An MMPI Study of Adolescents
I. Empirical Validity of the Standard Scales

This study is the 1st since the early work of the Minnesota Multiphasic Personality Inventory (MMPI) developer, Starke Hathaway, to present empirical descriptors for the standard MMPI clinical scales for a large adolescent sample. The 844 Ss (492 boys, 352 girls) ranged in age from 12 to 18 years and were obtained from substance-abuse inpatient (60%), psychiatric inpatient (24%), day treatment (8%), or special school (7%) settings. The Devereux Adolescent Behavior Rating Scale and the Child Behavior Checklist, as well as a thorough record review, were used to provide a list of descriptors. Procedures from previous MMPI correlate studies were used to determine the significant behavioral descriptors by gender for each of the standard MMPI scales. The descriptors found were consistent with those found in adult studies, with some replication across the 3 criterion measures. This provides support for the most frequently used MMPI interpretive strategy for adolescents of using MMPI scale descriptors from studies based on adults.

Much less is known about the assessment of emotional and behavioral problems in adolescents than in adults, a fact that hampers both epidemiological research and clinical practice. The Minnesota Multiphasic Personality Inventory (MMPI), although developed originally for adults, is used widely with adolescents ( Archer, 1984 , 1987 ; LeUnes, Evans, Karnei, & Lowry, 1980 ; Williams, 1986 ). Unlike the extensive research on the MMPI with adults, research with adolescents is more limited, and problems have been noted over the years. These problems are quite basic, with some even raising doubts about the appropriateness of the test for adolescents. Major early studies indicated promising validity of the MMPI with adolescents (e.g., Hathaway & Monachesi, 1963 ; Hathaway, Reynolds, & Monachesi, 1969 ; Marks, Seeman, & Haller, 1974 ). However, questions remain, including whether a separate norm set is needed for adolescents, the validity of scale and code type descriptors for adolescents, and whether additional adolescent-specific item content would enhance the MMPI's validity for this age group.

There seems to be a growing consensus that age-appropriate norms are needed for interpreting adolescents' profiles but that the most frequently used adolescent norms (i.e., those presented by Marks et al., 1974 ) are problematic in contemporary adolescent settings ( Archer, 1984 , 1987 ; Pancoast & Archer, 1988 ; Williams, 1986 ; Williams, Graham, & Butcher, 1986 ). A major problem with the adolescent norm set ( Marks et al., 1974 ) is that mean profiles on adolescent norms of adolescents in clinical settings do not reach clinical significance; this has led to the suggestion that an adolescent T score of 65, rather than the traditional adult T score of 70, be used as a cutoff for clinical significance. Williams et al. (1986) , after analyzing preliminary data from the MMPI Restandardization Project, concluded that new adolescent norms will likely be needed for the MMPI-2.

Very little research is available addressing the empirical validity of the standard MMPI scales in adolescent populations. MMPI scale validity in adults has been established through the study of behavioral descriptors and their differential associations with MMPI scales. MMPI scale descriptors have been defined further as personality descriptions and symptomatic behaviors that have been shown to be associated with various clinical scale elevations ( Graham, 1987 ). Only Hathaway and Monachesi (1963) presented such behavioral descriptors for the MMPI scales for a large sample of adolescents. They studied 10,104 valid profiles of normal ninth-graders from Minnesota. Williams (1986) summarized their findings for each of the 10 standard scales. Many of the scales (i.e., Scales 1, 2, 3, and 7, or those indicative of more neurotic or internalized problems in adults) had few descriptors. The descriptors associated with the other scales generally were valid compared to findings from adult populations (e.g., high Scale 4 scores were associated with acting-out behaviors, family problems, and school difficulties).

However, considerably fewer scale descriptors were identified for adolescents than adults, most likely because the adolescents were from the general population rather than clinical settings and because of limitations with the descriptors included in the study. A more recent study of psychiatric inpatients classified 73 adolescents into one of five high-point groups (i.e., elevated Scales 2,3,4,8, and 9) and reported descriptors for these scales that were generally consistent with the literature on adult-derived MMPI scale correlates ( Archer, Gordon, Giannetti, & singles, 1988 ).

The only extensive MMPI study of adolescents from clinical settings ( N = 834 outpatients) did not present descriptors for the standard MMPI scales; instead, it focused on the code type approach ( Marks et al., 1974 ). A study by Ehrenworth and Archer (1985) raised questions about the validity of the adolescent code type descriptors provided by Marks et al. (1974) . In the relative absence of empirical descriptors for MMPI scales and code types from adolescent samples, interpretations of MMPI profiles for adolescents are based on descriptors found with adult samples (e.g., Archer, 1987 ; Williams, 1986 ).

Some of the questions reviewed here will be addressed by the MMPI Restandardization Project. The Restandardization Project was started in 1982 by the University of Minnesota Press, and a committee ¹ was established to oversee the project. One of the initial decisions of the committee was to take a conservative approach to the restandardization by ensuring that all of the standard validity and clinical scales, and some special scales, be maintained in the revision. Further decisions included the development of a separate experimental booklet for adolescents that contained the items from the original standard scales as well as adolescent-specific item content, and plans were made for collecting normative data from 12—18-year-old subjects. No final decisions were made about whether a separate adolescent booklet or norms would be published. Instead, data collected in the project would guide those decisions.

The current two-study series is a collaborative project initiated with some MMPI Restandardization Committee members and represents the first step in the development of the MMPI-2 for adolescents. In keeping with the conservative philosophy established by the Restandardization Committee, this study seeks to determine the empirical validity of the standard MMPI scales in a large contemporary sample of adolescents in treatment before the instrument is changed substantially. This information can then be used to determine if, for example, the adolescent-specific content enhances the validity of the standard scales or if completely new scales using the adolescent items have greater validity for adolescents than the standard scales. The information presented here has two other uses as well, including giving researchers and clinicians information about the empirical validity of the MMPI scales that can be used during the developmental period when the MMPI-2 for adolescents is unavailable. Second, as is the case with the MMPI-2 for adults, even if a new adolescent MMPI-2 booklet is published, the standard MMPI scales most likely will be included in it and thus these validity data will remain relevant. Once the validity of the standard scales is determined, the next study in this series will examine code type classification procedures for adolescents.

A major challenge in designing a validity study of MMPI scales with adolescents is the selection of adequate criterion measures. As noted earlier, adequate assessment measures for adolescents are generally lacking, and, in fact, the MMPI is used frequently as a criterion or treatment outcome measure for adolescents! Because there is no stellar criterion measure for adolescents, we adopted a multimethod strategy, selecting two frequently used objective behavior rating scales and developing a third measure from record reviews. The Devereux Adolescent Behavior Rating Scale (DAB), developed by Spivack and Spotts (1967 ; Spivack, Haimes, & Spotts, 1967 ), and the Child Behavior Checklist ( Achenbach & Edelbrock, 1979 , 1983 ) are established measures of adolescent psychopathology that have been recommended for use in epidemiological and clinical research ( Orvaschel, Sholomskas, & Weissman, 1980 ). Neither measure is self-report, an advantage when validating a self-report instrument: The DAB uses treatment staff and the CBCL uses parents as raters. Both are multidimensional measures of psychopathology, consistent with the MMPI. The multimethod strategy allows us to replicate across the three measures in the current study as well as to compare our results with previous adult and adolescent research on scale descriptors. We will examine the clinically relevant behavioral descriptors from our three measures separately by gender to determine if gender contributes to differential descriptors.

The experimental adolescent form of the MMPI, developed by the University of Minnesota's MMPI Restandardization Project and the present authors, was used in this study. Form TX contains the original 550 MMPI items, 50 items assessing contemporary issues such as drug use. eating disorders, and treatment readiness, and 104 items covering adolescent concerns such as school problems, peer group influences, and identity issues. However, for the purposes of this study, only the original 550 items, with slight modifications to update the language, were used to score the 3 validity and 10 clinical scales. When T scores were used in the analyses, they were deter mined by the Marks et al. (1974) norms.

The DAB, developed by Spivack and colleagues ( Spivack et al., 1967 : Spivack & Spotts, 1967 ), is a broadbased behavior rating scale completed by treatment staff. The counselors or nursing staff in the present study were instructed not to look at the MMPI profiles until after they completed the DAB. Although it would have been preferable to guarantee absolutely blind DAB ratings (e.g., by removing the MMPI from the patients' hospital records), this was not possible in the naturalistic settings. However, we are confident, given the close monitoring and observations of our research assistants, that the treatment staff cooperated with the procedures, understood the importance of independent DAB ratings, and were motivated to produce uncontaminated data.

The 84 DAB items covered a wide range of problem behaviors disturbed adolescents in easily understood language requiring little inference. Because the original DAB scales were not developed from the instrument's present item list ( Spivack et al., 1967 ; Spivack & Spotts, 1967 ) and because our previous work with the DAB using subjects from the current sample revealed low internal consistencies and high interscale correlations in those original scales ( Ben-Porathh, Williams, & Uchiyama, 1989 ; Williams, Ben-Porath, & Weed, in press ), in these analyses we used the new scales we developed for the DAB. These scales were developed using item-level factor analysis ( Ben-Porath et al., 1989 ) and proved to be reliable and valid measures of Acting Out Behaviors (AOB scale), Withdrawn/Timid Behaviors (WTB scale), Psychotic Behaviors (PB scale), Neurotic/Dependent Behaviors (NDB scale), and Heterosexual Interests (HI scale; Williams, Ben-Porath, Uchiyama, Weed, & Archer, in press ).

The CBCL is a well constructed behavior checklist for children aged 6 to 16 years ( Achenbach & Edelbrock, 1979 , 1983 ). We obtained the author's permission to make slight wording changes to develop a form more appropriate for older adolescents (Achenbach, personal communication, January 1985). The computer scoring program provided by the author was used in this study, and raw scores were used in the analyses because norms were not available for our older subjects. Parents did not have information from their child's MMPI before making their ratings. The separate, narrow-band behavior problems scales for boys and girls on the CBCL were used as potential MMPI scale descriptors.

A standardized form to facilitate a thorough review of each subject's chart or school records was developed for this study. The charts or records were reviewed by nine research assistants who met weekly to ensure consistency across raters. Raters did not review MMPI profiles prior to completing the Record Review Form. We selected 34 record review variables with information about presenting behaviors and problems, suicide, sexual and physical abuse, court or social service involvement, and so forth, as potential MMPI descriptors. Seven of these variables (many dealing with bizarre behaviors) had a frequency of occurrence of less than 10% in the sample and were eliminated from subsequent analyses.

Interrater reliability for the Record Review Form was assessed by having raters independently fill out forms for the same subject in a total of 117 cases (14% of subjects for whom Record Review Forms were obtained). Kappas were computed using Fleiss's (1971) formula for nominal scale agreement with many raters. All but one (i.e., poor social skills) of the remaining 27 variables of interest had agreement exceeding the .05 level of significance. Although they were significant, the magnitude of some of the significant kappas was quite low (e.g., .25). We did not drop any of these items from the analyses because of limited reliability, but it is important to note that our ability to obtain significant empirical descriptors from the Record Review Form was hampered by low reliability.

In each of the treatment facilities the MMPI became part of the standard admission assessment battery, and patients and their parents were invited to participate in the study by treatment staff. Informed consent was obtained from parents before including subjects in the study. Subjects were administered the MMPI in supervised sessions, usually individually, by staff within the first few weeks of admission. In some cases the subjects completed the MMPI in several short sessions, rather than one long session. Social reinforcement and/or tangible rewards such as privileges were used to elicit cooperation.

In the special schools the school psychologist reviewed students' reading scores to identify those with a fifth-grade reading level and then sent letters to parents offering a special psychological assessment at no cost. The school psychologist and research assistants administered the MMPI to groups of no more than 4 to 5 subjects in supervised sessions. Students were excused from classes to complete the testing and received praise, points, and/or candy as reinforcement for cooperation.

Record reviews were completed by the research assistants at discharge or at the end of the school year. Day treatment and hospital staff completed DAB ratings after a period of observation and were instructed not to complete a DAB if they did not know the subject well enough. At the substance-abuse sites, the DAB was completed at the end of 2 weeks on the treatment units, because subjects did not stay long enough for assessment on the evaluation units. Some subjects do not have a DAB because they were discharged too soon or were not known well enough by staff. The schools did not have treatment counselors, and so DAB data were not collected there.

CBCL data were collected in the hospital settings. Parents were given the CBCL by the admissions staff and asked to complete it within a few days of the adolescent's admission. Staff in the day treatment and school settings were reluctant to request this information from parents, so CBCLs were not collected at these sites.

We used MMPI scale descriptor research from adult populations (e.g., Graham, 1987 ), adolescent MMPI research (e.g., Hathaway & Monachesi, 1963 ), and our recent work with the DAB ( Williams, Ben-Porath, Uchiyama, Weed, & Archer, in press ) to formulate predictions about the relationships of our criterion measures to the 10 MMPI scales. Because the number of predictions is long, we offer the following summary of our strategy for making hypotheses. We reviewed lists of empirically derived MMPI scale descriptors offered by Graham (1987) and then compared them with our measures to identify similar variables. For example, among descriptors for Scales 1, 2, 3, and 7 Graham (1987) included somatic complaints. Two of our similar measures, the CBCL Somatic Complaints scales and the Record Review Form variable of somatic complaints, were therefore predicted to be related to Scales 1, 2, 3, and 7. Likewise, Graham (1987) found prominent descriptors for Scales 6 and 8 to be psychotic behaviors, and we predicted our measures of these behaviors (i.e., the DAB PB scale and the CBCL Schizoid scale) would be related to these scales.

Hathaway and Monachesi's (1963) study indicated that Scales 0, 2, and 5 are "inhibitory scales," or associated with the lowest delinquency rates, whereas Scales 4, 8, and 9 are "excitatory scales," or associated with the highest delinquency rates. Thus we predicted that our measures of acting-out problems would be negatively related to Scales 0, 2, and 5 but positively related to Scales 4, 8, and 9. Our recent validity study of the DAB revealed the NDB scale to be a measure of both internalizing and externalizing behaviors ( Williams, Ben-Porath, Uchiyama, Weed, & Archer, in press ), thus prompting the prediction that it may be related to the excitatory MMPI scales. The DAB HI scale proved to be a stronger acting-out scale for girls than boys ( Williams, Ben-Porath, Uchiyama, Weed, & Archer, in press ), which suggested that it may be related to Scale 4 in girls, but not boys.

Although there are age and gender differences in adolescents' MMPI responses (e.g., Hathaway & Monachesi, 1963 ), MMPI descriptors typically are determined using age- and gender-corrected adolescent T scores and so are not presented separately by gender or age (e.g., Archer et al., 1988 ; Marks et al., 1974 ). However, others suggest that personality scale descriptors may differ by gender and age in adolescents, particularly in the measurement of depression ( Block & Gjerde, in press ; Gjerde, Block, & Block, 1988 ). Because of this, we used MMPI raw scores in our analyses to allow us to investigate whether different descriptors would be apparent by gender. Our sample size did not allow us to divide the sample further by age. However, previous MMPI work could not guide us in predicting different relationships by gender between the MMPI scales and criterion variables, other than the above prediction about the DAB HI scale and MMPI Scale 4. Our multimethod strategy allows us to determine if our descriptors replicateacross the three criterion measures in the present study.

In order to eliminate the possible confounding of MMPI scale scores with deviant or invalidating response sets, we used stringent exclusion criteria for the MMPI validity scales in the present study. Profiles with a Cannot Say ( CS ) scale raw score greater than 10, a Lie ( L ) scale adolescent T score greater than 70, an Infrequency ( F ) scale adolescent T score greater than 90, and a Defensiveness ( K ) scale adolescent T score greater than 70 were considered invalid and eliminated from the analyses. As shown in Table 1 , the application of these validity criteria resulted in elimination of 119 subjects from the analyses (14.1%). Almost twice as many girls (19.6%) as boys (10.2%) produced invalid MMPI profiles according to these criteria.

Previous MMPI correlate studies (e.g., Graham & McCord, 1985 ; Lewandowski & Graham, 1972 ) and recommendations by Green (1982) were used to determine the analysis procedures for the present study, although we made some modifications. One was the use of MMPI raw scores, instead of the non-K-corrected adolescent T scores ( Marks et al., 1974 ), in the analyses, allowing for the study of gender differences. Had we used the adolescent norms and the traditional adult cutoff T score of 70, 38% of our sample would have been classified as normal. We assumed this to be a false negative rate because the sample was predominantly hospitalized adolescents and because other assessment data indicated that the sample was seriously disturbed (e.g., Williams, Ben-Porath, & Weed, in press ). These results confirmed long-held suspicions of problems with the current adolescent norm set in clinical samples.

An alternate procedure was used to determine the appropriate cutoff scores for this sample. We split the sample by gender at the third quartile of subjects for each clinical scale distribution to define high-scoring boys and girls ² Male and female subjects scoring in the top 25% of each clinical scale were compared with the remaining subjects in their gender subsample (i.e., the comparison sample) according to procedures used in previous studies ( Graham & McCord, 1985 ; Lewandowski & Graham, 1972 ) and the recommendations of Green (1982) , Either/tests or chi-square analyses were used to determine statistical significance. The t tests were used for variables assumed to be continuous (i.e., DAB and CBCL scales), and chi-square analyses were used for the categorical variables (i.e., Record Review Form variables). One-tailed t tests were used when direction was predicted in our hypotheses, and two-tailed t tests were used when direction could not be predicted. A significance level of .01 was chosen for reporting empirical descriptors, based on Green's (1982) recommendations to control for both Type I and Type II errors in MMPI descriptor research. The decision to split the groups at the third quartile and perform significance tests, rather than using a dimensional analysis such as correlation or multiple regression, reflects our desire to report the results in a manner similar to the way the scales are used clinically, as a typology of problem behaviors. This seems particularly appropriate given that ours is a clinical sample. It is likely that some loss in statistical power is experienced through use of this contrast group procedure ( Cohen, 1965 ); thus, the results are likely to be conservative and may underestimate the strength of the relationships between the scales and external behaviors.

Table 2 presents the five MMPI scales demonstrating statistically significant differences between the MMPI-scale high scorers and their comparison groups on the five DAB scales. MMPI Scales 1, 2, 3, 5, and 7 are not included in Table 2 because they have no statistically significant DAB descriptors. The data in Table 2 are presented separately by gender and include the mean DAB scores of the comparison groups and of the high-scoring groups on the five MMPI scales, their standard deviations, and the t values. Table 2 reveals that MMPI Scales 4 and 0 receive the strongest support for their validity for both genders based on the DAB scales. Scale 4 is related significantly to the DAB acting-out scales, whereas the DAB descriptors for Scale 0 confirm its status as an inhibitory scale measuring withdrawn and timid behaviors in both genders. The DAB HI scale, a measure of sexual acting out, was related to high scores on Scale 4 only in girls, as predicted. The DAB analyses confirmed the validity of the MMPI scales associated with psychotic disorders (Scales 6 and 8) in boys only and suggest that Scale 6 may join Scales 4, 8, and 9 as an excitatory scale, again for boys only. Given Hathaway and Monachesi's (1963) difficulty in finding behavioral descriptors for the more neurotic MMPI scales (i.e., Scales 1, 2, 3, and 7), it is interesting to note the lack of a relationship in both genders between the DAB scales measuring similar problems and these MMPI scales. However, the DAB NDB and WTB scales are not exactly analogous to any MMPI scale, with the exception of the WTB scale's relationship with Scale 0, described above.

Table 3 presents similar information about the significant relationships between the CBCL scales and high scorers on eight of the MMPI scales, as did Table 2 for the DAB-MMPI relationships. Because the CBCL Schizoid, Immature Hyperactive, Aggressive, and Cruel scales for girls were not related significantly to any MMPI scales, they are not included in Table 3 . Likewise, MMPI Scales 5 and 7, with one significant CBCL description each, are not included in Table 3 either in order to conserve space. However, Scale 5 for boys was related to the CBCL Hostile Withdrawal scale: comparison group, M = 7.5, SD = 5.0, n = 210; high scorers, M = 9.4, SD = 4.6, n = 73; t (281) = -2.88, p <= .005, two-tailed test. Scale 7 for girls, as predicted, was related to the CBCL Somatic Complaints scale: comparison group, M = 3.5, SD = 3.1, n = 148; high scorers, M = 5.1, SD = 3.7, n = 45; t (191) = -2.91, p <= .005. An important difference between the CBCL and DAB is that the CBCL scales differ by gender, as is obvious in Table 3 . A striking difference in the validity data presented for the two measures is the number of significant CBCL correlates for MMPI Scales 1,2,3, and 7. This is to be expected because many of the CBCL scales are more directly related to these MMPI scales. For example, the CBCL Somatic Complaints scale, as predicted, is related significantly to MMPI Scales 1 and 3 in both genders and to Scales 2 and 7 in girls. Scales 4 and 9, again as predicted, are related to several of the CBCL acting-out scales in both genders. The excitatory nature of Scale 6 in boys, found in the DAB analyses, is replicated for boys with the CBCL data. The inhibitory nature of Scale 0 also is replicated with the CBCL data, but only for girls. The CBCL relationships suggest that MMPI Scales 1 and 3 in girls may have inhibitory effects, given their negative relationships with some of the CBCL acting-out scales. Some predicted relationships were not found, as revealed in Table 3 . For example, the CBCL Depressed Withdrawal scale for girls was not related to MMPI Scale 2, and Scale 5 was not found to be an inhibitory scale for either gender.

The chi-square analyses used to determine MMPI scale descriptors from the dichotomous Record Review Form variables are presented in Table 4 for eight MMPI Scales. As before, MMPI scales with fewer Record Review Form descriptions are not included in Table 4 because of space limitations. However, physical abuse was related significantly to Scale 5 in boys: χ ² 1 = 7.16, p <= .01 , n of comparison group = 331, n of high scorers = 102. Scale 6 had two significant Record Review Form describers in boys: poor social skills, χ ² 1 = 7.13, p < .01, n of comparison group = 324, n of high scorers = 109; and sexual abuse, χ ² 1 = 7.49, p <= .01, n of comparison group = 324, n of high scorers = 109. Overall, even given the limitations of the data available in hospital and school records and our somewhat low reliability, the Record Review Form provided clear empirical support for many MMPI scales and replicated the previous findings from the DAB and the CBCL. However, the findings did not always replicate across genders within the Record Review Form. Only the depression variable was related significantly to the predicted MMPI Scales 2 and 0 in both genders. Scales 1 and 3 were found to be associated with somatic complaints and eating problems among girls, and Scale 1 was associated among boys with having been sexually abused. Boys scoring high on Scale 4 had significantly more incidents of running away and suicide threats, and girls scoring high on Scale 4 were seen as being more sexually active than the comparison groups. Finally, boys scoring high on Scale 9 were rated as having more concentration difficulties, and girls scoring high on Scale 9 were rated as being more sexually active than the comparison groups.

The inhibitory nature of Scales 2 and 0 was confirmed with regard to the drug use variable for girls (i.e., girls scoring high on Scales 2 and 0 were less likely to be rated as using drugs). Girls scoring high on Scale 2 also were rated as less likely to act out. Scales 1 and 3 also demonstrated inhibitory effects in girls, as is apparent in Table 4 and which is consistent with the CBCL descriptors. Predictions for the excitatory scales were not confirmed by the Record Review Form data. Surprisingly, boys scoring high on Scale 8 were rated as less likely to have anger and fighting noted in their records, although the magnitude of this effect was relatively small.

Overall, the results of this study demonstrated that the MMPI standard scales were associated with clinically relevant behavior and symptoms in troubled adolescents, as rated by their parents and treatment staff and as found in their records. Scales 1 (Hypochondriasis), 2 (Depression), 3 (Hysteria), 4 (Psychopathic Deviate), 8 (Schizophrenia), 9 (Hypomania) and 0 (Social Introversion), as well as to a certain extent Scales 6 (Paranoia) and 7 (Psychasthenia), were shown to have descriptors consistent with the adult MMPI literature, providing support for the use of adult descriptors for adolescent MMPI interpretations, as previously suggested by Archer (1987) and Williams (1986) .

Scale 5 (Masculinity-Femininity) had a notable absence of clinically relevant descriptors across both genders and all measures used in this study. Scale 5 was developed originally by Hathaway and McKinley to identify male "sexual inverts" (they were unsuccessful in their attempts to create a female "sexual invert" scale, so convention has dictated using Scale 5 for both genders), and it was included in the original MMPI in a preliminary form ( Graham, 1987 ). Over the years, there has been controversy about what Scale 5 measures in adults. The lack of external correlates found in this study suggests that its usefulness in understanding adolescent psychopathology in clinical settings is limited. Given this low validity and its large number of items (60), its value needs further evaluation before inclusion in the MMPI-2 for adolescents.

The inhibitory/excitatory dimension of the MMPI scales, described by Hathaway and Monachesi (1963 ), was partially confirmed in the present study. Hathaway and Monachesi (1963) reported inhibitory and excitatory effects for both genders on these scales. In our study, Scales 0 and 2 were least frequently associated with our acting-out variables, but only among girls. Elevations on Scales 1 and 3 also seemed to inhibit expressions of acting out, again only among girls. Perhaps if our sample was as large as that of Hathaway and Monachesi and had included adolescents from the general population, we would have been able to detect these effects among boys as well. Scales 4, 8, and 9 were associated strongly with measures of acting out in both genders. Scale 6 may be an excitatory scale for boys. Achenbach and Edelbrock (1983) considered a similar dimension to adolescent psychopathology in their discussions about the internalizing/externalizing scales of the CBCL.

We found only weak evidence that symptoms of depression vary along the internalizing/externalizing dimension by gender in adolescents on the MMPI, as indicated by the work of Block and colleagues with other measures of psychopathology ( Block & Gjerde, in press ; Gjerde et al., 1988 ). Among boys scoring high on Scale 2, the only significant relationship with an externalizing variable was with the CBCL Hostile Withdrawal scale, which is described as a "mixed" scale ( Achenbach & Edelbrock, 1983 ). Further inspection of the content of its 14 items revealed only 3 "pure" externalizing items. Thus, it appears that the gender differences in the expression of depressive symptoms found in Block's longitudinal study of a sample selected from the general population are less salient in our clinical sample using the MMPI. Our results, in most respects, show a rather limited impact of gender on MMPI descriptors for adolescents from clinical settings when using established procedures for determining MMPI behavioral correlates. Further research, especially studies using T scores, could combine male and female subjects when sample size is a critical issue and when studying adolescents in treatment settings.

This study offers reassurance to clinicians and researchers using the MMPI with adolescents that traditional empirical descriptors for most MMPI scales are valid. These results apply to adolescents in psychiatric and substance-abuse inpatient, day treatment, or special school settings, but their generalizability to other settings needs to be demonstrated.

1

Original committee members included James N. Butcher, W. Grant Dahlstrom, and John R. Graham. Auke Tellegen joined the committee in 1986.

Table 1. Validity Patterns of the Adolescent Sample

Table 2. Empirical Descriptors for the MMPI Scales From the DAB

Table 3. Empirical Descriptors for the MMPI Scales From the CBCL

Table 4. Empirical Descriptors for the MMPI Scales From the Record Review Form