ADOLESCENTS AT RISK
CHARLTON J. COLES
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
I would like to acknowledge the assistance of my doctoral committee,
Drs. Eileen Fennell. Duane Dede, Gary Geffken, Faye Gary, Anthony Greene.
and David Miller. Acknowledgements are due to Lennie Weiss, Carrie
Roseberrv. Mr and Mrs Webb. John Rickicki. and Dr. Anderson, for assisting
me with sample information. I would like to thank my family, especially my
mother. Nlarian Pitts-Coles. my father. Cleo P. Coles, my sister. Cynthia
Robinson. and my brother, Christopher Coles, for their support throughout
this project and my academic career. I would like to dedicate this project
to my grandmother. Marian Pitts. Acknowledgements are also due to Charles
Isbell. L\ nn Sims. Walter Jacobs, Kerry-Ann Anderson. Samaria Jones. Sonja
Jonas, Cleo and Jane Coles. Yvonne Collins. and Charlene Armstrong for their
continuing support of me and my studies. I would like to thank my paternal
grandparents. Cleo and Louise Coles.
TABLE OF CONTENTS
ACKNOW LEDGM ENTS ......... ................................ .. iii
ABSTRACT ..................................................... v
CHAPTERS .............. ............................. .......... I
1. INTRODUCTION ........ .......... ......................... 1
Personality Characteristics of Adolescents At Risk ................. 13
Family Structure and Violence Among Adolescents ................ 20
Anger ........... ................................... 25
2. METHODS .............................................. 34
Subjects ........... .................................. 34
M measures and Procedures ............................. 34
MMPI-A .......................................... 36
STAXI ................................. ......... 40
STAI ......................................... 42
FES ............................................. 43
Hypotheses .......... ................................. 44
Statistical A nalyses ................................. 45
3. RESULTS ........ ........................................ 47
4. DISCUSSION ........................................... 68
APPENDIX ............ ............................... .......... 85
REFERENCES ............................................ ....... 86
BIOGRAPHICAL SKETCH ................ ......................... 92
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctorate of Philosophy
ADOLESCENTS AT RISK
Charlton J. Coles
Chairperson: Eileen B. Fennell
Major Department: Clinical and Health Psychology
The current literature on adolescents suggest that a multitude of
psychological and social factors. including personality, environment and
familial, can interact to place these adolescents at risk for aggression and
violence. It was hoped that this study could provide a heterogeneous
representation of adolescents for violence in terms of psychosocial
factors. The subjects in this study were 119 male adolescents who attended
schools targeted for children with behavioral difficulties. These
adolescents were given four self-report measures which examined
personality, affective and familial variables. A hierarchical,
agglomerative cluster analysis was conducted to see if adolescents at risk
for aggression presented a heterogeneous picture in terms of personality
characteristics. Results of the cluster analysis suggested three clusters
of adolescents at risk. Examination of the three cluster types by anger
expression. family control, trait anxiety and trait anger found
distinctions between groups for anger expression, trait anger and trait
anxiety. This suggests a possible relationship between personality
presentation and affective responding. Further analyses suggested a link
between anger control and the least elevated personality type. A series of
linear regression equations found relationships between high levels of
angry acting-out and the family variables of cohesion and conflict.
Adolescents at risk is a term used to describe sub-populations of
adolescents who are disadvantaged economically and/ or socially. As a
result, these adolescents are at risk for a number of developmental problems
including: juvenile delinquency, teenage pregnancy. sexually transmitted
diseases (STDs), substance abuse, and illiteracy. Current definitions of
adolescents at risk are usually fairly broad. Simply stated, adolescents at
risk are those juveniles who are at risk for poor developmental outcome.
Additionally. risk behavior can encompass a wide number of problems
including illicit substance experimentation/ substance abuse, sexual
promiscuity. and excessive violence. Risk behavior can have both short- and
long- term consequences. The factors that may predispose a youth to these
behaviors are complex and inter-related (Dryfoos. 1990: Ginzberg et al..
1988: Jessor and Jessor. 1977). Social factors may include parents, peers.
societal attitudes and media influences (Hawkins and Weiss, 1985: Dryfoos,
1990). Psychological influences include personality, affect, and
individual attitudes and beliefs (Quay, 1987: Dryfoos. 1990). Biological
factors may include psychiatric, neurological and non-neurological medical
factors. Finally, economic factors include unemployment and living in an
economically impoverished area and/ or a high crime area (Baldus and Tribe.
1978; Siegel. 1981). These at risk adolescents are likely to continue their
inappropriate behavior into adulthood and place additional burdens upon
society. It is difficult for a social science to separate the above
mentioned factors, determine which have the greatest impact on a developing
youth, and produce sufficient programs or interventions to sufficiently
impact these youth. Many theories have been generated to explain the high
prevalence of at risk adolescents in our society. Additionally. many
intervention programs have been developed over the years to impact these
behaviors. However, there exists a need for a larger empirical base about
these behaviors, the factors influencing them and what precipitates at risk
behaviors in order to properly serve this population (Dryfoos. 1990).
The present study was proposed in order to examine factors that may
predispose adolescents towards violent behavior, particularly expressive
violence. Expressive violence is violence that is predicated
interpersonally between individuals or groups. The youths who are prone to
expressive violence include juvenile delinquents, individuals who have
extensive histories of violence in settings such as school or the community.
but have not been officially arrested, and those who live in high crime/
economically impoverished areas. This study explores those factors that
may predispose youth to violence within the psychosocial domain:
personality, affect, and familial elements. This study utilized a more
heterogeneous representation of adolescents at risk in such a way as to
provide for more accurate assessment and possibly better intervention
measures for these adolescents.
The purpose of this study was to examine psychosociological factors
that may predispose male adolescents to violence. These factors include
personality, affect, and familial elements. By studying a more
heterogeneous group of adolescents at risk, this study was designed to
provide support for those factors in predicting violence in youths at risk.
If personality structure, affective symptoms and family cohesion do predict
increased risk. it may provide an empirical basis for more effective
intervention strategies. This study examined only male adolescents due the
relatively low frequency of female adolescents who are referred to agencies
such as alternative schools or detention centers compared to males.
However, future research may suggest a gender comparison on a number of
different psychosocial variables, including personality, affect and
family, for adolescents at risk for aggression.
Historically, psychological theories concerning adolescents at risk
were uni-modal in that these adolescents were viewed as a homogeneous group
with the same behavioral presentation (Vedder, 1954; McCann. 1957).
Traditionally, they were viewed as sociopathic with many of the traits
associated with adult sociopaths. Subsequent adolescent personality
theory, such as the adolescent personality theory by Hathaway and Monachesi
(1963), have suggested differential personality characteristics. However,
even today many institutions and individuals still use a uni-modal system to
intervene with adolescents at risk. A common example is the modern-day
school system. Children and adolescents with disruptive behaviors are
commonly classified as "emotionally handicapped (EH)," even though they
have significantly distinct behavioral presentations. personality factors
and familial components.
This study proposed that adolescents at risk are a heterogeneous
group in terms of personality structure. This personality structure is
important because it affects how a child views and interacts with the
outside world. This interaction with the outside world can be expressed
through emotional symptomatology which can be reflected in a child striking
out against the world externally through anger. or conducting harmful.
self-injurious behaviors internally through anxiety. Both of these
behaviors are completely distinct behavioral presentations. but under a
traditional uni-modal system they can simply be labeled as inappropriate
behavioral presentations. This is hypothesized to impair intervention
because no one specific form of symptomatology is targeted. It is further
hypothesized that there are sub-groups of adolescents at risk which have
significant anger and anxiety components occurring simultaneously. In
addition. it is proposed that the personality and behavioral presentations
of adolescents at risk are further affected by familial factors,
particularly those related to issues of attachment and warmth. It is
believed that an inverse relationship exists between low levels of family
cohesion and high levels of affective symptomatology and highly unstable
The Occurrence of Delinquency
One sub-population of adolescents at risk for violence includes
juvenile delinquents. Juvenile delinquency is defined by a finding of
violation of legal statutes relating to criminal codes, school (truancy)
and other behaviors (eg. runaway, sexual promiscuity. etc.). In many ways.
juvenile delinquency is less complicated to define because it is shaped by
legal statutes and official facts and figures compiled from a variety of
different agencies. One such agency is the United States Children's Bureau.
Since 1927. the U.S. Children's Bureau has circulated estimates on
the number of juvenile offenses committed within the United States. Reports
out of the Children's Bureau have indicated the number of crimes committed
by juveniles has shown a substantial increase every year since the Bureau
began tabulating juvenile offenses (Roberts, 1987, pg. 9).
The U.S. Children's Bureau in the 1930s estimated 200.000 juvenile
offenders in the United States. In 1950, the Bureau estimated the number to
be 435,000 juveniles. and by 1966 the estimate ballooned to over 1,000.000
delinquents arrested in the United States. During the last two decades, the
Federal Office of Youth Development has reported a significant increase in
the number of juvenile offenders. By 1984. the number of arrests for
adolescents less than the age of 18 had exceeded 2.000,000 (actual figure:
2,062.448). Consequently, this growth has significantly exceeded the
growth of the population of children in the 10- to 17-year-old age group.
According to the 1984 data. the delinquent most likely to be arrested was
between the ages of 15 and 18 and the most common offense was larceny/theft
(Roberts. 1987. pg. 10). Roberts (1987) indicated that the majority of the
crimes by juveniles were property-related, although there was a trend
towards violent crimes as the youths became older.
In terms of ethnicity. African-American males are more likely to be
arrested than any other group. Hindelang (198 1) analyzed information that
was collected in the American National Crime Survey. He compiled the data
from crimes against persons including rape. robbery. assault, and theft to
compute the offense rates for different population subgroups. The
estimates indicated that for juveniles aged 12-17. 2.1% of these offenses
were committed by white females, 8.6% by African-American females. 8.0% by
white males, and 43.2% by African-American males. There are various
theories which have been proposed to rationalize the differences in
delinquency rates for African-Americans and whites. The majority of these
theories suggest that ethnicity is not an important causal factor (Quay.
1987) but that. instead. African-Americans and whites differ on a number of
precursors to delinquency. These precursors may include low family income
and poor parental child-rearing practices. It has been argued that because
of the high proportion of black single-parent households, there is less
parental control and supervision in African-American families (Quay.
For African-American males, black-on-black homicide is the leading
cause of death for males between the ages of 15-34. The chances that an
African-American male will be murdered is roughly ten times that of whites
males. African-American females are five times as likely to be murdered as
white females (Quay. 1987).
These statistics may only represent the "tip of the iceberg" in terms
of actual reporting. Not all delinquent behavior is detected and the acts
which are officially recorded do not represent a random sample of juvenile
acts. Official statistics only provide a limited index of juvenile
delinquents. However, they may be useful in indicating trends or suggesting
trends for research and/ or intervention.
Public Health Responses to Adolescents At Risk For Violence
The conflicts of violent youth have received substantial attention
from the public via mass media. In response to public opinion that
interventions are needed for violence within our youth, the federal
government has responded with a number of programs targeting the youth. An
example of government intervention includes a response by the Centers for
In 1991, the Centers for Disease Control (CDC) defined violence among
adolescents as a public health dilemma. This agency worked on establishing
guidelines to help prevent the occurrence and the overall rate of violence
among teenagers. especially those of a minority population. ie. African-
Americans. Hispanics, and Native Americans. The CDC (1991) defined several
underlying assumptions about violence among youth in an effort to help
combat the problem. One of these assumptions was that violence is a learned
behavior that can be changed and prevented. Another assumption by the CDC
was that there is no simple or single solution to the problem of youth
violence. In the third assumption, the CDC indicated that a framework for
violence intervention exists. Within this framework, a community-based
program must have the following characteristics: 1) there must be
coordinated responses for many community organizations and groups: 2)
activities and interventions should be targeted at specific risk groups: 3)
community groups should provide societal support for individual behavioral
change via advocacy and policy development, and 4) it is necessary to apply
multiple methods in multiple settings (CDC, 1991).
The fourth assumption of the CDC centered around the idea that each
active participant in the intervention process be made held accountable for
whom and to whom activities are being directed. The CDC noted that citizens
should be given their own authorization to identify and choose their
priorities, activities, and retain control over what happens in their
communities, a process known as "empowerment." Empowerment is needed in
many of these communities due to the fact that the sense of powerlessness and
lack of control contributes to violence. Clear, conscientious efforts
should be attempted to return community control back to these communities.
Community participation, responsibility and revitalization were viewed as
important ingredients to empowerment (CDC. 1991).
The final assumption of the CDC was that meaningful change in a
complex. multisystemic problem takes a lengthy period of time. The final
measures of success in a program, such as decreases in mortality, injury.
and disability, will not be immediate, and an enduring commitment to provide
resources is critical to achieving these ends (CDC. 1991).
The CDC looked at targeting violence prevention strategies to
populations of adolescents at high-risk for violence. The factors that they
examined as contributing to the high rate of interpersonal violence among
adolescents at risk included unemployment, poverty, low educational
opportunities and achievement, drug/ alcohol abuse, and weapon carrying
The CDC (1991) identified five groups of high-risk teenagers
considered to be at highest priority. The first of these groups were those
that live in a geographically defined area in which rates of violent deaths.
injury and disability are extremely high. Children growing up in these
areas may be exposed to high rates of violence through such activities as
drug dealing and/ or gang activity. Children growing up in these areas may
be likely to model their self-defense and conflict-resolution skills on
those violent behaviors that they are exposed to on a routine basis.
The second of these groups were those youths who have been targeted as
belonging to a gang or youths that have been targeted as at risk for becoming
gang members. Studies have shown that the rate of violent offenses for gang
members is three times that of adolescents who are not gang members (CDC.
1991). It may be difficult to distinguish those youths who belong to a gang
from those who do not. Gang membership may span several generations within a
family or extended family. Children and adolescents may seek the
companionship, monetary rewards, and sense of belongingness that may be
associated with being in a gang (CDC, 1991).
The third group included those youths who are members of families that
have problems related to violence. The factors associated with this include
unsupervised children, parental drug abuse, child abuse, and the neglect or
absence of a parental role model. Studies (Loeber and Stouthamer-Loeber.
1986: Palmo and Palmo. 1993) have shown that children from violent families
and children with parents who show little affection tend to develop violent
behavior problems (CDC. 1991).
The fourth group consisted of those violent youths who have histories
of extreme violence and who have entered the court system due to their
violent behavior. This group also includes adolescents in the penal system.
Potential strategies to address this group would include skill training in
such areas as conflict resolution and anger management, peer counseling and
mentorships with respected members of the community, and group activities
to help bolster self-esteem. trust, and group support. The final group
includes those who were victims of violence, relatives of violence victims.
and witnesses to violence. It may be difficult to distinguish perpetrators
from victims in a high-risk violent situation.
The recommendations and effort to define groups of adolescents prone
to violence by the CDC are generally positive. However, there are a few
criticisms. One criticism of the CDC's guidelines is that the criteria for
selecting adolescents into groups may be too rigid and narrow. The CDC has
defined groups based on strict environmental factors and failed to examine
others factors which could account for violence. One such factor would be
psychological. The DSM-IV contains many diagnoses to account for acting-
out behavior (ADHD, Intermittent Explosive Disorder, Conduct Disorder,
etc.) and. when coupled with a harsh environment with few resources, could
make a child significantly at risk for aggression. In addition, there are
medical conditions to account for adolescent aggression such as
neurological injuries, genetic and endocrinological irregularities.
Another difficulty with defining groups based on environmental
factors is that one assumes a uni-modal theoretical orientation in which
children in the same category receive the same treatment. It never tests the
theory that within a category adolescents may have different reasons for
acting-out or being aggressive. Also, these adolescents may have different
behavioral or emotional presentations. For instance, one child with
violent tendencies may be violent without the slightest provocation while
another child has rigid controls over his aggression which break down on a
few occasions. However, when this child's control breaks down on those rare
occasions the result is extremely violent. If the theory of heterogeneity
within categories is true. would these two children benefit from the same
A third criticism is that the CDC defines groups by external criteria.
ie. by the number of violent incidents and histories of court referrals and
demographic information, such as community orientation and high-risk crime
areas. While such information is important. additional information which
could help target the population further may be missed. Such information
could include knowledge about internal characteristics of these
adolescents, such as the personality and affective states of adolescents at
risk. Moreover, we need to know if there is a relationship between affective
reactivity, personality and violent tendencies. Investigations should
examine the effect of being in these highly stressful environments on a
child's inter- and intrapersonal development. Specifically, we need to
know the functional role of affective reactivity within this population. If
emotional reactivity does serve a functional role, can it be altered?
Lastly, we need to begin to investigate the long-term costs of being
effectively reactive in a child's given environment. These are the
beginning of formulations which have arisen as a result of going through the
In order to fully service the need for intervention, it is vitally
important to have as much comprehensive information about the population to
be treated. This study examines the problem of adolescents at risk by
addressing areas such as personality variables and their interaction with
other variables such as family structure, substance abuse history, and
affective states,. such as anger and depression. This study tests an
hypothesis that adolescents at risk have differing psychosocial
characteristics and behavioral presentations. If this hypothesis is true.
then it may lead to more effective behavioral interventions. Additionally.
it would contribute to a comprehensive picture of these adolescents and how
they interact \\ ith their environment from a number of different factors.
Personality Characteristics Of Adolescents At Risk For Violence
This section will address the literature as to the most appropriate
personality measures for adolescence. The problems confronting
adolescents are generally less well known than those facing adults (Kazdin,
1992). Assessing the behavior and emotions of adolescents is a difficult
chore, and a complicating factor is that for a long time there was
significantly less research with adolescence compared to adults.
The Minnesota Multiphasic Personality Inventory (MMPI) is a well-
known personality measure that has been extensively studied with adult
populations, although it has also been used on adolescents as well. In
general, the research concerning adolescent personality has been extremely
limited and difficulty has arisen over a long number of years. One of the
first research studies to study the MMPI responses reported by adolescents
was a classic study by Hathaway and Monachesi in 1963.
Hathaway and Monachesi (1963) collected MMPIs from approximately
15,000 adolescents in the Minneapolis, Minnesota, area between the years
1945- 1963. In addition to the personality information, they collected
ratings on school adjustment. school conduct. and delinquency. Hathaway
and Monachesi stated that approximately 28.4 percent of the male sample was
delinquent compared to 9 percent of the female sample. Hathaway and
Monachesi placed delinquents on an inhibitory-excitatory continuum. They
concluded that Scales F. 4, 8, and 9 of the MMPI were excitatory scales and
thus successfully predicted high rates of delinquency. Conversely,
Hathaway and Monachesi reported that Scales 2, 3. 7. 0, and 5 were inhibitory
scales and were predictive of lower rates of delinquency. They stated that
the inhibitory scales were predictive of low rates of delinquency due to the
fact that they were "neurotic" scales and individuals who scored high on
these scales had low inclination to conduct delinquent acts. However, they
indicated that when inhibitory scales interacted with excitatory scales.
the excitatory scales served to have a stronger influence. An example is an
adolescent with a two-point code type of "42" or "24". Hathaway and
Monachesi (1963) predict that the constraining effects of depression would
be over-ruled by the stimulatory effects of reported sociopathy. Huesmann,
Lefkowitz. and Efron (1978) commented on Hathaway and Monachesi's (1963)
study and reported that the best predictors of adolescent aggression and
delinquent behavior were Scales F. 4, and 9.
Williams and Butcher (1989) made the argument that age-appropriate
norms may be needed for adolescents who are administered the MMPI.
Frequently. however, adolescent norms are problematic in adolescent age
groups. The major problem with this is that on the MMPI, the adolescent norm
set fails to reach clinical significance (T scores greater than 70).
Subsequently. this led to the suggestion that an adolescent cut-off score of
65, rather than 70 be used to determine clinical significance (Williams et
To address the problems of the MMPI with an adolescent sample,
Williams and Butcher (1989a and b) initiated a two-study project designed to
develop a personality inventory for use by adolescents. The first study was
an attempt to determine the empirical validity of the standard MMPI in a
large-scale sample of adolescents in treatment before the instrument was
changed. The second study was an attempt to determine if appropriate.
alternative code type classification procedures would lead to meaningful
descriptors based on adolescent responses.
The first study was conducted with 844 adolescents who ranged in age
from 12 to 18 years of age. This sample consisted of adolescents who were
admitted to several treatment facilities between 1985 and 1988. In general,
the authors stated that the MMPI clinical scales were associated with
clinically relevant behavior in adolescents at risk. They found that Scale
4, 8, and 9 were strongly associated with measures of acting out for both
genders. The authors concluded that Scale 6 may be an excitatory scale for
The Hathaway and Monachesi (1963) and Williams and Butcher (1989)
found similar research findings. For example, adolescents scoring high on
the excitatory continuum. "4. 8, 9." appear to display substantial evidence
for acting- out behaviors. However, the Williams and Butcher study found
additional evidence for adolescents with elevations on Scale 6 (Paranoia)
to have characteristics similar to those with elevations on the
"excitatory" dimension. The similarities in the findings of the two studies
over a 15- year period demonstrate that the inhibitory- excitatory theory of
the MMPI has been durable over the past several decades, although the scales
which contribute to this continuum have varied somewhat.
Weaver and Wootton (1992) conducted a MMPI study with delinquent
adolescents over a three-year time period using MMPI special scales. Their
study had three objectives. The first was to identify the MMPI special
scales that characterized the personality characteristics of delinquent
adolescents. The second objective was to examine whether certain special
scales could be used to discriminate between levels of recidivism, crime
severity, and category of crime committed. The final objective was to
discover whether special scales could be used to predict delinquent
behavior more accurately than the traditional MMPI scales F. 4. 8. and 9.
The subjects were 401 convicted male juvenile offenders who ranged in age
from 13 to 17. The subjects were tested within two weeks of court
conviction. A score was considered significant if it had a T-score of
greater than or equal to 65 or less than or equal to 35. These investigators
found that there was a significant difference between the T scores of the
delinquent and normative groups on the following special scales: Re (Social
Responsibility). MAC (MacAndrew Alcoholism Scale). Pdl (Familial Discord).
Pal (Persecutory Ideas), Ma2 (Psychomotor Acceleration). Si3 (Staid-
Personal Rigidity), and TSC4 (Depression and Apathy). Weaver and Wootton
(1992) also found significant differences between delinquents with high and
low rates of recidivism on 11 of the clinical and supplementary scales: F,
Pd (Psychopathic deviate), Re (Social responsibility), Mal (Amorality), Ma
(Mania). Pd2 (Authority problems), Aut (Authority conflict), MAC
(MacAndrew Alcoholism Scale), Ma3 (Social Imperturbability), Hos (Manifest
Hostility). and Pal (Persecutory Ideas).
Weaver and Wootton (1992) also utilized a crime index used by the
District Juvenile Courts in Utah to discriminate between adolescents who
engage in "serious" crimes from those who engage in less serious crimes. The
MMPI special scales that they reported having the largest differences
between these groups were Re (Social Responsibility), Es (Ego Strength), Cn
(Control). and St (Social Status). The delinquents who engaged in less
serious crimes scored higher on Mf I (Narcissism- Hypersensitivity) and Fam
(Family Conflicts). The authors also tested an hypothesis concerning those
delinquents who had committed assaultive behaviors against persons versus
those who had not. The high assaultive group scored significantly higher on
Re (Social Responsibility), Pdl (Family Discord), and Ma2 (Psychomotor
Acceleration). In contrast. those individuals who were in the low
assaultive group scored significantly high on MAC (MacAndrew Alcoholism
Scale). Mf I (Narcissism-Hypersensitivity). Sc IA (Social Alienation). Fam
(Family Problems). and TSC4 (Depression and Apathy).
Finally. Weaver and Wootton (1992) discussed four composite
"personalities." The first of these was the recidivist personality which
was differentiated from the others on the MAC (MacAndrew Alcoholism), Pd
(Psychopathic Deviate). and Re (Social Responsibility) scales. The second
"personality" was termed the property offender. This group was
characterized by high scores on Pd (Psychopathic Deviate). Pd2 (Authority
Problems). MAC (MacAndrew Alcoholism), Re (Social Responsibility), and Mal
(Amorality). The third personality group was the severe offender. This
group was distinguishable from others by high scores on Re (Social
Responsibility), Es (Ego Strength). Cn (Control), and St (Social Status).
The final group was the assaultive personality. The scales that best
discriminated between the high and low assaultive personalities were Ma3
(Imperturbability), Mf (Masculinity-Femininity), and Pa2 (Poignancy).
Weaver and Wootton (1992) stated that the combination of Scales F. 4. 8, and
9 may be the result of proportions of types of subgroups in each study's
delinquent sample. They concluded that this combination was not accurate in
discriminating types of offenders.
The Weaver and Wootton study (1992) was important in that it described
aggressive adolescents as more of a heterogeneous group than has been
reported in the earlier literature. However, the criteria by which they
differentiated groups was based more on results from special scales of the
MMPI rather than from the traditional scales.
In response to the controversies with the traditional MMPI, a more
specialized version was developed in 1991 and was termed the MMPI for
Adolescents (MMPI- A). In part, the MMPI- A developed from a number of
criticisms about the use of the original MMPI with adolescents. These
included the following: the test being too lengthy, the adolescent norms
were too outdated, the reading level may be too high, and, much of the
language may be outdated (Archer. 1984, 1987: Marks, 1974; Williams and
Butcher. 1989a, 1989b). Most of the research on the MMPI- A with adolescents
at risk for violence is still exploratory and speculative. It has been
hypothesized that use of the MMPI- A with aggressive adolescents will be
similar to that of the original MMPI: however, that has yet to be put to
empirical test. The differences between the two measures could account for
additional Scales to be added to the "excitatory" scales. particularly
Scale Pa (Paranoia).
FAMILY STRUCTURE AND VIOLENCE AMONG ADOLESCENTS
There have been a large number of studies over a 50-year period which
have attempted to look at the family structure of violent youths. These
studies have traditionally found that the family variables which contribute
most to aggressiveness and delinquency include parental criminality, poor
parental supervision, cruel or neglecting attitudes, erratic or harsh
discipline, marital conflict, and large family size (Bahr, 1979: McCord and
McCord. 1959; West and Farrington. 1973; Wilson, 1980). Studies have also
examined the role of family interaction patterns and the general emotional
environment of the family as it relates to delinquency. These studies have
tended to conclude that delinquents are more likely to be raised in families
which tend to exhibit more conflict and less stable family interaction
patterns (Alexander. 1973; Faunce and Riskin, 1970). In general, research
has found that the most violent adolescents are more likely to be found in
physically abusive households (Lewis et al., 1989). In fact. a history of
abuse and/ or family violence, were found to be one of the best predictors of
adult violent crimes (Lewis et al., 1989).
The emotional environment of the home is also important in
distinguishing delinquents from nondelinquents (Veneziano and Veneziano.
1992). In general. families of delinquents were more likely than families
of non-delinquents to express rejecting attitudes and to exhibit a lack of
warmth and affection (West and Farrington. 1973). In terms of parental
discipline styles. the parents of delinquents are more likely to be more
punitive, gave more vague commands, and were generally less effective in
stopping children's deviant behavior (Patterson, 1982). Singer (1974)
concluded that a configuration of very restrictive family policy, lax
policing, and lenient punishing were more common among the family of
There has also been a wealth of research on communication pattern
differences between delinquents and non-delinquents. Alexander (1973)
found that delinquent families evidenced higher rates of defensive
communication and lower rates of supportive communication than non-
delinquent families. He further noted that conversations in the families of
delinquents were disjointed and disorganized, and the actual communication
was unevenly distributed, with one member dominating during family
A recent study by Veneziano and Veneziano (1992) attempted to examine
the family functioning of male juvenile delinquents using the Family
Environment Scale (FES). The FES. developed by Moos and Moos (1981), is a
90-item psychological measure that asks respondents questions about their
perception of the family environment in a true-false format. This
instrument provides 10 subscales. with means of 50 and standard deviations
of 10. They also attempted to develop a typology of family social
environments using cluster analytic techniques. They examined the
resulting subgroups on a number of dimensions, including intellectual.
personality, and behavioral characteristics, in an attempt to determine if
the groups differed in ways that would have implications for research and
treatment. The subjects in their study consisted of 41 1 institutionalized
male adolescents who were committed as adjudicated delinquents to a state
department of corrections. The only exclusions were those delinquents who
transferred to another facility, or were released before they could
complete the five-workday classification process. Subjects ranged in age
from 11-7 to 16-4. with an average age of 14-9. The ethnic distribution of
the sample was 52% white and 49% African-American. Of these subjects, 67.6%
were serving their first offense and 57.9% were committed for one offense.
Approximately 67% were committed for property offenses, with the second
most common offense being of "probation violations." The resulting
analysis produced seven clusters, which classified 93.9% of the FES
profiles. They discovered that other analyses yielded clusters which were
either too small or too large to be statistically meaningful. They termed
Cluster 1 the Denial of Conflict. These subjects were similar to a norm
group except in the area of the open expression of anger. The subjects in
Cluster I were the oldest group and 95% were first time offenders. Cluster 2
was termed the Repression of Expression and they were consistently low on
measures of anger and hostility. However, they were highest in terms of
being aware of limited opportunity. Cluster 3 was termed the Suppression of
Independence. Families of these delinquents were attempting to establish
strong controls but that independence was discouraged and conformity was
encouraged. This group was the second highest in terms of state and trait
anger and were also high in terms of state and trait anxiety. The Cluster 4
(Unstructured Conflict) had very high Conflict scores and low scores on
other subscales. They were the youngest group and had the highest scores on
measures of anger and impulsivity, and were likely to be the group that
rejected middle-class values. Cluster 5 was termed Structured Conflict and
they had high Conflict Scores but also had high Control Scores and average
scores on the other scales. Cluster 6 was termed Unstructured Control. They
were characterized by little expression of freedom and with little emphasis
on independence outside of the household. Subsequently they typically had
little involvement in activities outside of the house. In addition, the
parental style was characterized by mild attempts at maintenance and
control. Collectively as a group. Cluster 6 fell within the middle of the
scores received in the assessment measures. Finally Cluster 7 was termed
Structured Control and were characterized by a high emphasis on
maintenance, control, and achievement, and low emphasis on expression of
feelings and independence. Cluster 7 had the lowest anger scores, but the
second highest anxiety scores.
Veneziano and Veneziano (1992) summarized that most of the
adolescents had relatively high scores on the scales of Organization and
Control, suggesting that their families were not characterized by an
absence of structure. They imply that the families of delinquents establish
rules and procedures, but they have a hard time implementing these
procedures due to inappropriate or ineffective monitoring and
disciplining. They also found that the subset of delinquents who live in
homes with a higher degree of conflict tend to exhibit more severe
difficulty. These findings are consistent with the literature that found
associations between abusive and/ or violent homes and aggressive behaviors
among delinquent adolescents (Lewis et al., 1988). Veneziano and Veneziano
(1992) concluded that delinquents are not a homogeneous population in terms
of presenting symptomatology, and need a multidimensional assessment
approach when complex constructs such as family dynamics are an issue.
The Veneziano and Veneziano (1992) study is interesting in pointing
out the affective states of the delinquents involved in the study. Many of
the clusters described in their study had abnormally high levels of anger
and anxiety. In terms of anger, the majority of the clusters had chronically
high levels. The researchers described their population well in terms of
heterogeneity, but it would be fascinating to assess how the dimension of
anger and anxiety in these groups differed. We know that anger can be
assessed in a number of distinct dimensions. Examples of differing anger
styles include hostility (trait anger). suppressing angry feelings for fear
of letting others know one is angry, overt anger expression, and frequent
attempts to consciously control the overt expression of anger. Since
Veneziano and Veneziano (1992) found a relationship between affect and
family structure, do these factors interact with personality variables
discussed before in the previous section? The inclusion of variables
assessing both family factors and anger in adolescents at risk in the
present study should help to clarify the potential interaction or role of
these factors in this group.
Anger is considered to be an affective state which can consist of
feelings of irritation, irritability, annoyance, fury. and rage. This
affective state is known to activate the autonomic nervous system (ANS) and
to involve the endocrine system as well (Johnson and Greene, 1992, pg. 25).
It is also known to cause tension in the skeletal musculature, antagonistic
thought patterns, and aggressive behaviors (Johnson and Greene, 1992). The
experience of anger is typically a complex emotional and physiological
pattern that can be elicited in interpersonal and social situations among
individuals who have a strong proneness (trait) to experience anger. When
discussing anger, it is typically important to distinguish between the
experience and expression of anger. Generally the experience of anger can
be characterized by frequency. intensity, and duration. The expression of
anger can be conceptualized as a single. bipolar entity which can range from
suppression of angry feelings to expression of anger through aggressive
behavior towards other individuals or objects in the environment
(Spielberger. 1988). It is important to understand the emotional experience
of anger before one can conceptualize the much broader concept of hostility.
Typically. hostility has been understood to contain angry feelings
but it may also have the added dimensions of negative and destructive
attitudes and beliefs. Thus hostility may motivate aggressive and violent
acts. particularly in interpersonal situations. Aggression is said to
describe destructive and punitive behaviors directed against another. The
feelings of anger often result in hostility and aggression, although anger
is not a necessary precursor in either of these states. Unfortunately, the
concepts of anger, hostility, and aggression are often used interchangeably
in research literature as with the lay public (Johnson, 1990).
Anger has been conceptualized by a model proposed by Spielberger
(1988). The model conceptualizes anger as a psychophysiologic response
that can be elicited in social situations where an individual feels (a) loss
or threat of loss of (b) something felt to be possessed (rights, job,
marriage, or physical objects) through (c) perceived arbitrary, unfair, and
unjustifiable acts by others (people, select groups, or society). The
behavioral manifestations of anger can involve overindulgence or overuse of
alcohol. cigarettes. illicit or over-the-counter drugs, and food.
Hostility Types and Anger-Coping Styles
The social learning theory of aggression proposed by Bandura (1973)
suggests that aversive stimulation can produce a general state of emotional
arousal that can promote a host of accompanying behaviors. Depending on a
person's cognitive appraisal of the situation and the environment, the
accompanying emotional arousal can be interpreted as anger (Hecker & Lunde.
1985. pg. 227). This anger can be expressed as aggressive behavior ranging
from a deviant thought to physical violence depending on an individual's
learned social skills. Usually this anger will dissipate and leave an
individual free for new emotional experiences; however, feelings of anger
can be prolonged or compounded on subsequent occasions by recalling or
focusing on provocative situations. Individuals who frequently become
angry in the absence of distinctive external stimuli may be seen as
exhibiting chronic hostility.
Chronic hostility is not formally recognized as psychopathology
according to the Diagnostic and Statistical Manual of Mental Disorders: 4th
Edition (DSM-IV): although chronic hostility has been implicated in a
number of psychiatric disorders which include Somatization Disorder.
Intermittent Explosive Disorder, Conduct Disorder. Oppositional Defiant
Disorder. Antisocial Personality Disorder. and Borderline Personality
Disorder (American Psychiatric Association, 1994). Common characteristics
of individuals with chronic hostility can include deficits in social
skills: lack of sensitivity in interpersonal situations; failure to
recognize the consequences of their behavior; and, being unable to see a
situation from another person's viewpoint (Spivack, Platt, & Shure. 1976).
The intense emotional arousal from anger-intensive individuals may impair
cognitive appraisal of aversive situations and can help to facilitate
impulsive behavior (Zillman. 1983).
The etiology of chronic hostility is largely unknown. Role-models
such as parents can influence the development of attitudes and behavior
patterns. Among these also would be unstable family patterns including but
not limited to abusive families or neglectful families. Other factors
influencing chronic hostility could include feelings of inadequacy or
insecurity,. unrealistic views and expectations, and insufficient skills
for dealing with stressful situations (Hecker & Lunde, 1985).
Some approaches to address the problem of chronic hostility have been
proposed (Novaco, 1975: Novaco. 1985; Glick & Goldstein, 1987). Usually
these approaches have addressed the issues of correcting behavioral and
cognitive deficits that may underlie chronic hostility. In addition,
relaxation training and systematic desensitization are used to manage the
physiological correlates of emotional arousal (Novaco. 1985). However,
chronically hostile individuals do not represent a clinically homogeneous
group. Although all chronically hostile individuals should be able to
benefit from stress inoculation training, the authors propose that
treatment should be tailored to meet the specific needs of individuals
(Hecker & Lunde. 1985).
Typology of Chronic Hostility:
Individuals prone to chronic hostility have been conceptualized by
Hecker and Lunde (1985) as belonging to three main types. each of which can
be further divided into two subtypes. The three main groups have been termed
the Uncontrolled Anger Type (Type I). the Overcontrolled Anger Type (Type
2), and the Suppressed Anger Type (Type 3).
The first type (Uncontrolled Anger) has been described as
experiencing anger emotionally and readily translating their angry
feelings into violent or aggressive behavior. These types were originally
described by Megargee (1966, 1982) as having weak inhibitions against
aggressive behaviors and exerting little or no control over instigation
towards aggression. This type was divided into two subtypes: the Impulsive
Undercontrolled Type and the Deliberate Undercontrolled Type.
The Impulsive Undercontrolled type experiences anger emotionally and
acts impulsively with little or no deliberation. When they are provoked.
this type will react quickly and many times will disregard personal safety
and impulsively cause accidents that may cause injuries to others or
themselves. For example, an adolescent may become angry at another and will
commence a "drive-by" accident with innocent bystanders in the line of fire.
The Deliberate Undercontrolled type experiences anger that does not
directly lead to aggressive behavior. They do not usually behave
impulsively, but will instead plan for aggressive behavior to be carried out
later. Because the aggressive behavior is separated in time from the angry
event, the aggressor is able to act anonymously (Hecker & Lunde. 1985).
Typical deliberate acts performed by them include conspiring with others to
defeat a competitor. plagiarism. embezzlement, and homicide.
The second main type (Overcontrolled Anger) commonly experiences the
emotion of anger, but their angry feelings are strongly suppressed and not
translated into aggre,,i e behavior. Megargee (1966: 1982) described this
type as having strong inhibitions against aggressive behavior; they exert
enormous control over provocations to aggression. This type can be further
divided into two subtypes: Stable or Unstable Overcontrolled types.
The Stable Overcontrolled type traditionally have strong controls
over the engaging of aggressive behaviors. Even under conditions of extreme
stress, they manage to control their responses to nearly every provocation
to aggression. When challenged, an individual of this type may adopt a
passive stance or withdraw so as to avoid an unpleasant emotional exchange.
However, the emotionally provoking incident is likely to be remembered for a
long time. The Stable Overcontrolled type have a tendency to become
preoccupied with their unexpressed thoughts and feelings that they may
neglect their health by smoking or eating too much, exercising too little,
or ignoring the early symptoms of physical disorders.
The Unstable Overcontrolled type also has strong inhibitions against
displaying emotionality yet. under conditions of extreme stress, the
inhibitions against aggression may fail for reasons which are not clearly
understood. Extremely violent criminal behavior may result, including
multiple homicides. Individuals who commit these crimes are usually
remorseful afterward, and are able to reconstruct their former. unstable
The third chronically hostile type has been termed the Suppressed
Anger type. The research suggests that this type does not experience anger
directly but rather the anger is suppressed and functions as a source of
anxiety. pain. or other maladaptive behavior (Hecker & Lunde. 1985). Since
their cognitions are separate. these individuals are usually surprised when
others around them call them angry or hostile. Their tendency may be to
suppress other emotions too, and their general responses to common
situations may lack genuine, appropriate affect. The two subtypes of the
Suppressed Anger type can be differentiated on the basis of whether or not
their interpretations of reality are normal or psychotic.
The Normal Suppressed type is a hostile type that will become anxious
and engage in unproductive behavior when confronted by a difficult or
frustrating situation. To obtain a sense of relief from their psychological
distress. these individuals may engage in unnecessary work or leisure-time
activities that will keep them constantly busy.
The second suppressed anger type is the Psychotic Suppressed anger
type. These individuals often present with a distorted perception of
reality, and their responses to situations and events are often
inappropriate. This may be present in psychotic depression or
schizophrenia. where one has to infer the presence of overt hostility.
It is believed that adolescents at risk are a heterogeneous group in
terms of personality and anger presentation. The three major hostility
types (Undercontrolled. Overcontrolled. and Suppressed) are hoped to be
found in adolescents at risk. The Undercontrolled Anger Types.
theoretically. would be the easiest to identify due to their frequent
presentations of anger. Particularly, they would be easy to identify by
self- report measures given the low energy invested to contain anger.
In contrast to the Undercontrolled Anger Types. the Overcontrolled
and Suppressed Anger Types would be more difficult to separate out on self-
report measures compared to normal adolescents. On self- report measures,
it is hypothesized that the Overcontrolled Anger Type. Suppressed Anger
Type and normals would all score within normal limits. To assist in further
dividing these groups into accurate subgroups would require additional
assessment devices, such as semi- structured or structured interviews. In
addition. retrieving information from secondary sources. ie. school
officials or parents. would be invaluable in providing an accurate portrait
of anger presentation over a period of years. For instance, one would
predict that the Overcontrolled Anger Type would have only a few instances
of violent presentations: however, these presentations would be extremely
violent in general and well out of proportion to the individual's usual
presentation. The Suppressed Anger Type could be hypothesized to have
frequent anger demonstrations, but would not attribute their outburst to
anger. For the Suppressed Anger Type. an interview could also provide
information about an individual's verbal display, such as verbal
stylistics. to better verify the existence of this anger type. Although
information, such as interview data, will not be gathered in this study.
hopefully the results of this study will guide the way for more expansive
Subjects were 119 male adolescents between the ages of 13 to 18 years
old. who were currently enrolled in alternative schools for behaviorally
disruptive children and juvenile detention centers located in the regions
of North Central Florida. South Central Alabama and Atlanta, Georgia. This
study collected self-report data from the students. After the data records
were collected the sample was classified according to demographic
information such as age. ethnic affiliation, family organization (single
parent. dual parent, etc.) and occupation of parent. The last 20 subjects
only have demographic information and completed MMPI-As. They were missing
the STAXI. STAI and the FES. A description of the demographic questionnaire
is included in Appendix A.
Prior to testing. the proper administration protocol and the nature
of the tests were explained to the subjects in a group format. If students
had difficulty understanding the official procedures detailed above, it was
explained to them on an individual basis. Support was obtained from the
identified schools and detention centers in the Florida, Alabama. and
Georgia areas prior to testing. All test procedures were scrutinized by an
Institutional Review Board (IRB) at the University of Florida before
The tests involved in this research project were administered in a
small group setting at a child's school. Subjects were assigned into groups
based on reading levels, which was accessed in a student's academic record
or by consultation with the child's teacher. Those research subjects with a
reading level above the sixth grade level were administered the tests in
their usual. standardized format. However, for those subjects with a
reading level below the sixth grade level, an audio-taped version of the
test measures was administered. The time required to finish all of the test
procedures was between 60 minutes to 120 minutes. Those children who
completed all of the test measures received a five-dollar gift certificate
for use at the McDonald's restaurant.
There were four assessment measures employed in this study. They were
the Minnesota Multiphasic Personality Inventory for Adolescents (MMPI- A),
the State- Trait Anger Expression Inventory (STAXI). the State- Trait
Anxiety Inventory (STAI), and the Family Environment Scale (FES).
Minnesota Multiphasic Personality Inventory- Adolescents (MMPI- A):
The MMPI-A was developed in 1989 in order for adolescents to have a
personality instrument that better reflected their outlook on life and
stage of human development than the original MMPI. It has been noted by
previous authors (Archer. 1984, 1987: Marks, 1974: Colligan and Offord.
1989: Williams and Butcher. 1989a. 1989b) that the original MMPI had a
number of limitations when the instrument was used by adolescents. One of
these limitations was the fact that many felt as though the item content was
inappropriate for use with adolescents: partly due to the fact that they
were written for adults. The problems may be subtle such as the verb tense
being inappropriate or there were was not enough content suitable for young
people. Another complaint was that there appeared to be a lack of content or
research scales suitable for adolescents. A third limitation concerned the
fact that many adolescents have a characteristic high responding on some
scales. such as the F Scale on the MMPI, which may in part reflect their
feeling of alienation or identity diffusion (Archer. 1984, 1987). The F
Scale in itself may be developmentally unsuitable for this group. A final
problem concerned the problem of selecting which norms to use with
adolescents since the MMPI uses both adult and adolescent norms. The
adolescent norms were developed by Dahlstrom, Welsh, and Dahlstrom (1972)
and Marks et al. (1974). Both of the adolescent norm sets were T-scale
conversion tables for adolescent scores. Authors have shown that the use of
adult norms tended to over-psychopathologize typical adolescents.
Colligan and Offord (1989) concluded that adolescent responding to the MMPI
was decidedly different from adult responding.
The committee to develop the MMPI- A was formed in 1989. The committee
was termed the MMPI Adolescent Project Committee and was appointed by
Beverly Kaemmer of the University of Minnesota Press. The committee had two
recommendations. The first recommendation the University of Minnesota
Press publish a separate form of the MMPI with separate norms for
adolescents. The second recommendation was that the MMPI- A have the same
standard scales as the MMPI.
The original form of the MMPI- A (Form TX) contained 704 items. The
704 item booklet was used for the adolescent normative data collection and
in a clinical sample described by Williams and Butcher (1989a. 1989b).
The committee recognized the need to reduce the total number of items
to make the instrument more appealing to adolescents. The final version of
the test had items comprising the original validity indicators L and K. new
indicators Fl and F2. the original clinical scales, new Content Scales.
existing supplementary scales MAC- R, A, and R. and a new scale. IMM
(Immaturity). were included in the MMPI- A. The final version that resulted
contained 478 items arranged such that all of the items on the basic scales.
as well as a number of items of new uniquely adolescent items. Content
Scales, and supplementary scales are not obtainable in the first 350 items,
but require administering the full MMPI- A.
MMPI- A Validity Indicators:
The MMPI- A Lie scale consists of 14 items and it was designed to
measure naive attempts by teenagers to put themselves in a favorable light.
The MMPI- A F scale is divided into a 33-item Fl scale and a 33-item F2
scale. This scale was developed to measure the frequency of endorsement of
infrequently used items. The Fl scale covers those items that occur in the
first 350 items of the test and F2 covers those that appear later in the test.
It was generated in this way to possibly identify adolescents who have
changed his or her test-taking approach in the later stages of the testing
The MMPI- A K scale consists of 30 items and is thought to be a measure
NMMPI- A Clinical Scales:
Scale 1 (Hs: Hypochondriasis): This scale consists of 33 items that
were selected to identify clients with a history of symptomatology
characteristic of hypochondriasis.
Scale 2 (D: Depression): The D scale is composed of 60 items that have
been chosen to reflect general feelings of dissatisfication, hopelessness,
and possibly suicidal ideation.
Scale 3 (Hy: Hysteria): This scale is comprised of 60 items that were
selected to identify individuals who respond to stress with hysterical
reactions that include sensory and motor impairment without an organic
Scale 4 (Pd: Psychopathic Deviate): This scale consists of 49 items.
It is developed on the basis of responses from adolescents with patterns of
lying, stealing, sexual promiscuity, and alcohol abuse.
Scale 5 (Mf: Masculinity-Femininity): The Mf scale is a measure of
stereotypic masculine or feminine interests.
Scale 6 (Pa: Paranoia): This scale is comprised of 40 items. It is
used to identify patients with a high degree of paranoid symptomatology.
Scale 7 (Pt: Psychasthenia): This scale was developed in order to
measure symptoms related to obsessive-compulsive disorder including
anxiety and rumination. It is composed of 48 items.
Scale 8 (Sc: Schizophrenia): This scale has 77 items. It is used to
reflect those individuals who endorse bizarre thought processes, peculiar
perceptions, social isolation, disturbances in mood and behavior, and
difficulties in impulse control.
Scale 9 (Ma: Hypomania): This scale consists of 46 items designed to
identify patients manifesting hypomanic symptoms.
Scale 0 (Si: Social Introversion): This is a scale designed to
measure scores on social relationship problems.
A test-retest was conducted on normative sample of 154 subjects.
These subjects completed one complete test questionnaire and then were
asked to complete the same questionnaire a week later. Pearson Product-
Moment test-retest correlations for the sample were calculated. The
correlations ranged from values of .65 to .84 for the three validity scales
and the ten clinical scales.
The State- Trait Anger Expression Inventory (STAXI):
The State- Trait Anger Expression Inventory (STAXI) is a measure
designed by Spielberger (1988) that is used to assess the experience and
expression of anger. One of the primary reasons the STAXI was developed was
to provide a technique to evaluate the components of anger that can be
utilized for evaluations of normal personalities and personality disorders
The STAXI is composed of 44 items. With these 44 items. one can derive
six major scales and two sub- scales. The six major scales are State Anger
(10 items r. Trait Anger (10 items). Anger- In (8 items). Anger- Out (8
items). Anger Control (8 items), and Anger Expression (24 items). The sub-
scales are Angry Temperament (4 items) and Angry Reaction (4 items).
The State Anger scale was designed to measure the magnitude of anger
felt at the time of testing. The Trait Anger scale assesses an individual's
disposition to experience anger in different situations. Subscales to the
Trait Anger Scale are Angry Temperament and Angry Reaction. Angry
Temperament is a measure of a general predisposition to experience anger
with little or no provocation. Angry Reaction assesses subject differences
in the experience of anger when unfairly evaluated or treated harshly by
Anger- In measures the frequency at which an individual is likely to
restrain angry feelings because the individual is uncertain of what to do
with the angry affect. Conversely. Anger- Out is an anger expression scale
which assesses how an individual expresses anger to other people or objects
in the environment. Anger Control professes to measure an individual's
frequency to control the feelings of anger.
Finally. Anger Expression is a general measure of the frequency at which
anger is expressed. Anger Expression assesses general angry affect
regardless of the direction of the expression.
Information about the internal consistency of the STAXI scales was
gathered from different samples (Spielberger. 1988). These samples
consisted of male and female college undergraduates. adolescents, adults
who were not college students and Navy recruits in the Tampa. FL area. For
all scales. Cronbach's coefficient alphas ranged from .70 to .87 reflecting
a high degree of internal consistency. Test-retest correlations of all of
the scales ranged from .58 to .75.
The State- Trait Anxiety Inventory (STAI):
The State- Trait Anxiety Inventory (STAI) was a measure designed by
Spielberger (1983) to evaluate an individual's experience of anxiety. The
measure wvas designed to be self- administering and may be administered
either individually or in groups. Completion time for the STAI ranges from
10 to 20 minutes. The STAI assesses two indices of anxiety: State- Anxiety
(S- Anxietvy) and Trait- Anxiety (T- Anxiety). S- Anxiety consists of 10
items. conversely T- Anxiety is composed of 11 items.
When administered. Spielberger recommends that the S- Anxiety Scale
should be administered first, followed by the T- Anxiety Scale, due to the
fact that the S- Anxiety scale is sensitive to the conditions by which the
test is administered and can be influenced by feelings created by the T-
Each STAI item is given a weighted score of 1 to 4. A rating of 4
suggests high anxiety while a rating of I implies low levels of anxiety.
Norms for the STAI were collected on working adults, college students, high
school students. and military recruits. The high school sample consisted of
424 tenth-grade students who were tested in their schools (Spielberger.
1983). Cronbach's coefficient alphas for the high school sample .90 for the
State and the Trait Anxiety scales. Test-retest reliability coefficients
were calculated for both scales and the values ranged from .65 to .75
The Family Environment Scale (FES):
The Family Environment Scale (FES) (Moos & Moos, 1994) is a 90- item
test that was designed to assess three dimensions of family functioning.
The first of these dimensions is relationships. The relationship dimension
is global and attempts to reflect several areas such as belonging, pride.
open expression, and conflict. Another dimension is personal growth and
development. Central to this theoretical dimensions are such areas as
autonomy, academics. competitions. family activities, and religious
emphasis. The final dimension is system maintenance which deals with
centers around the structure and organization of the family. The system
maintenance dimension also examines the amount of perceived control
exercised over each member. The FES is a self-report device which can be
given independently to individual family members. The ten subscales of the
FES are represented as T scores with means of 50 and standard deviations of
10. Internal consistencies were calculated for the ten scales of the FES
based on a normative sample of 591 individuals (Moos & Moos. 1994).
Cronbach's coefficient alphas ranged from .61 to .78. Test-retest
reliabilities were calculated at two-month intervals and values for the ten
scales ranged from .68 to .86.
1) The first hypothesis proposed that adolescents at risk are a
heterogeneous group in terms of personality presentation. It was predicted
that these adolescents will fall into four distinct personality groups
based on MNMPI scale scores. These personality groups could be postulated as
falling along the "inhibitory- excitatory" continuum originally proposed
by Hathaway and Monachesi (1963). One can expect an "inhibitory" group with
elevations on many of the inhibitory scales, an "excitatory" group with
elevations on many of the excitatory scales, a mixed group with features of
both and a non-elevated group.
2) The personality cluster groups which have been discussed in the
first hypothesis were postulated to differ on the following four variables:
anger expression, trait anxiety, maternal control, and trait anger. It was
proposed that each personality cluster group would have distinctive
patterns of responding to each of these variables. This is related to the
research in the literature of heterogeneous patterns of responding to
family environment and affective data which was influenced by the work of
Veneziano and Veneziano (1992).
It was further hypothesized that the personality groups would have
differing responses on the dimensions of anger-in, anger-out and anger
3) The third hypothesis proposed that there is a distinct
relationship between high levels of chronic anger (hostility), levels of
perceived family conflict, family cohesion, and delinquency. This
hypothesis seeks to help define the assessment and evaluation of the
Undercontrolled Hostility Type.
1) The first analysis consisted of a cluster analysis with the
following scales of the MMPI- A: F. 2. 3. 4. 5, 6. 7. 8, 9. 0. There were five
"inhibitory" scales that were investigated and they were 2. 3. 5, 7. and 0.
In addition, there were five "excitatory" scales and they consisted of F. 4.
6, 8. and 9. A hierarchical, agglomerative cluster analysis (Centroid
linkage method), which utilized squared Euclidean distances between scale
items. was conducted to maximize heterogeneity among cluster groups. This
followed the hypothesis which stated that adolescents-at-risk for violence
are a heterogeneous group in terms of personality presentation.
2) The second analysis consisted of running several one-way ANOVAs to
verify the evidence provided in the cluster analysis with personality
cluster group as the independent variable. Differences in the four
dependent variables (anger expression, trait anxiety, maternal control.
and trait anger) were examined. It was predicted that individual cluster
groups can be differentiated by the four dependent variables listed above.
Additional analyses were conducted with anger-in, anger-out and
anger control as dependent variables. Personality cluster group remained
the independent variable.
3) The third analysis was a series of independent linear regression
equations with Trait Anger. Anger-In, Anger-Out, and Anger Control as the
dependent variables. Maternal care, family conflict, and delinquency were
entered as independent variables.
Sample information collected is described below. Information was collected
from subjects (N = 119) from a number of testing sites in northwest Florida
(Alachua and Marion counties), central Alabama (Eufaula. AL), and northeast
Georgia (Atlanta. GA). The average age of the collection sample was 16 years
old (N= 119. standard deviation = 1.04). The minimum age was 13 years old and
the maximum age was 18 years old. The average grade level for the subjects
was 10th grade with a standard deviation of 1.2. From information collected
in the sample. 60.9 % of the sample described themselves as African-
American. 31.5 % of the sample as White. 5.2 % of the sample as Latino. 1.1 %
as Native American and 1.1 % as "Other." In addition. further breakdowns of
the sample revealed that 22.8 % of the sample resided with their biological
parents. 31.5 % resided with one biological parent and a step-parent. 33.6 %
with a single-parent (29.3% were single mothers). and 12 % had "Other"
arrangements. These other arrangements included additional family members
including grandparents, aunts or uncles, or frequently children were wards
of the state.
Of those children who lived with at least one of their biological
parents. 80.4% of the parents were employed and 19.6 % were described as
unemployed. Lastly. 84.8 % of the sample indicated that they had an arrest
history and 15.2 % had no arrest history. Subject characteristics are
described in Table 1.
Table 1: Mean subject characteristics (N= 119)
Both biological parents
Biological parent and step-parent
Parental Employment Status
Table 1 Cont.
Information from the MMPI-A is described below (See Table 2).
Table 2: Mean MMPI- A scores for the entire sample
Mean Standard Deviation
(N = 119)
The cut-off score for the MMPI-A was based on average subject scores
based on a normative sample. Information from Table 2 shows that no scale
had a clinical elevation (T score > 65) compared to the normative sample.
However, it should be noted that many of the highest scales had the greatest
standard deviations. Information on the mean sample MMPI-A distribution is
illustrated in Figure 1.
Figure 1: MMPI-A Mean Profile For Entire Sample (n= 1191
4- S5 -
F 3 4 5 6 7 9 10
Data was collected on the STAXI. Table 3 displays the mean score, standard
deviation and percentile rank for the STAXI. The percentile rank
comparisons for these scores were based on average subjects in a normative
Table 3: Mean STAXI scores for the entire sample
Mean Standard Deviation
State Anger 17.50 7.41
Trait Anger 22.47 6.12
Trait Anger/ Temperament 8.30 3.16
Trait Anger/ Reaction 9.66 3.04
Anger-In 17.77 4.13
Anger-Out 18.12 4.59
Anger Control 21.17 4.94
Anger Expression 30.78 8.89
On the STAXI there were clinical elevations (percentile rank > 75) on State
Anger only when the present sample was compared to a normative sample.
However. Anger Out and Anger Expression were substantially elevated as
Tables 4 and 5 illustrate the results of the STAI and the FES respectively.
Table 4: Mean STAI scores
N = 99
Mean Standard Deviation
Both State and Trait Anxiety scores for the subject group fell within
the normal range. The means for both the STAI and the FES were based on a
Table 5: Mean FES scores
Family Cohesion 44.09
Family Conflict 52.46
Family Control 52.25
T-tests were conducted between the normative sample means and standard
deviations and the means and standard deviations of our sample for the FES.
All t-tests were significant. Family cohesion was found be significantly
lower than the normative sample [t,,, 98 = -5.09]. Family Conflict [t 0, 98 =
2.83] and Family Control [to0, 98 = 2.36] were found to be significantly
higher than the normative sample.
MMPI-A Scores and Ethnicity
The majority of subjects within this sample were African-American.
Previous research on the MMPI has suggested differences between certain
validity and clinical scales by ethnicity. Comparison tests were conducted
between MMPI-A scales F. 2. 3. 4. 5. 6. 7. 8. 9. 0 and ethnicity (African-
American and White). See Table 6 below.
Table 6: NIMPI-A scores by ethnicity (African-American and White) N= 119
Scale African-American White Significance Level
Mean SD Mean SD
Scale F 61.78 10.37 60.79 10.37 .25
Scale 2 55.88 10.28 56.45 11.21 .94
Scale 3 51.61 9.40 52.38 10.09 .82
Scale 4 60.10 9.34 62.28 11.70 .35
Scale 5 45.72 8.50 45.73 6.42 .07
Scale 6 58.99 10.67 61.21 18.55 .00
Scale 7 53.07 7.70 54.38 12.50 .00
Scale 8 57.54 11.00 56.55 15.80 .02
Scale 9 58.06 11.53 55.79 14.23 .67
Scale 10 50.00 7.56 49.62 9.92 .06
According to this analysis. Whites were significantly higher than African-
Americans on Scale 6 (Paranoia) and Scale 7 (Psychasthenia). African-
Americans were significantly higher on Scale 8 (Schizophrenia).
Additionally. there was a trend towards African-Americans being higher on
Scale 5 (.Masculinity-Femininity) and Scale 10 (Social Introversion).
The first analysis consisted of a cluster analysis with the following
scales of the MMPI-A: F. 2, 3, 4. 5. 6, 7, 8. 9. 10. A hierarchical cluster
analysis (Centroid linkage method) which utilized squared Euclidean
distances between individual scale items was used. The Centroid linkage
method xwas utilized due to the nature by which it can maximize compact
clusters composed of similar cases. The Single linkage method was excluded
due to the fact that it has a tendency to form long, elongated clusters
(Aldenderfer & Blashfield. 1984. pg. 39). Consequently. Ward's method was
not utilized because it has difficulty when sample size exceeds 100 cases
and it generates solutions that are strongly influenced by profile
elevation, Aldenderfer & Blashfield, 1984. pg. 44). Squared Euclidean
distances %were utilized in order to maximize the dissimiliarity of unlike
clusters. The analysis was run on SPSS-Windows Version.
A dendrogram of the initial cluster analysis results strongly
suggested a three-cluster solution to the sample. Moreover, an inspection
of the fusion coefficients (the numerical value at which various cases merge
to form a cluster) suggested a trend in which there was a substantial "jump"
in the value of the coefficients at the three-cluster solution. However,
both of these methods are highly subjective and susceptible to experimenter
bias (Aldendefer & Blashfield, 1984. pg. 54). Mojena (1977) and Mojena and
Wishart (1980) developed a procedure to objectively evaluate the optimal
partition of a hierarchical clustering procedure. Their equation examines
the value of the fusion coefficient at a given stage of the clustering
process (z,,). the mean value of the fusion coefficients (z), the standard
deviate (k). and the standard deviation of the fusion coefficients (s,). For
the standard deviate, Mojena (1977) suggested a value of 3.00 to test for
significance. The equation is illustrated below.
Z > z + ks,
Based on the results of the initial cluster analysis, the following values
were derived: Z,+ (Z,,1)= 3189.87: z= 621.78: k= 3.00: sz= 631.16. The null
hypothesis would be a cluster of one if the value of Z,,, can not exceed the
value of [ z + ksz].
Z,,116 > z + ks
3189.87 > 621.78 +(3)(631.16)
3189.87 > 2515.27
Utilizing the information available in this equation, the best solution
suggested was the three cluster solution. Therefore, the null hypothesis
Table 7: Age and educational level for personality cluster types.
Cluster I Cluster 2 Cluster 3
(N= 58) (N= 25) (N= 36)
Age 16 16 16
Educational Level 10 10 10
There were no statistical differences between the three personality cluster
types on age or level of education.
MANOVA and MMPI- A Variables:
To determine if there was an effect of cluster type among MMPI- A
variables, a MANOVA was conducted with the MMPI- A scales as dependent
variables and cluster type as an independent variable.
Table 8: MANOVA of MMPI- A Variables and Cluster Type
Scale F-Score Significance Level
F Scale 4.98 .01
Scale 2 2.56 .08
Scale 3 8.19 .00
Scale 4 10.14 .00
Table 8 Cont.
Scale F-Score Significance Level
Scale 5 2.96 .06
Scale 6 3.52 .00
Scale 7 6.39 .00
Scale 8 6.35 .00
Scale 9 9.72 .00
Scale 10 .58 .56
The results of the MANOVA suggest that overall there was an effect for MMPI-
A scales and cluster type. Only Scale 2. Scale 5 and Scale 10 failed to reach
the .05 level of significance. To determine if there were further
differences between MMPI-A scales and cluster type. post-hoc comparisons
were conducted utilizing Bonferroni critical value to control for uneven
pair-wise comparisons. One-way ANOVAs were run with the MMPI-A scales as
dependent variables and personality cluster type as independent variables.
Differences are illustrated in Table 9. The three groups are statistically
distinct from one another on the F Scale. Scale 1 (Hypochondriasis), Scale 2
(Depression), Scale 6 (Paranoia), Scale 7 (Psychasthenia), and Scale 8
Table 9: Mean MMPI-A Scale T-scores for individual cluster types
M 46.22 M 48.00 M 44.33
SD 7.94 SD 6.40 SD 7.50
M 78.00' M 48.59'
SD 12.35 SD 5.49
M 65.20k' M 45.19'
SD 7.00 SD 5.86
M 75.47' M 47.48'
SD 10.56 SD 7.92
(Note:) Different superscripts denote significance at the .05 level.
A further illustration of the cluster types is provided in Figure 2.
Figure 2: MMPI Scale Profile for Cluster Types
Fig 2: Mean Validity Scales
S--- Cluster 2
- Cluster 3
2 3 4 5 6
MANOVA and STAXI variables:
7 8 9 10
A MANOVA was conducted with selected variables of the STAXI as
dependent variables and the three cluster types as independent variables.
The results are illustrated in Table 10.
Table 10: MANOVA and STAXI variables
The results indicate a significant effect for anger variables with the
exception of Anger-In and Anger-Out.
Cluster Type and Dependent Variables:
To address the second hypothesis that the individual cluster types
would be distinct on a number of affective and familial variables, a series
of one-w ay ANOVAs were run with cluster type as an independent variable. The
first one-way ANOVA was conducted with anger expression as a dependent
variable and cluster type (3-Cluster Solution) as an independent variable.
The results indicated that there was a statistically distinct difference
between the personality cluster types and the amount of anger expression,
[F(2. 96) = 7.65. p < .05]. Both Cluster Group 1 and 2 were significantly
elevated when compared to Group 3. However. Group 1 and 2 appear to not be
significantly distinct from each other.
A one-way ANOVA was conducted with Trait Anger as a dependent variable
and personality cluster type as an independent variable. The results
appeared to show a relationship between the personality cluster types and
Trait Anger. [F(2. 96) = 7.42. p < .05]. Groups 1 and 2 appear to be
distinctly elevated compared to Group 3. However, Group 1 and 2 are not
statistically distinct from each other in terms of elevation.
A one-way ANOVA was conducted with family control as a dependent
variable and personality cluster type as an independent variable. The
results suggested that there is not a strong relationship between the
personality cluster types and differing perceived levels of family control.
[F(2. 96) = .0911. p > .05]. The means for all three groups on family control
were statistically similar.
A one-way ANOVA was conducted with Trait Anxiety as a dependent
variable and personality cluster type was an independent variable. The
resulting analysis suggests a strong relationship between the personality
cluster type and the perception of anxiety, [F(2. 96) = 20.68. p < .05].
Group 1 and 2 are significantly elevated when compared to Group 3 in terms of
anxiety. However. Group 1 and 2 are not elevated when compared to each
other. The results to this analysis are provided in Table 11.
Table 11: Means for the Personality Cluster Types on Four Dependent
Cluster 1 Cluster 2 Cluster 3
Anger Expression M 33.02b M 32.54b M 25.283
SD 7.52 SD 10.40 SD 8.62
Trait An2er M 23.92b M 23.85' M 18.723
SD 5.31 SD 7.08 SD 5.81
Family Control M 51.92 M 52.08 M 53.00
SD 10.30 SD 10.14 SD 10.97
Trait Anxiety M 46.08b M 50.77" M 35.12"
SD 8.39 SD 7.04 SD 8.34
(Note:) Different superscripts denote differences at the .05 level of
Cluster T\ oe and Anger Variables:
To fully investigate the relationship between anger expression and
personality\ cluster type. separate one-way ANOVAs were run with Anger-In.
Anger-Out and Anger Control as dependent variables. When Anger-In was
entered a.s a dependent variable with personality cluster type as an
independent variable, the resulting analysis was non-significant for the
relationship between suppressing angry feelings and personality cluster
type. [ F(2. 96) = 2.13. p > .05]. The means for all three groups were
A one-way ANOVA was conducted with Anger-Out as a dependent variable
and personality cluster type as an independent variable. The resulting
analysis was non-significant for the relationship between expressing angry
feelings outwardly and personality cluster type, [F (2, 96) = 2.40. p > .05].
Finally. a one-way ANOVA was conducted with Anger Control as a
dependent variable and personality cluster type as an independent variable.
The resulting analysis was significant for the relationship between Anger
Control and personality cluster type. [ F(2, 96) = 4.23. p < .05]. Group 3
was statistically elevated compared to Group 1: however. Group 3 was
statistically non-distinct compared to Group 2.
Table 12: Means for the Personality Cluster Types on Three Anger Variables
Cluster 1 Cluster 2 Cluster 3
Anger-In M 18.40 M 17.92 M 16.36
SD 3.86 SD 4.09 SD 4.51
Anger-Out M 18.77 M 18.77 M 16.44
SD 3.86 SD 5.99 SD 4.93
Anger Control M 20.27' M 20.15b' M 23.52b
SD 4.21 SD 4.34 SD 5.93
(Note): Different superscripts denote significance at the .05 level
The third analysis was a series of independent linear regression
equations with various measures of the STAXI as dependent variables and
family cohesion and family conflict as independent variables.
When Trait Anger was entered as a dependent variable with family
cohesion and family conflict as independent variables, the regression
equation revealed a significant interaction for both independent
variables. [ F(2. 96) = 11.62. p < .05 ]. Further analysis indicated that
family cohesion had a negative, significant relationship with Trait Anger.
[ Beta = -.2313. t= -2.290. p < .05]. Consequently, family conflict had a
positive. significant relationship with Trait Anger [ Beta = .3271. t =
3.24, p < .05].
Anger-In was entered as a dependent variable with family cohesion and
family conflict entered as independent variables. The resulting regression
equation showed a non-significant relationship with either variable. [ F(2.
96) = 17. p > .05]. Further analysis demonstrated a negative, non-
significant relationship with family cohesion as a independent variable and
Anger-In as a dependent variable. [ Beta = -.0475, t= -.419, p > .05].
Additionally. family conflict displayed a positive, non-significant
relationship with Anger-In as a dependent variable, [ Beta = .0286, t= .252.
p > .05].
Anger-Out was entered as a dependent variable with family cohesion
and family conflict as independent variables. The resulting regression
equation showed a significant relationship with both independent variables
contributing significantly, [ F(2, 96) = 8.13. p < .05]. Further analysis
suggested that family conflict had a positive, significant relationship
with Anger-Out [ Beta = .2597. t = 2.49, p < .05]. Further analysis also
suggested a negative, significant relationship with Anger-Out and family
cohesion [ Beta = -.2258. t = -2.16. p < .05].
Anger Control was entered as a dependent variable with family
cohesion and family conflict as independent variables. The regression
equation suggested a significant relationship with one of the independent
variables contributing significantly. [ F(2, 96) = 4.35. p < .05]. Further
analysis suggests that family conflict had a positive, non-significant
relationship with the Anger Control. [ Beta = .0505. t = .46. p > .05].
Lastly. analysis revealed a positive, significant relationship with family
cohesion and Anger Control. [ Beta = .3158. t = 2.90. p < .05].
Further. the sample was sorted on arrest history so that only those
with a positive arrest history were analyzed. When Trait Anger was entered
as a dependent variable and family cohesion and family conflict were entered
as independent variables, the overall resulting relationship was
significant. [ F(2. 81) = 10.29. p < .05]. Further analysis suggests that
family conflict had a positive, significant relationship with Trait Anger,
[ Beta = .3007, t = 2.79, p < .05]. Family cohesion had a negative,
significant relationship with Trait Anger. [ Beta = -.2767, t = -2.563. p <
A regression equation with Anger-In as a dependent variable and
family cohesion and family conflict as independent variables suggested a
non-significant relationship, [ F(2. 81) = .0459, p > .05]. Examining the
variables independently, family conflict had a positive, non-significant
relationship with Anger-In. [ Beta = .1314. t= .352. p > .05]. Family
cohesion had a negative. non-significant relationship with Anger-In. [ Beta
=.0469. t = -. 126. p > .05.
When Anger-Out was entered as a dependent variable and family
conflict and family cohesion, the resulting regression equation suggested a
significant relationship. [ F(2. 81) = 7.11. p < .05]. Examining the
variables independently, family conflict had a positive, significant
relationship with Anger-Out. [ Beta = .2832. t = 2.534. p < .05]. Family
cohesion had a negative. non-significant relationship with Anger-Out:
however, a trend was still indicated, [ Beta = -.2112, t= -1.89. p = .06].
A regression equation was established with Anger Control as a
dependent variable and family cohesion and family conflict as independent
variables suggested a significant relationship, [ F(2, 81) = 4.7123. p <
.05]. Family conflict seemed to have a positive, non-significant
relationship with Anger Control, [ Beta = .0535, t = .463, p > .05]. Family
cohesion appeared to have a positive, significant relationship with Anger
Control. [ Beta = .3503. t = 3.034. p < .05].
This study was conducted in order to better understand possible
psychopathological personality patterns in adolescents at risk for
violence. In addition, this study sought to better understand the possible
relationships between affective variables, such as anger and anxiety,. and
familial components to these personality patterns in a select sample of
adolescents at risk for violence. Demographic information collected on the
present sample demonstrated that the majority of the sample was African-
American with a previous arrest history. In addition. the majority of the
sample either was raised in a single-family or other non-traditional family
arrangements. such as biological parent and step-parent. For those
children who lived with at least one biological parent. the vast majority of
these parents were employed. Since the majority of the subjects in this
study were African-American. the demographic information suggests a
sampling bias of ethnicity on referrals to alternative schools and
detention center referrals. The extent of this bias on test results is
unknown: however, future research with similar populations may want to
explore in more detail the effects of ethnicity and social class on paper-
and-pencil psychological inventories.
For this study. there were a number of pertinent variables that were
not examined. This variables include questions concerning substance abuse.
gang membership. number of incidences of violence and adolescent parentage.
These variables are important and need to be examined within the population
of adolescents at risk for violence. Future studies of this population
should examine the extent and scope of these variables with these children.
The child's place of residence was not examined in great detail within
the current study. Those children who were not living with a least one
biological parent were placed in an "Other" category. However, within this
category there was a great deal of range. Children could be living with
extended family members. such as aunts or grandparents. or be placed in
foster-care placement. There was not sufficient numbers of children living
in foster-care placement within this sample to conduct a statistical
analysis. However. hypothetical differences between children who live in
extended family situations and those in foster care placements on a number
of affective and familial variables should be examined in a future study.
Within this study. a comparison was conducted between white subjects
and African-American subjects on ten MMPI-A scales (F. 2. 3. 4. 5. 6, 7. 8, 9,
10). Significant differences were found between ethnicity and Scales 6, 7
and 8. Whites in the current sample were found to be higher on Scales 6
(Paranoia) and 7 (Psychasthenia) and African-Americans were found to be
higher on Scale 8 (Schizophrenia). In addition, there was a statistical
trend in whites being higher on Scale 5 (Masculinity-Femininity) and
African-Americans being higher on Scale 10 (Social Introversion). It can
be hypothesized that the white sample elevations on paranoia and
psychasthenia can be the result of coming from a majority culture and being
placed in a situation where they are minorities (ie., detention centers or
alternative schools). The elevation on Schizophrenia for African-
Americans may be explained by subtle cultural differences in language and
socialization. The current differences in MMPI-A scales for ethnic groups
is speculative. This current study suggests that a more detailed analysis
of ethnic differences in different subject populations needs to be
For the FES. t-tests were conducted between the sample means and the
means of a normative sample. The analyses showed that compared to a
normative sample. this sample was significantly lower on family cohesion
and significantly higher on family conflict and control. This findings
demonstrates dysfunctional family characteristics for the sample as a
whole. This finding suggests a more detailed study into the family
characteristics of adolescents at risk for violence utilizing more
variables from the FES.
The first hypothesis stated that the overall sample of adolescents at
risk for violence would be heterogeneous in terms of personality data
gathered from the MMPI-A. This hypothesis was tested utilizing a
hierarchical. agglomerative cluster analysis (Centroid linkage method)
with the following scales of the MMPI-A: F. 2, 3, 4, 5, 6, 7, 8, 9, 10.
Distances between clusters were examined using squared Euclidean
distances. A "stopping technique" formulated by Mojena (1977) and later by
Mojena and Wishart (1980) indicated the presence of a three-cluster
An examination of the three-cluster solution was conducted looking at
the overall MMPI Validity and Clinical Scales. The results were interpreted
based on the scores of a normative sample. Cluster I had no clinical
elevations on any of the validity scales or the clinical scales compared to a
normative sample. However, there were sub-clinical elevations on the F
Scale. Scale 4 (Psychopathic Deviate). and Scale 6 (Paranoia). This is in
direct contrast to Cluster 2. Cluster 2 had clinical elevations (scale
elevation > 65) on the F Scale. Scale 2 (Depression). Scale 6 (Paranoia),
Scale 7 (Psychasthenia). and Scale 8 (Schizophrenia). In addition, there
were sub-clinical elevations on Scale 3 (Hysteria). Scale 4 (Psychopathic
Deviate). and Scale 9 (Hypomania). Lastly, Group 3 had no clinical
elevations on any validity or clinical scales of the MMPI-A. From examining
the results of the means. it appears that the majority of the MMPI-A scales
for Cluster 3 are near the central T-Score mean of 50.
Cluster I had elevated, yet sub-clinical. ( > 60) T-scores on three
Hathaway and Monachesi (1963) "excitatory" scales (F, 4, and 6). For the
purpose of discussion, they were termed the Sub-excitatory group. In
contrast. Cluster 2 had elevated (>65) T-scores on three Hathaway and
Monachesi "excitatory" scales (F. 6 and 8), as well as sub-clinical (>60) T-
scores on two scales (4 and 9). Cluster 2 also had elevated T-scores on two
"inhibitory" scales (2 and 7). They were termed the Mixed group. Finally.
Cluster 3 had no clinical or sub-clinical elevations on any scale. They were
termed the Non-elevated group. Based on the original model suggested by
Hathaway and Monachesi. one would have expected an excitatory group to have
higher T-scores and hypothesized the presence of a purely Inhibitory group.
Since Hathaway and Monachesi (1963) and Williams and Butcher (1989) posited
that excitatory scales would have a greater influence over inhibitory
scales. one would predict that Cluster 2 would have high rates of acting-
out. Since Group 1 had minor elevations on excitatory scales. one would
predict higher rates of acting out than Group 3. which according to the
model, would have low rates of acting out. Acting-out was not measured
directly in this study by examining rates of violent behavior, although it
was examined in-directly by looking at the self-report of anger styles. One
would hypothesize that Group 1 and Group 2 would have higher Trait Anger
(Hostility) and Anger-Out scores compared with Group 3.
In examining the available personality cluster types, the only one
that clearly distinguishes itself is Cluster 2. See Figure 2 in the Results
Section. However, one-way ANOVAs were conducted to verify the separations
of the three personality cluster types.
One limitation of the first analysis is the use of a cluster analysis.
The majority of cluster analysis methods are heuristics which do not involve
a great deal of statistical reasoning (Aldenderer & Blashfield. 1984).
Different clustering techniques (Wards. Average Linkage. Single Linkage.
Centroid. etc.) have been known to generate different solutions to the same
data base (Aldenderer & Blashfield. 1984). While it is true that the
Centroid linkage method generated a three-cluster solution, it may have
been possible to generate a different result utilizing a different
technique. Researchers should be aware of the bias that accompany the
presentation and description of groups created by clustering techniques.
A number of one-way ANOVAs were conducted with a series of dependent
variables anger expression, trait anger, trait anxiety, and family
control) and personality cluster type as the independent variable. The
results indicated a significant relationship between anger expression and
personality cluster type in which Cluster Groups I and 2 were substantially
elevated compared to Group 3. However, the anger expression mean score was
not statistically distinct for Group 1 and 2. The same result occurred with
Trait Anger as a dependent variable and personality cluster type as an
independent variable. Cluster Groups I and 2 were statistically elevated
compared to Group 3. but not statistically elevated when compared to each
other. Trait Anxiety had a statistically significant relationship with
personality cluster type and Groups I and 2 were statistically elevated when
compared with Group 3. However, Groups 1 and 2 were not statistically
distinct from each other. Lastly, Family Control did not demonstrate a
statistically significant relationship with personality cluster type and
upon visual inspection of their means, all three personality cluster groups
were remarkably similar.
Theoretically, it is interesting that Group 1 and 2 were distinct on
personality variables yet similar on affective variables (anger
expression, trait anger, and trait anxiety). If compared to a normative
sample. Group 1 and 2 would be at the 69th and 65th percentile for anger
expression. One would expect that the personality cluster group with the
highest and greatest number of clinical elevations to have the highest anger
expression score. However, for this study this is not the case.
Hypothetically,. this could be explained in that the behavioral
manifestations of anger expression may be different for the two groups.
This would mean that Group I may have more socially acceptable outlets for
their anger than Group 2. This hypothesis could be developed further by
examining the groups in terms of a structured interview designed to gather
information about ways individuals react to situations when angry or upset.
Gathering behavioral information (such as voice characteristics, speech
patterns or eye contact) from each group could contribute more information
about possible distinct manifestations for groups with high levels of anger
Group 3 was relatively similar to Group 1 in terms of personality
presentation but distinct when it came to anger expression. Group 3 was low
on the majority of measures but had a degree of defensiveness as measured by
Scale K. It could be that the low anger expression scores could be the result
of defensiveness. Again additional information, such as information
provided in an interview would be able to examine if this hypothesis is
valid. The interview could tease out information about verbal style or
reactions to angry situations. Another possibility is that Group 3 is
indeed low on anger expression and that other factors present in a given
child's environment (poverty, poor role models, etc.) placed the child in an
unfortunate situation where they came to the attention of legal
Trait Anger was entered as a dependent variable and found to have a
significant relationship with the personality cluster types. Groups I and 2
were distinct from Group 3 but not dissimilar from each other. Trait Anger
is espoused to be a measure of hostility (sustained angry feelings over a
substantial period of time). Hypothetically, it would be interesting if
Group 1 and Group 2 differ in their presentations of hostility.
Specifically, does one group deal with angry feelings in differing, more
socially appropriate ways than the other. Questions arise as to why one
group that readily reveals psychological distress and no group in no
particular distress have the same level of affective responding to paper and
pencil inventories. If this question is asked then the other question would
be that how can two groups so similar in terms of personality variables have
such different levels of affective responding.
To examine the relationship more fully between anger, hostility and
personality cluster type. more analyses were conducted with Anger-In.
Anger-Out and Anger Control. Anger-In was examined to test the Suppressed
Anger Type hypothesis. Analyses were conducted with Anger-Out to test the
Uncontrolled Anger Type hypothesis. Finally. Anger Control was entered to
test the Overcontrolled Anger Type. Both Anger-In and Anger-Out proved to
be non-significant which was surprising given the differences in Trait
Anger and Anger Expression. Anger Control proved to be significantly higher
for Group 3 than for Group 1 or 2. This leads to a theory of Group 3 being an
Overcontrolled Anger type. Group 3 was found to be a more defensive group
than Group 1 or 2 just based on defensiveness (Scale K of the MMPI). For this
theory to be substantiated more. one would expect Group 3 to have higher
Anger Expression scores than it demonstrated. Also, one would have to get
behavioral samples for each of the three groups to test out theories of
Overcontrolled. Uncontrolled and Suppressed Anger Types. One might also
want to get bio-behavioral information as well as observer reliability
before confirming the hypothesis.
Familial levels of control were not found to be distinct among any of
the three personality cluster groups. The lack of variability suggests that
all three groups perceive the same levels of control. The familial
component of control seems to have little relationship to differing levels
of personality presentation. This result is unexpected given the results of
the Veneziano and Veneziano (1992) study which concluded that most of the
children they tested grew up in households high in control. This study
failed to find the same high levels of control within our sample. The sample
in this study appeared to have relatively consistent levels of control which
were not easily distinguishable from each other. One explanation for the
different findings could be subtle sample differences between this sample
and the sample of he Veneziano and Veneziano (1992) study.
Anxiety appeared to have a statistically significant relationship
with personality cluster type. Groups I and 2 had distinct elevations in
long-standing anxious feelings than Group 3. Coupled with the information
already gathered from anger expression and trait anger, the findings would
suggest that Groups 1 and 2 are effectively reactive and the nature of this
reactivit\ is long-standing.
Hathaway and Monachesi termed the "inhibitory" scales the neurotic
scales and predicted that if a child was elevated on anxiety dimensions, the
rates of acting-out would be lower. However, this study contradicts that
finding in that there are high levels of anxiety and anger for both groups.
This suggests a more complex relationship between anxiety, anger and
acting-out than anxiety being inhibitory and anger meaning expressive. The
results of this study suggest that both contribute to acting-out behavior.
More research needs to be conducted in this area to verify this finding.
One can postulate about the long-term nature of this reactivity in
terms of interpersonal relationships or the health of these children. As a
result of this reactivity would Groups I and 2 have a differing course than
Group 3. It would be interesting to undergo a longitudinal study to examine
if indeed those adolescents who indicated long-standing difficulty with
emotional reactivity had a differing course than those adolescents who did
not indicate such difficulties. This would contribute to the general body
of knowledge concerning the relationship between personality variables.
emotional reactivity and overall health.
Groups I and 2 are distinct on personality variables but not distinct
at all on certain affective variables. Group 1 and 3 are not extremely
distinct on personality variables but are extremely distinct on affective
variables. A conservative view would be that personality variables only
contribute to understanding a sub-group of this population but not the
majority of the population as a whole. Affective variables, particularly
those of a long-standing reactive nature, appear to add more to the nature of
understanding these children.
The final analyses were a series of independent, linear regression
equations which sought to examine relationships between anger and familial
variables. Trait Anger was found to have a negative, significant
relationship with family cohesion. Consequently, Trait Anger had a
positive, significant relationship with family conflict. The variable of
Trait Anger is said to be a reflection of Hostility (frequent experiences of
anger). so that as the variable of family cohesion decreases, the
experiences of anger increases. This result is not surprising in light of
the Veneziano and Veneziano (1992) study. One would suggest that good
family cohesion would somehow be a buffer to anger. Additionally, low
amounts of family cohesion could increase feelings of alienation or
inferiority which could indirectly heighten expressions of anger. This is
demonstrated further by family conflict having a positive relationship with
Trait Anger. Conceptually, high amounts of family conflict suggest a family
situation in which at least one member engages in frequent displays of anger
and/ or arguing, which could influence a child by simply modeling these
behaviors. The converse of family cohesion would be in effect here where
high amounts of family conflict influence frequent experiences of anger.
This finding suggests that a good family bond with average amounts of
conflict have an effect on lowering the experiences of anger.
Anger-Out (the outward expression of anger) was found to have a
negative. significant relationship with family cohesion and a positive.
significant relationship with family conflict. This finding reflects the
previous finding of a relationship between Trait Anger and family conflict
and cohesion. High amounts of family conflict increase the self-report of
anger outwardness. This result may reflect a modeling relationship in which
the child views a parent or guardian frequently engage in outward displays
of anger expressiveness and then follows the same pattern. The opposite
would be true of family cohesion in which a good. perceived bond would
moderate the display of anger. Regression results of the relationship
between Anger-Out and Trait Anger with family components of conflict and
cohesion have important treatment implications for ways to lower hostility
and outward displays of anger.
Anger Control also had a positive, significant relationship with the
variable of family cohesion. This relationship supports the notion of
family bondedness as a moderator of anger displays. The higher the
perceived feelings of family connectiveness the higher the feeling of
frequent emotional control. This may appear to be a positive result at
first. However, intense investment of energy in monitoring and preventing
the expression of anger may result in passivity or depression (Spielberger,
1988). We do not know the nature of the relationship between Trait Anger and
Anger Control for this population. Frequent instances of Anger Control
possibly coupled with Hostility according to Hecker and Lunde (1985) would
suggest the presence of the Overcontrolled Anger type. The evidence is
still tentative if the Overcontrolled Anger type was present in the sample:
however, a possible treatment option may be to encourage the healthy display
of anger to populations who believe that they must over-exert control of
There did not appear to be much of a relationship between perceived
levels of conflict and frequent attempts to control emotional expression.
In light of the previous finding of a relationship between family conflict
and overt emotional expression, the non-significance of this result appears
to make sense. The relationship appears to be that family cohesion appears
to moderate emotional expression while family conflict appears to escalate
the expression of emotion.
There did not appear to be much of a relationship between Anger-In and
the variables of family cohesion and family conflict within this study.
Anger-In is the suppression of angry feeling for fear that one does not know
what to do with these feelings. It may be that this study did not investigate
this variable fully within this population. More. comprehensive research
should be conducted to test whether the Suppressed Anger type is a viable
theory within an adolescent sample.
Due to the relatively low numbers of subjects without an arrest
history, the sample was sorted on arrest history so that only those who
indicated having a positive arrest history were analyzed. Those with an
arrest history were examined because of a theoretical postulation that
those with an arrest history would have a tendency to act out more and
support the Uncontrolled Anger type theory of Hecker and Lunde (1985).
However, the resulting relationships all maintained their significance and
direction. Again, family cohesion had a positive relationship with family
cohesion suggesting the link between perceived strong family bonds and
emotional control. These findings appear to have strong treatment
implications for adolescents at risk for aggression, especially those with
an arrest history.
Overall. the findings do confirm previous theoretical models that
adolescents at risk for aggression are a heterogeneous group in terms of
personality and affective presentations. However, the variables of
personality for this population did not appear to be as strong as was
indicated by previous experimenters (Hathaway and Monachesi, 1963:
Williams and Butcher. 1989). There was only a minority of subjects who had
extreme clinical elevations with most subjects only having a few clinical
elevations or no clinical elevations. This finding may reflect differences
between the original MMPI and the MMPI-A. The original MMPI was found to
have numerous criticisms including the test being too long, the adolescent
norms being outdated, and the reading level may have been too high (Archer.
1984, 1987: Williams and Butcher, 1989a. 1989b). However for the majority
of children in this sample that they could comprehend and respond to a
lengthy test protocol and respond in appropriate fashion. This is important
for designing treatment options for this population. The majority of these
adolescents may be suitable for a treatment program that is straight-
forward at least comprehensible at the 6th grade level) and relatively
The findings tentatively suggest the presence of the Uncontrolled and
Overcontrolled Anger type as postulated by Hecker and Lunde (1985).
However, a limitation of this current study is that only self-report
information was used. Further investigations of these Anger types should
include interview information and bio-behavioral measures to verify
utilizing more objective measures the theory of hostility in a population of
adolescents at risk for aggression. However, there was not sufficient
information to ug g:cst the presence of a Suppressed Anger type. Whether
this type can be verified in this population remains to be seen.
The findings as related to affective reactivity suggested that these
adolescen.:s were heterogeneous for emotional presentation as well. There
was evidence of high levels of hostility (Trait Anger) as well as evidence of
overt emouonal expression or stringent monitoring and control of emotional
expression. Future research may want to fully examine the role of emotional
reactivit\ in other behavioral areas such as substance abuse, unprotected
sex, thrill-seeking behavior, suicidality or homicidality.
The findings of this study suggest a treatment model for emotional
reactivit\ in adolescents at risk for aggression and violence. Since
emotional reactivity was found to be a salient variable within this
population, methods to reduce the biological manifestations of emotional
reactivity for this group need to be studied. Hypothetically. one could
study or compare various anxiety and anger reduction methods, such as
progressive muscle relaxation or medication. Other suggestions for
treatment include devising methods of treating elements of dysfunctional
family patterns within this population. A treatment model could be
systematically constructed based on the results of this study. An
examination of the current literature on available treatment methods needs
to be conducted before a treatment model is conceptualized.
Overall, this study suggested that there were differential levels of
anger expression, trait anger and trait anxiety for this population.
Whether or not these affective variables can be treated within a model
remains to be seen. Additionally. it is not known whether the behavior of
these adolescents can be corrected by addressing affective and familial
elements. However. we do know that the rates of adolescents acting out and
being violent is increasing and has been increasing with each subsequent
measurement. A treatment model based on good working hypotheses that can be
evaluated in a component fashion to determine what is effective and what is
ineffective is needed. It is the hope of this study that these treatment
implications can be tested and evaluated and more effective treatment
programs for aggression among adolescents can be found.
Please do not put your name on this form because we want to insure that
your answers are completely anonymous. The information you provide on this
form will be strictly confidential and will not be used against you in any
way. A number will be assigned to you so that feedback on other test results
can be provided.
1) What is your age?
2) What is your current grade in school?
3) What is your ethnic affiliation?
A) African- American B) Latino C) White D) American Indian E) Other
4) Describe your current family?
A) Both parents B) Mother and Stepfather C) Father and Stepmother
D) Mother only E) Father only F) Other-- Please explain below.
5) What is your parents) employment status?
A) Employed B) Unemployed
6) Please indicate whether or not you have an arrest history?
A) Yes B) No
Aldenderfer, M. S. and Blashfield, R. K. (1984).
Clusteranalvsis.BeverlyHills, CA: Sage Publications
Alexander. J. (1973). Defensive and supportive communication in normal and
deviant families. Journal of Consulting and Clinical Psychology, 40,
American Psychiatric Association. (1987). Diagnostic and statistical
manual of mental disorders (3rd ed.. rev.) Washington: Author.
American Psychiatric Association. (1994). Diagnostic and statistical
manual of mental disorders (4th ed.) Washington: Author.
Archer. R. P. (1984). Use of the MMPI with adolescents: A review of salient
issues. Clinical Psychology Review. 4. 241- 251.
Archer. R. P. (1987). Using the MMPI with adolescents. Hillsdale. NJ:
Archer. R. P. (1992). Assessing adolescent psychopathology. Hillsdale.
Bahr, S. (1979). Family determinants and effects of deviance. In Burr, W..
Hill. R., Nye. F. and Reiss. I. (eds.). Contemporary theories about
the family: research- based theories (Vol. 1). Free Press. New York.
Baldus. B. & Tribe. V. (1978). The development of perceptions and
evaluations of social inequality among public school children.
Canadian Review of Sociology and Anthropology. 15. 50- 60.
Bandura. A. (1973). Aggression: A social learning analysis. Englewood
Cliffs. NJ: Prentice- Hall.
Brook. J. S., Brook. D. W., Gordon, A. S., Whiteman, M., & Cohen. P. (1990).
The psychosocial etiology of adolescent drug use: A family
interactional approach. Genetic. Social, and General Psychology
Monographs. 116. 111- 267.
Centers for Disease Control. (1991). Application of principles of
community intervention. Public Health Reports. 106 (3). 244- 247.
Colligan. R. C.. & Offord, K. P. (1989). The aging MMPI: Contemporary norms
for contemporary teenagers. Mayo Clinic Proceedings, 64, 3- 27.
Coombs. R.. & Landsverk, J. (1988). Parenting styles and substance use
during childhood and adolescence. Journal of Marriage and Family,
50. 473- 482.
Dembo. R.. Grandon. G.. LaVoie, L., Schmeidler, J.. & Burgos, W. (1986).
Parents and drugs revisited: Some evidence in support of social
learning theory. Criminology, 24, 85- 104.
Dembo. R.. Williams. L., Getreu. A., Genung, L.. Schneidler, J.. Berry. E..
Wish. E., & LaVoie, L. (1991). A longitudinal study of the
relationships among marijuana/ hashish use, cocaine use, and
delinquency in a cohort of high risk youths. Journal of Drug
Issues. 21. 271, 312.
Dryfoos. J. G. (1990). Adolescents at risk. Oxford: Oxford University
Faunce. E.. & Riskin. J. (1970). Family interaction scales II. Data
analysis and findings. Archives of General Psychiatry. 22, 531- 526.
Ginzberg. E.. Berliner. H., & Ostow, M. (1988). Young people at risk: Is
prevention possible?. Boulder. CO: Westview Press.
Glick. B.. & Goldstein. A. P. (1987). Aggression replacement training.
Journal of Counseling and Development. 65. 356- 362.
Hathawa\. S. R. & Monachesi. E. D. (1963). Adolescent personality and
behavior: MMPI patterns of normal, delinquent, dropout. and other
outcomes. Minnesota: The University of Minnesota Press.
Hawkins. J. & Weiss. J. (1985). The social development model: An
integrated approach to delinquency prevention. Journal of Primary
Prevention. 6. 73- 97.
Hecker. M. H.. & Lunde. D. T. (1985). On the diagnosis and treatment of
chronically hostile individuals. In M. A. Chasrey & R. H. Rosenman.
Anger and hostility in cardiovascular and behavioral disorder [pp.
227- 240]. New York and Hemisphere/ McGraw- Hill.
Hindelang. M. J. (1981). Variations in sex-race-age-specific incidence
rates of offending. American Sociological Review, 46, 461- 474.
Hoffman. J. P. (1993). Exploring the direct and indirect family effects on
adolescent drug use. The Journal of Drug Issues. 23 (3), 535- 557.
Huesmann. L. R., Lefkowitz. M. M., & Efron, L. D. (1978). Sum of MMPI Scales
F. 4, and 9 as a measure of aggression. Journal of Consulting and
Clinical Psychology, 46, 1071- 1078.
Jessor. R.. & Jessor, R. (1977). Problem behavior and psychosocial
development: A longitudinal study of youth. New York: Academic
Kazdin. A. E. (1992). In MMPI- A Content Scales: Assessing psychopathology
in adolescents. ed. by Williams. C. E. (Foreword). Minnesota:
University of Minnesota Press.
Klimidis. S., Minas. I. H., & Ata. A. W. (1992). The PBI- BC: A brief current
form of the parental bonding instrument for adolescent research.
Comprehensive Psychiatry, 33 (6), 374-377.
Klimidis. S., Minas. I. H.. & Ata, A. W. (1992). Construct validation in
adolescents of the brief current form of the parental bonding
instrument. Comprehensive Psychiatry, 33 (6). 378- 383.
Johnson. B.. Wish. E.. & Schneidler. J. (1991). Concentration of delinquent
offending: Serious drug involvement and high delinquency rates.
Journal of Drug Issues. 21. 205- 229.
Johnson. E. H. (1990). The deadly emotions. New York: Praeger.
Johnson. E. H., & Greene. A. F. (1992). The interview method for assessing
anger: Development and validation. In E. H. Johnson, W. D. Gentry, &
S. Julius (Eds.). Personality, elevated blood pressure, and
essential hypertension [pp. 25- 66]. Washington D. C./ Hemisphere
Lewis. D.. Lovely, R.. Yeager, C., & dellaFemina, D. (1989). Towards a
theory of the genesis of violence: A follow- up study of delinquency.
Journal of the American Academy of Child and Adolescent Psychiatry.
28. 431- 436.
Loeber. P.. & Stouthamer-Loeber. M. (1986). Family factors as correlates
and predictors of juvenile conduct problems and delinquency. In M.
Tonry & N. Morris (Eds.). Crime and justice: An annual review of
research (Vol. 7) [pp. 29-150]. Chicago: University of Chicago
Marks. P. A.. Seeman. W., & Haller D. L. (1974). The actuarial use of the
MMPI with adolescents and adults. Baltimore. MD: Williams &
McCann. R. V. (1957). Delinquency: Sickness or sin?. New York: Harper &
McCord. W.. & McCord, J. (1959). Origins of crime: A new evaluation of the
Cambridae- Somerville youth study. New York: Columbia University
Megargee. E. I. (1966). Undercontrolled and overcontrolled personality
types in extreme antisocial aggression. Psychological Monographs.
80. (3). 611.
Megargee. E. 1. (1982). Psychological correlates and determinants of
criminal violence. In M. E. Wolfgang & N. Weiner (Eds.) Criminal
violence. Beverly Hills. CA: Sage.
Mojena. R. (1977). Hierarchical grouping methods and stopping rules- an
evaluation. Computer Journal. (20), 359- 363.
Mojena. R. & Wishart. R. (1980). Stopping rules for Ward's clustering
method. Computer Statistics, (3), 426- 432.
Moos. R. & Moos. B. (1994). Family Environment Scale Manual. Palo Alto. CA:
Consulting Psychologists Press.
Novaco. R. (1975). Anger control: The development and evaluation of an
experimental treatment. Lexington, MA: DC Health/ Lexington Books.
Novaco. R. (1985). In M. A. Chesney & R. H. Rosenman. Anger and hostility in
cardiovascular and behavioral disorder. New York and Hemisphere/
Olson. D. H.. Portner. J., & Lavee, Y. (1985). Family Adaptability and
Cohesion Evaluation Scales (FACES- III). St. Paul, MN: Family Social
Science/ University of Minnesota.
Palmo, A. J.. & Palmo. L. A. (1993). The harmful effects of dysfunctional
family dynamics. In D. Capuzzi & D. R. Gross. Youth at risk: A
resource for counselors, teachers and parents [pp. 43- 69].
Alexandria, VA: American Counseling Association.
Parker. G.. Tupling, H.. & Brown. L. B. (1979). A parental bonding
instrument. British Journal of Medical Psychology, 52, 1- 10.
Patterson. G. (1982). Family coercive processes. Eugene. OR: Castalia
Quay, H. C. (1987). ed. Handbook of juvenile delinquency. New York: John
Wiley & Sons.
Roberts. G. T. (1987). In Handbook of juvenile delinquency. ed. by Quay. H.
C. New York: John Wiley & Sons.
Seigel. M. (1984). Economic deprivation and the quality of parent-child
interactions: A trickle-down framework. Journal of Applied
Developmental Psychology. 5. 127- 144.
Singer. M. (1974). Delinquency and family disciplinary configuration.
Archives of General Psychiatry. 21. 795- 798.
Spielberger, C. D. (1983). State- Trait Anxiety Inventory (STAI) Manual.
Consulting Psychologists Press. Inc.
Spielberger. C. D. (1988). State- Trait Anger Expression Inventory (STAXI)
Manual. Psychological Assessment Resources. Inc.
Spielberger. C. D.. Johnson. E. H.. Russell. S. F.. Crane. R. S., & Wooden.
T. J. (1985). The experience and expression of anger: Construction
and validation of an anger expression scale. In M. A. Chasrey & R. H.
Rosenman Anger and hostility in cardiovascular and behavioral
disorder [pp. 5- 30]. New York: Hemisphere/ McGraw- Hill.
Spivack. G.. Platt, J. J., & Shure, M. B. (1976). The problem- solving
approach to adjustment. San Francisco: Jossey- Bass.
Vedder. C. B. (1954). The youthful offender: Perspective and readings.
New York: Random House.
Veneziano, C.. & Veneziano. L. (1992). A typology of family social
environments for institutionalized juvenile delinquents:
Implications for research and treatment. Journal of Youth and
Adolescence, 21 (5), 593- 607.
Weaver. G. M., & Wootton, R. R. (1992). The use of the MMPI special scales in
the assessment of delinquent personality. Adolescence. 27 (107),
West, D. & Farrington. D. (1973). Who becomes delinquent? London:
Williams. C. L. (1986). MMPI profiles from adolescents: Interpretive
strategies and treatment considerations. Journal of Child and
Adolescent Psychology, 3, 179- 193.
Williams. C. L., & Butcher, J. N. (1989a). An MMPI Study of adolescents: I.
Empirical validity of the standard scales. Psychological
Assessment: A Journal of Consulting and Clinical Psychology, 1 (4),
Williams. C. L., & Butcher, J. N. (1989b). An MMPI Study of Adolescents: II.
Verification and limitations of code type classifications.
Psychological Assessment: A Journal of Consulting and Clinical
Psychologv. 1 (4), 260- 265.
Wilson. H. (1980). Parental supervision: A neglected aspect of
delinquency. British Journal of Criminology, 20, 203- 235.
Winters. K. C.. Weller. C. L., & Meland. J. A. (1993). Extent of drug abuse
among juvenile offenders. The Journal of Drug Issues, 23 (3). 515-
Zillman. D. (1983). Arousal and aggression. In R. G. Geen & E. I.
Donnerstein (Eds.) Aggression: Theoretical and empirical reviews
(Vol. 1). New York: Academic.
Charlton James Coles was born in upstate New York on January 8. 1968.
He was the third child of Cleo and Marian Coles. He grew up in the suburbs of
Atlanta. GA. He attended high school at Benjamin E. Mays High School and
graduated on June of 1986. After high school he attended the University of
Georgia for one academic year. From there he transferred credit hours to
Georgia State University in 1987 and graduated with a bachelor's degree in
psychology in 1989. He enrolled in the clinical and health psychology
doctoral program in 1990 and received a Master of Science degree in 1993.
From September of 1995 to August of 1996. he was enrolled at an internship
program in clinical psychology. He completed internship requirements on
August 31. 1996. After completing his Doctor of Philosophy. Charlton is
planning on continuing a postdoctoral fellowship at the Morehouse School of
Medicine and working on professional licensure requirements.
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Eileen Fennell, Chair
Professor of Clinical and
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Professor of Foundations of
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Distinguished Service Professor of
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Gary Geflf k 17
Associate Professor of Clinical and
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philophy. \.
Assistant Professor of Clinical and
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Assistant Professor of Clinical and
This dissertation was submitted to the Graduate Faculty of the College of Health
Professions and to the Graduate School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
December, 1996 /C- 5 1 /G
Dean, College of Health Professions
Dean, Graduate School
UNIVERSITY OF FLORIDA
I I 1111111 1111II 1111111 III111 1 I
3 1262 08554 6199