Abbreviating MMPI Item Administration
What Can Be Learned From the MMPI for the MMPI—2 ?

Some researchers, concerned over the length of the MMPI , have attempted to reduce item administration yet provide equivalent, interpretable scales. This article reviews the research on reduced item-administration procedures for the MMPI , addressing issues related to the use of shortened forms. The MMPI has recently undergone a major revision, and the MMPI—2 is available; however, the issues discussed here have relevance to the revised instrument, as it is about the same length as the original. Three basic strategies for reducing item administration have been developed: abbreviating the administration by having the S respond only to items on the basic validity and clinical scales (about 399 in the original MMPI and 370 in the MMPI—2 ); actually reducing the number of items on the standard scales (short form); and using adaptive item-administration strategies to reduce the number of items presented. Future item-abbreviation issues and strategies are discussed.

Although the MMPI has become the most widely used personality assessment technique ( Lubin, Larsen, & Matarazzo, 1984 ), its usefulness is thought by some to be reduced because it contains a large number of items, and the administration requires about an hour and a half. A number of procedures have been developed to reduce the length of item administration; however, the use of these strategies is not without cost–either in terms of losing information or possibly using an invalid or unreliable scale that is not equivalent to the original version. This article explores strategies for reducing time for MMPI item administration, evaluating the research, and providing recommendations for the most effective approaches to reducing time for item administration if it is needed.

The MMPI has recently undergone an extensive redevelopment ( Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989 ) in which antiquated and objectionable items were dropped from and new item contents were included in the booklet. Because the revised version of the instrument (the MMPI—2 ) is approximately the same length (567 items) as the original inventory, some practitioners and researchers may be interested in abbreviated item administration for the MMPI—2 . Consequently, the problems with previous MMPI short forms and issues surrounding item abbreviation are likely to be relevant as psychologists make the transition to MMPI—2 . We now turn to a discussion of issues concerning reducing the number of MMPI items administered.

Varying procedures have been used by different investigators to develop a total of 14 MMPI short forms ( Dean, 1972 ; Faschingbauer, 1974 ; Ferguson, 1946 ; Grant, 1946 ; Holzberg & Alessi, 1949 ; Hugo, 1970/1971 ; Jorgenson, 1958 ; Kincannon, 1968 ; MacDonald, 1952 ; McLachlan, 1974 ; Olson, 1954 ; Overall & Gomez-Mont, 1974 ; Spera & Robertson, 1974 ; Srole, Langner, Michael, Opler, & Rennie, 1962 ). The three most widely researched MMPI short forms, the Mini-Mult, Faschingbauer's Abbreviated MMPI ( FAM ), and the MMPI —168 will be considered in some detail.

The Mini-Mult, containing 71 items, was devised originally by Kincannon, using item cluster groups based on published factor analyses of the scales ( Kincannon, 1968 ). A number of additional items were rationally selected to represent each cluster of items on each scale. For efficiency, Kincannon included items that were scored on the greatest number of MMPI scales. Test—retest correlations between the full form and the Mini-Mult, separately administered, were used to establish the equivalency. Although some loss in reliability was noted (i.e., 9% on average), this loss was considered less than expected.

The FAM , containing 166 items, was initially developed by using item factor clusters to develop short scales for F, Psychopathic Deviance, Masculinity and Femininity, Paranoia, Schizophrenia, Hypomania, and Social Introversion in addition to Kincannon's items for L, K, Hypochondriasis, Depression, Hysteria, and Psychasthenia. Additional items were selected by obtaining items for each scale that correlated with the initial item set ( Faschingbauer, 1974 ).

The MMPI —168 was developed by simply using the first 168 items in the booklet ( Overall & Gomez-Mont, 1974 ). This method of short-form development was chosen when the authors were unable to obtain publication permission from the MMPI distributor to use items out of the booklet context. After reviewing the first 168 items, Overall and Gomez-Mont concluded that there was sufficient representation of the MMPI clinical scales. Research conducted on the Mini-Mult, FAM , and MMPI —168 short forms is summarized in Tables 1 and 2 .

Research on the equivalence of MMPI short forms for predicting the full scale score is equivocal at best, and many researchers have concluded that the limitations of MMPI short forms are too great to use them for clinical prediction (e.g., Graham, 1987 ; Greene, 1982 ; Hart, McNeill, Lutz, & Adkins, 1986 ; Helmes & McLaughlin, 1983 ; Hoffmann & Butcher, 1975 ; Streiner & Miller, 1986 ; Willcockson, Bolton, & Dana, 1983 ).

The basic premise about the need for a short form of the MMPI is not well established. The idea that patients are too busy to respond to the full MMPI , or are otherwise unable to complete the inventory, is largely unfounded ( Graham, 1987 ). In many cases, such as inpatient or correctional programs, individuals often have more available time than they have activities to occupy them. Consequently, the administration of a full MMPI does not usually detract from ongoing patient activities. Even in outpatient mental health and medical settings, the situation can usually be arranged to obtain complete assessments if the staff and patients are informed of the importance of the information. Patients for the most part are willing to share information about themselves if they view the questions as relevant to understanding their problems. Rarely will clients object to testing if the clinician or staff person administering the test takes care to explain the purpose of the test. This is especially true if they are informed that the results will be valuable in the treatment planning.

Possibly some past objection to the MMPI 's length came in part from the frustration that some patients may have felt about the item redundancy (16 items were repeated) or the perceived irrelevant item content. However, as most of the redundant, objectionable, or obsolete items have been eliminated from the revised booklet, it is possible that subjects may not be as reluctant to complete the full MMPI—2 as the original version, where item duplications and irrelevant content may have added to their frustration.

Shortened scales are less reliable than the full version of the MMPI scales. The lowering of reliability is likely to lessen the validity of the scales. Inspection of the studies reporting alternate form reliability shows that the magnitude of correlations for the scales is often low. For example, low order correlations have been reported for the Mini-Mult by a number of investigators (e.g., Alfano & Finlayson, 1987 ; Armentrout, 1970 ; Armentrout & Rouzer, 1970 ; Harford, Lubetkin, & Alpert, 1972 ; Simono, 1975 ). Studies on the FAM and the MMPI —168 (e.g., Finch, Kendall, Nelson, & Newmark, 1975 ; Freeman, Calsyn, & O'Leary, 1977 ; Griffin & Danahy, 1982 ; Newby, Schroeder, & Hallenbeck, 1982 ) have reported somewhat higher correlations ranging from .52 to .97, but most studies have used a method of internal scoring that gives an overly optimistic picture of the relationship between the two measures.

Methods that have been used to derive MMPI short forms, based on correlational procedures, may not be appropriate for shortening empirically derived measures like the MMPI scales. The application of correlational procedures assumes that each item contributes to the total score, that is, is correlated with the total score. However, this is not true with some traditional MMPI clinical scales as some items contained on them are not correlated with the total score or with relevant external criteria. For example, the original scale-construction method incorporated in the development of scales included several items on each scale that are not related to the external correlates of the scale. These items, often referred to as "subtle items" actually lower the scales' internal consistency and external validity ( Butcher, 1989 ). If an MMPI short form relied on some or most of these items to carry the weight of the scale, both the internal consistency and the external validity would suffer even more than if the short-form scale incorporated more content-relevant items. The problem of differential item validity has not been addressed in the development of any MMPI short form.

The most problematic aspect of MMPI short-form scales is loss of validity. In most cases, the validation of the short forms involved correlating the estimated scores of the short form with full scale scores. Much of the earlier validation research (see Tables 1 and 2 ) involved correlating short-form scale scores with the full MMPI scale score. Because many studies simply scored the MMPI short form from a single administration of the full version, an overly optimistic appraisal of short forms was obtained, as correlations computed from internally scored short forms are artificially inflated. The early optimism over MMPI short forms was, however, dimmed by findings that short forms did not sufficiently allow for determination of profile validity ( Alfano & Finlayson, 1987 ; Armentrout, 1970 ; Armentrout & Rouzer, 1970 ; Bassett, Schellman, Gayton, & Tavormina, 1977 ; Edinger, Kendall, Hooke, & Bogan, 1976 ; Finch et al., 1975 ; Gayton, Bishop, Citrin, & Bassett, 1975 ; Hartman & Robertson, 1972 ; Jabara & Curran, 1974 ; Percell & Delk, 1973 ; Scott, Mount, & Kosters, 1976 ; Thompson, 1979 ), and they did not fare well when evaluated against clinically relevant criteria such as whether the short form actually predicted the profile code type ( Armentrout, 1970 ; Armentrout & Rouzer, 1970 ; Finch et al., 1975 ; Gaines, Abrams, Toel, & Miller, 1974 ; Harford et al., 1972 ; Hoffmann & Butcher, 1975 ; Huisman, 1974 ; Jabara & Curran, 1974 ; Palmer, 1973 ; Percell & Delk, 1973 ; Rybolt & Lambert, 1975 ; Svanum, Lantz, Lauer, Wampler, & Madura, 1981 ; and Thornton, Finch, & Griffin, 1975 ). Although one study reported very strong external correlates for the MMPI short forms ( Newmark, Ziff, Finch, & Kendall, 1978 ), further analyses of these data raised serious questions about the veracity of the results, and appropriate cautions in interpreting them were suggested ( Butcher, Kendall, & Hoffmann, 1980 ).

In general, shortened MMPI scales have been considered to be too inaccurate to use in clinical evaluations ( Dahlstrom, 1980 ; Graham, 1987 ; Greene, 1982 ; Hoffmann & Butcher, 1975 ; Lachar, 1979 ), although some limited research application has been seen to be potentially valuable ( Dahlstrom, 1980 ). Potential uses include, for example, assessing populations of physically impaired individuals, where full-scale MMPI results are difficult to obtain. However, it should be noted that resulting scores are likely to operate differently than full MMPI scales. Researchers developing or using a short form of the MMPI should consider it a new test ( Greene, 1982 ) and use "honesty in labeling" as shortened scales are not equivalent measures. Studies have been published that used the MMPI as a criterion of psychopathology where full MMPI scores were not used (e.g., Frame & Oltmanns, 1982 ). The results of such studies are later inappropriately cited in literature as evaluations of the full MMPI scale.

Clinically valuable information is lost when a short form is used in lieu of the full MMPI . The short forms that have been developed are limited to the clinical and validity scales and do not allow for the assessment of other supplementary measures in the assessment. For example, shortened scales do not allow for the application of useful measures such as the Harris—Lingoes subscales, which are valuable in assessing the meaning of a particular scale elevation and do not permit the scoring of the MAC—R scale or the MMPI—2 content scales.

No effort was made by the MMPI Restandardization Committee ( Butcher et al., 1989 ) to preserve any of the MMPI short forms in the MMPI—2 . Consequently, some of the items constituting previously developed MMPI short forms might have been deleted from the MMPI in the revision process. Researchers interested in scoring one of these short forms would need to verify their availability in the MMPI—2 item pool.

Although the MMPI short forms were well intended, they actually missed the mark in that they have not demonstrated sufficient power in estimating scale scores and profile codes to be confidently used in clinical prediction.

One approach ( Clavelle & Butcher, 1977 ) suggested that a core group of about 87 MMPI items could be administered to all subjects to estimate which MMPI code type the subject is likely to have. Then the remaining items contained in the scales making up that code type could be administered to the subject to obtain his or her likely level of scale elevation. The other items on less pertinent scales would not be administered. This approach was shown to be effective at reducing items while still producing full scale scores of relevant scales.

A second, highly effective approach for shortening item administration, the countdown method, was suggested by Butcher, Keller, and Bacon (1985) and empirically verified by BenPorath, Slutske, and Butcher (1989) . In this approach, a running tally is kept of the subjects' responses by the computer as items are administered. When it has been determined that a scale cannot reach a predetermined level of critical significance, say a T score of 65, then no more items are administered from the scale. Ben-Porath et al. (1989) demonstrated that considerable savings in item-administration time resulted with this procedure, yet no information was lost because the full scale score of the pertinent, most elevated scales was obtained.

A third approach to computer-derived scale abbreviation, item response theory ( IRT ), has been shown to be effective in appraising unidimensional variables such as dimensions of ability ( Weiss, 1985 ). In this approach, item-characteristic curves, based on properties such as difficulty level, are used to determine which items on a particular scale to administer depending on the individual's response to previous items. However, item response theory-based strategies are not effective with factorially complex (multidimensional) personality scales such as the MMPI empirical scales. IRT can be effectively used to estimate the full scores of homogeneous, unidimensional scales such as the MMPI—2 content scales ( Ben-Porathh, Waller, Slutske, & Butcher, 1988 ).

There are no practical, available adaptive MMPI—2 programs for use at this time; however, as their feasibility has been demonstrated, it is likely that effective adaptive programs will be available in the future.

A review of the research on various approaches to reducing MMPI item administration shows that there are a number of problems involved in using abbreviated or short-form versions of the inventory. The researcher or clinician, in choosing an alternative shorter version of the MMPI item pool, runs the risk of losing information or using an unreliable and unvalidated measure while believing it is equivalent to the standard MMPI measure. The least risky reduced item-administration procedure available at present involves the abbreviated administration of the instrument. The full traditional clinical and validity scales can be obtained by having the subject respond only to the first 370 items. Moreover, the new MMPI—2 booklet is more user friendly, even though it is about the same length as the original instrument. Because objectionable, redundant, and nonworking items have been eliminated from the inventory, subjects taking the revised version of the instrument have fewer complaints about the full test. Practitioners are likely to be more interested in obtaining responses to the full MMPI—2 item pool as there are a number of promising new scales whose items are scattered throughout the booklet ( Butcher, Graham, Williams, & BenPorath, 1990 ).

Short forms for the MMPI—2 have been found to be inadequate measures of the constructs assessed by the full scale score, and they are consequently likely to be of less interest to clinicians and researchers in the future for reasons already cited. However, if a particular research question or clinical application dictates that a short form be used, the interested researcher needs to determine if any previously reported short forms, such as the Mini-Mult, FAM , or MMPI —168, can be fully scored from the MMPI—2 item pool. In any case, if a short form is used, caution should be taken to communicate clearly what the measures actually entail. That is, if one uses a shortened version of the Depression scale, for example, care should be taken to describe the results so that readers are aware that the full MMPI—2 Depression score has not been obtained. Researchers may conclude that there is a need for an MMPI—2 short form for studying special populations; however, clear labeling in the resulting report, indicating that the assessment is not based on a full version of the instrument, is needed.

Adaptive administration strategies are being developed to reduce the number of items actually administered to the subject for some assessments without actually losing power in the measures needed. Adaptive personality assessment, that is, tailoring the item administration to the subject, contingent on problems, setting, or previous responses, is a promising avenue of pursuit for personality assessment.

In conclusion, a number of factors bear on whether to use an abbreviated item administration for the MMPI—2 and which type of abbreviated version is chosen. The cost, in terms of information lost or unreliability of measurement, of using an abbreviated version of the MMPI in clinical assessment or research might be too high to warrant use for most applications.

Table 1.

Table 2.