The Impact of Parenting on Risk Cognitions and Risk Behavior: A Study of
Mediation and Moderation in a Panel of African American Adolescents
Michael J. Cleveland, Frederick X. Gibbons,
Meg Gerrard, and Elizabeth A. Pomery
Iowa State University
Gene H. Brody
University of Georgia
Hypotheses concerning the extent to which adolescents’ cognitions mediate the relation between parenting
behaviors and adolescent substance use were examined in a panel of African American adolescents (N5714, M
age at Time 1510.51 years) and their primary caregivers. A nested-model approach indicated that effective
parenting (i.e., monitoring of the child’s activities, communication about substances, and parental warmth) was
related to adolescent substance use more than 5 years later. The parenting behaviors protected the adolescent
from subsequent alcohol, tobacco, and marijuana use through associations with two cognitive elements from the
prototype/willingness model: favorable risk images (prototypes) and behavioral willingness. Additional ana-
lyses indicated that these protective effects were strongest among families residing in high-risk neighborhoods.
It is generally agreed that a wide variety of risk
factors contribute to adolescents’ vulnerability to use
alcohol, tobacco, and other drugs. Hawkins, Cata-
lano, and Miller (1992) identified 17 known risk
factors for adolescent substance use, which they
discussed in terms of three general categories: (a)
contextual factors, such as norms, substance avail-
ability, and neighborhood disorganization; (b) in-
dividual factors, such as prodrug attitudes and
physiological factors; and (c) interpersonal factors
arising from children’s interactions in family, school,
and peer environments. Few studies, however, have
included factors from each of these categories or
examined how they may interact in determining
adolescent risk behaviors.
Moreover, research suggests that the risk factors
described by Hawkins et al. (1992) may not apply
equally to adolescents from different race and ethnic
backgrounds. For example, there is evidence that
African American adolescents have stronger bonds
to family than do White adolescents (Giordano,
Cernkovich, & Demaris, 1993). Past studies have also
revealed differences among racial and ethnic groups
in family management styles, particularly in terms of
parental monitoring and control of peer selection
(Peterson, Hawkins, Abbot, & Catalano, 1994). This
research suggests that for African American youth,
the family environment may be more protective
than for White youth (Wallace & Muroff, 2002).
Such discrepancies have led some to conclude that
additional race-specific research is needed to identify
the protective factors that reduce substance use
among African American adolescents (Wallace &
Parental Influence on Adolescent Substance Use
Several studies have concluded that peer sociali-
zation factors are the strongest predictors of adoles-
cent substance use (Brook, Whiteman, Czeisler,
Shapiro, & Cohen, 1997). However, some authors
have noted that peer influences, relative to parental
influences, may be overestimated (Aseltine, 1995;
Bauman & Ennett, 1996; Kandel, 1996). In fact, many
researchers have focused on the indirect effects of
parenting style on adolescent outcomes. This re-
search has provided evidence that family process
factors play a central role in determining associations
with deviant peers, which in turn predict adolescent
risk behaviors (Dishion, Capaldi, Spracklen, & Li,
1995; Whitbeck, 1999).
A substantial body of literature also suggests that
parents’ behaviors may directly affect their chil-
dren’s risk behaviors (Blanton, Gibbons, Gerrard,
Conger, & Smith, 1997; Brown, Mounts, Lamborn, &
Steinberg, 1993). Adolescents raised by parents who
are heavily involved in their lives (i.e., monitor their
behavior) are less likely to engage in risk behavior
(Leventhal & Brooks-Gunn, 2000; Li, Stanton, &
r 2005 by the Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/2005/7604-0010
Michael J. Cleveland is now at Department of Psychology,
University of Massachusetts Dartmouth.
This research was supported by National Institute of Mental
Health (Grant MH62668).
Correspondence concerning this article should be addressed to
Michael J. Cleveland, Department of Psychology, University of
Massachusetts Dartmouth, 285 Old Westport Road, North Dart-
mouth, MA 02747. Electronic mail may be sent to mcleveland@
Child Development, July/August 2005, Volume 76, Number 4, Pages 900–916
Feigelman, 2000). Similarly, provision of warmth and
support by parents is associated with less adolescent
substance use (Barnes, Reifman, Farrell, & Dintcheff,
2000; Barnow, Schuckit, Lucht, John, & Freyberger,
2002). There is also some evidence that parent–child
communication about substances and substance
use is associated with reduced risk of early-onset
use (Chassin, Presson, Todd, Rose, & Sherman, 1998;
Jackson & Henriksen, 1997), although that evidence
is mixed (see Ennet, Bauman, Foshee, Pemberton, &
Parents can also indirectly affect their children’s
behavior by influencing the attitudes and cognitions
that their children develop about substance use and
substance users. For instance, Sieving, Maruyama,
Williams, and Perry (2000) reported that parents’
attitudes toward underage drinking were indirectly
related to their children’s alcohol use through their
association with the children’s alcohol-related cog-
nitions (i.e., intentions to drink, refusal efficacy, and
perceived importance of reasons not to drink).
Likewise, research has shown that the link between
parental alcohol-use norms and subsequent adoles-
cent alcohol use is mediated through the children’s
alcohol-use norms (Brody, Ge, Katz, & Arias, 2000)
and that frequent and bidirectional parent–child
discussions were associated with less liberal (i.e.,
abstinence-based) alcohol-use norms (Brody, Flor,
Hollett-Wright, & McCoy, 1998).
Few researchers, however, have examined the
extent to which children’s cognitions mediate the
relation between (effective) parenting behaviors and
children’s substance use. Consequently, several
adolescent-risk researchers have suggested that there
is a need for research that looks at cognitive media-
tion of the relation between parenting behaviors and
substance use (e.g., Beyth-Marom & Fischhoff, 1997;
Gochman, 1992). Such research is especially im-
portant in light of findings that suggest that risk
factorsFparticularly cognitive risk factorsFmay
play different roles for different ethnic groups
(Ellickson & Morton, 1999). The current study
addressed this empirical need by incorporating
elements from a recently developed model of health
risk behavior, the prototype/willingness (prototype)
The Prototype/Willingness Model
Cognitive mediation is the focus of the prototype
model of adolescent health behavior (Gibbons &
Gerrard, 1995) used in the current study. The model
is described in detail elsewhere (Gibbons & Gerrard,
1997; Gibbons, Gerrard, & Lane, 2003); a brief over-
view is presented here. The model is based on two
primary assumptions about adolescent risk beha-
vior: (a) it is largely a social activity and (b) it is
often a reaction to risk-conducive circumstances
rather than deliberative or planned. These two
assumptions are reflected in the two focal constructs
for the model: risk images (prototypes) and behav-
Young people have clear social images of the type
of person their age who engages in a particular risk
behavior (e.g., the ‘‘typical’’ smoker or drinker; cf.
Chassin, Presson, Sherman, Corty, & Olshavsky,
1981). They also realize that if they engage in the
behavior in public, others will tend to associate them
with the behavior and the image. In this sense,
the images are social consequences of the behaviors.
The more acceptable the image is to the adolescent,
the more willing he or she is to engage in the behav-
ior if the opportunity arises (Blanton et al., 1997;
Gibbons, Gerrard, & Boney-McCoy, 1995). Risk
images (or prototypes) have been shown to predict a
variety of health risk behaviors, such as sexual risk
taking, alcohol consumption, smoking, and reckless
driving (Gerrard, Gibbons, Benthin, & Hessling,
1996; Gibbons et al., 1995; Thornton, Gibbons, &
Gerrard, 2002; see Gibbons et al., 2003, for a review).
When asked, most adolescents say they do not
intend to engage in risky behavior (Brown, Di
Clemente, & Reynolds, 1991); nonetheless, as statis-
tics indicate, many end up doing so (Johnston,
O’Malley, & Bachman, 2000). The prototype model
explains this incongruence by proposing that there
are two pathways to risk behavior instead of the
single path found in most models of health and so-
cial behavior. One path is intentional; the other is not.
The former pathway proceeds to behavior through
behavioral intention (Ajzen, 1985, 1991). The second
path proceeds through an additional proximal ante-
cedent, unique to the prototype modelFbehavioral
willingness. Behavioral willingness is defined as an
openness to risk opportunityFwhat an adolescent
might do under certain circumstances. Behavioral
willingness adds to the amount of variance in ado-
lescent risk behavior that can be predicted by the
antecedent of behavioral intention (Gibbons et al.,
2003). To maximize predictive power, both behav-
ioral willingness and intention were included in the
current study in a construct labeled susceptibility.
Impact of Parenting 901
Other Factors That Influence Adolescent Substance Use
Another domain of risk factors includes in-
dividual or dispositional factors such as sensation
seeking (risk-taking tendencies), which have been
shown to predict adolescent substance use (Shen,
Locke-Wellman, & Hill, 2001; Wickrama, Conger,
Wallace, & Elder, 1999; Wills, Windle, & Clearly,
1998; Wills et al., 2001). It has been suggested that a
risk-taking tendency has a predisposing role for
substance use (Wills, Vaccaro, & McNamara, 1994).
Controlling for this trait provides some protection
against spurious results arising from a common
disposition among adolescents willing to engage in
risky behaviors, including substance use.
Hawkins et al. (1992) also described several con-
textual factors that influence adolescent substance
use. In fact, considerable attention has recently been
devoted to the impact of context (neighborhood) on
children’s health (Rankin & Quane, 2000). Generally,
this research has found that adolescents who live in
more disadvantaged (high-risk) neighborhoods tend
to fare less well than those who reside in more ad-
Gunn, 2000). For instance, neighborhood risk has
been associated with decreased academic perfor-
mance (Gonzales, Cauce, Friedman, & Mason, 1996),
affiliation with deviant peers (Brody et al., 2003),
delinquent behavior (Peeples & Loeber, 1994), and
higher rates of substance use (e.g., Brook, Nomura, &
Cohen, 1988; Smart, Adlaf, & Walsh, 1994). Most of
this research has been conducted among urban and
inner-city youth; however, less is known about how
neighborhood risk factors influence adolescent sub-
stance use in rural communities or suburban areas.
There is also evidence that neighborhood risk
factors may moderate the effects of parenting behav-
iors (e.g., monitoring and supervision of children) on
several adolescent outcomes (Rankin & Quane,
2002). For example, Gonzales et al. (1996) found that
higher levels of parental control were prospectively
associated with higher academic achievement (GPA)
in neighborhoods perceived by the adolescent as
high risk; a negative relation between maternal
control and GPA was found in low-risk neighbor-
hoods. Nonetheless, both studies were limited to
samples of African American children residing in
only large urban areas.
The Current Study
The current study is part of a larger project, the
Family and Community Health Study (FACHS),
which is examining the impact of environmental
factors on the mental and physical health of African
American families living in contexts other than inner
cities. Several cross-sectional studies have examined
the influence of neighborhood context on negative
adolescent outcomes using the FACHS panel. Brody
et al. (2001) found that children’s (but not care-
givers’) reports of nurturant-involved parenting
significantly interacted with a measure of neighbor-
hood disadvantage to predict their affiliation with
deviant peers. Brody et al. (2003) concluded that
caregivers’ reports of their own parenting behaviors
were associated with children’s conduct disorder
symptoms, and again, this relation was strongest
among families residing in high-risk neighborhoods.
However, Simons et al. (2002) found that the relation
between caregivers’ appraisals of their parenting
behaviors and child conduct problems tended to
decrease as the level of community deviance, as
perceived by caregivers, increased.
Thus, although it is clear that a warm and in-
volved parenting style is associated with healthy
youth development, current knowledge about how
these processes unfold over time is limited, espe-
cially among nonurban African American adoles-
cents. Some clarification of this issue has been
provided by other research conducted with the
FACHS sample. For example, Gerrard, Gibbons,
Stock, Vande Lune, and Cleveland (2005) found that
parental behaviors were associated with risk images
and behavioral willingness, which predicted initia-
tion of smoking. However, these findings were lim-
ited because parenting, risk images, and behavioral
willingness were measured concurrently, constrain-
ing the interpretation of effective parenting as ante-
cedent to risk cognitions. The present study used
three waves of data to address this need.
The present research had two related objectives.
The primary goal was to determine the extent to
which cognitive factors mediated the relation be-
tween parenting behaviors and substance use among
African American adolescents, a population for
whom little evidence is available (Bachman et al.,
1991; Vaccaro & Wills, 1998). Our model is presented
in Figure 1. In addition to a direct influence, par-
enting behaviors at Time 1 were also hypothesized to
have indirect effects on Time 3 adolescent substance
use by way of associations with Time 2 risk images
and susceptibility. Because there is substantial
evidence that parenting behaviors influence adoles-
the model also allowed for indirect effects of par-
enting on Time 3 use through its association
with Time 2 friends’ use. The theoretical model
controlled for individual factors (risk-taking ten-
dencies, gender, and age) and contextual factors
902Cleveland, Gibbons, Gerrard, Pomery, and Brody
(parent substance use and neighborhood risk) as
well as the targets’ Time 1 use.
A second goal of this study was to investigate how
the relations among parenting behaviors, friends’
use, and risk cognitions may differ according to
neighborhood context. We hypothesized that effec-
tive parenting would have a greater impact on ado-
lescents’ risk cognitions in high-risk than in low-risk
A total of 897 families, 475 in Iowa and 422 in
Georgia, were recruited for participation in the
FACHS. Each family had an African American fifth-
grade target child ages 10 (52%), 11 (45%), or 12 years
old (3%) at Time 1. Slightly more than half (54%) of
the targets were female. Separate but concurrent in-
terviews were given to the target and his or her pri-
mary caregiver, defined as a person living in the same
household who was primarily responsible for the
target’s care. Of the 897 families, 779 (87%) remained
in the panel at Time 2, and 767 (86%) remained in the
panel at Time 3. The present analyses include the 714
families that were present at all three waves.
Of the 714 primary caregivers (parents), most
(84%) were the targets’ biological mothers (37% of
whom were married at Time 1); the rest were biolo-
gical fathers (5%), grandmothers (6%), or someone
else (5%). Almost all (91%) of the parents identified
themselves as African American; their mean age at
Time 1 was 37 (range523 to 80) and their educa-
tional backgrounds were diverse, ranging from less
than a high school diploma (19%) to a bachelor’s or
advanced degree (10%).
Sampling Strategy, Recruitment, and Interview
Families were recruited for FACHS from multiple
sites that varied considerably on demographic
characteristics, such as racial composition and eco-
nomic level. Sites included rural farm communities,
suburban areas, and small metropolitan areas; there
were no inner-city regions. Particular attention was
paid to sampling families from neighborhoods with
varying racial composition (e.g., percentage African
American) and economic level (percentage of fam-
ilies with children living below the poverty line).
Potential participants were chosen randomly from
lists of families living in neighborhoods with at least
10% African American population. The lists, com-
piled by community liaisons around Athens, Geor-
gia, and school officials in Des Moines and Waterloo,
Iowa, included all families with a 10-year-old or
fifth-grade African American child. The families re-
ceived an introductory letter, followed by a recruit-
T1 Sub Use
S-B χ2(160) = 304.56 (p < .001)
RMSEA = .04; 90% CI = [.03 − .04]
CFI = .97
N = 714
Figure1. Structural model estimated using LISREL 8.70 with robust standard errors. Coefficients represent completely standardized
maximum likelihood estimates; significant paths are in bold. Nghd risk5neighborhood risk; S–B w25Satorra–Bentler scaled chi-square
statistic; RMSEA5root mean square error of approximation; CI5confidence interval; CFI5comparative fit index.?po.05.??po.01.
Impact of Parenting 903
ment phone call and then a personal visit requesting
that the target child and his or her parent participate
in the study. In case a telephone was not available,
the letter included a toll-free number. Complete data
were gathered from 72% of the families on the re-
cruitment lists. The majority who declined cited the
amount of time the interview took as the reason not
to participate (for further description of the FACHS
sample and its recruitment, see Brody et al., 2001;
Cutrona, Russell, Hessling, Brown, & Murry, 2000;
Gibbons, Gerrard, Cleveland, Wills, & Brody, 2004;
Simons et al., 2002; Wills, Gibbons, Gerrard, & Brody,
2000). In general, the sample was representative of
the African American populations in the commu-
nities from which they were selected.
All interviewers were African American; most
resided in the communities where the study took
place. Interviews were conducted in participants’
homes, or in locations near their homes (e.g., a li-
brary or school). The interview required two sepa-
rate visits with two interviewers and lasted 90 min,
on average. It included a computer-assisted personal
interview (CAPI) and a structured psychiatric diag-
nostic assessment. In the CAPI, questions appeared
on the computer screen and, when necessary, were
read aloud to the participant. Parents received $100
and targets received $70 for their participation. The
second wave of data collection occurred approxima-
tely 2 years after the first wave (M525 months), and
the third wave of data collection occurred slightly
more than 3 years after the second (M538 months).
The instruments and procedures were exactly the
same at Times 1 and 2, but changes were im-
plemented at Time 3. Parents and targets were both
interviewed at Time 3 in one visit rather than two.
The parents also completed a pencil-and-paper
questionnaire that was mailed to them in advance
and turned it in to the interviewers at the time of the
home visit. In addition, the interviews and ques-
tionnaires were revised, discarding many old ques-
tions and adding new ones to reflect an increased
maturity of the targets. At Time 3, the targets were
given an accessory keypad to enter their own respon-
ses to sensitive questions. The participation stipend
at Time 3 remained the same for parents but increa-
sed to $80 for targets.
Most of the measures were adapted from previous
research with families and older adolescents (Ger-
rard et al., 1996; Gibbons, Gerrard, Blanton, & Rus-
sell, 1998). The variables were grouped together
according to four domains: control, parenting, med-
iating, and outcome. Parenting and control variables
were all measured at Time 1. The mediating vari-
ables (friends’ use, risk images, and susceptibility)
were measured at Time 2. The outcome variable (tar-
gets’ substance use) was measured at Times 1 and 3.
Parent substance use. The parent interview in-
cluded two measures of tobacco use (e.g., ‘‘On
average, how many cigarettes, cigars, pipes of to-
bacco do you usually smoke per day?’’). Parents also
completed the University of Michigan Composite
Kessler et al., 1994), which included three measures
of alcohol use (e.g., ‘‘In the past 12 months, did you
have at least 12 drinks of any kind?’’) and six mea-
sures of problematic use (e.g., ‘‘Have you ever been
arrested for DWI?’’). Parents were also provided
with a list of 21 drugs, plus a general use item (‘‘any
other drugs’’) and asked to indicate any that they
had used more than five times. The 33 items were
standardized and combined into a single index
Neighborhood context. Targets completed a six-item
neighborhood risk scale, which assessed the fre-
quency with which various acts (e.g., fights with
weapons, violent arguments between neighbors)
occurred in their neighborhood. The response format
for the items was a 3-point scale ranging from 1
(never) to 3 (often; a5.74). Parents completed a sev-
en-item community disorganization scale (adapted
from Sampson, Raudenbush, & Earls, 1997), which
asked the extent to which various indicators of
neighborhood problems (e.g., drinking in public,
people selling or using drugs, groups hanging out
and causing trouble) occurred in their neighbor-
hoods. The response format for these items was also
a 3-point scale ranging from 1 (not at all a problem) to
3 (a big problem; a5.90). The two scales were com-
bined into a single indicator by averaging the mean
of each scale.
scale, adapted from Eysenck and Eysenck’s (1977)
inventory, included items such as ‘‘You enjoy taking
risks’’ and ‘‘You would prefer doing something
dangerous rather than sitting quietly,’’ each accom-
panied by a 3-point scale ranging from 1 (not at all
true) to 3 (very true). The six items were combined to
create a single indicator of targets’ risk-taking ten-
904 Cleveland, Gibbons, Gerrard, Pomery, and Brody
Additional control variables. Two other control
variables were also included in the structural equa-
tion model (SEM)Ftargets’
15female) and age at Time 1 (10, 11, or 12 years).
Effective parenting. Separate subscales assessed the
targets’ perceptions of three aspects of effective
parenting at Time 1. Previous research has shown
that these measures correlate with observer ratings
of parental behavior and predict a variety of child
adjustment problems (Conger, Elder, Lorenz, Si-
mons, & Whitbeck, 1992; Simons, 1996). Monitoring
was assessed with five items (e.g., ‘‘How often does
your [parent] know what you do after school?’’),
each followed by a 4-point scale ranging from 1
(never) to 4 (always; a5.61). The communication
subscale contained three items that assessed ado-
lescents’ perceptions of the extent to which their
parents communicated with them about alcohol, cig-
arette, and marijuana use, each followed by a 4-point
scale ranging from 1 (never) to 4 (many times; a5.90).
The warmth measure included nine items (e.g.,
‘‘How often in the last 12 months did your [care-
giver] let you know she really cares about you?’’),
each followed by a 4-point scale ranging from 1
(never) to (always; a5.82). The three subscales were
used as indicators of the parenting latent construct
Friends’ substance use. Targets’ perceptions of their
friends’ use were assessed using the stem, ‘‘During
the past 12 months, how many of your close friends
. . . ’’ followed by a list of eight substances, each with
a 3-point scale ranging from 1 (none of them) to 3 (all of
them). Friends’ use of alcohol was indexed by com-
bining the items: ‘‘used alcohol’’ and ‘‘drunk a lot of
alcohol.’’ An index of friends’ other drug use was
created by combining items for illegal drugs, prescrip-
tion drugs, inhalants, nonprescription drugs, and other
drugs. These two indexes and the item ‘‘used tobacco’’
were used as three indicators of the friends’ substance
use latent construct (overall a5.82).
Risk images. To assess risk images, the child was
first asked to think about three specific prototypes,
presented separately in the following manner: ‘‘Take
a moment to think about the type of kid your age
who . . . ’’ ‘‘smokes cigarettes,’’ ‘‘frequently drinks
alcohol,’’ and ‘‘uses drugs.’’ In each case, targets
described their image of the person (i.e., image fa-
vorability) using six adjectives: popular, smart, cool,
good looking, childish, and dull (the last two reversed
so that higher scores indicated a more favorable
image), each accompanied by a 4-point scale ranging
from 1 (not at all) to 4 (very; Gibbons & Gerrard,
1995). The three prototypes were assessed separately
and then used as three indicators of the risk image
latent construct in the SEM (overall a5.89).
Susceptibility to use. Targets’ willingness was
measured with a pair of items for each substance,
worded as in previous studies (Gibbons et al., 1998).
The section began with a description of a hypothe-
tical scenario: ‘‘Suppose you were with a group of
friends and there were some [cigarettes/alcohol/
drugs] there that you could have if you wanted.’’
This statement was followed by a light and a heavier
use question (e.g., ‘‘How willing would you be to
take some and use it?’’ and ‘‘How willing would you
be to use enough to get high?’’ or ‘‘How willing
would you be to drink one drink?’’ and ‘‘How will-
ing would you be to drink more than one drink?’’),
each accompanied by a 3-point scale ranging from 1
(not at all) to 3 (very willing). Susceptibility also in-
cluded two measures of intentions for each sub-
stance: one a straight intention item and the other an
expectation or likelihood item (Warshaw & Davis,
1985). For smoking and drugs, the straight intention
items were: ‘‘Do you plan to [smoke cigarettes/use
drugs] in the next year?’’ followed by a 4-point scale
ranging from 1 (do not) to 4 (do plan to). The ex-
pectation items read: ‘‘How likely is it that you will
[smoke cigarettes/use drugs] in the next year?’’ fol-
lowed by a 4-point scale ranging from 1 (definitely
will not) to 4 (definitely will). For each of the drug
measures, a fifth option, ‘‘don’t know,’’ was added in
the middle slot. For alcohol, the two questions were:
‘‘How much alcohol [do you plan/are you likely] to
drink in the next year?’’ followed by a 3-point scale
ranging from 1 (none) to 3 (3 or more drinks at one
time). The tobacco and drug options were then col-
lapsed into three categories, using the following
code: 15do not, 25probably won’t, 35probably
will and plan to; for drugs, the ‘‘don’t know’’ option
was also included in the second category. Thus, as
with willingness, there were two 3-point scales (in
essence, no, maybe, and yes) for each of the three
substances for intention and expectation. The will-
ingness, intention, and expectation scales were then
combined for each of the three substances and used
as indicators of the latent susceptibility construct
At Time 1, targets were presented with compu-
terized versions of the Diagnostic Interview Sche-
Impact of Parenting905
dule for Children, Version 4 (DISC–IV; Shaffer et al.,
1993), which included dichotomous (05no, 15yes)
measures for lifetime and past year use of alcohol,
tobacco, and marijuana. The sum of the three lifetime
use items was used to create an index of Time 1
substance use (range50 to 3). The target interview
at Time 3 included frequency measures of alcohol,
tobacco, and marijuana use. Past year use of alcohol
and marijuana were measured by asking, ‘‘During
the past 12 months, how often have you [had a lot to
drink, that is 3 or more drinks at one time/used
marijuana to get high]?’’ Each of these items was
measured on a 6-point scale ranging from 1 (never) to
6 (several times per week [alcohol]) or 6 (more than once
a week [marijuana]). Tobacco use in the previous 3
months was measured by asking, ‘‘How many cig-
arettes have you smoked in the last 3 months?’’ This
item was measured on a 5-point scale ranging from
1 (I have not smoked in the last 3 months) to 5 (I
have smoked every day). The three items were used as
indicators of the latent Time 3 substance use con-
Four groups were created to test for differences
among those who were present at different waves.
Group 1 (n565) was present at only Time 1, Group 2
(n565) was present at Times 1 and 2, Group 3
(n553) was present at Times 1 and 3, and Group 4
(n5714) was present at Times 1, 2, and 3. Univariate
analyses of variance (ANOVAs), conducted on each
of the items used in the SEM, found no significant
differences among the groups. Next, a second set of
attrition analyses was conducted by collapsing the
first three groups and comparing the families who
were present at all three waves with those present for
only one or two waves. Independent samples t tests
indicated that compared with targets with complete
data, attriters reported lower levels of target mar-
ijuana prototypes, t(895)5 ?2.07, p5.039; target
susceptibility, t(895)5 ?2.16, p5.031; and lower
t(895)5 ?2.00, p5.045.
Table 1 presents the frequencies of targets re-
porting lifetime and past year use of each of the three
substances (at Time 1) alone and in combination. At
Time 1, 91% reported that they had never used any of
the three substances and 98% reported that they had
not used any substance in the past year. Alcohol was
the most commonly reported substance used. At
Time 3, the number reporting use increased to 18%
(alcohol or marijuana use in the previous year), and
12% reported tobacco use in the previous 3 months.
Frequent (more than once a week) use of marijuana
and tobacco was reported by 4% of the targets; 1%
reported using alcohol at least weekly.
Univariate ANOVAs were also performed to test
for mean differences of the study variables by gender
and neighborhood condition (using a median split).
Compared with females, males reported higher risk-
taking tendencies, F(1, 710)57.07, p5.008, and af-
filiations with substance-using peers at Time 1, F(1,
710)55.56, p5.019. At Time 2, female targets
reported more favorable risk images, F(1, 710)57.30,
p5.007, and greater susceptibility to use, F(1, 710)
58.94, p5.003. Targets residing in high-risk neigh-
borhoods were younger than their counterparts, F(1,
708)54.39, p5.036; and reported higher levels of
risk-taking tendencies, F(1, 710)58.13, p5.004; and
lower levels of effective parenting, F(1, 710)56.27,
p5.012. Targets in high-risk neighborhoods also re-
ported greater affiliations with substance-using
peers, more susceptibility to use, and more favorable
risk images, all Fs(1, 710)410.00, pso.001. At Time 2,
targets in high-risk neighborhoods reported higher
levels of friends’ use, F(1, 710)55.89, p5.016, and
more favorable risk images, F(1, 710)52.96, p5.086.
No significant Gender ? Neighborhood interac-
tions were found.
SEM: Plan of Analysis
SEM, using LISREL 8.70 (Jo ¨reskog & So ¨rbom,
2004a), was used to estimate the hypothesized model.
First, a confirmatory factor analysis (CFA) was
conducted to determine whether the observed mea-
sures loaded on the latent constructs as hypothe-
sized. We then compared our theoretical model with
three alternative models. The first model estimated
only a direct path from parenting behaviors to ado-
lescent substance use. All other paths in the baseline
model were constrained to zero, suggesting that the
Frequency of Respondents Reporting Lifetime and Past Year Use of Al-
cohol, Cigarettes, and Marijuana at Time 1
Substance usedEverPast year
One substance only
All three substances
906 Cleveland, Gibbons, Gerrard, Pomery, and Brody
influence of parenting on adolescent use was entirely
direct. The second model retained the direct path
from the baseline model and added the indirect in-
fluence of parenting on adolescent substance use
through its associations with friends’ use. The next
alternative model also retained the direct path from
the baseline model but allowed for indirect effects
only through the elements of the prototype model-
Frisk images and susceptibility. The final step in the
nested model approach represented our hypothe-
sized model. This model allowed for direct effects
of parenting in addition to both indirect pathwaysF
through friends’ use as well as elements of the pro-
totype model. We hypothesized that this model
would provide the best fit to the data, confirming
that the effects of parenting behaviors on adolescent
substance use were mediated through the elements
of the prototype model in addition to peer use.
Using normal-theory estimation techniques such
as maximum likelihood (ML) with non-normal data
may lead to overestimation of the chi-square good-
ness-of-fit test while underestimating standard er-
rors and indexes of fit (West, Finch, & Curran, 1995).
Thus, the data were examined for univariate and
multivariate normality using PRELIS 2.70 (Jo ¨reskog
& So ¨rbom, 2004b). Several measured items exhibited
violations of univariate normality (skewness43,
kurtosis410; Kline, 1998). Hence, the fit of all SEMs
was assessed using the Satorra–Bentler scaled chi-
square statistic (S–B w2; Satorra & Bentler, 1988) and
robust (adjusted for multivariate nonnormality)
standard errors. Nested models were compared by
using a difference test scaling correction, calculated
from the ratio of the normal theory to S–B w2test
statistics (Satorra & Bentler, 2001). A significant re-
duction in chi-square, relative to change in degrees
of freedom, indicates that the less constrained model
provides a better fit to the data than the more con-
strained model. Model fit was also evaluated with
the comparative fit index (CFI; Bentler, 1990) and the
root mean square error of approximation (RMSEA;
Browne & Cudek, 1993). CFI values greater than .90
and RMSEAvalues less than .05 indicate good fit. We
also calculated 90% confidence intervals (CI) for the
RMSEAvalues. Ideally, the lower limit of the 90% CI
includes or is very near zero and the upper limit is
less than .08.
SEM: The Measurement Model
All constructs were specified as latent with mul-
tiple indicators except the control variables, which
were manifest. The CFA was estimated using ML in
LISREL 8.70 with the asymptotic covariance matrix;
thus, robust standard errors and the S–B w2were
provided. The measurement model provided a good
fit to the data, as indicated by the fit indexes
(CFI5.98, RMSEA5.04, 90% CI5.03–.04) even
though the S–B w2test statistic was significant, S–B
w2(140, N5714)5266.32, po.001. All factor loadings
were significant; standardized values were ? .44
with one exceptionFthe communication subscale of
the parenting latent construct (l5.24). Table 2
presents the correlation matrix as well as means,
standard deviations, skewness, kurtosis, and stan-
dardized factor loadings of the items used in the SEM.
SEM: Hierarchical Testing of Alternative Models
Direct Effects Model
The first step in the hierarchical testing was to fit a
model that specified only a direct path from effective
parenting behaviors to Time 3 adolescent use. All
other paths in the model were constrained to zero.
The fit of this model (see Table 3) was adequate
(CFI5.92, RMSEA5.06, 90% CI5.05–.06), but the
relatively high chi-square value indicated that there
N5714)5587.87, po.001. The direct effect of par-
enting behaviors on Time 3 use in the model was
significant (b5 ?.19, t5 ?2.34, p5.020).
Friends’ Use-Only Model
Next, we fit a model that retained the direct path
from effective parenting to Time 3 use and added
additional paths from parenting to Time 2 friends’
use and from Time 2 friends’ use to Time 3 use. The
fit of this model was also adequate (CFI5.93,
RMSEA5.06, 90% CI5.05–.06) but again resulted
in a relatively high chi-square value, S–B w2(166,
N5714)5530.94, po.001. In this model, the direct
path from effective parenting to Time 2 friends’ use
was significant (b5 ?.13, t5 ?1.94, p5.053), and
Time 2 friends’ use significantly predicted Time 3 use
(b5.28, t53.54, po.001). Adding Time 2 friends’
use as an intervening variable resulted in a slight
decrease of the direct effect of parenting at Time 3
(b5 ?.13, t5 ?1.91, p5.057).
In the third step we fit a model that again retained
the direct path from effective parenting to Time 3 use
but added paths from parenting to Time 2 risk ima-
ges and susceptibility, Time 2 risk images to Time 2
susceptibility, and from Time 2 susceptibility to Time
Impact of Parenting907
Correlations, Means, Standard Deviations, Skewness, Kurtosis, and Standardized Factor Loadings for the Measurement Model
T1 Par Usea
T1 Sub use
T2 Image alc
T2 Image mar
T2 Image tob
T2 Susc alc
T2 Susc mar
T2 Susc tob
T2 Frd alc
T2 Frd mar
T2 Frd tob
T3 Use alc
T3 Use mar
T3 Use tob
Note. N5714. All variables coded such that high scores indicate more of the construct. T15Time 1; par use5parent self-report of substance use; nghd5parents’ and targets’ reports of
neighborhood risk; risk5targets’ self-report of risk-taking tendencies; gender50 refers to male, 1 refers to female; sub use5targets’ self-report of alcohol, tobacco, and marijuana use;
comm5targets’ reports communication with parents; monitor5targets’ report of parental monitoring; warmth5targets’ report of parental warmth; alc5alcohol; mar5marijuana;
tob5tobacco; image5risk image; susc5susceptibility to use; frd5targets’ reports of friends’ use.
aStandardized. All correlations ? .07, po.05; ? .10, po.01; ? .12, po.001.
908Cleveland, Gibbons, Gerrard, Pomery, and Brody
3 use. The fit of this model was slightly improved
(CFI5.95, RMSEA5.05, 90% CI5.04–.05) but once
more resulted in a relatively high chi-square value,
S–B w2(164, N5714)5426.16, po.001. In this model,
parenting had significant direct effects on both Time
2 risk images (b5 ?.14, t5 ?2.49, p5.013) and
susceptibility (b5 ?.20, t5 ?2.80, p5.005). In ad-
dition, risk images significantly predicted suscept-
ibility (b5.41, t56.24, po.001), which significantly
predicted Time 3 use (b5.23, t52.54, p5.011).
Adding the elements of the prototype model as in-
tervening variables resulted in the direct effect of
parenting on Time 3 again decreasing (b5 ?.11,
t5 ?1.67, p5.095).
The final step consisted of a model that allowed
the influence of parenting on Time 3 use to include
direct effects, and indirect effects through friends’
use as well as the risk cognitions. The fit of this
model was good (CFI5.97, RMSEA5.04, 90%
CI5.03–.04), although the chi-square value re-
mained significant, S–B w2(160, N5714)5304.56,
po.001. The hypothesized model represented a
better fit to the data when compared with either
the friends’ use-only model, DS–B w2(6)5157.79,
po.001, or the prototype-only model, DS–B w2(4)5
41.29, po.001. The fit of the hypothesized model was
also compared with the direct effects model. S–B w2
difference tests showed a dramatic improvement in
fit for the more constrained model, DS–B w2(8)5
As seen in Figure 1, parenting continued to have
(b5 ?.14, t5 ?2.48, p5.013) and susceptibility
(b5 ?.20, t5 ?2.85, p5.005). However, the direct
effect of parenting on friends’ use was now non-
significant, as was the effect of susceptibility on Time
on both riskimages
3 use. Targets’ risk images were associated with
vulnerability (b5.40, t56.47, po.001), which sig-
nificantly predicted friends’ use (b5.55, t57.49,
po.001). Compared with the friends’ use-only
model, the direct effect of friends’ use on Time 3 use
decreased slightly but remained significant (b5.22,
t52.39, p5.017). The direct effect of parenting on
Time 3 use was not significant; however, parenting
did have significant indirect effects on both friends’
use (b5 ?.15, t5 ?3.33, po.001) and T3 use (b5
?.06, t5 ?1.98, p5.048). Together, the direct and
indirect effects of parenting behaviors explained 11%
of the variance in Time 3 adolescent substance use.
Neighborhood and Parenting Interaction
The next step in our analyses concerned the hy-
pothesis that effective parenting would have a
greater impact on adolescents’ risk cognitions in
high-risk than in low-risk environments. Some re-
search has demonstrated that the S–B w2difference
test may not perform properly unless the sample size
is large, especially if high kurtosis is present in the
data (West et al., 1995). These conditions were pres-
ent in our data and precluded us from using multi-
sample SEM analyses to test the moderation hy-
pothesis. However, other empirical evidence has
shown that ordinary least squares regression has
more power to detect statistical interaction than
multisample SEM (e.g., Stone-Romero & Anderson,
1994). Thus, the anticipated moderating effects of
neighborhood context on parenting were examined
by conducting a series of hierarchical regressions
predicting risk images, susceptibility, friends’ use,
and targets’ own use.
In each case, the parenting and neighborhood risk
variables were centered, and an interaction term was
created by multiplying the two centered variables.
All three terms were then sequentially entered as
Chi-Square Statistics, Fit Indexes, and Direct Effects of Parenting for Alternative Models
Direct effect Variance explained
Direct effects only
Friends’ use only
Note. N5714. The Satorra–Bentler scaled chi-square statistic (S–B w2) was used to evaluate model fit. However, the normal chi-square (w2)
is also required when testing for differences between two nested models. Therefore, both are presented in the table. e5root mean square
error of approximation (RMSEA) point estimate; CI590% confidence interval for the RMSEAvalue. Direct effect values represent t values
of the direct effect of parenting on Time 3 adolescent substance use. Variance explained refers to the percentage of variance in Time 3
adolescent substance by the variables in the model.
Impact of Parenting909
predictors in separate models for the four outcome
variables at Times 1 and 2 (Time 3 for target use). As
seen in Table 4, effective parenting was a significant
predictor for all of the mediating variables (including
changes in risk images and susceptibility) and Time
1 use. Neighborhood risk had significant effects on
two of the outcomes: risk images and friends’ use.
The interaction term was significant for four of the
outcome variables: susceptibility, friends’ use, tar-
gets’ use at Time 1, and risk images at Time 2. In all
cases, the pattern was as expected: Targets’ were
more likely to report higher levels of risk cogni-
tions or behaviors if they lived in a high-risk neigh-
borhood and reported lower levels of effective
parenting. Figure 2 represents an example of the
interactions and displays the results of the regression
predicting Time 2 risk images, following the proce-
dures outlined by Aiken and West (1991). As seen in
the figure, the dashed line, which represents the re-
lationship between effective parenting and Time 2
risk images for adolescents residing in high-risk
neighborhoods (1 SD above the mean), has a steeper
slope than the solid line, which represents adoles-
cents residing in low-risk neighborhoods (1 SD be-
low the mean).
The current data provide evidence that parent-
ing behaviors (communication, monitoring, and
warmth) were associated with African American
adolescents’ substance use that occurred, on average,
5 years later and that this influence was primarily
indirect. Through a nested models approach, sup-
port for the mediation of effective parenting beha-
viors on subsequent use through both the child’s
peer associations and cognitions was noteworthy.
Our hypothesized model, which included indirect
effects through both of these pathways, resulted in
the direct effect of Time 1 parenting on Time 3 ado-
lescent use becoming nonsignificant. However, par-
enting was predictive of both elements of the
Effective Parenting and Neighborhood Risk as Predictors of Target Risk Images, Susceptibility, and Friends’ Use
Parenting NeighborhoodP ? N
Note. N5714. Parenting5Time 1 effective parenting behaviors; neighborhood5Time 1 neighborhood risk; P ? N5interaction term of
effective parenting by neighborhood risk; ?5standardized beta coefficient.
aValue remained significant (po.08) after controlling for outcome variable at Time 1.
T2 Risk image
Figure2. Graph depicting the association between effective par-
enting and Time 2 (T2) risk images for adolescents living in high-
and low-risk neighborhoods. Dashed lines represent adolescents
who reside in neighborhoods 1 SD above the mean on level of risk;
solid lines represent adolescents who reside in neighborhoods 1
SD below the mean on level of risk.
910 Cleveland, Gibbons, Gerrard, Pomery, and Brody
prototype model. Adolescents who reported receiv-
ing effective parenting had more negative risk ima-
themselves, 2 years later. Together, these indirect
relations accounted for a significant amount of the
total effect of parenting behaviors on both the ado-
lescents’ reports of Time 2 peer use and their own
These effects existed after controlling for several
individual- and contextual-level control variables.
Recent research has demonstrated that African
American adolescents are more likely than other
youth to be exposed to contextual-level risk fac-
tors, such as neighborhood risk (Wallace & Muroff,
2002). Therefore, these results provide one possible
explanation for the paradoxical finding that African
American adolescents report lower rates of sub-
stance use than their White counterparts, despite
their relatively higher exposure to such risk-pro-
moting factors (Bachman et al., 1991; Johnston et al.,
2000; Wallace & Muroff, 2002). In this sample, ef-
fective parenting predicted less favorable adolescent
substance-use cognitions, which in turn were related
to less risk behavior. Thus, adolescents who reported
that they received higher levels of parental commu-
nication, monitoring, and warmth were offered some
protection against the risk factors described by
Hawkins et al. (1992).
Additional evidence of effective parents’ protec-
tive role for African American youth was found in
the regression analyses, which also showed that such
behaviors were more powerful than neighborhood
risk factors in predicting several youth outcomes.
Furthermore, interaction analyses revealed that
among families who resided in high-risk neighbor-
hoods, these parenting strategies were more effective
at reducing contemporaneous measures of suscept-
ibility to use and targets’ reports of their friends’ use
as well as their own use. These results are consistent
with other research that has found that effective
parenting can act as a buffer against the deleterious
effects of high-risk neighborhoods (e.g., Brody et al.,
2001, 2003; Gonzales et al., 1996; Rankin & Quane,
2002). Perhaps most important, among families in
high-risk neighborhoods, effective parenting was
associated with desired changes in adolescents’ risk
images (i.e., becoming less favorable).
Such effects on cognitions are especially im-
portant because most adolescent substance use oc-
curs outside the home, when parents are not around.
It appears that adolescents with involved parents
approach such situations armed with risk-reducing
beliefs. Because adolescent substance use is largely a
social phenomenon (Gibbons & Gerrard, 1997), this
to use substances
relation (parenting and risk images) has much in-
tuitive appeal as an area worthy of further explora-
tion. Adolescents are very image conscious (Carroll,
Durkin, Hattie, & Houghton, 1997; Elkind, 1978;
Lloyd & Lucas, 1998), and previous studies with
older adolescents and young adults have indicated
that the images can be altered and that making them
more negative is associated with a decline in behav-
ioral willingness (Gerrard et al., in press; Gibbons
et al., 2004). Specifically, parents can help children
realize that they, like their friends, do not have very
favorable risk images. Therefore, altering their im-
agesFmaking them more negativeFis likely to re-
duce their willingness to engage in substance use.
Although this pathway represents just one of many
ways that parent behaviors can affect their children’s
use of substances, we believe the findings in this
study are important, particularly in light of evidence
that suggests that reducing African American ado-
lescents’ prorisk cognitions is likely to reduce their
later use of hard drugs (Ellickson & Morton, 1991).
The findings reported here replicate earlier cross-
sectional studies conducted with the FACHS panel,
which have used Census-level economic data to
study neighborhood effects on African American
adolescent outcomes (Brody et al., 2001, 2003). These
studies also found that effective parenting can act as
a buffer against the deleterious effects of high-risk
neighborhoods. This conclusion, however, contrasts
with the results by Simons et al. (2002), who found
that parental control behaviors tended to become
less effective in preventing childhood conduct dis-
orders as caregivers’ perceptions of neighborhood
These differences may be explained by studies
that show that parents and children often perceive
their neighborhoods differently (Burton, Price-Spra-
tlen, & Spencer, 1996), and other research that has
found discrepancies in parent and child assessments
of family processes (Brody & Sigel, 1990). More
specifically, research has also shown that the per-
ceptions of parents and their young adolescents
usually do not correlate well (Achenbach, McCo-
naughy, & Howell, 1987; Hartos & Power, 2000) and
that adolescents’ perceptions of parenting behaviors
(whether accurate or not) are often more effective
predictors of adolescent behaviors than are their
parents’ self-reports of parenting (Brody et al., 2001;
Smetana, Crean, & Daddis, 2002). The significant, yet
modest, correlation between the targets’ and parents’
report of neighborhood in this study (r5.33, po.001)
indicated that although there was some overlap
between the two sources of information, they were
Impact of Parenting911
Several limitations of this study should be ac-
knowledged. First, the mediating variables in the
SEM were all measured at Time 2. Thus, it is possible
that different ordering of these variables may yield
different path coefficients. Future analyses with this
panel, including additional waves of data, will allow
for estimation of models that help clarify the causal
direction of these relations. Furthermore, with the
exception of the parents’ report of neighborhood
risk, all other constructs in the model were measured
using adolescent self-reports. This introduces the
possibility that shared method variance may limit
the interpretation of these results, particularly with
regard to perceptions of parenting behaviors and
friends’ substance use. However, as noted earlier,
research suggests that adolescents’ perceptions of
their parents’ behaviors are more effective predictors
of their own behaviors than are their parents’ self-
reports (Brody et al., 2001; Smetana et al., 2002). We
also agree with the conclusion that parental behav-
iors are most meaningful to the extent that they are
filtered through the child’s perceptions (Fletcher,
Steinberg, & Williams-Wheeler, 2004). Future re-
search with the FACHS panel will allow us to ad-
dress some of these concerns by including measures
of peers’ actual substance use.
It should also be noted that the participants in this
study were very youngF97% of the targets were
either 10 or 11 years old at Time 1. Thus, there was
very little initial substance use (more than 90% re-
ported never trying alcohol, tobacco, or marijuana)
and not a lot of use to predict at Time 3. These rates
are consistent with other research that has examined
early-onset substance use (e.g., Kaplow, Curran,
Dodge, & The Conduct Problem Prevention Research
Group, 2002; Oxford, Harachi, Catalano, & Abbott,
2000; Simons-Morton, 2002). We used statistical
methods that account for such skewed data; how-
ever, it is important to note that our model accounted
for a modest 11% of the variance in target Time 3 use.
Although the 5-year lag in this study is a long time
for early adolescents, there remained a significant
amount of variation in adolescents’ substance use
that was not accounted for by the constructs in
Nonetheless, the influence of parental behaviors
on early teenage children is especially important to
consider given that research has shown the sig-
nificance of early use in terms of later use and abuse
(Anthony & Petronis, 1995; Guy, Smith, & Bentler,
1994) and the importance of preventing early onset
of use (e.g., Wills et al., 2001, 1998). In particular,
research has indicated that African American ado-
lescents report less substance use than White ado-
lescents (Johnston et al., 2000), whereas indicators of
substance abuse (e.g., clinic admissions) are usually
higher among African American adults than White
adults (Substance Abuse and Mental Health Service
Administration, 2002). This discrepancy has been
called the racial cross-over effect, and it has puzzled
researchers for some time (Biafora & Zimmerman,
1998). This phenomenon illustrates another reason
why research examining protective factors against
early use is needed in African American populations.
Another drawback associated with the use of
young adolescents is that their cognitions are not yet
well formed and tend to be unstable. This is reflected
in the low reliabilities of some of the constructs. As
indicated earlier, this instability is an advantage, in
terms of mutability, but it is also a methodological
disadvantage. It is important to bear in mind that the
current study was concerned with African American
families who reside in rural and suburban areas.
Although this group is underrepresented in the
adolescent risk literature, the results reported here
may not generalize to other ethnic and racial groups
or to African American families in urban areas. This
is an area ripe for exploration and deserves further
attention. Finally, because a significant majority of
the parents were biological mothers, the results may
not generalize to other types of parent–child re-
lationships, such as for fathers. This is another area
worthy of further study.
The present study represents one of the few to
examine the cognitions involved in early adolescent
substance use; fewer still have examined these rela-
tions among African American youth. We believe
that the findings in this study may provide some
clues to understand how to prevent early adolescent
use (and later abuse) among children in this under-
represented population. The current results are en-
couraging because the parenting construct consisted
of things that most parents can doFprovide warmth
and support, monitor their children’s behavior, and
communicate with their children about substances.
These everyday interactions were shown to be as-
sociated with African American adolescents’ cogni-
tions about substance use and actual substance use,
both concurrently and several years in the future.
Furthermore, these parental behaviors were effective
even in the presence of several well-documented risk
factors, and they were especially effective when risk
opportunities were (relatively) common.
912 Cleveland, Gibbons, Gerrard, Pomery, and Brody
Achenbach, T. M., McConaughy, S. H., & Howell, C. T.
(1987). Child/adolescent behavioral and emotional
problems: Implications of cross-informant correlations
for situational specificity. Psychological Bulletin, 101,
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing
and interpreting interactions. Newbury Park, CA: Sage.
Ajzen, I. (1985). From intentions to actions: A theory of
planned behavior. In J. Kuhl & J. Beckman (Eds.), Action
control from cognition to behavior (pp. 11–39). New York:
Ajzen, I. (1991). The theory of planned behavior. Organiza-
tional Behavior and Human Decision Processes, 50, 179–211.
Anthony, J. C., & Petronis, K. R. (1995). Early-onset drug
use and risk of later drug problems. Drug and Alcohol
Dependence, 40, 9–15.
Aseltine, R. H. (1995). A reconsideration of parental and
peer influences on adolescent deviance. Journal of Health
and Social Behavior, 36, 103–121.
Bachman, J. G., Wallace, J. M., O’Malley, P. M., Kurth, C. L.,
Johnston, L. D., & Neighbors, H. W. (1991). Racial/eth-
nic differences in smoking, drinking, and illicit drug use
among American high school seniors, 1976–1989.
American Journal of Public Health, 81, 371–377.
Barnes, G. M., Reifman, A. S., Farrell, M. P., & Dintcheff, B.
A. (2000). The effect of parenting on the development of
adolescent alcohol misuse: A six-wave latent growth
model. Journal of Marriage and the Family, 62, 175–186.
Barnow, S., Schuckit, M. A., Lucht, M., John, U., & Frey-
berger, H. J. (2002). The importance of a positive family
history of alcoholism, parental rejection and emotional
warmth, behavioral problems and peer substance use
for alcohol problems in teenagers: A path analysis.
Journal of Studies on Alcohol, 63, 305–315.
Bauman, K. E., & Ennett, S. T. (1996). On the importance of
peer influence for adolescent drug use: commonly ne-
glected considerations. Addiction, 91, 185–198.
Bentler, P. M. (1990). Comparative fit indices in structural
models. Psychological Bulletin, 107, 588–606.
Beyth-Marom, R., & Fischhoff, B. (1997). Adolescents’ de-
cisions about risks: A cognitive perspective. In J. Schu-
lenberg, J. L. Maggs, & K. Hurrelmann (Eds.), Health
risks and developmental transitions during adolescence (pp.
110–135). New York: Cambridge University Press.
Biafora, F., & Zimmerman, R. (1998). Developmental pat-
terns of African-American adolescent drug use. In W. A.
Vega & A. G. Gil (Eds.). Drug use and ethnicity in early
adolescence (pp. 149–175). New York: Plenum.
Blanton, H., Gibbons, F. X., Gerrard, M., Conger, K. J., &
Smith, G. E. (1997). Development of health risk proto-
types during adolescence: Family and peer influence.
Journal of Family Psychology, 11, 271–288.
Brody, G. H., Flor, D. L., Hollett-Wright, N., & McCoy, J. K.
(1998). Children’s development of alcohol use norms:
Contributions of parent and sibling norms, children’s
temperaments, and parent-child discussions. Journal of
Family Psychology, 12, 209–219.
Brody, G. H., Ge, X., Conger, R., Gibbons, F. X., Murry, V.
M., Gerrard, M., et al. (2001). The influence of neigh-
borhood disadvantage, collective socialization, and
parenting on African American children’s affiliations
with deviant peers. Child Development, 72, 1231–1246.
Brody, G. H., Ge, X., Katz, J., & Arias, I. (2000). A long-
itudinal analysis of internalization of parental alcohol-
use norms and adolescent alcohol use. Applied Develop-
mental Science, 4, 71–79.
Brody, G. H., Murry, V. M., Ge, X., Kim, S. Y., Simons, R. L.,
Gibbons, F. X., Gerrard, M., et al. (2003). Neighborhood
disadvantage moderates associations of parenting and
older sibling problem attitudes and behavior with con-
duct disorders in African American children. Journal of
Consulting and Clinical Psychology, 71, 211–222.
Brody, G. H., & Sigel, I. (Eds.). (1990). Research methods for
studying at-risk families. Hillsdale, NJ: Erlbaum.
Brook, J. S., Nomura, C., & Cohen, P. (1988). A network of
influences on adolescent drug involvement: Neighbor-
hood, school, peer, and family. Genetic, Social, and General
Psychology Monographs, 115, 125–145.
Brook, J. S., Whiteman, M., Czeisler, L. J., Shapiro, J., &
Cohen, P. (1997). Cigarette smoking in young adults:
Childhood and adolescent personality, familial, and peer
antecedents. Journal of Genetic Psychology, 158, 172–188.
Brown, B. B., Mounts, N., Lamborn, S. D., & Steinberg, L.
(1993). Parenting practices and peer group affiliation in
adolescence. Child Development, 64, 467–482.
Brown, L. K., Di Clemente, R. J., & Reynolds, L. A. (1991).
HIV prevention for adolescents: Utility of the health
belief model. AIDS Education and Prevention, 3, 50–59.
Browne, M. W., & Cudek, R. (1993). Alternative ways of
assessing model fit. In K. A. Bollen & J. S. Long (Eds.),
Testing structural equation models (pp. 136–162). New-
bury Park, CA: Sage.
Burton, L. M., Price-Spratlen, T., & Spencer, M. B. (1996).
On ways of thinking about measuring neighborhoods:
Implications for studying context and developmental
outcomes for children. In J. Brooks-Gunn, G. J. Duncan,
& J. L. Aber (Eds.). Neighborhood poverty: Context and
consequences for children (pp. 132–144). New York: Rus-
sell Sage Foundation.
Carroll, A., Durkin, K., Hattie, J., & Houghton, S. (1997).
Goal setting among adolescents: A comparison of de-
linquent, at-risk, and not-at-risk youth. Journal of Edu-
cational Psychology, 89, 441–450.
Chassin, L., Presson, C. C., Sherman, S. J., Corty, E., &
Olshavsky, R. W. (1981). Self-images and cigarette
smoking in adolescence. Personality and Social Psychology
Bulletin, 7, 670–676.
Chassin, L., Presson, C. C., Todd, M., Rose, J. S., & Sherman,
S. J. (1998). Maternal socialization of adolescent smoking:
The intergenerational transmission of parenting and
smoking. Developmental Psychology, 34, 1189–1201.
Conger, R. D., Elder, G. H., Lorenz, F. O., Simons, R.L, &
Whitbeck, L. B. (1992). A family process model of eco-
nomic hardship and influences on adjustment of early
adolescent boys. Child Development, 63, 526–541.
Impact of Parenting913
Cutrona, C. E., Russell, D. W., Hessling, R. M., Brown, P. A.,
& Murry, V. (2000). Direct and moderating effects of
community context on the psychological well-being of
African American women. Journal of Personality and So-
cial Psychology, 79, 1088–1101.
Dishion, T., Capaldi, D., Spracklen, K., & Li, F. (1995). Peer
ecology of male adolescent drug use. Development and
Psychopathology, 7, 803–824.
Ellickson, P. L., & Morton, S. C. (1999). Identifying ado-
lescents at risk for hard drug use: Racial/ethnic varia-
tions. Journal of Adolescent Health, 25, 382–395.
Elkind, D. (1978). Understanding the young adolescent.
Adolescence, 13, 127–134.
Ennett, S. T., Bauman, K. E., Foshee, V. A., Pemberton, M.,
& Hicks, K. A. (2001). Parent–child communication
about adolescent tobacco and alcohol use: What do
parents say and does it affect youth behavior? Journal of
Marriage and the Family, 63, 48–62.
Eysenck, S. B. G., & Eysenck, H. J. (1977). The place of
impulsiveness in a dimensional system of personality
description. British Journal of Social and Clinical Psychol-
ogy, 16, 57–58.
Fletcher, A. C., Steinberg, L., & Williams-Wheeler, M.
(2004). Parental influences on adolescent problem be-
havior: Revisiting Stattin and Kerr. Child Development,
Gerrard, M., Gibbons, F. X., Benthin, A. C., & Hessling, R.
M. (1996). The reciprocal nature of risk behaviors and
cognitions: What you think shapes what you do and vice
versa. Health Psychology, 15, 344–354.
Gerrard, M., Gibbons, F. X., Brody, G. H., Murry, V. M.,
Cleveland, M. J., & Wills, T. A. (in press). A theory-based
dual focus alcohol intervention program for pre-adolescents:
The Strong African American Families Program. Psychology
of Addictive Behaviors.
Gerrard, M., Gibbons, F. X., Stock, M. L., Vande Lune, L. S.,
& Cleveland, M. J. (2005). Images of smokers and will-
ingness to smoke among African American pre-adoles-
cents: An application of the prototype/willingness
model of adolescent health risk behavior to smoking
initiation. Journal of Pediatric Psychology, 30, 305–318.
Gibbons, F. X., & Gerrard, M. (1995). Predicting young
adults’ health-risk behavior. Journal of Personality and
Social Psychology, 69, 505–517.
Gibbons, F. X., & Gerrard, M. (1997). Health images and
their effects on health behavior. In B. P. Buunk & F. X.
Gibbons (Eds.), Health, coping, and well-being: Perspectives
from social comparison theory (pp. 63–94). Mahwah, NJ:
Gibbons, F. X., Gerrard, M., Blanton, H., & Russell, D. W.
(1998). Reasoned action and social reaction: Willing-
ness and intention as independent predictors of
health risk. Journal of Personality and Social Psychology, 74,
Gibbons, F. X., Gerrard, M., & Boney-McCoy, S. (1995).
Prototype perception predicts (lack of) pregnancy
prevention. Personality and Social Psychology Bulletin, 21,
Gibbons, F. X., Gerrard, M., Cleveland, M. J., Wills, T. A., &
Brody, G. H. (2004). Perceived discrimination and sub-
stance use in African American parents and their chil-
dren: A panel study. Journal of Personality and Social
Psychology, 86, 517–529.
Gibbons, F. X., Gerrard, M., & Lane, D. J. (2003). A social
reaction model of adolescent health risk. In J. M. Suls &
K. Wallston (Eds.), Social psychological foundations of
health and illness (pp. 107–136). Oxford, England:
Gibbons, F. X., Gerrard, M., Vande Lune, L. S., Wills, T. A.,
Brody, G. H., & Conger, R. D. (2004). Context and cog-
nition: Environmental risk, social influence, and ado-
lescent substance use. Personality and Social Psychology
Bulletin, 30, 1048–1061.
Giordano, P., Cernkovich, S., & Demaris, A. (1993). The
family and peer relations of Black adolescents. Journal of
Marriage and the Family, 55, 277–287.
Gochman, D. S. (1992). Here’s looking at you, kid! New
ways of viewing the development of health cognition. In
E. J. Susman, L. V. Feagans, & W. J. Ray (Eds.), Emotion,
cognition, health and development in children and adolescents
(pp. 9–23). Hillsdale, NJ: Erlbaum.
Gonzales, N. A., Cauce, A. M., Friedman, R. J., & Mason, C.
A. (1996). Family, peer, and neighborhood influences on
academic achievement among African-American ado-
lescents: One-year prospective effects. American Journal
of Community Psychology, 24, 365–387.
Guy, S. M., Smith, G. M., & Bentler, P. M. (1994). Conse-
quences of adolescent drug use and personality factors
on adult drug use. Journal of Drug Education, 24, 109–132.
Hartos, J. L., & Power, T. G. (2000). Association between
mother and adolescent reports for assessing relations
between parent-adolescent communication and adoles-
cent adjustment. Journal of Youth and Adolescence, 29,
Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk
and protective factors for alcohol and other drug pro-
blems in adolescence and early adulthood: implications
for substance abuse prevention. Psychological Bulletin,
Jackson, C., & Henriksen, L. (1997). Do as I say: Parent
smoking, antismoking socialization, and smoking onset
among children. Addictive Behaviors, 22, 107–114.
Johnston, L. D., O’Malley, P. M., & Bachman, J. G. (2000).
The Monitoring the Future national results on adolescent
drug use: Overview of key findings, 1999. Bethesda, MD:
National Institute on Drug Abuse.
Jo ¨reskog, K. G., & So ¨rbom, D. (2004a). LISREL (Version 8.70)
[Computer software]. Chicago: Scientific Software Inter-
Jo ¨reskog, K. G., & So ¨rbom, D. (2004b). PRELIS (Version
2.70) [Computer software]. Chicago: Scientific Software
Kandel, D. (1996). The parental and peer contexts of ado-
lescent deviance: An algebra of interpersonal influences.
Journal of Drug Issues, 26, 289–315.
914Cleveland, Gibbons, Gerrard, Pomery, and Brody
Kaplow, J. B., Curran, P. J., & Dodge, K. A. The Conduct
Problems Prevention Research Group. (2002). Child,
parent, and peer predictors of early-onset substance use:
a multisite longitudinal study. Journal of Abnormal Child
Psychology, 30, 199–216.
Kessler, R., McGonagle, K. A., Zhao, S., Nelson, C. B.,
Hughes, M., Eshleman, S., et al. (1994). Lifetime and 12-
month prevalence of DSM–III–R Psychiatric Disorders
in the United States. Archives of General Psychiatry, 51,
Kline, R. B. (1998). Principles and practice of structural equa-
tion modeling. New York: Guilford.
Leventhal, T., & Brooks-Gunn, J. (2000). The neighbor-
hoods they live in: The effects of neighborhood
residence on child and adolescent outcomes. Psycholo-
gical Bulletin, 126, 309–337.
Li, X., Stanton, B., & Feigelman, S. (2000). Impact of per-
ceived parental monitoring on adolescent risk behavior
over 4 years. Journal of Adolescent Health, 27, 49–56.
Lloyd, B., & Lucas, K. (1998). Smoking in adolescence: Images
and identities. New York: Rutledge.
Oxford, M. L., Harachi, T. W., Catalano, R. F., & Abbott, R.
D. (2000). Preadolescent predictors of substance initia-
tion: A test of both the direct and mediated effect of
family social control factors on deviant peer associations
and substance initiation. American Journal of Drug and
Alcohol Abuse, 27, 599–616.
Peeples, F., & Loeber, R. (1994). Do individual factors and
neighborhood context explain ethnic differences in ju-
venile delinquency? Journal of Quantitative Criminology,
Peterson, P., Hawkins, J., Abbot, R., & Catalano, R. (1994).
Disentangling the effects of parental drinking, family
management, and parental alcohol norms on cur-
rent drinking by African American and European
American adolescents. Journal of Research on Adolescence,
Rankin, B. H., & Quane, J. M. (2002). Social contexts and
urban adolescent outcomes: The interrelated effects of
neighborhoods, families, and peers on African-Amer-
ican youth. Social Problems, 49, 79–100.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997).
Neighborhoods and violent crime: A multlilevel study
of collective efficacy. Science, 277, 918–924.
Satorra, A., & Bentler, P. M. (1988). Scaling corrections for
chi-square statistics in covariance structure analysis. In
Proceedings of the Business and Economic Statistics Sections
(pp. 308–313). Alexandria, VA: American Statistical
Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-
square test statistic for moment structure analysis. Psy-
chometrika, 66, 507–514.
Shaffer, J. L., Fisher, P., Piancentmi, J., Conners, C. K.,
Schwab-Stone, M., Cohen, P., et al. (1993). The diagnostic
interview schedule for children, (DISC–R): I. Prepara-
tion, field testing, inter-rater reliability, and acceptability.
Journal of the American Academy of Child and Adolescent
Psychiatry, 32, 643–650.
Shen, S., Locke-Wellman, J., & Hill, S. Y. (2001). Adolescent
alcohol expectancies in offspring from families at high
risk for developing alcoholism. Journal of Studies on Al-
cohol, 62, 763–772.
Sieving, R. E., Maruyama, G., Williams, C. L., & Perry, C. L.
(2000). Pathways to adolescent alcohol use: Potential
mechanisms of parent influence. Journal of Research on
Adolescence, 10, 489–514.
Simons, R. L. (Ed.). (1996). Understanding differences between
divorced and two-parent families. Thousand Oaks, CA:
Simons, R. L., Lin, K., Gordon, L. C., Brody, G. H., Murry,
V., & Conger, R. D. (2002). Community differences in
the association between parenting practices and child
conduct problems. Journal of Marriage and Family, 64,
Simons-Morton, B. G. (2002). Prospective analysis of peer
and parent influences on smoking initiation among
early adolescents. Prevention Science, 3, 275–283.
Smart, R. G., Adlaf, E. M., & Walsh, G. W. (1994). Neigh-
borhood socio-economic factors in relation to student
drug use and programs. Journal of Child and Adolescent
Substance Abuse, 3, 37–46.
Smetana, J. G., Crean, H. F., & Daddis, C. (2002). Family
processes and problem behaviors in middle-class Afri-
can-American adolescents. Journal of Research on Adoles-
cence, 12, 275–304.
Stone-Romero, E. F., & Anderson, L. E. (1994). Relative
power of moderated multiple regression and the com-
parison of subgroup correlation coefficients for detect-
ing moderating effects. Journal of Applied Psychology, 79,
Substance Abuse and Mental Health Service Administra-
tion, Office of Applied Studies. (2002). National survey
on drug use and health. Retrieved June 2, 2005 from
Thornton, B., Gibbons, F. X., & Gerrard, M. (2002). Risk
perception and prototype perception: Independent
processes predicting risk behavior. Personality and Social
Psychology Bulletin, 28, 986–999.
Vaccaro, D., & Wills, T. A. (1998). Stress-coping factors in
adolescent substance use: Ethnic and gender differences
in samples of urban adolescents. Journal of Drug Educa-
tion, 28, 257–282.
Wallace, J. M., & Muroff, J. R. (2002). Preventing substance
abuse among African American children and youth:
race differences in risk factor exposure and vulnerability.
Journal of Primary Prevention, 22, 235–261.
Warshaw, P. R., & Davis, F. D. (1985). Disentangling be-
havioral intention and behavioral expectation. Journal of
Experimental Social Psychology, 21, 213–228.
West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural
equation models with nonnormal variables: problems
and remedies. In R. H. Hoyle (Ed.), Structural equation
modeling: concepts, issues, and applications (pp. 56–75).
Thousand Oaks, CA: Sage.
Impact of Parenting 915
Whitbeck, L. B. (1999). Primary socialization theory: It all
begins with the family. Substance Use and Misuse, 34,
Wickrama, K. A. S., Conger, R. D., Wallace, L. E., & Elder,
G. H. Jr. (1999). The intergenerational transmission of
health-risk behaviors: Adolescent lifestyles and gender
moderating effects. Journal of Health and Social Behavior,
Wills, T. A., Cleary, S., Filer, M., Shinar, O., Mariani, J., &
Spera, K. (2001). Temperament related to early-onset
substance use: Test of a developmental model. Preven-
tion Science, 2, 145–163.
Wills, T. A., Gibbons, F. X., Gerrard, M., & Brody, G. H.
(2000). Test of a model of factors relevant for early onset
adolescent substance use among African-American
children. Health Psychology, 19, 253–263.
Wills, T. A., Vaccaro, D., & McNamara, G. (1994).
Novelty seeking and risk taking as predictors of ad-
olescent substance use. Journal of Substance Use, 6,
Wills, T. A., Windle, M., & Cleary, S. D. (1998). Tempera-
ment and novelty seeking in adolescent substance use:
Convergence of dimensions of temperament with con-
structs from Cloninger’s theory. Journal of Personality and
Social Psychology, 74, 387–406.
916 Cleveland, Gibbons, Gerrard, Pomery, and Brody