 

 

 

THESiS

I
(-r

_ﬁ,o
(i

éJﬂVaoSé

This is to certify that the
dissertation entitled

FIRST ALCOHOL USE AND .THE DEVELOPMENT OF
ANTISOCIAL BEHAVIOR PROBLEMS FROM PRESCHOOL
THROUGH EARLY ADOLESCENCE

presented by
RONI MAYZER

has been accepted towards fulﬁllment
of the requirements for the

Doctoral degree in Psychology & Criminal Justice

 

 

9/“... 5, ,gw/

 

Major Professor’s Signature
i- / 7- ﬂ 7

Date

 

MSU is an Afﬁnnative Action/Equal Opportunity Institution

cu...-.—.------.n-p--r----~-u-.---.- -c---v-o--v-u-.-a--.-..

 

LIBRARY
Michigan State
University

 

 

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c-JCIRC/DateDuepss-sz

FIRST ALCOHOL USE AND THE DEVELOPMENT OF ANTISOCIAL BEHAVIOR
PROBLEMS FROM PRESCHOOL THROUGH EARLY ADOLESCENCE

By

Roni Mayzer

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Psychology and School of Criminal Justice

2004

ABSTRACT

FIRST ALCOHOL USE AND THE DEVELOPMENT OF ANTISOCIAL BEHAVIOR
PROBLEMS FROM PRESCHOOL THROUGH EARLY ADOLESCENCE

By
Roni Mayzer

Despite the fact that early Age of First Drink (AF D) is a robust predictor of
negative adult outcomes, few researchers have traced its occurrence to childhood
antecedents in order to fully delineate the longitudinal pathway leading to as well as from
this risk marker. The current study examined the relationship between early AF D and
antisocial behavior from preschool (ages 3-5) through early adolescence (ages 12-14),
hypothesizing that early drinking is associated with a long-term pattern of aggression and
delinquent behavior. It also attempted to explain the development of these behavior
problems, using risk factors from preschool (temperament, parental psychopathology)
and middle childhood (academic achievement, self-esteem, social problems, family
environment). Participants were 220 male children and their parents in the Michigan
Longitudinal Study. Early AFD was deﬁned as having had a ﬁrst drink by 12-14 years of
age.

Results showed that early AFD was associated with delinquent behavior, not
aggression. Compared to those who never tried alcohol, early drinkers were rated as
more delinquent at ages 3-5, 6-8, and 12-14 - with a temporary convergence at 9-11.
There were also group differences in the developmental trajectories for delinquent
behavior. Particularly interesting was the stronger direct effect of preschool
predispositions on adolescents’ delinquent behavior for the alcohol-initiated, supporting

the idea of hard-continuity in problem behavior for early AFD individuals. In a

simpliﬁed model, children were classiﬁed as high-high, low-low, low-high, or high-low
on each type of antisocial behavior at 3-5 and 12-14, respectively. Early drinking was
disproportionately likely for those classiﬁed as high-high in delinquent behavior (and
disproportionately uncommon for those classiﬁed as low-low).

Risk proﬁles during preschool and middle childhood were most favorable for the
low-low groups, least favorable for the high-high groups, and intermediate for the change
groups (i.e., low-high and high-low). Earlier antisocial behaviors were the best
predictors of later antisocial behaviors. Whereas delinquent behavior was self-
perpetuated, adolescent aggression could also be uniquely explained by personal (e.g.,
self-esteem) and contextual (e.g., family cohesion) characteristics during middle
childhood. In conclusion, there are important reasons to look at aggression and

delinquent behavior as often co-morbid but distinct — conceptually and developmentally.

C0pyright by
RONI MAYZER

2004

ACKNOWLEDGMENTS

I am grateful for the guidance and enthusiasm of my dissertation committee
members: Hiram Fitzgerald (co-chair, Psychology), Christina DeJong (co-chair, Criminal
Justice), Robert Zucker, Cynthia Perez McCluskey, Francisco Villarruel, and William
Davidson. You were always incredibly supportive of my research and multidisciplinary
interests — thanks for believing in me, and for making this dual-major possible. Thank
you to Alexander von Eye, statistics guru; to everyone involved with the Michigan
Longitudinal Study, for all your hard work; to Tanya Manning, best friend extraordinaire,
for sometimes bearing the weight of my world on your shoulders; and to my family, for
your love and understanding.

This research was supported by a Student Award Program grant from the Blue
Cross and Blue Shield of Michigan Foundation and, in part, by grants to R. A. Zucker
and H. E. Fitzgerald from the National Institute on Alcohol Abuse and Alcoholism

(NIAAA grant #R37 AAO7065).

TABLE OF CONTENTS

LIST OF TABLES ................................................................................. viii

LIST OF FIGURES .................................................................................. x
CHAPTER 1

INTRODUCTION .................................................................................... 1

Identiﬁcation of the Problem ............................................................... 1

Conceptual Framework ...................................................................... 1

Purpose of the Study ......................................................................... 7
CHAPTER 2

LITERATURE REVIEW ........................................................................... 10

Early Onset of Alcohol Use ................................................................ 10

Typologies of Alcoholism: An Antisocial Substrate ................................... 13

Developmental Theories on Antisocial Behavior ....................................... 17

Evidence of Continuity in Antisocial Behavior ......................................... 21

Evidence of Discontinuity in Antisocial Behavior ..................................... 25

Risk and Resiliency .......................................................................... 27

Risk Factors ......................................................................... 28

Protective/Promotive Factors ..................................................... 31

Initial Risk Trajectory for Antisocial Behavior ........................................ 38

Temperament ...................................................................... 38

Parental Psychopathology ........................................................ 44

Inﬂuences in Middle Childhood .......................................................... 45

Family Environment .............................................................. 46

Child Characteristics .............................................................. 50

Summary and Hypotheses ................................................................. 52
CHAPTER 3

METHODOLOGY ................................................................................. 6O

Participants ................................................................................. 60

Procedure .................................................................................... 62

Child Measures ............................................................................. 63

Parent Measures ............................................................................ 71

Contextual Measures ....................................................................... 73

Dropped Cases .............................................................................. 77

Missing Data Imputation .................................................................. 78

Analytical Strategy ......................................................................... 82
CHAPTER 4

RESULTS ............................................................................................ 87

Onset of First Drink ........................................................................ 87

vi

Hypothesis 1 .................................................................... 87

Hypothesis 2 .................................................................... 96
Summary ........................................................................ 98
The Development of Antisocial Behavior .............................................. 99
Hypothesis 3 .................................................................... 99
Aggression ................................................................ 103
Delinquent Behavior ..................................................... 109
Hypothesis 4 .................................................................. 114
Predicting Wave 1 Aggression and Delinquent Behavior... ......1 15
Predicting Wave 4 Aggression ........................................ 116
Predicting Wave 4 Delinquent Behavior ............................. 121
Predicting Change Scores ............................................. 123
Controlling for Contemporaneous Behaviors ....................... 128
Changing the Order of Variable Entry ............................... 132
Summary ...................................................................... 133
CHAPTER 5
DISCUSSION ...................................................................................... 136
Onset ofFirst Drink ...................................................................... 137
The Development of Antisocial Behavior ............................................. 144
Limitations and Future Directions ...................................................... 159
In Conclusion .............................................................................. 165
REFERENCES .................................................................................... 168

vii

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 11

Table 12

Table 13

Table 14

Table 15

LIST OF TABLES

Hypothesized Developmental Pathway of Enhanced Risk for

Substance Abuse and Co-morbid Psychopathology ........................... 2
Items on the CBCL Delinquent Behavior and Aggressive

Behavior Subscales ...... ‘ .......................................................... 66
Distribution of Groups Based on Sample Means at Wave 1 and

Wave 4 for Aggression and Delinquent Behavior ............................ 68
Correlations Between Aggression and Delinquent Behavior ............... 69

Distribution of Scores on Psychopathology Dimensions for

Mothers and Fathers .............................................................. 74
Proportion of Missing Cases by Variable .................................. 80-81
Means (and Standard Deviations) by Drinking Onset Group for

Aggression and Delinquent Behavior ........................................... 88
Covariance Matrix for Aggression by First Drink Onset Group ............ 90

Covariance Matrix for Delinquent Behavior by First Drink
Onset Group ......................................................................... 90

Direct and Indirect Effects of Paths from Non-Adj acent Waves in
Structural Equation Models for Aggression and Delinquent Behavior. . . .95

Conﬁgurations for Aggression Group and First Drink Onset ............... 97

Conﬁgurations for Delinquent Behavior Group and First Drink
Onset ................................................................................ 98

Means (and Standard Deviations) for Aggression by Aggression
Group and Delinquent Behavior by Delinquent Behavior Group. . . . . . l 02

Means (and Standard Deviations) by Aggression Group for
Wave 1 Variables .................................................................. 106

Means (and Standard Deviations) by Aggression Group for
Wave 2 Variables .................................................................. 107

viii

Table 16

Table 17

Table 18

Table 19

Table 20

Table 21

Table 22

Table 23

Table 24

Table 25

Table 26

Table 27

Table 28

Means (and Standard Deviations) by Aggression Group for
Wave 3 Variables .................................................................. 108

Means (and Standard Deviations) by Delinquent Behavior Group
for Wave 1 Variables ............................................................. 111

Means (and Standard Deviations) by Delinquent Behavior Group
for Wave 2 Variables ............................................................. 112

Means (and Standard Deviations) by Delinquent Behavior Group
for Wave 3 Variables ............................................................. 113

Hierarchical Regression Results for Aggression and Delinquent
Behavior at Wave 1 ................................................................ 116

Hierarchical Regression Results for Aggression at Wave 4 with
Preschool (Wave 1) and Middle Childhood (Wave 2) Predictors ........ 119

Hierarchical Regression Results for Aggression at Wave 4 with
Preschool (Wave 1) and Middle Childhood (Wave 3) Predictors ........ 120

Hierarchical Regression Results for Delinquent Behavior at
Wave 4 with Preschool (Wave 1) and Middle Childhood
(Wave 2) Predictors .............................................................. 124

Hierarchical Regression Results for Delinquent Behavior at
Wave 4 with Preschool (Wave 1) and Middle Childhood
(Wave 3) Predictors .............................................................. 125

Tolerances for Variables Predicting Wave 4 Aggression and
Delinquent Behavior with Preschool (Wave 1) and Middle
Childhood (Wave 2) Predictors ............................................... 126

Tolerances for Variables Predicting Wave 4 Aggression and
Delinquent Behavior with Preschool (Wave 1) and Middle
Childhood (Wave 3) Predictors ............................................... 127

Regression Results for Aggression and Delinquent Behavior at

Wave 4 with Preschool (Wave 1) and Middle Childhood (Wave 2)
Predictors Including Contemporaneous (Wave 2) Behavior

Problems .......................................................................... 130

Regression Results for Aggression and Delinquent Behavior at

Wave 4 with Preschool (Wave 1) and Middle Childhood (Wave 3)
Predictors Including Contemporaneous (Wave 3) Behavior

Problems .......................................................................... 131

ix

LIST OF FIGURES

Figure l Heuristic model of early etiology for Mofﬁtt’s (1993) life-course

persistent antisocial behavior type .............................................. 18
Figure 2 Heuristic model of the relationship between the initial risk

trajectory and middle childhood factors to antisocial behavior

problems from preschool through early adolescence ......................... 59
Figure 3 Continuity and discontinuity classiﬁcation groups for aggression

and delinquent behavior .......................................................... 67
Figure 4 Stacked (multi-group) model comparing early drinkers to

non-drinkers at 12-14 ............................................................. 83
Figure 5 Mean aggression scores from Waves 1—4 for ﬁrst drink versus

no ﬁrst drink group ................................................................. 89
Figure 6 Mean delinquent behavior scores from Waves 1-4 for ﬁrst drink

versus no ﬁrst drink group ....................................................... 89
Figure 7 Common metric standardized solution for aggression by

drinking onset group .............................................................. 93
Figure 8 Common metric standardized solution for delinquent behavior by

drinking onset group .............................................................. 93

Figure 9 Mean CBCL aggression scores from Wave 1 to Wave 4,
by aggression group ............................................................ 101

Figure 10 Mean CBCL delinquent behavior scores from Wave 1 to Wave 4,
by delinquent behavior group ................................................... 101

CHAPTER 1
INTRODUCTION

Identiﬁcation of the Problem

Although drinking during adolescence is not uncommon (and generally not a
precursor to future abuse or dependence), early onset of alcohol use may be a marker for
more serious developmental outcomes. Indeed, there is a substantial body of literature
demonstrating heightened risk for alcohol-related difficulties with an early age of ﬁrst
drink (Grant & Dawson, 1997; Gruber, DiClemente, Anderson, & Lodico, 1996; McGue,
Iacono, Legrand, Malone, & Elkins, 2001b; Prescott & Kendler, 1999; Schukit & Russell,
1983). An important question in trying to understand the developmental chain of events
underlying this association is whether early drinking onset reﬂects a longitudinal pattern
of behavior traceable from adolescence to younger ages. Considerable evidence exists
that a hard-continuity, risk-aggregation model may describe development for a small
subset of troubled individuals (Fitzgerald, Puttler, Mun, & Zucker, 2000; Zucker,
Fitzgerald, & Moses, 1995; Mofﬁtt, 1993, 1997).
Conceptual Framework

According to the hard-continuity paradigm (detailed in Table 1), risky
characteristics and risky environments from infancy onward funnel the individual into a
pathway of risk aggregation and cumulative disadvantage — leading to a greater
likelihood for early onset of alcohol and other drug use, delinquency, and depression in
adolescence as well as substance abuse and other psychopathology throughout the adult
years (Fitzgerald et al., 2000). Negative outcomes are not sudden but emergent, or

epigenetic, with dynamic structures that gain momentum as they organize over

developmental time (Zucker et al., 1995). Key etiological inﬂuences are already in place

well before drinking onset (Zucker et al., 1995).

Table 1

Hypothesized Developmental Pathway of Enhanced Risk for Substance Abuse and Co-

morbid Psychopathology

 

Age Period

Possible Expression

 

Core Diathesis
Prenatal
Infancy
Preschool

Childhood

Adolescence

Adulthood

Genetically mediated neural deﬁcits

Exposure to alcohol

Difﬁcult temperament

Externalizing behavior problems, Social withdrawal,
Behavior problems, Oppositional behavior, Irnpulsivity,
Social withdrawal, Poor school performance

Earlier onset of alcohol and other drug involvement,
Heavier alcohol and other drug use problems,
Delinquency, Depression

Antisocial personality disorder, Mood disorder,

Substance abuse disorder

 

bloke. Adapted with permission from Fitzgerald, H. E., Puttler, L. I., Mun, E-Y., &
Zucker, R. A. (2000). Prenatal and postnatal exposure to parental alcohol use and abuse.
In J. D. Osofsky & H. E. Fitzgerald (Eds). WAIMHMdbook of infant mental health:

Volume 4. Infant mental health in groups at high risk (pp. 123-159). New York: Wiley.

The concept of developmental contextualism, encompassed by Developmental
Systems Theory (DST; Ford & Lerner, 1992; see also Dixon & Lerner, 1992; Lerner,
1980; Lerner, Castellino, Terry, Villarruel, & McKinney 1995), can be used as a
framework for understanding the evolution of behavior problems. Of particular
importance is the idea of multiple inﬂuences on human development, from many levels
of human ecology (see also Bronfenbrenner, 1977). These inﬂuences range from genes
to culture, all inclusive (Ford & Lerner, 1992). Speciﬁcally,

There is no single cause of the individual’s functioning or development. Neither

within- person variables (e.g., biological or psychological ones), nor interpersonal

variables (such as peer group or personal relations), nor extrapersonal variables

(such as institutional or environmental ones) are sufficient in and of themselves.

(Dixon & Lerner, 1992, p. 37-8)

The fact that the developing individual is embedded in an ecological, multivariate,
multilevel constellation of inﬂuences means that research designs need to look at
contextual risk factors as well as individual predispositions.

There is a core genetic diathesis as well as contextual inﬂuences that predispose
towards negative developmental outcomes, but these are also nested within one another
(Zucker et al., 1995). Risky personal characteristics and behavior typically occur within
a socialization context that reinforces and models those attributes. Children with
behavior problems are frequently reared in a sub-optimal environment characterized by
high parental psychopathology (Mun, Fitzgerald, von Eye, Puttler, & Zucker., 2001;
Mofﬁtt, 1993). Their parents are the most likely to be the least capable of dealing with a

difﬁcult child, and may in fact share many of the same characteristics as their offspring

(e.g., in terms of temperament, personality, or cognitive ability) because of the
disproportionate prevalence of biopsychosocial deﬁcits in disadvantaged homes (Mofﬁtt,
1993)

A second contribution made by DST is the idea of dynamic interaction within and
between all levels of the organization that makes up human life: “the structure or pattern
of relations among these levels of analysis produces the individual’s behaviors; and
changes in the form (the conﬁguration) of these relations produce developmental change”
(Dixon & Lerner, 1992, p. 37-8). Person and context interact in a reciprocal and
transforming pattern of exchange (Ford & Lerner, 1992). Many inﬂuences are naturally
co-occurring anyway, or related to one another because individuals elicit and seek out
certain environmental responses or contexts (Wachs, 1996). Individuals serve as stimuli
in their own development (Lerner, 1985), with transactional feedback from the
interpersonal context that can exacerbate initial tendencies (Mofﬁtt, 1993) — as might be
the case if parents respond with hostility or neglect. Finally, each moment of
development exists not in isolation but as part of an emerging temporal history (e. g.,
Lerner, 1980) shaped by past events and future aspirations.

This person-in-context approach to developmental outcomes with its appreciation
of dynamic interaction (see Ford & Lerner, 1992; Lerner, 1985) emphasizes the fact that
biopsychosocial and environmental characteristics are inter-related in such a way as to
intensify risk. In fact, this aggregation and exacerbation of risk across a variety of
sources serves as a scaffold for development over time — becoming a crystallized

maintenance-structure for continuity in problem behaviors (Zucker et al., 1995).

A key point of this argument is that, although there is always the potential for
change, individuals who are chronically exposed to developmental risk factors ﬁnd it
increasingly difﬁcult to recover. The concept of experiential canalization (Gottlieb,
1991c) refers to hard-continuity that takes place not through the inheritance of species-
speciﬁc acquired characteristics (cf. Waddington, 1942), apropos of the original
conceptualization of “canalization,” but because of immediate inﬂuences. Canalization at
the behavioral level — driven by internal (e.g., genetic) or external (e.g., environmental)
experiences -— can constrain behavior over time as individuals are steered towards or
remain in an epigenetic rut. F rom this vantage point, “canalization is a narrowing of
responsiveness as a consequence of experience” (Gottlieb, 2001, p. 4). Human speech
perception provides one example: “During the ﬁrst year of postnatal life, human infants
are responsive to the universal range of phonemes that occur in all languages. However,
by the end of their ﬁrst year, they are responsive to only the phonemes of their native
language” (Gottlieb, 2001, p. 5). With regard to developmental psychopathology,
canalization refers to the rutting of maladaption over time (e.g., Zucker et al., 1995).

Another implication of the risk-cumulative, nested model (Zucker et al., 1995) is
that early drinking is likely to be associated with other behavior problems and
experiences that characterize the personal and ecological (e.g., family) landscape of
concurrent psychological inﬂuence. Jessor and Jessor (1975, 1977) were among the ﬁrst
to argue — according to what is now known as Problem Behavior Theory — that multiple
indicators of behavioral deviance (e. g., delinquency, substance use, smoking, sexual
activity, school failure) co-vary in adolescence and are so interrelated as to comprise a

syndrome, constellation, or class of conduct. They also suggested that substance use

occurring before age-based population norms could be described as early transition-
marking behavior and predicted from an extant pattern of psychosocial characteristics
called “transition-proneness.” For example, the order in which high school students
began to drink was found to be in parallel (and inversely related) to their ranking on
academic achievement values (J essor & Jessor, 1975, 1977).

Although Jessor and Jessor (1975, 1977) were early in articulating the theme of
co-occurring problem behaviors, the work of other researchers supports their perspective.
For example, Elliott, Huizinga, and Menard (1989) have described “multiple problem
youth” and the joint occurrence of delinquency, substance use, and mental health
problems. Similarly, Dryfoos (1990, 1998) has found that there is a great deal of overlap
among four risk behaviors: substance use, sexual activity, delinquency, and school
problems.

In addition to their co-occurrences, there are likely to be etiological connections
between these problem behaviors. For example, researchers have found that early
substance use increases the likelihood of precocious transitions to adult statuses such as
dropping out of school and parenthood (Krohn, Lizotte, & Perez, 1997; Perez
McCluskey, Krohn, Lizotte, & Rodriguez, 2002), and that — in turn — these precocious
transitions contribute to later substance use (Krohn et al., 1997).

This linkage of multiple problem behaviors may be at the core of what can be
described as an antisocial behavior pathway into early onset of alcohol use. Just as
substance use in adolescence is related to a cascading pattern of other problem behaviors
and substance use during young adulthood (Krohn et al., 1997; Perez McCluskey et al.,

2002), behavior problems from preschool onward may be precipitating factors.

Recalling the hard-continuity model presented in Table 1, the premise here is that early
drinking is but one marker on the developmental path followed by children who
demonstrate persistent difficulties with aggression and delinquent behavior.
Purpose of the Studv

The current study seeks to advance research on Age of First Drink (AFD) and
continuity in antisocial behavior, tying the two together. There are several weaknesses in
the extant literature that will be addressed. First, although numerous researchers have
examined the relative contributions of ecological, social, and individual factors to the
initiation and escalation of alcohol use, considerably less attention has been paid to AFD
as a risk marker. This is true despite the fact that the relationship between AF D and
alcoholic outcomes is quite robust (e.g., Grant & Dawson, 1997). Early onset of regular
drinking patterns is obviously more serious, but identiﬁcation of at-risk individuals
before drinking habits become ingrained would increase the likelihood of success for
intervention and prevention programs. Likewise, the argument can be made that
intervention and prevention programs would beneﬁt from a better understanding of the
risk factors that predispose very young children towards an early AF D. Unfortunately,
studies on AF D have primarily focused on downstream sequelae (e.g., drunkenness or
alcoholic diagnosis) rather than precursors to drinking onset. Is early AF D also a robust
indicator for early behavioral difﬁculties? Proving this would strengthen the argument
that there is an underlying risk trajectory for alcohol problems that can be observed even
before alcohol use becomes part of the behavioral repertoire.

Second and relatedly, although continuity in antisocial behavior has been

associated with greater risk for alcohol-related outcomes, samples are typically ﬁrst

observed during middle childhood or early adolescence. Because behavior is emergent,
there are likely to be important preschool substrates. It is therefore important to look at
behavioral difﬁculties in very young children, and to connect this with later behavior
problems as well as drinking onset.

Third, research on antisocial behavior has focused on continuity with relatively
less attention accorded to discontinuity. Although hard-continuity may be the stronger
trend, behavior is not predetermined. Risk is a process, not a stable trait or state. The
dynamic interaction that exists between individuals and their ecological contexts can
facilitate change. Risk adheres to the principles of probabilistic epigenesis. Epigenesis
is a developmental unfolding of events leading to increased complexity and organization
with the emergence of new structures and functions during the course of ontogeny
(Gottlieb, l99lc; Kitchener, 1978), and the fact that it is probabilistic means that
continuity cannot be assumed. As Werner (1986, p. 18) stated so strikingly, “risk factors
are not black boxes into which one ﬁts children to be neatly labeled and safely stored
away. They are probability statements, the odds of a gamble whose stakes change with

9,

time and place. There are many factors and processes at play, and it is the interplay
between them all — not just risk factors in isolation — that shapes developmental
trajectories. Resilience — the “dynamic process encompassing positive adaptation within
the context of signiﬁcant adversity” (Luthar, Cicchetti, and Becker, 2000b, p. 543) —
matters too. Moreover, this interplay is constant: Person and context are transformed,
minutely or dramatically, by their ongoing relationship with one another. The potential

for change (i.e., plasticity) means that the balance between risk and resiliency isn’t

immutable. Risk can increase or decrease; resiliency can be bolstered or weakened. In

turn, behavioral difﬁculties may remit or escalate during childhood. Prevention and
intervention programs can capitalize on a better understanding of the natural processes
that inhibit or bring about change.

To address these concerns, the current study will look at behavioral difﬁculties
upstream from serious alcohol-speciﬁc outcomes, beginning with preschool. It will also
recognize that — although continuity in aggression and delinquent behavior may be the
stronger trend with respect to predicting drinking onset — discontinuity is also possible.
The purposes of the current study are four-fold. Speciﬁcally, this research hopes to
demonstrate that 1) there are marked differences in antisocial behavior between early
drinkers and non-drinkers by early adolescence that are in evidence from preschool,
onward; 2) high and low antisocial behavior problem groups in preschool and early
adolescence (high-high, low-low, low-high, high-low) constitute a useful classiﬁcation
scheme, with strong continuity (high-high and low-low) as the main characteristic
distinguishing drinking onset from onset avoidance; 3) discontinuity patterns reﬂect a
less common but intermediate risk proﬁle, generally; and 4) antisocial behavior is

inﬂuenced by risk and promotive factors during middle childhood.

CHAPTER 2
LITERATURE REVIEW

Early On_set of Alcohol Use

“Early” drinking onset is typically deﬁned as alcohol use before age 14 (e.g.,
McGue et al., 2001b), but occasionally as young as 10-12 years of age (Gruber et al.,
1996). In general, the preadolescent period seems to be a particularly useful boundary
for distinguishing problematic drinking (i.e., early initiation) from normative drinking
behavior during the teenage years (i.e., during mid-to-late adolescence) —— and for
targeting those at the most elevated risk for alcoholism.

Alcohol consumption and alcohol-related difficulties in adulthood tend to vary
inversely with Age of First Drink (AFD), even in well-functioning university-based
samples (Schukit & Russell, 1983). Among more than 27,000 ever-drinking adults,
Grant and Dawson (1997) found that the rate of lifetime alcohol dependence was four
times higher for those who started to drink by age 14 compared to those who started to
drink after age 20. Other research studies (e. g., McGue et al., 2001b; Prescott &
Kendler, 1999) have reported similar ﬁndings. Early AF D has also been associated with
disinhibitory behaviors and diagnoses related to nicotine, drug use, and psychiatric
disorders (McGue et al., 2001b); as well as injuries, violence, absenteeism from school or
work, and drunk driving (Gruber et al., 1996). It should be noted that although having a
ﬁrst drink can be distinguished from “drinking onset” per se, which implies a regular
pattern of alcohol use, it is still a very robust indication of alcoholic risk.

Like Jessor and Jessor (1975, 1977), some researchers have used the idea of a
“syndrome” of problem behavior propensities to explain the relationship between onset

of drinking and a later alcoholism diagnosis. Speciﬁcally, early onset of drinking and

10

alcoholism may be common symptoms of the same underlying vulnerability (McGue et
al., 2001b; Prescott & Kendler, 1999). In a study of nearly 9000 adult twins, Prescott and
Kendler (1999; see also McGue, Iacono, Legrand, & Elkins, 2001a) found that genetic
factors and a shared family environment explained both. This supports the linkage of
early drinking with risk for later alcoholism as well as the person-in-context approach to
understanding risk.

Early drinking, for example, does not occur in a vacuum but, rather, is related to a
number of personal and contextual factors. Marked differences in the “matrix of social
relationships” have been described for adolescent drinkers versus non-drinkers
(Margulies, Kessler, & Kandel, 1977, p. 898). Modeling and the quality of interpersonal
relationships are both recognized as important correlates of drinking behavior (Margulies,
Kessler, & Kandel, 197 7). Exposure to alcohol use by parents and peers has been
associated with alcohol initiation (Ellickson & Hays, 1991; Felton, Parsons, Ward, Pate,
Saunders, Dowda, & Trost., 1999). Onset of regular drinking can be predicted from the
density of alcoholism within families (Hill & Yuan, 1999). Substance use has been
associated with concurrent levels of family conﬂict as well (Hops, Davis, & Lewin, 1999)
— with initiation less likely among adolescents from highly organized families (Hussong
& Chassin, 1997) and positive parent-child relationships (Cohen, Richardson, & LaBree,
1994). This may reﬂect ampliﬁed tensions/disruptions during adolescence in families
with problem children, or hard-continuity at the ecological level in terms of risky
individuals nested in risky environments over time. In general, although cognitions and

expectancies become more central to the development of regular and continued drinking

11

patterns, onset appears to be mediated by social factors and social processes (Ellickson &
Hays, 1991).

Personal characteristics also play an important role. Personality and temperament,
in particular, have been linked to adolescent substance use. Of relevance are traits such
as high activity level and less positive mood (Wills, DuHamel, & Vaccaro, 1995),
undercontrol (Caspi et al., 1996), high novelty seeking and low harm avoidance
(Cloninger, Sigvardson, & Bohrnan, 1988; Masse & Tremblay, 1997). Wills et al.
(1995) reported a unique contribution of temperamental factors on frequency of
substance use among seventh graders in metropolitan New York. In a longitudinal study
by Caspi et a1. (1996), undercontrolled boys at age three were 2.7 times as likely to be
diagnosed with alcohol dependence by age 21. High novelty-seeking (e.g., impulsive,
excitable, quick-tempered, distractible, curious) and low harm avoidance (e. g.,
uninhibited, carefree, unapprehensive) at age eleven were the most strongly predictive of
alcohol abuse at age 27 in a longitudinal study by Cloninger and colleagues (1988), with
75% of males with the highest scores on these two dimensions at observed risk for
alcohol abuse compared to only 4% of those with the lowest scores. Similarly, using data
from a longitudinal study of 1034 kindergarten students in Montreal, Masse and
Tremblay (1997) found a predictive relationship for teacher ratings of high novelty-
seeking and low harm avoidance at ages 6 and 10 with self-reported drunkenness in the
past 12 months at 10-15 years of age. Based on stability in personality ratings over time,
Masse and Tremblay (1997) even argued that personality must be relatively stable by age
6 and that children at risk for early onset of substance use can be identiﬁed at the time of

school entry.

12

The relationship between temperament and substance use is complex.
Temperament may aggravate the inﬂuence of environmental risk. For example,
researchers have suggested that the relationship between parental alcoholism and
adolescent drinking is mediated by personality or temperament (Chassin, Pillow, Curran,
Molina, & Barrera, 1993). In general, what matters are ongoing transactions between
temperamental characteristics and the social environment (Tarter & Vanyukov, 1994).
According to Tarter (1988), early temperament is part of a vulnerability structure for
substance abuse that depends on intervening factors to activate the pathway into problem
drinking and drug use. In this diathesis-stress model, a certain behavioral diathesis — or
predisposition (e.g., being highly excitable, emotionally intense) — will elicit negative
responses from others (e.g., attenuating the quality of parent-child relations). These
interactions place “stress” on the diathesis and increase the likelihood of a deviant style
of social adjustment. The environment is provocative (McGue, 1997) and exacerbating
(Mofﬁtt, 1993). This, in turn, predisposes towards negative developmental outcomes
related to alcohol use and other problem behaviors.

Typolggies of Alcoholism: An Antisoch Substrate

An early AF D is predictive of later alcohol problems (Grant & Dawson, 1997;
McGue et al., 2001b; Prescott & Kendler, 1999; Schukit & Russell, 1983). However,
alcoholism is a heterogeneous disorder. There are several etiological pathways into
diagnoses of abuse or dependence. Of particular relevance to a developmental
perspective is the distinction between individuals with transient manifestations of
alcoholism and individuals with long-tenn, persistent alcohol problems. Several

researchers (e.g., Zucker, 1987; Zucker et al., 1995; Cloninger, Sigvardsson, & Bohman,

13

1996) have proposed typologies to address this issue. These typologies demonstrate the
relevance of antisocial behavior problems to alcoholism characterized by early onset.

Cloninger et al. (1996) described two alcoholism trajectories in data from the
Stockholm Adoption Study. According to this scheme, Type I alcoholics are those with
both genetic (biological) and environmental (including adoptive) risks, onset after age 25,
and a personality structure characterized by high harm avoidance, low novelty seeking,
and high reward dependence (Cloninger et al., 1996). Type II alcoholics have early onset
and a primarily genetic diathesis that is often associated with a paternal history of severe
alcoholism and criminality. The personality proﬁle for type II alcoholics is the inversed
combination of traits (low harm avoidance, high novelty seeking, low reward
dependence) — traits that are also correlated with antisocial personality disorder
(Cloninger et al., 1996).

Zucker (1987, 1994) has proposed at least four types of alcoholism. Antisocial
alcoholism is deﬁned by early onset of alcohol problems combined with a history of
antisocial behavior (Zucker, 1987). Antisocial alcoholics tend to be male and from the
lower economic stratum, with a family history positive for alcoholism and greater
symptom severity. For these individuals, risk appears to be genetically mediated. From
conduct problems in childhood to social/personal/legal difﬁculties that continue into
adulthood, antisocial alcoholism most strongly reﬂects the hard-continuity model of
problem behavior. Developmentally limited alcoholism, on the other hand, involves
frequent, episodic heavy drinking that is life-stage speciﬁc (Zucker, 1987, 1994).
Speciﬁcally, alcohol use for this subtype may be a normative process of adolescence for a

large subset of individuals - peaking, along with other delinquent behaviors, during

14

young adulthood but followed by abatement to social drinking patterns when adult
responsibilities of work and family present themselves.

Risk factors for developmentally cumulative alcoholism (Zucker, 1987) are also
environmentally-located, with difﬁculties stemming from alcohol use rather than from a
co-morbid psychopathology and with abuse/dependence developing subsequent to heavy
drinking during adolescence. This subtype is the result of adolescent drinking patterns
that do not remit. To point,

In the term developmentally cumulative alcoholism the notion of developmental

cumulation implies that risk is more closely tied to normal, culturally prescribed

processes of drinking and problem drinking than in antisocial alcoholism but that
the additive process has, over the life course, become sufﬁciently cumulative so
that thereafter it has a different trajectory than if it were simply regulated by

normative developmental trends in the culture (Zucker, 1987, p. 67).

In his later work, Zucker (1994) renamed this subtype “Primary Alcoholism III” to reﬂect
more ﬁne distinctions in the etiologic processes leading to an environmentally mediated,
alcohol-speciﬁc (i.e., not driven by co-morbid psychopathology) disorder. Added to the

typology were isolated, single “incidents” of alcohol abuse (Primary Alcoholism I) that

 

often coincide with major life stresses such as divorce, and M symptomatology
(Primary Alcoholism II). A fourth major subtype described by Zucker et al. (1995),
negative affect alcoholism, is linked to anxiety and depression. Unlike the other
alcoholisms, internalizing symptomatology drives the etiologic process.

A hard-continuity model applies to Cloninger’s Type II and Zucker’s Antisocial

Alcoholism. Speciﬁcally, the long-term, persisting, and pervasive problem behaviors

15

characteristic of these alcoholism subtypes help to demonstrate the critical role of co-
morbid antisocial behavior in the development of risk for alcohol abuse or dependence.
Evidence for the existence of at least two alcoholisms comes from empirical studies
showing that Type II alcoholics have more severe symptoms, and that this severity is
predicted from antisociality in childhood and adulthood among Type II, but not Type 1,
men (Zucker, Ellis, & Fitzgerald, 1994). Zucker et a1. (1994) also classiﬁed individuals
by level and severity of child and adult antisocial symptoms into two groups: antisocial
alcoholics (AAs) and non-antisocial alcoholics (NAAs). This dimensional approach
revealed marked group differences. Compared to NAAs, AAs had an earlier onset of
alcohol-related difﬁculties, more symptom variety, lower socioeconomic achievement,
more depression, and a more dense family alcoholism history.

Findings from six longitudinal studies of childhood precursors to adulthood
alcoholism were summarized by Zucker and Gomberg (1986). Strong consistencies in
the extant literature were reported. During childhood and adolescence, individuals who
developed alcohol-related difﬁculties in adulthood had 1) patterns of antisocial behavior,
2) achievement-related diﬁiculties, 3) attenuated interpersonal relationships, 4) homes
with heightened marital conﬂict 5) higher activity levels, 6) inadequate parenting and
lack of parental contact, 7) alcoholic, antisocial, or sexually deviant parents, and 8) Irish
or Italian ethnic heritages. Together, these constituted a network of heightened risk that
was in evidence by childhood — in fact, as early as the preschool years (Zucker et al.,
1995, p. 694). Of key interest (e. g., Zucker, 1987), therefore, is the relationship between

early developmental processes and alcoholic etiology.

16

In addition, Zucker (1987) proposed a hypothetical model linking preschool
antecedents to adult outcomes. Speciﬁcally, intra-individual differences in temperament
and impulsivity combined with family socialization to aggression, negative affect, and
alcohol schemas are the earliest precursors. The effects of this infrastructure are
ampliﬁed by parents who model aggression, negative affect, and alcohol use. This
heightens the risk for poor school performance, involvement with a deviant peer group,
social problems, and problem behavior (including aggression, alcohol use, and other drug
use) in middle childhood and adolescence — with difficulties continuing into adulthood
and culminating in substance abuse (Zucker, 1987). Continuity in antisocial and other
problem behaviors is therefore central to the course of alcoholism for some individuals.
Developmental Theories on Antisocial Behavior

One model for individual variation in antisocial behavior and its timing over the
life course is offered by Terrie Mofﬁtt (1993, 1997). Speciﬁcally, Mofﬁtt suggests that
there are two distinct etiologic patterns: adolescent-limited (AL) and life-course-
persistent (LCP). Although indistinguishable during adolescence, individuals falling into
the two groups have decidedly different behavioral trajectories.

For a small (LCP) minority, juvenile delinquency is just a snapshot within a
lifetime of antisocial activity. Although behavioral manifestations may change (e.g.,
from conduct disorder to crime and substance abuse), there is a constant propensity. This
same argument is made by Gottfredson and Hirschi (1990), who then implicate low self-
control — resulting, in large part, from inadequate parenting — as the key underlying
predisposition. With a slightly different emphasis, Mofﬁtt focuses on neurological

deﬁcits in verbal ability and executive functioning (i.e., control over inattention and

17

impulsivity). LCP individuals are said to be born with neuro-psychological risk for
cognitive deﬁcits, difﬁcult temperament, hyperactivity, and other behavior problems. As
mentioned, their parents may share many of their characteristics and are often the least
capable of dealing with a difﬁcult child (see also Mun et al., 2001 , on the relationship
between parental psychopathology and behavior problems). The relationship between
problem behaviors and parent-child interactions is presumably reciprocal: Behavior
problems can evoke more harsh disciplinary styles from parents, and ineffective
parenting can exacerbate the child’s problem behaviors. Either way, behavioral
difﬁculties are nested in a family environment and that makes later antisocial behavior
more likely. The initial risk trajectory for LCP children is detailed in Figure 1.

Early vulnerabilities are aggravated by the way that others respond to difﬁcult
children — as well as the way that these individuals interpret social cues and select their

environments according to their own personal style (see also Lerner, 1985). This

 
  
     
 

 

  

_ . Cognitive deﬁcits
lNeuro—psychological risk] —9 Difﬁcult temperament
Hyperactivity

 

 

Antisocial Behavior

 

 

 

 

 

 

 

SOCIAL ENVIRONMENT
(family adversity, poor parenting)

Figure 1.
Heuristic model of early etiology for Mofﬁtt’s (1993) life-course persistent antisocial

behavior type.

18

accumulating feedback loop bears negative consequences for future opportunities,
interactions, and the scope of learned behaviors in social and academic settings (Mofﬁtt,
1993). Deﬁciencies in social and academic skills, from an early age, are difﬁcult to
overcome — contributing to developmental continuity and life-course persistent antisocial
behavior.

In contrast, most people engage in some form of delinquent behavior during
adolescence (i.e., the AL group). The impetus for this behavior, however, is thought to
be both proximal (i.e., immediately relevant in time and space) and temporary.
Adolescents seek individuality and “mature status.” Having biologically reached
adulthood but living in a culture that still deﬁnes them as immature, there is a sense of
cognitive discord. LCP peers appear independent and somehow more grown up, so their
behavior is mimicked by the AL group in a “kniﬁng-off childhood apron strings” gesture
(Mofﬁtt, 1993, p. 688). With an emphasis on motivation, reinforcement, and modeling,
theories of learning are therefore thought to lie at the core of AL behavior. Misbehavior
is situational and part of a tirne-limited developmental phase. Termination occurs - when
there is no longer a need to prove one’s “mature status” and with the acquisition of new
stakes in conformity (e. g., work). Never lacking in their cadre of socially-acceptable
behaviors and without cognitive deﬁcits to hinder them, they return to law-abiding
lifestyles with relative ease.

Another developmental perspective especially prominent in criminology is the
informal social control theory of Robert Sampson and John Laub (1993, 1997).
According to this paradigm, there is generally strong continuity across the lifespan or life

trajectory because of stable individual differences and a snowballing of risk (cumulative

19

disadvantage) which together link childhood misbehavior to adult crime. One’s
propensity for deviance is not seen as an invariant trait (e.g., like temperament or
intelligence), however, but rather varies with the ﬂow of personal and interpersonal
attachments.

Informal social controls are constraints against crime and delinquency (Hirschi,
1969; Sampson & Laub, 1993, 1997). During childhood and adolescence, family and
school are the primary socializing (or controlling) agents — with work, family, and the
military as age-salient institutions in adulthood (Sampson & Laub, 1993, 1997).
According to Sampson and Laub (1993, 1997), meaningful bonds can serve as “turning
points” or even termination points for misbehavior (i.e., there is both stability and change
in behavioral trajectories). It is the quality and strength of these social bonds (i.e., social
capital, stakes in conformity) - not just their presence or absence — that matter (Sampson
& Laub, 1993). However, although social bonds at each age explain contemporaneous
behaviors, the accumulation of consequences from past behaviors (“cumulative
continuity”) and the responses that are evoked from others (“interactional continuity”)
can repeatedly attenuate these bonds in what Sampson and Laub (1997) argue is an
inherently social process.

The perspective of Sampson and Laub is consistent with Mofﬁtt’s developmental
theory in terms of recognizing continuity in antisocial behavior. Both acknowledge the
fact that, for some individuals, early problems are both predictive of and contributing
factors to persistent deviant behavior. As mentioned, also shared is the idea of
cumulative consequences (Mofﬁtt, 1993) or cumulative disadvantage (Sampson & Laub,

1997): “Over time, accumulating consequences of the youngster’s personality problems

20

and academic problems prune away the options for change” (Mofﬁtt, 1993, p. 684). One
difference between perspectives is that Mofﬁtt largely implicates neurological deﬁcits as
the initial “cause” of this risk trajectory (although negative environmental inﬂuences can
play an exacerbating role) whereas Sampson and Laub (1993) focus on both parenting
processes and social structure inﬂuences as a core etiology. Most likely it is some
combination of psychological, biological, and social factors that helps to shape the
continuous or LCP pathway. With this stance, theoretical differences can be seen as
complementary rather than contradictory. Unlike Mofﬁtt, however, Sampson and Laub
(1993) also explicitly described inﬂuences that may contribute to desistence among even
the most at-risk individuals. With enough new social capital, informal social control can
disrupt the original risk trajectory.
Evidence of Continuity in Antisocial Bethor

The extant literature on continuity has primarily focused on conduct disorder
(e.g., Fergusson, Lynskey, & Nagin, 1996), crime (e.g., Farrington, 1995; Sampson &
Laub, 1993), and aggression (e.g., Campbell, Ewing, Breaux, & Szumowski, 1986;
Campbell & Ewing, 1990; Campbell, 1994; Campbell, Pierce, March, Ewing, &
Szumowski, 1994; Eron, Huesmann, Dubow, Romanoff, & Yarrnel, 1987). There is
robust support for continuity in antisocial behavior across samples (Robins, 1978). For
example, F ergusson et al. (1996) found that those with conduct problems at age eight
were 16.1 times more likely to meet diagnosis for adolescent conduct disorder or
oppositional deﬁant disorder at 15-16 years old. Reviews of the literature (e.g., Loeber,
1982) suggest that chronic delinquents tend to show persistent and pervasive patterns of

antisocial behavior with onset of delinquency by preadolescence.

21

The Cambridge Study in Delinquent Development (Farrington, 1995) was a
longitudinal survey of 411 inner-city South London boys that relied on information from
multiple sources (youth, parent, teacher, criminal records) to describe the development of
delinquent and criminal behavior. Antisocial behavior at age eight (e.g.,
troublesomeness, dishonesty, aggression) was the best predictor of ofﬁcial juvenile
delinquency at 10-16 and convictions up to age 32 (Farrington, 1995). Twenty-two
percent of the most troublesome boys were later convicted for violence, compared to only
4.9% of those in the least troublesome group (F arrington, 1997).

The longitudinal research of Sampson and Laub (1993) is also quite informative.
Within their follow-up studies on 500 ofﬁcial delinquents and 500 non-delinquents, there
were signiﬁcant relationships between antisocial behavior at ages 10-17 (ofﬁcial
delinquency, unofficial [self, parent, teacher] reports of delinquency, temper tantrums)
and adult misbehavior (arrest, alcohol/drug use, general deviance [e.g., gambling]) up to
thirty years later, at ages 17-45.

Other studies have implicated early aggressive tendencies and hyperactivity in the
etiologic pathway for delinquency. Campbell and colleagues (Campbell et al., 1986;
Campbell & Ewing, 1990; Campbell, 1994; Campbell et al., 1994) described patterns of
continuity in externalizing behavior, from preschool to middle childhood. Two cohorts
of children and controls were identiﬁed. In the ﬁrst cohort of parent-referred children
and a control group with no parental behavior problem complaints (Campbell & Ewing,
1990; Campbell et al., 1986), “hard-to-manage” three year olds showed strong continuity
in aggression and attention deﬁcit disorder. According to maternal report (Campbell et

al., 1986), 50% of those in the problem group continued to meet clinically-signiﬁcant

22

criteria at age six. Teacher-ratings corroborated these differences in externalizing
behavior problems. By age nine (Campbell & Ewing, 1990), children with persistent
problems from 3-6 years of age were more likely to meet criteria for an externalizing
disorder (67%) than children who had improved by age six (29%) and the comparison
group (16%).

In the second cohort (Campbell, 1994; Campbell et al., 1994), overactive,
inattentive, and impulsive preschoolers were identiﬁed by teachers and matched to
classroom controls. Also included in the sample were parent-referred preschoolers.
“Hard-to-manage” children at age four demonstrated problems in activity and impulsivity
that were pervasive across settings (home, preschool, laboratory; Campbell et al., 1994)
and a two-year time interval.

In general, early aggression tends to be the best predictor of later aggression (Eron
et al., 1987; Cairns, Cairns, Neckerrnan, Ferguson, & Gariépy, 1989) with a degree of
stability not much lower than that found in the literature on intelligence (Olweus, 1979).
Patterns of hard-continuity are often evident by preschool and school entry (e. g.,
Campbell, 1994; Campbell et al., 1994; Olweus, 1979), with antecedents traceable to
infancy (Keenan & Shaw, 1994).

Eron et a1. (1987; Huesmann, Eron, Lefkowitz, & Walder 1984) conducted a
study on aggressive behavior over a 22 year time period. In 1960, third graders were
interviewed and rated by peer nomination on aggressive behavior. Follow-up self-report
assessments were conducted in 1970 (age 19) and 1981 (age 30). Signiﬁcant correlations
between aggression at age 8 and antisocial behavior at age 30 were found — at least for

males (Eron et al., 1987). For male children, higher aggression ratings were associated

23

with greater contact with the criminal justice system two decades later (e. g., with a
correlation of r=.24 for convictions, r=.21 for moving violations, and r=.29 for driving
while intoxicated [DWI]) as well as severity of punishments for one’s own child (r=.24).
Structural equation modeling revealed relatively high stability coefﬁcients for aggression
(.50 for males, .35 for females, .46 overall). Finally, regression analyses demonstrated
the high predictability of early aggression for later aggression.

Cairns et a1. (1989) followed 220 fourth graders over six years in an investigation
of changes in aggression from childhood to early adolescence. Moderate correlations for
aggression were found, especially for males, from fourth grade and ninth grade in
teacher-ratings (r=.45 for males, r=.33 for females) and self-report (r=.36 for males, r=. 13
for females). Of an array of variables from the fourth grade, the best predictor of eighth
grade aggression was fourth grade aggression.

Interestingly, individual differences in the persistence of aggression may exist in
infancy. Shaw, Keenan, and Vondra (1994) documented the existence of developmental
precursors to early externalizing behavior problems, including child non-compliance at
18 months of age, in infants from low-income families. Stability from 18 to 24 months
was strongest for those with pervasive aggression (i.e., more than one type of aggression
exhibited in more than two situations) compared to those in the non-pervasive group —
with correlations ranging from .57 to .71 versus .05 to .32 (Keenan & Shaw, 1994).
Although Keenan and Shaw (1994) found few sex differences, other researchers (e. g.,
Cummings, Iannoti, and Zahn-Waxler, 1989) have reported strong stability in physical
aggression from age two to age ﬁve for males (F .76) but more moderate continuity for

females (r=.36).

24

Evidence of Discontinuity in AntisocﬁrkBerior

A wealth of studies have focused on continuity in antisocial behavior over the
course of development, but few have developed models to explain escalating or abating
behavioral difﬁculties. Most of the research on change patterns has compared early
onset, persistent offending to late onset, transitory offending (as discussed by Fergusson
et al., 1996). Alternately, researchers have attempted to delineate distinct developmental
trajectories that take other change patterns into account, but usually without attempting to
determine their underlying causal processes (e.g., Nagin & Tremblay, 1999). Fergusson
et a1. (1996) avoid both these limitations but begin to analyze disruptive behavior in
middle childhood, not earlier. Ackerman, Brown, and Izard (2003) covered an even
more circumscribed developmental time period: from ﬁrst grade (7 years old) to third
grade (9 years old).

Nagin and Tremblay (1999) identiﬁed a four-fold typology based on teacher-
ratings for three different externalizing problems: physical aggression, opposition, and
hyperactivity from ages 6 to 15. The four behavioral trajectories were Chronic, High
Level Near Desister, Moderate Level Desister, and Low. Less than 5% of the sample was
characterized as Chronic, with consistently high problem behavior scores.
Approximately 20-30% were classiﬁed as High Level Near Desisters due to initially high
levels of problem behavior that diminished (but didn’t disappear) with age. The majority
(50%) were Moderate Level Desisters, with modest levels of problem behavior that
largely subsided by 10-12 years of age. Problem behaviors were rare for the 10-25% who
fell into the Low group. The chronic trajectory for aggression and opposition led to a

greater risk of juvenile delinquency than did the chronic trajectory for hyperactivity.

25

In one of the few studies to include an escalation pattern (F ergusson, Lynskey, &
Horwood, 1996), researchers identiﬁed factors that distinguished between no onset, early
onset followed by remission, late onset, and persistent antisocial behavior groups.
Discontinuous pathways were associated with intermediate levels of risk. Risk factors
included social disadvantage, family dysfunction, attention deﬁcit behaviors, lower IQ,
poor school achievement, and low self-esteem. Children with no conduct problems at 7-9
and 14-16 were those with the most favorable proﬁles, children with persistent conduct
problems were those with the least favorable proﬁles, and children with remitting or late
onset patterns typically had proﬁles in-between these two extremes.

Finally, Ackerman et a1. (2003) sought to explain continuity and discontinuity in
teacher-rated externalizing behavior for children from impoverished homes. Children
were classiﬁed into one of four groups based on the presence or absence of high levels of
symptomatology in the ﬁrst and third grades: persistent problem (i.e., high at both
grades), irnprover (i.e., high only in ﬁrst grade), new problem (i.e., high only in third
grade) and unproblematic (i.e., low at both grades). Child, parent, and family variables
were explored as distinguishing characteristics. Unproblematic individuals had lower
levels of impulsivity compared to everyone else, and less parental maladjustment
compared to those with high levels of externalizing behaviors in the ﬁrst grade (i.e.,
improvers and persistent problems). Vocabulary scores were lowest for those with
problems in the third grade (persistent and new problem groups). Harsh punishment was
higher for the persistent problem group than for those without problems in the third grade
(i.e., improvers and the unproblematic group); the new problem group was at

intermediate risk. Some support was therefore found for the idea that verbal ability and

26

harsh parenting (along with family instability) were related to changes in externalizing
during the early school years - but with the emergence of new problems in particular.
Mam Resilienay

The topic of risk and resiliency in human development is one of the most
interesting and popular in the ﬁeld today. Moreover, as a framework for understanding
developmental processes, it is central to this discourse on continuity versus escalation or
abatement. Risk can be conceptualized as an internal or external adversity that increases
the likelihood of undesirable developmental outcomes (i.e., the probability of harm).
According to Luthar, Cicchetti, and Becker (2000b, p. 543), “resilience refers to a
dynamic process encompassing positive adaptation within the context of signiﬁcant
adversity.” To be labelled “resilient”, an individual must be exposed to “signiﬁcant
threat or severe adversity” and demonstrate positive adaptation despite that exposure
(Luthar et al., 2000b, p. 543; Masten, Best, & Garmezy, 1990; Masten & Coatsworth,
1998). Resiliency has often been used synonymously with competence (Garmezy,
Masten, & Tellegen, 1984; Luthar, 1993; Masten & Coatsworth, 1998; Masten, Hubbard,
Gest, Tellegen, Garmezy, & Ramirez, 1999; Sameroff, 2000) and “stress resistance”
(Garmezy et al., 1984; Masten et al., 1990), as well as with “buffers” and “protective
factors.”

Although discussions of risk have been historically ubiquitous, resiliency is a
relatively new concept that nonetheless has enjoyed much attention in recent years. It
complements studies of risk, because it is well known that not all individuals exposed to
adversity will suffer from the “push” towards maladaptive outcomes. For example, not

all children with difﬁcult temperament, neurological deﬁcits, or physical handicaps; and

27

not all children exposed to poverty, parental alcoholism, parental psychopathology,
domestic violence, violent neighborhoods, and so forth will manifest problematic
cognitive, social, physical, or emotional impairments. What differentiates those who do
from those who don’t? Studies of resiliency hope to clarify how and for whom risks do
and don’t exert their predicted effects on development. The theoretical and empirical
literature on risk and resilience is therefore a critical starting point for a comprehensive
understanding of individual differences in developmental trajectories.

Risk Factors. Risk factors in development were recognized even before
psychology became an established intellectual discipline. For example, Rutter (1985, p.
598) references the work of an early predecessor to Freud: Nearly 200 years ago, Pinel
wrote about the psychiatric risks associated with unexpected reverses or adverse
circumstances, and it is reported that his initial question to newly admitted psychiatric
patients was: “Have you suffered vexation, grief, or reverse of fortune?” Interestingly,
Pinel is sometimes referred to as the father of scientiﬁc psychiatry (Hothersall, 1995).
The idea of measuring risk actually originated with maritime insurance; merchants and
ship owners would negotiate a payment based on the risk of a disaster at sea (Werner,
1986). There is an important message carried over from these commercial bargains,
which is just as relevant for understanding the epigenesis of human behavior — namely,
that the “identiﬁcation of risk factors is an exercise in estimating probabilities”
(Sameroff, 2000, p. 7).

Only in recent decades has the concept of risk really come into the limelight of
scholarly inquiry, however (Rutter, 1985). It ﬁrst gained prominence in epidemiology

and preventative medicine (Werner, 1986). Researchers were interested in understanding

28

susceptibility to cardiovascular disease (e. g., in the Framingharn Heart Study; see
Sameroff, 2000) and other maladies. The disease model, applied to psychology, led to an
initial perspective of “reproductive casualty” in which risk was viewed as internal or
extending from the birth process (Sarneroff & Seifer, 1983). Eventually, “caretaking
casualty” was added to the risk model. Studies of infants and young children exposed to
pre-, peri-, and post-natal risks (e.g., chromosomal disorders, nutritional deprivation,
infections, accidents, illiteracy, poverty) emerged during the late 1950’s and 1960’s
(Werner, 1986). Because these forays examined constitutional and environmental
inﬂuences independently, ﬁndings lacked predictive power (Werner, 1986).

There is now general consensus among developmental psychologists that a more
ecological “person-in-context,” dynamic interaction approach is needed to address the
complexities of human development. Developmental Systems Theory (DST) is a guiding
paradigm in this respect. Within and across levels of inﬂuence, it is important to
recognize that there is usually not a single factor but a set of factors that contributes to
outcome (Sameroff, 2000). Given equiﬁnality in developmental trajectories, there may be
a different combination of variables at work for each individual (Sameroff, 2000) despite
the fact that outcomes are quite similar. Also, risk factors tend to co-occur (Masten, Best,
& Garmezy, 1990; Mofﬁtt, 1993; Zucker et al., 1995) and aren’t random (Rutter, 1987).
To varying degrees, these social forces shape the human ecology. In addition, individuals
select and elicit their own environments in active and evocative ways (Ford & Lerner,
1992; Lerner, 1985; Rutter, 1997).

Not surprisingly, accumulated risk appears to be most detrimental — as measured

by the number of risk factors to which individuals are exposed (Ackerman, Izard, Schoff,

29

Youngstrom, & Kogos., 1999; Barocas, Seifer, & Sameroff, 1985; Bry, McKeon, &
Pandina, 1982; Newcomb, Maddahian, & Bentler, 1986; Sameroff, 2000; Werner, 1986,
1989; Werner & Smith, 1982). For example, Bry and colleagues (1982) and Newcomb et
a1. (1986) found that risk load was the best predictor of drug use among high school
students. Of the children in Werner and Smith’s (1982) longitudinal study on the island
of Kauai, approximately one-third experienced chronic poverty and prenatal or congenital
insults and were raised by uneducated mothers in troubled families. Among those with
four or more risk factors present by age two, three-quarters “developed either serious
leaming or behavior problems by age 10, or had delinquency records and/or serious
mental health problems by age 18” (Werner, 1986, p. 13; Werner, 1989). With data
from the Rochester Longitudinal Study, Sameroff (2000) found that the number of risk
factors was the crucial measurement in predicting psychiatric disorder. It is worth
mentioning here that Sameroff (2000; Sameroff & Seifer, 1983; Barocas et al., 1985) is
one of the most outspoken advocates of looking at the quantity — rather than the quality or
exact type -— of risk, especially in ecologically oriented models.

Yet, given the complexities of understanding risk, there is a growing appreciation
of the processes leading to negative developmental outcomes (Rutter, 1987). Rather
than just identifying risk factors, researchers have sought an understanding of when and
how risks impinge upon human adaptation. For example, Rutter (1985) has looked at
family discord (which mediates parental psychopathology; see also Barocas et al., 1985);
parental loss (which exerts its effect through quality of care and family functioning); and
difﬁcult temperament (eliciting more parental criticism and hostility in already harsh

contexts; see also Hetherington, 1989), among other risks.

30

Risk factor studies are among the most prevalent in the literature on
developmental psychology and psychopathology. In fact, it is fair to say that the word
“risk” is ubiquitous therein. Apart from longitudinal studies focused on development in
general, a search for risk factors has been initiated with regard to a plethora of speciﬁc
outcomes of concern to public health and psychiatric teams. This includes alcohol
problems (Ellis, Zucker, & Fitzgerald, 1997; Wong, Zucker, Puttler, & Fitzgerald, 1999;
Zucker, 1994), binge drinking (Schulenberg, Wadsworth, O’Malley, Bachman, &
Johnston, 1996), drug use (Newcomb et al., 1986), depression (Petersen et al., 1993), and
problem behavior (Ackerman et al., 1999).

Protective/Promotive Factors. Not all individuals exposed to risk will
demonstrate developmental failures, however. In fact, many don’t. Is there a self-
righting tendency, located internally or in the care-taking environment, which maintains
adaptive behavior (canalized for the species, in the traditional sense of the word)? Who
are the winners, losers, and survivors of adversity (Hetherington, 1989)? For many
years, Garmezy and colleagues studied the effects of parental schizophrenia — ﬁnding that
a large number of offspring evidenced negligible signs of pathology and incompetence
(for an overview, see Garmezy, Masten, & Tellegen, 1984; Masten et al., 1990). Interest
in why some individuals thrive when others whither led to the birth of studies on
resiliency. From these early beginnings -- rooted deeply in the ﬁeld of developmental
psychopathology - the resiliency literature has burgeoned and diversiﬁed. Emmy
Wemer’s (1986; Werner & Smith, 1982) incredibly multifaceted investigations of risk
and resiliency among children in Kauai have often been credited for this diversiﬁcation

(Luthar et al., 2000b).

31

Early work often focused on individual characteristics thought to provide an
internal buffer against environmental adversity (Garmezy et al., 1984; Luthar et al.,
2000b; Rutter, 1985). This is consistent perhaps with organismic ideas about the nature of
development. More speciﬁcally, the dominant idea was that of “invulnerable” children
who were “so constitutionally tough that they could not give way under the pressures of
stress and adversity” (Rutter, 1985, p. 599; Luthar et al., 2000b). However, the
contextualizing of developmental psychology in the 1970’s led to recognition of
ecological resiliency. As with risk, resilience can be both biological/ genetic and
social/experiential (Rutter, 1985). Nature and nurture both matter; in fact, they are
enmeshed. ‘

Intelligence, personal competencies, and support in the caregiving environment
have been identiﬁed as primary proximate buffers against maladaptation (Egeland,
Carlson, & Sroufe, 1993; Masten & Coatsworth, 1998; Masten et al., 1999; Werner,
1986; Werner & Smith, 1982; Wyman, Cowen, Work, Hoyt-Meyers, Magnus, & F agen,
1999). Of the high-risk children in Wemer’s study, one of four “escaped the ill effects of
such multiple risks and developed into stable, mature, and competent young adults”
(Werner, 1986, p. 13; Werner, 1989). Resilient children tended to be ﬁrst-bom, well-
tempered infants who “met the world on their own terms” as toddlers, had many
hobbies/interests and effectively maximized their communication/prob]em-solving skills
during middle childhood, believed in their own self-efﬁcacy and self-worth during
adolescence, and were achievement-oriented (Werner, 1986, p. 14; Werner, 1989).

Resilient children also had a strong bond to a primary caretaker during infancy (Werner,

32

1986, 1989). This was often a parental substitute, such as a grandparent or older sibling
(Werner, 1989).

Similar reports are found elsewhere. In general, commonly identiﬁed resiliency
factors include having a close relationship with a competent adult, healthy self-esteem
and self-efﬁcacy, positive experiences at school, and a good-natured disposition (Masten
et al., 1990; Rutter, 1985, 1987). These buffers are also highly interrelated. For example,
individuals’ academic success can lead to “enhanced self-esteem and a feeling of self-
efﬁcacy, enabling them to cope more successfully with the subsequent life challenges and
adaptations” (Rutter, 1985, p. 604). Distal factors appear to work through more
proximate contexts — for example, with exosystem adversities (e. g., living in a violent
community) eroding the quality of functioning in the microsystem (e. g., the family)
(Richters & Martinez, 1993).

As with studies of risk and in recognition of probabilistic epigenesis, the
resiliency literature is also moving away from static conceptualizations of developmental
inﬂuence. Scholarly interest has evolved from discussions of “invulnerable” children
(i.e., absolute protection) to “resilient” children (i.e., relative protection) for whom risk
and protection can ﬂuctuate over the course of development (Luthar et al., 2000b; Rutter,
1985). This is consistent with Lerner’s (1980, p. 781) more global assertion that
a focus on the process and, particularly, on the process involved in the changing relations
between individuals and their contexts, is at the cutting edge of contemporary
developmental theory and, as such, is the prominent conceptual frame for research in the

study of human development.

33

Resiliency is a process, not a trait (Egeland et al., 1993; Luthar et al., 2000b;
Rutter, 1985). One’s resilience can change over time, may operate in behavioral chain
reactions, and may be dependent on the timing of risks throughout life course. For
example, parents’ divorce experienced during infancy is not the same as divorce
experienced during childhood, or adolescence, or young adulthood (Rutter, 1985). The
social environment also expands (e.g., Coie & Jacobs, 1993; Dekovic, 1999; Masten &
Coatsworth, 1998) as individuals become involved in institutions such as school and
work. Therefore, although initial scientiﬁc forays sought to identify protective factors,
researchers are increasingly interested in understanding how or when these factors exert
an effect (Luthar et al., 2000b; Gest, Neemann, Hubbard, Masten, & Tellegen, 1993), and
how or when these factors are acquired in the ﬁrst place. This parallels recent trends in
the risk literature. As researchers interested in the how’s of development (e.g., Anastasi,
1958) — not just the what’s - we ask: “Is it chance, the spin of the roulette wheel of life,
or did prior circumstances, occurrences, or actions serve to bring about this desirable
state of affairs?” (Rutter, 1987, p. 316). Looking at stability and change in multiple
levels over time also is consistent with life-span theories and longitudinal methods for
analyzing developmental phenomena (Luthar et al., 2000b; Staudinger, Marsiske, &
Baltes, 1993; von Eye & Schuster, 2000).

Because the ﬁeld is relatively new, there are still major disagreements about how
best to measure resiliency. The most controversial issue is whether to look for main
effects or interactions. Early conceptualizations referred to “protective” factors in
interaction terms — that is, as moderators (see Baron & Kenny, 1986) of the relationship

between risk and outcome. Speciﬁcally, these implied that resiliency modiﬁes (i.e.,

34

nulliﬁes or lessens) the effect that adversity would otherwise have on development (see
Luthar et al., 2000b; Rutter, 1985; Garmezy et al., 1984; Masten et al., 1990; Rutter,
1987). In other words, protective factors should be catalytic (Rutter, 1987). Interactions
are still the dominant focus of inquiries — as some argue they should be, by deﬁnition
(above; see also Roosa, 2000; Rutter, 1987). Although interaction effects tend to be
small and unstable (Luthar et al., 2000b), this is attributed to homogeneity on the risk
dimension within samples (e.g., Dekovic, 1999; Roosa, 2000).

The main effects of protective “factors” (or processes) have also been of interest
to some. Regardless of risk level, some inﬂuences may encourage more positive
adaptation. An interesting question is whether resiliency only helps those exposed to
adversity, helps those more than non-exposed others, or helps everyone regardless of risk
status (e. g., Luthar et al., 2000b; see also methodological issues as reviewed by von Eye
& Schuster, 2000). Some researchers have suggested, for example, that “resilience is an
integral part of normal development that every child must achieve” to the extent that
“resilience is a name for the capacity of the child to meet a challenge and use it for
psychological growth” (Baldwin, Baldwin, Kasser, Zax, Sameroff, & Seifer., 1993, p.
743). To clarify the semantic confusion, Sameroff (2000) suggested that modifying
factors be distinguished from factors that beneﬁt all children (i.e., that lie at the positive
pole of the risk dimension) by referring to the latter as “promotive” instead.

Risk and promotive factors would therefore be opposite ends of the same
spectrum, with protective factors having an interactive effect. This makes intuitive sense,
and may relieve some of the tension between researchers talking at each other rather than

to each other — using the same terminology but answering different questions. Other

35

solutions, such as adding elaborated labels to interactive effects (e.g., “protective-
enhancing”; Luthar, 1993), have been suggested but are no less confusing.

Generalizability across domains of resilience is another methodological concern.
Luthar et al. (2000b, p. 548) provided the most reasonable position on this issue: “In
studies of resilience, we believe that there should undoubtedly be some uniformity across
theoretically similar adjustment domains, but not across those that are conceptually
distinct” given that “unevenness in functioning across domains is a common occurrence
in the process of ontogenesis.” Researchers need to be speciﬁc in discussing outcomes
of relevance to which their data apply -— for example, referring to “educational

,, ‘6

resilience, emotional resilience,” “behavioral resilience,” and so forth (Luthar et al.,
2000b). Because adversity is often intrinsic to the deﬁnition of resiliency, care must be
taken to identify exactly what individuals are expected to be resilient against, and how
that would reveal itself. Social competency is not always associated with better
emotional adjustment. In fact, otherwise resilient individuals sometimes report more
emotional distress over time (Luthar, Doemberger, & Zigler, 1993).

Finally, how positive do adaptations need to be to qualify as resilience? Luthar et
al. (2000b) suggest that maintaining near-average functioning is sufﬁcient evidence when
extreme and severe stressors are at play, but that superior functioning may need to be
witnessed when the stressors are of a more moderate nature. However, superior
functioning may be asking too much (Robinson, 2000). A more appropriate standard
may be “trajectories that are ‘unexpectedly positive,’” with speciﬁc cut-offs determined

by the researcher (Luthar, Cicchetti, & Becker, 2000a, p. 574).

It is obvious that risk and resiliency are intertwined, by deﬁnition. Many risk

36

factors are distal in nature — exerting their effect through more intimate milieus.
Environmental (e.g., low socioeconomic status, social disorganization) or even
interpersonal (e. g., parental psychopathology) risks may be offset by proximal
“protections” such as having a supportive family or doing well in school. Not everyone
in a risk category will actually experience the same amount of adversity. For example,
high-risk children with easy temperaments may elicit more positive attention from
parents, meaning that they are potentially exposed to less adversity (Masten et al., 1990).
Children with difﬁcult temperaments may become the target of parental hostility,
criticism, and annoyance - possibly exacerbating an already aversive home life (Rutter,
1985, 1987). Consequently, “it cannot be assumed that any environmental risk factor —
wherever it falls on the distal-proximal continuum — carries equivalent levels of risk to all
children exposed to it” (Luthar, 1993, p. 443-4).

Researchers have even debated whether or not risk and protection/promotion are
actually distinct constructs, or opposite poles of the same phenomenon. Sometimes
empirical differences between risk and protective/promotive factors become blurred, to
the extent that reverse coding and categorizing would also make sense (e. g., see Costa,

J essor, & Turbin, 1999). However, it is not necessarily the case that resilience can be
equated with “lack of risk.” Having a coach who encourages positive self-esteem and
academic pursuits in the face of other adversity (e. g., parental psychopathology) is
equivalent to the presence of an inﬂuence; the absence of a caring and supportive adult
outside the family isn’t always as meaningful, especially when parent-child relationships
aren’t attenuated in any way (e.g., in poor, but cohesive, families). Even if assets are the

counterparts of risk (Gilgun, 2000), the term “protection” or “promotion” may be

37

important if only to explicitly acknowledge that main effects come from the positive end
of the variable (Rutter, 1987, p. 319) or that the two extremes are not equally potent.

“For instance, perhaps high religiosity is more related to low drug use than low religiosity
is associated with high drug use” (Newcomb & Felix-Ortiz, 1992, p. 281).

Initial Risk Traiectory for Antisocial Behavior

A distinction needs to be made between early predispositions (herein called the
initial “risk trajectory”) and factors within the personal and social surround that are
subsequently involved in risk or promotion. The former are assumed to be foundational;
they may interplay with life experience but also constitute the backdrop against which
later development can be understood. Temperament and lifetime parental
psychopathology (i.e., characteristics of parents that are long-standing, assumed to
predate the parenting experience, and possibly genetic) are two examples. Both also have
strong links to child behavior problems, as demonstrated in the extant literature, and are
therefore highly relevant to the present discourse.

Temperament. As a means of studying individual differences in risk
predisposition during the early years of child development (Rothbart, 1981), temperament
offers one framework for research. Temperament theory has been used in academic
inquiries and intervention (e. g., nursing) in terms of providing parents and investigators
with an appreciation of the unique characteristics of each child and insight into
expectations about child-rearing, specialized hospital care for children, the etiology of
problem behaviors, developmental trajectories, and maternal mental health/fatigue

(Tomlinson, Harbaugh, & Anderson, 1996).

38

Theories of temperament are both numerous and diverse. Unfortunately, different
conceptualizations abound about what “temperament” really means. Researchers have
focused on models with a variety of different theoretical emphases, including behavioral
style and the “how of behavior” (Chess & Thomas, 1989); stable, heritable, and
evolutionary-adaptive aspects of personality (Buss, 1991); and the psychobiological
nature of reactivity and self-regulation (Rothbart, 1981), among others.

The ﬁrst attempt to describe temperament is generally attributed to Thomas,
Chess, and colleagues through the New York Longitudinal Study (N YLS). Analyzing
parent interviews, these researchers ultimately postulated nine dimensions of
temperament: activity level, rhythmicity, approach, adaptability, threshold, intensity,
attention span, distractibility, and persistence (Chess & Thomas, 1989). Where children
fall on these nine dimensions is presumed to reﬂect a threshold characteristic that crosses
all emotional and behavioral processes across situational contexts: “the typical way in
which a child pursues activities and responds to the demands of the environment” (Chess
& Thomas, 1989, p. 166). Parenting responses can reinforce certain temperamental
tendencies, such that stability may even be the result of environmental inﬂuences rather
than a static psychological attribute (Chess & Thomas, 1989). In other words, there is
reciprocal, or dynamic, interaction between child temperament and its context.
Moreover, the environment evolves as the individual matures: “the changing context of
the child’s or adult’s behavior and the emergence of new forms of behavior at later
developmental periods may give the same characteristic very different forms of
expression” (Chess & Thomas, 1989). Tendencies or dispositions may remain quite

stable, but the ways in which they are manifested can vary with situation, maturation, and

39

interplay within changing social relationships (Strelau & Angleitner, 1991). Related to
this is “goodness of ﬁt” — the idea that what matters is match or mismatch between child
characteristics and parenting practices or the environment (e. g., Chess & Thomas, 1989).

A second perspective under the umbrella of temperament theory comes from Buss
and Plomin (Buss, 1991). Whereas some researchers (e. g., Chess & Thomas, 1989) have
argued that temperament may help to shape the individual’s personality structure - in
transaction with the environment — but that the two are not the same, Buss and Plomin
argued that temperament is a subset of personality traits (Buss, 1991; Worobey, 1999).
Speciﬁcally, it is deﬁned as such on the basis of three major assumptions: an appearance
during the ﬁrst year of life; persistence through adulthood, at least in terms of prediction;
and its inheritance as part of the genotype of humans and other animals (Buss, 1991).
Buss (1991) proposed three dimensions of temperament, known as EAS: emotionality
(distress and autonomic arousal accompanying both fear and anger), activity (movement
or physical expenditure of energy), and sociability (a preference for being with others);
activity is the only one of the three related to behavioral style. EAS only posits a limited
and restricting transactional approach: the child is the driving force - capable of
“choosing” environments, “setting the tone for others in social interaction,” and
modifying the interpersonal environment or its impact (Buss, 1991, p. 47).

A third perspective proposes a more biological, Pavlovian psychophysiology
deﬁnition of temperament. According to Rothbart (1981), temperament is related to the
autonomic and somatic responses of the nervous and endocrine systems (i.e., reactivity)
and the strategies adopted to modulate these psychobiological processes (i.e., self-

regulation). This has “a constitutional basis, with ‘constitutional’ deﬁned as the

40

relatively enduring biological makeup of the individual, inﬂuenced over time by the
interaction of heredity, life experience, and maturation” (Rothbart, 1981, p. 569). For
example, self-regulation becomes more and more organized over time (Strelau &
Angleitner, 1991), developing from “other” regulation shaped externally by the caregiver
into an internalized control structure that the child can then activate on his/her own
(Rothbart, 1991). Although the theoretical assumptions here are not generally compatible
with the NYLS model, two facets of temperament are recognized by both: its emergent
nature, and the importance of the caregiving relationship (for Rothbart, at least initially).

Despite theoretical differences in the conceptualization and measurement of
temperament, there is some consensus. In general, and at its most basic or synthesized
level, the concept of temperament describes individual differences with a constellation of
traits that constitute general dispositions to respond in certain ways, that have at least
some biological inﬂuence, that are relatively stable, and yet are subject to change
(Worobey, 1999). As such, temperament is a predisposition that precedes or underlies
behavior, and can be considered a precursor to outcomes in childhood, adolescence, and
adulthood.

For example, temperament has been linked to behavior problems (Campbell &
Ewing, 1990; Caspi et al., 1995; Jansen, Fitzgerald, Ham, & Zucker, 1995; Loukas,
Fitzgerald, Zucker, & von Eye, 2001; Wong, Zucker, Puttler, & Fitzgerald, 1999),
crime/delinquency (Caspi et al., 1996; Henry, Caspi, Mofﬁtt, & Silva, 1996; Tremblay,
Pihl, Vitaro, and Dobkin, 1994; White et al., 1990), and substance use (Masse and

Tremblay, 1997; Tarter, 1988, Tarter & Vanyokov, 1994) over the life-course. Details

41

from a few of these studies help to demonstrate the pervasiveness of temperamental
difﬁculties in predicting the persistence of antisocial behavior.

Most strikingly from a long-term perspective are reports from the Dunedin
Multidisciplinary Health and Development Study which suggest that adolescent
delinquency (White et al., 1990) and behavior problems (Caspi et al., 1995) as well as
adult psychiatric outcomes, crime, and alcohol dependence (Caspi et al., 1996) can be
predicted by the behavioral styles of children as young as age three. The Dunedin
Multidisciplinary Health and Development Study of Caspi and colleagues has been
following a complete cohort from New Zealand born between April 1, 1972, and March
31,1973, with assessments at ages 3, 5, 7, 9,11, 13, 15, 18, and 21. Caspi et a1. (1995)
found that temperament dimensions at 3 and 5 years of age were correlated with
externalizing behavior problems at 9, 11, 13, and 15 years of age. Speciﬁcally, there
were consistent positive correlations between externalizing symptomatology and lack of
control, a factor comprised of items related to emotional lability, restlessness, short
attention span, and negativism.

In other research by Caspi and colleagues (1996), children were classiﬁed into
one of 5 clusters at age three: undercontrolled (irritable, impulsive, impersistent, and
emotionally labile), inhibited (fearful, socially reticent), well-adjusted (within normal
limits), conﬁdent (zealous, eager to explore, easily adjusting), and reserved (timid,
somewhat uncomfortable during testing). Undercontrolled and inhibited preschoolers
were the most likely to develop an adult psychiatric disorder by age 21. Compared to
their well-adjusted peers, undercontrolled children were 2.9 times as likely to be

diagnosed with antisocial personality disorder, 2.2 times as likely to be a recidivistic

42

offender, 4.5 times as likely to be convicted for a violent offense, and 2.7 times as likely
to be diagnosed with alcohol dependence.

The predictive utility of early undercontrol is also demonstrated by Tremblay,
Pihl, Vitaro, and Dobkin (1994). For a large sample of boys attending French public
schools in Quebec, teacher-rated impulsivity in kindergarten was the personality
dimension most associated with self-reported delinquency at 10-13 years of age.

There are several possible explanations for the robust relationship between
temperament and behavior problems. Difficult temperament may be an early substrate of
behavior problems, reﬂecting the same individual differences but with one as a
subclinical predecessor to the other (Caspi et al., 1995). Altemately, the relationship can
be explained by dynamic interaction between person and environment (Caspi et al.,
1995). Children with difﬁcult temperament may evoke negative responses from the
interpersonal environment (e. g., if parents respond with harshness, neglect, or
impatience) and may react to that adversity in ways that elaborate their initial behavior
schemas. They may also be more susceptible to family adversity (e.g., parental
psychopathology) because of their core risk diathesis, with this environmental stress
activating the temperamental vulnerability to produce a behavior problem reaction. The
idea of temperament as a mediator of environmental risk and outcomes has been tested
and veriﬁed (Loukas et al., 2001). Problem behaviors may even be a coping mechanism
for dealing with unfavorable experiences. Interplay is also emphasized by Henry et a1.
(1996), who argued that risk for antisocial behavior is determined by two components of
behavioral regulation: social regulation provided by the family environment, and

individual differences in self-regulation (e.g., lack of control). Again, it is the nesting of

43

undercontrolled children in disorganized social contexts that characterizes the
developmental pathway into persistent antisocial behavior.

Parental Psychopathology. Parental psychopathology has also been extensively
studied in relation to the development of child behavior problems. Children of antisocial,
alcoholic, and depressed parents are at particularly heightened risk. In fact, these
psychopathologies are often co-morbid (Zucker, 1987 , 1994; Zucker et al., 1995).

Longitudinal studies of the development of delinquent behavior have often
implicated parental antisociality as an important risk factor. For example, juvenile
delinquents are more likely to have parents who have been convicted, have committed
crimes, or met diagnostic criteria for antisocial personality disorder (Elkins, Iacono,
Doyle, & McGue, 1997; Farrington, 1995; Loeber & Dishion, 1983; Sampson & Laub,
1993). In Fitzgerald, Zucker, and Yang (1995), both maternal and paternal antisociality
predicted child aggression at three to ﬁve years of age. Interestingly, the effect for
maternal antisociality was stronger.

Greater severity in behavior problems has also been found among children of
alcoholics (Fitzgerald, Sullivan, Ham, Zucker, Bruckel, & Schneider, 1993; Loukas,
1997). In Fitzgerald et a1. (1993), a higher percentage of children from alcoholic families
scored in the extreme clinical range for behavior problems. As mentioned, parental
alcoholism is also indirectly related to externalizing behavior problems through
temperament (Loukas et al., 2001). Fitzgerald, Zucker, & Yang (1995) reported
pervasive differences between male children of alcoholics and control children —
including deﬁcits in ﬁne-motor, personal-social, language and adaptive skills,

intelligence, behavior problems, and the reactivity dimension of temperament.

44

Another form of parental psychopathology associated with child behavior
problems in the extant literature is depression — especially maternal. Campbell (1994)
found that initial levels of maternal depression were higher for persistently “hard-to-
manage” boys two years after study entry. Fitzgerald et al. (1993) reported that mother’s
depression (and alcohol problems) predicted her ratings for son’s externalizing behavior
problems at age three.

In general, there may be a genetic vulnerability towards behavioral styles that
predispose towards antisocial behavior, and children may learn to cope with stress or
adversity through antisocial means as a result of modeling the behaviors they are exposed
to by parents (e.g., aggression, abuse, negative affect expression). Fuller et al. (2003), for
example, found evidence for continuity in aggression across three generations. As
discussed in the next section, however, many of these effects are mediated through the
more proximate family environment that accompanies parental psychopathology.
Inﬂuences in Middle Childhood

Early personal and environmentally-inﬂuenced risk trajectories do not exist in a
self-sustaining vacuum but interact with processes of risk and protection/promotion over
developmental time. As mentioned, among these proximate (and mediating) inﬂuences
are features of the family environment that maintain, exacerbate, or temper initial risk
status. Also of importance are proximate inﬂuences from other major domains. Schools,
peers, and self-concept are particularly relevant to developmental tasks of middle
childhood. For example, Erikson (1950) argued that the goal of development is to
achieve a positive ego identity. This is accomplished by successfully mastering each of

eightstages that, together, make up the life course (i.e., the “Eight Stages of Man”). Each

45

stage is marked by a conﬂict to be resolved. During middle childhood (i.e., Freud’s
latency period), the conﬂict is one of industry versus inferiority. The child “learns to win
recognition by producing things” (Erikson, 1950, p. 226) — that is, by being industrious.
Under this rubric, to not cultivate one’s “tools and skills” leads to despair, failure to
succeed at school, and later disappointments in the “total economy” (Erikson, 1950, p.
227). Similar issues are addressed by Havighurst (1972) in his discussion of
developmental tasks, which in middle childhood include getting along with age-mates
and building wholesome attitudes towards oneself. From a life-course perspective, it may
be that competencies (or deﬁcits) in these areas can buffer (or worsen) the effects of early
risk.

Family Environment. Family management techniques and family climate have
been widely studied as important predictors of child behavior problems. In Farrington’s
study (1995), future juvenile delinquents were more likely to have family environments
characterized by economic deprivation, physical neglect, harsh or erratic discipline,
parental conﬂict, and poor supervision. Delinquency was also associated with
erratic/harsh discipline, poor supervision, parental rejection, and weaker parental
attachment in the analyses conducted by Sampson and Laub (1993). Structural
background factors (e.g.. family disruption, low socioeconomic status, parental
criminality, parental alcohol use) had an effect on delinquency, but operated almost
entirely through these family process variables. The family-level risk factors for
continuity in behavior problems reported by Fergusson et a1. (1996) included social

disadvantage and family dysfunction. In Wong et al. (1999), parents’ negative affect

46

expression (spanking, bad mood, and sadness) was positively related to children’s
externalizing behavior problems.

Among the signiﬁcant predictors of persistent externalizing disorders in the
research of Campbell and colleagues were family stress, marital dissatisfaction, and
negative maternal control (Campbell & Ewing, 1990; Campbell, 1994). In fact, given
conditions of ongoing family adversity, Campbell (1994, p. 164) concluded that “ﬁndings
highlight the stability of problem behaviors in young boys, the association between high
symptoms and poor social functioning in young children, and the important role that
family context plays in both the stability of problems and their apparent onset.”

Patterson and colleagues (e. g., Patterson, DeBaryshe, & Ramsey, 1989; Larzelere
& Patterson, 1990) have described a developmental perspective on antisocial behavior
that focuses on the importance of the family context A three-step etiological process was
proposed to explain chronic delinquent behavior (Patterson et al., 1989), as follows:
Ineffective parenting (i.e., poor parental discipline and monitoring and little positive
parental involvement) leads to conduct problems, which lead to academic failure and peer
rejection, which lead to involvement with a deviant peer group.

Many distal risks inﬂuence child behavior through the more proximate social
environment. For example, Larzelere and Patterson (1990) found that parent
management skills — which include parental monitoring and discipline — mediated the
relationship between socioeconomic status and delinquency in early adolescence. One
measure of discipline was “Nattering”, or negative parental interactions toward the child.

Other research suggests that the effects of parental alcoholism and other parental

psychopathologies on child development are mediated through the quality of family

47

relationships and overall dynamics in the home environment (e.g., Barrera, Li, &
Chassin, 1995; Hawkins, 1997; Hecht, 1973; Henderson, Albright, Kalichman, &
Dugoni, 1994; Hyphantis, Koutras, Liakos, & Marselos, 1991; Jacob, Krahn, & Leonard,
1991; Roosa, Beals, Sandler, & Pillow, 1990). According to Hecht (1973), families that
seek professional help because of alcoholism do so because of conﬂicts at home or within
other social institutions (e.g., school, work, community), not because of drinking
behavior per se.

In general, alcoholism has been associated with a parenting style characterized by
more negative and less positive affect. Jacob et al. (1991) found that, compared to a
control group, fathers diagnosed as depressed or alcoholic demonstrated less congeniality
and fewer neutral problem-solving strategies when interacting with their sons during a
videotaped discussion about family issues. Notably, these authors described a general
atmosphere of distress, as characterized by alcoholism g depression, which deﬁnes
family “dysfunction.” Similarly, Henderson et al. (1994) suggested that co-morbidity in
psychiatric diagnoses lends itself to the oft-unwarranted conclusion that there is
something particular about substance use; rather, alcoholism seems to be one of several
mental health problems (e.g., schizophrenia, depression, generalized anxiety disorder)
that contribute to detrimental developmental outcomes among offspring.

In addition to attachment and affection, supervision and discipline are impacted
by parental psychopathology. For example, the alcoholic is unlikely to assume adequate
responsibility in this regard - expressing either detachment or inappropriate displays of
anger. In some cases, the non-alcoholic parent in an alcoholic household may be so

focused on dealing with the spouse that supervising and disciplining children becomes

48

lax (Hecht, 1973; cf. Hyphantis et al., 1991). At the other extreme is a harsh and
inﬂexible style of discipline (i.e., authoritarian). Malo and Tremblay (1997) found that
children of alcoholic fathers reported more frequent punishment than controls, but fewer
rules for proper conduct. Children may ﬁnd the inconsistencies inherent in a punitive but
undeﬁned system of punishment somewhat stressful (Malo & Tremblay, 1997).

Another correlate of parental alcoholism is family conﬂict or discord. In an
observational study, Jacob et a1. (1991) indicated that mother-father dyads in families
characterized by paternal alcoholism demonstrated more negative emotionality, less
positive affect, and heightened levels of general discord than their non-alcoholic
counterparts. Webb and Baer (1995) also found a signiﬁcant relationship between family
disharmony and parental alcohol use, with the quality of parent-child interactions having
a direct (and mediating) effect on outcomes such as attenuated social skills or adolescent
alcohol use. Unfortunately, the issue of causality is problematic in most of these studies.
Too often, given the data available, researchers cannot determine whether the drinking
behavior of parents leads to greater conﬂict in the family or whether conﬂict is the
impetus for alcohol use (e.g., Webb & Baer, 1995). Heightened conﬂict could also be a
reaction to acting-out behavior by the child, another potential confound. Reiterating
Hecht’s (1973; see also Henderson et al., 1994) observation, however, it is still extremely
important to consider the family environment in which children are embedded — not just
manifestations of individual-level parental psychopathology.

In summary, the emphasis needs to be on the “functional structure of the family”
(Hyphantis et al., 1991, p. 35), which is consistent with an ecological person-in-context

approach to risk. Alcoholics tend to be less demonstrative with affection, and more likely

49

to employ a permissive or authoritarian style of parenting. Alcoholic families may also
be less close-knit, with greater discord between family members. Although the emphasis
here has been on mediating parental alcoholism, other forms of parental psychopathology
(Henderson et al., 1994) and co-morbid psychopathology (Fitzgerald, Zucker, & Yang,
1995) have similar effects on the home environment.

Child Characteristics. As mentioned, developmental tasks for mastery of middle
childhood include healthy adjustment with respect to school, peers, and self-concept.
Competencies or deﬁcits in these domains do seem to play an important role in the
etiology of antisocial behavior.

In F ergusson et al. (1996), children with pervasive antisocial behavior had the
lowest self-esteem whereas non-problem behavior children had the highest self-esteem;
remitting or late onset children were intermediate. Using reciprocal effects analysis to
take into account the fact that involvement in delinquent behavior may shape self-esteem
at the same time that self-esteem inﬂuences delinquency risk (i.e., lagged and
contemporaneous inﬂuences), Rosenberg, Schooler, & Schoenbach (1989) found
evidence for both processes. Low self-esteem was signiﬁcantly related to an increase in
delinquency over two years and there was a trend, although non-signiﬁcant, towards an
association between earlier delinquency and increased self-esteem.

Campbell et a1. (1986) found that hard-to-manage children at age six were less
involved and skilled in social activities as indicated by social withdrawal, participation in
organized social groups, and play with peers. By age nine, children with persistent
problems from 3-6 years of age were still scoring lower in social competence and higher

on peer problems than those with desisting behavioral difﬁculties and a control group

50

(Campbell & Ewing, 1990). In general, Campbell (1994, p. 163) argued that children
with persisting behavioral difficulties are “not only more oppositional, impulsive,
inattentive, and disorganized, but also more socially anxious and incompetent” and that
this “may interfere with social development in the family and peer group.”

Academic competence is another important risk marker. In Farrington’s (1995)
longitudinal study, future juvenile delinquents were more likely to have low intelligence
and low school achievement (Farrington, 1995). Early aggression was also prognostic of
less achievement in academic domains (e. g., lower scores for spelling, reading,
arithmetic, education, and occupational status) for both males and females in the
longitudinal research of Eron et al. (1987). Similarly, the individual-level risk factors for
continuity in behavior problems reported by F ergusson et al. (1996) included lower IQ
and poor school achievement. Verbal IQ was consistently related to delinquency in ﬁve
major longitudinal studies (Elkins et al., 1997), and low IQ seems to be a useful predictor
of behavior problems that persist (Elkins et al., 1997). In Mofﬁtt (1990), delinquents
with attention deﬁcit disorder scored lower on verbal IQ and reading proﬁciency than
controls; however, there were no differences between controls, children with only
attention deﬁcit disorder, and children with only delinquent behavior patterns. Loeber
and Dishion (1983) reported on the predictive utility of educational attainment,
particularly low vocabulary and poor verbal reasoning. It should be noted that although
poor school performance is often cited as a contributor to later delinquency, high
scholastic achievement can also be considered a promotive factor. For example, doing
well in school may lead to greater “stakes in conformity” and strong conventional social

bonds (Hirschi, 1969) — with a decreased willingness to engage in problem behavior that

51

might jeopardize the promise of educational attainment or career success. Good grades
also contribute to higher levels of self-esteem (Rosenberg et al., 1989).
Summary and Hypotheses

Most likely, continuity in antisocial behavior over time is the result of
transactional processes that take place between a child’s constitutional nature, parental
characteristics, and the social environment (Keenan & Shaw, 1994; Huesmann et al.,
1984). In other words, researchers need to adopt an ecological and family systems
perspective to understand the origin and persistence of childhood behavior problems.
Within the studies reviewed here, key inﬂuences on the course and persistence of
antisocial behaviors include early undercontrol/impulsivity/hyperactivity, temperament,
intelligence, lower social competency, low educational achievement, family conﬂict,
negative control or harsh discipline, maternal depression, parental alcoholism, parental
antisociality and criminality. Altemately, intelligence, personal competencies, and
support in the caregiving environment have been identiﬁed as primary proximate buffers
against maladaptation (Egeland, Carlson, & Sroufe, 1993; Masten & Coatsworth, 1998;
Masten et al., 1999; Werner, 1986; Werner & Smith, 1982; Wyman, Cowen, Work, Hoyt-
Meyers, Magnus, & Fagen, 1999), as per the literature on resilience.

The current study will approach the issue of continuity with several key
assumptions based on ﬁndings from the extant literature. First, given the inﬂuences of
personal characteristics and environmental conditions, initial predispositions toward
antisocial behavior must consist of individual-level (e.g., temperament) and family-level
(e.g., parental psychopathology) components. Second, these risks appear to have their

effect on later behavior patterns through more proximate inﬂuences, such as family

52

environment and academic, social, or other personal competencies. This supports the
inclusion of middle childhood factors (e.g., family environment, discipline, academic
competence, self-esteem, social problems) in predicting outcome. However, and third,
middle childhood factors do not necessarily reﬂect initial risk levels. The fact that they
are inﬂuenced by what preceded them, developmentally (e.g., early temperament and
lifetime parental psychopathology), means that stability is probably the stronger trend;
old risks and protections can maintain the behavioral trajectory. However, new risks and
protections in the immediate context may serve as “turning points” (Sampson & Laub,
1993). Whether or not behavioral difﬁculties persist, remit, begin, or remain absent from
preschool to adolescence most likely depends on whether risk persists, remits, begins, or
remains absent as well as whether protection/promotion comes into play.

Finally, some factors can be both risks and protections— depending on their
valence and the initial behavioral trajectory. For example, low academic achievement
may be a risk factor for those without preschool adversity, whereas high academic
achievement can be a “turning point” for high-risk preschoolers. Similarly, high self-
esteem may protect at-risk youth from the effects of family dysfunction, whereas low
self-esteem may potentiate the effects of adversity. Even within behavior trajectories,
proximate inﬂuences can have a risk or promotive inﬂuence. This argument is made by
Jansen, Fitzgerald, Ham, & Zucker (1995): Children with difﬁcult temperament from
chaotic, stressful, conﬂicted homes are more likely to be at-risk for substance use,
whereas a stable, warm, non-alcoholic home can buffer the child from a risk trajectory.
As Rutter (1987, p. 318) remarked, “it seems helpful to use the term ‘protective

mechanism’ when what was previously a risk trajectory is changed to one with a greater

53

likelihood of an adaptive outcome... Conversely, the process would be labelled a
vulnerability one when a previously adaptive trajectory is turned into a negative one.”
Both processes are explored in the current study (e.g., for the high-low and low-high
antisocial behavior group classiﬁcations). Therefore, risk and protection will be viewed
as opposite ends of a single continuum, dependent on an individual’s circumstances
rather than a priori determinations, for ease of interpretation.

Four hypotheses will be tested in the current study — each supported by ﬁndings
from the extant literature. The ﬁrst two hypotheses test a strong continuity model of

problem behavior, and the idea of an antisocial behavior pathway into ﬁrst alcohol use.

Hypothesis 1. The primary pathway into early ﬁrst alcohol use begins with

heightened levels of antisocial behavior from as early as 3-5 years of age, onward.

This hypothesis is derived from literature suggesting that an early Age of First
Drink (AFD) and adolescent substance use is associated with later problem behaviors,
risky personality traits, and an earlier history of behavioral undercontrol (e.g., Caspi et
al., 1996; Cloninger et al., 1988; Gruber et al., 1996; Masse & Tremblay, 1997; Tarter,
1988; Wills et al., 1995); typologies linking childhood antisocial behavior to alcoholic
risk (e.g., Cloninger et al., 1996; Zucker, 1987, 1994) combined with evidence that AF D
is a marker for this risk (e.g., Grant & Dawson, 1997; Gruber et al., 1996; McGue et al.,
2001b; Prescott & Kendler, 1999; Schukit & Russell, 1983); and literature demonstrating
strong continuity in antisocial behavior problems for a subset of individuals (e. g., Calms

et al.,1989; Campbell, 1994; Campbell et al., 1986; Campbell et al., 1994; Campbell &

54

Ewing, 1990; Eron et al., 1987; Farrington, 1995, 1997; Fergusson et al., 1996; Robins,
1978; Sampson & Laub, 1993; Shaw et al., 1994) as early as infancy. Early drinkers are
expected to have higher stability coefﬁcients for aggression and delinquent behavior from

preschool through adolescence.

Hypothesis 2. Preschool and adolescent levels of antisocial behavior problems
are distinguishing characteristics along the developmental course leading to and
through drinking onset — and more speciﬁcally, antisocial behavior problems in
these two time periods can be used to categorize individuals into groups that

differ in the likelihood of early ﬁrst drink.

To determine whether this simpliﬁed classiﬁcation scheme is an adequate
representation of risk trajectories for early alcohol use, children will be grouped into one
of four classiﬁcations: high-high, low-low, low-high, and high-low in antisocial behavior
during preschool (3 -5 years of age) and early adolescence (12-14 years of age),
respectively. Early drinkers are expected to show persistent behavioral difficulties and
therefore be disproportionately found in the high-high group for both aggression and
delinquent behavior. It is expected that those without early drinking onset will be
disproportionately found in the low-low group for both aggression and delinquent
behavior. Both ﬁndings would provide support for the strong continuity model of
development described ﬁrst in Hypothesis 1.

Also of interest are change patterns. Not all difﬁcult children will become

delinquent adolescents and not all delinquent adolescents were difficult children.

55

Children with remitting antisocial behaviors and children with escalating antisocial
behaviors over the course of development may be at intermediate risk for early onset of
alcohol use. Another goal of the current study, therefore, is to determine the inﬂuences
that lead to continuity versus change in the development of antisocial behavior.

As mentioned in the literature review, predispositions may establish the initial risk
trajectory but risk and promotive factors over the developmental course can serve as
maintenance structures or “turning points” for behavior. Important distal risk factors for
behavior problem, crime/delinquency, and substance use trajectories — which, as
previously described, are heavily intertwined - include temperament (e. g., Campbell &
Ewing, 1990; Caspi et al., 1996; Henry et al., 1996; Jansen et al., 1995; Loukas et al.,
2001; Masse & Tremblay, 1997; Tarter, 1988; Tarter & Vanyokov, 1994; Tremblay et al.,
1994; White et al., 1994) and parental psychopathology (e.g., Elkins et al., 1997;
Farrington, 1995; Fitzgerald et al., 1993; Fitzgerald et al., 1995; Loeber & Dishion, 1983;
Loukas et al., 2001; Sampson & Laub, 1993). Important proximate middle-childhood
inﬂuences on behavior problems, crime/delinquency, and substance use trajectories
include facets of the family environment (e. g., Barrera et al., 1995; Campbell & Ewing,
1990; Campbell, 1994; Cohen et al., 1994; Farrington, 1995; Fergusson et al., 1996;
Hawkins, 1997; Hecht, 1973; Henderson et al., 1994; Hops et al., 1999; Hussong &
Chassin, 1997; Hyphantis et al., 1991; Jacob et al., 1991; Roosa et al., 1990; Sampson &
Laub, 1993; Zucker & Gomberg, 1986) and personal competencies or deﬁcits in domains
such as agtdemic achievement (e.g., Eron et al., 1987; Farrington, 1995; Fergusson et al.,
1996; Loeber & Dishion, 1983; Zucker & Gomberg, 1986), self-esteem (e.g., Fergusson

etal., 1996; Rosenberg et al., 1989; Werner, 1986), and social relationalaips/problems

56

(e.g., Campbell, 1994; Campbell & Ewing, 1990; Campbell et al., 1986; Zucker &
Gomberg, 1986). A third hypothesis therefore addresses continuity versus discontinuity
in antisocial behavior problems within the context of initial risk proﬁles and middle

childhood risk/promotive factors:

Hypothesis 3: On all measures of risk from preschool (temperament, parental
psychopathology) and middle childhood (academic achievement, self-esteem,
social problems, family cohesion, family expressiveness, family conﬂict, physical
discipline/negative control), scores will be least favorable for the high-high group,
moderate for the low-high and high-low groups, and most favorable for the low-
low group. If supported, this would corroborate the ﬁndings of Fergusson et al.

(1996).

The high-high group is expected to have the following characteristics relative to
the low-low group: more difficult temperament (i.e., high activity, short attention
span, more reactivity) and more parental psychopathology (mothers and fathers)
at 3-5 years of age; as well as less academic achievement (i.e., lower scores on
reading, spelling, and arithmetic), lower self-esteem, more social problems, less
family cohesion, less family expressiveness, more family conﬂict, and more
physical/negative punishment at 6-8 and 9-11 years of age. The high-low and
low-high groups are expected to have in-between scores on each of these

dimensions.

57

In addition, what are the relative contributions of these early risk and middle
childhood factors for explaining levels of antisocial behavior? Are early risks more
important than later inﬂuences, or do later — more proximate — inﬂuences take

precedence? The fourth hypothesis speaks to these issues:

Hypothesis 4: Initial risk trajectory variables will predict aggression and
delinquent behavior during the preschool years, and initial risk trajectory
variables along with middle childhood variables will predict adolescent
aggression and adolescent delinquent behavior. Middle childhood factors will
boost the explanatory power of the model (i.e., signiﬁcantly add to the R2) and
eclipse some of the early indicators of risk — recognizing both that risk is ﬂuid
(i.e., early indicators don’t always pan out), and that proximate inﬂuences are

likely to have greater effect.

With regard to the initial risk load, antisocial behavior in preschool and
adolescence will be predicted from more difﬁcult temperament (i.e., higher
activity levels, shorter attention span, and greater reactivity) and exposure to more
parental psychopathology. Adolescent antisocial behavior will also be predicted
by child characteristics (less academic achievement, lower self-esteem, and more
social problems) and the family environment (i.e., less cohesion, less

expressiveness, more conﬂict, and more physical punishment/negative control) in

middle childhood.

58

A heuristic model is depicted in Figure 2. In this model, early risks determine the

level of antisocial behavior observed during preschool. However, these behavior patterns

can either be maintained by experiences in middle childhood or modiﬁed by them (i.e.,

with respect to “turning points”). Middle childhood variables are therefore seen as

mediators in the relationship between early risks and later antisocial behaviors. Arrows

for antisocial behavior are intended to represent an ongoing developmental trajectory that

moves to and through middle childhood, into adolescence.

PRESCHOOL

 

Temperament
Parental Psychopathology

MIDDLE
CHILDHOOD

 

Academic Achievement
_ Reading Skills

 

 

 

 

Antisocial
Behavior
Problems

Figure 2.

 

Spelling Skills

Arithmetic Skills
Self-Esteem
Social Problems
Family Cohesion
Family Expressiveness
Family Conﬂict
Discipline/ Control

EARLY
ADOLESCENC E

 

 

 

V

Antisocial
Behavior
Problems

Heuristic model of the relationship between the initial risk trajectory and middle

childhood factors to antisocial behavior problems from preschool through early

adolescence.

59

CHAPTER 3
METHODOLOGY

Participants

Participants were 220 male adolescents and their parents from the prospective
Michigan Longitudinal Study (Zucker, Fitzgerald, Reﬁor, Puttler, Pallas, & Ellis, 2000).
This ongoing project utilized population-based recruitment strategies to identify alcoholic
and ecologically-matched comparison families. Families needed to be intact and
biological, with at least one male child between the ages of 3-0 and 5-11 at the time of
recruitment. Almost all participants in the Michigan Longitudinal Study are non-
Hispanic European. All were residents of the mid-Michigan area at the initial
assessment, with income levels mostly within the low-middle class range (see Zucker et
al., 2000, for additional details). The data used herein were collected at recruitment and
when the children were approximately 3-5 (Wave 1), 6-8 (Wave 2), 9-11 (Wave 3), and
12-14 (Wave 4) years of age.

Alcoholic families were identiﬁed on the basis of father’s drinking status.
Although maternal drinking was assessed, maternal alcoholism was neither a requirement
nor a basis for exclusion. However, families with children manifesting characteristics
associated with a diagnosis of fetal alcohol syndrome, or with mothers who consumed six
or more drinks per day during pregnancy, were excluded from participation. Alcoholic
fathers were recruited from drunk driving records or neighborhood recruitment.

Of all convicted drunk drivers in a four county Mid-Michigan area, all males
meeting the study criteria of a male child between 3-0 and 5-11 in an intact biological
household with a blood alcohol concentration (BAC) of 0. 1 5% (150 mg/ 100 ml) or

higher when arrested - or a BAC ofO. 12% with a history of prior alcohol-related driving

60

offenses — were asked permission for contact by study staff. At initial contact, a positive
alcoholism diagnosis was established using the short Michigan Alcohol Screening Test
(SMAST; Selzer, Vinokur, & van Rooijen, 1975) and later veriﬁed with the NIMH
Diagnostic Interview Schedule — Version 111 (DIS; Robins, Helzer, Croughan, & Ratcliff,
1980) and the Drinking and Drug History questionnaire (Zucker, Fitzgerald, & Noll,
1990). All of these men met a “deﬁnite” or “probable” criterion for alcoholism using the
Feighner Diagnostic Criteria (Feighner, Robins, Winokur, Guze et al., 1972).

Alcoholic fathers were also recruited from the neighborhoods in which the drunk
driver alcoholic fathers resided. Community alcoholics were identiﬁed during
neighborhood canvassing for non-alcoholic control families (see below) by positive
alcoholism screenings. These men also met Feighner criteria for probable or deﬁnite
alcoholism, and met the same inclusion criteria as the drunk-driving group for age of a
biological male target child and coupled status.

In addition to alcoholic families, a group of community control families were
recruited via door-to-door community surveys. These families were matched to a subset
of alcoholic families on the basis of age for the male target child (within six months) and
socio-economic status (from the same neighborhood or in a similar census tract). Neither
mother nor father met Feighner criteria for alcoholism or other drug abuse/dependence.

For inclusion in the current sample, there were two criteria. The ﬁrst was the
availability of behavior problem ratings at Wave 1 and Wave 4. These constituted major
dependent variables in analyses herein, so the decision was made to avoid missing data
imputation on these key measures. Second, families must have participated in at least

one of the middle childhood waves — Wave 2 and/or Wave 3. This requirement helped to

61

ensure that missing data imputation for middle childhood measures was at least partially
based on other middle childhood measures (i.e., using Wave 2 information to help impute
Wave 3 values, or using Wave 3 information to help impute Wave 2 values). Of the 333
families from whom some data have been collected as part of the larger Michigan
Longitudinal Study, 220 cases met the current criteria. All 220 valid cases had
information on age of ﬁrst drink, so it was not necessary to use the availability of onset
information as a criterion for inclusion or exclusion.

Unfortunately, sibling data collection did not begin at the time of family
recruitment. A majority of females with data on onset of drinking were ages 6 to 11
when they were invited to participate in the longitudinal study. Only a small number of
girls provided information at both Wave 1 and Wave 4 (approximate n= 20). This is
inadequate for a statistical analysis of gender differences. However, there are
documented gender differences in the development of antisocial behavior - for example,
in the stability of aggression (e.g., Cairns etal., 1989; Cummings et al., 1989; Eron et al.,
1987). It is for this reason that females were excluded.

Procedure

Data were collected by trained project staff blind to family alcoholism status. At
each wave, the visits involved approximately 15 hours of contact time for each parent and
seven hours of time for the target child. Contacts included questionnaire sessions, semi-
structured interviews, and interactive tasks. Families were compensated for their

participation.

62

Child Measures

First Drink. Alcohol use was measured by self-reports from adolescents on a
Drinking and Drug History Form for Children questionnaire at 12-14 years of age (Wave
4). This questionnaire is the child’s version of the Drinking and Drug History Form for
Adults (Zucker, Fitzgerald, & Noll, 1990). The adult questionnaire incorporates already
much tested items from the 197 8 NIDA Survey (Johnston et a1, 1979), from the American
Drinking Practices Survey (Cahalan, Cisson & Crossley, 1969), and from the VA.
Medical Center (University of California) San Diego, Research Questionnaire for
Alcoholics (Schuckit, 1978). All of the items have been extensively used in a variety of
survey and clinical settings. They provide data on quantity, frequency and variability of
alcohol consumption, frequency of drug use, and multiple questions on consequences and
troubles related to the use of these substances. Items on the child version have been
modiﬁed to be relevant for children and youth, but they cover the same substantive areas.
They also cover expectancies about use, extent of exposure to alcohol-related incidents
among parents and peers, and perceived availability.

Regarding onset, adolescents were asked at what age they ﬁrst tried alcohol,
excluding just a “sip” from an adult. If any age was listed, onset was categorized as
“yes” and adolescents were included in the “ﬁrst drink” group. Indications of not yet
having tried alcohol led to a “no ﬁrst drink” classiﬁcation. Onset is therefore a
dichotomous variable (0 = no ﬁrst drink by 12-14; 1 = ﬁrst drink by 12-14). Onset
information at Wave 4 was missing in two cases, but could be constructed from data
available at other waves. In the ﬁrst case, the individual reported having had a ﬁrst drink

at 12 years of age on the same instrument administered at Wave 5 (15-17 years) as well

63

as during separate annual assessments at ages 14 and 15; this resulted in a “ﬁrst drink”
classiﬁcation. In the second case, the individual reported not having had a ﬁrst drink at
the Wave 3 assessment as well as the annual assessment at age 14; this resulted in a “no
ﬁrst drink” classiﬁcation. In total, 25% of the sample (n = 55) had a ﬁrst drink by 12-14
years of age, whereas 75% had not (n = 165).

Antisocial Behavior. Child antisocial behavior was measured with the delinquent
behavior and aggressive behavior symptom subscales of the Child Behavior Checklist
(Achenbach, 1991; Achenbach & Edelbrock, 1983) at all waves. Maternal ratings were
used. Items in each subscale are listed in Table 2. The 118-item CBCL is a parent-report
measure of the prevalence and severity of child behavior problems. It is intended to
provide an objective assessment of the target child’s social and emotional functioning.
The CBCL yields raw and standardized scores on eight narrowband symptom scales, two
broadband subscales on externalizing and internalizing behaviors, and a total behavior
problems score. Narrowband subscales are comprised of items that co-occur in ratings
among clinically-referred children and load together in principal components analysis
(Achenbach, 1991).

The CBCL is a widely used instrument, with high validity and reliability
(Achenbach, 1991). Although the CBCL was normed on children 4 to 16 years old, it is
also administered to parents of children under four years of age because the Michigan
Longitudinal Study began before the publication of the CBCL for 2-3 year olds. Scores
for these younger children must be interpreted with caution, but Fitzgerald et al. (1993)
found no substantive differences in three-year-old children's scores and those obtained by

four- and ﬁve-year-olds (Reider, 1991). Test-retest reliability of item scores on the

64

CBCL range from .95 at a one-week interval, to .84 at a three-month interval
(Achenbach, 1991). Parent agreement in the item scores was .99. Similar reliabilities
were established on scale scores and the total problem score with one-week test-retest of
.89. The median parent agreement on scale scores was .66. Adequate construct validity
was established by correlations between CBCL scores and scores on a wide range of
other measures of child behavior problems.

Behavior problems are rated on a three-point scale (0 = Not true; 1 = Somewhat
or sometimes true; 2 = Very true or often true). Items are summed for each narrowband
subscale, with higher scores reﬂecting more behavior problems. One item (alcohol and
other drug use) was omitted from the delinquent behavior subscale to prevent artiﬁcial
inﬂation of the correlation between delinquent behavior and drinking onset status.

There are twenty aggressive behavior items (for a possible score range of 0-40) and 12
delinquent behavior items (for a possible score range of 0-24).

Continuity and discontinuity in aggressive behavior and delinquency were
operationalized, independently, by division of children into four groups: those who were
high-high, low-low, low-high, and high-low on the subscale score at Wave 1 (3-5) and
Wave 4 (12-14), respectively. Mean splits were used to classify subjects as high or low
on antisocial behavior at each wave. This allowed for adequate sample size in each cell

of the 2 x 2 classiﬁcation (see Figure 3).

65

Table 2

Items on the CBCL Delinquent BehaLvior and Aggpessive Behavior Subsca_la§

 

Delinquent Behavior

26.
39.
43.
63.
67.
72.
81.
82.
90.
96.
101.
106.

Doesn’t seem to feel guilty after misbehaving
Hangs around with others who get in trouble
Lying or cheating

Prefers being with older kids

Runs away from home

Sets ﬁres

Steals at home

Steals outside the home

Swearing or obscene language

Thinks about sex too much

Truancy, skips school

Vandalism

Aggressive Behavior

3.
7.
16.
19.
20.
21.
22.
23.
27.
37.
57.
68.
74.
86.
87.
93.
94.
95.
97.
104.

Argues a lot

Bragging, boasting

Cruelty, bullying, or meanness to others
Demands a lot of attention

Destroys his/her own things

Destroys things belonging to his/her family or others
Disobedient at home

Disobedient at school

Easily jealous

Gets in many ﬁghts

Physically attacks people

Screams a lot

Showing off or clowning

Stubborn, sullen, or irritable

Sudden changes in mood or feeling
Talks too much

Teases a lot

Temper tantrums or hot temper
Threatens people

Unusually loud

 

Note. One item from the delinquent behavior subscale was dropped (105. Uses
alcohol or drugs for non-medicinal purposes) because it is a confound in analyses
linking antisocial behavior problems with ﬁrst drink onset.

66

Behavior Problems in Adolescence

 

 

 

 

 

 

. Low High
Behavror Problems
in Preschool
Low Low-Low Low-High
High High-Low High-High
Figure 3.

Continuity and discontinuity classiﬁcation groups for aggression and delinquent

behavior.

Means for aggressive behavior were 10.65 at Wave 1 and 7.06 at Wave 4 (with
observed scores ranging from 0-27 and 0-26, respectively). Means for delinquent
behavior were 1.88 at Wave 1 and 1.85 at Wave 4 (with observed scores ranging from 0-
10 and 0-12, respectively). The distribution of groups is presented in Table 3 in the
marginals, along with the overlap between classiﬁcation categories. Numbers on the
diagonal represent consistency across behavior problem domains for severity at each time
period (e. g., low-low on both aggressive and delinquent behavior). Although
consistency is the stronger trend, there is also considerable disagreement of
classiﬁcations across domains. For example, twenty children classiﬁed as low-low on
aggressive behavior were classiﬁed as high-low on delinquent behavior. This supports

the separation of the two behavior problems for analysis.

67

Table 3

Distribution of Groups Based on Sample Means at Wave 1 and Wave 4 for Aggression

and Delingrent Behavior

 

Delinquent Behavior

 

Low-Low Low-High High-Low High-High Total N

 

 

Aggression
Low-Low 51 7 20 1 1 89
Low-High 6 1 1 1 9 27
High-Low 1 1 1 1 7 10 39
High-High 8 7 9 41 65
Total N 76 26 47 71 220

Table 4 shows the correlations between aggression and delinquent behavior
ratings at Wave 1 and Wave 4. From preschool to early adolescence, there are
moderately strong auto-correlations for aggression and delinquent behavior. Even higher
correlations are found within each time period between aggression and delinquent
behavior, especially during early adolescence. However, correlation isn’t equivalence.
For example, despite being conceptually different, height and weight are highly related;
ASPD and alcoholism are often co-morbid. Also, although there is considerable shared
variance (with r2 = .21, or 21%, at Wave 1 and r2 = .49, or 49%, at Wave 4), there is
considerable variance unique to each behavior problem as well; even in early
adolescence, where there is the most convergence, more than half is left unexplained (i.e.,

1-r2=.51,or 51%).

68

Table 4

Correkrtions Between Aggression and Delinquent Behayior

 

 

 

Variable 1 2 3 4

1. Wave 1 Aggression 1.00

2. Wave 1 Delinquent Behavior .456 ** 1.00

3. Wave 4 Aggression .444 ** .295 ** 1.00

4. Wave 4 Delinquent Behavior .341 ** .448 ** .701 ** 1.00
** p < .01

Temperament: Activity. Attention Span. Reactivity. Temperament was measured
by the Dimensions of Temperament Survey (DOTS; Lerner, Palermo, Spiro III, &
Nesselroade, 1982) at Wave 1 and Wave 2. The DOTS is a 34-item questionnaire that
measures ﬁve dimensions of temperament: activity level, attention span/distractability,
adaptability/approach-withdrawal, rhythmicity, and reactivity. The dimension scores are
based on sums of item ratings (1= True; 0 = False) for each scale. Ratings from mother
on child for activity (3 items; e.g., “My child moves a great deal in his/her sleep”),
attention span/distractibility (11 items; “My child persists at a task until it’s ﬁnished”),
and reactivity (6 items; “My child reacts intensely when hurt”) were used. These
dimensions were chosen because of their relevance to externalizing behavior problems in
other studies (e.g., Mun, Fitzgerald, von Eye, Puttler, & Zucker, 2001; Wong, Zucker,

Puttler, & Fitzgerald, 1999) and because they tap into constructs similar to attention

69

deﬁcit disorder with hyperactivity, which has been associated with aggression and other
problem behaviors in the extant literature (e. g., Campbell et al., 1986). Higher scores
reﬂect more activity, longer attention span, and greater reactivity.

At Wave 3, the revised DOTS-R (Windle & Lerner, 1986) was used to measure
temperament because of a change in protocol. The DOTS-R has 54 items rated on a four
point scale, from “usually false” to “usually true” about the child. There are nine
attributes assessed; of these, three overlap considerably with the DOTS attributes used in
this study: Activity-general (7 items), activity-sleep (4 items), and task orientation (8
items). Reactivity is not assessed by the DOTS-R but some of the DOTS items
measuring reactivity overlap with the DOTS-R dimension activity-general (e.g., “My
child can’t sit still for long”).

Agademic Achievement in Middle Childhood. Academic achievement was
measured with three subscales on the Wide Range Achievement Test - 3rd Edition
(WRAT; Wilkinson, 1993) from Wave 2 and Wave 3, using standardized scores. The
WRAT 3 assesses three areas of academic competence: Reading, Spelling, and
Arithmetic. It was standardized on 5000 individuals ranging in age from 5 to 74. Test-
retest reliability is .91 to .98. The WRAT 3 is an established screening tool for
identifying children at risk for learning disabilities.

Global Self-Esteem in Middle Childhood. The Harter Perceived Competence
Scale for Children (PCSC; Harter, 1978, 1979) was used to measure global self-esteem at
Wave 2 and Wave 3. This instrument was designed to measure the child's sense of his or
her own abilities and stature vis-a-vis other children. There are 36 items on this “What I

Am Like” questionnaire, six for each of six subscales: scholastic competence, social

70

acceptance, athletic competence, physical appearance, behavioral conduct, and global
self-worth. Responses are scored from 1 to 4. Statements represent feelings of
competency (e. g., “Some kids are very happy being the way they are BUT other kids
often wish they were different”). The child is asked to decide which type of kid is most
like him, and whether this is only sort of true or really true for him. Scale scores are
averages. Items were coded such that high scores reﬂect more positive self-evaluations.

The PCSC correlates low (.09) with a children’s social desirability scale (Crandall
et al., 1965). Internal consistency shows reliability alphas ranging from .73 to .86 across
all samples. Three month test-retest reliability ranges from .70 to .87 on the various
subscales. Over nine months, reliabilities were between .69 and .80.

Social Problems in Middle Childhood. Social problems, thought to represent peer
rejection or interpersonal awkwardness, were measured by maternal ratings on the social
problems subscale of the Child Behavior Checklist (Achenbach, 1991; Achenbach &
Edelbrock, 1983) for Wave 2 and Wave 3. Acting too young, not getting along with
peers, and gets teased are examples of the statements on the social problems subscale.
Items are rated on a three-point scale (0 = Not true; 1 = Somewhat or sometimes true; 2 =
Very true or often true) and summed.

Parent Meaames

Early/Lifetime Parental Psychopathology. Maternal and paternal
psychopathology was assessed with a psychopathology index that is a composite of three
measures from Wave 1: alcohol diagnosis, antisocial personality disorder, and
depression. Together they comprise a baseline measure reﬂecting the parental

component of an initial risk trajectory, whether this be environmentally or genetically

71

determined; the possibility of the latter is included via consideration of lifetime
diagnoses.

As described, alcoholism was screened with the short Michigan Alcohol
Screening Test (Selzer et al., 1975) and veriﬁed with the Diagnostic Interview Schedule —
Version 111 (DIS; Robins et a1, 1980) and Drinking and Drug History questionnaire
(Zucker et a1, 1990). On the basis of this information, a trained clinician made diagnoses
for alcohol abuse and dependence using several coding strategies. For this study, the
DSM IV diagnosis was used and diagnoses were scored as follows: 0 = no lifetime
diagnosis, l=alcohol abuse/dependence diagnosis but not in the last three years, and 2 =
alcohol abuse/dependence diagnosis in the last three years (see also Zucker, Wong,
Puttler, & Fitzgerald, 2003).

Parental lifetime antisocial personality disorder (ASPD) was assessed with the
DIS and the Antisocial Behavior Inventory (Zucker, Ellis, Fitzgerald, Bingham, &
Sanford, 1996; Zucker, Noll, Ham, Fitzgerald, & Sullivan, 1994). Information on the
Antisocial Behavior Inventory was used to supplement DIS data in establishing a
diagnosis. For this study, ASPD was scored as follows: 0 = no diagnosis, 2 = presence
of diagnosis (see also Zucker, Wong, Puttler, & Fitzgerald, 2003). Only the lifetime
indicator was used here because diagnosis for this personality disorder requires that a
long course of behavior over childhood and adulthood be established (Zucker et al.,
2003). A score of 2 was also deemed appropriate given the seriousness of ASPD and its
possible inﬂuence on child development; with this scoring system, having ASPD is

accorded the same weight as recent alcohol abuse/dependence.

72

Depression was measured by the Beck Depression Inventory (BDI; Beck et al.,
1961). The BDI is a widely used phenomenological measure of depressive state that has
received extensive validation (Carroll et al., 1973). The short form of the BDI is a 12-
item self-report instrument that assesses cognitive, emotional, motivational, and physical
manifestations of depression on a scale of 0 to 3. Beck reports a split-half reliability of
.93. Scores on the long and short forms of the BDI correlate between .89 and .97 (Beck,
Steer, & Garbin, 1988). A meta-analysis of 25 years of data on the BDI yielded an
internal consistency mean coefficient alpha of .86 for psychiatric patients and .81 for
non-psychiatric subjects. According to Beck and Beck (1972), scores of 0-4 demonstrate
no or minimal depression, scores of 5-7 demonstrate mild depression, scores of 8-15
demonstrate moderate depression, and scores greater than 15 demonstrate severe
depression. For this study, scores were recoded as follows, based on these guidelines: 0
= no or minimal depression, 1 = mild depression, and 2 = moderate depression. There
were no cases meeting criteria for severe depression in this sample.

The parental psychopathology index was computed by summing the alcohol
diagnosis score, ASPD score, and depression score for each parent. Consequently, scores
can range from 0 to 6 for mother and for father. Distributions are presented in Table 5.
Contextual Mea_sures

Family Environment in Middle Childhood: Cohesion. Expressivenesskaaia
@aﬂjp; Family environment was measured with the Moos Family Environment Scale
(FES; Moos, 1974; Moos & Moos, 1976) at Wave 2 and Wave 3. The FES is an

empirically based taxonomy of family social environments as perceived by family

73

Table 5

Distribution of Scores on Psychopatholmzy Dimensions for Mothers anﬁathers

 

 

 

 

Mother Father
N N
Alcohol Diagnosis
0 = no 148 66
1 = yes, not last three years 31 31
2 = yes, within last three years 41 123
Antisocial Personality Disorder (ASPD)
0 = no 214 185
2 = yes 6 35
Depression
0 = no 167 177
1 = mild 32 27
2 = moderate 21 16

 

members. Maternal reports were used because the youth-report version is not available
until Wave 3. Mothers had at least joint custody of the male target child in all but one
case. This is important because it means that the family environment being reported was
the one to which the child was primarily exposed. For the one case, maternal reports
were retained at Wave 2 because the mother reported living with the child for 2 to 2.5 of
the last three years — but were excluded at Wave 3 because the maternal report was for an
environment that too infrequently included the child (i.e., 7-12 months over the three year

time period).

74

The FES consists of ten scales that describe dimensions of family climate:
Cohesion, Organization, Expressiveness, Conﬂict, Independence, Control, Achievement,
Intellectual-Cultural Orientation, Active-Recreational Orientation, and Moral-Religious
Emphasis. Cohesion, expressiveness, and conﬂict were used here. Cohesion refers to the
extent to which family members are concerned about, committed to, and supportive of
one another (e.g., “There is a feeling of togetherness in our family”). In expressive
homes, family members are allowed to act openly and express their feelings (e. g., “There
are a lot of spontaneous discussions in our family”). Family conﬂict describes the degree
to which anger, aggression, and conﬂict interactions are characteristic of the home (e.g.,
“We ﬁght a lot in our family”).

Each scale has nine items, and respondents are asked whether the statement for
each item is more true or false for their family. Responses are summed, with higher
scale scores representing more cohesion, expressiveness, and conﬂict.

Discipline in Middle Childhood. The Parent Daily Report (PDR) was used to
measure physical discipline. The PDR is a revision of the same-named instrument at the
Oregon Social Learning Center (Chamberlain, 1980). Parents reported a “yes” or “no” for
the occurrence of a set of behaviors during the previous 24 hour during daily phone calls
over a six day time period (3 to mother and 3 to father, on alternating days). Information
is requested about 12 daily behaviors — six positive behaviors and six negative behaviors.
At Wave 2, the item used to measure negative control/harsh punishment was, “Within the
last 24 hours, did you spank or use other physical discipline with (your child)?” At Wave
3, the most similar item was, “Within the last 24 hours, did you use physical discipline

with (your child)?” Both items were scored as no-yes responses (0 = no; 1 = yes). A

75

.4_ i.

 

maximum score (1 or 0) was computed for each adult informant across their three
interview days. If any spanking/physical discipline was reported by an informant, this
resulted in an overall score of 1 = “yes” for that informant. Next, a maximum score (1 =
any spanking or physical discipline; 0 = none) was computed across informants,
including step-parents. If any spanking/physical discipline was reported on any day by
any respondent, the ﬁnal PDR score was a 1 = “yes”. Final PDR scores therefore reﬂect
whether or not the target child experienced physical discipline by any parent during the
report periods.

An additional instrument measured negative control exerted by parents in
relationships with their children but was only administered at Wave 3: the Parent
Perception Inventory. Because it provides more information about discipline style than
the PDR, it was also included in Wave 3 analyses.

The Parent Perception Inventory (PPI; Hazzard, Christensen, & Margolin, 1983)
for parents is an 18-item questionnaire developed to assess parenting styles. Half the
items describe parenting behaviors with a positive orientation (positive reinforcement,
comfort, talk time, involvement in decision making, time spent together, positive
evaluation, allowing independence, assistance, and non-verbal affection); these are
alternated with half that suggest a negative orientation (privilege removal, criticism,
command, physical punishment, yelling, threatening, time-out, nagging, and ignoring).
Responses are based on a ﬁve-point Likert scale (0=never; 4 = a lot). A summed total
for the negative orientation items on the PPI were used to measure negative control or

harsh punishment at Wave 3. The highest negative orientation scale score among all the

76

biological and step-parents who completed the PPI on the child and have adequate
contact with the target child were included in the analyses.
Dropp_ed Cases

Most of the missing data from the Michigan Longitudinal Study are missing by
design. Data from some families were initially not collected due to funding restraints;
when funding issues were resolved, data were then not collected from families with
children who were too old for the study wave in question (particularly at Wave 2 and
Wave 3). Much fewer data are missing due to non-administration or non-response.

Based on the criteria for inclusion in the study sample, described earlier, there
were 220 valid cases. Of the 113 cases with insufﬁcient data, 85 had less than three
waves of (at least partially) collected data, CBCL inclusive.l Seven of those meeting the
three-wave criteria were missing the CBCL at Wave 1, and 21 were missing the CBCL at
Wave 4.

Screening for differences between the study sample (n = 220) and those dropped
due to insufﬁcient data (n = 113) revealed some bias. Chi-square tests showed signiﬁcant
differences for paternal alcoholism, x2 (2, 333) = 8.93, p = .012; paternal antisocial
personality disorder, )6 (1, 322) = 15.24, p = .000; and maternal depression, x2 (3, 328) =
8.15, p = .043.2 Cases that were dropped were characterized by fathers who were more
alcoholic (85% with at least a lifetime diagnosis, versus 70% in this sample) and more

antisocial (35%, versus 16% in this sample) — as well as mothers who were more

 

I Two of the 220 cases included in the study were missing both middle childhood CBCL scores

(i.e., Waves 2 and 3), but otherwise had nearly-complete data for at least three waves. Because there were
actually very few missing data points for these cases, and the missing CBCL scores were ﬁ'om neither
Wave 1 nor Wave 4, the data were retained.

Non-signiﬁcant differences were found for maternal alcoholism, x2 (2, 333) = .28, p = .872;
maternal ASPD, x2 (r, 332) = 3.58, p = .058; and paternal depression, x2 (3, 326) = 6.89, p = .075.

77

depressed (17% of with at least moderate depression, versus 10% in this sample). Note
that sample size ﬂuctuations across analyses here derive from missing values in the
original data set by which the two groups can be compared (i.e., before imputation of
missing values for the study sample).

Regarding child behavior problems, there were also signiﬁcant differences in
maternal ratings of aggression at Wave 4, t(248) = 2.20, p = .029 — with a mean of 9.4
for those dropped and 7.1 in the study sample. However, the two groups were
statistically similar on Wave 1 aggression (mean of 10.93 versus 10.65, respectively;
t[319] = .39, p = .695), Wave 1 delinquent behavior (mean of 2.33 versus 1.88,
respectively; t[319] = 1.95, p = .105), and Wave 4 delinquent behavior (mean of 2.47
versus 1.85, respectively; L[248] = 1.44, p = .152). Nonetheless, for each of these
variables, the means for the dropped cases were higher.

In general, the 220 cases included in the study sample are less at-risk than the
families excluded from imputation due to insufﬁcient data. In some ways, then, this will
be a conservative test of the research hypotheses. Interpreted conversely, the study
sample is slightly more normative than the larger Michigan Longitudinal Study subject
pooL
Missing Data Imputation

For the 220 valid cases, the proportion of missing values for each variable is listed
in Table 6. The range of missing data was from O — 20%, with a total of 8.6%. The
average percentage of missing data was .3% at Wave 1, 18.1% at Wave 2, and 6.9% at
Wave 3. There were no missing data at Wave 4, by virtue of the criteria used for

inclusion in the sample (i.e., behavior problem ratings at Wave 1 and Wave 4 as well as

78

information on drinking onset). Recall that another criterion was participation in at least
one of the middle childhood waves (Wave 2 or Wave 3) to ensure that missing data
imputation for middle childhood measures were at least partially based on other middle
childhood measures.

Missing values were imputed using the EM (expectation-maximization) method
available in most statistical packages. The EM algorithm is an iterative procedure to ﬁnd
the maximum likelihood estimates of parameters for missing values using data from all
observed variables.

Each iteration involves an E step and an M step. In the E step, conditional
expectations for the missing data (i.e., a “best guess”) are derived from known values on
observed variables and the current estimate of parameters. In the M step, expected values
are substituted for the missing data and maximum likelihood estimation is used to
produce new parameter estimates. These updated parameter estimates are considered in
the next E step, with iterations repeating until convergence (i.e., when negligible change
in parameter estimates from iteration to iteration is obtained). The result is that set of
imputed values which best ﬁts the existing covariance matrix —- thereby maintaining
observed relationships between variables in the dataset. Stated another way, “the basic
principle of ML estimation is to choose as estimates those values that, if true, would
maximize the probability of observing what has, in fact, been observed” (Allison, 2002,
p. 13).

The EM method produces less bias in the results and is generally superior to
methods such as list-wise deletion, mean substitution, and regression imputation (Collins,

Schafer, & Kam, 2001; Graham &Hofer, 2000; Rovine & Delaney, 1990; Schafer &

79

Table 6

Proportion of Missing Cases by Vagable

 

 

Variable Number of Proportion
Intact Cases Missing

Wave 1

Aggression 220 --
Delinquent Behavior 220 --
Activity 21 8 .01
Attention 219 <.01
Reactivity 219 <.01
Age 220 --
Maternal Alcoholism and ASPD 220 --
Paternal Alcoholism and ASPD 220 --
Maternal Depression 21 8 .01
Paternal Depression 219 <.01
Wave 2

Aggression 1 83 . 17
Delinquent Behavior 183 .17
Reading 180 .18
Spelling 178 .19
Arithmetic 180 .18
Global Self-Esteem 176 .20
Social Problems 183 .17
Family Cohesion 179 .19
Family Expressiveness 179 .19
Family Conﬂict 179 .19
Spanking 180 .18
Activity 181 .18
Attention 181 .18
Reactivity l 8 1 . 1 8
Wave 3

Aggression 208 .05
Delinquent Behavior 208 .05
Reading 202 .08
Spelling 202 .08
Arithmetic 202 .08
Global Self-Esteem 206 .06

 

(table continues)

80

 

Table 6 (cont’d).

 

 

Variable Number of Proportion
Intact Cases Missing
Wave 3 (Continued)
Social Problems 208 .05
Family Cohesion 207 .06
Family Expressiveness 207 .06
Family Conﬂict 207 .06
Physical Discipline 195 .11
Negative Control 204 .07
Activity — General 205 .07
Activity - Sleep 205 .07
Task Orientation 205 .07
Wave 4
Onset Group 220 --
Aggression 220 --
Delinquent Behavior 220 --

 

Graham, 2002; West, 2001). In order to estimate missing values accurately, data must be

missing at random (MAR). Data are MAR for a variable if the pattern of missing values

does not depend on the value of that variable (Allison, 2002). Missing completely at

random (MCAR) involves an even stronger assumption: that the pattern of missing

values doesn’t depend on the value of the variable itself or any other variable in the data

set (Allison, 2002).

Little’s (1988) MCAR test uses all of the available data to produce a global chi-

square statistic which speaks to the issue of whether the data are MCAR. Because the

data in this study are presumed to be MAR (because of planned missingness and the role

of age), not MCAR, this is a conservative test. Little’s MCAR test resulted in a chi-

81

square = 1122.217 (df = 1078; p = .17) which indicates that no identiﬁable pattern exists
to the missing data; the null hypothesis against random missingness is rejected, and
missingness is considered ignorable for likelihood inferences (which require a minimum
of MAR; Little, 1988).
Analy_tical Strategy

To test Hypothesis 1 (that at the core of differences associated with drinking onset
status is a pattern of higher antisocial behavior from preschool thopgp adolescence),
levels of antisocial behavior from preschool to adolescence were compared between early
drinkers (i.e., ﬁrst drink group) and those who had not tried alcohol by 12-14 years of age
(i.e., no ﬁrst drink group). Next an autoregressive two-group (ﬁrst drink, no ﬁrst drink)
stacked model in LISREL 8.52 was run with aggression at Wave 1, Wave 2, Wave 3, and
Wave 4 measured by parental report on the CBCL (see Figure 4). This analysis was
repeated for delinquent behavior problems. Preliminary analyses suggested a unique
contribution of preschool antisocial behavior on later antisocial behavior — that is, that a
path from Wave 1 to Wave 4 would need to be included in the model for adequate ﬁt.
This ﬁts with the assumption of initial predispositions (internal or extemally-driven) that
contribute to the developmental process. Antisocial behavior becomes more normative in
adolescence, but relative severity rankings from preschool difficulties are expected to

remain stable from a hard-continuity perspective.

82

 

Behavior Problems, 3-5

\

Behavior Problems, 6-8

\

Behavior Problems, 9-11

\

: Behavior Problems, 12-14

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4.

Stacked (multi-group) model comparing early drinkers to non-drinkers at 12-14.

Conﬁgural frequency analysis (CFA; von Eye, 1990, 2002) was used to test
Hypothesis 2 (that classiﬁcation into high-high, low-low, low-high, and high-low groups
using preschool and early adolescent antisocial behavior scores signiﬁcantly overlap with
early drinking onset — such that early drinking is disproportionately associated with
strong continuity at the high end [i.e., high-high] and avoidance of drinking onset is
associated with strong continuity at the low end [i.e., low-low]). The four longitudinal
behavior classiﬁcations (e.g., recall Figure 3) were cross-tabbed with drinking onset in a
4 (low-low, low-high, high-low, high-high) x 2 (drinking onset: no, yes) conﬁguration.
Separate analyses were run for aggressive behavior classiﬁcations and the delinquent
behavior classiﬁcations.

Using CFA, researchers ask whether cells contain more or fewer cases than
expected from some chance model. A cell with more cases than expected is a “type;” a

cell with fewer cases than expected is an “antitype.” Data were analyzed under the

83

#—

   

assumption of total independence which dictates that the classiﬁcations are not related at
all; if this is true, neither types nor antitypes will be revealed. Lehmacher’s test (L,
Lehmacher, 1981) was used for signiﬁcance testing of types and antitypes with a
Bonferronni-adjusted alpha level (a* = 0.05/8 cells = .00625). If Hypothesis 2 is
supported, greater than expected frequencies will be found for the high-high and low-low
groups with respect to ﬁrst drink (i.e., there will be more cases than expected in the high-
high drinking onset cell, and there will be more cases than expected in the low-low no-
onset cell).

To test Hypothesis 3 (that discontinuity patterns are marked by intermediate
levels of risk), group differences on variables from preschool and middle childhood were
examined with Analysis of Variance (ANOVA). Analyses were run separately for
aggression and delinquent behavior classiﬁcations. Exact p-values are reported (except
where p < .001) so that the reader can evaluate the risk of Type 1 error.

Finally, regressions were employed to test Hypothesis 4 (that early risk load will
predict Wave 1 antisocial behavior, that early risk factors as well as middle childhood
variables will predict Wave 4 antisocial behavior, and that more proximate inﬂuences
will have a stronger effect than more distal inﬂuences).

Age at Wave 1 (from the CBCL administration) was included as a control
variable. Since participants were re-interviewed every three years, this is a proxy
measure for age at each wave. Recall that each wave covers a range (i.e., Wave 1, 3-5;
Wave 2, 6-8; Wave 3, 9-11; Wave 4, 12-14); if any of the measures have age-related

variance, including age in the model will help control for it. Regressions were run

84

separately for aggression and delinquent behavior, and separately within each for Wave 2
and Wave 3 predictors.

The Baseline Model includes child temperament and the composite parental
psychopathology measure for mothers and fathers. These variables reﬂect the initial risk
trajectory. It is especially important to include parental diagnostic variables in the model
because of the nature of the sample (i.e., non-alcoholic families, alcoholic families with
fathers convicted of drunk-driving, and community alcoholic families without a criminal
history). Antisocial behavior is likely to be elevated among children of alcoholics -
especially with co-morbid antisociality or depression.

The Contextual Model consists of risk and protective factors from the family
environment during middle childhood. Predictors include family cohesion,
expressiveness, and conﬂict, as well as discipline (i.e., negative control and/or physical
punishment). The Child Characteristics Model consists of individual-level risk and
protective factors from middle childhood. Predictors include academic achievement in
three domains (reading, spelling, arithmetic), self-esteem, and social problems.

The Baseline Model alone was used in a regression for Wave 1 aggression and
delinquent behavior. For Wave 4 aggression and delinquent behavior, it was tested in a
ﬁrst run to determine whether characteristics from preschool are sufﬁcient to predict later
behavioral trajectories. The other two models were then added in a second run. This
second run 1) controlled for the inﬂuence of initial characteristics on later characteristics
and 2) helped deterrrrine which variables are the most predictive of each antisocial
behavior pattern. If only baseline predictors are signiﬁcant in this second run, it would

suggest that risk for hard-continuity or change is established by preschool and that

85

subsequent inﬂuences do little to disrupt the risk trajectory.

Finally, three additional sets of control variables were added in a hierarchical
procedure. Temperament variables concurrent with the other middle childhood predictor
variables were entered to control for the possibility of temperamental variability over
time.3 Next, Wave 1 antisocial behavior was included to see whether early antisocial
behavior predicts later antisocial behavior above and beyond more proximate inﬂuences,
and to see whether other variables explain these outcomes above and beyond what is
known about behavior problems from as early as three years old. The new temperament
variables were expected to attenuate the relationship between Wave 1 temperament and
antisocial behavior, and the inclusion of Wave 1 antisocial behavior was expected to
make a signiﬁcant contribution to explained variance. Later, the effect of controlling for
contemporaneous (middle childhood) behavior problems was explored with similar
expectations.

The rule of thumb for statistical power with regression is at least 10 cases per
independent variable (V anVoorhis & Morgan, 2001). One alternate criteria is N > 50 +
8m (where m is the number of predictors) for testing the multiple correlation and N > 104
+ m for testing partial correlations for individual predictors, assuming a medium-sized
relationship between the dependent and independent variables (Green, 1991; VanVoorhis
& Morgan, 2001). The largest model in this study contained 21 variables, with 220 cases
(i.e., 10.5 cases per predictor, 50 + 8m = 218, and 104 + m = 125). Sample size therefore

should not be prohibitive, although some caution is warranted.

 

3 Parental psychopathology - the other initial risk trajectory component - may also change over

time but these changes should be accounted for by family environment characteristics (see literature
review).

86

CHAPTER 4
RESULTS

Onset of Fist Drig
Hypothesis 1. The primary pathway into early ﬁrst alcohol use begins with

heightened levels of antisocial behavior from as early as 3-5 years of age, onward.

Means and standard deviations for each variable by drinking onset status can be
found in Table 7. Graphical depictions can be found in Figure 5 for aggression means
and Figure 6 for delinquent behavior means; recall that the possible range of CBCL
scores was 0 to 40 for aggression (20 items), and 0 to 24 for delinquent behavior (12
items). T-tests revealed signiﬁcant differences in symptomatology by drinking onset
group for Wave 4 aggression and for delinquent behavior at three time periods: Wave 1,
Wave 2, and Wave 4 (see Table 7). Figures 5-6 show much more marked group
differences for delinquent behavior than for aggression.

In general, levels of aggression declined for both groups from 3-5 to 12-14 years
of age. The only group difference was for aggression at 12-14, with early drinkers rated
by their mothers as more aggressive. On the other hand, early drinkers were more
delinquent than their never-drinking peers at all ages except 9-11- when the level of
symptomatology for early drinkers appears to diminish and converge with that of those
who have never tried alcohol, before rising again in early adolescence. Early drinkers
therefore engaged in more delinquent activities at most time periods, with distinctly more
aggression and delinquent behavior in the transition from late childhood to early
adolescence.

To more precisely specify developmental differences in antisocial behavior

87

Table 7

Means (and Standard Deviations) by Drinking Onset Group for Aggression and

Delinquent Behavior

 

 

Variable Wave No First Drink First Drink
M (Sill _M_ (SD) t

Aggression 1 10.47 (5.57) 11.20 (5.51) -.85

2 8.61 (5.14) 10.15 (5.40) -1.90

3 8.21 (5.69) 8.31 (5.87) -.11

4 6.58 (5.12) 8.47 (5.96) -2.27*
Delinquent. 1 1.67 (1.28) 2.53 (1.96) -3.75 **
Behavior 2 1.76 (1.32) 2.96 (1.99) -5.01***

3 1.57 (1.43) 1.99 (1.96) -1.72

4 1.45 (1.77) 3.05 (2.70) -5.00 ***

 

Note. ”* p < .001. ** p < .01. * p < .05; df= 218 for all tests; the same pattern of
signiﬁcance was found when equal variances were not assumed.

between early drinkers and non-drinkers, structural equation modeling was used to

examine the inﬂuences of early antisocial behavior on later antisocial behavior. An

autoregressive two-group (ﬁrst drink, no ﬁrst drink) stacked model in LISREL 8.52 was

run separately for aggression and delinquent behavior problems at Wave 1 — 4, testing the

model from Figure 4. All parameters were ﬁrst allowed to vary across drinking onset

groups. When a tenable model was obtained, these parameters were constrained to be

equal in an invariant model to test for signiﬁcant group differences. Covariance matrices

for the ﬁrst drink and no ﬁrst drink groups are presented in Table 8 for aggression and

Table 9 for delinquent behavior.

88

 

 

 

 

 

 

 

 

 

 

 

 

 

+ First Drink Group

8 ..
2 + No First Drink

4 at _. -_*L- ,7..--.7,,-,7 - c. ,7 - -_.-__ Group

2 .. i m A f L 4., km _ __

O I I I

Wave 1 Wave 2 Wave 3 Wave 4
Figure 5.

Mean aggression scores from Waves 1-4 for ﬁrst drink versus no ﬁrst drink

group.

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

+ First Drink Group
8
E + No First Drink
Group
1 _
0.5
0 I I I
Wave 1 Wave 2 Wave 3 Wave 4
Figure 6.

Mean delinquent behavior scores from Waves 1-4 for ﬁrst drink versus no ﬁrst

drink group.

89

 

Table 8

Covariance Matrix for Aggression by First Dﬁak Onset Group

 

 

 

 

 

 

 

 

 

 

FIRST DRINK = NO Wave 1 Wave 2 Wave 3 Wave 4
Aggression Wave 1 30.9943

Aggression Wave 2 15.8370 26.4615

Aggression Wave 3 12.0781 21.0290 21.3235

Aggression Wave 4 11.4691 15.7427 19.9636 26.1991
FIRST DRINK = YES Wave 1 Wave 2 Wave 3 Wave 4
Aggression Wave 1 30.3481

Aggression Wave 2 19.9652 29.1121

Aggression Wave 3 17.6258 22.6642 34.5139

Aggression Wave 4 17.9778 16.9324 20.9610 35.5502
Table 9

Covariance Matrix for Delinquent Behavior by First Drink Onset Group

FIRST DRINK = NO Wave 1 Wave 2 Wave 3 Wave 4
Delinquent Wave 1 1.6260

Delinquent Wave 2 .7679 1.7528

Delinquent Wave 3 .6228 1.0534 2.0516

Delinquent Wave 4 .7012 1.2922 1.4708 3.1519
FIRST DRINK = YES Wave 1 Wave 2 Wave 3 Wave 4
Delinquent Wave 1 3.8465

Delinquent Wave 2 1.8846 3.9576

Delinquent Wave 3 1.3157 2.8008 3.8338

Delinquent Wave 4 2.7664 3.3493 2.6278 7.2862

 

The maximum likelihood (ML) method of estimating free parameters was

employed; ML is a standard procedure used throughout much of the literature. Several

criteria will be used to evaluate model ﬁt. Of primary interest is the chi-square (x2) test,
which measures discrepancies between the actual sample covariance matrix and the one
generated by the estimated (ﬁtted) model; x2 values should be low with p>.05 (relative to
the degrees of ﬁ'eedom), indicating that the data and model matrices are not signiﬁcantly
different. Notably, it was the normal theory weighted least squares chi-square that was
used for purposes of evaluation. If the model is tenable, the root mean square error of
approximation (RMSEA) — which is a measure of discrepancy — should be less than .05.
In addition, a value of at least .90 should be obtained for the normed ﬁt index (N FI) and
the comparative ﬁt index (CFI), which indicate how much better the model ﬁts (i.e.,
explains the observed variances and covariances) relative to a null model.

For aggression, the ﬁrst run tested a model in which paths exist between each
adjacent wave (i.e., without the path from Wave 1 to Wave 4). The result was poor
model ﬁt: x2= 16.41, df= 6, p = .01, RMSEA = .126, NFI = .96, CFI = .97. The path
from Wave 1 to Wave 4 aggression was then added, resulting in a model with acceptable
ﬁt statistics: x2 = 3.79, df = 4, p = .43, RMSEA = .000, NFI = .99, CFI = 1.00. This
speciﬁcation was a signiﬁcant improvement, A x2 = 12.62, A df = 2, p < .01.

For delinquent behavior, the ﬁrst run tested a model in which paths exist between
each adjacent wave (i.e., without the path from Wave 1 to Wave 4). The result was poor
model ﬁt: x2 = 36.17, df= 6, p < .001, RMSEA = .215, NFI = .87, CFI = .89. The path
from Wave 1 to Wave 4 delinquent behavior was then added, resulting in a model with

the following ﬁt statistics: 38 -= 24.81, df= 4, p < .001, RMSEA = .219, NFI = .92, CFI =

91

.93. This speciﬁcation was a signiﬁcant improvement, such that A x2 = 11.36, A df = 2, p
< .01, but additional changes were needed to meet acceptable ﬁt criteria. Modiﬁcation
indices suggested a direct path between Wave 2 and Wave 4 delinquent behavior. A
tenable model was achieved when this parameter was included in the model: )8 = 2.34,
df = 2, p = .31, RMSEA = .039, NFI = .99, CFI = 1.00.4 It was also a signiﬁcant
improvement in ﬁt with A x2 = 9.02, A df = 2, p < .05.

The common metric completely standardized solution can be found in Figure 7
for aggression and Figure 8 for delinquent behavior. Parameter estimates for the group
with no ﬁrst drink by 12-14 years of age are above each arrow; parameter estimates for
the ﬁrst drink group are below each arrow.5

For aggression, the severity of behavior problems at each wave signiﬁcantly
contributed to severity at adjacent waves for both groups. In addition, the severity of
problems at 3-5 signiﬁcantly predicted the severity of problems at 12-14 for both groups.
This link between aggression in preschool and early adolescence was needed to achieve
good model ﬁt.

For delinquent behavior, one additional path was required to obtain adequate
model ﬁt: a link between delinquent behavior at 6-8 and delinquent behavior at 12-14.

Although not hypothesized, this path makes sense given the developmental trajectories

 

4 When this path was added to the aggression model, x2 = 0.96, df = 2, p = .62, RMSEA = .000,

NFI - l. 00, CFI = 1 .00 It was not an improvement in ﬁt, however, with A x2= —.2 83, A df = 2, p > .05.
The path from Wave 2 -) Wave 4 also was not statistically signiﬁcant for either group.

Error terms are not depicted. The parameter estimates for the error terms were 1 .01, .,68 .,48 and
.47 for aggression at Waves 1-4, respectively, for the no first drink group and 98, .59 .,51 and .72 for
aggression at Waves 1-4, respectively, for the ﬁrst drink group. Parameter estimates for the error terms
were .75, .60, .57, and .45 for delinquent behavior at Waves 1-4, respectively, for the no ﬁrst drink group
and .1.77, 1.32, .74, and .95 for delinquent behavior at Waves 1-4 for the ﬁrst drink group.

92

 

[Aggressiom 3-5 I

. 54
JR

bggression, 6-8 1
72 *

.71\‘

LAggression, 9-1 1 I

. 60
AR

t l Aggression, 12-14]

 

*

 

 

 

 

4t

.16"
.35“

 

 

 

 

Figure 7.
Common metric standardized solution for aggression by drinking onset group.
Note: parameter estimates are above arrows for the no drink group and below arrows

for the ﬁrst drink group; * p < .05.

 

[ Delinquent Behavior, 3-5 I

\46"
.48* .32...

[ Delinquent Behavior, 6-8 I _41 :1:

\58‘*
.68 *

l Delinquent Behavior, 9-11 1

\33"
.ll 1'

t [Delinquent Behavior, 12-14]

 

 

 

 

 

 

 

. 03
.29 *

 

 

 

 

Figure 8.
Common metric standardized solution for delinquent behavior by drinking onset
group. Note: parameter estimates are above arrows for the no drink group and below

arrows for the ﬁrst drink group; * p < .05.

93

for delinquent behavior depicted in Figure 6; levels of delinquent behavior converged for
the two groups at Wave 3 (9-11). The added path bypasses this convergence and links
the elevated levels of delinquent behavior observed among early drinkers at 6-8 to their
elevated levels at 12-14. The decline in delinquent behavior during late middle childhood
among early drinkers also probably explains why the path from Wave 3 to Wave 4
delinquent behavior was not signiﬁcant for this group.

Interestingly, the path from Wave 1 to Wave 4 delinquent behavior was
signiﬁcant for early drinkers only. This direct path represents the inﬂuence of early
predispositions on adolescent outcomes and supports the idea of hard-continuity for this
subset of individuals.

Comparing each model to an invariance model (i.e., parameters between groups
constrained to be equal) revealed signiﬁcant differences in the covariance structure by
ﬁrst drink onset status for delinquent behavior but not aggression. For aggression, the
invariance model yielded )8 = 12.35, df = 12, p = .42, RMSEA = .012, NFI = .97, CFI =
1.00. Relative to the free parameter model, A x2 = 8.60, A df = 8, p > .05. For delinquent
behavior, the invariance model yielded x2 = 68.99, df = 11, p = .00, RMSEA = .220,

NFI = .82, CFI = .85. The free parameter model for delinquent behavior showed
signiﬁcantly better ﬁt: A x2 = 66.65, A df = 9, p < .01. Again, group differences in this
developmental model were much more marked for delinquent behavior than aggression.

Finally, the relative strength of indirect and direct effects were examined across
drinking onset groups for the added paths of Wave 2 on Wave 4 behavior (for delinquent)
and Wave 1 on Wave 4 behavior (for delinquent and aggressive). Standardized effects

are reported in Table 10. The most striking differences between groups were in the

94

effects of preschool symptomatology on delinquent behavior during early adolescence.
The total effect was much stronger among early drinkers (.51 versus .28). This disparity
was largely the result of differences in the direct effect (i.e., Wave 1 -) Wave 4) rather
than the indirect effect (i.e., Wave 1 through the middle childhood waves to Wave 4).
The direct effect of Wave 1 on Wave 4 delinquent behavior was .29 for early drinkers

compared to .04 for those who had not yet tried alcohol by 12-14 years of age.

Table 10
Direct and Indirect Effects of Paths from Non-Adjacent Wages in Structural Equation

Models for Aggression and Delinquent Behavior

 

 

 

 

Standardized Effect
Relationship Group Direct Indirect Total
Aggression
Wave 1 to Wave 4 No First Drink .16 .25 .41
First Drink .31 .21 .52
Delinquent Behavior
Wave 1 to Wave 4 No First Drink .04 .24 .28
First Drink .29 .23 .52
Wave 2 to Wave 4 No First Drink .32 .22 .54
First Drink .41 .07 .48

 

95

Hypothesis 2. Preschool and adolescent levels of antisocial behavior problems
are distinguishing characteristics along the developmental course leading to and
through drinking onset —— and more speciﬁcally, antisocial behavior problems in
these two time periods can be used to categorize individuals into groups that

differ in the likelihood of early ﬁrst drink.

Results from the Conﬁgural Frequency Analysis (CF A) are presented in Table 11
for aggression and Table 12 for delinquent behavior. The CFA was signiﬁcant for
delinquent behavior (Pearson’s x2 = 18.38 for df = 3, p < .001) but not for aggression
(Pearson’s x2 = 6.31 for df = 3, p > .05).

There were neither types nor antitypes for aggression. For delinquent behavior,
there were more cases than expected by chance for those who were low in delinquent
behavior during both preschool and early adolescence and who had not had a ﬁrst drink
by 12-14 years of age (i.e., the “1 1” type). There were fewer cases than expected by
chance for low-low early drinkers (i.e., the “12” antitype). The opposite pattern was
found for those who were high in delinquent behavior during both time periods.
Speciﬁcally, there were more cases than expected by chance for the high-high group with
drinking onset (i.e., the “42” type), and fewer cases than expected by chance for high-
high children who had not yet tried alcohol (i.e., the “41” antitype). In other words, those
in the high-high delinquent behavior group were more likely than chance to be early
drinkers, whereas those in the low-low delinquent behavior group were less likely than

chance to have tried alcohol by 12-14 years of age. This observed relationship between

96

continuity in delinquent behaviors and likelihood of early AF D is consistent with the

hard-continuity model of problem behavior.

Table 11

Configurations for Aggression Group and First Drink Onset

 

 

Observed Expected
AO Frequency Frequency L_ Type/Antitype
1 l 70 66.75 1.03 --
12 19 22.25 -1.03 --
21 20 20.25 -0.12 --
22 7 6.75 0.12 --
31 33 29.25 1.53 --
32 6 9.75 -1.53 --
41 42 48.75 -2.30 --
42 23 16.25 2.30 --

 

Note. A = aggression group; O = ﬁrst drink onset. Numerals in A0 column represent
ordered pairs of variable categories. Response categories for aggression were 1 =
low-low, 2 = low-high, 3 = high-low, and 4 = high-high. For ﬁrst drink onset, 1 = no
onset and 2 = onset. L stands for Lehmacher’s (1981) test statistic; Bonferroni
adjusted alpha, or" = .0062500 was used; * signiﬁcant at 01* = .0062500.

97

Table 12

Conﬁggations for Delingent Behavior Group and First Drink Onset

 

Observed Expected

 

DO Frequency Frequency L Type/Antitype
ll 65 57.00 2.61 * Type

12 11 19.00 -2.61 * Antitype

21 19 19.50 -0.24 --

22 7 6.50 0.24 --

31 40 35.25 1.80 --

32 7 11.75 -1.80 --

41 41 53.25 -4.07 * Antitype

42 30 17.75 4.07 * Type

 

Note. D = delinquent behavior group; O = ﬁrst drink onset. Numerals in D0 column
represent ordered pairs of variable categories. Response categories for delinquent
behavior were 1 = low-low, 2 = low-high, 3 = high-low, and 4 = high-high. For ﬁrst
drink onset, 1 = no onset and 2 = onset. L stands for Lehmacher’s (1981) test

statistic; Bonferroni adjusted alpha, or“ = .0062500 was used; * signiﬁcant at or* =
.0062500.

Summary. Hypothesis 1 was partially supported for delinquent behavior but not
supported for aggression. Contrary to expectations, there was a convergence of
delinquent behavior problem ratings between early drinkers and those who had not tried
alcohol by 12-14 years of age during late-middle childhood (9-11), and few differences
by drinking onset status for aggression from preschool to adolescence. Hypothesis 2 was

fully supported for delinquent behaviors only.

98

The Development of Antisocial Behavior
Hypothesis 3: On all measures of risk from preschool (temperament, parental
psychOpathology) and middle childhood (academic achievement, self-esteem,
social problems, family cohesion, family expressiveness, family conﬂict, physical
discipline/negative control), scores will be least favorable for the high-high group,
moderate for the low-high and high-low groups, and most favorable for the low-
low group. If supported, this would corroborate the ﬁndings of Fergusson et al.

(1996).

The high-high group is expected to have the following characteristics relative to
the low-low group: more difﬁcult temperament (i.e., high activity, short attention
span, more reactivity) and more parental psychopathology (mothers and fathers)
at 3-5 years of age; as well as less academic achievement (i.e., lower scores on
reading, spelling, and arithmetic), lower self-esteem, more social problems, less
family cohesion, less family expressiveness, more family conﬂict, and more
physical/negative punishment at 6-8 and 9-11 years of age. The high-low and
low-high groups are expected to have in-between scores on each of these

dimensions.

Before proceeding to risk proﬁle comparisons, it may be useful to examine
aggression and delinquent behavior groups on the variable that deﬁnes them — that is, on
aggressive and delinquent behavior, respectively. The developmental patterns of

aggression by group are depicted in Figure 9, and the developmental patterns of

99

delinquent behavior by group are depicted in Figure 10; again, the possible range of
CBCL scores was 0 to 40 for aggression (20 items), and 0 to 24 for delinquent behavior
(12 items). Means and standard deviations for aggression and delinquent behavior by
wave and behavior problem group can be found in Table 13.

There were behavior problem classiﬁcation group differences for all aggression
and delinquent behavior scores from Wave 1 to Wave 4, inclusive (see Table 13).
Speciﬁcally, there were signiﬁcant group differences between aggression groups at Wave
1 (E [3, 216] = 167.36, p <.001), Wave 2 (E [(3, 216] = 32.77, p <.001), Wave 3 (E [3,
216] = 27.84, p <.001), and Wave 4 (E [3, 216] = 138.53, p <.001). There were also
signiﬁcant group differences between delinquent behavior groups at Wave 1 (If [3, 216] =
98.76, p <.001), Wave 2 (E [(3, 216] = 32.29, p <.001), Wave 3 Q [3, 216] = 24.66, p
<.001), and Wave 4 (E [3, 216] = 91.88, p <.001).

Bonferroni post-hoe comparison tests revealed signiﬁcant differences between
pairs of groups as well (at or = .05). At Wave 1, all groups differed except for high-high
versus high-low aggression and for low-low versus low-high delinquent behavior. Stated
conversely, there were no signiﬁcant baseline differences between the high-high versus
high-low groups for aggression and the low-low versus low-high groups for delinquent
behavior.

High ratings of aggression and low ratings of delinquent behavior at Wave I
meant the same across classiﬁcations. On the other hand, it appears that the low-high
group for aggression was already more aggressive than the low-low group. Similarly, the

high-low group for delinquent behavior was already less delinquent than the high-

100

 

 

14

 

—-— High-High
—I— High-Low
+ Low-High
—x-— Low-Low

 

 

 

Mean

 

 

 

 

 

 

 

 

 

 

 

Wave 1 Wave 2 Wave 3 Wave 4

 

 

 

Figure 9.

Mean CBCL aggression scores from Wave 1 to Wave 4, by aggression group.

 

 

 

 

 

 

«mist-High
+ High-10W
+ Low-High

+ Low-Low

 

 

 

 

 

 

 

 

 

 

 

 

 

Wave 1 Wave 2 Wave 3 Wave 4

 

 

 

Figure 10.

Mean CBCL delinquent behavior scores from Wave 1 to Wave 4, by delinquent behavior

group.

101

move... .S.Vo: .Sovott

3. N m goutnwi mama? .35..ch a.
mo. N m got—Ami mama?» 264-304 m
3. N m £32254 mama? Boursoq o
3. m m .2322: Se? Enema 6
no. N m $5-23 assoc, BEBE a
3. N m .Boqtspoq mama? nwféwi <

 

 

102

Gas m; are me. am: ”mm 5.8 mm. hastens?

E: 42 as: e: so: so; A38 8. Estimates

se.: cum cm: a? can: :3 so: 8; 2:36.53

and as ﬁne 23 5.8 mm. are S. patios;
cogmnom BREE—on

33¢ 8.2 se.: 2.4 are 8.: ﬁne 3m hastens?

£59 8.2 ans as $22 :2: 5.3 86 22:28.63

23¢ cod 82; was and Bo 82¢ cow 2:268?

and 8.2 and 3.2 an: a: and an otloaa
ac_mmocwwos

Slmc 2 almv .2 aw .2 a 2

Ease: 23-3.3 s@323 23-23 osaea>

 

 

 

 

$6.5 Beacon Boson—zen >9 Sear—om €255.60 cam aide—U commmocwﬁx 3 eommmocwwd 8m Boast/0Q EMUSSm 28. 832

2 £an

high group. Some caution in interpreting differences between these group pairs is
therefore warranted as initial problem behavior severity was not truly equivalent.

At subsequent waves, classiﬁcations ﬁt with expected patterns. At Wave 2, all
groups differed except for low-high and high-low on both aggression and delinquent
behavior. These groups were also not statistically signiﬁcant at Wave 3. This reﬂects the
crossing over of these groups on behavior problem severity during middle childhood. At
Wave 3, two additional pairs of groups were not statistically different: high-high versus
low-high and low-low versus high-low for both aggression and delinquent behavior.
These groups were also statistically similar at Wave 4. The implication here is that the
cross-over has already taken place by Wave 3; groups high and low in behavior problems
at Wave 4 have converged by 9-11 years of age.

Means and standard deviations by aggression group on the risk factors can be
found in Table 14 (Wave 1 variables), Table 15 (Wave 2 variables), and Table 16 (Wave
3 variables). Corresponding information for the delinquent behavior groups are in Tables
17-19. Group differences on each variable were tested using a one-way analysis of
variance (AN OVA). Temperament variables from middle childhood were also examined
because of their inclusion as control variables in the testing of Hypothesis 4 (discussed
later).

Aggression. At Wave 1, there were signiﬁcant aggression group differences for
reactivity (E [3, 216] = 11.36, p <.001), maternal psychopathology (E [3, 216] = 5.60, p
=.001), and paternal psychopathology (E [3, 216] = 2.66, p = .049).6 At Wave 2, there

were signiﬁcant aggression group differences for reading (E [3, 216] = 2.78, p =.042),

 

6 Non-signiﬁcant differences for Wave 1 aggression groups were as follows: For age, If [3, 216] =

.37, p =.773; for activity, E [3, 216] = 1.23, p =.300; and for attention, E [3, 216] = 2.29, p =.079.

103

global self-esteem (E [3, 216] = 2.96, p =.033), social problems (E [3, 216] = 10.36, p
<.001), family conﬂict (E [3, 216] = 5.05, p =.002), and all three temperament variables:
activity (E [3, 216] = 2.66, p =.049), attention (E [3, 216] = 3.97, p =.009), and reactivity
(E [3, 216] = 14.61, p <.001).7 At Wave 3, there were signiﬁcant aggression group
differences for social problems (E [3, 216] = 7.06, p <.001), family conﬂict (E [3, 216] =
4.57, p =.004), negative control (E [3, 216] = 7.65, p <.001), and all three temperament
variables: activity-general (E [3, 216] = 10.15, p <.001), activity-sleep (E [3, 216] = 3.44,
p =.018), and task orientation (E [3, 216] = 3.61, p =.014).8

Pair-wise comparisons were investigated using Bonferroni post-hoe tests (with 01
= .05). At Wave 1 (see Table 14), more aggressive children (i.e., high-high and high-low
individuals) were more reactive than those with initially lower levels of aggression (i.e.,
low-low and low-high). Maternal psychopathology was lower for the low-low aggression
group compared to all others, and paternal psychopathology was lower for the low-low
aggression group than the high-high aggression group.

At Wave 2 (see Table 15), there were no pair-wise differences on reading or
activity despite signiﬁcant overall AN OVAs. Reactivity was higher in the high-high
group than for all others. The high-high group could also be distinguished from the low-
low group by more social problems and family conﬂict as well as a shorter attention span.
Fewer social problems were associated with a remitting pattern of aggressive behavior

(i.e., high-low versus high-high) and so good peer relations may be an important

 

7 Non-signiﬁcant differences for Wave 2 aggression groups were as follows: For spelling, E [3,

216] = 2.11, p =.100; arithmetic, E [3, 216] = 1.11, p =.346; family cohesion, E [3, 216] = 2.39, p =.070;
and family expressiveness, E [3, 216] = 1.31, p =.273.

8 Non-signiﬁcant differences for Wave 3 aggression groups were as follows: reading, E [3, 216] =
.57, p =.638; spelling, E [3, 216] = 1.36, p =.256; arithmetic, E [3, 216] = 1.15, p =.331; global self-esteem,

104

promotive factor — or, alternatively, poor peer relations may be a risk factor that
maintains the developmental trajectory among highly aggressive children. Lower self-
esteem was associated with an escalating pattern of aggressive behavior (i.e., low-high
versus low-low) and therefore may be an important risk factor.

At Wave 3 (see Table 16), the high-high aggression group could be distinguished
from the low-low group by more social problems, family conﬂict, negative parental
control, and general activity. Fewer social problems were again associated with a
remitting pattern of aggressive behavior (i.e., high-low versus high-high), as was a lower
activity-general level. More family conﬂict and activity-general combined with less task
orientation were characteristics of an escalating pattern of aggressive behavior (i.e., low-
low versus low-high).

Because of its dichotomous nature, mean group differences on spanking and
physical discipline could not be examined. Chi-square goodness-of-ﬁt tests were
therefore employed. At Wave 2, there was a negligible relationship between aggression
group and spanking, x2 (3, 220) = 2.13, p = .546. At Wave 3, the relationship between,
aggression group and physical discipline was signiﬁcant, )8 (3, 220) = 14.61, p = .002.
This was such that 4.5% of the low-low group, 25.9% of the low-high group, 5.1% of the
high-low group, and 18.5% of the high-high group had parents who reported using

physical discipline at 9-11 years of age.

 

E [3, 216] = 2.35, p =.074; family cohesion, E [3, 216] = 2.29, p =.079; and family expressiveness, E [3,
216] = .90, p =.444.

105

3. v m .264-nw5 £552 awn-4-264
mo. v a 264.43: 25.82 264-364
mo. v m Ema->64 £6.82 264-364
8. v a .26in one? hem-ea:
8. v a. .6523 means gem-ems:
mo. v m .264-264 9562 swam-Ami

mo.VQ* ._o.VQ..:.. ._oo.va 3.2..

<mUQmu.

 

 

4%.: mum on. a N3 an: o: 5.: of < ... .eoaa aeoaa
am. : :._ an s m: 48. c of 43.8 as a o a r. .2662 se.-ea:
an: 2:” so: can an: 23 so. 3 EN an a a :- creoaom
93.3 2.4 53 3.4 and New and a? 22222
as. c a: 3.: 63 24.: m2 5.: a? 2226.4
628 2.4 8o. 3 a? 95.8 4:. 828 N2. owe.
ﬁg .2 aw 2- 4.qu 2 Slmv .2-
2312: 23-223 N22.3.23 23-23 osaea>

 

 

maniac; _ 02m? 6.4 @380 common-«mac 2n AmeoquoQ 3.65% can 332

 

E 2an

106

mo. v m .3644me 6:662 awn-H-364 a
mo. V m 364-4%: mam—62 364-364 6
mo. v m BEE-364 £5.82 364-364 6
no. v 2 23.65 see? 226-226 6
8. v a 226.23 asses 226-226 a
mo. v a .364-364 2582 235-:me <

 

8m. 6 on,” 22.6 :3 an 6 in am. 6 SN 6 a a r... 326.com
$26 own 3.6 an 5.6 an 63.6 as a r. 2353..
£66 22 :66 2: $66 o: 3.6 z: .. 6226.4.
56.6 2.4 86.6 m3 636 8.4 43.6 82 a a... 6:266 2252
$6 No.6 $6 $6 666 8.6 5.6 3.6 aseozaoaxm 2256
84.6 a? 5.6 8.2 86.6 8.2 866 a: 86266 383
f6 8.». 866 SN 23.6 SN 3.6 a? or :- asozea 663
2.8.6 3.3 62.6 3.: 26.6 3.2 96.6 2.8 a .. eooameam 22.26
48.46 R? 5.26 3.8 6236 Sea $.26 8.8 2685.5.
66.6 3.3 826 2.8 3226 his 5.26 on? 62563
5.66 8.3 $3.6 8.8 62. _ 6 8.8 636 3.3 .. weeaom

 

Alma 2 8 m6 2 ﬁle 2 2166 2
3.3-26.3 23.263 3.323 23-23 63362

 

 

8368; N 6263 .66 96.5 66632124 26 $666269 6266285 6:3 682

2 636-4.

107

mo. v m 364-693 mam“? FEE-364
mo. v m .364-nwE mama-3 364-364
mo. v m h.35-364 9569» 364-364
8. v 6 .23-3: 82% 226-85
8. v m .8523 822, 85-2.6:
mo. v a .364-364 mama? :wE-nwf

<mUQl-ULL-

 

 

8.6 :1: 866 8.8 866 :6 $6 8.8 a _.. 865826 25
66.6 2.2 88.6 o; c: .6 8.2 63.6 :8 ... 885-2323.-
§6 8.2 8.6 8.2 46-6 N3: 666 2.2 a 3 :- 482562388.
636 8.6 48.6 8.: 8.6 8.6 :66 8.2 < :- 6286 02282
48.6 NS- 88. 6 £8 88.6 2:. 8. 6 22 a < ...... 3:28 2252
so. 6 MS 846 :6 A8. 6 8.6 8. 6 86 305232226 2:53
8.6 m: cm. 6 82 666 3.2 an. 6 3: 22886 2252
46.6 63 48. 6 :6 an 6 :.~. 48. 6 6: u < :- 28:on 28m
62.6 8.8 666 8.8 5.6 :6 42.6 8.8 8863-63 2326
336 8.8 $36 $8 $36 8.6: 66.26 8.2: 23655.
866 2.8 :36 8.8 $26 2.8 81.6 8.8 8525
£86 8.8 4866 2.8 4836 2.6: 836 8.2: 8632
436 2 ad 2 Ela 2 al.-6 2
266.26: 23.262 33.23 23-23 2823

 

 

 

832.6; m 9&3 66.6 66666 commmoﬂa‘ >6 86666360 6.86465 6666 £862

64 636-4-

108

Delinquent Behavior. At Wave 1, there were signiﬁcant delinquent behavior
group differences for attention (E [3, 216] = 3.08, p =.029), reactivity (E [3, 216] = 5.89,
p =.001), and maternal psychopathology (E [3, 216] = 4.54, p =.004).9 At Wave 2 there
were signiﬁcant delinquent behavior group differences for social problems (E [3, 216] =
6.77, p <.001), family conﬂict (E [3, 216] = 5.18, p =.002), and all three temperament
variables: activity (E [3, 216] = 3.02, p = .031), attention (E [3, 216] = 4.83, p =.003),
and reactivity (E [3, 216] = 6.36, p <.001).lo At Wave 3, there were signiﬁcant
delinquent behavior group differences for social problems (E [3, 216] = 3.30, p =.021),
family cohesion (E [3, 216] = 3.66, p =.013), negative control (E [3, 216] = 5.87, p
=.001), and activity-general (E [3, 216] = 8.63, p_ <.001).”

Pair-wise comparisons were investigated using Bonferroni post-hoe tests (with a
= .05). At Wave 1 (see Table 17), there were no group differences for attention despite a
signiﬁcant overall ANOVA. The high-high delinquent behavior group was more reactive
than those with initially low levels of delinquent symptomatology (i.e., low-low and low-
high) and experienced more maternal psychopathology than the low-low group.

At Wave 2 (see Table 18), there were no group differences on activity despite a
signiﬁcant overall ANOVA. The high-high group experienced more social problems and

family conﬂict, had a shorter attention span, and was more reactive compared to the low-

 

9 Non-signiﬁcant differences for Wave 1 delinquent behavior groups were as follows: age, E [3,

216] = 1.56, p =.201; activity, E [3, 216] = .30, p =.825; and paternal psychopathology, E [3, 216] = 1.31,

¥0=.27 1.

Non-signiﬁcant differences for Wave 2 delinquent behavior groups were as follows: reading, E [3,
216] = .76, p =.520; spelling, E [3, 216] = 1.51, p =.214; arithmetic, E [3, 216] = 1.16, p =.327; global-self-

esteem, E [3, 216] = .08, p =.972; family cohesion, E [3, 216] = 2.29, p =.080; and family expressiveness, E
[3, 216] = .35, p =.789.

109

low group. More family conﬂict and shorter attention spans were also associated with
persisting rather than remitting delinquent behaviors (i.e., high-high versus high-low),
and may therefore be risk factors that maintain the developmental trajectory for children
who already exhibit elevated symptomatology (or conversely, improvement is more
likely in the absence of these risks). The high-high group also had more social problems
compared to the low-high group.

At Wave 3 (see Table 19), the high-high group had more social problems,
negative control, and activity-general but less family cohesion than the low-low group.
More negative control and activity-general were also associated with the persisting rather
than remitting pattern (i.e., high-high versus high-low).

Because of its dichotomous nature, mean group differences on spanking and
physical discipline could not be examined. Chi-square goodness-of-ﬁt tests were
therefore employed. At Wave 2, there was a negligible relationship between delinquent
behavior group and spanking, x2 (3, 220) = 1.10, p_ = .778. At Wave 3, the relationship
between delinquent behavior group and physical discipline was signiﬁcant, )8 (3, 220) =
12.11, p_ = .007. This was such that 5.3% of the low-low group, 23.1% of the low-high
group, 4.3% of the high-low group, and 18.3% of the high-high group had parents who

reported using physical discipline at 9-11 years of age.

 

11 Non-signiﬁcant differences for Wave 3 delinquent behavior groups were as follows: reading, E [3,

216] = 1.25, p =.293; spelling, E [3, 216] = 1.90, p =. 130; arithmetic, E [3, 216] = 1.74, p =.159; global
self-esteem, E [3, 216] = .79, p =.502; family expressiveness, E [3, 216] = .21, p =.891; family conﬂict, E
[3, 216] = 2.24, p =.085; activity-sleep, E [3, 216] = 1.60, p =. 190; task orientation, E [3, 216] = 1.93,

p =. 126.

110

8.1.... 8.1....- 68.21....

3. v m .364-swmm 363 $5364 8.
mo. v a 364-32 3%? 364-364 6
mo. v m Ema-364 323 364-364 6
8. v a .23-22: 8.22 22:22: 8
8. v a £26.23 See, .2622: 2
mo. v a .364-364 383 :wE-nwmm <

 

 

866 33 26.6 NS 62.6 86 62.6 :6 .666 6833
62.6 26 62.6 8.8 666 a: 62.6 6.2 < r. .662 68322
5.6 82 E6 83 6.6 6.3 666 EN 3 2.. 622886
$6 a? 866 6% $66 8.6 8.6 N3 _. 8653.
A86 6: 5.6 22 866 86 62.6 8.6 6363..
8.6 6....- 86 8.6. 68. 6 8a 8.6 m:- 8....
R46 2 5'6 2 6'6 2 246 2
262-266 23-2.3 262-23 23-23 83853

 

 

85164.8 > 4 63.63 66.4 66660 66664634 Eo:u::oQ 2649666363 6:266me 663 8862

2 636-4.

111

mo.vo * .—o.Vo ”I. ._oo.Vo “I...

mo. v a $34.35 383 .35-302

 

 

3. v 6 $6433: mama?» Boa-Boa M.

mo. v m Ema-Bog 933 304-30-— 6

3. v a £3-33 8e? 33-33 u

8. v m 33-33 35> 33-33 3

mo. v m god->5..— msmb> .35-:92 <

696 m2 326 33 A36 33 3.6 83 < t... 3383

5.6 8.3 3.6 $6 3.6 3.0 23.6 8.6 o < 1. 86:33

826 :1 866 6.. $66 a; :66 63 .. $63.

95.6 m3 866 23 A86 33 E6 33 3 ...... 6380 3.5..“

336 36 3.6 6% :66 36 5.6 $6 30533233 383

3.6 23 £06 Gs 3.6 8.3 5.6 3: 83:8 363

5.6 m: 33. 6 33 an. 6 36 E. 6 m: m < t; 85383 38m

2 16 3.3 306 3.3 336 3.3 60.6 3.8 5023.63 320

336 3.3 8626 8.3 $36 8.3 326 3.3 03555.

A636 3.3 3.26 3.3 326 8.3 836 3.3 35:03

336 2.3 33.6 3.3 336 :3 33.6 3.3 352.3
816 2 ﬁe 2 g 2 ad 2

33-3.3 33.3.3 33-23 23-23 D33>

 

 

 

833.5» N 033 86 21380 LEE-Hum “5666:0356 Amcoummﬁa 636.3% 6:3 332

f 033-

112

mo. v m $64-:me 3535 632.304

 

 

mo. v 6 6504-665 was.» Bog-Bo:— M

mo. v m Ema-.304 mama? Boa-Bog a

8. v a .23-33 22.-2 33-33 0

8. v a .3323 222, 33-33 2

mo. v m SEA-Bog 333 632.55 <

66:6 63: 68.6 3.3 666 :3 666 3.3 239860 33

5.6 3:: 63.6 3.3 63.6 63 63.6 :3 306-623

:86 6:: :86 6.2 23.6 6.: 666 8.2 u < 2.... 6586-6263

866 3.6 3.6 3.: 63.6 8.6 666 6:: o < .2. 3656 23.-.2

63.6 8.3 63.6 63 :86 6; 366 8?. 83:8 2656

8. 6 38 82. 6 38 :8. 6 68 :3. 6 68 60838.33 268,-.

:66 38 :86 w: :36 8.3 63.6 6: < .. 268.86 625”:

366 33 6:6 :3 A86 8.: 3.6 3.: < .. €283 38m

36 3.3 83.6 66. a 23.6 3.3 3.6 8.3 8033-68 686

:36 8.3 8.66 3.6: 68.66 3.6: 68.26 3.6: 30555.

836 6.3 $26 8.6: :86 3.6: $36 8.3 6.603

:26 3.3 63.-0.6 8.6: 63.36 3.6: 6626 8.6: 3:686
8|6 2 3'6 2 S|6 2 86 2

3.3-3.3 23-3.3 33-23 23-23 285>

 

 

833.6; m 363 86 9:80 866636 8256on >6 36066326 636.8% 6:3 2on

2 2an

113

Hypothesis 4: Initial risk trajectory variables will predict aggression and
delinquent behavior during the preschool years, and initial risk trajectory
variables along with middle childhood variables will predict adolescent
aggression and adolescent delinquent behavior. Middle childhood factors will
boost the explanatory power of the model (i.e., signiﬁcantly add to the R2) and
eclipse some of the early indicators of risk — recognizing both that risk is ﬂuid
(i.e., early indicators don’t always pan out), and that proximate inﬂuences are

likely to have greater effect.

With regard to the initial risk load, antisocial behavior in preschool and
adolescence will be predicted from more difﬁcult temperament (i.e., higher
activity levels, shorter attention span, and greater reactivity) and exposure to more
parental psychopathology. Adolescent antisocial behavior will also be predicted
by child characteristics (less academic achievement, lower self-esteem, and more
social problems) and the family environment (i.e., less cohesion, less

expressiveness, more conﬂict, and more physical punishment/negative control) in

middle childhood.

Hierarchical regression results for Wave 1 variables predicting Wave 1 aggression

and delinquent behavior are presented in Table 20. Results for variables predicting Wave

4 aggression are in Table 21 (with Wave 2 middle childhood variables) and Table 22

(with Wave 3 middle childhood variables). Results for variables predicting Wave 4

delinquent behavior can be found in Table 23 (with Wave 2 middle childhood variables)

and Table 24 (with Wave 3 middle childhood variables). Given the large number of

114

independent variables in each ﬁnal model, multicollinearity diagnostics were examined.
Tolerance is a statistic that reﬂects the proportion of a variable’s variance not accounted
for by other predictors in the equation (i.e., 1 — R2 for the variable predicted by all
others). A variable with a tolerance close to 0 contributes little to the model and
represents a problem of multicollinearity. Tolerances for each variable in regression
equations using Wave 1 and Wave 2 predictors can be found in Table 25; tolerances for
each variable in regression equations using Wave 1 and Wave 3 predictors can be found
in Table 26. Tolerances ranged from .252 - .907; no multicollinearity problems were
detected.

PredictingWave 1 Aggression and Delinquent Behavior. Regression analyses
for Wave 1 behavior problems with Wave 1 predictors revealed that the model
signiﬁcantly predicted both aggression, F (6, 213) = 10.15, p < .001; and delinquent
behaviors, F(6. 213) = 7.79, p < .001. R2 was .22 for aggression and .18 for delinquent
behavior. Table 20 displays the standardized regression coefﬁcients for each variable. In
terms of individual relationships between the independent variables and behavior
problems, reactivity (for aggression t = 5.12, p < .001; for delinquent behavior; = 2.61, p
=.010) and maternal psychopathology (for aggression t = 2.39, p =.018; for delinquent
behavior t = 2.67, p = .008) were signiﬁcant predictors in both models. More reactivity
and higher levels of maternal psychopathology were associated with more behavior
problems during the preschool years. Age (1 = 4.21, p < .001) was also a signiﬁcant
predictor in the regression for delinquent behavior: As age within the wave increased

(i.e., a between-subjects distinction), delinquent behavior scores increased.

115

Table 20

Hierarchical Regression Results for Aggression and Delinquent Behavior at Wave 1

 

Standardized Regression Coefﬁcient (B)

 

 

 

Wave Variable Aggression Delinquent Behavior
1 Age .013 .267 ***
1 Activity .041 -. 1 08
1 Attention -.091 -.103
1 Reactivity .333 *** .174 *
1 Maternal Psych. .157 * .181 **
1 Paternal Psych. .120 .095
Total R2 .222 .180

 

*** p< .001, ** p< .01, * p< .05

Predicting Wave 4 Aggression. A hierarchical regression was run for Wave 4
aggression with predictors ﬁ'om preschool and Wave 2 middle childhood. Table 21
displays the standardized regression coefficients for each variable. Preschool measures
were entered as predictors in Step 1. The model signiﬁcantly explained aggression in
early adolescence, F (6, 213) = 2.96, p = .008. R2 was .08. Only maternal
psychopathology (t = 2.82, p_ = .005) at Wave I predicted aggression at Wave 4, with
more maternal psychopathology associated with more severe symptoms.

In Step 2, predictors from middle childhood (Wave 2) were entered. The model
signiﬁcantly predicted Wave 4 aggression (F [1 5, 204] = 2.96, p_ = .008), R2 was .28, and
adding these middle childhood predictors signiﬁcantly increased the amount of explained
variance (F [9, 204] = 6.48, p < .001) by 21%. Signiﬁcant predictors were maternal
psychopathology (t = 2.09, p = .038), global self-esteem (1= -3.05, p = .003), social

problems (1 = 4.13, p < .001), and family conﬂict (1= 3.12, p = .002). Adolescent

116

aggression was associated with more maternal psychopathology, social problems, and
family conﬂict but lower self-esteem.

In Step 3, temperament from Wave 2 was added to the model. The model
signiﬁcantly predicted Wave 4 aggression (F [18, 201] = 5.25, p < .001), R2 was .32, and
adding these predictors signiﬁcantly increased the amount of explained variance (F[3,
201] = 3.70, p =.013) by 4%. Signiﬁcant predictors were maternal psychopathology (t =
2.38, p = .018), global self-esteem (t = -2.15, p = .033), social problems (1 = 3.28, p =
.001), family conﬂict (t = 2.70, p = .008), and Wave 2 reactivity (t = 2.72, p = .007).
Greater reactivity at Wave 2 led to more aggressive symptomatology by adolescence.

Finally, Wave 1 (baseline) aggression was entered in Step 4. The model
signiﬁcantly predicted Wave 4 aggression (F[l9, 200] = 6.27, p < .001), R2 was .37, and
adding this predictor signiﬁcantly increased the amount of explained variance (F [1 , 200]
= 17.04, p <.001) by 5%. Wave 1 aggression was the strongest predictor of Wave 4
aggression, and controlling for it led to the effects of maternal psychopathology and
Wave 2 reactivity becoming statistically negligible (2 > .05). Signiﬁcant predictors were
global self-esteem (t = -2.49, p = .014), social problems (t = 2.63, p = .009), family
conﬂict (1 = 2.28, p = .024), and Wave 1 (baseline) aggression (t = 4.13, p < .001).

The same procedure was followed for predicting Wave 4 aggression with
predictors from preschool and Wave 3 middle childhood. Table 22 displays the
standardized regression coefﬁcients for each variable. Results for Step 1 are the same as
in the previous analysis. In Step 2, predictors from middle childhood (Wave 3) were
entered. The model signiﬁcantly predicted Wave 4 aggression (F[16, 203] = 6.24, p <

.001), R2 was .33, and adding these middle childhood predictors signiﬁcantly increased

117

the amount of explained variance (F[lO, 203] = 7.64, p < .001) by 25%. Maternal
psychopathology was no longer signiﬁcant. The signiﬁcant predictors were social
problems (t = 2.98, p = .003), family cohesion (t = -2.12, p = .035), family expressiveness
(1= 2.73, p = .007), physical discipline (1 = 2.49, p = .013), and negative control (1= 3.55,
p < .001). Adolescent aggression was associated with more social problems, family
expressiveness, physical discipline, and negative control but less family cohesion.

In Step 3, temperament from Wave 3 was added to the model. The model
signiﬁcantly predicted Wave 4 aggression (F[l9, 200] = 6.60, p = .001), R2 was .39, and
adding these predictors signiﬁcantly increased the amount of explained variance (F[3,
200] = 6.08, p =.001) by 6%. Signiﬁcant predictors were social problems (_t_ = 2.33, p =
.021), family cohesion (t = -2.44, p = .016), family expressiveness (t = 2.77, p = .006),
physical discipline (1 = 2.11, p = .036), negative control (_t_ = 3.45, p = .001), and Wave 3
general activity (1 = 3.52, p = .001). Higher general activity levels at Wave 3 predicted
more aggressive symptomatology in early adolescence.

Finally, Wave 1 (baseline) aggression was entered in Step 4. The model
signiﬁcantly predicted Wave 4 aggression (F[20, 199] = 8.17, p_ < .001), R2 was .45, and
adding this predictor signiﬁcantly increased the amount of explained variance (F[l, 199]
= 23.69, p <.001) by 7%. Wave 1 aggression was the strongest predictor of Wave 4
aggression. Social problems were no longer signiﬁcant. The signiﬁcant predictors were
family cohesion (1= -2.55, p_ = .012), family expressiveness (1= 2.83, p = .005), physical
discipline (1 = 2.28, p = .024), negative control (1 = 2.57, p = .011), Wave 3 general

activity (1= 3.33, p_ = .001, and Wave 1 (baseline) aggression (1= 4.87, p < .001).

118

Table 21

Hierarchical Regression Results for Aggression 8; Wave 4 with Preschool (Wave 1)

and Middle Childhood (Wave 2) Predictors

 

Standardized Regression Coefﬁcient ([3)

 

 

 

Wave Variable Step 1 Step 2 Step 3 Step 4

1 Age -.065 -.1 15 -.094 -.098

1 Activity .041 .042 .033 .022

1 Attention -.075 -.013 -.006 .012

1 Reactivity .090 . 1 14 .021 -.049

1 Maternal Psych. .202 ** .142 * .161 * .118

1 Paternal Psych. .014 -.063 -.082 -.095

2 Reading -.01 1 .009 .048

2 Spelling .134 .080 .077

2 Arithmetic -.148 -.107 -.112

2 Global-Self Esteem -.193 ** -.139 * -.155 "‘

2 Social Problems .260 *** .210 ** .164 **

2 Family Cohesion .026 -.006 -.014

2 Family Expressiveness .080 .080 .048

2 Family Conﬂict .235 ** .201 ** .165 *

2 Spanking .041 .001 -.031

2 Activity .064 .037

2 Attention -.004 -.01 3

2 Reactivity .220 ** .156

l Aggression .292 ***
Total R2 .077 .282 .320 .373
AR2 .205 *** .038 * .053 m

 

m p< .001, ** p< .01, * p< .05

119

Table 22

Hierarchical Regression Results for Aggression at Wave 4 with Preschool (Waveﬂ

and Middle Childhood (nge 3) Predictors

 

Standardized Regression Coefﬁcient (B)

 

 

 

Wave Variable Step 1 Step 2 Step 3 Step 4

1 Age -.065 -.084 -.100 -.100

1 Activity .041 .05 l .002 -.009

1 Attention -.075 -.003 .008 .034

1 Reactivity .090 .101 .057 -.036

1 Maternal Psych. .202 ** .073 .045 .013

1 Paternal Psych. .014 -.025 -.024 -.050

3 Reading .054 .098 .104

3 Spelling -.066 -.046 -.028

3 Arithmetic -.038 -.030 -.047

3 Global-Self Esteem -.009 -.008 -.031

3 Social Problems .194 ** .150 * .107

3 Family Cohesion -.166 * -.185 * -.184 *

3 Family Expressiveness .174 ** .172 ** .167 **

3 Family Conﬂict .116 .099 .077

3 Physical Discipline .156 * .128 * .131 *

3 Negative Control .233 *** .221 ** .159 *

3 Activity - General .262 ** .235 **

3 Activity - Sleep .071 .079

3 Task Orientation .061 .028

1 Aggression .305 ***
Total R2 .077 .330 .386 .451
AR2 .253 m .056 ** .065 ***

 

*** p< .001, ** p< .01, * p< .05

120

Predicting Wave 4 Delinquent BehJavior. A hierarchical regression was run for
Wave 4 delinquent behavior with predictors ﬁ'om preschool and Wave 2 middle
childhood. Table 23 displays the standardized regression coefﬁcients for each variable.
Preschool measures were entered as predictors in Step 1. The model signiﬁcantly
explained delinquent behavior in early adolescence, F(6. 213) = 4.24, p < .001. R2 was
.11. Only maternal psychopathology (t = 3.43, p = .001) at Wave I predicted delinquent
behavior at Wave 4, with more maternal psychopathology associated with greater
severity in adolescent’s delinquent behavior.

In Step 2, predictors from middle childhood (Wave 2) were entered. The model
signiﬁcantly predicted Wave 4 delinquent behavior (F [15, 204] = 3.81, p < .001), R2 was
.22, and adding these middle childhood predictors signiﬁcantly increased the amount of
explained variance (F[9, 204] = 3.26, p = .001) by 11%. Signiﬁcant predictors were
maternal psychopathology (t = 2.89, p = .004), social problems (t = 3.97, p < .001), and
family conﬂict (1 = 2.40, p = .017). More delinquent behavior was predicted from more
maternal psychopathology, social problems, and family conﬂict.

In Step 3, temperament from Wave 2 was added to the model. The model
signiﬁcantly predicted Wave 4 delinquent behavior (F[18, 201] = 3.44, p < .001), R2 was
.24, and adding these predictors did not signiﬁcantly increase the amount of explained
variance (F[3, 201] = 1.48, p >.10) at only 2%. Signiﬁcant predictors were maternal
psychopathology (t = 2.97, p = .003), social problems (t = 3.41, p = .001), and family
conﬂict (1 = 2.12, p = .036). The conﬁguration of signiﬁcant variables was therefore

unchanged from Step 2.

121

Finally, Wave 1 (baseline) delinquent behavior was entered in Step 4. The model
signiﬁcantly predicted Wave 4 delinquent behavior (F[l9, 200] = 5.072, p < .001), R2
was .33, and adding this predictor signiﬁcantly increased the amount of explained
variance (F[l, 200] = 26.51, p <.001) by 9%. Wave 1 delinquent behavior was the
strongest predictor of Wave 4 delinquent behavior. The effect of family conﬂict was no
longer signiﬁcant. Signiﬁcant predictors were maternal psychopathology (t = 2.26, p =
.025), social problems (1= 3.04, p = .003), and Wave 1 (baseline) delinquent behavior (t
= 5.15, p < .001).

The same procedure was followed for predicting Wave 4 delinquent behavior
with predictors from preschool and Wave 3 middle childhood. Table 24 displays the
standardized regression coefﬁcients for each variable. Results for Step 1 are the same as
in the previous analysis. In Step 2, predictors from middle childhood (Wave 2) were
entered. The model signiﬁcantly predicted Wave 4 delinquent behavior (F [16, 203] =
3.76, p < .001), R2 was .23, and adding these middle childhood predictors signiﬁcantly
increased the amount of explained variance (F[10, 203] = 3.20, p = .001) by 12%.
Signiﬁcant predictors were maternal psychopathology (t = 2.29, p = .023) and negative
control (1 = 2.14, p = .033). Delinquent behavior in early adolescence was predicted from
more maternal psychopathology and more negative control by parents.

In Step 3, temperament from Wave 3 was added to the model. The model
signiﬁcantly predicted Wave 4 delinquent behavior (F[l9, 200] = 4.06, p_ < .001), R2 was
.28, and adding these predictors signiﬁcantly increased the amount of explained variance
(F[3, 200] = 4.61, p = .004) by 5%. Signiﬁcant predictors were maternal

psychopathology (1= 2.16, p = .032), family cohesion (t = -2.18, p = .031), negative

122

control (t = 2.26, p = .025), Wave 3 general activity (1 = 2.73, p = .007), and Wave 3 task
orientation (1 = 2.07, p = .039). In addition to more maternal psychopathology and
negative control, more delinquent behavior in early adolescence was predicted by higher
activity levels, more task orientation, and less family cohesion.

Finally, Wave 1 (baseline) delinquent behavior was entered in Step 4. The model
signiﬁcantly predicted Wave 4 delinquent behavior (F[20, 199] = 5.155, p < .001), R2
was .34, and adding this predictor signiﬁcantly increased the amount of explained
variance (F [1, 199] = 19.04, p <.001) by 6%. Wave 1 delinquent behavior was the
strongest predictor. The effects of maternal psychopathology, family cohesion, negative
control, and Wave 3 task orientation were no longer signiﬁcant. The signiﬁcant
predictors were Wave 3 general activity levels (1= 2.17, p = .031) and Wave 1 (baseline)
delinquent behavior (t = 4.36, p_ < .001).

Predicting Change Scores. The previous regressions predicted raw aggression
and delinquent behavior scores using independent variables from preschool and middle
childhood. Would it make a difference if change scores were used instead? To answer
this question, Step 4 regression models were re-run with change as the dependent
variable. Change was computed by subtracting scores at Wave 1 from scores at Wave 4
(i.e., Wave 4 — Wave 1). Scores at either end of the range reﬂect discontinuity: High
change scores (>0) represent escalation and low change scores (<0) represent
improvement. Scores approximating zero reﬂect continuity (i.e., minimal change). The
same pattern of signiﬁcant predictors was found for both aggression and delinquent
behavior, using middle childhood predictors from either Wave 2 or Wave 3, as previously

reported in Step 4 of the hierarchical regressions.

123

Table 23

Hierarchical Regression Results for Delirlquent Belﬂor at Wave 4 with Preschool

(Ewe D and Middle Childhood (Wage 2) Predictors

 

Standardized Regression Coefﬁcient (B)

 

 

 

Wave Variable Step 1 Step 2 Step 3 Step 4

1 Age .089 .051 .064 -.022

1 Activity -.046 -.057 -.075 -.044

1 Attention -.121 -.095 -.094 -.073

1 Reactivity .097 .090 .039 -.004

1 Maternal Psych. .242 ** .205 ** .212 ** .154 *

1 Paternal Psych. .006 -.O41 -.050 -.069

2 Reading -.093 -.088 -.089

2 Spelling .202 .173 .166

2 Arithmetic -.082 -.058 -.041

2 Global-Self Esteem -.023 .014 .015

2 Social Problems .261 *** .231 ** .196 **

2 Family Cohesion .045 .022 .015

2 Family Expressiveness -.040 -.038 -.044

2 Family Conﬂict .189 "‘ .167 * .145

2 Spanking -.052 -.073 -.087

2 Activity .075 .071

2 Attention .000 .012

2 Reactivity .122 .087

1 Delinquent Behavior .339 ***
Total R2 .107 .219 .236 .325
AR2 .112 H .017 .089 W

 

*** p<.001, ** p< .01, * p<.05

124

Table 24

Hierarchical Regression Results for Delinquent Behavior at Wave 4 with Preschool

(Wave 1) and Middle Childhood (Wage 3) Predictors

 

Standardized Regression Coefﬁcient (B)

 

 

 

Wave Variable Step 1 Step 2 Step 3 Step 4
1 Age .089 .083 .058 -.027

1 Activity -.046 -.049 -. 102 -.066

1 Attention -.121 -.069 -.099 -.075

1 Reactivity .097 .103 .058 .018

1 Maternal Psych. .242 ** .161 * .150 * .112

l Paternal Psych. .006 -.016 -.033 -.057
3 Reading -.171 -.136 -.082
3 Spelling .152 . 173 .089

3 Arithmetic -.059 -.056 -.005
3 Global-Self Esteem -.026 -.042 -.076
3 Social Problems .107 .082 .075

3 Family Cohesion -.159 -.179 * -.109
3 Family Expressiveness .081 .081 .069

3 Family Conﬂict .000 -.006 .048

3 Physical Discipline .114 .085 .092

3 Negative Control .151 * .157 * .083

3 Activity - General .219 ** .170 *
3 Activity - Sleep .109 .097
3 Task Orientation .164 * .130

1 Delinquent Behavior .305 **"‘

Total R2 .107 .228 .278 .341
AR2 .121 ** .050 ** .063 ***

 

*** p< .001, ** p< .01, * p< .05

125

Table 25
Tolerances for Variables Predicting Wave 4 Aggression and Delinquent Behavior

with Preschool (Wave 1) aﬂ Middle Childhood (Wave 2) Predictors

 

 

 

Tolerance
Wave Variable Aggression Delinquent Behavior
1 Age .850 .794
1 Activity .777 .771
1 Attention .675 .676
1 Reactivity .609 .637
1 Maternal Psych. .725 .724
l Paternal Psych. .696 .696
2 Reading .302 .305
2 Spelling .252 .252
2 Arithmetic .528 .527
2 Global-Self Esteem .809 .812
2 Social Problems .802 .818
2 Family Cohesion .589 .589
2 Family Expressiveness .866 .881
2 Family Conﬂict .598 .606
2 Spanking .815 .826
2 Activity .773 .781
2 Attention .552 .552
2 Reactivity .497 .51 3
1 Aggression or Delinquent .624 .776

 

126

Table 26
Tolerances for Vanﬁles Predicting Wave 4 Aggressiongand Delinquent Behavior

with Preschool (Wave 1) and Middle Childhood (Wave 3) Predictors

 

 

 

Tolerance
Wave Variable Aggression Delinquent Behavior
1 Age .907 .823
1 Activity .755 .744
1 Attention .671 .672
1 Reactivity .697 .753
l Maternal Psych. .745 .742
l Paternal Psych. .726 .727
3 Reading .428 .419
3 Spelling .382 .367
3 Arithmetic .554 .543
3 Global-Self Esteem .770 .764
3 Social Problems .728 .743
3 Family Cohesion .532 .51 1
3 Family Expressiveness .794 .792
3 Family Conﬂict .609 .595
3 Physical Discipline .553 .544
3 Negative Control .693 .692
3 Activity .570 .569
3 Attention .833 .833
3 Reactivity .71 8 .704
l Aggression or Delinquent .701 .677

 

127

Controlling for Contempognegla Behaviors. It is possible that the effects of
discipline and home environment during middle childhood can be explained by
concurrent levels of antisocial behavior. For example, the greater level of family conﬂict
at Wave 2 (6-8 years) experienced by those who will later show heightened levels of
aggression symptomatology in early adolescence may be a function of heightened levels
of aggressiveness at 6-8 years of age.

To address this issue, the four ﬁnal regression models using middle childhood
variables to predict Wave 4 behavior problems were re-run with contemporaneous
behavior problem scores as an additional predictor (i.e., Wave 2 aggression or delinquent
behavior in the models with Wave 2 middle childhood predictors, and Wave 3 aggression
or delinquent behavior in the models with Wave 3 middle childhood predictors). In each
run, adding contemporaneous behavior problems signiﬁcantly increased the amount of
explained variance (2 < .001). Standardized regression coefﬁcients are in Table 27 for
models that include middle childhood predictors from Wave 2, and in Table 28 for
models that include middle childhood predictors from Wave 3.

For the regression with Wave 2 variables predicting Wave 4 aggression
controlling for Wave 2 aggression (see Table 27), the effects of social problems, family
conﬂict, and baseline aggression become statistically insigniﬁcant. Wave 2 aggression
was the strongest predictor (t = 5.16, p < .001), with age (_t_ = -2.75, p = .006) and global
self-esteem (1= -2.65, p = .009) also predicting aggression during early adolescence.

For the regression with Wave 3 variables predicting Wave 4 aggression
controlling for Wave 3 aggression (see Table 28), physical discipline and negative

control become insigniﬁcant - which suggests that contemporaneous behavior problems

128

may explain them. Wave 3 aggression was a signiﬁcant predictor (t = 5.97, p < .001) but
did not eclipse the effects of baseline aggression (t = 3.41, p = .001), Wave 3 general
activity level (1 = 2.34, p = .021), or two family environment variables: cohesion

(t = -2.25, p = .026) and expressiveness (t = 2.38, p = .018). Age also was marginally
signiﬁcant (1 = -.98, p = .049). Although punishments may overlap with
contemporaneous behavior problems, family cohesion and expressiveness at 9-11 years
of age make signiﬁcant contributions to understanding later aggression.

None of the contextual variables related to family environment or discipline were
still signiﬁcant by Step 4 of the hierarchical regressions predicting delinquent behavior in
early adolescence. However, the results do demonstrate the importance of more
temporally proximate behavior problems for predicting later outcome.

For the regression with Wave 2 variables predicting Wave 4 delinquent behavior
controlling for Wave 2 delinquent behavior (see Table 27), maternal psychopathology
and social problems were no longer signiﬁcant. In fact, the only two signiﬁcant
predictors were Wave 1 delinquent behavior (1 = 2.47, p = .015) and Wave 2 delinquent
behavior (3 = 7.08, p < .001).

For the regression with Wave 3 variables predicting Wave 4 delinquent behavior
controlling for Wave 3 delinquent behavior (see Table 28), Wave 3 activity level dropped
from signiﬁcance. Only baseline (t = 3.10, p = .002) and Wave 3 (1= 5.79, p < .001)

delinquent behavior were signiﬁcant predictors.

129

Table 27

Reggession Results forAggression and Delinquent Behavior at Wave 4 with Preschool

 

(Wave 1) and Middle Childhood (Wave 2] Predictors Including Contemmraneous

(LN ave 2) Behavior Problems

 

Standardized Regression Coefﬁcient (B)

 

 

 

Wave Variable Aggression Delinquent Behavior

1 Age -.161 ** -.052

1 Activity .044 -.073

1 Attention .001 -.062

1 Reactivity -.016 .029

1 Maternal Psych. .089 .121

1 Paternal Psych. -.088 -.087

2 Reading .025 -.062

2 Spelling .034 .131

2 Arithmetic -.129 -.045

2 Global-Self Esteem -.155 ** -.005

2 Social Problems .031 .035

2 Family Cohesion .007 .067

2 Family Expressiveness .064 -.052

2 Family Conﬂict .093 .107

2 Spanking -.054 -.056

2 Activity .034 .089

2 Attention .062 .05 1

2 Reactivity .076 .035

1 Aggression or Delinquent .126 .159 *

2 Aggression or Delinquent .443 *** .499 ***
Total R2 .447 .461
AR2 (from previous Step 4 model) .074 *** .136 ***

 

m p< .001, ** p<.01, *p< .05

130

Table 28
Ragression Results for Aggression and Delinquent Behavior at Wave 4 with Preschool
(“ﬁve 1) and Middle Childhood (Wave 3) Predictors Includig Contemporaneous

(Wave 3) Behavior Problems

 

Standardized Regression Coefficient (B)

 

 

 

Wave Variable Aggression Delinquent Behavior

1 Age -.101 * .004

1 Activity -.010 -.O67

1 Attention .023 -.088

1 Reactivity -.025 .045

1 Maternal Psych. .000 .098

1 Paternal Psych. -.037 -.057

3 Reading -.014 -.097

3 Spelling .039 .064

3 Arithmetic -.040 .040

3 Global-Self Esteem -.046 -.093

3 Social Problems -.033 -.004

3 Family Cohesion -.150 * -.091

3 Family Expressiveness .130 "' .053

3 Family Conﬂict -.006 -.013

3 Physical Discipline .081 .059

3 Negative Control .111 .063

3 Activity - General .155 * .119

3 Activity - Sleep .005 .047

3 Task Orientation .013 .132

1 Aggression or Delinquent .205 ** .207 **

3 Aggression or Delinquent .433 *** .379 ***
Total R2 .535 .437
AR2 (from previous Step 4 model) .084 *** .095 ***

 

*** p < .001, ** p < .01, * p < .05

131

Changing the Order of Variable Engy. Finally, the steps in the hierarchical
regression were reordered to answer one more question: Do middle childhood predictors
contribute signiﬁcantly to the model aﬁer we take into consideration what we know
about the development of the problem behaviors themselves? In other words, is there a
signiﬁcant R2 change when middle childhood predictors are added after Wave 1 and
middle childhood aggression or delinquent behavior scores?

To address this issue, variables were entered in the following order: Wave 1
aggression or delinquent behavior (Step 1), Wave 1 predictors (Step 2), middle childhood
aggression or delinquent behavior (Step 3), middle childhood predictors (Step 4), and
then middle childhood temperament control variables (Step 5).

When aggression was predicted from preschool and Wave 2 variables, adding the
Wave 1 predictors after the aggression scores from Wave 1 resulted in a non-signiﬁcant
R2 change for Step 2 (F[6, 212] = 1.04, p = .401). However, adding the Wave 2 middle
childhood predictors after the Wave 2 middle childhood aggression scores still led to a
signiﬁcant increase in explained variance at Step 4 (F[9, 202] = 2.15, p = .027). The
same pattern was found in the model predicting aggression from preschool and Wave 3
variables. Adding the Wave 3 middle childhood predictors after the Wave 3 middle
childhood aggression scores signiﬁcantly increased the R2 with Step 4 (F[10, 201] = 1.95,
p = .040). For aggression, then, middle childhood variables take precedence over
preschool contextual and temperamental factors — and add to our understanding of how
aggression develops beyond what we would understand from only looking at aggression

itself.

132

When delinquent behavior was predicted from preschool and Wave 2 variables,
adding the Wave 1 predictors after the delinquent behavior scores from Wave 1 resulted
in a non-signiﬁcant R2 change for Step 2 (F[6, 212] = 1.64, p = .137). Similarly, adding
the Wave 2 middle childhood predictors after the Wave 2 middle childhood delinquent
behavior scores revealed a non-signiﬁcant increase in explained variance for Step 4 (F[9,
202] = .79, p_ = .628). The same pattern was found in the model predicting delinquent
behavior from preschool and Wave 3 variables. Adding the Wave 3 middle childhood
predictors after the Wave 3 middle childhood delinquent behavior scores yielded a non—
signiﬁcant change in R2 for Step 4 (F[10, 201] = .96, p = .479). As in the
contemporaneous models, only earlier delinquent behavior predicted delinquent behavior
among early adolescents.

Summm. Hypothesis 3 was generally supported for both aggression and
delinquent behavior. The high-high and low-low groups were distinguishable from one
another on a number of different characteristics. Change groups were generally
intermediate in their risk proﬁles. However, these groups did not completely occupy the
“middle ground.” In some cases these groups were more similar to one continuity group
than the other — or conversely, more different from one continuity group than the other.

It is worth noting that classiﬁcation groups were as expected on the variable that
deﬁned them from Wave 2 through Wave 4 -— with equal “highs” (i.e., for high-high and
low-high) and equal “lows” (i.e., for low-low and high-low) by 9-11 years of age, and
with some intermediacy in scores for the change groups during middle childhood.
Interestingly, the results demonstrate that the cross-over from low to high or high to low

has already occurred by Wave 3; Wave 2 seems to be the important turn-point.

133

There were two unexpected ﬁndings at Wave 1. For aggression, those with
initially “low” levels of symptomatology were different from one another; for delinquent
behavior, those with initially “high” levels of symptomatology were different from one
another. Some caution is therefore warranted when describing the constellation of risks
associated with the escalating pattern for aggression (i.e., low-high versus low-low) and
the remitting pattern for delinquent behavior (i.e., high-low versus high-high) using
univariate tests like AN OVA; it may be that baseline differences in aggression and
delinquent behavior can account for some later disparities.

Hypothesis 4 was also partially supported. Of the initial trajectory risk factors,
only concurrent levels of maternal psychopathology and reactivity predicted preschool
levels of aggression and delinquent behavior. Adding in middle childhood
risk/promotive factors helped to explain aggression and delinquent behavior during
adolescence. In fact, middle childhood risk factors were much more useful than more
distal preschool risk factors for predicting adolescents’ antisocial behavior; maternal
psychopathology at Wave 1 was the only contributing factor from the baseline trajectory.
Many signiﬁcant middle childhood predictors remained signiﬁcant even after controlling
for baseline levels of antisocial behavior (but much more so for aggression).

Contrary to expectations, school achievement was relatively unimportant.

When baseline levels of aggression and delinquent behavior were included in the
model, the same predictors were identiﬁed as signiﬁcant regardless of whether raw scores
or change scores were used as dependent variables. Controlling for contemporaneous
behavior problems in middle childhood mitigated the inﬂuence of some variables (e. g.,

discipline at 9-11) but not others (e.g., family cohesion and expressiveness at 9-11) for

134

adolescent aggression. At this stage, only previous delinquent behaviors predicted later

delinquent behaviors.

135

CHAPTER 5
DISCUSSION

The present study examined early age of ﬁrst drink (AF D) from a perspective that
emphasized early alcohol use as a marker in a more longitudinal pattern of problem
behaviors — speciﬁcally, aggression and delinquent behavior from preschool onward.
This perspective, looking at early AF D as the outcome of an ongoing developmental
trajectory characterized by antisociality, stands in contrast to studies focusing on early
AF D as a precursor to other outcomes (e. g., alcoholism) as well as a literature on
drinking initiation that emphasizes concurrent or more proximate inﬂuences in the
personal and social surround of adolescents (e. g., research on peer inﬂuences or
concomitant levels of family conﬂict).

Consistent with a life-course or developmental perspective on problem behavior,
the present study explored the utility of aggression and delinquent behavior trajectories
for explaining early AFD. Hypothesizing that hard-continuity in aggression and
delinquent behaviors is associated with ﬁrst drink onset by 12 to 14 years of age, early
drinkers were expected to demonstrate higher levels of problem behavior
symptomatology from preschool to early adolescence — with levels of symptomatology in
preschool (3-5) and early adolescence (12-14) serving as particularly strong markers of
the developmental trajectory and early onset risk. Although continuity was assumed to
be the more common trend, change patterns were also recognized. It was hypothesized
that those with either remitting or escalating behavior problems would be at intermediate
risk for drinking onset and would experience intermediate levels of risk across a broad

domain of preschool and middle childhood risk (or, conversely, promotive) factors.

136

Finally, the role of these middle childhood variables — in combination with the initial risk
trajectory — was explored in terms of predicting severity of antisocial behaviors during
preschool and early adolescence.

The overarching premise is that if early alcohol use is associated with longitudinal
patterns of aggression and delinquent behavior, we might be able to prevent precocious
drinking by disrupting this behavioral trajectory. Efforts to do so should be informed by
a better understanding of the etiological factors relevant to antisocial behavior,
particularly from preschool and middle childhood (i.e., early in the developmental
process — before behaviors become too ingrained, before cumulative disadvantages from
school and social relationships snowball into an epigenetic rut, and before regular
drinking becomes part of the behavioral repertoire). Findings from the current study are
summarized and discussed here in terms of 1) onset of ﬁrst drink, 2) the development of
antisocial behavior, 3) limitations and future directions, then 4) concluding remarks.
Onset of First Drink

The results from testing Hypothesis 1 and Hypothesis 2 support the decision to
look separately at aggression and delinquent behavior. The idea that early drinkers can
be distinguished from those who have not yet tried alcohol by early adolescence using a
hard-continuity model of problem behavior was supported for delinquent behavior only.

In general, levels of aggression declined for both groups (early drinkers and non-
drinkers alike) from 3-5 to 12-14 years of age. This ﬁts with other literature describing a
decline in aggression from preschool through adolescence (Bongers, Koot, van der Ende,
& Verhulst, 2003; Stanger, Achenbach, & Verhulst, 1997; Tremblay, 2002). The only

group difference in the current study was for aggression at 12-14, with early drinkers

137

rated by their mothers as more aggressive. In general, however, early drinkers and those
who have not tried alcohol by 12-14 years of age showed more similarity than difference
in the developmental trajectory of aggression. Aggression levels during preschool and
early adolescence (i.e., low-low, low-high, high-low, high-high) also were not
signiﬁcantly related to onset of ﬁrst drink.

On the other hand, early drinkers were more delinquent than their never-drinking
peers at all ages except 9-11 -- when the level of symptomatology for early drinkers
appeared to diminish and converge with that of those who have never tried alcohol,
before rising again in early adolescence. This, too, is consistent with other research. For
example, Bongers et al. (2003; see also Stanger et al., 1997) found that scores on the
delinquent behavior scale of the CBCL were curvilinear from 4 to 18 years of age,
initially declining but then increasing in early adolescence. Interestingly, the current
study revealed a developmental increase in delinquent behavior for early drinkers but not
for those who hadn’t yet tried alcohol by 12-14 years of age (revisit Figure 6).

Bongers et a1. (2003; see also Stanger et al., 2003) tied the “normative”
adolescent-increase they observed to the “normative” delinquency of adolescents
described by theorists like Terrie Moffltt. This increase is also reﬂective of the
ubiquitous “age-crime curve”: At the aggregate level, there is a steep incline in antisocial
behavior between the ages of 7 and 17, followed by a steep decline from 17 to 30
(tapering off thereafter) (Mofﬁtt, 1993). For Mofﬁtt (1993), however, the age-crime
curve is a consequence of those who engage in delinquency only during adolescence (i.e.,
the adolescent-limited type) temporarily accompanying those with persistent delinquent

behavior problems (i.e., the life-course-persistent type) in antisocial involvement. As

138

Mofﬁtt (1993, p. 675) has stated, “Actual rates of illegal behavior soar so high during
adolescence that participation in delinquency appears to be a normal part of teen life.”

It was therefore surprising that, in the current study, there was no coinciding
increase seen for the non-drinker group. In retrospect, however, this study only begins to
look at problem behavior during adolescence, ending as it does with reports from 12-14
years of age. Those who have either persisted in their delinquent behaviors from
preschool or who escalate in delinquent activity very early in adolescence (i.e., by 12-14
years of age) are probably quite different from those who initiate delinquent behavior
during middle-to-late adolescence (i.e., the “normative” years). As a group, the early
drinkers were therefore precocious in their ﬁrst use of alcohol as well as in their
delinquent involvements going into adolescence.

An alternate explanation focuses on the dip in delinquent behavior for the early
drinker group at ages 9-11 rather than the subsequent increase at ages 12-14. It is
possible that the predisposition to engage in markedly more delinquent behaviors
remained constant but that control structures during middle childhood (e. g., socialization
by the school) temporarily reduced the enactment of these tendencies — or, alternately,
that mothers were less aware of these behaviors; if more autonomy is granted to pre-
adolescents than young children, the observed dip may be an artifact of less supervision
and parental oversight. In either case, the idea that delinquent behavior is “normative”
during adolescence may have led to its return during early adolescence despite control
structures or even because of them (i.e., in a “kniﬁng-off childhood apron strings” gesture
of rebelliousness; Mofﬁtt, 1993, p. 688).

Beyond simple statistics, complex analyses comparing early drinkers to those who

139

had not yet tried alcohol by 12-14 years of age also supported the idea of a delinquent
behavior substrate within’the early drinker group. Two ﬁndings speak to this conclusion.

First, there were omnibus differences in the developmental trajectories of
delinquent behavior for early drinkers and non-drinkers (revealed in a stacked auto-
regressive structural equation model). One of these differences was in the direct effect of
preschool delinquent behavior on delinquent behavior during early adolescence.
Speciﬁcally, the path from Wave 1 to Wave 4 delinquent behavior was signiﬁcant for
early drinkers only. The direct path represents the inﬂuence of early predispositions on
adolescent outcomes and supports the idea of hard-continuity for this subset of
individuals. One implication is that a high level of delinquent behavior as early as age
three should serve as a “red ﬂag” to parents. Children with this type of symptomatology
are prime candidates for early intervention programs that seek to disrupt the behavioral
trajectory at a young age (perhaps preventing early onset of alcohol use and related
sequelae in the process).

Second, conﬁgural frequency analysis revealed that those with high levels of
delinquent symptomatology in both preschool and adolescence (i.e., the high-high
delinquent behavior group) were more likely than chance to have had a ﬁrst drink,
whereas the opposite pattern was found for those with low levels of delinquent
symptomatology throughout (i.e., the low-low delinquent behavior group). In other
words, there was a relationship between continuity of delinquent behaviors and likelihood
of early AFD consistent with the hard-continuity model of behavior problems.

The ﬁndings presented thus far raise an important question: Why this relationship

with drinking onset for delinquent behavior but not aggression? Three explanations are

140

offered: two conceptual, one methodological.

Conceptually, there is a difference between aggression and delinquent behavior
that can account for both the distinct developmental patterns found in the current study
and by other researchers (e.g., Bongers et al., 2003) as well as the relationship with ﬁrst
alcohol use only for delinquent behavior. This difference is one of kind: Delinquent
behavior may be more “deviant” than aggression.

Aggression among preschool children is normative to some extent, as individuals
learn to internalize the norms and values of conventional society and begin to master the
skills of emotion regulation and behavior regulation. The ability to use language to
negotiate through interpersonal relationships is acquired in a developmental process
which also may parallel changes in aggression over time; indeed, aggression is expected
to decline during childhood as communication skills become more proﬁcient. CBCL
items used to measure aggression include boasting, arguing, screaming, showing off,
ﬁghting, and general disobedience.

Delinquent behavior, as assessed by the CBCL, is more intrinsically connected to
criminality (e. g., stealing, setting ﬁres, and vandalism). Consider the difference between
the following two individuals: One pushes people around, demands attention, destroys
things in a temper tantrum, and lacks the social skill of subtlety; the other commits theft,
sets ﬁres, lies, swears, and runs away from home. Although both individuals should be
of concern to parents, schools, researchers, and prevention workers, the second individual
engages in less common (i.e., more “deviant”) activities: Many children bite, push, and

have temper tantrums; fewer steal or set ﬁres.12 The development of delinquent behavior

 

‘2 This can also be demonstrated empirically. The sample mean for aggression was 10.65 at Wave 1

141

also better parallels the age-crime curve (see also Bongers et al., 2003; Stanger et al.,
1997).

An alternate conceptual distinction concerns developmental speciﬁcity. In other
words, the difference between aggression and delinquent behavior may be outcome-
dependent. Of key importance here is the distinction between the overt and covert
antisocial pathways described by Loeber and colleagues (Loeber & Hay, 1997; Loeber &
Stouthamer-Loeber, 1998).

According to Loeber (Loeber & Hay, 1997; Loeber & Stouthamer-Loeber, 1998),
antisocial behavior can be best understood with a triple-pathway model. Each path has a
unique developmental course leading to particular outcomes. The overt pathway begins
with minor aggression (e.g., bullying, annoying others) and escalates through physical
ﬁghting to violent crime (e. g., rape, assault). The covert pathway begins with minor
covert behaviors (e.g., shoplifting, lying) and escalates though property damage (e. g.,
vandalism, ﬁre-setting) to more serious forms of delinquency (e.g., fraud, burglary,
serious theﬁ). Finally, the authority conﬂict pathway begins with stubborn behavior and
escalates through deﬁant disobedience to problems associated with authority avoidance
(e.g., truancy, staying out past curfew). Overt behaviors tend to decrease with age,

whereas covert behaviors increase with age (Loeber & Stouthamer-Loeber, 1998).

 

and 7.06 at Wave 4; prorating these by a possible high score of 40 yields .226 for Wave 1 and .176 for
Wave 4. The sample mean for delinquent behavior was 1.88 at Wave 1 and 1.85 at Wave 4; prorating these
by a possible high score of 24 yields .078 for Wave 1 and .077 for Wave 4. Therefore, aggressive behavior
problems were almost three times more common than delinquent behavior problems at Wave 1 (i.e., .226

versus .078), and still more than twice as common at Wave 4 (i.e., .176 versus .077).

142

Overt behaviors appear to characterize the aggression subscale of the CBCL,
whereas covert behaviors seem to characterize the delinquent behavior subscale of the
CBCL — in both content and development trends. Of relevance to the current study is the
relationship of these behaviors to early ﬁrst alcohol use. Alcohol use is not part of
Loeber’s conceptual scheme, but is non-violent and likely to be associated with the covert
(rather than overt) pathway. If the outcome measure for the current study was date rape,
the result may have been a stronger association with aggression rather than delinquent
behavior. Rather than delinquent behavior being more “deviant,” perhaps it is just more
consistent with a developmental path that also includes substance use.

There is also a methodological explanation for the stronger relationship with ﬁrst
alcohol use found for delinquent behavior compared to aggression. The CBCL
delinquent behavior syndrome scale was initially constructed to include an item on
substance use. This item was dropped from the current set of analyses to avoid the
obvious confound with ﬁrst drink onset status. Nonetheless, CBCL scales were derived
from a factor analysis. The CBCL delinquent behavior scale used in the present study,
although not inclusive of substance use, therefore may be predisposed towards an
association with substance use — with this scale and ﬁrst drink onset status “hanging
together” by virtue of the way the delinquent behavior subscale was constructed.

Caution is certainly warranted in the interpretation of these ﬁndings, and future research
is needed to address the question of whether delinquent behavior is in fact more strongly
tied to ﬁrst alcohol use than aggression when separate measures of delinquent behavior
are used. However, the next section of this discussion returns to this issue of difference —

offering some support for the idea that delinquent behavior is, indeed, more consistent

143

with the hard-continuity paradigm than aggression.
The Development of Antisocial Behavior

Given the hard-continuity paradigm and problem behavior theory perspective
adopted to explain early ﬁrst alcohol use, the next step in the current study was to
examine the etiology of antisocial behavior itself. First, the four classiﬁcation groups for
aggression and delinquent behavior (i.e., low-low, low-high, high-low, high-high) were
compared on general risk proﬁles from preschool and middle childhood. These risk (or
promotive) factors were then used to predict severity of aggression and delinquent
behavior in preschool (Wave 1) and early adolescence (Wave 4).

“High” and “low” antisocial behavior scores were expected to mean the same
across classiﬁcations (e. g., with the low-low and low-high groups indistinguishable
during preschool, but the low-high and high-high groups indistinguishable during early
adolescence). Two exceptions were found at Wave 1. The low-high group for
aggression was already more aggressive than the low-low group - meaning that this
change pattern might be driven by more risky predispositions in addition to the presence
or absence of other risk/promotive factors during middle childhood. Similarly, the high-
low group for delinquent behavior was already less delinquent than the high-high group —
meaning that this change pattern might be driven by less risky predispositions in
conjunction with other developmental “turning points”. Some caution in interpreting
differences between these group pairs is therefore warranted as initial problem behavior
severity was not truly equivalent. At subsequent waves, classiﬁcations ﬁt with expected
patterns. By Wave 4, both high-high and low-high groups were indistinguishable, as

were both low-low and high-low groups.

144

Fergusson et al. (1996) found that those with no conduct problems had the most
favorable risk proﬁles on family and child characteristics, those with persistent conduct
problems had the least favorable proﬁles, and children with remitting or late onset
patterns showed proﬁles situated between these two extremes. These ﬁndings were
generally replicated in the current study.

Indeed, risk proﬁles in the low-low group were less serious than in the high-high
group for both aggression and delinquent behavior. For both aggression and delinquent
behavior, the low-low group presented with a less risky proﬁle with respect to reactivity
and maternal psychopathology at Wave 1; social problems, family conﬂict, attention, and
reactivity at Wave 2; as well as social problems, negative control, and activity-general at
Wave 3.

That the low-low versus high-high comparisons were so similar across antisocial
behavior subtypes speaks to the importance of these particular risk/promotive factors in
the development of behavior problems from preschool to adolescence. There were
several differences, however. For aggression, the low-low group also had less paternal
psychopathology at Wave 1 and less family conﬂict at Wave 3 than the high-high group.
For delinquent behavior, the low-low group had more family cohesion at Wave 3 than the
high-high group.

Change groups were generally intermediate in their risk proﬁles. However, these
groups did not completely occupy the “middle ground.” In some cases these groups were
more similar to one continuity group than the other. Conversely, the continuity group
they differed from in these instances reveals something about the change pattern itself

(e. g., which factors are related to the diverging trajectories of those with initially high or

145

initially low levels of symptomatology - and therefore, to remitting or escalating
antisocial behaviors).

The remitting pattern (i.e., high-low versus high-high) was associated with less
reactivity at 6-8 years and fewer social problems at both middle childhood waves for
aggression; less family conﬂict and longer attention spans at 6-8 years and less negative
control at 9-11 years for delinquent behavior; as well as less general activity at 9-11 years
for both aggression and delinquent behavior. At 9-11 years of age, both high-low groups
seemed most like the low-low groups (i.e., and less like the high-high groups) on
incidence of physical discipline.

Recall that the high-low and high-high delinquent behavior groups could be
differentiated from one another at 3-5 years of age. The already-higher level of
delinquent involvement among the high-high group suggests that one contributing factor
to persistence rather than improvement is one’s initial predisposition to some extent
(although the high-low group was more delinquent than those with low levels of
symptomatology during preschool — so signiﬁcant improvement did take place over the
developmental course nonetheless).

The ﬁnding of less difﬁcult temperament for remitting groups during middle
childhood compared to those with persistent behavioral difﬁculties suggests that
temperament and behavior problems may be reﬂections of the same underlying construct
or disposition. Both can change — and do, in the same direction. For example, although
the high-low aggression group was more reactive than those with initially low levels of
aggressive symptomatology during preschool, this difference had disappeared by 6-8

years of age. These results are consistent with some of the most popular perspectives on

146

temperament (e. g., Chess & Thomas, 1989; Worobey, 1999) and lend support to dynamic
views of the behavior-temperament nexus. Temperament isn’t immutable; its initial risk
trajectory is subject to alteration. However, it cannot be determined here whether change
in temperament precedes, follows, or is concordant with change in behavior problems.
Those who start of as highly antisocial but are not characterized as such by
adolescence also seem to experience more favorable family environments during middle
childhood (e. g., with regard to negative control, family conﬂict, and/or physical
discipline). This can be interpreted in several ways. Conﬂict, negative control, and
physical discipline may exacerbate the initial predispositions of aggressive or delinquent
children to the extent that an adverse home context and family dynamics (e. g., inadequate
parenting) become part of the maintenance structure for their behavior problems (e.g.,
Mofﬁtt, 1993; Patterson et al., 1989). Conversely, the absence of this familial risk may
enable some individuals to “phase out” of their early behavioral tendencies and allow
them to take advantage of “turning point” moments in their development. The alternate
explanation is also possible: ‘The remitting pattern emerges ﬁrst — leading to less
conﬂict, control, and physical discipline because the child behaves better, doesn’t
aggravate his parents as much, and isn’t involved in as many activities that require
disciplinary action compared to children with persistent antisocial behaviors (i.e., doesn’t
“evoke” the same style of parenting or parental aggravation as persisters). The fact that
the remitting pattern was associated with longer attention spans (at 6-8) for delinquent
behavior and less general activity (at 9-11) for both aggression and delinquent behavior
provides some support for the “less evocative” stance — because remitters do present with

less risky temperaments during middle childhood. However, it is impossible to

147

completely disentangle causal order in the current research design.

There is a similar quagmire with respect to the causal ordering of social problems
and aggression. Those with remitting aggressive behaviors had fewer social problems
than those with persistent aggression. This is interesting given that the two groups started
off in preschool with statistically indistinguishable levels of aggressive symptomatology.
These initial levels therefore cannot account for the divergence in the quality of peer
relationships; at least for aggression, preschool behaviors do not automatically “knife off”
opportunities for social engagement. There are two explanations for the divergence.
Either conventional peer relationships “protect” those in the remitting group and allow
for a “turning point” in behavioral trajectories (conversely, that poor social relationships
maintain the persistent behavior problem trajectory), or aggression begins to remit ﬁrst -—
leading to less peer rejection and better social relationships. The likelihood is that both
explanations apply: Behavior problems may evoke peer rejection; peer rej ection may
lead to rebellious or frustration-related problem behavior. Rejected children may also
join in with antisocial others in response to their more conventional social failures (e.g.,
see Patterson etal., 1989). Those with remitting behavior patterns for aggression break
the cycle at some point.

The remitting pattern is important to consider because it represents relative
desistence from aggression or delinquent behaviors. Prevention and intervention
programs can capitalize on a better understanding of the natural processes that inhibit or
bring about this change. Findings from the current study suggest that programs focused
on family dynamics and social problems may be beneﬁcial in either disrupting aggressive

and delinquent behavioral trajectories or disrupting their consequences socially and in the

148

home — thereby helping to break self-perpetuating maintenance structures of maladaption.

Prevention efforts, and future research endeavors, might also beneﬁt by looking
beyond these familial and social risk factors to those ecological variables that, in turn,
shape these interpersonal relationships (as well as individual behaviors) (Bronfenbrenner,
1977; Evans, 2003; Fraser, 1996; Jack, 2000; Tolan, Gorman-Smith, & Henry, 2003).

For example, neighborhood violence and poverty are two ecological variables that have
received much attention in this literature. An even broader appreciation of the
contextual, risk accumulation model can be adopted than studied here - with risky child
behaviors (e. g., aggression, delinquency) nested in risky immediate environments (e.g.,
familial, peer) which in turn are nested in risky ecological settings (e. g., community,
socioeconomic). Just as it is important to consider familial and social factors as
ecological inﬂuences with much explanatory power, there are wider ecological or
structural domains that impinge on these. To address family dynamics and social
problems, intervention/prevention efforts must keep in mind the role played by inﬂuences
at all levels of human ecology.

In addition to the remitting patterns, escalating patterns are also important because
they can provide some insight into what aggravates antisocial behavior. Looking beyond
the risk factors that maintain the developmental trajectory of children with behavioral
difﬁculties, which escalate the development of behavioral difﬁculties among others?

For aggression, the escalating pattern (i.e., low-high versus low-low) was
associated with more maternal psychopathology at 3-5 years, lower self-esteem at 6-8
years, as well as more family conﬂict, more general activity, and less task orientation at

9-11 years. For both aggression and delinquent behavior, the low-high group seemed

149

most like the high-high group on incidence of physical discipline. Interestingly, the low-
high delinquent group was similar to the low-low delinquent group in having fewer social
problems than the high-high group at Wave 2.

Again, it should be remembered that the low-high and low-low aggression groups
were already different from one another in preschool. For aggression, then, one
determinant of escalation may be this riskier predisposition compared to those who
remained relatively unproblematic.

Of interest is the fact that the risk/promotive factors included in the current study
did little to illuminate the processes associated with escalating problems for delinquent
behavior. Aside from the observation that those in the escalating delinquent behavior
group experienced more physical discipline at 9-11 than the low-low delinquent behavior
group (i.e., 23% versus 5%), there were no signiﬁcant differences between these two
classiﬁcations. The only other relevant ﬁnding was that the low-high group had fewer
social problems than the high-high group at 6-8 (and was similar to the low-low group in
this regard). For delinquent behavior, then, early social problems seem more dependent
on initial levels of symptomatology. The low-high group did become intermediate at 9-11
years of age, however. That the cross-over from “low” to “high” had already occurred by
Wave 2 for the group as a whole suggests that there is a lag between the escalation of
delinquent behaviors and escalation of social problems. This suggests causal order:

Peer rejection seems to be a consequence of, rather than an antecedent to, delinquent
symptomatology for those with emergent delinquent behavior problems.

Many more characteristics distinguished those with escalating aggression from

the low-low group. As previously mentioned, temperament often changed in sync with

150

behavior problems; those with escalating patterns of aggression were more active and less
task oriented by ages 9-11. There was more family conﬂict and physical discipline at 9-
11 years of age for the escalating aggression group compared to the low-low aggression
group (i.e., 26% versus 5%). The same issue of ambiguity in temporal ordering discussed
with respect to the remitting pattern applies here: Conﬂict and discipline may be a
consequence of, rather than a contributing factor to, the child’s aggressive behaviors. In
fact, that the family environment was not more adverse until Wave 3 supports the
“evocative” explanation; it was not until after the (aggregate) Wave 2 cross-over from
“low” to “high” that signiﬁcant differences in family environment or discipline style were
observed.

Those with escalating aggression also had lower self-esteem than the
unproblematic low-low group at 6-8, but not 9-1 1, years of age. Therefore, lower self-
esteem appears to temporarily accompany the transition from “low” to “high.” It’s
unclear from the current research design whether lower self-esteem is a consequence or
cause of escalating aggression. Nonetheless, this is an intriguing ﬁnding that deserves
future attention. If low self-esteem leads to aggression, programs geared toward fostering
positive feelings about one’s self might prevent this escalation. If increased aggression
leads to lower self-esteem (mediated, perhaps, by changes in the family environment or
parents’ reactions), this could fuel a child’s ﬁ'ustration and perpetuate his problems.

There are several interpretations for the ﬁnding that maternal psychopathology
was lower for the low-low aggression group than the low-high aggression group — and all
others. Maternal psychopathology was also lower for the low-low delinquent behavior

group than the high-high delinquent behavior group. It may be that the low-low groups

151

are at the least risk in terms of less exposure to alcoholic, antisocial, and/or depressed
mothers - and that this relative absence of maternal risk is what maintains the low-
aggression or low-delinquent behavioral trajectory. However, it may also be the case that
the other groups are rated high on aggression or delinquent behavior at some point
because psychopathology has affected mothers’ perceptions of their children’s behavior
problems (rather than because of the behavior problems per se). In other words, mothers
with more psychopathology may be more critical or less tolerant of their children’s
behavior than mothers with less psychopathology.

Finally, beyond speciﬁc issues related to remitting or escalating behavior
problems, the current study revealed an interesting developmental pattern in the
relationship between temperament and antisocial behavior. Among young children,
there seems to be a connection between reactivity and antisocial behavior. Those high on
aggression during preschool were more reactive than those low on aggression during
preschool. Those with persistent delinquent behaviors were more reactive than those low
in delinquent behaviors during preschool (although those with remitting delinquent
behaviors were not). There were no group differences on other temperament dimensions
at 3-5 years of age.

Other dimensions of temperament came into play during middle childhood.
Reactivity remained an important risk marker; those with persistent aggression problems
were more reactive than all others by Wave 2, and those with persistent delinquent
behavior problems were more reactive than the low-low group at Wave 2. Now attention
had emerged as another distinguishing characteristic. Those with persistent aggression

problems had shorter attention than those in the low-low aggression group; those with

152

persistent delinquent behavior problems had shorter attention than either group with low
levels of adolescent symptomatology. By Wave 3 (9-11), the major distinguishing
dimension of temperament was general activity. Those with persistent behavior problems
were more active than either group with low levels of adolescent symptomatology for
both aggression and delinquent behavior. In addition, those with escalating aggression
problems were more active and less task oriented than the low:low group.

To interpret these ﬁndings, remember that there was a change in instrumentation
between the second and third wave of data collection (from the DOTS to the DOTS-R).
The general activity subscale of the DOTS-R overlaps somewhat with the reactivity scale
of the DOTS. That said, persistent behavior problems were associated with more
reactivity over the developmental course as well as attenuated attention and higher
activity levels during middle childhood. Like communication skills and self-regulation,
the ability to focus one’s attention is learned with time. For all groups, scores for
attention increased from Wave 1 to Wave 2. Preschoolers are expected to have attention-
related deﬁcits because of their age. Differences in attention really become relevant after
school entry. By middle childhood, children should already be mastering attention skills.
It makes sense that those who have trouble with this developmental task are also those
with other behavioral difﬁculties, and that differences in attention become evident only
during this time period. That general activity became the risk marker at ages 9-11 can be
explained by its overlap with reactivity, its absence in earlier assessments (i.e., activity at
Wave 1 and Wave 2 was sleep-oriented), and the reﬁnement of self-regulatory skills over
time. As with attention deﬁcits, high activity levels are normative for young children but

become more restrained over the course of development. Again, it makes sense that

153

over—active children —- those who haven’t mastered self-regulation quite as well as other
children — would have other behavior problems.

In the ﬁnal set of analyses, the risk/promotive factors already discussed were used
to explain the initial risk trajectory and adolescent outcomes. These analyses address
some of the previous concerns about causal order in that regression procedures allowed
for an examination of the effect of each variable controlling for all others.

Reactivity and maternal psychopathology were the only signiﬁcant predictors of
initial levels of problem behavior severity. This reafﬁrms the importance of reactivity as
an early risk marker. More maternal psychopathology was associated with more
aggression and delinquent behavior during preschool, even controlling for maternal
ratings of child temperament — a possible substrate of behavior. Although this does not
negate the possible confound of maternal psychopathology and maternal behavior
problem ratings, it does provide some support for the contention that maternal
alcoholism, antisociality, and depression play a role in establishing the initial risk
trajectory. Similar ﬁndings have been reported by other researchers from the Michigan
Longitudinal Study (Fitzgerald et al., 1993; Fitzgerald et al., 1995).

Why maternal psychopathology but not paternal psychopathology? This may be
an artifact of sample characteristics. Paternal psychopathology was over-represented
among these children because the Michigan Longitudinal Study is a population-based
study of paternal alcoholism. A disproportionate number of alcoholic fathers (i.e.,
approximately two-thirds) were recruited for this reason. Therefore, there may be a
saturation effect with respect to paternal psychopathology. What probably makes a

difference is whether children are exposed to psychopathology from both parents versus

154

one or none. This echoes the conclusion drawn by Mun (2002) for the inﬂuence of only
one versus both parents’ smoking and alcoholism on early onset smoking by offspring.
Having a mother low in psychopathology may serve as an important protective factor in
the child’s deve10pment (directly buffering the effects of paternal deviance, or with an
alternate set of behaviors being modeled).

Early behavior problems were the best predictors of adolescent behavior problems
when predicting adolescent outcomes. This is strong evidence in support of the hard-
continuity model. In addition, risk/promotion factors from middle childhood were much
more inﬂuential than other facets of the initial risk trajectory. For example, preschool
reactivity was no longer a key inﬂuence when predicting this distal outcome. These
ﬁndings speak to the viability of prevention and intervention efforts which don’t get
started until school entry: Middle childhood matters; it’s not too late to mitigate some of
the risk associated with adolescent behavior problems.

Controlling for initial levels of aggression, key predictors of aggression in early
adolescence also included global self-esteem, social problems, and family conﬂict at ages
6-8; as well as family cohesion, family expressiveness, physical discipline, negative
control, and general activity at ages 9-11. When contemporaneous behavior problems
were taken into consideration, only global self-esteem at ages 6-8 also remained
signiﬁcant. At ages 9-11, the punishment variables (physical discipline, negative control)
dropped from signiﬁcance but all other predictors remained the same.

Controlling for initial levels of delinquent behavior, other key predictors of

delinquent behavior in early adolescence included maternal psychopathology and social

155

problems at ages 6-8; but only general activity at ages 9-11. Only the delinquent
behavior control variables remained signiﬁcant in the contemporaneous model.

Of these ﬁndings, several are particularly intriguing. Social problems at 6-8 were
a signiﬁcant predictor of later aggression and delinquent behavior. However, social
problems were no longer signiﬁcant at 9-11 years of age (once initial levels of aggression
and delinquent behavior were taken into account). This reafﬁrms the importance of peer
relations as a key marker in the developmental trajectory. As mentioned, social
problems have featured prominently in other research on behavioral difficulties
(Campbell, 1994; Campbell et al., 1986, Campbell & Ewing, 1990) and theories of
delinquency - including the work of Patterson et al. (1989) as well as other
criminological perspectives (e. g., social control theory and attachment to conventional
others, social learning theory and differential exposure to deﬁnitions for versus against
delinquent behavior, and general strain theory with peer rejection as a noxious
experience). However, what really seemed to matter here were early social problems.
Rather than focusing on social problems in adolescence or adolescent peer groups, then,
prevention/intervention efforts need to direct their attention developmentally upstream.
“First impressions” can have a lasting impact in the social circles of school children, and
peer rejection can exacerbate behavior problems above and beyond the effect of initial
behavior problem symptomatology (albeit not contemporaneous symptomatology).
Programs that redress deﬁcits in social skills might help — especially if they are
implemented in the ﬁrst few years of school, or earlier.

It is also noteworthy that lower self-esteem at ages 6-8 also predicted more

adolescent aggression (even controlling for contemporaneous aggression). The same was

156

not true for delinquent behavior. Self-esteem therefore seems to play a unique role in the
development of the aggressive behavior repertoire. Along with social skills training,
early prevention/intervention programs can capitalize on the importance of self-esteem by
fostering healthy, positive self-perceptions as well as healthy, positive social
relationships.

Regression results helped clarify the relationship between contextual variables
and antisocial behavior. In particular, controlling for contemporaneous aggression at
Wave 3 negated the effects of the punishment variables (physical discipline, negative
control) on later aggression. This suggests that level and style of punishment may in fact
have been in response to the severity of aggression exhibited by children — that is, the
evocative explanation. However, other family environment variables remained
signiﬁcant: family cohesion and family expressiveness. Therefore, even taking into
account concurrent levels of aggressive symptomatology, less cohesion and more
expressiveness at 9-11 predicted more aggression at 12-14 years of age. Lack of
cohesion and more expressiveness may be important risk factors that either maintain the
developmental trajectory for those with behavioral difﬁculties or “turn the tide” for
vulnerable others.

The fact that greater family expressiveness predicted more aggression during
early adolescence seems counterintuitive. Expressiveness was originally conceptualized
to reﬂect openness to discussion as characteristic of a warm home environment.
However, items such as “ We tell each other about our personal problems”, “Money and
paying bills is openly talked about in our family,” and “There are a lot of spontaneous

discussion in our family” may not uniquely capture the positive side of expressiveness.

157

In disadvantaged homes, homel with parental psychopathology, or homes with difﬁcult
children, expressiveness may actually contribute to more conﬂict and tension. In other
words, perhaps the valence for expressiveness is conditioned by the larger social context.
Finally, there is a very striking difference between aggression and delinquent
behavior in the models controlling for contemporaneous behavior problems. Although
some contextual and child characteristics remained important in the prediction of
aggression, only earlier delinquent behavior predicted later delinquent behavior. The
implication here is that aggression is partially mediated by middle childhood
risk/promotive factors whereas delinquent behavior is driven by its own core diathesis.
The results of the analyses linking onset of ﬁrst alcohol use to delinquent behavior but
not aggression led to the suggestion that delinquent behavior might be more “deviant.”
Whether or not it is in fact more serious, delinquent behavior does seem to follow the
hard-continuity paradigm more conservatively (i.e., have at its center a harder “core”).
Of relevance here may be the distinction Mofﬁtt (1993) makes between
cumulative consequences and contemporary consequences. Cumulative consequences
are those that snowball from early individual differences: Aggressive or delinquent
children may experience attenuated parent-child interactions, social rejection, and
academic difﬁculties that exacerbate their initial predispositions — leading to a
progressive deterioration of personal and social assets or more deep epigenetic ruts.
Contemporary consequences are those that result from ongoing or current behavior
problems. Cumulative consequences may be more important for aggression, with
interplay in the personal and social surround determining later outcomes; contemporary

consequences may be more important for delinquent behavior, with this underlying

158

constellation of traits serving as the dominant driving force in its own behavioral
trajectory.

One implication of this ﬁnding is that adolescent aggression may be more
amenable to prevention/intervention programs that address not only the use of aggression
itself but also personal and contextual factors that have a unique effect on later aggression
(above and beyond initial and contemporaneous symptomatology). On the other hand,
prevention/intervention programs for delinquent behavior would be best served by
focusing on the “hard-cored” nature of delinquent behavior itself.

Limitations and Future Direction_s

The current study has several limitations. These relate to 1) the nature of the
sample, 2) the operationalization of key constructs, 3) use of maternal report, and 4)
factors not included in analyses.

The sample used for this study was racially and geographically homogeneous.
Participants were mostly Caucasian and from the mid-Michigan area. It would be
desirable to replicate this research in other geographic locations and with a more diverse
population. In particular, there may be racial, ethnic, or cultural differences in the factors
that contribute to the maintenance of behavior problem trajectories or that contribute to
“turning points” in development. This was also a high-risk sample, heavily
characterized by paternal psychopathology. It would be useful to replicate these ﬁndings
in a general-population sample, such as the National Longitudinal Survey of Youth
(N LSY). In addition, families were intact at recruitment. Although some decoupling
occurred after the ﬁrst wave, and step-parents were added to the assessment protocol,

there may be important differences between these families and families with different

159

structures (e.g., single-parent, earlier-divorced, non-biological, etc.)

Another sample-related limitation involves the exclusive focus of this study on
male children. Do the same risk factors buttress the developmental process for girls and
boys? There were insufficient data on daughters to allow for inclusion in the sample or
to look at gender differences because of the importance of Wave 1 and Wave 4 variables
to the theoretical models tested. However, daughters from the Michigan Longitudinal
study were started at later assessments; some were young enough to be assessed at 3-5
years old and will eventually reach their own Wave 4 (12-14). This will enable the types
of analyses that were untenable for the current study. Alternatively, other samples could
be used to investigate sex differences in the development of antisocial behavior as well as
its relationship to ﬁrst alcohol use.

A second area of limitation involves the operationalization of key variables. One
issue is with the way that high and low levels of aggression or delinquent behavior were
deﬁned. Because of the relatively small sample size, mean splits were used to deﬁne
“high” and “low” levels of symptomatology. “High” and “low” therefore were not
extreme classiﬁcations. Different results might have obtained if more strict deﬁnitions of
“high” and “low” were used. For example, it would be useful to examine those who
reach clinical thresholds for aggression or delinquent behavior or who are at least one
standard deviation above the mean. Future research needs to look at the difference
between sub-clinical ratings and more conservative classiﬁcation schemes. In addition, it
would be useful to replicate these ﬁndings using a more pure delinquency scale, not
constructed originally to include substance use (as in the CBCL) and focusing on

delinquent acts more than a syndrome or constellation of behaviors (e. g., having older

160

friends) that “hang together” with delinquency.

Another operationalization issue concerns the use of the DOTS to measure
activity level. In the DOTS, items used to assess activity level are sleep-oriented.
Activity during sleep may not be the best indicator of general activity level (i.e.,
hyperactivity, which been implicated in the etiology of antisocial behavior; Moffltt,
1993). This may explain the lack of signiﬁcance of activity level as a key initial risk
trajectory marker in the current study. The DOTS-R includes both activity-general and
activity-sleep, and it is when the DOTS-R is employed that marked group differences in
activity-general were observed. Unfortunately, the DOTS-R was added to the assessment
protocol subsequent to the ﬁrst few waves (i.e., in the middle of Wave 2) and complete
DOTS-R data weren’t available until Wave 3.

The third major limitation of the current study involves the use of maternal
reports for describing child and family characteristics. Youth self-reports were generally
not available until Wave 4. Maternal reports were chosen over paternal reports because
this is common in the child development literature and because of a priori knowledge
about the prevalence of psychopathology among fathers in the Michigan Longitudinal
Study (e. g., paternal alcoholism was a criterion for recruitment in the majority of
families, but maternal alcoholism was neither a requirement for study inclusion nor a
basis for exclusion). In addition, there was decoupling in families after the ﬁrst wave of
data collection. Mothers were the primary caretakers in these decoupled families, and
therefore could provide more accurate information about the family environment to
which children were exposed during middle childhood. Also using data from the

Michigan Longitudinal Study, Mun (2002) found that maternal ratings of temperament

161

and behavior problems were more predictive of smoking onset in children than paternal
ratings. The use of averaged maternal and paternal reports was considered but deemed
unadvisable for several reasons. Speciﬁcally, there was no guarantee that mothers and
fathers would have the same relationships with their children or perceive their children
similarly; averaging their ratings might therefore create more “white noise” than
meaningful amalgarns.

There were indeed differences between maternal and paternal ratings in this study.
With 11 = 220 at Wave 1 and n = 186 at Wave 4, a retrospective analysis revealed that
mother-father correlations were moderately strong in early adolescence but weaker when
their children were of preschool age. For aggression, r = .36 at Wave 1 and r = .52 at
Wave 4. For delinquent behavior, r = .27 at Wave 1 and r = .40 at Wave 4.13 At most,
here, maternal and paternal CBCL scores shared 27% of their variance (i.e., r2 = .27 for
Wave 4 aggression). Note that sample size varies because these data were pre-imputation
(given that paternal data were not included in missing data estimation).

Future research on the difference between maternal and paternal reports would be
interesting and may help to illuminate any effect of rater on the results obtained. More
ideal would be a study utilizing child self-reports. It may be that the child’s own
perceptions drive his own behavior more — and would be more accurate for this reason —
than the feelings his parents have about family environment, discipline, or the social

problems experienced by their sons.

 

'3 Similar results were obtained when the correlations were run for all male target children in the

Michigan Longitudinal Study: for aggression, r = .39 at Wave 1 (n = 321) and r = .54 at Wave 4 (n = 209);

for delinquent behavior, r = .25 at Wave 1 (n = 321) and r = .46 at Wave 4 (n = 209).

162

Factors not included in the current study (e.g., parental monitoring, peer substance
use, peer delinquency, attachment to school and family, neighborhood violence, poverty)
represent a fourth limitation. Parental monitoring and delinquent peers have been
implicated in a number of criminological models (e.g., Patterson et al., 1989) but are not
directly assessed with this sample until Wave 4. The Michigan Longitudinal Study loads
heavily on the individual and parent-child causal aspects of risk accumulation, but
broader structural and environmental inﬂuences (e. g., violent neighborhoods, poverty)
might also help to explain the development of antisocial behavior. Neighborhood
violence was not assessed, and socioeconomic status was not included in the current
analyses. These variables likely exert their effect through more proximate processes
(e. g., parental psychopathology, parent-child relationships, family conﬂict, use of harsh
discipline, self-esteem, school achievement), but their inclusion in other research
endeavors would allow for further delineation of the underlying causal matrix. Finally,
attachments to school and family were not measured to the extent required for tests of
social control theory (Hirschi, 1969; Sampson & Laub, 1993). These are likely to be
important etiological inﬂuences in the development of aggression and delinquent
behaviors, and future research is needed to assess their predictive utility within the larger
framework adopted by the present study. Too often these variables are studied with a
cross-sectional research design or without consideration of the continuity versus change
perspective adopted here and in recent work within developmental criminology.

Despite these limitations, the ﬁndings from the current study are robust in their
consistency. Results point to several potentially viable foci for intervention. For

aggression especially, these include personal and contextual factors such as self-esteem,

163

social problems, and family functioning. The aggression and delinquent behaviors
themselves should also be addressed, given that their substrates are apparent quite early
and buttress an emergent, self-perpetuating developmental trajectory. A logical question
might now be: where should we go from here?

Ideally, the utility of different intervention strategies would be assessed in a
randomized clinical trial. However, in contrast to studies that compare an intervention
group to a follow-up only group, more complex research designs are recommended. For
example, one might randomly assign preschoolers to one of several conditions:
intervention addressing problem behaviors directly - perhaps through cognitive
restructuring, behavior modiﬁcation techniques, or training in self-regulation/self-control;
intervention addressing family functioning (e. g., the quality of parent-child relations,
discipline); intervention addressing personal competencies (e.g., self-esteem, social
skills); interventions combining two or three of the above; and no intervention/follow-up
only. This methodology would help researchers disentangle the many causal factors now
under consideration, and help to identify those most efﬁcacious to problem behavior
intervention/prevention.

The appeal of randomized trials in the developmental sciences is evidenced by
recent efforts to adopt this research design - including the Early Head Start National
Research and Evaluation Project (Love et al., 2002), among others (see Brooks-Gunn,
2003, for some examples). It is interesting to mentally map out how this model might be
used to test the implications of the current study, and a randomized trial would deﬁnitely
be the most scientiﬁcally rigorous option. More important than whether future research

adopts a randomized or quasi-experimental design, however, is that movement continue

164

toward a better understanding of prevention and intervention factors. This call to action
can be summarized as follows: “It is to be hoped that the next generation of studies will
get further inside the ‘black box’ of program design to tell us what features of these
programs are key to successful early intervention” (Currie, 2003, p. 5). Another
progressive step would be the creation of a national registry of prevention trials for the
behavioral sciences, to improve knowledge dissemination and facilitate meta-analyses
(Biglan, Mrazek, Carnine, & Flay, 2003).

In Conclusion

Early onset of drinking is more strongly tied to the developmental course of
delinquent behavior than aggression. It appears to be part of a longitudinal pattern of
delinquent behavior problems traceable to preschool — that is, as young as 3 years of age.
Preschool levels of delinquent behavior had a direct effect on later levels of
symptomatology for early drinkers only, supporting the hard-continuity model and the
role of initial predispositions on later outcomes for a small subset of individuals.

One implication of this ﬁnding is that intervention and prevention strategies
geared towards early alcohol use should focus on this longitudinal pattern of problem
behavior. Programs that focus only on trying to delay ﬁrst alcohol use will probably fail
to prevent its sequelae (e. g., alcohol problems) because it is likely that both are part of a
larger developmental trajectory. Unless this larger developmental trajectory is addressed,
intervention and prevention efforts will be mere ‘band-aids’ rather than “cures”. On the
other hand, a true disruption of the developmental trajectory may dismantle the structure
that supports early alcohol use.

A second consistent ﬁnding is in the importance of proximate versus distal risk

165

factors in the continuity of aggression and delinquent behavior from preschool to
adolescence. The initial risk trajectory or core diathesis is modiﬁed or maintained by
current inﬂuences in the personal and social surround of young children. For example,
temperament, social problems, and the family environment during middle childhood are
better predictors of an adolescent’s aggression or delinquent behavior than temperament
during preschool. This points to two key conclusions. First, risk is not immutable or set
in stone during infancy. What matters are ongoing person-environment interactions and
the developmental process. Risk may accumulate to become more deeply rutted, or
change and provide the opportunity for a behavioral “turning point.” Second,
prevention/intervention programs that include middle childhood risk/promotive factors
can have an effect on later outcomes; they may offer “windows of opportunity” for
change.

The hard-continuity model applied more to delinquent behavior than aggression.
The inﬂuence of preschool on adolescent levels of delinquent behavior was especially
pronounced among early drinkers. In addition, delinquent behavior during adolescence
was driven more exclusively by prior delinquent behavior than the parallel etiology for
aggression. Stated conversely, middle childhood risk/promotive factors have a stronger
inﬂuence on aggression (controlling for initial and contemporaneous levels of aggression)
than they do on delinquent behavior. Prevention efforts for delinquent behavior therefore
need to more directly attack the delinquent behavior trajectory, whereas prevention
efforts for aggression may beneﬁt from efforts to control aggression as well as efforts to
ameliorate other personal and contextual inﬂuences. Delinquent behavior (e. g., lying,

stealing, setting ﬁres) seems to be more ‘hard-cored” than aggression (e.g., bullying,

166

ﬁghting, showing off).

Finally, the current study reaches the same conclusion as Stanger et al. (1997, p.
55): “These ﬁndings suggest that failure to distinguish between these patterns of
problems in developmental research could generate misleading conclusions and limit the
generalizability of results.” Aggression and delinquent behavior are often aggregated
together under the overarching label of “externalizing” or “antisocial” problems, but
major developmental differences were observed here. Delinquent behavior and
aggression followed different age-related trajectories, were differentially associated with
other problem behaviors (i.e., early drinking), and were differentially predicted by
individual, interpersonal, and familial variables in this study. Ultimately, there may be
very important reasons to look at these behaviors as often co-morbid but distinct —

conceptually and developmentally.

167

REFERENCES

Achenbach, T. (1991). Manual for the child behavior checklist/ 4-18 and 1991
proﬁle. Burlington, VT: University of Vermont Department of Psychiatry.

Achenbach, T., & Edelbrock, C. (1983). Manual for the child behavior checklist
and revised child behavior proﬁle. Burlington, VT: University Associates in Psychiatry.

Ackerman, B. P., Brown, E., & Izard, C. E. (2003). Continuity and change in
levels of externalizing behavior in school of children from economically disadvantaged
families. Child Development. 74(3), 694-709.

Ackerman, B. P., Izard, C. E., Schoff, K., Youngstrom, E. A., & Kogos, J.
(1999). Contextual risk, caregiver emotionality, and the problem behaviors of six- and

seven-year-old children from economically disadvantaged families. Child Development.
3(6), 1415-1427.

Allison, PD. (2002). Missing data. Thousand Oaks, CA: Sage Publications.

Anastasi, A. (1958). Heredity, environment, and the question “how”.
Psychological Review, 65(4), 197-208.

Baldwin, A. L., Baldwin, C. P., Kasser, T., Zax, M., Sameroff, A., & Seifer, R.
(1993). Contextual risk and resiliency during late adolescence. Development ﬂ!
Psvclmathology, 5, 741-761.

Barrera, M., Jr., Li, S. A., & Chassin, L. (1995). Effects of parental alcoholism
and life stress on Hispanic and non-Hispanic Caucasian adolescents: A prospective study.

American Journal of Communig Psychology, 23(4), 479-507.

 

Barocas, R., Seifer, R., & Sameroff, A. J. (1985). Deﬁning environmental risk:
Multiple dimensions of psychological vulnerability. American Journal of Community
chhologv. 13(4), 433-447.

Beck, A. T., & Beck,. R. W. (1972). Screening depressed patients in family
practice. Postgraduate Medicine, 52, 81-85.

Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric properties of the
Beck Depression Inventory: Twenty-ﬁve years of evaluation. Clinigll Psychology
Review 8, 77-100.

 

Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An
inventory for measuring depression. Archives of General Psychiatry, 4, 561-571.

168

Biglan, A., Mrazek, P. J ., Carnine, D., & Flay, B. R. (2003). The integration of
research and practice in the prevention of youth problem behaviors. American
Psychologist. 58(6/7), 433-440.

Bongers, I. J ., Koot, H. M., van der Ende, J ., & Verhulst, F. C. (2003). The
normative development of child and adolescent problem behavior. Journal of Abnormal

Psychology, 112(2), 179-192.

Bronfenbrenner, U. (1977). Toward an experimental ecology of human
development. American Psychologist. 32. 513-531.

Brooks-Gunn, J. (2003). Do you believe in magic? What we can expect from
early childhood intervention programs. Social Policy Report. 17(1), 3-15

Bry, B. H., McKeon, P., & Pandina, R. J. (1982). Journal of Abnormal
Psychologyr9l(4), 273-279.

Buss, A. H. (1991). The EAS theory of temperament. In J. Strelau & A.
Angleitner (Eds.), Explorations in temperament: International perspectives on theory and
measurement (pp. 43-60). New York: Plenum Press.

Cahalan, D., Cisin, I. H., & Crossley, H. M. (1969). American drinkglg
practices: A national study of drinking behavior and attitudes. New Brunswick, NJ:
Rutgers Center of Alcohol Studies.

Cairns, R. B., Cairns, B. D., Neckerman, H. J ., Ferguson, L. L., & Gariépy, J-L.
(1989). Growth and aggression: 1. Childhood to early adolescence. Developmental
Psychology. 25(2), 320-330.

Campbell, S. B. (1994). Hard-to-manage preschool boys: Extemalizing
behavior, social competence, and family context at two-year follow-up. Journal of
Abnormal Child Psychology. 22(2), 147-166.

Campbell, S. B., & Ewing, L. J. (1990). Follow-up of hard-to-manage
preschoolers: Adjustment at age 9 and predictors of continuing symptoms. Journal of

Child Psychiagy, 31(6), 871-889.

Campbell, S. B., Ewing, L. J ., Breaux, A. M., & Szumowski, E. K. (1986).
Parent-referred problem three-year-olds: Follow-up at school entry. Journal of Child
Psychiatry. 27(4), 473-488.

Campbell, S. B., Pierce, E. W., March, C. L., Ewing, L. J ., & Szumowski, E. K.

(1994). Hard-to-manage preschool boys: Symptomatic behavior across contexts and
time. Child Development. 65. 836-851.

169

Carroll, B. J ., Fielding, J. M., & Blashki, T. G. (1973). Depression rating scales:
A critical review. Archives of General Psychiatry. 28. 361-366.

Caspi, A., Henry, B., McGee, R. O., Mofﬁtt, T. E., & Silva, P. A. (1995).
Temperamental origins of child and adolescent behavior problems: From age three to
age ﬁfteen. Child Development. 66, 55-68.

Caspi, A., Mofﬁtt, T. E., Newman, D. L., & Silva, E. A. (1996). Behavioral
observations at age 3 years predict adult psychiatric disorders: Longitudinal evidence
from a birth cohort. Archives of General Psychiatry. 53. 1033-1039.

Chamberlain, P. (1980). A parengaily report meaaure. Unpublished doctoral
dissertation, University of Oregon, Eugene, OR.

Chassin, L., Pillow, D. R., Curran, P. J., Molina, B. S., & Barrera, M., Jr. (1993).
Relation of parental alcoholism to early adolescence substance use: A test of three
mediating mechanisms. Journal of Abnormal Psychology, 102(1), 3-19.

Chess, S., & Thomas, A. (1989). Temperament and its functional signiﬁcance.
In 8.1. Greenspan & G. H. Pollock, (Eds), The course of life (volume 11): Early
childhood (pp. 163-227). Madison, Connecticut: International Universities Press, Inc.

Cloninger, C. R., Sigvardsson, S., & Bohman, M. (1988). Childhood personality
predicts alcohol abuse in young adults. Alcoholism: Clinical and Experimegtal
Research. 12(4), 494-505.

Cloninger, C. R., Sigvardsson, S., & Bohman, M. (1996). Type I and type II
alcoholism: An update. Alcohol Health and Resegch World, 20(1). 18-23.

Cohen, D. A., Richardson, J ., & LaBree, L. (1994). Parenting behaviors and the
onset of smoking and alcohol use: A longitudinal study. Pediatrics. 94(3), 368-375.

Coie, J. D., & Jacobs, M. R. (1993). The role of social context in the prevention
of conduct disorder. Developmentg and Psychopaghology, 5, 263-275.

Collins, L. M., Schafer, J. L., & Kam, C-M. (2001). A comparison of inclusive
and restrictive strategies in modern missing data procedures. bycholagical Methods,
6(4), 330-351.

Costa, F. M., Jessor, R., & Turbin, M. S. (1999). Transition into adolescent
problem drinking: The role of psychosocial risk and protective factors. Journal of
Studies on Alcohol. 60. 480-490.

Cummings, E. M., Iannotti, R. J ., & Zahn-Waxler, C. (1989). Aggression

between peers in early childhood: Individual continuity and developmental change.
Child Development. 60(4), 887-895.

170

Currie, J. (2003). What we can expect from early childhood intervention
programs. Social Policy Report, 17(1), 5.

Dekovic, M. (1999). Risk and protective factors in the development of problem
behavior during adolescence. Joumg of Youth and Adolescence. 28(6), 667-685.

Dixon, R. A., & Lerner, R. M. (1992). A history of systems in developmental
psychology. In M. H. Bomstein & M. E. Lamb (Eds), Developmental psychology: An
ﬂamed textboci (3-58). Hillsdale, NJ: Lawrence Erlbaum Associates.

Dryfoos, J. G. (1990). Adolescents at risk: Prevalence and prevention. New
York: Oxford University Press.

Dryfoos, J. G. (1998). Safe passage: Making it through adolescence in a risky
socieg. New York: Oxford University Press.

Egeland, B., Carlson, E., & Sroufe, L. A. (1993). Resilience as a process.
Developmentgand Psychopathology. 5. 517-528.

Elkins, 1. J ., Iacono, W. G., Doyle, A. E., & McGue, M. (1997). Characteristics
associated with the persistence of antisocial behavior: Results from recent longitudinal
research. Aggression and Violent Behavior. 2(2), 101-124.

Ellickson, P. L., & Hays, R. D. (1991). Antecedents of drinking among young
adolescents with different alcohol use histories. Journal of Studies on Alcohol. 52(5),
398-408.

Elliott, D. S., Huizinga, D., & Menard, S. (1989). Multiple problem youth:
Delinquency. substance use. and mental health problems. New York: Springer-Verlag.

Ellis, D. A., Zucker, R. A., & Fitzgerald, H. E. (1997). The role of family
inﬂuences in development and risk. Alcohol Health & Research World. 21(3), 218-226.

Erikson, E. H. (1950). Childhood and society. New York, NY: W. W. Norton &
Company, Inc.

Eron, L. D., Huesmann, L. R., Dubow, E., Romanoff, R., & Yarmel, P. W.
(1987). Aggression and its correlates over 22 years. In D. H. Crowell, I. M. Evans, & C.
R. O’Donnell (Eds). Childhood aggression and violence: Sources of inﬂuence.
ﬁvention, and control (pp. 249-262). New York: Plenum Press.

Evans, G. W. (2003). A multimethodological analysis of cumulative risk and
allostatic load among rural children. Developmental Pﬂcholqu, 39(5), 924-933.

F arrington, D. P. (1995). The development of offending and antisocial behaviour
from childhood: Key ﬁndings from the Cambridge Study in Delinquent Development.

171

Journal of Child Psychology__and Psychiatry. 360(6), 929-964.

Farrington, D. P. (1997). Early prediction of violent and non-violent youthful
offending. European Joumal on Criminal Policy and Research. 5(2), 51-66.

Feighner, J. P., Robins, E., Guze, S. B., Woodruff, R. A., Winokur, G., & Munoz,
R. (1972). Diagnostic criteria for use in psychiatric research. Archives of General
Psychiatry. 26(1), 57-63.

Felton, F., Parsons, M. A., Ward, D. S., Pate, R. R., Saunders, R. P., Dowda, M.,
& Trost, S. (1999). Tracking of avoidance of alcohol use and smoking behavior in a
ﬁfth grade cohort over three years. Public Health Nursing, 16(1), 32-40.

Fergusson, D. M., Lynskey, M. T., & Horwood, J. (1996). Factors associated
with continuity and changes in disruptive behavior patterns between childhood and
adolescence. Journal of Abnormal Child Psychology. 24(5), 533-553.

Fitzgerald, H. E., Puttler, L. I., Mun, E-Y., & Zucker, R. A. (2000). Prenatal and
postnatal exposure to parental alcohol use and abuse. In J. D. Osofsky & H. E. Fitzgerald
(Eds.). WAIMH handbook of infant mental health: Volume 4. Infant mental health in
groups at high risk (pp. 123-159). New York: Wiley.

Fitzgerald, H. E., Sullivan, L. A., Ham, H. P., Zucker, R. A., Bruckel, S., &
Schneider, A. M. (1993). Predictors of behavioral problems in three-year-old sons of
alcoholics: Early evidence for onset of risk. Child Development. 64. 110-123.

Fitzgerald, H. E., Zucker, R. A., & Yang, H-Y. (1995). Developmental systems
theory and alcoholism: Analyzing patterns of variation in high-risk families. Psychology
of Addictive Behaviors. 9(1), 8-22.

Ford, D. H., & Lerner, R. M. (1992). Developmental aystems theory: An
integgative approach. Newbury Park, CA: Sage Publications.

Fraser, M. W. (1996). Aggressive behavior in childhood and early adolescence:
An ecological-developmental perspective on youth violence. Social Work, 41(4), 347-
361.

Fuller, B. E., Chermak, S. T., Cruise, K. A., Kirsch, E., Fitzgerald, H. E., &
Zucker, R. A. (2003). Predictors of aggression across three generations among sons of
alcoholics: Relationships involving grandparental and parental alcoholism, child

aggression, marital aggression and parenting practices. Journal of Studies on Alcohol.
_6_4_(4), 472-483

Garmezy, N., Masten, A. S., & Tellegen, A. (1984). The study of stress and

competence in children: A building block for developmental psychopathology. Child
Development. 55(1), 97-111.

172

Gest, S. D., Neemann, J., Hubbard, J. J., Masten, A. S., & Tellegen, A. (1993).
Parenting quality, adversity, and conduct problems in adolescence: Testing process-
oriented models of resilience. Developmentg and Psychmathology, 5, 663-682.

Gilgun, J. F., Klein, C., & Pranis, K. (2000). The signiﬁcance of resources in
models of risk. Journal of Intgpersonal Violence. 15(6), 631-650

Gottfredson, Michael R. and Travis Hirschi. (1990). _A_genmtl theory of crime.
Stanford: Stanford University Press.

Gottlieb, G. (1991). Experiential canalization of behavioral development:
Theory. Developmentai Psychology. 27(1), 4-13.

Gottlieb, G. (2001). The relevance of developmental-psychobiological
metatheory to developmental neuropsychology. Developmental Neuropsycholqgm 19(1),
1-9.

Graham, J. W., & Hofer, S. M. (2000). Multiple imputation in multivariate
research. In T. D. Little, K. U. Schnabel, & J. Baumert (Eds), Modeling longitudinal and

multilevel data: Practicgal issues. applied approacheskand speciﬁc examples (pp. 201-
218).Mahwah, NJ: Lawrence Erlbaum Associates.

Grant, B. F., & Dawson, D. A. (1997). Age at onset of alcohol use and its
association with DSM-IV alcohol abuse and dependence: Results from the National
Longitudinal Alcohol Epidemiological Survey. Journal of Substance Abuse. 10. 59-73.

Green, S. B. (1991). How many subjects does it take to do a regression analysis?
Multivariate Behavioral Resaagch. 26(3), 499-510.

Gruber, E., DiClemente, R. J ., Anderson, M. M., & Lodico, M. (1996). Early
onset drinking and its association with alcohol use and problem behavior in adolescence.
Preventive Medicine. 25(3), 293-300.

Harter, S. (1978). Effectance motivation reconsidered: Toward a developmental
model. Human Development, 21, 34-64.

Harter, S. (1979). Manual for perceived competence scale for children. Denver,
CO: University of Denver.

Havighurst, R. J. (1972). Developmental tasks and education. New York, NY:
David McKay Company, Inc.

Hawkins, C. A. (1997). Disruption of family rituals as a mediator of the
relationship between parental drinking and adult adjustment in offspring. Addictive
Behaviors 22(2), 219-23 1.

 

173

Hazzard, A., Christensen, A., & Margolin, G. (1983). Children’s perceptions of
parental behaviors. Jougnal of Abnormal Child Psychology, 2, 49-60.

Hecht, M. (1973). Children of alcoholics. American Journal of Nursing. 73(10),
1764-1767.

Henderson, M. C., Albright, J. S., Kalichman, S. C., & Dugoni, B. (1994).
Personality characteristics of young adult offspring of substance abusers: A study
highlighting methodological issues. Jougnal of Personality Assesglent. 63(1), 117-134.

Henry, B., Caspi, A., Moffitt, T. E., & Silva, P. A. (1996). Temperamental and
familial predictors of violent and non-violent criminal convictions: Age 3 to age 18.
Developmental Psychology, 32(4), 614-623.

Hetherington, E. M. (1989). Coping with family transitions: Winners, losers,
and survivors. Child Development. 60. 1-14.

Hill, S. Y., & Yuan, H. (1999). Familial density of alcoholism and onset of
adolescent drinking. Journal of Studies on Alcohol. 60(1), 7-17.

Hirschi, T. (1969). Causes of delinquency. Berkeley, CA: University of
California Press.

Hops, H., Davis, B., & Lewin, L. M. (1999). The development of alcohol and
other substance use: A gender study of family and peer context. J ournal of Studies on
Alcohol Supplement No. 13, 22-31.

 

Hothersall, D. (1995). History of psychology. New York: McGraw-Hill, Inc.

Huesmann, L. R., Eron, L. D., Lefkowitz, M. M., & Walder, L. O. (1984).
Stability of aggression over time and generations. Developmental Psychology. 20(6),
1120-1134.

Hussong, A. M., & Chassin, L. (1997). Substance use initiation among

adolescent children of alcoholics: Testing protective factors. Journal of Studies on
Alcohol 58(3), 272-279.

 

Hyphantis, T., Koutras, V., Liakos, A., & Marselos, M. (1991). Alcohol and
drug use, family situation, and school performance in adolescent children of alcoholics.
International Journal of Social Psychiatry. 37(1), 35-42.

Jack, G. (2000). Ecological inﬂuences on parenting and child development.
British Joumagf Social WorQOw), 703-720.

174

Jacob, T., Krahn, G. L., & Leonard, K. (1991). Parent-child interactions in
families with alcoholic fathers. Journal of Coun_seling_and Clinical Psychology. 59(1),
176-181.

Jansen, R. E., Fitzgerald, H. E., Ham. H. P., & Zucker, R. A. (1995). Pathways
into risk: Temperament and behavior problems in three- to ﬁve-year-old sons of
alcoholics. Alcoholism: ClinicaL and Experimental Research. 19(2), 501-509.

Jessor, R., & Jessor, S. L. (1975). Adolescent development and the onset of
drinking: A longitudinal study. Journal of Studies on Alcohol. 36(1), 27-51.

Jessor, R., & Jessor, S. L. (1977). Problem btﬂvior and psychosocial
development: A longitudinalatudy of youth. New York: Academic Press.

Johnston, L. D., Bachman, J .G., & O’Malley, P. M. (1979). Drugs and the class
of ’78: Behaviors. attitudes. apd recent national trends. Washington, DC: National
Institute on Drug Abuse, Division of Research, US. Department of Health, Education,
and Welfare.

Keenan, K., & Shaw, D. S. (1994). The development of aggression in toddlers:
A study of low-income families. Journal of Abnormal Child Psychology, 22(1), 53-77.

Kitchener, R. F. (1978). Epigenesis: The role of biological models in
developmental psychology. Human Development, 21(3), 141-160.

Krohn, M. D., Lizotte, A. J ., & Perez, C. M. (1997). The interrelationship
between substance use and precocious transitions to adult statuses. Journal of Health and
Social Behavior, 38. 87-103.

Larzelere, R. E., & Patterson, G, R. (1990). Parental management: Mediator of
the effect of socioeconomic status on early delinquency. Criminology, 28(2), 301-323.

Lehmacher,W. (1981). A more powerful simultaneous test procedure in
conﬁgural frequency analysis. Biomem Journal. 23(5), 429-436.

Lerner, R. M. (1980). Concepts of epigenesis: Descriptive and explanatory
issues: A critique of Kitchener’s comments. Human Development. 23(1), 63-72.

Lerner, R. M. (1985). Adolescent maturational changes and psychosocial
development: A dynamic interactional perspective. Journal of Youth and Adolescence.
15(4), 355-372.

Lerner, R. M., Castellino, D. R., Terry, P. A., Villarruel, F. A., & McKinney, M.
H. (1995). Developmental contextual perspective on parenting. In M. H. Bomstein

(Ed.), Handbook of parenting. Vol. 2: Biology and ecology of parenting (pp. 285-309).
Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

 

175

Lerner, R. M., Palermo, M., Spiro III, A., & Nesselroade, J. R. (1982) Assessing
the dimensions of temperamental individuality across the life span: The Dimensions of
Temperament Survey (DOTS). Child Development, 53, 149-159.

Little, R. J. A. (1988). A test of completely missing at random for multivariate
data with missing values. Jourﬁnal of ma American StatisticaiAsso_ciation. 83(404), 1198-
1202.

Loeber, R. (1982). The stability of antisocial and delinquent behavior: A review.
Child Development. 53(6), 1431-1446.

Loeber, R., & Dishion, T. (1983). Early predictors of male delinquency: A
review. Psychological Bulletin. 94(1), 68-99.

Loeber, R., & Hay, D. (1997). Key issues in the development of aggression and
violence from childhood to early adulthood. Annual Review of PsychologY. 48. 371-410.

Loeber, R., & Stouthamer-Loeber, M. (1998). Development of juvenile
aggression and violence: Some common misconceptions and controversies. American
Psychologist. 53(2), 242-259.

Loukas, A. (1997). Problem behavior development in sons of alcoholics:
Individual, family, and contextual inﬂuences. Unpublished doctoral dissertation,
Michigan State University.

Loukas, A., Fitzgerald, H. E., Zucker, R. A., & yon Eye, A. (2001). Parental
alcoholism and co-occurring antisocial behavior: Prospective relationships to

externalizing behavior problems in their young sons. Journal of Abnormal Child
Psychology. 29(2), 91-106.

Love, J. M., Kisker, E. E., Ross, C. M., Schochet, P. 2., Brooks-Gunn, J .,
Paulsell, D., Boller, K., Constantine, J ., Vogel, C., Fuligni, A. S., & Brady-Smith, C.
(2002). Making a difference in the lives of infants and toddlers and their families: The
impacts of Early Head Start. Washington, DC: US. Department of Health and Human
Services. Retrieved from http://www.mathematica-mpr.com/pub]ications/PDFs/

ehsfmalsumm.ﬂf.

Luthar, S. S. (1993). Annotation: Methodological and conceptual issues in
research on childhood resilience. Journal of Child Psychology and Psychiam and Allied
Disciplines. 34(4), 441-453.

Luthar, S. S., Cicchetti, D., & Becker, B. (2000a). Research on resilience:
Response to commentaries. Child Development. 71(3), 573-575.

Luthar, S. S., Cicchetti, D., & Becker, B. (2000b). The construct of resilience: A
critical evaluation and guidelines for future work. Child Development. 71(3), 543-562.

176

Luthar, S. S., Doemberger, C. H., & Zigler, E. (1993). Resilience is not a
unidimensional construct: Insights from a prospective study of inner-city adolescents.
Development and Psychopathology, 5, 703-717.

Malo, J ., & Tremblay, R. E. (1997). The impact of paternal alcoholism and
maternal social position on boys’ school adjustment, pubertal maturation and sexual
behavior: A test of two competing hypotheses. Joya] of Child Psychology and
Psychiaﬂ and Allied Disciplines. 38(2), 187-197.

Margulies, R. Z., Kessler, R. C., & Kandel, D. B. (1977). A longitudinal study of
onset of drinking among high-school students. Journal of Studies on Alcohol. 38(5),
897-912.

Masse, L. C., & Tremblay, R. E. (1997). Behavior of boys in kindergarten and
the onset of substance use during adolescence. Archives of General Psychiatry. 54. 62-
68.

Masten, A. S., Best, K. M., & Garmezy, N. (1990). Resilience and development:
Contributions from the study of children who overcome adversity. Develmgment an_d
Psychopathology, 2, 425-444.

Masten, A. S., & Coatsworth, J. D. (1998). The development of competence in
favorable and unfavorable environments. American Psychologist, 53(2), 205-220.

Masten, A. S., Hubbard, J. J ., Gest, S. D., Tellegen, A., Garmezy, N., & Ramirez,
M. (1999). Competence in the context of adversity: Pathways to resilience and

maladaption from childhood to late adolescence. Development and Psychopathology. 11.
143-169.

McGue, M. (1997). A behavioral-genetic perspective on children of alcoholics.
Alcohol Health & Research World, 21(3), 210-217.

McGue, M., Iacono, W. G., Legrand, L. N., Malone, S., & Elkins, 1. (2001b).
Origins and consequences of age at ﬁrst drink, I. Associations with substance-use
disorders, disinhibitory behavior, and psychopathology, and P3 amplitude. Alcoholism:
Clinical and Experimental Researcla 25(8), 1156-1165.

Mofﬁtt, T. E. (1990). Juvenile delinquency and attention deﬁcit disorder: Boys’
developmental trajectories from age 3 to age 15. Child Development. 61. 893-910.

Mofﬁtt, T. E. (1993). Adolescence-limited and life-course-persistent antisocial
behavior: A developmental taxonomy. Psychological Review. 100(4), 674-701.

Mofﬁtt, T. E. (1997). Adolescence-limited and life-course-persistent offending:
A complementary pair of developmental theories. In T. P. Thomberry (Ed),

177

Developmental theories of crime agl delinquency (pp. 11-54). New Brunswick, NJ:
Transaction Publishers.

Moos, R. (1974). Family Environment Scale Preliminm Manual. Palo Alto,
CA: Consulting Psychologists Press.

Moos, R., & Moos, B. S. (1976). A typology of family social environments.
Family Process. 15, 357-371.

Mun, E. Y. (2002). Risk for early smoking onset in a high-risk population:
Contributions of early child behavioral risk and heterogeneity of parental smoking and
alcoholism. Unpublished doctoral dissertation, Michigan State University.

Mun, E. Y., Fitgerald, H. E., von Eye, A., Puttler, L. I., & Zucker, R. A. (2001).
Temperamental characteristics as predictors of externalizing and internalizing child

behavior problems in the contexts of high and low parental psychopathology. Infant
Mental Health Journal. 22(3), 393-415.

Nagin, D., & Tremblay, R. E. (1999). Trajectories of boys’ physical aggression,
opposition, and hyperactivity on the path to physically violent and nonviolent juvenile
delinquency. Child Development, 70(5), 1 181-1 196.

Newcomb, M. D., & Felix-Ortiz, M. (1992). Multiple protective and risk factors
for drug use and abuse: Cross-sectional and prospective ﬁndings. Journal of Personalig
and Social Psychology. 63(2), 280-296.

Newcomb, M. D., Maddahian, E., & Bentler, P. M. (1986). Risk factors for drug
use among adolescents: Concurrent and longitudinal analyses. American Journal of
Public Health, 76(5), 525-531.

 

Olweus, D. (1979). Stability of aggressive reaction patterns in males: A review.
Psychological Bulletin, 86(4), 852-875.

Patterson, G. R., DeBaryshe, B. D., & Ramsey, E. (1989). A developmental
perspective on antisocial behavior. American Psychologist. 44(2), 329-335.

Perez McCluskey, C., Krohn, M. D., Lizotte, A. J ., & Rodriguez, M. L. (2002).
Early substance use and school achievement: An examination of Latino, White, and
Aﬁican American youth. Journg of Drug Issues, 32(3), 921-944.

Prescott, C. A., & Kendler, K. S. (1999). Age at ﬁrst drink and risk for

alcoholism: A noncausal association. Alcoholism: Clinical and Experimental Research,
§(l),101-107.

178

Richters, J. E., & Martinez, P. E. (1993). Violent communities, family choices,
and children’s chances: An algorithm for improving the odds. Development aad
Psyclmathologyé, 609-627.

Robins, L. N. (1978). Sturdy childhood predictors of adult antisocial behaviour:
Replications from longitudinal studies. Psychological Medicine. 8(4), 611-622.

Robins, L., Helzer, J. Croughan, J ., & Ratcliff, K. S. (1980). The NIMH
Diagnostic Interview Schedule: Its history, characteristics, and validity. St. Louis, MO:
Washington University School of Medicine.

Robinson, J. L. (2000). Are there implications for prevention research from
studies of resilience? Child Development. 71(3), 570-572.

Roosa, M. W. (2000). Some thoughts about resilience versus positive
development, main effects versus interactions, and the value of resilience. Child
Development, 71(3), 567-569.

Roosa, M. W., Beals, J ., Sandler, I. N., & Pillow, D. R. (1990). The role of risk
and protective factors in predicting symptomatology in adolescent self-identiﬁed children
of alcoholic parents. American Journal of Communig Psychology, 18(5), 725-741.

Rosenberg, M., Schooler, C., & Schoenbach, C. (1989). Self-esteem and
adolescent problems: Modeling reciprocal effects. American Sociological Review, 54,
1004-1018.

 

Rothbart, M. K. (1981). Measurement of temperament in infancy. Child
Development, 52(2), 569-578.

Rothbart, M. K. (1991). Temperament: A developmental framework. In J.
Strelau & A. Angleitner (Eds), Explorations in temperament: Intematicmal perspectives
on theory an_d measurement (pp. 61-74). New York: Plenum Press.

Rovine, M. J ., & Delaney, M. (1990). Missing data estimation in developmental
research. In A. von Eye (Ed), mistical Methods in Longitudinal Research: Volume I.
Principles and Structuring Change (pp. 35-79). San Diego, CA: Academic Press, Inc.

Rutter, M. (1985). Resilience in the face of adversity: Protective factors and
resistance to psychiatric disorder. Britisl Journal of Psychiatry. 147. 598-611.

Rutter, M. (1987). Psychosocial resilience and protective mechanisms.
American Journal of Orthopsychiatry. 57(3), 316-331.

Rutter, M. L. (1997). Nature-nurture integration: The example of antisocial
behavior. American Psychologist, 52(4), 390-398.

179

Sameroff, A. (2000). Ecological perspectives on developmental risk. In J. D.
Osofsky & H. E. Fitzgerald (Eds), WAIMH Handbook of Infant Mental Health. Vol. 4:
Infant mental health in ggoups at high risk (pp. 4-33). New York, NY: John Wiley &
Sons, Inc.

Sameroff, A. J ., & Seifer, R. (1983). Familial risk and child competence. Child
Development. 54. 1254- 1 268.

Sampson, R. J ., and Laub, J. H. (1993). Crime in the making: Pathways and
turning point§ through life. Cambridge, MA: Harvard University Press.

Sampson, R. J ., and Laub, J. H. (1997). A life-course theory of cumulative
disadvantage and the stability of delinquency. In T. P. Thomberry (Ed), Developmental

Theories of Crime and Delinquency: Advances in Criminological Theogy (Volume 7; pp.
133-161). New Brunswick, NJ: Transaction Publishers, 1997.

Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of
the art. Psycholagical Methods, 7(2), 147-177.

Schuckit, M. A. (1978). Research questionna_ir_e_. San Diego, CA: Alcoholism
Treatment Program, V. A. Medical Center, University of California, San Diego.

Schulenberg, J ., Wadsworth, K. N., O’Malley, P. M., Bachman, J. G., & Johnston,
L. D. (1996). Adolescent risk factors for binge drinking during the transition to young

adulthood: Variable- and pattem-centered approaches to change. Developmental
Psychology. 32(4), 659-674.

Schukit, M. A., & Russell, J. W. (1983). Clinical importance of age at ﬁrst drink
in a group of young men. American Journal of Psychiatry. 140(9), 1221-1222.

Selzer, M. L., Vonokur, A., & van Rooijen, L. (1975). A self administered short
Michigan Alcoholism Screening Test (SMAST). Journal of Studies on Alcohol, E, 117-
126.

Shaw, D. S., Keenan, K., & Vondra, J. I. (1994). Developmental precursors of
externalizing behavior: Ages 1 to 3. Developmegtal Psycholggy. 30(3), 355-364.

Stanger, C., Achenbach, T. M., & Verhulst, F. C. (1997). Accelerated
longitudinal comparisons of aggressive versus delinquent syndromes. Development and
Psychopathology, 9, 43-5 8.

Staudinger, U. M., Marsiske, M., & Baltes, P. B. (1993). Resilience and levels of
reserve capacity in later adulthood: Perspectives from life-span theory. Development

and Psychopathology, 5, 541-566.

180

Strelau, J ., & Angleitner, A. (1991). Temperament research: some divergences
and similarities. In J. Strelau & A. Angleitner (Eds), Explorations in temperament:
International perspectives on theom and measurement (pp. 1-12). New York: Plenum
Press.

Tarter, R. E. (1988). Are there inherited behavioral traits that predispose to
substance abuse? Joumial of Consultingand Clinical Psychology, 56(2), 189-196.

Tarter, R. E., & Vanyukov, M. (1994). Alcoholism: A developmental disorder.
Journal of Consultingand Clinical Psychology, 62(6), 1096-1107.

Tolan, P. H., Gorman-Smith, D., & Henry, D. B. (2003). The developmental
ecology of urban males’ youth violence. Developmental Psychology,3 9(2), 274-291.

Tomlinson, P. S., Harbaugh, B. L., & Anderson, K. H. (1996). Children’s
temperament at 3 months and 6 years old: Stability, reliability, and measurement issues.
Issues in Pediatric Nursing, 19(1), 33-47.

Tremblay, R. E. (2002). Development of physical aggression from early
childhood to adulthood. In R. E. Tremblay, R. G. Barr, R. DeV. Peters (Eds),
Encyclopedia on early childhood development [Online]. Montreal, Quebec: Centre of
Excellence for Early Childhood Development, 1-6. Retrieved March 20, 2004, from
http://www.excellence-earh/childhood.ca/documents/TremblayANpr.pdf.

Tremblay, R. 13., Pihl, R. o., Vitaro, F., & Dobkin, P. L. (1994). Predicting early
onset of male antisocial behavior from preschool behavior. Archives of General
Psychiatry. 51(9), 732-73 9.

VanVoorhis, C. W., & Morgan, B. L. (2001). Statistical rules of thumb: What
we don’t want to forget about sample sizes [Electronic version]. Psi Chi Journal, 6(4).
Retrieved March 20, 2004, from http://www.psichi.org/pubs/articles/article_182.asp.

von Eye, A. (1990). Introduction to conﬁggal frequency analysis: The seath
for types and antitypes in cross-classiﬁcation. New York: Cambridge University Press.

von Eye, A. (2002). Conﬁgural frequency analysis. Mahwah, NJ: Lawrence
Erlbaum Associates, Inc.

von Eye, A., & Schuster, C. (2000). The odds of resilience. Child Development,
11(3), 563-566.

Wachs, T. D. (1996). Known and potential processes underlying developmentaal
trajectories in childhood and adolescence. Developmental Psychology. 32(4), 796-801.

Waddington, C. H. (1942). Canalization of development and the inheritance of
acquired characteristics. Nature 150(3811), 563-565.

 

181

Webb, J. A., & Baer, P. E. (1995). Inﬂuence of family disharmony and parental
alcohol use on adolescent social skills, self-efﬁcacy, and alcohol use. Addictive
Behaviors 20(1), 127-135.

 

Werner, E. E. (1986). The concept of risk from a developmental perspective.
In B. K. Keogh (Ed), Advances in specigal education (pp.1-23). Greenwich, Connecticut:
Jai Press, Inc.

Werner, E. E. (1989). High-risk children in young adulthood: A longitudinal
study from birth to 32 years. Amerllan Journal of Orthopsychiatry. 59(1), 72-81.

Werner, E. E., & Smith, R. S. (1982). VulneraLle but invincible: A longitudinal
study of resilient children and youth. New York: McGraw-Hill Book Company.

West, S. G. (2001). New approaches to missing data in psychological research:
Introduction to the special section. Psychological Methods. 6(4), 315-316.

White, J. L., Mofﬁtt, T. E., Earls, F., Robins, L., & Silva, P. A. (1990). How
early can we tell? Predictors of childhood conduct disorder and adolescent delinquency.

Criminology, 28(4), 507-533.

Wilkinson, G. S. (1993). Administration manual: WRAT3. Wilmington, DE:
Wide Range, Inc.

Wills, T. A., DuHamel, K., & Vaccaro, D. (1995). Activity and mood
temperament as predictors of adolescent substance use: Test of a self-regulation
mediational model. Journal of Personality and Social Psychology, 68(5), 901-916.

Windle, M., & Lerner, R. M. (1986). Reassessing the dimensions of
temperrnental individuality across the life span: the revised Dimensions of Temperament
Survey (DOTS-R). Journal of Adolescent Researcha1(2), 213-230.

Wong, M. M., Zucker, R. A., Puttler, L. I., & Fitzgerald, H. E. (1999).
Heterogeneity of risk aggregation for alcohol problems between early and middle

childhood: Nesting structure variation. Development and Psychopathology, 11(4), 727-
744.

Worobey, J. (1999). Assessment of temperament in infancy. In J. D. Osofsky &
H. E. Fitzgerald (Eds), WAIMH handbook of infant mental health. Volume 2: Eam
intervention. evaluation and assessment (pp. 477-514). New York: John Wiley & Sons,
Inc.

 

Wyman, P. A., Cowen, E. L., Work, W. C., Hoyt-Meyers, L., Magnus, K. B., &
Fagen, D. B. (1999). Caregiving and developmental factors differentiating young at-risk

urban children showing resilient versus stress-affected outcomes: A replication and
extension. Child Development. 70(3), 6445-659.

182

Zucker, R. A. (1987). The four alcoholisms: A developmental account of the
etiologic process. In P. C. Rivers (Ed), Alcohol and addictive behavior (pp. 27-83).
Lincoln: University of Nebraska Press.

Zucker, R. A. (1994). Pathways to alcohol problems and alcoholism: A
developmental account of the evidence for multiple alcoholisms and for contextual
contributions to risk. In R. Zucker, G. Boyd, & J. Howard (Eds), The development of
alcohol problems: Exploring the biopsychosocia_l_pr_atrix of risk (N IAAA Resear_ch
Monograph No. 26) (pp. 255-289, Chapter 13). Rockville, MD: US. Department of
Health and Human Services.

Zucker, R. A., & Barron, F. H. (1973). Parental behaviors associated with
problem drinking and antisocial behavior among adolescent males. In M. E. Chafetz
(Ed), Research on alcoholis_m: I. Clinical problems and special populations. (DHEW
Publication No. 74-675, pp. 276-296). Washington, DC: Government Printing Ofﬁce.

Zucker, R. A., Ellis, D. A., Fitzgerald, H. E., Bingham, C. R., & Sanford, K.
(1996). Other evidence for at least two alcoholisms 11: Life course variation in

antisociality and the heterogeneity of alcoholic outcome. Development and
Psychopathology. 8. 831-848.

Zucker, R. A., Ellis, D. A. & Fitzgerald, H. E. (1994). Developmental evidence
for at least two alcoholisms, I: Biopsychosocial variation among pathways into
symptomatic difﬁculty. Annals of the New York Agdemy of Sciences, 708, 134-146.

Zucker, R. A., & Fillmore, K. M. (1968, September). Motivational factors and
problem drinking arnongadolescents. Paper presented at the 28‘h International Congress
on Alcohol and Alcoholism. Washington, DC.

Zucker, R. A., Fitzgerald, H. E., & Moses, H. D. (1995). Emergence of alcohol
problems and the several alcoholisms: A developmental perspective on etiological theory
and life course trajectory. In D. Cicchetti & D. J. Cohen (Eds), Developmental
psychopathology. Volume 2: Risk. disorder. and adaptation (pp. 677-711). New York:
John Wiley & Sons, Inc.

Zucker, R. A., Fitzgerald, H. E., & Noll, R. B. (1990). Drinking and drug history
(revised edition, ver. 4). Unpublished instrument, Michigan State University, East
Lansing, MI.

Zucker, R. A., Fitzgerald, H. E., Reﬁor, S. K., Puttler, L. I., Pallas, D. M., & Ellis,
D. A. (2000). The clinical and social ecology of childhood for children of alcoholics:
Description of a study and implications for a differentiated social policy. In H. E.
Fitzgerald, B. M. Lester, & B. S. Zuckerrnan (Eds), Children of addiction: Research.

health, and public policy issues (Chapter 6, pp. 109-141). New York: Routledge
Publishers.

183

Zucker, R. A., & Gomberg, E. S. (1986). Etiology of alcoholism reconsidered:
Case for a biopsychosocial process. American Psychologist, 41(7), 783-793.

Zucker, R. A., & Noll, R. B. (1980). The Antisocial Behavior Checkl_iag.
Unpublished instrument, Michigan State University, Department of Psychology, East
Lansing.

Zucker, R. A., Noll, R. B., Ham, H. P., Fitzgerald, H. E., & Sullivan, L. S.
(1994). Assessing antisociality with the Antisocial Behavior Checkl_ist: Regliability an_d
validity studies. East Lansing & Ann Arbor, MI: Michigan State University and
University of Michigan.

Zucker, R. A., Wong, M. M., Puttler, L. I., & Fitzgerald, H. E. (2003).
Resilience and vulnerability among sons of alcoholics: Relationship to developmental
outcomes between early childhood and adolescence. In S. Luthar (Ed), Resilience and
vulnerability: Adaptation in the context of childhood adversities (Chapter 4, pp. 76-103).
New York: Cambridge University Press. New York: Cambridge University Press.

184