Psychosocial development from college through midlife: a 34-year sequential study.

Abstract

Two cohorts of alumni, leading-edge and trailing-edge baby boomers, first tested in their college years, were followed to ages 43 (N = 136) and 54 (N = 182) on a measure of Erikson's theory of psychosocial development. Hierarchical linear modeling was used to model the trajectory of growth for each psychosocial issue across middle adulthood. As predicted, the early psychosocial issues (trust, autonomy, and initiative) showed patterns of slow and steady increases in favorable resolution, as did the midlife issue of generativity. Industry, found in earlier investigations on the samples to change to differing degrees by cohort, continued to show cohort differences through midlife. The quadratic terms indicated that growth was curvilinear for both cohorts on identity and intimacy, and ego integrity showed variations by cohort, with the older cohort showing steeper patterns of increases. Gender differences were observed on intimacy, with women receiving higher initial scores, but the curves showed deceleration through midlife. Tests of variations in growth curves by the life history variables of educational attainment, occupational prestige, commitment to a long-term relationship, and parenthood status showed variations by cohort, but a general pattern of catching up emerged in which those who entered early adulthood at a relative disadvantage in terms of psychosocial development were able to attain favorable outcomes by midlife.

Full text

Page 1

Psychosocial Development From College Through Midlife:
A 34-Year Sequential Study
Susan Krauss Whitbourne
University of Massachusetts Amherst
Joel R. Sneed
Queens College, City University of New York; Columbia
University; and New York State Psychiatric Institute
Aline Sayer
University of Massachusetts Amherst
Two cohorts of alumni, leading-edge and trailing-edge baby boomers, first tested in their college years,
were followed to ages 43 (N ? 136) and 54 (N ? 182) on a measure of Erikson’s theory of psychosocial
development. Hierarchical linear modeling was used to model the trajectory of growth for each
psychosocial issue across middle adulthood. As predicted, the early psychosocial issues (trust, autonomy,
and initiative) showed patterns of slow and steady increases in favorable resolution, as did the midlife
issue of generativity. Industry, found in earlier investigations on the samples to change to differing
degrees by cohort, continued to show cohort differences through midlife. The quadratic terms indicated
that growth was curvilinear for both cohorts on identity and intimacy, and ego integrity showed variations
by cohort, with the older cohort showing steeper patterns of increases. Gender differences were observed
on intimacy, with women receiving higher initial scores, but the curves showed deceleration through
midlife. Tests of variations in growth curves by the life history variables of educational attainment,
occupational prestige, commitment to a long-term relationship, and parenthood status showed variations by
cohort, but a general pattern of catching up emerged in which those who entered early adulthood at a relative
disadvantage in terms of psychosocial development were able to attain favorable outcomes by midlife.
Keywords: adulthood, psychosocial, personality, midlife, generativity
Erik Erikson (1963) provided the first widely accepted perspec-
tive in developmental psychology to hold that personality changes
throughout life. However, with its basis in clinical observation and
emphasis on diverse processes spanning many decades, it has
proven notoriously difficult to put to a rigorous empirical test. As
an offshoot of psychodynamic theory, Erikson’s work is subject to
challenges that face its larger perspective and, in a time of
computerized assessments and manualized therapy, has waned
in popularity within the larger fields of personality and clinical
psychology.
Despite these challenges, Erikson’s theory retains its appeal
within areas of personality psychology that focus on specific
developmental periods. The area of identity development in ado-
lescence, for example, seems to have benefited from the Erikso-
nian perspective, which has stimulated widespread research on the
emergence and subsequent shaping of the individual’s sense of self
from the teenage years to the early 20s (Schwartz, 2006), if not
throughout adulthood (Helson & Srivastava, 2001; Marcia, 2002;
Zucker, Ostrove, & Stewart, 2002), and has been extended to
explain the development of ethnic and racial identity (Sneed,
Schwartz, & Cross, 2006). Studies that focus on other Eriksonian
concepts, such as generativity and ego integrity, have provided
valuable insights as well as empirical data regarding developmen-
tal processes in midlife (Bradley & Marcia, 1998; de St. Aubin,
McAdams, & Kim, 2004; James & Zarrett, 2005).
Erikson’s Psychosocial Theory
The thrust of Eriksonian theory is that personality continually
evolves throughout life, a view that contrasts sharply with the
increasing emphasis in the field on trait theory, which proposes
that personality remains largely stable as an expression of consti-
tutionally or innately determined predispositions (Terracciano,
Costa, & McCrae, 2006). In Eriksonian theory, the ego, or con-
scious self, is the central structure of personality. As the ego
evolves in early childhood, it begins to establish certain qualities
that can enhance the range of the individual’s adaptive responses.
Although Erikson’s theory is often characterized as proposing
that there are a fixed set of stages through which the ego develops,
a more accurate representation of the theory is that it proposes a set
Susan Krauss Whitbourne and Aline Sayer, Department of Psychology,
University of Massachusetts Amherst; Joel R. Sneed, Department of Psy-
chology, Queens College, City University of New York; Department of
Psychiatry, Columbia University; and New York State Psychiatric Insti-
tute, New York, New York.
We wish to thank Mark Weinberger, Yariv Hofstein, and Allison Branch
for their preliminary work on this study and Karyn M. Skultety, Kelly M.
Jones, and Taryn Patterson for their assistance in maintaining the partici-
pant database. We also wish to thank the alumni office of the University of
Rochester for its help in locating participants.
Correspondence concerning this article should be addressed to Susan
Krauss Whitbourne, Department of Psychology, University of Massachu-
setts Amherst, Amherst, MA 01003-7710. E-mail: swhitbo@psych
.umass.edu
Developmental Psychology
2009, Vol. 45, No. 5, 1328–1340
© 2009 American Psychological Association
0012-1649/09/$12.00DOI: 10.1037/a0016550
1328

Page 2

of issues that are most likely to be characterized by certain strug-
gles among the biological, psychological, and social forces acting
on the individual and that these issues can become salient at any
point in life. When an issue reaches greatest salience for an
individual, it is referred to as a crisis. The outcome of each crisis
may be either favorable (as in the attainment of identity) or
unfavorable (as in the failure to achieve a coherent identity), and
having arrived at this resolution, at least for the time being, the
individual has more or less of a new ego strength to add to his or
her repertoire of adaptive capacity. Erikson proposed that the
resolution of earlier psychosocial issues forms the basis for reso-
lution of subsequent issues.
In the matrix of psychosocial issues and age periods proposed
by Erikson, the diagonal shows the typical pattern of development
in which particular issues match particular ages (e.g., identity in
adolescence). However, what is usually not emphasized in discus-
sions of his theory is that earlier issues may arise at a later point in
life, and the later ones may move to the forefront in earlier periods
if “hazards of existence” (E. H. Erikson, 1963, p. 274) develop that
stimulate the individual to confront those particular issues. Our
operationalization of Erikson’s theory therefore is based on mea-
suring all psychosocial issues simultaneously, allowing individuals
to be characterized by more or less favorable resolutions on all
eight issues throughout life.
The eight Eriksonian psychosocial issues are as follows: trust
versus mistrust, autonomy versus shame and doubt, initiative ver-
sus guilt, industry versus inferiority, identity achievement versus
identity diffusion, intimacy versus isolation, generativity versus
stagnation, and ego integrity versus despair. According to Erikson,
the first five issues are found in the diagonal of the matrix up
through adolescence; the final three in the middle and later adult
years.
Previous Research From the Rochester Adult Longitudinal
Study (RALS)
As discussed earlier, research based on Erikson’s theory has
tended to focus on specific psychosocial issues, but several inves-
tigations have examined the entire sequence of all eight. The most
extensive study of the entire Eriksonian model is the RALS, begun
in 1966 when Constantinople (1969) administered a questionnaire
measure of psychosocial development to a sample of over 349
students in the classes of 1965–68. The original sample was
followed up 11 years later in 1977 when a new sample of 299
undergraduates was added to allow for sequential analyses in
which the two cohorts were compared (Whitbourne & Waterman,
1979). In the second follow-up in 1988–89, yet another under-
graduate sample was added. Sequential comparisons were then
made among three cohorts of college students and two cohorts of
adults in their early 30s. Longitudinal follow-up analyses were also
conducted on adults from college up to age 42. Additional data
were collected on other measures of identity and life events at each
of the follow-up testings (Van Manen & Whitbourne, 1997; Whit-
bourne, Zuschlag, Elliot, & Waterman, 1992).
The heart of the previous findings from the RALS regarding
psychosocial development was the consistency of age changes
across cohorts in the two psychosocial issues theorized to be most
sensitive to biopsychosocial forces in college and early adulthood:
identity achievement versus identity diffusion and intimacy versus
isolation. Analysis of data from another measure given in 1988 to
respondents from all three cohorts specifically intended to assess
identity development showed convergent support for the notion of
its continued growth during adulthood (Whitbourne & Van Manen,
1996).
The Present Investigation
Earlier follow-ups of the RALS used the statistical techniques
readily available at the time (i.e., repeated measures analysis of
variance) that only allowed for analysis of mean differences rather
than investigation of individual patterns of age and cohort effects
in the sequential design. The present investigation uses hierarchi-
cal linear modeling (HLM; Raudenbush & Bryk, 2002) to track
changes in each of the Eriksonian psychosocial issues over the 34
years of the study. This approach has a number of advantages over
the previous analytic methods. First, the spacing and number of
repeated measurements can vary across individuals. This feature of
HLM was important in the present study because participants
varied by age within cohort. Second, missing data, as existed in the
present study, can be accommodated with the multilevel frame-
work because estimation occurs for each individual separately
using all available data for that particular person. Third, because
multilevel approaches model change at the individual level, they
do not assume that the overall mean pattern over time generalizes
to all individuals in the sample. Consequently, individual differ-
ences in changes are not relegated to the “bin of random error,” as
they are in repeated measures analysis of variance (Helson, Jones,
& Kwan, 2002, p. 756). The theoretical proposition of Erikson’s
theory is that personality change in adulthood varies from person
to person depending on the particular biopsychosocial issues that
bear on the individual, a proposition that was supported in prelim-
inary work on the data from the oldest cohort (Sneed, Whitbourne,
& Culang, 2006). We therefore used a data analytic approach that
accommodates the possibility of individual differences in the level
and rate of change over time.
As in the previous investigations, the early issues of trust versus
mistrust, autonomy versus shame and doubt, and initiative versus
guilt were expected to show stability or at least slow patterns of
growth from young adulthood to midlife. The issue of industry
versus inferiority should theoretically show a similar pattern of
slow growth. However, as is well known in the vocational devel-
opment literature (Robitschek & Woodson, 2006), the process of
identifying with society’s work ethic, the core of this issue in
Erikson’s scheme, continues to evolve through early adulthood.
For identity achievement versus identity diffusion, intimacy versus
isolation, and generativity versus stagnation, we expected initial
increases as these emerged through adulthood, followed by a
leveling off as individuals confront and resolve the developmental
challenges of forming an identity, establishing intimate partner-
ships, and making family and career decisions (Arnett, 2000).
Confronting mortality, making meaning out of life’s accomplish-
ments, and gaining self-acceptance, associated with ego integrity
versus despair (Torges, Stewart, & Duncan, 2008), were expected
to evolve more gradually, showing initial stability and subsequent
increases in mature adulthood.
Thus, our first task was to characterize the trajectories of change
for each psychosocial issue. Given sufficient interindividual dif-
1329
PSYCHOSOCIAL DEVELOPMENT THROUGH MIDLIFE

Page 3

ferences in change, our next task was to examine the impact of
potential moderator variables on these trajectories. In line with the
previous investigations, we examined the effect of birth cohort on
personality development. Our focus in the present investigation on
midlife development led us to include only the two older cohorts,
who were followed from college through ages 43 and 54, respec-
tively. We expected that the differences in social context experi-
enced by these two cohorts would be manifest as differences in
progress through the psychosocial issues (Whitbourne & Willis,
2006). The older cohort (average birth year ? 1946), known as the
“leading edge” baby boomers, reached early adolescence during
the 1950s and early 1960s, which was the height of the Cold War,
a period of increasing prosperity and characterized by traditional
social and work roles. By the time they reached college, though,
the old social order was crumbling, and they were exposed to
widespread questioning and challenging of authority in connection
with the Kennedy assassination, the height of civil rights protests,
the hippie movement, and the Vietnam War. In contrast, the
younger cohort (average birth year ? 1957), known as the “trailing
edge” baby boomers, grew through their early adolescence during
that same tumultuous era but in their own college years were
exposed to a far more conservative and traditional set of values.
Such fundamental differences in social context would be expected
to translate into cohort differences in psychosocial development
(Schoon, Martin, & Ross, 2007).
In particular, we expected the older cohort, whose childhood
was marked by a more traditional upbringing but whose college
years involved rebellion and questioning of the traditional social
order, would have had greater difficulty resolving issues of identity
and intimacy while in college. Ultimately, with the older cohort
having been stimulated to question and challenge authority while
in college, their midlife growth would have a stronger and better
elaborated foundation. In contrast, the younger cohort while in
college may have appeared more mature psychosocially, but per-
haps would have prematurely foreclosed on issues related to career
and family. Hence, in middle adulthood, these trailing-edge baby
boomers would be faced with greater challenges to their earlier,
less well considered life decisions. Within these general social
trends, individual differences in patterns of change were also
expected; the present analyses allowed us to test both variations by
person and variations by cohort.
We also know from the previous RALS follow-up (Whitbourne
et al., 1992) that ego integrity versus despair is an issue sensitive
to sociohistorical effects, and therefore scores for both cohorts
were expected to reflect time-of-measurement effects. Thus, cut-
ting across cohorts, the materialism associated with the 1980s
seemed to have a dampening effect on all participants, not only
from these two cohorts but also from their younger counterparts,
then in college.
The next moderator variable we examined was gender.
Throughout the course of the RALS, gender differences have been
investigated, but previous analyses were limited to comparisons of
means. Using HLM, we were able to examine gender differences
in trajectories of change rather than be restricted to mean-level
analyses. Given the important differences in socialization that have
profound effects on men’s and women’s sense of self (Cross &
Madson, 1997), and the strong gender differences that have been
observed in longitudinal research on personality development
through early adulthood (Kasen, Chen, Sneed, Crawford, &
Cohen, 2006), we expected to see gender differences in both level
and rate of change, with women receiving higher scores on inti-
macy versus isolation and showing a steeper rate of change than
men and the converse pattern holding for men in the more tradi-
tionally masculine qualities of initiative and industry and, to a
lesser extent, identity.
The third set of variables investigated in association with
changes in psychosocial development related to the life history
variables of education, occupation, long-term relationships, and
parenthood. In particular, following up on the RALS findings
reported by Van Manen and Whitbourne (1997), and the research
of others showing the importance of considering life history vari-
ables as predictors of individual differences in change (Helson &
Roberts, 1994; Mroczek & Spiro, 2005), we wished to determine
whether individuals would differ by cohort in their trajectories
through midlife on the psychosocial issues according to the paths
they had set for themselves in the early adult years in terms of
education, prestige, and family variables (Sneed, Johnson, et al.,
2006).
On the basis of the previous findings from the RALS, we
hypothesized that more positive circumstances in the life history
variables at the testing conducted when participants were, on
average, age 31 years would be related to more favorable patterns
of change over time through the midlife years. Thus, we examined
whether participants who were more successful early in life occu-
pationally in terms of having achieved higher levels of education
and prestige would show more favorable patterns of resolution on
the relevant psychosocial issues of initiative versus guilt, industry
versus inferiority, and identity achievement versus identity diffu-
sion than those who were less successful. Second, we investigated
whether individuals who established a stable family life early in
adulthood would show more favorable patterns of resolution for
issues of intimacy versus isolation and generativity versus stagna-
tion. Finally, in keeping with the previous findings from the
RALS, we examined whether gender differences would be found
in these patterns of relationships.
Method
Participants
Alumni who attended the University of Rochester from the
1960s through the 1980s served as the participants in this study,
forming two distinct cohorts, separated in age by an average of 11
years. The portion of the sample who attended the university in the
1960s was first assessed in 1966–68, and those assessed from the
1970s were first tested in 1977–78. The distribution of participants
by cohort and gender for each time of testing is shown in Table 1.
Information on race and ethnicity was not obtained, and therefore
the sample distribution on these variables is not known.
Of the original 349 participants in the 1946 cohort, 124 (35.5%)
had completed their participation only in college, 98 (28.1%) on
two occasions, 42 (12%) on three occasions, and 85 (24.4%) on all
four test occasions. Of the 299 originally in the 1957 cohort, 142
(47.5%) were tested only in college, 95 (31.8%) twice, and 62
(20.7%) at all three follow-ups.
Measures
Inventory of Psychosocial Development (IPD).
nople (1969) developed the IPD in the mid-1960s as a question-
Constanti-
1330
WHITBOURNE, SNEED, AND SAYER

Page 4

naire measure of the first six issues in Erikson’s psychosocial
theory of development. The scales for generativity versus stagna-
tion and ego integrity versus despair were developed in subsequent
research and implemented for the first time in the 1977 testing
(Whitbourne & Waterman, 1979). The IPD yields eight scores,
with five positive items and five negative items contributing to
each score. Respondents are asked to indicate how characteristic or
uncharacteristic a given item is of them on a 7-point Likert scale.
Difference sores are obtained for each scale score by subtracting
the summed score of the negative items from the summed scores
of the positive items. For each psychosocial issue, the range of
possible scores is thus from ?30 to 30. The items reflect the key
components of each issue, such as “never get what I want” (mis-
trust), “comfortable in intimate relationships” (intimacy), and “feel
productive in my work” (generativity). As in previous reports on
the IPD, results are presented in terms of the positive scale names,
reflecting the fact that the difference scores are based on subtract-
ing negative from positive subscale scores.
Demographic information.
Participants were asked to com-
plete a brief open-ended questionnaire that requested information
on their educational, work, and family history since college grad-
uation including dates and names of educational institutions and
employers. For family history, participants provided dates of major
family life events. The form of these questions was identical to the
form of the questions included in the 1977 and 1989 follow-ups.
Analyses involving the relationship between demographic vari-
ables and psychosocial development were made based on status in
early adulthood (i.e., the age 31 testing). Because more partici-
pants were reached in the 2000–2 testing than had been contacted
on previous occasions, and because the data they provided spanned
the entire postcollege period, it was possible to obtain a larger
number of ratings on education, occupation, and family history
from prior years than had been available in earlier testings.
Educational attainment was rated according to a scale of 1 to 6:
1 ? bachelor’s degree only, 2 ? master’s degree, 3 ? Ph.D., 4 ?
degree in dentistry or other non-M.D. health field (e.g., osteo-
pathic medicine), 5 ? law degree, 6 ? M.D. In subsequent
analyses, the sample was divided at the value closest to the median
into low (bachelor’s or master’s degree) versus high (doctorate or
professional degree). Additionally, for reporting purposes of sam-
ple means, the above categories were converted roughly into years
of education as follows, based on the average length of time that
participants reported they had taken to obtain their degrees: 1 ?
16, 2 ? 18, 3 ? 21, 4 ? 20, 5 ? 19, and 6 ? 20.
Occupational prestige scores were obtained with the Nakao–
Treas scale (1994), which ranges from a low of 21 at the level of
blue-collar workers to 97, the prestige rating of physicians. For the
purposes of subsequent analyses, the median prestige rating of 73
(managerial level) was used to group participants into high- and
low-prestige occupations.
Categories of long-term relationships were coded as follows:
1 ? married, 2 ? cohabitating, 3 ? divorced, 4 ? remarried, 5 ?
widowed, 6 ? single, 7 ? separated. In line with Van Manen and
Whitbourne (1997), for the purposes of subsequent analyses, re-
lationship status was coded into the categories of not being in-
volved in a long-term relationship (separated, divorced, single) and
being involved in a long-term relationship (married, remarried, or
cohabitating). Widowed individuals who had not remarried were
not included in these analyses.
Parental status was coded into the categories of parent and
nonparent. Participants who had become stepparents through re-
marriage or cohabitation were considered to be parents if the
children had lived in the home of the participant at some point
during their upbringing.
Data for both cohorts on all demographic variables included in
the present study are summarized in Table 2.
Procedure
The 2000–2 RALS follow-up was initiated in November 1999,
when records were obtained from the most recent edition of the
University of Rochester alumni directory. To make it possible to
request participation from individuals whose information was not
listed in the directory, a fee-based service (Find A Friend) was used
to generate possible addresses from the existing information in the
files.TheseaddresseswereusedinMay2000asthebasisforaninitial
mailing of questionnaire packets. In this first mailing, respondents
were sent a letter describing the study along with the instruments and
a self-addressed stamped envelope to be used for returning the ques-
tionnaires. The initial round of requests for participation yielded 105
participants from the 1946 cohort and 73 from the 1957 cohort. In the
subsequent efforts to locate participants that took place from June
2000 to October 2002, additional resources were used, including
online alumni directories from the university, assistance from the
alumni office, another low-fee name search service (Online Detec-
tive), and other free search services including AltaVista, Google,
Yahoo! Qwest, and Lycos. These efforts produced an additional 77
respondents from the 1946 cohort and 63 from the 1957 cohort. Full
Table 1
Percentages of Original Samples by Sex at Each Year of Testing
Cohort
19661977 19892000
N
% of original
N
% of original
N
% of original
N
% of original
1946
Male
Female
Total
1957
Male
Female
Total
180
169
349
100
100
100
80
76
44
45
45
62
37
99
34
22
28
104
78
182
58
46
52156
143
156
299
100
100
100
43
40
83
30
26
28
57
79
40
51
46136
1331
PSYCHOSOCIAL DEVELOPMENT THROUGH MIDLIFE

Page 5

informationonparticipationbycohortisshowninTable3.Therewereno
significant differences in attrition by cohort, ?2(1) ? .83, p ? .05.
Of the 182 participants in the 1946 cohort, almost all (90%)
were found through the university resources; the remainder was
located through a Web search. A large proportion of the 1957
cohort (80%) was also located through the university; the remain-
der included those obtained through Find A Friend (15%) and
Web-based searches (5%).
Analytic Method
HLM (Version 6.04) was used to fit a series of growth curve
models for each dimension of psychosocial development. As noted
earlier, mixed effects or growth curve models have a number of
advantages over more traditional analyses of repeated measures
data because they allow for fuller exploitation of the available data.
They permit inclusion of individuals not assessed at all time points,
are tolerant of unequal intervals between data points, and allow
time to be treated flexibly so that the intercept can correspond to
any time point in the study. For the present investigation, these
advantages were important, given that not all participants were
available for each test occasion. Moreover, participants were of
differing ages within cohorts. Using HLM made it possible to take
maximum advantage of the age structure of the data.
As applied in this study, growth curves provide estimates for
each scale of the average level of psychosocial resolution (inter-
cept) as well as the linear and quadratic changes over time that
characterize the growth of each issue. In addition, growth curve
models provide estimates of the factors (e.g., cohort, gender, and
demographic variables) that potentially explain variations among
participants in both the level and rate of change in the scale scores.
We were interested in comparing predicted values of the outcome
when people were in early adulthood (age 31). We therefore
centered on this value, giving us a model parameter that we could
compare across individuals. Differences in scores in early adult-
hood could then be predicted from our key moderator variables:
cohort, gender, education, occupational prestige, relationship sta-
tus, and parental status.
Modeling occurred in several stages. First, baseline models were
fit that characterized the level (intercept) and rate of change (slope)
of each scale over the duration of the study (as well as variation
between participants in the intercept and slope). Next, we added a
quadratic term to the model to specify further the shape of the
trajectories of each scale. Finally, we added cohort and demo-
Table 2
Sample Characteristics by Cohort
Variable
1946 cohort 1957 cohort
1977 testing 1989 testing2000 testing 1989 testing2000 testing
Age
N
M
SD
156
30.59
1.05
99
42.59
0.99
182
55.09
1.47
83
31.8
1.36
136
44.21
1.6
Education (in years)
N
M
SD
Occupational prestige
N
M
SD
Relationship status (%)
Married
Cohabitating
Remarried
Single
Divorced
Separated
Widowed
Total
Parenthood status (%)
Nonparent
Parent
Total
273
18.58
1.97
165
19.41
1.56
193
18.88
1.94
163
18.13
1.67
105
19.09
1.38
208
74.62
14.9
207
75.57
15.74
197
75.34
16.46
16.25
77.92
16.25
26.26
74.29
16.26
159 (71)
11 (5)
6 (3)
35 (16)
9 (4)
4 (2)
0 (0)
224
132 (68)
12 (6)
18 (9)
14 (7)
16 (8)
0 (0)
1 (1)
194
120 (62)
18 (9)
29 (15)
6 (3)
14 (7)
1 (1)
3 (2)
194
103 (66)
6 (4)
4 (3)
36 (23)
5 (3)
0 (0)
1 (1)
155
102 (70)
8 (6)
9 (6)
12 (8)
9 (6)
1 (1)
2 (1)
145
51 (30)
118 (70)
169
48 (23)
160 (77)
208
33 (16)
167 (84)
200
38 (33)
77 (67)
115
37 (24)
115 (76)
152
Table 3
Response Rates for 2000–2 Testing by Cohort
Sample component
1946 1957
N
%
N
%
Original sample
Deceased
Was not contacted
Was not an alum
Contact information available
Percent of those with information who
participated
Percent of those with information who
did not participate
Reasons for nonparticipation among those
for whom information was available
Did not respond
Agreed but did not complete
Refused
349
12
100299 100
4.15
0.35
0.00
82.81
5
3
2
2.02
1.21
0.81
82.61
1
0
289 247
62.9855.06
37.0244.94
91
16
31.49
5.54
2.77
96 38.87
1.21
0.81
3
28
1332
WHITBOURNE, SNEED, AND SAYER

Page 6

graphic variables sequentially to test whether these predictors
account for variability in the level and rate of change over time in
the scales. Improvement in fit between two successive models was
formally tested with a chi-square test of the difference between the
?2 log-likelihoods of the two models with degrees of freedom
equal to the difference between the number of parameters in the
two models. (Technical note: Fixed and random effects were
estimated for the intercept, linear slope, and quadratic slope for all
models. However, because only three time points were available
for the last two psychosocial issues [added in 1977], we were
unable to estimate the random component of the quadratic for
these scales.) All models used full maximum likelihood estimation
and treated missing data as missing at random.1
Results
Attrition Analyses
To determine whether there was a systematic bias in those who
participated in the 2000–2 testing compared with those who did
not, we conducted multivariate analyses of variance for each
cohort comparing the college IPD scores of those who remained in
the study in 2000–2 with those who had dropped out of the study
in each prior phase of testing. Thus, three analyses were conducted
for the 1946 cohort (comparing scores of follow-ups vs. dropouts
at 1966–67, 1977–78, and 1988–89) and two for the 1957 cohort
(1977–78 and 1988–89). None of the attrition analyses yielded
significant main effects of attrition status on specific scales, al-
though for the 1957 cohort, there was a significant multivariate
effect for attrition status comparing those who completed testing in
1989 with those who did not, Wilks’s ? ? .728, F(8, 74) ? 3.455,
p ? .002. Because there were no significant univariate main
effects on the specific scales, this multivariate effect was not
further interpreted.
There was differential attrition by gender for the 1946 cohort
with men more likely to have completed the 2000–2 follow-up,
?2(1) ? 4.49, p ? .034. This sex difference was due primarily to
the fact that men in this cohort were more likely to be listed in the
alumni directory and therefore could more readily be contacted.
There were no sex differences in attrition status for the 1957
cohort.
Alumni who had taken part in the study on previous occasions
were more likely to complete their participation in 2000–2. Of the
99 in the 1946 cohort who participated in 1988–89, 85 (85%)
provided data in 2000–2, and of the 83 in the 1957 cohort with
1988–89 data, 62 (75%) participated in 2000–2. Of those from the
1946 cohort who participated in 1977–78, a similarly large pro-
portion, 113 (72%), remained with the study in 2000–2.
Multivariate analyses of variance comparing the IPD scores of
those who completed the questionnaires in 2000 and 2002 pro-
duced no differences at any of the prior or current times of
measurement. There were also no differences on the IPD scales
between participants who were contacted through the alumni di-
rectory and those whose names were obtained through other meth-
ods, either in their 2000–2 scores or their earlier IPD scores.
Growth Curve Estimates of IPD Scale Scores by Cohort
Table 4 presents the summaries of the best-fitting models for
each IPD scale. Cohort differences were inferred when the Level
2 analyses yielded a significant fixed effect of the intercept (linear
or quadratic) for cohort. As hypothesized, the first three IPD scale
scores (trust, autonomy, and initiative) demonstrated a slow, up-
ward direction of growth across both cohorts, as indicated by the
existence of significant positive linear slopes in the fixed-effects
estimates. The random-effects estimates show that there was sig-
nificant variability in the slopes for these scales, indicating that
individuals differed in their patterns of change over time (e.g., see
Figure 1). On industry, as hypothesized, differing patterns of
change were observed by cohort. At the beginning of the period of
testing, the 1946 cohort’s scores were significantly lower than the
1957 cohort’s, but the rise in scores of the older cohort was steeper
over the period of the study than the increase for their younger
counterparts.
Identity and intimacy, as indicated by the significant fixed
quadratic slope, increased across adulthood, but a visual inspection
of the graph indicated that deceleration is most prominent from
early adulthood to midlife. These two psychosocial issues subse-
quently showed slow and steady growth toward increasingly fa-
vorable resolution, patterns that were similar for both cohorts.
Generativity scores increased steadily throughout midlife, but
we did not observe the anticipated curvilinear change in generat-
ivity with age. No differences between the cohorts in either the
level or rate of change were found. Ego integrity, however, fol-
lowed a curvilinear change pattern, one that differed between
cohorts. Although for both cohorts growth followed a U-shaped
pattern, based on a visual inspection of the fitted trajectories, the
1946 cohort’s increase was significantly greater than that of the
1957 cohort in the latter portion of the growth curve.
The HLM growth models provide estimates of the covariances
among the trajectory parameters (see Table 5). For example, the
covariance between the intercept (age 31 trust score) and the linear
rate of growth is .152, with a standard error of .096. This indicates
that increases in age 31 trust scores are weakly associated with
steeper rates of change.
Gender Differences in Growth Curves
The only psychosocial issue to show gender differences was
intimacy (see Table 6). In line with the findings from the previous
RALS follow-ups, women had higher intimacy scores than men.
However, the patterns of change over time differed by gender.
Men showed a consistent steady rise, as indicated by a significant
linear slope. Although overall women increased in intimacy
scores, their significant negative quadratic slope indicated that
their rate of growth slowed over time. The peak rise in intimacy for
women occurred up through the early 40s, after which it tapered
off.
Demographic Variables and Psychosocial Resolution
In addition to cohort and gender, the relevant demographic
variables of educational attainment, occupational prestige, length
1In addition to the analyses presented here, we ran models on only the
longitudinal sample (i.e., those who participated at all occasions) to check
if the pattern of results would be similar to those reported in the main
analyses. Although this analysis severely limited the numbers, there were
significant linear effects for cohort on industry and ego integrity.
1333
PSYCHOSOCIAL DEVELOPMENT THROUGH MIDLIFE

Page 7

and types of long-term relationships, and parenthood status were
expected to explain significant variability in level and rate of
change in psychosocial development over time. To investigate the
interactions of these variables with time, we dummy-coded scores
on the demographic variables as 0 and 1 based on median splits at
age 31. In order to avoid multicollinearity in calculating the
interaction terms, the means of each demographic variable were
subtracted from the dummy-coded variable for that particular
group and used in place of the dummy codes. In plotting these
interactions, however, the uncorrected coefficients were used (i.e.,
0s and 1s).
Educational and occupational status.
early socioeconomic success would predict more favorable psy-
chosocial resolution was operationalized by using educational and
occupational status at age 31 as predictors of change on initiative,
industry, and identity. As shown in Table 7, growth curve analyses
The hypothesis that
by cohort, sex, and educational group produced Education ?
Cohort interactions that were significantly different from baseline
for initiative scores only, ?2(13) ? 25.23, p ? .021, but not for
industry or identity. The Education ? Cohort interaction was
significant for the linear and quadratic slopes. This model is
illustrated in Figure 2, which shows that the 1957 cohort partici-
pants with high levels of education began college with the highest
initiative scores. Their growth curves then began a decline to the
Figure 1.
observed (raw) trajectories for 15% of the participants selected at random.
Average model-fitted trajectory for industry superimposed on
Table 4
Growth Curve Estimates for Inventory of Psychosocial Development Scales, Best-Fitting Models
Scale
Final estimation of fixed effects
Final estimation of random
effects
Residual
?2vs. linear
baseline
Intercept at age
31 (SE)
Linear slope
(SE)
Quadratic
slope (SE) Intercept
Linear
slope
Quadratic
slopea
Trust
Autonomy
Initiative
Industry (1946 cohort)
Industry (1957 cohort)
Identity
Intimacy
Generativity
Ego integrity (1946 cohort)
Ego integrity (1957 cohort)
10.81??(0.28)
8.00??(0.18)
11.10??(0.25)
13.45??(0.48)
14.37 (0.81)
9.46??(0.30)
13.19??(0.34)
8.23??(0.23)
7.36??(0.58)
3.96??(0.96)
.08??(.01)
.03??(.01)
.03??(.01)
.46??(.03)
.29??(.04)
.14??(.02)
.12??(.02)
.03?(.01)
?.56??(.09)
?.12??(.08)
32.40??
11.03??
24.63??
30.51??
.03??
.01??
.01??
.05??
24.63
15.68
26.39
24.01
NA
NA
NA
?.01??(.002)
?.002 (.004)
?.003??(.001)
?.002?(.001)
.000 97.26??
21.11??
30.77??
17.38??
30.79??
.03??
.04??
.01??
.01?
.000
.000
18.51
19.79
14.20
27.51
14.69??
9.54?
NA
38.99??
.02??(.004)
.003?(.006)
Note.
intimacy ? 3144.90; generativity ? 5921.91; ego integrity ? 6521.39 (all df ? 6). NA ? could not be estimated.
aThe quadratic slope for generativity and ego integrity could not be determined.
?p ? .05.
N ? 382. Absolute fit statistics for all models: trust ? 3242.10; autonomy ? 3006.36; initiative ? 3196.98; industry ? 3241.70; identity ? 3082.34;
??p ? .01.
Table 5
Covariance Terms From Unconditional Models for Each
Dependent Variable
Scale
Intercept ?
Linear Slope
Intercept ?
Quadratic Slopea
Linear ?
Quadratic Slopea
Trust
Covariance
SE
Autonomy
Covariance
SE
Initiative
Covariance
SE
Industry
Covariance
SE
Identity
Covariance
SE
Intimacy
Covariance
SE
Generativity
Covariance
SE
Ego integrity
Covariance
SE
.152
.096
.001
.042
.108
.077
?.106
?.155
.002
.011
?.002
.001
.058
.112
?.001
.008
?.001
.000
.234
.136
?.008
.010
?.001
.001
?.003
.066
.187
.119
aCovariance could not be estimated for ego integrity.
1334
WHITBOURNE, SNEED, AND SAYER

Page 8

40s, which continued through the 50s. Their counterparts from the
1957 cohort with low levels of education showed a steady down-
ward trajectory, following a slight upturn through the 30s. The
effect of education was not pronounced for the 1946 cohort par-
ticipants, although those below the median educational level even-
tually reached the same levels of initiative as their more highly
educated peers by the 50s.
Next, the interactions of occupational prestige with initiative,
industry, and identity were examined with prestige classified into
0 and 1 at the median. As shown in Table 8 and Figures 3 and 4,
the 1957 cohort men who were in low-prestige jobs at the age of
31 showed an inverted U shape of gains in industry. They peaked
in the late 30s but then continued to slide downward thereafter.
Conversely, men in the 1946 cohort with low levels of prestige,
whose industry scores were the lowest in college, continued a
steady growth upward in industry throughout the late 50s. A
strikingly different pattern was obtained for the 1957 cohort
women, as it was those women in high-prestige occupations whose
industry scores peaked in the 30s and then declined steadily after
reaching that high point. However, the low-prestige women in the
1946 cohort showed a similar pattern as the low-prestige men in
their cohort, that is, low scores in college and steady gains up
through the 50s. All groups other than women in the 1957 cohort
with high-prestige occupations showed gains throughout midlife in
this psychosocial quality.
Intimacy and long-term relationships.
outcomes in the world of relationships is not easily defined, for the
purposes of the present study, as in the previous RALS follow-up,
a long-term committed relationship was considered to represent
greater success. However, in the present study, unlike the previous
follow-up, involvement in a committed relationship was defined as
living with or married to a person of the same or other sex. As
noted above, the ratings for long-term relationships were based on
the individual’s status at the age of 31.
This definition led to an uneven distribution, however, of rela-
tionship commitment, with only one third of participants in the
noncommitment category. However, the alternative of grouping
participants on the basis of married versus unmarried failed to
capture the quality sought in these analyses. As shown in Table 9
and Figure 5, there was an interaction of cohort and relationship
status. The 1957 cohort individuals in committed relationships,
though higher in intimacy in college, peaked in the 30s and
declined continuously thereafter. The opposite pattern was shown
for the 1957 cohort individuals who were not in a committed
relationship by early adulthood. For the 1946 cohort, a similar
pattern was obtained, with the committed peaking and then de-
clining and the noncommitted continuing a slow steady pace of
growth.
Parenthood status.
As noted earlier, the majority of partici-
pants in both cohorts were parents. Although there was little
variation in parenthood status, there was variation in the ages at
which participants became parents. Holding parental status con-
trolled by selecting only those participants who were parents, we
then investigated the predictive effect of parental status by the age of
31 on generativity growth through midlife. As shown in Table 10
and Figure 6, there was a significant interaction of cohort and
parental status at age 31. Parents in Cohorts 1 and 2 showed
opposite trajectories of growth over adulthood, with the 1946
cohort parents declining and the 1957 cohort parents showing a
pattern of increases. However, perhaps most striking about these
growth curves was the increase in generativity among the 1946
cohort participants who became parents after the age of 31. Al-
though initially lower in generativity scores, these individuals
continued to show gains until by the 50s their generativity scores
nearly equaled those of parents in the 1957 cohort and slightly
exceeded those of their peers in the 1946 cohort who had become
parents by early adulthood.
Although favorable
Table 7
Fixed- and Random-Effects Estimates for Growth Curve Models
of Initiative
VariableIntercept
Linear
component
Quadratic
component
Fixed-effects estimates
Intercept (age 31)
Time ? Cohort
Time ? Education
Time ? Cohort ?
Education
Random-effects estimates
Intercept
Linear slope
Quadratic slope
Residual
?2vs. linear baseline
(df ? 13)
11.94??(0.37)
1.24 (0.81)
.31 (0.76)
.04?(.02)
?.03 (.04)
?.03 (.03)
?.005 (.001)??
?.008 (.004)
.004 (.003)
?1.16 (1.64)
?.18?(.08)
.02 (.009)?
24.06??
.03??
.00
23.74
25.23?
Note.
education: 0 ? below the median for sex. Degree of freedom for random-
effects estimates ? 185.
?p ? .05.
Values in parentheses are standard errors. Cohort: 0 ? 1946 cohort;
??p ? .01.
Table 6
Growth Curve Estimates for Men and Women, Intimacy Scores
Gender
Final estimation of fixed effects
Final estimation of variance
components
Residual
?2vs. linear
baseline
Intercept at age
31 (SE)
Linear
slope (SE)
Quadratic slope
(SE)Intercept
Linear
slope
Quadratic
slope
Men
Women
11.56??(0.51)
14.82??(0.67)
.08??(.03)
.16?(.03)
.000228.48??
.04??
.00019.7834.88??
?.005?(.002)
Note.
?p ? .05.
N ? 323 men, 325 women.
??p ? .01.
1335
PSYCHOSOCIAL DEVELOPMENT THROUGH MIDLIFE

Page 9

Discussion
The growth curves modeled in this sequential analysis of per-
sonality change from college through midlife provide evidence
that personality continues to evolve throughout the adult years.
This overall conclusion is consistent with an emerging literature
documenting the complexity of midlife personality development
(Lachman, 2004) and the potential for changes in personality over
the life course (Roberts, Walton, & Viechtbauer, 2006). In cases
such as trust, autonomy, and initiative, theorized to be most typ-
ically associated with early childhood, growth continued through
the mid-50s. Although the rate of change was gradual, the fact that
any change was observed to occur at all in these qualities would
seem to challenge the view that personality is fixed by early life.
Conversely, the observed increases in ego integrity that occurred
throughout early to middle adulthood indicate that later psychos-
ocial issues can become salient at earlier points in the life course.
The present findings thus support the interpretation that Erik-
son’s model of psychosocial development describes not a ladder,
often used to depict it, but a matrix in which issues that are
associated with earlier ages on the matrix are continuously revis-
ited throughout life, and issues that Erikson associated primarily
with later adulthood can be confronted at earlier ages. In fact, some
have argued that complete resolution is impossible at the “typical”
age, given the changes that are experienced throughout life, and
that, by necessity, each psychosocial issue must be continually
reworked in the context of later issues (J. M. Erikson, 1991).
In contrast to other longitudinal studies of personality through
midlife, which follow single cohorts, the present analyses also
provided support for the importance of examining changes over
time in multiple cohorts. In the present study, we were able to
compare generational differences in personality growth particu-
larly as these interacted with key events in the lives of the partic-
ipants. Although the preponderance of data supported the gener-
ality of the growth models, cohort differences continued to be
demonstrated in industry and ego integrity, as was true in the
previous investigation carried out in the late 1980s. Moreover, the
interaction of cohort with education, prestige, relationship status,
and parenthood status points to the differential socialization expe-
riences that can alter the course of psychosocial development
throughout adulthood.
Industry, identity, and intimacy tended to change over time in a
linear fashion, but each was subject to the influence of life history
factors. In the two prior follow-ups, possible interpretations were
raised for the dramatically lower scores of participants in the 1946
cohort on industry compared with those of their younger counter-
parts during college. In the present study, with the availability of
long-term follow-up data and the modeling capability of HLM, it
was possible to gain additional insight into the processes affecting
industry development during adulthood. It would appear from the
present analyses that the very lowest industry scores were obtained
in college by participants who, in early adulthood, had jobs with
relatively low prestige. However, they managed to exceed their
peers in industry scores throughout the course of the study. In
interpreting the scores of the women in particular, the findings
support those of other studies of midlife women (Stewart, Ostrove,
& Helson, 2001) in which gains in self-confidence and determi-
nation occurred through involvement in their jobs, allowing them
to feel increasingly competent and effective. It is possible that for
Figure 2.
Model-fitted trajectories for initiative scores by education status at age 31 and cohort.
Table 8
Fixed- and Random-Effects Estimates for Growth Curve Models
of Industry
VariableIntercept
Linear
component
Quadratic
component
Fixed-effects estimates
Intercept (age 31)
Sex
Cohort
Prestige
Sex ? Cohort
Sex ? Prestige
Cohort ? Prestige
Sex ? Cohort ? Prestige
Random-effects estimates
Intercept
Linear slope
Quadratic slope
Residual
?2vs. linear baseline
(df ? 25)
14.00??(0.44)
2.24?(0.88)
1.43 (0.90) ?.18??(.05) ?.001 (.004)
.84 (0.88)
?.06 (.05)
.79 (1.80).05 (.10) ?.004 (.008)
.13 (1.77)
?.04 (.10)
.50 (1.81).10 (.10) ?.004 (.009)
3.20 (3.61).31 (.20) ?.04?(.02)
.35??(.02)
.02 (.05) ?.003 (.004)
?.01 (.002)??
.001 (.004)
?.01 (.008)
27.91??
.04??
.00
23.89
99.96?
Note.
0 ? 1946 cohort; prestige: 0 ? below the median. Degree of freedom for
random-effects estimates ? 169.
?p ? .05.
Values in parentheses are standard errors. Sex: 0 ? male; cohort:
??p ? .01.
1336
WHITBOURNE, SNEED, AND SAYER

Page 10

these leading-edge baby boomer women, feelings of competence
were suppressed in college, when it seemed as though their careers
would not play an important role in their future success. With
transitions in women’s roles, they then became better able to fulfill
their desires for making contributions to the world of work outside
the home.
Other findings including those for initiative, intimacy, and gen-
erativity support the idea that individuals can overcome early
psychosocial deficits to catch up with, if not exceed, their initially
more advantaged peers. In the area of intimacy, participants who
were not in a committed relationship early in adulthood showed
continued gains throughout the period of the study and in fact
moved toward an increasingly favorable resolution of this issue
that exceeded those of their peers who early in adulthood were in
a committed relationship. This effect was particularly noteworthy
for the 1957 cohort, perhaps reflecting the changing norms be-
tween the 1960s and 1970s regarding age of marriage. A similar
trend of enhanced developmental gains was also shown for par-
ticipants who became parents after the age of 31. Perhaps by
waiting until their mid-30s or beyond, when their careers were
already established, they were best able to enjoy their new parent-
hood status to the fullest. In the case of men, there can be a number
of possible reasons that they delay fatherhood, each of which can
have different consequences for generativity development (Snarey,
Son, Kuehne, Hauser, & Vaillant, 1987). However, the findings
indicate that gains in psychosocial resolution can occur and per-
haps even be relatively greater when individuals enter this status at
a later age than when they do so in their 20s.
For both men and women, an important set of sociocultural
factors influence ego integrity, the psychosocial issue involving
questions regarding the meaning of existence and the sense of
connectedness with social values. The previous follow-up
raised the question of whether participants would continue to
show the downward trend in ego integrity that was attributed to
the “me generation” mentality of the 1980s (Whitbourne et al.,
1992). In the present follow-up, participants in both cohorts did
show increases in their ego integrity scores. If the time of
measurement interpretation still holds, this finding suggests that
the decade of the 1990s, exemplified by movements such as
President Bill Clinton’s initiation of AmeriCorps in 1993, may
have helped stimulate participants in the RALS to feel more
connected with the world outside themselves and their own
immediate concerns.
Sociocultural influences may also have contributed to the
pattern of cohort differences, given that, as noted earlier, the
1946 and 1957 cohorts represent, respectively, the leading-edge
and trailing-edge baby boomers. Demographers note that intra-
cohort differences in the baby boom generation are important to
Figure 3.
Model-fitted trajectories for industry scores by prestige at age 31 and cohort, men.
Figure 4.
Model-fitted trajectories for industry scores by prestige at age 31 and cohort, women.
1337
PSYCHOSOCIAL DEVELOPMENT THROUGH MIDLIFE

Page 11

consider in socioeconomic status, given that the leading-edge
baby boomers were forced to compete economically with a
larger number of age peers (Easterlin, 1987). Although the
findings from the present study were consistent with large-scale
demographic analyses showing only modest differences in so-
cioeconomic status between leading-edge and trailing-edge
baby boomers, exposure to the differing value systems preva-
lent during their college and early adult years may have played
a role in differentially affecting their values, attitudes, and
relationship qualities and dynamics (Eggebeen & Sturgeon,
2006). These factors, in turn, may have had an impact on their
patterns of change through middle age in the psychosocial
issues examined here.
There were several constraints limiting the conclusions that can be
drawn from the present investigation. First, it is possible that attrition
has substantially compromised the internal validity of the findings.
According to Roberts and DelVecchio (2000), high attrition rates bias
research findings in longitudinal studies of personality toward stabil-
ity, because the people who remain in longitudinal studies are by
definition stable. Furthermore, attrition rates increase as a function of
the length of the follow-up interval, with smaller intervals finding
greater stability and longer intervals finding greater change. Roberts
and DelVecchio did not find any effect of attrition in their meta-
analysisof152studiesandconcludedthatresearchersshouldquestion
the assumption that attrition is a major distorting influence in longi-
tudinal studies of personality development. In the present study,
significant effects were not observed in the overall multivariate anal-
yses used to test attrition effects for the 1946 cohort or on individual
scales for the 1957 cohort, indicating that the pattern of scores for
those who remained in the study and those who did not was not
significantly different.
This study was also limited because it exclusively relied on a single
self-report measure, thus potentially producing surplus construct ir-
relevancies that cannot be dissociated from the measurement of the
target construct (Cook & Campbell, 1979). The sample of the present
study also allows for limited generalization, as it consisted of primar-
ily White, educated, middle-class and upper-middle-class men and
women attending a private university in the 1960s and 1970s. Al-
though this kind of sample is typical of longitudinal studies, homo-
geneity with respect to ethnicity and social class does not allow for
Table 9
Fixed- and Random-Effects Estimates for Growth Curve Models
of Intimacy
VariableIntercept
Linear
component
Quadratic
component
Fixed-effects estimates
Intercept (age 31)
Relationship status
Cohort
Relationship Status ?
Cohort
Random-effects estimates
Intercept
Linear slope
Quadratic slope
Residual
?2vs. linear baseline
(df ? 13)
13.60??(0.38)
3.81??(0.96)
2.55??(0.78)
.12??(.02)
.06 (.04)
?.02 (.04)
?.005??(.002)
?.013??(.001)
?.007 (.004)
4.88 (1.96)
?.004 (.09)
?.17?(.009)
27.28?
.04??
.00
18.72
43.13?
Note.
unmarried at age 31; cohort: 0 ? 1946 cohort. Degree of freedom for
random-effects estimates ? 185.
?p ? .05.
Values in parentheses are standard errors. Marital status: 0 ?
??p ? .01.
Figure 5.
Model-fitted trajectories for intimacy scores by relationship status at age 31 and cohort.
Table 10
Fixed- and Random-Effects Estimates for Growth Curve Models
of Generativity
VariableIntercept
Linear
component
Fixed-effects estimates
Intercept (age 31)
Parental status
Cohort
Parental Status ? Cohort
Random-effects estimates
Intercept
Linear slope
Residual
?2vs. linear baseline (df ? 6)
8.66??(0.38)
1.72 (1.03)
?.02??(1.14)
?.43 (1.37)
.08 (.05)
?.12?(.05)
?.09 (.05)
.19??(.07)
14.68??
.01?
14.14
12.70?
Note.
a parent by age 31; cohort: 0 ? 1946 cohort. Degree of freedom for
random-effects estimates ? 212.
?p ? .05.
Values in parentheses are standard errors. Parental status: 0 ? not
??p ? .01.
1338
WHITBOURNE, SNEED, AND SAYER

Page 12

generalization beyond the traditional 4-year White college student.
Although having two cohorts studied sequentially increases the
generalizability compared with single-cohort studies, the find-
ings may also be limited to the historical eras in which they
attended college.
Social desirability may have also played a role in the present
findings such that individuals are willing to acknowledge only
those aspects of the self that they believe to be favorable. Given the
positive and negative items that contribute to each IPD scale score,
it seems possible that socially desirable responding may have
been an issue. However, Costa and McCrae (1994) maintained
that social desirability is not as serious a contaminant as re-
searchers once thought. According to these authors, a substan-
tial body of work has demonstrated that individual differences
in the propensity to endorse socially desirable items are weak
compared with individual differences in traits. Indeed, social
desirability shows very little effect on the factor structures of
personality measures when groups of honest and socially de-
sirable responders are compared (Ellingson, Smith, & Sackett,
2001). Although we were not in fact measuring traits, Whit-
bourne and Waterman (1979) assessed the impact of socially
desirable responding by correlating IPD scale scores with the
Marlowe–Crowne Social Desirability Scale in the 1977 follow-
up. Correlations between the Marlowe–Crowne Social Desir-
ability Scale and the first six scales on the IPD ranged from .16
to .28 for the 1957 cohort and from .05 to .20 for the 1946
cohort, suggesting that the IPD is relatively insensitive to social
desirable response bias. Thus, social desirability did not appear
to limit significantly this investigation’s findings.
The present study is unique in that it is the only investigation to
date involving systematic comparisons of Eriksonian psychosocial
issues in adulthood across several cohorts and several times of
measurement. The findings have provided additional evidence to
support the notion that resolutions of psychosocial issues evolve
continuously and cumulatively over the years of adulthood. This
evolution continues at least through the mid-50s in a pattern of
steady progress toward enhanced growth on issues typically linked
to childhood and in a pattern of more rapid growth for the psy-
chosocial issues that typically confront individuals through the
decades of adulthood.
References
Arnett, J. J. (2000). Emerging adulthood: A theory of development from
the late teens through the twenties. American Psychologist, 55, 469–
480.
Bradley, C. L., & Marcia, J. E. (1998). Generativity-stagnation: A five-
category model. Journal of Personality, 66, 39–64.
Constantinople, A. (1969). An Eriksonian measure of personality devel-
opment in college students. Developmental Psychology, 1, 357–372.
Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design
and analysis issues for field settings. Boston: Houghton Mifflin.
Costa, P. T., Jr., & McCrae, R. R. (1994). Set like plaster? Evidence for the
stability of adult personality. In T. F. Heatherton & J. L. Weinberger
(Eds.), Can personality change? (pp. 21–40). Washington, DC: Amer-
ican Psychological Association.
Cross, S. E., & Madson, L. (1997). Models of the self: Self-construals and
gender. Psychological Bulletin, 122, 5–37.
de St. Aubin, E., McAdams, D. P., & Kim, T.-C. (Eds.). (2004). The
generative society: Caring for future generations. Washington, DC:
American Psychological Association.
Easterlin, R. A. (1987). Birth and fortune: The impact of numbers on
personal welfare (2nd ed.). Chicago: University of Chicago Press.
Eggebeen, D. J., & Sturgeon, S. (2006). Demography of the baby boomers.
In S. K. Whitbourne & S. L. Willis (Eds.), The baby boomers grow up:
Contemporary perspectives on midlife (pp. 3–22). Mahwah, NJ: Erl-
baum.
Ellingson, J. E., Smith, D. B., & Sackett, P. R. (2001). Investigating the
influence of social desirability on personality factor structure. Journal of
Applied Psychology, 86, 122–133.
Erikson, E. H. (1963). Childhood and society (2nd ed.). New York: Norton.
Erikson, J. M. (1991). Wisdom and the senses: The way of creativity. New
York: Norton.
Helson, R., Jones, C., & Kwan, V. S. (2002). Personality change over 40
years of adulthood: Hierarchical linear modeling analyses of two longi-
tudinal samples. Journal of Personality and Social Psychology, 83,
752–766.
Helson, R., & Roberts, B. W. (1994). Ego development and personality
change in adulthood. Journal of Personality and Social Psychology, 66,
911–920.
Helson, R., & Srivastava, S. (2001). Three paths of adult development:
Conservers, seekers, and achievers. Journal of Personality and Social
Psychology, 80, 995–1010.
James, J. B., & Zarrett, N. (2005). Ego integrity in the lives of older
women: A follow-up of mothers from the Sears, Maccoby, and Levin
Figure 6.
Model-fitted trajectories for generativity scores by parental status at age 31 and cohort for parents.
1339
PSYCHOSOCIAL DEVELOPMENT THROUGH MIDLIFE

Page 13

(1951) Patterns of Child Rearing Study. Journal of Adult Development,
12, 155–167.
Kasen, S., Chen, H., Sneed, J., Crawford, T., & Cohen, P. (2006). Social
role and birth cohort influences on gender-linked personality traits in
women: A 20-year longitudinal analysis. Journal of Personality and
Social Psychology, 91, 944–958.
Lachman, M. E. (2004). Development in midlife. Annual Review of Psy-
chology, 55, 305–331.
Marcia, J. E. (2002). Identity and psychosocial development in adulthood.
Identity, 2, 7–28.
Mroczek, D. K., & Spiro, A., III. (2005). Change in life satisfaction during
adulthood: Findings from the Veterans Affairs Normative Aging Study.
Journal of Personality and Social Psychology, 88, 189–202.
Nakao, K., & Treas, J. (1994). Updating occupational prestige and socio-
economic scores: How the new measures measure up. Sociological
Methodology, 24, 1–72.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models:
Applications and data analysis methods (2nd ed.). Thousand Oaks, CA:
Sage.
Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency
of personality traits from childhood to old age: A quantitative review of
longitudinal studies. Psychological Bulletin, 126, 3–25.
Roberts, B. W., Walton, K. E., & Viechtbauer, W. (2006). Patterns of
mean-level change in personality traits across the life course: A meta-
analysis of longitudinal studies. Psychological Bulletin, 132, 1–25.
Robitschek, C., & Woodson, S. J. (2006). Vocational psychology: Using
one of counseling psychology’s strengths to foster human strength.
Counseling Psychologist, 34, 260–275.
Schoon, I., Martin, P., & Ross, A. (2007). Career transitions in times of
social change. His and her story. Journal of Vocational Behavior, 70,
78–96.
Schwartz, S. J. (2006). Predicting identity consolidation from self-
construction, eudaimonistic self-discovery, and agentic personality.
Journal of Adolescence, 29, 777–793.
Snarey, J., Son, L., Kuehne, V. S., Hauser, S., & Vaillant, G. (1987). The
role of parenting in men’s psychosocial development: A longitudinal
study of early adulthood infertility and midlife generativity. Develop-
mental Psychology, 23, 593–603.
Sneed, J. R., Johnson, J. G., Cohen, P., Gilligan, C., Chen, H., Crawford,
T. N., & Kasen, S. (2006). Gender differences in the age-changing
relationship between instrumentality and family contact in emerging
adulthood. Developmental Psychology, 42, 787–797.
Sneed, J. R., Schwartz, S. J., & Cross, W. E., Jr. (2006). A multicultural
critique of identity status theory and research: A call for integration.
Identity, 6, 61–84.
Sneed, J. R., Whitbourne, S. K., & Culang, M. E. (2006). Trust, identity,
and ego integrity: Modeling Erikson’s core stages over 34 years. Journal
of Adult Development, 13, 148–157.
Stewart, A. J., Ostrove, J. M., & Helson, R. (2001). Middle aging in
women: Patterns of personality change from the 30s to the 50s. Journal
of Adult Development, 8, 23–37.
Terracciano, A., Costa, P. T., Jr., & McCrae, R. R. (2006). Personality
plasticity after age 30. Personality and Social Psychology Bulletin, 32,
999–1009.
Torges, C. M., Stewart, A. J., & Duncan, L. E. (2008). Achieving ego
integrity: Personality development in late midlife. Journal of Research
in Personality, 42, 1004–1019.
Van Manen, K.-J., & Whitbourne, S. K. (1997). Psychosocial development
and life experiences in adulthood: A 22-year sequential study. Psychol-
ogy and Aging, 12, 239–246.
Whitbourne, S. K., & Van Manen, K.-J. (1996). Age differences and
correlates of identity status from college through middle adulthood.
Journal of Adult Development, 3, 59–70.
Whitbourne, S. K., & Waterman, A. S. (1979). Psychosocial development
in young adulthood: Age and cohort comparisons. Developmental Psy-
chology, 15, 373–378.
Whitbourne, S. K., & Willis, S. L. (Eds.). (2006). The baby boomers grow
up: Contemporary perspectives on midlife. Mahwah, NJ: Erlbaum.
Whitbourne, S. K., Zuschlag, M. K., Elliot, L. B., & Waterman, A. S.
(1992). Psychosocial development in adulthood: A 22-year sequential
study. Journal of Personality and Social Psychology, 63, 260–271.
Zucker, A. N., Ostrove, J. M., & Stewart, A. J. (2002). College-educated
women’s personality development in adulthood: Perceptions and age
differences. Psychology and Aging, 17, 236–244.
Received November 26, 2007
Revision received February 2, 2009
Accepted February 25, 2009 ?
1340
WHITBOURNE, SNEED, AND SAYER