-being measures of depression, physical well-being, or
life satisfaction, nor did they differ on personality resources.
In an effort to provide a context for differences in rates of traditional
role regret in the Michigan and Radcliffe cohorts, we conducted further
analyses examining differences between them on early work and
family-life outcomes (at ages 36 and 37, respectively; see Table 3).
279
MIDLIFE REVIEW AND WELL-BEING
These analyses indicated that the Radcliffe sample had significantly
higher occupational status than the Michigan sample at the time when
traditional role regrets were measured. However, the samples did not
differ on any other work or family-life outcomes in their late thirties,
including marital status, motherhood, number of children, educational
achievement, or personal income.
Analysis Plan
Because we wanted to replicate our findings from the Michigan sample,
we conducted identical analyses, using nonparametric statistics, bivariate
correlations, and planned comparisons where appropriate. Because this is a
replication, we report these findings in terms of one-tailed significance
levels. Because we also wanted to extend the findings to capitalize on the
longitudinal data available only for the Radcliffe sample, we conducted
path analyses assessing the role of age 43 personality as a mediator of the
impact on age 48 well-being of regret at age 37 and change by age 43
(Baron & Kenny, 1986), controlling for age 37 well-being.
Results
In order to examine our first hypothesis in the slightly older and
less regretful Radcliffe women, we analyzed the relationship between traditional role regrets and career goals in the Radcliffe
sample at age 37, as well as the relationship between regrets,
contextual barriers, and career-relevant changes from age 37 to
age 43 (see Table 4). As was true for the Michigan sample, these
analyses indicated that having traditional role regrets was related
to setting career-related goals, )f(2, N = 16) = 6.70, p < .05, but
not to making desired career-relevant changes, ^ ( l , N = 76)
= 0.62, ns. However, unlike the Michigan sample, Radcliffe
women with regrets at age 37 who made desired career changes
had more total contextual barriers at age 37 than women with
regrets who did not make changes, r(24) = —3.52, p < .01.
Obviously, the presence of "barriers" did not account for women's
change-making.
Our next step was to examine differences among the women
grouped according to regrets about traditional roles and careerrelevant changes on the various indices of midlife well-being and
personality (see Table 5 for group means and standard deviations).
In contrast to the Michigan sample, the two groups of women with
traditional role regrets (who either did or did not make later
career-relevant changes) in the Radcliffe sample did not differ in
terms of depression or anxiety. However, like the Michigan sample, women with traditional role regrets who did not make careerrelevant changes reported lower levels of physical health, f(67)
= 1.82, p < .05, in midlife. Finally, in this sample, women with
traditional role regrets who did not make later career-relevant
changes scored lower on adult adjustment than women with such
regrets who did make later changes, t(72) = 1.94, p < .05.
Again, this overall pattern of results was repeated when women
with traditional role regrets who did not make later career-relevant
changes were compared with the other three groups as a whole.
That is, they did not differ on measures of depression or anxiety.
However, they felt significantly less healthy compared with others,
?(67) = 2.28, p < .01, and they scored lower on adult adjustment,
f(72) = 2.29, p< .01.
To rule out the impact of earlier well-being in this sample, we
performed these same analyses using life satisfaction at age 37 as
a covariate for all well-being variables at age 43 and age 48.
Again, using earlier life satisfaction as a covariate did not change
the results of the well-being analyses reported above. In this
sample we also had depression and anxiety scores at age 33 for a
subsample of women (n = 46). Using these variables as covariates
did not change the results of the well-being analyses reported
above. Moreover, the regret-no change group was not different
from the two comparison groups on these measures of earlier
well-being.
In terms of personality resources (see Table 5 for group means
and standard deviations), the results of the planned comparisons
indicated that women in this sample with traditional role regrets at
age 37 who did not make desired career-relevant changes from
age 37 to 43 scored significantly lower on effective instrumentality
at age 43, f(72) = 2.68, p < .01, and significantly higher on
rumination, r(72) = —3.02, p < .01, than women with traditional
role regrets who did make such changes over that period. This
Table 4
Radcliffe Sample: Relationships Among Traditional Role Regret, Career-Related Goals, Desired Career-Relevant Changes,
and Contextual Barriers to Change
Career-related goals (age 37)
Variable (age 37)
Absence of
traditional role
regret
Presence of
traditional role
regret
Mean contextual
barriers to
change
Desired career-relevant changes
(age 37 to age 43)
None
Incremental
change desired
Major change
desired
No desired
changes made
Any desired
changes made
50% (25)
30% (15)
20% (10)
44% (22)
56% (28)
35% (9)
19% (5)
46% (12)
35% (9)
65% (17)
Traditional role regret (age 37)
by desired career-relevant changes
(age 37 to age 43)
Regret-no change
(n = 9)
Regret with change
(n = 17)
0.60(0.69)
1.50(0.62)
Note. For career-related goals and desired career-relevant changes, row percentages add up to 100; numbers in parentheses indicate the number of
participants associated with each percentage. For contextual barriers to change, numbers in parentheses indicate the standard deviation associated with each
280
STEWART AND VANDEWATER
Table 5
Radcliffe Sample: Planned Comparisons of Traditional Role Regret (at Age 37) by Desired Career-Relevant Change Groups
(at Age 37 to 43) on Well-Being and Personality in Midlife
Comparison 1
Regret-no change
Variable
Well-being"
Depression
Anxiety
Adult adjustment
Physical well-being
Life satisfaction
Personality (age 43)
Effective instrumentality
Rumination
Comparison 2
Regret with change
Regret-no change
All other groups
M
SD
M
SD
M
SD
M
SD
.41
.42
3.70a
-.65a
2.33
.13
.11
1.88
1.01
0.50
.44
.42
4.66b
2.56
.07
.06
1.18
0.70
0.62
.41
.42
3.70a
-0.65 a
2.33
.13
.11
1.88
1.01
0.50
.40
.39
4.84,,
0.08h
2.52
.09
.07
1.14
0.85
0.57
5.28a
4.62a
1.81
1.62
6.55h
3.44,,
1.30
0.78
5.28a
4.61 a
1.81
1.62
7.05b
3-31,,
1.11
0.84
o.oih
Note. Means with different subscripts differ significantly at p < .05. Regret-no change, n = 9; regret with change, n = 17; no regret-no change,
n = 22; no regret with change, n = 28.
a
Depression, anxiety, and adult adjustment measured at age 43; physical well-being and life satisfaction measured at age 48.
pattern of results was repeated when women with traditional role
regrets who did not make desired career changes were compared
with the other three groups as a whole: effective instrumentality,
r(72) = 4.17, p < .01; rumination, ((72) = -3.84, p < .01.
Finally, because the age 48 well-being results did suggest lower
well-being among women who had regrets and did not make
changes by age 43, we performed regression analyses to evaluate
the hypothesized role of personality as a mediator of the relationship between regret with or without desired career-relevant change
and well-being. In Figure 1, we present results of two sets of path
analyses, in which the role of each of the two personality variables
(effective instrumentality and rumination) was assessed as a mediator of the relationship between a dummy variable indicating
regret at age 37 without change by age 43, and age 48 well-being
(assessed in terms of physical health), controlling for age 37
well-being. Using Baron and Kenny's (1986) procedure for estimating the mediational effect of a variable, we found that once the
personality variables were inserted into the paths, the direct relationship between traditional role regret without change and wellbeing was reduced, whereas strong relationships between the
dummy variable and personality, and between personality and
well-being, remained.5
Discussion
Results of these analyses did largely replicate the results with
the Michigan sample. First, we found exactly the same pattern of
results for goal setting. Early midlife regret seemed to motivate
goal setting but was not associated with actual life change by
age 43. Moreover, those who made changes based on their regrets
did not face fewer contextual barriers to change than those who did
not. We had weaker results for well-being in the Radcliffe sample,
with differences emerging for physical well-being and adult adjustment, but not psychological distress. Two differences between
the Michigan and the Radcliffe studies seem likely to account for
these weaker results in the Radcliffe sample: Overall, there were
significantly fewer women with traditional role regrets in the
Radcliffe sample, and the absolute number was small (26); and the
Radcliffe sample overall was significantly higher in most measures
of well-being than the Michigan sample. Finally, with respect to
personality resources, the regret-no change group (in this sample
as in the Michigan sample) was higher in rumination and lower in
effective instrumentality than the other three groups and than the
women who had regrets and did make changes.
The path analyses provide support for the hypothesis that the
women with regrets who did not successfully make changes are at
least partially impeded by personality characteristics that block
them. Specifically, the relationship between regret-no change and
physical well-being was mediated by both rumination and effective instrumentality.6
General Discussion
The results of these two studies provide strong support for the
notion that acknowledgement of life regrets in early middle age
(the result, we propose, of a process of midlife review) can
motivate goal setting. However, it is also clear that regret alone is
insufficient to bring about actual life changes, just as external
barriers are insufficient to prevent them. If desired changes are
brought about, though, it is also clear that there is a payoff in
well-being in later middle age. For that reason, it is important to
identify the conditions that may either interfere with or enhance
5
As has been noted several times, at age 37 the only well-being
indicator was a global rating of life satisfaction. Although this measure did
not differentiate between the group of women with regrets who did not
make changes and other women, it was significantly correlated with age 48
physical well-being (r = .32, p < .01). For that reason, though it is
certainly an imperfect control, it seemed useful.
6
Analyses revealed that because there was no direct path between regret
with no change and life satisfaction at age 48, personality was not a
mediator of their connection. However, regret with no change was indirectly related to later life satisfaction through personality. Having regrets
and not making changes was related to more rumination and less effective
instrumentality, respectively, which were in turn related to lower life
satisfaction in later midlife.
281
MIDLIFE REVIEW AND WELL-BEING
Effective
Instrumentality N
Age43
\
Regret Age 37 with
No Changes Made
Age 37-43
B
-.21' /-.07n.s.
Rumination
Age 43
.42*
Regret Age 37 with
No Changes Made
Age 37-43
-.21* /-.06n.s.
.28**/.25*
Physical
Well-being
Age 48
-.37*** / -.35*'
Physical
Well-being
Age 48
Figure 1. Radcliffe sample: Personality as a mediator between traditional
role regret at age 37 with no subsequent changes and later physical
well-being. Panel A = effective instrumentality as the mediator; Panel B =
rumination as the mediator. Path coefficients were generated from hierarchical regressions controlling for life satisfaction at age 37. Regret-change
was dummy coded (1 = regret with no subsequent changes, 0 = all other
groups). Paired coefficients indicate results of mediation test steps (first
coefficient = path with no mediator; second coefficient = path with
personality as mediator). 'p < .07 (marginally significant). *p <
.05. * * / ? < . 0 1 . ***/>< .001.
the probability that midlife regrets will be transformed into desired
life changes.
We explored the role of two personality characteristics posited
to make a difference: effective instrumentality and rumination.
Both were correlated (in opposite directions, as predicted) with
successful transformation of regret into change; path analyses with
longitudinal data from the Radcliffe sample suggested that both
variables played a role in mediating that relationship. Consistent
with these findings, Lyubomirksy and Nolen-Hoeksema (1995)
found that people who ruminate not only are pessimistic about
their ability to overcome obstacles but also generate less effective
solutions to problems. Although it may be difficult to influence the
tendency to ruminate directly, it may be more feasible to increase
effective problem solving.
It should of course be noted that we cannot be sure of the causal
role of any of the variables we have examined. The availability of
longitudinal data, and some control variables, strengthens our
confidence in the causal relations we have hypothesized. However,
it is entirely possible that unmeasured variables actually account
for the relationships among having regrets, not making changes,
and later midlife unhappiness. Thus, for example, poor health
might prevent a woman who had acknowledged regrets from
making life changes that she desired; it might be that poor health
in turn (rather than the fact of regret without changes) that was
related to the ill health we assessed a decade later. We attempted
to rule out as many of these explanations as possible (by drawing
on age 37 well-being data and assessing contextual barriers to
change), but we certainly cannot rule them all out.
We also do not know whether these findings are limited mainly
to the kinds of regrets these women expressed: regrets about
sacrificing involvement in school and work roles in favor of family
roles. Previous research suggests that these kinds of regrets are the
most common ones American women have expressed in the past,
but of course this may be changing for future generations. It would
be very interesting, in a sample or cohort with sufficient numbers,
to compare the consequences of regrets about sacrificing career in
favor of family with regrets about, for example, sacrificing family
for career or about failing to develop certain interests or talents.
This comparison has been impossible in past and current generations of women but may well become more feasible in future ones,
as well as in studies of men.
Educated women who came of age in the late 1960s and early
1970s were the first American middle-class women to be able to
take advantage of new opportunities for education and careers. At
the same time, this generation of women was socialized in childhood and early adulthood to a traditional female role as child rearer
and culture bearer. It is clear that some women in this generation
managed the shift in norms and values without regrets, some
without making changes. Still others were able to use regrets about
their past traditional role choices to make changes in middle age.
There is, though, a group of women who had regrets about their
past choices but did not make changes in their lives; they seemed
quite discontented in later middle age. The evidence from these
two studies suggests that these women did not face more external
obstacles to change in their lives, but they faced internal psychological ones, particularly a tendency to ruminate. It is unclear
whether these women will be able to create a more satisfactory
adaptation in their fifties or later. Future research should assess this
possibility, as well as the degree to which these findings apply to
men, to other generations of women, and to women with fewer
economic and social resources.
References
Andrews, F., & Robinson, J. P. (1991). Measures of subjective well-being.
In J. P. Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.), Measures
of personality and social psychological attitudes (pp. 61-110). San
Diego, CA: Academic Press.
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral
change. Psychological Review, 84, 191-215.
Bandura, A. (1986). The explanatory and predictive scope of self-efficacy
theory. Journal of Social and Clinical Psychology, 4, 359—373.
Bandura, A., & Schunk, D. H. (1981). Cultivating competence, selfefficacy and intrinsic interest through proximal self-motivation. Journal
of Personality and Social Psychology, 41, 586-598.
Baron, R. M , & Kenny, D. A. (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182.
Baruch, G., Barnett, R., & Rivers, C. (1983). Lifeprints. New York:
McGraw-Hill.
Bird, C. E., & Fremont, A. M. (1991). Gender, time use, and health.
Journal of Health and Social Behavior, 32, 114-129.
282
STEWART AND VANDEWATER
Block, J. (1978). The Q-sort method in personality assessment and psychiatric research. Palo Alto, CA: Consulting Psychologists Press. (Original work published 1961)
Block, J., & Haan, N. (1971). Lives through time. Berkeley, CA: Bancroft
Books.
Boninger, D. S., Gleicher, F., & Strathman, A. (1994). Counterfactual
thinking: From what might have been to what may be. Journal of
Personality and Social Psychology, 67, 297-307.
Butler, R. (1974). Successful aging and the role of the life review. Journal
of the American Geriatric Society, 22, 529-535.
Cantor, N. (1990). From thought to behavior: "Having" and "doing" in the
study of personality and cognition. American Psychologist, 45, 735-750.
Chiriboga, D. (1989). Mental health at the midpoint: Crisis, challenge or
relief? In S. Hunter & M. Sundel (Eds.), Midlife myths (pp. 116-144).
Newbury Park, CA: Sage.
Cole, E. R., & Stewart, A. J. (1996). Meanings of political participation
among Black and White women: Political identity and social responsibility. Journal of Personality and Social Psychology, 71, 130-140.
Coleman, P. G. (1974). Measuring reminiscence characteristics from conversation as adaptive features of old age. Journal of Aging and Human
Development, 5, 281-294.
de Charms, R. (1976). Enhancing motivation: Change in the classroom.
New York: Irvington.
DeGenova, M. K. (1992). If you had your life to live over again, what
would you do differently? International Journal of Aging and Human
Development, 34, 135-143.
Erikson, E. H., Erikson, J. M., & Kivnick, H. Q. (1986). Vital involvement
in old age. New York: Norton.
Erskine, H. (1973). The polls: Hopes, fears, and regrets. Public Opinion
Quarterly, 37, 132-145.
Franz, C. E., McClelland, D. C , & Weinberger, J. (1991). Childhood
antecedents of conventional social accomplishment in midlife adults: A
36-year prospective study. Journal of Personality and Social Psychology, 60, 586-595.
Ginzberg, E., & Associates. (1966). Educated American women. New
York: Columbia University Press.
Helson, R. (1992). Women's difficult times and the rewriting of the life
story. Psychology of Women Quarterly, 16, 331-347.
Helson, R., & Picano, J. (1990). Is the traditional role bad for women?
Journal of Personality and Social Psychology, 59, 311-320.
Hollingshead, A., & Redlich, C. (1956). Social class and mental illness: A
community study. New York: Wiley.
Hulbert, K. D., & Schuster, D. T. (Eds.). (1993). Women's lives through
time: Educated American women of the twentieth century. San Francisco: Jossey-Bass.
Jenkins, S. R. (1989). Longitudinal prediction of women's careers: Psychological, behavioral, and social-structural influences. Journal of Vocational Behavior, 34, 204-235.
Jenkins, S. R. (1994). Need for power and women's careers over 14 years:
Structural power, job satisfaction, and motive change. Journal of Personality and Social Psychology, 66, 155—165.
Jung, C. G. (1954). Marriage as a psychological relationship. In W.
McGuire (Ed.), The development of personality: The collected works of
C. G. Jung (Vol. 17, pp. 187-201). New York: Pantheon Books. (Original work published 1925)
Karoly, P. (1993). Mechanisms of self-regulation: A systems view. Annual
Review of Psychology, 44, 23-52.
Landman, J. (1993). Regret: The persistence of the possible. New York:
Oxford University Press.
Landman, J., & Manis, J. (1992). What might have been: Counterfactual
thought concerning personal decisions. British Journal of Psychology, 83, 473-477.
Landman, J., Vandewater, E. A., Stewart, A. J., & Malley, J. E. (1995).
Missed opportunities: Psychological ramifications of counterfactual
thought in midlife women. Journal of Adult Development, 2, 87—97.
Langer, E. J. (1983). The psychology of control. Beverly Hills, CA: Sage.
Lecci, L., Okun, M. A., & Karoly, P. (1994). Life regrets and current goals
as predictors of psychological adjustment. Journal of Personality and
Social Psychology, 66, 731-741.
Levinson, D. (1978). The seasons of a man's life. New York: Ballantine
Books.
Levinson, D. (1996). The seasons of a woman's life. New York: Knopf.
Lyubomirsky, S., & Nolen-Hoeksema, S. (1995). Effects of self-focused
rumination on negative thinking and interpersonal problem solving.
Journal of Personality and Social Psychology, 69, 176-190.
Meacham, J. (1977). A transactional model of remembering. In N. Datan
& H. Reese (Eds.), Life-span developmental psychology: Dialectical
perspectives on experimental research (pp. 261—283). New York: Academic Press.
Metha, A. T., Kinnier, R. T., & McWhirter, E. H. (1989). A pilot study of
the regrets and priorities of women. Psychology of Women Quarterly, 13, 167-174.
Niedenthal, P. M., Tangney, J. P., & Gavanski, I. (1994). "If only I
weren't" versus "If only I hadn't": Distinguishing shame and guilt in
counterfactual thinking. Journal of Personality and Social Psychology, 67, 585-595.
Nolen-Hoeksema, S. (1991). Responses to depression and their effects on
the duration of depressive episodes. Journal of Abnormal Psychology,
100, 569-582.
Nolen-Hoeksema, S., Parker, L., & Larson, J. (1994). Ruminative coping
with depressed mood following loss. Journal of Personality and Social
Psychology, 67, 92-104.
'
Okun, M., & George, L. (1984). Physician and self ratings of health:
Neuroticism and subjective well-being among men and women. Personality and Individual Differences, 5, 533—539.
Piaf, E*. (No date). Non, je ne regrette rien. On The voice of the sparrow:
The very best of Edith Piaf [CD]. New York: Capitol Records.
Picano, J. (1986). Development and validation of a life history index of
adult adjustment for women. Journal of Personality Assessment, 53,
308-318.
Robinson, J. P., Shaver, P. R., & Wrightsman, L. S. (1991). Measures of
personality and social psychological attitudes. San Diego, CA: Academic Press.
Roese, N. J. (1994). The functional basis of counterfactual thinking.
Journal of Personality and Social Psychology, 66, 805-818.
Rotter, J. B. (1966). Generalized expectancies for internal versus external
control of reinforcement. Psychological Monographs, 80{\, Whole No.
609).
Ryff, C. (1989). Happiness is everything or is it? Explorations on the
meaning of psychological well-being. Journal of Personality and Social
Psychology, 57, 1069-1081.
Sears, P. S. (1979). The Terman genetic studies of genius, 1922-1972. In
A. H. Passow (Ed.), The gifted and talented: Their education and
development (pp. 75-96). The 78th Yearbook of the National Society for
the Study of Education. Chicago: University of Chicago Press.
Seligman, M. E. P. (1975). Helplessness. San Francisco: Freeman.
Stewart, A. J. (1974). Longitudinal prediction from personality to life
outcomes among college-educated women. Unpublished doctoral dissertation, Harvard University.
Stewart, A. J. (1978). A longitudinal study of coping styles of self-defining
and socially defined women. Journal of Consulting and Clinical Psychology, 46, 1079-1084.
Stewart, A. J. (1980). Personality and situation in the prediction of women's life patterns. Psychology of Women Quarterly, 5, 195-206.
Stewart, A. J., & Salt, P. (1981). Life-stress, life styles, depression and
illness in adult women. Journal of Personality and Social Psychology, 40, 1063-1069.
MIDLIFE REVIEW AND WELL-BEING
Stewart, A. J., & Vandewater, E. A. (1993). The Radcliffe Class of 1964:
Career and family clocks in a transitional cohort. In K. Hulbert & D.
Schuster (Eds.), Women's lives throughtime:Educated American women of
the twentieth century (pp. 235-258). San Francisco: Jossey-Bass.
Tangri, S. S. (1969). Role-innovation in occupational choice among college women. Unpublished doctoral dissertation, University of Michigan.
Tangri, S. S. (1972). Determinants of occupational role-innovation among
college women. Journal of Social Issues, 28, 177-200.
Tangri, S. S., & Jenkins, S. R. (1986). Stability and change in role
innovation and life plans. Sex Roles, 13, 647—662.
Tangri, S. S., & Jenkins, S. R. (1993). The University of Michigan Class
of 1967: The Women's Life Paths Study. In K. Hulbert & D. Schuster
(Eds.), Women's lives through time: Educated American women of the
twentieth century (pp. 259-281). San Francisco: Jossey-Bass.
Thoreau, H. D. (1949). The Journals of Henry D. Thoreau (Vol. 1). Boston:
Houghton Mifflin. (Original work published 1906)
Vaillant, G. (1977). Adaptation to life. Boston: Little, Brown.
Vandewater, E. A., Ostrove, J. M., & Stewart, A. J. (1997). Predicting
283
women's well-being in midlife: The importance of personality development and social role involvements. Journal of Personality and Social
Psychology, 72, 1147-1160.
Yohalem, A. (1993). Columbia University graduate students, 1945-51:
The vanguard of professional women. In K. Hulbert & D. Schuster
(Eds.), Women's lives through time: Educated American women of the
twentieth century (pp. 140-157). San Francisco: Jossey-Bass.
Zung, W. K. (1965). A self-rating depression scale. Archives of General
Psychiatry, 12, 63-70.
Zung, W. K. (1971a). A differentiation of anxiety and depressive disorders:
A biometric approach. Psychosomatics, 12, 380—384.
Zung, W. K. (1971b). A rating instrument for anxiety disorders. Psychosomatics, 12, 371-379.
Received February 25, 1998
Revision received August 3, 1998
Accepted August 22, 1998 •