Psychiatric Annals

Revitalizing the MMPI: The Development of Contemporary Norms

Robert C Colligan, PhD; Kenneth P Offord, MS

Abstract

(ProQuest: ... denotes formulae omitted.)

The MMPI has aged. Since its inception in 1957, dramatic changes have taken place in our society. It is apparent that our improved standard of living and increased level of education, the impact of the feminist movement, the questioning, discussing, and liberalizing of moral, religious, and ethical views, the increase in family mobility and changes in composition of the typical nuclear and extended families, the shift from a rural agrarian economic base to our present technologic society, and other transitions have had an impact on everyone in this country. However, in spite of the significant changes that have occurred in our society since 1937, there has not been a systematic re-examination of the original normative data of the MMPI to determine whether they are still valid for contemporary use. Thus, whether we are evaluating a middle-aged female business executive, a teen-age male high school dropout, or the resident of a retirement home, the MMPI profiles from the responses of these people would still be based on the original sample of normal Midwestern subjects obtained during the late 1930s and early 1940s. These original normal subjects were of limited socioeconomic status and had about 8 to 10 years of education; typically they were married, about 35 years old (with ages ranging from lb to b5), and not under a physician's care at the time of their participation in the study.

The MMPI has clearly demonstrated its clinical and research utility during the past 45 years, but the need for ^standardization is equally obvious. Age changes were apparent in the original sample but the numbers of subjects at each age level were insufficient for the development of age-specific norms. In addition, although it originally was developed for use with a psychiatric population, the MMPI has been found to be of considerable value in assessing personality characteristics or functioning over a much broader spectrum of educational, occupational, consultation, and counseling applications. This value underscores the need for a normative base developed from a contemporary sample of normal people.

For all of these reasons, therefore, a study to develop contemporary norms was undertaken.

DEVELOPING NEW NORMS AND NEW TRANSFORMATION PROCEDURES

Method

There were a number of clinical and research issues that needed to be considered (lor example, the impact of age and sex on the MMPI profile) but the broadest of the basic issues was simply whether or not the original MMPI norms were outdated for contemporary use. In order to approach this problem it was decided to administer the MMPI to a large contemporary sample of men and women who met criteria comparable to those required for participation in the original normative work. MMPI responses from this sample would then be systematically compared with the response patterns from the original normative group as represented by the refined sample reported by Hathaway and Briggs1 (the total data base of original subjects is no longer available).

Preliminary Investigations

Some large-scale normative studies have already been carried out with medical patients,2·3 mean profiles for 50 normal parents have been reported,4 and some as yet unpublished work within a single community has also been undertaken.5 Because we were interested in studying normal healthy people, at first we thought the study could be carried out in a medical center so that empirical documentation of normal physical and mental status could be provided. However, study of a large sample of medical patients (854 males. 742 females) who received the diagnosis of "normal" or "negative'" on physical examination revealed significant elevations on several MMPI scales- typically 1 (Hs). 2 (D). and 3 (Hy). forming the…

(ProQuest: ... denotes formulae omitted.)

The MMPI has aged. Since its inception in 1957, dramatic changes have taken place in our society. It is apparent that our improved standard of living and increased level of education, the impact of the feminist movement, the questioning, discussing, and liberalizing of moral, religious, and ethical views, the increase in family mobility and changes in composition of the typical nuclear and extended families, the shift from a rural agrarian economic base to our present technologic society, and other transitions have had an impact on everyone in this country. However, in spite of the significant changes that have occurred in our society since 1937, there has not been a systematic re-examination of the original normative data of the MMPI to determine whether they are still valid for contemporary use. Thus, whether we are evaluating a middle-aged female business executive, a teen-age male high school dropout, or the resident of a retirement home, the MMPI profiles from the responses of these people would still be based on the original sample of normal Midwestern subjects obtained during the late 1930s and early 1940s. These original normal subjects were of limited socioeconomic status and had about 8 to 10 years of education; typically they were married, about 35 years old (with ages ranging from lb to b5), and not under a physician's care at the time of their participation in the study.

The MMPI has clearly demonstrated its clinical and research utility during the past 45 years, but the need for ^standardization is equally obvious. Age changes were apparent in the original sample but the numbers of subjects at each age level were insufficient for the development of age-specific norms. In addition, although it originally was developed for use with a psychiatric population, the MMPI has been found to be of considerable value in assessing personality characteristics or functioning over a much broader spectrum of educational, occupational, consultation, and counseling applications. This value underscores the need for a normative base developed from a contemporary sample of normal people.

For all of these reasons, therefore, a study to develop contemporary norms was undertaken.

DEVELOPING NEW NORMS AND NEW TRANSFORMATION PROCEDURES

Method

There were a number of clinical and research issues that needed to be considered (lor example, the impact of age and sex on the MMPI profile) but the broadest of the basic issues was simply whether or not the original MMPI norms were outdated for contemporary use. In order to approach this problem it was decided to administer the MMPI to a large contemporary sample of men and women who met criteria comparable to those required for participation in the original normative work. MMPI responses from this sample would then be systematically compared with the response patterns from the original normative group as represented by the refined sample reported by Hathaway and Briggs1 (the total data base of original subjects is no longer available).

Preliminary Investigations

Some large-scale normative studies have already been carried out with medical patients,2·3 mean profiles for 50 normal parents have been reported,4 and some as yet unpublished work within a single community has also been undertaken.5 Because we were interested in studying normal healthy people, at first we thought the study could be carried out in a medical center so that empirical documentation of normal physical and mental status could be provided. However, study of a large sample of medical patients (854 males. 742 females) who received the diagnosis of "normal" or "negative'" on physical examination revealed significant elevations on several MMPI scales- typically 1 (Hs). 2 (D). and 3 (Hy). forming the psychosomatic conversion or "conversion V" profile.0 Thus, although these were normal people from the medical and psychiatric standpoints, their MMPI profiles, whether considered as a group or by age and sex, varied little from the mean profile of the typical medical patient referred to Mayo Clinic. Therefore, plans were made for obtaining an entirely new, population-based, random sample.

Hathaway's original subjects were primarily from rural areas or small towns. Our sample included parts of Minnesota, Iowa, and Wisconsin. More specifically, a random sample of 1.919 household listings was obtained from current telephone directories in an area encompassing about 8,000 square miles within a 50-mile radius centered on Rochester, Minnesota. An informational invitational letter was then sent to each household. Subsequently, each was contacted and a structured telephone interview was carried out by trained personnel. This interview was to obtain information from all potential subjects about age. sex. education, ethnic group membership, and mentally or physically handicapping conditions that might be a significant source of response bias. Both husband and wife were permitted to participate but, because ol the possibility of genetic influence, no combination of blood relatives was accepted.7-9

Figure 2. Card deck for Mayo Clinic automated MMPI system. (From Colligan RC, Osborne D, Swenson WM, Offord KP The MMPIA Contemporary Normative Study. New York. Praeger Publishers, 1983.)

Figure 2. Card deck for Mayo Clinic automated MMPI system. (From Colligan RC, Osborne D, Swenson WM, Offord KP The MMPIA Contemporary Normative Study. New York. Praeger Publishers, 1983.)

Persons with a chronic disease, chemical dependency, learning disability, mental retardation or undergoing psychotherapy were excluded from the study. However, a concept of relative wellness was also utilized. For example, a 75-year-old woman who may have described herself as having a "touch of rheumatism'* or a 68-year-old man who stated that "I'm not as spry as I used lobe" may have been included in the study. Determination was made on the basis of the subjects' self-report about the interference of the potentially handicapping condition with work, recreation, or the carrying out of basic self-care tasks or household chores. It should be noted that a random sample of subjects meeting our selection criteria was obtained, not a random sample of the general population. Subjects were systematically excluded for the reasons described above.

In this manner we obtained 1,71 1 subjects, 18 through 99 years old, who agreed to participate in the study (Figure I). The procedures just described thus yielded a group from the same geographic area as the original Minnesota normal sample and, like Hathaway and McKinley 's10 group, our subjects were not under the care of a physician or other health-care professional lor any illness or handicapping condition. Each of the 1,711 persons agreeing to participate in the study was sent a packet of materials including an MMPI. a consent form, and a return envelope. Of this number 1,467 MMPIs were received, an 86% return rate.

The MMPI used in this study was from the Mayo Clinic automated system and is composed of 2b standard data cards on which the MMPI items are printed (figure 2). Items follow the sequence of the booklet form, and the subjects respond to the items in the usual manner but mark their answers directly on the data cards. The data cards are bound together at the top with a plastic binder that is snapped off before the cards are processed. The cards are read by an optical scanner and. alter scoring, the profile is printed on special computer paper having the basic scales listed in the usual MMPI format (Figure 3). On the basis of the pattern of scale elevations, selected statements from an interpretive library are also printed at the bottom of the form.

Rescreening of the telephone interview dala subsequently led to the exclusion of 53 subjects because it was believed they had mentally or physically handicapping conditions that might have biased their responses. At that point, the MMPI profiles were also reviewed; 6 subjects had left 250 or more MMPI items blank and thus were excluded. It is important to note that no subject was excluded on the basis of scale F. the F-minus-K index, or any of the other measures of response bias commonly used by MMPI clinicians. The final sample of 1,408 subjects, all white adults, was divided into seven age groups- 18-19. 20-29. 50-59. 40-49. 50-59. 60-69. and ≥ 70 (the 2=70 group was composed of the following: 70-79 yr, 66 F. 6 1 M ; 80-89 yr. 1 7 F. 1 2 M ; 90-99 yr. 5 F. 1 M).

Figure 3. Typical printout of automated MMPI scoring (From Colligan RC. Osborne D. Swenson WM. Offord KP The MMPIA Contemporary Normative Study. New York. Praeger Publishers, 1983)

Figure 3. Typical printout of automated MMPI scoring (From Colligan RC. Osborne D. Swenson WM. Offord KP The MMPIA Contemporary Normative Study. New York. Praeger Publishers, 1983)

Figure 4. MMPI profiles (mean and mean + 2 SD) for 1 980 US census-matched sample. A, For 335 women. B,For 305 men. (From Colligan RC. Osborne D. Swenson WM. Offord KP: The MMPIA Contemporary Normative Study New York, Praeger Publishers, 1983.)

Figure 4. MMPI profiles (mean and mean + 2 SD) for 1 980 US census-matched sample. A, For 335 women. B,For 305 men. (From Colligan RC. Osborne D. Swenson WM. Offord KP: The MMPIA Contemporary Normative Study New York, Praeger Publishers, 1983.)

Next, a subset of persons matching the age and sex distributions of the white adult population of the United States as determined by the 1980 census was randomly selected from the total sample. This yielded a sample of 335 women ranging in age from 18 through 99 and 505 men ranging in age from 18 through 85. This censusmatched sample was used to compare the response patterns obtained from our group with those from the Hathaway and Briggs1 sample, the only remaining pool of data still available from the original normal sample.

Findings

Although one of our goals had been the development of a normal sample comparable to the original Minnesota normal group, the educational level of our society has increased significantly. This was reflected by the mean educational level of 1 3 years, for both women and men. in our census-matched sample compared with the 8- IO years of education described for the original group. More specifically, in our sample 38% had completed their education with 4 years of high school (compared with 37% in the I980 census for the U.S. white population), and 21% reported 4 or more years of college education (compared with 17% in the 1980 census for the U.S. white population).

Our society as a whole is gradually getting older. The mean age of the contemporary group (45 years for women, 45 years formen) is significantly higher than that of the original sample: however, it should be recalled that Hathaway and McKinley did not include people over age 65 in their sample.

Are the original MMPl norms outdated for contemporary use? Have the changes in society had an impact on the response patterns of normal people? Examination of the mean profile for each sex (Figure 4) provides a visual response to these questions. The illustrations have one line representing the mean score and a second line representing 2 standard deviations above the mean, based on the responses from our contemporary census-matched sample but making use of the old norms for the MMPl.

Except for scale L (P =£0.04), all of the mean T scores for the men (Figure 4B) are markedly higher (P =£0.0001) than 50. the T score mean from the original standardization sample. For the women (Figure 4A). the mean values for scales F, K, 2 (D). 5 (Hy). 4 (Pd), 5 (Mf). 6 (Pa). 7 (Pt), 8 (Sc), and 0 (Si) are markedly different (P ≤0.0001): there was no significant difference on scale L.

Linear correlation analysis detected numerous significant relationships between age and MMPI response patterns for both sexes (Table I). When plotted, these age changes were most evident on scales L, 1 (Hs), 2 (D). 4 (Pd), 5 (MO. and 9 (Ma). For both sexes, scores on scale L increased with age. suggesting greater feelings of conservativism among older people. Expressions of somatic concern und dysphoric emotional tone also increased with age for both men and women. Scores on scale 4 decreased with age, for both sexes, a reflection of fewer feelings of impulsivity and rebelliousness. For women, scores on scale 5 increased with age. suggesting a wider range ol interests among younger women than was apparent in the original normal sample. For men. scores on scale 5 decreased with age. probably reflecting the somewhat lower educational level in the older age range. Scores on scale 9 decreased with age for both sexes, quite likely an indication of gradually decreasing energy levels after adolescence and young adulthood. Tables 2 and 3 provide mean T-score values, using the old norms, for the 13 basic MMPI scales for each sex by age group and also for the census-matched sample.

Evaluation of the age correlations and the mean scale values strongly suggested that two kinds of new normative tables should be used to evaluate MMPI profiles: first, contemporary norms reflecting, for each sex, the response patterns of the adult population in general; and second, a different set of tables, for each sex, specific for age. These tables have recently been prepared."

Development of New MMPI Normative Tables

The clinical scales of the MMPl vary in the number of items each contains. In order to interpret and to compare scores on the different MMPI scales it is necessary to transform the raw scores into standard units. This transformation allows comparisons to be made across scales because the use of the same standard unit means that clinicians can be assured that scores of equal numerical value will reflect the same degree of deviation from the mean, regardless of the scale being considered. The standard scores (T scores) from the original MMPI tables are frequently discussed as if they could be viewed in this manner: however, they cannot. The standard scores used on the original MMPI scales are not T scores by the usual statistical definition. Rather, according to the early manual for the MMPl,12 standard score equivalents for the raw scores obtained from the basic clinical scales were calculated by the linear transformation

...

in which ... is the raw score to be transformed, ... is the mean, and ... is the standard deviation of the normative group. However, a careful evaluation of the HathawayBriggs data13 revealed significant degrees of skewness on nearly all of the MMPI scales. A linear transformation does not correct for this underlying skewness. Thus, transformed T scores for MMPI scales based on a linear transformation may represent markedly different deviations from the mean (eg. in terms of percentile rank), even though the scale scores are numerically equivalent in value. A T score of 70 has traditionally been viewed as a point of clinical significance on the MMPl because it represents a departure of 2 standard deviations from the mean in a normal distribution. However, a T score of 70 obtained from the linear transformation used in the old normative tables may not fall at the 97.7th percentile, in contrast to what would be expected if the scores were normally distributed. Thus, for example, a male subject with a T score of 70 on scale 2 ( D, depression) would not be in a relatively exclusive group of approximately 2% of the population obtaining scores this high but rather would be among the relatively large group of 8% of the normal population who fall above this score (using the old norms) because the underlying distribution of raw scores on scale 2 is significantly skewed to the right.

Table

TABLE 1LINEAR CORRELATION COEFFICIENTS BETWEEN CHRONOLOGIC AGE AND MMPI SCALES (NON K CORRECTED RAW SCORES)

TABLE 1

LINEAR CORRELATION COEFFICIENTS BETWEEN CHRONOLOGIC AGE AND MMPI SCALES (NON K CORRECTED RAW SCORES)

NEW PROCEDURES FOR TRANSFORMING RAW SCORES TO T SCORES

In order to correct the problems associated with linear methods, it was decided to select raw score transformations that would yield new values having a normal distribution. This would allow us to express, as deviations from the mean of the normal reference sample, raw scores that could be much higher than ones typically observed in a normal group. Furthermore, because of the significant age changes evident on most scales, it was decided to develop two types of tables: one that would allow comparisons with people of the same age and sex. and one that would allow comparisons to be made with the general population of adults.

Table

TABLE 2MEAN T SCORES (OLD NORMS) FOR SELECTED MMPI SCALES IN WOMEN BY AGE GROUP AND FOR CENSUS-MATCHED SAMPLE

TABLE 2

MEAN T SCORES (OLD NORMS) FOR SELECTED MMPI SCALES IN WOMEN BY AGE GROUP AND FOR CENSUS-MATCHED SAMPLE

Table

TABLE 3MEAN T SCORES (OLD NORMS) FOR SELECTED MMPI SCALES IN MEN BY AGE GROUP AND FOR CENSUS-MATCHED SAMPLE

TABLE 3

MEAN T SCORES (OLD NORMS) FOR SELECTED MMPI SCALES IN MEN BY AGE GROUP AND FOR CENSUS-MATCHED SAMPLE

To find a raw .score transformation thai would yield a normal distribution, the Box-Cox power transformation technique was used.14 Results were graphically presented for each of the raw scores considered, and the optimal transformation was selected. When logarithmic transformations were indicated. 1 was added to each score to avoid the problem of the undefined log,n (0). Occasionally, two different power transformations yielded equally good results. In this ease, the simpler transformation, or the one that also could be applied to the opposite sex, was selected.

These procedures were carried out on the censusmatched sample and the tables that were developed, for all possible K -corrected raw scores, are presented separately for males and females in Tables 4 and 5. These tables allow comparisons to be made with the contemporary sample of normal adults.

Norms for specific ages were developed separately for each sex by using the data from the total sample of 1. 408 subjects. The relationship between MMPI score and the age was evaluated for each scale by calculating the four polynomial regression equations (linear, parabolic, cubic, and quartic). Subsequently, one of the four models and the power transformation that resulted in the normal distribution of residuals were calculated simultaneously. To carry out the task, the Box-Cox procedure was again utilized in evaluating the more than 1.000 possibilities that result from considering four models and nine or more possible transformations. This process was carried out separately by sex for each scale.

Table

TABLE 4NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: FEMALE ADULTS

TABLE 4

NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: FEMALE ADULTS

Table

TABLE 4 (continued)NORMALIZED T SCORES FOR K CORRECTED RAW SCORES: FEMALE ADULTS

TABLE 4 (continued)

NORMALIZED T SCORES FOR K CORRECTED RAW SCORES: FEMALE ADULTS

After selection, the appropriate model and power transformation were used to calculate extensive tables separately for each sex, to provide raw score conversion tables forages 18-19, and then by 5-year intervals through age 69, with a single category for persons age 7Ü and older.

Thus, using these new norms, clinicians can be assured that, for example, a person who obtains a T score of 60 on Si (social introversion-extroversion) has endorsed enough items to fall at the 85lh percentile (in the direction of introversion) in comparison with a contemporary group of normal adults. Clinicians can also be more sure that elevations are due to actual differences in patterns of item endorsement, compared to use of the old norms in which differences may have been inappropriately magnified because of skewness.

SUMMARY

MMPI response patterns among normal people have changed. Analysis of data from 1,048 normal healthy people ranging in age from 18 through 99 years yielded scores and profiles higher than those obtained from the original standardization group. The differences are more apparent for men than for women and are of clinical as well as statistical significance. Although these relatively elevated profiles might be viewed as suggesting somewhat increased psychologic or physical stress, the response pattern changes are probably due more to changes in social attitudes and in perceptions than to a change in mental health status. Our sample left fewer items unanswered than the original normative group did, which may also have contributed to higher scores. Age trends are apparent on many scales for both sexes.

The data suggest that two kinds of new normative tables should be used to evaluate MMPI profiles: first, norms reflecting the contemporary response pattern of the adult population in general; and second, a set of tables based on age. Such tables for each sex have been prepared and those offering comparisons with adults in general are presented in this article.

In addition to the new raw score conversion tables, the manner in which the raw scores are transformed to T scores has also been changed. Traditional scoring procedures utilized a linear transformation which maintains any underlying skewness in the distribution of raw scores. The linear transformation used for the old normative tables has been replaced by power transformations to yield normalized T scores. This means that the T scores for all MMPI scales have equivalent meanings in terms of departure from the mean, thus making comparisons of profile elevations across the various scales more valid and reliable. Furthermore, the normalized T scores meet the assumptions underlying parametric statistical tests more closely than do those obtained from a linear transformation and therefore may be preferable for research purposes.

The MMPI has been a valuable clinical tool for practitioners in the health care professions for more than 45 years. Hathaway and McKinley had envisioned a viable diagnostic instrument that would undergo developmental changes as our psychologic knowledge and sophistication increased. This research effort is intended as a first step toward the goals of stimulating new research and revitalizing the MMPl to maintain its clinical relevance.

Table

TABLE 5NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: MALE ADULTS

TABLE 5

NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: MALE ADULTS

Table

TABLE 5 (continued)NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: MALE ADULTS

TABLE 5 (continued)

NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: MALE ADULTS

REFERENCES

1. Hathaway SR, Briggs PF: Some normative dala ?? new MMPI scales. J CIm Psychol 1957; 13:364-368.

2. Swenson WM. Pearson JS. Osborne D: An MMPI Source Book: Basic Item, Scale, and Rittern Data on 50,000 Medical Patients. Minneapolis, University of Minnesota Press, 1973

3. Colligan RC. Osborne D: MMPI profiles from adolescent medical patients. J Clin Psychol 1977; 33:186-189.

4. Anderson LM: Personality characteristics of parents of neurotic, aggressive, and normal pre;adolescent boys. J Consult Clin Psychol 1969; 33:575-581.

5. Diehl LA: The Relationship Between Demographic Factors. MMPl Scores and the Social Readjustment Rating Scale. PhD dissertation. Ohio University. 1977 (Library of Congress Card No. Mie. 77-23.461). Xerox University Microfilms, Ann Arbor, Michigan.

6. Polley HF. Swenson WM, Steinhilber RM: Personality characteristics of patients with rheumatoid arthritis. Psychosomatws 1970; 11:45-49.

7. Gottesman II: Heritabihty of personality: A demonstration. Psychological Monographs 1963; 7715721:1-21.

8. Gottesman II: Genetic variance in adaptive personality traits. J Child Psychol Psychiatry 1966; 7:199-208.

9. Hill MS, Hill RN: Hereditary influence on the normal personality using the MMPt. I. Age-corrected parent-offspring resemblances. Behav Genet 1973; 3:133-144.

10. Hathaway SR. McKinley JC: A multiphasic personality schedule (Minnesota): I. Construction of the schedule. J Psychol 1940; 10:249-254.

11. Colligan RC. Osborne D, Swenson WM. et al: The MMPI: A Contemporary Normative Study. New York. Praeger Publishers. 1983.

12. Hathaway SR, McKinley JC: Manual for the Minnesota Multiphasic Personality Inventory. New York, The Psychological Corporation, 1943.

13. Colligan RC. Osborne D. Offord KP: Linear transformation and the interpretation of MMPI T scores. J Clin Psychol 1980; 36:162-165.

14. Box GEP, Cox DR: An analysis of transformations. J R Stat Sw iSeries B) 1964; 26:211-243.

TABLE 1

LINEAR CORRELATION COEFFICIENTS BETWEEN CHRONOLOGIC AGE AND MMPI SCALES (NON K CORRECTED RAW SCORES)

TABLE 2

MEAN T SCORES (OLD NORMS) FOR SELECTED MMPI SCALES IN WOMEN BY AGE GROUP AND FOR CENSUS-MATCHED SAMPLE

TABLE 3

MEAN T SCORES (OLD NORMS) FOR SELECTED MMPI SCALES IN MEN BY AGE GROUP AND FOR CENSUS-MATCHED SAMPLE

TABLE 4

NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: FEMALE ADULTS

TABLE 4 (continued)

NORMALIZED T SCORES FOR K CORRECTED RAW SCORES: FEMALE ADULTS

TABLE 5

NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: MALE ADULTS

TABLE 5 (continued)

NORMALIZED T SCORES FOR K-CORRECTED RAW SCORES: MALE ADULTS

10.3928/0048-5713-19850901-10

Sign up to receive

Journal E-contents