h O Oh o •Si 0) II Q *2 o o 3 ' S CO u. JJ C 3 co .2, c X o a. o <5 u j r i 2 Id fi » ?h -2 •a > E cd (U « W W J X 0 S < ss Q o a V E c u o T3 01 C > II KJ J> £ c n O X Oh CO W 3 Q >>*3 O H B ?. .> o _; cd S i-T o G O Q "H o S .21 II ° 2 5 ft « T3 K J *3 B "u .ti *o .& D T3 o "o C c o •c U P-I
48 Psychological Discomfort and Depression (r = .37). An unexpected finding was that the Friends content scale from the CAS did not correlate with corresponding scales from the other two measures, particularly the Social Skills scales on both the PIY and PAS. To examine discriminant validity, convergent validities should be higher than both heterotrait-heteromethod and heterotrait-monomethod correlations (see Table 9). In this study, the convergent validity coefficients (M = -39) for the PIY and PAS scales did exceed substantially the heterotrait-heteromethod coefficients between the PIY and PAS (M = -25). However, the heterotraitmonomethod correlations for the PIY were moderately high (r = .46) and the relationship between the PIY and CAS (range from -.15 to .37; M = -12), was similar to the relationship between the PAS and CAS (range from -.10 to .33; M = .09).
In addition, convergent validities did not significantly exceed heterotrait-
monomethod coefficients (range from .02 to .72; M = -27). PIY convergent and discriminant validities were also examined by diagnoses.
These correlations are displayed in Table 10.
The discriminant validity coefficient for ADHD was slightly higher than the convergent validity, which indicates that the ADHD construct is not specific to the scale.
The ODD/CD construct, likewise, displayed little difference between
convergent and discriminant validity.
On the other hand, Anxiety, Depression
and Social Skills showed moderate convergent validity and sufficient discriminant validity as per the Fiske and Campbell (1995) model.
49 Table 10 PTY Convergent and Discriminant Validity bv Diagnosis Convergent Validity
Note: PIY = Personality Inventory for Youth. a = Discriminant validity as measured on the CAS and PAS. b = Oppositional Defiant Disorder and Conduct Disorder as measured by the Delinquency Scale. 0 = Anxiety as measured by the Somatic Concerns Scale. d = Depression as measured by the Psychological Discomfort scale, and Depression subscale on the Psychological Discomfort scale.
The Campbell and Fiske (1959) model also allows examination of the relative validity (convergent and discriminant) of individual measures, therefore, although problems are apparent with an overall construct, a particular measure may yield satisfactory validity. To examine this possibility, method effects were also determined between the three measures (See Table 11). M validity coefficients for the PIY were compared to the CAS and PAS. For the PIY, the M validity coefficients were generally significant for convergent validity (.39), however, the intercorrelations of the PIY subscales nullifies the discriminant validity (M - -46). Even with removal of those scales which are
50 Table 11 Method Effects for the PIY. CAS, and PAS
Discriminant Validity HeterotraitHeterotraitMonomethod Heteromethod
Note: PIY = Personality Inventory for Youth; CAS = Child Assessment Schedule; PAS = Personal Attitude Scale. a = M coefficients with removal of scales that are predicted to be highly correlated (PIY = Psychological Discomfort and Somatic Concerns; PAS = Aggression, Violence, and Gang Involvement).
predicted to be highly correlated, the discriminant validity is vitiated (M = -43). Examining the heterotrait-heteromethod only, the PIY showed good discriminant validity (r = .17) from the CAS and PAS. To isolate possible relationships between particular measures in a study, removal of lower correlated measures can be done to explore the relationship between the remaining measures. To better examine the relationship between the PIY and PAS, method effects were determined between the two measures (See Table 12). The convergent validity coefficients (M = -52) between the PIY and PAS scales did exceed substantially the heterotrait-heteromethod coefficients between them (M = .25).
51 Table 12 Method Effects for the PIY and PAS
Discriminant Validitv HeterotraitHeterotraitMonomethod Heteromethod
Note: PIY = Personality Inventory for Youth; PAS = Personal Attitude Scale. a = M coefficients with removal of scales that are predicted to be highly correlated.
Contrary to the first hypothesis, a relationship between the Delinquency scale on the PIY and conduct disorder (CD) as measured on the CAS was not found. In addition, no relationship was found between CD as measured by the CAS and any scale on the PIY or PAS which measured acting out behavior (e.g., the aggression scale on the PAS).
A relationship was found between
oppositional defiant disorder (ODD) as measured on the CAS and the Delinquency scale on the PIY (r = .30). The second hypothesis predicted a relationship between the Family Dysfunction scale on the PIY and the Family Section of the CAS.
relationship (see Table 9) was found with a coefficient of .55. The relationship between the Delinquency scale on the PIY and the Depression subscaie on the PIY, as predicted in the third hypothesis, did not reach significance (r = .21).
However, convergence was demonstrated
52 between the PIY Delinquency scale and the PIY Psychological Discomfort scale on the (r = .43).
Additionally, a relationship was found between the
Delinquency scale and Major Depressive Disorder on the CAS (r = .37) and the Delinquency scale and the Depression scale on the PAS (r = .44).
significant correlation was also found between the Delinquency scale and the Somatic Concerns scale on the PIY.
This scale measures physical
discomfort and anxiety. To examine the relationship between the age of the juvenile and the type of disruptive behavior displayed, analyses of variance were conducted between the juveniles' age, the PIY scores on the Delinquency scale and the category assigned to the juveniles at Gainesville based on their type of offense. As discussed previously (see Table 6), when adjudicated into the Texas Youth Commission, juveniles are classified according to one of six categories, based on their criminal behavior or type of offense. The results for the analyses of variance are listed in Tables 13 and 14. The mean PIY Delinquency scale score for this sample was 57.08, and the mean category rating was 3.10. Based on the following results, the category score was a better predictor of juvenile delinquency classification. Statisically, the F value of 1.02 reached significance at the .01 level when comparing Category and age of the offender. A post-hoc analysis of the ANOVA results using the
53 Table 13 Age of Offender by Category
Main Effects of Category
Note: Category delineates Category Ratings based on Categories 1 through 6, as provided by TYC.
Table 14 Age of Offender bv Delinquency Score
Main Effects of Delinquency Score
Note: The Delinquency score is taken from the PIY and is based on a T score of 50, with a standard deviation of 10.
Duncan Multiple-Range Test indicated that no two groups were significantly different at the .05 level.
With ANOVA testing, using the comparison of
Delinquency scores from the PIY and age of the offenderstatisticalal significance was not obtained, with an F value of 3.06 (See Table 14).
54 However, in the post-hoc analysis using the Duncan Multiple-Range Test, the 15 year old group showed the highest delinquency scores (M = 61.88, SD = 10.14), which were significantly different from the other groups at the .05 level. Additional Information This research also sought to examine the correlations between type of home situation and variables, such as ethnic background and type of offender. The home situation was classified in one of five categories: Both biological parents; biological parent and stepparent; single parent, either mother or father; any other form of caregiver, such as aunt, uncle or grandparent; and a child with no adult caregiver. The majority of juveniles fell into the single parent classification (40%), with both biological parents representing 24% of the juveniles.
three classifications were biological and stepparent (18%), other caregiver (15%), and no caregiver (2%).
Both of the juveniles that fell into the no caregiver
classification had been on the streets for a. period of at least one year prior to their arrests.
The largest percentage of juveniles (31%) were Hispanic Americans
and African Americans raised in single parent homes.
The offense classifications
of VOB and GEN showed the highest frequencies in single parent homes (36% and 31% respectively).
The majority of juveniles raised by both biological
parents fell into the VOB classification (41%).
Few assessment instruments are available which provide accurate and consistent clinical information on juvenile delinquents. Additionally, few effective and appropriate assessment tools have been established to identify and assess more severe behavior disorders with the delinquent offender population. Typically, clinicians, case workers, and others who evaluate individuals in the delinquent population rely on case histories, police reports, and diagnostic criteria, in determining mental disorders in the adolescent offender. Historically, assessment research on juvenile delinquency has utilized selfreport inventories, such as the Minnesota Multiphasic Personality InventoryAdolescent (MMPI-A); the Millon Adolescent Personality Inventory (MAPI); the Basic Personality Inventory (BPI); and the Devereux Adolescent Behavior Rating Scales (DAB) The newest of the measures, the MMPI-A, was developed in 1992 strictly for the adolescent population to assess varying degrees of psychopathology and personality characteristics (Butcher et al., 1992).
Current research on the MMPI-
A consists mainly of the reliability and validity data reported in its manual, as few external validity studies have been conducted.
56 The MAPI, which was also constructed for the adolescent population, was developed to measure pathological functioning and enduring personality styles (Reidy & Carstens, 1990). Research utilizing the MAPI (Pantle, Barger, Hamilton & Thorton, 1994) has identified Scale 6 (Forceful), which was based on the Antisocial scale of the MCMI, as identifying symptomatology in adolescents which is associated with conduct disturbance. Thus, the MAPI is reported to have construct validity in the identification of conduct disorders, disruptive behaviors, and depression in adolescence. The BPI, which was developed by Jackson in 1989, provides separate norms for male and female adolescents.
Jackson (1989), using the BPI,
identified homogeneous subgroups of young offenders: mental health maladjustment, interpersonal maladjustment, antisocial delinquency, somatic complaints, and high-risk rebelliousness. The above studies illustrate the efficacy of using a self-report measurement with adolescents with disruptive disorders.
However, these assessment
instruments are lacking both (a) specific applications to a delinquent population, and (b) validation of a measure that addresses the more extreme, disruptive behaviors, such as assault and other forms of violence. With the exception of the MMPI-A, previous instruments were developed to assess psychiatric disorders in the adult population, and later extended to the adolescent population on the basis of norms, but not by developing appropriate content. The PIY was recently developed by Lachar and Kline (1994), and was
57 based upon PIC items, which were rephrased as a self-report inventory specifically for children and adolescents of the ages nine to 18. Based upon its specific item content, the PIY was chosen for this study for the purpose of determining if it possessed construct validity in its application with a delinquent population. Construct Validity Determining construct validity includes demonstrating convergent validity (obtaining a correlation between scores which are predicted to show a relationship), as well as discriminant validity (obtaining little or no correlation between scores which are predicted to show no relationship).
construct validity of the assessment instrument is supported if it demonstrates moderate to high correlations with instruments designed to measure the same constructs, while demonstrating low to no relationship with instruments designed to measure different constructs (Golden, Satwicki, & Franzen, 1984), assuming the instruments used for comparison have demonstrated construct validity. Campbell and Fiske (1959) devised an experimental design which addresses both types of validity in a multitrait-multimethod matrix. When using the multitrait-multimethod matrix, it should be noted that successful validity coefficients are often modest, typically in the .30 to .50 range (Fiske & Campbell, 1992). Convergent Validity of the PTY Overall, the first hypothesis, which describes predicted relationships
58 between constructs, was supported between the PIY and the PAS, but not all of the relationships with the CAS were supported.
Specifically, the diagnosis of
ADHD, as measured on the CAS and the ADHD scale on the PIY, showed no significant convergence. PAS ADHD scales.
However, a relationship was found between the PIY and
This finding may be attributed to the fact that both scales
also address concepts of anxiety.
Additionally, the PIY ADHD scale measures
purposeful intrusiveness and acting out behavior commonly associated with disorders such as conduct disorder and oppositional defiant disorder (i.e., "I often nag and bother other people," "Nothing scares me," and "I do not learn from my mistakes").
It appears that both the PIY and PAS ADHD scales are measuring
more than just symptoms of ADHD, and therefore do not correspond with a strict diagnosis of ADHD, as is found on the CAS measure. A significant relationship was found between the Delinquency scale on the PIY, Aggression, Violence, and Gang Involvement scales on the PAS, and CAS symptoms of oppositional defiant disorder, but not for conduct disorder.
finding indicates the CAS oppositional defiant symptoms are more related to the PIY and PAS measures of delinquency than CD.
Thus, the PIY and PAS
emphasize the oppositional defiant component of conduct disorder more so than the more aggressive and physically violent behaviors evident in the diagnosis of conduct disorder. A positive relationship was found between the Psychological Discomfort scale on the PIY, the Depression scale on the PAS, and CAS symptoms of
59 major depressive episode, but not for the Somatic Concerns scale on the PIY. The Psychological Discomfort scale on the PIY does contain a Depression subscale. Therefore, the Psychological Discomfort scale on the PIY, the Depression scale on the PAS, and the CAS depressive symptoms are describing psychological discomfort and dysphoric mood.
The Somatic Concerns scale
appears to be measuring anxiety and worries about psychosomatic concerns, rather than depression. Thus these results, as reported in Table 9, support the convergent validity of these scales.
The convergent validity coefficients, with the exceptions noted
previously, were between .37 to .62, which falls within the range of successful validity coefficients as noted by the Fiske and Campbell (1992) model. Moderate convergence was found between the Social Skills Deficits scales on the PIY and PAS, however, the Friends content scale on the CAS was not significantly related.
It would appear that the Friends content scale is measuring
something other than social skills or social interaction. The questions on the Friends content scale are directed more to identification of friendships, versus social interaction skills.
This finding is consistent with the poor social skills and
lack of close friendships found among delinquents. An interesting finding was the relationship between the Delinquency Scale on the PIY and the diagnosis of Oppositional Defiant Disorder on the CAS.
the same relationship did not occur with the diagnosis of Conduct Disorder, it is hypothesized that the Delinquency Scale on the PIY taps into behavior more
indicative of children and adolescents diagnosed with Oppositional Defiant Disorder.
On the PIY, the majority of juveniles did not me, the cutoff for the
delinquency scale, and therefore werP
The low scores
on the delinquency scale su gg ests that the test manual, Cinically referred population sample, which was not screened for delinquency, 3
D U m b W
i s l i k e l y (() h a v e
when used with a d e c e n t population.
™ » the test manual how this cutoff score was determiued.
explanation for the lower than expected Delinquency Scale sco r e s —
D e c e n c y Scale taps into is
delinquent behavtor by more severe juvenile delinquents and, therefore, is „ 0 , endorsed by them.
This consideration would be likely
normalized their less severe acting „ u , behavior e tssues addressed on the D e I i n ( J u e i l c y
if the JuveniIes
A fourth ourth
•„ consideration may be
s e n i l e s are not attending schoo,, an,or no, l M n g a t h o m e
* support of these fmdings, ,he Delinquency scale does possess a lar g e number of „ems relafcd ,o behaviors more consistent with ODD.
Of the 41
verbally aggressive behavior found with ODD (i.e., "sharing things has been a problem for me," "At times I hurt other people's feelings," and "I am pretty stubborn,"). I„ addition, the Delinquency scale has three subscales, Antisocial Behavtor, Dyscontrol, and Noncompliance, all of which contain some items related to ODD.
However, the Noncompliance scale consists mainly of ODD
type behaviors. Discriminant Validity of the PTY
Campbell and Fiske (1959) proposed that each convergent coefficient should surpass the congelations in the corresponding row and column in its discriminant validity block.
The mean convergent and discriminant coefficients
for the ADHD scale on the PIY were similar.
ADHD scale does not appear to
be measuring specifically a construct of ADHD, but rather a composite of several different constructs.
As discussed previously, this scale also addresses anxiety
related behaviors and symptoms, as well as acting out behaviors associated with ODD or CD.
Therefore, based on the items on the ADHD scale, it shows
considerable item overlap with other scales on the PIY. For the PIY Delinquency scale, the convergent coefficient exceeded the discriminant coefficient. The Delinquency scale does appear to be measuring a construct of acting out behavior. For the PIY Somatic Concerns Scale, the convergent coefficient significantly exceeded the discriminant coefficient. Likewise, for the Psychological Discomfort scale on the PIY, the convergent coefficient significantly
62 exceeded the discriminant coefficient. Both the Somatic Concerns and Psychological Discomfort scales are measuring those traits consistent with bodily concerns and psychological distress.
Finally, for the Social Skills scale on
the PIY, the convergent coefficient (r = .33) with the PAS Social Skills scale exceeded the discriminant coefficient (M = .14).
The Social Skills scale
describes a construct relating to social interaction and interpersonal skills. Based on these findings, construct validation was adequate for the PIY as compared to the guidelines from the Fiske and Campbell (1992) model. Additionally, monomethod convergent and discriminant validity was established for some of the described scales (Delinquency, Somatic Concerns, Psychological Discomfort, and Social Skills).
However, the ADHD scale was not found to
have adequate discriminant validity.
The ADHD scale was highly correlated with
the PIY scales of Delinquency, Somatic Concerns and Psychological Distress, which indicates considerable overlap between these constructs.
Thus, within a
juvenile offender population, this scale does not show good ability to measure this construct. Methods Effect The means of the heterotrait-monomethod coefficients are higher than the means of the heterotrait-heteromethod coefficients by .26 (PIY), .01 (CAS), and .19 (PAS).
This demonstrates that there is very little method effect on the CAS,
however, moderate amounts on the PIY and PAS.
According to the Marsh
63 (1990) guidelines, the PIY has adequate convergent validity, however its discriminant validity (as noted above) is not supported. As was noted previously, the discriminant validity coefficient was higher than the convergent validity coefficient when the CAS was included in the calculations, which indicates that the measurement of the identified constructs is contaminated by other effects. Since the CAS is focusing on diagnosis and is not a self-report measure, it was removed from the calculations as an exploratory approach to compare the convergent validity of the PIY and PAS.
removal the convergent validity increased significantly, which suggests the CAS is not measuring the same constructs, and possibly obscuring the validity of the other two instruments.
Further, with the exclusion of the CAS, the relationship
between the convergent and discriminant validity of both the PIY and PAS changed in a more favorable direction, and the estimates of construct validity improved.
In examining those constructs which are not expected to be similar,
the PIY demonstrated good discriminant validity across measures (Campbell & Fiske, 1959). Heterotrait-Monomethod Correlations of the PIY The PIY manual validation study is based on two sample populations: a regular education sample (represented youth in the United States in the 4th through 12th grades); and a clinically referred sample (represented youth from inand outpatient clinics, school special service clinics, hospitals, and private practice
For comparison purposes with the present study, the clinically referred
sample was utilized. In comparing the heterotrait-monomethod correlations of the PIY from this study with those reported in the PIY test manual, similar relationships between the constructs are noted.
For the PIY test manual clinically referred population,
the heterotrait-monomethod correlations ranged from .04 to .63 (M = .37).
the present study, the heterotrait-monomethod correlations ranged from .18 to .77 (M = .46).
This finding indicates that the PIY test manual showed better
discriminant validity in the clinical population as compared to the present study. This is probably a result of the more homogenous group found in the delinquent population, which are predominantly conduct disordered.
In the following
correlation comparisons, the numbers in parentheses represent the present study an
PIY test manual, respectively. Approximately 50% of the compared correlations were within several
decimal points of each other: ADHD and Delinquency (.67Z.63); ADHD and Social Skills (.22/. 17); Delinquency and Somatic Concerns (.35A33); Delinquency and Psychological Discomfort (.437.43); and Somatic Concerns and Social Skills (.23/. 17).
The heterotrait-monomethod coefficients from the present study strongly
correspond to the heterotrait-monomethod coefficients from the PIY test manual. Heterotrait-Heteromethod Correlations of the PIY The PIY manual, in addition to providing heterotrait-monomethod coefficients, also provides a portion of the heterotrait-heteromethod coefficients in
65 the comparison of the PIY to .he PIC-R.
Of those comparisons that are
provided by the manual, only two constructs (delinquency and social skills) are relevant to the present study.
The correlation between the PIY Delinquency scale
and the PIC-R Delinquency scale was moderate (.52).
The correlation between
the PIY Social Skills scale, and the PIC-R Social Skills scale was modest (.42). In relation to the present study, similar correlations were found between the PIY and PAS for the Delinquency (M = .53) scale, however the correlation for the Social Skills scale was weaker (.33). the Aggression ft = .54), Violence
The Delinquency scale was correlated with = .61), and Gang Involvement
scales from the PAS. For the remaining scale comparisons, the manual does not provide the coefficients necessary.
Therefore, an adequate comparison of the scales from the
PIY and PIC-R is not possible.
Of the two scales which can be compared, the
heterotrait-heteromethod validity appears to be modest. Con-elation of Conduct Disorder and Depression The comorbidity between conduct disorder and depression has been estimated between 1 0 * and 35% (Kashani, Reid, & Rosenberg, 1989; Kovacs, Paulaskas, Gatsonis, & Richards, 1988; Rohde, Lewinsohn, & Seeley 1991).
was previously reported by Rohde et al. (1991), depressed adolescents were significantly more likely than nondepressed to have a concurrent conduct or oppositional disorder. In keeping with past research, the c u r a t study also found a relationship ft . .37) between acting out behavior on the Delinquency scale on
66 the PIY and Major Depressive Disorder on the CAS.
A similar relationship was
found (i = .44) between the Delinquency scale on the PIY and the Depression scale on the PAS.
Interestingly, of the three scales on the PAS developed to
measure forms of acting out behavior (Aggression Scale, Violence Scale, Gang Involvement Scale), only the Violence scale significantly correlated with measures of depression on both the PIY (Somatic Concerns Scale, r = .25; Psychological Discomfort Scale,
= .21) and the CAS (Depressive Disorder,
Violence scale on the PAS, in addition to containing questions about violence, also contains questions which are associated with exposure to violence and which are related to abuse and victimization, rather than acting out violence. The diagnosis of conduct disorder on the CAS correlated negatively with both the PIY Somatic Concerns (i = -.15) .13) scales.
Psychological Discomfort (r = -
Also, the diagnosis of Depression as measured on the CAS showed
a poor con-elation with both the PAS Aggression scale fc = .,0) and Gang Involvement scale (i = .16), which indicates that delinquent juveniles tend to deny or minimize emotional disturbance and psychological discomfort. The relationship between CAS Depressive Disorder and the PAS Violence scale .26) showed a modest correlation.
This correlation with depression would be
expected to be associated with exposure to abnse and victimization. The PIY Delinquency scale and its correlations with measures of depression supports the validity of the co-occurence of conduct disorder and depression.
While the PIY Delinquency scale correlations are modes, (M = .39),
67 they meet the criteria set forth from previous studies indicating the comorbidity between conduct disorder and depression ranging from 10% to 35%. Age and Severity of Offenses The ability of age to be able to discriminate type of acting out behavior (see Table 14) was supported by the scores on the PIY Delinquency scale, but the pattern was different than predicted by the fourth hypothesis.
these scores, the 15 year olds of this population reported the greatest amount of delinquent behavior, whereas the 16 and 17 year olds reported the least amount of acting out behavior.
One reason for these findings may be due to the
increasing normalization of deviant behaviors as the individual ages and is involved in a greater number of crimes.
A second factor may be a function of
institutionalization due to repeated arrests and incarcerations. This process desensitizes them to delinquent nature of their behaviors.
During the assessment
process, it was noted by this examiner that the younger members of this sample presented as more aware of the criminalistic nature of their behaviors, as well as the wrongfulness of their crimes.
The oldest members (18 year olds) of this
sample did present with the second highest average of scores, which is likely due to recommitment as a result of repeated offenses. Analysis of the representation of age in the Category groupings support this statement as well.
serious classification, the Violent A Offender category, consisted of only four juveniles, ages 15, 16, 17, and 18. Additionally, no 13 or 14 year olds were represented in the Firearms Offender, Controlled Substance Dealer, Chronic
68 Serious Offender, or Violent A Offender categories.
The majority of the 18 year
olds (60%) were represented in the two most serious categories (VOB and VOA). These findings display a stronger upward trend in level of category assigned (and therefore, severity of crimes) with the age of the offender. This indicates that the category assigned by TYC, based partly on the offenders' past criminal record, appears to be a better predictor of type of acting out behavior in correlation with age.
As the PIY Delinquency scale scores are based on self-
report, it appears that the juveniles' self-perceptions of their crimes are not as accurate as the perceptions of the mental health and detention professionals who work with them. Comparison of the PIY and PAS Not surprisingly, stronger convergent validity was found between the PIY and PAS, which utilize the same methodology: scaled responses to standardized questions designed to measure general constructs and behaviors, as opposed to diagnostic categories.
Between these two highly similar assessment methods,
validation was established.
The trait variance was higher than the method
variance; convergent validity values are in general higher than discriminant values in both the heteromethod and monomethod blocks. The modest convergent validity of the CAS to these measures can be partially attributed to the methodology of the CAS which utilizes standardized questions to determine diagnostic categories.
These differences become apparent
in Table 9, which shows the low convergent correlations between the CAS and
69 the other two measures.
Additionally, the CAS scores measure different subtypes
of a diagnosis (e.g., Conduct disorder, group type), whereas the scale scores from the PIY and PAS measure overall constructs, such as delinquency or depression. Thus, the PIY and PAS appear to be measuring more similar constructs than the CAS.
Additionally, the PIY appears to be a useful measure of those constructs
assessed, particularly delinquency, psychological distress, and somatic concerns, when used with a delinquent population. A promising finding was the potential of the PAS as an assessment instrument for the juvenile offender population.
The development of the PAS
was designed to specifically address populations similar to this sample.
content includes extensive questions addressing gang violence, physical aggression, drug and alcohol use, criminal activity, and severe family dysfunction. Due to the newness of the PAS, and its lack of external validation studies, further research is warranted.
The convergent validity (see Table 9) of the PAS is
moderate (r = .33), however, it demonstrated poor discriminant validity (r = .43) using heterotrait-monomethod comparisons.
This finding may be due to either
item overlap or the homogeneity of common behaviors found in delinquency. Those scales which are described as measuring different constructs were overall more highly correlated than those scales described as measuring similar constructs.
With removal of those validity coefficients that are predicted to be
highly correlated (gang involvement, violence, and aggression), the within measure discriminant validity improved slightly (r = .37). It did demonstrate good
70 between measures discriminant validity (r = .18) in comparison with the PIY and CAS. Concerns Regarding the PIY The PIY was lengthy in administration, due to its 270 items.
scoring was very difficult, using the templates which come attached underneath the answer sheet and utilize a number of carbons, which are easily smudged and unable to be clearly corrected.
Also, the time required to score takes between
20 to 30 minutes per test, which makes it impractical for testing large populations, as is found in most agencies dealing with juvenile delinquents.
in a clinical practice with a smaller caseload would be more practical. The reading level appears higher than that stated in the manual.
The questions are
written to allow for the understanding by children with a language comprehension of a typical 9-year-old. However, words like "constipated," "seizures," and "scolding" were unknown many the juvenile offenders who are adolescents. Clearly, a restriction of utilizing the PIY with a more severe delinquent population is the lack of questions on the PIY which are directed to highly troublesome behaviors of juvenile delinquents. In general, a significant proportion of incarcerated juveniles have been involved in felonies. The majority of the questions on the Delinquency scale on the PIY address the less severe behavior (e.g., status offenses) more commonly found in juveniles labeled with behavioral problems, rather than juveniles who have been incarcerated.
As stated by Lachar
and Gruber (1995), the PIY items are written to specifically address those
71 interests and issues of young people through adolescence.
Therefore, the PIY
questions do not adequately address, both in terms of qualitative and quantitative measurement, the more severe issues of juvenile delinquency, such as gang involvement, substance abuse, and physical aggression. Limitations of the Current Research A limitation of the current research is the exclusive use of males.
failure in endorsement on the Delinquency Scale on the PIY, in addition to being a lack of items which tap into more severe acting out behavior, may also be a gender-related issue.
Research (Sullivan & Wilson, 1995) has shown that female
delinquents tend to display less physical, and more verbal and emotional acting out behavior, such as status offenses. Therefore, females may endorse those items on the Delinquency Scale that the male population from this sample failed to endorse. Based on the results of the current study, future research utilizing the PIY, along with other measures, with a female juvenile delinquent population may provide information to support the gender differences in delinquency. A second limitation is the purpose of the two different types of assessment instruments utilized in this research.
Two of the instruments, the PIY
and PAS, are both true/false, self-report measures, designed to describe behavior rather than diagnose or classify symptoms. They provide a standardized measure, but do not allow for additional inquiry or clarification of answers.
instrument, the CAS, is a semi-structured interview, which allows for probes and
72 clarification of answers, but is less descriptive of general behaviors outside of the diagnostic categories. Thus, developing a measure to describe the behavior of a person diagnosed as conduct disordered involves a different conceptual approach then diagnosing the conduct disorder.
The use of structured interviews tend to be
more comprehensive and less likely to miss diagnoses than other assessment methods, but they are not as practical for providing descriptions of personality functioning or general behavioral constructs.
The scoring method, based upon
absolute diagnostic categories, and type of information (diagnosis versus test items) gleaned from the CAS inherently complicates the methodology of determining convergent and discriminant validity.
The CAS scores appear to
represent a nominal scale, whereas the PIY and PAS are both interval scales. Comparisons of mixed measures can weaken the validity coefficients outlined in the multi-trait/multi-method matrix.
Conversely, "robust" constructs can be
measured by dissimilar methods. A third limitation of this study was that the population sample was adjudicated delinquents, which would limit the scoring range to having a much larger proportion of high scoring individuals. This delinquent population was adjudicated by state courts, and assessed and assigned to maximum security care level 5 by psychologists at the state reception center.
adolescents were not included in this sample, the correlations and validity coefficients would likely be lower because of the restricted range of responses.
73 Another limitation of this study was the time interval between the administration of tests for each participant. as close in time as possible.
Ideally, test administration should be
However, due to the difficulties of scheduling
participants, the interval between testing sessions ranged between two to four weeks.
During this time period some changes in behavior or symptomatology
may have occurred in the participants. The gender of the interviewer in relation to the interviewee may have also posed a limitation of this study. Conflict and acting out behavior was more likely to occur when the interviewer was a male, based upon personal experiences of the male and female interviewers.
This observation became
apparent during discussions between the interviewers, which occurred at the end of each testing period. Whether this conflict was a function of the personality dynamics of the parties involved, or was more simply a function of the gender of the interviewer was not determined. Directions for Future Research The most important issue for future research focuses on the identification of the most troublesome behavior.
As clinicians, we are constantly asked to
make assessments of risk behavior, and whether an individual is a danger to self or others.
Mental health professionals, in the past, have often not been involved
in research regarding juvenile offenders due to the common misconception that criminal behavior and violent acting out were not pathological, therefore not clinical issues (Holliman, 1995).
Additionally, as Borum (1996) pointed out, few
74 investigators have focused on developing instruments which would provide validity for making clinical decisions regarding individuals' risk for violent behavior.
He specifically stated that no explicit national psychological or mental
health standards exist for the assessment of violence risk. Instruments are needed that effectively measure external expressions (aggressive or violent acting out) of mental disorders.
instruments make little attempt to ask the necessary questions about actual behaviors which juvenile delinquents would endorse, and thus are unable to accurately classify the realistic level of aggression and violence found in society. Current demographic information suggests that the age of juvenile delinquents is decreasing and younger juveniles are committing more serious offenses (Kernberg & Chazen, 1991), however, the progression of age with increasing severity of the offense remains stable.
As suggested by Holcomb and Kashani (1991), a better
comprehension of the conduct disordered juvenile may assist in better classification and description of this heterogeneous group.
REFERENCES American Ps y c h i a t ri c Association m m # of
(3rd e d }
American Psychiatric Association
American Psychiatric Association
,. Washington DC:
(1987). Oamastic and ^
manual of mental disorders (3 r[ j e c j r p ^ & p r a ed., rev.).
Macmillan Publishing Company. ArCher
Lawrence Erlbaum Associates, Inc.
Archer, R. p., & Jacobson, J. M 'he
v enseal items "critical" for 54? 556
Family and peer socia,
support as specific c o l l a t e s of adoiescent depressive symptom, Abnormal Child Pgycholopv ?n(i) j _ | 6 Bellack, A. S., & Hersen, M. mchQlQgX.
y g , E. R.
p0 u ~ e s e a r c h methrvk ; n r i j n i y a |
d u s ( e r a n a l y s j s Qf B a s . c
(BPD adolescent profiles.
persona]jtv jnventory 265-272.
76 Borum, R. assessment.
Improving the clinical practice of violence risk
American Psychologist. 51 945-956.
Brent, D. A., Kalas, R., Edelbrock, C„ Costello, A. J., Dulcan, M. K„ & Conover, N.
Psychopathology and its relationship to suicidal ideation in
childhood and adolescence.
Journal of the American
A c a d e m y r»f rw,\A
Psychiatry. 95 666-673. Butcher, J. N., Williams, C. L., Graham, J. R., Archer, R. P., Tellegen, A., Ben-Porath, T. S., & Kaemmer, B. Personality Inventory - Adolescent.
(1992). Minnesota MnlHpWi.
Minneapolis: University of Minnesota.
Campbell, D. T., & Fiske, D. W.
validation by the multitrait-multimethod matrix.
Convergent and discriminant
105. Cansler, D. Inventory.
Review of the Millon Adolescent Personality
Journal of Counseling and Development,
Dumas, J. E. (1989). family approaches.
Treating antisocial behavior in children:
Clinical Psychology Review q 197.222
Edelbrock, D., Costello, A. M., Dulcan, M. K., Conover, N. C., & Kala, R.
Parent-child agreement on child psychiatric symptoms assessed via
Journal of Child Psychology and P . v r h i ^ , 0-7
Ehrenberg, M. F., Cox, D. N., & Koopman, R. F.
Adolescent Personality Inventory profiles of depressed adolescents. 25(98), 415-424.
The Millon Adolesce
77 Fiske, D. W., & Campbell, D. T. problems.
Citations do not solve
Psychological Bulletin, 119/3^ 393-395.
Golden, C. J., Sawicki, R. F., & Franzen, M. D. construction. assessment.
In G. Goldstein, & M. Hersen's (Eds.), Handbook of psychological New York: Pergamon Press.
Holcomb, W. R., & Kashani, J. H.
Personality characteristics of
a community sample of adolescents with conduct disorders.
579-586. Hart, L. R.
Diagnosis of disruptive behavior disorders using the
Millon Adolescent Personality Inventory. Hodges, K.
Psychological Reports- 73 895-914.
Structured interviews for assessing children.
of Child Psychology and Psychiatry and Allied Disciplines 3 4 m 49-68. Hodges, K., Cools, J., & McKnew, D.
Test-retest reliability of a
clinical research interview for children: The Child Assessment Schedule. Psychological Assessment \
(1990). Parent-child agreement
on symptoms assessed via a clinical research interview for children: The Child Assessment Schedule (CAS). Allied Disciplines. 31 m
Journal of Child Psychology and Psvchiatrv and
Hodges, K., Kline, J., Stern, K„ Cytryn, K., & McKnew, D.
The development of a child assessment interview for research and clinical use. Journal of Abnormal Child Psychology. 10r,>'t 173-189.
78 Hodges, K., McKnew, D., Burbach, D.J., & Roebuck, L.
Diagnostic concordance between the Child Assessment Schedule (CAS) and the Schedule for Affective Disorders and Schizophrenia for School-Age Children (KSADS) in an outpatient sample using lay interviewers.
Journal of the American
Academy of Child and Adolescent Psychiatry. 26 654-661. Hodges, K., McKnew, D„ Cytryn, L., Stern, L., & Kline, J. Child Assessment Schedule (CAS) diagnostic interview: and validity.
A report on reliability
Journal of the American Academy of Child Psychiatry. 21 468-
473. Hodges, K., & Plow, J.
Intellectual ability and achievement in
psychiatrically hospitalized children with conduct, anxiety and affective disorders. Journal of Consulting and Clinical PsychologyHodges, K., & Saunders, W.B.
589-595. Internal consistency of a
diagnostic interview for children: The Child Assessment Schedule.
Abnormal Child Psychology 1 7 ^ 691-701. Holcomb, W. R., & Kashani, J. H.
Personality characteristics of
a community sample of adolescents with conduct disorders.
26(103), 579-586. Holliman, N.
(1995). [Personal Attitude Scale],
Unpublished raw data.
Johnson, C., Archer, R. P., Sheaffer, C. I., & Miller, D.
Relationships between the MAPI and the MMPI in the assessment of adolescent psychopathology.
Journal of Personality Assessment. W2^ 277-286.
79 Kashani, J. H., Deuser, W., & Reid, J. C. anxiety: A new look at an old notion.
Journal of the American Academy of
Child and Adolescent Psychiatry. 30(T). 218-223. Kashani, J. H., Reid, J., & Rosenberg, T.
(1989). Levels of hopelessness
in children and adolescents: A developmental perspective. Journal of Consulting and Clinical Psychology. 57. 496-499. Kazdin, A. E.
Treatment of antisocial behavior in children:
Current status and future directions. Kazdin, A, E. treatment programs.
Adolescent mental health:
American Psychologist. 48 127-141.
Kernberg, P., & Chazan, S. New York:
Psychological Bulletin in? 187-203.
Children with conduct disorders.
Kovacs, M., Paulauskas, S., Gatsonis, C., & Richards, C. Depressive disorders in childhood: risk for conduct disorders. Lachar, D.
A longitudinal study of comorbidity with and
Journal of Affective Disorders. 15m 205-217. Symptom checklists and personality inventories. In
T. R. Kratochwill, & R. J. Morris (Eds.), Handbook of psychotherapy for children and adolescents (pp. 38-58). New York: Allyn & Bacon. Lachar, D., & Gruber, C. P. Inventory for Youth: Children.
Development of the Personality
A self-report companion to the Personality Inventory for
Journal of Personality Assessment. 61H V 81-98.
80 Lachar, D., & Kline, R. B. and Personality Inventory for Youth.
Personality Inventory for Children
In M. Maruish (Ed.), The use nf
psycholopiral testing for treatment phnning and ontrnm. 516).
Levin, J., & Montag, I.
Relationship between the Basic
Personality Inventory and the NEO-Personality Inventoty in a nonpatient sample. Psychological Reports, ftQ 1176-1178.
Maguire, K., & Flanagan, T. J. (Eds.) iaake.
Washington, DC: U.S. Department of Justice, Bureau of Justice
Statistics. Marsh, H. W.
Confirmatory factor analysis of multitrait-
multimethod data: The construct validation of multidimensional self-concept responses.
Journal of Personality sg/vp 661-692.
Nottelman, E. D.
from childhood to adolescence.
Competence and self-esteem during transition Developments! P , . , . V , „ n . „
Pantle, M. L„ Evert, J. M„ & Treneny, M. R. the MAPI in the Assessment of depression.
The Utility of ^
55, (3 & 4), 673-682. Pantle, M. L., Barger, C. G„ Hamilton, M. C., Thornton, S. S., & Piersma, H. L. treatment.
Persistent MAPI scale 6 elevations after inpatient
JournaL_0f Personality Acc e s s m p n f
81 Piersma, H. L„ Pantle, M. L„ Smith, A., Boes, J., & Kubiak, J. The MAPI as a treatment outcome measure for adolescent inpatients.
Clinical Psycholopy 40^1 709-714. Reich, W„ Herjanic, B„ Welner, A., & Gandhy, P. R,
Development of a structured psychiatric interview for children: diagnosis comparing child and parent interviews.
Journal of Ah
Psychology in 325-336. Reidy, T. J., & Carstens, C.
Stability of the Millon Adolcscent
Personality Inventory in an incarcerated delinquent population. Personality Assessment i Rogers, R. for p s v c h o l o f i ^ .
Jo,,m a l of
Diagnostic and striirtiired intervi^winp- A
Psychological Assessment Resources.
Rohde, P., Lewinsohn, P. M„ & Seeley, J. R.
unipolar depression: Comorbidity with other mental disorders in adolescents and adults.
Journal of Abnormal
Sullivan, R„ & Wilson, M. F.
New directions for research in
prevention and treatment of delinquency: A review and proposal.
A H o l ^
30(1171. 1-16. Texas Youth Commission (TYC). Operating Policy Manual.
Austin, TX: Author.
Thompson, R. J., Hodges, K„ & Hamlett, K. W.
( 1 990).
comparison of adjustment in children with cystic fibrosis and psychiatrically
82 referred and nonreferred children. Journal of Pediatric PsvrhnWy
759. Thompson, R. J., Merritt, K. A., Keith, B. R„ Murphy, L. B., & Johndrow, D. A.
Mother-child agreement on the Child Assessment
Schedule with nonreferred children:
A research note.
Psychology and Psychiatry and Allied Discipline
Tolan, P., Ryan, K., & Jaffe, C.
Journal of rhild 813-820.
Adolescents' mental health
service use and provider, process, and recipient characteristics.
Clinical Child Psychology. 17 229-236. Treneny, M. R. & Pantle, M. L. inpatient crisis-unit sample.
MAPI code types in an
Journal of Personality A«~.nv»..
4)> 6 8 3 .
691. Verhulse, F. C., Althaus, M., & Berden, G. F. M. G.
Assessment Schedule: Parent-child agreement and validity measures. Child Psychology and Psychiatry and Allien Discipline V°gt,
(1993). Dictionary of statistics and methndnWy A
nontechnical pnirle for the social sciences. Newbury Park, CA:
Publications, Inc. Walsh, W. B„ & Bee. N. E. Jersey:
83 Williams, C. L., Ben-Porath, Y.S., Uchiyama, C. & Weed, N.C.
External validity of the new Devereux Adolescent Behavior Rating Scales. Journal of Personality Assessment. W 1 & 2), 73-85. Wrightsman, L. S„ Nietzel, M. T, & Fortune, W. H. Psvcholofv ami the lrfnl
California: Brooks/Cole Publishing Company.