The Minnesota Mutliphasic Personality Inventory (MMPI) is one of the most widely used instruments for assessing personality characteristics of older individuals in clinical settings (
Botwinick, 1970
;
Lawton, Whelihan, & Belsky, 1980
). This measure of personality has been researched extensively with normal aging samples and with clinical populations, and its values and limitations with older samples have been broadly explored (
Taylor, Strassberg, & Turner, 1989
).
The MMPI has recently undergone a major revision in which outmoded language and obsolete items were deleted, new item content was added to assess additional problem areas, and new representative national norms were obtained (
Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989
). The original MMPI validity and clinical scales have been preserved relatively intact in the second version (MMPI—2) in order to maintain continuity with the extensive research on the instrument for the past 50 years. In addition, several new scales were developed, incorporating new item content, to assess additional clinical problems. The MMPI—2 content scales (
Butcher, Graham, Williams, & Ben-Porath, 1990
) expand the areas assessed using a content-oriented assessment strategy.
Because of its ties with the original MMPI and its expanded assessment foci, the MMPI—2 is likely to be used extensively with older populations in the future. Because the MMPI—2 will be used for making important treatment or dispositional decisions with the elderly, it is important to examine whether the test scores and norms in older populations are similar to those found in younger populations. The purpose of this study is to explore factors relevant to using the MMPI—2 with older individuals and to determine if older adults differ from younger ones on MMPI—2 variables.
MMPI Scores of Older Respondents
As a group, older adults respond somewhat differently to original MMPI measures than younger adults on 6 of the 10 clinical scales. Several studies have shown that older adults in aggregate tend to have higher scores on the Hypochondriasis, Hysteria, Depression, and Social Introversion scales and lower scores on the Psychopathic Deviate and Hypomania scales (
Aaronson, 1958
,
1960
,
1964
;
Botwinick, 1970
;
Brozek, 1955
;
Calden & Hokanson, 1959
;
Dye, Bohm, Anderten, & Cho, 1983
;
Ginsburg & Goldstein, 1974
;
Gynther & Shimkunas, 1966
;
Hardyck, 1964
;
Hibbs, Kobos, & Gonzalez, 1979
;
Hyer, Harkey, & Harrison, 1986
;
Kornetsky, 1963
;
Neugarten, 1977
;
Postema & Schell, 1967
;
Schenkenberg, Gottfredson, & Christensen, 1984
;
Schwartz, Osborne, & Krupp, 1972
;
Swenson, 1961
).
Although the consistency across studies is impressive, there are a number of difficulties with this research. The observed differences are usually small and probably do not reflect genuine changes in psychopathology as individuals age (
Swenson, 1985
). The magnitude of the observed differences does not allow for predictive or descriptive inferences about older compared with younger adults.
In addition, many of the studies are cross-sectional and confound cohort differences with age changes. Furthermore, most of the scales on which differences are seen contain significant numbers of items reflecting physical health status, namely, somatic complaints, hysteria, depression, and mania. Thus, any observed age differences may relate primarily to differences in physical health rather than in psychopathology per se (cf.
Aldwin, Spiro, Levenson, & Bossé, 1989
). Thus, only lower scores on Psychopathic Deviate and higher scores on Social Introversion may be indicative of age-related personality change. Even then, most of the evidence is cross-sectional and could be due to cohort differences.
Other researchers have focused on the stability of personality dimensions over the adult years (e.g.,
Schaie & Parham, 1976
;
Slater & Scarr, 1964
).
Costa and McCrae (1977)
found extraversion to be stable over time, and
Leon, Gillum, Gillum, and Gouze (1979)
found that people maintain their relative position on MMPI scales over a 30-year period, with the Social Introversion scale having the highest test—retest stability.
Finn (1986a
,
1986b)
also found substantial evidence of greater stability on many traits in older than in younger age cohort groups.
Costa and McCrae (1977)
concluded that personality was "remarkably stable" over time.
Swenson (1985)
, reviewing his 25 years of work with the MMPI and aging populations, concluded that:
There is certainly no objective evidence familiar to me to show that individuals admit to more psychopathology as they age. With very few exceptions, the gerontic end of the continuum holds up very well in terms of stability, conformity, and lack of psychopathology. ... (p. 257)
Method
Samples and Procedures
Two samples were used in this study, the MMPI—2 Restandardization sample and the NAS panel. The procedures also varied by study.
The MMPI—2 Restandardization Study.
The sample used in the MMPI Restandardization Study comprised 2,600 respondents, of whom 1,138 were men. Their ages ranged from 18—84 (
M
= 41.71,
SD
= 15.32). Respondents were randomly solicited from several regions of the United States: California, Minnesota, North Carolina, Ohio, Pennsylvania, Virginia, and Washington (
Butcher et al., 1989
). Respondents were contacted by mail, informed of the study, and invited to come to a prearranged testing site to complete the test battery. An effort was made to balance the sample for age, gender, social class, education, and ethnic group membership. Special targeted mailings were incorporated into the sampling procedures to obtain representative numbers of minority subjects. The restandardization participants were tested in individual or small-group sessions using standard instructions. The entire sample of 1,398 men was used to generate the male norms.
The experimental booklet (Form AX), a 704-item version of the instrument, was used for all subjects. Eventually, this item pool was winnowed down to 567 items. The original validity and clinical scales were retained nearly intact and retained their original names. The three validity scales are Lie (
L
), Fake Bad (
F
), and Subtle Defensiveness (
K
), and the clinical scales are Hypochondriasis (
Hs
), Depression (
D
), Hysteria (
Hy
), Psychopathic Deviate (
Pd
), Masculinity—Femininity (
Mf
), Paranoia (
Pa
), Psychasthenia (
Pt
), Schizophrenia (
Sc
),
Ma,
and Social Introversion (
Si
).
Fifteen new content scales were added, including Anxiety (
ANX
), Fears (
FRS
), Obsessiveness (
OBS
), Depression (
DEP
), Health Concerns (
HEA
), Bizarre Mentation (
BIZ
), Anger (
ANG
), Cynicism (
CYN
), Antisocial Practices (
ASP
), Type A (
TPA
), Low Self-Esteem (
LSE
), Social Discomfort (
SOD
), Family Problems (
FAM
), Work Interference (
WRK
), and Negative Treatment Indicators (
TRT
). An additional
F
score for the content scales was constructed,
F(B).
(See
Butcher et al., 1990
, for a complete description of the development and uses of these scales.)
Only valid records were used to develop the norms. Respondents were eliminated who had incomplete records or MMPI invalidity (Cannot Say score > 30;
F
raw score > 25;
F(B)
raw score > 25).
The Normative Aging Study (NAS).
The NAS sample initially comprised 2,280 men who were selected for the absence of serious chronic physical or mental disease between 1961 and 1968. The men in the sample were predominantly White, middle-class individuals, consisted of roughly equal numbers of blue- and white-collar workers, and were fairly representative of the population of Boston at the time the study was begun. They were selected in part because of their geographic stability and their willingness to participate in a longitudinal research program (
Bossé, Ekerdt, & Silbert, 1984
). There has been a less than 1% annual attrition from the study, and 79% are still participating (
Aldwin et al., 1989
).
The MMPI Restandardization project forms were modified for data collection with the NAS respondents. Special booklets were printed with larger type size and with a
T
and
F
adjacent to the item. The men responded by circling the
T
or
F
on the booklet itself rather than using separate answer sheets. This was done because pilot testing revealed that the oldest NAS men had both perceptual and mechanical difficulties with the scantron sheets, and we wished to make responding as easy as possible to encourage participation among the oldest age groups. The data were later either transferred to scantron sheets or keyed directly onto a tape.
Data collection followed standard procedures used by the NAS (see
Bossé, Aldwin, Levenson, & Ekerdt, 1987
). Over the past 25 years, NAS men have become accustomed to responding to questionnaires sent to them in the mail, and response rates typically exceed 80%. For this study, questionnaires were mailed to 1,881 men participating in the NAS in June of 1986. Reminder postcards were sent 3 weeks later to the nonrespondents, and 3 weeks after that a second questionnaire was sent. Of the 1,881 potential respondents, 1,550 returned questionnaires for an 82.4% response rate (12 men were deceased from various causes and 319 failed to respond). However, 63 records were incomplete and 28 MMPI—2s were invalid, so the final NAS sample included 1,459 men. Their ages ranged from 40 to 90 (
M
= 61.27,
SD
= 8.37).
Results
The primary goal of this study was to determine whether separate norms are needed when the MMPI—2 is used with older individuals and samples. To do this, we conducted three sets of analyses. The first compared the NAS's clinical and content scale scores to the new Restandardization sample's scores. The second set examined differences in item endorsements using a cutoff score of 20% endorsement difference between the two groups, following standard procedures for item comparison across samples in the MMPI (cf.
Butcher & Pancheri, 1976
). Finally, we examined and compared age-group differences within each of the two samples to identify possible age-related elevations on the MMPI—2.
Comparison of the NAS and Restandardization Samples on Scale Scores
Figures 1
and
2
, in which the scores of the NAS sample are plotted against the new MMPI—2 norms on the clinical and content scales, respectively, present an overall picture of our results. These figures show quite clearly that the men in the NAS sample were remarkably similar in their clinical and content scale profiles to the men in the MMPI—2 Restandardization sample. Of the 10 clinical and 15 content scales, only 1 scale deviated by more than 5 points from the Restandardization mean of 50: NAS men had a mean score of 44 on
Mf.
In most cases, the mean scores for the NAS sample were within two
T
-score points of the MMPI—2 Restandardization sample. The standard deviations of the NAS subjects on the clinical and content scales were quite close to 10, which is the standard deviation for the MMPI—2 Restandardization sample. Thus, members of the NAS sample resembled the men of the MMPI—2 Restandardization sample both in the average and in the variance of their scores on the MMPI—2.
Item Comparison Between the NAS and MMPI—2 Restandardization Samples
Having found no appreciable differences between the two samples on the scale level, we proceeded to compare the NAS and MMPI—2 Restandardization samples on the item level. Item-level differences were analyzed by comparing the endorsement percentages of the two samples on all of the MMPI—2 items. Of the 567 items, only 14 differed by more than 20% in their endorsement percentages between the two samples. These items are presented in
Table 1
.
The small number of discrepancies in item endorsement provided further support for the conclusion that the two samples' responses to the MMPI—2 are remarkably similar. The content of those items that did differentiate the two samples reflects attitudes and behaviors that would be expected to vary with age group (e.g., the use of marijuana). However, as seen in
Figures 1
and
2
, these minor discrepancies did not have a noticeable effect at the scale level.
Age-Group Comparisons of MMPI—2 Validity and Clinical Scale Scores
A third goal of this study was to identify and examine the replicability of age-group differences within the two samples. For the purpose of these analyses, subjects in each sample were grouped according to age decade, that is, 40—49, 50—59, 60—69, and a final group of subjects 70—91 years old (Groups 1—4, respectively). For these analyses, only the 519 men from the Restandardization Study were used who were at least 40 years old.
Comparisons were then conducted between age groups within each sample. First, multivariate analyses of variance (MANOVAs) were conducted to determine whether there were overall age-group differences on the clinical and content scales. If the overall
F
was significant, univariate
F
s were computed to determine age-group differences on individual scales. Because of the large samples involved, we decided to correct for family-wise error in significance testing using a Bonferroni correction; the significance level for alpha was set at .004. Scheffé's post hoc range tests (
p
< .05) were also calculated to determine which groups were significantly different from each other.
Table 2
presents the means and standard deviations of the clinical scales by age group for the NAS sample. There were significant overall age-group differences, Wilks' l = .88,
F
(39, 4273.78) = 4.93,
p
< .0001. Four of the six MMPI scales that are typically sensitive to age differences (on the original MMPI) showed significant age-group differences in the expected directions on the MMPI—2:
D, Pd, Ma,
and
Si.
There was also a trend for
Hs
(
p
< .05), and only
Hy
did not reveal the expected age-group differences. In addition, the older age groups in the NAS sample had significantly higher
L
scores.
As shown in
Table 3
, there were also significant age-group differences in the MMPI—2 Restandardization sample, Wilks' l = .79,
F
(39, 1490.24) = 3.20,
p
< .0001. Univariate
F
s revealed significant differences in
D, Pd, Hy,
and
Ma
scores in the expected directions.
Age-Group Comparisons of MMPI—2 Content Scale Scores
We also examined age-group differences on the content scales, using the same procedure (e.g., MANOVAs followed by Bonferroni-corrected [
p
< .004] univariate
F
s). In the NAS sample, the multivariate
F
revealed significant age-group differences, Wilks' l = .84,
F
(45, 4281.62) = 5.95,
p
< .0001 (see
Table 4
). Univariate
F
s revealed that 5 of the 15 scales showed significant age-group differences. Older groups scored higher on
FRS, OBS, LSE,
and
TRT,
whereas younger groups scored higher on
ANG
and
FAM.
In the Restandardization sample, the multivariate
F
for the content scales was also significant, Wilks' l = .75,
F
(45, 1489.12) = 3.37,
p
< .0001. However, none of the univariate
F
s achieved significance with the Bonferroni correction, although similar trends were seen across age groups (see
Table 5
).
The smaller number of significant age-group effects identified in the MMPI—2 Restandardization sample is probably a reflection of the somewhat reduced power of the analyses in this sample, due to the smaller number of subjects. Thus, a greater difference was required to achieve significance in this sample. Nevertheless, when the raw-score differences in the MMPI—2 restandardization sample were translated into
T
scores, they did not exceed 5—6 points. Thus, although these findings point to some interesting age-group differences, we do not feel that they are clinically significant enough to indicate the necessity of separate norms for the elderly.
A final goal of this study was to establish a baseline for the examination of the effects of aging on scores on the MMPI—2. The cross-sectional comparisons of age groups just reported clearly confound the effects of age and cohort. However, because the NAS is a longitudinal study, the results of future administrations of the MMPI—2 to the NAS men will be compared to age-group differences identified at baseline to tease out the effects of age and cohort.
Figures 3
and
4
present comparisons among 5 age groups on the MMPI—2 clinical and content scales, respectively, for the subjects in the NAS. As depicted in these figures, scores on certain MMPI—2 scales, (e.g.,
D, LSE,
and
TRT
), are higher in older age groups, whereas scores on other scales (e.g.,
Pd
) are lower in older groups. As mentioned, the exact nature of these changes will be discernible only after future administrations of the MMPI—2 to the NAS sample.
Finally, as suggested by a reviewer, we conducted some exploratory analyses, contrasting subjects 80 years and older with the other four age groups. Given that there were only 35 men 80 years and older, these results should be interpreted with caution. Analyses of variance with Scheffé's post hoc range tests (
p
< .05) on the clinical scales revealed that the 80-year-olds had significantly lower scores on
Pd
than the 40- and 50-year-olds (as did the 60- and 70-year-olds). With the content scales, however, the 80-year-olds showed significantly higher scores on 4 of the 15 scales,
OBS, CYN, LSE,
and
TRT.
On both the clinical and content scales, the 80-year-olds differed most from the youngest groups and never differed from the 60- and 70-year-olds, although on two of the scales (
OBS
and
CYN
), only the 80-year-olds differed from the younger groups. Furthermore, when translated into
T
scores, the largest differences between the significantly different groups were again no more than 5—6 points. Nonetheless, these findings are suggestive, and more research on personality within old—old age groups is needed.
Discussion
The results of this study indicate that there are few MMPI—2 differences between older NAS respondents and the MMPI Restandardization sample on the validity, clinical, and content scales. Contrasting the means on the clinical scales between the two samples revealed that only one scale (
Mf
) differed by more than 5 points (one half of a standard deviation) between the two samples. In general, the scale profiles on both samples were remarkably similar, and only 14 of the 567 items (4%) differed by more than 20 percentage points in endorsement rates.
However, significant differences emerged in the within-sample age-group analyses. For the most part, these were in the expected directions (e.g., the differences observed were quite similar to the age-group differences generally found on the original MMPI), but somewhat attenuated. In both samples, the older groups were higher on
D,
and lower on
Pd
and
Ma.
Older age groups in the NAS sample showed higher scores on
Si,
whereas, in the Restandardization sample, older groups showed higher scores on
Hs.
The higher depression scores are expected on the MMPI—2
D
scale, but these findings run counter to studies using other depression measures, which often show lower rates of depression in older groups (cf.
Aldwin et al., 1989
;
Feinson & Thoits, 1986
;
Lawton & Albert, 1990
;
Myers et al., 1984
; but see
Gatz & Hurwicz, 1990
). The higher scores we found may reflect the use of items indicating somatic complaints, which can be expected to increase with age.
Of particular interest are the consistently lower scores on
Ma
and
Pd.
On
Ma
and
Pd,
this may reflect generally lower self-reported emotionality in older samples (cf.
Labouvie-Vief, DeVoe, & Bulka, 1989
;
Lawton & Albert, 1990
;
Schulz, 1985
). However, the few differences obtained were small and, in part, a function of statistical power afforded by the large sample sizes used.
For the MMPI—2, new content scales have been developed (see
Butcher et al., 1990
), and we conducted exploratory age-group analyses on these scales. In the NAS sample, 5 of the 15 scales showed significant age-group differences. Older groups scored higher on
OBS, LSE,
and
TRT,
whereas younger groups scored higher on
ANG
and
FAM.
The higher scores on obsessiveness have also been observed in the NAS using other instruments (
Bossé et al., 1987
) and may reflect perceived changes in memory function with age. That is, many older adults report memory lapses and may check and double-check their own actions as a guard against these lapses.
Aldwin (1990)
has also found that the incidence of family problems decreased with age, probably as a result of children reaching adulthood and living away from home. The finding of low self-esteem scores in the older groups is contrary to what is generally found in the literature (cf.
Lachman, 1986
), and it is not clear why these scores should be lower in the older NAS participants. However, because these findings on age differences in the content scales in the NAS sample were not replicated in the restandardization sample, they should be considered tentative.
There are two major limitations to this study. First, the NAS sample includes mostly White, middle-class men; thus, the results of the study may not generalize to women or to other ethnic groups. Second, the data presented here are cross-sectional and not longitudinal. It was possible neither to separate age and cohort effects (
Adam, 1978
;
Schaie, 1977
) nor to examine individual differences in change over time. As
Aldwin et al. (1989)
noted, comparisons of means between age groups may be insensitive to change over the life span. It assumes first, that change is linear, and second, that all individuals change in similar ways. Often, personality change over time emerges only when individual differences in baseline levels are taken into account and when enough data points have been collected to reveal nonlinear change. Thus, ipsative analyses of data collected over long periods are required to determine who changes over time and which characteristics are most likely to change. Consequently, the data presented here will serve as a baseline for studying possible changes over time in future assessments of the NAS men.
