Content uploaded by Brian B Boutwell
Author content
All content in this area was uploaded by Brian B Boutwell
Content may be subject to copyright.
http://cjb.sagepub.com
Criminal Justice and Behavior
DOI: 10.1177/0093854808326992
2009; 36; 41 Criminal Justice and Behavior
Barnes
Kevin M. Beaver, J. Eagle Schutt, Brian B. Boutwell, Marie Ratchford, Kathleen Roberts and J.C.
Peer Affiliation: Results from a Longitudinal Sample of Adolescent Twins
Genetic and Environmental Influences on Levels of Self-Control and Delinquent
http://cjb.sagepub.com/cgi/content/abstract/36/1/41
The online version of this article can be found at:
Published by:
http://www.sagepublications.com
On behalf of:
International Association for Correctional and Forensic Psychology
can be found at:Criminal Justice and Behavior Additional services and information for
http://cjb.sagepub.com/cgi/alerts Email Alerts:
http://cjb.sagepub.com/subscriptions Subscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
http://cjb.sagepub.com/cgi/content/refs/36/1/41 Citations
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
GENETIC AND ENVIRONMENTAL INFLUENCES
ON LEVELS OF SELF-CONTROL AND
DELINQUENT PEER AFFILIATION
Results From a Longitudinal Sample
of Adolescent Twins
KEVIN M. BEAVER
Florida State University
J. EAGLE SHUTT
University of Louisville
BRIAN B. BOUTWELL
MARIE RATCHFORD
KATHLEEN ROBERTS
J. C. BARNES
Florida State University
Despite the fact that low self-control and exposure to delinquent peers are two of the most robust and consistent predictors
of crime, delinquency, and antisocial behavior, much remains unknown about what causes self-control to develop and what
causes youths to befriend antisocial peers. This study estimated the relative effects of environmental and genetic factors on
levels of self-control and contact with delinquent peers in a sample of twins from the National Longitudinal Study of
Adolescent Health (Add Health). DeFries-Fulker analysis of the Add Health data revealed that both self-control and contact
with drug-using friends were influenced by genetic factors and the nonshared environment, whereas the shared environment
exhibited relatively small and inconsistent effects. Implications for self-control theory and social learning theory are
discussed.
Keywords: Add Health; delinquent peers; genetics; self-control; twins
N
umerous criminological studies have tested the central propositions of self-control and
social learning theories. The results of these studies, which have analyzed hundreds of
different samples and used very different analytical strategies, have converged to reveal that
low self-control and self-regulation and exposure to delinquent peers are among the strongest
predictors of conduct disorders, aggression, and criminal and delinquent involvement (Akers
& Jensen, 2006; Morgan & Lilienfeld, 2002; Pratt & Cullen, 2000; Séguin, Boulerice,
Harden, Tremblay, & Pihl, 1999; Séguin, Nagin, Assaad, & Tremblay, 2004; Warr, 2002). At
the same time, much less empirical attention has been devoted to identifying what causes vari-
ation in levels of self-control and what causes adolescents to befriend delinquent peers.
Most criminological research examining the antecedent causes of self-control and delin-
quent peer associations has identified social and environmental factors as particularly
41
CRIMINAL JUSTICE AND BEHAVIOR, Vol. 36 No. 1, January 2009 41-60
DOI: 10.1177/0093854808326992
© 2009 International Association for Correctional and Forensic Psychology
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
important. For example, empirical work has revealed that parental socialization, neighbor-
hoods, and schools can all contribute to variation in levels of self-control (Gibbs, Giever, &
Higgins, 2003; Gibbs, Giever, & Martin, 1998; Pratt, Turner, & Piquero, 2004; Turner,
Piquero, & Pratt, 2005). Similar conclusions have been reached in regard to the correlates
to delinquent peer associations (Akers, 1998; Cairns & Cairns, 1994; Warr, 2002).
Although criminologists from a sociological perspective have, in general, dismissed the
importance of genetics (Robinson, 2004), other research has indicated that a complex
arrangement of social and genetic factors are involved in the etiology of self-control and
self-regulation (Lahey & Waldman, 2003; Nigg & Huang-Pollock, 2003; Rhee & Waldman,
2003) as well as the formation of delinquent peer groups (Beaver, Wright, & DeLisi, 2008;
DiLalla, 2002). Against this backdrop, the purpose of the current article is to examine the
relative roles of genetic and environmental factors in the development of self-control and
in exposure to delinquent peers.
THE DEVELOPMENT OF SELF-CONTROL
In their book A General Theory of Crime, Gottfredson and Hirschi (1990) argued that low
self-control was the unitary cause of crime, delinquency, and analogous behaviors. Although
this proposition formed the crux of their theory, they also advanced an explanation regard-
ing the development of self-control. According to Gottfredson and Hirschi, individual levels
of self-control are determined mainly through three different parental management tech-
niques. Specifically, parents who supervise their children, who recognize their child’s mis-
behavior, and who punish or correct their child’s misconduct will, on average, raise children
with high levels of self-control. In contrast, parents who fail to engage in these three par-
enting tactics will, on average, raise children with low levels of self-control.
While Gottfredson and Hirschi highlighted the importance of socialization effects on the
emergence of self-control, they simultaneously rejected the possibility that levels of self-
control were influenced by genetic and biological factors. In the words of Gottfredson and
Hirschi (1990), “the magnitude of the ‘genetic effect’ is near zero” (p. 60). Given that most
tests of the parental management thesis have failed to estimate genetic effects, it is difficult to
assess whether the “magnitude of the ‘genetic effect’ is near zero.” However, there are at least
three reasons to suspect that low self-control is at least partially explained by biogenic factors.
First, in a recent study, J. P. Wright and Beaver (2005) analyzed a sample of twins from
the Early Childhood Longitudinal Study–Kindergarten Class data to examine parental
socialization effects on self-control. In their first set of analyses, they did not control for
42 CRIMINAL JUSTICE AND BEHAVIOR
AUTHORS’ NOTE: The authors wish to thank the anonymous reviewers and the editor for their insightful
comments and suggestions. This research uses data from the National Longitudinal Study of Adolescent Health
(Add Health), a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris
and funded by Grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and
Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due to
Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining
data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street,
Chapel Hill, NC 27516-2524; e-mail: addhealth@unc.edu. No direct support was received from Grant P01-
HD31921 for this analysis. Correspondence concerning this article should be addressed to Kevin M. Beaver,
Florida State University, College of Criminology and Criminal Justice, 634 West Call Street, Tallahassee, FL
32306-1127; e-mail: kbeaver@fsu.edu.
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
genetic influences, and the results of their models revealed that the parenting measures had
statistically significant effects on measures of self-control. They then recalculated the
models and controlled for genetic effects. Once genetic influences were held constant, most
of the parenting measures were reduced to statistical insignificance, indicating that the rela-
tionship between parenting and self-control is accounted for by shared genetic factors. The
results of this study thus suggested that genetic influences are important in the development
of self-control.
Second, behavioral genetic research has examined the genetic and environmental influ-
ences on impulsivity, hyperactivity, attention deficit hyperactivity disorder, and other atten-
tion deficiencies. Behavioral genetic research typically decomposes variance into three
different components: a heritability component, a shared environmental component, and a
nonshared environmental component. Heritability captures the proportion of variance
accounted for by genetic factors. Shared environments refer to environments that are the
same between siblings from the same household, such as economic well-being and parental
socialization techniques. Nonshared environments, in contrast, capture any environment
that is different between siblings from the same household. It is also important to point out
that measurement error is subsumed within nonshared environmental effects. Shared envi-
ronments work to make siblings similar to each other, whereas nonshared environments
work to make siblings dissimilar to each other. Behavioral geneticists typically analyze
samples of monozygotic (MZ) and dizygotic (DZ) twins to estimate genetic, shared envi-
ronmental, and nonshared environmental effects on a broad range of behaviors, and they
have also used twin-based research designs to estimate the relative importance of genetic
factors and environmental factors in the formation of delinquent peer groups.
The results of behavioral genetic studies have been quite consistent and have revealed that
genetic factors account for between 50% and 90% of the variance in self-control, self-regulation,
and impulsivity, whereas the remaining variance is attributable to the nonshared environment
and measurement error (Barkley, 1997; Price, Simonoff, Waldman, Asherson, & Plomin, 2001;
Reiss, Neiderhiser, Hetherington, & Plomin, 2000; Rietveld, Hudziak, Bartels, van Beijsterveldt,
& Boomsma, 2003; J. P. Wright, Beaver, DeLisi, & Vaughn, 2008). Findings garnered from
these studies stand in direct contradiction to Gottfredson and Hirschi’s (1990) claim that self-
control is insulated against biogenic effects (J. P. Wright & Beaver, 2005).
Third, research has revealed that levels of self-control are influenced, in part, by the structure
and functioning of the prefrontal cortex of the brain (Barkley, 1997; Beaver, Wright, & DeLisi,
2007; Cauffman, Steinberg, & Piquero, 2005; Ishikawa & Raine, 2003; Raine, 2002). The for-
mation of the brain, as well as brain functioning, is due in large part to genetic influences
(Pfefferbaum, Sullivan, Swan, & Carmelli, 2000; Thompson et al., 2001; Toga & Thompson,
2005). Thus, variations in brain structure and functioning, which reflect genetic differences,
could translate into variations in individual levels of self-control. If this is the case, then genes
may have an effect on self-control by producing variability in regions of the brain that are tied
to impulse control, aggression, and self-regulation (Meyer-Lindenberg et al., 2006).
It is important to point out that a behavioral genetic explanation of self-control is not
necessarily incompatible with an environmental/socialization explanation of self-control.
Indeed, researchers working from a behavioral genetic perspective have argued that a com-
bination of biogenic factors and social factors act independently and interactively to pro-
duce antisocial dispositions (Caspi et al., 2002; Moffitt, 1993; Robinson, 2004; Rutter,
2006). Even so, the construct of low self-control, as defined by criminologists, has been
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 43
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
subjected to very little behavioral genetic analysis. As a result, much remains unknown
about what role, if any, genetic influences play on the development of self-control (but see
J. P. Wright et al., 2008; J. P. Wright & Beaver, 2005). The current article seeks to shed
some light on this issue.
DELINQUENT PEER GROUP FORMATION
Although a long line of literature has examined the effects that associating with antisocial
friends has on delinquent involvement (Akers, 1998; Akers & Jensen, 2006; Warr, 2002),
there is a rather limited amount of research that has explored the correlates to delinquent
peer group formation. Much of this research has investigated the role that the social context
has on adolescent contact with deviant friends. Results from this area of empirical work have
revealed that one of the most important predictors of antisocial friendship formation is prox-
imity to delinquent peers. All else equal, adolescents who are in close contact with antiso-
cial youths will be more apt to befriend delinquent others (Cairns & Cairns, 1994; Warr,
2002). Parents, too, are often hypothesized to affect their child’s choice of peer groups, either
directly through close supervision and location of residence (J. R. Harris, 1998) or indirectly
through the development of socioemotional attachment (Hirschi, 1969). Taken together, the
common theme cutting across this body of criminological research is that adolescents
become embedded in antisocial friendship networks mainly because of social factors.
An additional explanation to the formation of antisocial peer networks has been offered
by some behavioral geneticists (Cleveland, Wiebe, & Rowe, 2005; DiLalla, 2002; Scarr,
1992; Scarr & McCartney, 1983; Walsh, 2002b). According to this line of reasoning, ado-
lescents actively seek out certain friends, such as delinquent friends, because of genetic
propensities. Youths who are genetically predisposed to be impulsive, to be risk seekers,
and to be antisocial will tend to select peer groups that reinforce these propensities. As a
result, genetic influences can partially explain why individuals sort themselves into one
particular peer group over another. Behavioral geneticists refer to the close nexus between
genetic tendencies and the environment (in this case the environment is delinquent peers)
as a Gene × Environment correlation.
A Gene × Environment correlation explanation of delinquent peer associations is very sim-
ilar to self-selection arguments. Advocates of a self-selection perspective (e.g., Gottfredson &
Hirschi, 1990) argue that adolescents choose to associate with antisocial friends because of
some underlying propensity (e.g., low self-control). Advocates of a Gene × Environment cor-
relation perspective use similar logic, except that instead of focusing on a particular trait they
focus on genetic effects. The difference, then, is the unit of analysis. But it should be pointed
out that behavioral geneticists do not think that there is a gene for delinquent peers. Instead,
behavioral geneticists take a more realistic and nuanced approach and argue that personality
traits (e.g., low self-control) are heavily influenced by genetic factors (J. P. Wright & Beaver,
2005). Thus, whereas sociologically oriented criminologists explore self-selection effects by
examining personality factors and underlying propensities (Baron, 2003; Chapple, 2005;
Gottfredson & Hirschi, 1990; B. R. E. Wright, Caspi, Moffitt, & Silva, 1999), behavioral
geneticists reduce the unit of analysis to the genetic level (Cleveland et al., 2005).
To illustrate, Iervolino et al. (2002) analyzed data from the Nonshared Environment in
Adolescent Development (NEAD) research and from the Colorado Adoption Project (CAP)
to estimate the relative importance of genes and the environment on deviant peer affiliations.
44 CRIMINAL JUSTICE AND BEHAVIOR
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
The results of the biometric model-fitting techniques revealed two somewhat contradictory
findings. For the NEAD sample, Iervolino et al. found that almost all of the variance in the
measure of peer delinquency was accounted for by the shared environment (20%) and the
nonshared environment (77%), whereas genetic effects were close to zero (3%). Remarkably
dissimilar results were garnered when the CAP data were analyzed. For this analysis, genetic
factors accounted for 65% of the variance in peer delinquency, the nonshared environment
explained 35% of the variance, and the shared environment explained none of the variance.
In a similar vein, Cleveland and his colleagues (2005) analyzed a sample of sibling pairs
from the National Longitudinal Study of Adolescent Health (Add Health) to estimate
genetic and environmental effects on substance-using friends. Analysis of the data revealed
that 64% of the variance in delinquent peer affiliations was accounted for by genetic fac-
tors, 36% of the variance was attributable to the nonshared environment, and the shared
environment explained 0% of the variance. In the most recent study, Kendler et al. (2007)
analyzed a sample of male twin pairs from the Virginia Twin Registry and found that
genetic factors account for between 30% and 50% of the variance in peer-group deviance.
Most of the remaining variance was due to nonshared environmental factors. Collectively,
the results of these studies provide mixed evidence on the extent to which genetic factors
influence the formation of antisocial friendship networks. These divergent results are per-
haps due to differences in sample characteristics, the operationalization of delinquent peers,
or some other unidentified factor or factors. Regardless, these studies implicate genetic
factors—to varying degrees—in the befriending of antisocial peers.
THE CURRENT STUDY
The goal of the current study was to determine the relative effects of genetic and envi-
ronmental factors in the development of self-control and in delinquent peer associations. It
is important to note that prior research by Beaver (2008) and J. P. Wright et al. (2008) also
used twins drawn from the Add Health data to examine issues related to the development of
self-control and the formation of antisocial peer groups. However, their studies differ from
the present one in three ways. First, Beaver examined difference scores only in the sample
of MZ twins to estimate the effects that specific nonshared environments had on differences
in low self-control and delinquent peers. His research did not estimate the heritability of
these outcomes. Second, J. P. Wright and his colleagues examined the genetic and environ-
mental effects on low self-control and delinquent peers, but they used a different modeling
strategy than the one used in the present study. The different modeling strategy utilized
herein helps determine whether the findings are robust or whether they may be explained
away as statistical artifacts. Third, and unlike these other two studies, our research simulta-
neously estimates the effects of genetic and shared environmental factors and then estimates
the effects of specific nonshared environments in a sample of MZ and DZ twins.
METHOD
DATA
Data for this study come from the Add Health (Udry, 2003). The Add Health used a school-
based research design to select a nationally representative sample of adolescents enrolled in
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 45
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
7th through 12th grade. Using stratified sampling techniques, a total of 80 high schools and
52 middle schools were included in the study. On a specified day in 1994, all students attend-
ing these schools were administered a self-report survey that asked a variety of questions
about their lives, their daily activities, and their peer and familial relationships. Overall, more
than 90,000 adolescents submitted completed questionnaires (K. M. Harris et al., 2003).
A subsample of adolescents was then chosen to be interviewed in their homes along with
their primary caregiver (usually their mother). The in-home questionnaires included more
sensitive items that tapped the adolescents’ involvement in delinquent and criminal behav-
iors, their involvement in risky behaviors, and the quality of their relationships with their
parents and friends, among others (K. M. Harris et al., 2003). Questions were also asked
that measured certain dimensions of the respondents’ temperament, such as whether they
have a temper, whether they are impulsive, and whether they are risk seekers. In addition,
their primary caregivers also provided detailed information about various aspects of the
adolescent’s life. Altogether, 20,745 adolescents and 17,700 of their primary caregivers
took part in the Wave 1 in-home survey (K. M. Harris et al., 2003).
Approximately 1 to 2 years after the Wave 1 in-home survey, respondents were adminis-
tered a second wave of questionnaires. Given that relatively little time had lapsed since Wave
1, and given that most of the respondents were still adolescents, many of the same questions
asked at Wave 1 were retained on the Wave 2 surveys. For example, adolescents were asked
about their relationships with family and friends, their involvement in delinquency, and their
involvement in sexual behaviors. Interviews were completed with 14,738 respondents at
Wave 2, for a response rate of about 71%. Nearly 7 years after the first wave of data was col-
lected, the third round of interviews was conducted. Most of the participants were between
the ages of 18 and 26 years old; thus, the questionnaires had to be redesigned to include more
age-appropriate items. The surveys now included questions that were geared toward indexing
the respondents’ childbearing history, their marital status, and their lifetime contact with the
criminal justice system. At Wave 3, 15,197 of the original Wave 1 participants were reinter-
viewed, which corresponds to a response rate of about 73% (K. M. Harris et al., 2003).
Embedded within the entire sample of adolescents is a subsample of sibling pairs. During
Wave 1 interviews, respondents were asked to indicate whether they were part of a twin pair.
If the adolescent indicated he or she was a twin, then the co-twin was added to the sample.
In addition, half siblings, genetically unrelated siblings (e.g., stepsiblings), and cousins were
oversampled for inclusion in the study. A probability sample of full siblings was also
retained in the sample. Subsequent analyses testing for potential selection biases in the
sibling-pairs sample did not reveal any significant differences in the characteristics between
the nationally representative sample and the sample of siblings (Jacobson & Rowe, 1998).
To calculate conservative parameter estimates, we included only MZ twin pairs and
same-sex DZ twin pairs. Because the items assessing delinquent peer associations were not
collected at Wave 3, and because the low self-control items were changed at Wave 3, our
analysis was restricted to the first two waves of data. With these criteria in place, and once
missing cases were removed and after deleting twins whose zygosity was undetermined, we
were left with a final analytical sample size that ranged between N = 662 and N = 914 twins.
DEPENDENT VARIABLES
Low self-control. Even though Gottfredson and Hirschi’s (1990) general theory has been
one of the most empirically scrutinized theories in recent years, there is still a considerable
46 CRIMINAL JUSTICE AND BEHAVIOR
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
amount of disagreement concerning the most reliable and valid way to measure self-
control (DeLisi, Hochstetler, & Murphy, 2003; Longshore, Stein, & Turner, 1998;
Longshore, Turner, & Stein, 1996; Marcus, 2003, 2004; Piquero & Rosay, 1998). Although
Gottfredson and Hirschi are quite clear that people with low levels of self-control are risk
seekers, are impulsive, are self-centered, prefer physical activities to mental ones, prefer
simple tasks, and have a temper, there is wide variability in the scales that have been used
to tap self-control. Perhaps the most frequently used measurement strategy is the scale
developed by Grasmick, Tittle, Bursik, and Arneklev (1993). Unfortunately, the Add Health
surveys did not include items that could be used to construct the Grasmick et al. scale.
There are, however, a series of items available in the Add Health data that researchers
have used to create a composite measure of self-control. Respondents were asked questions
that tapped whether they have trouble paying attention, problems finishing their homework,
troubled relationships with their teachers, and a difficult time staying focused. Perrone,
Sullivan, Pratt, and Margaryan (2004) argued that “these questions tap into the simple tasks,
physical activities, and impulsivity components of self-control” (p. 302). Last, respondents
were asked whether they felt they do everything just right. This item indexed the self-
centeredness dimension of self-control (Perrone et al., 2004). The exact same items were
available at Wave 1 and Wave 2. Responses to each of the items were summed together to
form the Wave 1 Low Self-Control scale (alpha = .65) and the Wave 2 Low Self-Control
scale (alpha = .62). Higher values on both scales reflected lower levels of self-control.
We calculated a series of additional statistical specifications to examine the psychometric
properties of the low self-control scales. We conducted principal components analysis, and
only one component was present. Structural equation models were also calculated to deter-
mine whether the observable indicators all loaded on the same factor. Again, the results sug-
gested that all of the items were significant indicators of one unobservable construct. In
addition, we examined the predictive validity of the self-control scales. To do so, we created
a Wave 1 and a Wave 2 Delinquency scale. The correlations between the Wave 1 Low Self-
Control scale and the two delinquency scales (r = .29, p < .05 for the Wave 1 Delinquency
scale; r = .24, p < .05 for the Wave 2 Delinquency scale) were statistically significant, and
the correlation between the Wave 2 Low Self-Control scale and the Wave 2 Delinquency
scale (r = .32, p < .05) was statistically significant. The effect sizes for the two self-control
scales were very similar to those reported in Pratt and Cullen’s (2000) meta-analysis. Finally,
the two self-control scales were correlated, r = .50, p < .05, suggesting that the two scales
have wave-to-wave consistency and are measuring the same latent construct over time.
Taken together, there is reason to believe that the low self-control scales available in the Add
Health study are valid and reliable measures of individual variation in self-control.
Drug-using peers. The Add Health data contain a number of questions that index each
respondent’s contact with, and exposure to, drug-using peers. At Wave 1 and Wave 2 inter-
views, respondents were asked three questions about their friends’ drug and alcohol use.
Specifically, adolescents were asked how many of their three closest friends smoked at least
one cigarette each day, smoked marijuana at least once a month, and got drunk at least once
per month. Responses to each question were coded as follows: 0 = 0 friends,1 = 1 friend,
2 = 2 friends, and 3 = 3 friends. The three items were then summed together to create the
Wave 1 Drug-Using Peers scale (alpha = .76) and the Wave 2 Drug-Using Peers scale
(alpha = .77). Higher scores indicated more contact with drug-using peers.
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 47
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
Similar to the low self-control scales, we also subjected the drug-using peers scales to
additional statistical testing to examine the psychometric properties of the scales. First, we
calculated principal components analyses, and the results revealed that the covariance
structure of the three items was accounted for by a single component. Second, structural
equation models were estimated, and once again the results confirmed the presence of one
latent factor. Third, and in line with prior research examining stability in drug-using friends
over time (Warr, 1993), the wave-to-wave stability in drug-using peers was high (r = .60,
p < .05). Finally, it should also be pointed out that researchers analyzing the Add Health
data have used identical scales and found them to be some of the strongest correlates to
delinquent involvement (Beaver & Wright, 2005a; Bellair, Roscigno, & McNulty, 2003).
SOCIALIZATION VARIABLES
Maternal disengagement. Adolescents who have cold and withdrawn parents are at risk for
engaging in delinquency, using drugs and alcohol, and affiliating with delinquent peers
(Gottfredson & Hirschi, 1990; Loeber & Stouthamer-Loeber, 1986; Patterson, 1982). To take
these findings into account, we created a Maternal Disengagement scale. During Wave 1 inter-
views, adolescents were asked five different questions that measured maternal disengagement.
For example, respondents were asked whether they are satisfied with their relationship with
their mother, whether their mother is warm and loving toward them, and whether they are sat-
isfied with the way their mother communicates with them. Responses to the items were added
together, and higher scores indicated more maternal disengagement (alpha = .83).
Maternal attachment. Research has indicated that parents who are emotionally con-
nected and attached to their children are more apt to raise prosocial offspring (Sampson &
Laub, 1993). In line with previous researchers analyzing the Add Health data (Haynie,
2001; Schreck, Fisher, & Miller, 2004), we included a two-item measure of maternal
attachment. During Wave 1 interviews, adolescents were asked how much they thought
their mother cares about them and how close they felt to their mother. Responses to these
two questions were added to create the Maternal Attachment scale (alpha = .70). Higher
scores indicated greater maternal attachment.
Maternal involvement. Parents who are heavily involved in their children’s lives act as a
protective factor against antisocial outcomes (Loeber & Stouthamer-Loeber, 1986). As a
result, and similar to the work of Crosnoe and Elder (2004), we developed a 10-item measure
of maternal involvement from Wave 1 interviews with the adolescent. Specifically, partici-
pants were presented with a series of activities and asked to indicate which ones they had
taken part in with their mother in the past 4 weeks. Items that were endorsed were assigned
a value of 1; otherwise, they were coded 0. The items were then summed to form the Maternal
Involvement index, where higher scores indicated more maternal involvement (alpha = .53).
Parental permissiveness. Parents who fail to monitor and supervise their children or who
fail to set appropriate boundaries for their children are more likely to raise disruptive and
troublesome offspring (Gottfredson & Hirschi, 1990; Loeber & Stouthamer-Loeber, 1986).
The Add Health data contain seven different items that index parental permissiveness.
During Wave 1 interviews, adolescents were asked whether their parents let them make
their own decisions about their bedtimes, curfews, friends, clothes, and what they eat. All
48 CRIMINAL JUSTICE AND BEHAVIOR
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
of the items were coded dichotomously (0 = no,1 = yes). Responses to the questions were
summed to create the Parental Permissiveness scale, where higher scores reflected more
parental permissiveness (alpha = .65).
CONTROL VARIABLES
Age. Levels of self-control are age graded, and contact with drug-using peers varies as a
function of age (Gottfredson & Hirschi, 1990; Warr, 1993). As a result, and to help prevent
misspecification of the statistical models, we included age as a continuous variable mea-
sured in years.
Gender. To control for potential gender differences in levels of self-control and exposure
to drug-using friends, gender was included as a dichotomous dummy variable in all of the
statistical models (0 = female,1 = male).
Race. We included race as a control variable to account for potential racial differences
in self-control and drug-using peers (0 = White,1 = non-White).
ANALYSIS
To estimate genetic and environmental effects on low self-control and on drug-using
peers, we use DeFries-Fulker (DF) analysis. DF analysis was originally developed by
DeFries and Fulker (1985) to be used when one twin had an extreme score on some mea-
sure (e.g., having severe language deficits). The original DF formula, however, has been
adjusted and extended, and now it can be used with samples drawn from the general pop-
ulation (Rodgers & Kohler, 2005; Rodgers, Rowe, & Li, 1994). The newly amended DF
analysis, referred to as the “augmented” DF model, has been used quite frequently by a
range of researchers from different disciplines including psychology, behavioral genetics,
and criminology (Cyphers, Phillips, Fulker, & Mrazek, 1990; Haynie & McHugh, 2003;
McCartan, 2007; Rodgers, Buster, & Rowe, 2001; Rodgers et al., 1994).
DF analysis is a regression-based analytical approach designed to analyze data consist-
ing of sibling pairs. The basic DF equation takes the following form:
K
1
= b
0
+ b
1
K
2
+ b
2
R + b
3
(R
*
K
2
) + e, (1)
where K
1
is the score for one twin on a particular measure, K
2
is his or her co-twin’s score
on that same measure, R is the coefficient of genetic relatedness (R = 1.0 for MZ twins, R =
.5 for DZ twins), and R
*
K
2
is the multiplicative interaction term between R and K
2
. For
this equation, b
0
is the constant, b
1
is the proportion of variance in K
1
explained by shared
environmental influences (c
2
), b
2
is not usually interpreted in the DF model, and b
3
is the
proportion of variance in K
1
explained by genetic effects (h
2
). The effects of the nonshared
environment (e
2
) and measurement error are captured by the error term, e.
Recently, Rodgers and Kohler (2005) reformulated the DF model presented in Equation
1 to help overcome some inconsistencies inherently built into the DF equation. Their newly
transformed DF equation is
K
1
= b
0
+ b
1
(K
2
− K
m
) + b
2
(R
*
(K
2
− K
m
)) + e, (2)
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 49
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
where K
1
is still the score for one twin on a particular measure, K
2
is still the co-twin’s score
on that same measure, and R is still the coefficient of genetic relatedness. The main differ-
ence is that in this equation the term K
m
is introduced into the model. K
m
is the mean of the
measure for K
2
. In contrast to Equation 1, K
2
is being centered on its mean, whereas in
Equation 1 it was left untransformed. The main effect of R is also removed in Equation 2.
The interpretation of the coefficients, however, remains unchanged. For example, b
1
is the
shared environmental effects and b
2
is the genetic effects. The unexplained variance is
attributable to measurement error and to nonshared environmental effects.
1
Equation 2 effectively removes the variance in K
1
that is accounted for by the shared envi-
ronment and by genetic factors. The terms b
1
and b
2
are in many ways like latent factors
because it is not possible to determine which shared environments are important or which
genetic factors are important. Although some researchers have introduced shared environ-
mental variables into DF models to try to identify salient shared environments, this proce-
dure is probably unadvisable because, as Turkheimer, D’Onofrio, Maes, and Eaves (2005)
argue, “by adding the shared family-level covariate to this model [i.e., the DF model], one
is attempting to predict shared twin variability in a model that already has a term specifically
designed to account for all of it” (p. 1224). However, it is possible to introduce measures of
the nonshared environment into the equation in order to determine their effects on K
1
once
the effects of the shared environment and genetic influences are held constant.
Rodgers et al. (1994) presented an expanded DF equation to demonstrate how to include
nonshared environmental effects. This expanded DF equation can be used in conjunction
with Equation 2 to yield the following formula:
K
1
= b
0
+ b
1
(K
2
− K
m
) + b
2
(R
*
(K
2
− K
m
)) + b
3
ENVDIF + e. (3)
This equation introduces one additional term, ENVDIF. ENVDIF is the difference score on
a particular environmental measure between two twins from the same twin pair. For
example, Twin 2’s score on the Maternal Disengagement scale can be subtracted from Twin
1’s score on the Maternal Disengagement scale to produce a difference score in maternal
disengagement. In the above equation, this difference score could be entered as the
ENVDIF variable. Essentially, ENVDIF captures differences between twins and uses those
difference scores to tap the nonshared environment. Although Equation 3 shows the inclu-
sion of only one ENVDIF, more than one ENVDIF variable can be entered into the model
without biasing the coefficients. All of the remaining regression coefficients can be inter-
preted the same way that they were in Equation 2.
2
Although it might seem that nonshared environmental effects are detected if the
ENVDIF coefficient is statistically significant, such a conclusion would be too hasty
because the possibility exists that the ENVDIF coefficient may be estimating nonshared
genetic influences rather than nonshared environmental effects. To determine whether a
significant ENVDIF coefficient is being driven by the nonshared environment or by non-
shared genetic effects, an additional term needs to be introduced into Equation 3. The inclu-
sion of this term results in the following equation:
K
1
= b
0
+ b
1
(K
2
− K
m
) + b
2
(R
*
(K
2
− K
m
)) + b
3
ENVDIF + b
4
(ENVDIF
*
R) + e. (4)
The only difference between this equation and Equation 3 is the inclusion of the interaction
term, ENVDIF
*
R. If the coefficient for ENVDIF
*
R is significant, then the effect of
50 CRIMINAL JUSTICE AND BEHAVIOR
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
ENVDIF (b
3
) is attributable to nonshared genetic effects. On the other hand, if the ENVDIF
*
R coefficient is not statistically significant, then the effect of ENVDIF (b
3
) is attributable
to nonshared environmental effects. It is important to note that Equation 4 is estimated only
if the ENVDIF variable is statistically significant in Equation 3.
Researchers using DF analysis are confronted with the decision of which twin from each
twin pair should be selected to be used as the dependent variable (K
1
) and which twin
should be selected to be used as the independent variable (K
2
). This problem is often
resolved by “double entering” the twins, where each twin is entered into the data twice:
One time the twin is the dependent variable and the co-twin is the independent variable, and
one time the twin is the independent variable and the co-twin is the dependent variable.
Indeed, this is the most common approach when using the augmented DF equations, and
we use double entry in the current study (Haynie & McHugh, 2003; Kohler & Rodgers,
2001; Rodgers et al., 2001; Rodgers & Rowe, 1987). Although double entering twins
is advantageous, it also violates one of the main assumptions of ordinary least squares
regression—namely, the observations are no longer independent from each other. Although
nonindependence does not bias the regression slopes, it does deflate the standard errors and
thus biases tests of statistical significance for the coefficients (Hanushek & Jackson, 1977).
To take the clustering of observations into account, all of the standard errors were calcu-
lated using Huber/White variance estimators.
DF analyses were calculated for the Wave 1 and Wave 2 Low Self-Control scales and for
the Wave 1 and Wave 2 Drug-Using Peers scales. All four of the parental socialization
scales were transformed into difference scores by subtracting the two twins’ scores on each
measure. These difference scores were then entered into the DF equation as measures of the
nonshared environment. Any nonshared environmental measures that were statistically sig-
nificant using Equation 3 were then reexamined using Equation 4 to determine whether
nonshared genetic factors or nonshared environment influences were driving this effect.
RESULTS
We begin our analyses by examining cross-twin correlations for the low self-control
scales and the delinquent peers scales. Zero-order cross-twin correlations allow for an ini-
tial estimation of whether genetic influences are potentially important to self-control and to
associating with drug-using peers. In general, if the MZ cross-twin correlation is signifi-
cantly larger than the DZ cross-twin correlation, then there is at least a minimal genetic
effect on the measure. Table 1 reveals the cross-twin correlations for the entire sample of
twins and separately for MZ and DZ twins. As can be seen, all of the cross-twin correla-
tions were statistically significant, but the MZ cross-twin correlations were all larger than
the DZ cross-twin correlations. These findings provide initial evidence that low self-
control and contact with drug-using peers may be at least partially genetically influenced.
However, to provide a more rigorous examination of the relative effects of environmental
influences and genetic forces on the self-control and drug-using peers scales, we next turn
to the results generated from DF analysis.
Table 2 presents the results of the DF equations using the low self-control scales as the
dependent variable. Model 1 displays the results of a baseline DF model that estimated
genetic and shared environmental effects on the Wave 1 Low Self-Control scale with race,
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 51
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
age, and gender introduced as control variables. Remember the slope of (R
*
(K
2
− K
m
)) is
interpreted as the proportion of variance accounted for by genetic factors (i.e., heritability)
and the slope of (K
2
− K
m
) is interpreted as the proportion of variance accounted for by the
shared environment. To facilitate readability, these two terms have been labeled as “heri-
tability” and “shared environment” in Table 2. The results of this model revealed that
genetic influences accounted for 56% of the variance in low self-control. The coefficient
for the shared environment was negative, but this does not mean that the shared environ-
ment explained a negative amount of variance in low self-control. Instead, because the
slope for the shared environment is not statistically significant, the results simply mean that
the confidence interval for the slope of the shared environment includes the value of 0.
When a coefficient is not significantly different from 0, the sign and the value of the coef-
ficient should be ignored and subsequently equated with 0. In this case, the slope of the
shared environment (b =−.16, p > .05) should really be interpreted as b = 0. Applying this
logic to the current analyses means that the proportion of variance in self-control that was
accounted for by the shared environment is 0. The remaining variance (44%) in low self-
control was explained by the nonshared environment and measurement error.
52 CRIMINAL JUSTICE AND BEHAVIOR
TABLE 1: Cross-Twin Correlations for the Low Self-Control Scales and the Delinquent Peers Scales
All Twins MZ Twins DZ Twins
Low Self-Control (Wave 1) .290* .406* .122*
Low Self-Control (Wave 2) .232* .316* .118*
Drug-Using Peers (Wave 1) .588* .666* .435*
Drug-Using Peers (Wave 2) .551* .674* .416*
Note.
MZ = monozygotic; DZ = dizygotic.
*
p
< .05, two tailed.
TABLE 2: DeFries-Fulker (DF) Analysis of Low Self-Control at Waves 1 and 2
Low Self-Control at Wave 1 Low Self-Control at Wave 2
Model 1 Model 2 Model 3 Model 4
b SE b SE b SE b SE
DF analysis components
Heritability .56* .13 .52* .14 .40* .15 .44* .15
Shared environment –.16 .11 –.11 .12 –.09 .12 –.09 .12
Nonshared environments
Maternal disengagement .14* .03 .11* .03
Maternal attachment –.03 .08 .12 .10
Maternal involvement .05 .05 .09 .05
Parental permissiveness .05 .06 .14* .06
Control variables
Age .01 .06 .00 .06 –.03 .07 –.03 .07
Gender .38 .20 .46* .20 .19 .21 .21 .21
Race –.08 .20 –.05 .21 –.27 .22 –.18 .22
Note.
Huber/White standard errors are presented.
*
p
< .05, two tailed.
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
Recall that if the nonshared environment accounts for a significant amount of variance,
then it is possible to estimate the effects of specific nonshared environments. We next turn
our attention to Model 2, which introduces four different nonshared environmental vari-
ables into the DF equation. The reader is reminded that the nonshared environmental vari-
ables were constructed by calculating difference scores. Three findings warrant attention in
this model. First, genetic influences accounted for more than one half (52%) of the vari-
ance in low self-control. Second, the shared environment accounted for none (0%) of the
variance in low self-control. Third, of the four measures of the nonshared environment,
only maternal disengagement had a statistically significant effect on low self-control.
3
The
association between maternal disengagement and low self-control was positive, meaning
that the twin who received more maternal disengagement had lower levels of self-control.
This effect was observed even after holding genetic effects constant.
The next set of analyses is identical to those previously calculated, except that the Wave
2 Low Self-Control scale was used as the dependent variable. As shown in Model 3, genetic
effects accounted for 40% of the variance in low self-control, the shared environment had
no effect on low self-control (0% of the variance), and the nonshared environment and mea-
surement error accounted for the remaining 60% of the variance.
In Model 4, the four nonshared environmental variables (all measured at Wave 1) were
used to predict variance in the Wave 2 Low Self-Control scale. After partitioning out the
effects of genetic influences, two nonshared environmental variables were statistically sig-
nificant: maternal disengagement and parental permissiveness. The association for both of
these variables was positive, meaning that the twin who received more maternal disen-
gagement or more parental permissiveness had lower levels of self-control. It is important
to note that in Model 4, genetic effects accounted for 44% of the variance in low self-
control, whereas the remaining 56% of variance was attributable to the nonshared environ-
ment and measurement error. Shared environmental effects were zero. Taken together, the
DF models calculated for the low self-control scales indicated that genetic and nonshared
environmental effects accounted for 100% of the variance, whereas the shared environment
had no effect on individual levels of self-control.
The last set of analyses used the drug-using peers scales as dependent variables in the
DF equations controlling for the effects of age, gender, and race. Model 1 of Table 3 con-
tains the findings for the baseline DF model where the Wave 1 Drug-Using Peers scale is
the outcome measure. Results gleaned from this model indicated the genetic effects
accounted for 37% of the variance, the shared environment accounted for 27% of the vari-
ance, and the nonshared environment and measurement error accounted for the remaining
36% of variance. As with the low self-control scales, the environment and genes both con-
tributed to the explained variance in drug-using peers; however, and in contrast to the low
self-control models, the shared environment explained a statistically significant amount of
variance (27%).
Because a significant portion of the variance in the Wave 1 Drug-Using Peers scale was
accounted for by the nonshared environment, we also calculated an additional equation
where the four nonshared environmental variables were entered into the equation. As shown
in Model 2, the introduction of the nonshared environmental variables increased the per-
centage of explained variance attributable to genetic effects to 40%, whereas the percentage
of variance attributable to the shared environment decreased to 22%. Most important,
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 53
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
however, was that only one of the nonshared environmental variables, maternal attachment,
was significantly associated with the Wave 1 Drug-Using Peers scale. The effect of the
maternal attachment variable was negative, meaning that the twin who received more mater-
nal attachment had fewer drug-using peers. The remaining three nonshared environmental
variables were not related to the Drug-Using Peers scale in the DF equation.
Models 3 and 4 of Table 3 were estimated using the Wave 2 Drug-Using Peers scale as
the dependent variable. The results presented in Model 3 revealed that 53% of the variance
in the Drug-Using Peers scale was accounted for by genetic factors, 0% of the variance was
attributable to the shared environment, and 47% of the variance was accounted for by the
nonshared environment and measurement error.
Last, we entered the four nonshared environmental variables into the DF equation to
determine whether they were associated with the Wave 2 Drug-Using Peers scale. In this
model, 62% of the variance in the Drug-Using Peers scale was due to genetic influences,
0% of the variance was due to the shared environment, and 38% of the variance was attrib-
utable to the nonshared environment. Of particular importance are the results for the non-
shared environmental variables, which indicated that none of the four variables was
significantly associated with the Wave 2 Drug-Using Peers scale. The results presented in
Table 3 revealed the importance of both genetic and environmental effects on the two drug-
using peers scales.
DISCUSSION
Social learning theory and low self-control theory are two of the most empirically sup-
ported criminological theories (Pratt & Cullen, 2000). Extant empirical tests of these per-
spectives have focused on linking low levels of self-control to delinquent involvement and
linking contact with delinquent peers to antisocial behaviors. Comparatively less research
54 CRIMINAL JUSTICE AND BEHAVIOR
TABLE 3: DeFries-Fulker (DF) Analysis of Drug-Using Peers at Waves 1 and 2
Drug-Using Peers at Wave 1 Drug-Using Peers at Wave 2
Model 1 Model 2 Model 3 Model 4
b SE b SE b SE b SE
DF analysis components
Heritability .37* .12 .40* .13 .53* .12 .62* .13
Shared environment .27* .10 .22* .11 .12 .11 .02 .11
Nonshared environments
Maternal disengagement .03 .02 .03 .02
Maternal attachment –.16* .07 –.09 .06
Maternal involvement .04 .04 .02 .04
Parental permissiveness –.01 .04 .01 .05
Control variables
Age .18* .04 .20* .05 .14* .05 .18* .05
Gender .12 .15 .13 .15 .28 .16 .37* .17
Race –.30 .16 –.37* .16 –.38* .17 –.39* .18
Note.
Huber/White standard errors are presented.
*
p
< .05, two tailed.
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
has attempted to identify the causes of low self-control and the forces that lead youths to
befriend delinquent peers. The current article addressed this gap in the literature by exam-
ining whether environmental and genetic factors are associated with individual levels of
self-control and with the formation of drug-using friendship networks. To address these
issues, a sample of MZ and DZ twins from the Add Health study was analyzed by using DF
regression techniques. The results of the DF models revealed three broad findings.
First, once genetic effects were held constant, the shared environment had no effect on
either of the two low self-control scales and no effect on the Wave 2 Drug-Using Peers
scale. Indeed, the shared environment had a significant effect on only the Wave 1 Drug-
Using Peers scale, where it accounted for between 22% and 27% of the variance. Second,
and relatedly, across all of the models, the nonshared environment accounted for a large
proportion of variance in the dependent variables. However, the specific measures of the
nonshared environment—that is, the ones that were tapping differential parental treatment—
had small and relatively inconsistent effects. The one exception was maternal disengage-
ment, which was significantly associated with the Wave 1 and Wave 2 Low Self-Control
scales. However, the only other two nonshared environmental measures that were significant
were parental permissiveness and maternal attachment, both of which were statistically
significant in only one of the four equations.
Third, the Wave 1 and Wave 2 Low Self-Control scales and the Wave 1 and Wave 2
Drug-Using Peers scales were under substantial genetic influence. Genetic factors
accounted for between 40% and 56% of the variance in low self-control and between 37%
and 62% of the variance in drug-using peers. Taken together, analysis of the Add Health
data suggests that contact with drug-using peers and levels of self-control are strongly
affected by the nonshared environment and genetic factors, whereas the shared environ-
ment is relatively unimportant.
The results garnered from the DF models have important implications for Gottfredson
and Hirschi’s (1990) parental management thesis and for social learning theory. Recall that
Gottfredson and Hirschi argued that levels of self-control were determined largely by the
way in which parents socialize their children. Analysis of the Add Health data revealed a
very different picture—namely, that once genetic effects were removed, the shared envi-
ronment did not exert any effect on the development of self-control. We also found very
limited evidence that differential parental treatment (i.e., the nonshared environment) was
associated with low self-control. The skeptical reader will be quick to point out that there
is a body of research that has revealed some evidence linking parental socialization to indi-
vidual levels of self-control (Gibbs et al., 1998; Gibbs et al., 2003; Polakowski, 1994;
Unnever, Cullen, & Pratt, 2003). Unfortunately, none of these studies has controlled for
genetic influences, raising the very real possibility that the models are misspecified and the
findings are spurious (J. P. Wright & Beaver, 2005).
Our findings also add to an emerging body of empirical research revealing that bio-
genetic factors are a cause—perhaps even the dominant cause—of problems with self-
control and self-regulation (Beaver & Wright, 2005b; Beaver et al., 2007; Tremblay, 2003;
J. P. Wright & Beaver, 2005). We are, of course, not saying that the environment is unim-
portant. Rather, we are pointing to the likelihood that the environments that are most likely
to be linked to self-control are rarely studied by criminologists. For example, there is now
convincing research revealing that prenatal exposure to toxins, such as drugs, alcohol, nico-
tine, and lead, interferes with normal brain development (Karr-Morse & Wiley, 1997; Sood
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 55
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
et al., 2001; Yolton, Dietrich, Auinger, Lanphear, & Hornung, 2005). Given that levels of
self-control are closely tied to the structure and functioning of the brain, any event or
pathogen that may lead to neurological deficits may also reduce levels of self-control
(Barkley, 1997; Beaver & Wright, 2005b; Beaver et al., 2007). Future research should
begin to unravel the complex ways that these nonshared environmental effects may ulti-
mately lead to maladaptive behaviors and antisocial traits.
Our findings also partially stand in direct contrast to arguments that parents are able to
influence and affect their child’s choice of peer groups. As our findings revealed, most of the
variance in exposure to drug-using peers was accounted for by genetic factors. Behavioral
geneticists have long theorized that individuals create their own environments based in large
part on genotype—a phenomenon that is referred to as a Gene × Environment correlation
(Scarr, 1992; Scarr & McCartney, 1983). The logic of Gene × Environment correlations can
easily be applied to delinquent peer group formation (Cleveland et al., 2005). For example,
individuals with a genetic propensity to engage in delinquency, or to use illegal substances,
are likely to seek out peers with these same preferences. In this case, genes are the driving
force behind why some youths associate with antisocial friends whereas other adolescents
associate with prosocial peers. From a theoretical standpoint, these findings would be in line
with a self-selection argument and would cast doubt on a purely social causation explana-
tion to the delinquent peers–delinquent involvement nexus.
Before concluding, it is important to touch on three limitations of the analysis. First, there
were only a limited number of parenting measures available in the Add Health, which ham-
pered our ability to provide a more stringent test of Gottfredson and Hirschi’s (1990) parental
socialization thesis. For example, there were no questions that measured parental recognition
of deviance or parental punishment. This problem, however, is not unique to the current study;
many prior studies examining the effect parents have on the development of self-control did
not include all of the parenting dimensions outlined by Gottfredson and Hirschi (Unnever
et al., 2003; J. P. Wright & Beaver, 2005). Second, the Add Health data did not contain items
that could be used to construct Grasmick et al.’s (1993) Low Self-Control scale. Even so, we
were able to reconstruct the Low Self-Control scale used by Perrone and her colleagues
(2004). Third, the analysis was based on a sample of MZ and DZ twins, leaving open the pos-
sibility that the findings reported here may not generalize to the larger population.
Nonetheless, the results of our study provide a serious challenge to some of the central
tenets from low self-control theory and from social learning theory. But does this mean that
we should abandon these theories and start afresh? Of course not, nor are we advocating
such a radical position. Instead, we are in agreement with Walsh (2000, 2002a), who has
recently argued that many of the dominant criminological theories should be revised to
incorporate findings from biology and genetics. Not only would this type of theoretical
integration link criminology with other disciplines, but also it would provide a much more
complete explanation of crime, delinquency, and antisocial behavior.
NOTES
1. In describing the use of Equation 2, Rodgers and Kohler (2005) suggest that “in cases where trait means significantly
differ across kinship categories (e.g., across MZ [monozygotic] and DZ [dizygotic] twins), K
m
should be calculated separately
by kinship category, and the reformulated DeFries-Fulker [DF] analysis presented here [i.e., Equation 2] based on centered
scores should use category-specific means” (p. 212). We calculated independent-samples t tests to examine whether there
were mean differences between MZ and DZ twins on the low self-control scales and the drug-using peers scales. The results
56 CRIMINAL JUSTICE AND BEHAVIOR
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
of the t tests did not reveal any average differences in the dependent variables between MZ and DZ twins. As a result, in all
of our analyses K
m
is the average combined score for MZ and DZ twins.
2. Model misspecification is not a problem with DF analysis. That is, the omission of variables will not bias the results
because all shared environmental effects (e.g., neighborhood conditions and family influences) are captured by the shared
environmental term (b
1
) and the omission of nonshared environmental effects is captured by the error term (e). Including par-
ticular measures of the nonshared environment helps to remove some of the unexplained variance, but there is no reason to
include measures of the shared environment. Thus, the lack of control variables in the models will not affect the substantive
findings. We do, however, include control variables for age, gender, and race. All of the models were recalculated without
these control variables, and the results remained virtually identical.
3. It is important to note that for all models estimated, any time that a nonshared environmental variable was statistically
significant, we reestimated the models using Equation 4. Recall that Equation 4 is calculated to determine whether nonshared
environmental effects are really capturing nonshared genetic effects. The results of the models revealed that the effects were
due to the nonshared environment, not to genetic differences. Therefore, we do not present the results garnered from using
Equation 4.
REFERENCES
Akers, R. L. (1998). Social learning and social structure: A general theory of crime and deviance. Boston: Northeastern
University Press.
Akers, R. L., & Jensen, G. F. (2006). The empirical status of social learning theory of crime and deviance: The past, present,
and future. In F. T. Cullen, J. P. Wright, & K. R. Blevins (Eds.), Taking stock: The status of criminological theory
(Advances in Criminological Theory Series Vol. 15, pp. 37-76). New Brunswick, NJ: Transaction Publishers.
Barkley, R. A. (1997). ADHD and the nature of self-control. New York: Guilford.
Baron, S. W. (2003). Self-control, social consequences, and criminal behavior: Street youth and the general theory of crime.
Journal of Research in Crime and Delinquency, 40, 403-425.
Beaver, K. M. (2008). Nonshared environmental influences on adolescent delinquent involvement and adult criminal behav-
ior. Criminology, 46, 341-370.
Beaver, K. M., & Wright, J. P. (2005a). Biosocial development and delinquent involvement. Youth Violence and Juvenile
Justice, 3, 168-192.
Beaver, K. M., & Wright, J. P. (2005b). Evaluating the effects of birth complications on low self-control in a sample of twins.
International Journal of Offender Therapy and Comparative Criminology, 49, 450-471.
Beaver, K. M., Wright, J. P., & DeLisi, M. (2007). Self-control as an executive function: Reformulating Gottfredson and
Hirschi’s parental socialization thesis. Criminal Justice and Behavior, 34, 1345-1361.
Beaver, K. M., Wright, J. P., & DeLisi, M. (2008). Delinquent peer group formation: Evidence of a gene X environment cor-
relation. Journal of Genetic Psychology, 169, 227-244.
Bellair, P. E., Roscigno, V. J., & McNulty, T. L. (2003). Linking local labor market opportunity to violent adolescent delin-
quency. Journal of Research in Crime and Delinquency, 40, 6-33.
Cairns, R. B., & Cairns, B. D. (1994). Lifelines and risks: Pathways of youth in our time. Cambridge, UK: Cambridge
University Press.
Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I. W., et al. (2002). Role of genotype in the cycle of violence
in maltreated children. Science, 297, 851-854.
Cauffman, E., Steinberg, L., & Piquero, A. R. (2005). Psychological, neuropsychological and physiological correlates of seri-
ous antisocial behavior in adolescence: The role of self-control. Criminology, 43, 133-175.
Chapple, C. L. (2005). Self-control, peer relations, and delinquency. Justice Quarterly, 22, 89-106.
Cleveland, H. H., Wiebe, R. P., & Rowe, D. C. (2005). Sources of exposure to smoking and drinking friends among adoles-
cents: A behavioral-genetic evaluation. Journal of Genetic Psychology, 166, 153-169.
Crosnoe, R., & Elder, G. H., Jr. (2004). Family dynamics, supportive relationships, and educational resilience during adoles-
cence. Journal of Family Issues, 25, 571-602.
Cyphers, L. H., Phillips, K., Fulker, D. W., & Mrazek, D. A. (1990). Twin temperament during the transition from infancy to
early childhood. Journal of the American Academy of Child and Adolescent Psychiatry, 29, 392-397.
DeFries, J. C., & Fulker, D. W. (1985). Multiple regression analysis of twin data. Behavior Genetics, 15, 467-473.
DeLisi, M., Hochstetler, A., & Murphy, D. S. (2003). Self-control behind bars: A validation study of the Grasmick et al.
scale. Justice Quarterly, 20, 241-263.
DiLalla, L. F. (2002). Behavior genetics of aggression in children: Review and future directions. Developmental Review, 22,
593-622.
Gibbs, J. J., Giever, D., & Higgins, G. E. (2003). A test of Gottfredson and Hirschi’s general theory using structural equation
modeling. Criminal Justice and Behavior, 30, 441-458.
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 57
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
Gibbs, J. J., Giever, D., & Martin, J. S. (1998). Parental management and self-control: An empirical test of Gottfredson and
Hirschi’s general theory. Journal of Research in Crime and Delinquency, 35, 40-70.
Gottfredson, M., & Hirschi, T. (1990). A general theory of crime. Palo Alto, CA: Stanford University Press.
Grasmick, H. G., Tittle, C. R., Bursik, R. J., Jr., & Arneklev, B. J. (1993). Testing the core empirical implications of
Gottfredson and Hirschi’s general theory of crime. Journal of Research in Crime and Delinquency, 30, 5-29.
Hanushek, E. A., & Jackson, J. E. (1977). Statistical methods for social scientists. San Diego, CA: Academic Press.
Harris, J. R. (1998). The nurture assumption: Why children turn out the way they do. New York: Free Press.
Harris, K. M., Florey, F., Tabor, T., Bearman, P. S., Jones, J., & Udry, J. R. (2003). The National Longitudinal Study of
Adolescent Health: Study design. Chapel Hill, NC: Carolina Population Center. Available at http://www.cpc.unc.edu/
projects/addhealth/design
Haynie, D. L. (2001). Delinquent peers revisited: Does network structure matter? American Journal of Sociology, 106, 1013-1057.
Haynie, D. L., & McHugh, S. (2003). Sibling deviance: In the shadows of mutual and unique friendship effects. Criminology,
41, 355-392.
Hirschi, T. (1969). Causes of delinquency. Berkeley: University of California Press.
Iervolino, A. C., Pike, A., Manke, B., Reiss, D., Hetherington, E. M., & Plomin, R. (2002). Genetic and environmental influ-
ences in adolescent peer socialization: Evidence from two genetically sensitive designs. Child Development, 73, 162-174.
Ishikawa, S. S., & Raine, A. (2003). Prefrontal deficits and antisocial behavior. In B. B. Lahey, T. E. Moffitt, & A. Caspi
(Eds.), Causes of conduct disorder and juvenile delinquency (pp. 277-304). New York: Guilford.
Jacobson, K., & Rowe, D. C. (1998). Genetic and shared environment influences on adolescent BMI: Interaction with race
and sex. Behavior Genetics, 28, 265-278.
Karr-Morse, R., & Wiley, M. S. (1997). Ghosts from the nursery: Tracing the roots of violence. New York: Atlantic Monthly.
Kendler, K. S., Jacobson, K. C., Gardner, C. O., Gillespie, N., Aggen, S. A., & Prescott, C. A. (2007). Creating a social world:
A developmental twin study of peer-group deviance. Archives of General Psychiatry, 64, 958-965.
Kohler, H. P., & Rodgers, J. L. (2001). DF-analysis of heritability with double-entry twin data: Asymptotic standard errors
and efficient estimation. Behavior Genetics, 31, 179-191.
Lahey, B. B., & Waldman, I. D. (2003). A developmental propensity model of the origins of conduct problems during child-
hood and adolescence. In B. B. Lahey, T. E. Moffitt, & A. Caspi (Eds.), Causes of conduct disorder and juvenile delin-
quency (pp. 76-117). New York: Guilford.
Loeber, R., & Stouthamer-Loeber, M. (1986). Family factors as correlates and predictors of juvenile conduct problems and
delinquency. In M. Tonry & N. Morris (Eds.), Crime and justice: A review of research (Vol. 7, pp. 29-149). Chicago:
University of Chicago Press.
Longshore, D., Stein, J. A., & Turner, S. (1998). Reliability and validity of a self-control measure: Rejoinder. Criminology,
36, 175-182.
Longshore, D., Turner, S., & Stein, J. A. (1996). Self-control in a criminal sample: An examination of construct validity.
Criminology, 34, 209-228.
Marcus, B. (2003). An empirical examination of the construct validity of two alternative self-control measures. Educational
and Psychological Measurement, 63, 674-706.
Marcus, B. (2004). Self-control in the general theory of crime: Theoretical implications of a measurement problem.
Theoretical Criminology, 8, 33-55.
McCartan, L. M. (2007). Inevitable, influential, or unnecessary? Exploring the utility of genetic explanation for delinquent
behavior. Journal of Criminal Justice, 35, 219-233.
Meyer-Lindenberg, A., Buckholtz, J. W., Kolachana, B., Hariri, A. R., Pezawas, L., Blasi, G., et al. (2006). Neural mecha-
nisms of genetic risk for impulsivity and violence in humans. Proceedings of the National Academy of Sciences of the
United States of America, 103, 6269-6274.
Moffitt, T. E. (1993). Adolescence-limited and life-course persistent antisocial behavior: A developmental taxonomy.
Psychological Review, 100, 674-701.
Morgan, A. B., & Lilienfeld, S. O. (2002). A meta-analytic review of the relation between antisocial behavior and neuropsy-
chological measures of executive function. Clinical Psychology Review, 20, 113-136.
Nigg, J. T., & Huang-Pollock, C. L. (2003). An early-onset model of the role of executive functions and intelligence in con-
duct disorder/delinquency. In B. B. Lahey, T. E. Moffitt, & A. Caspi (Eds.), Causes of conduct disorder and juvenile
delinquency (pp. 227-553). New York: Guilford.
Patterson, G. R. (1982). Coercive family process. Eugene, OR: Castilia.
Perrone, D., Sullivan, C. J., Pratt, T. C., & Margaryan, S. (2004). Parental efficacy, self-control, and delinquency: A test of a
general theory of crime on a nationally representative sample of youth. International Journal of Offender Therapy and
Comparative Criminology, 48, 298-312.
Pfefferbaum, A., Sullivan, E. V., Swan, G. E., & Carmelli, D. (2000). Brain structure in men remains highly heritable in the
seventh and eighth decades of life. Neurobiology of Aging, 21, 63-74.
Piquero, A. R., & Rosay, A. B. (1998). The reliability and validity of Grasmick et al.’s self-control scale: A comment on
Longshore et al. Criminology, 35, 157-174.
58 CRIMINAL JUSTICE AND BEHAVIOR
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
Polakowski, M. (1994). Linking self- and social control with deviance: Illuminating the structure underlying a general theory
of crime and its relation to deviant identity. Journal of Quantitative Criminology, 10, 41-78.
Pratt, T. C., & Cullen, F. T. (2000). The empirical status of Gottfredson and Hirschi’s general theory of crime: A meta-
analysis. Criminology, 38, 931-964.
Pratt, T. C., Turner, M. G., & Piquero, A. R. (2004). Parental socialization and community context: A longitudinal analysis
of the structural sources of low self-control. Journal of Research in Crime and Delinquency, 41, 219-243.
Price, T. S., Simonoff, E., Waldman, I., Asherson, P., & Plomin, R. (2001). Hyperactivity in preschool children is highly her-
itable. Journal of the American Academy of Child and Adolescent Psychiatry, 40, 1362-1364.
Raine, A. (2002). Annotation: The role of prefrontal deficits, low autonomic arousal, and early health factors in the develop-
ment of antisocial and aggressive behavior in children. Journal of Child Psychology and Psychiatry, 43, 417-434.
Reiss, D., Neiderhiser, N. M., Hetherington, E. M., & Plomin, R. (2000). The relationship code: Deciphering the genetic and
social influences on adolescent development. Cambridge, UK: Cambridge University Press.
Rhee, S. H., & Waldman, I. D. (2003). Testing alternative hypotheses regarding the role of development on genetic and envi-
ronmental influences underlying antisocial behavior. In B. B. Lahey, T. E. Moffitt, & A. Caspi (Eds.), Causes of conduct
disorder and juvenile delinquency (pp. 305-318). New York: Guilford.
Rietveld, M. J. H., Hudziak, J. J., Bartels, M., van Beijsterveldt, C. E. M., & Boomsma, D. I. (2003). Heritability of attention prob-
lems in children: Cross-sectional results from a study of twins, age 3 to 12 years. Neuropsychiatric Genetics, 1176, 102-113.
Robinson, M. (2004). Why crime? An integrated systems theory of antisocial behavior. Upper Saddle River, NJ: Pearson
Prentice Hall.
Rodgers, J. L., Buster, M., & Rowe, D. C. (2001). Genetic and environmental influences on delinquency: DF analysis of
NLSY kinship data. Journal of Quantitative Criminology, 17, 145-168.
Rodgers, J. L., & Kohler, H. P. (2005). Reformulating and simplifying the DF analysis model. Behavior Genetics, 35, 211-217.
Rodgers, J. L., & Rowe, D. C. (1987). IQ similarity in twins, siblings, half-siblings, cousins, and random pairs. Intelligence,
11, 199-206.
Rodgers, J. L., Rowe, D. C., & Li, C. (1994). Beyond nature versus nurture: DF analysis of nonshared influences on prob-
lem behaviors. Developmental Psychology, 30, 374-384.
Rutter, M. (2006). Genes and behavior: Nature-nurture interplay explained. Malden, MA: Blackwell.
Sampson, R. J., & Laub, J. H. (1993). Crime in the making: Pathways and turning points through life. Cambridge, MA:
Harvard University Press.
Scarr, S. (1992). Developmental theories for the 1990s: Development and individual differences. Child Development, 63, 1-19.
Scarr, S., & McCartney, K. (1983). How people make their own environments: A theory of genotype → environment effects.
Child Development, 54, 424-435.
Schreck, C. J., Fisher, B. S., & Miller, J. M. (2004). The social context of violent victimization: A study of the delinquent
peer effect. Justice Quarterly, 21, 23-47.
Séguin, J. R., Boulerice, B., Harden, P. W., Tremblay, R. E., & Pihl, R. O. (1999). Executive functions and physical aggres-
sion after controlling for attention deficit hyperactivity disorder, general memory, and IQ. Journal of Child Psychology
and Psychiatry, 40, 1197-1208.
Séguin, J. R., Nagin, D., Assaad, J. M., & Tremblay, R. E. (2004). Cognitive-neuropsychological function in chronic physi-
cal aggression and hyperactivity. Journal of Abnormal Psychology, 113, 603-613.
Sood, B., Delaney-Black, V., Covington, C., Nordstrom-Klee, B., Ager, J., Templin, T., et al. (2001). Prenatal alcohol expo-
sure and childhood behavior at age 6 to 7 years: I. Dose-response effect. Pediatrics, 108, 34-43.
Thompson, P. M., Cannon, T. D., Narr, K. L., van Erp, T., Poutanen, V., Huttunen, M., et al. (2001). Genetic influences on
brain structure. Nature Neuroscience, 4, 1-6.
Toga, A. W., & Thompson, P. M. (2005). Genetics of brain structure and intelligence. Annual Review of Neuroscience, 28, 1-23.
Tremblay, R. E. (2003). Why socialization fails: The case of chronic physical aggression. In B. B. Lahey, T. E. Moffitt, &
A. Caspi (Eds.), Causes of conduct disorder and juvenile delinquency (pp. 182-224). New York: Guilford.
Turkheimer, E., D’Onofrio, B. M., Maes, H. H., & Eaves, L. J. (2005). Analysis and interpretation of twins studies including
measures of the shared environment. Child Development, 76, 1217-1233.
Turner, M. G., Piquero, A. R., & Pratt, T. C. (2005). The school context as a source of self-control. Journal of Criminal
Justice, 33, 327-339.
Udry, J. R. (2003). The National Longitudinal Study of Adolescent Health (Add Health), Waves I and II, 1994-1996; Wave
III, 2001-2002 [Machine-readable data file and documentation]. Chapel Hill: Carolina Population Center, University of
North Carolina at Chapel Hill.
Unnever, J. D., Cullen, F. T., & Pratt, T. C. (2003). Parental management, ADHD, and delinquent involvement: Reassessing
Gottfredson and Hirschi’s general theory. Justice Quarterly, 20, 471-500.
Walsh, A. (2000). Behavior genetics and anomie/strain theory. Criminology, 38, 1075-1107.
Walsh, A. (2002a). Biosocial criminology: Introduction and integration. Cincinnati, OH: Anderson.
Walsh, A. (2002b). Essay review: Companions in crime: A biosocial perspective. Human Nature Review, 2, 169-178.
Warr, M. (1993). Age, peers, and delinquency. Criminology, 31, 17-40.
Beaver et al. / GENETIC AND ENVIRONMENTAL EFFECTS 59
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from
Warr, M. (2002). Companions in crime: The social aspects of criminal conduct. New York: Cambridge University Press.
Wright, J. P., & Beaver, K. M. (2005). Do parents matter in creating self-control in their children? A genetically informed test
of Gottfredson and Hirschi’s theory of low self-control. Criminology, 43, 1169-1202.
Wright, J. P., Beaver, K. M., DeLisi, M., & Vaughn, M. G. (2008). Evidence of negligible parenting influences on self-
control, delinquent peers, and delinquency in a sample of twins. Justice Quarterly, 25, 544-569.
Wright, B. R. E., Caspi, A., Moffitt, T. E., & Silva, P. A. (1999). Low self-control, social bonds, and crime: Social causation,
social selection, or both? Criminology, 37, 479-514.
Yolton, K., Dietrich, K., Auinger, P., Lanphear, B. P., & Hornung, R. (2005). Exposure to environmental tobacco smoke and
cognitive abilities among U.S. children and adolescents. Environmental Health Perspectives, 113, 98-103.
Kevin M. Beaver is currently an assistant professor in the College of Criminology and Criminal Justice at Florida State
University. His research examines the ways in which the environment intersects with biological and genetic factors to pro-
duce antisocial outcomes. He is the author of Biosocial Criminology: A Primer (Kendall/Hunt, 2009).
J. Eagle Shutt, AB (Harvard University), JD, MCJ, PhD (University of South Carolina), is an assistant professor at the
University of Louisville’s Department of Justice Administration. His research interests include evolutionary psychology, sub-
cultures, and law.
Brian B. Boutwell is a doctoral student in the College of Criminology and Criminal Justice at Florida State University. His
research and teaching interests include life-course and biosocial criminology, including the gene-environment basis to anti-
social behaviors. His recent publications have appeared in Criminal Behaviour and Mental Health and Criminology.
Marie Ratchford is a graduate student in the College of Criminology and Criminal Justice at Florida State University. Her cur-
rent research interests include biosocial criminology with an emphasis on identifying the genetic correlates to criminal behavior.
Kathleen Roberts is a doctoral student in the College of Criminology and Criminal Justice at Florida State University. Her
research examines the biosocial and genetic correlates to antisocial behavior.
J. C. Barnes is a doctoral student in the College of Criminology and Criminal Justice at Florida State University. His research
interests include the etiology of delinquency, especially the biosocial correlates to antisocial behaviors. His work has appeared
in the Criminal Justice Policy Review, American Journal of Criminal Justice, and Journal of Criminal Justice Education.
60 CRIMINAL JUSTICE AND BEHAVIOR
at SAGE Publications on February 10, 2009 http://cjb.sagepub.comDownloaded from