Citation
Family structure and juvenile delinquency

Material Information

Title:
Family structure and juvenile delinquency the mediating role of social learning variables
Creator:
Page, Edwin R., 1966-
Publication Date:
Language:
English
Physical Description:
ix, 173 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Delinquency ( jstor )
Family structure ( jstor )
Learning ( jstor )
Mathematical variables ( jstor )
Observational learning ( jstor )
Parents ( jstor )
Social learning theory ( jstor )
Social structures ( jstor )
Violence ( jstor )
Young offenders ( jstor )
Dissertations, Academic -- Sociology -- UF ( lcsh )
Sociology thesis, Ph.D ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 1998.
Bibliography:
Includes bibliographical references (leaves 164-172).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Edwin R. Page.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
029545892 ( ALEPH )
40257769 ( OCLC )

Downloads

This item has the following downloads:


Full Text











FAMILY STRUCTURE AND JUVENILE DELINQUENCY:
THE MEDIATING ROLE OF
SOCIAL LEARNING VARIABLES














By


EDWIN R. PAGE
















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA

1998


























Copyright 1998

by

Edwin R. Page














ACKNOWLEDGMENTS


I express my most heartfelt gratitude to Natalee Waters

for the love, friendship, and support that she gave me during

my work on this research.














TABLE OF CONTENTS


ACKNOWLEDGMENTS ....................... .................. iii

LIST OF TABLES ............................................. vi

ABSTRACT ................................................. viii

CHAPTERS

1 INTRODUCTION ............................................. 1

2 FAMILY STRUCTURE AND DELINQUENCY ..........................5

Family Relationships ...................................... 5
Parental Criminality ...................................6
Parental Discipline and Punishment .....................7
Abuse in the Family ....................................9
Other Family Relationships ............................15
Family Structure and Broken Homes ........................18
Conceptual and Methodological Issues ..................19
Research on Family Structure ..........................23
Conclusions.............................................37

3 SOCIAL LEARNING THEORY OF CRIME AND DELINQUENCY ..........41

Social Learning Theory ..................................42
Differential Association ..............................44
Definitions ...........................................45
Imitation .............................................47
Differential Reinforcement ............................48
Temporal Sequence and Reciprocal Effects in Social
Learning Theory ....................................49
Social Structure and Social Learning: The Case of
the Family and Delinquency .........................51
Research on Social Learning Theory.......................53
Social Learning Theory and Delinquency ................53
Family, Social Learning Theory and Delinquency .......60
Hypotheses............................................... 65
The Present Research .....................................68

4 METHODOLOGY ..............................................71

Data..................................................... 71
Measurement of Variables .................................76
Delinquency Index ....................................76
Family Structure ......................................78








Social Learning Variables ............................80
Sociodemographic Variables ............................87
Plan of Analysis.........................................95

5 ANALYSIS AND RESULTS .................................... 100

Hypothesis One: Family Structure and Delinquency........101
Hypothesis Two: Family Structure and Social Learning... 102
Hypothesis Three: Social Learning and Delinquency....... 104
Hypothesis Four: Family Structure, Social Learning,
and Delinquency ...................................... 106
Additional Analyses ..................................... 115
Structural Effects of Sociodemographic Variables .....115
Logistic Regression ..................................116
Controlling for Sex ..................................121

6 CONCLUSIONS .............................................130

Summary .................................................130
Conclusions............................................136
Implications for Policy .................................140
Limitations and Suggestions for Future Research......... 144

APPENDICES

A VARIABLE MEASUREMENT ITEMS .............................150

B ADDITIONAL REGRESSION TABLES .................... ..........156

REFERENCES ................................................164

BIOGRAPHICAL SKETCH ....................................... 173













LIST OF TABLES


Table page

1. Correlation Matrix .................................... 93

2. Analysis Plans for Family Structure, Social
Learning, and Delinquency: Years in which
Dependent, Independent, and Sociodemographic
Variables are Measured ............................96

3. Percentage Distributions of Self-Reported
Delinquency (SRD) ................................100

4. Crosstabulation Summary: Percent Nondelinquent by
Family Structure for Each Analysis Plan ..........101

5. Zero-Order Correlation Coefficients for Self-
Reported Delinquency (SRD) and Family Structure
in Four Analysis Plans ...........................102

6. Zero-Order Correlation Coefficients for Family
Structure and Social Learning Variables in Four
Analysis Plans .................................. 103

7. Crosstabulation Summary: Self-Reported Delinquency
(SRD) across Averaged Social Learning Scale
Items for Each Social Learning Variable .......... 104

8. Zero-Order Correlation Coefficients for Self-
Reported Delinquency and Social Learning
Variables in Four Analysis Plans ................. 105

9. Family Structure, Social Learning, and Delinquency:
One-Year Lagged Analysis Plan (1) with Self-
Reported Delinquency Measured in Year Two ........ 107

10. Family Structure, Social Learning, and Delinquency:
One-Year Lagged Analysis Plan (2) with Self-
Reported Delinquency Measured in Year Three ......109

11. Family Structure, Social Learning, and Delinquency:
Two-Year Lagged Analysis Plan (3) with Self-
Reported Delinquency Measured in Years Two and
Three ............................................ 111








12. Family Structure, Social Learning, and Delinquency:
Temporal Sequence Analysis Plan (4) with Self-
Reported Delinquency Measured in Year Three ......112

13. Family Structure, Social Learning, and Delinquency:
Analysis Plan 1 Using Logistic Regression ........ 118

14. Percentage Distributions of Self-Reported
Delinquency (SRD) for Males and Females .......... 122

15. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 for Male Subsample ...............123

16. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 for Female Subsample ............. 124

17. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 for Male Subsample ...............125

18. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 for Female Subsample .............127

19. Sociodemographic Variables, Family Structure,
Social Learning, and Delinquency: OLS
Regression ....................................... 156

20. Family Structure, Social Learning, and Delinquency:
Analysis Plan 2 Using Logistic Regression ........ 157

21. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 Using Logistic Regression ........ 158

22. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 Using Logistic Regression ....... 159

23. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 Using Logistic Regression for
Male Subsample ...................................160

24. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 Using Logistic Regression for
Female Subsample .................................161

25. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 Using Logistic Regression for
Male Subsample ...................................162

26. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 Using Logistic Regression for
Female Subsample .................................163


vii













Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


FAMILY STRUCTURE AND JUVENILE DELINQUENCY:
THE MEDIATING ROLE OF
SOCIAL LEARNING VARIABLES

By

Edwin R. Page

August, 1998




Chairman: Ronald L. Akers
Major Department: Sociology


There has been relatively little research to date

connecting the behavioral process in social learning theory

to elements of social structure in accounting for criminal

and delinquent behavior. The linking of social structure to

crime and delinquency through the social learning process is

the next step in the development of social learning theory.

The theory would hypothesize that the variables specified in

social learning theory mediate the effects of social

structure on deviant behavior.

Seeking to establish intervening variables in the

relationship between family structure and delinquency, a

three-level theoretical model is presented with six measures

representing three components of social learning theory. A


viii








longitudinal analysis is conducted on three waves of a

nationally representative sample, permitting generalizability

of findings. The panel data of the National Youth Survey

enable the observation of particular individuals in terms of

their family structure and their scoring on social learning

measures, and their subsequent involvement in serious

delinquent or criminal behavior. Correlational and

regression analyses are used to establish relationships

between the three levels of measurement, and to test the

central hypothesis that social learning variables mediate the

effects of family structure on delinquency.

Results provide support for the hypotheses that family

structure is associated with delinquency, that family

structure is associated with social learning variables, that

social learning variables are associated with delinquency,

and that selected social learning variables generally reduce

the strength and significance of the effect of family

structure on delinquency. These findings persist after

controlling for sex. It is concluded that the social

learning variables significantly mediate the impact of family

structure on delinquency.













CHAPTER 1
INTRODUCTION


Although crime rates have moderated and declined in the

past few years, between 1965 and 1990 the arrest rate for

violent crime by juveniles rose from approximately 130 per

100,000 to 430 per 100,000, according to the FBI (U.S.

Department of Justice, 1992). During the same period of time

there also has been considerable change in the structure of

families. The divorce rate (per 1,000 married women), for

example, rose from 9.2 in 1960 to 21 in 1988, having peaked

in 1981 at 22.6 (U.S. Department of Commerce, Bureau of the

Census, 1994a). The rising divorce rate has accompanied a

general change in family structures. The amount of single-

parent families, for example, increased from 12 percent of

families with dependent children in 1970 to 27 percent in

1993 (U.S. Department of Commerce, Bureau of the Census,

1994b). Many researchers have sought to specify a causal

link between the increase in crime and the changes in

families. Families are an intuitively appealing place to

look for an explanation of delinquency because of the

prominent role that they occupy in the process of

socialization.

Family has long been an important variable in the

etiology of delinquency in both social and biological








research (see Akers, 1994; Siegel and Senna, 1991; Hagan and

Sussman, 1988; Vold and Bernard, 1986). The sociological

literature suggests a variety of ways by which family

variables affect delinquency (Jensen and Rojek, 1992). For

example, strain theories position the family as an important

determinant of social class and thereby the likelihood of

delinquency. Social learning theory suggests that families

are an important shaper of conforming or deviant attitudes

and behavior. For control theories, families are an

important source of bonding by which youths are directed

toward conformity and away from delinquency.

It is easy enough to understand how the family would

usually have a considerable impact on delinquency. "The

family has typically been viewed as the most crucial

institution in our society for shaping a child's personality,

attitudes, and behavior," (1992, p. 262) according to Jensen

and Rojek. Children typically spend most of their early

years in the presence of other family members. Even after

reaching school age the family remains one of the most

important institutions of socialization, along with schools

and peer groups.

Socialization in the family remains significant today

even if the social functions of families have declined and

individuals spend less time in the family setting. With the

increased complexity of modern life and the variety of

available leisure activities, individuals today are more

likely than their forebears to live a greater portion of








their lives outside of the home. The typical household size

itself declined as individuals increasingly moved away from

their family of origin and the nuclear family became more

common. Census data show that the average household size

declined from 5.4 persons in 1790 to 4.2 in 1900, to 3.3 in

1940, and to 2.6 in 1988 (Eshleman, 1991). Despite changes

such as these, families remain central socializing forces

today. "Although the family may have been a more important

social institution at earlier points in history, most

scholars are still willing to accept that it remains a major

setting for socialization in American society" (Jensen and

Rojek, 1992, p. 263).

The present study adds to the empirical and theoretical

body of knowledge relating to the connection between family

structure and delinquency in a variety of ways. While it

will be shown that the empirical validity of social learning

theory has been well demonstrated in the literature, there

has been relatively little research to date connecting the

behavioral emphasis of the theory to elements of social

structure. The linking of social structure to crime and

delinquency by the social learning process is the next step

in the development of social learning theory (Akers, 1994).

In looking to establish intervening variables in the

relationship between family structure and delinquency, a

three-level theoretical model will be presented with six

measures representing three components of social learning

theory. The theory would hypothesize that the variables in








social learning theory mediate the effects of social

structure on deviant behavior.

Additionally, the methodological design employed here

attempts to improve on the shortcomings of much previous

research on family and delinquency. In a review of the

literature, Free (1991) concluded that a major weakness of

previous studies is their overreliance on cross-sectional

data and that there has been an overemphasis on minor

offenses. A longitudinal analysis will be employed similar

to that used by Heimer and Matsueda (1994) but using more

recent data spanning three years. The use of a nationally

representative sample will allow a broad generalizability of

findings. The panel data of the National Youth Survey and

the use of longitudinal analysis will enable the observation

of particular individuals who have experienced a change in

family structure and their reactions to that change in terms

of serious delinquent or criminal behavior.














CHAPTER 2
FAMILY STRUCTURE AND DELINQUENCY


In this chapter a variety of literature pertaining to

family and delinquency will be reviewed. First, the impact

of family relationships will be considered, including the

effects that parental criminality, parental discipline and

punishment, various types of abuse, and other family

relationships have on delinquency. This part of the review

will provide a background for the subsequent review of the

literature on family structure and delinquency. As will be

shown, some researchers maintain that family structure

affects delinquency, at least in part, through family

relationships such as those to be discussed here. As the

exogenous independent variable in this study, the concept of

family structure and the implications of its measurement will

be discussed. Finally, research pertaining to the

relationship between family structure and delinquency will be

reviewed.


Family Relationships


Given that families are a key socializing institution,

they may be implicated in delinquency to the extent that the

socialization they provide is inadequate or deviant.

Parental criminality is one way that delinquency may be based








in the family. Delinquency has also been investigated as a

product of the method used by parents to discipline and

punish their children. Family violence and abuse is another

way in which socialization may result in delinquent behavior.

Furthermore, other nonviolent and seemingly healthy family

relationships may conceal pathological interaction and

involve socialization toward delinquency.


Parental Criminality


Parental criminality has been found to be related to an

increased likelihood of delinquency. Using official records

from the Cambridge Study in Delinquency Development, West

(1982) found that criminality in both fathers, mothers, and

brothers was highly predictive of delinquent behavior. Sons

with criminal fathers were approximately twice as likely to

engage in criminal activities by age 22 compared to those

with noncriminal fathers. The presence of multiple criminal

influences further increased the likelihood of a son engaging

in delinquency. "The chances of a man being a delinquent

were some three and a half times greater if he had more than

two other family members with a record than if he belonged to

a conviction-free family," (1982, p. 45) according to West.

In a recent article using data from the Cambridge Study, Rowe

and Farrington (1997) suggested a genetic explanation for the

familial transmission of convictions. Their findings

indicated that parental criminality affected child

criminality directly and not via the effects of family








environment. However, they could not rule out the role of

family environment altogether due to a small family

environment effect in their findings and the possibility of

other unmeasured effects.

Lattimore et al. (1995) found parental criminality to be

one among several family variables, including family

violence, that was significantly related to violent

recidivism in a cohort of male parolees. In a Danish cohort

study, Mednick et al. (1987) found that fathers' criminality

was directly associated with general crime in their young

adult sons. Mednick et al. (1990) found that fathers'

criminality influenced adult sons' criminality, in part

through an increased risk of family instability. Laub and

Sampson (1988; see also Sampson and Laub, 1993) found that

criminality in either parent, and especially substance abuse,

had a strong indirect influence on delinquency by way of

disrupted social control in the family. Analyzing data

collected in the 1940s by Sheldon and Eleanor Glueck, Laub

and Sampson found that significant effects exhibited by

parental criminality and drunkenness on serious delinquency

were largely mediated by family process variables such as

attachment, supervision, and erratic parental discipline.


Parental Discipline and Punishment


Over 90 percent of American parents use physical

punishment to correct their children's misbehavior, according

to Straus (1991). Most research is in agreement that








inconsistent parental discipline, whether too harsh or too

lenient, is related to delinquent behavior (Jensen and Rojek,

1992; Siegel and Senna, 1991). In a research review, Snyder

and Patterson (1987) report strong evidence of a link between

ineffective disciplinary practices and the families of

delinquents. "In both lax and enmeshed discipline styles,

punishment is not used consistently, contingently, and

effectively to discourage the aggressive, antisocial behavior

of the child," (1987, p. 223) according to Snyder and

Patterson. Agnew (1983) found that while physical punishment

slightly decreased the likelihood of self-reported

delinquency among children who perceived their parents as

consistent, where discipline was inconsistent children were

more likely to report delinquency. Krohn et al. (1992) found

that consistent punishment and spanking were significantly

related to a decrease in self-reported delinquency. McCord

(1991) found that consistent, nonpunitive discipline by

mothers helped to insulate sons from delinquent influences.

Investigating the sex differences in delinquency, Jensen and

Eve (1976) suggested that the more restrictive supervision

received by females may partly account for their lower

involvement in delinquency.

Using the Glueck data, Sampson and Laub (1993) also

found that the use of erratic and punitive discipline or

punishment by mothers and fathers affected official and

unofficial delinquency. Moreover, they found that background

variables, such as father's alcoholism and criminality,








affected discipline and punishment by either parent,

suggesting "the possibility that some, if not all, of the

effect of father's criminality on the son's delinquency may

be attributable to family discipline" (Sampson and Laub,

1993, p. 79). Finally, they found that discipline and

punishment, among other family process variables, mediated

the effects of structural background variables on official

and unofficial delinquency.


Abuse in the Family


violence in the family may take the form of child abuse,

including sexual abuse, as well as violence between parents

or siblings. Child abuse may refer to the overt physical

aggression against a child or to neglect, wherein children

are deprived of attention, food, shelter, or other care. Any

type of violence or abuse in the family, directly involving a

child or not, may increase the likelihood of that child

engaging in delinquency.

The issue of child abuse has gained much recognition

since the 1960s. Although the true extent of abuse is not

known due to the difficulty in obtaining accurate data and

the methodological differences of various reports, most

researchers agree that child abuse is a significant problem

(Gelles and Conte, 1991; Pagelow, 1984). Straus and Gelles

(1986), for example, found that parents reported incidents of

severe violence toward their children at an annual rate of 19

per 1,000 children aged three to 17. Their measure of severe








violence included kicking, biting and hitting with a fist;

beating up; and using a gun or knife. The findings by Straus

and Gelles were considered conservative because the sample

included only those children living with both parents and the

violence was self-reported, and there may have been an

increased consciousness of the unacceptability of child abuse

by the mid-1980s. Additionally, this finding includes only

the most serious type of abuse. Although Straus and Gelles

report that this figure represented a decline in violent

abuse compared to their findings in 1975, other studies

suggest that child abuse has increased (see Gelles and Conte,

1991).

Sexual abuse is a more specific form of child abuse that

poses serious social-psychological threats to children

(Gelles and Conte, 1991). It may range from rewarding a

naive child for inappropriate sexual behavior, to fondling

and other inappropriate touching, to using force or the

threat of force for the purposes of sex (Siegel and Senna,

1991). Findings on the amount of sexual abuse vary but

typical estimates are that 19 percent of females and nine

percent of males are sexually abused before age 18

(Finkelhor, 1984). Sex abuse within the family usually

occurs across generations. The victims are more often

female, although sex abuse of boys is more likely

underreported, and the abusers are overwhelmingly male

(Pagelow, 1984; Finkelhor, 1984).








In their national survey of family violence, Gelles and

Straus report that children from violent homes are "three to

four times more likely than children from nonviolent homes to

engage in illegal acts" (1988, p. 129). While they were not

able to establish causality, Gelles and Straus suggest that

their research and other studies show evidence of a cycle of

violence existing in many homes. "Violent experiences set

the stage for the individual and social traits that lead to

trouble. The trouble is responded to with more violence"

(1988, p. 129).

The effects of child abuse and its relationship with

delinquency are not clear even though it is generally agreed

that abuse does have detrimental long-term consequences.

"Definitive support for a direct causal relationship between

an abusive childhood and delinquency is lacking, but

researchers generally agree that abuse and neglect may have a

profound effect on behavior in later years," (1991, p. 265)

according to Siegel and Senna. The authors said that this

agreement is partly based on a number of studies showing that

many delinquents and criminals have a history of child abuse.

In a cohort study comparing serious, officially

recognized abuse cases with a control group Widom (1989a)

found that 29 percent of the abused group later had adult

criminal records for a nontraffic offense compared to 21

percent of the control group. Widom's measures of abuse

included violent physical abuse, sex abuse, and neglect.

Considering demographic variables, Widom found that those who








were older at the time of abuse, male, and black were at

highest risk of criminality while being young, white, and

female was associated with lower risk. While abused youths

were more likely to become criminal than nonabused youths,

the relatively small difference between the two groups,

compared to what might be expected, points to the indirect

nature of the relationship between family variables and

delinquency.

Widom (1989b) reported that children who were neglected,

and especially, violently abused had a significantly greater

risk of becoming delinquents, criminals, and violent

criminals. More recently, Widom (1996) also found a greater

likelihood of arrest as juveniles and as adults for the

victims of childhood abuse or neglect. These findings

applied to arrests for both drug-related offenses and

property crimes. Widom also found that "the victims of

sexual abuse are not more likely than other victims of

physical abuse or neglect to become involved with crime"

(1996, p. 50).

Widom (1989b) suggested that her findings, and similar

findings by other researchers, have a mixed message for the

cycle of violence hypothesis. Considering the large number

of abused children that do not become delinquent or criminal,

Widom (1989b, 1996) suggested that many victims apparently

manifest their abuse in more subtle ways. This may be the

case particularly for females, who are both more likely to be

sexually abused and less likely to become delinquent. Widom








said that attention should focus on possible mediating

variables that buffer or protect abused or neglected children

from delinquent outcomes.

Reviewing the literature on the effects of childhood sex

abuse, Finkelhor (1984) reported that a variety of studies

have shown high proportions of various troubled populations,

including delinquents, were sexually victimized as children.

While he said the case apparently may be easily made that sex

abuse leads to injurious outcomes, Finkelhor cautioned, "It

is possible that the long-term effects seen in these cases

are a function not of the sexual abuse but of other

pathological elements, such as psychological abuse, parental

neglect, or family disorganization" (1984, p. 189).

Similarly, Pagelow (1984) warned that the finding of a

strong relationship between childhood violence and later

deviant behavior may overlook the issue of causality. "Such

a blanket assumption would be hazardous because there are too

many possible intervening variables that can influence

outcomes not included in simple 'cause and effect' models"

(1984, p. 229). Pagelow also suggested that official records

may be misleading. "Connections between child protection

agencies and the juvenile justice system are clear. Entering

the first, by becoming an identified case, can serve to

increase an abused child's chances of entering the second"

(1984, p. 255).

Reviewing the literature on family violence and neglect,

and its relationship with children's deviant behavior, Koski








(1988) reported a complex and often contradictory array of

findings. The research "remains largely descriptive and

piecemeal and leads to only one fairly obvious conclusion:

unhealthy families tend to produce deviant children, although

not always in the ways expected" (1988, p. 43). Koski also

said that research appears to suggest that a child's whole

family environment, where more than one type of abuse may

exist, may be more critical to understanding deviance and

delinquency than any single type of abuse.

The family environment includes not only the direct

physical abuse and neglect of a child by a parent discussed

above, but also violence involving other family members.

Straus and Gelles (1986) reported that cases of severe

violence by husbands toward their wives numbered 30 per 1,000

couples in a 1985 national survey. Controversially, they

also found that the rate of severe violence by wives toward

husbands was 44 per 1,000 couples. Straus and Gelles added

that a number of studies show that violence on the part of

wives is often committed in retaliation or self-defense, and

that this explanation is supported by the fact that females

are relatively nonviolent outside of the home. Some studies

have shown that a child's exposure to spousal violence as

well as violence committed by other family members is related

to the child's use of violence (Pagelow, 1984) and to later

family and nonfamily violence by the grown child (Fagan and

Wexler, 1987).








Other Family Relationships


While much research has focused on overt, physical abuse

and neglect, delinquent or other deviant behavior may result

from more routine experiences within the family setting.

Social groups such as families exert considerable influence

over the individuals within that group. "The relationships

of persons in a [family] nexus are characterized by enduring

and intensive face-to-face reciprocal influence on each

other's experience and behavior" (Laing and Esterson, 1971,

p.7). An individual may derive from others within their

group a sense of self-fulfillment or they may derive feelings

of alienation, even to the point of madness (Laing, 1962).

Laing and Esterson (1971), for example, make the case that

schizophrenia is socially based within dysfunctional family

relationships.

Satir (1972) distinguishes between "troubled" and

"nurturing" families. "Troubled families make troubled

people and thus contribute to crime, mental illness,

alcoholism, drug abuse, poverty, alienated youth, political

extremism, and many other social problems" (1972, p. 18).

The factors that determine whether a family is nurturing or

troubled--individual feelings of self-worth, communication

with each other, family rules about proper feelings and

behavior, and the way that the family's members relate to

other people and institutions outside the family--are largely

learned, primarily within the family, Satir said. The more








that self-worth is high, the more direct and honest the

communication, the more flexible the family's rules, and the

more open a family is to society, then the more likely a

family will be nurturing. These variables are important to

family quality whatever the nature of family structure--

whether it is a family with two biological parents, a single-

parent family, a family including stepparents, or an

institutional family not involving parents, according to

Satir (1972).

Snyder and Patterson (1987) describe positive parenting

as interactions between parent and child which foster growth

skills and encourage the development of normative values and

standards of behavior. They said these ends are fostered in

part "when the parents demonstrate and communicate interest

in the child in a positive, noncritical fashion, when the

parents communicate support and caring to the child, and when

the parents and child share mutually pleasurable leisure

activities" (1987, p. 223). According to Snyder and

Patterson, a variety of research suggests a link between the

absence of positive parenting and delinquent behavior. "A

failure by parents to foster a child's skills, to model and

encourage normative values, and to provide a caring

environment, places a child at risk for delinquency" (1987,

p. 225).

McCord (1991) found that a combination of family

variables which she labeled "poor child rearing" increased

the risk of officially recorded delinquency. This








combination included variables from three sets of items

measuring mother's competence, father's interaction with the

family, and family expectations. Of these three groupings,

McCord found that mother's competence and family expectations

were strongly related to serious criminality as a juvenile,

while father's interaction with the family had a more

important role in affecting adult criminality.

McCord's data consisted of case studies on male youths

from homogeneous socioeconomic backgrounds collected in the

Boston area in the 1940s and tracked in the late 1970s.

McCord's measures of mother's competency included self-

confidence in handling problems, affection for her son,

nonpunitive discipline, and family leadership role. Father's

interaction with the family included affection for son,

esteem for mother, esteem in the eyes of the mother.

Father's interaction also included, negatively, parental

conflict and father's aggressiveness. Finally, the family

expectations cluster included maternal restrictiveness,

parental supervision, and demands placed on the son.

In a meta-analysis of the research on families and

delinquency, Loeber and Stouthamer-Loeber (1986) reported a

significant relation between marital discord and children's

delinquency and aggression. The measures of marital

relations include ratings of happiness, amount of conflict,

lack of warmth, and hostility. Generally, Loeber and

Stouthamer-Loeber found that variables such as parent-child

involvement, parental supervision, and parental rejection had








the largest impact on crime and delinquency. They also

reported that many studies have shown that broken homes have

an impact on delinquency, although not as strongly as in the

case of marital discord.

Nevertheless, much research has focused on the question

of family structure. The following section will review the

research literature on the delinquent outcomes related to

broken homes and to family structure. Despite the large

amount of research in this area there remains some

inconclusiveness regarding the relationship between these

measures of family and delinquency.


Family Structure and Broken Homes


Research addressing family structure has often been

grouped under the label "broken homes." Thus, the two terms

often have been used interchangeably in previous literature.

In order to avoid confusion, a distinction should be made

between the terms family structure and broken homes. A

broken home is one in which a two-parent structure has been

reduced to a single-parent structure, whether by divorce,

separation, or death. Family structure refers more generally

to the number or type of parents living in a household.

Family structure may be measured by the presence of mothers

or fathers, and biological parents or stepparents. Thus, a

broken home more specifically describes a particular family

situation, compared to the broader description of family

structure.








As will be seen in the remainder of this chapter and in

the review of the research literature pertaining to social

learning theory in Chapter 3, many researchers have focused

on an increase in divorce and a growing number of single-

parent families as correlates of delinquency. The assumption

is that the disintegration of the family is a significant

cause of inadequate socialization, which in turn should have

an effect on the likelihood of delinquency. In fact,

however, family disintegration manifests itself in a variety

of forms which may be captured with varying success by the

dichotomy of "intact" and "broken" homes or by measures of

family structure. The first part of this section will

address the potential for confusion about the concepts of

broken homes and family structure, and the implications of

this distinction for measurement. Second, empirical research

on the relationship between family structure and delinquency

will be reviewed.


Conceptual and Methodological Issues


In the section above on family relationships it was seen

that "intact," or two-parent family structures, may be

variously integrated. "Broken," or single-parent families as

well are variously disintegrated, and perhaps not broken at

all. For example, single parenthood may result not simply

from divorce or other family conflict. There are single

parents, such as some who adopt, who have not at any time

been united with another as a couple and neither parent nor








children experience the stress of divorce. Furthermore,

family conflict may result in divorce, separation, or in no

recognizable change in family structure. Stepfamilies are

another source of complication for the study of families, as

they may vary in the quality of life they provide for

children.

According to Wells and Rankin, "Broken homes are

invariably defined from the reference point of children vis-

a-vis their relationship with parents, whereas sibling

structures are ignored" (1986, p. 70). A broken home refers

to an impaired or disrupted family structure. Broken homes

are typically measured as those in which the mother and

father are divorced (e.g. Furstenberg and Teitler, 1994;

Paternoster and Brame, 1997). Such a measure does not

necessarily distinguish the sex of a single parent. Using

single parenthood as a proxy for divorce would overlook the

alternative possibilities of separation, death of a parent,

or a never-married single parent.

Measuring divorce itself is suited to a focus on the

process and outcome of marital breakdown with the implication

that children suffer adverse consequences as a result of

their parents' divorce or separation. According to Wells and

Rankin (1986), the sudden change in living routines which

accompany divorce can generate stress and conflict that may

result in antisocial behavior. Thus, delinquency may result

not from the change in family structure but from "parental

conflict that precedes and attends the separation, and the








sudden change in circumstances affected by the separation"

(1986, p. 77). Furstenberg and Teitler (1994) suggested that

the quality of earlier family life may have significant

consequences for children whether parents separate or not.

They found that some family characteristics typically thought

to occur with divorce, such as marital conflict, economic

stress, and poor parenting practices, actually existed prior

to divorce.

The conditions associated with divorce that have an

influence on delinquency may only be temporary, existing

until individuals adapt to new routines. The results of

Furstenberg and Teitler (1994) suggest, however, that the

impact of divorce may be more long-lasting. They found that,

compared to their counterparts from intact families, children

of divorce were more likely to express discontent with life,

more likely to experience teen pregnancy, less likely to

finish high school and attend college, more likely to engage

in substance abuse and delinquency, and more likely to be

arrested.

In contrast to the focus on divorce, many researchers

alternatively look to the number of adults in the household,

hypothesizing that a change in family structure that results

in fewer parents will be associated with a decreased level of

social and economic resources and conventional socialization

compared to what would be expected from two-parent families.

Thus, single-parent homes are thought to be at a social and

economic disadvantage due to the reduced resources and








opportunities available to single, and typically female,

parents (Wells and Rankin, 1986; see also Smith and Krohn,

1995). In such cases, divorce leads to a reduction of

economic resources and perhaps to poverty which, through a

variety of pathways, may lead to delinquency (Curran and

Renzetti, 1996). For example, poorer families are more

vulnerable to stress and conflict that may result in less

effective parenting and child behavior problems. According

to Gove and Crutchfield, "Single-parent families tended to be

of lower socioeconomic status than the intact families, and

children of single-parent families were also more likely to

be delinquent" (1982, p. 307). In a review of social

psychological research pertaining to female-headed families

published between 1970 and 1980, Cashion (1982) found that

after controlling for socioeconomic status children from

female-headed families fared as well as children from two-

parent families, in terms of juvenile delinquency rates and

other measures.

Measures of family structure also may take into account

the presence of stepparents. Some researchers divide their

family structure measure into categories for two biological

parents, a biological parent combined with a stepparent, and

a single-parent (e.g. Coughlin and Vuchinich, 1996;

Cernkovich and Giordano, 1987). Another approach is to

elaborate on this typology by distinguishing each type of

parent by sex (e.g. Johnson, 1986). Another way of measuring

family structure is to simply count the number of parents in








the household, whether they are biological parents or

stepparents (e.g. Van Voorhis et al., 1988). This measure

differs significantly from the previously described measure

in that it focuses on the number of parents rather than on

the type of parents.


Research on Family Structure


The varying ways in which family structure may be

defined and measured suggest the difficulties that may be

encountered when dealing with the concept. In fact, the

empirical evidence on family structure and delinquency

reveals varied and sometimes contradictory findings. The

following two paragraphs illustrate the point.

In favor of a family structure effect, in their

nationally representative self-report survey of youths aged

12 to 17, Dornbusch et al. (1985) consistently found a

greater number of delinquents among mother-only households

compared to households with both biological parents. This

relationship was consistent for a variety of self-reported

and officially recorded delinquency, including law contacts,

as well as across sex and social class. Reviewing the

literature on family and delinquency, Gove and Crutchfield

(1982) cite overall findings consistent with Dornbusch et al.

"An intact home with harmonious marital and family

relationships will be closely associated with a lack of

delinquency and a lack of harmonious marital and family

relationships and/or a single-parent home will be associated








with delinquency" (1982, p. 304). More recently, Wells and

Rankin systematically analyzed 50 studies and concluded that

there is "a consistent and real pattern of association"

between broken home and delinquency variables, showing a

bivariate correlation of .10 to .15 (1991, p. 87). They said

this association is stronger for milder forms of juvenile

misconduct and weakest for serious and violent crimes.

On the other hand, in their study of high school

students in a large midwestern city, Hennessy et al. (1978)

found no effect of broken homes on a wide range of self-

reported delinquency, including being picked up by police and

appearances in court. They measured a broken home as any one

that did not include both a biological mother and a

biological father. While finding no broken-home effect, they

did suggest that other patterns of interactions within the

family may be related to delinquency. For example, Salts et

al. (1995) found in their study of males aged 12 to 19 that

family structure variables were not related to self-reported

violence, theft, or other delinquency. However, they found

that other family variables such as conflict and cohesion

were related to violence in the expected directions. Jensen

and Rojek concluded in their text, "At the present time there

is little conclusive evidence to suggest that the broken home

is a critical variable in the understanding of delinquency"

(1992, p. 270). In another text, Curran and Renzetti claim

there is "little evidence to support the broken home

hypothesis," while at the same time they acknowledge that








"research indicates that parents' relationships and

interactions with their children are correlated with the

children's involvement in delinquency" (1996, p. 349).

Consistent with Curran and Renzetti's contention and the

findings of Salts et al., many researchers have addressed the

complicated findings relating to broken homes and delinquency

by drawing a distinction between the number and type of

parents in a household (family structure), and the quality of

parenting in a household (often referred to as family

function). "It is not family structure in itself that

affects delinquency but the personal relationships within the

family unit which result from family disorganization in

single-parent families," (1984, p. 363) suggested Farnworth.

For example, Van Voorhis et al. (1988) compared the

effects of structure versus quality of relationships in the

family on self-reported delinquency. Using a sample of high

school students from a small midwestern town, they found in

both correlational analyses and multivariate tests that

family structure had weak effects for status offenses only,

while measures of home quality were more strongly related to

a variety of delinquency. Family structure was measured in

two ways: intact versus broken homes, and two-parent versus

single-parent homes. Family quality measures included scales

on supervision, affection, enjoyment of home, conflict, and

abuse of children.

Using data from the National Survey of Youth, a 1972

self-report survey of youths aged 11 to 18, Rankin and Kern








(1994) found that the likelihood of delinquency was the same

for all children who were strongly attached to one parent,

regardless of whether they lived in an intact or single-

parent family. They also found the likelihood of delinquency

to be lower for those living in intact families and attached

to both parents than for those living with and attached to a

single parent. Rankin and Kern suggested that single-parent

families may have higher rates of delinquency due to more

sporadic parental supervision, monitoring, and discipline.

Thus, measures of family quality may mediate family

structure. Attachment was measured as a composite of

intimate communication, family activities, supervision, and

affection. Rankin and Kern's results applied to a variety of

delinquency but not to property offenses such as theft and

vandalism.

In their study of youths aged 12 to 19 in a single

metropolitan area, Cernkovich and Giordano (1987) found that

family structure was unrelated to a wide range of self-

reported delinquency. Family quality variables, on the other

hand, were significantly related to delinquency. Their

measure of family structure was divided into responses for

living with both biological parents, with the mother only,

and with the mother and a stepfather. Significant family

quality variables were supervision or monitoring, identity

support, parental caring and trust, instrumental

communication, parental disapproval of peers, and conflict

with parents. Only intimate communication was not related to








delinquency. Cernkovich and Giordano said that their family

quality variables were similarly related to delinquency

across different types of family structure. "While this does

not necessarily mean that home status is an unimportant

variable in delinquency involvement, it does suggest that

similar family dynamics are operating within various types of

family structure" (1987, p. 312). Noting their finding of no

broken-home effect, Cernkovich and Giordano said their

results do not suggest that family interaction mediates

broken homes, as many have assumed.

Indirect and direct effects of family structure

Other research has substantiated that family structure

does have an effect on delinquency. A number of studies have

suggested that family quality variables mediate the

delinquent effects of family structure. Using a larger

sample from the Richmond Youth Project (a 1965 self-report

survey of junior and senior high school students in

California's Contra Costa County) Matsueda and Heimer (1987)

showed that the effect of family structure on delinquency may

be mediated by intervening variables that relate to family

function, or quality. Investigating general self-reported

delinquency, they found for both blacks and whites that

"broken homes influence delinquency by impeding the

transmission of antidelinquent definitions and increasing the

transmission of prodelinquent patterns" (1987, p. 835).

Their broken home variable measured the absence of either the

mother or the father from the respondent's household. As








mentioned above, in their reanalysis of the Gluecks' data

comparing delinquent and nondelinquent adolescent males, Laub

and Sampson (1988) found that significant effects exhibited

by parental criminality and drunkenness on serious

delinquency were largely mediated by family quality variables

including attachment, supervision, and erratic or threatening

parental discipline.

Using a longitudinal, nationally representative data

set, McLeod et al. (1994) also investigated the extent to

which family structure affects antisocial behavior.

Specifically, they found that single-parent motherhood (both

never-married and previously married) affected mother's use

of alcohol, use of physical punishment, and frequency of

affection. These variables in turn were related to

antisocial behavior on the part of the children, such as

disobedience, bullying, and destructiveness. While their

measures of family quality also affected delinquency, McLeod

et al. (1994) report that their family structure measures

directly affected antisocial behavior. Living with a never-

married mother led to antisocial outcomes for blacks while

for whites antisocial behavior depended on living with a

previously married mother.

Johnson's (1986) findings from a sample of high school

sophomores also support the direct effect of family structure

on delinquency. He reported that while white males with

stepfathers self-report significantly more delinquent

behavior than those from intact homes, attachment to a parent








did not mediate the relationship. Also, the relationship for

white females between mother-only family structure and self-

reported official trouble was not mediated by any variables,

including the quality of parent-child relationship. Findings

such as those by McLeod et al. (1994) and Johnson (1986)

suggest either that family structure, as measured by broken

homes or number of parents in the family, has some direct

relation to delinquency.

Stepparents

As we have seen, there has been an increasing focus on

variables measuring quality of family relationships and

delinquency. Other researchers have maintained their focus

on the impact of family structure, considering not just

mother-only or single-parent families but also stepfamilies

and other variations on the two-parent family. For example,

Johnson said that "family structure, independent of the

effects of the quality of parent-child relationships,

predicts certain measures of delinquency for certain

categories of adolescents" (1986, p. 78). He found that the

presence of a stepfather tended to increase self-reported

delinquency for males, while family structure in general made

little difference for female self-reported delinquency.

Coughlin and Vuchinich (1996) reported that males from

mother-only families and stepfamilies were both more than

twice as likely to be arrested than those from families with

two biological parents. Similarly, Dornbusch et al. (1985)

found similar levels of male self-reported delinquency and








self-reported arrests in stepfamilies and mother-only

families. For females, those from two-parent stepfamilies

were less delinquent than those from mother-only families,

although they were still more delinquent than those living

with both biological parents. For males, they found similar

levels of delinquency in stepfamilies and mother-only

families. The authors concluded that their data provided

little support for the belief that two adults in the

household are more likely to prevent adolescent deviance than

are single parents.

Given the findings reported above from Johnson (1986),

however, the similar levels of male delinquency in

stepfamilies and mother-only families may result from

different influences. That is, males in stepfamilies may

experience conflict with their stepfathers while males in

mother-only families may have less parental contact. In

their review of the literature on broken homes, Wells and

Rankin (1985) found results on the stepparent issue to be

inconsistent and inconclusive.

Additional research

In this section a variety of research will be reviewed,

including studies of the delinquent effects of living with a

single mother compared to living with a single father and

investigations of whether the timing or nature of parental

disruption affect delinquency. Also reviewed here is

research on the effects of sociodemographic variables in the

relationship between family structure and delinquency. The








section will conclude with a look at several aggregate-level

studies that address the question of how family structure is

related to delinquency.

Comparisons about the delinquent consequences of mother-

only families versus father-only families have been made

infrequently due to the relatively small percentage of

father-only families. Approximately 12 percent of the

children living in single-parent families live with a father,

and this figure was much lower a generation ago (U.S.

Department of Commerce, Bureau of the Census, 1994a). But in

their research review, Wells and Rankin concluded that "which

parent is absent makes no difference in predisposing to

delinquency" (1985, p. 260). However, Wells and Rankin noted

that some researchers have reported an interaction between

gender of parent and gender of child. That is, father-

absence may be more harmful for males while mother-absence

may be more harmful for females. Research in this area has

yielded mixed results and the issue remains unsettled.

Some researchers have examined the extent to which the

timing and the nature of family breakups are relevant to

delinquent outcomes. Mednick et al. (1987), dividing timing

of divorce into three levels based on the child's age, found

that it did not relate to number of arrests. West (1982)

found no difference in delinquency whether family disruption

occurred early or later in the child's life. Similarly,

using nationally representative, longitudinal data,

Furstenberg and Teitler (1994) reported no difference in the








effects of divorce whether parents separated early in the

child's life or later. Mednick et al. (1990), on the other

hand, did find that instability during adolescence was a

comparatively stronger predictor of young male crime than

instability earlier in childhood. The authors used a more

elaborate measure of stability by combining two scores on

family constellation changes, one score for prior to age 12

and the other score for age 12 and older.

Turning to the nature of parental disruption, it might

be reasoned that divorce or separation would have a greater

influence on delinquency than the death of a parent. This

might be the case because a divorce likely signifies conflict

and stress in family relationships. Otherwise, as some

research has suggested, the presence of stepparents may

result in a greater likelihood of conflict, especially

between sons and stepfathers. On the other hand, family

stress and the arrival of stepparents may also follow the

death of a parent. Stepparents aside, whatever the nature of

family disruption, at least a temporary result will be a

single-parent family. If delinquency is related to the

number of parents in the household, then one might expect the

nature of family disruption to have little effect on

delinquency. In their analysis of previous research, Wells

and Rankin reported, "The association with delinquency is

slightly (albeit not significantly) stronger for families

broken by divorce or separation than by death of a parent,"

(1991, p. 88).








Researchers have also examined the extent to which the

relationship between family structure and delinquency varies

by sex and race. Looking first at sex, Wells and Rankin

(1991) report that research generally shows no consistent

differences between males and females in the impact of family

structure. Wells and Rankin noted that their use of broken

homes as a measure of family structure was likely complicated

as a result of their research being an analysis of previous

studies, given the variety of ways by which broken homes may

be measured. In contrast to the findings of Wells and

Rankin, Gove and Crutchfield (1982) found that parental

marital status had little impact on delinquency for females

while males with unmarried parents were more likely to be

delinquent than those from two-parent homes. Gove and

Crutchfield's data differed from most self-report studies in

that they measured parents' reports of their children's

behavior. Using self-report data on youths aged 13 to 16,

Thomas et al. (1996) also found that the deleterious effects

of family structure were concentrated among males. Johnson

(1986) found family structure to be more significant for

males' self-reported delinquency as a result of the

stepfather relationship. For females, Johnson said family

structure did not appear to be as important. While girls

from mother-only families were more likely to be officially

delinquent, Johnson said this finding may reflect bias by

school officials and police.








Other researchers reporting sex differences locate these

differences within the mediating variables of family quality.

Among her small sample of blacks at age 15, Farnworth (1984)

found that while family structure had only a small impact on

delinquency for males and no effect for females, variables

based on parents' expectations and perceptions of the parent-

child relationship were significant for females, suggesting

that home socialization is more important for females than

males. Pointing to a similar conclusion in their nearly all-

white sample, van Voorhis et al. (1988) found that the

intervening effect of home quality in the relationship

between broken homes and delinquency was significant for

females across a wider variety of self-reported delinquency

than was the case for males.

Although Wells and Rankin (1991) reported that previous

research generally does not indicate racial differences in

the relationship between broken homes and delinquency, Salts

et al. (1995) found that family variables, including time

spent away from the home and with friends, predicted self-

reported violence among blacks and whites. For both racial

categories, family structure was not a significant predictor

of violence. Similarly, Smith and Krohn (1995) found in a

longitudinal study of urban teens that family structure was

not related to general self-reported delinquency for either

blacks or whites, although it did have a direct effect on

Hispanic delinquency. Overall, Smith and Krohn reported that

family measures, including family structure and family








quality, explained about 10 percent of the variance in

delinquency for both blacks and whites, and about 20 percent

for Hispanics.

However, Rosen (1985) concluded that both family quality

and family structure were related to official delinquency

among blacks, whereas only structural variables were

important for white delinquency. As mentioned previously, in

their nationally representative, longitudinal study, McLeod

et al. (1994) found that the particular measure of family

structure affecting antisocial behavior depended on race.

For blacks, antisocial outcomes were associated with living

with a never-married mother. For whites antisocial behavior

was related to living with a previously married mother.

However, Thomas et al. (1996) found a higher rate of black

self-reported delinquency among those in mother-only families

that included father involvement. Whites, on the other hand,

were more delinquent in mother-only families where there was

no father involvement. Delinquency was reduced for blacks

and whites living with two biological parents.

Matsueda and Heimer (1987) found for both blacks and

whites that broken homes influence delinquency indirectly

through delinquent definitions and delinquent peers, but the

relationship is considerably stronger for blacks. Smith and

Krohn (1995) found that parent-child attachment and the

perception of parental control directly affected delinquency

for blacks and whites, but they reported no indirect effects

involving family structure as an exogenous variable.








Generally, although greater differences were found for

Hispanics, the relationships between family variables and

delinquency appear to be similar for blacks and whites.

A variety of ecological research has been carried out on

the relationship between family structure and delinquency,

contributing additional support for the association between

family structure and delinquency. However, any inference of

causality for corresponding individual-level measures would

not be warranted due to the use of aggregate measures in this

research. In a series of studies, Sampson and various

associates have shown a connection between family structure

and delinquency while identifying antecedent variables that

pertain to the relationship, such as inequality, joblessness,

sex ratio, and racial segregation. Sampson (1985) found that

family structure, measured by the percentage of female-headed

families in a neighborhood, strongly affected serious

property and violent victimization. Analyzing data from 171

U.S. cities, Sampson (1987) and Messner and Sampson (1991)

found that official rates of serious violence were strongly

affected by the percentage of female-headed households with

children under age 18, and more strongly affected by

percentage of female-headed households. Smith and Jarjoura

(1988) found that neighborhoods with larger percentages of

single-parent households had higher victimization rates of

violent crime (robbery and assault) and burglary.

Additionally, family structure helped explain the significant

relationship between poverty and rates of burglary. Using








census data for 158 cities, Shihadeh and Steffensmeier (1994)

found that the percentage of female-headed black households

had a significant effect on officially reported black

violence. Female householding itself was affected by income

inequality within black communities.

These studies and others suggest a relationship between

family structure and delinquency using a variety of

variables, while pointing to various antecedent variables

that are associated with family structure. While much of the

research reviewed earlier positions family structure as an

exogenous variable that affects delinquency through various

mediating variables, this research hypothesizes a weakened

family structure as an endogenous variable intervening

between economic difficulties and crime (see Petras and

Davenport, 1991).


Conclusions


Wherever family structure is positioned in analytical

models, throughout the literature review it has been seen

that the broken-home measure of family structure is actually

an imperfect measure of a more general family process and

that the broken-home effect is an unclear one which interacts

with other variables and depends upon context (Gove and

Crutchfield, 1982; Hennessy et al., 1978). Such ambiguity

necessitates a precise definition of family measures such as

broken home and single parenthood, and the delineation of

potentially intervening variables that are affected by a








broken home or single parenthood and in turn produce

delinquent outcomes for children.

One cannot conclude from the existing literature that

any one measure of family structure is the best. The only

conclusion is that the measure used should be clearly defined

and appropriate for the purposes of the research. The

present study will employ a measure emphasizing the

structural makeup of the family rather than the occurrence or

conditions of divorce. This measure is chosen with the

assumption that two parents will be better able than one

parent to provide a consistent learning environment for

children.

For example, Dornbusch et al. (1985) found that

adolescents in mother-only households are more likely to be

delinquent as a result of making decisions without parental

input (see also McLanahan and Booth, 1989). Using a similar

measure of family structure, Rosen (1985) found significant

relationships with delinquency for both blacks and whites.

Gove and Crutchfield (1982) included a similar measure in

their study that found family structure to be more important

for male delinquency, while family interaction was more

central for females. But for their equivalent family

structure measure, Van Voorhis et al. (1988) found no

significant effects on a variety of delinquency measures.

McLeod et al. (1994) also investigated single-parent families

compared to two-parent families, distinguishing between

previously married and never-married parents. Both single-








parent family types were related to antisocial behavior.

Thomas et al. (1996) offered another approach to measuring

family structure by the number of parents present, comparing

families with two biological parents, single-mother families

that had nonresident father involvement in child

socialization, and single-mother families without nonresident

father involvement. White males were more delinquent the

lower the level of father's involvement, while black males

were more delinquent where nonresident fathers were involved

in the family.

Dornbusch et al. (1985) and other researchers have also

indicated that the presence of stepparents may increase the

likelihood of delinquency, especially in the case of

stepfathers and their stepsons. Johnson (1986) distinguished

between both number and type of parents, finding that males

living with a mother and a stepfather were most delinquent

while females in mother-only families were most delinquent.

Coughlin and Vuchinich (1996) found that youths from

stepfamilies and mother-only families were more likely to be

arrested. These findings call for the coded separation of

two biological parents from those two-parent families

consisting of one biological parent and one stepparent.

Much of the research reviewed in this chapter points to

the variables that intervene between family structure and

crime and delinquency. Many researchers have called for the

continued explication of these and other potentially

intervening variables (e.g. Smith and Jarjoura, 1988; Wells








and Rankin, 1986). According to Van Voorhis et al. (1988),

family structure usually is considered to be an indirect

cause of delinquency, predisposing family members to other,

more direct influences such as lack of affection, poor

supervision, shortage of economic and time resources, and

limited opportunities for modeling, among others. Wells and

Rankin said, "Few if any analyses presume ... a direct and

automatic causal link. Rather, family structure affects

delinquent behavior through intervening interactional

processes (engendered by structural conditions)" (1986, p.

87). While acknowledging that the relationship between

family structure and delinquency is a weak one, Wells and

Rankin (1985) argue that it is consistent and there is a need

to clarify the process by which the family relates to

delinquency.

The present study will focus on social learning as

intervening processes between family structure and

delinquency. Specifically, it draws upon the social learning

theoretical perspective to explain the effects of family

structure on delinquent behavior. Various components of

social learning theory will be hypothesized to mediate the

relationship between family structure and delinquency. In

the next chapter, social learning theory will be discussed

and additional research focusing specifically on the theory

will be reviewed.














CHAPTER 3
SOCIAL LEARNING THEORY OF CRIME AND DELINQUENCY


The last chapter discussed a variety of research

findings that link family variables to delinquent behavior.

In this chapter, social learning theory will be discussed and

considered as a useful approach to understanding the

relationship between the family and the delinquent behavior

of adolescents. The theory is well suited to explaining the

relationship between family variables, including family

structure, and delinquency because it focuses on the process

of interaction, particularly within primary groups. The

family normally is the first primary group that individuals

encounter. "Social learning models emphasize the importance

of parents as reinforcers and models of socially appropriate

behavior. Parental availability, supervision, and affection

are important factors in influencing the effectiveness of

their modeling and reinforcement strategies," (1988, p. 239)

according to Van Voorhis et al. Moreover, the family

typically has a considerable impact on other primary groups

that gain significance later, such as the peer group (Jensen

and Rojek, 1992). Thus, it is reasonable to expect that

changes in family structure may affect youths via social

learning processes in their interaction with parents, and

subsequently, with friends. The nature of the social








learning that occurs in these important primary groups may be

expected to significantly affect whether behavior is

conforming or delinquent.


Social Learning Theory


Social learning theory was developed by Akers and

Burgess (Burgess and Akers, 1966) as a reformulation of

Sutherland's (1947) theory of differential association.

Akers (1973; 1977; 1985; 1997; 1998) has continued to develop

and test the theory. Central to the theory is the idea that

delinquency is learned in interaction with others, primarily

within intimate, personal groups such as family and friends.

According to Sutherland, "A person becomes delinquent because

of an excess of definitions favorable to violation of law

over definitions unfavorable to violation of law" (1947, p.

6). Sutherland said that this process of differential

association varies by frequency, duration, priority, and

intensity. Thus, the commission of a criminal act results

from a complex ratio wherein the influence of criminal

patterns and definitions exceeds the influence of conforming

patterns.

Sutherland's theory was one of the first to suggest that

deviance results from the effects of environmental conditions

on biologically and psychologically normal individuals (Vold

and Bernard, 1986). According to Sutherland, "The process of

learning criminal behavior by association with criminal and

anti-criminal patterns involves all of the mechanisms that








are involved in any other learning" (1947, p. 7). If

criminal behavior is learned as is other behavior, then

Sutherland's theory of differential association also directs

us toward the social context or structure (differential

social organization as Sutherland would put it) in which

learning takes place for a further explanation of crime and

delinquency. Family structure would be one context that may

indirectly affect delinquency via the process of differential

association.

Burgess and Akers (1966) broadened Sutherland's

differential association by integrating it with behavioral

learning theory, particularly the concept of differential

reinforcement, whereby behavior is conditioned by rewards and

punishments. Their theory originally was called differential

association-reinforcement, but now is usually referred to as

social learning theory.

The theory is intended to apply to any form of deviant

behavior, including criminal and delinquent behavior. "The

basic assumption in social learning theory is that the same

learning process, operating in a context of social structure,

interaction, and situation, produces both conforming and

deviant behavior" (Akers, 1998, p. 50). Akers said that the

difference between conforming and deviant behavior lies in

the balance of influences on behavior. Social learning

theory focuses on four central concepts to account for these

influences. These are differential association, definitions,

imitation of models, and differential reinforcement. While








these four concepts influence conformity in one direction, in

his central proposition of social learning theory Akers

describes the way in which they combine to instigate and

strengthen deviance in the other direction:


The probability that persons will engage
in criminal and deviant behavior is
increased and the probability of their
conforming to the norm is decreased when
they differentially associate with others
who commit criminal behavior and espouse
definitions favorable to it, are
relatively more exposed in-person or
symbolically to salient criminal/deviant
models, define it as desirable or
justified in a situation discriminative
for the behavior, and have received in
the past and anticipate in the current or
future situation relatively greater
reward than punishment for the behavior.
(Akers, 1998, p. 50; emphasis added)

Rewards and punishment are the alternate ways in which

differential reinforcement is distributed. The four concepts

and other aspects of the theory will be explained in greater

detail below.


Differential Association


Differential association in social learning theory

encompasses interaction with others in intimate primary

groups, as well as in secondary groups and indirect

association with reference groups of individuals throughout

the community and the mass media. Modalities of association

combine to determine the effects of various associations.

"Those associations which occur first (priority), last longer

(duration), occur more frequently (frequency), and involve








others with whom one has the more important or closer

relationships (intensity) will have the greater effect"

(Akers, 1994, p. 96). In addition to this interactional

dimension there is a normative dimension of differential

association, which more widely refers to the different

patterns of norms and values to which individuals are

exposed.

"Differential association refers to the process whereby

one is exposed to normative definitions favorable or

unfavorable to illegal or law-abiding behavior" (Akers, 1994,

p. 96). Additionally, it is through differential association

that one is exposed to other mechanisms of social learning,

including models to imitate and differential reinforcement

for criminal or conforming behavior. These three concepts--

definitions, imitation, and differential reinforcement--will

be discussed next.


Definitions


Definitions are evaluative norms, attitudes, or

orientations regarding a particular behavior as good or bad.

A positive definition exists when behavior is considered good

while a negative definition exists when behavior is seen as

undesirable. A neutralizing definition exists when behavior

is justified or excusable. This type of definition includes

techniques of neutralization and rationalizations that are

used to disclaim responsibility or fault.








Conventional definitions about criminal behavior are

negative toward deviant behavior. However, the more one

holds ideas favorable to crime or deviance, the more one is

likely to engage in criminal or deviant behavior. In this

case the definitions about criminal behavior are positive.

Thus, the use of the terms positive and negative should not

be mistaken as synonymous with conforming and deviant,

respectively. For example, at the decisive moment a

conforming person may define conformity as positive and

deviance as negative while a nonconforming person may define

deviance as positive and conformity as negative. More

typically however, definitions favorable to crime and

delinquency are "conventional beliefs so weakly held that

they provide no restraint or are positive or neutralizing

attitudes that facilitate law violation in the right set of

circumstances" (Akers, 1994, p. 98).

Definitions are beliefs that may be general or specific.

General beliefs include religious, moral, and other

conventional ideas favorable to conformity and unfavorable to

deviance or criminality (Akers, 1994). Specific beliefs are

held about particular acts or types of acts. Thus, one may

believe that it is wrong to do harm to others but that it is

all right to harm those who are the source of frustration or

disagreement.

Definitions are learned through imitation and

reinforcement, and once learned, they serve as discriminative

stimuli for conforming or deviant behavior. According to








Akers, "Discriminative stimuli are stimuli which become

associated with reinforcement. In addition to the

reinforcers, other stimuli are ordinarily present when

behavior is reinforced--the physical surroundings, one's own

feelings, others' behavior, one's own and others' spoken

words, and so on" (1985, p. 49). Discriminative stimuli

occur before or as the behavior occurs, signaling the actor

as to the appropriateness of the behavior and whether

reinforcement or punishment is forthcoming.


Imitation


Imitation is the process of acquiring social behavior,

whether conforming or deviant, by observing and modeling the

behavior of others. Models, like definitions, may be drawn

from primary, secondary, or reference groups. Additionally,

imitation may occur immediately, at some later time, or not

at all. Imitation is a complex process as its likelihood of

occurrence depends on the characteristics of the model as

well as the model's observed behavior and the outcomes

produced by that behavior. "[Imitation] is more important in

the initial acquisition and performance of novel behavior

than in the maintenance or cessation of behavioral patterns

once established, but it continues to have some effect in

maintaining behavior" (Akers, 1994, p. 99).








Differential Reinforcement


The primary learning mechanism in social learning theory

is operant conditioning, or instrumental learning, wherein

behavior is shaped by the stimuli which follow the behavior.

Operant conditioning is distinguished from respondent

conditioning, wherein behavior is shaped by antecedent or

eliciting stimuli, in that it is "developed, maintained, and

strengthened depending on the feedback received or produced

from the environment" (Akers, 1985, p. 42). Thus, while a

respondent is controlled by the preceding stimulus and

unaffected by the outcome it creates for the environment, an

operant is controlled by the stimulating reinforcement and

punishment following or contingent on it.

Positive reinforcement occurs as rewards, such as

approving words, money, or pleasant feelings, increase the

probability that an act will occur or be repeated. The

likelihood of an act occurring or being repeated also is

increased through negative reinforcement, when one avoids

unpleasant consequences. For example, a motorist who exceeds

the speed limit will likely continue to do so, and perhaps to

a greater degree, so long as a citation is not received.

Punishment also may be positive or negative. Positive

punishment occurs when unpleasant consequences are attached

to a behavior, such as when the speeding motorist receives

that citation. Negative punishment occurs when pleasant








consequences are removed. For example, a worker receives

negative punishment when a paycheck is docked.

Deviant or conforming behavior may be continuously

reinforced but is more likely reinforced on an intermittent

schedule, the uncertainty of which helps account for the

variety, complexity, and stability of behavior.

Reinforcement depends on the values of those with whom one

interacts, as behavior is regarded differently across time

and place. Modalities of reinforcement--amount, frequency,

and probability--affect the extent of reinforcement. The

greater the amount of reinforcement for one's behavior, the

more frequently it is reinforced, and "the higher the

probability that behavior will be reinforced (as balanced

against alternative behavior), the greater the likelihood

that it will occur and be repeated" (Akers, 1994, p. 98).


Temporal Sequence and Reciprocal Effects in Social Learning
Theory


Social learning theory has most often been posited as

preceding delinquent behavior. However, Akers maintains that

the relationship between social learning variables and

delinquent behavior does not simply occur in one direction.

Akers stresses that "social learning is a complex process

with reciprocal and feedback effects" (1994, p. 99).

Akers (1998) explains how these reciprocal and feedback

effects are integrated into social learning theory. First,

association with family members and friends typically








precedes the commission of delinquent acts, and

reinforcement, modeling, and exposure to definitions occur

within these groups prior to the onset of delinquency.

Second, after the initial event imitation becomes less

important while the balance of the other social learning

variables continue to be used in determining repetition.

Third, actual consequences of the event (social and nonsocial

reinforcers and punishers) affect the probability and

frequency of repeated events. Consequences include the

actual effects of the behavior itself, the actual reactions

of others present at the time or who find out about it later,

and the anticipated reactions of others not present or

without knowledge of the behavior. Fourth, both the overt

behavior and the definitions favorable or unfavorable to it

are affected by these consequences, which in turn affect

further behavior. Definitions, for example, may be applied

retroactively as justification for a previous act. At the

same time, such justification precedes future acts as it

variably mitigates sanctioning (punishing reinforcement) by

others or oneself. Finally, progression into more frequent

or sustained behavior again occurs in the context of

reinforcement, modeling, and definitions.

Differential association shares a complex relationship

with delinquency, according to Akers (1998). Social learning

theory recognizes that youths may choose to associate with

friends based on a similarity in delinquent behavior that

already exists. Similarly, youths confined for delinquent








offenses will have little choice but to associate with

delinquent peers. However, because associations are

typically formed around circumstances and attractions such as

neighborhood proximity that have little to do with co-

involvement in deviant behavior, delinquent behavior most

often results from delinquent peer association (Akers, 1994).

Akers (1998) concludes that there is both a tendency for

persons to choose to interact with others based on behavioral

similarities (known as "flocking" or "selection") and a

tendency for persons who interact to have mutual influence on

each other's behavior (known as "feathering" or

"socialization"). Both are part of the social learning

process and both are explained by the same variables.


Social Structure and Social Learning: The Case of the Family
and Delinquency


Akers (1998) has presented a Social Structure-Social

Learning (SSSL) theory of crime and delinquency. The theory

basically assumes that social learning is the primary process

linking social structure to individual behavior. It

primarily proposes that "variations in the social structure,

culture, and locations of individuals and groups in the

social system explain variations in crime rates, principally

through their influence on differences among individuals on

the social learning variables" (Akers, 1998, p. 322). Social

structural variables that have an effect on individual

behavior include various broad measures of society and








community; common demographic variables such as age, sex,

class, and race; and other more immediate variables such as

family and peer group.

Social learning theory is useful in the explanation of

the effects of family structure on delinquency because the

family and peer groups are identified in the theory as the

most significant primary groups implicated in differential

association. Thus, the components of social learning theory

may be presented as intervening variables between the social

structural variable of family structure and the individual-

level variable of delinquent behavior. Differences in the

effects of family structure on rates of delinquency depend on

the extent to which each type of family is able to "provide

socialization, learning environments, and immediate

situations conducive to conformity or deviance" (Akers, 1994,

p. 101). For example, a single-parent family may be less

likely than a two-parent family to provide a consistent

learning environment, and a single parent may be less able

than two parents to oversee children's behavior. More

precisely, a two-parent family may offer greater

opportunities for differential association, the transmission

of values or definitions, the imitation of parental role

models, and differential reinforcement for conforming

behavior, compared to smaller family structures. The linking

of family structure to delinquent behavior, therefore, can be

seen as a special case of Akers' SSSL model. The present

study is one way of testing the SSSL model, as it will focus








on social learning theory as a link between social structure

and individual delinquent behavior by hypothesizing a

mediating effect of social learning variables on the

relationship between family structure and delinquency.


Research on Social Learning Theory


A variety of researchers writing on crime and

delinquency have cited social learning theory as a valuable

perspective for explaining deviance, and the findings of

numerous studies have supported or are consistent with the

theory (see Akers, 1998).


Social Learning Theory and Delinquency


Akers and various associates have demonstrated support

for the relationship between all four social learning

concepts and delinquency in a variety of studies, including

tests of social learning theory by itself and in comparison

with other theories (Akers, 1994).

Akers et al. (1979) provided a thorough test of the

explanatory power of social learning theory in relation to

self-reported alcohol and marijuana use and abuse among

adolescents. A combined 15 social learning variables

measuring all four concepts of social learning theory--

differential association, definitions, differential

reinforcement, and imitation--explained 55 percent of the

variance in use of alcohol and 68 percent of the variance in

use of marijuana in a sample of more than 3,000 adolescents.








For alcohol and marijuana abuse, the variance explained by

the social learning variables was 32 percent and 39 percent,

respectively.

Akers et al. (1979) distinguished between social

differential reinforcement (actual and anticipated) and a

second group of differential reinforcement measures which

combined social with nonsocial factors. Comparing the five

subsets of variables, differential association was most

effective in explaining variance in alcohol and marijuana use

and their abuse. Consistent with the findings of numerous

other studies, the most important single variable for

explaining both the use and abuse of alcohol and marijuana

was differential peer association. For substance use, the

second-most effective social learning concept was

definitions, followed by the combined social and nonsocial

subset of differential reinforcement measures, social

differential reinforcement, and imitation. A similar ranking

was found for abusers, although social-nonsocial

reinforcement was more important than definitions among

alcohol abusers. According to Akers et al., "In substance

abuse the user comes more and more to respond to direct

reinforcement, especially from the drug effects themselves"

(1979, p. 650).

Of the items measuring definitions, the respondent's own

approval or disapproval of use had the strongest effect on

both alcohol and drug use, and drug abuse, over neutralizing

definitions and attitudes about the law. For alcohol abuse,








attitudes about the law was more important. Of the items

measuring social differential reinforcement, friends'

rewarding or punishing reactions had a stronger effect on

substance use than parents' rewards or punishment and

deterrence items (being caught by parents or the police).

Friends' reactions was also important for substance abusers,

although marijuana abusers were equally concerned about the

interference it would have on other activities such as school

and athletics. Additionally, parental reactions was found to

be curvilinearly related to abuse, with a lower probability

of abuse being associated with parents who react moderately.

With the exception of imitation, four of the five

subsets explained a substantial proportion of variance in

both alcohol and marijuana use. The authors took these

findings as support for the theory as a whole. "Friends

provide social reinforcement or punishment for abstinence or

use, provide normative definitions of use and abstinence,

and, to a lesser extent, serve as admired models to imitate"

(1979, p. 644).

Akers and Lee (1996) found that social learning theory

explained adolescent smoking. Using five-year longitudinal

data, measures of differential association, definitions, and

differential reinforcement significantly affected future

smoking. Also consistent with social learning theory, the

authors identified reciprocal effects of smoking on

differential association, although similar effects were weak

or nonsignificant for definitions and differential








reinforcement. According to Akers and Lee, "As smoking

behavior develops, it is shaped by association with peers,

exposure to their normative definitions of smoking, and

social reinforcement, but then over time, one's own smoking

behavior comes to exert influence over patterns of

association with friends" (1996, p. 336). Menard and Elliott

(1994) identified reciprocal relationships between delinquent

peer association, attitudes toward delinquency, and self-

reported minor offending and index offending. However, the

stronger causal direction was clearly indicated by the

effects of social learning on delinquency.

Boeringer and Akers (1993) found that social learning

variables also explained violent behavior in the form of

sexual aggression among college males. They reported that

social learning models, including measures of all four of the

theory's central concepts, accounted for over 20 percent of

the variance in sexually aggressive behavior. This outcome

variable included nonphysical coercion, use of drugs or

alcohol as a coercive sexual strategy, and attempted or

completed rape with or without force. The social learning

models also accounted for 40 to 55 percent of the variance in

rape proclivity.

Additional support for social learning theory has come

from other researchers. In a sample of ninth-grade students,

Winfree et al. (1994) found that pro-gang attitudes

contributed significantly to the prediction of several self-

report crime measures, including alcohol and drug use,








property offenses, and violence. Analyzing eight waves of

the National Youth Survey, Elliott (1994) found that

delinquent peers, attitudes toward deviance and neutralizing

definitions (Elliott's "peer normlessness") all had strong

direct effects on the onset of violent offending.

Also using eight waves from the National Youth Survey,

Esbensen and Elliott (1994) found that social learning

variables accounted for the initiation of drug use. A

measure of differential association, exposure to drug-using

peers, was most strongly related to the initiation of

alcohol, marijuana, and other drugs such as amphetamines,

cocaine, and heroin. Differential reinforcement variables

also were significant in their findings. The anticipated

disapproval of friends decreased the likelihood of initiating

alcohol and marijuana use, while the anticipated disapproval

of parents decreased the likelihood of beginning alcohol use.

A similar measure of the anticipated disapproval of parents

was found to increase the likelihood of initiating alcohol

use, although it also slightly decreased the probability of

onset of marijuana and other drug use. Attitudes against

drug use, a measure of definitions, decreased the probability

of starting to use drugs such as heroin and cocaine. Another

measure of definitions, attitudes against delinquent

behavior, was not significant for drug use.

Esbensen and Elliott (1994) found that social learning

variables generally were not as significant in terms of the

termination of drug use, although exposure to drug using








peers reduced the likelihood of discontinuing the use of

marijuana and other drugs, and negative attitudes toward

delinquency increased the likelihood of termination of

marijuana and other drugs. In summary, Esbensen and

Elliott's results indicate variably significant effects on

drug use for a variety of social learning concepts, including

differential association, definitions, and differential

reinforcement. Measures of neutralizing definitions were not

significant.

Using two waves of the National Youth Survey, Warr and

Stafford (1991) found that differential association, as

measured by friends' delinquency, was strongly related to

self-reported cheating, petty larceny, and marijuana use.

Friends' delinquency also affected self-reported delinquency

indirectly through respondents' definitions. Friends'

attitudes, an indirect measure of differential reinforcement,

indirectly affected delinquency through respondents'

attitudes (definitions), but the direct relationships between

friends' attitudes and delinquency were weak or

nonsignificant.

Elliott et al. (1989), using the first three waves of

the National Youth Survey, found involvement with delinquent

peers to have the strongest effect on both general

delinquency and index offending rates. Other theoretical

variables affected both types of delinquency only indirectly

through involvement with delinquent peers. Attitudes toward

delinquency and neutralizing definitions ("normlessness" in








their usage), for example, were found to affect involvement

with delinquent peers but had no direct effect on

delinquency. Elliott et al. (1989) concluded in the test of

their integrated theoretical model that learning theory

provides the principal explanation for delinquency and

substance use. Using waves three through five of the NYS,

Menard and Elliott (1994) also found delinquent peer

association to significantly affect minor offending and index

offending, while the effects of attitudes toward delinquency

were weak and inconsistent.

In this section it has been shown with a variety of

cross-sectional and longitudinal research that social

learning variables are related to various delinquent

outcomes, including smoking, alcohol and drug use, general

delinquency and violent behavior. Differential (peer)

association, typically measured as having delinquent or

substance using friends, emerges as having the clearest

relationship with delinquency of the four social learning

concepts. In nearly every study reviewed, friends'

delinquency or substance use was most strongly related to

one's own delinquency or substance use. Indeed, in their

review of the empirical literature on broken homes and

delinquency, Wells and Rankin note, "Peer influence

constitutes the variable most consistently and strongly

predictive of delinquency across a variety of studies" (1985,

p. 265).








Various measures of definitions and differential

reinforcement also indicate direct effects on delinquent

behavior. A few studies suggest that some social learning

variables affect delinquency indirectly through other social

learning variables. For example, Elliott et al. (1989) found

that measures of definitions, including neutralizing

definitions, affected delinquent peer association. For Warr

and Stafford (1991), however, delinquent peer association and

the differential reinforcement of peers (friends' attitudes)

affected one's own definitions.


Family, Social Learning Theory, and Delinquency


Research has demonstrated the intervening effects of

social learning variables between various structural

variables and delinquency. For example, Krohn et al. (1984)

found that social learning variables explained much of the

variance in self-reported alcohol and marijuana use across

urban and rural community types. Akers and Lee (1997)

reported that social learning variables explained the effects

of age for adolescent marijuana users. Boeringer et al.

(1991) found that social learning variables mediated the

effects of fraternity membership on sexually aggressive

behavior among college males. In this section research will

be reviewed which links family variables, including family

structure, to social learning variables and delinquency.

Thereby it will be shown how social learning provides the

process through which the social structure of the family has








an impact on individual behavior. "As the main conventional

socializer against delinquency and crime, the family provides

anti-criminal definitions, conforming models, and

reinforcement for conformity through parental discipline,"

according to Akers (1998, p. 165). Later, he said, family

socialization remains important as it competes against or is

undergirded by socialization in peer groups.

Consistent with the findings reported above on the

primacy of differential peer association, Elliott et al.

(1985) found delinquent peer association to have a strong

effect on self-reported serious and nonserious delinquency.

Using the first three waves of the National Youth Survey, the

authors also found that weak and antisocial relationships

with family members and friends contributed to an increased

involvement with delinquent peers. Definitions, measured as

attitudes toward deviance, had little effect on delinquency.

Neutralizing definitions did not have a direct effect on

delinquency although the variable was associated with an

increased involvement with delinquent peers.

Heimer (1997), also using the first three waves of the

National Youth Survey, found that definitions relating to

interpersonal violence were associated with subsequent

violent delinquency among males. Definitions mediated the

delinquent effect of concurrent association with delinquent

peers, which mediated the effect of prior parental

supervision of friendships. Patterson and Dishion (1985)

likewise suggested that parents' monitoring practices








contribute to the parents' failure to discipline antisocial

or delinquent behavior and to their lack of control over a

child's association with delinquent peers. Applying social

learning principles with an emphasis on the roles that

parental reinforcement and punishment play in socialization,

Patterson and Dishion found that parental monitoring had a

direct effect in reducing male delinquency, as well as an

indirect effect by ways of delinquent peer association and

social skills. According to Patterson and Dishion, "The

parents' failure to track provides ample opportunity for the

child to engage in delinquent acts and to seek out deviant

peers, who in turn further exacerbate the problem" (1985, p.

75).

Warr (1993) also reported a strong relationship between

association with delinquent peers and self-reported

delinquency. His findings showed that the strength of this

relationship was significantly reduced by the quantity of

time spent with one's family. Compared with what has been

seen to be the typical sequencing, in this case the variables

representing family and social learning are inversely

ordered. Warr concluded that peer influence may be overcome

only by avoiding the company of delinquent peers altogether.

"This may be achieved either by inhibiting the formation of

delinquent friendships in the first place (as attachment to

parents seems to do) or by reducing the time that adolescents

spend with their delinquent friends" (1993, p. 259).








The studies reviewed to this point have employed family

variables relating to family quality. A few studies have

investigated the relationships between family structure,

social learning, and delinquency. In a cross-sectional test

of differential association theory using Richmond Youth

Project data, Matsueda and Heimer (1987) found that

delinquent peer association had the largest total effect on

self-reported male delinquency. Their findings showed that

broken homes, measured as mother or father absence, influence

delinquency by directly fostering an excess of prodelinquent

definitions. Matsueda and Heimer said that broken homes also

influence delinquency more indirectly "by attenuating

parental supervision, which in turn increases delinquent

companions, prodelinquent definitions, and ultimately,

delinquent behavior" (1987, p. 836).

Heimer and Matsueda (1994) found that delinquent peer

association and attitudes toward delinquency both were

related to self-reported delinquency among males. In a

longitudinal analysis using the first three waves of the

National Youth Survey (NYS), they reported that these social

learning variables (at wave two) mediated the effect of

family structure (measured as father absence at wave one) on

serious and nonserious delinquency (at wave three). Heimer

(1997), also using the first three waves of the NYS, found

support for the mediating effects of social learning on male

violence. Her findings indicated that broken homes (those

without both biological parents present) were related to less








supervision, which led to an increase in subsequent

delinquent peer association and then violent definitions,

which finally led to subsequent violent delinquency.

Using longitudinal data from Houston-area junior high

schools, Chen and Kaplan (1997) found that delinquent peer

association and definitions, including attitudes toward

deviance and neutralizing definitions, mediated the effect of

family structure on self-reported delinquency. Measured as

having both biological parents present, family structure also

maintained a modest, direct effect on delinquency. Chen and

Kaplan reported that separate analyses yielded similar

results for both serious delinquency and general delinquency,

which included drug use, theft, and violence. A unique

feature of their research was the eight to 16 years that

elapsed between their measurement of the independent

variables and delinquency. This arrangement permitted the

investigation of the effects of family structure and the

mediating variables in early adolescence on deviance in young

adulthood.

The literature reviewed in this section has

demonstrated, both cross-sectionally and longitudinally, the

mediating effects of social learning variables between family

variables, including family structure, and delinquency.

Again, delinquent peer association emerges as the dominant

social learning variable, playing a significant mediating

role in each of the studies reviewed. Definitions also

played a significant intervening role in about half of these





65

studies, sometimes mediating the effect of delinquent peer

association on delinquency. The self-reported delinquency

measures in these studies included nonserious and serious

general delinquency, and in one case, violent delinquency.


Hypotheses


Based on the empirical findings of the literature

reviewed to this point and the reasoning of social learning

theory, hypotheses may be deduced about the concepts to be

investigated. Below are four hypotheses specifying the

expected relationships between family structure, social

learning, and delinquency.

(1) Research has demonstrated the importance of families

as socializing agents both in the initial stages of life and

continuing through childhood, adolescence and into adulthood.

Although less clear, research also has suggested that a

reduced number of adult socializers increases the likelihood

of a reduced quality of socialization for a child. This

relationship is suggested by empirical evidence showing that

children from single-parent homes are more delinquent than

children from two-parent homes.

Therefore it is first hypothesized that family structure

predicts delinquency. Family structure as used here refers

to the number and type of parents in the household, as

contrasted to measures focusing on the occurrence of a

divorce or the absence of a particular parent. Specifically,

youths living with a single parent are expected to be more








delinquent than youths living with two parents. It also is

expected that youths living with neither parent in a non-

parental household will be more delinquent than youths living

with a single parent. Additionally, youths living with one

biological parent and one stepparent are predicted to have

delinquency levels between those of youths living with two

biological parents and those living with one parent.

(2) The literature review has shown that family

structure has an impact on social learning variables. For

example, Heimer and Matsueda (1994) found that father absence

increased the level of definitions favorable to delinquency

and increased the extent of associations with delinquent

peers. Other studies suggested that a reduced number of

parents resulted in reduced supervision, which resulted in an

increase in association with delinquent peers and the

adaptation of delinquent attitudes. Social learning theory

recognizes the family as an important source of both

conforming and deviant reinforcement, models, attitudes, and

association. The theory would suggest that living with fewer

parents would mean a weakened parental association with

conforming patterns, the absence of parental reinforcement,

the reduction of parental influence, and the loss of parental

role models. Thus, single parents will be less likely than

two parents to instill conforming attitudes and discourage

delinquent ones, guide their children toward conforming

associations and away from deviant ones, and to reinforce

conformity and punish delinquency.








Accordingly, it is hypothesized that family structure

affects the social learning process conducive to delinquency.

It is expected that having fewer biological parents will be

correlated with scores on social learning measures in the

delinquent direction, such as having more delinquent friends

and adhering to delinquent attitudes.

(3) Perhaps the strongest of the research findings

regarding social learning theory was that delinquent peer

association leads to delinquent behavior. Research also

indicated that having prodelinquent definitions increases the

likelihood of delinquency. Social learning theory holds that

one who has an excess of delinquent associations, adheres to

an excess of prodelinquent definitions, is reinforced toward

delinquent behavior and away from conformity, and has

delinquent role models will be more likely to engage in

delinquency.

Specifically, it is hypothesized that the more

delinquent one's friends are, and the more delinquent friends

one has, the more likely the respondent will also be

delinquent. Similarly, the more delinquent definitions one

possesses, the more likely he or she will be to engage in

delinquent behavior. Moreover, the weaker the punishment for

delinquent behavior, and the stronger the reinforcement for

it, the greater the likelihood of delinquency.

(4) Akers proposes that social learning theory is

capable of explaining how social structures such as the

family shape individual behavior (1997; 1998). The last








section of the preceding literature review cited several

studies that demonstrated the mediating role of social

learning theory. In particular, differential peer

association and definitions were shown to be important

mediating concepts between family structure and delinquency.

Therefore the fourth and final hypothesis, and the one

most central to the present research, is that social learning

variables will mediate the effect of family structure on

delinquency. Specifically, the introduction of social

learning variables into the analysis will reduce the strength

and significance of the relationship between family structure

and delinquency. It is expected that an attenuated family

structure will be associated with an increase in delinquent

friends, an increase in delinquent attitudes, and an increase

in rewards for delinquency along with a decrease in

punishment for delinquency. It is expected that these

conditions of social learning will be associated with an

increase in delinquent behavior, and will substantially

reduce or render nonsignificant the relationship between

family structure and delinquency.


The Present Research


The intent of the present research is to further

investigate the mediating effects of social learning between

family structure and delinquency. This will be accomplished

by expanding upon previous research in a variety of ways.

Where social learning variables have been included in








previous research on this topic, they typically have appeared

one or two at a time and usually as part of another

theoretical construction. As an explicit test of social

learning theory, this study will employ six social learning

measures of three of the theory's four central components to

provide a test of more complete social learning/social

structure models.

Secondly, most of the previous research in this area has

combined nonserious and serious offenses into a general

delinquency measure. Indeed, measures of serious delinquency

are much less common in self-report studies than nonserious

or general delinquency measures. One exception is the

research by Heimer (1997), reviewed above, that focuses on

violence. The present study will employ an index of serious

delinquency combining serious property and violent offenses.

Thirdly, the present study will expand on previous

research by using a different measure of family structure.

Previous studies in this area have used dichotomous measures

based on the presence of two biological parents or on father

absence. Here family structure will be based on the number

of parental figures in the household, be they biological or

stepparents. The distinction is noteworthy because this type

of measure differentiates between single-mother families and

two-parent, mother-stepfather families. Given the importance

of parental monitoring or supervision suggested by several of

the studies reviewed in the previous section, and a presumed

difference between the monitoring ability of a single parent






70

versus two parents, the number of parental figures in the

household may have significant consequences for social

learning and delinquency.

Finally, the present study differs from previous

research by using more recent longitudinal data on males and

females. While a few of the studies cited above employed the

first three waves of the National Youth Survey, this study

will use waves three through five of the NYS.














CHAPTER 4
METHODOLOGY


Data


The purpose of the present study is to conduct a

nationally representative, longitudinal test of social

learning as a mediating process between family structure and

serious, self-reported delinquency. The study employs wave

one (1976) and waves three through five (1978, 1979, 1980) of

the National Youth Survey (Elliott et al., 1985; Elliott et

al., 1989). Hereafter, waves three through five will be

referred to as years one through three, respectively. Use of

the first wave, 1976, is limited and separately identified

(see below).

In addition to the extensive work of Elliott and his

associates, various waves of the NYS data set have been used

by different researchers. These include Heimer (1997),

Paternoster and Brame (1997), Heimer and Matsueda (1994),

Lauritsen (1993), Warr (1993), and Warr and Stafford (1991).

For the present research, tape data on the NYS are obtained

through the Inter-University Consortium for Political and

Social Research (ICPSR).

The NYS uses a panel design with a national probability

sample of households in the continental United States based








upon self-weighting, multistage, cluster sampling. The

sample contained 2,360 eligible youth aged 11-17 in 1976, the

year of the initial wave. Of those eligible, 1,725 (73%)

agreed to participate in the study and completed interviews,

which were conducted for each year in the first three months

of the following year. Elliott et al. (1989) report that the

representativeness of the sample with respect to reported

delinquency and other variables was not significantly

affected by nonparticipating youth or by panel attrition.

The age range for the present study is 13-19, as

determined by the waves selected and as a result of excluding

those respondents aged 18 and over at the time of family

structure measurement in year one or year two. The dependent

variable of delinquent behavior is variously measured in year

two (1979) and year three (1980). The effective sample size

for the present study ranges from 1,020 to 1,200 (see Table

3, below). There are roughly 250 respondents at each age.

Further details regarding the design of the present study and

sampling distributions will be presented below.

Elliott and Ageton (1980) note that self-report measures

and official statistics typically offer different findings on

delinquent behavior and its social correlates. While both

types of data reveal significant age and sex differences, the

magnitude of these differences is often smaller with self-

report data than with official statistics. Also, self-report

studies generally have found fewer differences in delinquent








behavior by race or class than are found in studies of

official delinquency.

Based on these differences in findings, self-report data

have long been considered useful for eliminating the

potential bias of official statistics wherein minority or

lower-class youths are more likely to be labeled delinquent

(see Hirschi, 1969). Likewise, youths from broken homes have

been found to be overrepresented in official statistics

(Jensen and Rojek, 1992; Wells and Rankin, 1985) and it may

be that these reflect systematic disparities in official

decisions about youth from broken homes versus those from

intact homes. Thus, the self-report survey would be a more

valid approach to the study of the relationship between

family structure and delinquency.

However, Elliott and Ageton (1980), using wave one of

the NYS, found significant race and class differences with

self-reported general delinquency. They attributed these

findings to the extended frequency range used on their

delinquency measures. Hindelang, Hirschi, and Weis (1979)

similarly suggest that the discrepancy between official

statistics and self-report surveys, with respect to sex,

race, and class, is largely illusory, at least for the

relatively minor types of delinquency typically found in

self-report data. They concluded that the apparent

differences between self-report data and official statistics

are a result of these methods effectively measuring different

behavior. Addressing these claims, Akers et al. (1981)








distinguished between serious and nonserious self-reported

delinquency, and between low- and high-frequency offenders.

Their findings suggested that previous self-report studies

were correct in their findings of nonsignificant differences

in delinquency across sociodemographic variables such as race

and class.

Whatever the extent of discrepancy between the two

methods of data collection, one inherent strength of self-

report surveys is their ability to avoid the potential bias

of which official statistics may be susceptible. They also

are useful for the measurement of "hidden" crime and

delinquency that escapes the notice of the justice system

altogether and therefore does not show up in police reports.

Self-reports also permit the avoidance of a self-selective

sample of all delinquent youths wherein the dependent

variable has little variance. On the other hand, self-report

surveys have been criticized for a number of weaknesses

involving issues of reliability and validity.

With the National Youth Survey, Elliott and his

associates attempted to overcome the weaknesses common to

self-report data. Elliott et al. (1989) suggest that while

self-report measures generally appear to be reasonably valid

and reliable, the reporting of trivial events may be the

greatest threat to the validity of self-report surveys.

Typically a general delinquency measure will include many

trivial acts that result in the unrepresentativeness of the

measure. The NYS includes serious index offenses of property








and violent delinquency and these are the exclusive domain of

this study. The reporting of trivial events is likely to be

a smaller problem when measuring strictly the most serious

offenses.

Another threat to the validity of self-reports is recall

error. The NYS has recall periods of one year with the

Christmas holiday providing specific, memorable reference

points for the respondents. Additionally, the serious

delinquency that is the focus of this study may be better

remembered than less serious acts. The year-long periods

also eliminate the possibility of seasonal variation.

Another threat to validity is deliberate falsification.

Elliott et al. (1989) report that much evidence on self-

report measures indicates that this is a relatively rare

problem. Akers et al. (1983), for example, used a

biochemical measure of smoking behavior to confirm that

respondents reported their smoking quite accurately. Given

the present focus on serious delinquency, deliberate

falsification may reasonably be considered a greater

potential problem. The NYS personal interviews typically

took place in the privacy of the respondents' homes. Privacy

and confidentiality were protected and guaranteed by official

certificates from the United States government.

A variety of methods were used to analyze the validity

of delinquency scales from wave one through six of the NYS

(Elliott et al., 1989). These included an analysis of

detailed follow-up questions, an analysis of the proportion








of trivial responses that did not qualify as delinquency, and

a comparison of self-reported delinquency with arrest

records. Each of the methods used generally confirmed the

validity of the delinquency scales. These scales included a

variety of item groupings based on type and seriousness of

offense. The felony theft scale and the felony assault

scale, comprised of a combined seven offenses, contained five

of the six serious delinquency measures included in the

present study.

The reliability of self-report measures is generally

well established (Elliott et al., 1989; Hindelang, Hirschi,

and Weis, 1979). Using the test-retest method of reliability

assessment on wave five (1980) of the NYS, Huizinga and

Elliott (1986) reported the highest reliability for index

offenses. Elliott et al. reported that for their delinquency

scales from waves one through six of the NYS, "reliabilities

were adequate to excellent for the total sample and all

demographic subgroups (e.g., test-retest reliabilities ranged

from .7 to .9 in the total sample)" (1989, p. 15).



Measurement of Variables



Delinquency Index


The dependent variable is self-reported delinquent

behavior (SRD), combining self-reports of violent and

property offenses. The violent offenses include personal








attack, sexual assault, and use of force on another person.

The property offenses include motor vehicle theft, theft

involving a value of $50 or more, and breaking into a

building or vehicle. The delinquency index is a three-point

scale coded as follows: no reported property or violent

offenses (0); at least one reported occurrence of either

property or violent delinquency (1); and at least one

occurrence of both property delinquency and violent

delinquency (2). See Appendix A for a complete description

of the delinquency items. Alternatively coding delinquency

0-6, representing the total number of delinquent activities

engaged in by respondents, produced no more variation in the

dependent variable and yielded a weaker correlation with

family structure. Subsequent analyses indicated that

findings were not significantly different for this variation

of the dependent variable compared to the three-category,

ordinal coding described above and employed in the analyses

reported below.

The scale used here as a measure of serious delinquency

follows the constructions of Elliott and other researchers

employing NYS data. For example, the three items measuring

property delinquency were included by Elliott and Huizinga

(1983, p. 156) in their "felony theft" category, while the

three items measuring violent delinquency were used by

Elliott (1994, p. 3) to characterize "serious violent

offenders."








Family Structure


The primary independent variable is the respondent's

family structure, using a measure focusing on the parental

makeup of his or her household. Recall from the literature

review that a measure of family structure stands in contrast

to measuring the occurrence or effects of divorce.

Furthermore, the particular way in which the family structure

measure will be coded here looks at not simply father absence

but at the number of parents living in the household.

Responses to the question, "With whom are you now

living?" (LIV) were used to measure family structure (see the

question and response categories in Appendix A). The

responses were collapsed into three categories: (1) living

with at least two parents, either two biological parents or

one biological parent and one stepparent (original codings 1,

4, 5, 14, and 15; see Appendix A); (2) living with a single-

parent, either male or female (original codings 2 and 3); and

(3) living with neither parent (original codings 7, 8, 9, and

16). This third category includes living with a boyfriend or

girlfriend, living alone, living with a roommate, or living

with nonparental relatives. Several response categories,

such as "mother and other relativess)" were deleted in order

to maintain clear distinctions between the recorded

categories. Deleted original codings include 6, 10, 11, 12,

13, and 17. Age is limited to 17 and younger at the time

family structure is measured (usually in year one, but in








year two for Model 2) so that the measure would not be

complicated by respondents moving away from home at age 18.

These three categories of living with two parents, with

one parent, and living with no parents are coded and treated

as a three-point ordinal scale of an underlying continuum of

family structure ascending from the most complete, parent-

intact family to the most incomplete, parent-absent family.

Initially this was intended as a four-point scale

differentiating between two-parent family living arrangements

with two biological parents and those with one biological

parent and one stepparent. However, analysis showed that

there were no differences in delinquency between youth living

with two biological parents and those living with a

biological and a stepparent (r=.05, p=.16). Further,

consideration was given to whether the sex of the parent in

single-parent families would make for any difference in

delinquent behavior by the children. Consistent with

previous research (Wells and Rankin, 1985), no difference in

delinquency for those respondents living with only a father

compared to those living with only a mother (r=-.02, p=.77)

was found in the NYS data. Thus, separate categories for sex

of single parent were not included in further analyses.

There were too few cases in these data to investigate the

effects of male or female respondents living with a single

parent of the same sex compared to that of a different sex.

But the data did show that of those living with a single








parent, 85 percent of males lived with their mother and 92

percent of females lived with their mother.

In a longitudinal analysis spanning three years, any

change in family structure in the second year may presumably

have a significant impact on delinquent outcomes in the third

year. However, preliminary correlational analyses determined

that in these data there were few changes in family structure

in the time period included. Family structure at year one

was strongly correlated with family structure at year two

(r=.78, p<.01) and at year three (r=.71, p<.01), suggesting

little change in the family structure variable over time.

Moreover, comparisons were made of the effect on

delinquency of whatever changes in family structure did

occur. A variable, change in family structure, was created

based on the change in the number of parents in the household

from year one to year three. Change in family structure was

coded as one=increase in number of parents in the household,

2=no change in number of parents in the household, and

3=decrease in number of parents in the household. This

variable was unrelated to subsequent delinquent behavior

(r=.006, p=.89).


Social Learning Variables


Six measures are employed to represent three major

concepts in social learning theory: differential

reinforcement, definitions, and differential association.

The fourth major component of social learning theory,








imitation or modeling, is not included in this study due to

the absence of an appropriate measure in the National Youth

Survey. See Appendix A for a listing of the NYS items used

here as social learning variables.

Differential reinforcement

Recall that differential reinforcement refers to the

balance of rewards and punishment that is received in

response to behavior, and it depends on the values of those

with whom one interacts. The two groups of people with whom

youths most often interact are their parents and their peers.

Thus, one variable will be used to measure the differential

reinforcement of each of these two groups. The measures tap

the reaction that the respondent anticipates from their

parent or peer in response to given delinquent acts, be that

reaction approving (rewarding) or disapproving (punishing).

Variables based on the same measures from the National Youth

Survey, although with varying scale items, were used by

Esbensen and Elliott (1994).

The two reinforcement variables are: the anticipated

parental reactions for one's own delinquency (APR) and the

anticipated friends' reactions for one's own delinquency

(AFR). Both variables are measured on five-point Likert

scales with 1 being strong disapproval and 5 being strong

approval. The APR index is an eight-item scale, measuring

the respondent's anticipated parental reactions to the

following acts: stealing something worth less than $5,

selling hard drugs, using marijuana, stealing something worth








more than $50, hitting or threatening to hit someone without

any reason, using alcohol, purposely damaging others'

property, and breaking into a vehicle or building to steal

something (alpha = .85). The AFR index includes nine items,

measuring the respondent's anticipated reactions of friends

to the following acts: stealing something worth less than

$5, selling hard drugs, using marijuana, stealing something

worth more than $50, hitting or threatening to hit someone

without any reason, using alcohol, pressuring or forcing

someone sexually, purposely damaging others' property, and

breaking into a vehicle or building to steal something (alpha

= .90).

Items measuring actual consequences experienced by

youths committing delinquent behavior are not available in

the NYS data set. However, both APR and AFR are adequate

measures of differential reinforcement in that they measure

anticipated rewards or punishments that would follow from

delinquent acts. The more a respondent anticipates, for

example, that breaking into a vehicle or stealing something

will be met with tolerance or even approval by parents and

peers (rewards), then the more likely the respondent will be

to commit those acts. The more the respondent anticipates

parental or peer disapproval (punishment) for those acts, the

more likely the respondent will be to refrain from

delinquency.








Definitions

In the last chapter definitions were defined as

evaluative norms, attitudes, or orientations regarding a

particular behavior as good or bad. A definition is positive

or negative, depending on whether behavior is seen as

appropriate or undesirable, respectively. Alternatively, a

neutralizing definition exists when behavior is considered

justified or excusable. The respondent's attitudes toward

delinquent acts (ATT) and the respondent's attitudes toward

interpersonal violence (IPV) measures one's positive/negative

definitions. Similar variables based on the same measures

from the National Youth Survey were used in Heimer (1997),

Heimer and Matsueda (1994), Matsueda and Heimer (1987), and

Elliott et al. (1985). The respondent's endorsement of

deviant neutralizations (EDN) measures neutralizing

definitions. Similar variables based on this measure were

used by Esbensen and Elliott (1994) and Elliott et al.

(1985). (See Appendix A for a complete description of all

three measures of definitions.)

Attitudes toward delinquent acts are measured on a four-

point Likert scale with 1 being "very wrong" and 4 being "not

wrong at all," while IPV is measured on a five-point Likert

scale with 1 being strong disagreement and 5 being strong

agreement. Attitudes toward deviant neutralizations also is

measured on a five-point Likert scale with 1 being strong

disagreement and 5 being strong agreement.








The ATT index includes ten items: purposely damaging

others' property, using marijuana, stealing something worth

less than $5, hitting or threatening to hit someone without

any reason, using alcohol, breaking into a vehicle or

building to steal something, selling hard drugs, stealing

something worth more than $50, getting drunk, and giving or

selling alcohol to minors (alpha = .89).

The IPV index includes six items: it is all right to

beat up people if they started the fight, it is all right to

beat up people who call you names, television violence is

effective, it is all right to beat up people if they make you

mad, it is all right to hit someone to get them to do what

you want, and physical force prevents others from walking

over you (alpha = .80).

Items measuring attitudes toward delinquent acts ask

respondents whether particular delinquent acts are right or

wrong. Items measuring attitudes toward interpersonal

violence likewise ask respondents whether they agree with

statements that are favorable to violent behavior. Both of

these items serve as measures of definitions in that they

measure attitudes attached by respondents to particular

behavior. Respondents who hold attitudes positive or

favorable to delinquency will be more likely to commit

delinquent acts while those who hold attitudes negative or

unfavorable to delinquency will be less likely to commit

delinquent acts. Attitudes toward delinquency focuses on

definitions about specific acts, while attitudes toward








interpersonal violence focuses on general definitions about

violence and deviance.

The EDN index includes seven items: lying to teachers

is sometimes necessary to avoid trouble, to win at school it

is sometimes necessary to play dirty, lying is okay if it

keeps friends out of trouble, gaining the respect of friends

sometimes means beating up other kids, a willingness to break

rules is necessary to be popular with friends, lying to

parents is sometimes necessary to keep their trust, and it

may be necessary to break some of your parents' rules in

order to keep friends (alpha = .82).

Endorsement of deviant neutralizations is measured by

asking respondents about their agreement with attitudes

justifying deviant behavior. Respondents who are in greater

agreement with justified deviance or delinquency will be more

likely to engage in delinquency themselves. Attitudes toward

deviant neutralizations focuses on general definitions about

violence and deviance.

Differential association

Recall from the last chapter that differential

association refers to interaction with others in primary and

secondary groups as well as in broad reference groups via the

mass media. Initial, enduring, frequent, and close

associations have greater influence on one's behavior than

those which lack these characteristics. The peer group is an

example of a primary group in which youths typically form

associations characterized to some extent by these traits.








Those who associate with others who steal or behave

violently, for example, will receive normative definitions

favorable to delinquent behavior and be more likely

themselves to engage in this type of behavior. Associations

also provide the opportunity for other social learning

mechanisms, such as reinforcement, to have an impact.

Thus, a variable of the respondent's association with

delinquent peers will be used to measure differential

association. Research literature consistently shows that

this variable is one of the strongest predictors of juvenile

delinquency. Other studies using a similar variable based on

the same measure from the National Youth Survey include

Heimer (1997), Heimer and Matsueda (1994), Warr (1993),

Matsueda and Heimer (1987), and Elliott et al. (1985).

A five-point Likert scale is used to measure delinquent

peer association (DPA) with 1 being "none of them" engaging

in delinquent acts and 5 being "all of them." The DPA index

includes ten items: purposely damaging others' property,

using marijuana, stealing something worth less than $5,

hitting or threatening to hit someone without any reason,

using alcohol, breaking into a building or vehicle to steal

something, selling hard drugs, stealing something worth more

than $50, getting drunk, and giving or selling alcohol to

minors (alpha = .88). (See Appendix A for a complete

description of the variable.)








Sociodemographic variables


Six additional variables are considered, including age,

sex, race, class (as measured by income and education), and

family size. Any examination of the relationship between

family structure and delinquency without consideration of

these variables might result in the oversight of significant

subgroup differences. The common sociodemographic variables

of age, sex, race, and class were selected for their widely

known impact on delinquency. Number of siblings is a less

common variable but some studies have suggested that it has

an impact on delinquency (Hirschi, 1969). Correlations

indicating the strength and significance of the relationship

between self-reported delinquency and each of the

sociodemographic variables may be found below in Table 1, on

pp. 93-4. Each sociodemographic variable is discussed in

greater detail below. (See Appendix A for a description of

the original items employed in this section.)

Age

The relationship between age and delinquency is well

known in criminology and can be seen across various types of

data collection. Generally, delinquent behavior increases

near the start of the teen years, peaks at around age 17, and

declines through the 20s as offenders age-out. Both official

statistics and self-report data show that serious criminal

behavior peaks around ages 16 and 17 (Siegel and Senna, 1991;

Elliott, 1994).








The results of a cross-sectional analysis of NYS data at

wave three were consistent with previous research, showing

ages 16-18 to be significantly more delinquent than ages 13-

15. The relationship was weaker in a longitudinal analysis

but there remained significant differences in delinquency by

age cohort. When coded at year one as ages 13-15 (1) and

ages 16-17 (2), age was significantly correlated with

delinquency in years two through three (r=.06, p=.049). An

interval-level coding of age yielded a weaker correlation

with delinquency. Therefore, age is coded dichotomously as

indicated and entered into the equations.

Sex

A variety of data indicate that males are more

delinquent than females. Official statistics typically show

that the ratio for serious violent offending by males to that

of females is approximately eight to one, and for property

crime approximately four to one in favor of males (Siegel and

Senna, 1991). Sex differences in self-report data typically

have been smaller than those seen in official statistics,

while remaining significant (Elliott and Ageton, 1980). The

current research confirms this literature. These data show

that males accounted for 76 percent of all delinquents while

comprising 52 percent of the sample. Self-reported

delinquency, in years two through three, and sex were found

not to be independent of each other (r=-.18, p<.01).

Therefore sex is included as a control variable, coded males

(1) and females (2).








Race

While official statistics indicate that there are

significant differences by race, some self-report studies

have shown race differences to be virtually nonexistent,

suggesting the presence of bias in the justice system (Siegel

and Senna, 1991). However, using wave one of the National

Youth Survey (NYS), Elliott and Ageton (1980) found

significant race differences for self-reported general

delinquency and for crimes against property, due to blacks

reporting higher frequencies than whites. Later, Elliott

(1994) reported that serious, self-reported violence differed

very little by race. In the present research, blacks

accounted for 16 percent of the sample. Self-reported

delinquency, in years two through three, and race were found

to be significantly correlated (r=.09, p<.01). Race is

retained as a control variable, coded white (1) and black

(2).

Class

Class differences also tend to be more common in

official statistics while self-report studies often report

smaller class differences or none at all, again suggesting

the presence of bias in the justice system (Siegel and Senna,

1991). As class may be measured in a variety of ways,

conclusions about its effects on delinquency may be harder to

draw, compared to conclusions about race differences.

Nevertheless, as they found with race, Elliott and Ageton

(1980) reported that class differences may be more








substantial than has been suggested by many self-report

studies, due to lower-class youths reporting higher

frequencies than middle-class youths. Using NYS waves one

through five (1976-1980), Elliott and Huizinga (1983)

likewise found that middle class youth of both sexes are less

likely to be involved in serious offenses, and they commit

fewer offenses when they are involved, compared to lower and

working classes. Both of these studies using NYS data employ

the Hollingshead (1958) index of social class.

The present study employs both family income level and

parents' educational attainment as measures of class. Both

of these measures were obtained from the initial NYS wave in

1976, which included a survey of parents in addition to the

youth survey that would be repeated for subsequent waves.

Thus, these measures were obtained from a time two years

earlier than the other sociodemographic variables, which were

measured at wave three. It is expected that this earlier

measurement will not be problematic because family income and

parental education level are variables that are not likely to

change significantly in that time span. Income was coded as

follows: low ($10,000 and less) (1); medium ($10,001-30,000)

(2); and high (more than $30,000) (3). Twenty-four percent

of the sample reported low family income, 61 percent reported

medium income, and 15 percent reported high income. Income

was found to be significantly correlated with delinquency in

years two through three (r=-.10, p<.01), indicating that the








two variables are not independent. Therefore family income

is retained as a control variable.

Parents' educational attainment was coded as follows:

high (approximating college graduate) (1); medium

(approximating high school graduate) (2); and low

(approximating less than high school graduate) (3). The

scale combines measures of both mother's and father's

schooling. Sixteen percent of the respondents' parents

scored high on education, 44 percent scored medium on

education, and 40 percent had low levels of education.

Similar to family income, education measured in 1976 was

found to be significantly related to respondents' delinquency

in years two (1978) through three (1979) (r=.09, p<.01) and

is retained as a control variable.

Number of Siblings

The influence of siblings on delinquent outcomes has

received relatively little attention although the number of

siblings may be an important variable for a variety of

reasons, including decreased parent-child interaction. Where

there are a greater number of children in a household,

parent-child interaction might be expected to decrease as it

does where there are fewer parents. Farnworth (1984)

reported no significant effects for family size and family

density on a wide range of delinquent behavior among black

youths. But in a review of the literature, Loeber and

Stouthamer-Loeber (1986) reported significant effects for

even small differences in family size. They suggest,




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EDOJZ3YZ3_NSF3L3 INGEST_TIME 2013-10-10T00:47:26Z PACKAGE AA00017633_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


FAMILY STRUCTURE AND JUVENILE DELINQUENCY:
THE MEDIATING ROLE OF
SOCIAL LEARNING VARIABLES
By
EDWIN R. PAGE
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1998

Copyright 1998
by
Edwin R. Page

ACKNOWLEDGMENTS
I express my most heartfelt gratitude to Natalee Waters
for the love, friendship, and support that she gave me during
my work on this research.

TABLE OF CONTENTS
ACKNOWLEDGMENTS ÜÍ
LIST OF TABLES vi
ABSTRACT viii
CHAPTERS
1 INTRODUCTION 1
2 FAMILY STRUCTURE AND DELINQUENCY 5
Family Relationships 5
Parental Criminality 6
Parental Discipline and Punishment 7
Abuse in the Family 9
Other Family Relationships 15
Family Structure and Broken Homes 18
Conceptual and Methodological Issues 19
Research on Family Structure 23
Conclusions 37
3 SOCIAL LEARNING THEORY OF CRIME AND DELINQUENCY 41
Social Learning Theory 42
Differential Association 44
Definitions 45
Imitation 47
Differential Reinforcement 48
Temporal Sequence and Reciprocal Effects in Social
Learning Theory 49
Social Structure and Social Learning: The Case of
the Family and Delinquency 51
Research on Social Learning Theory 53
Social Learning Theory and Delinquency 53
Family, Social Learning Theory and Delinquency 60
Hypotheses 65
The Present Research 68
4 METHODOLOGY 71
Data 71
Measurement of Variables 76
Delinquency Index 76
Family Structure 78
iv

Social Learning Variables 80
Sociodemographic Variables 87
Plan of Analysis 95
5 ANALYSIS AND RESULTS 100
Hypothesis One: Family Structure and Delinquency ...101
Hypothesis Two: Family Structure and Social Learning.... 102
Hypothesis Three: Social Learning and Delinquency 104
Hypothesis Four: Family Structure, Social Learning,
and Delinquency 106
Additional Analyses 115
Structural Effects of Sociodemographic Variables 115
Logistic Regression 116
Controlling for Sex 121
6 CONCLUSIONS 130
Summary 130
Conclusions 136
Implications for Policy 140
Limitations and Suggestions for Future Research 144
APPENDICES
A VARIABLE MEASUREMENT ITEMS 150
B ADDITIONAL REGRESSION TABLES 156
REFERENCES 164
BIOGRAPHICAL SKETCH 173
v

LIST OF TABLES
Table gage
1. Correlation Matrix 93
2. Analysis Plans for Family Structure, Social
Learning, and Delinquency: Years in which
Dependent, Independent, and Sociodemographic
Variables are Measured 96
3. Percentage Distributions of Self-Reported
Delinquency (SRD) 100
4. Crosstabulation Summary: Percent Nondelinquent by
Family Structure for Each Analysis Plan 101
5. Zero-Order Correlation Coefficients for Self-
Reported Delinquency (SRD) and Family Structure
in Four Analysis Plans 102
6. Zero-Order Correlation Coefficients for Family
Structure and Social Learning Variables in Four
Analysis Plans 103
7. Crosstabulation Summary: Self-Reported Delinquency
(SRD) across Averaged Social Learning Scale
Items for Each Social Learning Variable 104
8. Zero-Order Correlation Coefficients for Self-
Reported Delinquency and Social Learning
Variables in Four Analysis Plans 105
9. Family Structure, Social Learning, and Delinquency:
One-Year Lagged Analysis Plan (1) with Self-
Reported Delinquency Measured in Year Two 107
10. Family Structure, Social Learning, and Delinquency:
One-Year Lagged Analysis Plan (2) with Self-
Reported Delinquency Measured in Year Three 109
11. Family Structure, Social Learning, and Delinquency:
Two-Year Lagged Analysis Plan (3) with Self-
Reported Delinquency Measured in Years Two and
Three Ill
vi

12. Family Structure, Social Learning, and Delinquency:
Temporal Sequence Analysis Plan (4) with Self-
Reported Delinquency Measured in Year Three 112
13. Family Structure, Social Learning, and Delinquency:
Analysis Plan 1 Using Logistic Regression 118
14. Percentage Distributions of Self-Reported
Delinquency (SRD) for Males and Females 122
15. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 for Male Subsample 123
16. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 for Female Subsample 124
17. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 for Male Subsample 125
18. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 for Female Subsample 127
19. Sociodemographic Variables, Family Structure,
Social Learning, and Delinquency: OLS
Regression 156
20. Family Structure, Social Learning, and Delinquency:
Analysis Plan 2 Using Logistic Regression 157
21. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 Using Logistic Regression 158
22. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 Using Logistic Regression 159
23. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 using Logistic Regression for
Male Subsample 160
24. Family Structure, Social Learning, and Delinquency:
Analysis Plan 3 Using Logistic Regression for
Female Subsample 161
25. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 Using Logistic Regression for
Male Subsample 162
26. Family Structure, Social Learning, and Delinquency:
Analysis Plan 4 Using Logistic Regression for
Female Subsample 163
vii

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
FAMILY STRUCTURE AND JUVENILE DELINQUENCY:
THE MEDIATING ROLE OF
SOCIAL LEARNING VARIABLES
By
Edwin R. Page
August, 1998
Chairman: Ronald L. Akers
Major Department: Sociology
There has been relatively little research to date
connecting the behavioral process in social learning theory
to elements of social structure in accounting for criminal
and delinquent behavior. The linking of social structure to
crime and delinquency through the social learning process is
the next step in the development of social learning theory.
The theory would hypothesize that the variables specified in
social learning theory mediate the effects of social
structure on deviant behavior.
Seeking to establish intervening variables in the
relationship between family structure and delinquency, a
three-level theoretical model is presented with six measures
representing three components of social learning theory. A
viii

longitudinal analysis is conducted on three waves of a
nationally representative sample, permitting generalizability
of findings. The panel data of the National Youth Survey
enable the observation of particular individuals in terms of
their family structure and their scoring on social learning
measures, and their subsequent involvement in serious
delinquent or criminal behavior. Correlational and
regression analyses are used to establish relationships
between the three levels of measurement, and to test the
central hypothesis that social learning variables mediate the
effects of family structure on delinquency.
Results provide support for the hypotheses that family
structure is associated with delinquency, that family
structure is associated with social learning variables, that
social learning variables are associated with delinquency,
and that selected social learning variables generally reduce
the strength and significance of the effect of family
structure on delinquency. These findings persist after
controlling for sex. It is concluded that the social
learning variables significantly mediate the impact of family
structure on delinquency.

CHAPTER 1
INTRODUCTION
Although crime rates have moderated and declined in the
past few years, between 1965 and 1990 the arrest rate for
violent crime by juveniles rose from approximately 130 per
100,000 to 430 per 100,000, according to the FBI (U.S.
Department of Justice, 1992). During the same period of time
there also has been considerable change in the structure of
families. The divorce rate (per 1,000 married women), for
example, rose from 9.2 in 1960 to 21 in 1988, having peaked
in 1981 at 22.6 (U.S. Department of Commerce, Bureau of the
Census, 1994a). The rising divorce rate has accompanied a
general change in family structures. The amount of single¬
parent families, for example, increased from 12 percent of
families with dependent children in 1970 to 27 percent in
1993 (U.S. Department of Commerce, Bureau of the Census,
1994b). Many researchers have sought to specify a causal
link between the increase in crime and the changes in
families. Families are an intuitively appealing place to
look for an explanation of delinquency because of the
prominent role that they occupy in the process of
socialization.
Family has long been an important variable in the
etiology of delinquency in both social and biological
1

2
research (see Akers, 1994; Siegel and Senna, 1991; Hagan and
Sussman, 1988; Void and Bernard, 1986). The sociological
literature suggests a variety of ways by which family
variables affect delinquency (Jensen and Rojek, 1992). For
example, strain theories position the family as an important
determinant of social class and thereby the likelihood of
delinquency. Social learning theory suggests that families
are an important shaper of conforming or deviant attitudes
and behavior. For control theories, families are an
important source of bonding by which youths are directed
toward conformity and away from delinquency.
It is easy enough to understand how the family would
usually have a considerable impact on delinquency. "The
family has typically been viewed as the most crucial
institution in our society for shaping a child's personality,
attitudes, and behavior," (1992, p. 262) according to Jensen
and Rojek. Children typically spend most of their early
years in the presence of other family members. Even after
reaching school age the family remains one of the most
important institutions of socialization, along with schools
and peer groups.
Socialization in the family remains significant today
even if the social functions of families have declined and
individuals spend less time in the family setting. With the
increased complexity of modern life and the variety of
available leisure activities, individuals today are more
likely than their forebears to live a greater portion of

3
their lives outside of the home. The typical household size
itself declined as individuals increasingly moved away from
their family of origin and the nuclear family became more
common. Census data show that the average household size
declined from 5.4 persons in 1790 to 4.2 in 1900, to 3.3 in
1940, and to 2.6 in 1988 (Eshleman, 1991). Despite changes
such as these, families remain central socializing forces
today. "Although the family may have been a more important
social institution at earlier points in history, most
scholars are still willing to accept that it remains a major
setting for socialization in American society" (Jensen and
Rojek, 1992, p. 263).
The present study adds to the empirical and theoretical
body of knowledge relating to the connection between family
structure and delinquency in a variety of ways. While it
will be shown that the empirical validity of social learning
theory has been well demonstrated in the literature, there
has been relatively little research to date connecting the
behavioral emphasis of the theory to elements of social
structure. The linking of social structure to crime and
delinquency by the social learning process is the next step
in the development of social learning theory (Akers, 1994).
In looking to establish intervening variables in the
relationship between family structure and delinquency, a
three-level theoretical model will be presented with six
measures representing three components of social learning
theory. The theory would hypothesize that the variables in

4
social learning theory mediate the effects of social
structure on deviant behavior.
Additionally, the methodological design employed here
attempts to improve on the shortcomings of much previous
research on family and delinquency. In a review of the
literature. Free (1991) concluded that a major weakness of
previous studies is their overreliance on cross-sectional
data and that there has been an overemphasis on minor
offenses. A longitudinal analysis will be employed similar
to that used by Heimer and Matsueda (1994) but using more
recent data spanning three years. The use of a nationally
representative sample will allow a broad generalizability of
findings. The panel data of the National Youth Survey and
the use of longitudinal analysis will enable the observation
of particular individuals who have experienced a change in
family structure and their reactions to that change in terms
of serious delinquent or criminal behavior.

CHAPTER 2
FAMILY STRUCTURE AND DELINQUENCY
In this chapter a variety of literature pertaining to
family and delinquency will be reviewed. First, the impact
of family relationships will be considered, including the
effects that parental criminality, parental discipline and
punishment, various types of abuse, and other family
relationships have on delinquency. This part of the review
will provide a background for the subsequent review of the
literature on family structure and delinquency. As will be
shown, some researchers maintain that family structure
affects delinquency, at least in part, through family
relationships such as those to be discussed here. As the
exogenous independent variable in this study, the concept of
family structure and the implications of its measurement will
be discussed. Finally, research pertaining to the
relationship between family structure and delinquency will be
reviewed.
Family Relationships
Given that families are a key socializing institution,
they may be implicated in delinquency to the extent that the
socialization they provide is inadequate or deviant.
Parental criminality is one way that delinquency may be based
5

6
in the family. Delinquency has also been investigated as a
product of the method used by parents to discipline and
punish their children. Family violence and abuse is another
way in which socialization may result in delinquent behavior.
Furthermore, other nonviolent and seemingly healthy family
relationships may conceal pathological interaction and
involve socialization toward delinquency.
Parental Criminality
Parental criminality has been found to be related to an
increased likelihood of delinquency. Using official records
from the Cambridge Study in Delinquency Development, West
(1982) found that criminality in both fathers, mothers, and
brothers was highly predictive of delinquent behavior. Sons
with criminal fathers were approximately twice as likely to
engage in criminal activities by age 22 compared to those
with noncriminal fathers. The presence of multiple criminal
influences further increased the likelihood of a son engaging
in delinquency. "The chances of a man being a delinquent
were some three and a half times greater if he had more than
two other family members with a record than if he belonged to
a conviction-free family," (1982, p. 45) according to West.
In a recent article using data from the Cambridge Study, Rowe
and Farrington (1997) suggested a genetic explanation for the
familial transmission of convictions. Their findings
indicated that parental criminality affected child
criminality directly and not via the effects of family

7
environment. However, they could not rule out the role of
family environment altogether due to a small family
environment effect in their findings and the possibility of
other unmeasured effects.
Lattimore et al. (1995) found parental criminality to be
one among several family variables, including family
violence, that was significantly related to violent
recidivism in a cohort of male parolees. In a Danish cohort
study, Mednick et al. (1987) found that fathers’ criminality
was directly associated with general crime in their young
adult sons. Mednick et al. (1990) found that fathers'
criminality influenced adult sons' criminality, in part
through an increased risk of family instability. Laub and
Sampson (1988; see also Sampson and Laub, 1993) found that
criminality in either parent, and especially substance abuse,
had a strong indirect influence on delinquency by way of
disrupted social control in the family. Analyzing data
collected in the 1940s by Sheldon and Eleanor Glueck, Laub
and Sampson found that significant effects exhibited by
parental criminality and drunkenness on serious delinquency
were largely mediated by family process variables such as
attachment, supervision, and erratic parental discipline.
Parental Discipline and Punishment
Over 90 percent of American parents use physical
punishment to correct their children's misbehavior, according
to Straus (1991). Most research is in agreement that

8
inconsistent parental discipline, whether too harsh or too
lenient, is related to delinquent behavior (Jensen and Rojek,
1992; Siegel and Senna, 1991). In a research review, Snyder
and Patterson (1987) report strong evidence of a link between
ineffective disciplinary practices and the families of
delinquents. "In both lax and enmeshed discipline styles,
punishment is not used consistently, contingently, and
effectively to discourage the aggressive, antisocial behavior
of the child," (1987, p. 223) according to Snyder and
Patterson. Agnew (1983) found that while physical punishment
slightly decreased the likelihood of self-reported
delinquency among children who perceived their parents as
consistent, where discipline was inconsistent children were
more likely to report delinquency. Krohn et al. (1992) found
that consistent punishment and spanking were significantly
related to a decrease in self-reported delinquency. McCord
(1991) found that consistent, nonpunitive discipline by
mothers helped to insulate sons from delinquent influences.
Investigating the sex differences in delinquency, Jensen and
Eve (1976) suggested that the more restrictive supervision
received by females may partly account for their lower
involvement in delinquency.
Using the Glueck data, Sampson and Laub (1993) also
found that the use of erratic and punitive discipline or
punishment by mothers and fathers affected official and
unofficial delinquency. Moreover, they found that background
variables, such as father's alcoholism and criminality,

9
affected discipline and punishment by either parent,
suggesting "the possibility that some, if not all, of the
effect of father's criminality on the son's delinquency may
be attributable to family discipline" (Sampson and Laub,
1993, p. 79). Finally, they found that discipline and
punishment, among other family process variables, mediated
the effects of structural background variables on official
and unofficial delinquency.
Abuse in the Family
Violence in the family may take the form of child abuse,
including sexual abuse, as well as violence between parents
or siblings. Child abuse may refer to the overt physical
aggression against a child or to neglect, wherein children
are deprived of attention, food, shelter, or other care. Any
type of violence or abuse in the family, directly involving a
child or not, may increase the likelihood of that child
engaging in delinquency.
The issue of child abuse has gained much recognition
since the 1960s. Although the true extent of abuse is not
known due to the difficulty in obtaining accurate data and
the methodological differences of various reports, most
researchers agree that child abuse is a significant problem
(Gelles and Conte, 1991; Pagelow, 1984). Straus and Gelles
(1986), for example, found that parents reported incidents of
severe violence toward their children at an annual rate of 19
per 1,000 children aged three to 17. Their measure of severe

10
violence included kicking, biting and hitting with a fist;
beating up; and using a gun or knife. The findings by Straus
and Gelles were considered conservative because the sample
included only those children living with both parents and the
violence was self-reported, and there may have been an
increased consciousness of the unacceptability of child abuse
by the mid-1980s. Additionally, this finding includes only
the most serious type of abuse. Although Straus and Gelles
report that this figure represented a decline in violent
abuse compared to their findings in 1975, other studies
suggest that child abuse has increased (see Gelles and Conte,
1991).
Sexual abuse is a more specific form of child abuse that
poses serious social-psychological threats to children
(Gelles and Conte, 1991). It may range from rewarding a
naive child for inappropriate sexual behavior, to fondling
and other inappropriate touching, to using force or the
threat of force for the purposes of sex (Siegel and Senna,
1991). Findings on the amount of sexual abuse vary but
typical estimates are that 19 percent of females and nine
percent of males are sexually abused before age 18
(Finkelhor, 1984). Sex abuse within the family usually
occurs across generations. The victims are more often
female, although sex abuse of boys is more likely
underreported, and the abusers are overwhelmingly male
(Pagelow, 1984; Finkelhor, 1984).

11
In their national survey of family violence, Gelles and
Straus report that children from violent homes are "three to
four times more likely than children from nonviolent homes to
engage in illegal acts" (1988, p. 129). While they were not
able to establish causality, Gelles and Straus suggest that
their research and other studies show evidence of a cycle of
violence existing in many homes. "Violent experiences set
the stage for the individual and social traits that lead to
trouble. The trouble is responded to with more violence"
(1988, p. 129).
The effects of child abuse and its relationship with
delinquency are not clear even though it is generally agreed
that abuse does have detrimental long-term consequences.
"Definitive support for a direct causal relationship between
an abusive childhood and delinquency is lacking, but
researchers generally agree that abuse and neglect may have a
profound effect on behavior in later years," (1991, p. 265)
according to Siegel and Senna. The authors said that this
agreement is partly based on a number of studies showing that
many delinquents and criminals have a history of child abuse.
In a cohort study comparing serious, officially
recognized abuse cases with a control group Widom (1989a)
found that 29 percent of the abused group later had adult
criminal records for a nontraffic offense compared to 21
percent of the control group. Widom's measures of abuse
included violent physical abuse, sex abuse, and neglect.
Considering demographic variables, Widom found that those who

12
were older at the time of abuse, male, and black were at
highest risk of criminality while being young, white, and
female was associated with lower risk. While abused youths
were more likely to become criminal than nonabused youths,
the relatively small difference between the two groups,
compared to what might be expected, points to the indirect
nature of the relationship between family variables and
delinquency.
Widom (1989b) reported that children who were neglected,
and especially, violently abused had a significantly greater
risk of becoming delinquents, criminals, and violent
criminals. More recently, Widom (1996) also found a greater
likelihood of arrest as juveniles and as adults for the
victims of childhood abuse or neglect. These findings
applied to arrests for both drug-related offenses and
property crimes. Widom also found that "the victims of
sexual abuse are not more likely than other victims of
physical abuse or neglect to become involved with crime"
(1996, p. 50).
Widom (1989b) suggested that her findings, and similar
findings by other researchers, have a mixed message for the
cycle of violence hypothesis. Considering the large number
of abused children that do not become delinquent or criminal,
Widom (1989b, 1996) suggested that many victims apparently
manifest their abuse in more subtle ways. This may be the
case particularly for females, who are both more likely to be
sexually abused and less likely to become delinquent. Widom

13
said that attention should focus on possible mediating
variables that buffer or protect abused or neglected children
from delinquent outcomes.
Reviewing the literature on the effects of childhood sex
abuse, Finkelhor (1984) reported that a variety of studies
have shown high proportions of various troubled populations,
including delinquents, were sexually victimized as children.
While he said the case apparently may be easily made that sex
abuse leads to injurious outcomes, Finkelhor cautioned, "It
is possible that the long-term effects seen in these cases
are a function not of the sexual abuse but of other
pathological elements, such as psychological abuse, parental
neglect, or family disorganization" (1984, p. 189).
Similarly, Pagelow (1984) warned that the finding of a
strong relationship between childhood violence and later
deviant behavior may overlook the issue of causality. "Such
a blanket assumption would be hazardous because there are too
many possible intervening variables that can influence
outcomes not included in simple ’cause and effect' models"
(1984, p. 229). Pagelow also suggested that official records
may be misleading. "Connections between child protection
agencies and the juvenile justice system are clear. Entering
the first, by becoming an identified case, can serve to
increase an abused child's chances of entering the second"
(1984, p. 255).
Reviewing the literature on family violence and neglect,
and its relationship with children's deviant behavior, Koski

14
(1988) reported a complex and often contradictory array of
findings. The research "remains largely descriptive and
piecemeal and leads to only one fairly obvious conclusion:
unhealthy families tend to produce deviant children, although
not always in the ways expected" (1988, p. 43). Koski also
said that research appears to suggest that a child's whole
family environment, where more than one type of abuse may
exist, may be more critical to understanding deviance and
delinguency than any single type of abuse.
The family environment includes not only the direct
physical abuse and neglect of a child by a parent discussed
above, but also violence involving other family members.
Straus and Gelles (1986) reported that cases of severe
violence by husbands toward their wives numbered 30 per 1,000
couples in a 1985 national survey. Controversially, they
also found that the rate of severe violence by wives toward
husbands was 44 per 1,000 couples. Straus and Gelles added
that a number of studies show that violence on the part of
wives is often committed in retaliation or self-defense, and
that this explanation is supported by the fact that females
are relatively nonviolent outside of the home. Some studies
have shown that a child's exposure to spousal violence as
well as violence committed by other family members is related
to the child’s use of violence (Pagelow, 1984) and to later
family and nonfamily violence by the grown child (Fagan and
Wexler, 1987).

15
Other Family Relationships
While much research has focused on overt, physical abuse
and neglect, delinquent or other deviant behavior may result
from more routine experiences within the family setting.
Social groups such as families exert considerable influence
over the individuals within that group. "The relationships
of persons in a [family] nexus are characterized by enduring
and intensive face-to-face reciprocal influence on each
other's experience and behavior" (Laing and Esterson, 1971,
p.7). An individual may derive from others within their
group a sense of self-fulfillment or they may derive feelings
of alienation, even to the point of madness (Laing, 1962).
Laing and Esterson (1971), for example, make the case that
schizophrenia is socially based within dysfunctional family
relationships.
Satir (1972) distinguishes between "troubled" and
"nurturing" families. "Troubled families make troubled
people and thus contribute to crime, mental illness,
alcoholism, drug abuse, poverty, alienated youth, political
extremism, and many other social problems" (1972, p. 18).
The factors that determine whether a family is nurturing or
troubled—individual feelings of self-worth, communication
with each other, family rules about proper feelings and
behavior, and the way that the family’s members relate to
other people and institutions outside the family—are largely
learned, primarily within the family, Satir said. The more

16
that self-worth is high, the more direct and honest the
communication, the more flexible the family's rules, and the
more open a family is to society, then the more likely a
family will be nurturing. These variables are important to
family quality whatever the nature of family structure—
whether it is a family with two biological parents, a single¬
parent family, a family including stepparents, or an
institutional family not involving parents, according to
Satir (1972).
Snyder and Patterson (1987) describe positive parenting
as interactions between parent and child which foster growth
skills and encourage the development of normative values and
standards of behavior. They said these ends are fostered in
part "when the parents demonstrate and communicate interest
in the child in a positive, noncritical fashion, when the
parents communicate support and caring to the child, and when
the parents and child share mutually pleasurable leisure
activities” (1987, p. 223). According to Snyder and
Patterson, a variety of research suggests a link between the
absence of positive parenting and delinquent behavior. "A
failure by parents to foster a child's skills, to model and
encourage normative values, and to provide a caring
environment, places a child at risk for delinquency" (1987,
p. 225).
McCord (1991) found that a combination of family
variables which she labeled "poor child rearing" increased
the risk of officially recorded delinquency. This

17
combination included variables from three sets of items
measuring mother's competence, father's interaction with the
family, and family expectations. Of these three groupings,
McCord found that mother's competence and family expectations
were strongly related to serious criminality as a juvenile,
while father's interaction with the family had a more
important role in affecting adult criminality.
McCord's data consisted of case studies on male youths
from homogeneous socioeconomic backgrounds collected in the
Boston area in the 1940s and tracked in the late 1970s.
McCord's measures of mother's competency included self-
confidence in handling problems, affection for her son,
nonpunitive discipline, and family leadership role. Father's
interaction with the family included affection for son,
esteem for mother, esteem in the eyes of the mother.
Father's interaction also included, negatively, parental
conflict and father's aggressiveness. Finally, the family
expectations cluster included maternal restrictiveness,
parental supervision, and demands placed on the son.
In a meta-analysis of the research on families and
delinquency, Loeber and Stouthamer-Loeber (1986) reported a
significant relation between marital discord and children’s
delinquency and aggression. The measures of marital
relations include ratings of happiness, amount of conflict,
lack of warmth, and hostility. Generally, Loeber and
Stouthamer-Loeber found that variables such as parent-child
involvement, parental supervision, and parental rejection had

18
the largest impact on crime and delinquency. They also
reported that many studies have shown that broken homes have
an impact on delinquency, although not as strongly as in the
case of marital discord.
Nevertheless, much research has focused on the question
of family structure. The following section will review the
research literature on the delinquent outcomes related to
broken homes and to family structure. Despite the large
amount of research in this area there remains some
inconclusiveness regarding the relationship between these
measures of family and delinquency.
Family Structure and Broken Homes
Research addressing family structure has often been
grouped under the label "broken homes." Thus, the two terms
often have been used interchangeably in previous literature.
In order to avoid confusion, a distinction should be made
between the terms family structure and broken homes. A
broken home is one in which a two-parent structure has been
reduced to a single-parent structure, whether by divorce,
separation, or death. Family structure refers more generally
to the number or type of parents living in a household.
Family structure may be measured by the presence of mothers
or fathers, and biological parents or stepparents. Thus, a
broken home more specifically describes a particular family
situation, compared to the broader description of family
structure.

19
As will be seen in the remainder of this chapter and in
the review of the research literature pertaining to social
learning theory in Chapter 3, many researchers have focused
on an increase in divorce and a growing number of single¬
parent families as correlates of delinquency. The assumption
is that the disintegration of the family is a significant
cause of inadequate socialization, which in turn should have
an effect on the likelihood of delinquency. In fact,
however, family disintegration manifests itself in a variety
of forms which may be captured with varying success by the
dichotomy of "intact" and "broken" homes or by measures of
family structure. The first part of this section will
address the potential for confusion about the concepts of
broken homes and family structure, and the implications of
this distinction for measurement. Second, empirical research
on the relationship between family structure and delinquency
will be reviewed.
Conceptual and Methodological Issues
In the section above on family relationships it was seen
that "intact," or two-parent family structures, may be
variously integrated. "Broken," or single-parent families as
well are variously disintegrated, and perhaps not broken at
all. For example, single parenthood may result not simply
from divorce or other family conflict. There are single
parents, such as some who adopt, who have not at any time
been united with another as a couple and neither parent nor

20
children experience the stress of divorce. Furthermore,
family conflict may result in divorce, separation, or in no
recognizable change in family structure. Stepfamilies are
another source of complication for the study of families, as
they may vary in the quality of life they provide for
children.
According to Wells and Rankin, "Broken homes are
invariably defined from the reference point of children vis-
á-vis their relationship with parents, whereas sibling
structures are ignored” (1986, p. 70). A broken home refers
to an impaired or disrupted family structure. Broken homes
are typically measured as those in which the mother and
father are divorced (e.g. Furstenberg and Teitler, 1994;
Paternoster and Brame, 1997). Such a measure does not
necessarily distinguish the sex of a single parent. Using
single parenthood as a proxy for divorce would overlook the
alternative possibilities of separation, death of a parent,
or a never-married single parent.
Measuring divorce itself is suited to a focus on the
process and outcome of marital breakdown with the implication
that children suffer adverse consequences as a result of
their parents' divorce or separation. According to Wells and
Rankin (1986), the sudden change in living routines which
accompany divorce can generate stress and conflict that may
result in antisocial behavior. Thus, delinquency may result
not from the change in family structure but from "parental
conflict that precedes and attends the separation, and the

21
sudden change in circumstances affected by the separation"
(1986, p. 77). Furstenberg and Teitler (1994) suggested that
the quality of earlier family life may have significant
consequences for children whether parents separate or not.
They found that some family characteristics typically thought
to occur with divorce, such as marital conflict, economic
stress, and poor parenting practices, actually existed prior
to divorce.
The conditions associated with divorce that have an
influence on delinquency may only be temporary, existing
until individuals adapt to new routines. The results of
Furstenberg and Teitler (1994) suggest, however, that the
impact of divorce may be more long-lasting. They found that,
compared to their counterparts from intact families, children
of divorce were more likely to express discontent with life,
more likely to experience teen pregnancy, less likely to
finish high school and attend college, more likely to engage
in substance abuse and delinquency, and more likely to be
arrested.
In contrast to the focus on divorce, many researchers
alternatively look to the number of adults in the household,
hypothesizing that a change in family structure that results
in fewer parents will be associated with a decreased level of
social and economic resources and conventional socialization
compared to what would be expected from two-parent families.
Thus, single-parent homes are thought to be at a social and
economic disadvantage due to the reduced resources and

22
opportunities available to single, and typically female,
parents (Wells and Rankin, 1986; see also Smith and Krohn,
1995). In such cases, divorce leads to a reduction of
economic resources and perhaps to poverty which, through a
variety of pathways, may lead to delinquency (Curran and
Renzetti, 1996). For example, poorer families are more
vulnerable to stress and conflict that may result in less
effective parenting and child behavior problems. According
to Gove and Crutchfield, "Single-parent families tended to be
of lower socioeconomic status than the intact families, and
children of single-parent families were also more likely to
be delinquent" (1982, p. 307). In a review of social
psychological research pertaining to female-headed families
published between 1970 and 1980, Cashion (1982) found that
after controlling for socioeconomic status children from
female-headed families fared as well as children from two-
parent families, in terms of juvenile delinquency rates and
other measures.
Measures of family structure also may take into account
the presence of stepparents. Some researchers divide their
family structure measure into categories for two biological
parents, a biological parent combined with a stepparent, and
a single-parent (e.g. Coughlin and Vuchinich, 1996;
Cernkovich and Giordano, 1987). Another approach is to
elaborate on this typology by distinguishing each type of
parent by sex (e.g. Johnson, 1986). Another way of measuring
family structure is to simply count the number of parents in

23
the household, whether they are biological parents or
stepparents (e.g. Van Voorhis et al., 1988). This measure
differs significantly from the previously described measure
in that it focuses on the number of parents rather than on
the type of parents.
Research on Family Structure
The varying ways in which family structure may be
defined and measured suggest the difficulties that may be
encountered when dealing with the concept. In fact, the
empirical evidence on family structure and delinquency
reveals varied and sometimes contradictory findings. The
following two paragraphs illustrate the point.
In favor of a family structure effect, in their
nationally representative self-report survey of youths aged
12 to 17, Dornbusch et al. (1985) consistently found a
greater number of delinquents among mother-only households
compared to households with both biological parents. This
relationship was consistent for a variety of self-reported
and officially recorded delinquency, including law contacts,
as well as across sex and social class. Reviewing the
literature on family and delinquency, Gove and Crutchfield
(1982) cite overall findings consistent with Dornbusch et al.
"An intact home with harmonious marital and family
relationships will be closely associated with a lack of
delinquency and ... a lack of harmonious marital and family
relationships and/or a single-parent home will be associated

24
with delinquency" (1982, p. 304). More recently, Wells and
Rankin systematically analyzed 50 studies and concluded that
there is "a consistent and real pattern of association"
between broken home and delinquency variables, showing a
bivariate correlation of .10 to .15 (1991, p. 87). They said
this association is stronger for milder forms of juvenile
misconduct and weakest for serious and violent crimes.
On the other hand, in their study of high school
students in a large midwestern city, Hennessy et al. (1978)
found no effect of broken homes on a wide range of self-
reported delinquency, including being picked up by police and
appearances in court. They measured a broken home as any one
that did not include both a biological mother and a
biological father. While finding no broken-home effect, they
did suggest that other patterns of interactions within the
family may be related to delinquency. For example, Salts et
al. (1995) found in their study of males aged 12 to 19 that
family structure variables were not related to self-reported
violence, theft, or other delinquency. However, they found
that other family variables such as conflict and cohesion
were related to violence in the expected directions. Jensen
and Rojek concluded in their text, "At the present time there
is little conclusive evidence to suggest that the broken home
is a critical variable in the understanding of delinquency"
(1992, p. 270). In another text, Curran and Renzetti claim
there is "little evidence to support the broken home
hypothesis," while at the same time they acknowledge that

25
"research indicates that parents' relationships and
interactions with their children are correlated with the
children's involvement in delinquency" (1996, p. 349).
Consistent with Curran and Renzetti's contention and the
findings of Salts et al., many researchers have addressed the
complicated findings relating to broken homes and delinquency
by drawing a distinction between the number and type of
parents in a household (family structure), and the quality of
parenting in a household (often referred to as family
function). "It is not family structure in itself that
affects delinquency but the personal relationships within the
family unit which result from family disorganization in
single-parent families," (1984, p. 363) suggested Farnworth.
For example, Van Voorhis et al. (1988) compared the
effects of structure versus quality of relationships in the
family on self-reported delinquency. Using a sample of high
school students from a small midwestern town, they found in
both correlational analyses and multivariate tests that
family structure had weak effects for status offenses only,
while measures of home quality were more strongly related to
a variety of delinquency. Family structure was measured in
two ways: intact versus broken homes, and two-parent versus
single-parent homes. Family quality measures included scales
on supervision, affection, enjoyment of home, conflict, and
abuse of children.
Using data from the National Survey of Youth, a 1972
self-report survey of youths aged 11 to 18, Rankin and Kern

26
(1994) found that the likelihood of delinquency was the same
for all children who were strongly attached to one parent,
regardless of whether they lived in an intact or single-
parént family. They also found the likelihood of delinquency
to be lower for those living in intact families and attached
to both parents than for those living with and attached to a
single parent. Rankin and Kern suggested that single-parent
families may have higher rates of delinquency due to more
sporadic parental supervision, monitoring, and discipline.
Thus, measures of family quality may mediate family
structure. Attachment was measured as a composite of
intimate communication, family activities, supervision, and
affection. Rankin and Kern's results applied to a variety of
delinquency but not to property offenses such as theft and
vandalism.
In their study of youths aged 12 to 19 in a single
metropolitan area, Cernkovich and Giordano (1987) found that
family structure was unrelated to a wide range of self-
reported delinquency. Family quality variables, on the other
hand, were significantly related to delinquency. Their
measure of family structure was divided into responses for
living with both biological parents, with the mother only,
and with the mother and a stepfather. Significant family
quality variables were supervision or monitoring, identity
support, parental caring and trust, instrumental
communication, parental disapproval of peers, and conflict
with parents. Only intimate communication was not related to

27
delinquency. Cernkovich and Giordano said that their family
quality variables were similarly related to delinquency
across different types of family structure. "While this does
not necessarily mean that home status is an unimportant
variable in delinquency involvement, it does suggest that
similar family dynamics are operating within various types of
family structure" (1987, p. 312). Noting their finding of no
broken-home effect, Cernkovich and Giordano said their
results do not suggest that family interaction mediates
broken homes, as many have assumed.
Indirect and direct effects of family structure
Other research has substantiated that family structure
does have an effect on delinquency. A number of studies have
suggested that family quality variables mediate the
delinquent effects of family structure. Using a larger
sample from the Richmond Youth Project (a 1965 self-report
survey of junior and senior high school students in
California’s Contra Costa County) Matsueda and Heimer (1987)
showed that the effect of family structure on delinquency may
be mediated by intervening variables that relate to family
function, or quality. Investigating general self-reported
delinquency, they found for both blacks and whites that
"broken homes influence delinquency by impeding the
transmission of antidelinquent definitions and increasing the
transmission of prodelinquent patterns" (1987, p. 835).
Their broken home variable measured the absence of either the
mother or the father from the respondent’s household. As

28
mentioned above, in their reanalysis of the Gluecks' data
comparing delinquent and nondelinquent adolescent males, Laub
and Sampson (1988) found that significant effects exhibited
by parental criminality and drunkenness on serious
delinquency were largely mediated by family quality variables
including attachment, supervision, and erratic or threatening
parental discipline.
Using a longitudinal, nationally representative data
set, McLeod et al. (1994) also investigated the extent to
which family structure affects antisocial behavior.
Specifically, they found that single-parent motherhood (both
never-married and previously married) affected mother's use
of alcohol, use of physical punishment, and frequency of
affection. These variables in turn were related to
antisocial behavior on the part of the children, such as
disobedience, bullying, and destructiveness. While their
measures of family quality also affected delinquency, McLeod
et al. (1994) report that their family structure measures
directly affected antisocial behavior. Living with a never-
married mother led to antisocial outcomes for blacks while
for whites antisocial behavior depended on living with a
previously married mother.
Johnson's (1986) findings from a sample of high school
sophomores also support the direct effect of family structure
on delinquency. He reported that while white males with
stepfathers self-report significantly more delinquent
behavior than those from intact homes, attachment to a parent

29
did not mediate the relationship. Also, the relationship for
white females between mother-only family structure and self-
reported official trouble was not mediated by any variables,
including the quality' of parent-child relationship. Findings
such as those by McLeod et al. (1994) and Johnson (1986)
suggest either that family structure, as measured by broken
homes or number of parents in the family, has some direct
relation to delinquency.
Stepparents
As we have seen, there has been an increasing focus on
variables measuring quality of family relationships and
delinquency. Other researchers have maintained their focus
on the impact of family structure, considering not just
mother-only or single-parent families but also stepfamilies
and other variations on the two-parent family. For example,
Johnson said that "family structure, independent of the
effects of the quality of parent-child relationships,
predicts certain measures of delinquency for certain
categories of adolescents" (1986, p. 78). He found that the
presence of a stepfather tended to increase self-reported
delinquency for males, while family structure in general made
little difference for female self-reported delinquency.
Coughlin and Vuchinich (1996) reported that males from
mother-only families and stepfamilies were both more than
twice as likely to be arrested than those from families with
two biological parents. Similarly, Dornbusch et al. (1985)
found similar levels of male self-reported delinquency and

30
self-reported arrests in stepfamilies and mother-only
families. For females, those from two-parent stepfamilies
were less delinquent than those from mother-only families,
although they were still more delinquent than those living
with both biological parents. For males, they found similar
levels of delinquency in stepfamilies and mother-only
families. The authors concluded that their data provided
little support for the belief that two adults in the
household are more likely to prevent adolescent deviance than
are single parents.
Given the findings reported above from Johnson (1986),
however, the similar levels of male delinquency in
stepfamilies and mother-only families may result from
different influences. That is, males in stepfamilies may
experience conflict with their stepfathers while males in
mother-only families may have less parental contact. In
their review of the literature on broken homes. Wells and
Rankin (1985) found results on the stepparent issue to be
inconsistent and inconclusive.
Additional research
In this section a variety of research will be reviewed,
including studies of the delinquent effects of living with a
single mother compared to living with a single father and
investigations of whether the timing or nature of parental
disruption affect delinquency. Also reviewed here is
research on the effects of sociodemographic variables in the
relationship between family structure and delinquency. The

31
section will conclude with a look at several aggregate-level
studies that address the question of how family structure is
related to delinquency.
Comparisons about the delinquent consequences of mother-
only families versus father-only families have been made
infrequently due to the relatively small percentage of
father-only families. Approximately 12 percent of the
children living in single-parent families live with a father,
and this figure was much lower a generation ago (U.S.
Department of Commerce, Bureau of the Census, 1994a). But in
their research review, Wells and Rankin concluded that "which
parent is absent makes no difference in predisposing to
delinquency" (1985, p. 260). However, Wells and Rankin noted
that some researchers have reported an interaction between
gender of parent and gender of child. That is, father-
absence may be more harmful for males while mother-absence
may be more harmful for females. Research in this area has
yielded mixed results and the issue remains unsettled.
Some researchers have examined the extent to which the
timing and the nature of family breakups are relevant to
delinquent outcomes. Mednick et al. (1987), dividing timing
of divorce into three levels based on the child's age, found
that it did not relate to number of arrests. West (1982)
found no difference in delinquency whether family disruption
occurred early or later in the child's life. Similarly,
using nationally representative, longitudinal data,
Furstenberg and Teitler (1994) reported no difference in the

32
effects of divorce whether parents separated early in the
child's life or later. Mednick et al. (1990), on the other
hand, did find that instability during adolescence was a
comparatively stronger predictor of young male crime than
instability earlier in childhood. The authors used a more
elaborate measure of stability by combining two scores on
family constellation changes, one score for prior to age 12
and the other score for age 12 and older.
Turning to the nature of parental disruption, it might
be reasoned that divorce or separation would have a greater
influence on delinquency than the death of a parent. This
might be the case because a divorce likely signifies conflict
and stress in family relationships. Otherwise, as some
research has suggested, the presence of stepparents may
result in a greater likelihood of conflict, especially
between sons and stepfathers. On the other hand, family
stress and the arrival of stepparents may also follow the
death of a parent. Stepparents aside, whatever the nature of
family disruption, at least a temporary result will be a
single-parent family. If delinquency is related to the
number of parents in the household, then one might expect the
nature of family disruption to have little effect on
delinquency. In their analysis of previous research, Wells
and Rankin reported, "The association with delinquency is
slightly (albeit not significantly) stronger for families
broken by divorce or separation than by death of a parent,"
(1991, p. 88).

33
Researchers have also examined the extent to which the
relationship between family structure and delinquency varies
by sex and race. Looking first at sex, Wells and Rankin
(1991) report that research generally shows no consistent
differences between males and females in the impact of family
structure, wells and Rankin noted that their use of broken
homes as a measure of family structure was likely complicated
as a result of their research being an analysis of previous
studies, given the variety of ways by which broken homes may
be measured. In contrast to the findings of Wells and
Rankin, Gove and Crutchfield (1982) found that parental
marital status had little impact on delinquency for females
while males with unmarried parents were more likely to be
delinquent than those from two-parent homes. Gove and
Crutchfield's data differed from most self-report studies in
that they measured parents' reports of their children's
behavior. Using self-report data on youths aged 13 to 16,
Thomas et al. (1996) also found that the deleterious effects
of family structure were concentrated among males. Johnson
(1986) found family structure to be more significant for
males' self-reported delinquency as a result of the
stepfather relationship. For females, Johnson said family
structure did not appear to be as important. While girls
from mother-only families were more likely to be officially
delinquent, Johnson said this finding may reflect bias by
school officials and police.

34
Other researchers reporting sex differences locate these
differences within the mediating variables of family quality.
Among her small sample of blacks at age 15, Farnworth (1984)
found that while family structure had only a small impact on
delinquency for males and no effect for females, variables
based on parents’ expectations and perceptions of the parent-
child relationship were significant for females, suggesting
that home socialization is more important for females than
males. Pointing to a similar conclusion in their nearly all-
white sample, Van Voorhis et al. (1988) found that the
intervening effect of home quality in the relationship
between broken homes and delinquency was significant for
females across a wider variety of self-reported delinquency
than was the case for males.
Although Wells and Rankin (1991) reported that previous
research generally does not indicate racial differences in
the relationship between broken homes and delinquency, Salts
et al. (1995) found that family variables, including time
spent away from the home and with friends, predicted self-
reported violence among blacks and whites. For both racial
categories, family structure was not a significant predictor
of violence. Similarly, Smith and Krohn (1995) found in a
longitudinal study of urban teens that family structure was
not related to general self-reported delinquency for either
blacks or whites, although it did have a direct effect on
Hispanic delinquency. Overall, Smith and Krohn reported that
family measures, including family structure and family

35
quality, explained about 10 percent of the variance in
delinquency for both blacks and whites, and about 20 percent
for Hispanics.
However, Rosen (1985) concluded that both family quality
and family structure were related to official delinquency
among blacks, whereas only structural variables were
important for white delinquency. As mentioned previously, in
their nationally representative, longitudinal study, McLeod
et al. (1994) found that the particular measure of family
structure affecting antisocial behavior depended on race.
For blacks, antisocial outcomes were associated with living
with a never-married mother. For whites antisocial behavior
was related to living with a previously married mother.
However, Thomas et al. (1996) found a higher rate of black
self-reported delinquency among those in mother-only families
that included father involvement. Whites, on the other hand,
were more delinquent in mother-only families where there was
no father involvement. Delinquency was reduced for blacks
and whites living with two biological parents.
Matsueda and Heimer (1987) found for both blacks and
whites that broken homes influence delinquency indirectly
through delinquent definitions and delinquent peers, but the
relationship is considerably stronger for blacks. Smith and
Krohn (1995) found that parent-child attachment and the
perception of parental control directly affected delinquency
for blacks and whites, but they reported no indirect effects
involving family structure as an exogenous variable.

36
Generally, although greater differences were found for
Hispanics, the relationships between family variables and
delinquency appear to be similar for blacks and whites.
A variety of ecological research has been carried out on
the relationship between family structure and delinquency,
contributing additional support for the association between
family structure and delinquency. However, any inference of
causality for corresponding individual-level measures would
not be warranted due to the use of aggregate measures in this
research. In a series of studies, Sampson and various
associates have shown a connection between family structure
and delinquency while identifying antecedent variables that
pertain to the relationship, such as inequality, joblessness,
sex ratio, and racial segregation. Sampson (1985) found that
family structure, measured by the percentage of female-headed
families in a neighborhood, strongly affected serious
property and violent victimization. Analyzing data from 171
U.S. cities, Sampson (1987) and Messner and Sampson (1991)
found that official rates of serious violence were strongly
affected by the percentage of female-headed households with
children under age 18, and more strongly affected by
percentage of female-headed households. Smith and Jarjoura
(1988) found that neighborhoods with larger percentages of
single-parent households had higher victimization rates of
violent crime (robbery and assault) and burglary.
Additionally, family structure helped explain the significant
relationship between poverty and rates of burglary. Using

37
census data for 158 cities, Shihadeh and Steffensmeier (1994)
found that the percentage of female-headed black households
had a significant effect on officially reported black
violence. Female householding itself was affected by income
inequality within black communities.
These studies and others suggest a relationship between
family structure and delinquency using a variety of
variables, while pointing to various antecedent variables
that are associated with family structure. While much of the
research reviewed earlier positions family structure as an
exogenous variable that affects delinquency through various
mediating variables, this research hypothesizes a weakened
family structure as an endogenous variable intervening
between economic difficulties and crime (see Petras and
Davenport, 1991).
Conclusions
Wherever family structure is positioned in analytical
models, throughout the literature review it has been seen
that the broken-home measure of family structure is actually
an imperfect measure of a more general family process and
that the broken-home effect is an unclear one which interacts
with other variables and depends upon context (Gove and
Crutchfield, 1982; Hennessy et al., 1978). Such ambiguity
necessitates a precise definition of family measures such as
broken home and single parenthood, and the delineation of
potentially intervening variables that are affected by a

38
broken home or single parenthood and in turn produce
delinquent outcomes for children.
One cannot conclude from the existing literature that
any one measure of family structure is the best. The only
conclusion is that the measure used should be clearly defined
and appropriate for the purposes of the research. The
present study will employ a measure emphasizing the
structural makeup of the family rather than the occurrence or
conditions of divorce. This measure is chosen with the
assumption that two parents will be better able than one
parent to provide a consistent learning environment for
children.
For example, Dornbusch et al. (1985) found that
adolescents in mother-only households are more likely to be
delinquent as a result of making decisions without parental
input (see also McLanahan and Booth, 1989). Using a similar
measure of family structure, Rosen (1985) found significant
relationships with delinquency for both blacks and whites.
Gove and Crutchfield (1982) included a similar measure in
their study that found family structure to be more important
for male delinquency, while family interaction was more
central for females. But for their equivalent family
structure measure, Van Voorhis et al. (1988) found no
significant effects on a variety of delinquency measures.
McLeod et al. (1994) also investigated single-parent families
compared to two-parent families, distinguishing between
previously married and never-married parents. Both single-

39
parent family types were related to antisocial behavior.
Thomas et al. (1996) offered another approach to measuring
family structure by the number of parents present, comparing
families with two biological parents, single-mother families
that had nonresident father involvement in child
socialization, and single-mother families without nonresident
father involvement. White males were more delinquent the
lower the level of father's involvement, while black males
were more delinquent where nonresident fathers were involved
in the family.
Dornbusch et al. (1985) and other researchers have also
indicated that the presence of stepparents may increase the
likelihood of delinquency, especially in the case of
stepfathers and their stepsons. Johnson (1986) distinguished
between both number and type of parents, finding that males
living with a mother and a stepfather were most delinquent
while females in mother-only families were most delinquent.
Coughlin and Vuchinich (1996) found that youths from
stepfamilies and mother-only families were more likely to be
arrested. These findings call for the coded separation of
two biological parents from those two-parent families
consisting of one biological parent and one stepparent.
Much of the research reviewed in this chapter points to
the variables that intervene between family structure and
crime and delinquency. Many researchers have called for the
continued explication of these and other potentially
intervening variables (e.g. Smith and Jarjoura, 1988; Wells

40
and Rankin, 1986). According to Van Voorhis et al. (1988),
family structure usually is considered to be an indirect
cause of delinquency, predisposing family members to other,
more direct influences such as lack of affection, poor
supervision, shortage of economic and time resources, and
limited opportunities for modeling, among others. Wells and
Rankin said, "Few if any analyses presume ... a direct and
automatic causal link. Rather, family structure affects
delinquent behavior through intervening interactional
processes (engendered by structural conditions)" (1986, p.
87). While acknowledging that the relationship between
family structure and delinquency is a weak one, Wells and
Rankin (1985) argue that it is consistent and there is a need
to clarify the process by which the family relates to
delinquency.
The present study will focus on social learning as
intervening processes between family structure and
delinquency. Specifically, it draws upon the social learning
theoretical perspective to explain the effects of family
structure on delinquent behavior. Various components of
social learning theory will be hypothesized to mediate the
relationship between family structure and delinquency. In
the next chapter, social learning theory will be discussed
and additional research focusing specifically on the theory
will be reviewed.

CHAPTER 3
SOCIAL LEARNING THEORY OF CRIME AND DELINQUENCY
The last chapter discussed a variety of research
findings that link family variables to delinquent behavior.
In this chapter, social learning theory will be discussed and
considered as a useful approach to understanding the
relationship between the family and the delinquent behavior
of adolescents. The theory is well suited to explaining the
relationship between family variables, including family
structure, and delinquency because it focuses on the process
of interaction, particularly within primary groups. The
family normally is the first primary group that individuals
encounter. "Social learning models emphasize the importance
of parents as reinforcers and models of socially appropriate
behavior. Parental availability, supervision, and affection
are important factors in influencing the effectiveness of
their modeling and reinforcement strategies," (1988, p. 239)
according to Van Voorhis et al. Moreover, the family
typically has a considerable impact on other primary groups
that gain significance later, such as the peer group (Jensen
and Rojek, 1992). Thus, it is reasonable to expect that
changes in family structure may affect youths via social
learning processes in their interaction with parents, and
subsequently, with friends. The nature of the social
41

42
learning that occurs in these important primary groups may be
expected to significantly affect whether behavior is
conforming or delinquent.
Social Learning Theory
Social learning theory was developed by Akers and
Burgess (Burgess and Akers, 1966) as a reformulation of
Sutherland’s (1947) theory of differential association.
Akers (1973; 1977; 1985; 1997; 1998) has continued to develop
and test the theory. Central to the theory is the idea that
delinquency is learned in interaction with others, primarily
within intimate, personal groups such as family and friends.
According to Sutherland, "A person becomes delinquent because
of an excess of definitions favorable to violation of law
over definitions unfavorable to violation of law" (1947, p.
6). Sutherland said that this process of differential
association varies by frequency, duration, priority, and
intensity. Thus, the commission of a criminal act results
from a complex ratio wherein the influence of criminal
patterns and definitions exceeds the influence of conforming
patterns.
Sutherland's theory was one of the first to suggest that
deviance results from the effects of environmental conditions
on biologically and psychologically normal individuals (Void
and Bernard, 1986). According to Sutherland, "The process of
learning criminal behavior by association with criminal and
anti-criminal patterns involves all of the mechanisms that

43
are involved in any other learning" (1947, p. 7). If
criminal behavior is learned as is other behavior, then
Sutherland's theory of differential association also directs
us toward the social context or structure (differential
social organization as Sutherland would put it) in which
learning takes place for a further explanation of crime and
delinquency. Family structure would be one context that may
indirectly affect delinquency via the process of differential
association.
Burgess and Akers (1966) broadened Sutherland's
differential association by integrating it with behavioral
learning theory, particularly the concept of differential
reinforcement, whereby behavior is conditioned by rewards and
punishments. Their theory originally was called differential
association-reinforcement, but now is usually referred to as
social learning theory.
The theory is intended to apply to any form of deviant
behavior, including criminal and delinquent behavior. "The
basic assumption in social learning theory is that the same
learning process, operating in a context of social structure,
interaction, and situation, produces both conforming and
deviant behavior" (Akers, 1998, p. 50). Akers said that the
difference between conforming and deviant behavior lies in
the balance of influences on behavior. Social learning
theory focuses on four central concepts to account for these
influences. These are differential association, definitions,
imitation of models, and differential reinforcement. While

44
these four concepts influence conformity in one direction, in
his central proposition of social learning theory Akers
describes the way in which they combine to instigate and
strengthen deviance in the other direction:
The probability that persons will engage
in criminal and deviant behavior is
increased and the probability of their
conforming to the norm is decreased when
they differentially associate with others
who commit criminal behavior and espouse
definitions favorable to it, are
relatively more exposed in-person or
symbolically to salient criminal/deviant
models, define it as desirable or
justified in a situation discriminative
for the behavior, and have received in
the past and anticipate in the current or
future situation relatively greater
reward than punishment for the behavior.
(Akers, 1998, p. 50; emphasis added)
Rewards and punishment are the alternate ways in which
differential reinforcement is distributed. The four concepts
and other aspects of the theory will be explained in greater
detail below.
Differential Association
Differential association in social learning theory
encompasses interaction with others in intimate primary
groups, as well as in secondary groups and indirect
association with reference groups of individuals throughout
the community and the mass media. Modalities of association
combine to determine the effects of various associations.
"Those associations which occur first (priority), last longer
(duration), occur more frequently (frequency), and involve

45
others with whom one has the more important or closer
relationships (intensity) will have the greater effect"
(Akers, 1994, p. 96). In addition to this interactional
dimension there is a normative dimension of differential
association, which more widely refers to the different
patterns of norms and values to which individuals are
exposed.
"Differential association refers to the process whereby
one is exposed to normative definitions favorable or
unfavorable to illegal or law-abiding behavior" (Akers, 1994,
p. 96). Additionally, it is through differential association
that one is exposed to other mechanisms of social learning,
including models to imitate and differential reinforcement
for criminal or conforming behavior. These three concepts—
definitions, imitation, and differential reinforcement—will
be discussed next.
Definitions
Definitions are evaluative norms, attitudes, or
orientations regarding a particular behavior as good or bad.
A positive definition exists when behavior is considered good
while a negative definition exists when behavior is seen as
undesirable. A neutralizing definition exists when behavior
is justified or excusable. This type of definition includes
techniques of neutralization and rationalizations that are
used to disclaim responsibility or fault.

46
Conventional definitions about criminal behavior are
negative toward deviant behavior. However, the more one
holds ideas favorable to crime or deviance, the more one is
likely to engage in criminal or deviant behavior. In this
case the definitions about criminal behavior are positive.
Thus, the use of the terms positive and negative should not
be mistaken as synonymous with conforming and deviant,
respectively. For example, at the decisive moment a
conforming person may define conformity as positive and
deviance as negative while a nonconforming person may define
deviance as positive and conformity as negative. More
typically however, definitions favorable to crime and
delinquency are "conventional beliefs so weakly held that
they provide no restraint or are positive or neutralizing
attitudes that facilitate law violation in the right set of
circumstances" (Akers, 1994, p. 98).
Definitions are beliefs that may be general or specific.
General beliefs include religious, moral, and other
conventional ideas favorable to conformity and unfavorable to
deviance or criminality (Akers, 1994). Specific beliefs are
held about particular acts or types of acts. Thus, one may
believe that it is wrong to do harm to others but that it is
all right to harm those who are the source of frustration or
disagreement.
Definitions are learned through imitation and
reinforcement, and once learned, they serve as discriminative
stimuli for conforming or deviant behavior. According to

47
Akers, "Discriminative stimuli are stimuli which become
associated with reinforcement. In addition to the
reinforcers, other stimuli are ordinarily present when
behavior is reinforced—the physical surroundings, one's own
feelings, others' behavior, one's own and others' spoken
words, and so on" (1985, p. 49). Discriminative stimuli
occur before or as the behavior occurs, signaling the actor
as to the appropriateness of the behavior and whether
reinforcement or punishment is forthcoming.
Imitation
Imitation is the process of acquiring social behavior,
whether conforming or deviant, by observing and modeling the
behavior of others. Models, like definitions, may be drawn
from primary, secondary, or reference groups. Additionally,
imitation may occur immediately, at some later time, or not
at all. Imitation is a complex process as its likelihood of
occurrence depends on the characteristics of the model as
well as the model's observed behavior and the outcomes
produced by that behavior. "[Imitation] is more important in
the initial acquisition and performance of novel behavior
than in the maintenance or cessation of behavioral patterns
once established, but it continues to have some effect in
maintaining behavior" (Akers, 1994, p. 99).

48
Differential Reinforcement
The primary learning mechanism in social learning theory
is operant conditioning, or instrumental learning, wherein
behavior is shaped by the stimuli which follow the behavior.
Operant conditioning is distinguished from respondent
conditioning, wherein behavior is shaped by antecedent or
eliciting stimuli, in that it is "developed, maintained, and
strengthened depending on the feedback received or produced
from the environment" (Akers, 1985, p. 42). Thus, while a
respondent is controlled by the preceding stimulus and
unaffected by the outcome it creates for the environment, an
operant is controlled by the stimulating reinforcement and
punishment following or contingent on it.
Positive reinforcement occurs as rewards, such as
approving words, money, or pleasant feelings, increase the
probability that an act will occur or be repeated. The
likelihood of an act occurring or being repeated also is
increased through negative reinforcement, when one avoids
unpleasant consequences. For example, a motorist who exceeds
the speed limit will likely continue to do so, and perhaps to
a greater degree, so long as a citation is not received.
Punishment also may be positive or negative. Positive
punishment occurs when unpleasant consequences are attached
to a behavior, such as when the speeding motorist receives
that citation. Negative punishment occurs when pleasant

49
consequences are removed. For example, a worker receives
negative punishment when a paycheck is docked.
Deviant or conforming behavior may be continuously
reinforced but is more likely reinforced on an intermittent
schedule, the uncertainty of which helps account for the
variety, complexity, and stability of behavior.
Reinforcement depends on the values of those with whom one
interacts, as behavior is regarded differently across time
and place. Modalities of reinforcement—amount, frequency,
and probability—affect the extent of reinforcement. The
greater the amount of reinforcement for one's behavior, the
more frequently it is reinforced, and "the higher the
probability that behavior will be reinforced (as balanced
against alternative behavior), the greater the likelihood
that it will occur and be repeated" (Akers, 1994, p. 98).
Temporal Sequence and Reciprocal Effects in Social Learning
Theory
Social learning theory has most often been posited as
preceding delinquent behavior. However, Akers maintains that
the relationship between social learning variables and
delinquent behavior does not simply occur in one direction.
Akers stresses that "social learning is a complex process
with reciprocal and feedback effects" (1994, p. 99).
Akers (1998) explains how these reciprocal and feedback
effects are integrated into social learning theory. First,
association with family members and friends typically

50
precedes the commission of delinquent acts, and
reinforcement, modeling, and exposure to definitions occur
within these groups prior to the onset of delinquency.
Second, after the initial event imitation becomes less
important while the balance of the other social learning
variables continue to be used in determining repetition.
Third, actual consequences of the event (social and nonsocial
reinforcers and punishers) affect the probability and
frequency of repeated events. Consequences include the
actual effects of the behavior itself, the actual reactions
of others present at the time or who find out about it later,
and the anticipated reactions of others not present or
without knowledge of the behavior. Fourth, both the overt
behavior and the definitions favorable or unfavorable to it
are affected by these consequences, which in turn affect
further behavior. Definitions, for example, may be applied
retroactively as justification for a previous act. At the
same time, such justification precedes future acts as it
variably mitigates sanctioning (punishing reinforcement) by
others or oneself. Finally, progression into more frequent
or sustained behavior again occurs in the context of
reinforcement, modeling, and definitions.
Differential association shares a complex relationship
with delinquency, according to Akers (1998). Social learning
theory recognizes that youths may choose to associate with
friends based on a similarity in delinquent behavior that
already exists. Similarly, youths confined for delinquent

51
offenses will have little choice but to associate with
delinquent peers. However, because associations are
typically formed around circumstances and attractions such as
neighborhood proximity that have little to do with co¬
involvement in deviant behavior, delinquent behavior most
often results from delinquent peer association (Akers, 1994).
Akers (1998) concludes that there is both a tendency for
persons to choose to interact with others based on behavioral
similarities (known as "flocking" or "selection") and a
tendency for persons who interact to have mutual influence on
each other's behavior (known as "feathering" or
"socialization"). Both are part of the social learning
process and both are explained by the same variables.
Social Structure and Social Learning: The Case of the Family
and Delinquency
Akers (1998) has presented a Social Structure-Social
Learning (SSSL) theory of crime and delinquency. The theory
basically assumes that social learning is the primary process
linking social structure to individual behavior. It
primarily proposes that "variations in the social structure,
culture, and locations of individuals and groups in the
social system explain variations in crime rates, principally
through their influence on differences among individuals on
the social learning variables" (Akers, 1998, p. 322). Social
structural variables that have an effect on individual
behavior include various broad measures of society and

52
community; common demographic variables such as age, sex,
class, and race; and other more immediate variables such as
family and peer group.
Social learning theory is useful in the explanation of
the effects of family structure on delinquency because the
family and peer groups are identified in the theory as the
most significant primary groups implicated in differential
association. Thus, the components of social learning theory
may be presented as intervening variables between the social
structural variable of family structure and the individual-
level variable of delinquent behavior. Differences in the
effects of family structure on rates of delinquency depend on
the extent to which each type of family is able to "provide
socialization, learning environments, and immediate
situations conducive to conformity or deviance" (Akers, 1994,
p. 101). For example, a single-parent family may be less
likely than a two-parent family to provide a consistent
learning environment, and a single parent may be less able
than two parents to oversee children's behavior. More
precisely, a two-parent family may offer greater
opportunities for differential association, the transmission
of values or definitions, the imitation of parental role
models, and differential reinforcement for conforming
behavior, compared to smaller family structures. The linking
of family structure to delinquent behavior, therefore, can be
seen as a special case of Akers' SSSL model. The present
study is one way of testing the SSSL model, as it will focus

53
on social learning theory as a link between social structure
and individual delinquent behavior by hypothesizing a
mediating effect of social learning variables on the
relationship between family structure and delinquency.
Research on Social Learning Theory
A variety of researchers writing on crime and
delinquency have cited social learning theory as a valuable
perspective for explaining deviance, and the findings of
numerous studies have supported or are consistent with the
theory (see Akers, 1998).
Social Learning Theory and Delinquency
Akers and various associates have demonstrated support
for the relationship between all four social learning
concepts and delinquency in a variety of studies, including
tests of social learning theory by itself and in comparison
with other theories (Akers, 1994).
Akers et al. (1979) provided a thorough test of the
explanatory power of social learning theory in relation to
self-reported alcohol and marijuana use and abuse among
adolescents. A combined 15 social learning variables
measuring all four concepts of social learning theory—
differential association, definitions, differential
reinforcement, and imitation—explained 55 percent of the
variance in use of alcohol and 68 percent of the variance in
use of marijuana in a sample of more than 3,000 adolescents.

54
For alcohol and marijuana abuse, the variance explained by
the social learning variables was 32 percent and 39 percent,
respectively.
Akers et al. (1979) distinguished between social
differential reinforcement (actual and anticipated) and a
second group of differential reinforcement measures which
combined social with nonsocial factors. Comparing the five
subsets of variables, differential association was most
effective in explaining variance in alcohol and marijuana use
and their abuse. Consistent with the findings of numerous
other studies, the most important single variable for
explaining both the use and abuse of alcohol and marijuana
was differential peer association. For substance use, the
second-most effective social learning concept was
definitions, followed by the combined social and nonsocial
subset of differential reinforcement measures, social
differential reinforcement, and imitation. A similar ranking
was found for abusers, although social-nonsocial
reinforcement was more important than definitions among
alcohol abusers. According to Akers et al., "In substance
abuse the user comes more and more to respond to direct
reinforcement, especially from the drug effects themselves"
(1979, p. 650).
Of the items measuring definitions, the respondent's own
approval or disapproval of use had the strongest effect on
both alcohol and drug use, and drug abuse, over neutralizing
definitions and attitudes about the law. For alcohol abuse,

55
attitudes about the law was more important. Of the items
measuring social differential reinforcement, friends'
rewarding or punishing reactions had a stronger effect on
substance use than parents' rewards or punishment and
deterrence items (being caught by parents or the police).
Friends ' reactions was also important for substance abusers,
although marijuana abusers were equally concerned about the
interference it would have on other activities such as school
and athletics. Additionally, parental reactions was found to
be curvilinearly related to abuse, with a lower probability
of abuse being associated with parents who react moderately.
With the exception of imitation, four of the five
subsets explained a substantial proportion of variance in
both alcohol and marijuana use. The authors took these
findings as support for the theory as a whole. "Friends
provide social reinforcement or punishment for abstinence or
use, provide normative definitions of use and abstinence,
and, to a lesser extent, serve as admired models to imitate"
(1979, p. 644).
Akers and Lee (1996) found that social learning theory
explained adolescent smoking. Using five-year longitudinal
data, measures of differential association, definitions, and
differential reinforcement significantly affected future
smoking. Also consistent with social learning theory, the
authors identified reciprocal effects of smoking on
differential association, although similar effects were weak
or nonsignificant for definitions and differential

56
reinforcement. According to Akers and Lee, "As smoking
behavior develops, it is shaped by association with peers,
exposure to their normative definitions of smoking, and
social reinforcement, but then over time, one's own smoking
behavior comes to exert influence over patterns of
association with friends" (1996, p. 336). Menard and Elliott
(1994) identified reciprocal relationships between delinguent
peer association, attitudes toward delinquency, and self-
reported minor offending and index offending. However, the
stronger causal direction was clearly indicated by the
effects of social learning on delinquency.
Boeringer and Akers (1993) found that social learning
variables also explained violent behavior in the form of
sexual aggression among college males. They reported that
social learning models, including measures of all four of the
theory's central concepts, accounted for over 20 percent of
the variance in sexually aggressive behavior. This outcome
variable included nonphysical coercion, use of drugs or
alcohol as a coercive sexual strategy, and attempted or
completed rape with or without force. The social learning
models also accounted for 40 to 55 percent of the variance in
rape proclivity.
Additional support for social learning theory has come
from other researchers. In a sample of ninth-grade students,
Winfree et al. (1994) found that pro-gang attitudes
contributed significantly to the prediction of several self-
report crime measures, including alcohol and drug use,

57
property offenses, and violence. Analyzing eight waves of
the National Youth Survey, Elliott (1994) found that
delinquent peers, attitudes toward deviance and neutralizing
definitions (Elliott's "peer normlessness") all had strong
direct effects on the onset of violent offending.
Also using eight waves from the National Youth Survey,
Esbensen and Elliott (1994) found that social learning
variables accounted for the initiation of drug use. A
measure of differential association, exposure to drug-using
peers, was most strongly related to the initiation of
alcohol, marijuana, and other drugs such as amphetamines,
cocaine, and heroin. Differential reinforcement variables
also were significant in their findings. The anticipated
disapproval of friends decreased the likelihood of initiating
alcohol and marijuana use, while the anticipated disapproval
of parents decreased the likelihood of beginning alcohol use.
A similar measure of the anticipated disapproval of parents
was found to increase the likelihood of initiating alcohol
use, although it also slightly decreased the probability of
onset of marijuana and other drug use. Attitudes against
drug use, a measure of definitions, decreased the probability
of starting to use drugs such as heroin and cocaine. Another
measure of definitions, attitudes against delinquent
behavior, was not significant for drug use.
Esbensen and Elliott (1994) found that social learning
variables generally were not as significant in terms of the
termination of drug use, although exposure to drug using

58
peers reduced the likelihood of discontinuing the use of
marijuana and other drugs, and negative attitudes toward
delinquency increased the likelihood of termination of
marijuana and other drugs. In summary, Esbensen and
Elliott's results indicate variably significant effects on
drug use for a variety of social learning concepts, including
differential association, definitions, and differential
reinforcement. Measures of neutralizing definitions were not
significant.
Using two waves of the National Youth Survey, Warr and
Stafford (1991) found that differential association, as
measured by friends' delinquency, was strongly related to
self-reported cheating, petty larceny, and marijuana use.
Friends' delinquency also affected self-reported delinquency
indirectly through respondents' definitions. Friends'
attitudes, an indirect measure of differential reinforcement,
indirectly affected delinquency through respondents'
attitudes (definitions), but the direct relationships between
friends' attitudes and delinquency were weak or
nonsignificant.
Elliott et al. (1989), using the first three waves of
the National Youth Survey, found involvement with delinquent
peers to have the strongest effect on both general
delinquency and index offending rates. Other theoretical
variables affected both types of delinquency only indirectly
through involvement with delinquent peers. Attitudes toward
delinquency and neutralizing definitions ("normlessness" in

59
their usage), for example, were found to affect involvement
with delinquent peers but had no direct effect on
delinquency. Elliott et al. (1989) concluded in the test of
their integrated theoretical model that learning theory
provides the principal explanation for delinquency and
substance use. Using waves three through five of the NYS,
Menard and Elliott (1994) also found delinquent peer
association to significantly affect minor offending and index
offending, while the effects of attitudes toward delinquency
were weak and inconsistent.
In this section it has been shown with a variety of
cross-sectional and longitudinal research that social
learning variables are related to various delinquent
outcomes, including smoking, alcohol and drug use, general
delinquency and violent behavior. Differential (peer)
association, typically measured as having delinquent or
substance using friends, emerges as having the clearest
relationship with delinquency of the four social learning
concepts. In nearly every study reviewed, friends'
delinquency or substance use was most strongly related to
one's own delinquency or substance use. Indeed, in their
review of the empirical literature on broken homes and
delinquency, Wells and Rankin note, "Peer influence
constitutes the variable most consistently and strongly
predictive of delinquency across a variety of studies" (1985,
p. 265).

60
Various measures of definitions and differential
reinforcement also indicate direct effects on delinquent
behavior. A few studies suggest that some social learning
variables affect delinquency indirectly through other social
learning variables. For example, Elliott et al. (1989) found
that measures of definitions, including neutralizing
definitions, affected delinquent peer association. For Warr
and Stafford (1991), however, delinquent peer association and
the differential reinforcement of peers (friends' attitudes)
affected one's own definitions.
Family, Social Learning Theory, and Delinquency
Research has demonstrated the intervening effects of
social learning variables between various structural
variables and delinquency. For example, Krohn et al. (1984)
found that social learning variables explained much of the
variance in self-reported alcohol and marijuana use across
urban and rural community types. Akers and Lee (1997)
reported that social learning variables explained the effects
of age for adolescent marijuana users. Boeringer et al.
(1991) found that social learning variables mediated the
effects of fraternity membership on sexually aggressive
behavior among college males. In this section research will
be reviewed which links family variables, including family
structure, to social learning variables and delinquency.
Thereby it will be shown how social learning provides the
process through which the social structure of the family has

61
an impact on individual behavior. "As the main conventional
socializer against delinquency and crime, the family provides
anti-criminal definitions, conforming models, and
reinforcement for conformity through parental discipline,”
according to Akers (1998, p. 165). Later, he said, family
socialization remains important as it competes against or is
undergirded by socialization in peer groups.
Consistent with the findings reported above on the
primacy of differential peer association, Elliott et al.
(1985) found delinquent peer association to have a strong
effect on self-reported serious and nonserious delinquency.
Using the first three waves of the National Youth Survey, the
authors also found that weak and antisocial relationships
with family members and friends contributed to an increased
involvement with delinquent peers. Definitions, measured as
attitudes toward deviance, had little effect on delinquency.
Neutralizing definitions did not have a direct effect on
delinquency although the variable was associated with an
increased involvement with delinquent peers.
Heimer (1997), also using the first three waves of the
National Youth Survey, found that definitions relating to
interpersonal violence were associated with subsequent
violent delinquency among males. Definitions mediated the
delinquent effect of concurrent association with delinquent
peers, which mediated the effect of prior parental
supervision of friendships. Patterson and Dishion (1985)
likewise suggested that parents' monitoring practices

62
contribute to the parents 1 failure to discipline antisocial
or delinquent behavior and to their lack of control over a
child's association with delinquent peers. Applying social
learning principles with an emphasis on the roles that
parental reinforcement and punishment play in socialization,
Patterson and Dishion found that parental monitoring had a
direct effect in reducing male delinquency, as well as an
indirect effect by ways of delinquent peer association and
social skills. According to Patterson and Dishion, "The
parents' failure to track provides ample opportunity for the
child to engage in delinquent acts and to seek out deviant
peers, who in turn further exacerbate the problem” (1985, p.
75).
Warr (1993) also reported a strong relationship between
association with delinquent peers and self-reported
delinquency. His findings showed that the strength of this
relationship was significantly reduced by the quantity of
time spent with one's family. Compared with what has been
seen to be the typical sequencing, in this case the variables
representing family and social learning are inversely
ordered. Warr concluded that peer influence may be overcome
only by avoiding the company of delinquent peers altogether.
"This may be achieved either by inhibiting the formation of
delinquent friendships in the first place (as attachment to
parents seems to do) or by reducing the time that adolescents
spend with their delinquent friends" (1993, p. 259).

63
The studies reviewed to this point have employed family
variables relating to family quality. A few studies have
investigated the relationships between family structure,
social learning, and delinquency. In a cross-sectional test
of differential association theory using Richmond Youth
Project data, Matsueda and Heimer (1987) found that
delinquent peer association had the largest total effect on
self-reported male delinquency. Their findings showed that
broken homes, measured as mother or father absence, influence
delinquency by directly fostering an excess of prodelinquent
definitions. Matsueda and Heimer said that broken homes also
influence delinquency more indirectly "by attenuating
parental supervision, which in turn increases delinquent
companions, prodelinquent definitions, and ultimately,
delinquent behavior" (1987, p. 836).
Heimer and Matsueda (1994) found that delinquent peer
association and attitudes toward delinquency both were
related to self-reported delinquency among males. In a
longitudinal analysis using the first three waves of the
National Youth Survey (NYS), they reported that these social
learning variables (at wave two) mediated the effect of
family structure (measured as father absence at wave one) on
serious and nonserious delinquency (at wave three). Heimer
(1997), also using the first three waves of the NYS, found
support for the mediating effects of social learning on male
violence. Her findings indicated that broken homes (those
without both biological parents present) were related to less

64
supervision, which led to an increase in subsequent
delinquent peer association and then violent definitions,
which finally led to subsequent violent delinquency.
Using longitudinal data from Houston-area junior high
schools, Chen and Kaplan (1997) found that delinquent peer
association and definitions, including attitudes toward
deviance and neutralizing definitions, mediated the effect of
family structure on self-reported delinquency. Measured as
having both biological parents present, family structure also
maintained a modest, direct effect on delinquency. Chen and
Kaplan reported that separate analyses yielded similar
results for both serious delinquency and general delinquency,
which included drug use, theft, and violence. A unique
feature of their research was the eight to 16 years that
elapsed between their measurement of the independent
variables and delinquency. This arrangement permitted the
investigation of the effects of family structure and the
mediating variables in early adolescence on deviance in young
adulthood.
The literature reviewed in this section has
demonstrated, both cross-sectionally and longitudinally, the
mediating effects of social learning variables between family
variables, including family structure, and delinquency.
Again, delinquent peer association emerges as the dominant
social learning variable, playing a significant mediating
role in each of the studies reviewed. Definitions also
played a significant intervening role in about half of these

65
studies, sometimes mediating the effect of delinquent peer
association on delinquency. The self-reported delinquency
measures in these studies included nonserious and serious
general delinquency, and in one case, violent delinquency.
Hypotheses
Based on the empirical findings of the literature
reviewed to this point and the reasoning of social learning
theory, hypotheses may be deduced about the concepts to be
investigated. Below are four hypotheses specifying the
expected relationships between family structure, social
learning, and delinquency.
(1) Research has demonstrated the importance of families
as socializing agents both in the initial stages of life and
continuing through childhood, adolescence and into adulthood.
Although less clear, research also has suggested that a
reduced number of adult socializers increases the likelihood
of a reduced quality of socialization for a child. This
relationship is suggested by empirical evidence showing that
children from single-parent homes are more delinquent than
children from two-parent homes.
Therefore it is first hypothesized that family structure
predicts delinquency. Family structure as used here refers
to the number and type of parents in the household, as
contrasted to measures focusing on the occurrence of a
divorce or the absence of a particular parent. Specifically,
youths living with a single parent are expected to be more

66
delinquent than youths living with two parents. It also is
expected that youths living with neither parent in a non-
parental household will be more delinquent than youths living
with a single parent. Additionally, youths living with one
biological parent and one stepparent are predicted to have
delinquency levels between those of youths living with two
biological parents and those living with one parent.
(2) The literature review has shown that family
structure has an impact on social learning variables. For
example, Heimer and Matsueda (1994) found that father absence
increased the level of definitions favorable to delinquency
and increased the extent of associations with delinquent
peers. Other studies suggested that a reduced number of
parents resulted in reduced supervision, which resulted in an
increase in association with delinquent peers and the
adaptation of delinquent attitudes. Social learning theory
recognizes the family as an important source of both
conforming and deviant reinforcement, models, attitudes, and
association. The theory would suggest that living with fewer
parents would mean a weakened parental association with
conforming patterns, the absence of parental reinforcement,
the reduction of parental influence, and the loss of parental
role models. Thus, single parents will be less likely than
two parents to instill conforming attitudes and discourage
delinquent ones, guide their children toward conforming
associations and away from deviant ones, and to reinforce
conformity and punish delinquency.

67
Accordingly, it is hypothesized that family structure
affects the social learning process conducive to delinquency.
It is expected that having fewer biological parents will be
correlated with scores on social learning measures in the
delinquent direction, such as having more delinquent friends
and adhering to delinquent attitudes.
(3) Perhaps the strongest of the research findings
regarding social learning theory was that delinquent peer
association leads to delinquent behavior. Research also
indicated that having prodelinquent definitions increases the
likelihood of delinquency. Social learning theory holds that
one who has an excess of delinquent associations, adheres to
an excess of prodelinquent definitions, is reinforced toward
delinquent behavior and away from conformity, and has
delinquent role models will be more likely to engage in
delinquency.
Specifically, it is hypothesized that the more
delinquent one's friends are, and the more delinquent friends
one has, the more likely the respondent will also be
delinquent. Similarly, the more delinquent definitions one
possesses, the more likely he or she will be to engage in
delinquent behavior. Moreover, the weaker the punishment for
delinquent behavior, and the stronger the reinforcement for
it, the greater the likelihood of delinquency.
(4) Akers proposes that social learning theory is
capable of explaining how social structures such as the
family shape individual behavior (1997; 1998). The last

68
section of the preceding literature review cited several
studies that demonstrated the mediating role of social
learning theory. In particular, differential peer
association and definitions were shown to be important
mediating concepts between family structure and delinquency.
Therefore the fourth and final hypothesis, and the one
most central to the present research, is that social learning
variables will mediate the effect of family structure on
delinquency. Specifically, the introduction of social
learning variables into the analysis will reduce the strength
and significance of the relationship between family structure
and delinquency. It is expected that an attenuated family
structure will be associated with an increase in delinquent
friends, an increase in delinquent attitudes, and an increase
in rewards for delinquency along with a decrease in
punishment for delinquency. It is expected that these
conditions of social learning will be associated with an
increase in delinquent behavior, and will substantially
reduce or render nonsignificant the relationship between
family structure and delinquency.
The Present Research
The intent of the present research is to further
investigate the mediating effects of social learning between
family structure and delinquency. This will be accomplished
by expanding upon previous research in a variety of ways.
Where social learning variables have been included in

69
previous research on this topic, they typically have appeared
one or two at a time and usually as part of another
theoretical construction. As an explicit test of social
learning theory, this study will employ six social learning
measures of three of the theory's four central components to
provide a test of more complete social learning/social
structure models.
Secondly, most of the previous research in this area has
combined nonserious and serious offenses into a general
delinquency measure. Indeed, measures of serious delinquency
are much less common in self-report studies than nonserious
or general delinquency measures. One exception is the
research by Heimer (1997), reviewed above, that focuses on
violence. The present study will employ an index of serious
delinquency combining serious property and violent offenses.
Thirdly, the present study will expand on previous
research by using a different measure of family structure.
Previous studies in this area have used dichotomous measures
based on the presence of two biological parents or on father
absence. Here family structure will be based on the number
of parental figures in the household, be they biological or
stepparents. The distinction is noteworthy because this type
of measure differentiates between single-mother families and
two-parent, mother-stepfather families. Given the importance
of parental monitoring or supervision suggested by several of
the studies reviewed in the previous section, and a presumed
difference between the monitoring ability of a single parent

70
versus two parents, the number of parental figures in the
household may have significant consequences for social
learning and delinquency.
Finally, the present study differs from previous
research by using more recent longitudinal data on males and
females. While a few of the studies cited above employed the
first three waves of the National Youth Survey, this study
will use waves three through five of the NYS.

CHAPTER 4
METHODOLOGY
Data
The purpose of the present study is to conduct a
nationally representative, longitudinal test of social
learning as a mediating process between family structure and
serious, self-reported delinquency. The study employs wave
one (1976) and waves three through five (1978, 1979, 1980) of
the National Youth Survey (Elliott et al., 1985; Elliott et
al., 1989). Hereafter, waves three through five will be
referred to as years one through three, respectively. Use of
the first wave, 1976, is limited and separately identified
(see below).
In addition to the extensive work of Elliott and his
associates, various waves of the NYS data set have been used
by different researchers. These include Heimer (1997),
Paternoster and Brame (1997), Heimer and Matsueda (1994),
Lauritsen (1993), Warr (1993), and Warr and Stafford (1991).
For the present research, tape data on the NYS are obtained
through the Inter-University Consortium for Political and
Social Research (ICPSR).
The NYS uses a panel design with a national probability
sample of households in the continental United States based
71

72
upon self-weighting, multistage, cluster sampling. The
sample contained 2,360 eligible youth aged 11-17 in 1976, the
year of the initial wave. Of those eligible, 1,725 (73%)
agreed to participate in the study and completed interviews,
which were conducted for each year in the first three months
of the following year. Elliott et al. (1989) report that the
representativeness of the sample with respect to reported
delinquency and other variables was not significantly
affected by nonparticipating youth or by panel attrition.
The age range for the present study is 13-19, as
determined by the waves selected and as a result of excluding
those respondents aged 18 and over at the time of family
structure measurement in year one or year two. The dependent
variable of delinquent behavior is variously measured in year
two (1979) and year three (1980). The effective sample size
for the present study ranges from 1,020 to 1,200 (see Table
3, below). There are roughly 250 respondents at each age.
Further details regarding the design of the present study and
sampling distributions will be presented below.
Elliott and Ageton (1980) note that self-report measures
and official statistics typically offer different findings on
delinquent behavior and its social correlates. While both
types of data reveal significant age and sex differences, the
magnitude of these differences is often smaller with self-
report data than with official statistics. Also, self-report
studies generally have found fewer differences in delinquent

73
behavior by race or class than are found in studies of
official delinquency.
Based on these differences in findings, self-report data
have long been considered useful for eliminating the
potential bias of official statistics wherein minority or
lower-class youths are more likely to be labeled delinquent
(see Hirschi, 1969). Likewise, youths from broken homes have
been found to be overrepresented in official statistics
(Jensen and Rojek, 1992; Wells and Rankin, 1985) and it may
be that these reflect systematic disparities in official
decisions about youth from broken homes versus those from
intact homes. Thus, the self-report survey would be a more
valid approach to the study of the relationship between
family structure and delinquency.
However, Elliott and Ageton (1980), using wave one of
the NYS, found significant race and class differences with
self-reported general delinquency. They attributed these
findings to the extended frequency range used on their
delinquency measures. Hindelang, Hirschi, and Weis (1979)
similarly suggest that the discrepancy between official
statistics and self-report surveys, with respect to sex,
race, and class, is largely illusory, at least for the
relatively minor types of delinquency typically found in
self-report data. They concluded that the apparent
differences between self-report data and official statistics
are a result of these methods effectively measuring different
behavior. Addressing these claims, Akers et al. (1981)

74
distinguished between serious and nonserious self-reported
delinquency, and between low- and high-frequency offenders.
Their findings suggested that previous self-report studies
were correct in their findings of nonsignificant differences
in delinquency across sociodemographic variables such as race
and class.
Whatever the extent of discrepancy between the two
methods of data collection, one inherent strength of self-
report surveys is their ability to avoid the potential bias
of which official statistics may be susceptible. They also
are useful for the measurement of "hidden" crime and
delinquency that escapes the notice of the justice system
altogether and therefore does not show up in police reports.
Self-reports also permit the avoidance of a self-selective
sample of all delinquent youths wherein the dependent
variable has little variance. On the other hand, self-report
surveys have been criticized for a number of weaknesses
involving issues of reliability and validity.
With the National Youth Survey, Elliott and his
associates attempted to overcome the weaknesses common to
self-report data. Elliott et al. (1989) suggest that while
self-report measures generally appear to be reasonably valid
and reliable, the reporting of trivial events may be the
greatest threat to the validity of self-report surveys.
Typically a general delinquency measure will include many
trivial acts that result in the unrepresentativeness of the
measure. The NYS includes serious index offenses of property

75
and violent delinquency and these are the exclusive domain of
this study. The reporting of trivial events is likely to be
a smaller problem when measuring strictly the most serious
offenses.
Another threat to the validity of self-reports is recall
error. The NYS has recall periods of one year with the
Christmas holiday providing specific, memorable reference
points for the respondents. Additionally, the serious
delinquency that is the focus of this study may be better
remembered than less serious acts. The year-long periods
also eliminate the possibility of seasonal variation.
Another threat to validity is deliberate falsification.
Elliott et al. (1989) report that much evidence on self-
report measures indicates that this is a relatively rare
problem. Akers et al. (1983), for example, used a
biochemical measure of smoking behavior to confirm that
respondents reported their smoking quite accurately. Given
the present focus on serious delinquency, deliberate
falsification may reasonably be considered a greater
potential problem. The NYS personal interviews typically
took place in the privacy of the respondents' homes. Privacy
and confidentiality were protected and guaranteed by official
certificates from the United States government.
A variety of methods were used to analyze the validity
of delinquency scales from wave one through six of the NYS
(Elliott et al., 1989). These included an analysis of
detailed follow-up questions, an analysis of the proportion

76
of trivial responses that did not qualify as delinquency, and
a comparison of self-reported delinquency with arrest
records. Each of the methods used generally confirmed the
validity of the delinquency scales. These scales included a
variety of item groupings based on type and seriousness of
offense. The felony theft scale and the felony assault
scale, comprised of a combined seven offenses, contained five
of the six serious delinquency measures included in the
present study.
The reliability of self-report measures is generally
well established (Elliott et al., 1989; Hindelang, Hirschi,
and Weis, 1979). Using the test-retest method of reliability
assessment on wave five (1980) of the NYS, Huizinga and
Elliott (1986) reported the highest reliability for index
offenses. Elliott et al. reported that for their delinquency
scales from waves one through six of the NYS, "reliabilities
were adequate to excellent for the total sample and all
demographic subgroups (e.g., test-retest reliabilities ranged
from .7 to .9 in the total sample)” (1989, p. 15).
Measurement of variables
Delinquency Index
The dependent variable is self-reported delinquent
behavior (SRD), combining self-reports of violent and
property offenses. The violent offenses include personal

77
attack, sexual assault, and use of force on another person.
The property offenses include motor vehicle theft, theft
involving a value of $50 or more, and breaking into a
building or vehicle. The delinquency index is a three-point
scale coded as follows: no reported property or violent
offenses (0); at least one reported occurrence of either
property or violent delinquency (1); and at least one
occurrence of both property delinquency and violent
delinquency (2). See Appendix A for a complete description
of the delinquency items. Alternatively coding delinquency
0-6, representing the total number of delinquent activities
engaged in by respondents, produced no more variation in the
dependent variable and yielded a weaker correlation with
family structure. Subsequent analyses indicated that
findings were not significantly different for this variation
of the dependent variable compared to the three-category,
ordinal coding described above and employed in the analyses
reported below.
The scale used here as a measure of serious delinquency
follows the constructions of Elliott and other researchers
employing NYS data. For example, the three items measuring
property delinquency were included by Elliott and Huizinga
(1983, p. 156) in their "felony theft" category, while the
three items measuring violent delinquency were used by
Elliott (1994, p. 3) to characterize "serious violent
offenders.

78
Family Structure
The primary independent variable is the respondent's
family structure, using a measure focusing on the parental
makeup of his or her household. Recall from the literature
review that a measure of family structure stands in contrast
to measuring the occurrence or effects of divorce.
Furthermore, the particular way in which the family structure
measure will be coded here looks at not simply father absence
but at the number of parents living in the household.
Responses to the question, "With whom are you now
living?" (LIV) were used to measure family structure (see the
question and response categories in Appendix A). The
responses were collapsed into three categories: (1) living
with at least two parents, either two biological parents or
one biological parent and one stepparent (original codings 1,
4, 5, 14, and 15; see Appendix A); (2) living with a single¬
parent, either male or female (original codings 2 and 3); and
(3) living with neither parent (original codings 7, 8, 9, and
16). This third category includes living with a boyfriend or
girlfriend, living alone, living with a roommate, or living
with nonparental relatives. Several response categories,
such as "mother and other relative(s)," were deleted in order
to maintain clear distinctions between the recoded
categories. Deleted original codings include 6, 10, 11, 12,
13, and 17. Age is limited to 17 and younger at the time
family structure is measured (usually in year one, but in

79
year two for Model 2) so that the measure would not be
complicated by respondents moving away from home at age 18.
These three categories of living with two parents, with
one parent, and living with no parents are coded and treated
as a three-point ordinal scale of an underlying continuum of
family structure ascending from the most complete, parent-
intact family to the most incomplete, parent-absent family.
Initially this was intended as a four-point scale
differentiating between two-parent family living arrangements
with two biological parents and those with one biological
parent and one stepparent. However, analysis showed that
there were no differences in delinquency between youth living
with two biological parents and those living with a
biological and a stepparent (r=.05, p=.16). Further,
consideration was given to whether the sex of the parent in
single-parent families would make for any difference in
delinquent behavior by the children. Consistent with
previous research (Wells and Rankin, 1985), no difference in
delinquency for those respondents living with only a father
compared to those living with only a mother (r=-.02, p=.77)
was found in the NYS data. Thus, separate categories for sex
of single parent were not included in further analyses.
There were too few cases in these data to investigate the
effects of male or female respondents living with a single
parent of the same sex compared to that of a different sex.
But the data did show that of those living with a single

80
parent, 85 percent of males lived with their mother and 92
percent of females lived with their mother.
In a longitudinal analysis spanning three years, any
change in family structure in the second year may presumably
have a significant impact on delinquent outcomes in the third
year. However, preliminary correlational analyses determined
that in these data there were few changes in family structure
in the time period included. Family structure at year one
was strongly correlated with family structure at year two
(r=.78, p<.01) and at year three (r=.71, pc.01), suggesting
little change in the family structure variable over time.
Moreover, comparisons were made of the effect on
delinquency of whatever changes in family structure did
occur. A variable, change in family structure, was created
based on the change in the number of parents in the household
from year one to year three. Change in family structure was
coded as one=increase in number of parents in the household,
2=no change in number of parents in the household, and
3=decrease in number of parents in the household. This
variable was unrelated to subsequent delinquent behavior
(r=.006, p=.89).
Social Learning Variables
Six measures are employed to represent three major
concepts in social learning theory: differential
reinforcement, definitions, and differential association.
The fourth major component of social learning theory,

81
imitation or modeling, is not included in this study due to
the absence of an appropriate measure in the National Youth
Survey. See Appendix A for a listing of the NYS items used
here as social learning variables.
Differential reinforcement
Recall that differential reinforcement refers to the
balance of rewards and punishment that is received in
response to behavior, and it depends on the values of those
with whom one interacts. The two groups of people with whom
youths most often interact are their parents and their peers.
Thus, one variable will be used to measure the differential
reinforcement of each of these two groups. The measures tap
the reaction that the respondent anticipates from their
parent or peer in response to given delinquent acts, be that
reaction approving (rewarding) or disapproving (punishing).
Variables based on the same measures from the National Youth
Survey, although with varying scale items, were used by
Esbensen and Elliott (1994).
The two reinforcement variables are: the anticipated
parental reactions for one's own delinquency (APR) and the
anticipated friends' reactions for one's own delinquency
(AFR). Both variables are measured on five-point Likert
scales with 1 being strong disapproval and 5 being strong
approval. The APR index is an eight-item scale, measuring
the respondent's anticipated parental reactions to the
following acts: stealing something worth less than $5,
selling hard drugs, using marijuana, stealing something worth

82
more than $50, hitting or threatening to hit someone without
any reason, using alcohol, purposely damaging others'
property, and breaking into a vehicle or building to steal
something (alpha = .85). The AFR index includes nine items,
measuring the respondent's anticipated reactions of friends
to the following acts: stealing something worth less than
$5, selling hard drugs, using marijuana, stealing something
worth more than $50, hitting or threatening to hit someone
without any reason, using alcohol, pressuring or forcing
someone sexually, purposely damaging others' property, and
breaking into a vehicle or building to steal something (alpha
= .90).
Items measuring actual consequences experienced by
youths committing delinquent behavior are not available in
the NYS data set. However, both APR and AFR are adequate
measures of differential reinforcement in that they measure
anticipated rewards or punishments that would follow from
delinquent acts. The more a respondent anticipates, for
example, that breaking into a vehicle or stealing something
will be met with tolerance or even approval by parents and
peers (rewards), then the more likely the respondent will be
to commit those acts. The more the respondent anticipates
parental or peer disapproval (punishment) for those acts, the
more likely the respondent will be to refrain from
delinquency.

83
Definitions
In the last chapter definitions were defined as
evaluative norms, attitudes, or orientations regarding a
particular behavior as good or bad. A definition is positive
or negative, depending on whether behavior is seen as
appropriate or undesirable, respectively. Alternatively, a
neutralizing definition exists when behavior is considered
justified or excusable. The respondent's attitudes toward
delinquent acts (ATT) and the respondent's attitudes toward
interpersonal violence (IPV) measures one's positive/negative
definitions. Similar variables based on the same measures
from the National Youth Survey were used in Heimer (1997),
Heimer and Matsueda (1994), Matsueda and Heimer (1987), and
Elliott et al. (1985). The respondent's endorsement of
deviant neutralizations (EDN) measures neutralizing
definitions. Similar variables based on this measure were
used by Esbensen and Elliott (1994) and Elliott et al.
(1985). (See Appendix A for a complete description of all
three measures of definitions.)
Attitudes toward delinquent acts are measured on a four-
point Likert scale with 1 being "very wrong" and 4 being "not
wrong at all," while IPV is measured on a five-point Likert
scale with 1 being strong disagreement and 5 being strong
agreement. Attitudes toward deviant neutralizations also is
measured on a five-point Likert scale with 1 being strong
disagreement and 5 being strong agreement.

84
The ATT index includes ten items: purposely damaging
others' property, using marijuana, stealing something worth
less than $5, hitting or threatening to hit someone without
any reason, using alcohol, breaking into a vehicle or
building to steal something, selling hard drugs, stealing
something worth more than $50, getting drunk, and giving or
selling alcohol to minors (alpha = .89).
The ipv index includes six items: it is all right to
beat up people if they started the fight, it is all right to
beat up people who call you names, television violence is
effective, it is all right to beat up people if they make you
mad, it is all right to hit someone to get them to do what
you want, and physical force prevents others from walking
over you (alpha = .80).
Items measuring attitudes toward delinquent acts ask
respondents whether particular delinquent acts are right or
wrong. Items measuring attitudes toward interpersonal
violence likewise ask respondents whether they agree with
statements that are favorable to violent behavior. Both of
these items serve as measures of definitions in that they
measure attitudes attached by respondents to particular
behavior. Respondents who hold attitudes positive or
favorable to delinquency will be more likely to commit
delinquent acts while those who hold attitudes negative or
unfavorable to delinquency will be less likely to commit
delinquent acts. Attitudes toward delinquency focuses on
definitions about specific acts, while attitudes toward

85
interpersonal violence focuses on general definitions about
violence and deviance.
The EDN index includes seven items: lying to teachers
is sometimes necessary to avoid trouble, to win at school it
is sometimes necessary to play dirty, lying is okay if it
keeps friends out of trouble, gaining the respect of friends
sometimes means beating up other kids, a willingness to break
rules is necessary to be popular with friends, lying to
parents is sometimes necessary to keep their trust, and it
may be necessary to break some of your parents' rules in
order to keep friends (alpha = .82).
Endorsement of deviant neutralizations is measured by
asking respondents about their agreement with attitudes
justifying deviant behavior. Respondents who are in greater
agreement with justified deviance or delinquency will be more
likely to engage in delinquency themselves. Attitudes toward
deviant neutralizations focuses on general definitions about
violence and deviance.
Differential association
Recall from the last chapter that differential
association refers to interaction with others in primary and
secondary groups as well as in broad reference groups via the
mass media. Initial, enduring, frequent, and close
associations have greater influence on one's behavior than
those which lack these characteristics. The peer group is an
example of a primary group in which youths typically form
associations characterized to some extent by these traits.

86
Those who associate with others who steal or behave
violently, for example, will receive normative definitions
favorable to delinquent behavior and be more likely
themselves to engage in this type of behavior. Associations
also provide the opportunity for other social learning
mechanisms, such as reinforcement, to have an impact.
Thus, a variable of the respondent's association with
delinquent peers will be used to measure differential
association. Research literature consistently shows that
this variable is one of the strongest predictors of juvenile
delinquency. Other studies using a similar variable based on
the same measure from the National Youth Survey include
Heimer (1997), Heimer and Matsueda (1994), Warr (1993),
Matsueda and Heimer (1987), and Elliott et al. (1985).
A five-point Likert scale is used to measure delinquent
peer association (DPA) with 1 being "none of them” engaging
in delinquent acts and 5 being "all of them.” The DPA index
includes ten items: purposely damaging others' property,
using marijuana, stealing something worth less than $5,
hitting or threatening to hit someone without any reason,
using alcohol, breaking into a building or vehicle to steal
something, selling hard drugs, stealing something worth more
than $50, getting drunk, and giving or selling alcohol to
minors (alpha = .88). (See Appendix A for a complete
description of the variable.)

87
Sociodemographic Variables
Six additional variables are considered, including age,
sex, race, class (as measured by income and education), and
family size. Any examination of the relationship between
family structure and delinquency without consideration of
these variables might result in the oversight of significant
subgroup differences. The common sociodemographic variables
of age, sex, race, and class were selected for their widely
known impact on delinquency. Number of siblings is a less
common variable but some studies have suggested that it has
an impact on delinquency (Hirschi, 1969). Correlations
indicating the strength and significance of the relationship
between self-reported delinquency and each of the
sociodemographic variables may be found below in Table 1, on
pp. 93-4. Each sociodemographic variable is discussed in
greater detail below. (See Appendix A for a description of
the original items employed in this section.)
Age
The relationship between age and delinquency is well
known in criminology and can be seen across various types of
data collection. Generally, delinquent behavior increases
near the start of the teen years, peaks at around age 17, and
declines through the 20s as offenders age-out. Both official
statistics and self-report data show that serious criminal
behavior peaks around ages 16 and 17 (Siegel and Senna, 1991;
Elliott, 1994).

88
The results of a cross-sectional analysis of NYS data at
wave three were consistent with previous research, showing
ages 16-18 to be significantly more delinquent than ages 13-
15. The relationship was weaker in a longitudinal analysis
but there remained significant differences in delinquency by
age cohort. When coded at year one as ages 13-15 (1) and
ages 16-17 (2), age was significantly correlated with
delinquency in years two through three (r=.06, p=.049). An
interval-level coding of age yielded a weaker correlation
with delinquency. Therefore, age is coded dichotomously as
indicated and entered into the equations.
Sex
A variety of data indicate that males are more
delinquent than females. Official statistics typically show
that the ratio for serious violent offending by males to that
of females is approximately eight to one, and for property
crime approximately four to one in favor of males (Siegel and
Senna, 1991). Sex differences in self-report data typically
have been smaller than those seen in official statistics,
while remaining significant (Elliott and Ageton, 1980). The
current research confirms this literature. These data show
that males accounted for 76 percent of all delinquents while
comprising 52 percent of the sample. Self-reported
delinquency, in years two through three, and sex were found
not to be independent of each other (r=-.18, pc.01).
Therefore sex is included as a control variable, coded males
(1) and females (2).

89
Race
While official statistics indicate that there are
significant differences by race, some self-report studies
have shown race differences to be virtually nonexistent,
suggesting the presence of bias in the justice system (Siegel
and Senna, 1991). However, using wave one of the National
Youth Survey (NYS), Elliott and Ageton (1980) found
significant race differences for self-reported general
delinquency and for crimes against property, due to blacks
reporting higher frequencies than whites. Later, Elliott
(1994) reported that serious, self-reported violence differed
very little by race. In the present research, blacks
accounted for 16 percent of the sample. Self-reported
delinquency, in years two through three, and race were found
to be significantly correlated (r=.09, pc.01). Race is
retained as a control variable, coded white (1) and black
(2).
Class
Class differences also tend to be more common in
official statistics while self-report studies often report
smaller class differences or none at all, again suggesting
the presence of bias in the justice system (Siegel and Senna,
1991). As class may be measured in a variety of ways,
conclusions about its effects on delinquency may be harder to
draw, compared to conclusions about race differences.
Nevertheless, as they found with race, Elliott and Ageton
(1980) reported that class differences may be more

90
substantial than has been suggested by many self-report
studies, due to lower-class youths reporting higher
frequencies than middle-class youths. Using NYS waves one
through five (1976-1980), Elliott and Huizinga (1983)
likewise found that middle class youth of both sexes are less
likely to be involved in serious offenses, and they commit
fewer offenses when they are involved, compared to lower and
working classes. Both of these studies using NYS data employ
the Hollingshead (1958) index of social class.
The present study employs both family income level and
parents' educational attainment as measures of class. Both
of these measures were obtained from the initial NYS wave in
1976, which included a survey of parents in addition to the
youth survey that would be repeated for subsequent waves.
Thus, these measures were obtained from a time two years
earlier than the other sociodemographic variables, which were
measured at wave three. It is expected that this earlier
measurement will not be problematic because family income and
parental education level are variables that are not likely to
change significantly in that time span. Income was coded as
follows: low ($10,000 and less) (1); medium ($10,001-30,000)
(2); and high (more than $30,000) (3). Twenty-four percent
of the sample reported low family income, 61 percent reported
medium income, and 15 percent reported high income. Income
was found to be significantly correlated with delinquency in
years two through three (r=-.10, p<.01), indicating that the

91
two variables are not independent. Therefore family income
is retained as a control variable.
Parents' educational attainment was coded as follows:
high (approximating college graduate) (1); medium
(approximating high school graduate) (2); and low
(approximating less than high school graduate) (3). The
scale combines measures of both mother's and father's
schooling. Sixteen percent of the respondents' parents
scored high on education, 44 percent scored medium on
education, and 40 percent had low levels of education.
Similar to family income, education measured in 1976 was
found to be significantly related to respondents ' delinquency
in years two (1978) through three (1979) (r=.09, pc.Ol) and
is retained as a control variable.
Number of Siblings
The influence of siblings on delinquent outcomes has
received relatively little attention although the number of
siblings may be an important variable for a variety of
reasons, including decreased parent-child interaction. Where
there are a greater number of children in a household,
parent-child interaction might be expected to decrease as it
does where there are fewer parents. Farnworth (1984)
reported no significant effects for family size and family
density on a wide range of delinquent behavior among black
youths. But in a review of the literature, Loeber and
Stouthamer-Loeber (1986) reported significant effects for
even small differences in family size. They suggest,

92
however, that it is not simply the number of siblings, but
having delinquent siblings that is related to delinquency.
Indeed, Brownfield and Sorenson (1994) found that the
relationship between number of siblings and self-reported
delinquency was fully accounted for by exposure to delinquent
siblings. Consistent with this finding, the current data
showed no significant relationship between number of siblings
and general delinquency in years two through three (r=.04,
p=.09). As for the two measures of class, number of siblings
had to be obtained from the parental survey at wave one
(1976). The variable was coded into categories of one (16
percent of the sample), two or three (53 percent of the
sample), and four or more (31 percent of the sample). Based
on its weak correlation with delinquency, number of siblings
is excluded from further analysis.
Table 1 presents a correlation matrix for the variables
employed throughout this research: delinquency, family
structure, six social learning variables, and six
sociodemographic variables. The correlations are based on
the measurement in Analysis Plan 3 (see below). The first
column shows the correlations between delinquency and each of
the independent variables, indicating significant
correlations with family structure and all six social
learning variables, as expected. Also as expected, the table
indicates significant correlations between family structure
and each of the social learning variables. Among the
sociodemographic variables, all but number of siblings are

Table 1
Delinquency SRD
Parents' Reactions APR
Friends' Reactions AFR
Delinq. Attitudes ATT
Violence Attitudes IPV
Neutralizations EDN
Delinq. Peers DPA
Family Structure LIV
Age
Sex
Race
Income
Education
Siblings
Correlation Matrix
SRD
APR
AFR
1.0
.20
**
.31
O
CN
**
1.0
.49
.31
**
.49
**
1.0
.25
**
.44
**
.69
.26
* *
.33
**
.52
.25
* *
.38
* *
.62
.34
**
.28
**
.67
.13
* *
.20
* *
.14
.06
★
.12
**
.16
.18
* *
-.19
**
-.29
o
VO
* *
.14
**
.01
â– â– 10
**
-.14
**
-.07
.09
**
.13
**
.05
.04
.05
.02
significant at the .05 level
significant at the .01 level
ATT
IPV
EDN
DPA
*
.25 **
.26 **
.25 **
.34
*
.44 **
.33 **
.38 **
.28
.69 **
.52 **
. 62 **
.67
*
1.0
.41 **
. 56 **
.66
*
.41 **
1.0
.57 **
.35
*
. 56 **
.57 **
1.0
.51
*
.66 **
.35
.51 **
1.0
*
.06 *
.11 **
.08 **
.12
*
.27 **
-.03
.07 *
.29
*
-.15 **
-.41 **
-.23 **
-.14
-.12 **
. 15 **
.04
-.08
.07 *
-.14 **
-.03
-.00
-.04
.18 **
.05
.04
-.04
.06
.04
-.02

Table 1—continued
Delinquency SRD
Parents' Reactions APR
Friends' Reactions AFR
Delinq. Attitudes ATT
Violence Attitudes IPV
Neutralizations EDN
Delinq. Peers DPA
Family Structure LIV
Age
Sex
Race
income
Education
Siblings
LIV
AGE
SEX
.13 **
.06 *
-.18
.20 **
.12 **
-.19
.14 **
.16 **
-.29
.06 *
.27 **
-.15
.11 ♦*
-.03
-.41
.08 **
.07 *
-.23
.12 **
.29 **
-.14
1.0
.05
-.06
.05
1.0
-.06
-.06
-.06
1.0
.36 **
-.02
-.04
•.31 **
.08 **
.05
.21 **
.03
-.04
o
o
-.06 *
.01
significant at the .05 level
significant at the .01 level
RACE
INC
EDUC
SIBS
.09 **
-.10 **
. 09 **
.04
.14 **
-.14 **
.13 **
.05
.01
-.07 *
.05
.02
.12 **
.07 *
-.04
-.04
.15 **
-.14 **
.18 **
.06
.04
-.03
.05
.04
.08 *
-.00
.04
-.02
.36 **
-.31 **
.21 **
.00
.02
.08 **
.03
-.06 *
.04
.05
-.04
.01
1.0
-.34 **
.28 **
. 09 **
.34 **
1.0
-.29 **
-.02
.28 **
-.29 **
1.0
.12 **
.09 **
-.02
.12 **
1.0

95
seen to be significantly correlated with delinquency, as
discussed above.
Plan of Analysis
The intent of the analysis is to test four hypotheses
about the relationships among the variables described above,
and thereby determine whether there is support for the
mediating effect of social learning theory in the
relationship between family structure and delinquent
behavior. It has been hypothesized that family structure
affects delinquent behavior (Hypothesis 1); that family
structure is related to various social learning variables
(Hypothesis 2); that social learning variables predict
delinquent behavior (Hypothesis 3); and that social learning
variables mediate the relation of family structure to
delinquent behavior (Hypothesis 4). Figure 1 illustrates the
hypothesized relationships between the three levels of
variables. The hypotheses will be tested using four
longitudinal data analysis plans (see Table 2.)
Family
Structure
Social
-â–º Learning â–º
Variables
Delinquency
Figure 1. Theoretical Model of Social Learning
Variables Mediating the Effects of Family
Structure on Delinquency
The term "analysis plan", or simply "plan", will be used
throughout to refer to the four different ways in which the

96
longitudinal data will be used to test the hypotheses. The
analysis plans specify in which years each of the variables
is measured. Each of the four hypotheses will be tested with
the data specified in each plan, providing for replication
tests of the hypotheses. As indicated below, with each of
these plans the first three hypotheses are tested with zero-
order correlations, while the fourth and main hypothesis is
tested with multivariate regression models. Most of the
statistical analysis therefore is centered on testing
Hypothesis 4. This will entail both OLS and logistic
regression models and testing separate models for male and
female respondents.
Table 2. Analysis Plans for Family Structure, Social
Learning, and Delinquency: Years in which
Dependent, Independent, and Sociodemographic
Variables are Measured
Plan 1
Plan 2
Plan 3
Plan 4
S-R Delinauencv
1979
1980
1979-80
1980
Family Structure
1978
1979
1978
1978
Social Learnino
Diff. Reinf.
Parents 1 Reactions
1978
1979
1978
1979
Friends' Reactions
1978
1979
1978
1979
Definitions
Delinq. Attitudes
1978
1979
1978
1979
Violence Attitudes
1978
1979
1978
1979
Neutralizations
1978
1979
1978
1979
Diff. Assoc.
Delinq. Peers
1978
1979
1978
1979
Sociodemoaraohics
Age
1978
1979
1978
1978
Sex
1978
1979
1978
1978
Race
1978
1979
1978
1978
Education
1976
1976
1976
1976
Income
1976
1976
1976
1976

97
(1) Analysis Plan Is In this plan, family structure,
social learning, and the sociodemographic variables will all
be measured at year one (1978) and the delinquency variables
will be measured at year two (1979), providing for a test of
one-year lagged effects of the variables.
(2) Analysis Plan 2: All of the independent variables
will be measured at year two (1979) and the dependent
variables will be measured at year three (1980). This allows
a replication of the one-year lagged effects but for a
different set of years.
(3) Analysis Plan 3: In this plan measurement of the
dependent variable will be summed across two years instead of
just one year. In this case social learning, family
structure, and the other independent variables will be
measured in the same year (1978) while the dependent variable
is measured across years two and three (1979 and 1980). This
changes measurement of the dependent variable and will
provide a greater frequency of delinquency in the sample.
Given the relatively small amount of self-reported
delinquency in this sample, extending measurement an
additional year will result in a less skewed dependent
variable and thereby provide an additional check on the
reliability of the findings.
(4)Analysis Plan 4: In Plan 4, family structure will
be longitudinally separated from the intervening social
learning variables. Family structure will be measured at

98
year one (1978), the social learning variables will be
measured at year two (1979), and the dependent variable will
be measured at year three (1980). This will provide the most
precise temporally ordered plan in which family structure
precedes the social learning variables, which in turn precede
self-reported delinquency. Such a temporal separation of
variables is important given the theoretical predictions
about the causal relationships between the variables as
stated in the fourth hypothesis. If family structure affects
delinquency to a significant degree through the mediating
effects of social learning variables, measuring the three
concepts separately across three consecutive years should
most closely match the empirical analysis with the
theoretical design.
The first three hypotheses posit bivariate relationships
between the three levels of variables—family structure,
social learning, and delinquency. The hypothesized
relationships between these sets of variables must be
established first because they are presupposed in Hypothesis
4. Using the SAS statistical package, crosstabulations and
zero-order correlations will be employed to test the first
three hypotheses.
Also using SAS, ordinary least squares (OLS) multiple
linear regression will be used to test Hypothesis 4, which
states that the social learning variables will mediate the
effect of family structure on delinquency, controlling for
sociodemographic variables. Multiple regression permits the

99
prediction of values on the dependent variable based on
knowledge of values on independent variables. It also
permits the assessment of the relative degree to which each
independent variable accounts for variance in the dependent
variable. Thus, the procedure will also allow for the
detection of the mediating effects as hypothesized. To the
extent that the standard regression assumption of normal
distribution is violated, logistic regression analyses will
provide a further investigation of these data.
Support for social learning theory will be indicated if
the introduction of social learning variables into the
regression models reduces the main or initial effect of
family structure on delinquency; the more substantial the
reduction the greater the support for the theory.

CHAPTER 5
ANALYSIS AND RESULTS
The distributions of the dependent variable,
delinquency, for each of the four analysis plans are given in
Table 3. Note that the sample size in Plan 2 is
substantially reduced compared to the other three plans.
This results from there being one less cohort of respondents
under age 18 at year two, when age and family structure were
measured in that plan, compared to year one, the time when
those variables were measured for the other three plans. The
number of missing cases in each plan results from listwise
deletions of nonresponses on any variable in any year (Berk,
1983). Plan 3 indicates a greater frequency of delinquency
because delinquent events are summed across two years instead
of one.
Table 3. Percentage Distributions of Self-Reported
Delinquency
(SRD)
Plan 1
Plan 2
Plan 3
Plan 4
Zero SRD
90%
91%
86%
91%
LOW SRD
8
7
11
7
High SRD
2
2
3
2
N
1,031
806
986
1,006
Missing
138
121
183
163
Sample Size
1,169
927
1,169
1,169
100

101
Hypothesis One: Family Structure and Delinquency
Crosstabulations provided a look at the distribution of
delinquency across family structure, indicating that the
variables are related in the hypothesized direction.
Respondents living with two parents were more likely than
those living with a single parent or no parents to be
nondelinquent. Table 4 provides a summary of the
crosstabulations for family structure and delinquency for
each of the four analysis plans, indicating the percentage of
respondents in each type of family structure who were
nondelinquent. While the crosstabulations indicate little
difference in delinquent outcomes between those respondents
living with a single parent and those living with no parents,
they generally suggest the presence of a linear relationship
between family structure and delinquency.
Table 4. Crosstabulation Summary: Percent Nondelinquent
by Family Structure for Each Analysis Plan
Family Structure
Two-parent
Sinale-parent
No-parent
Plan 1
% Nondelinquent
92%
81%
82%
Plan 2
% Nondelinquent
93
86
84
Plan 3
% Nondelinquent
88
77
76
Plan 4
% Nondelinquent
92
87
86

102
Table 5 presents zero-order correlations for self-
reported delinquency and family structure for each of the
four plans. The zero-order correlations indicate moderate
support for the first hypothesis, that delinquency is related
to differences in family structure; delinquency is least
likely to be found in two-parent families and most likely to
be found in living arrangements with no custodial parents
present.
Table 5. Zero-Order Correlation Coefficients for
Self-Reported Delinquency (SRD) and
Family Structure in Four Analysis Plans
Family Structure
Plan
1
SRD
.12
**
Plan
2
SRD
.10
**
Plan
3
SRD
.13
**
Plan
4
SRD
.08
* *
* significant at the .05 level
** significant at the .01 level
Hypothesis Two: Family Structure and Social Learning
Table 6 presents zero-order correlations for family
structure and the social learning variables for each of the
four analysis plans. For each plan, the table indicates weak
to moderate relationships between family structure and the
social learning variables, providing support for the second
hypothesis. That is, a family structure of fewer parents is
associated with higher scores (in the delinquent direction)
on the measures of social learning theory. Anticipated

103
parental reactions (APR), anticipated friends' reactions
(AFR), and delinquent peer association (DPA) were most
strongly correlated with family structure. For example,
those respondents living with fewer parents were more likely
to anticipate less disapproving parental and peer reactions
to delinquent acts. Those respondents living with fewer
parents also were more likely to associate with delinquent
peers.
Table 6. Zero-Order Correlation Coefficients
for Family Structure and Social Learning
Variables in Four Analysis Plans
Plan 1
Plan 2 Plan 3
Family Structure
Plan
4
Social Learning
Diff. Reinf.
APR3
.20 **
.12 **
.20 **
.13
**
AFRb
.14 **
.11 **
.14 **
.11
**
Definitions
ATTC
.06 *
.04
.06 *
.06
*
IPVd
.11 **
.09 *
.11 **
.12
**
EDNe
.08 **
.10 **
.08 **
.08
*
Diff. ASSOC.
DPAf
.12 **
.19 **
.12 **
.13
* *
* significant at the .05 level
** significant at the .01 level
a anticipated parental reactions
b anticipated friends' reactions
c attitudes toward delinquency
d attitudes toward interpersonal violence
e endorsement of deviant neutralizations
f delinquent peer association

104
Hypothesis Three: Social Learning and Delinquency
Table 7. Crosstabulation Summary: Self-Reported Delinquency
(SRD) across Averaged Social Learning Scale Items
for Each Social Learning Variable
Social Learning Variables
APR3
AFRa
ATTb
IPVC
EDNC
DPAd
Zero SRD
98.0%
83.1%
59.0%
69.8%
74.4%
69.5%
Low SRD
94.6
65.8
36.2
50.0
60.9
45.7
High SRD
91.7
54.5
30.5
42.1
48.3
37.2
a anticipated "disapproval" or "strong disapproval" of
delinquency
b defines delinquent act as "very wrong"
c "disagrees" or "strongly disagrees" with delinquency
d number of delinquent peers: "none of them"
Crosstabulations provided a look at the distribution of
delinquency across the social learning variables, indicating
that the variables are related in the hypothesized direction.
The percentages reported in Table 7 were obtained by
averaging for each value of the delinquency variable the
percentage of responses in the nondelinquent direction (as
indicated in the footnotes) on all of the items that make up
each of the social learning variables (see Appendix A). For
example, the first cell of the table indicates that 98
percent of nondelinquents and 92 percent of those scoring
high on delinquency anticipated parental disapproval or
strong disapproval of delinquency. These findings suggest a
linear relationship, with nondelinquent youths being more
likely to score in the nondelinquent direction on social
learning measures, such as having fewer delinquent friends or

105
anticipating parental or peer disapproval of delinquent acts.
These findings are based on the measurement of delinquency
across two years and the measurement of social learning in
the preceding year.
Table 8 presents zero-order correlations for the social
learning variables and self-reported delinquency for each of
the four analysis plans. The table indicates a strong
relationship between the two sets of variables. All six
social learning variables are significantly correlated with
delinquency in the expected direction. Also as expected,
delinquent behavior is most strongly related to differential
peer association in each plan.
Table 8. Zero-Order Correlation Coefficients
for Self-Reported Delinquency and Social
Learning Variables in Four Analysis Plans
Plan 1 Plan 2 Plan 3 Plan 4
Self-Reported Delinquency
Social Learning
Diff. Reinf.
Parents' Reactions
.22 **
.16 **
.20 **
.16
**
Friends' Reactions
.29 **
.29 **
.31 **
.28
*★
Definitions
Delinq. Attitudes
.25 **
.30 **
.25 **
.26
* *
Violence Attitudes
.25 **
.23 **
.26 **
.24
* *
Neutralizations
.26 **
.24 **
.25 **
.23
* *
Diff. Assoc.
Delinq. Peers
.35 **
.36 **
.34 **
.32
**
significant at the .05 level
significant at the .01 level

106
Hypothesis Four; Family Structure, Social Learning, and
Delinquency
Thus far, we have seen support for the first three
bivariate hypotheses in that delinquency is related to both
family structure and the social learning variables, and that
the social learning variables are related to family
structure. This brings us to the fourth and major
hypothesis, which states that social learning variables will
mediate the relationship that has been demonstrated between
family structure and delinquent behavior. The testing of
this hypothesis involves a series of ordinary least squares
(OLS) multiple regression analyses. These analyses are shown
for each of the four analysis plans in Tables 9 through 12.
The tables present standardized estimates of delinquency
regressed first on the social learning variables (a), and
then with added sociodemographic variables (b). Results are
shown in both full and reduced forms. The full models
include all of the independent variables regardless of
significance while the reduced models include family
structure and the sociodemographic and social learning
variables found to be significantly predictive of delinquency
in the full models. Thus, the reduced models provide a
clearer look at the social learning and sociodemographic
variables that significantly affect delinquency.
Recall from Table 5 that the zero-order correlations
between family structure and delinquency were modest but
statistically significant. If the data fit the hypothesis,

107
in each regression analysis the net effect of family
structure on delinquency should be less than the zero-order
effect. When the social learning variables are entered into
the regression models, the effect of family structure on
delinquency should be reduced toward zero and become
nonsignificant.
Table 9. Family Structure, Social Learning, and Delinquency:
One-Year Lagged Analysis Plan (1) with Self-
Reported Delinquency Measured in Year Two (N=l,031)
Full
Models
Reduced
Models
(a)
(b)
(a)
(b)
Familv Structure
.04
.02
.05
.05
Social Learning
Parents' Reactions
.11 **
.09 **
.10 **
.11 **
Friends' Reactions
-.00
-.02
Delinq. Attitudes
-.05
-.02
Violence Attitudes
.12 **
.08
.13 **
Neutralizations
.04
.04
Delinq. Peers
.28 **
.28 **
.27 **
.30 **
Sociodemoaraohics
Age
-.00
Sex
-.07 *
-.10 **
Race
.02
Education
.02
Income
-.05
R2=.15
R2=. 16
R2=.15
R2=. 15
* significant at the .05 level
** significant at the .01 level

108
Table 9 presents the results for Plan 1, in which the
family structure and social learning variables were measured
at year one and delinquency was measured at year two. Recall
from Table 5 that the zero-order correlation for family
structure and delinquency was .12 (p<.01). Note that in each
model the net effect of family structure on delinquency is
reduced considerably and rendered nonsignificant, thus
providing support for Hypothesis 4.
Table 9 also indicates the relative strength of the
independent variables. Among the social learning variables,
anticipated parental reactions (APR), attitudes toward
interpersonal violence (IPV), and delinquent peer association
(DPA) significantly mediated the relationship between family
structure and delinquency, although IPV was rendered
nonsignificant when the sociodemographic variables were
included in the model. Among the sociodemographic variables
only sex was significant, suggesting that family structure
has less effect (net of the social learning effects) on
delinquent behavior among boys than it does among girls.
Table 10 presents the results for Plan 2, in which the
family structure and social learning variables were measured
at year two and delinquency was measured at year three.
Recall from Table 5 that the zero-order correlation for
family structure and delinquency in Plan 2 was .10 (pc.01).
As for Plan 1, in each model the net effect of family
structure on delinquency is reduced considerably and rendered
nonsignificant, again providing support for Hypothesis 4.

109
Among the social learning variables, delinquent peer
association (DPA) and attitudes toward interpersonal violence
(IPV) were significant in reducing the net effect of family
structure.
With the addition of the sociodemographic variables to
the model, delinquent peer association and age had the only
significant effects on delinquency. However, the observed
Table 10. Family Structure, Social Learning, and Delinquency:
One-Year Lagged Analysis Plan (2) with Self-
Reported Delinquency Measured in Year Three (N=806)
Full
Models
Reduced
Models
(a)
(b)
(a)
(b)
Family Structure
.05
.04
.02
.02
Social Learning
Parents' Reactions
-.00
-.01
Friends' Reactions
.01
-.00
Delinq. Attitudes
.08
.10
Violence Attitudes
.10 *
.07
.14 **
Neutralizations
.03
.02
Delinq. Peers
.27 **
.28 **
.31 **
.36 **
SociodemoqraDhics
Age
-.07 *
Sex
-.05
Race
.03
Education
.05
Income
.04
R2= .16
R2= .17
R2=.15
R2=. 13
* significant
** significant
at the .05
at the .01
level
level

110
effect of age is in the opposite direction than might be
expected, indicating when the effects of other variables are
controlled by entering them in the full regression model that
younger adolescents were actually more delinquent than older
teens. This is an unusual finding and suggests that the age
effects in this sample are uncertain enough that they deserve
a closer look. A crosstabulation of age and delinquency
revealed that the levels of delinquent involvement for
younger teens was very close to that of older adolescents.
For whatever reason, the age variable in Plan 2—the only
plan in which age had a significant negative effect—does not
perform reliably. Accordingly, even though its net effect is
significant, the variable was omitted from the corresponding
reduced model (b).
Table 11 presents the results for Plan 3, in which the
family structure and social learning variables were measured
at year one and delinquency was measured cumulatively across
years two and three. Recall from Table 5 that the zero-order
correlation for family structure and delinquency in Plan 3
was .13 (pc.01). Again, the net effect of family structure
on delinquency was reduced in models which included the
social learning and sociodemographic variables. Although
family structure remained significant in the reduced models,
the degree to which the strength of the effect was reduced in
these models and in the full models may be taken as support
for Hypothesis 4. Among the social learning variables,
delinquent peer association (DPA) and attitudes toward

Ill
interpersonal violence (IPV) were significant in reducing the
net effect of family structure on delinquency. As in
Analysis Plan 1, sex was the only significant
sociodemographic variable, suggesting again that family
structure has less effect (net of the social learning
effects) on delinquency among boys than it does among girls.
Table 11. Family Structure, Social Learning, and Delinquency:
Two-Year Lagged Analysis Plan (3) with Self-
Reported Delinquency Measured in
Three (N=986)
Years Two and
Full
Models
Reduced
Models
(a)
(b)
(a)
(b)
Family Structure
.05
.03
.07 *
.07 *
Social Learninq
Parents' Reactions
.05
.04
Friends' Reactions
.07
.06
Delinq. Attitudes
-.04
-.01
Violence Attitudes
.15 **
.10 *
.17 **
.13 **
Neutralizations
-.01
-.01
Delinq. Peers
.25 **
.26 **
.27 **
.27 **
Sociodemoqraohics
Age
-.02
Sex
-.08 *
-.09 **
Race
.04
Education
.05
Income
.01
R2=. 15
R2=. 16
R2= .14
R2=.15
* significant
** significant
at the .05
at the .01
level
level

112
Table 12 presents the results for Plan 4, in which the
measurement of the three sets of variables was temporally
ordered across years one through three with family structure
measured in year one, social learning in year two, and
delinquency in year three. Recall from Table 5 that the
zero-order correlation for family structure and delinquency
Table 12. Family Structure, Social Learning, and Delinquency:
Temporal Sequence Analysis Plan (4) with Self-
Reported Delinquency Measured in Year Three
(N=l,006)
Full
Models
Reduced
Models
(a)
(b)
(a)
(b)
Familv Structure
.02
-.00
.03
.04
Social Learninq
Parents' Reactions
.02
.02
Friends' Reactions
.01
-.01
Delinq. Attitudes
.07
.11 *
.13 **
Violence Attitudes
.09 *
.04
.12 **
Neutralizations
.01
.00
Delinq. Peers
.26 **
.27 **
.31 **
.26 **
Sociodemoaraohics
Age
-.05
Sex
-.06
Race
.06
Education
.04
Income
.02
R2= .14
R2=. 15
R2= .14
R2= .13
significant at the .05 level
significant at the .01 level

113
in Plan 4 was .08 (pc.Ol). In each model the net effect of
family structure on delinquency is reduced considerably and
rendered nonsignificant, again supporting the hypothesis that
social learning variables mediate the effects of family
structure on delinquency. Among the social learning
variables, delinquent peer association (DPA) and attitudes
toward interpersonal violence (IPV) were significant in
reducing the net effect of family structure. With the
addition of the sociodemographic variables to the model,
delinquent peer association and attitudes toward delinquency
(ATT) had the only significant impact in reducing the net
effect of family structure on delinquency.
Generally, analyses of data in all four analysis plans
provided support for Hypothesis 4. That is, social learning,
even when taking into account sociodemographic variables,
reduced the effect of family structure on delinquency. In
every case the net effect of family structure on delinquency
was lower than the zero-order effect; in most cases the net
effect of family structure was rendered nonsignificant and in
some cases disappeared. Delinquent peer association (DPA)
consistently had the strongest net effect on delinquency,
ranging from r=.26 to .28 and in each case highly significant
(p<.01). Attitudes toward interpersonal violence (IPV) also
had a significant net effect on delinquency in every plan,
while anticipated parental reactions (APR) significantly
affected delinquency in Plan 1 and attitudes toward
delinquency (ATT) significantly affected delinquency in Plan

114
4. Among the sociodemographic variables, sex had a
significant net effect on delinquency in Plans 1 and 3. A
further investigation into the impact of sex on the
relationship between family structure and delinquency will be
carried out below. The findings indicate that it is through
these variables that the effect of family structure on
delinquency is mediated.
Social learning variables theoretically reflect the same
underlying process. Therefore, one would expect them to be
intercorrelated. Moreover, prior research has found and the
present research finds that they are intercorrelated.
Therefore, there is some cause for concern regarding the
possibility of collinearity among these variables. The
problem of collinearity arises when two or more predictor
variables are very highly correlated with each other and
thereby distort measures of explained variance and other
coefficients. However, the correlation matrix presented in
Table 1 suggests that while the social learning variables are
strongly related, the magnitude of the correlations suggests
that the data are not beset by the collinearity problem.
Nevertheless, SAS collinearity diagnostics conducted on the
data failed to uncover any serious collinearity problem among
the explanatory variables (see Belsley et al., 1980).

115
Additional Analyses
Structural Effects of Sociodemographic variables
The regression analyses presented above begin with
family structure as a single independent variable, then add
social learning variables to the equation, and finally add
sociodemographic variables as control variables. Ordering
the regression models in this way leaves open the question of
what structural role the sociodemographic variables may play
and whether their effects on delinquency may also be mediated
by the social learning variables. To answer this question a
separate regression analysis was conducted, using data from
Analysis Plan 3 (see Table 19 in Appendix B).
Delinquency was first regressed on family structure
alone in the tests of Hypothesis 4; here delinquency is first
regressed on the sociodemographic variables alone. Findings
show that two of these variables, sex (beta—.17, pc.Ol) and
education (beta=.07, p<.05), exert significant effects on
delinquency. Next, family structure is added to the
equation. While education drops out of the model, the
variable of sex persists with significant effects.
Additionally, family structure exerts a significant,
independent effect on delinquency (beta=.09, pc.Ol). In the
next model, the added social learning variables particularly
mediate the delinquent effects of family structure while also
having a mediating effect on the sociodemographic variables.

116
That is, the effect of family structure is weakened and
rendered nonsignificant, and the effect of sex is
considerably weakened. Among the social learning variables,
attitudes toward interpersonal violence and delinquent peer
association have significant net effects on delinquency.
From these findings it may be concluded that two of the
sociodemographic variables, sex and education, do exert
structural effects on delinquency as does family structure.
Further, as observed above in the case of family structure
(and is seen again in this analysis), the social learning
variables, namely delinquent peer association and attitudes
toward violence, have a mediating effect on the relationship
between the structurally located sociodemographic variables
and delinquent outcomes.
Logistic Regression
The OLS regression models employed above to test
Hypothesis 4 were conducted on a sample including both males
and females. Given the lower level of delinquency among
females relative to males (see Table 14, below), using a
combined sample of females and males decreases the level of
delinquency and produces a greatly skewed distribution on the
dependent variable in the overall sample compared to that of
delinquency among males only. As indicated above in Table 3,
from nine to 14 percent of the respondents in the combined
sample reported any delinquent behavior. Where the dependent
variable is highly skewed, the standard regression assumption

117
of normal distribution is violated. Logistic regression
permits the use of multivariate analysis without the
assumption of a normal distribution by modeling how a
proportion of the dependent variable, rather than the mean,
depends on the independent variables (e.g. Sampson and Laub,
1993).
Additional analyses are conducted for Analysis Plans 1
through 4 using logistic regression and a dichotomous
recoding of the dependent variable. Delinquency is now coded
as follows: no reported property or violent offenses (0); at
least one reported occurrence of either property or violent
delinquency (1). Logistic regression findings consistent
with the preceding OLS regression findings would indicate
that the latter were not altered by the skewness of the
dependent variable.
Table 13 presents standardized estimates for delinquency
regressed on family structure (1), with social learning
variables in the model (2), and with sociodemographic
variables in the equation (3). Model (2) in the table shows
that the effect of family structure on delinquency is reduced
from beta=.17 (pc.01) to .07 (p=.22). As hypothesized the
effect of family structure is mediated by the social learning
variables. The effects are reduced toward zero and become
nonsignificant. As was the case in the OLS analysis,
delinquent peer association (DPA) and attitudes toward
interpersonal violence (IPV) have the strongest effects on
delinquency while anticipated parental reactions (APR) also

118
has a significant effect on delinquency. The addition of
sociodemographic variables further reduces the net effect of
family structure on delinquency (beta=.02, p=.78), with sex
.being the only one of these variables to significantly affect
delinquency (beta=-.22, p<.01).
Table 13. Family Structure, Social Learning, and
Delinquency: Analysis Plan 1 Using Logistic
Regression (N=l,029)
(1) (2) (3)
Familv Structure
.17 **
.07
.02
Social Learnincr
Parents' Reactions
.15 *
.11
Friends’ Reactions
.14
.11
Delinq. Attitudes
-.01
.04
Violence Attitudes
.27 **
.19 *
Neutralizations
.10
.09
Delinq. Peers
.27 **
.31 **
Sociodemograohics
Age
.04
Sex
-.22 **
Race
.11
Education
-.02
Income
-.10
X2=11.5
1 d.f.
X2=131.3
7 d.f.
X2=146.1
12 d.f.
* significant at
** significant at
the .05
the .01
level
level
Logistic analyses also were conducted for Plans 2
through 4, and these produced similar findings. In Plans 1

119
through 3, a significant net effect of family structure on
delinquency was weakened and rendered nonsignificant by the
addition of the social learning variables and the control
variables. In Plan 4, although family structure by itself
did not have a significant net effect on delinquency
(beta=.09, p=.ll), its effect was reduced when social
learning variables were added to the model. Anticipated
friends' reactions (AFR) had a significant effect on
delinquency in Plan 3, while attitudes toward delinquency
(ATT) significantly affected delinquency in Plan 4.
Delinquent peer association and attitudes toward
interpersonal violence were significant social learning
variables in each of the analysis plans. Among
sociodemographic variables, in Plans 2 through 4 sex
significantly affected delinquency. As in the OLS regression
analysis, age had a significant, negative effect on
delinquency in Plan 2. Again, it is concluded that age is
performing unreliably here given the contrary findings in
crosstabulation analysis. The complete tabulated results of
these logistic analyses are presented in Tables 20 through 22
in Appendix B.
The classification analysis procedure in SAS provides an
intuitively appealing way to summarize the results of a
logistic regression model (Hosmer and Lemeshow, 1989). A
crosstabulation is created with the dependent variable,
delinquency, and a dichotomous independent variable whose
values are derived from the predicted logistic probabilities.

120
Thus, the table provides a comparison of predicted and
actually observed responses, and measures the percentage of
all cases that are classified in the correct cells on the
dependent variable by the predicted probabilities.
Classification analyses conducted on the most complete
logistic regression models (3) in each analysis plan indicate
a moderately high degree of correlation between predicted and
observed responses, with the model in each analysis plan
producing an overall correct classification of between 75 to
77 percent. This finding, that the logistic models correctly
predict about three-fourths of all cases offers additional
evidence that they are good models which fit the data well.
In sum, the logistic regression analyses generally
corroborate the findings of the OLS regression models and
thereby provide additional support for the fourth hypothesis,
adjusting for skewed distributions on the dependent variable
by dichotomizing it and conducting logistic regression
produces no differences in findings. That is, social
learning variables are found to mediate the effect of family
structure on delinquency. Specifically, the introduction of
social learning variables into the logistic regression models
reduces the strength and significance of the effects that
family structure exerts on delinquency in the same way as in
the OLS regression models.

121
Controlling for Sex
Recall that gender had a significant effect on
delinquency, net of the effects of the social learning
variables. This suggests that while the social learning
variables mediate the relationship of delinquency to family
structure, they may do so somewhat differently for male and
female adolescents. In light of this finding, in addition to
well-known differences in male and female delinquency
reported in the literature, analyses patterned after Plans 3
and 4 were conducted separately for a subsample of males and
a subsample of females.
These two plans were chosen because they represent
various extremes. Plan 3 provides the greatest variance in
the dependent variable while Plan 4 has the least variance in
the dependent variable. Plan 3 also provides the strongest
correlation between family structure and delinquency while
the weakest correlation between the two variables appears in
Plan 4. Finally, Plan 3 includes the measurement of
delinquency over a two-year period, while Plan 4 uniquely
separates family structure, the social learning variables,
and delinquency in a longitudinal sequence. For these
reasons, controlling for sex in Plans 1 and 2 would not
likely yield considerably different results from those
reported here. Table 14 gives the distributions for
delinquency by sex for Plans 3 and 4.

122
Table 14. Percentage Distributions of Self-Reported
Delinquency (SRD) for Males and Females
Males Females
Plan 3
Plan 4
Plan 3
Plan
Zero SRD
79%
87%
93%
95%
Low SRD
16
10
6
4
High SRD
5
3
1
1
N
516
529
470
477
Missing
110
97
76
69
Sample Size
626
626
546
546
Table 15
presents
standardized
estimates for
male
delinquency, in Plan 3, regressed on family structure (1),
with social learning variables in the model (2), and with
sociodemographic variables in the equation (3). The table
shows that among males, family structure by itself has a
significant net effect on delinquency (beta=.09, pc.05),
which is weakened and rendered nonsignificant by the addition
of social learning and sociodemographic variables to the
model. Delinquent peer association (DPA) is the only one of
these independent variables to have a significant net effect
on delinquency (beta=.29, pc.01). In a reduced model
including only family structure and delinquent peer
association (not shown), the former was slightly stronger but
still nonsignificant (beta=.05, p=.23) while the latter was
slightly stronger (beta=.31, pc.01).

123
Table 15. Family Structure, Social Learning, and
Delinquency: Analysis Plan 3 for Male
Subsample
(N=516)
(1)
(2)
(3)
Familv Structure
.09 *
.02
-.00
Social Learninq
Parents' Reactions
.07
.05
Friends' Reactions
.05
.06
Delinq. Attitudes
-.08
-.06
Violence Attitudes
.06
.05
Neutralizations
.02
.01
Delinq. Peers
.29 **
.29 **
Sociodemoaranhics
Age
.00
Race
.08
Education
.06
Income
.05
R2=. 01
R2=.12
R2= .13
* significant
at the .05
level
** significant at the .01 level
Table 16 presents standardized estimates for female
delinquency, in Plan 3, regressed on family structure (1),
with social learning variables in the model (2), and with
sociodemographic variables in the equation (3). The net
effect on delinquency of family structure by itself is
stronger for females (beta=.19, p<.01) than for males.
However, a comparison of Tables 15 and 16 indicates that the
mediating impact of the social learning variables is weaker
for females than for males. For females, the addition of

124
social learning variables to the model weakens the
relationship between family structure and delinquency but it
remains significant. Among the social learning variables in
Table 16, attitudes toward interpersonal violence (IPV) and
delinquent peer association exert significant effects on
delinquency and thereby mediate the effect of family
structure on delinquency. No sociodemographic variables are
Table 16. Family Structure, Social Learning, and
Delinquency: Analysis Plan 3 for Female
Subsample (N=470)
(1)
(2)
(3)
Familv Structure
.19 **
.11 *
.10 *
Social Learninq
Parents' Reactions
.01
.01
Friends’ Reactions
.06
.05
Delinq. Attitudes
.07
.11
Violence Attitudes
.20 **
.16 **
Neutralizations
-.07
1
o
00
Delinq. Peers
.21 **
.22 **
Sociodemoqraohics
Age
-.07
Race
-.03
Education
.02
Income
-.09
R2=.04
R2=. 18
R2= .19
* significant at
** significant at
the .05
the .01
level
level

125
significantly related to delinquency. In reduced versions of
these analyses (not shown) the only notable difference is
that delinquent peer association has a stronger net effect on
delinquency.
Table 17. Family Structure, Social Learning, and
Delinquency: Analysis
Subsample (N=529)
Plan 4
for Male
(1)
(2)
(3)
Familv Structure
.06
.01
-.01
Social Learninq
Parents' Reactions
.05
.04
Friends’ Reactions
.01
-.02
Delinq. Attitudes
.08
.11
Violence Attitudes
.01
-.00
Neutralizations
-.02
-.03
Delinq. Peers
.26 **
.28 **
Sociodemoqraohics
Age
-.05
Race
.10 *
Education
.03
Income
.06
R2=. 00
R2=.ll
R2= .12
* significant
at the .05 level
** significant at the .01 level
Table 17 presents standardized estimates for male
delinquency, in Plan 4, regressed on family structure (1),
with social learning variables in the model (2), and with
sociodemographic variables in the equation (3). The table

126
shows that the net effect of family structure on delinquency
is beta=.06 (p=.19). Thus, male delinquency in Plan 4 is
less affected by family structure than in Plan 3, and the
relationship between the two variables is weaker than that
for females in either plan. The addition of social learning
and sociodemographic variables to the model reduces the net
effect of family structure. Delinquent peer association
(DP A) is the only social learning variable to significantly
affect delinquency (beta=.26, p<-01). Race is the only
sociodemographic variable to have a significant effect on
delinquency (beta=.10, p<.05), with blacks being slightly
more delinquent than whites. In a reduced model (not shown)
that includes family structure, delinquent peer association,
and race, the effect of family structure disappears (beta=-
.00, p=1.0) while DPA has a stronger effect on delinquency
(beta=.33, p<.01). Race becomes nonsignificant (beta=.06,
p=.21).
Family structure by itself has a significant net effect
(beta=.12, p<.01) on female delinquency in Plan 4. Table 18
presents standardized estimates for female delinquency in
Plan 4 regressed on family structure (1), with social
learning variables in the model (2), and with
sociodemographic variables in the equation (3). The weaker
findings in Plan 4 do not indicate the mediating differences
between males and females that were observed in Plan 3. The
findings do show the mediating impact of social learning
variables, as delinquent peer association (DPA) and attitudes

127
toward interpersonal violence (IPV) have moderate effects on
delinquency: beta=.26 (p<.01) and beta=.13 (pc.05),
respectively, where the sociodemographic variables are
included. The sociodemographic variables are not
significant, a reduced model (not shown) resulted in beta
coefficients of .04 (p=.37) for family structure, .19 (pc.01)
for attitudes toward interpersonal violence, and .32 (pc.01)
for delinquent peer association.
Table 18. Family Structure, Social Learning, and
Delinquency: Analysis Plan 4 for Female
Subsample (N=477)
(1)
(2)
(3)
Family Structure
.12 **
.03
.02
Social Learnina
Parents' Reactions
-.03
o
1
Friends’ Reactions
.00
.00
Delinq. Attitudes
.10
.12
Violence Attitudes
.16 **
.13 *
Neutralizations
.06
.05
Delinq. Peers
.26 **
.26 **
Sociodemoqranhics
Age
-.06
Race
o
1
Education
.05
Income
-.06
R2=. 01
R2=. 19
R2= .20
* significant at
** significant at
the .05
the .01
level
level

128
As indicated above in Table 14, from 13 to 21 percent of
males engaged in delinquent behavior while only five to seven
percent of females were delinquent. Again, due to the
skewness of these distributions, especially that of females,
these analyses were repeated using logistic regression and a
dichotomous coding of the dependent variable. Again,
classification analyses were conducted on the most complete
logistic regression model (3), separately for males and
females. The overall rate of correct classification for
males ranged from 70 to 73 percent, while the rate for
females ranged from 80 to 85 percent. These findings again
show that the variables in the logistic regression models do
a good job of predicting delinquent behavior for both sexes,
but somewhat better for the girls than the boys in the study.
The results of the logistic regression analyses
generally follow the patterns of the findings discussed above
where sex is controlled, including stronger family structure
effects on delinquency among females. As was seen in Table
16, however, family structure retains a significant net
effect on female delinquency after the addition of the social
learning variables to the model in Plan 3 (see Table 24 in
Appendix B). For males (see Tables 23 and 25 in Appendix B)
the effects of family structure on delinquency are
nonsignificant both before and after the social learning and
sociodemographic variables are added to the model, and in
Plan 4 there is no bivariate relationship between family
structure and delinquency. There are a few other differences

129
regarding the significance of specific variables. For
example, attitudes toward interpersonal violence (IPV) and
race have significant effects on male delinquency in Plan 3.
For females, anticipated friends' reactions toward delinquent
behavior (AFR) has a significant effect on delinquency in
Plan 3.
In sum, controlling for sex confirms that the
relationship between family structure and delinquency is
stronger for girls than for boys but that the social learning
variables have a mediating effect on family structure for
both. The findings also reveal, however, that the
intervening effects of the social learning variables on the
relationship between family structure and delinquency is
weaker for girls than for boys. The fourth hypothesis is
generally supported but it is confirmed more for the male
than the female subsample.

CHAPTER 6
CONCLUSIONS
Summary
Three central issues were addressed in this study. One
was the explication of the relationship between family
structure and delinquent behavior. The second was the
relationship of delinquent behavior to the social learning
variables. The third and most significant issue was the
examination of a specific instance of Akers' (1998) Social
Structure-Social Learning model, namely a family structure
and social learning model.
In Chapter 2 a variety of literature on family variables
and delinquency was reviewed. Research has shown that the
likelihood of delinquent behavior may be increased as a
result of parental criminality, parents' patterns of
discipline and punishment, and family violence, including sex
abuse and other types of abuse. Research also showed that
delinquency may result from more routine and seemingly
innocuous family relationships. But perhaps the most
dominant focus relating to the delinquent effects of family
has been in the area of divorce and single parenthood. Given
the increase in divorce and other changes in family structure
during the last thirty years, a topic of considerable
130

131
interest in the research literature has been the potentially
deleterious effects of these changes on children. While some
studies have focused on children of divorce, this research
focused on the more general concept of family structure.
In Chapter 3 social learning theory was introduced to
provide a potential explanation for whatever impact family
structure has on the delinquent behavior of children. The
central role typically played by parents as primary
socializers of children makes social learning a useful
perspective for explaining the effects of family structure.
Specifically, the usual centrality of family life provides a
setting for differential association in which children are
exposed to parental patterns of norms, values, and behavior.
Parents also provide differential reinforcement in the form
of rewards and punishment for children's behavior. Through
such parental reinforcement, as well as imitation, children
learn norms and attitudes and formulate beliefs about
conforming or deviant behavior. Social learning theory
recognizes these as positive, negative, or neutralizing
definitions. Youths with more delinquent associations and
models, those who are exposed to reinforcement toward
delinquency, and those who learn delinquent definitions will
more likely engage in delinquent behavior.
The theory also accounts for the impact of peers, a
group consistently found to have a strong influence on
behavior. Family discipline, supervision of children, and
the attitudes and beliefs learned by the children within the

132
family affect the choice and influence of peers with whom the
children differentially associate. These associations
provide the basis for the learning of other norms and values,
and they provide other settings in which differential
reinforcement may occur. Behavior is rewarded and punished
by one's peers as it is by one’s parents. Through this
reinforcement, as well as from imitation, youths learn
additional definitions about conforming or deviant behavior.
Again, the more delinquent one's social learning among peers
the greater the likelihood of engaging in delinquency.
Social learning theory was purported to provide the link
between social-structural elements such as family structure,
and individual-level outcomes such as delinquent behavior.
If families are central in social learning as described
above, then it is reasonable to suppose that variations in
the structure of families would result in variations in
conforming or delinquent individual behavior. Hypotheses
were therefore posited linking family structure, social
learning, and serious delinquency.
Longitudinal data from the National Youth Survey, a
nationally representative data set focusing on self-reported
delinquent behavior, were analyzed. First, crosstabulations
provided evidence of relationships between independent
variables of family structure and social learning, and the
dependent variable of delinquency. Second, zero-order
correlations were used to test Hypotheses 1-3 and ordinary
least squares (OLS) multiple linear regression models were

133
used to test the fourth hypothesis. Given the significance
of sex in two of the analysis plans, indicating differences
between males and females in the way that the family
structure-delinquency relationship is mediated, regression
analyses were repeated while controlling for sex. Logistic
regression models were run to take into account the skewed
distributions of serious delinquency. The findings using
logistic regression essentially confirm those from the OLS
regression analyses. Generally, all four hypotheses were
supported by the findings.
The first hypothesis, that family structure predicts
delinquency, is supported by findings indicating that those
respondents with only one parent or no parents at all in
their household were more likely to engage in delinquent
behavior. Correlational analyses indicated that overall
relationships were statistically significant but varied from
weak to moderate across the four analysis plans.
The second hypothesis, that family structure affects the
social learning process conducive to delinquency, is
supported in all four analysis plans. Correlational findings
indicated that those respondents with one or no parent in
their household were more likely to score in a delinquent-
prone direction on social learning variables than those in
two-parent families. Five of the six social learning
variables, excepting attitudes toward delinquency (ATT), were
significantly related to family structure across all four
analysis plans. Family structure was most strongly and

134
significantly correlated with anticipated parental reactions
to delinquent acts (APR), anticipated friends' reactions to
delinquent acts (AFR), and delinquent peer association (DPA).
The third hypothesis, that the social learning variables
predict delinquency, is strongly supported in all four
analysis plans. Crosstabulations indicated that respondents
scoring in the delinquent direction for each of the six
social learning variables also scored higher on self-reported
delinquency. Correlational findings indicated that those
respondents who scored higher on various pro-delinquency
social learning measures were more likely to engage in
delinquent behavior. All six social learning variables were
strongly correlated with delinquency at a high level of
significance across all four analysis plans.
The results from the first three hypotheses established
the relationships between family structure, the social
learning variables, and delinquency, thereby setting up the
fourth and major hypothesis. Hypothesis 4, which states that
the introduction of the social learning variables in the
relationship between family structure and delinquency will
reduce the strength and significance of family structure' s
effect on delinquency, also is supported. In each of four
OLS regression models, both the strength and the significance
of the effects of family structure on subsequent delinquency
were substantially reduced from the zero-order correlations
and from models without the social learning variables when
the social learning variables were added to the model. These

135
results were confirmed with the addition of sociodemographic
variables to the models. Using logistic regression, the
strength of family structure was reduced in each of the
models, but in Plan 4 family structure by itself was not
significantly related to delinquency so there was little
effect to mediate.
Sex was identified, in Plans 1 and 3, as the only
sociodemographic variable to have a significant effect on
delinquent behavior. Thus, separate analyses were conducted
for boys and girls. The results generally confirmed earlier
findings, while revealing that the weak relationship between
family structure and delinquency in Plan 4, as discussed
above in relation to the first hypothesis, held only for
males. The social learning variables reduced the net effect
of family structure on delinquency for both sexes, although
family structure remained significant for females in Plan 3.
As mentioned above, delinquent peer association and
attitudes toward interpersonal violence were the social
learning variables most frequently found to play a
significant mediating role in the relationship between family
structure and delinquent behavior. Anticipated parental
reactions was significant in Plan 1 for both the OLS and
logistic regressions, and anticipated friends' reactions was
significant in Plan 3 for the logistic models, both before
controlling for sex and for females alone. Attitudes toward
delinquent acts was significant in Plan 4 for both the OLS
and logistic regressions. Among other sociodemographic

136
variables, education and income were not significant in any
case, and race significantly affected male delinquency only
in one OLS model and one logistic model. Age was significant
in Plan 2 but it was determined that this finding was
unreliable given contradictory findings in crosstabulation
analysis and the absence of significance in other analysis
plans.
Conclusions
The conclusions that may be drawn from this research are
both practical and theoretical. The findings showed that
children living in homes without two parents are more likely
to engage in delinquency than children living with two
parents. This is the case whether the two parents are both
biological or one of the two is a stepparent. Beyond this
distinction, it does not appear to affect delinquency whether
a child lives with a single parent or with nonparental
adults. These findings provide mixed support for the idea
that family structure affects delinquency as the relationship
is not found across all values of family structure. However,
the observed difference between two parents of whatever type
and a single parent is consistent with the consensus of
previous research suggesting that family structure and
divorce are two distinct concepts that measure different
phenomena. In this case, two-parent families, regardless of
the occurrence of divorce, are less conducive to delinquency
than are single-parent families.

137
Controlling for sex revealed that the relationship
between family structure and delinquency is stronger for
females than for males, contradicting much previous research
(e.g. Wells and Rankin, 1991; Gove and Crutchfield, 1982).
This discrepancy may result from the combination in the
present research of two-parent stepfamilies and families with
two biological parents in a single category of family
structure. Johnson (1986) concluded that the stronger
relationship between family structure and delinquency among
males resulted largely from the relationship between males
and their stepfathers. The coding of family structure may
have obscured a similar pattern in the present research.
Nevertheless, the current findings indicate a stronger
relationship between family structure and delinquency for
females than for males. In Analysis Plan 3 family structure
retained a significant net effect on delinquency even after
the social learning variables were added to the equation.
But given the general finding of a reduction in the net
effect of family structure on male and female delinquency, it
appears that this may result simply from the stronger
relationship between family structure and female delinquency,
particularly in Plan 3.
The findings suggest that females may be more dependent
upon their parents for learning attitudes consistent with
conformity than are their male counterparts. While
delinquent peer association had a significant net effect on
delinquency for both sexes, attitudes toward interpersonal

138
violence had significant net effects on female delinquency.
Additionally, logistic regression indicated that anticipated
friends' reactions toward delinquent behavior had a
significant net effect on delinquency only for females.
Thus, it may be through a greater variety of social learning
variables that family structure plays a role in determining
female delinquency, compared to that which occurs among
males. This would be consistent with previous research (e.g.
Farnworth, 1984; Van Voorhis et al., 1988) which found the
effects of other mediating variables between family structure
and delinquency to be greater for females than for males.
The findings relating to the fourth hypothesis may be
taken as significant support for social learning theory. The
results indicate that social learning is the process that
explains the relationship between family structure and
delinquency. This finding provides empirical support for the
theoretical argument posited by Akers (1998) in his Social
Structure-Social Learning (SSSL) model, that social learning
theory is capable of linking macro and micro levels by
explaining how the social structure shapes individual
behavior. Specifically, the social-structural element of
family structure explains the individual-level outcome of
delinquent behavior via the behavioral process of social
learning. Presumably, social learning theory holds the
potential for explaining how other elements of social
structure, such as society or community, affect delinquency
and other types of deviant behavior.

139
Differential association, definitions, and differential
reinforcement each had significant mediating effects on the
relationship between family structure and delinquency. For
example, the significance of anticipated parental reactions
indicates that the social learning process of differential
reinforcement is mediating the delinquent effects of family
structure. The significance of attitudes toward
interpersonal violence indicates that adherence to delinquent
definitions toward violence mediates the relationship between
family structure and delinquency.
But it is most notably through the social learning
process of differential association, as measured here by
delinquent peer association, that social learning theory
explains the relationship between family structure and
delinquency. For example, a youth living in a single-parent
household will be more likely than a youth living in a two-
parent home to associate with other youths who have engaged
in delinquent behavior. This association may then result in
one's own delinquency. Social learning theory also
recognizes reciprocal effects whereby a reversed sequence may
occur. That is, one's delinquency may lead a youth to
befriend other delinquents, and increasing delinquent
patterns may lead to the disintegration of the youth's
family. However, Akers (1998) suggests that the former
sequence, where delinquent associations precede delinquency,
is the more typical one.

140
The interdependence between family and peer variables
has been noted by Jensen and Rojek. The stronger the
relationship between a youth and parents, they said, "the
weaker the chances that he or she will acquire delinquent
friends or choose to go along with peers in situations of
conflict with authority" (1992, p. 306). While the influence
of delinquent peers often has been found to be a strong
predictor of delinquency, and the association with delinquent
peers is believed to be related to family relationships and
family structure, the causal ordering of family and peer
variables is less clear. Using longitudinal data, the
current study supports the contention that delinquent peer
association is a strong predictor of delinquency, while
suggesting that delinquent peer association is influenced by
family structure.
Implications for Policy
This research finds that single-parent and no-parent
family structures have an indirect impact on serious self-
reported delinquency. This finding suggests, however, that
an increased divorce rate will result in more delinquency
unless the remarriage rate is also high. Similar levels of
delinquency were found among children living with two
biological parents and those living with one biological
parent and one stepparent. Thus, the evidence suggests that
divorce will have an effect on delinquency if it induces one-
parent families. It appears from these findings that if the

141
two-parent family is re-established for the children, its
delinquency-inhibiting effects will also be re-established.
It follows that family intervention and social policy
designed to encourage or provide incentives for the stability
of two-parent households (e.g. prevent divorce in the first
place and encourage remarriage) will contribute to a decrease
in delinquency. Conversely, economic and social policies
which set conditions promoting the formation of single-parent
households may also have the effect of contributing to an
increase in delinquency.
While the results indicate that a two-parent family
structure will more likely provide the conditions that are
less conducive to delinquency, the significance of the
intervening social learning variables reveals that
associating with delinquent peers and learning deviant
attitudes are the variables directly implicated in
delinquency causation. Thus, the findings suggest that
social learning variables may be affected by conditions other
than family-related variables, and that policies which take
into account the role of social learning variables—
especially differential peer association and definitions—in
the formation of delinquency will contribute to the reduction
of delinquency for all youths, regardless of family
structure.
Compared to structural family variables, process
variables such as differential peer association and
definitions that occur within the usually intimate

142
interaction of peer groups and families may require a more
complicated response in terms of social policy.
Nevertheless, programs such as organized sports, peer
counseling, pro-social peer activities, and social skill
development take advantage of pro-conformity peer
associations while at the same time providing the setting for
peer interaction that may counter delinquent influences.
Informational campaigns should aim toward increasing
awareness of the ways by which delinquent behavior is
socially learned. This may better provide parents with an
appreciation for the importance of remaining aware of and
shaping their children's attitudes, providing reinforcement
and punishment in favor of nondelinquency and against
delinquency, and closely monitoring their children's peer
interaction. Parents know intuitively that who their
children's friends are is very important. Social learning
provides theoretical support for this knowledge. Similarly,
given that social learning variables directly affect
delinquency regardless of family structure, teachers and
other adults who occupy prominent places in the lives of
adolescents should also be aware of the ways by which
delinquent behavior is socially learned.
As stated in the introduction, single-parent families
have been an increasing reality in recent decades.
Furthermore, there is no indication that the trend will
reverse itself. In Chapter 2, the point was also made that
single-parent families may be variously disintegrated. In

143
other words, some single-parent families may prove equal or
superior to two-parent families in pro-social socialization.
This is likely to be the case when the choice is a competent
single parent or a two-parent family in which there is
considerable conflict and alienation of children in the
family. Given these considerations, it is appropriate that
care be taken in the implementation of policy so that single¬
parent families are not marginalized. Where possible,
policies should concentrate on providing community resources
for day-care and other support for single parents along with
opportunities to enhance parenting skills.
This research has found that social learning variables
play an important role in the process of becoming delinquent
or conforming, and that single-parent families are more at
risk of delinquency. Additionally, it is consistent with
social learning theory to suggest that efforts made to
increase the level of pro-conformity social learning among
children from single-parent families may contribute to the
prevention of delinquent outcomes. Where parent-child
interaction is compromised, especially in the case of working
single parents, mentoring programs, big-brother programs,
after-school programs, and similar efforts may successfully
serve a surrogate function. Programs aimed at making parents
(both in two-parent and single-parent homes) aware of the
level of risk of delinquency that their children may face,
and educating them about the ways in which that risk can be

144
minimized, can be a valuable step toward preventing
delinquency.
Limitations and Suggestions for Future Research
A number of limitations should be noted regarding the
present research. First, the waves of data employed in this
study are approximately twenty years old, years during which
trends in juvenile delinquency may have changed. Osgood et
al. (1989), for example, found that self-reported assaults
increased from 1975 to 1985. In the 1990s juvenile violence
has been an increasing source of concern (Elliott, 1994).
The rise in divorce and nontraditional family structures may
have resulted in more recent years in a decrease in their
deleterious effects (Jensen and Rojek, 1992). Whatever
changes in juvenile delinquency and norms of family structure
may have occurred in the last two decades, the National Youth
Survey's (NYS) nationally representative sample and
longitudinal data collection make it a valuable data set
despite the passage of time. Moreover, while delinquency
trends may change, the social learning mechanisms that are at
the center of this research are likely more constant.
Nevertheless, analysis of more recent longitudinal data would
address this limitation.
The longitudinal analysis employed here enabled the
analysis of the effects of family structure and social
learning variables across time. The longitudinal measurement
of variables is one way to test the theoretical ordering of

145
variables. However, as noted by Van Voorhis et al. (1988), a
social learning perspective suggests that the effects of
changes in family structure may be cumulative over a number
of years. The one-year lags employed here, although
preferable to cross-sectional analysis, are relatively short
periods of time for detecting the effects of family structure
and social learning. A more extended longitudinal analysis
would provide a stronger test of the effects of family
structure and of social learning theory.
Another limitation of the present research is the
unavailability of an appropriate item measuring the social
learning concept of imitation, while the inclusion of all
four central concepts of social learning theory would be
desirable in a study focusing on the theory, the six items
that were used provided good measures of differential
association and definitions. The measures of differential
reinforcement are adequate, although they are limited by
their measurement of the anticipated reactions of parents and
peers to hypothetical delinquency, rather than the actual
rewards and punishment by parents and peers to actual
delinquent behavior carried out by the respondents.
Moreover, empirical findings have shown that imitation is the
weakest of the four social learning concepts. But a more
complete test of social learning variables as mediators of
the effect of family structure on delinquency would result
from an analysis including an imitation measure, as well as
better measures of differential reinforcement.

146
This study was hindered by a relatively low level of
delinquency in the sample and the resulting statistical
limitations. One cause of the low level of delinquent
behavior in this sample is the relatively small amount of
serious delinquency that occurs compared to other types of
delinquency. As seen in Table 3, there was considerable
skewness in the dependent variable. This shortcoming was
addressed through the use of logistic regression, and
indirectly by controlling for sex (see Table 14).
Future research also might include an investigation of
the reciprocal effects between variables. Most researchers
agree that the temporal ordering presented here is the more
likely and reasonable arrangement. However, some research
(see Akers and Lee, 1996) has found that, in addition to the
effects of social learning variables on delinquent behavior,
delinquency has an impact on social learning variables.
Although less likely, there may also be some impact by
delinquency and social learning variables on family
structure. According to Akers (1998), the two sequences are
not mutually exclusive but simply the social learning process
operating at different times.
Primary research would be one way to address the
limitations of the social learning measures used in this
research. Variables would be created for each of the four
social learning concepts. The concepts of differential
association and definitions could be measured in a way
similar to that found in the National Youth Survey. Measures

147
of differential reinforcement by parents and peers would be
constructed to measure actual reinforcement of delinquent
acts rather than anticipated reinforcement as found in the
NYS. Lastly, a measure of imitation would be included to
test a more complete social learning model.
The relationships between family structure, social
learning, and delinquency might be investigated in a sample
of high school students. In addition to seeking to replicate
the general findings of the present research particular
attention would be given to the findings relating to sex,
given the finding in the present research that the effect of
family structure on delinquency is stronger for females.
Indeed, sex may be investigated as an exogenous variable
indicating location in the social structure that, like family
structure, has an impact on delinquency through the process
of social learning. Empirically, the findings presented in
this research suggest that differences in delinquency between
boys and girls may be accounted for by social learning
variables. Theoretically, in his Social Structure-Social
Learning (SSSL) model, Akers (1998) lists sex as one among
several variables that may affect delinquency through a
mediating process of social learning. Other structural
variables that may serve as exogenous variables in future
research on the SSSL model include broad measures of society
and community, sociodemographic variables such as age, race,
and class, and more immediate measures of one's primary
groups such as family and peers.

148
Another direction for future inquiry may be to address
the limitations of the present research in another study
using NYS data. The focus on serious delinquency in the
present research was driven by an interest in more serious
offenses as well as the relative paucity of research on
serious delinquency. However, the NYS offers a great variety
of delinquency measures, ranging from small acts of deviance
such as cheating at school to alcohol and drug abuse to the
serious offenses used in this research. Including less
serious offenses in the measurement of the dependent variable
would increase the frequency of delinquency. This would
permit more flexibility in the coding of the dependent
variable and in statistical analysis, particularly in the use
of ordinary least squares regression.
Such a measure of delinquency would be no less adequate
for testing the mediating effects of social learning
variables. A measure of delinquency broadened to include NYS
measures of substance abuse would also allow a more complete
test of social learning theory's central concepts. There is
an item in the survey measuring exposure to substance abuse
by parents, representing the imitation, or modeling, concept
of social learning theory that was missing from the current
research.
Employing a broader range of longitudinal data would
serve both to include more recent data on juvenile
delinquency (and at later ages) and to permit analysis of the
effects of change in family structure across more extended

149
lags than the one- and two-year lags used here. In addition
to the first five waves of NYS data collected between 1976
and 1980, data were collected for each succeeding year at
three-year intervals into the 1990s. These waves of data are
becoming available through the Inter-University Consortium
for Political and Social Research (ICPSR).
Finally, a longitudinal analysis could also include an
investigation of reciprocal relationships between the social
learning variables and delinquency, and perhaps family
structure. This would be achieved simply by including in an
analysis measurement of each of the variables from each wave
of data. Akers has defended social learning theory from the
suggestion that "peer associations take place only or largely
after adolescents have already separately and individually
established a pattern of deviant behavior and then choose
delinquent peers simply because they have the same behavior
in common" (Akers, 1998, p. 120). Rather, Akers maintains
that evidence indicates that youths, based on prior family
and other variables, differentially associate with delinquent
peers or those who are tolerant of delinquency, learn or
strengthen delinquent definitions, and are exposed to models
that reinforce delinquency, and then initiate or increase
their own delinquent behavior. This behavior goes on to
influence further associations and definitions. An analysis
of NYS data could contribute to the clarification of the
relationships between these levels of variables.

APPENDIX A
NATIONAL YOUTH SURVEY ITEMS OPERATIONALIZED AS MEASURES OF
VARIABLES USED IN THIS STUDY
Self-Reported Delinquency (SRD)
Property Offenses
1. Motor vehicle theft: "How many times in the last
year have you stolen or tried to steal a motor
vehicle, such as a car or motorcycle?"
2. Theft of over $50: "How many times in the last year
have you stolen or tried to steal something worth
more than $50?"
3. Breaking into building: "How many times in the last
year have you broken or tried to break into a
building or vehicle to steal something or just to
look around?"
Violent Offenses
1. Personal attack: "How many times in the last year
have you attacked someone with the idea of seriously
hurting or killing him or her?”
2. Sexual assault: "How many times in the last year
have you had or tried to have sexual relations with
someone against their will?"
3. Use of force on another: "How many times in the
last year have you used force or strong-arm methods
to get money or things from other students?"
Family Structure
Living arrangement: "With whom are you now living?
1. mother and father
150

151
2. mother only
3. father only
4. mother and stepfather
5. father and stepmother
6. spouse
7. other (specify)
8. boyfriend/girlfriend (opposite sex)
9. alone
10. mother and other relative(s)
11. mother and other non-related adult(s)
12. father and other relative(s)
13. father and other non-related adult(s)
14. mother and father plus relative(s)
15. mother and father plus other non-related adult(s)
16. relative(s) (not parents)
17. other
Two-parent: 1, 4, 5, 14, 15
Single-Parent: 2, 3
No-Parent: 7, 8, 9, 16
Omitted: 6, 10, 11, 12, 13, 17
Social Learning Variables
Differential Reinforcement
Differential reinforcement of parents (APR): "How would
your parents react if you ...?"
l=strongly disapprove
2=disapprove
3=neither approve nor disapprove
4=approve
5=strongly approve
1. stole something worth less than $5
2. sold hard drugs such as heroin, cocaine and LSD
3. used marijuana or hashish
4. stole something worth more than $50
5. hit or threatened to hit someone without any reason
6. used alcohol
7. purposely damaged or destroyed property that did not
belong to you
8. broke into a vehicle or building to steal something
Differential reinforcement of friends (AFR): "How would
your friends react if you ...?"
l=strongly disapprove
2=disapprove

152
3=neither approve nor disapprove
4=approve
5=strongly approve
1. stole something worth less than $5
2. sold hard drugs such as heroin, cocaine and LSD
3. used marijuana or hashish
4. stole something worth more than $50
5. hit or threatened to hit someone without any reason
6. used alcohol
7. pressured or forced someone to do more sexually than
they wanted to do
8. purposely damaged or destroyed property that did not
belong to you
9. broke into a vehicle or building to steal something
Definitions
Attitudes toward delinquent acts (ATT)s "How wrong is
it for someone your age to ... ?"
l=very wrong
2=wrong
3=a little bit wrong
4=not wrong at all
1. purposely damage or destroy property that does not
belong to him or her
2. use marijuana or hashish
3. steal something worth less than $5
4. hit or threaten to hit someone without any reason
5. use alcohol
6. break into a vehicle or building to steal something
7. sell hard drugs such as heroin, cocaine, and LSD
8. steal something worth more than $50
9. get drunk once in awhile
10. give or sell alcohol to kids under 18
Attitudes toward interpersonal violence (IPV):
l=strongly disagree
2=disagree
3=neither agree nor disagree
4=agree
5=strongly agree
1. It is all right to beat up people if they started
the fight.
2. It is all right to physically beat up people who
call you names.

153
3. Since the people on TV often get what they want by
using violence, it's probably all right for you to
use it too.
4. If people do something to make you really mad, they
deserve to be beaten up.
5. It's OK to hit someone to get them to do what you
want.
6. If you don't physically fight back, people will walk
all over you.
Deviant neutralizations (EDN):
l=strongly disagree
2=disagree
3=neither agree nor disagree
4=agree
5=strongly agree
1. To stay out of trouble, it is sometimes necessary to
lie to teachers.
2. At school it is sometimes necessary to play dirty in
order to win.
3. It's okay to lie if it keeps your friends out of
trouble.
4. In order to gain the respect of your friends, it's
sometimes necessary to beat up on other kids.
5. You have to be willing to break some rules if you
want to be popular with your friends.
6. Sometimes it's necessary to lie to your parents in
order to keep their trust.
7. It may be necessary to break some of your parents'
rules in order to keep some of your friends.
Differential Association
Delinquent peer association (DPA): "During the last
year how many of (your friends) have ...?”
l=none of them
2=very few of them
3=some of them
4=most of them
5=all of them
1. purposely damaged or destroyed property that did not
belong to them
2. used marijuana or hashish
3. stolen something worth less than $5
4. hit or threatened to hit someone without any reason
5. used alcohol
6. broken into a vehicle or building to steal something

154
7. sold hard drugs such as heroin, cocaine, and LSD
8. stolen something worth more than $50
9. gotten drunk once in awhile
10. sold or given alcohol to kids under 18
Sociodemoqraphic Variables
Age
"How old are you now?" Possible ages range from 13 to
20 at wave three (1978).
Sex
Interviewer notes the sex of the respondent.
1. Males
2. Females
Race
"Which one of these groups best describes you?" This
question was asked only if the interviewer could not visually
determine the answer.
1. Anglo or White
2. Black
3. Mexican-American
4. Spanish-American
5. Chicano
6. American Indian
7. Asian
8. Puerto Rican
9. Other (specify)
The NYS subsequently collapsed the Mexican-American,
Spanish-American, Chicano, and Puerto Rican categories into a
single category labeled "Hispanic."
Class
Income:
"Using the gold card, what would you say was the
approximate total family income last year, including all
sources before taxes?" This question was asked of parents.

155
1.
$ 6,
,000
or
under
2.
6,
,001
to
10,
000
3.
10,
,001
to
14,
000
4.
14,
,001
to
18,
000
5.
18,
,001
to
22,
000
6.
22,
,001
to
26,
000
7.
26,
,001
to
30,
000
8.
30,
,001
to
34,
000
9.
34,
.001
to
38,
000
10.
38,
,001
or
more
Education:
"How far did you go in school?" This question was asked
of each parent.
1. post-graduate degree
2. completed college
3. some college; completed specialized training or
education; some specialized training or education
4. completed high school (grade 12 or GED)
5. some high school (grades 9-12)
6. completed grade school (grades K-8)
7. some grade school
The following recoding was used for the present study:
1. (High) a summed parental score of 5 or less
2. (Medium) a summed parental score of 6 to 8
3. (Low) a summed parental score of 9 or more
Number of Siblings
"How many children or youth under 18 (including subject)
live in this home?" This question was asked of parents.
Possible responses range from 1 to 9 or more.

APPENDIX B
ADDITIONAL REGRESSION TABLES
Table 19. Sociodemographic Variables, Family Structure,
Social Learning, and Delinquency: OLS
Regres sion (N= 9 8 6)
(1)
(2)
(3)
Sociodemoqraphics
Age
.06
.05
-.02
Sex
-.17 **
-.17 **
-.08 *
Race
.05
.03
.04
Education
.07 *
.06
.05
Income
-.02
.00
.01
Family Structure
.09 **
.03
Social Learnina
Parents’ Reactions
.04
Friends’ Reactions
.06
Delinq. Attitudes
-.01
Violence Attitudes
.10 *
Neutralizations
-.01
Delinq. Peers
.26 **
R2=.047
R2=. 055
R2=. 159
* significant at
** significant at
the .05
the .01
level
level
156

157
Table 20. Family Structure, Social Learning, and
Delinquency: Analysis Plan 2 Using Logistic
Regression (N=806)
(1)
(2)
(3)
Family Structure T2
.13 *
.07
.10
Social Learnino T2
Parents' Reactions
-.06
-.08
Friends' Reactions
.13
.09
Delinq. Attitudes
.11
.18
Violence Attitudes
.24 *
.18
Neutralizations
.10
.10
Delinq. Peers
.33 **
.37 *’
Sociodemoaraohics
Age
-.17 *
Sex
-.19 *
Race
.06
Education
.05
Income
.08
X2=4.6
1 d.f.
X2=101.1
7 d.f.
X2=108.
12 d.f.
* significant at the .05 level
** significant at the .01 level

158
Table 21. Family Structure, Social Learning, and
Delinquency: Analysis Plan 3 Using Logistic
Regression (N=983)
(1)
(2)
(3)
Family Structure T1
.15 **
.07
.03
Social Learnina T1
Parents' Reactions
.06
.03
Friends' Reactions
.26 **
.24 **
Delinq. Attitudes
-.03
.01
Violence Attitudes
.31 **
.23 **
Neutralizations
-.08
-.10
Delinq. Peers
.26 **
.28 ’*
SociodemoqraDhics
Age
-.00
Sex
-.19 **
Race
.11
Education
.06
Income
.03
X2=ll. 4
1 d.f.
X2=130.7
7 d.f.
X2=142.8
12 d.f.
* significant
** significant
at the .05
at the .01
level
level

159
Table 22. Family Structure, Social Learning, and
Delinquency: Analysis Plan 4 Using Logistic
Regression (N=l,005)
(1)
(2)
(3)
Family Structure Tl
.09
-.02
-.06
Social Learninq T2
Parents’ Reactions
-.04
-.04
Friends' Reactions
.17
.09
Delinq. Attitudes
.13
.23 *
Violence Attitudes
.18 *
.12
Neutralizations
.05
.03
Delinq. Peers
.32 **
.34 **
Sociodemoqraphics
Age
-.10
Sex
-.19 *
Race
.12
Education
.01
Income
.03
X2=2.4
1 d.f.
X2=109.9
7 d.f.
X2=117.8
12 d.f.
* significant
at the .05
level
** significant at the .01 level

160
Table 23. Family Structure, Social Learning, and
Delinquency: Analysis Plan 3 Using Logistic
Regression for Male Subsample (N=514)
(1)
(2)
(3)
Family Structure T1
.08
-.00
-.04
Social Learnina T1
Parents' Reactions
.07
.05
Friends' Reactions
.13
.16
Delinq. Attitudes
-.08
-.06
Violence Attitudes
.17 *
.16
Neutralizations
-.04
-.06
Delinq. Peers
.32 **
.32 **
Sociodemooraohics
Age
.02
Race
.15 *
Education
.06
Income
.12
X2=2.1
1 d.f.
X2=50.2
7 d.f.
X2=56.3
11 d.f.
* significant at
** significant at
the .05
the .01
level
level

161
Table 24. Family Structure, Social Learning, and
Delinquency: Analysis Plan 3 Using Logistic
Regression for Female Subsample (N=469)
(1)
(2)
(3)
Family Structure T1
.28 **
.19 *
.15
Social Learning T1
Parents' Reactions
.04
-.01
Friends' Reactions
.46 *
.41 *
Delinq. Attitudes
.09
.24
Violence Attitudes
.38 **
.35 *
Neutralizations
-.20
-.19
Delinq. Peers
.18
.18
Sociodemoqraphics
Age
-.07
Race
.04
Education
.07
Income
-.25
X2=12.8
1 d.f.
X2=68.6
7 d.f.
X2=74.9
11 d.f.
* significant
** significant
at the .05
at the .01
level
level

162
Table 25. Family Structure, Social Learning, and
Delinquency: Analysis Plan 4 Using Logistic
Regression for Male Subsample (N=528)
(1)
(2)
(3)
Family Structure T1
-.00
-.09
-.12
Social Learning T2
Parents' Reactions
-.01
-.01
Friends' Reactions
.09
.02
Delinq. Attitudes
.14
.22
Violence Attitudes
.01
.01
Neutralizations
-.01
-.02
Delinq. Peers
.35 **
.39 **
Sociodemograchics
Age
-.11
Race
.15
Education
.00
Income
.10
X2= .00
1 d.f.
X2=48.8
7 d.f.
X2=52.7
11 d.f.
* significant at the .05 level
** significant at the .01 level

163
Table 26. Family Structure, Social Learning, and
Delinquency: Analysis Plan 4 Using Logistic
Regression for Female Subsample (N=477)
(1)
(2)
(3)
Family Structure T1
..24 **
.14
.05
Social Learning T2
Parents' Reactions
-.11
-.16
Friends' Reactions
.28
.32
Delinq. Attitudes
.13
.18
Violence Attitudes
.35 *
.31
Neutralizations
.18
.17
Delinq. Peers
.31 *
.29 *
Sociodemoqraphics
Age
.01
Race
.09
Education
.02
Income
-.16
X2=6.8
1 d.f.
X2=60.2
7 d.f.
X2=62.0
11 d.f.
* significant
** significant
at the .05
at the .01
level
level

REFERENCES
Agnew, Robert. 1983. "Physical Punishment and Delinquency."
Youth and Society 15:2:225-36.
Akers, Ronald L. 1973. Deviant Behavior: A Social Learning
Approach. Belmont, CA: Wadsworth.
Akers, Ronald L. 1977. Deviant Behavior: A Social Learning
Approach. 2 ed. Belmont, CA: Wadsworth.
Akers, Ronald L. 1985. Deviant Behavior: A Social Learning
Approach. 3 ed. Belmont, CA: Wadsworth. Reprinted 1992.
Fairfax, VA: Techbooks.
Akers, Ronald L. 1994. Criminological Theories: Introduction
and Evaluation. Los Angeles: Roxbury.
Akers, Ronald L. 1997. Criminological Theories: Introduction
and Evaluation. 2 ed. Los Angeles: Roxbury.
Akers, Ronald L. 1998. Social Learning and Social Structure:
A General Theory of Crime and Deviance. Boston:
Northeastern University Press.
Akers, Ronald L., Marvin D. Krohn, Lonn Lanza-Kaduce, and
Marcia Radosevich. 1979. "Social Learning and Deviant
Behavior: A Specific Test of a General Theory." American
Sociological Review 44:636-55.
Akers, Ronald L., Marvin D. Krohn, Marcia Radosevich, and
Lonn Lanza-Kaduce. 1981. "Social Characteristics and
Self-Reported Delinquency: Differences in Extreme
Types." Pp. 48-62 in Sociology of Delinquency: Current
Issues. Ed. Gary F. Jensen. Beverly Hills, CA: Sage.
Akers, Ronald L. and Gang Lee. 1996. "A Longitudinal Test of
Social Learning Theory: Adolescent Smoking." Journal of
Drug Issues 26(2) :317-43.
Akers, Ronald L. and Gang Lee. 1997. "Age, Social Learning,
and Social Bonding in Adolescent Substance Use."
Unpublished paper. Department of Sociology, University
of Florida.
164

165
Akers, Ronald L., James Massey, William Clarke, and Ronald M.
Lauer. 1983. "Are Self-Reports of Adolescent Deviance
Valid? Biochemical Measures, Randomized Response, and
the Bogus Pipeline in Smoking Behavior.” Social Forces
62:1:234-51.
Belsley, David A., Edwin Kuh, and Roy E. Welsch. 1980.
Regression Diagnostics: Identifying Influential Data and
Sources of Collinearity. New York: John Wiley and Sons.
Berk, Richard A. 1983. "Applications of the General Linear
Model to Survey Data." Pp. 495-546 in Handbook of Survey
Research. Eds. Peter H. Rossi, James D. Wright, and Andy
B. Andersen. New York: Academic Press.
Boeringer, Scot and Ronald L. Akers. 1993. "Rape and Rape
Proclivity: A Comparison of Social Learning, Social
Control, and Relative Deprivation Models." Unpublished
paper. Department of Sociology, University of Florida.
Boeringer, Scot, Constance L. Shehan, and Ronald L. Akers.
1991. "Social Contexts and Social Learning in Sexual
Coercion and Aggression: Assessing the Contribution of
Fraternity Membership." Family Relations 40:58-64.
Brownfield, David and Ann Marie Sorenson. 1994. "Sibship Size
and Sibling Delinquency." Deviant Behavior 15:45-61.
Burgess, Robert L. and Ronald L. Akers. 1966. "A Differential
Association-Reinforcement Theory of Criminal Behavior."
Social Problems 14:128-47.
Cashion, Barbara G. 1982. "Female-Headed Families: Effects on
Children and Clinical Implications." Journal of Marital
and Family Therapy (April):77-85.
Cernkovich, Stephen A. and Peggy C. Giordano. 1987. "Family
Relationships and Delinquency." Criminology 25:295-321.
Chen, Zeng-yin and Howard B. Kaplan. 1997. "The Impact of
Family Structure during Adolescence on Deviance in Early
Adulthood." Deviant Behavior 18:4:365-91.
Coughlin, Chris and Samuel Vuchinich. 1996. "Family
Experience in Preadolescence and the Development of Male
Delinquency." Journal of Marriage and the Family 58:491-
501.
Curran, Daniel J. and Claire M. Renzetti. 1996. Social
Problems: Society in Crisis. 4 ed. Needham Heights, MA:
Allyn and Bacon.

166
Dornbusch, Sanford M., J. Merrill Carlsmith, Steven J.
Bushwall, Philip L. Ritter, Herbert Leiderman, Albert H.
Hastorf, and Ruth T. Gross. 1985. "Single Parents,
Extended Households, and the Control of Adolescents."
Child Development 56:326-41.
Elliott, Delbert S. 1994. "Serious Violent Offenders: Onset,
Developmental Course, and Termination—The American
Society of Criminology 1993 Presidential Address."
Criminology 32:1:1-21.
Elliott, Delbert S. and Suzanne S. Ageton. 1980. "Reconciling
Race and Class Differences in Self-Reported and Official
Estimates of Delinquency." American Sociological Review
45:95-110.
Elliott, Delbert S. and David Huizinga. 1983. "Social Class
and Delinquent Behavior in a National Youth Panel."
Criminology 21:2:149-77.
Elliott, Delbert S., David Huizinga, and Suzanne S. Ageton.
1985. Explaining Delinquency and Drug Use. Beverly
Hills, CA: Sage.
Elliott, Delbert S., David Huizinga, and Scott Menard. 1989.
Multiple Problem Youth: Delinquency, Substance Use, and
Mental Health Problems. New York: Springer-Verlag.
Esbensen, Finn-Aage and Delbert S. Elliott. 1994. "Continuity
and Discontinuity in Illicit Drug Use: Patterns and
Antecedents." The Journal of Drug Issues 24:1:75-97.
Eshleman, J. Ross. 1991. The Family: An Introduction. 6 ed.
Needham Heights, MA: Allyn and Bacon.
Fagan, Jeffrey and Sandra Wexler. 1987. "Crime at Home and in
the Streets: The Relationship between Family and
Stranger Violence." Violence and Victims 2:1:5-23.
Farnworth, Margaret. 1984. "Family Structure, Family
Attributes, and Delinquency in a Sample of Low-Income,
Minority Males and Females." Journal of Youth and
Adolescence 13:4:349-64.
Finkelhor, David. 1984. Child Sexual Abuse: New Theory and
Research. New York: Free Press.
Free, Marvin D. 1991. "Clarifying the Relationship between
the Broken Home and Juvenile Delinquency: A Critique of
the Current Literature." Deviant Behavior 12:109-67.
Furstenberg, Frank F., Jr. and Julien 0. Teitler. 1994.
"Reconsidering the Effects of Marital Disruption."
Journal of Family Issues 15:173-90.

167
Gelles, Richard J. and Murray A. Straus. 1988. Intimate
Violence. New York: Simon and Schuster.
Gelles, Richard J. and Jon R. Conte. 1991. "Domestic Violence
and Sexual Abuse of Children: A Review of Research in
the Eighties." Pp. 327-40 in Contemporary Families:
Looking Forward, Looking Back. Ed. Alan Booth.
Minneapolis, MN: National Council on Family Relations.
Gove, Walter R. and Robert D. Crutchfield. 1982. "The Family
and Juvenile Delinquency." The Sociological Quarterly
23:301-19.
Hagan, Frank E. and Marvin B. Sussman. 1988. "Deviance and
the Family: Where Have We Been and Where Are We Going?"
Marriage and Family Review 12:1-2:1-22.
Heimer, Karen. 1997. "Socioeconomic Status, Subcultural
Definitions, and Violent Delinquency." Social Forces
75:3:799-833.
Heimer, Karen and Ross L. Matsueda. 1994. "Role-Taking, Role
Commitment, and Delinquency: A Theory of Differential
Social Control." American Sociological Review 59:365-90.
Hennessy, Michael, Pamela J. Richards, and Richard A. Berk.
1978. "Broken Homes and Middle Class Delinquency."
Criminology 15:4:505-528.
Hindelang, Michael J., Travis Hirschi, and Joseph G. Weis.
1979. "Correlates of Delinquency: The Illusion of
Discrepancy Between Self-Report and Official Measures."
American Sociological Review 44:995-1014.
Hirschi, Travis. 1969. Causes of Delinquency. Berkeley:
University of California Press.
Hollingshead, August B. and Fredrick C. Redlich. 1958. Social
Class and Mental Illness: A Community Study. New York:
Wiley.
Hosmer, David W., Jr. and Stanley Lemeshow. 1989. Applied
Logistic Regression. New York: John Wiley and Sons.
Huizinga, David and Delbert S. Elliott. 1986. "Reassessing
the Reliability and Validity of Self-Report Delinquency
Measures." Journal of Quantitative Criminology 2:4:293-
327.
Jensen, Gary F. and Raymond Eve. 1976. "Sex Differences in
Delinquency." Criminology 13:4:427-48.

168
Jensen, Gary F. and Dean G. Rojek. 1992. Delinquency and
Youth Crime. 2 ed. Prospect Heights, IL: Waveland Press.
Johnson, Richard E. 1986. "Family Structure and Delinquency:
General Patterns and Gender Differences." Criminology
24:1:65-80.
Koski, Patricia R. 1988. "Family Violence and Nonfamily
Deviance: Taking Stock of the Literature." Marriage and
Family Review 12:1-2:23-46.
Krohn, Marvin D., Lonn Lanza-Kaduce, and Ronald L. Akers.
1984. "Community Context and Theories of Deviant
Behavior: An Examination of Social Learning and Social
Bonding Theories." The Sociological Quarterly 25:353-71.
Krohn, Marvin D., Susan B. Stern, Terence P. Thornberry, and
Sung Joon Jang. 1992. "The Measurement of Family Process
Variables: The Effect of Adolescent and Parent
Perceptions of Family Life on Delinquent Behavior."
Journal of Quantitative Criminology 8:287-315.
Laing, Ronald D. 1962. The Self and Others: Further Studies
in Sanity and Madness. Chicago: Quadrangle Books.
Laing, Ronald D. and Aaron Esterson. 1971. Sanity, Madness,
and the Family: Families of Schizophrenics. 2 ed. New
York: Basic Books.
Lattimore, Pamela K., Christy A. Visher, and Richard L.
Linster. 1995. "Predicting Rearrest for Violence Among
Serious Youthful Offenders." Journal of Research in
Crime and Delinquency 32:1:54-83.
Laub, John H. and Robert J. Sampson. 1988. "Unraveling
Families and Delinquency: A Reanalysis of the Gluecks'
Data." Criminology 26:355-80.
Lauritsen, Janet L. 1993. "Sibling Resemblance in Juvenile
Delinquency: Findings from the National Youth Survey."
Criminology 31:387-409.
Loeber, Rolf and Magda Stouthamer-Loeber. 1986. "Family
Factors as Correlates and Predictors of Juvenile Conduct
Problems and Delinquency." Pp. 29-149 in Crime and
Justice, Vol. 7. Eds. Michael Tonry and Norval Morris.
Chicago: University of Chicago.
Matsueda, Ross L. and Karen Heimer. 1987. "Race, Family
Structure, and Delinquency: A Test of Differential
Association and Social Control Theories." American
Sociological Review 52:826-40.

169
McCord, Joan. 1991. "Family Relationships, Juvenile
Delinquency, and Adult Criminality." Criminology
29:3:397-417.
McLanahan, Sara and Karen Booth. 1989. "Mother-Only Families:
Problems, Prospects, and Politics." Journal of Marriage
and the Family 51:557-80.
McLeod, Jane D., Candace Kruttschnitt, and Maude Dornfeld.
1994. "Does Parenting Explain the Effects of Structural
Conditions on Children's Antisocial Behavior? A
Comparison of Blacks and Whites." Social Forces
73:2:575-604.
Mednick, Birgitte R., Robert L. Baker, and Linn E. Carothers.
1990. "Patterns of Family Instability and Crime: The
Association of Timing of the Family's Disruption with
Subsequent Adolescent and Young Adult Criminality."
Journal of Youth and Adolescence 19:3:201-20.
Mednick, Birgitte, Charlotte Reznick, Dennis Hocevar, and
Robert Baker. 1987. "Long-Term Effects of Parental
Divorce on Young Adult Male Crime." Journal of Youth and
Adolescence 16:1:31-45.
Menard, Scott and Delbert S. Elliott. 1994. "Delinquent
Bonding, Moral Beliefs, and Illegal Behavior: A Three-
Wave Panel Model." Justice Quarterly 11:2:173-88.
Messner, Steven F. and Robert J. Sampson. 1991. "The Sex
Ratio, Family Disruption, and Rates of Violent Crime:
The Paradox of Demographic Structure.” Social Forces
69:3:693-713.
Osgood, D. Wayne, Patrick M. O'Malley, Jerald G. Bachman, and
Lloyd D. Johnston. 1989. "Time Trends and Age Trends in
Arrests and Self-Reported Illegal Behavior." Criminology
27:3:389-416.
Pagelow, Mildred Daley. 1984. Family Violence. New York:
Praeger Press.
Paternoster, Raymond and Robert Brame. 1997. "Multiple Routes
to Delinquency? A Test of Developmental and General
Theories of Crime." Criminology 35:1:49-84.
Patterson, Gerald R. and Thomas J. Dishion. 1985.
"Contributions of Families and Peers to Delinquency."
Criminology 23:1:63-79.
Petras, James and Christian Davenport. 1991. "Crime and the
Transformation of Capitalism." Crime, Law, and Social
Change 16:155-75.

170
Rankin, Joseph H. and Roger Kern. 1994. "Parental Attachments
and Delinquency." Criminology 32:495-515.
Rosen, Lawrence. 1985. "Family and Delinquency: Structure or
Function?" Criminology 23:553-73.
Rowe, David C. and David P. Farrington. 1997. "The Familial
Transmission of Criminal Convictions." Criminology
35:1:177-201.
Salts, Connie J., Byron W. Lindholm, H. Wallace Goddard, and
Stephen Duncan. 1995. "Predictive Variables of Violent
Behavior in Adolescent Males." Youth and Society 26:377-
99.
Sampson, Robert J. 1985. "Neighborhood and Crime: The
Structural Determinants of Personal Victimization."
Journal of Research in Crime and Delinquency 22:1:7-40.
Sampson, Robert J. 1987. "Urban Black Violence: The Effect
of Male Joblessness and Family Disruption." American
Journal of Sociology 93:2:348-82.
Sampson, Robert J. and John H. Laub. 1993. Crime in the
Making: Pathways and Turning Points through Life.
Cambridge, MA: Harvard University Press.
Satir, Virginia. 1972. Peoplemaking. Palo Alto, CA: Science
and Behavior Books.
Shihadeh, Edward S. and Darrell J. Steffensmeier. 1994.
"Economic Inequality, Family Disruption, and Urban Black
Violence: Cities as Units of Stratification and Social
Control." Social Forces 73:2:729-51.
Siegel, Larry J. and Joseph J. Senna. 1991. Juvenile
Delinquency: Theory, Practice, and Law. 4 ed. St. Paul,
MN: West Publishing Co.
Smith, Carolyn and Marvin D. Krohn. 1995. "Delinquency and
Family Life among Male Adolescents: The Role of
Ethnicity." Journal of Youth and Adolescence 24:69-93.
Smith, Douglas A. and G. Roger Jarjoura. 1988. "Social
Structure and Criminal Victimization." Journal of
Research in Crime and Delinquency 25:1:27-52.
Snyder, James and Gerald R. Patterson. 1987. "Family
Interaction and Delinquent Behavior." Pp. 216-43 in
Handbook of Juvenile Delinquency. Ed. Herbert C. Quay.
New York: Wiley.

171
Straus, Murray A. 1991. "Discipline and Deviance: Physical
Punishment of Children and Violence and Other Crime in
Adulthood." Social Problems 38:2:133-54.
Straus, Murray A. and Richard J. Gelles. 1986. "Societal
Change and Change in Family Violence from 1975 to 1985
As Revealed by Two National Surveys." Journal of
Marriage and the Family 48:465-79.
Sutherland, Edwin H. 1947. Criminology. 4 ed. Philadelphia:
Lippincott.
Thomas, George, Michael P. Farrell, and Grace M. Barnes.
1996. "The Effects of Single-Mother Families and
Nonresident Fathers on Delinquency and Substance Abuse
in Black and White Adolescents." Journal of Marriage and
the Family 58:884-94.
U.S. Department of Commerce, Bureau of the Census. 1994a.
Statistical Abstract of the United States. Washington,
D.C.: U.S. Government Printing Office.
U.S. Department of Commerce, Bureau of the Census. 1994b.
Current Population Reports. Series P-70-34. Washington,
D.C.: U.S. Government Printing Office.
U.S. Department of Justice, Federal Bureau of Investigation.
1992. Crime in the United States, 1991. Washington,
D.C.: U.S. Government Printing Office.
Van Voorhis, Patricia, Francis T. Cullen, Richard A. Mathers,
and Connie Chenoweth Garner. 1988. "The Impact of Family
Structure and Quality on Delinquency: A Comparative
Assessment of Structural and Functional Factors."
Criminology 26:2:235-61.
Void, George B. and Thomas J. Bernard. 1986. Theoretical
Criminology. 3 ed. New York: Oxford University.
Warr, Mark. 1993. "Parents, Peers, and Delinquency." Social
Forces 72:247-64.
Warr, Mark and Mark Stafford. 1991. "The Influence of
Delinquent Peers: What They Think or What They Do?"
Criminology 29:4:851-66.
Wells, L. Edward and Joseph H. Rankin. 1985. "Broken Homes
and Juvenile Delinquency: An Empirical Review." Criminal
Justice Abstracts 17:249-72.
Wells, L. Edward and Joseph H. Rankin. 1986. "The Broken
Homes Model of Delinquency: Analytic Issues." Journal of
Research in Crime and Delinquency 23:1:68-93.

172
Wells, L. Edward and Joseph H. Rankin. 1991. "Families and
Delinquency: A Meta-Analysis of the Impact of Broken
Homes.” Social Problems 38:1:71-93.
West, Donald J. 1982. Delinquency, Its Roots, Careers, and
Prospects. Cambridge, MA: Harvard University Press.
Widom, Cathy Spatz. 1989a. "Child Abuse, Neglect, and Violent
Criminal Behavior." Criminology 27:2:251-71.
Widom, Cathy Spatz. 1989b. "The Cycle of Violence." Science.
244:160-6.
Widom, Cathy Spatz. 1996. "Childhood Sexual Abuse and Its
Criminal Consequences." Society (May/June):47-53.
Winfree, L. Thomas, Jr., Teresa Vigil Backstrom, and G. Larry
Mays. 1994. "Social Learning Theory, Self-Reported
Delinquency, and Youth Gangs." Youth and Society
26:2:147-77.

BIOGRAPHICAL SKETCH
Edwin R. Page was born in 1966 in Bradenton, Florida, to
parents Suzanne Robson and William Eugene Page III. He
received his Bachelor of Science degree in journalism at the
University of Florida in 1988, and his Master of Arts degree
in sociology at the University of Florida in 1991.
173

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Ronáld L. Akers, Chair
Professor of Sociology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree oJ_Deetpr of Philosophy.
Professor of Sociology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,^_as
a dissertation for the degree ojf'DocTBor of Philosophy. ^
(P2C1
Lonn M. Lariza-Kadi
Associate Professor of
Sociology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
- /<¡róhn H.' Scanzoni
Professor of Sociology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctorj&f Philosophy.
Hernán Vera
Professor of Sociology

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Allan F. Burns
Professor of Anthropology
This dissertation was submitted to the Graduate Faculty
of the Department of Sociology in the College of Liberal Arts
and Sciences and to the Graduate School and was accepted as
partial fulfillment of the requirements for the degree of
Doctor of Philosophy.
August, 1998
Dean, Graduate School

ID
mo
msr
uwveSSIlSi.nlwilM
3 1262 0B556 6676