Vous êtes sur la page 1sur 27

Development and Psychopathology 17 ~2005!

, 99–125
Copyright © 2005 Cambridge University Press
Printed in the United States of America
DOI: 10.10170S0954579405050066

Inattention0hyperactivity and aggression


from early childhood to adolescence:
Heterogeneity of trajectories and
differential influence of family
environment characteristics

JENNIFER M. JESTER,a JOEL T. NIGG,b KENNETH ADAMS,a


HIRAM E. FITZGERALD,b LEON I. PUTTLER,a MARIA M. WONG,a
and ROBERT A. ZUCKER a
a University of Michigan; and b Michigan State University

Abstract
Inattention0hyperactivity and aggressive behavior problems were measured in 335 children from school entry
throughout adolescence, at 3-year intervals. Children were participants in a high-risk prospective study of substance
use disorders and comorbid problems. A parallel process latent growth model found aggressive behavior decreasing
throughout childhood and adolescence, whereas inattentive0hyperactive behavior levels were constant. Growth
mixture modeling, in which developmental trajectories are statistically classified, found two classes for
inattention0hyperactivity and two for aggressive behavior, resulting in a total of four trajectory classes. Different
influences of the family environment predicted development of the two types of behavior problems when the other
behavior problem was held constant. Lower emotional support and lower intellectual stimulation by the parents in
early childhood predicted membership in the high problem class of inattention0hyperactivity when the trajectory of
aggression was held constant. Conversely, conflict and lack of cohesiveness in the family environment predicted
membership in a worse developmental trajectory of aggressive behavior when the inattention0hyperactivity
trajectories were held constant. The implications of these findings for the development of inattention0hyperactivity
and for the development of risk for the emergence of substance use disorders are discussed.

A cluster of child problems including inatten- cally as attention-deficit0hyperactivity disor-


tion, disorganization, hyperactivity, and im- der ~ADHD; American Psychiatric Associa-
pulsivity present serious dilemmas for our tion @APA#, 2000!. Clinical diagnosis of ADHD
society. These problems tend to co-occur and, is associated with serious impairment in mul-
when sufficiently severe, are diagnosed clini- tiple domains ~Johnston, Murray, Hinshaw, Pel-
ham, & Hoza, 2002! and risk for poor long-
term outcome ~Klein & Mannuzza, 1991!
including development of later substance use
This work was supported by National Institute on Alco-
hol Abuse and Alcoholism grants ~RO1 AA12217 to R.A.Z. disorders ~Zucker & Gomberg, 1986!. Treat-
and J.T.N. and R37 AA07065 to R.A.Z. and H.E.F.!. We ment of attention problems and hyperactivity
are indebted to Bengt Muthén for his advice on the statis- has led to increased rates of medication of chil-
tical analyses for this study. dren in the United States ~Robison, Sclar,
Address correspondence and reprint requests to: Jen-
Skaer, & Galin, 1999! and associated contro-
nifer M. Jester, University of Michigan, Department of
Psychiatry, Addiction Research Center, 400 E. Eisen- versy ~Searight & McLaren, 1998!. As a re-
hower Parkway, Suite 2A, Ann Arbor, MI 48108; E-mail: sult, understanding the antecedents of these
jjester@umich.edu. problems is an important priority.

99
100 J. M. Jester et al.

Yet the determinants of this cluster of child- verse home environments featuring hostility,
hood problems are far from well understood. conflict, and other characteristics ~Ge, Conger,
Twin and adoption studies indicate that heri- Cadoret, Neiderhiser, Yates, Troughton, &
tability of normal and severe levels of these Stewart, 1996!, which interact ~Cadoret, Yates,
behavior problems is substantial ~Loney, Pa- Troughton, Woodworth, & Stewart, 1995!.
ternite, Schwartz, & Roberts, 1997! and neuro- Success in understanding family correlates
psychological correlates are becoming more of aggression, however, has conversely spelled
clearly delineated ~Nigg, 2001!. At the same difficulty in understanding the family environ-
time, it is recognized that epigenetic pro- ment specific to inattention 0hyperactivity.
cesses, including family context effects, must Child aggression tends to take up all the vari-
also be better understood to fully account for ance in family adversity, leaving no variance
the developmental trajectory of these difficul- to be related to child inattention0hyperactivity.
ties ~Whalen & Henker, 1999!. Indeed, even Most family risks associated with ADHD are
heritable processes are likely to exert at least notable primarily when children have comor-
part of their effects through family processes bid conduct problems or aggression ~Johnston
~Scarr & McCartney, 1983!. & Mash, 2001! and studies that associate fam-
ily adversity with ADHD ~e.g., Biederman,
Milberger, Faraone, Kiely, Guite, Mick, Ablon,
Role of the Family Context in Warburton, & Reed, 1995!, often fail to con-
Development of Inattention/ trol for comorbid child aggression or conduct
Hyperactivity Versus Aggression problems. Indeed, some family context vari-
ables that at one time were thought to be re-
One concern in the field has been that insuffi- lated to inattention0hyperactivity now appear
cient energy has been expended to understand to be primarily related to co-occurring child
the family context for development of atten- aggression0conduct problems ~see Frick, 1994,
tion problems and hyperactivity, despite am- for a review!. As a result, few family context
ple theoretical and empirical grounds for effects relate to inattention0hyperactivity spe-
doing so ~Greenberg, Kusche, & Speltz, 1991; cifically. However, family processes have not
Johnston & Mash, 2001!. Perhaps the major been studied sufficiently deeply or broadly in
difficulty in isolating family context factors relation to inattention0hyperactivity to solid-
specific to inattention0hyperactivity is that the ify any conclusions ~Johnston & Mash, 2001!.
latter so often co-occur with aggression and Moreover, most studies have been cross-
other disruptive behaviors ~Frick, 1994!. Work sectional, and few have considered multiple
conducted two decades ago established the family process variables while systematically
partial independence of the related child prob- assessing both inattention0hyperactivity and
lem domains of inattention 0hyperactivity antisocial child behaviors over time ~Johnston
and aggression0conduct problems ~Hinshaw, & Mash, 2001!.
1987!. Substantial progress ensued in under- In fact, a few studies have suggested
standing family processes related to the devel- family context variables specific to child
opment of other disruptive behaviors such as inattention0hyperactivity. For example, par-
aggression, conduct problems, and opposi- ent attention problems appear to be specific to
tional behavior. These include parent hostility, child attention problems apart from child
lack of monitoring, and coercive interchange aggression or conduct problems ~Frick, Kam-
with children in the development of aggres- phaus, Lahey, Loeber, Christ, Hart, & Tannen-
sion ~Patterson, 1982! along with low paren- baum, 1991!. Parenting problems also may be
tal competence and inconsistent discipline in related distinctly to attention problems in ad-
oppositional behaviors ~Kendziora & O’Leary, dition to their link to aggression ~Loeber,
1993!. These effects on the development of Brinthaupt, & Green, 1990; Nigg & Hinshaw,
antisocial behavior and aggression are medi- 1998!.
ated by the joint contribution of shared genes Further, we should note that a substantial
~vulnerability to antisocial behavior! and ad- literature addresses the development of atten-
Trajectories of inattention0hyperactivity and aggression 101

tion and self-regulation per se, without regard ies have found that parenting behaviors im-
to family correlates. For example, compo- prove when disruptive children are medicated
nents of attentional operations as described in forADHD ~Barkley & Cunningham, 1979!. Fur-
the cognitive literature mature at varying rates thermore, family correlates can reflect both
throughout childhood and into adolescence ~for heritable and experiential processes, so that con-
review and recent data, see Huang–Pollock, clusions must be drawn carefully from family
Carr, & Nigg, 2002! and a substantial devel- data with regard to specific causal mechanisms.
opmental literature considers the broader con- Nonetheless, understanding of family corre-
struct of self-regulation from the point of view lates is a key kind of data needed to understand
of temperament and relates those constructs to the processes by which child inattention and hy-
behavioral problems ~Eisenberg, Cumber- peractivity develop and are maintained over
land, Spinrad, Fabes, Shepard, Reiser, Mur- time. By measuring home and family environ-
phy, Losoya, & Guthrie, 2001; Kochanska, ment factors at a time prior to the measurement
Coy, & Murray, 2001; Kopp, 1982; see Nigg, of child behavior problems, we reduce the risk
2000, for a general overview!. Thus, the field that the relationship we study is actually caused
is scarcely devoid of understanding about de- by the effect of the child behavior on the home
velopmental processes that are related in a and family environment.
broad sense to child inattention and hyper-
activity0impulsivity. Nonetheless, develop-
Advantages of a Longitudinal,
mental relations of family context measures to
High-Risk Sampling Strategy
psychopathological symptoms of inattention0
hyperactivity, as distinct from antisocial be- To study the problem of how family context
haviors, remain sorely in need of further might relate to development of child behav-
clarification in long-term, longitudinal studies. ioral problems with inattention and activity, it
Whereas one could nominate several fam- is essential to have longitudinal data that per-
ily domains for further study ~Frick, 1994!, mit assessment of the development of behav-
we elected to emphasize characteristics of the ior problems over time. Doing so requires that
early family and home environment. Several children be identified before being selected
reviewers and theorists ~Greenberg et al., 1991; for clinical diagnoses. In the current study, we
Winsler, 1998! have proposed a relation be- followed a “high-risk” strategy, identifying
tween responsive parenting and the develop- children who would be at elevated risk for both
ment of self-regulation in children. Further, aggression0conduct problems and inattention
parenting style, such as captured by the and hyperactivity0impulsivity problems, due
authoritative0authoritarian dichotomy ~Baum- to parental alcoholism. By not selecting spe-
rind, 1968!, may be more strongly related to cifically for risk for ADHD or aggression, we
child behavior than more discrete measures increase the generality of results with regard
of parenting ~Darling & Steinberg, 1993!. to the nature of family correlates with each
Thus, one might expect that parenting style or behavioral domain. Moreover, the strategy al-
responsiveness of the parent would relate lows us to study a sample of children with a
specifically to development of child inatten- full range of inattention0hyperactivity during
tion0hyperactivity. Conceptually, some family development, not just children with diagnos-
characteristics may be specifically associated able symptom levels. Doing so is consistent
with development of aggression ~e.g., family with a dimensional rather than a categorical
conflict, coercive exchange, cohesiveness, formulation of etiology of these problems ~Fer-
monitoring!, while others are specific to the gusson & Horwood, 1995!.
development of inattention0hyperactivity ~e.g.,
parental responsiveness and stimulation!, even
Children of Alcoholics
though some covariation of those contexts and
associated problems is expected. In selecting a high-risk sample, we chose a
In cross-sectional studies, the direction of sample of children of alcoholics; we briefly
effects also needs be considered. Several stud- note some established risk characteristics in
102 J. M. Jester et al.

this population. Although the relation of fam- hypothesized that the distinct trajectory groups
ily context and child behavior problems has describing the development of inattention0
often been studied in normal populations, the hyperactivity and aggression could be iden-
major questions driving this research origi- tified over this 9-year period, and that
nated in the observation that problems of both membership in a particular trajectory group
hyperactivity and aggressiveness were over- would be anticipated by family context factors
represented among children of alcoholics measured in the preschool period.
~Zucker & Gomberg, 1986!. Early studies This study extends the work of Loukas and
found greater prevalence of alcoholism in bio- colleagues from our group ~Loukas, Zucker,
logical mothers and fathers of adopted hyper- Fitzgerald, & Krull, 2003!. The latter used
active children, but not the legal parents hierarchical linear modeling to examine the
~Cantwell, 1975!, indicating a substantial bio- effects of parent psychopathology, family con-
logical component mediating this relation- flict, and behavioral undercontrol on the de-
ship. Retrospective reports of hyperactivity in velopment of children’s disruptive behavior.
alcoholic men predicted earlier onset of alco- Their outcome of interest was disruptive be-
holism and more family history of alcohol havior, as measured by the aggression sub-
problems ~Tarter, McBride, Buonpane, & scale of the Child Behavior Checklist ~CBCL;
Schneider, 1977!. More recently, Kuperman, Achenbach, 1991a! from ages 3 to 12. Child
Schlosser, Lidral, and Reich ~1999! found that undercontrol ~measured via the items: inatten-
the children with alcoholic parents had a higher tive; demands must be met immediately, eas-
likelihood of having ADHD after controlling ily frustrated; gets overexcited easily; easily
for family factors such as low income and bored by a repetitive activity; fails to finish
child–parent conflict. The family environ- things he started! was controlled as a child
ment of children of alcoholics may be an temperament variable at one time point, the
important factor in the development of in- initial contact with the children. Hence, Lou-
attention0hyperactivity. For instance, Reich, kas et al. ~2003! treated undercontrol as a co-
Earls, and Powell ~1988! found that the home variate of aggression problems and measured
environment of children of alcoholics was char- undercontrol only at one point in time. In the
acterized by increased marital conflict and current study, the outcomes of interest were
parent–child conflict. However, marital con- the developmental trajectories of inattention0
flict is also associated with child antisocial hyperactivity, with the trajectory of aggres-
behaviors ~Grych & Fincham, 2001! and spec- sive behavior examined so as to isolate effects
ificity to ADHD symptoms again remains un- of child-rearing measures on inattention 0
clear because of mixed findings across studies hyperactivity. In the present study, however,
~Johnston & Mash, 2001!. inattention0hyperactivity was measured across
the full time course of the longitudinal study
from childhood to adolescence. Furthermore,
The Current Study
because we also measured aggression longi-
The aims of our study were to establish the tudinally and a parallel process model was
developmental trajectories of inattention 0 developed, the current study allows for inves-
hyperactivity behavior problems, and to ex- tigation of the effects of home and family
amine these trajectories in relation to family environments on each of these trajectories in-
context, over a 9-year time span from school dependent of the other. These effects were not
entry to adolescence. To avoid the problem of examined by Loukas et al. ~2003!.
confounding of inattention0hyperactivity and
aggression, we examined the development of
Method
aggression in parallel with inattention0hyper-
activity. Then we could hold constant the de-
Participants
velopment of aggression and compare family
context effects in children who only differed The present work is part of an ongoing multi-
in inattention 0hyperactivity trajectories. We wave prospective study ~Zucker, Fitzgerald,
Trajectories of inattention0hyperactivity and aggression 103

Refior, Puttler, Pallas, & Ellis, 2000! that is there were two or more waves of teacher- and
following a community sample of families parent-report data.
with high levels of alcohol use disorders and Successful follow-up procedures in this
other drug use especially among the fathers, study have limited the number of dropouts and
along with a community contrast sample of missed waves. However, because of complex-
families that were drawn from the same neigh- ities in the study design, not all children had
borhoods but that did not have the high sub- data available for all waves. Most of the un-
stance abuse profile. Substance abusing men available data were attributable to children who
were initially identified through the courts were recruited into the study at a time follow-
when they were convicted of drunk driving ing the initial startup ~e.g., the sisters of the
with a high blood alcohol level ~at least 0.15%, original target boys were added to the study
or at least 0.12% and a previous drinking- after the initial assessment had been made!.
related legal problem!. To be included in the Some of these children were added to the study
study, the fathers were required ~a! to have a when they were already past the age of 8, so
Feighner diagnosis for probable or definite they missed the Wave 2 ~age 6–8! assessment
alcoholism ~Feighner, Robins, Guze, Woo- entirely. Other children were not followed at
druff, Winokur, & Munoz, 1972!, ~b! to have Wave 2 due to funding limitations. With the
at least one son between 3 and 5 years of exception of these “missing by design” chil-
age, and ~c! to be living with both the child dren, very few children are missing wave data.
and his biological mother at the time of re- At Wave 2, only 9 children ~2.7%! were miss-
cruitment. Later funding allowed the inclu- ing data for other reasons; at Wave 3 ~age
sion of female siblings from the same families. 9–11!, 13 children ~3.9%! were missing data
A contrast0control group of families who re- for other reasons. At Wave 4 ~age 12–14!, 22
sided in the same neighborhoods as the alco- children ~6.6%! and Wave 5 ~age 15–18!, 6
holic families but with no substance abuse children ~1.8%! were missing data for other
history was also recruited using door to door reasons. Similarly, extensive measures were
canvassing. In addition, an intermediate risk taken to collect all of the teacher report data
group was provided by recruiting all families that was available, but not all of the annual
with an alcohol abuse0dependence diagnosis and wave data were available from all the chil-
who were found during the community can- dren. For instance, the annual collection was
vass. For a more detailed description see started in 1998. At that time, some of the chil-
Zucker et al. ~2000!. dren were not yet 11, and some of the target
The child and both parents were assessed children who were assessed earliest in the study
extensively in their home following the initial were already out of school. High success rates
recruitment ~Wave 1, child age 3–5! with as- of collection of teacher report data for the chil-
sessment repeated every 3 years for a total of dren of the appropriate age can be seen in these
five assessment waves. The information ob- example rates: in school year 2000–2001, 435
tained from the parents included, among other Teacher Report Forms were sent out and 400
issues, substance use, mental health, child rear- returned ~92%!; in 2001–2002, 385 were sent
ing practices and ratings of children’s behav- out and 346 returned ~90%!; and in 2002–
ior problems. In addition, at each wave of data 2003, 369 were sent out and 335 returned
collection following school entry, teachers rated ~91%!.
children’s behavior problems. Ratings were The 335 children with two or more waves
completed in elementary school by the regular of data were in 248 families: 85 of the fami-
teacher and in middle school and high school lies included 2 participating siblings, 1 family
by the English teacher or another teacher who had 3 participating children, and the remain-
had significant contact with the child. Begin- ing 162 families had only 1. Of the children in
ning in 1998, annual collection of teacher re- this sample, 238 ~71%! are boys, because girls,
ports was begun for children 11 years and older, who were all siblings of the initial boys in the
and continued until the child left school. The sample, were included in the study well after
present analysis included all children for whom the first assessments. The total sample sizes
104 J. M. Jester et al.

for parent reports were 256 for ages 6–8, 299 if available. These forms are a widely used
for ages 9–11, 295 for ages 12–14, and 205 and recognized empirical measure of child be-
for ages 15–17; and for teacher report data havior problems, with excellent reliability and
there were 151 for ages 6–8, 251 for ages 9–11, validity ~Achenbach, 1991a, 1991b!. Items are
289 for ages 12–14, and 231 for ages 15–17. rated on a 3-point scale: not true, somewhat or
sometimes true, or very true or often true. The
Attention Problems subscale includes 20
Measures
items such as “can’t concentrate, can’t pay at-
DSM-IV ~APA, 2000! alcohol abuse0depen- tention for long,” “can’t sit still, restless, or
dence diagnoses in the parents were assessed hyperactive,” and “poor school work.” It thus
with the Diagnostic Interview Schedule ~DIS- includes features of both inattention and over-
III; Robins, Helzer, Croughan, & Ratcliff, activity, the two core domains associated with
1980! as well information provided from the ADHD as defined in DSM-IV ~APA, 2000! and
Drinking and Drug History ~Zucker, Fitzger- high scores have relatively good convergence
ald, & Noll, 1990!, at the first wave of data with ADHD diagnoses made from structured
collection. Only an alcohol diagnosis in adult- interviews ~Chen, Faraone, Biederman, & Tsu-
hood ~over age 20! was considered, because ang, 1994!. The Aggression subscale contains
remitted adolescent alcoholism would be less 20 items, including “gets in many fights,” “ar-
likely to affect children via the family envi- gues a lot,” “physically attacks people,” and
ronment. For the purposes of analysis, either “temper tantrums or hot temper.” It thus re-
parent’s diagnosis was sufficient for a family flects antisocial behaviors and aggressive be-
classification of “alcoholic”; however, in 25% haviors. For each of the problem behavior
of the families, both parents met criteria for a subscales, the range of possible scores was
diagnosis of alcoholism. Overall, there were 0– 40, with 0 being no problems reported and
225 children in families that met the defini- 40 being often true for all 20 items. Actual
tion of family alcoholism: 129 children in fam- scores ranged from 0 to 37 for inattention0
ilies with only the father meeting alcohol hyperactivity and 0 to 28 for aggression.
diagnosis, 93 children in families in which Given the known variations in child behav-
mother and father were lifetime alcoholics, and ior as a function of setting ~Nesselroade, 1992!,
3 children in families with only the mother the choice of whether to use parent or teacher
having lifetime alcohol diagnosis; there were ratings ~or both! is a conceptual as well as a
110 children in families with no parent having practical issue. This is especially the case for
lifetime alcohol diagnosis. behaviors with a socially undesirable compo-
Socioeconomic status ~SES! was evaluated nent, such as aggressiveness, for which damp-
based on occupational prestige, using the Dun- ening is more likely to occur in public ~e.g.,
can TSEI2 Socioeconomic Index ~Stevens & school! than in private ~e.g., home! settings.
Featherman, 1981!. Family income was esti- The problem is also evident in circumstances
mated by taking the average of each parent’s when the social environment is less evocative
report of family annual income, adjusted to of the behavior being evaluated. Thus, for
1998 dollars. inattention0hyperactivity, the school setting
Child behavior was rated by parents on the in contrast to the home should be more evoc-
Attention Problems subscale of the CBCL ative of difficulties because of the nature of
~Achenbach, 1991a! and by teachers on the school task demands. Indeed, in our data, we
Teacher Report Form ~Achenbach, 1991b! found that teachers rated the children as hav-
at each assessment period. Both parents re- ing more inattention0hyperactivity than the
sponded and their reports were averaged ~in- parents, whereas parents rated the children as
tercorrelations between mother and father having more aggressive behaviors. Previous
report at each time point ranged from 0.40 to research has also found that the teacher re-
0.53 across the four waves of data!. Data from ports of inattention0hyperactivity yielded more
biological parents was used when available; discrimination of these problems than did par-
otherwise, we used ratings from step-parents ent reports ~Loeber et al., 1990!. Second, we
Trajectories of inattention0hyperactivity and aggression 105

hoped to evade negative halo effects, in which four scales therefore were omitted herein. To
a child with one type of problem is rated in- isolate the factors relevant to our study, we
accurately as having additional problems ~e.g., conducted a factor analysis of the remaining
Stevens, Quittner, & Abikoff, 1998!. For these 12 scores ~mother and father report on the six
reasons, we selected different informants for other FES subscales!, resulting in a four-
measurement of the dependent variables. We factor solution. The first factor consisted of
used parent ratings to assess aggression, and mother’s and father’s reports of Cohesion, Or-
teacher ratings to assess inattention0hyper- ganization, and Conflict ~reverse coded!, which
activity. We note later that this choice did not accounted for 27% of the variance, and was
yield different results than the more concep- the factor of substantive interest to our study
tually questionable alternative of parent– due to our interest in family stability and co-
teacher composites. hesiveness. ~The other three factors, of less
The child-rearing environment was evalu- interest and so omitted from our subsequent
ated in three ways. First, the preschool ver- analysis, were moral–religious emphases,
sion of the Home Observation for Measurement active–recreation orientation, and intellectual–
of the Environment ~HOME; Bradley, Cald- cultural orientation!. This factor analysis re-
well, Rock, Hamrick, & Harris, 1988! was used sult parallels the previously reported factor
to measure intellectual stimulation and social structure for this instrument ~Gondoli & Ja-
and emotional support provided by parents as cob, 1993!. The three scales loading on the
well as the physical environment. A trained first factor were thus averaged ~across mother
interviewer conducted an interview of the and father report! to form a single scale that
mother in the child’s home during the first wave we called FES Cohesiveness ~a 5 .76!.
of assessment ~child between ages 3 and 5! Third, information on parenting styles was
and completed the HOME inventory based on obtained at Wave 1 via maternal ratings on the
the interview as well as observation of the Child Rearing Practices Report ~Block, 1981!.
home. The HOME scale has been used exten- This is a Q-sort instrument in which the mother
sively and relates well to child developmental ranks various aspects of her parenting into seven
outcomes, including IQ and externalizing be- categories from most important to least impor-
havior ~Duncan, Brooks–Gunn, & Klebanov, tant. Authoritarian and Authoritative parenting
1994; Luster & McAdoo, 1994!. The alpha scales were derived from this instrument fol-
reliability of the scale total score was .68. lowing Kochanska, Kuczynski, and Radke–
Second, the Family Environment Scale Yarrow ~1989!. Mothers with an authoritarian
~FES; Moos & Moos, 1994! was completed style of parenting seek to control the child
by both parents at the first wave of data through physical punishment and fear, with a
collection. The FES is comprised of 10 sub- great deal of supervision of the child. Moth-
scales, grouped into three dimensions: Re- ers using authoritative styles of parenting are
lationship ~Cohesion, Expressiveness, and more likely to discipline with reasoning and
Conflict!, Personal Growth ~Independence, praise, to encourage the child’s independence
Achievement Orientation, Intellectual–Cultural and to provide a warmer environment.
Orientation, Active–Recreational Orientation, Symptoms of ADHD ~inattention0hyper-
and Moral–Religious Emphasis!, and System activity! in the parents are a possible con-
Maintenance ~Organization and Control!. Re- founding factor in linking child inattention0
lation of FES subscales to child outcomes has hyperactivity with family environment or
been shown, for example, by the finding that parenting. Parents with inattention 0hyper-
alcoholic families who scored higher on Co- activity are more likely to have children with
hesion and Organization had children with inattention0hyperactivity ~Frick, Kamphaus,
greater self-esteem and less anxiety ~Moos & et al., 1991! and may also provide a less opti-
Moos, 1994!. Sanford, Bingham, and Zucker mal parenting and family environment. Parent
~1999! found poor psychometric properties on ratings of their own childhood symptoms of
four of the FES scales: Expressiveness, Con- inattention and hyperactivity were obtained by
trol, Independence, and Achievement. Those interview using the DIS.
106 J. M. Jester et al.

Parents were asked to report if their child scores. Then, the analyses were repeated with
was on a medication program for hyperactiv- the HOME score imputed for the children miss-
ity as well as if they were on medication for ing the HOME. Imputation in this case was
any other chronic problem. Between the ages accomplished using the estimation maximiza-
of 6 and 17, 20 children ~6.1%! were reported tion algorithm, which is a type of maximum
to be taking medication for hyperactivity or likelihood method for data imputation. These
were taking a drug that may have been pre- methods are considered a superior option ver-
scribed for inattention, impulsivity, or hyper- sus either listwise deletion or mean or regres-
activity. The analysis was repeated excluding sion based estimation ~Schafer & Graham,
these children to check effects of medication 2002!. Further detail on these analyses is pro-
usage on the analysis. vided later.

Handling siblings and nonindependence. Be-


Data analysis
cause multiple children from the same family
Handling missing data. The Mplus program ~up to two in most cases, with three siblings
was used to handle missing data in the models from only one family! were included in the
~Muthén & Muthén, 1998!. The missing data study, the assumption of independence was vi-
method used by this program estimates the pa- olated. To estimate the effect of this noninde-
rameters with a full information maximum like- pendence, the design effect was estimated as
lihood estimator using all observations in the ~Muthén, 2000!
data set. In mixture models, the Mplus pro-
gram allows missing data for the dependent d 5 1 1 r~c 2 1!,
variables ~in this analysis, child behaviors!,
but not the predictors ~in this analysis, the where r is the intraclass correlation and c is
child-rearing0family environment variables!. the common cluster size, a measure of the av-
To use as much of the data as possible, we erage number of units per cluster ~in our case
used the entire sample of 335 for the latent this is the average number of children per fam-
growth model and the unconditional growth ily!. Generally, a design effect of ,2.0 is con-
mixture models ~i.e., models without the child- sidered acceptable and small enough to be
rearing predictor variables in them!. When pre- ignored based on simulation data ~Muthén &
dictors were included in the subsequent models Satorra, 1995!. For the 335 children in 248
~i.e., HOME, Child-Rearing Practices, and FES families, the common cluster size in our data-
Cohesiveness!, missing data for them was han- set was 1.35, indicating that the maximum pos-
dled as follows. First, the growth mixture mod- sible design effect was also 1.35. Because this
els with predictors were estimated for the was within the range considered small enough
smaller sample that had no missing predic- to ignore, we proceeded to fit models with all
tors. This allowed us to check whether there children. As a simple check on the effect of
were changes in predicted class membership the violation of independence, models were
and growth parameters between the full sam- also run with boys only, because all the boys
ple and the sample limited by the predictor in the sample are from different families and
variables. Because there were no major differ- all the girls are siblings of these boys. The
ences between the limited data set models and basic results held, as detailed later.
the full models, and because we judged the
amount missing small enough to ignore, we Latent growth modeling. Latent growth mod-
accepted this listwise deletion for the FES and eling is a statistical procedure that develops
Child-Rearing Practices analyses. The predic- latent variables to describe different facets of
tor with the most missing data was the HOME, the trajectories. In this case, we used a linear
for which 113 ~33.7%! of the children did not model, which consists of slope and intercept.
have data. To address this, for analyses with Furthermore, inattention0hyperactivity and ag-
HOME as a predictor, models were first run in gression were modeled as a parallel growth
Mplus with the 222 children who had HOME process, as shown in Figure 1. This type of
107

Figure 1. The parallel process model of inattention0hyperactivity and aggression. The bold arrows that have a number above the arrow represent model parameters that
are fixed to the number indicated; the light arrows represent model parameters that are estimated. Residual variances ~not shown! were estimated for each measured
variable ~in rectangles!.
108 J. M. Jester et al.

analysis is analogous to examining aggression be equal. For instance, growth parameters of


as a covariate of inattention0hyperactivity, aggression were held equal across classes of
while taking into account the developmental inattention0hyperactivity to examine effects
nature of the aggressive behavior as well as of of child-rearing factors on inattention0hyper-
the inattention0hyperactivity ~Muthén, 2001!. activity trajectories. Figure 2 shows the sche-
The four latent variables, shown in ovals, de- matic for this model. Each behavior has two
scribe the trajectories of the two types of be- growth parameters as in Figure 1. Now for
havior problems. The indicators of these latent each behavior there is also a latent categorical
variables are the measured aggression and variable called the trajectory class indicator.
inattention0hyperactivity at each time point, These are the latent variables that define class
shown in rectangles. membership. For these analyses, the model
chosen as best fitting, based on substantive
Primary analyses of latent classes. and statistical considerations detailed in Re-
Latent trajectory class analysis was performed sults, was one in which attention and aggres-
using Mplus software ~Muthén & Muthén, sion problem trajectories were each modeled
1998!. This type of analysis, also called growth with two classes so the trajectory class indica-
mixture modeling, allows one to find underly- tors would each have two possible values. This
ing heterogeneity of trajectories over several resulted in a 2 3 2 or four-class solution overall.
time points in a sample. The method builds After this unconditional four-class model
upon latent growth modeling, in which a set had been developed, it was used as a basis to
of growth parameters is found for an entire evaluate our hypotheses about the relative con-
sample. Growth mixture modeling extends this tribution of various early predictors to the de-
analysis by creating a categorical latent vari- velopment of behavior problems. This was
able, which allows the trajectories in the sam- done independently for each of our three early
ple to be defined by multiple sets of growth predictors ~the HOME total score, the FES Co-
parameters. For instance, one class may have hesiveness Scale, and the Child Rearing Prac-
a high intercept and a positive slope ~indicat- tices authoritarian and authoritative scales!.
ing a high level of initial problem behaviors Each model was the same as the uncondi-
that is worsening over time!, whereas a differ- tional model shown in Figure 2, with the ad-
ent class could have a lower intercept and a neg- dition of child-rearing variables as predictors
ative slope ~lower level of initial problems, of the trajectory class indicators.
improving over time!. Given a certain number
of classes, the estimation of the model involves
Results
determining ~a! the sets of growth parameters
that best define those classes, and ~b! how many
Preliminary overview of the data
and which individuals are best fit into each class.
To arrive at the optimal number of classes to Despite careful matching, alcoholic families
describe the data, models are estimated for an had lower income and SES than nonalcoholic
increasing number of classes in succession. The families. The mean income for alcoholic fam-
best number of classes is determined by exam- ilies was $44,400 ~SD 5 21,500!, whereas that
ination of substantive and statistical issues, de- for nonalcoholic families was $54,000 ~SD 5
scribed later with our results. 19,700, t 5 4.1, p , .001, Satterthwaite test,
In these analyses, the trajectories of parallel not assuming equal variances!. For alcoholic
growth of inattention0hyperactivity and aggres- families mean SES was 320.0 ~SD 5 115! and
sion were examined for fit into latent classes. A for nonalcoholic families mean SES was 385.0
limited set of trajectory shapes was allowed for ~SD 5 145, t 5 4.1, p , .001!.
each behavior problem, so that variations in de- Table 1 presents zero-order correlations be-
velopment of one type of behavior problem tween family environmental factors and child
could be examined in cases where the other type inattention0hyperactivity and aggression, for
of behavior problem was essentially “held con- a first-pass descriptive examination of the re-
stant” by requiring the growth parameters to lationships between the variables in this analy-
Figure 2. The latent trajectory class analysis model of parallel development of inattention0hyperactivity and aggression.
109

Table 1. Correlations between family and parenting factors and child attention and aggression problem scores

Intellectual
Stimulation0 Parenting
Inattention0 Emotional Parental Cohesiveness
M SEM Hyperactivity a Aggression a Support Alcoholism Factor Authoritarian Authoritative

Inattention0hyperactivity a 7.2 0.37 — .37*** 2.42*** .27*** 2.15** .10 2.03


Aggression a 8.1 0.21 — 2.24*** .16** 2.26*** .08 .03
Intellectual stimulation0emotional support 44.9 0.33 — 2.19** .19** 2.10 .19**
Parental alcoholism .67 0.03 — 2.15** 2.04 .18**
FES Cohesiveness factor 51.8 0.19 — .07 .12
Parenting
Authoritarian 10.9 0.11 — 2.30***
Authoritative 16.6 0.08 —

a Inattention0hyperactivity and aggression were averaged across the course of the study to form a summary score.

Two-tailed significance: *p , .05. **p , .01. ***p , .001.


110 J. M. Jester et al.

sis. The summary problem behavior scores are the schematic represents a parameter in the
averages across all time points throughout the model. For instance, the arrow connecting “Ag-
study for each problem behavior. As expected, gression Intercept” and “Aggression Slope”
behavior problems in the two domains were represents the model parameter of the covari-
correlated. Higher level of intellectual stimu- ance of these two parameters. The growth pa-
lation and emotional support ~HOME! and rameters are defined by setting the loadings of
higher scores on the FES Cohesiveness factor the values measured at each age to 1.0 ~indi-
were associated with lower levels of both cated by the number above the arrow! for the
inattention0hyperactivity and aggression in intercept latent variable and proportional to
children. There was no correlation between par- the time elapsed for the slope variable ~e.g.,
enting style and child behavior problems. As the loading of slope on Aggression Age 9–11
expected, parental alcoholism was associated is 3, because there were three years from the
with more inattention 0hyperactivity and start of the study to this point!. Error vari-
aggression in children. For the most part, al- ances for each indicator were estimated, but
coholic families presented a lower level of are not shown in the figure.
functioning of the home environment as well. This initial latent growth model estimated
They scored lower on both the HOME and the for all children ~again, no classes estimated at
FES Cohesiveness factor. The Child Rearing this stage! fit well. The chi-square was not
Practices Authoritarian factor did not differ be- significant, x 2 ~22! 5 28.4, p 5 .16, as is de-
tween alcoholic and nonalcoholic families, al- sired in a well-fitting model. Other fit indices
though the Authoritative factor was higher in were well within the acceptable range. The
the alcoholic families ~ p , .01!. comparative fit index and Tucker–Lewis in-
At each data collection wave, some chil- dex, for which a value of .95 or higher is de-
dren scored in the clinical range for inattention0 sirable, were .99. The root mean square error
hyperactivity. We defined the clinical range of of approximation, for which desirable values
the data as the score of the 98th percentile of are less than .06, was .029. This model pro-
the sample control group ~Drotar, Stein, & Per- duces estimates of inattention0hyperactivity
rin, 1995!. The number of children in the clin- and aggression trajectories for the entire sam-
ical range was 18 ~8.3%! for Wave 2, 14 ~5.6%! ple, defined by mean and variance of the
for Wave 3, 17 ~6.9%! for Wave 4, and 10 growth parameters and shown in Figure 3 ~bold
~6.9%! for Wave 5. lines!. For inattention0hyperactivity, the mean
intercept was 6.8, and mean slope was 0.04
~ns!. The mean aggression intercept was 8.6
Preliminary parallel process latent growth
and mean slope of aggression was 20.30. Con-
model of inattention0hyperactivity
sistent with our expectation that this was not a
and aggression
problem-free risk sample, the intercept terms
As explained earlier, the development of for inattention0hyperactivity and aggression
inattention0hyperactivity and aggression in the were significantly different from zero. The
children was modeled with a parallel process slope of aggression problems was signifi-
latent growth model, in which two latent cantly negative, indicating improvement over
growth models are developed simultaneously time. These intercepts are given in the units of
and the growth factors for each are allowed to the original raw score on these scales, and thus
covary. Recall that this initial model did not reflect relatively low to moderate average lev-
try to sort children into classes, but simply els of behavior problems in the sample as a
described the growth parameters in the entire whole. A score of 6 could indicate, for exam-
sample. We used a linear model, which esti- ple, a rating of very often true ~2! for three
mates a trajectory for each child, identified items or a sometimes true ~1! for six items.
with two growth parameters: an intercept and Figure 3 shows the results of this parallel pro-
a slope for each type of problem behavior cess latent growth model, with the bold lines
~McArdle & Epstein, 1987!. Figure 1 shows showing the mean trajectories and the light
the schematic for this model. Each arrow in lines each representing an individual’s devel-
Trajectories of inattention0hyperactivity and aggression 111

Figure 3. The trajectories of inattention0hyperactivity and aggression development for the latent growth
model. The heavy lines show mean trajectories of inattention0hyperactivity and aggression, and each
thin line represents the trajectory of an individual.

opment of behavior problems. There was differed in one type of behavior problem. Ac-
significant variance in all of the growth pa- cordingly, in the four-class model, two classes
rameters, indicating enough variability in had equal trajectories of aggression devel-
growth parameters between subjects that it was opment and two had equal trajectories of
reasonable to conduct further analyses to ex- inattention0hyperactivity development. This
plain more variability. We next attempted such was accomplished by setting the parameters
explanation using latent trajectory class analy- for the trajectories equal in the classes for
sis, in effect assuming multiple populations of which equivalent trajectories were desired.
children with different trajectories of behav-
ior problems. Choosing the number of classes. Choosing the
number of classes is based on several criteria,
including statistical considerations, such as the
Latent trajectory class analysis models Vuong–Lo–Mendell–Rubin likelihood ratio
test ~VLMR LRT!, which provides a statisti-
Latent trajectory class analysis is a technique cal test of the fit of a model compared to a
that seeks to explain some of the variance in model with one less class. Another criterion is
the trajectories in a sample by assuming that the robustness of the model to find the same
the sample is drawn from different popula- solution with different starting values. Inspec-
tions, each having a different set of growth tion of the trajectories of the different classes
parameters. In addition, in the current model- provides a substantive evaluation of the neces-
ing we sought to identify classes that only sity of using more or fewer classes.
112 J. M. Jester et al.

Table 2. Latent trajectory class analysis of parallel process model for


inattention0hyperactivity and aggression

Aggression
Inattention0Hyperactivity
Problem Problem
Class Trajectory Intercept Slope Trajectory Intercept Slope n %
Comorbid High 11.0* 20.04 High 12.4* 20.17 82 24
Aggressive Low 1.2* 0.08 High 12.4* 20.17 12 4
Inattentive0hyperactive High 11.0* 20.04 Low 6.97* 20.39* 110 33
Healthy Low 1.2* 0.08 Low 6.97* 20.39* 131 39

Note: Inattention0hyperactivity growth parameters were fixed equal in the Comorbid and Inattentive0Hyperactive
classes ~high! and in the Aggressive and Healthy classes ~low!. Aggression problem growth parameters were fixed
equal in the Comorbid and Aggressive classes ~high! and in the Inattentive0Hyperactive and Healthy classes ~low!.
*Significantly different from zero, p , .05.

Each individual in the model is assigned a of this study was the developmental nature of
probability of membership in each of the inattention0hyperactivity controlling for ag-
classes and can be said to be “in” the class for gression, as a further check we also tested a
which he or she is assigned the highest prob- model in which three classes of trajectories
ability of membership. Classification quality were allowed for inattention0hyperactivity ~re-
is determined by averaging the probability of sulting in a 3 3 2 or six-class model overall!.
membership in each class for all the individu- However, this six-class alternative model was
als. Values that are close to 1 for the individ- not able to converge on a solution, even when
uals in the assigned class and close to 0 for multiple starting values were attempted. We
individuals in the other classes are desired. therefore accepted the four-class solution as
For instance a value close to 1 for the Class 1 the best fit to these longitudinal data.
element for the class one individuals indicates These four classes were quite interpret-
that all of the individuals in Class 1 have a able. We labeled the classes as follows: ~a! the
high probability of being in Class 1 and a low Comorbid class had a high intercept and level
probability of being in any other class. slope on both inattention0hyperactivity and ag-
When we followed the procedure of increas- gression, ~b! the Inattentive class had a high
ing the number of classes in successive mod- intercept on inattention0hyperactivity but a low
els and examining the various relative fit intercept on aggression, ~c! the Aggressive class
indices, the four-class model was clearly cho- was high on aggression but not inattention0
sen by the statistical criteria over a three-class hyperactivity, and ~d! the Healthy class exhib-
model, with a significant VLMR LRT ~ p 5 ited the lowest level of both aggression and
.03!, indicating a significantly better fit for inattention0hyperactivity. To isolate effects of
the four-class than the three-class model. The inattention0hyperactivity and aggression, in the
substantive difference between trajectories for two high attention classes ~Comorbid and In-
both inattention0hyperactivity and aggression attentive! the inattention0hyperactivity trajec-
was similarly clear for choosing two classes tories ~parameters of intercept and slope! were
over one class for each of these behavior prob- fixed to be equal ~compare mean trajectories
lems. The four-class model was also robust for inattention 0hyperactivity in these two
with respect to different start values for the classes in Figure 4!. For the classes with high
intercepts and slope of the problem behavior inattention0hyperactivity trajectories, Inatten-
trajectories. The classification quality was ac- tive and Comorbid, one set of parameters de-
ceptable, with average class probabilities of scribed the growth of inattention0hyperactivity
Class 1 5 .89, Class 2 5 .84, Class 3 5 .84, of the two classes, forcing them to be equal.
and Class 4 5.93. Because the primary focus A different set of parameters was used for
Trajectories of inattention0hyperactivity and aggression 113

inattention0hyperactivity trajectories of the two predictors of class membership. Parental alco-


low inattention0hyperactivity classes, the Ag- holism and the three measures of the child-
gressive and Healthy classes. Similarly, one rearing environment were used to predict
set of parameters was used for aggression in trajectory class membership in the parallel pro-
the high aggression classes ~Aggressive and cess latent trajectory class model. Each pre-
Comorbid! and another set was used for the dictor was tested independently of the others,
two low aggression classes ~the Inattentive and in a multinomial regression of the class mem-
the Healthy classes!. This procedure allowed bership on the child-rearing variable, within
us to examine in isolation classes that only the Mplus model. Odds ratios were then com-
differed with respect to one type of behavior puted to test the effect of the predictor on the
problem while holding the other constant. odds of being in the particular class of inter-
The mean growth parameters are shown in est. The first item to examine when adding
Table 2 and define the mean trajectories shown predictors to a model is whether the model
in bold in the graphs in Figure 4 for inattention0 still finds the same solution. We found that
hyperactivity and aggression respectively. For that the solution remained quite similar for
each class, we show a graph of inattention0 any of the predictors added to the model. The
hyperactivity and then one of aggression prob- models all resulted in a four-class solution in
lems. Sample individual trajectories ~light lines, which the most prevalent class was the Healthy
Figure 4! are individual raw scores grouped into class and the membership of this class varied
classes according to the outcome of the un- between 37 and 39%. The second most prev-
conditional growth model. Both inattention0 alent class was always the Inattentive class;
hyperactivity trajectories ~high and low! were membership for this class varied from 34 to
flat, one with an intercept of 11.0 and a nonsig- 38%. The Comorbid class membership varied
nificant slope of 20.04 ~the high inattention0 between 24 and 25%. Finally, the Aggressive
hyperactivity, characterizing both the Comorbid class was always rare, comprising only 1.2– 4%
and the Inattentive class! and the other with an of the sample. The growth parameters esti-
intercept of 1.2 and a nonsignificant slope of mated by the model also remained relatively
0.08 ~the Aggressive and Healthy classes!. For unchanged by the addition of the predictors.
aggression, the high problem trajectory class ~in- Thus, the inattention0hyperactivity intercept
tercept 5 12.4, characterizing the Aggressive ranged from 11.0 to 11.5 for the high classes
and the Comorbid classes! had a constant level and from 1.2 to 1.6 for the low classes. Ag-
of problems over time ~i.e., with a nonsignifi- gression intercept ranged from 12.4 to 13.2
cant negative slope!, whereas the low problem for the high classes and 7.0 to 7.2 for the low
class ~intercept 5 6.97, characterizing the In- classes. Slope estimates were also quite simi-
attentive and Healthy classes! had a decreasing lar. In all cases, the parameter estimates that
problem trajectory over time ~ p , .001!. With were not significantly different from zero re-
regard to prevalence of each class, the Healthy mained so and the ones that were significantly
class was the most common, with 39% of the different from zero also remained so. These
sample. This group displayed the lower trajec- results show that the model was quite stable to
tories for both types of problem behaviors, and the addition of predictors and gave us confi-
decreasing aggression over time. The Inatten- dence in our ability to meaningfully interpret
tive class was the next most common, with 33% these results.
of the children falling in that group. The Co- To separate the effects of inattention 0
morbid class had 24% of the children, and the hyperactivity and aggression trajectories in re-
Aggressive had 4%. lation to home environment predictors, a series
of two-group comparisons were conducted
using multinomial regression within the same
Family factors as predictors of trajectory
modeling environment. Comparisons were
class membership
made between classes, which only differed in
With this basic four-class developmental model the trajectory of one type of problem behavior
in place, we proceeded to examine familial at a time ~i.e., either inattention0hyperactivity
Figure 4.

114
Figure 4. The trajectories of inattention0hyperactivity and aggression development for the four-class latent trajectory
class model. For each class, two graphs are displayed: one of development of inattention0hyperactivity over time, and
the other the development of aggression over time. Heavy lines show mean trajectories for each class, and each of the
thin lines represent the development of behavior problems for one individual. Inattention0hyperactivity growth pa-
rameters were fixed equal in the Comorbid and Inattentive0Hyperactive classes and in the Aggressive and Healthy
classes. Aggression problem growth parameters were fixed equal in the Comorbid and Aggressive classes and in the
Inattentive0Hyperactive and Healthy classes.

115
116 J. M. Jester et al.

Table 3. Odds ratios for predicting membership in higher vs. lower


inattention0hyperactivity trajectory classes

Emotional Support
and Intellectual Parental
Stimulation Alcoholism Cohesiveness

Within low aggression trajectory group


~Inattentive0Hyperactive vs. Healthy
classes! 0.86** ~0.78–0.95! 2.1* ~1.1– 4.1! 0.87 ~0.64–1.2!
Within high aggression trajectory group
~Comorbid vs. Aggressive classes! 0.87** ~0.80–0.95! 11.7** ~1.6–84.8! 1.2 ~0.67–2.0!

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

Table 4. Odds ratios for predicting membership in higher vs. lower


aggression problem trajectory classes

Emotional Support
and Intellectual Parental
Stimulation Alcoholism Cohesiveness

Within low inattention0


hyperactivity trajectory group
~Aggressive vs. Healthy class! 0.94 ~0.82–1.0! 0.25 ~0.04–1.6! 0.44* ~0.24–0.84!
Within high inattention0
hyperactivity trajectory group
~Comorbid vs. Inattentive0Hyperactive
class! 0.95 ~0.88–1.0! 1.4* ~1.1– 4.1! 0.59* ~0.43–0.82!

Note: Odds ratio ~95% confidence interval!.


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

or aggression!, in effect holding the other be- high aggression trajectories were Aggressive
havior constant. For each behavior problem, and Comorbid. Odds ratios for continuous pre-
two sets of comparisons were made, one for dictors, such as HOME scores and FES Cohe-
each level of the other type of behavior prob- siveness, show the relative change in likelihood
lem. For instance, as shown in Figure 4, the in membership in the higher behavior trajec-
Comorbid class and the Aggressive class both tory classes for a one unit change in the pre-
have the same high mean trajectory of aggres- dictor. As seen in Table 3, for example, for
sive behavior ~this is by design; they are fixed each one unit higher HOME score, children
to equality as described earlier!. These classes had 86% lower odds of being in the higher
therefore only differ in the trajectory of inattention0hyperactivity class, compared to
inattention0hyperactivity, which is higher in the lower inattention0hyperactivity class. The
the Comorbid class. Tables 3 and 4 show the interpretation for a binary variable is simpler,
results of the multinomial regressions predict- with the odds ratio being the ratio of the odds
ing membership in trajectory classes. Table 3 for the two classes. For example, Table 3 shows
shows odds ratios for predicting membership that for the group of children with higher ag-
in the higher trajectory class of inattention0 gression trajectories, those with an alcoholic
hyperactivity compared to low inattention0 parent have 11.7 times higher odds of being in
hyperactivity, for the two sets of classes in the high inattention0hyperactivity trajectory
which the aggression trajectory was the same. class. Table 4 shows odds ratios for predicting
Classes with low aggression trajectories were membership in the higher trajectory class of
Healthy and Inattentive, whereas those with aggression compared to low aggression, for
Trajectories of inattention0hyperactivity and aggression 117

the two sets of classes in which the inattention0 described. For the HOME score, there was a
hyperactivity trajectory was the same. large percentage of the sample missing ~33.8%!.
In relation to the primary study hypotheses, Listwise deletion, while widely used in psycho-
consistent with a model in which stimulation logical research, has many known shortcom-
and support contribute to attention develop- ings ~Graham, Cumsille, & Elek Fisk, 2003;
ment, lower HOME scores predicted that the Schafer & Graham, 2002!. To test whether this
children were more likely to be in classes with methodology had biased the results, the analy-
higher inattention0hyperactivity trajectories sis was repeated using imputation of the HOME
~with aggression trajectories held constant, data by the EM algorithm as described in the
shown in Table 3!. Membership in lower ag- Method section. As in the original data set,
gression trajectory classes was not predicted by HOME scores predicted membership in trajec-
the HOME total score when inattention 0 tory classes that only differed in inattention0
hyperactivity trajectory classes were held con- hyperactivity trajectory ~odds ratio @OR# 5
stant ~Table 4!. Conversely, a cohesive family 0.78!, but not aggression ~OR 5 1.003!. These
environment ~FES Cohesiveness! predicted results provide some assurance that the effects
membership in lower aggression trajectory of the HOME on trajectory class membership
classes when inattention0hyperactivity trajec- were not merely a result of an inferior missing
tories were held constant ~Table 3! but did not data handling algorithm.
predict attention class membership. As a check on the assumption that these
With regard to other predictors, having an results were not attributable to nonindepen-
alcoholic parent was associated with member- dence of the siblings used in the analyses,
ship in the higher inattention0hyperactivity tra- the unconditional model and the model with
jectory classes. It was also associated with the HOME as predictor were estimated using only
higher aggression trajectory group, but only the boys, all of whom are in different fam-
for high attention trajectories ~i.e., Comorbid ilies. The intercept of inattention0hyperactivity
vs. Inattentive, but not Aggressive vs. Healthy!. in the two classes with lower inattention 0
Authoritarian and authoritative parenting styles hyperactivity trajectories ~Aggressive and
~Child Rearing Practices! did not predict tra- Healthy! was 35% higher than before, but all
jectory class membership ~data not shown!. of the other parameters were within 10% of
the full sample model. Consistent with the pre-
vious results, those children with higher HOME
Potential confounds
scores were less likely to be in classes with
The possible confound of parent’s inattention0 higher inattention0hyperactivity trajectories
hyperactivity in childhood was tested by using ~OR 5 0.86, p , .005!.
it as a predictor of trajectory class. Although Next, we considered that the source of pre-
father’s inattention0hyperactivity did not pre- dictor report was confounded with the behav-
dict membership in any of the different tra- ior domain ~parents reported on aggression and
jectory classes, mother’s report of her own teachers on inattention0hyperactivity!. To eval-
childhood inattention0hyperactivity problems uate whether results depended on this proce-
predicted membership in aggression trajec- dure, we repeated the entire analysis sequence,
tory classes ~again, while holding inattention0 from unconditional model through predicting
hyperactivity trajectories constant!. It did membership with child-rearing measures, using
not predict membership in inattention 0 a composite measure of teacher reports and
hyperactivity trajectory classes. Thus, paren- parent reports for each behavior problem. That
tal history of inattention0hyperactivity did not analysis again yielded a four-class solution as
account for the primary findings. the best solution, and reproduced the basic as-
As we noted earlier, the Mplus modeling pro- sociation of HOME with attention, and FES
gram does not allow missing data for predic- Cohesiveness with aggression, as reported
tors ~the child-rearing environment variables!. above.
Hence, when data is missing for predictors, list- Finally, we noted that 20 children were re-
wise deletion was performed in the results just ported to have been on ADHD medication at
118 J. M. Jester et al.

some point during the course of the study. We to remain constant throughout the course
estimated both the unconditional model and of childhood while aggression problems were
the model with HOME as a predictor, exclud- decreasing.
ing these children from the analysis. In the An important element of the study was
unconditional model, class membership was that our methodology isolated inattention 0
within 2% of the model with all children. For hyperactivity from overlapping aggression.
the inattention0hyperactivity intercepts, the es- This confound has rendered interpretation of
timates were lower for both high and low tra- family influences on inattention0hyperactivity
jectory classes ~by 9 and 10%, respectively!. problematic in much of the literature. We did
For the intercept of aggressive problems, the this by modeling the growth of inattention0
estimates were within 5% of the original esti- hyperactivity and aggression as a parallel pro-
mates. All of the significant slopes were within cess and creating multiple trajectory groups in
2% of the original model estimates. When which the trajectory of one type of behavior
HOME was used as a predictor of class mem- was held constant, while the trajectory of the
bership, it remained a significant predictor for other type of problem varied. A four-class
classes differing in inattention0hyperactivity model was chosen as the best fitting model,
trajectory only ~OR 5 0.86, p , .01!, whereas with two classes for inattention0hyperactivity
FES scores significantly predicted member- trajectories and two for aggression trajecto-
ship in classes differing in aggression only ries. The class with the largest number of chil-
~OR 5 0.44, p , .05!. dren had stable, low levels of inattention0
hyperactivity and moderate and decreasing
levels of aggression. The next most prevalent
Discussion
class had higher, stable levels of inattention0
This is one of the first attempts to examine hyperactivity while their level of aggression
effects of family and home environment on started out higher than the normative class and
the development of children’s inattention 0 was decreasing throughout childhood and ad-
hyperactivity with growth mixture modeling. olescence. Even though this group was remit-
The two most striking findings were the emer- ting from aggression, in the normative way,
gence of four classes or pathways of develop- they were not improving in their attention
ment, and the double dissociation with regard problems0hyperactivity. The third most prev-
to family environment, in which early stimu- alent class was the most problematic class, with
lation ~HOME Score! predicted attentional high levels of aggression that were only slightly
trajectories but not aggression, and family Co- remitting throughout childhood and adoles-
hesiveness ~FES! predicted aggression but not cence and with high stable levels of inatten-
inattention0hyperactivity. We discuss each of tion0hyperactivity. This is obviously the most
these in turn. disruptive group and continued to experience
First, let us comment on the nature of the greater problems of both types throughout ad-
trajectory paths we obtained. To recap, we ob- olescence. Only a small group of children were
tained reports of inattention0hyperactivity and in the class with high levels of aggression and
aggression from the early school years through- low levels of inattention0hyperactivity. This
out childhood into adolescence, using contin- is consistent with findings that there are few
uous measures of these important externalizing children with aggressive problems who do not
behavior problems. With growth mixture mod- have inattention0hyperactivity in most popu-
eling, we were then able to categorize the chil- lation samples ~Nagin & Tremblay, 2001!.
dren in a statistically supported way, into Moreover, it is noteworthy that the in-
groups based on the trajectory of development attention0hyperactivity and aggression did not
of these problems. The first step in such mod- evolve in parallel for many of the children.
eling was to develop a latent growth model of All of the trajectory class had flat levels of
the development of the two behavior prob- inattention0hyperactivity, but 72% of the chil-
lems for the entire sample. This model showed dren were in a decreasing aggression trajec-
that inattention0hyperactivity overall tended tory group. These trajectory findings were quite
Trajectories of inattention0hyperactivity and aggression 119

comparable to those of Nagin and Tremblay lel decrease in delinquency scores and aggres-
~1999!. They found that 80% of a community sion from early to late childhood, followed by
sample of 1,037 boys, drawn from low socio- a sharp increase in delinquency, with aggres-
economic areas of Montréal and assessed sion continuing to decline, at the advent of
repeatedly from 6 to 15 years of age, were adolescence. Delinquency may indeed be re-
in decreasing aggression trajectory classes, lated differently to the child-rearing factors
whereas only 4% were in a high stable group. studied here.
Their results were quite similar for hyper- When it came to identifying family corre-
activity. Their finding of substantial but not lates, we identified membership in classes that
complete, overlap between problem domain only differed in the trajectory of inattention0
trajectories was in agreement with our results. hyperactivity. For both high and low aggres-
In another report from the same study, Nagin sion trajectory groups, lower levels of emo-
and Tremblay ~2001! used modeling to jointly tional support and intellectual stimulation
estimate hyperactivity and physical aggres- predicted membership in the high inattention0
sion development. They found four classes for hyperactivity trajectory groups. Conversely, co-
each type of behavior yielded the best fit for hesiveness of the family environment predicted
their model. Their larger sample size may have high aggression trajectories, but not higher
allowed finer differentiation of classes. Group- inattention0hyperactivity trajectory classes.
ing their data into a high0low trajectory as This is one of the first clear demonstrations of
ours, 62% of the sample showed the low tra- specific family environment effects on atten-
jectories for both problem behaviors. This is tion problem and hyperactivity development,
comparable to the 39% found in the current independent of aggressive development. As
study’s Healthy class. The lower percentage such, it may provide clues for theories of how
in the present study could reflect that ours was these problems in particular develop.
a high-risk sample. They identified 1.8% as The Moos Family Environment scales
following high aggression, low hyperactivity ~which were related to aggression! used here
trajectories, similar to our placement of 4% in were Cohesion, Organization, and Conflict ~re-
the Aggressive class. verse coded!. These reflect the overall quality
Our finding of a decreasing trajectory of of the family’s interactions; high scores are
aggressive behavior is consistent with that of found in highly functional families that dis-
a number of other investigators, besides those play structure and warmth toward all the mem-
just mentioned ~cf. Loeber, 1982, pp. 1439– bers of the family and a minimum of conflict
1440, for a review!. For instance, over the ages and in-fighting. Violent behavior in the home,
of 9–14, there was a decrease in teacher- and which is related to aggressive behavior in chil-
self-report aggression and fighting in a sam- dren, would also be reflected in lower scores
ple of 115 girls and 104 boys ~Cairns, Cairns, on this measure. The HOME ~which was re-
Neckerman, Ferguson, & Gariépy, 1989!. It lated to inattention0hyperactivity! is a mea-
should be noted however, that a different pat- sure that is collected specifically with respect
tern might have appeared had we instead as- to a particular child and was done during the
sessed other antisocial misbehaviors related preschool period. Several of the subscales are
to delinquency, given that the transition to ad- related to different types of intellectual stim-
olescence is frequently marked by an increase ulation, such as stimulation through toys and
in problem behavior ~Stanger, Achenbach, & games, academic behavior, and the variety of
Verhulst, 1997!. Similar to the aggression and stimulation. These types of stimulation may
delinquency trajectories found by Stanger et al. reflect parenting that encourages children’s
~1997!, an early examination of the trajectory development of self-regulation in the pre-
of delinquency in our own data, as measured school years through early learning experi-
on the Delinquency subscale of the Achen- ences. This may be one determinant of the
bach CBCL ~Mayzer, Moñtanez, Wong, Put- child’s ability to be attentive in the school
tler, Fitzgerald, & Zucker, 2002! and based on environment. Because this measure does not
a heavily overlapping sample, shows a paral- take into account the overall quality of the
120 J. M. Jester et al.

family environment, it may be less important Kessler, McGonagle, Zhao, Nelson, Hughes,
for aggressive behavior. Eshleman, Wittchen, & Kendler, 1994!. To re-
Findings with regard to the Block Child- duce the possible confounding of teratogenic
Rearing Practices Report were not as ex- effects of drinking during pregnancy upon de-
pected. The scores on Authoritarian parenting, velopment, children with evidence of fetal al-
which were expected to be higher in alcoholic cohol effects were excluded from the sample.
families, were not different between alcoholic Given what was already known about chil-
and nonalcoholic families. Neither Authoritar- dren of alcoholics when the study began, it
ian nor Authoritative parenting were related was anticipated that these children would be
to the trajectory classes of aggression or in- at elevated risk for behavioral problems. In-
attention0hyperactivity. These results were all deed, children in the alcoholic families in this
based on responses from the mothers of the study did display higher levels of inattention0
children. Perhaps ratings were biased. Alter- hyperactivity and aggressive behaviors than
natively, they may strive to compensate for children in the control group. Family alcohol-
the alcoholic father by increasing warmth and ism status also predicted membership in classes
rational guidance for the child, but these ef- with higher levels of inattention0hyperactivity,
forts may be overwhelmed by the stress in the when aggressive behavior trajectories were
family due to the father’s psychopathology. held constant. This finding strengthens the ex-
In a previous analysis from our group, Lou- isting literature showing a relationship be-
kas et al. ~2003! found that family conflict tween alcoholism in the parent and higher
predicted the development of aggressive levels of child ADHD symptomatology ~Ku-
behavior in boys, controlling for the boys’ perman et al., 1999!.
undercontrol. Although the methodology is Family alcoholism was also associated with
somewhat different, their results were similar lower SES and lower quality of the child-
to ours in that a measure of conflict, cohesion, rearing environment. Lower SES in alcoholic
and organization predicted a worse trajectory families has been often noted in previous lit-
of aggressive behavior, when concurrent in- erature ~Fitzgerald & Zucker, 1995!. We chose
attention0hyperactivity were held constant. not to control for SES in our analyses because
Loukas et al. ~2003! did not use the longitudi- lower SES is also an important consequence
nal measure of inattention0hyperactivity and of the problem of alcoholism, especially for
did not include the measure of intellectual stim- alcoholism with antisocial behavior ~Zucker,
ulation and warmth. Loukas also found that Ellis, Fitzgerald, Bingham, & Sanford, 1996!.
paternal alcoholism was a predictor of child Controlling for a variable that is an important
aggressive behavior, whereas we found that part of the syndrome under investigation is
alcoholism predicted only inattention0hyper- not advised ~Miller & Chapman, 2001!. Lower
activity, not aggression. They measured of al- SES, in turn, is associated with diminished
coholism throughout the study as a time- quality of child-rearing environment ~Garrett,
varying covariate that would strengthen the Ng’andu, & Ferron, 1994!. As the child-rearing
relation between child problems and paternal environment also predicted inattention0hyper-
alcoholism, as it more accurately reflected the activity and aggression trajectories, it may be
presence of alcohol problems in the fathers that these variables are mediators of the rela-
concurrent with the measures of aggressive be- tionship between family alcoholism and child
havior in the children. behaviors.
This study utilized a sample with high prev- Several potential sources of error in the anal-
alence of alcoholism in the fathers. Alcohol- yses were examined. Two potential confounds
ism in the women was free to vary. Due to were children’s medication for ADHD, and the
assortative mating, the sample also has a sub- inclusion of siblings in the data set. In each
stantially higher than population-rate preva- case, the model results confirmed the robust-
lence of women with alcohol use disorders ness of our approach. Another possible source
~28.7% in this population, compared to the na- of error was that the child-rearing measures
tional norm of 14.6% for this age range; were obtained from the parents, the same re-
Trajectories of inattention0hyperactivity and aggression 121

porter as for aggressive behavior in the cross- Sherman, Iacono, & McGue, 1997!. Rather,
informant model. Shared method variance for our results usefully supplement this other work.
these might cause a greater apparent relation- What the data suggest is that mechanisms spe-
ship between aggressive behavior and the child- cific to maintaining and moderating problems
rearing environmental variables. The ability of with inattention0hyperactivity behaviors over
the model to differentiate the influence of par- time do occur in the home. This is an impor-
enting and family environment variables on ag- tant step in a field that is lacking a clear pic-
gressive behavior suggests that this reporter ture of such mechanisms ~Johnston & Mash,
confound does not dominate the predictions of 2001!. It must be recognized that in a family
the model. Further, a model that used cross- study, genetic and environmental effects are
informant composites of aggressive behavior not dissociable. Thus, even the measure of
and inattention0hyperactivity rated by teach- home environment is likely to be mediating
ers and parents yielded exactly equivalent re- genetic effects via parent temperament or per-
sults of predicting aggressive behavior with sonality traits that we did not measure ~Nigg
family environment and inattention 0hyper- & Goldsmith, 1998; O’Connor, Deater Deck-
activity with intellectual stimulation. This re- ard, Fulker, Rutter, & Plomin, 1998!. Further,
inforces our findings and provides evidence that it is possible that, although the home environ-
the results are not a function of this shared ment was measured before the measurement
method variance. of the inattention0hyperactivity, these prob-
The identification of family factors related lems may have been present in children start-
to inattention0hyperactivity in children may ing at a very early age. The ability of the parents
be due to several differences between our study to provide intellectual stimulation to the child
and prior studies. First, previous studies may could therefore have been compromised due
not have measured the same elements of the to the behavior of the child. This does not
environment that we did. For example, Pater- change the importance of understanding the
nite and Loney ~1980! used self- and spouse mediating mechanisms in the home environ-
reports of “parental shortcomings.” Loney, ment by which such effects may be actualized.
Langhorne, and Paternite’s ~1978! measures The finding that there are differential influ-
included only love0hostility and control 0 ences of characteristics in the environment of
autonomy continuums. Second, the current the child has implications for the development
sample was high risk because of the high of the risk of substance disorders in these chil-
prevalence of alcoholism in the parents, but it dren. The children most at risk for substance
was not selected specifically for inattention0 disorders and serious behavioral consequences
hyperactivity or conduct disorder and was a are those in the most dysfunctional families,
community, as opposed to clinical, sample. for whom both high levels of conflict and low
Third, the children were studied over an levels of cohesion in the family are present as
interval of at least 6 and up to 12 years, as well as low levels of emotional support and
contrasted with many studies that are cross- intellectual stimulation from the parents. These
sectional or consist of only two time points. families may be ones in which both parents
Information on child behavior was obtained are experiencing difficulties with drugs of
from two independent reporters, the child’s abuse themselves, which would further in-
teachers as well as the parents. It was notable crease the likelihood of the child’s having later
that the home environment, which was as- substance use disorders.
sessed by interviewing and observing the par- In contrast, we found little relationship be-
ent, was related to the development of teacher- tween development of children’s inattention0
reported inattention0hyperactivity. hyperactivity and the parents’ retrospective
These findings do not necessarily contra- reports of their own inattention0hyperactivity
dict the substantial literature indicating that problems in childhood. Although there were
inattention0hyperactivity problems are sub- significant zero-order correlations between
stantially heritable ~but not exclusively so, non- inattention 0hyperactivity of the child and
genetic effects remain notable in most studies; the mother’s report of her own inattention 0
122 J. M. Jester et al.

hyperactivity in childhood, her inattention0 findings of home environment effects on child


hyperactivity did not predict trajectory class inattention0hyperactivity development.
membership in a higher inattention 0hyper- Both the home environment and parental al-
activity class. However, there was a relation- coholism were measured at the first time point
ship between the child’s aggression level and of the study ~children aged 3–5 years!, whereas
the mother’s report of inattention0hyperactivity the developmental trajectories span the age
in childhood. This stands in some contrast to a range from 6 to 8 years of age up to 15 to 17
sizeable literature indicating associations be- years. Therefore, these early indicators are pre-
tween ADHD in children and in parents ~Frick, dicting development over a span of up to 12
Lahey, Christ, Loeber, & Green, 1991!. Retro- years. Early prediction of which children are
spective reports from parents reporting on their more likely to experience more inattention0
childhood problems of attention and aggres- hyperactivity may make it possible to direct in-
sion are problematic, because they are report- tervention to those children most in need.
ing on a time period up to several decades prior. In summary, by applying recently devel-
In the Frick study ~1991b!, for the most part the oped statistical techniques that create groups of
mothers were reporting on behavior problems children based on developmental trajectories,
of the children’s biological fathers in child- we were able to find factors in the family and
hood. In the present study, two different mod- home environment that predicted development
els of inattention0hyperactivity were utilized: of inattention0hyperactivity problems indepen-
children were assessed with theAchenbach ~sin- dent of the development of aggression prob-
gle factor! model, while parents were assessed lems. In addition, different home environment
with the DSM-IV ~two-factor! model. Further- factors predicted aggression development, in-
more, it is possible that the mothers who re- dependent of differences in the development of
ported childhood inattention 0hyperactivity inattention0hyperactivity problems. Disentan-
problems also experienced elevated aggres- gling the development of inattention0hyper-
sion problems in childhood and the relation- activity and aggression problems has allowed
ship reflects this comorbidity. Nonetheless, it us to find distinct predictors of aggression and
was notable that parent history of inattention0 inattention0hyperactivity trajectories through-
hyperactivity problems did not account for our out childhood and adolescence.

References
Achenbach, T. M. ~1991a!. Manual for the Child Behav- Bradley, R. H., Caldwell, B. M., Rock, S. L., Hamrick,
ior Checklist04–18 and 1991 profile. Burlington, VT: H. M., & Harris, P. ~1988!. Home observation for mea-
University of Vermont Department of Psychiatry. surement of the environment: Development of a home
Achenbach, T. M. ~1991b!. Manual for the teacher’s re- inventory for use with families having children 6 to
port form and 1991 profile. Burlington, VT: Univer- 10 years old. Contemporary Educational Psychology,
sity of Vermont Department of Psychiatry. 13, 58–71.
American Psychiatric Association. ~2000!. Diagnostic and Cadoret, R. J., Yates, W. R., Troughton, E., Woodworth,
statistical manual of mental disorders ~4th ed., text G., & Stewart, M. A. ~1995!. Genetic–environmental
revision!. Washington, DC: Author. interaction in the genesis of aggressivity and conduct
Barkley, R. A., & Cunningham, C. E. ~1979!. The effects disorders. Archives of General Psychiatry, 52, 916–924.
of methylphenidate on the mother–child interactions Cairns, R. B., Cairns, B. D., Neckerman, H. J., Ferguson,
of hyperactive children. Archives of General Psychi- L. L., & Gariépy, J.-L. ~1989!. Growth and aggres-
atry, 36, 201–208. sion: I. Childhood to early adolescence. Developmen-
Baumrind, D. ~1968!. Authoritarian vs. authoritative pa- tal Psychology, 25, 320–330.
rental control. Adolescence, 3, 255–272. Cantwell, D. ~1975!. Familial-genetic research with hy-
Biederman, J., Milberger, S., Faraone, S. V., Kiely, K., peractive children. In D. Cantwell ~Ed.!, Hyperactiv-
Guite, J., Mick, E., Ablon, S., Warburton, R., & Reed, ity, diagnosis, management, and current research. New
E. ~1995!. Family–environment risk factors for York: Spectrum.
attention–deficit hyperactivity disorder: A test of Rut- Chen, W. J., Faraone, S. V., Biederman, J., & Tsuang,
ter’s indicators of adversity. Archives of General Psy- M. T. ~1994!. Diagnostic accuracy of the Child Be-
chiatry, 52, 464– 470. havior Checklist scales for attention–deficit hyper-
Block, J. H. ~1981!. The child rearing practices report. activity disorder: A receiver–operating characteristic
Berkeley, CA: University of California, Institute of analysis. Journal of Consulting & Clinical Psychol-
Human Development. ogy, 62, 1017–1025.
Trajectories of inattention0hyperactivity and aggression 123

Darling, N., & Steinberg, L. ~1993!. Parenting style as D. Cicchetti ~Ed.!, Internalizing and externalizing ex-
context: An integrative model. Psychological Bulle- pressions of dysfunction ~pp. 21–55!. Hillsdale, NJ:
tin, 113, 487– 496. Erlbaum.
Drotar, D., Stein, R. E. K., & Perrin, E. C. ~1995!. Meth- Grych, J. H., & Fincham, F. D. ~Eds.!. ~2001!. Interpa-
odological issues in using the Child Behavior Check- rental conflict and child development: Theory, re-
list and its related instruments in clinical child search, and applications. New York: Cambridge
psychology research. Journal of Clinical Child Psy- University Press.
chology, 24, 184–192. Hinshaw, S. P. ~1987!. On the distinction between atten-
Duncan, G. J., Brooks–Gunn, J., & Klebanov, P. K. ~1994!. tional deficits0hyperactivity and conduct problems0
Economic deprivation and early childhood develop- aggression in child psychopathology. Psychological
ment. Child Development, 65, 296–318. Bulletin, 101, 443– 463.
Eisenberg, N., Cumberland, A., Spinrad, T. L., Fabes, R. A., Huang–Pollock, C. L., Carr, T. H., & Nigg, J. T. ~2002!.
Shepard, S. A., Reiser, M., Murphy, B. C., Losoya, Development of selective attention: Perceptual load
S. H., & Guthrie, I. K. ~2001!. The relations of regu- influences early versus late attentional selection in chil-
lation and emotionality to children’s externalizing and dren and adults. Developmental Psychology, 38,
internalizing problem behavior. Child Development, 363–375.
72, 1112–1134. Johnston, C., & Mash, E. J. ~2001!. Families of children
Feighner, J. P., Robins, E., Guze, S. B., Woodruff, Jr., with Attention–Deficit0Hyperactivity Disorder: Re-
R. A., Winokur, G., & Munoz, R. ~1972!. Diagnostic view and recommendations for future research. Clin-
criteria for use in psychiatric research. Archives of ical Child & Family Psychology Review, 4, 183–207.
General Psychiatry, 26, 57– 63. Johnston, C., Murray, C., Hinshaw, S. P., Pelham, W. E.,
Fergusson, D. M., & Horwood, L. J. ~1995!. Early disrup- & Hoza, B. ~2002!. Responsiveness in interactions of
tive behavior, IQ, and later school achievement and mothers and sons with ADHD: Relations to maternal
delinquent behavior. Journal of Abnormal Child Psy- and child characteristics. Journal of Abnormal Child
chology, 23, 183–199. Psychology, 30, 77–88.
Fitzgerald, H. E., & Zucker, R. A. ~1995!. Socioeconomic Kendziora, K. T., & O’Leary, S. G. ~1993!. Dysfunctional
status and alcoholism: The contextual structure of de- parenting as a focus for prevention and treatment of
velopmental pathways to addiction. In H. E. Fitzger- child behavior problems. In T. H. Ollendick ~Ed.!,
ald, B. M. Lester, & B. Zuckerman ~Eds.!, Children of Advances in clinical child psychology ~Vol. 15,
poverty: Research, health, and policy issues ~pp. 125– pp. 175–206!. New York: Plenum Press.
148!. New York: Garland Publishing. Kessler, R. C., McGonagle, K. A., Zhao, S., Nelson, C. B.,
Frick, P. J. ~1994!. Family dysfunction and the disruptive Hughes, M., Eshleman, S., Wittchen, H. U., & Kend-
behavior disorders: A review of recent empirical ler, K. S. ~1994!. Lifetime and 12-month prevalence
findings. In T. H. Ollendick & R. J. Prinz ~Eds.!, Ad- of DSM-III-R psychiatric disorders in the United
vances in clinical child psychology ~Vol. 16, pp. 203– States. Results from the National Comorbidity Sur-
226!. New York: Plenum Press. vey. Archives of General Psychiatry, 51, 8–19.
Frick, P. J., Kamphaus, R. W., Lahey, B. B., Loeber, R., Klein, R. G., & Mannuzza, S. ~1991!. Long-term out-
Christ, M. A. G., Hart, E. L., & Tannenbaum, L. E. come of hyperactive children: A review. Journal of
~1991!. Academic underachievement and the disrup- the American Academy of Child & Adolescent Psychi-
tive behavior disorders. Journal of Consulting & Clin- atry, 30, 383–387.
ical Psychology, 59, 289–294. Kochanska, G., Coy, K. C., & Murray, K. T. ~2001!. The
Frick, P. J., Lahey, B. B., Christ, M. A., Loeber, R., & development of self-regulation in the first four years
Green, S. ~1991!. History of childhood behavior prob- of life. Child Development, 72, 1091–1111.
lems in biological relatives of boys with attention– Kochanska, G., Kuczynski, L., & Radke–Yarrow, M.
deficit hyperactivity disorder and conduct disorder. ~1989!. Correspondence between mothers’ self-
Journal of Clinical Child Psychology, 20, 445– 451. reported and observed child-rearing practices. Child
Garrett, P., Ng’andu, N., & Ferron, J. ~1994!. Poverty Development, 60, 56– 63.
experiences of young children and the quality of Kopp, C. B. ~1982!. Antecedents of self-regulation: A de-
their home environments. Child Development, 65, velopmental perspective. Developmental Psychology,
331–345. 18, 199–214.
Ge, X., Conger, R. D., Cadoret, R. J., Neiderhiser, J. M., Kuperman, S., Schlosser, S. S., Lidral, J., & Reich, W.
Yates, W., Troughton, E., & Stewart, M. A. ~1996!. ~1999!. Relationship of child psychopathology to pa-
The developmental interface between nature and rental alcoholism and antisocial personality disorder.
nurture: A mutual influence model of child antisocial Journal of the American Academy of Child Adoles-
behavior and parent behaviors. Developmental Psy- cence & Psychiatry, 38, 686– 692.
chology, 32, 574–589. Loeber, R. ~1982!. The stability of antisocial and delin-
Gondoli, D. M., & Jacob, T. ~1993!. Factor structure within quent child behavior: A review. Child Development,
and across three family-assessment procedures. Jour- 53, 1431–1446.
nal of Family Psychology, 6, 278–289. Loeber, R., Brinthaupt, V. P., & Green, S. M. ~1990!. At-
Graham, J. W., Cumsille, P. E., & Elek Fisk, E. ~2003!. tention deficits, impulsivity, and hyperactivity with or
Methods for handling missing data. In J. A. Schinka & without conduct problems: Relationships to delin-
W. F. Velicer ~Eds.!, Handbook of psychology: Re- quency and unique contextual factors. In R. J. McMa-
search methods in psychology ~Vol. 2, pp. 87–114!. hon & R. D. Peters ~Eds.!, Behavior disorders of
New York: Wiley. adolescence: Research, intervention, and policy in clin-
Greenberg, M. T., Kusche, C. A., & Speltz, M. ~1991!. ical and school settings ~pp. 39– 61!. New York: Ple-
Emotional regulation, self-control, and psychopathol- num Press.
ogy: The role of relationships in early childhood. In Loney, J., Langhorne, J. E., & Paternite, C. E. ~1978!. An
124 J. M. Jester et al.

empirical basis for subgrouping the hyperkinetic0 Nigg, J. T. ~2001!. Is ADHD a disinhibitory disorder?
minimal brain dysfunction syndrome. Journal of Ab- Psychological Bulletin, 127, 571–598.
normal Psychology, 87, 431– 441. Nigg, J. T., & Goldsmith, H. H. ~1998!. Developmental
Loney, J., Paternite, C. E., Schwartz, J. E., & Roberts, psychopathology, personality, and temperament: Re-
M. A. ~1997!. Associations between clinic-referred flections on recent behavioral genetics research. Hu-
boys and their fathers on childhood inattention- man Biology, 70, 387– 412.
overactivity and aggression dimensions. Journal of Nigg, J. T., & Hinshaw, S. P. ~1998!. Parent personality
Abnormal Child Psychology, 25, 499–509. traits and psychopathology associated with antisocial
Loukas, A., Zucker, R. A., Fitzgerald, H. E., & Krull, J. L. behaviors in childhood attention–deficit hyperactiv-
~2003!. Developmental trajectories of disruptive be- ity disorder. Journal of Child Psychology & Psychia-
havior problems among sons of alcoholics: Effects of try & Allied Disciplines, 39, 145–159.
parent psychopathology, family conflict, and child un- O’Connor, T. G., Deater Deckard, K., Fulker, D., Rutter,
dercontrol. Journal of Abnormal Psychology, 112, M., & Plomin, R. ~1998!. Genotype–environment cor-
119–131. relations in late childhood and early adolescence: Anti-
Luster, T., & McAdoo, H. P. ~1994!. Factors related to the social behavioral problems and coercive parenting.
achievement and adjustment of young African Amer- Developmental Psychology, 34, 970–981.
ican children. Child Development, 65, 1080–1094. Paternite, C. E., & Loney, J. ~1980!. Childhood hyperkine-
Mayzer, R., Moñtanez, M., Wong, M. M., Puttler, L. I., sis: Relationships between symptomatology and home
Fitzgerald, H. E., & Zucker, R. A. ~2002!. Onset of environment. In C. K. Whalen & B. Henker ~Eds.!,
alcohol use: A comparison between early drinkers and Hyperactive Children: The Social Ecology of Identi-
non-drinkers in the continuity of aggression and de- fication and Treatment. New York: Academic Press.
linquency from preschool to adolescence. Paper pre- Patterson, G. R. ~1982!. Coercive family process. Eugene,
sented at the Annual meeting of the American Society OR: Castalia.
of Criminology, Chicago. Reich, W., Earls, F., & Powell, J. ~1988!. A comparison of
McArdle, J. J., & Epstein, D. ~1987!. Latent growth curves the home and social environments of children of al-
within developmental structural equation models. Child coholic and non-alcoholic parents. British Journal of
Development, 58, 110–133. Addiction, 83, 831–839.
Miller, G. M., & Chapman, J. P. ~2001!. Misunderstand- Robins, L. N., Helzer, J. E., Croughan, J. L., & Ratcliff,
ing analysis of covariance. Journal of Abnormal Psy- K. S. ~1980!. The NIMH Diagnostic Interview Sched-
chology, 110, 40– 48. ule: Its history, characteristics and validity. St. Louis,
Moos, R. H., & Moos, B. S. ~1994!. Family Environment MO: Washington University School of Medicine.
Scale manual ~3rd ed.!. Palo Alto, CA: Consulting Robison, L. M., Sclar, D. A., Skaer, T. L., & Galin, R. S.
Psychologists Press. ~1999!. National trends in the prevalence of attention–
Muthén, B. ~2000!. Methodological issues in random co- deficit0hyperactivity disorder and the prescribing of
efficient growth modeling using a latent variable frame- methylphenidate among school-age children: 1990–
work: Applications to the development of heavy 1995. Clinical Pediatrics, 38, 209–217.
drinking. In J. Rose, L. Chassin, C. Presson, & J. Sher- Sanford, K., Bingham, C. R., & Zucker, R. A. ~1999!.
man ~Eds.!, Multivariate applications in substance use Validity issues with the Family Environment Scale:
research: New methods for new questions ~pp. 113– Psychometric resolution and research application with
140!. Hillsdale, NJ: Erlbaum. alcoholic families. Psychological Assessment, 11,
Muthén, B. ~2001!. Latent variable mixture modeling. In 315–325.
G. A. Marcoulides & R. E. Schumacker ~Eds.!, New Scarr, S., & McCartney, K. ~1983!. How people make
developments and techniques in structural equation their own environments: A theory of genotype r en-
modeling ~pp. 1–33!. Mahwah, NJ: Erlbaum. vironment effects. Child Development, 54, 424– 435.
Muthén, B., & Satorra, A. ~1995!. Complex sample data Schafer, J. L., & Graham, J. W. ~2002!. Missing data: Our
in structural equation modeling. Sociological Method- view of the state of the art. Psychological Methods, 7,
ology, 25, 267–316. 147–177.
Muthén, L. K., & Muthén, B. O. ~1998!. Mplus user’s Searight, H. R., & McLaren, A. L. ~1998!. Attention–
guide. Los Angeles: Muthén & Muthén. deficit hyperactivity disorder: The medicalization of
Nagin, D., & Tremblay, R. E. ~1999!. Trajectories of misbehavior. Journal of Clinical Psychology in Med-
boys’ physical aggression, opposition, and hyperactiv- ical Settings, 5, 467– 495.
ity on the path to physically violent and nonviolent Sherman, D. K., Iacono, W. G., & McGue, M. K. ~1997!.
juvenile delinquency. Child Development, 70, Attention–deficit hyperactivity disorder dimensions:
1181–1196. A twin study of inattention and impulsivity–hyper-
Nagin, D. S., & Tremblay, R. E. ~2001!. Analyzing devel- activity. Journal of the American Academy of Child &
opmental trajectories of distinct but related behaviors: Adolescent Psychiatry, 36, 745–753.
A group-based method. Psychological Methods, 6, Stanger, C., Achenbach, T. M., & Verhulst, F. C. ~1997!.
18–34. Accelerated longitudinal comparisons of aggressive
Nesselroade, J. R. ~1992!. Adult personality develop- versus delinquent syndromes. Development and Psy-
ment: Issues in assessing constancy and change. In chopathology, 9, 43–58.
R. A. Zucker ~Ed.!, Personality structure in the life Stevens, G., & Featherman, D. L. ~1981!. A revised so-
course: Essays on personology in the Murray tradi- cioeconomic index of occupational status. Social Sci-
tion ~pp. xiv, 382!. New York: Springer. ence Research, 10, 364–395.
Nigg, J. T. ~2000!. On inhibition0disinhibition in devel- Stevens, J., Quittner, A. L., & Abikoff, H. ~1998!. Factors
opmental psychopathology: Views from cognitive and influencing elementary school teachers’ ratings of
personality psychology and a working inhibition tax- ADHD and ODD behaviors. Journal of Clinical Child
onomy. Psychological Bulletin, 126, 220–246. Psychology, 27, 406– 414.
Trajectories of inattention0hyperactivity and aggression 125

Tarter, R. E., McBride, H., Buonpane, N., & Schneider, sociality and heterogeneity of alcoholic outcome. De-
D. U. ~1977!. Differentiation of alcoholics: Child- velopment and Psychopathology, 8, 831–848.
hood history of minimal brain dysfunction, family Zucker, R. A., Fitzgerald, H. E., & Noll, R. B. ~1990!.
history, and drinking pattern. Archives of General Psy- Drinking and drug history ~4th ed.!. Unpublished
chiatry, 34, 761–768. manuscript, Michigan State University.
Whalen, C. K., & Henker, B. ~1999!. The child with Zucker, R. A., Fitzgerald, H. E., Refior, S. K., Puttler,
attention–deficit0hyperactivity disorder in family con- L. I., Pallas, D. M., & Ellis, D. A. ~2000!. The clinical
texts. In Handbook of disruptive behavior disorders and social ecology of childhood for children of alco-
~pp. 139–155!. New York: Kluwer Academic0Plenum holics: Description of a study and implications for a
Press. differentiated social policy. In H. E. Fitzgerald, B. M.
Winsler, A. ~1998!. Parent–child interaction and private Lester, & B. S. Zuckerman ~Eds.!, Children of addic-
speech in boys with ADHD. Applied Developmental tion: Research, health and policy issues ~pp. 174–
Science, 2, 17–39. 222!. New York: Garland Press.
Zucker, R. A., Ellis, D. A., Fitzgerald, H. E., Bingham, Zucker, R. A., & Gomberg, E. S. ~1986!. Etiology of al-
C. R., & Sanford, K. ~1996!. Other evidence for at coholism reconsidered: The case for a biopsycho-
least two alcoholisms II: Life course variation in anti- social process. American Psychologist, 41, 783–793.

Vous aimerez peut-être aussi