Psychological Reports, 1989, 65, 707-726.

O Psychological Reports 1989

CHANGES IN DRUG INVOLVEMENT:

A LONGITUDINAL STUDY OF CHILDHOOD AND
ADOLESCENT DETERMINANTS '

JUDITH S. BROOK MARTIN WHITEMAN

Department of Psychiatry, School of Social Work,
Division of Child and Adolescent Psychiatry, Columbia Univer~ity
Mount Sinai School of Medicine

ANN SCOVELL GORDON PATRICIA COHEN

Department of Psychiatry, New York Psychiat7ic Instiiute,
Diuision of Child and Adolescent Psychiahy, Columbia University
Mount Sinai School of Medicine

Summary.-Previous research has identified childhood and adolescent personality

determinants of early adolescent drug involvement. The purpose of the present study
was to examine the determinants of increased involvement over time and to compare
these results with previous findings regarding early involvement. Data were available
on 654 white males and females at three paints, Time 1 (T,) at ages 1-10 yr, Time
2 (TJ at ages 9-18 yr., and T i e 3 (T,) at ages 11-20 yr. The subjects (at T, and
T,) and their mothers (at all three points) were given structured interviews assessing
the child's personality and behavior. Results indicated that T, traits of conventionality
and emotional control were associated with similar traits at T,, which, in turn, were
related to lower drug involvement over time (T, and T,). Interactive effects indicated
first that T, adolescent protective (nondrug-conducive) factors weakened the effect of
childhood-risk (drug-conducive) characteristics resulting in lower drug involvement.
Second, high levels of earlier drug use interacted synergistically with personality risk
leading to increased levels of involvement. Over-all, the personality Factors implicated
in increased involvement were similar to those related to earlier involvement.

In the last two decades, adolescent drug use has received as much
research attention as almost any other health-risk behavior, if not more. At
the present time, we have accumulated a great deal of evidence about the
prevalence and correlates of adolescent drug behavior. At the same time
there have been a number of well-designed, theoretically based studies
dealing with the effects of drug behavior. In addition, there have been
widespread media campaigns, massive prevention programs instituted in
the schools, and increased legislative restrictions dealing with drug use.

'This study was supported by Research Grant DA03188 from the National Institute on Drug
Abuse, by Grant MH36971 from the National Institute on Mental Health, and in part by
Research Scientist Development Award DA00094 Emm the National Institute on Drug Abuse
to Judith S. Brook. The authors acknowledge Jesse Smith, Claudia Hartmark, and
Erika Waly for subject recruitment and &r their help in carrying out the study. They also
express appreciation to Coryl Jones for advice, and to Carolyn Nomura and Sandy Stillman for
research assistance. The original study from which the childhood data were obtained was
conducted by Leonard Kogan. Corres ondence concerning this article should be addressed to
Judith Brook, Mount Sinai School o P ~ e d i c i n e ,One Gustave Levy Place. Amenberg 22-74,
Box 1229, New York, New York 10029.
708 J. S. BROOK, ET AL.

Nevertheless, a significant proportion of our youth continues to engage in

drug behavior. This state of affairs leads to a concern with two related, but
distinct, issues; the first deals with the factors that underlie the beginning
use of drugs, the second with the factors that are associated with changes in
drug use once the behavior has been initiated. In our view, it is not clear
whether the factors related to earlier drug use are similar to, or different
from, those implicated in later use. The present study was specifically
designed to compare the childhood and adolescent personality characteristics
associated with increased use with those implicated in earlier use. It is of
both theoretical and practical value to differentiate between these two
aspects of drug use because such understanding can provide guidelines not
only for prevention but also for ways to help adolescents who have begun to
use drugs and are at the greatest risk for escalation.
In an earlier paper, the authors examined the childhood antecedents
(Time 1) and adolescent (Time 2) correlates of Time 2 drug use (Brook,
Whiteman, Gordon, & Cohen, 1986). In that study we reported on young-
sters who were between the ages of 5 and 10 yr. at initial Time 1
assessment and who were 13 to 18 yr. old at Time 2. The current study is
also based on this data set but includes a third data collection (Time 3)
when the youngsters were between the ages of 15 and 20 yr. In addition,
we have added a number of children who were 1-4 yr. at T,, 9-12 yr, at
T,, and 13-16 yr. at T,. We did not include them in our previous T, study
because many of the youngsters were too young to have initiated drug use at
that time. The findings of this previous study suggested that a set of T,
childhood personality factors was associated with a set of T, adolescent per-
sonality factors which, in turn, was associated with the adolescent's drug use
(T,), supporting a mediational model. Based on this previous research, it is
hypothesized that a mediational model will also best describe the interrela-
tionship of childhood and adolescent personality traits and increased drug
involvement over time (T, to T,). More specifically, a mediational model
would indicate that childhood personality traits lead to similar, but not nec-
essarily identical, adolescent personality traits which, in turn, would be
associated with the adolescents' becoming more heavily involved in drug
use.
Prospective and long-term investigations of the childhood antecedents
of adolescent drug use are quite sparse. At the present time, there are only a
few studies that meet this double criterion, and none examines the child-
hood factors associated with change in levels of drug use over time. One
long-term prospective study which has focused on the early antecedents of
adolescent drug use is that of Block, Block, and Keyes (1988) who reported
that a configuration of childhood attributes including less social competence,
less control, and more rebelliousness was associated with more adolescent
 CHANGE OVER TIME IN DRUG INVOLVEMENT 709

drug use. In another prospective longitudinal study, Kellam and his col-
leagues (Kellam, Branch, Agrawal, & Ensminger, 1975; Kellam, Brown,
Rubin, & Ensminger, 1983), using a sample of lower-class black youths liv-
ing in Chicago, found that aggressive first graders were most likely to use
drugs in adolescence. In our own previous research (Brook, et al., 1986), we
found that a childhood set of personality factors including such dimensions
as temper tantrums, anger, and nonpersistence was positively associated with
adolescent drug use. In the present investigation, we hoped to determine
whether the childhood characteristics just described are as reliable predictors
of changes in drug involvement over time as they were of earlier levels of
drug use.
In contrast to the studies dealing with the childhood predictors of ado-
lescent drug use, there have been a number of studies of adolescent
personality correlates. However, these studies have been, for the most part,
cross-sectional or short-term. Their results have shown a great deal of consis-
tency and indicate that aspects of adolescent unconventionality, such as
deviant behavior, rebelliousness, lack of social conformity, and low law abid-
ance, are associated with drug use (Brook, Lukoff, & Whiteman, 1977,
1980; Huba & Bentler, 1983; Jessor & Jessor, 1977, 1978). Findings have
further shown that adolescent intrapsychic distress and poor control of emo-
tions are related to adolescents' use of drugs (Block, et al., 1988; Brook,
Whiteman, Brook, & Gordon, 1981; Brook, Whiteman, Gordon, & Brook,
1984). Unfortunately, the various findings derived from these investigations
have been somewhat constrained as they have included individuals who were
studied at one point in time or who were seen over relatively short periods
of time. For the most part, the subjects were in the adolescent stage of
development. Since a number of the studies are cross-sectional, it is difficult
to establish causal priorities of the independent and dependent variables. By
having childhood (T,) and adolescent (T,) measures that clearly precede T,
drug involvement, we will be able to address this issue of causality to a
greater extent.
As indicated earlier, we are particularly concerned with how childhood
and adolescent personality traits interrelate in terms of drug use. To our
knowledge, no studies have examined such interrelationships in terms of
both earlier levels of drug use and increased involvement over time.
A final aspect of the study involves an investigation of the interactive
effects of individual childhood and adolescent personality traits on changes
in drug involvement over time. In our previous work, such interactive
effects were quite important for early levels of drug use. In the present
study, we focus on one type of interactive mechanism, risk/protective, which
deals with the extent to which childhood-risk (drug-conducive) personality
traits can be ameliorated by the existence of protective (nondrug-conducive)
710 J. S. BROOK, ET AL.

adolescent characteristics leading to less drug involvement over time. F o r a

more complete description of this and other interactive mechanisms, see
Brook, et a/. (1986).]
Finally, we examine the interactions of T, drug use and the childhood/
adolescent personality traits to see how they affect the adolescents' drug
involvement over time. I n brief, are there certain personality traits that will
make youngsters more (or less) prone to drug use escalation once they have
begun to use drugs?

Subjects
The sample consisted of 654 youngsters and their mothers. The subjects
were predominantly white and were from wide-ranging socioeconomic
backgrounds. The distribution of male and female youth was approximately
equal. The mothers were seen at three points in time, Time 1 (T,) when the
children were 1-10 yr. old, Time 2 (T,) when the children were 9-18 yr.,
and Time 3 (T,), when the children were 11-20 yr. At T, and T,, the
youngsters were also interviewed. Regarding attrition rates, over the three
points in time the loss has been less than 20% In our view, this is a modest
figure for an investigation that has spanned so many years. In addition, de-
mographic comparisons of the T I , T,, and T, data show that the sample has
been essentially the same at all three points demographically. The relatively
low rates of attrition in our study are probably due to our constant efforts
to maintain and update contacts with the subjects. For example, we commu-
nicate with the sample at least semiannually, requesting information on
possible address changes. In addition, we send the families occasional news-
letters describing general study results to maintain their interest and
involvement. (See Brook, et al., 1986 for a more detailed description of the
sample and demographic comparisons over time. See Kogan, Smith, and
Jenkins, 1977, for details on the sampling procedures used at T,.)
Procedures and Instruments
At T,, the mothers were given a structured interview in their own
homes by trained interviewers. The interview contained questions about the
child's personality/behavior, parental childrearing techniques, family struc-

2Socioeconomic status was assessed by combining parental education, father's occupation, and
farmly income. A proximately 42% of the parents were high school graduates and another
40% had some coiege education. Forty-five percent of the fathers were skilled workers or less,
18% held clerical, technical, or sales positions, and the remaining 36% were business adminis-
trators or owners, semiprofessionals, or professionals. Fifty-one percent of the family incomes
were below $15,000 a year and 41% were above (8% did not respond to this question). For
this study, we used the TI (1975) measure of socioeconomic status, which explains the low (by
current standards) family income figures. The correlations of T, SES with T, and T, SES
were .85 and 3 2 , respectively.
 CHANGE OVER TIME IN DRUG INVOLVEMENT 711

ture, family health, and demographics. The mothers were given a parallel
version of this instrument at T, and T,. At T, and T, the adolescents were
given a self-administered questionnaire assessing their own personality/be-
havior and drug use. A number of scales were developed separately from the
mother interviews and the adolescent questionnaires measuring four major
aspects of the child's personahty-conventionality, control of emotions, in-
trapsychic functioning, and interpersonal relatedness. Please note that the
term "personality" is used as a convenience in this paper to cover a wide
range of personality, attitudinal, and behavioral characteristics. [A complete
description of these scales, their sources, and their Cronbach alphas can be
found in Brook, et al. (1986).1
Because of the large number of scales at each point in time, we at-
tempted to reduce the data set in two ways. First, we eliminated scales with
alphas below .46. Second, after examining their intercorrelations, we com-
bined scales that assessed similar traits as follows: mother and child reports
of the youngster's achievement were combined into one scale, as were their
reports of the child's deviant behavior; separate scales measuring the child's
intrapsychic functioning were combined into one "pathology" scale; and two
items on educational expectations and aspirations were combined into one.
The complete list of T, and T, personality scales used in this paper can be
found in Table 1.
Dependent Drug Use Variable
Measurement of drug involvement has developed from two main tradi-
tions. The first, following Kandel's (1975) demonstration of Guttman
Scalability of such data, describes the stage of drug use as progressing from
legal drug to marijuana to other illicit drugs. These data are usually gener-
ated from "ever used" questions. The second set of measures are usually
drug- or drug-group-specific. These measures tap frequency and sometimes
dosages of use of drugs. Each of these strategies has its own advantages and
disadvantages. The most serious disadvantage of the stage approach to meas-
uring drug involvement is that it ignores information regarding the intensity
of involvement as well as multi-drug use. Since recent evidence has called
into question the long-term consequences of experimental drug use, investi-
gations limited to drug-stage analysis may be missing that aspect that is most
consequential, both to society and to the individual. Furthermore, an indi-
vidual's heavy involvement with alcohol may be a much more significant
problem than a simultaneous or previous experimental use of an illicit drug.
Measures of intensity of involvement with individual drugs, on the
other hand, ignore the fact that individuals are often multidrug users. One
solution to this is to divide samples to take multidrug use into account with
a resulting serious undermining of the sample size and statistical power.
Furthermore, each drug must be treated as a separate problem, in spite of
712 J. S. BROOK, ET AL.

TABLE I
CORREU~ONS OF T,AND T, VARIABLES(PARTIALING AGE)WITH T,
INCREASED
LEVELSOF DRUGINVOLVEMENT AND T, EARLY LEVELS OF DRUGINVOLVEMENT

Variable r (Partialin~rAee) With:

T, Increased T, Early
Levels of Drug Levels of Drug
Involvement' Involvement

Conventionality
Achievement (less)
Compliance (less)
Noncodorming behavior (more)
School behavior
Control of emotions
Task persistence
Anger (more)
Crying/demanding (more)
Temper (more)
Intrapsychic functioning
(No T , measures)
Interpersonal relatedness
Aggression to peers
Aggression to siblings (more)
Social inhibition (less)
Other
Eating disorders
T, measures
Conventionality
Achievement (less)'
Church attendance
Compliance (lesslb
Deviancy (more)'
Educational goals
Rebelliousness ( m o d d
Responsibility (lessld
School behavior (poor)b
Task persistence (less)b
Tolerance of deviance (more)d
Control of emotions
Anger (more)b
Cryingldemanding (more)b
Impulsivity (more)d
Temper (morelb
Intrapsychic functioning
Ego integration ( l e d d
Pathology (moreld
Self-esteem (lessld
(continued on next page)
 CHANGE OVER TIME IN DRUG INVOLVEMENT

TABLE 1 (CONT'D)
CORRELATIONS
OF TI AND T, VARIABLES(PARTIALING
AGE)WITH TJ
INCREASED
LEVELSOF DRUGINVOLVEMENT AND T, E m y LEVELSOF DRUGINVOLVEMENT

Variable r (Partialing Age) With:

T, Increased T, Early
Levels of Drua Levels of Drua

T, measures (continued)
Interpersonal relatedness
Aggression to peers (more)'
Aggresrion to siblingsb
Interpersonal difficulty (moreld
Isolation fmm peersb
Social inhibition Oesslb
Other
Eating- -problemsb
Note.-Three TI measures were not obtained for the very youngest children because they
were inappropriate-Achievement, Nonconforming Behavior, and School Behavior. Similarly,
at T, the younger children (ages 9-11 yr.) were not given the Sensation Seeking scale. Mean
scores were assigned For the missing data.
'The T, increased drug involvement variable is with control on T, involvement. This variable
is more fully explcated in the text. 'Scales based on maternalgcrception. 'Scales based on
combined adolescent and maternal perception. dScales based on a olescents' perception.
*p<.05. t p c . 0 1 . $p<.OOl.

evidence that risk factors are overwhelmingly parallel for most kinds of drug
involvement.
We used the accumulated evidence on drug use to devise a combined
stage-intensity of involvement measure in the following- way. .
Use of each
drug (beerlwine, alcohol, cigarettes, marijuana, other illicit drugs) was meas-
ured on a frequency scale. These frequency scales were then weighted by the
seriousness of drug as indicated by its position on the stage of drug use
measure. Since drugs at higher stages also have lower prevalence of use in
the population, these weights also serve to approximately equalize the vari-
ance of the individual drug frequency measures. These weighted frequency
measures were then summed to produce an over-all drug involvement score.
Two indices were created, one for T, drug and one for T,. The T, index was
based on one's ever use of drugs. The T, index assessed drug use over the 2-
yr. span from T, to T,. The means and standard deviations for the two
indices were as follows; for T,: M = 21.16, SD = 11.33; T,: M = 27.03,
SD = 15.01.
The major thrust of this paper is to examine whether T, and T, person-
ality factors are related to increased levels of drug involvement over time (T,
to T,). To determine this, the impact of T, and T, personality on T, drug
involvement will be measured controhng for T, drug involvement. By con-
trolling for drug history prior to and including T,, we can determine
7 14 J. S. BROOK, ET AL.

whether T , and T, personahty traits d o predict T, increased drug involve-

ment. More specifically, because we know that there has been an o v e r - d
increase in drug involvement from T, to T, (see means reported above), once
we have controlled for T, drug use, if the T, and T, personality scales are
significantly correlated with T, drug use in the expected direction, we can
assume that prior personality characteristics d o predict increased drug in-
volvement over time.

Antecedents of Increased Drug Involvement

To examine the relation between earlier (T, and T,) personality vari-
ables and greater involvement in drug use (T, to T,) we obtained correlations
of the persondty variables with T, drug use partialing T, drug use effects.
(We also partialed age effects because of the strong relationship between age
and drug use. I n our sample, socioeconomic status was not related to drug
use and sex was only weakly related at T, l.071, with males reporting slightly
more T, drug use than did females.) The results are presented in the first
correlational column of Table 1. I n the second column of Table 1, the par-
tial correlations of the personality variables and T, drug use with only age
partialed are presented so that the factors related to greater drug
involvement over time can be compared with the factors associated with
earlier levels of drug use. As shown in column one of the table, children (at
T,) who were unconventional (low achievement, less compliance, greater
nonconforming behavior), had difficulty in controlling their emotions (more
anger and temper tantrums), and who were aggressive toward their siblings
were more likely to increase drug involvement during adolescence. In con-
trast to these childhood predictors of greater involvement, the adolescent
factors were considerably stronger. Those adolescents who were unconven-
tional (tolerant of deviance, rebellious, not responsible, and deviant) were
likely to become more drug involved. The inability to control one's emotions
(greater anger, crying, impulsivity, and temper) was also of importance, but
less so than the presence of unconventional attitudes and behavior.
Intrapsychic distress (less ego integration and greater pathology) aIso re-
sulted in greater involvement in drug use. Finally, in the interpersonal
sphere, aggression toward peers and interpersonal difficulty led to greater
drug escalation.
We then compared the T, and T, factors related to increased drug use
with those related to earlier use (column one vs two, respectively, in Table
1). It should be noted that in terms of increased drug use, the causal priori-
ties of the variables are clear; that is, the T, and T, personality variables
have preceded T, drug use. With earlier usage, a reciprocal relationship
might have existed in that the ~ersonalityfactors could have led to greater
 CHANGE OVER TIME I N DRUG INVOLVEMENT 715

drug use and greater drug use could then have resulted in changes in person-
ality, especially in the direction of unconventionahty. Over-all, the majority
of the partial correlations between the childhood/adolescent personality attri-
butes and greater drug involvement (column one) were lower than the ones
between these attributes and earlier drug use (column two). I n particular,
the correlations between several unconventionahty attributes and greater
drug involvement were substantially lower than those between these same
measures and earlier usage (e.g., tolerance of deviance and rebelliousness).
One might conclude from these findings that the correlations of the uncon-
ventionality measures and earlier drug use may reflect some reciprocal
causation. It is also possible that unconventionality may play a more impor-
tant role in earlier use than it does later on. However, it is important to
note that dimensions of u n c o n ~ e n t i o n a h tstill
~ led to greater drug involve-
ment even though there was some drop in the magnitude of the correlations.
I n contrast to the unconventionality measures, the pattern of relations
between the variables assessing control of emotions, intrapsychic function-
ing, and interpersonal relatedness and greater drug involvement was similar
to that obtained for earlier drug use. Therefore, one might conclude that the
relation of these factors to drug use is less likely to be due to mutual reci-
procity.
I n sum, the findings show that, although the correlations between the
dimensions of unconventionality and increased drug use dropped in compari-
son to earlier use, the overwhelming majority were s t ~ Usignificant. Further,
the magnitude of the correlations between the control of emotion, intrapsy-
chic distress, and interpersonal relatedness variables, although quite modest,
was similar for both earlier and increased use.
Domain Interrelationships and Increased Drug Involvement
We hypothesized that the influence of the T, personality domain would
be mediated by the T, personality domain which, in turn, would be related
to increased drug involvement (a mediational model). To test this model,
hierarchical regressions were performed. Each personality domain was exam-
ined alone (bur with control on T, drug use and the demographic factors of
age, sex, and socioeconomic status)' and with partialing the effects of the
other personality domain (also with control on T, drug use and the demo-
graphic factors). We included in the regressions those T, and T, variables
that were significantly related to either T,or T, drug involvement; see Table
1. The dependent variable in these analyses was T, drug involvement. The
results are shown in Table 2 . As can be seen in this table, the T, personality
domain was related to greater drug involvement without but not with par-

'Separate dummy variables were created to control lor lour types of systematic assignment of
mean scores to missing data (Cohen & Cohen, 1983). All regressions were run with control on
these dummy variables.
716 J. S. BROOK, ET AL.

tiahng the T, domain. In contrast, the T, domain was significant both with-
out and with partialing the T, domain. These findings d o support a media-
tional model in that the T, domain was related to the T, domain which was
then associated with greater involvement in drug use. For comparison with
earlier drug use, we ran the same regressions with T, drug use as the depen-
dent variable, partialing only the demographic effects. A mediational model
was again observed (statistics not presented), indicating that the causal link-
ages among the domains were s i d a r for both greater drug involvement and
for earlier use. It should be noted that this replicated our previous findings
regarding T, and T, personality effects on T, drug use (Brook, et al., 1986).
We tested these effects again in the current study because this study differs
from the previous one in having an increased sample size and some variation
in the personality variables included in the T, and T, sets.
TABLE 2
MULTIPLE FOR TI AND TI ~ R S O N A L I T YDOMAINS
CORRELATIONS AND T, DRUG INVOLVEMENT
PART~ALTNGDF~OGRAPIUCS AND T, DRUG INVOLVEMENT
Domain Partialing Demographics Partialing Demographics, T, Drug
and T, Drug Involvement Involvement, and Other Domain
R F df R F d/
T, personality .14 2.00' 7,638 .06 .36 7,618
T, personality .33 3.75t 20,625 .31 3.19t 20,618
Note.-The variables within each domain are those significantly related to either T, or T,
drug involvement.
* p c . 0 5 . tp<.01.

To determine which specific variables explained the association be-

tween the childhood and adolescent personality sets, a canonical analysis was
run. The findings were similar to results reported in our previous paper on
earlier drug use (Brook, et al., 1986) and are presented in detail there. Here,
only the first three canonical correlations will be briefly summarized. (Seven
canonical correlations were significant in all.) The results of the first canoni-
cal correlation indicated that school-related achievement in childhood was
related to achievement during adolescence. The second canonical suggested
that emotional control during childhood was linked with similar behavior
during adolescence. Finally, the third canonical indicated that lack of social
inhibition, noncompliance, and problem behavior in childhood persisted into
adolescence.
Interactive Effects of T, and T, Risk and Protective Factors on
Increased Drug Involvement
When we examined T, levels of drug use in our earlier work (Brook, et
al., 1986), we found that drug use was affected by the interaction of risk
(drug-conducive) and protective (nondrug-conducive) personality traits.
 CHANGE OVER TIME IN DRUG INVOLVEMENT 717

Based on this research, we hypothesized that adolescent protective dimen-

sions (mainly reflecting conventionality) would serve to offset the effects of
childhood personality risk factors on increased drug involvement. To exam-
ine this risk/protective interactive mechanism, herarchical regression
analyses were computed. The significance of the interaction term of T, by T,
personality with T, drug involvement was determined after the main effects
of the two components of the interaction term and T, drug use were par-
tialed out. We also partialed the effects of a control ser of the demographic
variables. An example of a significant interaction is illustrated graphically in
Fig. 1. I n t h s graph, the slope of the line for adolescent achievement and

I TEMPER

Low High
(risk1 (prolsnive)
TI ACHIEVEMENT

FIG.1. Ulustrative example of a significant risk/proteccive interaction (T,high temper

ameliorated by T, high achievement)

increased drug involvement is greater for high childhood temper (risk) than
for low cMdhood temper (protective). Thus, school achievement during ado-
lescence appears to be capable of offsetting a T, personality risk factor. For
brevity, the remainder of these risk/protective interactions are presented in
tabular form; see Table 3. In a broad sense, the findings indicated quite
clearly that adolescent protective factors (particularly conventionality attri-
butes) were capable of weakening the effect of poor control of emotions and
unconventionalit.y during childhood leading to less involvement in drugs
over time. To ensure that these interactive effects were not just a chance
phenomenon, a regression was run in which a set consisting of the
significant interaction terms was examined partiding the main effects and T,
drug use. The interaction set was significant after partialing (F,,,,,= 2.00,
p < .01).
A related line of inquiry was to examine how personality traits inter-
acted with T, drug use to affect the adolescents' further involvement in drug
use over time. For these analyses, we examined, in separate regressions, the
718 J. S. BROOK, ET AL.

interaction term of each T, and T, personality trait by T, drug use. Again, T,

drug use was the dependent variable and there were controls on the main
effects and demographics. The results are presented in Table 4. Two types
of interactions emerged as significant: risklrisk in which personality risk

TABLE 3
~ R S O N A L I T Y INTERACTIONS
INVOLVING
AMELIORATIONOF CHILDHOOD (T,)
RJSKFACTORS
BY ADOLESCENT
PROTECTIVE (T,)
FACTORS
INLOWERDRUGINVOLVEMENT
RESULTING OVERT ~ (T,E TO T,)
T, Personality Risk Factor T, Protective Effect F,,m
Ameliorating T , Risk Factor For Two-way
Interaction
High aggression to siblings' Low rebelliousnessb
High anger Good school behavior'
High school achievementc
High temper Good school behavior'
High intolerance of devianceb
Low rebelliousnessb
Low impulsivityb
High school achievement'
Low deviancy'
Low social inhibition Low deviancy'
Low task persistence High educational goalsb
Low compliance Good school behavioi
High school achievementc
Low deviancyc
High nonconforming bellavior Low aggression to peers'
Low anger'
Good school behavior'
High educational goalsb
High school achievement'
Low achievement Low anger'
Note.-All interactions were examined partialing the effects of sex, age, socioeconomic status,
and T, drug involvement.
'Scales based on maternal report (all T, scales are maternal report). 'scales based on adoles-
cents' report. 'Scales based on combined maternal and adolescent reports.
*p<.05. t p c . 0 1 .

traits (e.g., temper tantrums, low achievement) were seen to exacerbate the
effect of high T, drug use leading to the greatest increase in drug involve-
ment (see Panel A of Table 4); and protectivelprotective, in which aspects
of adolescent conventionality (intolerance of deviance, Iow rebellion, etc.)
served to enhance the effect of low T, drug use leading to the least amount
of drug use at T, (see panel B of Table 4). Again, a regression was run ex-
amining a set of these interactions (partiding the main effects) to ensure
that the interactions were making a significant contribution. The interaction
set was significant after parti&ng (F,,6,, = 7.78, p < .01).
 CHANGE OVER TIME IN DRUG INVOLVEMENT 7 19
TABLE 4
INTERACTIONS INVOLVING T I A N D T2 PERSON~~~ITY
FACTORS BY T, DRUGUSE
AND THEIR EFFECTS ONT, CHANGES INDRUGINVOLVEMENT
A. Risklrisk interactions resulting in T, increased drug involvement
T, drug risk factor Penonality risk factor F1.64,
exacerbating T, drug risk
High levels of use T, high temper' 22.07t
T , low achievementb 8.97t
T, high temper' 18 507
T, high aggression to sibs' 6 48'
B. Protective/protective interactions resulting in T, decreased drug involvement
T, drug protective factor Personality protective factor
strengthening T, drug protecuon
Law levels of use T, high social inhib~tlon' 3.95'
T, high intolerance of deviance' 36.40t
T, low rebellion' 6.48*
T, high church attendancec 3.70'
T, low deviancyb 3.99'
Note.-Adl interactions were examined partialing the effects of sex, age, and socioeconomic
status.
Scales based on maternal report, bScales based on combined adolescent and maternal reports.
'Scales based on adolescents' report.
*p< .05. t p < .Ol.

DISCUSSION
In previous work, Brook, et al. (1986) reported findings regarding the
interrelationships and interactions of childhood and adolescent personality
traits in terms of their effects on adolescent drug use measured at just one
point in time (what we have deemed "earlier levels of drug use"). The cur-
rent study was undertaken to examine how childhood and adolescent traits
affect the adolescents' increased involvement in drug- use over time. These
results will now be discussed with a focus on comparisons between factors
related to earlier levels of drug use and those related to later levels.
The results of the current study support the view that childhood per-
sonality risk traits are associated with adolescent personality risks which, in
turn, lead to greater drug involvement-a mediational model. We should
emphasize that, although the childhood effects were of low magnitude and
were fully mediated by the variables in adolescence, these earlier effects
were still of importance. First, they were significantly related to drug use at
T,, even when the powerful impact of T, drug use was partialed. Second,
childhood personality formed the foundation for the subsequent adolescent
personality characteristics which then had a more direct influence on drug
involvement. I n brief, the childhood effects operated through the medium
of the adolescent personality. S i d a r mediational findings from our own
work (Brook, et al., 1986) and a related study by Block, et al. (1988) were
720 J. S. BROOK, ET AL.

reported for adolescent drug use measured at one point in time. The impor-
tance of personahty continuity is common to a wide variety of theoretical
perspectives regarding development, but one for which there has been little
specific empirical support. The studies just cited help to fill this gap by dem-
onstrating that childhood factors do lay the foundation for the development
of adolescent personality characteristics.
Over-all, the individual childhood personality attributes related to in-
creased drug involvement as well as to earlier involvement are similar to
adolescent characteristics previously implicated in drug use (e.g., Hawkins,
et al., 1985; Jessor, 1979; Kandel, 1980). More specifically, two types of
childhood variables appear to be of importance. First, there are the child-
hood precursors of unconventionality, as reflected by our measures of poor
achievement, less compliant behavior, and nonconforming behavior. Block
and his colleagues (1988) also reported that children who are less compliant
are more likely to use drugs during adolescence. A second type of childhood
variable associated with greater drug use reflects the emotional and impul-
sive life of the individual. Our childhood measures reflecting difficulty in
controlling one's emotions (e.g., temper and anger) and aggression to siblings
resulted in greater adolescent drug involvement. Consistent with these re-
sults, Block, et al. (1988) found that childhood ego under-control was
strongly associated with subsequent drug use.
The childhood personality qualities just described appear to affect drug
use in two ways. First they predispose the individual to initial levels of drug
involvement. Second, after the adolescents have begun drug use, they fur-
ther increase the probability of greater drug involvement over time. One
generalization that could be made from the present study and from the pre-
vious work of Block, et al. (1988) and Brook, et al. (1986) is that childhood
personality qualities are important not only for early levels of use but for in-
creased use as well.
Theories of drug use have long espoused the fundamental etiological
role of adolescent personality characteristics (Jersild, Brook, & Brook, 1978;
Lettieri, 1975). To hold that childhood personality traits are also of signifi-
cance in terms of drug use does not necessitate the claim that these
childhood traits act in any direct or automatic way. As our findings have
shown, their effects on drug use are mediated by adolescent variables. On
the other hand, to maintain that adolescent personality traits are fundamen-
taI to the etiology of adolescent drug use does not mean that cMdhood
plays no role in the process. Indeed, our canonical analyses clearly indicated
the ways in which individual childhood factors were related to specific ado-
lescent risk factors of drug use (both earlier use and increased involvement).
In brief, our canonical findings suggested that children who have school-
related difficulties, poor impulse control, and who are socially uninhibited,
 CHANGE OVER TIME IN DRUG INVOLVEMENT 72 1

noncompliant, and engage in nonconforming behavior are likely to exhibit

similar traits as adolescents. Despite this evidence of considerable personal-
ity stability over time (whch supports the results of a number of
investigators, e.g., Brim & Kagan, 1980; Brook, Gordon, & Whiteman,
1985; Kagan & Moss, 1962; Robins, 1974, 1978), there are ample data (our
own, as well as others') indicating that personality traits can be altered, per-
haps as a result of organism-environmental interaction factors (Rutter,
1982).
Having discussed the childhood personahty traits that lead to adoles-
cent drug use via adolescent personality traits, we now turn to a more
detailed description of the specific adolescent characteristics implicated in
increased drug involvement. To begin with, our findings showed that uncon-
ventional adolescent behavior (such as lack of responsibility, rebelliousness,
tolerance of deviance, and delinquency) predicted increased drug use. These
results are in accord with those of previous investigators (e.g., Newcomb &
Bentler, 1988) who also found longitudinal effects of socially distant atti-
tudes on changes in drug use. It may be that measures of unconventionality
assess a more general attitude of rebellion against conventional figures, sym-
bols, and values of society. In this sense, drug use is a specific manifestation
of a more general orientation. The fact that among the unconventionality
measures at T,, rebelliousness and delinquency showed the highest correla-
tions with increased drug use would be consistent with this viewpoint.
Our results further showed that aspects of adolescent intrapsychic dis-
tress, difficulty in controlling emotions, and difficulty in the interpersonal
sphere were all related to more drug involvement. According to Kaplan, et
al. (1986), increased psychological distress is associated with inadequate pat-
terns of coping and adaptation which increase the probabdity the individual
will "adopt alternative and deviant mechanisms." In line with this, Kaplan,
et al. (1986) found that psychological distress affected escalated drug use
and noted that these results are consistent with previous ones indicating
that substance abuse is used to cope with ongoing life stress or to "assuage"
the inner turmoil that is accompanied by these stresses (Duncan, 1977;
Khantzian, 1980; Mellinger, 1978; Wurmser, 1980). Kandel and her col-
leagues (Paton, Kessler, & Kandel, 1977) also reported that emotional
distress affected initial drug use.
As regards comparisons between adolescent personality characteristics
related to increased levels of drug involvement and those implicated in
earlier levels of use, our findings show that the correlations involving the
unconventionality measures were generally higher for earlier than for later
levels of drug use. One reason for this may be that the correlations between
unconventionality measures and earlier drug use may reflect to some extent
reciprocal causation; that is, T, drug use and T, unconventionality may both
722 J. S. BROOK, ET AL.

result from the same prior causal factors and, as such, would be highly cor-
related with each other. However, it is important to note that even after
minimizing the possibility of reciprocal causation by controlling for T,drug
use, several measures of unconventionality (poor school achievement, rebel-
liousness, delinquency, and irresponsibility) were still highly correlated with
T,drug involvement. An exception was tolerance of deviance, an attitudinal
measure, which served as a strong predictor of earlier drug use, but was only
minimally, although significantly, associated with TI drug escalation. I n sup-
port of this finding, Kaplan, et al. (1986) reported that conventional values
"serve to delay trying marijuana and, as such, are indirectly a social control
against any disposition to increase use" (p. 54).
There are other possible explanations as to why unconventionality was
found to be more powerful in predicting earlier as opposed to later drug in-
volvement. First, since unconventionality is related to drug use, by
partialing drug use effects at T,, one controls for that "portion" of uncon-
ventionality that is related to T,drug use. Therefore, only the residual
unconventionality (that which is not incorporated in T, drug use) is left to
predict T, drug use. I n brief, the relationship between T, unconventionahty
and T, drug use is stronger because none of the unconventionality variance
has been taken out as is the case when one controls for T, drug use before
examining the relationship between T, unconventionality and TI drug use. A
second possible explanation is that as one makes a commitment to drugs at
T, that lasts until T,, this begins to outweigh the effects of personality attri-
butes operative at T, that lead to T, drug use. More specifically, before a
commitment to drug use is made, unconventionality is important. Once one
uses drugs, unconventionality is less important and drug use takes on its
own influence.
Intrapsychic distress, difficulty in controlling one's emotions, and inter-
personal difficulties predicted greater involvement in drug use about as well
as they did earlier involvement. We alluded previously to the fact that ado-
lescents may use drugs to help alleviate emotional stress. Using drugs as a
form of self-medication would apply equally to the earlier phases of use as to
continued or increased use, assuming that drugs only temporarily mask emo-
tional pain. As the problems recur, so probably would the use of drugs to
deal with them.
I n sum, there appears to be ample evidence that similar childhood and
adolescent traits are of importance for both early and increased
levels of involvement in drug use and that stability in these personahty char-
acteristics exists from childhood to adolescence. These findings have
practical as well as theoretical implications. First, it appears that focusing on
certain crucial personality attributes may not only serve to inhibit initial
levels of drug use but may also help prevent the adolescent from increasing
 CHANGE OVER TIME IN DRUG INVOLVEMENT 723

his involvement over time. Second, it has been demonstrated that preven-
tion efforts begun in childhood would be of great value in that certain
childhood traits seem to be precursors of adolescent drug-risk characteristics.
Such preventive efforts might take the form of suggestions to parents (either
via the schools, health professionals such as pedatricians, or media cam-
paigns) to be especially attentive to young children who show signs of
having early drug-risk personalities (in particular, precursors of unconven-
tionality and problems with emotional or impulse control). If these signs
persist, parents should be encouraged to consult with school or health pro-
fessionals. In addition, the school and health professionals themselves might
recommend help for such children. Early intervention with very young chil-
dren might effect personality changes that would not only reduce the risks
of future drug use or abuse but should also foster general development.
Because school-based drug prevention programs have not proven to be very
effective in reducing drug use [see Bangert-Drowns (1988) and Tobler
(1986) for reviews], we feel the more direct and individual approach just de-
scribed might prove more helpful. However, extreme sensitivity must be
shown in targeting very young children for intervention so that parents are
not unduly alarmed and the children themselves are not made to feel they
have "problems." We should add further that, although the focus of this
paper has been on intrapersonal factors, our own work and that of others
(e.g., Block, et al., 1988; Brook, et al., 1983; Jessor, 1979) has shown that
the child's social milieu (e.g., farmly, peers, the school) must also be ad-
dressed in any drug prevention regimen. Finally, it would be of considerable
interest to conduct empirical tests by following children longitudinally to see
if alterations in precursors of drug use result in decreased drug use later on.
We would like to note that, in general, our results related to increased
drug involvement accounted for a relatively small portion of the variance.
However, as mentioned previously, these effects were measured after the
powerful impact of T, drug use was partialed. By itself, T, drug use ac-
counted for 42% of the variance in T, drug use. Over-all, the two
personality domains, T, drug use, and the demographic variables explained
fully 51% of the variance in T, drug use.
Another focus of the present study and one which, to our knowledge
has never been studied, was to examine whether protective personality traits
in adolescence could offset childhood risks in terms of drug involvement
over time. This type of "risk-protective" mechanism had been shown in
earlier work to be effective in lessening the likelihood of adolescent drug use
(Brook, et af., 1986). The present results, similar to the previous ones,
demonstrated that adolescent traits of conventionality were capable of weak-
ening the risk effects of poor control of emotions and unconventionality
during childhood leading to lower levels of drug involvement over time.
724 1. S. BROOK, ET AL.

This is a key developmental finding to emerge from this study and suggests
that protective adolescent factors are capable of offsetting childhood risk
not only in terms of early levels of drug use, but also with regard to decreas-
ing drug use once it has begun. The implication of these findings is that the
presence of childhood risk factors does not have to lead inexorably to high
levels of early drug use or to greater involvement over time if one can assist
the adolescent in developing protective personality traits, particularly those
adolescents with childhood vulnerabilities.
A second type of interaction was found to be of significance in the pres-
ent study, namely, the interaction of early levels of drug use with childhood
and adolescent personality traits. Two types of synergistic interactions
emerged. In the first, risk factors of temper and low achevement interacted
with high levels of drug use at T, leading to even greater drug involvement
at T,. In the second, adolescent low levels of early drug use in combination
with adolescent conventionality led to the least probability of increased in-
volvement in drugs. These results suggest that more sophisticated and
adequate accounts of change in drug use need to take cognizance of the role
of personality as a factor interacting with drug use. Personality factors do
influence the course of drug use, as the results of the present study docu-
ment. Not only do they have main effects, but interactive ones as well.
Clearly, the more subtle question on which efforts should focus is exactly
how and in what manner personality factors lead individuals to modify the
course of their drug-taking behavior once that behavior has been initiated.
To summarize, the findings of this study clearly demonstrate that cer-
tain childhood and adoIescent personality dispositions are implicated in
increased levels of drug use as they were in earlier levels. A number of com-
plex personality traits tapping unconventionality, poor control of emotions,
interpersonal difficulty, and intrapsychic distress were related to drug escala-
tion. It was also shown that childhood vulnerabilities could be mitigated by
protective adolescent personality traits leading to a decrease in drug escala-
tion, suggesting that childhood vulnerab5ty need not invariably result in
greater adolescent drug involvement. We further found that early drug use
interacted with adolescent risk traits (mainly unconventionality) leading to a
strong likelihood of increased drug involvement over time. These results sug-
gest that special attention should be paid to those adolescents who have
high levels of use and who also have an at-risk personality, as these young-
sters are in the most danger of further escalating their involvement.
In conclusion, research must concern further delineation of the causaI
function and the moderating circumstances for the influence of these
personality/behavioral variables if we are to increase our comprehension of
why some adolescents become seriously involved in drug use while others do
not.
726 J. S. BROOK, ET AL.

KAPLAN,H. B., ARTIN IN, S. S., JOHNSON,R. ., & ROBBINS,C. A. (1986) Escalation of mari-
h
juana use: application of a general t eory of deviant behavior. Journal of Health and
Social Behavior, 27, 44-61.
KELLAM,S. G., BRANCH,J. D., AGRAWAL, K. C., & ENSMINGER, M. E. (1975) Mentol health
and going to school: the Woodlawn program of assessment, early interuention, and eualua-
tion. Chicago, IL: Univer. of Chicago Press.
KELLAM, S. G., BROWN, C. H., RUBIN,B. R., & ENSMINGER, M. E. (1983) Paths leading to
sychotic symptoms and substance use: developmental epidemiological studies
F%:dPawn. I n S. B. Guze, E J. Earles, & J. E. Barrett (Eds.), Chikdhood psychopa-
thology and development. New York: Raven Press. Pp. 17-47.
KHANTZIAN, E. J. (1980) An ego/self theory of substance dependence: a concemporar psycho-
analytic perspective. In D. J. Lettieri, M. Sayers, & H . W. Pearson (Eds.), Tleories on
drug abuse: selected contemporary perspectiues. (Research Monograph No. 30, Pp. 29-33)
Rockville, MD: National Institute on Drug Abuse.
KOGAN, L. S., SMITH,/., & JENKINS, S. (1977) Ecological validity of indicator data as predic-
tors of survey mdmgs ]ourno1 of Social Service Research, 1, 117-132.
LABOWVIE,E . W., & MCGEE,C. R. (1986) Relation of person3l1ry to alcohol and drug use in
adolescence. Journal of Consulting and Clinical Psychology, 54, 289-293.
LETIIERI, D. J. (Ed.) (1975) Predicting adolescent drug abuse a review of issues, methods, and
correlates. (Research Issue No. 11) Rockville, MD: Nat~onalInstitute on Drug Abuse.
MELLINGER, G . D. (1978) Use of licit drugs and other coping alternatives: some ob-
servations on the hazards of living. In D. J. Lettieri (Ed.), Drugs and suicide: when other
coping strategiesJai1. Beverly Hills, CA: Sage. Pp. 249-278.
NEWCOMB, M. D., & BENTLER,l? M. (1988) The impact of family context, deviant attitudes,
and emotional distress on adolescent drug use: longitudmal latent-variable analyses of
mothers and their chddren. Journal of Research in Personality, 2 2 , 154-176.
PATON,S., KESSLER,R., & KANDEL, D. B. (1977) Depressive mood and illegal drug use: a lon-
gitudinal analysis. Journal of Genetic Psychology, 131, 267-289.
ROBINS, L. N. (1974) Deviant children grown up. Huntington, N Y : R. E. Krieger.
ROBINS,L. N. (1978) Sturdy childhood predictors of adult antisocial behavior: replications
from longitudinal studies. Psychological Medicine, 8, 611-622.
RUTER, M. (1982) Emotional development. In M. Rutter (Ed.), Scientific foundations of &el-
opmental psychiatry. London: Univer. Park Press. Pp. 306-321.
RUTTITER, M. (1982) Epidemiolo ical-longitudinal approaches to the study of development. In
W. A. Collins (Ed.), Tfe concept of development: the Minnesota Symposia on Child
Psychology: Vol. 15. Hillsdale, NJ: Erlbaum. Pp. 105-144.
STEM,J. A,, NEWCOMB, M. D., & BENTLER,P. M. (1986) StabiIi and change in personality:
a longitudinal study from early adolescence to young a d d x o o d . lourno1 of Rexearrh in
Personalie, 20, 276-291.
TOBLER,N. S. (1986) Meta-analysis of 143 adolescent drug prevention programs: quantitative
outcome results of program participants compared to a control or comparison group.
The Journal of Drug Issues, 16, 537-567.
WURMSER,L. (1980) Drug use as a protective system. I n D. J. Lettieri, M. Sayers, & H . W.
Peanon (Eds.), Theories on drug abuse: selected contemporary perspectiues. (Research
Monograph No. 30, Pp. 71-74) Rockville, MD: National Institute on D N Abuse. ~

Accepted October 23, 1989.

This article has been cited by:

1. Staci M. Zolkoski, Lyndal M. Bullock. 2012. Resilience in children and youth: A review. Children and
Youth Services Review 34, 2295-2303. [CrossRef]
2. Kristin L. Tomlinson, Sandra A. Brown. 2012. Self-medication or social learning? A comparison of
models to predict early adolescent drinking. Addictive Behaviors 37, 179-186. [CrossRef]
3. Jennifer A. Epstein, Heejung Bang, Gilbert J. Botvin. 2007. Which psychosocial factors moderate or
directly affect substance use among inner-city adolescents?. Addictive Behaviors 32, 700-713. [CrossRef]
4. Neo K. Morojele, Judith S. Brook. 2001. Adolescent Precursors of Intensity of Marijuana and Other
Illicit Drug Use Among Adult Initiators. The Journal of Genetic Psychology 162, 430-450. [CrossRef]
5. Joseph Biederman, Timothy E Wilens, Eric Mick, Stephen V Faraone, Thomas Spencer. 1998. Does
attention-deficit hyperactivity disorder impact the developmental course of drug and alcohol abuse and
dependence?. Biological Psychiatry 44, 269-273. [CrossRef]
6. Judith S. Brook, Martin Whiteman, Elinor B. Balka, Pe T. Win, Michal D. Gursen. 1998. Similar
and Different Precursors to Drug Use and Delinquency Among African Americans and Puerto Ricans.
The Journal of Genetic Psychology 159, 13-29. [CrossRef]
7. Judith S. Brook, Martin Whiteman, Elinor B. Balka, Patricia Cohen. 1997. Drug Use and Delinquency:
Shared and Unshared Risk Factors in African American and Puerto Rican Adolescents. The Journal
of Genetic Psychology 158, 25-39. [CrossRef]
8. Alfred S. Friedman, Arlene Terras, Cheryl Kreisher. 1995. Family and client characteristics as predictors
of outpatient treatment outcome for adolescent drug abusers. Journal of Substance Abuse 7, 345-356.
[CrossRef]
9. Thomas R. Zastowny, Edgar H. Adams, Gordon S. Black, Kirke B. Lawton, Anne L. Wilder. 1993.
Sociodemographic and Attitudinal Correlates of Alcohol and Other Drug Use Among Children and
Adolescents: Analysis of a Large-Scale Attitude Tracking Study. Journal of Psychoactive Drugs 25,
223-237. [CrossRef]
10. Donald M. Gallant, Rochelle Head-Dunham. 1991. ALCOHOL AND DRUG ABUSE
PREVENTION IN ADOLESCENTS. Alcoholism: Clinical and Experimental Research 15:10.1111/
acer.1991.15.issue-2, 308-308. [CrossRef]