University of Southern California Dissertations and Theses

Social self-control and adolescent substance use

(USC Thesis Other)

Social self-control and adolescent substance use

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content SOCIAL SELF-CONTROL AND ADOLESCENT SUSBTANCE USE

by

Pallav Pokhrel

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PREVENTIVE MEDICINE)

December 2009

Copyright 2009 Pallav Pokhrel

ii
Dedication

To my father, Puru Pokhrel (1947-1998)

iii
Acknowledgments

I would like to thank my committee members, Drs Jennifer B. Unger, Lawrence
Palinkas, Louise A. Rohrbach, Ping Sun, and Steve Sussman for their guidance, support,
and above all, for their patience. I am grateful to Dr Sussman for his unstinting
mentorship, in matters academic and beyond.

iv
Table of Contents

Dedication................................................................................................... ii

Acknowledgments...................................................................................... iii

List of Tables ............................................................................................. vi

List of Figures........................................................................................... vii

Abstract.................................................................................................... viii

Chapter 1: Introduction...............................................................................1
Specific Aims...............................................................................................1
Background and Significance ......................................................................3
Overview of the three studies ....................................................................15

Chapter 2: Prospective associations of social self-control
with substance use ....................................................................................19
Abstract......................................................................................................19
Introduction................................................................................................20
Methods......................................................................................................23
Analyses & Results....................................................................................24
Discussion..................................................................................................31

Chapter 3: Social self-control versus impulsive sensation-seeking...........36
Abstract......................................................................................................36
Introduction................................................................................................36
Methods..................................................................................................... 39
Statistical Analyses ....................................................................................43
Results........................................................................................................47
Discussion.....................................................……………….....................50

Chapter 4: Structural models examining relations among social
self-control, sense of coherence, and substance use ..................................56
Abstract......................................................................................................56
Introduction................................................................................................57
Methods......................................................................................................60
Statistical Analyses ....................................................................................62
Results........................................................................................................65
Discussion..................................................................................................75

Chapter 5: Conclusion................................................................................81
Limitations and future directions...............................................................82

v
Implications for prevention........................................................................88

References..................................................................................................92

Appendix A..............................................................................................104

Appendix B ............................................................................ .................108

vi
List of Tables

Table 1. Baseline demographics ................................................................26

Table 2. Follow-up demographics .............................................................27

Table 3. Social self-control as a predictor of substance use ......................29

Table 4. Substance use as a predictor of social self-control ......................30

Table 5. Sample characteristics at baseline................................................41

Table 6. Multitrait/multi-item correlation matrix ......................................46

Table 7. Correlations between social self-control, sensation
seeking, and impulsivity indices....................................................48

Table 8. Social self-control and sensation seeking as
predictors of substance use ...........................................................50

Table 9. List of indicators ..........................................................................67

Table 10. Factor loadings...........................................................................68

Table 11. Model fit indices ........................................................................69

vii
List of Figures

Figure 1. Structural model 1: substance use as outcome ...........................71

Figure 2. Structural model 2: social self-control as outcome ....................72

Figure 3. Structural model 3: sense of coherence as outcome...................73

viii
Abstract

The present dissertation project examined social self-control in the context of
adolescent substance use behavior. Social self-control is a relatively understudied
dispositional variable that measures individuals’ self-control in social situations or
interpersonal interactions. One of the primary aims of the present project was to examine
whether the lack of social self-control predisposed adolescents for higher substance use
(cigarette, alcohol, marijuana, and hard drug use). A lack of social self-control may
adversely affect adolescents’ ability to form prosocial bonds, engage in prosocial
activities, and build an adaptive social support network. In addition, adolescents lacking
in social self-control may experience higher interpersonal conflicts and negative life
events. Thus, lacking social self-control may make adolescents vulnerable to a number of
proximal risk factors of substance use such as deviant peer affiliation and stressful life
events. We collected two sets of 1-year prospective data from adolescents representing
Regular and Continuation (alternative) high school students in Southern California. In
addition to examining the relation between social self-control and substance use, we
examined the construct validity of social self-control and examined the relationships of
social self-control with sensation seeking and sense of coherence. Across both datasets,
we found that higher social self-control was a unique predictor of lower substance use 1
year later. For regular high school students, a higher social self-control at baseline
predicted higher sense of coherence 1 year later. Hence, it appears that social self-control
is an important predictor of adolescent health and well-being.

1
Chapter 1

Introduction

Specific Aims

The project examines the construct of social self-control in relation to adolescent
cigarette, alcohol, marijuana, and hard drug use. Social self-control is a variable that
measures adolescents’ self-restraint abilities in social situations (e.g., interpersonal
interactions). The ten Likert-type items of social self-control intend to tap adolescents’
abilities to control impulse and delay gratification in social situations. Studies examining
the relations between self-control and adolescent substance use tend to agree that lack of
self-control is a robust concurrent and prospective predictor of higher substance use;
suggesting that teaching self-control skills to adolescents may be protective against
substance use.
Social self-control is a relatively new measure of self-control, in that the measure
has not been extensively studied yet. No study so far has examined the longitudinal
relationship between social self-control and adolescent substance use. Although social
self-control appears to be a relatively stable construct, studies so far have not tested
whether earlier substance use can predict later social self-control. This dissertation
project examines the possible bidirectional relationship between social self-control and
substance use among youths representing high and low school-based risk-contexts (i.e.
Continuation/Alternative vs. Regular High Schools). In addition, the project examines
whether or not social self-control diverges in construct validity from two other commonly
studied indicators of self-control, namely, impulsivity and sensation-seeking. Finally, the

2
project examines social self-control in relation to sense of coherence, a construct that
signifies resilience and is related to mental and physical well-being. The specific aims of
the project are as follows:
Study 1
1. Examine whether social self-control at baseline predicts substance use (i.e., past-
30 day cigarette, alcohol, marijuana, and hard drug use) one year later, after
controlling for baseline substance use and demographic variables.
2. Examine whether substance use at baseline predicts social self-control one year
later after controlling for baseline social self-control and demographic variables.
3. Examine whether the bidirectional relationship between social self-control and
substance use is moderated by high- and low-risk school contexts (i.e.,
Continuation/Alternative and Regular High School status).
Study 2
4. Examine the convergence and divergence of the items of the 10-item social self-
control scale by comparing them against the items from Zuckerman’s impulsive
sensation seeking scale, using the multitrait scaling method.
5. Examine the prospective relationship between social self-control and substance
use (cigarette, alcohol, marijuana, and hard drug use), controlling for impulsivity
and sensation-seeking.
Study 3
6. Use structural equation modeling to examine whether the latent social self-control
factor at baseline predicts substance use and sense of coherence one year later,
controlling for baseline substance use, sense of coherence and demographic
variables.

3
7. Use structural equation modeling to examine whether the substance use at
baseline predicts social self-control and sense of coherence one year later,
controlling for baseline sense of coherence, baseline social self-control, and
demographic variables.
8. Use structural equation modeling to examine whether sense of coherence at
baseline predicts social self-control and substance use one year later, controlling
for baseline substance use, social self-control, and demographic variables.
9. Perform multiple group comparison between regular and continuation high school
samples to examine whether school context moderates the above relationships.

Background and significance

Problem Importance
Substance abuse remains a significant health risk behavior among American
secondary school students. According to the recent Monitoring the Future Survey
(MTFS; 2007), about 50% of adolescents are likely to have tried an illicit drug (i.e.,
marijuana and hard drugs) by the time they graduate from high school; 46% are likely to
have tried cigarettes; and about 75% are likely to have drank more than a few sips of
alcohol (Johnston, O’Malley, Bachman, & Schulenberg, 2008). Among youth surveyed
by the Youth Risk Behavior Surveillance (YRBS; 2005), the prevalence of past 30-day
cigarette use, alcohol use, and marijuana use was 23%, 43.3%, and 20.2%, respectively
(Eaton, Kahn, Kinchen et al., 2006). Although the overall prevalence of cigarette
smoking in the secondary school population has been decreasing since the early 1990s,
the rate of decline has been slower in recent years (Johnston et al., 2008). MTFS (2007)
found no decline in cigarette smoking among 11
th
and 12
th
graders in 2007 (Johnston et

4
al., 2008). In addition, the past-12-month marijuana usage, which had been declining
among 12
th
graders previously, has leveled at about 32% in recent years (Johnston et al.,
2008).
Although binge drinking in recent times has been on a decline relative to the
1980s and 1990s (Johnston et al., 2008), teenage drinking still remains a problem of
serious concern. Fifty five percent of high school students tend to report having been
drunk once in a lifetime (Johnston et al., 2008). Further, the recent decrease in hard drug
use is not common for all types of hard drugs (i.e., inhalants and illicit drugs other than
marijuana). While amphetamine, Ritalin, and methamphetamine uses show decreasing
trends, prevalence of LSD, other hallucinogens, cocaine, and heroin use has not changed,
and usages of inhalants and ecstasy seem to be on the rise (Johnston et al., 2008).
Health Risks of Adolescent Drug Use
Seventy one percent of deaths among youth and young adults (10-24 years) in the
United States are caused by motor vehicle accidents (31%), unintentional injuries (14%),
homicide (15%), and suicide (11%) (Eaton et al., 2006). Youth are more likely to die of
these causes when they are under the influence of alcohol, marijuana, and hard drugs. For
example, alcohol use is likely to be involved in about one-third of all teenage suicide
cases (Karch, Crosby, & Simon, 2006). In addition, individuals who initiate using
cigarettes, alcohol, and other drugs as adolescents are more likely to suffer from cancer
and cardiovascular diseases later in life. Currently, cancer and cardiovascular diseases
account for 61% of all deaths in the United States among adults 25 years and older (Eaton
et al., 2006). Further, adolescents who develop substance use dependence are also likely
to develop other psychiatric disorders (Deas, 2006; Shrier, Harris, Kurland, & Knight,

5
2003). Some of the psychiatric disorders comorbid to adolescent substance use
dependence include anxiety, depression, attention-deficit disorder, conduct disorder,
mania, eating disorders, and hallucinations or delusions (Shrier et al., 2003).
Etiology of Adolescent Drug Use
Sussman & Ames discuss (2008) seven multivariate models of substance abuse
etiology. The “risk and protective factors” approach describes substance abuse etiology
in terms of interplay among several intrapersonal, interpersonal, social, and
environmental risk and protective factors. For example, some of the known risk factors
associated with adolescent substance use are impulsiveness, aggression, school failure,
negative home environment, and low socioeconomic status (SES); and some of the
known protective factors include social competence, higher attachment with parents,
higher academic achievement, and better neighborhood cohesiveness. The
biopsychosocial model recognizes three domains of casual influence: biological (e.g.,
genes that control neurotransmission), psychological (e.g., cognitive and affective states),
and social (social influence, built-environment). These domains may affect substance use
behavior directly or interactively. The Problem Behavior Theory (PBT; Jessor & Jessor,
1977) attempts to explain adolescent problem behavior in terms of interactions among
adolescents’ personality system (e.g., motivation, beliefs, and control), behavioral system
(e.g., prosocial vs. problem behavior), and perceived environmental system (e.g., family
conflict). Perceived environmental system consists of “distal” and “proximal” factors.
Proximal factors (e.g., family drug use role models) would have more immediate effects
on individuals than distal factors (e.g., family conflict). PBT views adolescent problem

6
behaviors as adolescents’ response to the challenges (e.g., identity formation, peer
influence) of their developmental stage.
The Triadic Influence Theory (TIT; Petraitis, Flay, & Miller, 1995) integrates
several theories of health behavior to propose a triadic model based on three substantive
domains of influence, namely, interpersonal, attitudinal/cultural, and intrapersonal.
Further, TIT recognizes that variables within each of these domains can be differentiated
as “distal,” “proximal,” and “ultimate,” based on their closeness of relation with
substance use behavior. For example, proximal, distal, and ultimate interpersonal
variables would include drug use role models, social-related drug beliefs (e.g., social
approval of drug use), and home stress, respectively. Hence, as suggested by PBT and
TIT, another way to approach drug use etiology is in terms of proximal and distal factors.
Among the etiologic models discussed by Sussman & Ames (2008), TIT is the most
integrative. Theories integrated into TIT include cognitive-affective theories [e.g.,
theories of planned behavior (Azjen, 1985) and reasoned action (Fishbein & Azjen,
1975)], social-learning theories (e.g., Bandura, 1977), commitment and social attachment
theories [e.g., social control theory (Hirschi, 1969), social skill theories], and
comprehensive theories [e.g., PBT].
The stage modeling (e.g., Flay et al., 1983) approach views substance abuse
behavior as a stage in a succession of stages, such that each stage is likely to have its own
set of risk and protective factors. The first stage is a preparation stage in which non-users
are exposed to substance use risk factors related to genes, personality (e.g.,
impulsiveness), family, and peers. In the second stage, non-users become initiators and
risk factors such as peer pressure and drug availability can be active at this stage. In the

7
third stage, initiators become experimenters and consume drugs more frequently. Some
of the risk factors during this stage may include positive outcome expectancies, pro-drug
use implicit association, and use habits. In the fourth stage, experimenters become
abusers and may become subject to physiological and psychological dependence.
The PACE stage model (Sussman & Unger, 2004) proposes that substance use
behavior is determined somewhat sequentially by variables that can be labeled pragmatic,
attraction, communication, and expectation. Pragmatic variables indicate how easily a
person may acquire drugs (e.g., ease of drug availability). Attraction variables indicate
the extent to which a person he or she enjoys the drug. Communication indicates how
comfortable a person is talking about the drug across private and public contexts.
Communication variables also suggest whether or not a person is associated with a drug-
specific subculture. Expectation variables indicate the user’s feelings and beliefs about
whether or not the drug is fulfilling the inner or social experiences expected off the drug.
Self-control and Adolescent Drug Use
Self-control is an important intrapersonal variable that has been consistently
associated with substance use. Mischel (1974) has defined self-control as the ability to
delay gratification. Others have defined self-control as the tendency to act without
thinking (Tarter, 1988) or act on immediate small rewards in preference over delayed
large rewards (Gottfredson & Hirschi, 1990; O’Donoghue & Rabin, 1999; Vollmer et al.,
1999). Discussed below are some key self-control theories that have relevance to
adolescent substance use behavior.
Epigenetic and transactional models of self-control development and substance
use. According to the epigenetic model of self-control development, temperament

8
dimensions provide a foundation on which a more complex dispositional attribute such as
self-control develop (Wills, Sandy, & Yaeger, 2000). Wills et al. (see Wills, Sandy, &
Yaeger, 2000) found temperamental characteristics such as physical activity level (i.e.,
tendency to move about rather than sit still) and negative emotionality (e.g., touchiness)
during childhood to predict substance use during adolescents. Conversely, task attentional
orientation (i.e., ability to focus attention) and positive emotionality (i.e., frequent
experience of positive affect) during childhood were found to act as protective factors
against substance use during adolescence (Wills, Sandy, & Yaeger, 2000).
Wills and colleagues (Wills, Sandy, & Yaeger, 2000; Wills & Dishion, 2004)
suggest that the temperament dimensions that act as protective and risk factors in
childhood serve as foundations for good self-control (e.g., planfulness) and poor self-
control (e.g., impulsivity) development during adolescence, respectively. They maintain
that self-control requires a sophisticated level of cognitive and social development that is
not common in childhood (Wills, Sandy, & Yaeger, 2000; Wills & Dishion, 2004). Some
of the indicators of self-control examined by Wills and colleagues include impulsivity,
distractability, planfulness, problem-solving ability, and abilities to control anger and
sadness (e.g., Wills, Walker, Mendoza, & Ainetter, 2006).
Wills et al.’s (see Wills & Dishion, 2004) transactional model of emerging self-
control considers self-control development to be dependent, to a large extent, on the
“transaction” between parenting practices and the child’s temperament. According to
Wills & Dishion (2004), early socialization processes dependent on parenting is likely to
have a direct effect on later self-control. For example, parenting practices based on
positive parent-child attachment, parent-child communication, and parental monitoring

9
are likely to result in better self-control characteristics in the child. However, the
socialization processes are also likely to be dependent on the child’s temperament, to
some extent (e.g., negative temperament of the child is likely to result in poorer parent-
child attachment). Wills & Dishion (2004) further propose that self-control is not often
directly related to substance use but the proximal factors affecting substance use such as
negative life events and deviant peer-group affiliation. Self-control in the model is
considered to be likely to moderate the relation between social adaptation and substance
use (Wills & Dishion, 2004).
Wills and colleagues (e.g., Wills, Gibbons et al., 2003; Wills, Resko, Ainitter, &
Mendoza, 2004; Wills, Murry, Brody et al., 2007) have found positive parent-child
relation, parent support, and involved parenting predictive of good self-control. Harsh
parenting, on the other hand, has been found to be positively associated with poor self-
control and negatively associated with good self-control (Wills et al., 2007). Poor self-
control has been found to be predictive of sex-willingness and sex engager prototypes
among preadolescents (Wills et al., 2007). Among young adolescents, sex abstainer
prototypes have been found to mediate the effects of positive and negative effects of good
and poor self-control on sexual behavior, respectively (Wills et al., 2003).
Self-control, self-regulation, and executive functioning. Self-control involves a
complex self-regulatory system in that self-control is concerned not only with controlling
immediate impulses but also with exerting self-directed efforts over a longer period of
time. For example, self-control for a recovering addict may involve careful self-
monitoring and planning. In Bandura’s (1991) social cognitive theory of self-regulation,

10
the self-regulatory system is dependent on the following three subfunctions: self-
monitoring, judgmental, and self-reactive.
Self-monitoring is the first step toward attaining a desired behavior and serves the
functions of self-diagnosis or observing one’s pattern of behavior (e.g., knowing how
different social contexts affect one’s thoughts and behavior) and initiating an internal
system of motivation (e.g., setting goals, evaluating one’s progress toward the goals)
(Bandura, 1991). Self-evaluation of one’s progress toward a goal involves several
judgmental processes and is guided by one’s personal standards (i.e., evaluating one’s
attainments against one’s own standards) and social referencing (i.e., evaluating one’s
attainments against the attainments of others) (Bandura, 1991). These judgments would
in turn serve as a basis for establishing a self-reactive subsystem that would manipulate
rewards and punishments in order to encourage the desired behavior and discourage the
undesired ones, respectively (Bandura, 1991). In Bandura’s theory (1991), an individual’s
self-efficacy determines the functional operation of his or her self-regulatory system.
Barkley (1997) has proposed a neuropsychological model of self-control and self-
regulation. According to Barkley (1997), behavioral disinhibition is the primary
characteristic of individuals with low self-control. Behavioral disinhibition in the model
is further linked with impairments in four executive functions controlled by the prefrontal
cortex of the brain: working memory, self-regulation of affect-motivation-arousal,
internalization of speech, and reconstitution (i.e., behavioral analysis and synthesis)
(Barkley, 1997). The impairments in executive functions in turn lead to decreased motor
control (e.g., execution of goal-directed responses) (Barkley, 1997).

11
Working memory deals with faculties that help in the rational organization of
actions; for example, the ability to hold events in mind, hindsight, foresight, and
manipulation of actions (Barkley, 1997). Self-regulation of affect-motivation-arousal
helps to generate arousal and motivational states conducive to goal-directed performance
(Barkley, 1997). Internalization of speech facilitates reflection over past actions,
instruction for imminent actions, moral reasoning, rule-setting and rule-guided behaviors,
and problem solving (Barkley, 1997). Reconstitution refers to analysis and synthesis of
behavior (i.e., the ability to relate causes and effects in relation to actions and goals) and
goal-directed behavioral creativity (i.e., the ability to be innovative in behavior in order to
achieve specific goals) (Barkley, 1997).
Research on adolescents shows that lower ability to self-regulate is associated
with higher levels of drinking (e.g., Carey, Carey, Carnike, & Meisler, 1990; Glassman,
Werch, & Jobli, 2007) and greater likelihood of practicing unprotected sex (e.g.,
Hernandez & Diclemente, 1992). Among adolescent girls, Mezzich et al. (1997) have
found behavioral and emotional dysregulation to be strongly associated with substance
use and risky sexual behavior. Novak & Clayton (2002) have suggested that adolescents
with lower emotional regulation more likely to initiate smoking in schools with poor
levels of discipline and involvement; and students with high emotional regulation less
likely to initiate smoking despite high risk context. Cognitive-behavioral skills of self-
regulation appear to be effective in reducing one’s levels of drinking (e.g., Werch &
Gorman, 1986; McMurran & Whitman, 1990).

12
Social Self-Control
Sussman’s social self-control scale measures adolescents’ self-control in general
social situations and consists of 10 Likert-type items, which were created based on
previous program development works on adolescent substance abuse prevention (Sussman,
McCuller, & Dent, 2003). The items intend to tap attitudes and behaviors that favor
immediate gratification of urges at the cost of possible social alienation (e.g. “I enjoy
arguing with people”; Sussman, McCuller, & Dent, 2003). In addition, the items are though
to tap such tendencies as speaking one’s mind without thinking, insensitiveness, and self-
centrism (e.g., “If I think something someone says is stupid I tell them so”; Sussman,
McCuller, & Dent, 2003).
Social self-control is an important form of dispositional self-control because it is
likely to influence the quality of adolescents’ social interactions. Good social self-control
may help adolescents to learn and perform prosocial behaviors. Lack of social self-control,
on the other hand, may expose adolescents to higher levels of social conflicts, deviant peer
affiliation, and drug use (Wills, Sandy, & Yaeger, 2000; Wills & Filer, 1996).
Adolescents with higher social self-control are more likely to be liked by their
peers. Social competence is known to be an important correlate of popularity (e.g.,
Gottlieb, Gonso, & Rasmussen, 1975). Sociometric research on adolescent peer social
status has identified, mainly, four types of peer groups: “popular” (i.e., highly liked by
most), “controversial” (i.e., highly liked by some and highly disliked by others),
“neglected” (i.e., neither liked nor disliked), and “rejected” (i.e., highly disliked by most)
(Coie, Dodge, Coppotelli, 1982). Based on this system of grouping, one could argue that
adolescents lacking in social self-control might be classified as either “controversial” or

13
“rejected.” Controversial adolescents with lower social self-control would be those who are
popular among peers like themselves (who also lack social self-control), but are disliked by
peers from other groups. Whereas, rejected adolescents with lower social self-control
would be those who are disliked by most of their peers, primarily because of their lower
social self-control. Research shows that rejected adolescents are often victims to peers’
relational and physical aggression (Crick & Grotpeter, 1996). In addition, controversial
adolescents of the type discussed here are likely to experience intragroup conflicts. Thus,
when adolescents lacking in social self-control interact with peers the result is likely to be
conflict or rejection. When adolescents with lower social self-control interact with
adolescents with higher social self-control, the result is likely to be rejection and relational
and/or physical aggression; and when adolescents with lower social self-control interact
with adolescents like themselves, the result is likely to be conflict.
Social Self-Control and Adolescent Drug Use
The relations between adolescent substance use and a variety of indicators of
general dispositional self-control (e.g., impulsivity, hyperactivity, inattention) have been
relatively well documented in the literature (Pokhrel et al., unpublished). However, the
social aspect of self-control has not been well-studied. Sussman, McCuller, & Dent (2003)
examined the cross-sectional association of social self-control with adolescent substance
use behavior, controlling for 12 personality disorder indices and 4 demographic variables
(White ethnicity, Latin ethnicity, socioeconomic status, and male gender) in a sample of
1050 continuation (alternative) high school youth. The 12 personality disorders assessed
were the following subscales from the Personality Diagnostic Questionnaire (PDQ; e.g.,
Hyler, 1996): paranoid, schizoid, schizotypal, histrionic, narcissistic, borderline, antisocial,

14
avoidant, dependent, obsessive-compulsive, and negativistic (Sussman, McCuller, & Dent,
2003).
The study found social self-control, male gender, and antisocial personality to be
associated with 30-day cigarette smoking, alcohol use, marijuana use, and hard drug use
(Sussman, McCuller, & Dent, 2003); suggesting that social self-control might be a unique
predictor of drug use among continuation high school (CHS) youth. [CHS in California is
attended by youth that are not able to remain in mainstream education for functional
reasons (e.g., truancy, lack of credits, drug use; Sussman, Dent, & Stacy, 2002); see below
for more on Regular High School (RHS) and CHS students]. Moreover, the study
suggested that lack of self-control might be more than just an expression of problem
personality (Sussman, McCuller, & Dent, 2003).
A causal relationship between social self-control and adolescent drug use seems
plausible. As a dimension of self-control, poorer social self-control is likely to reflect
poorer ability to delay gratification. Hence, one might expect adolescents with lower social
self-control to use more drugs in order to immediately gratify the positive outcome
expectancies associated with drug use (Gottfredson & Hirschi, 1990). In addition,
adolescents with poor self-control tend to adopt avoidant coping strategies such as drug use
more often as opposed to effortful types of coping (Wills & Filer, 1996; Wills, Sandy, &
Yaeger, 2000). Further, due to poor social skills and tendency to alienate others,
adolescents with lower social self-control are more prone to experiencing negative life
events (e.g., trouble with school authority, fights with peers; Wills, Cleary et al., 2001).
Negative life events due to poor self-control have been associated with deviant peer
affiliation and increased drug use (Wills, Cleary, et al., 2001; Wills, Resko, Ainette, &

15
Mendoza, 2004). Negative life events may result in deviant peer affiliation as youth who
become alienated or deviant in reaction to negative life events are likely to identify with
each other (Wills, Cleary, et al., 2001). In fact, adolescents who lack self-control and have
difficulty making or keeping friends in general tend to aggregate together (Gottfredson &
Hirschi, 1990; Akers et al., 1979). A peer group formed of such adolescents is likely to
promote deviancy through social learning and differential reinforcement (Akers et al.,
1979).
Currently, it is unclear how stable social self-control is over time. Since it tends to
represent the dispositional type of self-control (Pokhrel et al., unpublished) one might
expect social self-control to be relatively stable. However, some research in
neuropsychology indicates that persistent use of drugs, including nicotine, may affect a
person’s self-control abilities over time (e.g., Robinson & Berridge, 2003; Baler &
Volkow, 2006). The drugs-induced changes in self-control abilities may occur due to
neuroplastic adaptations (Robinson & Berridge, 2003; Baler & Volkow, 2006). Drug abuse
may disrupt some of the neuro-circuitries involved in reward, motivation, inhibitory
control, and memory consolidation (Baler & Volkow, 2006).
Overview of the Present Three Studies

Clearly, social self-control is an understudied aspect of general self-control. Apart
from Sussman, McCuller, & Dent (2003), so far no other study has examined the social
self-control construct as assessed by Sussman’s 10-item measure of social self-control
(see Appendix A, Table 1). Across three studies, the present dissertation examined the
causal impact of social self-control on adolescent substance use and sense of coherence

16
(SOC), a variable that has been consistently linked with mental and physical well-being
(see Chapter 4 for more on SOC).
Studies. The first of the three studies attempted to replicate the findings of the
previous cross-sectional (Sussman, McCuller, & Dent, 2003) on a longitudinal sample.
That is, we examined the prospective associations between baseline social self-control
and later substance use. In addition, the first study examined whether the substance use at
baseline affected the development of social self-control 1-year later. The second study
examined the construct of social self-control at the item level. We used multtrait scaling
method (Hays & Hayashi, 1990) to examine the item discrimination of the social self-
control items in comparison with the items of Zuckerman’s impulsive sensation seeking
scale (Zuckerman, Kuhlman, Thornquist, & Kiers, 1991). Sensation seeking tendency or
the tendency to novel and thrilling experience is considered an attribute of one’s self-
control disposition and is a well known predictor of adolescent substance use (Bardo,
Donohew, & Harrington, 1996). Thus, discriminating the items of social self-control
from the items of impulsive sensation seeking would to some extent validate the non-
redundancy of the social self-control construct. The second study also examined the
longitudinal association between baseline social self-control and substance use one year
later adjusting for impulsive sensation seeking.
The third study first established a confirmatory factor analysis (CFA) model of
social self-control, with its items hypothesized to load on a latent social self-control
construct. Next, the third study used three structural equation models (SEM) to examine
the associations among social self-control, sensation seeking, sense of coherence, and
substance use at two time-points (1 year apart), across low- and high-risk samples of

17
youth. In addition, the present dissertation tested the generalizability of the social self-
control construct across ethnicities and its stability over time by testing the CFA model of
social self-control for measurement invariance across 4 ethnicities (Whites, African
Americans, Hispanics, and Asian Americans) and 3 time-points (baseline, 6-week follow
up, and 1-year follow-up; see Appendix B).
Subjects. The subjects across the three studies involved regular and continuation
high school students (i.e., RHS and CHS). The first and the third studies were based on
the same dataset, which included both RHS and CHS subjects. The second study included
CHS subjects only. In the present set of studies, regular high schools represented
traditional or mainstream public high schools in Southern California; whereas,
continuation high schools represented alternative high schools in the same region.
Continuation high schools in California were originally established to provide an
alternative for students who needed a more flexible schedule (e.g., school day or week)
and educational program than those provided by the traditional high schools [California
Continuation Education Educational Association (CCEA), 2009]. There are over 500
continuation high schools in California that serve more than 70,000 youths (CCEA,
2009). According to the CCEA brochure (CCEA, 2009), the continuation high schools
offer an educational program “that has emphasis on occupational orientation or a work-
study schedule, intensive guidance services to meet students’ special needs, and a
program that will lead to completion with a diploma.” Unlike the RHS students, the CHS
students are not assigned a distinct grade level. A CHS provides a higher teacher-student
ratio compared to a RHS (15:1 versus 30:1) and allows students to complete the required
number of course credits at their own convenient pace (Sussman, Dent, & Stacy, 2002).

18
Students might join continuation high schools for various reasons that make it
difficult for them to remain in regular high schools. Such reasons may include difficulties
in attending school during regular hours or days due to family or work-related
circumstances, inability to complete required number of academic credits within a
specified time, problem behaviors such as substance use and delinquency, and unplanned
pregnancy (Sussman, Dent, & Stacy, 2002). Previous research shows that compared with
a RHS sample, a CHS sample is likely to represent higher percentages of males and self-
identified Hipanics (Latino) (Sussman, Dent, & Stacy, 2002). In addition, the levels of
past-30-day cigarette, alcohol, and marijuana use among the CHS students are likely to
be almost twice as high as compared with the RHS students (Sussman, Dent, & Stacy,
2002). The CHS students are also more likely to report past-12-month violence
perpetration and victimization (Sussman, Dent, & Stacy, 2002). Based on these data, it is
clear that the CHS subjects represent an adolescent population that is at a higher risk for
problem behavior, including substance use. Hence, by including both RHS and CHS
samples the present set of studies were able to test hypotheses across high and low risk
samples of adolescents.

19
Chapter 2
Prospective associations of social self-control with substance use among
youths from regular and continuation high schools
Abstract
In a previous cross-sectional study, poor social self-control was found to be
associated with higher substance use, controlling for 12 personality disorder categories. In
this study, we attempted to find out (a) whether lack of social self-control predicted
substance use one year later, and (b) whether drug use at baseline predicted social self-
control one year later. We surveyed 2081 older adolescents from 9 regular (N=1529) and 9
continuation (alternative) (N=552) high schools in the Los Angeles area. Data were
collected at two time points in an interval of approximately 1 year. Past 30-day cigarette
smoking, marijuana use, hard drug use, and problem drug use at baseline were found to
predict lower social self-control at follow-up, controlling for baseline social self-control
and demographic variables. The effect of problem drug use as a 1-year predictor of social
self-control was found to be moderated by school type (regular or continuation high
school), such that the relationship was significant for continuation high school students
only. Conversely, social self-control was found to predict past 30-day alcohol use,
marijuana use, and problem drug use, controlling for baseline substance use and
demographic variables. For alcohol use, marijuana use, and problem drug use outcomes,
school type was not found to moderate the effects of social self-control, though an
interaction effect was found regarding cigarette smoking. Social self-control was a
significant predictor of cigarette use only at regular high school.

20
Introduction
The present study provides a prospective extension of Sussman, McCuller, & Dent
(2003) and also examines whether substance use variables at baseline predict social self-
control 1 year later. In addition, the present study examines whether relationships between
social self-control and substance use variables are similar for both Regular High School
(RHS) and Continuation High School (CHS) students. Two hypotheses are proposed. First,
we hypothesize that youths low in social self-control would be more likely to use drugs 1
year later, controlling for baseline drug use and demographic variables, and regardless of
type of school. Our assumption is that youth who tend to alienate others through lack of
social self-control skills may differentially associate with peers who are more tolerant of
their deviant behavior (e.g., drug users). [The theory behind the proposition that lower
social self-control is likely to lead to higher substance use over time has been discussed in
Chapter 1]. Second, we hypothesize that baseline drug use will fail to predict social self-
control 1 year later, controlling for baseline social self-control and demographic variables.
Empirical evidence in the literature suggesting a predictive relation between
substance use and low self-control among adolescents does not seem compelling.
Gottfredson & Hirschi’s (1990) consider individuals’ social self-control to be relatively
stable across time. According to Gottfredson & Hirschi (1990), under similar
circumstances, compared to individuals with high self-control, individuals with low self-
control are more likely to commit crime and other analogous behaviors (e.g., substance
use) all through life. They propose that the quality of social environment (i.e., family
environment) that children are exposed to when they are young determines their self-
control characteristics (Gottfreson & Hirschi, 1990), which is likely to remain stable across

21
developmental stages. Studies have found empirical support for Gottfredson & Hirschi’s
(1990) claim that an individual’s self-control does not change with his or her age or
differing contexts of residence (e.g., Ameklev, Grasmick, & Bursik, 1999). Thus, if self-
control is similar to a personality trait, then drug use or any other environmental influence
(e.g., deviant peer association) should not alter social self-control over time.
Methods
Subjects and Data Collection
Nine school districts from two counties in southern California were recruited for
participation in this study. In order to be eligible to participate, districts needed to
contain at least one RHS and one CHS, each with an enrollment between 50 and 2000
students. The CHS and RHS were selected from the same school districts (one RHS and
one CHS from each district) as 9 pairs of schools.
All students in the randomly selected classrooms first completed project
participation information cards that contained their name, address and phone number.
Then, students were provided with parental consent forms to take home for their parents’
signature, indicating approval or refusal of student participation in the study surveys.
Subjects also assented to their involvement. The homes of those students for whom no
consent form was returned were called by project staff to request verbal parental consent
for survey participation. The informed consent procedure was carried out as per the
proposal approved by the University of Southern California’s Institutional Review Board
(IRB).
A total of 3908 high school students were enrolled in the selected classrooms. Of
these, 2751 (70.4% of the enrollment roster) completed pretest questionnaires (1902 RHS

22
and 849 CHS subjects). There were several reasons why we did not have access and,
therefore, could not collect data from all enrolled students, including chronic absenteeism
(i.e., participant information cards could not be filled out after three weeks of daily
attempts; approximately 80% of those not enrolled in the study), either the parent or the
student declined participation (approximately 5% of those not enrolled in the study), and
students were absent on testing days (which involved a full week of daily attempts at
reaching students for pretest and posttest, respectively; approximately 15% of those not
enrolled in the study). In general populations, those absent on testing day report more
problem prone characteristics (Sussman, Dent, Stacy, Burton, & Flay, 1995); however,
among populations at higher risk, those not surveyed at posttest fail to differ greatly from
the full sample (Dent, Sussman, & Stacy, 1997). No incentive was provided for
participation.
Of the students who completed pretest questionnaires, 2081 also completed the
follow-up questionnaire an average of 16.5 months later. The sample of 2081 students
(N
RHS
= 1529; N
CHS
= 552) constitutes our analysis sample. As with previous studies, our
follow-up sample for continuation high school was 65.0% of students assessed at baseline
(Sussman, Dent, & Stacy, 2002). Subject retention at follow-up was higher for the RHS
sample: 80.4% of the baseline RHS sample. The follow-up surveys were conducted by
telephone using an interview format. The interviewers (previously unknown to the
subjects) contacted the subjects at home, read the questionnaire, and recorded the
responses in a survey form. Survey forms and response categories were identical to the
in-school questionnaire format used at baseline, and test not to show differences in
response patterns (see discussion in Sussman, Dent, Stacy, & Craig, 1998).

23
Measures
Demographics were assessed using an ethnic indicator (e.g., “what is your ethnic
background?” Followed by 6 response options, including an open-ended “Other” option),
gender, and parental education indicators (the highest educational level reached across
father/step-father or mother/step-mother was measured using 6-point scale, ranging from
not completed elementary school to completed graduate school).
Ten social self-control items were used (see Appendix A, Table 1). The items
were set on a 4-point scale from “never” to “always,” and tap behavior in which one
seems driven to social excitement even though it distances oneself from social harmony,
involving open expression of whatever it is that one feels at the moment which is likely to
alienate others, and a desire for the social world to adjust to one's behavior. Example
items include: "I enjoy arguing with people," "If I think something someone says is
stupid I tell them so," and "My mouth gets me in trouble a lot" (alpha=.73 for entire
sample at baseline; alpha=.74, for RHS; alpha=.68, for CHS).
Substance use was measured using single fill-in-the blank item measures.
Participants were asked “how many times have you used cigarettes in the last 30 days?”
Thirty-day use was also asked in regards to alcohol, marijuana, and 6 categories of hard
drugs (which were summed together to form an index: cocaine, stimulants, inhalants,
hallucinogens, ecstasy, and other). Subjects were provided with eight response categories,
which included 0-10, 11-30, 31-50, 51-60, 61-70, 71-80, 81-90, and 91-100+ times.
These drug use behavior items were adapted from previous self-report questionnaires
(e.g., Graham, Flay, Johnson et al., 1984; Johnston, O’Malley, & Bachman, 1999; Stacy,
Newcomb, & Bentler, 1991; Sussman et al., 1995) and have been generally found to have

24
2-week test-retest reliability of above 0.75 (Sussman, Dent, & Leu, 2000). An index of
past 30-day problem drug use was composed of the 11-item Problem Consequences
Subscale of the Personal Experience Inventory (PEI; these 4-point items were averaged;
Winters & Henly, 1989). This subscale gives a good discriminant validity between
interview-derived diagnostic groups (biserial correlation = 0.72; see Sussman, Dent, &
Leu, 2000).
Attrition analysis
Single sample t-tests and chi-square test for specified proportions were calculated
for all variables examined in the study comparing the analysis sample with the full
baseline sample for each school type (RHS and CHS). For either school type, no
statistically significant differences were detected between the samples on age (p=0.14, for
RHS; p=0.14, for CHS), gender (p=0.13, for RHS; p=0.14, for CHS), ethnicity (p=0.63,
for RHS; p=0.69, for CHS), parental education (p=0.70, for RHS; p=0.76, for CHS),
cigarette use (p=0.25, for RHS; p=0.42, for CHS), alcohol use (p=0.38, for RHS; p=0.96,
for CHS), marijuana use (p=0.28, for RHS; p=0.54, for CHS), hard drug use (p=0.15 for
RHS; p=0.63, for CHS), problem drug use (p=0.87, for RHS; p=0.71, for CHS), and
social self-control (p=0.78, for RHS; p=0.66, for CHS). These results imply that attrition
bias did not pose serious threats to the present study.
Analyses and Results
Table 1 and Table 2 show the demographic and drug use characteristics of the
analysis sample at baseline and follow-up for RHS and CHS students. The Pearson
correlation coefficients among cigarette, alcohol, marijuana, and hard drug use for RHS
subjects at baseline ranged from 0.25 (between alcohol and hard drug use; p<0.0001) to

25
0.41 (between marijuana and cigarette use; p<0.0001); for CHS subjects at baseline
correlations ranged from 0.29 (between alcohol and hard drug use; p<0.0001) to 0.46
(between alcohol and marijuana use; p<0.0001). At follow-up, the correlations among
the drugs ranged from 0.28 (between alcohol and hard drug use; p<0.0001) to 0.49
(between hard drug and marijuana use; p<0.0001) for RHS subjects and from 0.25
(between alcohol and hard drug use; p<0.0001) to 0.39 (between marijuana and cigarette
use; p<0.0001). For both RHS and CHS samples, a statistically significant increase in
social self-control was found to have taken place between baseline and follow-up
(p<0.0001).
Ten models were examined using PROC MIXED and PROC GLIMMIX on SAS
statistical software (Version 9.1). PROC MIXED and PROC GLIMMIX analytical
techniques account for random model effects in regression analyses with continuous and
dichotomous outcome variables respectively. Accounting for random model effects was
important to our analyses as our subjects were nested within schools. The first five
models examined the prediction of drug use one year later (cigarette smoking, alcohol
use, marijuana use, hard drug use, or problem drug use) from baseline social self-control,
controlling for baseline drug use, age, gender, Latino ethnicity, school type (RHS or
CHS), and parents’ education. In addition, each model also tested for the interaction
effect of school type on the relationship between social self-control and specific drug use
by including an interaction term in the model. All variables were centered on their means
and specified a standard deviation of 1. The interaction term was created after
standardizing the variables (Aiken & West, 1991). Since the drug use variables were

26
highly skewed towards non-use, they were dichotomized into having used a specific drug
in the past 30 days or not.
Table1. Demographic, baseline drug use, and social self-control characteristics of
the analysis sample.

RHS (N=1529) CHS (N=552)

Mean
age***

14.8 (SD=0.86) 16.8 (SD=0.74)
Gender***

% Male 51.6 57.2

% Female

48.4 42.8
Ethnicity***

% Latino

60.6 71.2

% Non-Latino

39.4 28.8
Parents
education***

% Some college
or below
69.7 80.0

% Full college or
above
30.3 20.0
% using cigarette***

8.3 35.4
% using alcohol ***

31.3 58.0
% using marijuana***

13.7 41.0
% using hard drug***

4.90 17.6
Mean Problem drug use***

1.09 (SD=0.24) 1.21 (SD=0.33)
Mean social self-control

2.87 (SD=0.50) 2.88 (SD=0.51)
Note: *p<0.05; **p<0.001; ***p<0.0001; SD = Standard deviation; % using drug
pertains to any use in last 30 days; RHS = Regular High School; CHS = Continuation
High School

27
Table 2. Drug use and social self-control characteristics of the analysis sample at one
year follow-up.

RHS (N=1529)

CHS (N=552)
% using cigarette*** 11.8

39.0
% using alcohol*** 34.0

58.1
% using marijuana*** 12.9

31.5
% using hard drug*** 5.59

12.6
Mean Problem drug use*** 1.06 (SD=0.20)

1.14 (SD=0.30)
Mean social self-control 3.08 (SD=0.46)

3.07 (SD=0.45)
Note: *p<0.05, **p<0.001, ***p<0.0001; SD = Standard deviation; % using drug
pertains to any use in last 30 days; RHS = Regular High School; CHS = Continuation
High School.

The results based on the first five analytical models are shown in Table 3. Social
self-control predicted alcohol use, marijuana use, and problem drug use controlling for
baseline drug use and the five demographic variables. Although not statistically
significant, social self-control at baseline was inversely associated with both cigarette use
and hard drug use at follow-up. No significant interaction was detected between social
self-control and school type for alcohol use (p=0.10), marijuana use (p=0.21), hard drug
use (p=0.44), and problem drug use (p=0.39). However, the interaction between school
type and social self-control was found to be statistically significant for cigarette use
(p=0.01). When examined separately by school type, social self-control at baseline was
found to predict cigarette use significantly at follow-up only among RHS subjects
(OR=0.79; 95% CI = 0.67-0.92).

28
The second five models examined the prediction of social self-control at one year
follow-up from the level of drug use (cigarette smoking, alcohol use, marijuana use, hard
drug use, or problem drug use) at baseline, controlling for baseline social self-control,
school type, female gender, Latino ethnicity, and parents’ education. All variables were
standardized prior to running the models. The term for interaction between specific drug
variable and school type was also included in each model. The results of this set of
analyses are summarized in Table 4. The inverse association between drug use at baseline
and social self-control at follow-up was found to be significant for cigarette use (p=0.05),
marijuana use (p=0.04), hard drug use (p=0.04), and problem drug use (p=0.002). No
interaction was found taking place between school type and cigarette use (p=0.26),
alcohol use (p=0.38), marijuana use (p=0.84), and hard drug use (p=0.48). However,
school type was found to modify the effect of problem drug use at baseline on social self
control at follow-up (p=0.03). Further analysis showed that higher problem drug use at
baseline was significantly associated with lower social self-control at follow-up for CHS
students (standardized β=-0.13; p<0.001) but not for RHS students (β=-0.03; p=0.23).

29
Table 3: Social self-control at baseline as a predictor of drug use one year later,
after controlling for age, female gender, Latino ethnicity, parents’ education, school
type, and baseline drug use.
Odds Ratio (95% Confidence Interval) Standardized
β

Cigarette
smoking

Alcohol use Marijuana
use
Hard drug
use
Problem
drug use
Baseline
drug
variable

2.13***
(1.94-2.35)
1.82***
(1.65-2.01)
1.92***
(1.74-2.10)
2.05***
(1.87-2.25)
0.35***
Social self-
control

0.91
a

(0.79-1.04)
0.80***
(0.72-0.89)
0.81***
(0.71-0.91)
0.89
(0.76-1.05)
-0.09**
Age

1.01
(0.89-1.14)

0.99
(0.90-1.09)
0.92
(0.82-1.04)
0.74**
(0.62-0.88)
-0.07**
Female
gender

0.86**
(0.76-0.96)
0.83**
(0.76-0.92)
0.80**
(0.71-0.90)
0.94
(0.82-1.07)
-0.06*
Latino
ethnicity

1.05
(0.91-1.20)
1.09
(0.99-1.20)
1.02
(0.91-1.15)
1.13
(0.98-1.33)
0.02
Parents’
education

1.04
(0.93-1.17)
0.93
(0.85-1.02)
1.00
(0.89-1.12)
1.04
(0.91-1.20)
-0.02
School type
(RHS=0;
CHS=1)

7.69
b
***
(3.87-15.27)
2.94***
(2.67-3.24)
3.86***
(2.55-5.94)
4.01***
(1.45-11.12)
0.35**
Note: *p<0.05; **p<0.001; ***p<0.0001; All independent variables were centered on
their means.
a
Interaction between school type and social self-control statistically significant (p=0.01).
The reported OR is when school type=0 (school type was centered on its mean).
b
Interaction between school type and social self-control statistically significant
(p=0.01).The reported OR is when social self-control=0 (social self-control was centered
on its mean).

30
Table 4: Effects baseline drug use on social self-control one year later after
controlling for baseline social self-control, age, female gender, Latino ethnicity,
parents’ education, and school type.

Note: *p<0.05; ***p<0.0001; All independent variables were centered on their means;

a
Interaction between school type and problem drug use was found to be statistically
significant (p<0.03). The reported β is when school type=0 (school type was centered on
its mean).

Standardized β

Cigarette
smoking

Alcohol use Marijuana
use
Hard drug
use
Problem
drug use
Baseline
social self-
control

0.50*** 0.50*** 0.50*** 0.50*** 0.49***
Drug use at
follow-up

-0.04* -0.03 -0.04* -0.04* -0.07
a
*
Age

0.06* 0.06* 0.06* 0.06* 0.05*
Female
gender

-0.04* -0.03 -0.04* -0.03 -0.04*
Latino
ethnicity

-0.001 -0.006 -0.007 -0.005 -0.004
Parents’
education

0.005 0.004 0.004 0.007 0.002
School type

0.01 0.008 0.007 0.007 0.0008

31
Discussion
Higher levels of social self-control may be protective against substance use
experimentation (e.g., Sussman, Dent, & Stacy, 2002). In Sussman, McCuller, & Dent
(2003), which was a cross-sectional study, high social self-control was inversely related
to drug use, controlling for relatively unchangeable disorders of personality. This result
suggested to us that social self-control is not merely a facet of a problem personality. The
present study attempted to examine whether the findings of the previous study would
replicate longitudinally in a larger, more varied sample. Extending the sample from the
previous study which included only the “high risk youth” (i.e. CHS students), the present
study included both regular and continuation high school students. In addition, the
present study also examined whether drug use had prospective effects on social self-
control.
The findings regarding the differences in demographic and substance use
characteristics between the CHS and RHS subsamples generally supported our previous
findings in that we found the CHS subsample to represent higher percentages of males,
Hispanics, lower-SES subjects, and substance users (e.g., Sussman, Dent, & Stacy, 2002).
However, the baseline differences in past-30-day substance use prevalence between RHS
and CHS students in the present sample were greater than found in previous studies
(Sussman, Dent, & Stacy, 2002). We found that compared to the RHS students, the CHS
students tended to report 4 times higher cigarette use, 3 times higher marijuana use, and 3
times higher hard drug use. However, relative to baseline, the school-type differences in
substance use were found to have narrowed down at follow-up, even though by small
amounts. It is interesting to note that somewhat drastic changes were noticed in marijuana

32
and hard drug use prevalence among CHS students between baseline and follow-up. The
almost 10% and 5% decreases in recent marijuana and hard drug use prevalence,
respectively, between baseline and 1-year follow-up among CHS students may require
closer attention. One may speculate that the alternative high school system may be
protective against students’ substance use behavior. Although future studies need to
explore this possibility in detail, it seems that providing higher risk adolescents with a
more congenial learning environment might have protective effects on their problem
behavior.
In the previous cross-sectional study (Sussman, McCullar, & Dent, 2003),
Hispanic ethnicity was found to be strongly associated with higher cigarette use. In
addition, higher level of parental occupation was found to protective against alcohol use.
In the present study, we did not find ethnicity or parental education to predict substance
use 1 year later, after controlling for baseline substance use. However, even after
controlling for baseline use, male gender predicted higher levels of cigarette, alcohol, and
marijuana use.
Contrary to our assumption that social self-control might be relatively stable over
time, our findings indicated that the mean social self-control for both RHS and CHS
samples increased significantly from baseline to follow-up. This increase in social self-
control suggests that rather than being an immutable personality characteristic, social
self-control is likely to be a complex cognitive-behavioral attribute that might improve as
one gets older. Wills and colleagues (Wills, Sandy, & Yaeger, 2000; Wills & Dishion,
2004) have argued that self-control requires a sophisticated level of cognitive and social
development which takes place through adolescence.

33
Our hypothesis that drug use at baseline would fail to predict social self-control 1
year later was supported only in case of alcohol use. In cases of cigarette smoking,
marijuana use, and hard drug use, our results did not support the hypothesis. After
controlling for baseline social self-control and demographic variables, cigarette smoking,
marijuana use, and hard drug use were found to be inversely related to social self-control
at follow-up 1 year later across both school types. Research in neuropharmacology
suggests that persistent use of drugs, including nicotine, may affect the neurophysiology
associated with behavioral inhibition through neuroplastic changes (Robinson & Kolb,
2004). Drug use behaviors are also likely to affect social self-control through differential
deviant peer selection and subsequent imitation of socially inappropriate behavior (Akers,
1979). It is possible that alcohol use failed to predict social self-control because most of
the alcohol users may not be abusing the drug. Currently, the extent of research
examining substance use as a prospective predictor of self-control seems to be limited.
Future research on adolescent self-control and drug use behavior may need to examine
the relationship more closely in order to explain the mediators linking drug use to social
self-control. As regards the finding that problem drug use at baseline had significant
effects on the social self-control of CHS subjects, but not RHS subjects, the implication
might be that negative interpersonal consequences of substance use (e.g., trouble with
teachers, peers, and parents due to drug use) may further alienate the high risk youth
socially and encourage them to behave more irresponsibly in social situations later on.
Our hypothesis that social self-control at baseline would predict drug use a year
later was partly supported. Social self-control at baseline showed significant inverse
associations with alcohol, marijuana use, and problem drug use 1 year later, controlling

34
for the specific baseline drug variable and demographic covariates. The effects of
baseline social self-control on these substance use outcomes were found to be similar
across RHS and CHS. However, school type was found to moderate the effects of social
self-control on cigarette use. That is, social self-control had a significant inverse effect on
follow-up cigarette use for RHS students only. Our failure to detect significant
association between social self-control and hard drug use might be due to relatively lower
prevalence of hard drug use among the subjects. The results were in the expected
direction in the present investigation.
Failure to find a longitudinal association between social self-control and cigarette
use among CHS youth could be interpreted as an implication that cigarette use in the
CHS context is considered less deviant (or more socially acceptable) than in the RHS
context. If it is possible that lack of social self-control predisposes one to alienate others
and use drugs as a shared activity that others who also lack social self-control may tend to
engage in (Allison, Leone, & Spero, 1990) then perhaps the students lacking in social
self-control in the CHS social environment do not gravitate toward deviant peers whose
deviancy is attributed exclusively to smoking.
To see whether the association between social self-control and cigarette use
differed by school type cross-sectionally at baseline, we conducted a regression analysis
following the same procedures outlined above (see Results & Analyses) for the
longitudinal analyses. No significant interaction was detected between school type and
social self-control (p=0.47). In addition, social self-control was inversely and
significantly associated with cigarette use among both RHS and CHS subjects,
controlling for demographic covariates (p<0.001) when the analysis was run separately

35
by school type. This cross-sectional finding was similar to the findings of our previous
study. Thus, only further longitudinal studies may be able to elucidate whether lack of
social self-control skills is causally related to cigarette use, or that cigarette smoking
might reflect contextual differences in appropriateness.
There are several limitations of this study. Although we tested all the relevant
variables for indications of a bias due to attrition, we didn’t test several others that were
not measured in the study. For example, indicators of SES other than parental education,
such as parental income and number of persons living in a house were not measured in
the study. In addition, our data may not generalize to all Los Angeles area RHS and CHS.
Although students were randomly selected at the classroom level, some selection bias
might have been introduced to the data at the level of school, which was based on
convenience sampling. Furthermore, about 30% of the students enrolled in the selected
classes did not fill out the baseline survey. The fact that these students could have been
more likely to show problem behavior (Sussman et al., 1995) might have also biased our
data.

36
Chapter 3
Social self-control versus impulsive sensation-seeking: item
discrimination and one-year prospective prediction of adolescent
substance use
Abstract
The present study used a multitrait scaling analysis method to examine the
construct validity of a social self-control measure as compared with the impulsivity and
sensation-seeking subscales of Zuckerman’s Impulsive Sensation Seeking Scale. In
addition, the study examined social self-control as a longitudinal predictor of adolescent
substance use 1 year later, controlling for sensation-seeking and demographic variables
(i.e., age, gender, ethnicity, and parental education). Data were collected at baseline from
894 adolescents (mean age= 16.32; SD=1.34) from 14 continuation high schools in
California, of whom 566 were followed-up 1 year later. The results indicated that 8 of the
10 hypothesized social self-control items showed adequate evidence of convergence and
item discrimination. Furthermore, social self-control was found to be a significant
predictor of cigarette use 1 year later, adjusting for sensation-seeking, baseline use, and
demographic variables.
Introduction
So far it is not clear whether the 10-item social self-control measure represents a
unique self-control variable or taps the same constructs as represented by some of the
other dimensions of generalized self-control such as impulsivity and sensation-seeking
(e.g., Gottfredson & Hirschi, 1990). Sensation-seeking has been defined as the tendency

37
to seek varied, novel, and stimulating experiences or take risks to undergo such
experiences (Zuckerman, 1979). Impulsivity has been defined variously as the inability to
inhibit behavior (e.g., Milich & Kramer, 1984); tendency to act before thinking or act
without plan or foresight (e.g., Eysenck & Eysenck, 1977; Schalling, 1978); inability to
control temptations or urges (e.g., Gordon, 1979); and the tendency to act in situations
where inhibition of such actions would provide benefit (Martin et al., 1994). The present
study examines how well the items of the social self-control scale converge with each
other and discriminate from the items of the impulsivity and sensation-seeking subscales
of Zuckerman’s impulsive sensation-seeking scale (Zuckerman, Kuhlman, Thornquist, &
Kiers, 1991).
Zuckerman’s impulsive sensation-seeking scale consists of 19 items that tend to
factor out into 11-item sensation-seeking and 8-item impulsivity subscales (Zuckerman et
al., 1991; Ames, Zogg, & Stacy, 2002). Impulsivity and sensation-seeking are likely to
represent different constructs (Schalling, 1978; Petry, 2001; Magid, MacLean, & Colder,
2007) with separate pathways to adolescent drug use behavior (Magid et al., 2007).
Individuals high in sensation seeking may use drugs to maintain a functionally optimal
level of cortical arousal (Zuckerman, 1979). In addition, it has been suggested that
individuals high in sensation-seeking generally tend to show higher biological sensitivity
towards the reinforcing effects of pleasurable stimuli associated with drugs of abuse
(Cloninger, 1994; Cloninger, Svrakic, & Przybeck, 1993; Hegerl, Lipperheide, Juckel, &
Schmidt, 1995; Zuckerman, Kuhlman, Joireman, Teta, & Kraft, 1993). Impulsive
individuals tend to choose short-term reinforcers such as drugs over long-term reinforcers
(Colinger, 1994). Moreover, impulsive individuals are prone to adopt avoidant coping

38
styles, which include the use of drugs in order to cope with stress (Wills, Sandy, &
Yaeger, 2000). Both impulsivity and sensation seeking have been consistently linked
with higher adolescent cigarette use (e.g., Burt, Dinh, Peterson, & Sarason, 2000; Brook,
Whitman, & Gordon, 1981), alcohol use (e.g., Donohew et al., 1999; Colder & Chassin,
1997; Earleywine and Finn, 1991), marijuana use (e.g., Crawford, Petnz, chou, Li, &
Dwyer, 2003; Donohew et al., 1999), and hard drug use (e.g., Parrott, Sisk, & Turner,
2000).
In the present study, we analyze the data in two steps. First, we use Hays &
Hayashi’s (1990) method of multitrait scaling to analyze the item convergence and
discrimination of the social self-control, impulsivity, and sensation-seeking scales. Item
convergence and discrimination can be considered minimal requirements of convergent
validity. Convergent validity refers to how well the different ways of measuring the same
trait are intercorrelated to each other whereas discriminant validity refers to the degree to
which traits are distinct (Campbell & Fiske, 1959). Multitrait scaling (Hays & Hayashi,
1990) applies the logic of multitrait-mutlimethod (MTMM) procedures to analyze item
convergence and discrimination when the Likert-type scales are the only method used to
assess multiple traits. Using these procedures, we test the hypothesis that the items
thought to measure social self-control converge toward each other and diverge from the
impulsivity and sensation seeking items.
Second, we extend the findings of the multitrait scaling analysis by comparing the
prospective effects of the validated scales on adolescent substance use 1 year later.
Provided that the items hypothesized to measure each respective trait show adequate
convergence and divergence in the expected directions, scales were created for each trait

39
and the following hypothesis was tested: higher social self-control predicts lower
substance use (i.e., cigarette, alcohol, marijuana, and hard drug use) 1 year later, both
before and after adjusting for sensation-seeking and impulsivity.
Methods
Subjects and Data Collection
Data were collected at baseline and one year later as part of a school-based teen
drug abuse prevention study [see Valente et al. (2007) for information on study design
and program components]. The baseline data were collected from 894 students from 14
continuation high schools that represented 7 school districts and 3 counties in southern
California. Trained data-collectors administered the self-report questionnaires in the
classroom at baseline. The survey questionnaire was designed so as not to take more than
one class period (about 45 minutes) to complete. The study followed the informed
consent protocol approved by the university’s Institution Review Board (IRB). To
participate in the study, the minor subjects were required to have parental consent.
Student assent was required for all students, including minors. Students were explained
that they could choose to discontinue participation at any time. The 1-year follow-up data
collection was conducted on site or by phone. For the phone interview, trained
interviewers (previously unknown to the subjects) contacted the subjects at home, read
the questionnaire, and recorded the responses in a survey format (as letters or numbers).
Of the 894 students who completed the baseline survey, 566 participated in the 1-
year follow-up. That is, about 63% of the baseline subjects were retained at follow-up.
According to past research, this rate of attrition is expected among CHS students
(Sussman, Dent, & Stacy, 2002). Single sample t-tests and chi-square tests for specified

40
proportions were conducted for all variables examined in the present study comparing the
total longitudinal sample with the total baseline sample on the baseline data. No
statistically significant differences were detected between the samples on age [t (558) = -
0.96, p=0.34], gender [χ
2
(1, N=563) = 0.04, p=0.84], ethnicity [χ
2
(1, N=546) = 0.36,
p=0.55], parental education [χ
2
(1, N=566) = 1.66, p=0.89], cigarette use [χ
2
(1, N=552)
= 0.10, p=0.75], alcohol use [χ
2
(1, N=555) = 0.0009, p=0.98], marijuana use [χ
2
(1,
N=553) = 0.11, p=0.73], hard drug use [χ
2
(1, N=546) = 0.43, p=0.51], social self-control
[t (518) = -0.38, p=0.70], sensation-seeking [t (554) = -0.89, p=0.38], and impulsivity [t
(557) = -0.42, p=0.67]. Hence, it may be concluded that the total sample size at follow-up
was comparable to a random sub-sample of the baseline subjects. Baseline descriptive
statistics on the demographics, substance use, and other relevant variables are presented
in Table 5.

41

Table 5. Sample characteristic at baseline (N=894).

Measures
Demographics were assessed using an ethnic indicator (e.g., 6 response options,
including an open-ended “Other” option), gender, and parental education indicators. The
highest educational level reached across father/step-father or mother/step-mother was
Means age (SD)

16.32 (1.34)
% Female

39.8
Ethnicity
% African-American 6.30
% Asian 2.68
% Hispanic/Latin 69.0
% American Indian 0.82
% Non-Hispanic White 12.6
% Other

8.63
Parental education
% Below elementary
school
11.8
% Below high school 25.6
% Completed high
school
21.7
% Some college 22.0
% Completed college 13.1
% Completed graduate
school

6.0
% using cigarettes

39.9
% using alcohol

60.6
% using marijuana

46.1
% using hard drugs

26.4
Mean social self-control
(SD)

2.82 (0.57)

42
measured using a 6-point scale, ranging from not completed elementary school to
completed graduate school (see Table 1; Hollingshead & Redlich, 1958).
Social self-control was assessed using the 10 items of Sussman’s social self-
control scale (Sussman, McCuller, & Dent, 2003). This measure of social self-control
includes items such as: “I enjoy arguing with people” (Cronbach’s alpha at baseline in the
present sample=0.79).
Sensation-seeking and impulsivity were assessed using the 8 impulsiveness and 11
sensation-seeking subscale items from the Zuckerman-Kuhlman Personality
Questionnaire (Zuckerman et al., 1991). The impulsivity and sensation-seeking items
were distinguished based on the past research (e.g., Ames et al., 2002). Examples of the
impulsivity items include: “I tend to begin a new job without much advance planning on
how I will do it,” and “I often do things on impulse” (Cornbach’s alpha= 0.57; see Table
6). The sensation-seeking items involve general novelty-seeking/ risk-taking tendencies
(Cornbach’s alpha=0.75). Examples of the sensation-seeking items include “I like to have
new and exciting experiences and sensations even if they are a little frightening,” and “I
like doing things just for the thrill of it” (see Table 6). For both sets of items, participants
were asked to respond true or false to statements that they might use to describe
themselves.
Substance use was assessed as self-reports of past 30-day substance use behavior
with an 11-item rating scale. Participants were asked how many times they have used
various drugs in the last month. The list of drugs included 11 drug categories. The
response choices ranged between "1" or “never used” to “11” or "91-100+ times” in

43
intervals of 10 (e.g., "1-10 times", "11-20 times"). The reliability and predictive validity
of these types of items have been previously established (Graham et al., 1984). Single
items assessed use of cigarettes, alcohol, and marijuana. A past-30-day hard drug use
index was created by summing past-30-day responses across cocaine, ecstasy,
hallucinogens (e.g., LSD), tranquilizers (e.g., valium), stimulants or amphetamines (e.g.,
speed, crystal meth), opiates (e.g., Heroin), inhalants, and other club drugs (e.g.,
ketamine).
Statistical Analyses
Multitrait Scaling Analyses. Hays & Hayashi’s (1990) method was used to
examine the convergent and discriminant validity of the social self-control measure based
on the baseline data. This method involves calculation of item/total scale correlations.
First, indexes of social self-control, impulsivity, and sensation-seeking were created by
summing up each set of items Next, correlations between each item across the three sets
of items and each of the three indexes were calculated. The correlation between an item
and the index it is hypothesized to measure (item/same-index correlation) was corrected
for item/index overlap. As shown in Table 6, the rows of the correlation matrix
represented the correlations of an item with each of the three indexes and the columns
represented the correlations of an index with each item across the three sets of items.
Next, the item/same-index correlations were compared with the correlations
between the item and the indices it is hypothesized not to measure (item/different-index
correlation), examining whether the differences in coefficients between item/same-index
correlation and item/different-index correlation were more than two standard errors (Hays
& Hayashi, 1990). If the coefficient for item/same-index correlation was higher by more

44
than two standard errors compared with the other two correlation coefficients in the row
(i.e., item/different index correlations), then the item was considered sufficiently
discriminating (Hays & Hayashi, 1990). Convergence among the items of the same index
was established by the size of the item/same index correlations: items that showed
corrected item/same-index correlations of 0.40 or above were considered sufficiently
convergent (Hays & Hayashi, 1990). Discriminant validity between indices also was
tested by examining zero-order correlations and correlations adjusted for unreliability of
measurement between the indices. These two correlations should be less than unity for
the indexes to be sufficiently discriminant (Hays & Hayashi, 1990).
Regression Analyses. All regression analyses were carried out using the
GLIMMIX procedure in SAS (Version 9.1). PROC GLIMMIX performs estimations for
generalized linear mixed models, which incorporates normally distributed random effects
(Schabenberger, http://www2.sas.com/proceedings/sugi30/196-30.pdf). The procedure is
suitable for the proposed analyses primarily for two reasons. First, the individual subjects
(units of analysis) in the present study are nested within schools. Second, since the
distributions of the substance use outcome variables were skewed toward non-use, the
variables were dichotomized in the present study, thus resulting in a binomial distribution
of the dependent variables.
Three regression models were analyzed for each substance use outcome (i.e.,
cigarette, alcohol, marijuana, and hard drug). Model 1 examined the effects of baseline
social self-control on 1-year follow-up substance use, controlling for baseline substance
use and the following four demographic variables: age, gender, ethnicity, and parents’
education. Model 2 examined the effects of baseline sensation-seeking on follow-up

45
substance use, controlling for baseline substance use and the four demographic variables.
Model 3 examined the effects of baseline social self-control and sensation-seeking on
follow-up substance use, controlling for each other, baseline substance use and the four
demographic variables. (Impulsiveness was excluded from Models 2 and 3, due to
inadequate item discrimination, which is described in the Results.) To make the
interpretation of the results easier, all continuous independent variables were
standardized. All three models adjusted for program effects. One-tailed test of
significance at alpha=0.05 was used to test the unidirectional hypotheses involving social
self-control, sensation-seeking, and substance use. Two-tailed test of significance at
alpha=0.05 was used to test the relationships between demographic variables and
substance use.

46
Table 6. Multitrait/multi-item correlation matrix for social self-control, sensation-
seeking, and impulsivity.
Social self-
control
Impulsi
vity
Sensation-
seeking
Social self-control
I enjoy arguing with people. 0.49* 0.21 0.15
I express all of my feelings. 0.32* 0.01 0.07
If I think something someone says is stupid I tell
them so.
0.45* 0.10 0.17
If I am angry I act like it. 0.56* 0.18 0.15
My mouth gets me in trouble a lot. 0.61* 0.26 0.19
I do things just to get attention. 0.51* 0.27 0.18
Sometimes I provoke people just for the fun of it. 0.53* 0.31 0.30
My feelings get hurt easily. 0.30* 0.11 -0.002
I hate being wrong. 0.43* 0.17 0.12
I say things that I regret later. 0.48* 0.27 0.15
Impulsivity
I tend to begin a new job without much advance
planning on how I will do it.
0.16 0.23* 0.18†
I usually think about what I am going to do before
doing it.
0.17 0.34* 0.18
I often do things on impulse. 0.22 0.37* 0.24
I very seldom spend much time on the details of
planning ahead.
0.05† 0.11* 0.08†
Before I begin a complicated job, I make careful
plans.
0.14† 0.19* 0.13†
I tend to change interests frequently. 0.13 0.25* 0.27†
I often get so carried away by new and exciting
things and ideas that I never think of possible
complications.
0.24 0.35* 0.39†
I am an impulsive person. 0.29 0.39* 0.38†

47
Table 6 (Continued)
Sensation-seeking
I like to have new and exciting experiences and
sensations even if they are a little frightening.
0.18 0.22 0.41*
I would like to take off on a trip with no
preplanned or definite routes or timetable.
0.07 0.22 0.33*
I enjoy getting into new situations where you can't
predict how things will turn out.
0.16 0.26 0.42*
I like doing things just for the thrill of it. 0.20 0.27 0.52*
I sometimes like to do things that are a little
frightening.
0.16 0.18 0.46*
I'll try anything once. 0.12 0.26 0.33*
I would like the kind of life where I’m on the
move and traveling a lot, with lots of change and
excitement.
0.04 0.13 0.29*
I sometimes do "crazy" things just for fun. 0.23 0.27 0.47*
I like to explore a strange city or section of town
by myself, even if it means getting lost.
0.04 0.13 0.31*
I prefer friends who are excitingly unpredictable. 0.19 0.27 0.39*
I like "wild" uninhibited parties. 0.17

0.31 0.44*
Cronbach’s alpha 0.79 0.57 0.75
Note. Standard error of correlation= 0.03. *Item-scale correlations for hypothesized
scales (corrected for item overlap). †Correlation within two standard errors of the
correlation of the item with its hypothesized scale.

Results
Multitrait scaling analyses. Table 6 shows the results of the multitrait scaling
analyses. All 10 social self-control items showed discriminant validity. That is, none of
the item/same-index correlations for the social self-control items was within two standard
errors of the item/different-index correlations. However, two of the 10 items were not
found to be adequately convergent; the item/same-index correlations for these two items
were less than 0.40 (see Table 6). Six of the 8 impulsivity items failed to show
discriminant validity (see Table 6). In addition, none of the 8 impulsivity items showed
adequate convergence. Although all sensation-seeking items were sufficiently

48
discriminant, 5 of the 11 items were not found to be convergent (see Table 6). However,
the social self-control, impulsivity, and sensation-seeking indexes were found to be
discriminant: both the zero-order correlation and the unreliability-adjusted correlation
among them were found to be less than unity (see Table 7).
Table 7. Correlations between social self-control, sensation-seeking, and impulsivity
indices.
Social self-control Sensation-seeking Impulsivity
Social self-
control

1.0 0.25 0.31
Sensation-
seeking

0.29 1.0 0.40
Impulsivity
0.43
0.47 1.0
Note. Zero-order correlations are provided above the diagonal; correlations adjusted for
unreliability of measurement are provided below the diagonal.

Regression analyses. Based on the findings of the multitrait scaling method
analyses, only social self-control and sensation-seeking scales were included in the
regression analyses. Since the impulsivity items failed to show discriminant and
convergent validity, impulsivity was considered a redundant variable and excluded from
the prediction models. The social self-control and sensation-seeking scales were created
using those items only that demonstrated sufficient convergent and discriminant validity
(Hays & Hayashi, 1990). Table 8 presents the results of the regression analyses.
Controlling for baseline substance use and demographic variables, social self-
control was found to be a significant predictor of cigarette, alcohol, and marijuana use
one year later (Model 1, see Table 8), but not hard drug use. After adding sensation-
seeking to the model, social self-control was a significant predictor of cigarette use but

49
was only a marginally significant predictor of alcohol and marijuana use (Model 3, see
Table 8). Thus, the results suggested that adolescents with one standard deviation higher
social self-control at baseline were 0.77 times (95% CI: 0.63, 0.95) less likely to use
cigarettes one year later, even after controlling for sensation-seeking, baseline substance
use, and potential demographic confounders. Sensation-seeking was the strongest
predictor of later alcohol and hard drug use, both before and after controlling for social
self-control (see Models 2 and 3, Table 8). However, after controlling for social self-
control, the effect of sensation-seeking on cigarette use one year later was only
marginally significant. Sensation-seeking was not found to predict marijuana use either
before or after controlling for social self-control.

50
Table 8. Social self-control and sensation-seeking as one-year predictors of
substance use, controlling for demographic variables.
Odds Ratio (95% Confidence Interval)

Cigarette
use
Alcohol
use
Marijuana
use
Hard drug
use

Model 1 Social self-
control

0.72 (0.59,
0.88)**
0.77 (0.63,
0.93)**
0.81 (0.67,
0.98)*
0.82 (0.67,
1.02) †
Model 2

Sensation-
seeking

1.28 (1.06,
1.54)*
1.36 (1.15,
1.62)**
1.09 (0.90,
1.31)
1.90 (1.41,
2.56)***
Model 3

Social self-
control

0.77 (0.63,
0.95)*
0.84 (0.69,
1.02) †
0.83 (0.69,
1.01) †
1.05 (0.83,
1.32)
Sensation-
seeking

1.21 (0.99,
1.48) †
1.31 (1.09,
1.58)**
1.04 (0.85,
1.28)
2.45 (1.71,
3.51)***
Note. ***< 0.001; **<0.01; *<0.05; †<0.1 (one-tailed). All models controlled for
baseline substance use, age, gender, parents’ education, ethnicity, and program condition.

Discussion
We examined the construct validity of the social self-control (Sussman, McCuller,
& Dent, 2003) which has previously been found to show significant cross-sectional and
longitudinal associations with adolescent substance use. We used the “multitrait scaling
analysis method” (Hays & Hayashi, 1990) to examine the item-level convergence and
divergence of the social self-control measure in comparison with the sensation-seeking
and impulsivity subscales of Zuckerman’s impulsive sensation-seeking scale (Zuckerman
et al., 1991). In addition, we examined whether baseline social self-control explained a
significant variance in future substance use after accounting for the variance explained by
sensation-seeking, baseline substance use and four demographic variables: age, gender,
parental education, and ethnicity.

51
The multitrait scaling analysis results revealed that the 10 social self-control items
showed adequate discrimination; however, only 8 of the 10 items showed sufficient
convergence. Moreover, the social self-control index only moderately correlated with the
impulsivity and the sensation-seeking indices, with and without correcting for the
unreliability of measurement. These findings have two important implications. First, the
social self-control measure appears to tap a unique construct that is not redundant with
the constructs of impulsivity or sensation-seeking. Second, the 8 social self-control items
found to be adequately convergent are more likely to better represent the social self-
control construct than the original 10 items. Hence, the two non-convergent items (i.e.,
“My feelings get hurt easily”; “I express all my feelings”) were dropped from the
subsequent regression analysis.
When the 8 social self-control items were combined to form a scale and used in
the regression analyses, we found baseline social self-control to be a statistically
significant predictor of adolescent cigarette, alcohol, and marijuana use 1 year later,
controlling for baseline substance use and demographic variables. When sensation
seeking was added to the model as a predictor, the predictive association between social
self-control and cigarette use remained statistically significant, whereas the associations
between social self-control and alcohol and marijuana use were found to be only
marginally significant. Conversely, after adjusting for social self-control the effect of
sensation-seeking on future cigarette use was found to be only marginally significant.
However, sensation-seeking was found to be a strong predictor of later alcohol and hard
drug use, before and after adjusting for social self-control. Note that the sensation-
seeking scale combined 8 of the original 11 items that the multitrait scaling analysis

52
found to be adequately convergent and divergent. The results of the regression analyses
suggest that social self-control is likely to be a unique predictor of adolescent substance
use, especially cigarette use, even after accounting for the effects due to sensation-
seeking.
Lack of social self-control is likely to represent poor emotional control whereas
higher sensation-seeking and impulsivity are likely to reflect poor behavioral self-control
(Bates & Labouvie, 1997; Wills et al., 2006; Verdejo-Garcia, Bechara, Recknor, & Perez-
Garcia, 2007). Compared to behavioral self-control, emotional self-control has been
studied less frequently in the context of adolescent health behavior research. In their high
school sample, Wills and colleagues (2006) found emotional self-control to affect
adolescents’ substance use behavior through proximal risk factors such as coping
motives, negative life-events, and peer substance use. In particular, good emotional self-
control was found to be protective against deviant peer affiliation whereas poor emotional
self-control was found to promote negative coping motives (Wills et al., 2006). Among a
sample of substance-dependent individuals, Verdejo-Garcia et al. (2007) found that those
who showed difficulty regulating negative emotions were less likely to be able to delay
substance use gratification.
In adolescent substance use research, emotional self-control has mostly been
assessed in terms of anger management, affective lability, and sadness management (e.g.,
Wills et al., 2006; Novak & Clayton, 2001; Simons & Carey, 2002; Brown et al., 2002).
One way to conceptualize social self-control is as one’s control over one’s emotions in
interpersonal communications. Although emotional self-control in interpersonal
communications has received some attention in communication research (e.g., Booth-

53
Butterfield & Booth-Butterfield, 1994; McCroskey & Richmond, 1993; McCroskey &
Richmond, 1995), this kind of construct has been relatively less explored in substance use
research. Traditionally, social skills training in adolescent substance use has focused
more on enhancing assertiveness than affect orientation or emotional regulation (Pentz,
1983; Botvin, 1983).
In communication research, affect orientation has been defined as the ability to
process emotional information in such as way as to guide one’s interpersonal interaction
(Booth-Butterfield & Booth-Butterfield, 1994). Individuals with higher affect orientation
are likely to be aware of their own emotional states and the feelings of others they are
interacting with (Booth-Butterfield & Sidelinger, 1997). Thus, individuals lacking in
social self-control are likely to have lower affect orientation. In addition, individuals
lacking social self-control may be prone to compulsive communication (McCroskey &
Richmond, 1995), verbal aggression (Infante & Wigley, 1986), and proactive
argumentativeness (Infante & Rancer, 1982); attributes which are likely to make them
less prosocial, which in turn may act as a risk factor for substance use. Clearly, more
research is needed to explore the relations between social self-control and
communications variables such as affect orientation and compulsive communication. In
addition, future research is needed to identify the possible mediators of the effects of
social self-control on adolescent substance use.
In the present study, the Model 1 results mostly support the findings of the
previous study (Chapter 2), although the present social self-control scale comprised 8 of
the original 10 items, all of which were used to compose the social self-control scale in
the previous study. As in Model 1, we found social self-control to be significantly

54
associated with higher marijuana and alcohol use 1 year later for the CHS subjects in the
previous study. In additional, as in the previous study, social self-control was not found
to be strongly associated with hard drug use in the present study. However, unlike the
previous study, social self-control was found to be associated with higher later cigarette
use in the present study. Given that in the present study the association between social
self-control and cigarette use remained strong even after controlling for sensation
seeking, the failure to find an association between these two variables in the CHS
subsample of the previous study is surprising.
In the previous study, the lack of association was explained in terms of the
possibility of cigarette use being considered a less deviant type of substance use among
CHS students. The findings of the present study make this conclusion questionable.
Given the fact that the CHS subsample in the previous study and the sample of the
present study were similar, in terms of sample size as well as demographics, it seems that
excluding the two non-convergent items enhances the predictive ability of the social self-
control construct. Clearly, future studies are needed to better understand the relation
between social self-control and cigarette use among CHS students.
Perhaps, our findings related to the differential association of sensation seeking
across cigarette, alcohol, marijuana, and hard drug use may be explained in terms of
personality-based differences in the motivations for substance use. Some previous
research has found higher sensation seekers to show higher enhancement motives for
alcohol use than for cigarette or marijuana use (Comeau, Stewart, & Loba, 2001). In fact,
Comeau et al. (2001) did not find sensation seeking to be associated with any of the four

55
motives of cigarette or marijuana use that they analyzed: coping, enhancement,
conformity, and social approval.
There are several limitations of this study. Although we tested all the relevant
variables for indications of a bias due to attrition, we didn’t test several others that were
not measured in the study. For example, indicators of SES other than parental education,
such as parental income and number of persons living in a house were not measured in
the study. In addition, our data may not generalize to all Los Angeles area CHS.
Although students were randomly selected at the classroom level, some selection
bias might have been introduced to the data at the level of school, which was based on
convenience sampling. Nevertheless, the findings clearly support the distinctiveness of
the social self-control scale and its prospective effects on drug use, even after adjusting
for previous use and other potential confounders.

56
Chapter 4
Structural models examining relations among social self-control, sense
of coherence, and substance use among youths from regular and
continuation high school students
Abstract
We used structural equation modeling (SEM) to test multiple hypotheses related
to social self-control, sensation-seeking, sense of coherence (SOC), and substance use in
a prospective sample of regular and continuation high school (RHS and CHS,
respectively) youths. In particular, we tested three structural equation models with
demographic variables (age, gender, and ethnicity) as exogenous variables. The first
model tested social self-control, sensation seeking, SOC, and substance use at baseline as
predictors of substance use 1 year later. The second model tested the latent social self-
control variable at 1 year follow-up as an outcome with the baseline social self-control,
sensation seeking, SOC, and substance use as predictors. The third model tested the
baseline social self-control, sensation seeking, SOC, and substance use as predictors of
the latent SOC variable 1 year later. For each model we conducted a multiple group
comparison to test whether the path regression coefficients differed significantly across
school type (i.e., RHS versus CHS). We found the latent variables of social self-control
and sensation seeking to be significant predictors of substance use 1 year later across both
school types, such that a higher social self-control was associated with lower substance
use and a higher sensation seeking predicted higher substance use. In addition, a higher
social self-control at baseline was found to predict higher SOC 1 year later, but only

57
among RHS students. We did not find baseline substance use or SOC to predict social
self-control 1 year later.
Introduction
The present study, which is based on the dataset discussed in Chapter 2, extends
the findings of the previous two chapters. The second chapter examined the bidirectional
relations between the social self-control scale composed of the original 10 items
(Sussman, Dent, & McCuller, 2003) and adolescent substance use. The third chapter
examined the convergence and divergence of the 10 social self-control items in
comparison with the items of Zuckerman’s impulsive sensation seeking (Zuckerman et
al., 1991). In addition, the third chapter examined the longitudinal relation between the
social self-control scale composed of 8 social self-control items (the 8 items that were
found to be sufficiently convergent and discriminant) and substance use 1-year later,
controlling for sensation-seeking and demographic variables. In the present chapter we
utilize structural equation modeling to test a number of hypotheses concerning the
prospective relations among social self-control, sensation seeking, SOC, and adolescent
substance use. Hence, the present study furthers the findings of the previous two studies
in three major ways. First, in addition to sensation seeking, the present study uses SOC as
a covariate of baseline social self-control in the prediction of future substance use.
Second, the study examines baseline social self-control, substance use, and sensation
seeking as possible predictors of SOC 1 year later. Third, the structural regression models
used in the present study incorporate a measurement model involving social self-control,
sensation seeking, and SOC.

58
Sense of Coherence (SOC) is central to the concept of salutogenesis, which
argues that healthfulness should be viewed as the maintenance of health and well-being
rather the absence of diseases (Antonovsky, 1987). SOC reflects an individual’s ability to
successfully cope with the various stressors of life (Antonovsky, 1987). Antonovsky has
defined SOC as a generalized orientation reflecting the extent to which a person is in
terms with his or her internal and external environments, and the extent to which he or
she believes that things in life can be reasonably predicted (Antonovsky, 1987).
According to Antonovsky (1987), a person’s SOC can be assessed in terms of
comprehensibility, manageability, and meaningfulness. Comprehensibility refers to one’s
confidence in one’s ability to understand life events, outside environment, and
interpersonal relationships (Antonovsky, 1987). Comprehensibility reflects the extent to
which a person believes that his or her environment is structured, predictable, and
explainable (Antonovsky, 1987). Manageability refers to one’s beliefs that one
possesses adequate resources to deal successfully with the internal and external stressors
of life (Antonovsky, 1987). Meaningfulness reflects one’s beliefs that life is meaningful
and worth preserving despite having to face recurring stressors (Antonovsky, 1987).
Research among adolescents as well as adults has shown that SOC is a consistent
predictor of health and well-being (Eriksson & Lindstrom, 2005). According to
Antonovsky (1987), SOC develops mainly during adolescence.
Although SOC has been examined extensively in relation to adult health issues
(see Erriksson & Lindstrom, 2005 for review), to our knowledge only a few cross-
sectional studies have examined the construct with respect to adolescent substance use.
These studies suggest that higher SOC is related to lower cigarette and alcohol use (e.g.,

59
Glanz, Maskarinec, & Carlin, 2005; Ayo-Yusuf, Reddy, & van den Borne, 2008; Nilsson,
Starrin, Simonsson, & Leppert, 2007). It is likely that SOC is also related with illicit drug
use among adolescents. For example, Lundqvist (1995), found an inverse association
between SOC and marijuana use in a sample of adults.
Adolescents with higher SOC may be more resourceful against the risk factors of
substance use (Glantz et al., 2005). Adolescents higher in SOC may be less likely to use
drugs for self-medication (Khantizian, 1985) or avoidant coping (Wills & Cleary, 1996).
In fact, individuals with a low SOC are likely to report higher levels of perceived stress
(e.g., Torsheim, Aaroe, & Wold, 2001; Bowman, 1996), depression (e.g., Koposov,
Ruchkin, & Eisemann, 2003), and psychosomatic complaints (e.g., Simonsson, Nilsson,
Leppert, & Diwan, 2008; Torsheim et al., 2001). In addition, adolescents with lower
SOC may have lesser resources to resist peer pressure, pro-substance use media
influences or perceived social norms (Glantz et al., 2005).
It is likely that social self-control may affect the development of SOC. A variable
such as social self-control is crucial for an adolescent’s proper social functioning (Elliot
& Gresham, 1993). Better social functioning in turn may result in higher SOC.
Individuals with better social functioning are likely to be more resourceful in terms of
using social support for coping purposes. In a cross-sectional sample of adolescents,
Margalit & Eysenck (1990) found a strong association between social skills and SOC,
controlling for age, gender, and scales of Neuroticism, Extraversion, and Psychoticism.
Furthermore, a systematic review of more than 450 health-related worldwide SOC studies
suggests that SOC is highly and positively correlated with social skills, among other
variables (Eriksson & Lindstrom, 2005). In the present study, we expect that higher

60
baseline social self-control will predict higher SOC one year later, controlling for
baseline SOC, sensation seeking, substance use, and demographic variables.
To our knowledge, the possible causal effects of substance use or sensation
seeking on SOC have not been examined previously. A previous cross-sectional study on
adolescents found sensation seeking and SOC to be inversely related (Glantz et al., 2005).
One may speculate that adolescents higher in sensation seeking and substance use at
baseline may show lower SOC over time, perhaps due to persistent exposure to negative
physical and/or social consequences (e.g., drug use dependence, interpersonal conflicts).
Conversely, we expect higher baseline SOC to predict lower substance use 1 year later,
controlling for baseline substance use, social self-control, sensation seeking, and
demographic variables.
Methods
Subjects and data collection
This study used the same dataset discussed in Chapter 2. Subjects, data collection
procedure, and attrition analyses have been described in detail in the methods section of
Chapter 2.
Measures
Ethnicity was assessed using the following ethnic indicator: “What is your ethnic
background? Please circle one category that best applies.” The response options included
a) Black or African American; b) Asian/Pacific Islander (subcategories: Chinese,
Japanese, Filipino, Korean, and open-ended “other”); c) Latino/Hispanic (subcategories:
Mexican-American, Mexican, Central American, South American, and an open-ended
“other”) ; d) White/Non Latino; e) Native American; and f) Other ethnic group (open-

61
ended). Since the present sample was predominantly Hispanic, for analysis purposes
ethnicity was binary coded as Hispanic (1) or non-Hispanic (0). Other demographic
indicators used in the present analysis included gender (coded as Female=1 and male=0)
and age.
Sense of Coherence was assessed in the study using a 13-item version of
Antonovsky’s Sense of Coherence (SOC-13) scale (1987). Although a shortened version
of the original 29-item scale, the 13 items are considered to tap the essential aspects of
comprehensibility (e.g., “Has it happened in the past that you were surprised by the
behavior of people whom you thought you knew well?”), manageability (e.g., “Do you
have the feeling that you’re being treated unfairly?”) and meaningfulness (e.g., “How
often do you have the feeling that there’s little meaning in the things you do in your daily
life?”; see Appendix A, Table 3 for complete list of items). According to Antonovsky
(1987), the three domains of SOC are supposed to reflect one SOC factor. Hence, in the
present study, all items were treated as potential indicators of a single SOC factor
[Cronbach’s alpha at baseline= 0.83 (RHS); 0.80 (CHS)].
Social self-control. Social self-control was assessed using the following 8 items:
“I enjoy arguing with people”; “If I am angry I act like it”; “Sometimes I provoke people
just for the fun of it”; “My mouth gets me in trouble a lot”; “If I think something
someone says is stupid I tell them so”; “I do things just to get attention”; “I hate being
wrong”; and “I say things that I regret later” [Cronbach’s alpha at baseline=0.75 (RHS);
0.74 (CHS)]. For each question, the response options included (1) Always to (4) Never.
Sensation seeking. The sensation-seeking measure included 6 of the 11 items from
the Impulsive Sensation Seeking subscale of the Zuckerman-Kuhlman Personality

62
Questionnaire (Zuckerman et al., 1991). This shorter version of the 11-item was based on
the 6 items that loaded highest on the sensation-seeking factor in our previous studies
(e.g., Sussman, Dent, & Galaif, 1997). Participants were asked to respond true or false to
statements that they might use to describe themselves: “I like to have new and exciting
experiences and sensations even if they are a little frightening”; “I like doing things just
for the thrill of it”; “I sometimes like to do things that are a little frightening”; “I
sometimes do ‘crazy’ things just for fun”; “I prefer friends who are excitingly
unpredictable”; and “I like ‘wild’ uninhibited parties” [Cronbach’s alpha at baseline=
0.72 (RHS); 0.74 (CHS)].
Substance use. The substance use measures included self-reported past-30-day
cigarette, alcohol, marijuana, and hard drug use behaviors that were assessed on 8-point
scales (0, 1-10, 11-20, …,91-100, over 100 times) (for information on reliability and
validity see Chapter 2). Hard drug use was an index composed of the following 7
categories of hard drugs: cocaine (crack), hallucinogens (LSD, acid, mushrooms),
stimulants (ice, speed, amphetamines), inhalants (rush, nitrous, glue), ecstasy (MDMA,
XTC, Adam), opiates (vicodin, oxycontin, morphine, heroin, opium), and other
(depressants, PCP, steroids, heroin, etc.) (Cronbach’s alpha=0.86). For the analyses
involved in the present study, a substance use index was created summing across the
standardized scores for cigarette, alcohol, marijuana, and hard drug use.
Statistical Analysis
Confirmatory factor analyses (CFA) and SEM were conducted on Mplus Version
5.1 (Muthen & Muthen, 1998-2008) using the weighted least squares means and variance
(WLSMV) adjusted estimator, which allows for the modeling of ordered categorical

63
variables. Model fit for the latent variable models was assessed using the mean- and
variance-adjusted chi-square statistic
1
(Mutehn, du Toit, & Spisic, 2007), root-mean-
square error of approximation index (RMSEA; Browne & Cudeck, 1993), and the
comparative fit index (CFI; Bentler, 1990). As Chi-square values tend to inflate due to
large sample sizes, the model fits were additionally evaluated using other fit indices such
as the root mean squared error of approximation (RMSEA; Steiger, 1990) and the
comparative fit index (CFI; Bentler, 1990). RMSEA values less than 0.05 indicate close
approximate fit whereas values between 0.05 and 0.08 indicate a reasonable fit (Hu &
Bentler, 1999). CFI values above 0.95 are considered to indicate a reasonably good fit
(Hu & Bentler, 1999).
The analysis employed a 3-step procedure. The first step involved CFA to test the
measurement model with the baseline and follow-up social self-control, sensation
seeking, and SOC constructs. A measurement model describes the nature of the
relationship between a number of latent variables and the observed indicators
corresponding to each latent variable. The latent constructs of baseline social self-control,
sensation seeking, SOC, and follow-up social self-control and SOC, were estimated
together and allowed to correlate. The second step tested three separate structural models

1
Both model chi-square and degrees of freedom for the model fit chi-square test are
mean- and variance-adjusted when the WLSMV estimator is used. Thus, the degrees of
freedom used for the significance test do not correspond directly with the numbers of
measured variables and estimated parameters. Hence, while comparing nested models
(e.g., in multiple group analysis) the estimated degrees of freedom for the constrained
model are the same or fewer than for the unconstrained model. Differences in model fit of
nested models were based on the derivatives difference test (Satorra, 2000; Satorra &
Bentler, 1999). The derivatives difference test does not correspond directly with the
differences in estimated chi-square and degrees of freedom between the constrained and
unconstrained models.

64
for follow-up substance use, social self-control, and SOC outcomes for RHS and CHS
separately. Each model specified age, Hispanic ethnicity, and female gender as
exogenous variables (i.e., not predicted by any prior construct in the model) and baseline
social self-control, sensation seeking, and SOC were specified as endogenous variables
(i.e., predicted by prior constructs in the model), with covariances of their residual terms.
In all three models, paths were specified from each demographic variable to each
endogenous variable and the outcome variable. In addition, paths were specified from
each endogenous variable to the outcome variable. Figures 1, 2, and 3 depict the three
models: Model 1, 2, and 3, respectively. Note, however, that the figures represent a
reduced form of the models [i.e., for simplicity of presentation, the models exclude
structural paths with statistically non-significant path coefficients at p<0.05 (two-tailed);
and indicator factor loadings and error terms]. The third step involved carrying out a two-
group comparison for Models 1, 2, & 3 to test the equivalence of relations among
constructs across RHS and CHS students.
Several steps were followed to conduct the multiple group comparison. First, each
model was separately estimated for RHS and CHS sub-samples. Next, each model was
estimated with the sub-samples combined, grouping specified (RHS versus CHS), and the
factor loadings (of indicators on latent constructs), the indicator thresholds, and the
structural path coefficients constrained to be equal across the two groups. This
represented the restrictive model. Next, a less restrictive combined model was estimated
with the structural path coefficients freed across the two groups. Next, a nested
comparison was carried out between the restrictive model and less restrictive model. This
procedure involved testing for significance the difference in Chi-square value between

65
the restrictive model and the less restrictive one using the difference in the degrees of
freedom (DF) between the two [i.e., Chi-square value (restrictive) – Chi-square value
(less restricted); DF (restrictive)- DF (less restrictive)]. If the Chi-square difference was
statistically significant at p<0.05, it was concluded that restricting the parameters across
the groups worsened the model. Conversely, if the Chi-square difference was statistically
non-significant at p>0.05, it was concluded that restricting the parameters did not worsen
the model significantly. When placing equality constraints on all structural path
coefficients was found to worsen the model, we attempted to find out whether certain
specific paths differed across the groups while others remained equal. This was done by
consulting the model modification indices. Subsequently, we released the equality
constraints from paths suggested by the model modification indices and carried out
another round of nested model comparison.
Results
Measurement Model
The measurement model examined in the study consisted of five latent variables:
baseline social self-control, sensation seeking, baseline SOC, follow-up social self-
control, and follow-up SOC (see Table 9). First, a preliminary CFA on all hypothesized
indicators corresponding to a latent construct was conducted separately for each
construct. Only the standardized factor loadings that were greater or equal to 0.50 were
considered meaningful and retained for the final measurement model (Wills et al., 2006).
As a result, the final measurement model had six indicators per construct (for the items
that were selected see Table 9). The measurement model showed a reasonable fit to the
data for RHS (Chi-square=782.5; DF=183; CFI=0.94; RMSEA=0.05) and CHS (Chi-

66
square=319.4; DF=129; CFI=0.94; RMSEA=0.05) separately and combined (with factor-
loadings constrained to equal across RHS and CHS; Chi-square=1031.2; DF=315;
CFI=0.94; RMSEA=0.05). The standardized parameter estimates (factor loadings)
relating indicators to their corresponding latent constructs in the final measurement
model (RHS and CHS combined sample) are presented in Table 10. All factor loadings
were statistically significant at p<0.0001.

67
Table 9. List of items used as indicators in the measurement model
Indicators Items
Baseline Social self-control/
Follow-up Social self-
control

assc1/cssc1 I enjoy arguing with people.
assc2/cssc2 If I am angry I act like it.
assc3/cssc3 My mouth gets me in trouble a lot.
assc4/cssc4 I do things just to get attention.
assc5/cssc5 Sometimes I provoke people just for the fun of it.
assc6/cssc6 I say things that I regret later.
Sensation seeking
ss1

I like to have new and exciting experiences and
sensations even if they are a little frightening
ss2 I like doing things just for the thrill of it
ss3 I sometimes like to do things that are a little
frightening
ss4 I sometimes do ‘crazy’ things just for fun

ss5 I prefer friends who are excitingly unpredictable
ss6 I like ‘wild’ uninhibited parties
Baseline Sense of
Coherence/ Follow-up
Sense of Coherence

asoc1/bsoc1

Has it happened that people whom you counted on
disappointed you?
asoc2/bsoc2 Do you have the feeling that you’re being treated
unfairly?
asoc3/bsoc3 Do you have the feeling that you are in a strange
situation and don’t know what to do?
asoc4/bsoc4 Do you have very mixed-up feelings and ideas?
asoc5/bsoc5

Does it happen that you have feelings inside you
would rather not feel?
asoc6/bsoc6

How often do you have the feeling that there’s little
meaning in the things you do in your daily life?

68
Table 10. Parameter estimates (factor loadings) for the hypothesized measurement
model.
Latent construct and
indicators
Standardized loading*
Baseline social self-control
assc1 0.58
assc2 0.50
assc3 0.73
assc4 0.59
assc5 0.70
assc6 0.61
Sensation seeking
ss1 0.60
ss2 0.77
ss3 0.76
ss4 0.82
ss5 0.61
ss6 0.66
Baseline sense of coherence
asoc1 0.62
asoc2 0.72
asoc3 0.75
asoc4 0.78
asoc5 0.79
asoc6 0.73
Follow-up social self-
control

cssc1 0.61
cssc2 0.51
cssc3 0.79
cssc4 0.54
cssc5 0.64
cssc6 0.57
Follow-up sense of
coherence

csoc1 0.59
csoc2 0.63
csoc3 0.68
csoc4 0.74
csoc5 0.75
csoc6 0.67
*Equal across RHS and CHS; Each loading differed significantly from zero at p<0.0001

69
Structural Model
Figure 1, 2, and 3 present the results of the SEM analysis. The figures show the
standardized estimates for all of the statistically significant regression paths and
covariance relating the constructs. The models presented in the figures represent a
combined model for RHS and CHS that resulted after multiple group comparison. Before
combining the groups, the models were run separately for RHS and CHS. For each
outcome, the separate group analysis showed reasonable model fit, thus enabling us to
combine the sub-samples and conduct a multiple group analysis (see Table 11). Model
specific results are as follows.
Table 11. Model fit indices for SEM on RHS and CHS separately.
Model outcome Chi-square
(DF)

CFI RMSEA
RHS (N=1514)

723.2 (121) 0.94 0.05 Substance use 1
year later
CHS (N=534)

271.9 (105) 0.93 0.05
RHS (N=1514)

1124.0 (181) 0.91 0.06 Social self-
control 1 year
later CHS (N=534)

409.2 (140) 0.91 0.06
RHS (N=1514)

864.7 (175) 0.94 0.05 SOC 1 year
later
CHS (N=534)

318.4 (137) 0.93 0.05
Note. DF= Degrees of freedom; CFI= Comparative fit index; RMSEA= Root mean
square error of approximation; RHS= Regular high school; CHS= Continuation high
school; SOC= Sense of coherence. Note that due to different outcomes and WLSMV
estimation DF’s are not same for same school type across models.

70
Model 1. Nested comparison of the restrictive model (i.e., all factor loadings
constrained to be equal across school type) with the less restrictive (i.e., all factor
loadings free across school type) model suggested that the Chi-square difference was
statistically significant. Hence, based on model modification indices, the coefficients for
the paths from Hispanic ethnicity to baseline substance use and from gender to follow-up
substance use were released from equality constraints step by step. The resulting model
was not significantly worse from the unrestricted model (∆Chi-square= 23.61; DF=16;
p>0.05). Figure 1 represents this model, which found the latent constructs of social self-
control and sensation seeking at baseline to significantly predict substance use 1 year
later across both school types, even after controlling for baseline substance use and
demographic variables. Hispanic ethnicity was found to be negatively associated with
baseline substance use for CHS students only. In addition, female gender was protective
against follow-up substance use for CHS students only. As expected, female gender was
negatively associated with sensation seeking. Interestingly, age and gender were found to
negatively associate with baseline SOC. SOC did not predict follow up substance use but
showed a significant negative covariance with baseline substance use.
Model 2. Nested comparison between restrictive and less restricted models
suggested that equality constraints on all structural path coefficients significantly
worsened the nested model (p<0.05). Hence, based on the model modification indices,
three regression coefficients were systematically released from equality constraints across
groups. These three coefficients corresponded to the paths from Hispanic ethnicity to
baseline social self-control and baseline substance use, and from age to baseline
substance use. The Chi-square difference test suggested that the resulting model

71
Figure 1. Structural model representing follow-up substance use as outcome (RHS versus CHS)

Note: Rectangles indicate observed variables and ovals indicate latent constructs. Included in the model but excluded from the figure for graphical
simplicity are the indicators of latent constructs and error notations. Straight single-headed arrows indicate standardized path coefficients (only
those paths that were statistically significant are shown in the figure. *denotes that the path was significant only for CHS students.

72
Figure 2. Structural model representing follow-up social self-control as an outcome (RHS versus CHS)

Note: Rectangles indicate observed variables and ovals indicate latent constructs. Included in the model but excluded from the figure for graphical
simplicity are the indicators of latent constructs and error notations. Straight single-headed arrows indicate standardized path coefficients (only
those paths that were statistically significant are shown in the figure. *denotes that the path was significant only for CHS students. +denotes that
the path was significant only for RHS students.

73
Figure 3. Structural model representing follow-up sense of coherence as an outcome (RHS versus CHS)

Note: Rectangles indicate observed variables and ovals indicate latent constructs. Included in the model but excluded from the figure for graphical
simplicity are the indicators of latent constructs and error notations. Straight single-headed arrows indicate standardized path coefficients (only
those paths that were statistically significant are shown in the figure. *denotes that the path was significant only for CHS students. +denotes that
the path was significant only for RHS students.

74
(presented in Figure 2) was not significantly worse from the unrestricted model (∆Chi-
square= 23.7; DF=16; p>0.05). Model 2 results suggested that the latent baseline social
self-control was the strongest predictor of the future social self-control. In addition,
higher age at baseline was found to predict higher social self-control 1 year later.
Sensation seeking and SOC at baseline had no effects on social self-control 1 year later.
Hispanic ethnicity was associated with higher social self-control at baseline but only for
CHS students.
Model 3. Nested comparison between restrictive and less restrictive models
suggested that equality constraints on all structural path coefficients across RHS and CHS
significantly worsened the nested model (p<0.05). Hence, based on the model
modification indices, five regression coefficients were systematically released from
equality constraints. These five coefficients corresponded to the paths from baseline
social self-control to follow-up SOC, from Hispanic ethnicity to follow-up SOC and
baseline substance use, from gender to follow-up SOC, and from age to baseline
substance use. Chi-square difference test suggested that the partially freed model
(presented in Figure 3) was not significantly worse from the unrestricted model (∆Chi-
square= 23.23; DF=15; p>0.05). Model 3 results suggested that among RHS students
higher baseline social self-control was a significant predictor of higher later SOC. Also,
among RHS students, Hispanic ethnicity and female gender showed significant negative
relationship with SOC one year later. As expected, baseline SOC was the strongest
predictor of future SOC across both groups. Interestingly, adjusting for baseline SOC,
age showed positive significant association with follow-up SOC. We did not find
baseline substance use to have any significant effect on SOC 1 year later.

75
Discussion
The present study was based on the data collected from 1514 regular high school
youths and 534 continuation high school youths at two time-points approximately 1 year
apart. The longitudinal design of the study and the use of SEM allowed us to test several
hypotheses concerning the development of substance use behavior, social self-control,
and SOC over time. In particular, we examined three SEM models across subsample of
adolescents from two types of high schools: regular and continuation. The first model
tested the longitudinal relations between baseline social self-control, sensation seeking,
and SOC on substance use behavior 1 year later while accounting for the effects of
subjects’ demographic characteristics (age, ethnicity, and gender) and their baseline
substance use behavior on the later substance use. The second model tested the
longitudinal relations between baseline substance use, sensation seeking, SOC and social
self-control 1 year later, adjusting for the subjects’ social self-control at baseline and the
demographic variables. The third model tested the longitudinal relations between baseline
substance use, social self-control, sensation seeking, and SOC 1 year later, adjusting for
the subjects’ SOC at baseline and the demographic variables.
Thus, the present study attempted to reexamine the questions examined in the
previous two chapters through a more sophisticated and thorough approach. In addition to
examining the bidirectional relationship between social self-control and substance use (as
in Chapter 2) while adjusting for sensation seeking (as in Chapter 3), these models
examined the latent social self-control factor and adjusted for SOC, an additional
dispositional variable. To our knowledge, the present study is one of the first few studies

76
to examine the effects of demographic, substance use, and self-control-related variables
on future SOC among US adolescents.
Of the 8 items that were retained based on the findings from Chapter 3, two items
(“I hate being wrong”; “If I think something someone says is stupid I tell them so”) were
excluded from the measurement model used in the present study because these items
showed lower standardized factor loading values than the cut-off of 0.50. The findings
based on the latent social self-control factor as indicated by the remaining six items
supported the findings of the previous studies where the social self-control indices
composed of 10 (Chapter 2) and 8 (Chapter 3) were used. The latent social self-control
factor was found to be a significant predictor of later substance use, even after controlling
for baseline substance use and sensation seeking. In addition, the relatively high
standardized path coefficient relating the latent social self-control variable at baseline to
the social self-control construct at 1 year follow-up (i.e., 0.73) suggests that social self-
control is likely to be a reasonably stable construct. That, however, does not mean that
social self-control does not change over time at all. In fact, the Model 2 results also
suggest that social self-control may increase with age.
Model 2 results somewhat conflicts with what we found in Chapter 2. In Chapter
2 we found cigarette, marijuana, and hard drug use at baseline related with lower social
self-control 1 year later, controlling for baseline social self-control. However, in the
present analysis we did not find a significant relationship between the baseline substance
use index and follow-up social self-control factor. One of the reasons for the discrepancy
might be the use of a general substance use index in the present study compared with the
individual substance use scales in the previous study. Another reason might be the

77
stronger association between the baseline and follow-up social self-control factors in the
present study: the standardized parameter estimate relating baseline and follow-up social
self-control scales in Study 2 was 0.50 compared with the 0.73 in the present study.
The results of the present study did not support our hypotheses regarding the
possible protective effects of higher SOC on later substance use or the possible negative
effects of higher baseline substance use on later SOC. However, the fact that we did not
find a significant association between baseline SOC and follow-up substance use does not
necessarily mean that our results did not support the findings of some of the cross-
sectional studies on SOC and adolescent substance use (e.g., Glantz et al., 2005; Nilsson
et al., 2007). It should be noted that our model testing the relation between SOC and
substance use also included social self-control as a predictor, which shared a relatively
high co-variance with SOC and appears to be a stronger predictor of substance use. Based
on the strength of association between baseline and follow-up SOC factors, it appears that
SOC is a less stable construct than social self-control.
Our hypothesis regarding the positive effects of higher social self-control on the
development of future SOC was supported but only for RHS students. Individuals with
higher social self-control may develop higher SOC overtime because such individuals are
more likely to form prosocial bonds and engage in prosocial activities and have a positive
outlook towards life in general. In addition, someone with a higher social self-control is
more likely to be resourceful in terms of social support. Hence, such individual is likely
to have higher perceived ability to cope with stress. However, it appears that the positive
effects of social self-control on SOC are valid only for youths in a low-risk context,
perhaps because in a high-risk context social self-control is not as important in acquiring

78
a self-affirming, optimistic worldview or forming a resourceful social network. Our
findings regarding the negative effects of female gender on SOC could be important.
Given that SOC is strongly associated with depressive and psychosomatic symptoms
(e.g., Koposov et al., 2003; Simonsson et al., 2008), girls may benefit from interventions
with SOC or resiliency-related components.
Across the three models, we found Hispanic ethnicity to be associated with
baseline substance use and social self-control among CHS students, and follow-up SOC
among RHS students. Hispanic ethnicity was found to be negatively associated with
baseline substance use in all three models among CHS students (but not for RHS
students). It appears that compared to non-Hispanic adolescents, Hispanic adolescents are
less likely to be at a CHS because of their substance use behaviors. Currently, a thorough
study of the various reasons why adolescents might join alternative high schools is
lacking. Moreover, it is not clear if such reasons vary according to adolescents’ ethnicity.
However, it is possible that Hispanic adolescents, more so than, for example, White or
Asian adolescents, are likely to join alternative high schools because of the difficulties
they might face in keeping up with the schedule and curriculum of a regular high school.
For Hispanic adolescents, especially those who are recent immigrants or first generation
immigrants, the stress associated with acculturating to the mainstream US culture,
including the education system, may result in negative academic and/or social
consequences, which may affect their success at school (Vega, Khoury, Zimmerman, Gil,
& Warheit, 1995). In fact, our finding that the Hispanic adolescents at regular high
schools are more likely to report lower SOC 1 year later supports this proposition. In
addition, since Hispanics in the US represent one of the lower socio-economic groups,

79
Hispanic adolescents are more likely to work (Fry, 2002), either to supplement their
family’s income or to support themselves. Thus, these youth are more likely to join CHS
because of the problem that work might pose against their maintaining a regular school
attendance
In the second model, we found Hispanic ethnicity to be associated positively with
baseline social self-control among CHS students (but not among RHS students). As Wills
& Dishion (2004) have pointed out, the development of self-control characteristics to
some extent depend on the quality of interactions that individuals have as children with
their parents. This could be especially true in the case of social self-control development,
as children are likely to learn self-control in social communication based on feedback
from adults (Patterson, DeBaryshey, & Ramsey, 1989). Given the importance placed on
family and collective living in Hispanic culture (Triandis, 1989), Hispanic adolescents
may show a higher level of social self-control because of the greater likelihood of their
having had spent more time with adults while growing up. It is possible that in the
present sample the CHS Hispanic adolescents were less acculturated to the mainstream
US culture compared with the RHS Hispanic adolescents: a possibility which might
explain why the positive association between Hispanic ethnicity and social self-control
was significant for CHS youths only.
The present study was limited in a number of ways. First, in order to reduce the
amount of missing data, we did not include parental education as an exogenous variable
in any of the models. Socioeconomic status could be an important confounder, especially
when drug use is an outcome. Second, only 6 out of 13 SOC items were included in the
measurement model as indicators of SOC. The selected items represented the three

80
dimensions of SOC (i.e., comprehensibility, meaningfulness, and manageability)
disproportionately. Three items represented comprehensibility; two items represented
manageability, whereas only one item represented meaningfulness. Third, since our
sample was predominantly Hispanic, we were not able to examine the effects of other
ethnic identification on the endogenous and outcome variables.

81
Chapter 5
Conclusion
This dissertation project sought to clarify the construct of social self-control and
to examine the relations of the construct with adolescent substance use and SOC while
adjusting for potential confounders such as sensation seeking and demographic variables.
Primarily, the social self-control construct was examined using multitrait scaling (Hays &
Hayashi, 1990) and confirmatory factor analysis (CFA). The results of the multitrait
scaling analysis suggested that although all 10 of the original social self-control items
diverged from the items of established impulsivity and sensation seeking scales
(Zuckerman et al., 1991), 2 of the 10 items did not converge adequately. When on a
separate sample, a measurement model of social self-control was tested using CFA with
the 8 items selected based on the mutlitrait scaling result, only 6 of the items showed
standardized factor loadings of greater than 0.50. This measurement model of social self-
control was found to be moderately stable over time and reasonably generalizable across
ethnicities (see Appendix B).
Thus, for the most part, the 6 items of social self-control seem to assess the same
construct over time and across adolescents from different ethnic backgrounds. In
addition, Chapter 4 results suggested that among adolescents across both high- and low-
risk contexts, the latent social self-control construct at baseline was very strongly
associated with social self-control 1 year later. In summary, our results across the three
studies suggested that 1) social self-control is a unique predictor of future substance use
among youths from high- and low-risk samples of youth; and that 2) social self-control is

82
likely to affect future SOC among youths from regular high schools, which suggests that
social self-control might aid in healthy youth development.
Limitations and future directions
Social self-control and substance use. In the first study we found that a higher
social self-control at baseline predicted lower alcohol and marijuana use across both RHS
and CHS students and lower cigarette use among RHS students. The findings of the first
study also suggested that higher levels of cigarette, marijuana, and hard drug use at
baseline might predict lower social self-control 1 year later. In the second study, which
was based entirely on a CHS sample, we found that social self-control was a statistically
significant predictor of future cigarette use, but not of alcohol, marijuana, or hard drug
use, after controlling for sensation seeking. In the third study, which was based on the
same dataset as the first, we found that a higher latent social self-control variable at
baseline predicted lower generalized substance use across both RHS and CHS samples,
controlling for sensation seeking and SOC; but the generalized substance use at baseline
did not predict social self-control 1 year later.
Hence, although the findings of the current set of studies strongly suggest that a
longitudinal relationship between social self-control and substance use exists, much more
research is needed to understand the various aspects of this relationship. In particular,
future research is needed to further explore 1) how the relationship varies across different
substance use types among youths from different risk contexts; and 2) whether substance
use affects social self-control over time. The present three studies each used a different
set of social self-control items, which might to some extent explain the inconsistencies
across their findings. Furthermore, each study used a different analytical approach or

83
examined a different set of covariates. Hence, future research needs to examine the
models studied in each of the present studies in new samples of CHS and RHS students
and compare the results.
Social self-control and culture. Developmentally, social self-control is likely to
have strong cultural underpinnings. One’s culture of upbringing is likely to affect one’s
social attitudes and behavior (Triandis, 1989). For example, each culture tends to have its
own peculiar codes that distinguish between good and bad manners. Since social self-
control is an aspect of social behavior, culture is likely to have an important impact on
the development of social self-control. In the present set of studies, we examined
Hispanic/non-Hispanic ethnicity as a predictor of social self-control. We coded the
ethnicity variable as Hispanic/non-Hispanic because our samples were predominantly
(i.e., more than 60%) Hispanic. Due to a relatively small number of subjects in each of
the Asian, African American, and Other ethnic groups across school types, we opted to
code the ethnicity variable as Hispanic/non-Hispanic in the regression analyses.
However, we carried out an analysis to see if the social self-control measure assessed the
same construct across ethnicities, as youths from different cultures might interpret the
same set of items differently (see Appendix B). Future longitudinal studies need to
examine how ethnicity might affect the development of social self-control.
Culture may impact social self-control in two ways at the least. First, individuals
from cultures that encourage stricter codes of interpersonal conduct may show higher
levels of social self-control. For example, in certain cultures (e.g., Hindus; Pokhrel,
Regmi, & Piedade, 2008) individuals are not encouraged to speak their mind before
people who are older to them in age. In cultures where age and caste hierarchies are

84
strongly prevalent (e.g., Hindus; Pokhrel, Regmi, & Piedade, 2008), a variable such as
social self-control may depend greatly on who the individual is interacting with: someone
who behaves uninhibitedly among peers might show a higher level of social self-control
among older family members. Additionally, in certain cultures women might be expected
to show higher personal restraints in social interactions than males (e.g., Hindus; Pokhrel,
Regmi, Piedade, 2008).
Second, one might speculate children or adolescents from collectivistic cultures
who receive authoritarian parenting and grow up in joint families or larger families with
more adults may show a higher level of social self-control. Conversely, children or
adolescents from individualistic cultures who grow up in nuclear families with a
relatively limited intercourse with adults in their day-to-day lives may show a lower level
of social self-control. In the former type of culture, provided that adults encourage social
self-control, children and adolescents have a greater opportunity to learn social self-
control through modeling, instruction, and coaching. Moreover, in collectivist societies or
communities where individuals are more interdependent in terms of resource sharing,
higher social self-control might help one to acquire resources more successfully. Perhaps,
having a higher social self-control has even greater practical benefits among members of
low-SES immigrant communities who have to rely heavily on social support to manage
daily life. Hence, it is important for future research to examine culture as a determinant of
social self-control.
Social self-control and neurocognition. The present set of studies did not use any
overt measures of social self-control. One potential avenue for future research would be
to explore the nature of the correlation between social self-control and the executive

85
cognitive functions (i.e., attention, memory, working memory, inhibitory control,
planning, decision-making, and monitoring). Executive functions make the self-
regulation of thoughts, emotion, and behavior possible. Conversely, deficiencies in
executive function may result in poor impulse control, poor judgment, and disinhibited
behavior (Barkley, 1997). Lack social self-control might have roots in the deficiencies in
executive cognitive functions. Among adolescents, poor executive functioning has been
consistently associated with higher rates of drug use (e.g., Grekin & Sher, 2006; Tarter et
al., 2003; Mezzich et al., 1997).
Understanding social self-control in relation to executive cognitive functions
among adolescents is of special importance. Research suggests that executive function
develops in sophistication at the same rate as the structural maturation of the prefrontal
cortex; and age-related social and cognitive maturation during adolescence may be
attributed to the concomitant structural changes in the brain (Fuster, 2008; Steinberg,
2005). For example, improvements in planning and decision-making have been linked
with the structural developments in the dorsolateral and ventrolateral prefrontal cortex,
respectively (Steinberg, 2005). Most notable developmental changes in the forebrain
region have been observed as changes in grey and white matter volumes. Recent
neuroimaging studies suggest that there occurs a continuous increase in the brain white
matter volume during adolescence (Giedd et al., 1999; Paus, 2005). For example, a
significant growth is noticed in the posterior corpus callosum, the collection of over 200
million nerve-fibers that allow communication between right and left hemispheres of the
brain (Giedd et al., 1999; Paus, 2005). In addition, the grey matter volume, which

86
increases substantially during childhood, appears to decrease during adolescence in
certain cortical structures [e.g., the prefrontal cortex; (Giedd, 2004)].
Reduction in cortical gray matter volume might occur due to increased intra-
cortical myelination and/or due to synaptic pruning (Giedd et al., 1999). Increased
myelination of neurons results in a more efficient propagation of action potentials.
Synaptic pruning involves selective removal of synapses that “do not efficiently transmit
information pertaining to accumulating experience” (Chambers, 2003). Synaptic pruning
appears to serve a number of functions that facilitate cognitive development. For
example, the process appears to stabilize the firing patterns of cortical neurons, which in
turn is thought to enhance working memory performance (Chambers, 2003). In general,
both myelination and synaptic pruning are believed to enhance the efficiency of cortical
information processing as well as the connectivity between cortical and subcortical
regions (Steinberg, 2005; Chambers, 2003; Giedd et al., 1999). Thus, since adolescent
prefrontal cortex is not yet fully developed, the associated executive functions are
expected to be inadequately developed. As a result, adolescents tend to have lower
regulatory competence, which makes them highly susceptible to drug use risk factors
such as rash impulsiveness and poor decision-making, and perhaps lower social self-
control (Tarter, 2002).
Attrition. Sample attrition may be viewed as one of the chief limitations of the
present studies. Both of the present datasets were longitudinal in design and, therefore,
subject to sample attrition. A high attrition rate may result in a selective sample over time
and introduce a bias into the research (Campbell & Stanley, 1966). Research on
adolescent substance use shows that attrition is likely to be selective on “high risk”

87
characteristics such as lower socioeconomic status, ethnic minority status, and higher
levels of problem behavior (Snow, Tebes, & Arthur, 1992). In other words, adolescents at
a higher risk for substance use are usually the ones who are more prone to be lost to
follow-up compared with the low-risk adolescents.
In both datasets, to examine the possibility of attrition bias, we conducted attrition
analyses on all observed variables of interest. We did not find any significant difference
between the full baseline sample and sample retained at follow-up on the baseline
measures (see the “Attrition Analyses” sections in Chapter 2 and 3). However, it should
be noted that in one of our datasets (see Chapter 2 and 4) the subsamples (i.e., RHS vs.
CHS) differed significantly on 1-year retention rates. CHS students were more likely to
lose subjects 1-year later compared with the RHS subjects (i.e., 35% vs. 20%). Although
based on the attrition analyses on the observed variables, attrition in each subsample
seem to be random, one limitation of our current analytical approach could be the failure
to address the problem of differential attrition across the subsamples. If based on the
unobserved variables, the attrition in one of the subsamples were to be biased, then that
bias might affect the assumed characteristics of the subsamples (e.g., high-risk vs. low-
risk). One way to address this problem in future research would be to calculate and
control an attrition propensity score in the regression models across the subsamples (Sun,
Sussman, Dent, & Rohrbach, 2008). An attrition propensity score is calculated for all
subjects, including the ones retained at follow-up, based on the observed baseline
variables that predict follow-up attrition (Sun et al., 2008).

88
Implications for Prevention
It appears that training social self-control skills may be useful in order to prevent
adolescent substance use, including the misuse of various types of prescribed drugs for
behavioral disorders (Sussman, Pentz, Sruitz-Metz, & Miller, 2006). What is clear from
our findings is that youth who learn not to alienate others frequently in social interactions
may be less likely to revert to substance use. The body of evidence linking personality
and pathological forms of behavioral undercontrol in adolescents to substance use has
been growing (e.g., Elkins, King, McGue, & Iacono, 2006; King, Iacono, & McGue,
2004; McGue, Iacono, Legrand, & Malone, & Elkins, 2001).
Although a measure of behavioral undercontrol, social self-control is not
necessarily a trait such as novelty seeking nor a form of psychopathology. Sussman et al.
(2003) found that social self-control was non-redundant compared with the measure of
antisocial personality disorder. However, despite some fundamental differences, the
measure of social self-control does share some similarities with DSM-IV diagnostic
criteria for Oppositional Defiant Disorder (ODD) (DSM-IV-TR, 2000). ODD primarily
concerns a youth’s or a child’s behavioral disinhibition as reflected through his or her
interaction with adults, whereas social self-control concerns his or her interaction with
both peers and adults. More importantly, compared to the social self-control measure the
ODD measure focuses more on the defiant and vindictive aspects of personality (e.g.,
“Often blames others for his or her mistakes or misbehavior”, “Is often spiteful or
vindictive”). Currently, it is not known whether lack of social self-control may over time
develop into a psychopathogical condition in certain types of personality. However, given
the fact that social self-control shares similarities with ODD and that it is associated with

89
future substance use, it seems worthwhile to address the problem of behavioral
disinhibition at the level of social self-control, especially among younger teens.
Although we found social self-control to be fairly stable over a 1-year period,
social self-control skills can be acquired through training, as has been long suggested by
successful cognitive-behavioral skills training programs (e.g., Botvin & Wills, 1985). The
social self-control skills training material could be designed with the view of promoting
prosocial adolescent-peer and adolescent-adult interactions. That is, providing
adolescents with skills to act in social situations so as not to alienate others. The material
could utilize elements of social problem solving and impulse control (e.g., Spivak &
Shure, 1982; Meichenbaum, 1985; Weissberg, 1985), empathy training (e.g., Frey,
Hirschstein, & Guzzo, 2000), and anger management (e.g., Novaco, 1975).
Social problem solving consists of a series of steps such as identifying the
problem, determining alternative ways of reacting to the problem, predicting
consequences of each reaction (e.g., “How might people feel?”), and selecting the most
prosocial reaction (Elliott & Gresham, 1993). A lesson on social problem solving can
have students go through these steps as a response to a hypothetical situation. Student
may also be taught how these steps can be incorporated into a “self-talk” to control their
impulses (Meichenbaum, 1985). Empathy training involves recognizing one’s and others’
feelings, giving consideration to others’ perspectives, and responding sensitively to others
(Frey, Hirschstein, & Guzzo, 2000). Lessons on empathy training may involve exercises
on interpreting emotional expressions and discriminating between accidental and
intentional actions (Frey, Hirschstein, & Guzzo, 2000). Anger-management techniques

90
may include recognizing personal triggers that arouse anger and practicing strategies
(e.g., self-talk) to inhibit or subdue impulsive responses (Novaco, 1975).
An example of a teen drug abuse prevention program that utilizes social self-
control is Project Toward No Drug Abuse (Project-TND). Project TND has proven to
reduce drug use significantly among teens of ages 14-19 years and diverse ethnicity (see
SAMHSA, 2009). Project TND uses the theoretical elements of motivation, skills, and
decision-making to prevent adolescents from drug abuse (Sussman, 1996; Sussman et al.,
2004). The “skills” aspect of TND addresses coping, communication, and social self-
control (Sussman, 1996). TND consists of a 12 session curriculum which is implemented
in the classroom by a trained teacher or health educator over a 4 week period. Each
session is provided in lieu of a standard health education class and lasts the length of a
normal class period (40 to 50 minutes). At present, one of the 12 TND sessions deals
specifically with self-control skills, one session deals with coping skills, 3 sessions
concern decision-making, and the rest deal with providing drug use relevant knowledge
and correcting drug use specific cognitive misperceptions.
Two of the sixteen adolescent drug abuse prevention principles enlisted by
National Institute of Drug Abuse (NIDA) advocate that programs for elementary school
children should focus on imparting self-control skills and interpersonal skills (NIDA,
2009). The present study suggested that self-control and social/interpersonal skills are not
entirely independent entities, but related in the form of social self-control, which needs to
be incorporated in prevention programs designed for high school students. Although
future studies would help in order to understand social self-control better, present policies

91
are likely to benefit from recognizing the protective effects of social self-control on drug
use.

92
References

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). Heidelberg: Springer.

Akers, R. L., Krohn, M. D., Lanza, -. K., L.,, & Radosevich, M. (1979). Social learning
and deviant behavior: A specific test of a general theory. American Sociological
Review, 44, 636-655.

Allison, K., Leone, P. E., & Spero, E. R. (1990). Drug and alcohol use among
adolescents: Social context and competence. In P. E. Leone (Ed.), Understanding
troubled and troubling youth. Newbury Park: Sage Publications.

American Psychiatric Association: Diagnositc and statistical manual of mental disorders
(DSM IV-TR). Washington, DC: American Psychiatric Association; 2000

Ames, S. L., Zogg, J. B., & Stacy, A. W. (2002). Implicit cognition, sensation seeking,
marijuana use and driving behavior among drug offenders. Personality and
Individual Differences, 33, 1055-1072.

Antonovsky, A. (1987). Unraveling the mystery of health: How people manage stress
and stay well. San Francisco: Jossey-Bass.

Arneklev, B. J., Grasmick, H. G., Tittle, C. R., & Bursik, R. J. (1993). Low self-control
and imprudent behavior. Journal of Quantitative Criminology, 9, 225-247.

Ayo-Yusuf, O. A., Reddy, P. S., & van den Borne, B. W. (2008). Adolescents' sense of
coherence and smoking as longitudinal predictors of self-reported gingivitis
Journal of Clinical Periodontology 35, 931-937.

Baler, R. D., & Volkow, N. D. (2006). Drug addiction: the neurobiology of disrupted
self-control. Trends in Molecular Medicine, 12(12), 559-566.

Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice Hall.

Bandura, A. (1991). Social cognitive theory of self-regulation. Organizational behavior
and human decision processes, 50, 248-287.

Bardo, M. T., Donohew, R. L., & Harrington, N. G. (1996). Psychobiology of novelty
seeking and drug seeking behavior. Behavior and Brain Research 77(1-2), 23-43.

93

Barkley, R. A. (1997). Behavioral inhibition, sustained attention, and executive functions:
constructing a unifying theory of ADHD. Psychol Bull, 121(1), 65-94.

Bates, M. E., & Labouvie, E. W. (1997). Adolescent risk factors and the prediction of
persistent alcohol and drug use into adulthood. Alcoholism, Clinical and
Experimental Research, 21(5), 944-950.

Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological
Bulletin, 107, 238-246.

Booth-Butterfield, M., & Booth-Butterfield, S. (1994). The affective orientation to
communication: Conceptual and empirical distinctions. Communication
Quarterly, 42(4), 331-344.

Booth-Butterfield, M., & Sidelinger, R. J. (1997). The relationship between parental traits
and open family communication: Affective orientation and verbal aggression.
Communication Research Reports, 14(4), 408-417.

Bowman, B. (1996). Cross-cultural validation of Antonovsky's sense of coherence scale.
Journal of Clinical Psychology, 52(5), 447–549.

Brook, J. S., Whitman, M., & Gordon, A. S. (1981). Maternal and personality
determinants of adolescent smoking behavior. Journal of Genetic Psychology,
139, 185-193.

Brown, R. A., Lejuez, C. W., Kahler, C. W., & Strong, D. R. (2002). Distress tolerance
and durantion of smoking cessation attempts. Journal of Abnormal Psychology,
111, 180-185.

Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A.
Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136-162).
Newbury Park, CA: Sage.

Burt, R. D., Dinh, K. T., Peterson, J., A.V. , & Sarason, I. G. (2000). Predicting
adolescent smoking: A prospective study of personality variables. Preventive
Medicine, 30, 115-125.

California Continuation Education Association. Retrieved on October 2, 2009, from
www.cceanet.org/brochure.

Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the
multitrait-multimethod matrix. Psychological Bulletin, 56, 81-105.

94
Carey, M. P., Carey, K. B., Carnrike, C. L., Jr., & Meisler, A. W. (1990). Learned
resourcefulness, drinking, and smoking in young adults. Journal of Psychology,
124(4), 391-395.

Chambers, R. A., Taylor, J. R., & Ponteza, M. N. (2003). Developmental neurocircuitry
of motivation in adolescence: A critical period of addiction vulnerbality. The
American Journal of Psychiatry, 160, 1041-1052.

Cloninger, C. R. (1994). The Temperament and Character Inventory (TCI): a guide to its
development and use. Washington University, St Louis, Missouri: Centre for
Psychobiology of Personality.

Cloninger, C. R., Svrakic, D. M., & Przybeck, T. R. (1993). A psychobiological model of
temperament and character. Archives of General Psychiatry, 50, 975–990.

Colder, C. R., & Chassin, L. (1997). Affectivity and impulsivity: Temperament risk for
adolescent alcohol involvement. Psychology of Addictive Behaviors, 11(2), 83-97.

Comeau, N., Stewart, S.H., & Loba, P. (2001). The relations of trait anxiety, anxiety
sensitivity, & sensation seeking to adolescents’ motivations for alcohol, cigarette,
and marijuana use. Addictive Behaviors, 26 (6): 803-825.

Coie JD, Dodge KA, Coppotelli H. Dimensions and types of social status: A cross-age
perspective. Developmental Psychology, 1982; 18(4):557-570.

Crawford, A. M., Pentz, M. A., Chou, C.-P., Li, C., & Dwyer, J. H. (2003). Parallel
developmental trajectories of sensation seeking and regular substance use in
adolescents. Psychology of Addictive Behaviors, 17(3), 179-192.

Crick, N.R., & Grotpeter, J.K.(1996). Children's treatment by peers: Victims of relational
and overt aggression. Development and Psychopathology, 8, 367-380.

Deas, D. (2006). Adolescent substance abuse and psychiatric comorbidities. Journal of
Clinical Psychiatry, 67(Suppl. 7), 18-23.

Dent, C. W., Sussman, S., & Stacy, A. W. (1997). The impact of a written parental
consent policy on estimates from a school-based drug use survey. Evaluation
Review, 21, 698-712.

Donohew, R. L., Hoyle, R. H., Clayton, R. R., Skinner, W. F., Colon, S. E., & Rice, R. E.
(1999). Sensation Seeking and Drug Use by Adolescents and Their Friends:
Models for Marijuana and Alcohol. Journal of Studies on Alcohol, 60.

95
Earleywine, M., & Finn, P. R. (1991). Sensation seeking explains the relation between
behavior disinhibition and alcohol consumption. Addictive Behaviors, 16(3-4),
348-361.

Eaton, D. K., Kahn, L., Kinchen, S., Ross, J., Hawkins, M. A., Harris, W. A., et al.
(2006). Youth Risk Behavior Surveillance--United States, 2005. Mortality and
Morbidity Weekly Report, 55(SS-5).

Elkins IJ, King SM, McGue M, Iacono WG: Personality traits and the development of
nicotine, alcohol, and illicit drug disorders: prospective links from adolescence to
young adulthood. J Abnorm Psychol. 2006 115: 26-39.

Elliott, S. N., & Gresham, F. M. (1993). Social skills interventions for children. Behavior
Modification, 17, 287-311.

Eriksson, M., & Lindstrom, B. (2005). Validity of Antonovsky's sense of coherence
scale: a systematic review. Journal of Epidemiology and Community Health, 59,
460-466.

Eysenck, S. B., & Eysenck, H. J. (1977). The place of impulsiveness in a dimensional
system of personality description. British Journal of Social and Clinical
Psychology, 16(1), 57-68.

Flay, B. R., d'Avernas, J. R., Best, J. A., Kersell, M. W., & Ryan, K. B. (1983). Cigarette
smoking: Why young people do it and ways of preventing it. In P. McGrath & P.
Firestone (Eds.), Pediatric and Adolescent Behavioral Medicine (pp. 132-183).
New York: Springer.

Frey, K. S., Hirschstein, M. K., & Guzzo, B. A. (2000). Second step: preventing
aggression by promoting social competence. Journal of Emotional and
Behavioral Disorders, 8(2), 102-112.

Fry, J. (2003). Hispanic youths dropping out of U.S. schools: Measuring the challenge.
The Pew Hispanic Center.

Fuster, J. M. (2008). The prefrontal cortex. London: Elsevier.

Giedd, J. N., Blumenthal, J., Jeffries, N. O., Castellanos, F. X., Liu, H., Zijdenbos, J., et
al. (1999). Brain development during childhood and adolescence: a longitudinal
MRI study. Nature Neuroscience, 2, 861 - 863.

Giedd, J. N. (2004). Structural magnetic resonance imaging of the adolescent brain.
Annals of the New York Academy of Sciences, 1021, 77-85.

96
Glanz, K., Maskarinec, G., & Carlin, L. (2005). Ethnicity, sense of coehrence, and
tobacco use among adolescents. Annals of Behavioral Medicine, 29(3), 192-199.

Glassman, M., Werch, C. E., & Jobli, E. (2007). Alcohol self-control behaviors of
adolescents. Addictive Behaviors, 32, 590-597.

Gordon, M. (1979). The assessment of impulsivity and mediating behaviors in
hyperactive and non-hyperactive boys. Journal of Abnormal Child Psychology, 7,
317-326.

Gottfredson, M. R., & Hirschi, T. (1990). A General Theory of Crime. Stanford, CA:
Stanford University Press.

Gottman, J., Gonso, J., & Rasmussen, B. (1975). Social interaction, social competence,
and friendship in children. Child Development, 46: 709-718.

Graham, J. W., Flay, B. R., Johnson, C. A., Hansen, W. B., Grossman, L., & Sobel, J. L.
(1984). Reliability of self-report measures of drug use in prevention research:
Evaluation of the project SMART questionnaire via the test-retest reliability
matrix. Journal of Drug Education, 14, 175-193.

Hays, R. D., & Hayashi, T. (1990). Beyond internal consistency reliability: Rationale and
user's guide for multitrait analysis program on the microcomputer. Behavior
Research Methods, Instruments, and Computers, 22(2), 167-175.

Hegerl, U., Lipperheide, K., Juckel, G., Schmidt, L. G., & Rommelspacher, H. (1995).
Antisocial tendencies and cortical sensory-evoked responses in alcoholism.
Alcoholism, Clinical and Experimental Research, 19(1), 31-36.

Hernandez, J. T., & Diclemente, R. J. (1992). Self-control and ego identity development
as predictors of unprotected sex in late adolescent males. Journal of Adolescence,
15(4), 437-447.

Hirschi, T. (1969). Causes of Delinquency. Berkeley , CA: University of California Press.

Hollingshead, A. B., & Redlich, F. C. (1958). Social class and mental illness: a
community study. American Journal of Public Health, 97(10), 1756-1757.

Hu, L. T., & Bentler, P. M. (1999). Fit indices in covariance structure modeling:
sensitivity to underparameterized model misspecification. Psychological Methods,
3, 424-453.

Infante, D. A., & Rancer, A. S. (1982). A conceptualization and measure of
argumentativeness. Journal of Personality Assessment, 46(1), 72-80.

97

Infante, D. A., & Wigley III, C. J. (1986). Verbal aggressiveness: An interpersonal model
and measure. Communication Monographs, 53, 62-69.

Jessor, R., & Jessor, S. (1977). Problem-behavior and psychosocial development: A
longitudinal study of youth. New York: Academy Press.

Johnston, L. D., O'Malley, P. M., & Bachman, J. G. (1999). National survey results on
drug use from the Monitoring the Future Study, 1975-1998. Rockville, MD:
U.S.DHHS (NIH Publication Nos. 99-4660 and 99-4661).

Johnston, L. D., O'Malley, P. M., Bachman, J. G., & Schulenberg, J. E. (2008).
Monitoring the Future national results on adolescent drug use: Overview of key
findings, 2007 (No. (NIH Publication No. 08-6418)). Bethesda, MD: National
Institute on Drug Abuse.

Karch, D., Crosby, A., & Simon, T. (2004). Toxicology testing and results for suicide
victims-13 States. Mortality and Morbidity Weekly Report, 55, 1245-1248.

King SM, Iacono WG, McGue M: Childhood externalizing and internalizing
psychopathology in the prediction of the early substance use. Addiction 2004, 99:
1548-1559.

Koposov, R. A., Ruchkin, V. V., & Eisemann, M. (2003). A mediator between violence
exposure and psychopathology in Russian juvenile delinquents. The Journal of
Nervous and Mental Disease, 191(10), 638-644.

Lundqvist, T. (1995). Chronic cannabis use and the sense of coherence. Life Science,
56(23/24), 2145-2150.

Magid, V., MacLean, M. G., & Colder, C. R. (2007). Differentiating between sensation
seeking and impulsivity through their mediated relations with alcohol use and
problems. Addictive Behaviors, 32, 2046-2061.

Margalit, M., & Eysenck, S. (1990). Prediction of coherence in adolescence: gender
differences in social skills, personality, and family climate. Journal of Research in
Personality, 24, 510-521.

Martin, C. S., Earlywine, M., Blackson, T. C., Vanyukov, M. M., Moss, H. B., & Tarter,
R. E. (1994). Aggressivity, inattention, hyperactivity, and impulsivity in boys at
high and low risk for substance abuse. Journal of Abnormal Child Psychology,
22(2), 177-203.

98
McCroskey, J. C., & Richmond, V. P. (1993). Interpersonal processes, cognitions, and
predispositions. Communication Research Reports, 10(2), 107-114.

McMurran, M., & Whitman, J. (1990). Strategies of self-control in male young offenders
who have reduced their alcohol consumption without formal intervention. Journal
of Adolescence, 13(2), 115-128.

McGue M, Iacono WG, Legrand LN, Malone S, Elkins I: Origins and consequences of
age at first drink. I. Associations with substance-use disorders, disinhibitory
behavior and psychopathology, and P3 amplitude. Alcohol Clin Exp Res 2001, 25:
1156-1165.

Meichenbaum, D. (1985). Stress inoculation training. New York: Plenum.

Mezzich, A. C., Tarter, R. E., Giancola, P. R., Lu, S., Kirisci, L., & Parks, S. (1997).
Substance use and risk sexual behavior in female adolescents. Drug and alcohol
dependence, 44, 157-166.

Milich, R., & Kramer, J. (1982). Socialization and peer relations in hyperactive children.
In K. Gadow & I. Bialer (Eds.), Advances in learning and behavioral disabilities
(Vol. 3, pp. 283-339). Greenwich, CT: JAI Press.

Mischel, W. (1974). Processes in delay of gratification. Advances in Experimental and
Social Psychology, 7, 249-292.

Muthen, B. O., du Toit, S. H. C., & Spisic, D. (1997). Robust inference using weighted
least squares and quadratic estimating equations in latent variable modeling with
categorical and continuous outcomes. Retrieved from
http://www.gseis.ucla.edu/faculty/muthen/articles/Article_075.pdf

National Institute on Drug Abuse. NIDA facts. Lessons from prevention rese
Retrieved December 20, 2006, from
http://www.drugabuse.gov/pdf/infofacts/prevention04.pdf

Nilsson, K. W., Starrin, B., Simonsson, B., & Leppert, J. (2007). Alcohol-related
problems among adolescents and the role of a sense of coherence. International
Journal of Social Welfare, 16, 159-167.

Novaco, R. W. (1975). Anger control: The development and evaluation of an
experimental treatment. Lexington, MA: D.C. Health.

Novak, S. P., & Clayton, R. R. (2001). The influence of school environment and self-
regulation on stages of cigarette smoking. Health Psychology, 20(196-207).

99
O’Donoghue, T., & Rabin, M. (1999). Doing It Now or Later. American Economic
Review, 89(103-124).

Parrott, A. C., Sisk, E., & Turner, J. J. D. (2000). Psychobiological problems in heavy
''ecstasy'' (MDMA) polydrug users. Drug and Alcohol Dependence, 60(1), 105-
110.

Patterson, G.R., DeBaryshe, B.D., & Ramsey, E. (1989). A developmental perspective on
antisocial behavior. American Psychologist, 44(2): 329-335.

Paus, T. (2005). Mapping brain maturation and cognitive development during
adolescence. Trends in Cognitive Sciences, 9(2), 60-68.

Pentz, M. A., & Chou, C.-P. (1994). Measurement invariance in longitudinal clinical
research assuming change from development and intervention. Journal of
Consulting and Clinical Psychology, 62(3), 450-462.

Petriatis, J., Flay, B. R., & Miller, T. Q. (1995). Reviewing theories of adolescent
substance use: Organizing the pieces in the puzzle. Psychological Bulletin,
117(67-86).

Petry, N. (2001). Substance abuse, pathological gambling, and impulsiveness. Drug and
Alcohol Dependence, 63(1), 29-38.

Robinson, T. E., & Berridge, K. C. (2003). Addiction. Annu. Rev. Psychol., 54, 25-53.

Robinson, T. E., & Kolb, B. (2004). Structural plasticity associated with exposure to
drugs of abuse. Neuropharmacology, 47, 133-146.

SAMHSA model programs. Retrieved December 20, 2006, from
http://nrepp.samhsa.gov/programfulldetails.asp?PROGRAM_ID=62

Satorra, A. (2000). Scaled and adjusted restricted tests in multi-sample analysis of
moment structures. In R. D. H. Heijmans, D. S. G. Pollock & A. Satorra (Eds.),
Innovations in multivariate statistical analysis: A Festschrift for Heinz Neudecker
(pp. 233-247). London: Kluwer Academic Publishers.

Satorrra, A., & Bentler, P. M. (1999). A scaled difference chi-square test statistics for
moment structure analysis. Los Angeles: University of California, Los Angeles.

Schalling, D. (1978). Psychopathy-related personality variables and the
psychophysiology of socialization. In R. D. Hare & D. Schalling (Eds.),
Psychopathic behaviour: Approaches to research (pp. 85-105). New York: Wiley.

100
Shrier, L. A., Harris, S. K., Kurland, M., & Knight, J. R. (2003). Substance use problems
and associated psychiatric symptoms among adolescents in primary care.
Pediatrics, 111(6 Pt 1), e699-705.

Simons, J. S., & Carey, K. B. (2002). Risk and vulnerability for marijuana use problems:
The role of affect dysregulation. Psychology of Addictive Behaviors, 16, 72-75.

Simonsson, B., Nilsson, K. W., Leppert, J., & Diwan, V. K. (2008). Psychosomatic
complaints and sense of coherence among adolescents in a county in Sweden: a
cross-sectional school survey. Biopsychosocial Medicine, 2(4).

Snow, D.L., Tebes, J.K., Arthur, M.W. (1992). Panel attrition and external validity in
adolescent substance use research. Journal of Consulting and Clinical Psychology,
60 (5): 804-807.

Spivak, G., & Shure, M. B. (1982). The cognition of social adjustment: Interpersonal
cognitive problem-solving thinking. In B. B. Lahey & A. E. Kazdin (Eds.),
Advances in clinical child psychology (Vol. 5, pp. 323-372). New York: Plenum.

Stacy, A. W., Newcomb, M. D., & Bentler, P. M. (1991). Cognitive motivation and
problem drug use: A 9-year longitudinal study. Journal of Abnormal Psychology,
100, 502-515.

Steenkamp, J.-B. E. M., & Baumgartner, H. (1998). Assessing measurement invariance in
cross-national consumer research. The Journal of Consumer Research, 25(1), 78-
90.

Steiger, J. H. (1990). Structural model evaluation and modification: An interval
estimation approach. Multivariable Behavioral Research, 25, 173-180.

Steinberg, L. (2005). Cognitive and affective development in adolescence. Trends in
Cognitive Sciences, 9(2), 69-74.

Sun, P., Sussman, S., Dent, C.W., Rohrbach, L.A. (2008). One-year follow-up evaluation
of Project Toward No Drug Abuse (TND-4). Preventive Medicine, 47 (4): 438
442.

Sussman, S., & Ames, S. L. (2008). Drug abuse concepts, prevention, and cessation.
New York: Cambridge University Press.

Sussman, S., Dent, C., Burton, D., Stacy, A., & Flay, B. (1995). Developing school-based
tobacco use prevention and cessation programs. Thousand Oaks, CA: Sage
Publications.

101
Sussman, S., Dent, C. W., & Galaif, E. R. (1997). The correlates of substance abuse and
dependence among adolescents at high risk for drug abuse. Journal of Substance
Abuse Treatment, 9, 241-255.

Sussman, S., Dent, C. W., & Leu, L. (2000). The one-year prospective prediction of
substance abuse and dependence among high-risk adolescents. Journal of
Substance Abuse, 12, 373-386.

Sussman, S., Dent, C. W., Stacy, A. W., & Craig, S. (1998). One-year outcomes of
Project Towards No Drug Abuse.[erratum appears in Prev Med 1998 Sep-
Oct;27(5 Pt 1):766]. Preventive Medicine, 27(4), 632-642.

Sussman, S., Dent, C. W., & Stacy, S. (2002). Project Towards No Drug Abuse: A
review of the findings and future directions. Am. J. Health Behav., 26(5), 354-
365.

Sussman, S., McCuller, W. J., & Dent, C. W. (2003). The associations of social self
control, personality disorders, and demographics with drug use among high risk
youth. Addictive Behaviors, 28, 1159-1166.

Sussman S, Pentz MA, Spruijt-Metz D, Miller T:. Misuse of "study drugs:" prevalence,
consequences, and implications for policy. Subst Abuse Treat Prev Policy 2006,
1: 15.

Sussman, S., & Unger, J. B. (2004). A "Drug Abuse" Theoretical Integration: A
Transdisciplinary Speculation. Substance Use & Misuse, 39(10-12), 2055-2069.

Tarter, R. E. (2002). Etiology of Adolescent Substance Abuse: A Developmental
Perspective American Journal on Addictions, 11(3), 171-191.

Torsheim, T., Aaroe, L. E., & Wold, B. (2001). Sense of coherence and school-related
stress as predictors of subjective health complaints in early adolescence:
interactive, indirect or direct relationships? Social Science & Medicine, 53, 603-
614.

Triandis, H.C. (1989). The self and social behavior in differing cultural contexts.
Psychological Review, 93 (3): 506-520.

Valente, T. W., Ritt-Olson, A., Stacy, A., Unger, J. B., Okamoto, J., & Sussman, S.
(2007). Peer acceleration: effects of a social network tailored substance abuse
prevention program among high-risk adolescents. Addiction, 102(11), 1804-1815.

102
Verdejo-Garcia, A., Bechara, A., Recknor, E. C., & Perez-Garcia, M. (2007). Negative
emotion-driven impulsivity predicts substance dependence problems. Drug and
Alcohol Dependence, 91, 213-219.

Vega, W. A., Khoury, E. L., Zimmerman, R. S., Gil, A. G., & Warheit, G. J. (1995).
Journal of Community Psychology, 23: 167-179.

Vollmer, T. R., Borrero, J. C., Lalli, J. S., & Daniel, D. (1999). Evaluating self-control
and impulsivity in children with severe behavior disorders. Journal of Applied
Behavior Analysis, 32(4), 451-466.

Weissberg, R. P. (1985). Designing effective social problem-solving programs for
classroom. In B. H. Schneider, K. H. Rubin & J. E. Ledingham (Eds.), Children's
peer relations: Issues in assessment and intervention (pp. 225-242). New York:
Springer-Velag.

Werch, C. E., & Gorman, D. R. (1986). Factor analysis of internal and external self-
control practices for alcohol consumption. Psychological Reports, 59(3), 1207-
1213.

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.
Prevention Science, 2(3), 145-163.

Wills, T. A., & Dishion, T. J. (2004). Temperament and adolescent substance use: a
transactional analysis of emerging self-control.[see comment]. Journal of Clinical
Child & Adolescent Psychology, 33(1), 69-81.

Wills, T. A., & Filer, M. (1996). Stress-coping model of adolescent substance use. In T.
H. Ollendick & R. J. Prinz (Eds.), Advances in clinical child psychology (Vol. 18,
pp. 91-132). New York: Plenum.

Wills, T. A., Gibbons, F. X., Gerrard, M., Murry, V. M., & Brody, G. H. (2003). Family
communication and religiosity related to substance use and sexual behavior in
early adolescence: a test for pathways through self-control and prototype
perceptions. Psychology of Addictive Behaviors, 17(4), 312-323.

Wills, T. A., Murry, V. M., Brody, G. H., Gibbons, F. X., Gerrard, M., Walker, C., et al.
(2007). Ethnic prode and self-control related to protective and risk factors: Test of
the theoretical model for the strong African American families program. Health
Psychology, 26(1), 50-59.

103
Wills, T. A., Resko, J. A., Ainette, M. G., & Mendoza, D. (2004). Role of parent support
and peer support in adolescent substance use: a test of mediated effects.
Psychology of Addictive Behaviors, 18(2), 122-134.

Wills, T. A., Sandy, J. M., & Yaeger, A. (2000). Temperament and adolescent substance
use: an epigenetic approach to risk and protection. Journal of Personality, 68(6),
1127-1151.

Wills, T. A., Walker, C., Mendoza, D., & Ainette, M. G. (2006). Behavioral and
emotional self-control: Relations to substance use in samples of middle and high
school students. Psychology of Addictive Behaviors, 20(3), 265-278.

Winters, K. C., Stinchfield, R. D., & Henly, G. A. (1989). Further validation of new
scales measuring adolescent alcohol and other drug abuse. Journal of Studies on
Alcohol, 54, 534-541.

Zuckerman, M. (1979). Sensation seeking: Beyond the optimal level of arousal. Hillsdale,
NJ: Erlbaum.

Zuckerman, M., Kuhlman, D. M., Joireman, J., Teta, P., & Kraft, M. (1993). A
comparison of three structural models for personality : the Big Three, the Big
Five, and the Alternative Five. Journal of Personality and Social Psychology,
65(4), 757-768.

Zuckerman, M., Kuhlman, D. M., Thornquist, M., & Kiers, H. (1991). Five (or three)
robust questionnaire scale factors of personality without culture. Personality and
Individual Differences, 12, 929-941.

104
Appendix A
Measure
Table1. Social self-control items (Sussman, McCuller, & Dent, 2003)

Social Self-control Items

Response options
I enjoy arguing with people.

I express all of my feelings.

If I think something someone says is stupid I
tell them so.

If I am angry I act like it.

My mouth gets me in trouble a lot.

I do things just to get attention.

Sometimes I provoke people just for the fun of
it.

My feelings get hurt easily.

I hate being wrong.

I say things that I regret later.

1. Always
2. Usually
3. Sometimes
4. Never

105

Table 2: Sensation-seeking items (Zuckerman, Kuhlman, Thornquist, & Kiers,
1991)
Items Response Options

I like to have new and exciting experiences and
sensations even if they are a little frightening.

I would like to take off on a trip with no
preplanned or definite routes or timetable.

I enjoy getting into new situations where you
can't predict how things will turn out.

I like doing things just for the thrill of it.

I sometimes like to do things that are a little
frightening.

I'll try anything once.

I would like the kind of life where I’m on the
move and traveling a lot, with lots of change
and excitement.

I sometimes do "crazy" things just for fun.

I like to explore a strange city or section of
town by myself, even if it means getting lost.

I prefer friends who are excitingly
unpredictable.

I like "wild" uninhibited parties.

1. True
2. False

106

Table 3. Impulsivity Items (Zuckerman, Kuhlman, Thornquist, & Kiers, 1991)
Items Response Options

I tend to begin a new job without much
advance planning on how I will do it.

I usually think about what I am going to do
before doing it.

I often do things on impulse.

I very seldom spend much time on the details
of planning ahead.

Before I begin a complicated job, I make
careful plans.

I tend to change interests frequently.

I often get so carried away by new and exciting
things and ideas that I never think of possible
complications.

I am an impulsive person.

1. True
2. False

107
Table 4. Sense of Coherence items (Antonovsky, 1987)
Items

Response options
Do you have the feeling that you don’t really
care about what goes on around you?

1 (Never have this feeling) to 7 (Always have
this feeling)
Has it happened in the past that you were
surprised by the behavior of people whom you
thought you knew well?

1 (Never happened) to 7 (Always happened)
Has it happened that people whom you counted
on disappointed you?

1 (Never happened) to 7 (Always happened)
Until now your life has had… 1 (No clear goals or purpose at all) to 7 (Very
clear goals and purpose)

Do you have the feeling that you’re being
treated unfairly?

1 (Never) to 7 (Always)
Do you have the feeling that you are in a
strange situation and don’t know what to do?

1 (Never) to 7 (Always)
Do the things you do every day is… 1 (A source of pain and boredom) to 7 (A
source of deep pleasure and satisfaction)

Do you have very mixed-up feelings and ideas?

1 (Never) to 7 (Always)
Does it happen that you have feelings inside
you would rather not feel?

1 (Never) to 7 (Always)
Many people-even those with a strong
character—sometimes feel like sad sacks
(losers) in certain situations. How often have
you felt this way in the past?

1 (Never) to 7 (Always)
When something happened, you generally
found that…
1 (You overestimated or underestimated its
importance) to 7 (You saw things the way it
really was)

How often do you have the feeling that there’s
little meaning in the things you do in your daily
life?

1 (Never) to 7 (Always)

108
Appendix B
Confirmatory Factor Analysis and testing the measurement invariance
of social self-control across ethnicities and three time-points
Abstract
Confirmatory factor analysis (CFA) was used to test a measurement model of
social self-control. First, a CFA was conducted on the baseline sample of 2495
adolescents on the 8 social self-control items that were selected based on the findings
from Chapter 2. To achieve a reasonable model fit, two items that showed standardized
factor loadings of less than 0.50 were excluded from the model. Second, the measurement
model consisting of the 6 remaining social self-control items was tested for invariance
across three time-points. The first and the second time-points were 6 weeks apart (i.e., T1
and T2). The time interval between T2 and the third time-point (T3) was 1-year. Third,
the measurement model which was tested for invariance across 4 ethnic categories (Non-
Hispanic White, Hispanic, African-American, and Asian-American) at each time-point.
Although the measure was not found to be fully measurement invariant across ethnicities
and over time, social self-control was found to be partially measurement invariant such as
to suggest acceptable levels of cross-ethnic validity and over-time stability.

109

Introduction
Currently, it is not known whether the self-report items of social self-control
relate to the latent social self-control construct similarly across different ethnic groups in
the United States. Assuming that the high school students from ethnic minority groups in
the US are highly acculturated to the mainstream US culture, it may be hypothesized that
social self-control means the same for Hispanics, African-Americans, and Asian-
Americans, compared with non-Hispanic Whites. The present study attempted to test this
hypothesis. In addition, the present study examined the longitudinal stability of the social
self-control construct by comparing the relationships between the social self-control
indicators (i.e., items) and the latent social self-control construct at baseline, 6-week
follow-up, and 1-year follow-up. It is reasonable to assume that the mean level of social
self-control may increase over time: as adolescents grow older they become socially more
mature. However, it is not clear currently whether the latent constructs of social self-
control measured at two time-points separated by an interval of as long as a year are
comparable.
First, the present study used confirmatory factor analysis (CFA) to establish a
measurement model of social self-control. Next, we tested the measurement invariance of
the model across time-points and ethnicities. Measurement invariance of the social self-
control construct across the 4 ethnic groups and the 3 time-points would suggest that the
construct is comparable across ethnic groups and is stable over time. Measurement
invariance is thought to be achieved when parameters of the measurement model are

110
equivalent across groups (Pentz & Chou, 1994). The present chapter examined two
parameters in particular: the intercepts and the factor loadings, thus testing scalar and
metric invariance (Steenkamp & Baumgartner, 1998), respectively. When an item is
measured as a function of a latent construct the factor loading represents the regression
coefficient relating the latent construct to the item and the intercept represents the mean
the observed item score when the latent construct is zero.
Methods
Subjects and data collection
Data collection procedure is described in the methods section of Chapter 2. In
addition to using the data collected at baseline (T1) and 1-year follow-up (T3), the
present study also used data collected at the second wave (T2), which was not discussed
in Chapter 3. Total 2734 subjects were surveyed at T1, of whom 2331 and 2064 were
followed up at T2 (6-weeks later) and T3 (1-year later), respectively. Only those subjects
who clearly identified themselves as non-Hispanic White, Hispanic, African-American,
or Asian-American were included in the present study. Subjects with missing ethnicity
data or subjects who identified themselves as “Other” were excluded from the study. The
total sample sizes analyzed in the present study for T1, T2, and T3 were 2395, 2218, and
1950, respectively. Table 1a lists the ethnic distribution at each time-point.
Measures
Ethnicity. Ethnicity was assessed using the following ethnic indicator: “What is
your ethnic background? Please circle one category that best applies.” The response
options included a) Black or African American; b) Asian/Pacific Islander (subcategories:

111
Chinese, Japanese, Filipino, Korean, and open-ended “other”); c) Latino/Hispanic
(subcategories: Mexican-American, Mexican, Central American, South American, and an
open-ended “other”) ; d) White/Non Latino; e) Native American; and f) Other ethnic
group (open-ended). Since only 15 subjects (0.52%) subjects identified themselves as
Native American, they were excluded from the present analysis together with subjects
who did not provide their ethnic information or identified themselves as Other.
Social self-control. The same social self-control measure used in Chapters 2 and 3
was used in the present analysis. Based on the findings from Chapter 2, the following two
items were excluded from the present analysis: “I express all my feelings”; “My feelings
get hurt easily.” For the complete list of social self-control items see Appendix I.

Table 1a. Ethnic composition at each time-point.
T1 (N=2395)

T2 (N=2218) T3 (N=1950)
Non-Hispanic White

458 418 347
Hispanic

1532 1422 1279
African-American

190 183 143
Asian-American

215 195 181

112
Table1b. The six social self-control items
Item code Item
SSC1 I enjoy arguing with people

SSC2 If I am angry I act like it

SSC3 My mouth gets me in trouble a lot

SSC4 I do things just to get attention

SSC5 Sometimes I provoke people just for the fun of it

SSC6 I say things that I regret later

Analyses & Results
Mplus (Muthen & Muthen, 1998-2009) was used to test all CFA models in the
present study using maximum likelihood estimation. As Chi-square values tend to inflate
due to large sample sizes, the model fits were additionally evaluated using other fit
indices such as the root mean squared error of approximation (RMSEA; Steiger, 1990)
and the comparative fit index (CFI; Bentler, 1990). RMSEA values less than 0.05
indicate close approximate fit whereas values between 0.05 and 0.08 indicate a
reasonable fit (Kline, 2005). CFI values above 0.95 are considered to indicate a
reasonably good fit (Hu & Bentler, 1999).
The initial CFA model was tested on the baseline data (T1; N=2395) with the 8
social self-control items as indicators of the latent social self-control construct. The
model fit was only marginally reasonable (Chi-square= 224.2; DF=18; p<0.0001;
RMSEA= 0.07; CFI= 0.93). Hence, setting the cut-off value for a meaningful
standardized factor loading at 0.50 (Wills et al., 2006), two items were excluded from the
model. The resulting measurement model with 6 indicators showed a good model fit

113
(Chi-square= 34.4; DF=7; p<0.0001; RMSEA= 0.04; CFI=0.99). Also, the 6 indicator
measurement model showed a good fit at T2 and T3 (see Model 1, Table 2) and for each
ethnicity at T1, T2, and T3 (see Model 1; Tables 3, 4, & 5).
Several steps recommended by Pentz & Chou (1994) were taken to test
measurement invariance using the multi-group CFA approach. Time invariance was
tested by comparing the model across three groups represented by T1, T2, and T3. Ethnic
invariance was tested by comparing the basic model across the 4 ethnic groups at T1, T2,
and T3 separately. After ascertaining that the basic measurement model was statistically
adequate, we followed the next step of testing for equality of intercepts across groups
(Pentz & Chou, 1994). To test equality of intercepts we imposed equality constraints on
intercepts across groups. If the imposition of equality constraints deteriorated the model
fit, equality constraints for individual intercept parameter were released as guided by
Mplus model modification indices. Significant deterioration of model fit was tested using
the Chi-Square difference test. According to this test, if the difference in Chi-Square
between the restricted model and the less restricted model is statistically significant then
the null hypothesis that the restricted model is as good as the less restricted model is
rejected. In other words, the constraints imposed on the model are considered to have
significantly worsened the model fit.
After testing for equality of intercepts we tested for equality of factor loadings
across groups (Pentz & Chou, 1994). As with intercepts, first we imposed equality
constraints on factor loadings across groups. If the imposition of equality constraints

114
deteriorated the model fit, we released the equality of constraints for individual factor-
loading parameter as guided by the model modification indices.
Testing for time invariance. First, equality constraints were imposed on all
indicator intercepts across T1, T2, and T3, while the factor loadings remained free. The
resulting model fit was significantly worse compared with the baseline model (see Model
1; Table2) in which both the intercepts and factor loadings were feely estimated across
the 3 time groups (∆Chi-square=105.2; ∆DF= 10; p<0.0001). As guided by the model
modification indices, three T2 intercept parameters (corresponding to items SSC3, SSC4,
and SSC6; see Table 1b) and two T3 intercept parameters ( corresponding to SSC3 and
SSC5) were freed. The resulting model was not significantly worse than the baseline
model (Model 1) (∆Chi-square=10.52; ∆DF= 5; p>0.05). Next, equality constraints were
imposed on all factor loadings across T1, T2, and T3 with only the intercepts found to be
non-invariant in the last model were released to be free. Model 2 in Table 2 shows the
resulting factor loadings and model fit indices. As seen in Table 2, placing the equality
constraints on factor loadings across three time groups significantly worsened the
baseline model. Again, as guided by model modification indices, the equality constraints
were released from two T3 factor loadings (SSC4 and SSC5; see Table 1b & Table 2).
The resulting model (Model 3, Table 2) was not significantly worse than the baseline
model (see Table 2). As shown by Model 3 in Table 2, all factor loadings were invariant
across T1 and T2. However, only 4 out of the 6 factor loadings were invariant across T1,
T2, and T3.

115
Table 2. Parameter estimates and fit indices for testing measurement invariance
models across three time-points (baseline, 4-week follow-up, & 1-year follow-up).
Parameters Model 1 (M1) Model 2
(M2)
Model 3 (M3)
T1
(N=2495)
T2
(N=2308)
T3
(N=2033)
T1
(N=2495)
T2
(N=2308)
T3
(N=2033)
λ
SSC1

1.00 1.00 1.00 1.00 1.00 1.00 1.00
λ
SSC2
(se)

1.19
(0.08)

1.14
(0.07)
1.35
(0.10)
1.17
(0.04)
1.18
(0.04)
1.18
(0.04)
1.18
(0.04)
λ
SSC3

(se)
1.80
(0.11)

1.79
(0.11)
2.05
(0.14)
1.82
(0.06)
1.84
(0.06)
1.84
(0.06)
1.84
(0.06)
λ
SSC4

(se)
0.77
(0.05)

0.83
(0.06)
0.69
(0.06)
0.75
(0.03)
0.81
(0.04)
0.81
(0.04)
0.66
(0.04)
λ
SSC5

(se)
1.16
(0.06)

1.16
(0.06)
0.92
(0.06)
1.08
(0.03)
1.17
(0.04)
1.17
(0.04)
0.88
(0.05)
λ
SSC6

(se)
0.96
(0.07)

0.83
(0.06)
0.98
(0.08)
0.89
(0.04)
0.89
(0.04)
0.89
(0.04)
0.89
(0.04)
Chi-square

96.92 142.54 114.87
DF

21 36 34
CFI

0.99 0.98 0.99
RMSEA
(95% CI)

0.040 (0.032-0.048)

0.036
(0.030,
0.042)
0.032 (0.026, 0.039)
∆ Chi-
square

M2-M1=
45.62
M3-M1= 17.95
∆ DF

M2-M1=
15
M3-M1= 13
p-value

<0.0001 p>0.05
Note. λ= Item factor loading; se=standard error; Model 1= λ fully free across T1, T2, and
T3; Model 2= λ fully constrained to be equal across 3 time-points; Model 3= λ partially
constrained across 3 time-points; λ in bold indicated λ was released from equality
constraint; for SSC1-SSC6 see Table 1b; CFI= Comparative Fit Index; RMSEA= Root
mean squared error of approximation.

116
Testing for ethnic invariance. Measurement invariance of social self-control
across non-Hispanic Whites, Hispanics, African Americans, and Asian Americans was
tested at T1, T2, and T3 separately. Model 1 in Table 3 represents the baseline model
with factor loadings that were freely estimated across ethnicities. In the next step,
intercept parameters were constrained to be equal across ethnicities. The fit of the
resulting model was significantly worse compared with the baseline model (∆Chi-
square= 52.27; ∆DF= 15; p<0.0001). Based on the model modification indices, 4
intercept parameters were subsequently freed: 1 for Hispanics (SSC6), 2 for African
Americans (SSC6 and SSC4), and 1 for Asian Americans (SSC10). The resulting model
was not significantly different from the baseline model (∆Chi-square= 16.12; ∆DF= 11;
p>0.05). Next, with the freed intercepts intact, each factor loading was constrained to be
equal across the 4 ethnic groups. The factor loadings and fit indices of this model are
presented in Table 3 (Model 2). As shown in Table 3, the Chi-square difference test
suggested that Model 2 was not significantly worse compared with Model 1. Thus we
found that the factor loadings of the 6 social self-control indicators were invariant across
ethnicities at T1.
The same steps were repeated for T2 (see Table 4) and T3 (see Table 5). First,
intercepts were constrained to be equal across ethnic groups, which significantly
worsened the baseline models for both T2 (∆Chi-square= 92.05; ∆DF= 15; p<0.0001)
and T3 (∆Chi-square= 64.13; ∆DF= 15; p<0.0001). Based on model modification indices,
equality constraints were released from 3 intercept parameters within each of the
following ethnic groups in T2: Hispanic (SSC6, SSC10, and SSC7), African American

117
(SSC6, SSC10, SSC4), and Asian American (SSC4, SSC5, SSC10). Similarly, in T3, 1
intercept was released for Whites (SSC5), 2 for Hispanics (SSC6 and SSC4), and 1 for
African Americans (SSC6). The resulting models were not significantly worse compared
with the baseline model for either T2 (∆Chi-square= 10.33; ∆DF= 6; p>0.05) or T3
(∆Chi-square= 19.06; ∆DF= 11; p>0.05). Next, factor loadings were constrained to be
equal across ethnicities. The resulting model for T2 was not significantly worse from the
baseline model (see Table 4; Model 2). For T3, however, the constrained model was
significantly worse (see Model 2, Table 5). Hence, based on the model modification
indices, the equality constraints were released from three factor loadings, one for Whites
and two for Hispanics (see Model 3, Table 5). The resulting model was found not to be
different from the baseline model (see Table 5). Hence, we found that 4 out of the 6
factor-loadings were invariant across ethnic groups at T3. Of the remaining two, the
factor loading for SSC5 was different for Whites and Hispanics but same for African
Americans and Asian Americans, whereas the factor loading for SSC4 was different for
Hispanics but same for Whites, African Americans, and Asian Americans.

118
Table 3. Parameter estimates and fit indices for testing measurement invariance
models across ethnicities at Time 1 (baseline)
Parameters Model 1 Model 2

W
(N=458)

H
(N=1532)

B
(N=190)
A (N=215)
λ
SSC1

1.00 1.00 1.00 1.00 1.00
λ
SSC2

1.04 (0.19) 1.19
(0.09)
1.12 (0.28) 1.62 (0.53)

1.18 (0.08)
λ
SSC3

1.96 (0.32) 1.66 (0.12) 1.90 (0.45) 2.67 (0.87)

1.80 (0.11)
λ
SSC4

0.76 (0.14) 0.82 (0.07) 0.63 (0.17) 0.63 (0.28)

0.79 (0.06)
λ
SSC5

1.17 (0.16) 1.14 (0.07) 1.18 (0.23) 1.13 (0.31)

1.16 (0.06)
λ
SSC6

1.14 (0.19) 0.91 (0.08) 0.92 (0.26) 1.49 (0.47)

0.99 (0.07)
Chi-square

61.58 95.69
DF

28 54
CFI

0.99 0.98
RMSEA
(95% CI)

0.045 (0.03, 0.06) 0.036 (0.024, 0.048)
∆ Chi-
square

M2-M1= 34.11
∆ DF

M2-M1= 26
p-value

p>0.05
Note. λ= Item factor loading; Model 1= λ fully free across ethnicities; Model 2= λ fully
constrained to be equal across ethnicities; SSC1-SSC6 (see Table 1b); A=Asian
American, B= African American; H= Hispanic; W= White. CFI= Comparative fit index;
RMSEA= Root mean squared error of approximation.

119
Table 4. Parameter estimates and fit indices for testing measurement invariance
models across ethnicities at Time 2.
Model 1 Model 2

W
(N=418)
H
(N=1422)
B (N=183) A
(N= 195)

λ
SSC1

1.00 1.00 1.00 1.00 1.00
λ
SSC2

1.18
(0.23)

1.10 (0.08) 1.24 (0.34) 1.35 (0.39) 1.12 (0.07)
λ
SSC3

2.25
(0.40)
1.64 (0.11) 1.54 (0.41) 2.26 (0.58)

1.75 (0.10)
λ
SSC4

1.24
(0.22)
0.73 (0.06) 0.85 (0.24) 1.02 (0.29)

0.81 (0.06)
λ
SSC5

1.23
(0.19)
1.10 (0.07) 1.20 (0.25) 1.21 (0.26)

1.13 (0.06)
λ
SSC6

1.06
(0.20)
0.79 (0.07) 0.69 (0.22) 1.55 (0.43)

0.86 (0.06)
Chi-square

74.87 105.34
DF

28 49
CFI

0.98 0.98
RMSEA
(95% CI)

0.055 (0.040, 0.070) 0.047 (0.034, 0.060)
∆ Chi-
square

M2-M1= 30.47
∆ DF

M2-M1=21
p-value

p>0.05
Note. λ= Item factor loading; Model 1= λ fully free across ethnicities; Model 2= λ fully
constrained to be equal across ethnicities; SSC1-SSC6 (see Table 1b); A=Asian
American, B= African American; H= Hispanic; W= White; CFI= Comparative fit index;
RMSEA= Root mean squared error of approximation.

120
Table 5. Parameter estimates and fit indices for testing measurement invariance
models across ethnicities at Time 2 (1-year follow-up)
Model 1

Model 2 Model 3
W
(N=
347)
H
(N=12
79)
B
(N=14
3)
A
(N=18
1)
W H B A
λ
SSC1

1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
λ
SSC2

1.39
(0.28)

1.30
(0.11)
1.82
(0.61)
1.25
(0.47)
1.35
(0.10)
1.33
(0.10)
1.33
(0.10)
1.33
(0.10)
1.33
(0.10)
λ
SSC3

2.68
(0.52)

1.82
(0.14)
4.54
(1.69)
2.36
(0.84)
2.04
(0.14)
2.02
(0.14)
2.02
(0.14)
2.02
(0.14)
2.02
(0.14)
λ
SSC4

1.07
(0.21)

0.59
(0.06)
0.78
(0.32)
1.13
(0.34)
0.70
(0.06)
0.90
(0.10)
0.62
(0.06)
0.90
(0.10)
0.90
(0.10)
λ
SSC5

0.68
(0.15)

0.95
(0.07)
1.06
(0.35)
0.90
(0.26)
0.92
(0.06)
0.62
(0.12)
0.98
(0.07)
0.89
(0.15)
0.89
(0.15)
λ
SSC6

1.17
(0.22)

0.95
(0.09)
0.60
(0.35)
1.38
(0.46)
1.02
(0.08)
1.01
(0.08)
1.01
(0.08)
1.01
(0.08)
1.01
(0.08)
Chi-
square
55.76 95.16 83.28
DF 28 43 47
CFI 0.98 0.97 0.98
RMSE
A
(95%
CI)
0.045 (0.027, 0.062) 0.050
(0.04,
0.06)
0.040 (0.025, 0.054)
∆ Chi-
square
M2-M1=
39.4
M3-M1= 27.52
∆ DF 15 M3-M1=19
p-value

p<0.001 p>0.05
Note. λ= Item factor loading; Model 1= λ fully free across ethnicities; Model 2= λ fully
constrained to be equal across ethnicities; SSC1-SSC6 (see Table 1b); A=Asian
American, B=African American; H= Hispanic; W= White; CFI= Comparative fit index;
RMSEA= Root mean squared error of approximation.

121
Discussion
The present study examined the measurement invariance of a social self-
control measure across 1) three waves of data: baseline, 6-week follow-up, and 1-year
follow-up; and 2) 4 ethnic groups at each wave of data. In particular, we examined scalar
(intercept) and metric (factor-loading) invariance. The results suggested that our
measurement model of social self-control was not fully invariant either across time-points
or across ethnic groups at each time point. However, we found the model to show partial
measurement invariance, thus suggesting that the social self-control measure is fairly
generalizable across ethnicities and moderately stable over time.
In terms of factor-loadings, no significant difference was found between T1 and
T1 and across ethnicities at T1 and T2. This suggests that the relations between the social
self-control items and the social self-control latent construct are similar across two time-
points 4 weeks apart. In addition, the relations between the items and the latent construct
seem to be similar across ethnicities when measured 6-weeks apart. As suggested by the
differences in some of the intercept parameters across groups, the measurement bias at T1
and T2 seems mostly to be at the level of item means. Non-invariant intercept parameters
across groups suggest that the item mean differences do not reflect latent mean
differences in the same way across groups (Pentz & Chou, 1994). Hence, this finding
suggests that the results involving the differences in social self-control at the manifest
level across ethnicities or time-points may be limited in the sense that such differences
may not truly reflect the differences in latent social self-control construct.
For two items (SSC4 and SSC5; see Table 1b for description), the factor-loadings
at T3 differed from factor-loadings at T1 and T2. Further, these same items showed

122
ethnic differences in factor-loadings at T3. The factor loadings for SSC4 and SSC5
showed an overall decrease at T3 compared with the corresponding values at T1 and T2.
Among Hispanics, the factor loading for SSC4 at T3 was lower compared to the value
that was equal across Whites, African Americans, and Asian Americans. In addition, the
factor loadings for SSC5 among Hispanics and Whites at T3 were higher than and lower
than the value that was equal for Asian Americans and African Americans, respectively.
Hence, it appears that the relations of certain social self-control items with the latent
social self-control construct may change differentially across ethnicities over a 1 year
time period. The basis of this change may be developmental, cultural, or an interaction
between the two. Future research is needed to determine the factors that influence the
relations between the social self-control items and the latent construct across ethnicities
over time.
The present study was subject to two major limitations. First, our sample, which
was predominantly Hispanic, was not distributed proportionately across ethnicities.
Although we had enough subjects in each ethnic group necessary for single-group CFA,
the possibility that the unequal sample distribution might have adversely affected the
multigroup comparison cannot be ruled out. Second, our three data collection points were
not separated by equal time intervals, which may make comparisons among the three
time-points less meaningful. However, despite these limitations, the present study is
important in two important regards. First, we established a measurement model of social
self-control that fitted well across time-points and across 4 ethnic categories at each time
point. Second, by systematically testing measurement invariance we were able to

123
establish a model that suggested partial invariance of social self-control across ethnicities
and time-points.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Sociocultural stress, coping and substance use among Hispanic/Latino adolescents

PDF

Contextualizing social network influences on substance use among high risk adolescents

PDF

Exploring the role of peer influence, linguistic acculturation, and social networks in substance use

PDF

The role of depression symptoms on social information processing and tobacco use among adolescents

PDF

Energy drink consumption, substance use and attention-deficit/hyperactivity disorder among adolescents

PDF

The influence of contextual factors on the processes of adoption and implementation of evidence-based substance use prevention and tobacco cessation programs in schools

PDF

Adolescent social networks, smoking, and loneliness

PDF

The dynamic relationship of emerging adulthood and substance use

PDF

A network analysis of online and offline social influence processes in relation to adolescent smoking and alcohol use

PDF

The effects of mindfulness on adolescent cigarette smoking: Measurement, mechanisms, and theory

PDF

Role transitions, past life events, and their associations with multiple categories of substance use among emerging adults

PDF

An examination of the association between spousal support and type 2 diabetes self-management

PDF

Motivational interviewing with adolescent substance users: a closer look

PDF

Post-traumatic growth among high-risk youth: predictors, impact of stressful life events, and relationship with changes in substance use behaviors

PDF

The Internet activities, gratifications, and health consequences related to compulsive Internet use

PDF

Genetic variants and smoking progression in Chinese adolescents

PDF

The role of social support in the relationship between adverse childhood experiences and addictive behaviors across adolescence and young adulthood

PDF

Intrapersonal and environmental factors associated with Chinese youth alcohol use experimentation and binge drinking behaviors

PDF

Affective decision-making predictive of Chinese adolescents drinking behaviors: a longitudinal study

PDF

Prospective associations of stress, compulsive internet use, and posttraumatic growth among emerging adults

Asset Metadata

Creator Pokhrel, Pallav (author)

Core Title Social self-control and adolescent substance use

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Preventive Medicine (Health Behavior)

Publication Date 12/04/2009

Defense Date 09/14/2009

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Adolescents,OAI-PMH Harvest,social self-control,substance use

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sussman, Steven Y. (committee chair), Palinkas, Lawrence A. (committee member), Rohrbach, Louise A. (committee member), Sun, Ping (committee member), Unger, Jennifer B. (committee member)

Creator Email pallav429@gmail.com,pokhrel@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2785

Unique identifier UC1310846

Identifier etd-Pokhrel-3323 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-280990 (legacy record id),usctheses-m2785 (legacy record id)

Legacy Identifier etd-Pokhrel-3323.pdf

Dmrecord 280990

Document Type Dissertation

Rights Pokhrel, Pallav

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu